Forest Flammability Modelling and Managing a Complex System Philip Zylstra A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy School of Physical, Environmental and Mathematical Sciences The University of New South Wales Australian Defence Force Academy To the Bemerangal, Wolgalu and Ngarragu people of the Snowy Mountains and Monaro for your inspiration and willingness to share with me some of the knowledge and passion you have for this country. 2 ACKNOWLEDGEMENTS This project was funded jointly by the Bushfire Cooperative Research Centre, the NSW Department of Environment, Climate Change and Water and the University of New South Wales Australian Defence Force Academy. I am grateful to fire and operations management staff in the National Parks and Wildlife Service – Ian Dicker, Megan Bowden, Anthony Evans and Dave Darlington for providing me with the time and support to complete the work. I am particularly grateful to my primary supervisor Rod Weber for his preparedness to take on both me and my project, and for his ongoing encouragement and patient assistance with those areas of mathematics and physics that I struggled with but he did not. Also to my cosupervisory team – Roger Good who took up my cause from the beginning and gained me the support of NPWS, Ross Bradstock who set my course in the right direction with the right people, Geoff Cary who helped so much to organise and focus me as well as providing useful references and review and Malcolm Gill who was always so generous with thoughts, discussion, equipment and assistance in writing. Both Joanne Chapman and Peter Hayes provided valuable advice on statistics. Mike Erskine was instrumental in helping me construct my original concept by sharing his engineering perspectives on flammability and by providing very constructive review, and Ron Needham gave me the breakthrough I needed in geometry to continue. Rod Mason, Alice Williams and Mike Young have devoted great amounts of time discussing Aboriginal fire and land management with me, and I am particularly grateful to Rod and Alice for their assistance in helping me to see the world from a new perspective. Marta Yebra Alvarez has been a great support in discussing ideas on fire and live fuel moisture as well as facilitating many worthwhile introductions and interactions and being a good friend throughout. Domingos X. Viegas and Gavril Xanthopoulos devoted a considerable amount of time to provide useful feedback and information. 3 Others that have provided insight and encouragement through thoughtful and informed discussion include the fellow firefighters I have discussed practicalities with over a rake hoe in remote corners of the country, in particular Dave Hipwell and Grant Brewer. Also Paul Wadland for his help with Excel, Jason Sharples for his conversations on flame attachment, Mariano García for bringing me up to speed on LIDAR and Steve Nossiter, Mike Fairbairn and Nicole Shotter for providing very helpful perspectives on a number of occasions. Nicky also put in considerable time helping me wordsmith important sections. The experimental burns I conducted could not have happened without the help of Ian Dicker, Steve Wright, Michael Austin, Ailish Pope, Andrew Miller, Tim Greville, Fiona Solomon and Roy Hedger. Thanks too Ian for the wind vane you made me. Last and most importantly I want to thank my wife Donwen and son Zak for their great patience and support for me during the many hours I spent looking at the laptop rather than their faces, and for the assistance they and Kelsey Jamieson have given me in field and lab work. It was good burning things in the kitchen with you Donwen. 4 ABSTRACT Wildland fire has considerable influence on both natural and anthropogenic environments and consequently, the ability to understand, predict and manage it has become a growing priority as human populations have increased their influence upon and awareness of the natural, fire prone environment. Despite this, in Australia a disconnection and failure to transfer traditional understandings of the subject into the modern context has coupled with an inadequate and frequently simplistic grasp of fuels and their implications for management. This has been manifested in ineffective fuel management, a low level of confidence in fire behaviour predictions and a paucity of peer-reviewed science on the subject. Concepts of forest flammability still do little to address concerns and observations that were raised in some priority fuel arrays over a century ago. This thesis attempts to address the dilemma by identifying where the gaps in Australian fire knowledge lie and by offering a unique approach to the situation using complex systems modelling. A structure is proposed that by taking a literal approach to modelling the propagation of flame through and between plants and fuel strata creates a transparent framework for understanding and analysing fire behaviour. Fire spread is conceptualised as the interaction of the three aspects of flammability – ignitability, combustibility and sustainability with the geometry of the fuels. Fire spread occurs when a critical state is satisfied; once the new fuels are alight the resulting flame dimensions are calculated and the process is repeated. Fuels are defined as any dead or live plant material that undergoes combustion and thereby contributes to the fire behaviour. This definition excludes a priori measurements as the three dimensional structure of a forest imposes a circular feedback between flame dimensions and fuel availability. This level of complexity is critical as the same vegetation that acts as fuel may also act to suppress fire spread by affecting drying processes or reducing wind speed at different heights in 5 the forest. As a result, the concept of simple fuel reduction or the reduction of carbon storage or biomass as a proxy for managing flammability is shown to be fundamentally flawed. It is proposed that effective management of the flammability of forests or any other fuel array is only possible by understanding the complex relationships between potential fuels and forest structure. To achieve this, the necessary sub-models required to create a fully operational fire behaviour model are identified, sourced from the literature or developed where necessary. New models were constructed to describe the ignitability, combustibility and sustainability of flame in individual leaves, as well as describing external influences on fire behaviour such as the effect of slope on flame angle and the leaf area index of the forest on wind speeds at different heights. Plant moisture models were developed for six species to enable validation of the model, and the Keetch-Byram Drought Index was tested as a model of soil moisture. The models used are provided as “first generation” models – capable of fulfilling the role while identifying where future work is required to improve on them. The complete fire behaviour model was constructed in an Excel spreadsheet and validated with an extreme condition test, predictive validation and a credibility analysis. The extreme condition test demonstrated that the model was not subject to the same domain considerations of empirical models but that modelled fire behaviour was inherently limited by the physical processes which underlie the model. Model accuracy was compared with the performance of three widely used Australian models and provided a improvement of 4 to 12 times greater accuracy to all three on rate of spread, which was statistically significant for two of the models It also demonstrated up to 12 times greater accuracy when estimating flame heights, although this was only significant against one model. The credibility analysis tested all four models against decision thresholds for prescribed burn planning, determining initial attack success against unplanned fires, determining the attack method to be employed against unplanned fires and as tools for forward planning in wildfire management. The model was slightly out performed on one test but performed more reliably than two of the other three on 6 all other counts except for forward planning in incidents, where it significantly out-performed all other models. Suitable applications of the model were examined and examples given, demonstrating that new areas of research into fuel management, fuel-weather interactions and feedbacks between fire and climate change were now possible due to the model. The post-fire succession of an area of sub-alpine forest was used to model the changes in flammability with time since fire as an example application. The results demonstrated that in these conditions, the fuel-age paradigm of fire management was an inappropriate simplification that if followed would produce counter-productive results in some environments. Further examples of fire-weather interactions were given and the implications for fire management and climate change examined. 7 CONTENTS Dedication Acknowledgements Abstract Publications by the author contributing to this thesis Introduction 1 The Gaps in Australian Fire Knowledge 1.1 Traditional knowledge and the fire myth 1.11 Knowledge transfer in indigenous society 1.12 The European-Australian approach to Aboriginal knowledge 1.2 The current state of fire behaviour modelling in Australia 1.21 Identifying priority fuel arrays 1.22 Australian fire behaviour modelling for priority fuel arrays 1.23 Existing alternatives for Australia 1.24 Summary 2 Fuels and Flammability 2.1 Defining flammability 2.2 Describing and classifying fuels 2.21 Fuel dynamics 2.3 Succession and fire ecology 2.31 Fuel ecology 2.4 Fuel age and flammability 2.41 Conclusion 3 A Conceptual Model of Fire Behaviour 3.1 Introduction 3.2 Methods 3.21 Defining the critical state 8 3.22 Fire growth patterns 3.23 A conceptual model of complex fire behaviour 3.3 The model 3.31 Steps of ignition 3.4 Discussion 4 Leaf Flammability 4.1 Introduction 4.2 Temperature of ignition 4.3 Ignition delay time 4.31 Introduction 4.32 Methods 4.33 Results 4.34 Conclusion 4.4 Combustibility and sustainability 4.41 Introduction 4.42 Methods 4.43 Results 4.44 Discussion 4.5 Leaf Flammability 5 Plant Flammability 5.1 Plant architecture 5.11 Angles and dimensions 5.2 Plant ignitability 5.21 Temperature of a plume 5.22 Plume angle 5.23 The critical state test 5.24 Dimensions of the plume pathway 5.25 Preheating 5.26 Finding depth of ignition 5.3 Plant combustibility 5.31 Branch architecture 5.32 The number of leaves burning simultaneously 9 5.33 Flame merging 5.4 Plant flame sustainability 5.5 Summary 6 Forest Flammability 6.1 Forest structure and geometry 6.2 Interactions of forest structure with wind and moisture 6.21 Wind speed 6.22 Dead fuel moisture 6.23 Live fuel moisture 6.24 Soil moisture 6.3 Fire spread in surface fuels 6.4 Forest fire behaviour 6.41 Fire spread within a stratum 6.42 Fire spread between strata 6.43 Flame dimensions 6.44 Rate of spread 6.5 Discussion 7 Model Validation 7.1 Overview 7.11 Validating models 7.12 The approach adopted 7.2 Extreme condition test 7.21 Methods 7.22 Results 7.23 Discussion 7.3 Predictive validation and credibility analysis 7.31 Methods of validation 7.32 Experimental studies 7.33 Guthega, 26th January 2003 7.34 Tooma Dam, 13th October 2006 7.35 Kilmore east Mountain Ash, 7th February 2009 10 7.4 Model strengths and weaknesses 7.41 Predictive validation 7.42 Credibility analysis 7.43 Discussion 8 Implementation and applications 8.1 Implementation of the model as an operational tool 8.2 Application of the model 8.21 Fuel management 8.22 Weather and climate interactions Conclusions References Appendices I Symbols II Equations and assumptions III Ignition delay time of leaves IV Length and duration of flames from different leaf species V Moisture content (% ODW) and corresponding soil/weather conditions VI KBDI and measured/modelled soil moisture values VII Operating the Forest Flammability Model spreadsheet. 11 PUBLICATIONS BY THE AUTHOR The author was either the sole or co-author in each of the following works written as part of this research thesis. Gill AM, Zylstra P (2005) Flammability of Australian Forests. Australian Forestry. 68(2), 88-94 Zylstra P (2006b) Physical properties determining flammability in sclerophyllous leaves and flame propagation within shrubs and trees in the Australian Alps. From V International Conference on Forest Fire Research. Figueira da Foz: Portugal. Forest Ecology and Management. 234S, S81 The following publications drew from the work presented in this thesis. Zylstra P (2006a) ‘Fire History of the Australian Alps’. (AALC: Canberra) Zylstra P (2008) Live fuels and forest flammability. Bushfire CRC Research Poster, http://bushfirecrc.com/publications/B_Zylstra.pdf Zylstra P (2008) Perceptions and evidence of pre-European fire in the Australian Alps. Bushfire CRC Research Poster, http://www.bushfirecrc.com/publications/B_Zylstra1.pdf Zylstra P (2009) How fire works and what it means for fuel control. Bushfire CRC Firenote Issue 49. http://bushfirecrc.com/publications/downloads/0912_firenote 49_lowres.pdf 12 INTRODUCTION Quantifying and managing wildland flammability is an issue of growing importance, particularly as it has become clear that anthropogenic activity has influenced the global climate (IPCC 2007). The effect of this has been to increase temperatures and wind speeds and to reduce rainfall in fire prone areas (Hennessy et al 2005, Lucas et al 2007); factors that add to the likelihood of fire occurrence and potentially increase fire severity. Other anthropogenic influences on fire include directly altered fire regimes (Zylstra 2006a) and both direct and indirect changes to forest structure and cover (Berry and Roderick 2002, Dodson and Mooney 2002, Griffiths 2002, Lunt 2002, Zylstra 2006a). Forest flammability affects biodiversity (e.g. Possingham et al 1995, Williams et al 2008) and can severely impact human life (Viegas et al 2005, Haynes et al 2008, Taylor 2009) and economic values (e.g. WFLC 2009). Despite this reality, flammability is a concept that is not clearly defined and understood (Martin et al 1994, Behm et al 2004). Our understanding of forest fire behaviour in Australia relies extensively on models that have not been peer-reviewed and rely on empirical observations derived from a limited range of conditions and often extrapolated well beyond these conditions (Cruz and Gould 2009a). While laboratory definitions of flammability have been produced (e.g. Gill and Moore 1996), transferring these to field and management scenarios is difficult (Gill and Zylstra 2005) and complexity within the system means that the relationship between fuels and flammability is not a simple one. 13 In the absence of a clear understanding, fuel management in Australia is not aimed at directly managing flammability but at suggested indicators such as surface fuel loads (McArthur 1967), fuel scores (McCarthy et al 1999) and time since fire (McArthur 1967, Gould et al 2007). Despite their broad-scale application across the Australian landscape, these indicators have only been connected to flammability in isolated occurrences or broad-scale generalisations (e.g. McCarthy and Tolhurst 1998), although these have also as a rule been in the absence of peer-review. In some cases they have been explicitly disproven by peerreviewed science (Burrows 1994, 1999a,b) but are still in use. Until we better understand flammability across the full range of fuel arrays and conditions, it is difficult to generalise whether these practices are effective in their goals of reducing fire risk (Fernandes and Botelho 2003). In response to this reality, three objectives are identified for this thesis: 1. To identify the factors governing the flammability (ignitability, combustibility and sustainability) of leaves and the way in which these affect the flammability of plants 2. To develop a fire behaviour model capable of examining the way that all fuels in an array affect wildland fire behaviour through flammability of plants and the geometry of the fuel array. The priority for this is the forest environment and preference given to accuracy over simplicity. 3. To demonstrate the validity of this model as a tool in fuel and land management The intent is that the model produced is not limited to one area or particular context, but will provide a framework for the understanding of the role of different fuels in fire behaviour in many environments. While the priority is for understanding of forest fuels and the context of the study is Australia, it is intended that the framework developed will eventually facilitate use of the model in other fuel arrays internationally. The model also provides a “first generation” formulation intended to enable the identifications of shortfalls in the sub-models or modelling framework. Consequently, the structure is both adaptable to new advances in fire science and transparent in the way calculations are made. Because it quantifies the roles of forest structure and live fuels, the model provides entirely new avenues for research into fuel management and environmental impacts on forest flammability, such as the influence of climate change. 14 Thesis structure The broad structure of the thesis is to begin by describing the current understanding of forest flammability modelling and management. This discussion highlights some serious flaws in the underlying assumptions, and then explains how these can be addressed. The main body of the thesis describes a modelling approach designed to achieve this that quantifies flammability at the scales of leaf, plant and fuel array, then concludes by testing the model and demonstrating that it effectively addresses the issues. Chapter one Chapter one provides a theoretical background to the issues, establishing the need for an improved understanding of fire behaviour that is free of major unsupported assumptions. This is achieved by an examination of the argument in Australia that forest fire management is sufficiently understood due to an adequate retention of effective traditional knowledge and a solid scientific basis. Chapter two The scientific basis is further scrutinised in chapter two with an examination of fuel dynamics, identifying the fact that the ‘fuel-age paradigm’ is grounded in the artificial classification of certain components of a forest as “fuel” while excluding others. The ways in which fuel dynamics vary between plant communities and in response to various disturbances are discussed using a broad range of examples, illustrating that flammability particularly in a multi-stratum environment such as a forest, is determined not simply by fuel quantity but also by the structure of the forest and the fuel environment. Chapter three The ways in which forest structure and fuel quantity interact to produce fire behaviour are described and discussed in detail, explaining that these are evidence of a complex, dynamical system which cannot be characterised by a trend, an equation or an index as the propagation of fire through a forest is a successive phenomenon dependent on its own history. A conceptual model is proposed, whereby the aspects of fire physics and chemistry are allotted 15 roles as sub-models in the main model framework. This is an attempt to quantify the interactions between complex geometries within the fuel array and the individual flammability properties of its components. Chapter four Chapter four commences to populate the model with the necessary sub-models, presenting experimental studies to quantify the ignitability, combustibility and sustainability of individual leaves. The concepts of flammability are examined in detail, and empirical models are produced describing the flammability of leaves from their physical characteristics. Chapter five Chapter five examines the geometrical considerations for modelling flame within an isolated plant. Aspects of flame angles and the trajectory of the flame through a plant in relation to an idealised canopy shape are described mathematically, and equations are produced that link this with the models for flammability from chapter four to quantify the process and rate of ignition and fire propagation through a plant, and the flame characteristics produced. Chapter six Chapter six ties together the work of the previous three chapters into an operational fire behaviour model. Studies in the chapter quantify the earlier assertion that forest flammability is influenced by other aspects of the forest as well as the fuels, addressing issues such as air movement and moisture dynamics theoretically and experimentally. The geometry of a fuel array is described mathematically and the principles and process identified in the previous chapter are extended to the full forest environment so that a complete fire behaviour model is produced. A worked example of a fire behaviour calculation is produced, and the completed model is provided on an attached CD. Chapter seven This chapter provides an analysis of the way the model performs when tested against other models and against observed fire behaviour. In addition to some historic approaches, a new suite of tests is presented to describe whether any model is sufficiently valid for use in fire operations and land management. The chapter closes by identifying the strengths and weaknesses of the new model that were observed in the validation process, and describes further work required to improve the model. 16 Chapter eight Having presented a model and embarked upon a validation, chapter eight returns to the issues of fuel dynamics examined in chapter two. The newly developed model is used to examine one of the examples raised in chapter two to determine whether it is capable of providing greater insight than the fuel-age paradigm based approaches. Other aspects raised in chapter two are discussed briefly in light of the new model, examining the changes in approach that this new work might entail. The practicalities of implementing the model into fire and land management are discussed, and areas for further work identified. Conclusions The three objectives identified in this introduction are re-examined to determine whether they have been satisfied, and specific outcomes of the project are summarised. Appendices The seven appendices initially summarise the symbols, equations and assumptions used in the new model, then the following four appendices provide the raw data for the experimental studies into leaf flammability, plant moisture and soil moisture. The final appendix provides instructions on the use of the newly developed model. The prototype model itself is provided as an Excel spreadsheet along with the thesis so that the reported model predictions can be more easily repeated. 17 Chapter 1 The Gaps in Australian Fire Knowledge Fire and fuel management in Australia is directed both by science and by public opinion based on our traditional perceptions. Both of these areas are examined and significant inadequacies are identified. 1.1 Traditional knowledge and the Fire Myth It has been proposed that traditional European use of fire in Australia is a continuation of Aboriginal fire management and is therefore not to be questioned as it represents an unbroken continuation of an effective approach by an ancient culture. This is expressed in statements such as: “Aborigines burnt landscape continually and, when there was sufficient fuel to carry fire, fires burnt unchecked. Fire intensities were relatively mild even on extreme days because the fuels were light” (Cheney 1993). “Consequently, the Aborigines did not just burn now and again, or only in autumn, or when the birds were not nesting. They burnt all the time” (Ryan et al 1995). 18 “The graziers were following the practice of the Aboriginal people by using fire to keep the forests open” (Attiwill et al 2009). “What we do know, without any doubt whatever [sic], is that Aboriginal burning prior to European settlement of Australia was all-pervasive over the continent” (Bush Fire Front 2009). As in each of the cases quoted, the claim is used to argue for a particular course of fire management, specifically more frequent introduction of prescribed fire or more extensive use of wildland fire for fuel management. The assertion is that the early European-Australian understanding of Aboriginal fire management accurately portrays management that was effective in preventing extreme fire behaviour and that because that management has been in practice for many millennia it is also the management under which the full diversity of Australian species have survived. If such a claim is true, it is possible that any further work on fuel management is purely academic. It suggests that disastrous fires are preventable; that the answers are known and all that is required is the will to act. This thesis disputes the claim, asserting that the popular understanding of traditional fuel management is inadequate for modern usage and that there are complexities in indigenous fire management that are not captured by simple generalisations applied across the country such as those quoted above. It is argued that the inadequacies in understanding arose because the scope of most observations did not allow for the collection of sufficient detail, and that such detail was frequently not sought because the perception of Aboriginal people by the European observers was affected by cultural paradigms. It is concluded that as a result, the claim has little to offer modern fire management because it is grounded in the same cultural paradigms, and that further scientific input is required to fill the gaps. 1.11 Knowledge transfer in indigenous society Due to its lack of a written language, by definition Aboriginal knowledge transfer arises from a ‘primary orality’ (Ong 2002). Because of this, much of the information that is presented 19 here on the culture has been taken from conversations with Bemerangal Storyteller Rod Mason rather than from written sources. Western thinking has a long history with the written word, and western minds are accustomed to thinking in this way. Learning from the written word has the convenience that it can be ignored or put away; oral learning has a compulsion to it in that if the words are ignored, they are missed (Ong 2002). Consequently, oral tradition dictates a respect for the teacher; a Master and Student or Rabbi and Disciple relationship must be established. Put simply, unless the listener is prepared to humble themselves into student role, they will not learn the relevant message. Another characteristic of primary orality is its use of story, song and tradition as mnemonics or memory tools to convey information (Ong 2002). Rather than teaching students the details of fire, weather and survival explicitly, these central tenets were captured in numerous novel approaches. Non-Aboriginal listeners often receive such information as “Dreamtime” stories, however many traditional storytellers reject the name as a poor substitute for traditional names such as burbung (Hartland 1898), tjurkupa or darama on the basis that people automatically think of the Dreamtime as another time long ago (Pers. Comms. R Mason 2009). Inherent within darama is warru – the specific lore explaining how to live in country. In the Ngarragu country of south-eastern New South Wales, the warru for the Burrungubudgee hunting grounds in the Tidbillaga high country is different than it is for the Wadbillaga ribbon gum country or the open Nullaga grasslands (Pers. Comms. Mason 20082009), so the only way to know the right way to live in those places is by learning the warru through listening to and participating in darama. In Young’s (2005) record of the Monaro Ngarragu story of the brown snake and the turtle for instance, the story does not simply describe the origins of the brown snake’s poison. The ritual of the story involves drawing the path of the snake in the dust along with the paths of the other animals. Although listeners may focus on the interesting and sometimes humorous or frightening aspects of the story, through the simple act of listening and watching the map being drawn, they learn from this and other stories the lay of the mountains in the area, the places where various important species may be found and important travel routes and boundaries (Pers. Comms. R. Mason 2008, A. Williams 2010). Another form of mnemonic integral to darama is the place of bagal or kinship, which has been erroneously referred to as totem. Bagal means that at birth, all people are named for a 20 particular plant or animal and given responsibility for that species. Their association with the species automatically connects them with one of the three central spirits in Ngarragu theology – wind, rain or fire (Pers. Comms. R. Mason 2005- 2009). This has several effects. By association with one spirit you are excluded from the warru of the other two, so knowledge is divided and essentials to survival are only possible within a cohesive community structure where there is mutual support by those with skills in different areas. The connection with the species itself provides a further specialisation, as each species has its own requirements and serves as an indicator of other things. Maliyan or Meerung (eagle) people for instance will have a closer association with areas that are recognised as Meerung country. Meerung are associated with the Fire Spirit, so there are laws regarding fire in these areas and travellers that see the signs of a place being Meerung country are immediately reminded that their own knowledge may be insufficient in this area. Consequently, they will refer to someone with a more appropriate bagal. The converse is also true – Meerung people could not enter other country and bring their fire; other country had its own warru. The reference to the spirits wind, rain and fire introduces the next aspect of darama – the inseparability of religion and daily life. Just as bagal divides responsibility, encourages specialisation in different fields and provides indicators in the wildlife; religion gives meaning to bagal, interprets the landscape and provides motivation for appropriate actions. An illustration of the way religion, bagal and other forms of darama combine to produce a highly specific use of fire is given by a Bemerangal tradition in part of the Wadbillaga country (provided by Pers. Comms. R Mason 2005). Many of the Ngarragu clans including the Bemerangal moved to and fro annually between coastal areas and the high mountains of Tidbillaga (Kosciuszko area). One of the first campsites once they had gained the Monaro plateau was in the Wadbillaga country at the top of the escarpment. This was an important camp as it was situated near the base of a mountain named for the rainmaker Djillagamberra, who presided over the country. As a result the Bemerangal families stayed at the site for some time. The camp was also visited on the return journey during the autumn, but in this case travellers brought with them fruits from several of the Geebung (Persoonia) and other food plants of the Tidbillaga area as stores for the crossing of the open Nullaga grasslands. This practice 21 over centuries or millennia has meant the movement of large quantities of seed across a natural grassland barrier into a new environment, and Mason points out that the relevant food plants are concentrated around the campsites or ‘gardens’. The duration of the spring visit to Wadbillaga was determined by the women through bagal ties that connected them to the bipinnate wattles or Matruk of the area. When the women announced that they were satisfied the flowers had fallen from the Matruk, the time to move on was close. The actual day was then indicated by Djillagamberra, shown by clouds resting on the top of the mountain. On the morning of these spring days, the Bemerangal broke camp and walked to the top of the hill looking back down onto their gardens. “They sing out on top of the hill up there, and they said ‘hey, Boodjeree Djillagamu’, and they’d chuck a fire-stick into the bush and burn it, then they’d go. And next year when they’d come back, they’d sing out ‘hey, Boodjeree Djillagamu’, and they’d come back into the area.” The story is quite simple to the casual listener, but it is the simplicity of the tradition that gives it strength as a mnemonic. In one sense all they did was to throw a fire stick into the bushes and walk away, but the role of bagal, religion and specifics in the landscape gave a precision to the ritual that would be very difficult to achieve without them. A. The use of fire ensures the regeneration of important food plants localised to the site. Introducing fire was an act of gratitude to Djillagamberra, indicated by the words “Boodjeree Djillagamu”, which translate roughly as “Djillagamberra’s country is good.” Fire was seen as a gift from the country to the people, so throwing a fire stick back into the country is the giving of a gift in turn. B. The burn was conducted only at a certain time of the year, based on the flowering patterns of the local species and therefore their relative vulnerability to fire. This time was determined through the bagal relationships. C. Only forest with wattles mature enough to flower was burnt, defining a maximum fire frequency for each patch. Bagal also identified the areas. D. The location of the cloud base on top of the mountain defined an exact dew point. The prescription was given by the fact that Djillagamberra was the rainmaker, so the association of his mountain with clouds identified the timing. 22 E. The ignition pattern was specifically a single point on a hill top, dictated by the tradition itself. Given the conditions, the result would have been a very low intensity fire burning downhill, picking out the rare patches that were dry enough to carry fire on such a cool, humid morning and effectively treating a very small area of vegetation before it self-extinguished. If the prescription had been for a drier time, the fire would have burnt the entire area, removing the food plants and making the site unsuitable as a camp ground on the return journey. Although the single fire stick was the tool in the Wadbillaga; warru for other places along the journey dictated other practices. When crossing the Nullaga grasslands for instance, the timing of late spring was linked to the first of the three great spirits wind, rain and fire. In recognition of the Wind Spirit, fires were lit in the grasslands as the strong spring frontal winds were coming through (Pers. Comms. R Mason 2009). In the current context of southeastern Australia, this is the classic “bad” fire weather as it produces high intensity fires before a 90 degree wind change in the lee of the front turns the long fire flank into a front. Such situations have been the cause of great loss of life in Victoria during the Ash Wednesday and Black Saturday fires (Haynes et al 2008, Taylor 2009). The risk taken by the Bemerangal was not the same however, as the direction of travel by the people was against the wind and the fires blew backward in their wake. The wind changes that followed the passage of the spring fronts caused no safety issues as there were no people on the flanks of the fire attempting to control them and running the risk of being overrun; rather the change in spring frequently brought rain or snow and at the very least, cold temperatures and high humidity. The Monaro grasslands that form Nullaga are predominantly composed of tussock-dominated groups such as the Themeda triandra - Poa sieberiana – Crysocephalum apiculatum – Acaena ovina community (Benson 1994). In such discontinuous fuels, fire will only spread effectively if wind speeds are greater than a certain threshold determined by the size of the clumps and their spacing (Gill 1990, Burrows and van Didden 1991, Bradstock and Gill 1993). This arrangement necessitates the prescription of wind, but as mentioned, the following change in the weather placed limits on the size of the fires. Unlike the forest fires of Ash Wednesday and Black Friday, the fires were also burning in grassy fuels rather than 23 forests, so the finer nature of the fuels and the lack of coarse woody material together with the greener spring growth of inter-tussock grasses and forbs meant that changes in humidity and temperature probably had very rapid effects of fire extinction, with minimal chance of reignition. Benson (1994) notes that grassland diversity in this area is strongly influenced by the size of the tussocks; if tussocks become too large the canopy becomes closed and the inter-tussock forbs and smaller grasses begin to decline. Conversely, loss of the tussocks removes their protective influence from the wind and negates their role as lookout points for small hunting reptiles such as the Grassland Earless Dragon Tympanocryptis lineata pinguicolla (Smith 1994). Wind-driven fires produce long, narrow burn patterns and thereby allow diversity in the age-classes of the grassland. Examples from other areas Langton (1999) provides a record from Arnhem Land Elder Yibarbuk: “But there is one kind of burning which is men’s business alone – and it is dangerous work. This is the fire drive mainly for macropods (the larger ridge dwelling species like kalkberd, djukerre, kandakidj, karndayh) rather than the agile wallabies (kornobolo) which favour monsoon jungle and thicker forest. Emus (wurrbbarn or ngurrurdu) may also be a target for this specialised hunting technique. “…When the most senior landowner [sic] from the area where the fire drive (kunj ken manwurrk = fire for kangaroo) is to be held sees that the time is getting close, he will talk with his senior djunkay. They sit down and discuss how the djunkay will organise the drive – where it will be held, when it will be held (expressed by reference to floral seasonal indicators and moon phases) and who will be invited.” This quote demonstrates consideration of the following factors for a routine fire drive: 1. Suitability of those conducting the burn. Not everyone takes part in such a burn; those involved have been hand picked. 2. Chain of command. This fire requires the work of the senior djunkay under the authority of the most senior landowner 24 3. Consideration as to whether the vegetation community should be burnt. “Monsoon jungle and thicker forest” were not burnt to capture kornobolo, other methods than fire were preferred 4. Consideration as to where the burn should happen. There was no sudden decision resulting in burns taking place across a broad landscape. The reasons for this are not given but may involve the locations of other camps, the distribution of wildlife and threats to areas they did not want to burn 5. Consideration as to the timing of the burn. Environmental indicators were used to specify the best time for the burn. Once again, the interaction of kinship with different species, signs in the weather and landscape and ritual produce very specific outcomes. Practicing the right lore for country is central to Aboriginal management across Australia. Although the names vary depending on the language, the tjurkupa or darama for that country defines where and how fire should be used. In some areas this demanded complete fire exclusion. Jones (1975) described the role of religion in fire exclusion for some parts of northern Australia. In one area of jungle, it was taught that there were beings living in the forest that would blind someone that burnt it. Similarly, Gould (1971) described an encounter with some men who were deeply distressed upon finding that part of a totemic trail in South Australia had been burnt, and that there would be serious punishment for whoever burnt the area. When discussing the ularaga or fire lore of the Wongkonguru people of Lake Eyre, Elkin (1938) identified the role of ritual in defining the way fire was used. Like the warru of the Ngarragu, the ularaga contains a set of prescriptions that must be followed for even the simplest use of fire, these include particular songs and rites to be performed when fire is being generated by ‘twirling’ or even carried on the fire stick. Only one initiated to the ularaga had the right to deal with fire. As Elkin explained: “…the myth and rite do not merely provide an explanation for the making and appearance of fire, as though mere friction were not sufficient of itself; they also 25 witness to its great social importance and to the necessity of guarding it most carefully, on the one hand against its careless use, lest bushfires should be caused, and on the other hand, against loss.” The way in which fire itself was managed (and still is in some areas of the country) demonstrates a keen understanding of fire behaviour. Finlayson (1936) described an expedition in which he employed the services of a local Aboriginal group to assist him in capturing wildlife. In one instance, animals were flushed from their hiding places by lighting fire in an arc surrounding the area, such that the direction of the wind and the convective activity of the fire caused it to draw inward around the target rather than spreading. Jones (1975) in his observation of Gidgingali fire management in the Northern Territory said: “People used their fires accurately, aiming them into a natural break such as an old fire scar or swamp, timing the fire so that predictable wind changes later in the day would blow them back onto their own track, or so that the evening dew would dampen them down.” Thompson (1949) also explained: “This is not a random business, it is well organised, and is carried out by the men as a communal enterprise, although in a restricted and controlled manner with ‘drives’ for kangaroos, wallabies and other game. The actual burning of grass is directed by the old men of the clan, or by others who have an [sic] hereditary right.” The purpose of these examples is to illustrate the complexity of Aboriginal fire management and the fact that it is not simply a matter of “burning the bush”. They also go some small way to explaining the deep connection between indigenous Australians and their country. If the Bemerangal people for example were taken from their land and given entirely different country, their knowledge necessary for survival would have been taken from them because they would no longer have had Djillagamberra Mountain to identify the correct dew point for ignition. The Nullaga grasslands would not lie along the path of their travels to be crossed in late spring when the time was right for burning. In fact all of the strength of their darama would be negated because they would not know the landscape, the stories or the songs that taught them how to live in it. Mason (2003) says: 26 “For non-Aboriginal people, the most difficult challenge in appreciating Aboriginal knowledge and use of fire is that it is intimately linked with some of the most central and important tenets of Aboriginal lore, common throughout Australia. Unless an individual passes through the various stages of learning, the detail of these spiritual beliefs is not accessible.” 1.12 The European-Australian approach to Aboriginal knowledge Given this context, if the hypothesis is to be supported that as a generalisation (rather than in specific exceptions) rural European-Australian traditional knowledge of fire has its foundations in Aboriginal fire management, it is necessary to demonstrate that: 1. The overall relationship of European settlers, graziers and others having first contact with the Aboriginal nations was one of sufficient respect that those moving into the country were prepared to take the role of student and persist in their learning from the local elders. 2. The tjurkupa or darama of the area was appreciated and understood by those settlers. A full examination of the history of interactions between Aboriginal Australians and the European settlers will not be conducted here as it is beyond the scope of a study on fire behaviour and management; however the literature documenting the dispossession of Aboriginal people from their land and culture is both extensive and conclusive. The rate at which indigenous nations were either driven from their land or killed (e.g. Pepper and De Araugo 1985, Harris 1990, Young 2005) was not indicative of a culture of respect. The destructive approach toward Aboriginal people and their knowledge was compounded in the nineteenth century by the rise of Social Darwinism in the scientific community and its assertion that indigenous Australians were sub-human (Galton 1864-5, 1869). While the violence and overt persecution of the people may have come from the broad populace and its push for new land, the intellectual community was specifically inclined toward dismissal of Aboriginal knowledge. Even much-acclaimed anthropologists such as Howitt appear to have at times not recognised the significance of the knowledge transfer mechanisms inherent in the 27 oral tradition and dismissed them as a lack of intelligence: “…they have the minds of children and the bodies of adults” (Howitt 1869). By the twentieth century, this approach had become institutionalised and enshrined in government policy, with the express intent of eliminating indigenous identity, culture and oral tradition (Lavarch 1997). Although there were notable exceptions to this part of Australian history, this was nevertheless the pervading culture of the time and it is in this context that we must view the claims made in recent years that rural European Australia learnt its fire management from Aboriginal Australia. It is also significant that the front line of conflict was the conquest of land and the replacement of kangaroos, bustard and other staple foods of the Australians with the sheep and cattle of the settlers. In this conflict, it appears that the graziers more often played the role of soldiers and hunters than willing students to the knowledge of the Aboriginal nations. In regard to the first test then, despite the individuals in Australian history who stood out from their time due to their respectful relationship with the Aboriginal nations, the net effect of the period was the decimation of Aboriginal culture and, all too often, the people themselves. Even where intentions may have been at least humane, the displacement of Aboriginal people from their land directly or by the introduction of exotic species and shooting of native competition, the erecting of fences or the forbidding of fire such as the Nullaga fires which would have negatively impacted stock and fences; the overall effect was the estrangement of the people from their land, their darama and thereby their knowledge. In the second case, there is little evidence that early European Australians comprehended even the basics of the Aboriginal world view. Once again, there were notable exceptions; however it appears from the historical record that the majority of early European Australians mistook primary orality to be unintelligence. In a well known example of the way indigenous fire was interpreted by European settlers, Townsend in March 1846 (Clarke 1860) recorded the following observation: “The Blacks had visited the Snowy Mountains, a short time previously to us, for the purpose of getting ‘Bogongs,’ a species of moth, about an inch long, of which they are 28 particularly fond; to obtain them they light large fires, and the consequence was, the country throughout the whole survey was burnt…” Note that there was no claim to have seen the fires lit, but only that the area had been burnt before the arrival of Townsend and his party. Townsend assumed at the time that the fire was a result of Aboriginal burning, and readily believed that multitudes travelled great distances into this rugged country, ignited large bushfires every year at the height of summer to catch moths, and then set up camp amongst the bushfires. This assumption can be tested by comparison with an eyewitness account of moth collection provided by Helms (1895): “Their method of catching the insects was both simple and effective. With a burning or smouldering bush in the hand the rents in the rocks were entered as far as possible, when the heat and smoke would stifle the thickly congregated moths, that occupied nearly every crack, and make them tumble to the bottom of the cleft. Here an outstretched Kangaroo skin or a fine net made of Kurrajong fibre would receive most of the stupefied and half-singed insects, which were then roasted on hot ashes.” In the eyewitness account, the use of fire is proportional to the need, carefully controlled and makes rational sense. This contrasts with Townsend’s speculation, and provides a far more plausible picture. With this context in mind, consider how a European migrant, believing themself more intelligent than indigenous Australians might interpret the explanation of the Wadbillaga ritual by a Bemerangal Fire Man. When the question was initially asked, the response would have been a simple answer, informing the enquirer that they burnt the area every year for instance. The simple answer would have been dictated by the consideration of the Storyteller – in a primary orality, information is passed along by persistent education over time. If it is all given in one instance it will be forgotten. The teacher assumes that the student will return to learn more if they are sincere about learning. Would the migrant have understood this and returned, or would they, believing that they had their answer or perhaps believing the teacher to be unintelligent have assumed that they now understood fire management for the Wadbillaga? How would this “knowledge” translate into modern management for the area? 29 Attempts at annual block burning perhaps? Without the benefit of the prescription, would the blocks be burnt in hot, dry or windy weather to facilitate spread in one-year-old fuels? Without the additional information provided by the Warru for each place, it is quite possible that a partial knowledge may actually produce the opposite result to that intended by the Aboriginal custodians. It is apparent from this that the popular European-Australian understanding of Aboriginal fire does not accurately translate the oral tradition into practical management. Critical information was lost from the tradition because the relevant detail was beyond the scope of casual observations or because it was not looked for or understood due to the cultural paradigms of the period. The combination of this inability to comprehend the complexities of Aboriginal culture in the 19th century combined with the 20th century attempts to systematically eliminate it from the country has meant that the majority of modern Australians are largely unaware of or unable to relate to it. In this light, it is proposed that a simple assumption that Aboriginal burning was indiscriminate, frequent, and without planning or prescription is a social myth more representative of the racially ignorant culture of early Australia than it is of Aboriginal culture. Aboriginal burning was (and is in areas where the culture is still well intact) a targeted, controlled practice with different approaches for different areas. An outcome of such an approach which provides different management for specific sites is that less area would be burnt overall than might be achived today with for example the use of aerial ignition across large areas. The available evidence on the subject suggests that this is indeed the case, as a sharp increase in fire frequency is evident with European settlement, particularly in south eastern Australia where the majority of evidence is available (Pryor 1939, Raeder-Roitzsch and Phillips 1958, Banks 1982, 1986, 1989, Pulsford 1991, Richards et al 2001, Kershaw et al 2002, Mooney et al 2010). “Today fire is not being well looked after. Some people, especially younger people who don’t know better or who don’t care, sometimes just chuck matches anywhere without thinking of the law and culture of respect that we have for fire.” Yibarbuk, quoted by Langton (1999) 30 1.2 The current state of fire behaviour modelling in Australia Understanding flammability and its management is synonymous with understanding fuels and fire behaviour; however the science in this area in Australia is dominated by community specific empirical models that have not been peer reviewed. A large proportion of peer-reviewed fire behaviour models are not designed or suitable for operational use as they are either focused on addressing a particular theoretical issue or they are computation intensive and cannot operate at a speed useful in an incident management setting (Sullivan 2009a,b). Operational models for many priority areas have been developed (Table 1.1), however the coverage of priority fuel types in Australia is limited, confidence in some of the more universally applicable models such as Rothermel (1972) is low (Burrows 1991b) and in practice, fire management is underpinned by material that has not been peerreviewed and/or is not applicable to the location (Table 1.2). In NSW for instance, the Rural Fire Service standard for bushfire risk assessment makes use of four separate fire behaviour models for different fuel environments (Tan et al 2007). These are grassland (McArthur 1966a), forest (McArthur 1967), alpine (Marsden-Smedley and Catchpole 1995b) and heathland (Catchpole et al1998). All models are empirical, applied well beyond the range of conditions within which they were developed and only MarsdenSmedley and Catchpole 1995b) has been peer-reviewed. This model however was developed specifically for Tasmanian Button Grass (Gymnoschoenus sphaerocephalus) moorlands, rather than the heath, herbfield and feldmark environments found in alpine areas. Table. 1.1. Peer reviewed fire behaviour models designed for operational use Author Type Fuel type Rothermel (1972) Quasi-physical Various Van Wagner (1976) Empirical Conifer forest crown Validation fire Albini (1985, 1986, 1996) Quasi-physical Various Partial; ROS 14% max error in single stratum (1986), 350% max error in crown fire (1996), Fh 31 none Cheney et al (1988) Empirical Grassland Xanthopoulos (1990) Quasi-physical Conifer forest crown fire initiation Burrows et al (1991) Empirical Spinifex Burrows (1994, Empirical Jarrah forest 1999a,b) Marsden-Smedley and Under predicted ROS under severe conditions Empirical Buttongrass moorland Empirical Mallee Heath Catchpole (1995b) McCaw (1997) “Good agreement” (Sullivan 2009b) in ROS for slower fires, over-predicted an extreme fire by 30% Fernandes (2001) Empirical Shrub ROS R2 0.91 with data, “close agreement” with other datasets (Sullivan 2009b) Beaza et al (2002) Empirical Gorse heath Only intended for prescribed burns Alexander (1998) Quasi-physical Conifer forest crown fire Bilgili and Saglam Empirical Maquis Quasi-physical Various (2003) Vaz et al (2004) Partial; ROS 46% max error, Fh poor Cruz et al (2006) Tanskanen et al (2007) Conifer forest Empirical Pinus sylvestris and Picea abies Table. 1.2. Non Peer reviewed fire behaviour models designed for operational use Author Type Fuel type McArthur (1962) Empirical Dry Sclerophyll Validation forest McArthur (1966) Empirical Grassland McArthur (1967) Empirical Dry Sclerophyll forest Sneeuwjagt and Peet Empirical Jarrah forest Empirical Coniferous, (1985) Forestry Canada Fire 32 Danger Group (1992) deciduous and mixedwood forests, logging slash and grasslands Catchpole et al (1998) Empirical Heathland Performed “reasonably well” for ROS (Catchpole et al 1999) Fernandes et al (2002) Empirical Maritime Pine Gould et al (2007a) Empirical Jarrah forest ROS gives values mostly within ± 25 % error margins of selected wildfires using assumption of no slope effect; Fh has R2 of 0.81 with experimental data but no outside validation 1.21 Identifying priority fuel arrays There are many factors involved in identifying the priority areas for understanding fire behaviour and fuel dynamics. Considerations range from biodiversity and ecosystem function through to utility value such as catchment water yield and quality, built structures and human life. Loss of human life from bushfires in Australia is minimal when compared to other causes; from the years 1900 to 2007/08 for instance there were 105 fatalities from bushfires in New South Wales (Haynes et al 2008). This equates to an average of 0.98 deaths per annum in that state. By contrast, 2262 people died from coronary heart disease in NSW in 2008 alone (NSW Health 2009). Of the 552 deaths from wildfire that have happened in Australia between 1900 and 2007/08 (Haynes et al 2008) however, 63% occurred in only 10 of those fire seasons. Including the 173 deaths from the February 2009 “Black Saturday” fires (Taylor 2009), causes of death can be attributed to forest fires, grass fires, heath fires and others by generalising as to the predominant type of fire in each of the 11 seasons. The results (table 1.3) are not precise as many fires burnt across multiple fuel types, but the results do suggest that the majority of fatalities (82%) have been caused by forest fires and provide an informative approximation for the purposes of prioritisation. 33 Table 1.3. Most fatal fire years in Australia, generalised according to the dominant fuel type burning. Statistics for the number of fatalities each year taken from Haynes et al (2008) and Taylor (2009), fuel type as listed. Year 1926 1939 1944 1952 1967 1969 1983 2003 2005 2006 2009 TOTAL FUEL TYPE Forest Grassland 39 79 46 20 64 21 60 6 9 5 173 426 96 State VIC VIC VIC VIC TAS VIC VIC ACT/VIC SA VIC VIC Source Haynes et al (2008) Zylstra (2006) Zylstra (2006) DSE (2009), uncertain Haynes et al (2008) Haynes et al (2008) Haynes et al (2008) Zylstra (2006) Haynes et al (2008) Haynes et al (2008) Taylor (2009) 1.22 Australian fire behaviour modelling for priority fuel arrays In this context, it is notable that for Australian fuel complexes there is one peer reviewed model each for grassland, spinifex, and Mallee heath, but the one peer-reviewed model for forest (Burrows 1994, 1999a,b) was developed specifically for Jarrah (Eucalyptus marginata) which does not occur in Victoria, Tasmania or South Australia where the fatalities have occurred. For these areas, both fuel management and fire behaviour prediction have been modelled predominantly using sources that have not been peer-reviewed – notably McArthur (1967) and more recently with Gould et al (2007a). McArthur is believed to have constructed his fire behaviour model on data taken from over 800 fires (Noble et al 1980), although this data was never presented and his statistical analysis has not been verified. The majority of these fires are believed to have been carried out on Black Mountain in the ACT (AM Gill Pers. Comm. 2009), with the exception of a number of fires which were conducted in Jarrah forest. Nine of these are shown on a graph in his report used to illustrate his principle that doubling the “fuel load” (dead surface fuels) will double the rate of spread. Since that time however, the peer-reviewed studies of Burrows (1994, 1999a,b) in that same fuel complex established that surface fuels had no influence on headfire rate of spread. Yet this focus on removal of surface fuels remains a dominant paradigm in fuel management across Australia (e.g. RFS 2007, Ku-ring-gai 2008, McCaw et 34 al 2008, Melville 2008). McArthur also addresses weather parameters by combining all into the one index so that a cool, windy day produces the same modelled fire behaviour as a hot, still day. Although this index may have value for determining general levels of fire danger, it remains to be shown that all such weather factors can be combined in this way to meaningfully predict fire behaviour. In recognition of some of these limitations, Gould et al (2007a) conducted 99 experimental burns in Jarrah forest, and although the model was never submitted for peer-review a proportion of the data and some statistics were presented in a more significant report. The study confirmed Burrow’s findings that surface fuels had little influence on rate of spread, and examined the changes in fire behaviour with time since fire for two sites with different understoreys. The study did not investigate the effect of slope on fire, but accepted the exponential relationship proposed by McArthur (1967). This was also limited by assuming that the downhill rate of spread was equal to rate of spread on flat ground (Gould et al 2007b), and that slopes greater than 30o did not increase the rate of spread any further (Gould et al 2007a p93). The model also separated the roles of the various weather parameters rather than combining all into an index. The report concluded (pp. iv-v) that: • Near surface fuels primarily determine rates of spread • The height of the shrub layer along with the rate of spread determines the flame height • Hazard reduction by prescribed burning will reduce the rate of spread, flame height, intensity, spotting distance and firebrand intensity of future fires • After near surface fuels and shrubs have developed to maturity, ongoing increases in loose bark will continue to increase the difficulty of fire suppression • “Stimulation of understorey shrub regeneration after burning will not increase the rate of spread of a fire until such time as a significant near-surface fuel layer accumulates.” • Younger fuels produce fewer firebrands. 35 Although some of these conclusions are intuitive, a number of them are not supported by data or statistical analyses, and due to the prominence of this model and its influence on the Australian understanding of fuel management, the most significant discrepancies are examined below. 1) The elevated fuel ‘hazard score’ had the strongest correlation with rate of spread (P = 0.57), although near surface fuels were almost as high (P = 0.56, p199). Although it was stated that elevated and near surface fuels were correlated, no explanation was given as to why near surface fuels (predominantly suspended leaf litter in the studies) were chosen for the model rather than elevated fuels. This is significant as suspended leaf material requires the presence of elevated fuels in which to be suspended, so the functional change in the fuel array is the growth of shrubs and the concept that dead suspended litter affects rate of spread has therefore only been hypothesised, not investigated. The role of shrubs in determining rates of spread appears to have been prematurely dismissed. 2) Rate of spread continued to increase with time since fire at one site (McCorkhill), but at the other site (Dee Vee) rate of spread increased until six years had passed, then decreased (McCaw et al 2008), although the actual values are not given in the report. The report also does not give flame heights with time since fire, however shrub height increased at McCorkhill and decreased after six years and as the flame height model shows a positive relationship between flame height, and rate of spread and shrub height, this suggests that flame height also followed this pattern. This data from the report indicates that in one of the two sites rate of spread and flame height increased until it reached equilibrium, but decreased at the other site. 3) Despite a slower rate of spread and smaller flames following the onset of shrub senescence at six years, fire intensity was separately reported as continuing to increase (McCaw et al 2008). This report was based not on measured values of intensity, but on approximations using Byram’s (1959) fire intensity and surface fuel load – a fuel shown to have very little relationship to fire behaviour (p. 199 and Burrows 1994, 1999a,b). Byram’s calculation also assumes complete combustion of the included weight of fuels at a rate equal to the fire’s rate of spread, however the infire camera evidence produced by the study revealed that the fire front moved ahead of the surface fire, lighting the surface fuels from the top so that these burnt in a broad area of swirling flames after the front had passed. Those fuels burning after the 36 passage of the fire front should therefore not be added to the intensity calculations as the energy released from these does not add to that of the fire front. As the approximated intensity gives opposite results to the physical fire behaviour, it appears that the approximation does not adequately reflect the observed fire behaviour, 4) The report does not provide any data to support the statement that spotting distance increases with time since fire, but the model for spotting distance uses rate of spread with surface and bark fuels to estimate spotting distance and both fuels were shown to increase with time since fire (pp27, 30). The model presented in the report was not developed from data gathered by the experiments however, but was an alteration of Ellis (2000) relying on multiple unsupported assumptions (pp131-133) and providing no data in support of the alteration. The original model (p118) taken in conjunction with the data on shrub heights over time and the model connecting this to flame heights suggests that spotting distance at McCorkhill will increase with age, but at Dee Vee will increase up until six years and decrease thereafter. The altered model however predicts that spotting at Dee Vee will continue to increase with time since fire. The data provided shows that the longest distance spotting in the experiments occurred from five year-old fuels (p131), although this may have been due to weather conditions that were not specified. As no data were provided to support the claim that spotting distance increased with age and as the spotting model only predicted this after alteration and without evidence or support for the assumptions; the only evidence presented in the report shows that spotting distance increases with age at one site but decreases after approximately six years at the other. 5) The results on firebrand intensity were provided for two pairs of fires burnt in five and 22 year-old old fuels. In one set of results, firebrand intensity was greater in the older fuels, but in the other set firebrand intensity was greater in the younger fuels (pp124-125). These results do not provide a consistent argument that firebrand intensity increases with time since fire. As however the bark ‘hazard score’ was shown to increase with age (p30), it is a reasonable hypothesis that there will be more bark firebrands available for spotting as the fuels age. 6) No sites were studied where stimulation of shrub growth by fire was a factor, so it is not clear how the study refuted this observation. It is true that shrubs do not form an input to the rate of spread equation, however as explained at point one, the decision not to use it was not adequately justified. There is no support for the statement that “stimulation of understorey shrub regeneration after burning will not increase the rate 37 of spread of a fire until such time as a significant near-surface fuel layer accumulates”. In fact, rates of spread increased in both sites as shrub heights increased, then decreased at the Dee Vee site as shrub heights decreased even though near surface fuels continued to increase. These discrepancies are significant in that their implications for fuel management have not been adequately interpreted. The two sites McCorkhill and Dee Vee had opposite relationships to time since fire; the McCorkhill site continued to support faster fires with larger flames the longer it was left unburnt; reaching a state of equilibrium after approximately 10 years. In contrast, the Dee Vee site reached a maximum at around six years and declined thereafter. This evidence suggests that broad-scale prescribed burning while effective at the McCorkhill site is not effective or is counter-productive fuel management at the Dee Vee site unless it is used very frequently. Although both sites were Jarrah forest, the Dee Vee site lacks an understorey of shrub species which continue to grow after six years. The report suggests that the two species present at McCorkhill and not Dee Vee were the Fine Ti-tree (Taxandra parviceps) and Persoonia elliptica (pp143-144). A valid conclusion from the data in the report is that no generalisation can be made as to the efficacy of prescribed burning, but that depending on the dynamics of the component vegetation, this practice may be helpful in some environments but counter-productive in others. Validation Gould et al (2007a) provided validation for their rate of spread model; however despite the impressive coverage of fires the reliability of the validation was reduced because of two factors: 1) Slope was not included in the calculations 2) Predictions were only compared to fires that were known to have spread at high intensity, rather than fires that may have been expected to spread at high intensity but did not. The removal of slope from fire behaviour calculations was presumably based on a statement by McArthur (1967), where he said that slope could be ignored on high intensity fires after the first 30 minutes or so as the effect of spotting from ridge to ridge would bypass the downhill components of the landscape (p13). This assertion has mathematical problems, as 38 removal of the negative slopes leaves only the positive slopes, so that generalisation should actually result in a positive slope to the entire fire area rather than a zero value. If the assumption is accepted that fire spread rates on negative slopes are the same as for flat ground, then even without spotting, all slopes will either be neutral or positive so that the final result is still an overall (albeit reduced) positive value. The same confusion is evident in case studies used for validation. McCaw et al (2009) in their validation of McArthur (1967) and Gould et al (2007a) against the Black Saturday fires of 2009 for instance state that “In forest fuels, the process of spotting (fire brands being carried ahead of the main fire that start new fires) allows the fire to spread rapidly across the topography and overcomes the retarding effect of a negative (in the direction of the wind) slope” (p7). Having said that however, the report later states that “Predicted rates of spread were not corrected for slope on the basis that the fires traversed both positive and negative slopes during their run” (p55). The exclusion of slope calculations in the validation studies greatly alters the outcomes as the effect of the slope equation used in both models is to double the rate of spread for every doubling of the slope. The second consideration is important from the perspective of the incident manager. If a fire is expected to spread quickly because of perceived ‘bad’ fire weather, suppression efforts will be more conservative, falling back to less aggressive forms of attack or into property defence. If the fire does not spread with the expected intensity however, an opportunity is lost for gaining better control of the fire. Such incidents can produce a “cry wolf” syndrome, where fire managers have reduced confidence in the value of the predictions. A model that predicts high rates of spread when they do not occur may appear correct on those days when fire behaviour is extreme, but inclusion of the other days will greatly increase the average error of the model. These factors highlight the need for validation attempts to be tied to operationally meaningful outcomes. Ultimately, fire behaviour predictions are used to inform decisions around the placement of fire crews and the nature of suppression activities during an incident, so the level of accuracy provided by any model should ultimately reflect the number of effective decisions it produces. 39 Summary Australian fire behaviour modelling for south-eastern forests where fire has historically posed the greatest threat is limited by a number of simplifications and assumptions that have not been exposed to peer-review. In particular, McArthur (1967) has been influential in constructing a view of fuel management that runs contrary to the peer-reviewed literature on the topic. Gould et al (2007a) provides two major improvements to McArthur (1967) by delineating some differences in the roles of different fuel strata and by separating the influences of the weather parameters considered. Despite this, the value of the model as a tool for fire and fuel management is limited by a number of unsupported simplifications. The data collected has potential to inform managers as to where the introduction or exclusion of fire might be strategically valuable in Jarrah forests, but this value is lost due to the way the data is interpreted and combined. The attempt is made via validation to extend these generalisations to other fuel arrays including the high priority south-eastern Australian forests, however it is impossible to say whether this is successful or not due to the approach taken. The study did however demonstrate a case for the influence of shrub dynamics on flame height, a factor which although reduced by the lack of shrub inclusion in the rate of spread model should still be considered in fuel management operations where reduced flame heights are advantageous. 1.23 Existing alternatives for Australia Neither of the peer-reviewed forest fire behaviour models (Rothermel 1972 and Albini 1985, 1986, 1996) have been widely adopted for use in Australian fuel arrays. Gould et al (2007a) stated that the “practice of building artificial fuel models to adjust parameters so that a reasonable fit to field data could be made was rejected in Australia because it assumed that the relationships established in the wind tunnel could be transferred directly to field fires, and it did not predict the changes that occur by manipulation of the fuel bed, say by prescribed burning.” No systematic validation of either model has occurred in Australian forests, although Catchpole (1987) reported that Rothermel (1975) was the most appropriate model for use in heathland fuel and fire behaviour modelling, and Burrows (1999b) reported that the performance of Rothermel in Jarrah forest was poor. Until these two models have been properly examined and validated in Australian fuels, it is not reasonable to dismiss them on the grounds that they are not expected to be accurate. Both 40 models however share the characteristic that they model all fuels as a ‘bed’, a single contiguous stratum without information on the fuel arrangement. Burgan and Rothermel (1984) suggested a mechanism for modelling strata separately, although this does not appear to have continued in use and no longer forms part of the BEHAVE system of modelling (Andrews et al 2008). No systematic validation of the approach is apparent (although Catchpole 1987 did test the approach in two types of heathland), and it remains limited by the underlying assumptions about the beds. Sandberg et al (2007) introduced multipliers to account for beds that were heterogeneous, which combined with the multi-stratum model may go some way toward addressing the issues. An important factor lacking in the single stratum approach is the ability to differentiate between the effects of live and dead fuels. Whereas dead surface fuels can have high moisture contents following rain events or due to factors such as high humidity, low temperature or limited solar radiation (Matthews 2006), plant moisture may be affected by other factors that do not have such quick response times. Neither model presents evidence to support the assumption that moisture can be adequately represented by a simple average of the two values, therefore both models are limited in their ability to model fire behaviour when for example live fuel moisture is unusually high or low. 1.24 Summary The dominant fuels that require understanding for the protection of human life in Australia are the forests of the south-eastern areas, where approximately 82% of deaths since 1900 from wildland fires have occurred. Although a number of peer-reviewed models have been developed for other fuel complexes, this priority area has been represented by a single empirical model that has not been peer reviewed. This model has been used for operational predictions of fire spread despite a lack of validation as to an acceptable level of accuracy, and it has underpinned fuel management for risk reduction despite the fact that its central premise has been refuted by peer-reviewed evidence. In response to some of these observed inadequacies, Gould et al (2007a) have developed another fire behaviour model. This model however while providing some improvements in transparency and validation is subject to many of the same limitations. Most significantly the report gives a lack of support for its central conclusions in a way which may potentially result in fuel management that is at times inadequate or even counter-productive. 41 The dependence upon empirical data also limits the usefulness of the model for informing fuel management. In the case of McArthur (1967), surface fuels at best could be seen as a proxy for all other fuels, where greater density of vegetation produced more dead leaf material (Walker 1981). In the case of Gould et al (2007a), the rate of spread for fires burning elevated, midstorey and canopy fuels is modelled without any information about those fuels, so that surface and near surface fuels are still used as a proxy value. Such proxy values cannot inform decision making about which fuels should be targeted for reduction in fuel management as they incorporate no information about them. Gould et al however have included mean shrub height in their model for flame height, so that the impact of these fuels via flame height on suppression effectiveness can be more effectively assessed. Two peer-reviewed forest fire behaviour models may be possible candidates for use in southeastern Australia as their performance in this area has not yet been tested adequately. These models however have the limitation that they do not consider fuel structure and provide an inadequate simplification of the influence of plant moisture on fire behaviour. These limitations will affect both the accuracy of the models in the context of incident management, and the value of the models for informing fuel management. It is proposed at this point that the current state of knowledge regarding fire behaviour and fuel management in south eastern Australia is inadequate, and that further research is required. Two main shortfalls are identified: 1) Fire behaviour is not understood outside of a small range of fuel and weather conditions; a physical understanding of the interaction of fuels and fire behaviour is therefore required before adequate fire behaviour predictions can be extended to the range of fuel environments found in south-eastern Australia. 2) The two models in use provide very little supported theory upon which to base decisions about fuel management, although Gould et al (2007a) have identified the role of shrubs in affecting flame height. As discussed in section 1.1, the lack of scientific understanding cannot be compensated for with traditional knowledge, as the European-Australian understanding of Aboriginal fire has been irreconcilably altered from the original understanding. Collectively, these facts suggest 42 that the current state of Australian fire knowledge is not sufficient for underpinning effective fire and fuel management. It is intended that by developing a model which explicitly provides a physical understanding of the role of various fuels in a fire, this thesis will make significant advances toward addressing these shortfalls. 43 Chapter 2 Fuels and Flammability Chapter one established the need for an improved understanding of the relationship between fire behaviour and fuels, demonstrating that both the traditional understanding in Australia and the modern scientific basis are inadequate to explain the connection. Chapter two builds on this basis to establish what is known so far, examining two central assumptions about fuels and flammability and thereby identifying where the knowledge deficit lies. 2.1 Defining Flammability Flammability can be understood most simply as the issue of whether the forest “burns well” (Gill and Zylstra 2005). The difficulty arises when the attempt is made to define what burning well means. Consider two different forest types as an illustration. Forest type one is wet sclerophyll found in steep, cool mountainous country and has a dense understorey of shrubs and green ferns, a deep, wet litter layer and an extensive midstorey under tall, close canopied trees. Forest type two is grassy woodland found in dry, hot inland areas. It has a continuous cover of grasses with no understorey or midstorey and the trees are small and open. In most seasons, forest type one is very difficult to ignite due to the wet surface fuels and the protection that the dense structure affords against wind at the ground surface. When fires do ignite, they tend to be small and will often self extinguish. Forest type two by contrast is dry enough to burn far more frequently, so while the time between fires at any one point in forest one may be measurable in multiple decades or even centuries, fires can occur every few years to decades in forest two. At this point it may appear straightforward to say that forest type 44 two is the more flammable of the examples, but this considers only one aspect of flammability – the likelihood that fires will ignite and spread. If instead we define whether the forest burns well by the potential intensity of the fire, we may arrive at an entirely different conclusion. Forest type two may produce frequent and occasionally fast moving fires; however the flame heights are often low enough that local fire crews are able to contain the fire spread in their initial attack. Forest type one however is susceptible to the development of rare but severe crown fires, where all fuels from the ground to the canopy of the tallest trees are burning and producing enormous flames with much faster rates of spread. If this factor alone is considered, it would seem that forest type one burns better. While the connection between flammability and fire behaviour is clearly not straightforward, it may be understood a little more easily if its components are examined. Anderson (1970) defined flammability as the ignitability, combustibility and sustainability of fire in a fuel. These terms refer to how easily the fuel ignites, how well it burns in the sense of intensity and for how long it burns (Gill and Zylstra 2005). In the example given, forest type one is less ignitable but is more combustible and sustainable. A complete definition of flammability for a forest requires both a means of estimating the three aspects of flammability and of putting them back together again; the first question is one of fire behaviour modelling, but the second is far less straightforward. The three characteristics have different implications for the physical world; highly ignitable fuels may produce more frequent fire but this will be less of an issue if the intensity or combustibility is sufficiently low that the fire can be easily suppressed or that the fire does less damage to the values of concern. Long flame residence times allow greater heat penetration into the cambium of the trees or the soil, and can add to the suppression difficulty during the “blacking out” stage of fire suppression. Any attempt to combine the components then involves a subjective weighting based upon the objectives of the manager. Because of this, a fully objective comparison of flammability between forest types is not achievable. Three types of indices may be developed, and these are discussed below. In order to incorporate ignitability, each index is an indication of the percentage of time for which the relevant threshold is breached rather than a single average value. 45 Suppression difficulty. Where the primary consideration is fire suppression, the components of flammability may be weighted by their theoretical impact on suppression difficulty. The considerations and specific fire behaviour are listed in Table 2.1 using the rationale given there; each of these has a threshold beyond which suppression resources are ineffective. The threshold for suppression failure may be tailored specifically to varying levels of resources, and a number of indices produced to accommodate the varying scenarios. Table 2.1. Fire behaviour considerations for suppression effectiveness Consideration Aspect of fire behaviour Rationale The capacity for fire fighting resources to - Rate of spread Fire suppression occurs at a rate suppress the active edge at a faster rate - Spotting limited by the resources available. If the growth in the perimeter of than it is increasing. the fire or the capacity of the fire to breach fuel breaks is greater than the speed of suppression, suppression is ineffective. McCarthy and Tolhurst (1997), Weber and Sidhu (2006) The ability of fire fighting resources to - Flame height suppress any part of the active edge Standard approach used in fire suppression limits suppression tactics by flame height (e.g. Australasian Fire Authorities Council 1996) The capacity for fire fighting resources to - prevent re-ignition of fire perimeters Flame residence in Factors that increase flame course fuels residence in the longest burning fuels (logs, trees etc) along control lines extend the opportunity for reignition of fire edges following initial suppression efforts. The ability of the fire to cross fuel breaks - Spotting distance McCarthy and Tolhurst (1997), as influenced by spotting distance and - Firebrand intensity Gould et al (2007a), Gill (2008) firebrand intensity, and by flame length - Flame length and angle - Flame angle 46 Environmental impact. Environmental concerns need to be measured separately to suppression interests as the two goals should not be assumed to coincide (Clarke 2008). Any number of indices may also be produced to reflect different environmental considerations and a listing is not given here as the considerations are very broad ranging. Economic impact. This relates to the destruction of economic assets and is therefore most affected by the heat load imposed on the asset (e.g. scorch of tree crowns in forestry land or radiant heat damage to buildings), the propensity for firebrands to ignite the asset (Leonard and Bowditch 2003, Sullivan et al 2003, Tan et al 2007) or the residence time of flames in timber plantations where there is a likelihood of cambium damage. 2.2 Describing and classifying fuels By definition, anything that can potentially burn in a bushfire can be classified as a fuel. In this sense, the ‘fuel load’ of a forest consists of the forest’s entire biomass from leaves through to tree trunks. Such a definition is important when considering carbon cycles and the energy balance of an ecosystem, but has little relevance to the behaviour of a bushfire unless all these fuels are going to combust simultaneously and at the same rate. Understanding the conditions that allow these fuels to burn and the way in which they burn is therefore central to understanding fire behaviour Although there are logical divisions in the vertical structure of fuels such as the classification of surface, near surface, elevated, mid-storey, bark and canopy fuels (McCarthy et al 1999, Gould et al 2007a) , the concept of a fuel ‘stratum’ can erroneously suggest the idea of a homogenous layer. With the exception of surface fuels and some plant forms such as grasses or heaths that form closed swards, bushfire fuels are naturally discontinuous in nature with the majority of plants growing as isolated individuals. The scale of a fuel array The term “fuel array” refers to all of the potential fuels in a given location. These can be classed approximately into the following categories: 47 1. Fuel particles – individual leaves, portions of bark etc. 2. Fuel units – a collection of fuel particles into a natural grouping that can effectively burn as a unit, for example a tree or shrub. There are often many smaller fuel units within a large unit, for instance one branch of a tree may burn without the entire tree catching fire. 3. Fuel strata – fuel units located at approximately the same height in the fuel array so that horizontal fire spread can potentially occur between them. Fuels, flammability and fire behaviour will be examined in this thesis at the scales of leaf, plant and array, with the understanding gained from the simpler fine-scale levels used to build the more complex levels. 2.21 Fuel dynamics The term ‘fuel dynamics’ refers to the way in which the quantities of different components in a fuel array change over time, or to the way in which changes in fuels influence the flammability of the array. If we consider fire behaviour and therefore flammability to be the interaction of fuel, weather and terrain, then the change in fuels represents a potential change in forest flammability. A direct relationship should not be assumed however as other factors such as the arrangement of the fuel or the way it interacts with weather variables may introduce feedbacks and/or complexity. McCarthy et al (2001) described five different models of the way in which flammability changes over time due to fuels. These were: 1. Stasis – flammability remains constant 2. Linear model – flammability increases without limit 3. Olsen model – rapid initial increase eventually reaching equilibrium 4. Logistic model – slow initial increase followed by a rapid climb and eventual equilibrium 5. Moisture model – rapid initial increase to a peak followed by decrease and eventual equilibrium due to plant senescence or succession to wetter or less flammable species. 48 A sixth possibility may also be proposed: 6. Succession model – flammability rises and falls over time depending on the structure of the array; a state of equilibrium may or may not be reached. The moisture model may be seen as a special case of this model. These models are illustrated below in Figure 2.1. Flammability 2 4 3 6 5 1 Time since fire Figure 2.1. Theoretical models of fuel accumulation, 1. Stasis, 2. Linear, 3. Olsen, 4. Logistic, 5. Moisture, 6. Succession. It is not assumed in this thesis that overall fuel quantity equates to flammability as this is the subject being investigated. The rates of change for various fuel descriptors can however be described with some level of generalisation. 2.211 Surface fuel accumulation Surface fuel accumulation is most often modelled with the Olsen Model (Olsen 1963, equation 2.1). This produces an initially steep curve of fuel accumulation which levels out into a quasi-steady-state over time. The assumptions are that all litter is consumed in a fire, and that the rate of litter fall is constant, as is the rate of decomposition k. If there were no decomposition, the increase in fuel load would be linear, but as the litter load increases the 49 habitat for decomposers improves and therefore the rate of accumulation decreases over time. The steady-state reached (Xss) is the point where litter is decomposing at the same rate that it is accumulating (Cary and Golding 2002). ( W = X ss 1 − e − kY ) Equation 2.1 Where W is the weight of litter (t.Ha-1) at Y years since fire, and e is the natural logarithm. The assumptions described above simplify the situation due to the fact that not all litter is always consumed in a fire, and that litter accumulation and decomposition are not constant in reality. Depending upon the impact of the fire on the forest canopy and understorey, the immediate litter fall may be accelerated as scorched leaves fall following a fire, or delayed if a forest canopy is consumed by crown fire. Similarly, decomposition rates are determined by factors such as the moisture content, fungal biomass, diversity and abundance of litter dwelling invertebrates, soil nutrient status and the temperature of the litter (Raison et al 1986, Cary and Golding 2002); all factors which vary seasonally and with the temperature and frequency of fire. The rate at which litter is added La is related to the Projected Foliage Cover FP – “the proportion of ground that would be shaded if sunshine came from directly overhead” (Walker 1981) according to: La = 7.45 + 0.123 FP − 0.004 FP2 Equation 2.2 Having made these qualifications, the limitations given do not significantly affect the accuracy of the model; rather their effect is more a reduction in precision. While an Olsen curve is unlikely to give a precise fuel load measurement for a given point on a given day, the curve will give adequate average fuel load values expected over the general area. Surface fuel accumulation values are given for a number of different forest communities in Table 2.2. 50 TABLE 2.2 Surface fuel accumulation values for various Australian forest communities, using an Olsen curve (Equation 2.1) Vegetation Alliance Alpine Ash (E. delegatensis) k Xss Samples Visits Source 0.23 23.45 Hutchings and Oswald (1975) 0.21 14.36 48 8 Zylstra (2007) 0.70 13.0 70 7 Raison et al (1986). 0.31 22.9 120 12 Raison et al (1986). Banksia low woodland 0.318 2.45 Burrows and McCaw (1990) Brown Barrel (E. fastigata) 0.34 24.56 Hamilton (1964) 0.41 10.50 Hutchings and Oswald (1975) 0.47 15.85 84 14 Zylstra (2007) Dry Sclerophyll 0.57 11.33 210 35 Zylstra (2007) High Montane 0.25 12.71 198 33 Zylstra (2007) Jarrah, Northern (E. marginata) 0.204 13.16 Gould et al (2007a) 0.168 12.36 Gould et al (2007a) Mallee Heath 0.064 6.64 McCaw (1997) Moist Forest 0.44 13.84 Mountain Ash (E. regnans) 0.35 22.25 Ashton (1975) 0.26 12.75 Hutchings and Oswald (1975) 0.77 17.45 24 4 Zylstra (2007) Peppermint (E. dives) Forest 0.60 11.76 90 15 Zylstra (2007) Peppermint (E. dives) Forest 0.42 14.9 80 8 Raison et al (1986). Peppermint (E. dives) – 0.52 9.31 108 18 Zylstra (2007) forest Alpine Ash (E. delegatensis) forest Alpine Ash (E. delegatensis) (pole stand) Alpine Ash (E. delegatensis) (mature stand) forest Brown Barrel (E. fastigata) forest Candlebark (E. rubida) Woodland forest – low shrub understorey Jarrah, Southern (E. marginata) forest – Taxandria parviceps understorey 198 33 Zylstra (2007) forest Mountain Gum (E. dalrympleana) forest Peppermint (E. dives) – Box (E. polyanthemos) Woodland Mountain Gum (E. dalrympleana) Forest 51 Red Stringybark (E. 0.28 11.00 Hutchings and Oswald (1975) 0.10 16.10 Shalders (2007) 0.28 13.10 0.27 22.34 0.29 12.13 0.43 11.30 macrorhyncha) Forest Red Stringybark (E. macrorhyncha) Forest Snowgum (E. pauciflora) – 204 34 Zylstra (2007) Mountain Gum (E. dalrympleana) Forest Snowgum (E. pauciflora) – Hamilton (1964) Mountain Gum (E. dalrympleana) Forest Snowgum (E. pauciflora) 126 21 Zylstra (2007) Woodland Spotted Gum (E. maculata) McColl (1966) forest 2.212 Dynamics of fuels in the higher strata Gould et al (2007a) divided the higher fuel strata into near surface fuels consisting of grasses, low shrubs and suspended litter up to 50cm above ground level, elevated fuels (shrubs), intermediate (midstorey), overstorey (canopy) and bark. These strata were ranked in the report using a subjective visual estimate of density, an average measure of plant height and a decision “based on the perceptions of experienced fire personnel” that the quantity of dead material in the plants was the primary factor determining their flammability (Wilson 1993). The combination of these factors was used to produce a fuel ‘hazard score’ similar to the approach of McCarthy et al (1999). This thesis divides fuel strata in the same manner, using the names ‘midstorey’ and ‘canopy’ instead of ‘intermediate’ and ‘overstorey’, however no subjective scores are used. The dynamics of all fuel strata above the surface fuels are dependent upon the ecology of the plants. Suspended dead leaves categorised as near-surface fuels are still dependent upon the presence of shrubs within which to be suspended, without these, the leaves reach the ground surface and become surface fuels. As a number of studies have established that surface fuels have very little influence on fire behaviour (Burrows 1994, McAlpine 1995, Burrows 1999a,b, Gould et al 2007a), it may be that the dynamics of many fuels actually follow a complex ‘succession curve’ such as number six in Figure 2.1. In this case, any management of fuels is uninformed if it does not consider the ecology and physiological characteristics of the plant species such as their growth rates. 52 2.3 Succession and fire ecology Richard Helms (1893) in his examination of grazing leases in the Australian Alps observed: “A common and in my opinion a very improvident practice, which will probably be continued, is the constant burning of the forest and scrubs. This proceeding has only a temporary beneficial effect in regard to the improvement of the pasture by the springing up of young grass in places so cleared, for after a year or so the scrub and underwood spring up more densely than ever… however I have seen some very detrimental effects from this practice here, because the heavy rains wash the soil away from the steep declivities, and it is either carried into the creeks and rivers and lost entirely, or it accumulates in the boggy places, and thus becomes useless. The more or less constant diminution of humus in the soil of the slopes is a danger not generally recognised.” The response of “scrubs” to fire by dense colonisation was a common observation for many years throughout the mountains of south eastern Australia. Speaking of the Mountain Ash (Eucalyptus regnans) forests of Victoria, A.E. Kelso, the engineer in charge of the water supply for the Board of Works (Stretton 1939) stated in evidence to the Royal Commission into the 1939 “Black Friday” fires: “In the natural forest there is not much scrub. I can show you forests in which there is not much scrub and in which you can ride uninterruptedly without trouble. I am not going to suggest the forest floor is clean by any means; but it does not carry dogwood scrub and this other kind of bush which normally follows after fire. “We have parts of our areas where there is such scrub, and we will not attempt to deny it. There are parts to which fires have had access and we have put them out. In those parts I have no doubt you would find bracken and dogwood. “It is our experience that if fire is excluded that kind of bush will gradually revert to the cleaner type of bush.” 53 In reference to the Alpine Ash (E. delegatensis) forests of Victoria, the 1957 report by the Australian Academy of Science (Turner et al 1957) stated: “The forest floor of the surrounding Alpine Ash (Woollybutt) forest (E. gigantea) was formerly grassy and was regarded as the choicest of the summer grazing lands, especially as the trees provided shelter for stock in the uncertain summer climate. These forests, however, have been regularly burnt-off in autumn over a long period of years and this has undoubtedly resulted in extensive sheet erosion throughout the Murray and Snowy catchments, such as has been reported over and over again. The result of this burning has also been that scrub now dominates these areas to the exclusion of grass. Not only are they, therefore, a greater fire menace, but cattle and sheep now tend to concentrate for grazing on the High Plains, finding shelter but little to eat in the adjacent forests.” Graziers of the area were also aware of the dynamic. Roy Hedger, a grazier from the Snowy Plains area of the NSW Alps and former holder of a snow lease prior to declaration of the Kosciusko State Park wrote (Hedger 2003): “We would never burn bush country as it would make it worse and bring suckers up.” And: “In my experience, bush country should not be burnt above 5000 ft. It’s the clear country that should be burnt every five or six years.” Harry Reid observed the effects of an intense fire run in 1939 (Moriarty 1993): “Early in this century there were old White Gum [E. pauciflora] trees 30 to 40 feet apart on the slopes above the Snowy River near Island Bend. It was easy and pleasant to ride through those woodlands. “After the intensely hot 1939 fires thick impenetrable scrub grew and it became very difficult to ride through it.” 54 Or in the words of Ken Kidman (Neal 1988): “In 1939, a big fire swept through the mountains. The burnt country remained open for a few years but afterwards the scrub grew back thicker than ever – too thick for a dog to bark in!” Rogers, another grazier from the Victorian part of the Alps said (Wakefield 1970): “My father came to Black Mountain in 1902. In those days John O’Rourke of Wulgulmerang and others used to tell of the open, clean-bottomed, park-like state of the forests of this tableland and adjacent areas, which they could well remember from earlier days.” In order to encourage grass growth, the Rogers family introduced fire to all areas except the “Box country”, as this naturally had good grazing. “The practice was to burn the country as often as possible, which would be every three or four years according to conditions. One went burning in the hottest and driest weather in January and February, so that the fire would be as fierce as possible, and thus make a clean burn. As a general practice, in the valleys we would light along the rivers and creeks so that the fire would roar up the steep slopes on either side, making a terrific inferno and sweeping all before it. The hotter the fire, the sweeter and better the feed for the cattle after the new growth came.” (Wakefield 1970) The effects of this however were unexpected: “It would seem that the long-followed practice of regularly burning the bush in the hot part of the year has resulted in a great increase of scrub in all timbered areas except the Box country.” Rogers described the effects in the main forest types of his area, and stated that the most pronounced effect was in the Peppermint forests: 55 “By three years, shrubbery had matured and ceased growth, grass coverage was smothered and reduced in bulk, and the eucalypt scrub was about seven or eight feet high. At this stage the area was re-burnt.” As discussed earlier, the available fire behaviour models can not provide any insight as to the effect of these fuel changes on rates of spread, although the flame height model of Gould et al (2007a) suggests that if the shrubs had matured at three years then this also marked the point of peak flame heights. It is likely however that as the eucalypt regrowth was still only seven or eight feet (approximately 2.5m) high, this may also have acted as a fuel under some circumstances. The “crown fire” produced would almost certainly have produced greater flame lengths than expected from the shrub height or the three years of litter accumulation, and probably far greater rates of spread. Cruz et al (2005) suggest that rates of spread in crown fires are often twice the speed of the surface fire. Despite this inability to quantify the changes, the influence of shrub fuels on fire behaviour is recognised in some areas of Australia through tools such as the “Overall Fuel Hazard Guide” (McCarthy et al 1999); however in absence of an understanding of the fuel ecology, the response to the observation has been pressure to return to a fire regime similar to that described by Rogers, where the forest is re-burnt when shrub re-growth reaches a given density (McCarthy and Tolhurst 1998, RFS 2007a). The discussion of the potential effectiveness of this approach has been held outside of the peer-reviewed literature, and the ecological responses of the relevant species are at times explicitly ignored or even dismissed as “stupidity” (e.g. Attiwill et al 2009). 2.31 Fuel ecology The examples above demonstrate live fuels following a trend which is counter to what has been referred to as the ‘fuel-age paradigm’ (Zedler and Seiger 2000) where the fuel load always increases with time since fire, at least until it reaches a point of equilibrium. Following the 1939 “Black Friday” fires in Victoria, Judge Leonard Stretton recognised that there were two possible cures for damaging fire in their tall wet forests – either exclude fire long enough for the flammable understoreys to develop into less flammable arrays, or to introduce regular, controlled fire. Due to the fire-loving culture of Australians however, he 56 concluded the former would be impossible in European Australia so the latter was the only alternative (Griffiths 2002). It should not be assumed however that this is the case in all circumstances or indeed that colonisation by fire-germinated species automatically increases the flammability of a fuel array. This assumption that increasing fuels directly increases the flammability of the array was first articulated by McArthur (1967), but as discussed in section 1.22, this ‘fuelflammability paradigm’ has been thoroughly investigated and shown theoretically and empirically to have no basis in the community where it was first developed. The paradigm is a theoretical construct which ignores the complexities of heat transfer and fuel structure. Changes in live fuels are dictated by the ecology of the plants involved, displaying variability between communities. Also, while increased vegetation density in an array may increase the total weight of fuels present, it may be too simplistic to assume that all species are equally flammable or that the effects of increased shading on lower fuels or increased vegetation density on wind speeds in lower strata are negligible. Sturtevant et al (2004) for instance modelled the influence of fire suppression on understorey composition in northern hardwood and boreal species for northern Wisconsin; finding that the suppression of low intensity fires encouraged the establishment of sugar maple (Acer saccharum), whereas more frequent fire favoured conifer establishment and more frequent stand-replacing fires. This contrasts with other studies from the western United States, where reduced fire frequency can lead to dominance of fir species and thereby encourage more stand-replacing fires through improved ladder fuels (e.g. Stephensen 1999). In both cases, fire suppression encouraged biomass accumulation contrary to the examples given earlier, but added ‘fuel load’ increased high intensity fire likelihood in one community while decreasing it in another. Plucinski (2003) examined the accumulation of heathland biomass from a number of studies, demonstrating the dynamic nature of the process as a result of pyric succession. Tables 2.3 and 2.4 list a number of studies for a variety of communities globally to illustrate the variety of responses shown by life fuels to disturbances. This is by no means an exhaustive list of vegetation responses, but is intended to address the hypothesis that fire will always either reduce or increase the biomass of live fuels (fuel-age paradigm) by demonstrating that either possibility can occur, depending on the community, forest structure and component species. 57 Table 2.3. Responses of a range of Australian plant communities to disturbance Australian vegetation Atmospheric Increasing growth of woody vegetation found to correlate with Berry and composition increasing CO2 concentrations. Roderick (2002) Frequent fire Biomass increased in subalpine grasslands for 26 years with Fallon (2008) and sheep removal of disturbance, increase accelerated over the next 20 grazing years then declined over the next 15 years. Frequent fire Biennial fire in Themeda triandra – Poa sieberiana grassland Prober et al caused increased soil compaction, decreased infiltration and (2008) Grassland decreased soil biological activity causing poor sward recovery. Fire and sheep Decrease in Poa spp. in subalpine grasslands after repeated fire grazing due to topsoil loss Newman (1954) Heathland Fire Tall shrubs in the Mt Wellington alpine area tend to be more Kirkpatrick et al fire-germinated than other growth forms. (2002) Fire and Cattle Heathland plots burnt in 1939 were colonised by shrubs on flat Wahren et al grazing to gently sloping ground. Grazed areas were colonised by taller (1994) shrubs than ungrazed areas and shrub cover was still increasing 55 years later. Shrubs on ungrazed land began to senesce after 40 years. Frequent fire Very frequent fire in a Sydney sandstone heathland markedly Bradstock et al reduced the density of the large shrub Banksia ericifolia and (1997) increased the frequency of the grass Entolasia stricta. Open Forest and woodland Fire Germination of leguminous shrubs in Sydney sandstone influenced by the weight of surface fuels burning. 6t.Ha Newman (1954), -1 likely to germinate seeds to a depth of 1cm, whereas 20t.Ha Dickinson and -1 Kirkpatrick 58 likely to germinate seeds down to 3cm. Unrelated to Byram (1985), Hobbs intensity. and Atkins (1990), Decrease in Bossiaea foliosa, Cassinia sp., Helichrysum spp. Hodgkinson and on flat ground in subalpine areas with repeated fire, increase on Oxley (1990), slopes due to soil loss. Bradstock and Auld (1995), Hill General trend apparent in Tasmanian forests where dry habitats and French (2004) favour species with low ash contents, high energy levels, high volatile oil contents and low moisture contents. Spring burns in Banksia woodland of Western Australia were found to produce faster and more dense regrowth than autumn burns. The large shrub Eremaea pauciflora covered increasing area with time since fire, although the number of individuals declined with age as did the leafy biomass. Fire promoted germination of the arid-zone species Acacia aneura and to a lesser extent Cassia nemophila where heavy surface litter produced slow-burning, low intensity fires; whereas faster grass fires did not induce germination. Germination of Dodonaea viscosa was not affected by either treatment. Shrub diversity and abundance was promoted by fire in Cumberland Woodland, and summer fires were found to promote eucalypt recruitment. Fire and sheep Biomass increased in E. pauciflora woodland for 26 years with Wimbush and grazing removal of disturbance, increase slowed over the next 20 years Costin (1979a), then biomass declined over the next 15 years. Fallon (2008) Woody biomass in transition between woodland and grassland areas increased very strongly for 46 years, then significantly less so in the next 15 years. Grazing prevented lignotuberous regrowth following high intensity fire in E. pauciflora woodland. Cessation of fire caused decrease in Pimelea spp. and increase in the grasses Poa fawcettiae, P. costiniana. Also increase in Bossiaea 59 foliosa, Oxylobium ellipticum, Helichrysum hookeri; particularly strong increase in Bossiaea foliosa compared to ungrazed control. Frost Extreme frost events killed Bossiaea foliosa shrubs in areas of Leigh et al (1987) cold air drainage Soil loss / Strong increase in the shrubs Oxylobium ellipticum, Hovea Wimbush and cover loss purpurea var. Montana, Grevillea australis and decrease in Costin (1979b), Helichrysum alpinum in subalpine woodland following topsoil Williams and loss. Ashton (1987), Leigh et al (1987) Creation of bare ground in subalpine areas due to grazing/fire promoted growth of Asterolasia trymaloides, Grevillea australis, Phebalium squamulosum, Prostanthera cuneata. Tree clearing Tree removal in Stringybark open forest promoted growth of Gibbs et al (1999) deep-rooted, summer active grasses such as Aristida ramosa in preference to year-long green, shade tolerant grasses such as Microlaena stipoides Rainforest Fire Low intensity humus fires in Tasmania had species-specific Hill (1982), Hill effects on plant mortality, killing the trees Nothofagus and Read (1984) cunninghamii and Eucryphia lucida in particular. Fire extensively germinated Acacia melanoxylon, although numbers thinned rapidly over the next nine months. Regeneration of all rainforest species followed low intensity fire in pure stands in Tasmania, however fire in mixed rainforest – wet sclerophyll forest encouraged dominance of sclerophyllous understorey species. Rainforest E. regnans senesce around 300 yo, shift toward rainforest. Ashton (1981, formation Canopy opening of old-growth due to storm damage or tree 2000), Brown and death causes increase in ground ferns and tree ferns with shift Podger (1982), from Pomaderris aspera to Olearia argophylla. Final stage is Ellis (1985) Atherospermum moschatum, Nothofagus cunninghamii and 60 Olearia argophylla rainforest. Closed rainforest formation in Tasmania was found to be related to time since fire, with significantly greater occurrence of Nothofagus cunninghamii dominated forest occurring in areas unburnt for 250 years or more. Wet Sclerophyll Forest Fire In dry Alpine Ash forest, E. delegatensis seedlings require Floyd (1966), removal of understorey competition and litter to establish. Ellis (1985), Mature trees resistant to fire, trees <40cm DBH highly fire Bowman and sensitive. Kirkpatrick (1986) Frequent fire used by indigenous people to maintain grassy disclimax communities in flat or gently sloping areas within Tasmanian wet sclerophyll forest. Differing fire rates of spread were found to favour different leguminous species in NSW wet sclerophyll forests. Prolonged soil heating encouraged the vine Kennedia rubicunda strongly but only moderately germinated the midstorey species Acacia binervata, A. irrorata and the shrub Dodonaea triquetra. Fast moving fires which provided less soil heating had the reverse affect. Low intensity Dense colonisation of E. regnans forest by herbs such as fire Senecio and grasses e.g. Dryopoa following surface fire, light Ashton (1981) re-burn in next few years produces Cassinia aculeata; otherwise matures to Pomaderris aspera and Olearia argophylla. High intensity Pteridium esculentum, E. regnans saplings and Cassinia fire aculeata or Pomaderris aspera thickets colonise E. regnans Ashton (1981) forest after crown fire. Repeated fire within 15 to 20 years can remove E. regnans. 61 Table 2.4 Responses of a range of international plant communities to disturbance Heath Fire Fire encourages exotic species at the expense of native herbs in Muñoz and Chilean Matorral Keeley (1986, 1987), Fuentes Laboratory studies in Chilean Matorral show significant fire (1989), Delitti et induced germination of coloniser shrubs in comparison to al (2005), Gómez- climax species, limited field studies show fire temperatures to González and kill rather than germinate soil stored seed. Cavieres (2009) Increasing fire frequency caused increases in herbs and subshrubs in Spanish Garrigues, decreasing shrub cover (Ulex parviflorus) and significant reduction in biomass and Leaf Area Index of the canopy tree Kermes oak Quercus coccifera. Germination of shrubs in Californian chaparral is induced by fire. Conifer forests Fire Increased density of understorey saplings and smaller trees Dodge (1972), observed in ponderosa pine forests following fire exclusion. Cocke et al (2005), Hart et al Biennial fires over a period of 20 years were found to reduce (2005), Espelta et the below-ground biomass of fine roots and mycorrhizal al (2008), Veg et symbionts in ponderosa pine forests. al (2008), Zalba et al (2008) Fire exclusion promoted the spread and density of mesic species along with increasing tree density in mixed conifer forest on the San Francisco Peaks. Repeated fire produced decline in tree density, height and reproductive ability of Pinus halapensis, whereas an isolated fire event promoted invasion of the species into an Argentine grassland. Increased burn severity in Galician Pinus pinaster forests 62 encouraged greater seedling germination and establishment. Savannah Forest Fire exclusion Increased cover and height of litter and grasses in Thai Stott (1986) savannah woodlands Tropical rainforest Fire Fire causes positive feedbacks in future fire susceptibility, fuel Cochrane et al loading and fire intensity in Amazonian rainforest. (1999), Cochrane (2003), Cochrane and Messina (2006) 2.4 Fuel age and flammability It is apparent from these examples that although surface fuels may accumulate in a predictable Olsen curve, there is no straightforward rule defining the way in which live fuels accumulate following fire. In some communities species are germinated by fire and produce a greater biomass in the younger fuels than the older communities; in others succession may cause changes on the scale of centuries with biomass rising and falling in that time. Fire intensity, burn season and other factors affect the germination of some species but not others. Feedback relationships between species (Freckleton 2004) can also introduce complexity, so that stochastic disturbance events in the course of a community’s development may cause bifurcations in the succession from that point onward (e.g. Ashton 1981). While the accumulation of biomass relates directly to the accumulation of potential fuel, it may not relate directly to flammability. Variability between some of the flammability characteristics of different species (e.g. King and Vines 1969, Dickinson and Kirkpatrick 1985, Hogenbirk and Sarrazin-Delay 1995, Gill and Moore 1996, Dimitrakopoulos 2001, 63 Tran et al 2001, Boer and Stafford Smith 2003, Bradstock and Kenny 2003, Gill and Zylstra 2005, Keith et al 2007, Dibble et al 2007, Sandberg et al 2007) may cause the biomass of some species to have a greater effect on fire behaviour than others. Taller plants or increased vegetation density also reduces the speed of the wind in the lower strata (McArthur 1966, Tran and Pyrke 1999, Gould et al 2007a), thereby directly affecting fire behaviour. Shading reduces solar radiation and along with reduced wind speeds reduces the temperature and raises humidity at the forest floor (Byram and Jemison 1943, Van Wagner 1969, Viney 1991, 1992, Matthews 2006, Matthews et al 2010). This increases dead fuel moisture content (McArthur 1962, McArthur 1967, Sneeuwjagt and Peet 1985, Van Wagner and Picket 1985, Péch 1989, Matthews 2006) and reduces evaporation (Penman 1948) from soils. The presence of a closed canopy in some forests has also been shown to maintain a cooler, higher humidity microclimate which inhibits or prevents the spread of fire (Cochrane et al 1999, Cochrane 2003, Cochrane and Messina 2006). These changes in biomass, plant flammability and the impacts of community structure on air movement, moisture and energy flow mean that the ignitability, combustibility and sustainability of a fuel array may fluctuate widely over time. By contrast, the fuel-age paradigm proposes that ‘fuel load’ will increase with time since fire, perhaps stabilising after a period. The fuel-flammability paradigm builds on this, suggesting that increases in fuel equate to increases in flammability. Because the implication is that re-burning fuels will always return the array to a period of time for which it will be less flammable, these two assumptions form the basis of prescribed burning theory. Prescribed burning has been advocated across numerous fuel environments, although the basis for this broad-scale application may be at this time little more than the untested assumption that seral changes as described above will have little or no effect on flammability. Fernandes and Botelho (2003) alluded to this by stating that “the efficiency of prescribed fire in reducing wildfire hazard is frequently mentioned as a matter of fact, but the basic premise is seldom questioned.” Studies examining the effectiveness of prescribed fire rely heavily on anecdotal evidence (Fernandes and Botelho 2003), selected examples (e.g. McArthur et al 1966, Peet and Williamson 1968, McCaw et al 2008) and are rarely subject to peer-review. The use of selected examples as opposed to rigorously systematic sampling introduces bias in the tendency to ‘cherry pick’ favourable results and ignore those instances where young fuels did not have an expected effect. 64 Systematic studies are rare and a comparatively recent phenomenon (e.g. Seydack et al 2007, Boer et al 2009, Price and Bradstock 2010), although an increasing number have been conducted in US fuels using a modelling approach rather than direct observations of the effectiveness of a treatment in stopping a fire. This approach measures the reduction in key fuel variables identified by relevant models, removing the bias of anecdotal evidence but creating dependence on the validity of the model in the given fuel array. Studies such as Schwilk et al (2009) and Vaillant et al (2009) established that prescribed fire was effective in reducing these variables across a large range of US forests, but only considered the immediate effectiveness of the treatment rather than its value over time. The successional model of forest flammability described above does not challenge this premise; rather it proposes that while flammability may be expected to be less immediately after a fire, in some environments the follow-on effects of plant growth and changing forest structure may in time produce a greater level of flammability compared to the long-unburnt fuel array. In considering these factors, Fernandes and Botelho (2003) said “…post-treatment recovery can be so fast that fuel management may be futile or even counter-productive in some fuel types”. Their findings were that on consideration of the available case studies and reports, the average effectiveness of prescribed fire in assisting suppression efforts was a post-treatment period of 2-4 years. The value of studies that integrate systematic findings into mean values across community types is reduced by the fact that the mean value disguises the exceptional cases. Zylstra (2006) for instance examined 169 instances where unplanned fires had burnt into areas that had been burnt within the previous 16 years by either natural or prescribed fire, comparing the number of instances where fire had either self-extinguished or been extinguished for each age class. The data showed a statistically significant effect of fuel age, the strongest comparison being where fires burning into fuels up to five years of age were 1.34 times more likely to be extinguished than in fuels 10 to 15 years of age (p=0.1). McCarthy and Tolhurst (2001) collated anecdotal evidence from fire operational personnel for 114 fires across Victoria as to the effectiveness of prescribed burns in assisting suppression, concluding that prescribed burns may provide assistance for up to 10 years. In both studies, results across widely varying fuel arrays were pooled to provide statistical strength. No statistics for the variability in suppression success were provided by McCarthy and Tolhurst (2001); however Zylstra (2006) gave the variance for one to five year old fuels as 42.7% and 186.6% for 10 to 15 year-old fuels. Linear regression was performed for a smaller data set examining large 65 burnt areas, with an R2 of 0.10. If the dataset for the smaller blocks is analysed, this gives an R2 of 0.07. These values demonstrate that for the range of fuel arrays in Kosciuszko National Park, approximately 90% of the variability in suppression success is due to factors other than fuel age (e.g. weather, terrain, access, availability of resources). This suggests that in some arrays young fuels provided greater assistance than the average, whereas in other areas they provided less. An illustration of this can be drawn from the chaparral fires of California where since 1970, 12 of the 15 most destructive fires in the United States have occurred (Halsey 2004). Contrary to the much publicised dynamics of the ponderosa pine forests where indicators of flammability have been shown to increase for many years after fire (Dodge 1972, Schwilk et al 2009, Vaillant et al 2009), fuel age has been shown to have no effect on chaparral flammability (Dunn 1989, Keeley et al 1999, Zedler and Seiger 2000, Keeley and Fotheringham 2001) although fire is necessary for the germination and persistence of the shrubs (Muñoz and Keeley 1986, 1987). Any study that used an average derived from these two important Californian environments would produce a bland generalisation of the effect of fuel age. To the detriment of forest health, too little emphasis would be placed on introducing fire to the ponderosa pine, and fire would be introduced to chaparral when in some cases exclusion is warranted. After discussing this matter, Fernandes and Botelho concluded that it “leads to the conclusion that the fuel/age paradigm is a simplification, and that the hazard reduction effectiveness of prescription burning will vary by ecosystem (or fuel type) and according to the relative impacts of fuels and weather on fire behaviour.” 2.41 Conclusion The relationship between fuels and fire behaviour is a poorly understood science. In the absence of systematic scientific study, two assumptions have been used to define the relationship, the fuel-age paradigm and the fuel-flammability paradigm. The fuelflammability paradigm has been brought into question with physical arguments and in some cases has been soundly refuted by empirical and theoretical evidence. Likewise, the fuel age paradigm has been demonstrated not to hold true in all environments; so much so that in some environments there is an inverse relationship between fuel age and flammability. As a result, the dynamics of flammability and implications for its management are highly complex and general trends observed in one fuel array should not be applied indiscriminately 66 across ecosystems. Studies that determine a mean relationship with time since fire across a range of communities may have uses from a landscape planning perspective but should not be used to determine fire management at given sites. Hazard reduction effects in some communities may be longer lasting than expected from mean values, whereas complexities in secondary succession and other factors may cause other communities to recover their flammability at a rate that provides no benefit proportional to the cost or may even be counter-productive. Consequently, quantifying the effectiveness of fuel management operations relevant to specific communities and their limits in relation to weather factors is a priority area of research (Schmoldt et al 1999). Empirical observation of the effectiveness of fuel treatments is fraught with difficulties; a possible alternative is the use of fire behaviour models to interpret changes in the fuel array over time, however the value of this is dependent on the ability of the model to quantify the influence of different fuels, flammability properties and forest structure on fire behaviour. As discussed in the previous chapters the existing Australian models are not capable of this. 67 Chapter 3 A Conceptual Model of Fire Behaviour The preceding chapters have identified gaps in the current understanding of fuels and flammability. Both fuel management and fire behaviour modelling are frequently premised on the fuel-age and fuel-flammability paradigms, despite the fact that both assumptions are beset by theoretical and observed contradictions. The validity of the fuel-age assumption has been shown to be dependent upon the ecology of the fuel array, and the fuel-flammability paradigm has been refuted in the peer-reviewed literature. Removing the faulty assumption has however left unexplained the relationship between fuel and flammability classically expressed in the ‘fire triangle’ of fuel, weather and terrain. Consequently, the remainder of this thesis sets out to reduce the knowledge gap by replacing the fuel-flammability paradigm with a model that can quantify the respective roles of fuel quantity and associated factors such as fuel arrangement. This chapter presents the conceptual framework for the model and the following three chapters construct the model in increasing complexity from the scale of a particle to that of an entire fuel array, so that a fully operational fire behaviour model is given in chapter six and included on the attached CD. After providing some validation of the model, it is applied as a corrected understanding of the fuel-flammability relationship and used to examine the flammability-age relationship for a forest array. This is contrasted with the predictions of three empirical fire behaviour models that are premised upon the earlier assumptions. 68 Overview of the chapter Wildland fire behaviour is described as a complex system based on observations of fire behaviour, and a conceptual model for this is presented. Flame dimensions and rate of spread are defined as emergent behaviour resulting from the capacity of flame to cross spaces and ignite new fuel elements at different scales. Fire behaviour is therefore related to the geometry of the fuel array and to the theory of flammability, where the ignitability of a fuel element increases the likelihood that flame will propagate to it, and where higher combustibility and sustainability values increase the likelihood that it will produce flame capable of propagating to neighbouring fuel elements. Positive and negative feedbacks between flame characteristics and fuel availability are explored, and the use of an iterative approach is proposed to accommodate these in the model. The model provides a framework whereby theoretical work on component aspects of fire behaviour can be integrated into an operational fire behaviour model, and produces a physical mechanism for fire behaviour that could be used to objectively inform fuel management. 3.1 Introduction Sudden escalations in fire behaviour have been responsible for a significant number of firefighter fatalities (e.g. Weick 1993, Butler et al 1998, Stevenson 2001, Weick 2001, Rothermel and Mutch 2003 and Viegas 2006, 2009) and knowledge of their likelihood is therefore important from an operational perspective. Observations of sudden escalations and their underlying causes are frequently recorded in the literature, but due to modelling difficulties, most mainstream fire behaviour models do not address many aspects and rely instead on the assumption of a persistent increase in a given behaviour pattern as the causative weather conditions intensify (McArthur 1967, Rothermel 1972, Gould et al 2007a). This chapter presents the case that wildland fire behaviour does not intensify along a simple linear pathway, but displays sudden escalations in an irregular step-wise pattern that a statistical approach will smooth over. These escalations represent critical points resulting from many component ‘decisions’. Every space between fuel elements represents a potential decision – fire will either grow by crossing the space and igniting new fuels or it will be 69 limited to the fuels already burning. The interaction between flammability of fuel elements and geometry of the fuels is encapsulated in the concept of an ignition isotherm (Weber 1990a) – the distance from a heat source to the point where temperatures are sufficient to ignite a fuel. Complex systems modeling of fire spread has at times focused on simplistic geometric relationships while greatly simplifying the actual mechanism of spread itself (e.g. Green 1989, Resnick 1994, Malamud et al 1998). While providing some useful insights, oversimplifying the mechanism of spread has serious limitations (Karafyllidis and Thanailakis 1997). The concept of the ignition isotherm proposes a physical mechanism as a hypothetical means of fire spread that is based upon the core concepts of flammability previously discussed. Because available fuel is determined through a multitude of choices located between each potential fuel element, the primary difficulty in modelling fire is that of quantifying the fuel input. Fuel inputs determine flame characteristics, but flame characteristics also affect fuel inputs. The reciprocal dependencies between fuel and fire behaviour mean that it is often not possible to adequately quantify fuel inputs based purely on a priori knowledge. Consequently, although it may be possible to identify a generalised trend of fire behaviour from empirical data specific to a set of conditions, it is not possible to calculate the amount of fuel that will contribute to the fire or the resulting fire behaviour at a point without addressing the complex processes within the system. While a generalised trend may prove to be acceptable when averaged across a broad range of conditions, specific predictions made from the trend will frequently display large error margins because they are unable to identify the physical triggers within the system that initiate sudden escalations of intensity. It is this propensity for a system to develop due to the influence of its components rather than from the influence of an external guiding trend that is essentially the defining characteristic of a complex system (Bak et al 1988, Ottino 2003). The constraints to effective modelling are described based upon observed fire behaviours, and a conceptual model for a complex system using flammability characteristics and fuel geometry to define the ignition isotherm is proposed as the basis for developing an operational model. In this chapter, a number of thresholds and feedback systems are identified then organised into a conceptual framework. Freedom is allowed for individual 70 processes (eg the combustion of a leaf) to be modelled using fundamental physics and chemistry or via empirical simplifications. Processes of heat transfer Heat is transferred by the processes of conduction, convection and radiation. As conduction requires contact between heated elements, this can be generally disregarded as an agent of fire spread through the discontinuous fuels of a natural environment. Radiation – the transfer of thermal infra red energy has been accepted by some to be the main mechanism for fire spread (e.g. Rothermel 1972, Albini 1985), however this has been questioned by others who have pointed out a stronger role for convection – the movement of hot gases, especially in discontinuous fuels (e.g. Van Wagner 1967, Steward 1971, Frandsen and Andrews 1979, Nelson and Adkins 1986, Cheney 1990, Weber 1990b, Weber and deMestre 1991). Burrows (1999a,b) delineated the roles of these two processes in surface fuel beds, concluding that radiation was the mechanism causing fire spread in backing fires, but that convection became dominant when wind or slope permitted the flame to lean forward over the fuel bed. Both Van Wagner (1967) and Weber and deMestre (1991) have produced compelling evidence that convection is the dominant heat transfer mechanism in discontinuous fuels, and this approach is adopted through this thesis when discontinuous fuels such as the leaves within a branch or the plants within a forest are considered. There is no fuel without a fire The broad agreement amongst fire behaviour models is that flame dimensions and rate of spread are the products of the amount of fuel burning simultaneously its arrangement in the fuel profile and the way in which weather factors influence this (Sullivan 2009a, 2009b). The external factors weather and terrain serve to impose limits to the availability of that fuel through their influence on moisture and the configuration of the flame. Knowledge of flame dimensions and rate of spread are then contingent upon knowledge of both weather and available fuel, defined here as the potential fuel that is consumed by the fire. Other fuels may be dry enough to burn, but if factors such as their position in relation to the fire render them inaccessible for ignition (e.g. leaves in tree crowns suspended well above a small fire), then they do not contribute as fuels to the behaviour of the fire and cannot be called ‘available fuel’. 71 The following discussion examines observations of fire behaviour and identifies mechanisms whereby flame dimensions and rate of spread have a mutually dependent relationship with available fuel. Available fuel determines fire behaviour, and fire behaviour influences available fuel. Non reciprocal mechanisms are also examined. It is apparent from these observations that it is not possible to measure the available fuel prior to burning without knowing which fuels will burn. Complexity means that fire behaviour and fuel availability are not determined by an initial trend, but that processes integral to fire spread build on one another successively to determine which fuels will ignite. As flames increase in size, they ignite fuels that were previously inaccessible and thereby increase flame height, or may ignite already available fuels more quickly thereby increasing rate of spread. This process of development is limited by the presence or absence of a critical state at each potential point of growth. If the flames are not large enough, they will not ignite the potential fuels and fire intensity will not increase. Negative feedbacks also act to retard fire development at various points. 3.2 Methods 3.21 Defining the critical state Beyond the continuous surface layer, fire propagation is as much concerned with the capacity of flame to cross the spaces between fuel elements as it is with the total quantity of fuel (Martin et al 1994, Cornelissen et al 2003), because this is what determines which strata and how much of them will be available as fuel. Buchanan (2000) said in reference to Malamud et al (1998): “What counts in the critical state are not complex details but extremely simple underlying features of geometry that control how influences can propagate.” Spaces between fuels range in scale from those between leaves and branches to spaces between plants and fuel strata; every space represents a critical point at which fire may or may not cross and the resulting fire is an emergent behaviour resulting from a multitude of individual outcomes. This is the mechanism by which complexity operates in wildland fires. The proposed approach is based upon the geometry of the fuel array and the flammability of the fuel components, where geometry determines the size of the spaces and flammability 72 determines the capacity of the fuel elements both to be ignited by an existing flame and to produce flame capable of further ignition. Flammability is a concept that has been seen to be poorly defined (Martin et al 1994, Behm et al 2004). An adequate description of flammability for a fuel element must include the three parameters ignitability, combustibility and sustainability which refer respectively to how easily the element ignites (ignition delay time and temperature of the endotherm), how well it burns (flame length), and how for how long it burns (Anderson 1970; see also Gill and Zylstra 2005). This definition is adequate when considering individual fuel elements, but it cannot be applied to an entire discontinuous fuel array because not all of the component fuel elements will necessarily be exposed to a significant heat source due to their arrangement. A plant for instance may score well on all three aspects, but only part of the plant may be ignited. Martin et al (1994) proposed a fourth aspect of flammability to accommodate this, which he termed consumability. Consumability refers to the amount of fuel which may be consumed, as limited by the ability of fire to cross the spaces in discontinuous fuels. When a fuel is burning, its combustibility demonstrated by the length of flame it produces (Gill et al 1978, Gill and Zylstra 2005) produces a temperature field oriented along the plume of buoyant convection. Longer flames raise the temperature to a greater distance (Weber et al 1995) and are therefore capable of igniting fuels across greater spaces, but this also depends upon the ignitability of those fuels. Ignitability has two components – the minimum temperature of piloted ignition or endotherm P (Philpot 1970) and the time taken to ignite or ignition delay time (Montgomery and Cheo 1971, Trabaud 1976, Xanthopoulos and Wakimoto 1993, Gill and Moore 1996). Whether P is reached or not depends upon what temperature the potential fuel is being exposed to. The ignition delay time however is a component of both temperature and time. Temperature influences the length of (Xanthopoulos and Wakimoto 1993) so both and fuel spacing are relevant here as for P, but for to be exceeded the isotherm must be maintained for a sufficient period of time. As ignition delay time relates to the time taken for water to be evaporated from the fuel allowing it to reach its endotherm, it is also dependent upon the surface area to volume ratio of the fuel and its moisture content (Montgomery and Cheo 1971, Trabaud 1976, Xanthopoulos and Wakimoto 1993, Gill and Moore 1996, Pellizzaro et 73 al 2007). The conduction properties of the fuel probably also exert some effect, however the differences from one species to another are probably minor. The duration of flaming tr is a measure of a fuel’s sustainability (Gill and Zylstra 2005). Just as flame length determines to what distance an isotherm exceeding P will be extended as well as influencing the length of , flame duration determines whether will be exceeded under those conditions. The ignition isotherm produced by a flame of length producing a temperature of T at distance P is therefore the point at which this temperature (T P ) exceeds the temperature of the endotherm P and the residence time of the flame tr exceeds ignition delay time ; that is: TP P, tr Equation 3.1. This is the central process of fire propagation at all scales, the critical decision points where potential fuels become available fuels and by which flammability (P, , and tr) interacts with fuel geometry (P ). The conceptual model is a system constructed around this concept using sub-models to calculate each of these parameters for the given fuel array and the temperature fields produced. 3.22 Fire growth patterns The following discussion describes a number of processes that occur and factors to be considered in the development of a fire. Some of these processes can be captured with submodels, whereas others are emergent behaviours resulting from the process just described. 3.221 Fuel ladder McArthur (1967) described fires as accelerating from the point of ignition until the point of full development. The acceleration does not follow a smooth line, but develops in steps with sudden increases where new strata become available as fuel. The description by McArthur was specific to the acceleration of fire from the point of ignition until it reached a quasiequilibrium point in its rate of spread; however, it illustrates the fact that changes such as fire escalation into higher strata result in stepwise changes to fire behaviour. The same 74 observation underlies the theory of crown fire modelling (e.g. Van Wagner 1976, Rothermel 1991, Cruz et al 2002), where the transition from a surface fire to a “fully fledged” crown fire can result in “at least double” the rate of spread (Cruz et al. 2005). To clarify terms - flame length refers to the longitudinal dimension of the flame itself and flame height refers to the height above a level surface of the tip of a flame which may or may not be originate at that surface. The ignition of higher strata increases the flame length and intensity (by definition) of the fire, but this in itself increases the likelihood that even higher strata will ignite. There is a reciprocal dependence between flame height or length and the amount and arrangement of fuel available; this positive feedback will be referred to as the fuel ladder effect. 3.222 Spread acceleration The flame depth or horizontal distance of fuels burning perpendicular to the fire front is determined by the rate of spread of the fire and the flame residence time of the fuels. As flame depth is equal to rate of spread multiplied by the residence time (Rothermel and Deeming 1980), greater rates of spread produce greater flame depths if residence time is constant. The total volume of a stratum that is burning is equal to the width of the fire front (x-axis) multiplied by the flame depth (y-axis) and the vertical penetration of the fuel by the flames (z-axis). Increases in rate of spread therefore cause increases in available fuel, but as rate of spread is dependent upon available fuel the reverse may also be true so that this relationship can be a positive feedback (Figure 3.1). Flame depth Flame depth Figure 3.1. Spread acceleration. The larger and faster flame has a greater flame depth than the smaller flame, but the greater flame depth means that more fuel is contributing to the flame; i.e. a positive feedback is apparent. 75 3.223 Head development The size of flames determines the heat flux from them and therefore the rate of spread. Spread does not occur only in the forward dimension, however; fire is able to propagate in any direction, increasing the amount of the stratum that is burning and thereby the total quantity of available fuel. In addition to spread acceleration which is development in the y dimension, development of the head fire occurs by widening the fire front and deepening the vertical profile of fuel that is burning (x and z dimensions). As these dimensions represent the area of fire facing the fuels ahead of the fire, they potentially increase the rate of spread due to increased flame area and thereby increased radiation (Wotton et al 1999, Knight and Sullivan 2003). As the head fire develops in width and vertical penetration of the fuel strata, the rate of spread is increased. 3.224 Flame merging Head development and spread acceleration increase the total volume of fuel burning in the head fire. However, the relationship is limited by the fact that all heated gases across a fire front do not necessarily coalesce into a single flame. Rather, flames from surrounding fuels are drawn together to produce larger flames within a limited field defined by the dimensions of the flames; that is, larger flames are able to merge with other flames within a wider radius. The two processes that lead to this are the blocking of air entrainment by neighbouring flames and heat feedback (Thomas, 1963, Thomas et al 1965, Huffman et al 1967, Steward 1970, Chigier and Apak 1975, Tewarson 1980, Gill 1990, Heskestad 1998, Weng et al 2004, Liu et al 2009). Whereas blocked air entrainment increases flame length by blocking oxygen supply and thereby slowing the rate of combustion of particles within the plume, heat feedback refers to accelerated pyrolysis of neighbouring fuels due to increased radiative heat transfer from the larger flame. As the effect of flame merging is to increase flame length and therefore T P , the critical state is more readily achieved in all directions and both rate of spread and vertical development are accelerated. In this way flame merging influences fire intensity in a positive feedback, but also imposes limits by defining the volume of fuels that are contributing to the one flame. Flame merging, head development and spread acceleration are three positive feedbacks that interact to determine the quasi-steady rate of spread for a fuel stratum. Wotton et al. (1999) found that surface fire spread rates were affected by the width of the fire front, so that fires continued to increase in rate of spread until an optimal head width was reached. Cheney et al. 76 (1993) and Gould et al (2007a, p5) described the same phenomenon in higher intensity fires for both grasslands and forests. It is likely that in higher intensity fires other atmospheric effects also come into play. 3.225 Heat feedback Flame depth Df is determined by both rate of spread R and flame residence time tr (Rothermel and Deeming 1980) according to: D f = tr × R Equation 3.2. As fuels are heated by adjacent flames, the effect of heat feedback acts to increase the rate of combustion and thereby decrease tr (Liu 2009). Reducing tr thereby reduces Df which is the dimension y on fig. 3.1; thereby reducing the amount of fuel that is potentially contributing to the flame. As discussed earlier, heat feedback can also contribute to positive feedback via improved flame merging; in either case the size of the effect may be negligible in most situations as it only affects fuels in close proximity to the flame (Liu 2009). 3.226 Flame angle damping Flame length dictates the vertical flux of the plume, which in conjunction with the horizontal flux provided by the wind dictates the flame angle (Van Wagner 1973). Increased fuel availability produces longer flames which are thereby more upright, but more upright flames produce less of a forward heat flux as the primary heat transfer via convection (the ‘plume pathway’ Pp or central trajectory of the heated plume) is directed above the stratum rather than along or through it. In this way, fuel availability interacts with wind speed to determine flame angle, which in turn impacts on fuel availability within the stratum as a negative feedback. Flame angle damping was observed by Burrows (1999) for Eucalyptus marginata forest and a threshold wind speed was measured for that fuel array (Gould et al 2007a, p.91). This threshold will vary in other arrays with different plant spacing and sizes. Weber (1990b) for instance found that in a laboratory study of vertically mounted fuel elements the capacity for flame to spread across spaces increased with taller fuels. These experiments were conducted 77 under still conditions so that convective heat transfer relied upon turbulent diffusion alone; where wind is present and flame angles are tilted flame angle damping will also be affected by other flame angle constraints as described below. 3.227 Optimal Angle Flame angle has a significant impact on rate of spread when fire is burning through discontinuous fuels such as shrubs or isolated tussocks. In such cases, the ability of the flame to propagate from one plant to the next is dependent upon the spacing between the plants, the length of flame that can be produced by an individual plant and the flame angle (Gill 1990, Burrows and Van Didden 1991, Bradstock and Gill 1993). The flame from an individual plant is determined primarily by the quantity of the plant burning at any one time (Etlinger and Beall 2004). Figure 3.2 shows a simplified upright rectangular plant intersected by two different plume pathways determined by the flame angle. The pathway marked as ‘a’ allows maximum fuel availability, whereas the pathway marked as ‘b’ allows only a small crosssection of the plant to burn at a time. While the remainder of the plant may subsequently burn, it is not part of the initially available fuel and therefore the flame length produced by this pathway will be shorter than that produced by the more upright flame. Under still conditions, flames are vertical and convective heat transfer a passes above the fuel stratum. Using laser interferometry, Weber and deMestre (1990,1991) demonstrated that even in continuous fuel beds, fire spread occurs via buoyant convection, so spread will not occur b through a discontinuous stratum under still conditions unless the fuels Figure 3.2. The are sufficiently close for turbulent diffusion to spread the convective effect of flame heat to adjacent plants. For spread to occur in a discontinuous bed angle on available separated beyond this distance, the flame must be tilted sufficiently for fuel within an convective heat transfer to pass through adjacent plants, and the length upright rectangular plant. of the flame must be sufficient to raise the temperature at those plants to their required ignition isotherms (Gill 1990, Burrows and Van Didden 1991, Bradstock and Gill 1993, Cheney et al 1998). The limits for flame connection so that one burning plant can ignite its neighbour can be derived geometrically to define a range of flame angles that will permit connection, expressed as: 78 H u − Oy tan −1 S f − Ox ≥ (θ f − θ g ) ≥ tan −1 H b − Oy S f − Ox Equation 3.3 Where Hu is the height of the Hu top of the potential fuel, Hb is the height of the base of the Hu potential fuel, Sf is the distance Hb Ox Hb from one plant centre to the f Oy next, Oy is the height of the Sf flame origin above the base of the plant in which it is burning, g Figure 3.3. Dimensions for determining flame connection. of the base of the plant that is burning, f Ox is the distance to the right is the flame angle and g is the slope (Figure 3.3). When the flame angle is outside of these limits the plume pathway Pp is equal to 0. As wind speed increases beyond the point where the upper flame angle limit is satisfied, the flame length from individual plants will reach a maximum determined by the geometry of the plant and the flame angle. After this point flame length will begin to decrease and may reach the point where insufficient flame is produced to ignite adjacent plants and contribute to the rate of spread. Alternatively, the flame angle may become so small that the plume passes beneath the next plant. This pattern may be referred to as the ‘optimal angle’ concept, and the reduction in fire intensity outside of the optimal angle may be referred to as ‘optimal angle damping’. McArthur (1969) examined rates of spread from four grassfires and compared these to wind speed. Rates of spread consistently rose until a wind velocity of approximately 13 ms-1 was reached, beyond which spread rates dropped at a slightly faster rate. Such behaviour is consistent with the optimal angle concept, but was modelled with an exponential function by McArthur which overrode the drop in velocity beyond the optimal angle. Since that time, Cheney et al (1998) have argued that the drop in rate of spread for the fires under higher wind conditions was due to this fire burning through “both dry eucalypt forest and open grasslands and across quite divided terrain”, rendering the data unrepresentative of grassland fuels. That 79 being the case, the data cannot properly be compared to the grass fire data at lower wind speeds, however a strong reduction in rate of spread with increasing wind speed is apparent for that one fire, and this may still be an observation of optimal angle damping. Where flame is propagating through a stratum, the vertical depth of the fuel bed adds an additional level of complexity. Figure 3.4 shows the effect of flame angle on the plume pathway Pp determining available fuel for the stratum. The plume pathway through the stratum is related to the depth of the fuel bed Df and the flame angle f and tends toward infinity at low flame angles according to: Pp = Df sin (θ f ) Equation 3.4 Figure 3.5 shows the effect of the same flame angle on a deeper fuel bed. The deeper fuel bed allows a longer plume pathway through the stratum, creating a greater flame depth and quantity of fuels burning simultaneously. We can assume that all fuels burning along the plume pathway will contribute to the same flame as all gases have been entrained along the same line. Pp y Figure 3.4. Effect of flame angle on flame depth for a shallow fuel bed Pp y Figure 3.5. Effect of flame angle on flame depth for a deeper fuel bed 80 Where more than one discontinuous stratum is present, the effect is amplified. The shortest distance between two strata separated by distance h results from a vertical plume. A vertical plume, however, can only occur in still conditions, and as discussed will produce the shortest plume pathway through the stratum and therefore the shortest flame length. The introduction of wind or slope will tilt the flame relative to the stratum, increasing Pp and thereby the flame dimensions and the distance to T P but also increasing P to the upper stratum by the sine of the flame angle according to: Pα = h sin (θ f ) Equation 3.5 Ignition of the upper stratum will occur if the distance to the ignition isotherm exceeds P . As both Pp and P change with flame angle along a sine curve, their changes will occur at the same rate. If flame length is proportional to Pp and temperature is proportional to flame length, flame angle cannot affect the ignition isotherm because the separation will change at the same rate as the temperature. This is not the case however, as the flame length produced by the amount of fuels given by Pp is subject to the effects of air entrainment and therefore flame merging, which has a power relationship (Gill 1990). When optimal angle and fuel ladder effects are combined, a small variation in wind speed or slope affecting the angle of the flame relative to the stratum may result in the transformation of a surface fire to a crown fire. 3.228 Non linearity of fuel contributions The contribution of fuels to fire behaviour is not linear. In a linear system ignition of higher fuel strata will display additivity as it is indeed reported to do by McArthur (1967): “As the fuel quantity doubles so will rate of spread…” Field data such as that underlying Gould et al (2007a, p91), however, have shown that this is not the case. The rate of spread equation presented in the report is: 81 Rad = 30 + 3.102(u10 − u t ) 0.904 exp(0.279S fhs + 0.611NS fhs + 0.013NS h ) Equation 3.6 Where Rad is the rate of spread adjusted to 7% fine fuel moisture and 0o slope, u10 is the wind speed (km/h) at 10m in the open, ut is a threshold wind speed for fire spread estimated at 5km/h, Sfhs is the “surface fuel hazard score”, NSfhs is the “near-surface fuel hazard score” and NSh is the near surface fuel height (cm). Fuel hazard scores for the two strata do not relate to the same quantities of fuel, rather the equivalent score in surface fuel represents four times as much fuel it does for the near surface fuel layer (Gould et al. 2007b pp9-11) In addition, near surface fuels are given an added weighting based upon their depth (NSh). Without this additional weighting, near surface fuels are given 0.611 × 4 or 8.76 times the influence on rate of spread per weight that surface fuels 0.279 have. If however near surface fuel is used with a fuel depth of 50cm (the maximum value given on the tables in Gould et al 2007b pp22-25), the weighting in relation to surface fuels is now 0.013 × 50 + 0.611 × 4 or 18.08 times the influence on rate of spread per weight that 0.279 surface fuels have. Although this is the same site that McArthur studied to illustrate additivity in fuels, these higher intensity experiments have shown that doubling the fuel load by adding an equivalent amount of near surface fuel would not double rate of spread, but increase it between 8.76 and 18.08 times as much. Gould et al have demonstrated that the contribution of fuels to fire behaviour has more to do with their location than it does their weight; that contrary to McArthur’s assertion, the contribution of fuel load is not linear. While it is reasonable to assume as Gould et al (2007a) have done that the proximity of near surface fuels to the ground means that they will ignite from a fire burning in surface fuels, optimal angle and fuel ladder considerations mean that the same cannot be said of higher fuel strata. Elevated, midstorey and canopy fuels will not ignite unless the critical state defined in equation 3.1 is satisfied; both for that stratum and for the fuels beneath it. The non linearity of fuels when coupled with this means that the behaviour of fires burning in fuels beyond 82 surface and perhaps near surface will become increasingly difficult to predict from an initial trend. 3.229 Summary of fire growth patterns The presence of complexity in fire behaviour means that fire growth is successive, that each event depends upon those preceding it (Buchanan 2000) and the final fire characteristics are emergent behaviours which cannot be predicted from an initial trend alone or a central controller such as a pre-defined fuel load. Negative feedbacks impose limits on fire growth, but where such limits are finite a point will be reached where either the feedback is overcome or a threshold is crossed in the process and sudden change will occur in the behaviour of the fire. Such threshold changes explain the stepwise nature of fire growth and may account for significant variability in fire behaviour. By themselves, the feedbacks described do not necessarily indicate complexity in fire behaviour; complexity derives from the dependence of fire development on its history. Satisfaction of equation 3.1 at any point where separation P between flame and potential fuel is present causes fire growth, and the resulting change in flame length and angle either presents new possibilities for growth such as through the fuel ladder effect, or removes existing possibilities such as through flame angle damping. The location of fuels has a far greater influence on fire behaviour than the quantity of fuel. Coupled with the dependence of fire growth on successively satisfying the critical state defined in equation 3.1, this means that high intensity fire cannot be modelled based upon an initial trend because doing so assumes: 1. ignition of new fuels without knowing whether the required successive critical states have been reached, and 2. that all fuels beyond those measured will contribute an equal part to fire behaviour. As the mechanisms described are emergent behaviours, it is not necessary or advised that they should be dealt with using individual models. Rather it is proposed that these behaviours will emerge as the process of fire spread is modelled as a complex system. A possible exception is the phenomenon of flame merging, as this relies on computation-intensive 83 atmospheric interactions but has also been described by simpler empirical models (e.g. Gill 1990). The following discussion describes a conceptual model relying on the flammability properties of the leaves and the geometry of the fuels. This model is then parameterised in the following chapters until a final model is given in chapter six as well as on the attached CD. The model is given some validation and sensitivity analysis in chapter seven. 3.23 A conceptual model of complex fire behaviour The following section describes a model framework to accommodate the complexities described. The intention is not to produce a merely theoretical model but the basis for an operational model which can be constructed using existing equations for specific facets of fire behaviour as sub-models. Because a complex systems model arrives at an outcome based upon successive calculations that build upon each other, any errors present in each iteration can at times be compounded. If the sub-models that compose it are inaccurate, the final model may be significantly wrong. The advantage of the framework, however, is its transparency. Any logical errors in the conceptual model can be addressed and relevant sub-models can be replaced with better equations as these are developed. In this way, the model framework provides research priorities for component aspects of fire behaviour as well as a means of identifying the precision required for each area by demonstrating how the sub-model will affect the predicted fire behaviour as a whole. 3.231 Assumptions used in the model Two main assumptions are used in the model to provide a level of abstraction and thereby produce faster calculations. These are: Fire spread through discontinuous fuels occurs via convective heat transfer. Assumption 3.1 84 Flames and the resulting plumes can be simplified to a central vector without significant loss of accuracy. The effects of turbulent diffusion and flame depth can be approximated using simple means rather than using complex fluid dynamics. Assumption 3.2 All assumptions are listed in Appendix II along with the equations used for the model, so that future advances in the model can specifically address them. Errors in the model introduced by these assumptions can be identified in the process of model validation. Disregarding one of the two dominant means of heat transfer (radiation and convection) is not a unique approach; the widely used fire behaviour model developed by Rothermel (1972) for instance is based entirely on radiative heat transfer. That decision was not made due to evidence that convective transfer was irrelevant, but by an implicit assumption. Similarly, Albini (1985) used only radiation as the propagating heat flux in his fire spread model. The basis for using convective rather than radiative heat transfer is defensible using both conceptual arguments and empirical observations. Nelson and Adkins (1986) for instance reported that the key mechanism for fire spread was the “horizontal displacement of flame at the fuel surface by a convection mechanism related to windspeed”. Based on extensive laboratory studies of fires in surface fuels, Burrows (1999a) concurred with this. Van Wagner (1967) compared previous laboratory studies of fire spread in wooden cribs with spread in pine needle beds, and concluded that convective heat transfer was more important in the crib fires but radiation the governing factor in the pine needle fires. Thomas et al (1961), Fons et al (1962) and Thomas (1970) however argued that the emissivity of flames was not sufficient to propagate fire via radiation and proposed that heat radiated through the fuel bed itself was more important. Weber (1990a) examined the proposed mechanisms for short-range heat transfer in 13 different physical models and concluded that the assumed role of radiation did not adequately explain the speed at which temperatures rose in close proximity to the flame, arguing that buoyant convection provided greater promise as an explanatory process (Weber and de Mestre 1990). Weber and de Mestre (1991) used laser interferometry to study the heat transfer processes directly and found that the spread of fire within a fuel bed was via buoyant convection rather than radiation. In the context of flames extending above the fuel bed, experiments by de Mestre et al (1985) demonstrated that when radiation from flames was excluded with a radiation shield, the rate of spread was almost unaffected. Weber (1990b) 85 further demonstrated that flame propagation between components of a vertically mounted, discontinuous fuel array was well explained by convective heat transfer. This is of particular relevance to the propagation of flames between discontinuous fuels such as plants. In the context of this evidence, assumption 3.1 is considered a sound starting point. Where fuels are continuous such as in surface fuel beds, backing fires are accepted to spread via radiation and will be modelled using existing methods suited to that purpose. Due to the modular nature of the model, it is possible to include the effects of radiation as a transfer mechanism between discontinuous fuels at a later time if further research establishes a role for it in this area. Assumption 3.2 presents some significant issues to be overcome, as turbulent diffusion results in the horizontal dispersion of heat from the central vector as well as producing non-linear pathways of convective heat transfer. Even when a flame has minimal depth, shear stresses at the boundary of the convective stream and the surrounding air result in the dispersion of heat through eddies (Figure 3.6). Turbulent diffusion within the flame and plume can be modelled using the tools of fluid dynamics, particularly the Figure 3.6. Turbulent diffusion evidenced by eddies in a flame Navier-Stokes equations (Constantin and Foias 1988, Baum et al 1994). The necessary gridded three-dimensional calculations required for this however are extraordinarily computation- intensive and would negate the possibility of this model’s operational use as they have for the other physical models in which they have been employed (Sullivan 2009a). Without an analysis of the fluid dynamics however, the model must compensate for the lack of dispersion using other means such as empirical sub-models or further assumptions. As stated earlier, the validity of this approach will be shown in the model validation, so in this sense assumption 3.2 may be treated as a hypothesis. The effect of treating flames as a vector rather than something with depth such as an inclined triangle may be less of an issue. Assumption 3.2 amounts to a hypothesis that flame propagation occurs at the front of the flame, so if the effect of flame depth on emissivity is discounted as unimportant through assumption 3.1, the trailing depth of the flame has little effect on its propagating front. The effects on convection may be another issue however, as 86 the depth of a flame potentially affects the angle of the flame and the vertical transfer of convective heat by providing a wider column of air and therefore a greater period of time before the heat is fully dispersed into the surrounding air. This issue may possibly be addressed through empirical modelling, with appropriate models used for flame angle and plume temperature as they become available. 3.232 Components of the model Surface fire spread At the lowest levels of intensity when fire is burning only one shallow continuous stratum such as leaf litter, the level of complexity is low because the separation P between fuels is not present and the critical state is therefore continuously satisfied. The limits to growth will be defined by the flammability properties of the fuels (affected by factors such as moisture content) and the processes of flame merging and heat feedback. The range of fire intensities possible will be minimal without the inclusion of higher strata and the predictability of the fire should be comparatively high due to minimal involvement of optimal angle feedback and the absence of flame angle damping and fuel ladder effects. This is apparent from studies of low intensity fire in leaf litter such as those of McArthur (1967). Although no statistics were presented or analysed, Figure 5 in Leaflet 107 displays the data for nine simultaneous experimental fires in E. marginata forest. Visual appraisal of the graph suggests an excellent fit of rate of spread against surface fuel load, although the leaflet suggests that the complete number of experimental fires conducted by the author was in the hundreds and it is unknown how well all of these fit to the model. Because of the more predictable nature of fires burning in surface litter alone, it is feasible to use a simple empirical model for this stratum, provided the model explicitly differentiates between surface litter and other strata. Complexity need only be considered in reference to fire propagation to and within higher fuel strata. This applies most notably to discontinuous strata such as tussock forming grasses, shrubs or trees. Ignition of plants The separation between the flame base and the potential fuel P represents the limitation imposed on fire spread by the fuel geometry and therefore along with flammability it defines where the critical state will be met in the fuel array. Pp and P together define the role of fuel 87 arrangement and geometry and provide the inputs necessary to calculate consumability (Martin et al 1994). P can be found for different situations as follows. The dimensions for ignition of a plant base from a fire burning in surface fuels are shown in Figure 5, and the length of P can be found using the sine rule from: (O Pα = + δ )sin (90 + θ g ) y sin (θ f − θ g ) Equation 3.7 Where Oy is the height above the plant stem base at which the plume is intersecting the plant, and is the height of this point above the ground minus Oy. For backing flames, Equation 3.7 becomes: (O Pα = y + δ )sin (90 − θ g ) [ sin (180 − θ f ) + θ g ] Equation 3.8 Oy Plant stem P f 1 g Figure 3.6. Schematic diagram showing the dimensions and angles involved in calculating P for ignition of a plant base. The plant is shown as a thick light grey line, the plume as a thick dark grey line and the slope as a thick black line. Oy is the height above the plant stem base at which the plume is heating the plant, and is the height of this point above the ground minus Oy. 88 Together with the characteristics of the flame that define the decrease in temperature along the convective plume, the distance P can be used to find T P and thereby examine whether ignition of the lower leaves has occurred according to equation 3.1. After the initial leaves that caught fire have burnt out to a distance P into the plant along the line of the plume, the values Ox and Oy are then adjusted to Oxn and Oyn by finding the horizontal and vertical dimensions of P as below. The subscript n is the number of times Ox and Oy have been examined in the plant since ignition, for example if we examine them at one second intervals, then n will equal 5 after the fifth second. At the same time, the fuels that have been ignited are now burning within the plant, and P measured from the flame produced by these leaves is smaller than the external value. Two values of P are present along with two different flame lengths, so T P needs to be calculated for two different possibilities. The two possibilities are: 1) The external flame length. 2) The plant flame. This is the length of the newly ignited flame; P is measured from the origin of this flame with the coordinates Ox2 and Oy2 measured laterally to the right and vertically respectively from the point where the plant stem connects with the ground. For example, an origin that is 30cm to the left of the base of the stem and 50cm above it has the coordinates Ox2 = -30cm, Oy2 = 50cm. The diagonal P can be converted to values of x and y as follows: O xn = O xn −1 + Pλ cos (θ f ) Equation 3.9 O yn = O yn −1 + Pλ sin (θ f ) Equation 3.10 In the first case described above, P can be found for fire fronts using the sine rule as follows: 89 Pα = (O + δ )sin (90 + θ g ) yn sin (θ f − θ g ) Equation 3.11 And for backing fires: Pα = (O yn [ + δ )sin (90 − θ g ) sin (180 − θ f ) + θ g ] Equation 3.12 In the second case, P is equal to: Pα = (O − O y ) + (Oxn − Ox ) 2 yn 2 Equation 3.13 Equation 3.5 described P for the plume of a burning stratum heating a stratum above separated by a vertical distance h. If the flame origin in the burning stratum is renamed Osx, Osy and the point being heated in the upper stratum is designated by the coordinates Oxn, Oyn, then the value of P at any point in the upper stratum on a slope of Pα = g is: O yn − Osy sin (θ f − θ g ) Equation 3.14 Flame length As discussed earlier, fuel contribution is not additive; rather fuels contribute differing amounts based on their position in the array and flame length is not connected purely to the weight of fuels burning but is also influenced by factors such as location and the chemistry and dimensions of the fuel element (Zylstra 2006b). 90 A flame is essentially a collection of oxidising soot particles heated to incandescence (Lyons 1985); the length of the flame is the distance those particles can travel before they burn out. While adding more particles to the flame by burning more fuel makes the flame larger, so does reducing the quantity of oxygen available. In this way, flame length is dictated by the fuel-air mix. This phenomenon is described by the concept of flame merging (Thomas, 1963, Thomas et al 1965, Huffman et al 1967, Steward 1970, Chigier and Apak 1975, Tewarson 1980, Gill 1990, Heskestad 1998, Weng et al 2004, Liu et al 2009). Flame merging theory relates flame length to the number of component flame sources able to contribute to the one flame rather than the weight of these. The magnitude of the contribution of each is determined from the length of flame produced by each individually. This concept is incomplete in the context of flames burning in a three dimensional matrix such as a plant, however if it is accompanied by further sub-models or explicit assumptions to cover the knowledge gaps it provides a starting position to model this aspect. To find flame length then for a branch, plant, a live fuel stratum or an entire fuel array requires knowledge of for each individual fuel element and the number N of elements producing flames capable of merging together. can be found using empirical or physical models for individual elements. When considering merging from lateral sources, N can be found from models describing the radius within which flames will merge (e.g. Gill 1990). For longitudinal sources such as the number of elements along a plume pathway, all flames can be considered to merge, so the essential information is the length of the plume pathway. This in turn requires knowledge of tr for the component elements, and rate of spread through that fuel. tr can be found from existing models such as McArthur (1967), Anderson (1969), Cheney et al (1990), Burrows (2001) or Zylstra (2006b) for different types of fuel elements. Some of these models are not specific as to whether the model applies to elements burning in the presence of a heat source such as a ground fire beneath, or in the absence of such a source such as in the case of an independent crown fire. Unless the difference is shown to be insignificant, a precautionary approach is to assume that elements with a large moisture component such as fresh leaves will self extinguish much quicker in the absence of an external heat source than they might otherwise. 91 Rate of spread Flame length and rate of spread are interdependent. The fuel ladder effect coupled with spread acceleration, head development, flame merging and heat feedback means that flame length determines available fuel, which in turn may accelerate rate of spread or slow it through optimal angle effects. This interdependence means that calculation of flame length and rate of spread must be done iteratively. The mechanism proposed is to use time steps such as one second intervals, calculating the number of fuel elements ignited in that period. In each successive period, the number of new ignitions is added to the existing total, and the number of elements extinguishing as they pass tr is subtracted. The resulting flame length and angle is recalculated along with the new coordinates for the flame base and the process repeated from the new coordinates until the flame pathway reaches the edge of the plant or stratum. Equation 3.1 can be adjusted to take into account the two new terms introduced here, time elapsed since initial ignition te and calculation interval t. TP P, (tr – t) , (tr – te) t Equation 3.15 The spread of fire through a single plant may or may not affect the overall rate of spread, but more significantly it does affect the flame length and angle produced by that plant. These in turn can lead to fire spread through the stratum if the flame angle is within the optimal range defined at equation 3.3 (flame connection) and the ignition isotherm extends into the potential fuel as defined at equation 3.15. Rate of spread R for a stratum can be calculated from the depth of fuel ignited in the stratum over a given period of time. Fuels will be ignited along the plume pathway at an angle to the slope. The rate of spread parallel to the slope is therefore equal to the length of the plume pathway ignited Ppi multiplied by the cosine of the angle between the flame and the slope, all divided by the time taken for the measurement tm: 92 R= Ppi cos(θ f − θ g ) tm Equation 3.16 This is illustrated in Figure 3.7. The influence of both wind speed and slope on Ppi rate of spread can be seen immediately, where increasing wind speeds that reduce f g f or increasing slopes that increase both reduce the value of ( Figure 3.7. Dimensions for calculating rate of spread through a stratum f – g) g so that rate of spread becomes greater. When the rate of spread for a stratum exceeds that of those below it, there is potential for independent spread. The factors that limit this are flame depth and angle. Flame depth is determined by rate of spread and flame residence time. Weber (1990c) demonstrated that sustainability is strongly dependent upon moisture and that when fuels were placed so that convective heat transfer was maximised, some of this effect was overcome. This suggests that flame residence time will be significantly shortened if the independent fire burns ahead of the ground fire and its supply of convective heat, so flame depth can only be maintained if the rate of spread is faster than the rate at which the trailing edge of the crown fire is going out. If the flame angle is not parallel to the slope (flame attachment), the plume will pass through the top of the crowns and there will be no forward convective heat transfer. Optimal angle and flame angle damping considerations are pronounced in the higher strata as an increase in wind speed affects flame angle in all strata below it. For example, the flame length from elevated fuels may be sufficient to support an active crown fire, but the increase in wind speed necessary to approximately align the crown fire to the slope may also tilt the ground fire so that its flames are unable to ignite the canopy and start such a fire. For these reasons, greater rates of spread in the crown than those in the lower strata may produce short burst of crown fire ahead of the main, but may not affect the overall rate of spread unless wind velocity is sufficiently high to reduce the flame angle or the slope is sufficiently steep to minimise the difference between slope and flame angle. 93 It is possible that by spreading the convective heat outward from its central axis, turbulent diffusion of the flame will allow independent fire spread when flame angle is slightly greater than slope. It is also possible that when the spacing between strata is small relative to the flame length, that radiation from the flame will act to pre-heat lower fuels and facilitate faster rate of spread in lower strata. Removal of the forest canopy even by a temporary independent crown fire moving a limited distance ahead of the surface fire will also allow higher wind speeds in the lower strata (Gill et al 1996). Potentially, this may mean that an independent crown fire could facilitate faster fire spread in the lower strata, so that even if the independent spread cannot be maintained due to the difference between flame angle and slope, the surface fire may still spread at a speed close to that of the independent fire. Fire progress through the fuel The progress of flame through a fuel array refers both to the development of the flame length, and to progress of the flame origin. When a given depth of fuel within a plant has been ignited (Ppi), the fuels burn for a period equal to tr specific to those leaves. After this period has passed, the flame origin Oxn, Oyn is adjusted into the plant for a distance equal to Ppi at the angle of the flame f. The new values of Oxn and Oyn are: O xn = O xn −1 + Ppi cos θ f Equation 3.17 O yn = O yn −1 + Ppi sin θ f Equation 3.18 Fire spreading through a single plant is a point measurement and as the plant can be considered to be vertical, flame angle is not affected by slope. When fire is spreading through a stratum however, the flame angle is reduced by the slope so that equations 3.18 and 3.19 are adjusted to: O xn = O xn −1 + Ppi cos (θ f − θ g ) Equation 3.19 94 O yn = O yn −1 + Ppi sin (θ f − θ g ) Equation 3.20 3.3 The Model As described earlier, fire spread occurs in successive time steps which build upon each other. Fire spread for a single time step within a plant is outlined in the flowchart shown in Figure 3.8. All input processes feed into the decision at the centre of the model; if these criteria are not satisfied fire spread in that direction will not occur. If they are satisfied, the flame length changes and the new length is examined again by the same process, also taking into account the new flame origin coordinates. Each cell in the chart represents a sub-model, except for the grey cells, which show model outputs. The submodels are not described in this chapter but will be explored in the following chapters. The output of a single time step is a set of coordinates for the new flame origin and a new flame length . The calculation process is also described in steps in the text next to the chart. Pp P TP 3.31 Steps of ignition 1. Calculate flame angle f. 2. Designate a point of origin in the plant TP (Ox, Oy) 3. Oy f P, (tr – t) T , (tr – te) t No Growth F Calculate the length of the plume pathway through the plant from the flame angle, point of origin and dimensions of the plant tr > t Ignition depth Ppi No. leaves ignited Leaf No. leaves extinguished using equation 3.4 4. or 3.13 depending on the situation 5. Flame merging Calculate P using equation 3.7, 3.8, 3.11 Oxn, Oyn Calculate the temperature within the plume at point P Figure 3.8. Model for a single time-step ignition of a plant. All inputs lead to the critical state test, highlighted with a bold border. 95 6. Calculate the ignition delay time from T P . 7. Perform the critical state test (equation 3.15); if no ignition occurs there is no growth in the flame dimensions. 8. If ignition occurs, the ignition depth Ppi is the distance P 9. The flame origin is unchanged if no leaves have extinguished in that time step, otherwise it can be calculated using the depth of extinction as a vector stretching from the previous origin at angle f 10. Calculate the number of leaves ignited from the depth of ignition and the leaf density in the plant 11. Calculate the number of leaves extinguished from the last time step based on the flame residence time in the leaves and the time for which those leaves have been burning 12. Find the total number of leaves burning from the number already alight plus the number newly ignited, minus the number newly extinguished 13. Calculate the merged flame length from the number of leaves burning and the flame length of an individual leaf. Because of the dynamic nature of the calculations, there is no single flame length that represents a burning plant, stratum or array; rather new flame lengths are produced at each time step. The exception is if an empirical model is used for the surface litter stratum; in this case flame length is considered constant until the residence time of the fuel is exceeded. When a stratum other than the surface litter is exposed to flame from the burning strata beneath it, this external flame length is a combination of the component flames for each time step. This is not a straightforward combination; it may at times be more accurate to model the effects of individual flames as it is discussed under ‘Finding P ’. Doing so would greatly increase the number of calculations involved however, so a simplification is proposed in Figure 3.9 that assumes all flames contributing from lower strata combine into one flame, and subtracts the length of overlap and spaces between each from the total. Figure 3.10 shows the way the different model components are combined to produce flame length and rate of spread measurements for a stratum. To model an entire fuel array, this process is carried out for a plant in the stratum then repeated for the whole stratum considering the plant flame as the “external flame” rather than the flame burning from lower 96 strata. The process is then repeated for the next stratum above, updated with the new combined flame length as per Figure 3.9. 3.311 Producing the model and considerations for validating Producing the model will consist of utilising sub-models or developing new ones to fill the model cells in Figures 3.8 to 3.10. The symbols for all equations in this thesis are given in Appendix I and the equations used for the fire behaviour model are given in Appendix II. Aside from comparisons of model predictions against observed fire behaviour, a component of model validation will be to determine whether the model reproduces the emergent behaviours described earlier. 97 t s Tr ≥ t F s1 T =0 s1 = s s1 s1 + + ns1 ns1 ns1 + + overlap Tr ≥ t s1 + e1 overlap e1 + m1 overlap F s2 T =0 s2 = s s2 s2 s2 + + + ns2 ns2 ns2 + + overlap e2 overlap e2 + m2 overlap Tr ≥ t F sn = T 0 sn = s sn sn sn + + + nsn nsn nsn + + overlap en overlap en + mn overlap Figure 3.9. Model for determining external between strata for n time steps. Flame lengths for each stratum are represented by with the subscript indicating the stratum and the number of the time period. Fuel strata are: Surface = s, Near Surface = ns, Elevated = e and Midstorey = m. The surface fuel stratum maintains a constant flame length until it passes tr, but different flame lengths are calculated each time period for the higher fuel strata. 98 External t Oy f s Pp P Tr ≥ t Time step 1 TP F Ppi1 TP P, (tr – t) , (tr – te) T =0 s1 s1 = s s1 t s1 s1 + + + ns1 ns1 ns1 + + overlap Tr ≥ t e1 overlap e1 + m1 overlap F T T No Growth F =0 s2 tr > t Ignition depth Ppi s2 = s s2 External s2 + + ns2 ns2 ns2 + + e2 overlap No. leaves ignited Leaf s2 + overlap e2 + m2 overlap No. leaves extinguished Combined Flame merging Tr ≥ t Oxn, Oyn Time step 2 Combined F sn =0 T sn = s sn sn Time step model sn + + + nsn nsn nsn + + overlap en overlap en + mn overlap Ppi2 External External Combined Time step n Ppin External Combined Ppi R= Ppi cos(θ f − θ g ) tm Figure 3.10. Ignition of a plant or stratum over n time steps. The grey time step boxes indicate the Time Step Model in fig. 6 and the black boxes indicate the External model from fig. 7. The broken line before time step n indicates a discontinuity; where the 4 step process within the grey box is repeated until n time steps. 99 3.4 Discussion This chapter has demonstrated that fire behaviour is not controlled by a central process but by multiple decisions at the interface of the flame and the fuels. The absence of a central process means that flame dimensions and rate of spread are emergent behaviours – each individual decision builds on those preceding it. The individual decisions relate to the ability of fire to cross the spaces between fuels, whether they are leaves, branches, plants or fuel strata. Fire spread operates in three dimensions – forward (rate of spread), lateral (head fire width) and vertically (fuel strata contributing). In each instance spread can only occur if the flame of the burning material is angled so that convective heat is directed into the potential fuels, and if flame length is sufficient to raise the temperature at the potential fuel high enough to exceed the endotherm and lasts long enough to exceed the ignition delay time of the potential fuel. The interdependence of flame dimensions and available fuel causes a number of both positive and negative feedbacks in the system, and fuel geometry imposes limits that interact with these so that fire intensity does not always increase with weather or terrain according to a smooth trajectory. Under some circumstances, fire behaviour will remain almost unchanged in relation to the governing weather, at other times it will undergo sudden escalations or reductions in intensity with little change in the weather, or even begin to decrease in flame height or rate of spread when it might be expected to increase. In order to address these complexities, a conceptual model was described which can form the basis for an operational model. Development of this model may provide a framework whereby physical explanations for eruptive fire behaviour and the conditions that facilitate it can be explored. The following three chapters describe the submodels required for the model to operate, producing a complete model in chapter six and the attached CD. The model produced in this chapter does not rely on an assumed direct relationship between fuel quantity and flammability, but provides a framework whereby this relationship may be studied and quantified. 100 Chapter 4 Leaf Flammability Context Section 3.31 outlined the process of fire spread described in chapter three. As explained in that chapter, the model operates by calculating whether fire is capable of crossing the spaces between the fuels. This requires a combination of geometry and a physical model of ignition based on the principles of flammability. This model was provided in equation 3.1 and extended in equation 3.15; but before it can be used it requires parameterisation. The model states that ignition will occur within a time step if the flame length can extend the ignition isotherm into the potential fuels. The ignition isotherm is a temperature defined by the endotherm of the fuels, where that temperature can be maintained for long enough to exceed the ignition delay time of the fuels. All three aspects of flammability affect this. Chapter four provides an experimental study of the factors that determine flammability in leaves to provide submodels that can fit the conceptual model from chapter three. The submodels are studies of the physical characteristics of leaves – their dimensions, moisture content and aspects of their chemistry. The intention is to produce models that are applicable across species. 4.1 Introduction Given the importance of forest flammability, it is necessary that we have a clear understanding of its characteristics. In a study of historic crown-fire incidence across the 101 United States, Sturtevant et al (2004) found that management-induced increases in the density of understorey vegetation increased fuel characteristics correlated with the incidence of crown fire in some forest types, whereas it decreased them in others. In the first case, the understorey was dominated by fir species (Abies spp.), whereas in the second case it was dominated by deciduous species. Although the trend in forest structural change was the same, the different species involved produced opposite trends in flammability. To understand the change in forest flammability, it is necessary to explain why one set of species were apparently more flammable than the other. To compare the flammability of species, however, we must first define the term in parameters relevant to the field situation. To achieve this, the approach used in this thesis is study flammability from the simplest level of an individual leaf through plant flammability then to an entire forest. The models developed at each stage will be used as building blocks to construct the model for the next scale of complexity. The physically observable aspects of forest flammability are the likelihood of ignition, and once ignited, the flame dimensions and the rate at which these flames propagate. Flame dimensions affect the probability of ignition of tree crowns from lower fuel strata (Van Wagner 1976, Alexander 1998, Cruz et al 2006, Morvan 2007), therefore affecting fire spread vertically as well as horizontally. While it is often the flame length from an array of fuels or a fuel bed that is of most interest, flame length can be studied at many scales ranging from an isolated burning leaf to single branches or entire plant crowns. The model developed in this thesis and presented in the following chapters is constructed on the approach that the flammability of a forest, a branch or plant is a product in some way of its parts, and therefore to model forest flammability it is necessary to begin with the smallest units, the leaves. The flammability of a plant depends on both the characteristics of the material that compose its parts, and the architecture of these parts in the way that they influence heat conductivity through the plant (Cornelissen et al 2003). In this context, flame length is a primary consideration as flame dimensions affect radiative and convective heat fluxes (Weber 1989b, Weber et al 1995, Sullivan et al 2003, Morvan 2007), the buoyancy of the fire plume (Byram 1959) and flame contact with other fuels – all fundamental processes affecting fire spread. To 102 quantify heat conductivity through a plant and therefore its flammability as a whole, it is necessary to quantify the characteristics of the flames produced by its parts. Flammability and observable flame characteristics Anderson (1970) defined the concept of flammability by dividing it into the three components of ignitability, combustibility and sustainability. These components refer to the time taken and temperature required for ignition, the rate of combustion, and the duration of combustion, respectively. In the case of a defined stage of combustion such as the period of flaming, combustibility and sustainability are inversely proportional to each other as sustainability is the time taken (tr) for a quantity of fuel (Wf) to burn tr / Wf, and combustibility is the rate of burning or the quantity of fuel that will burn in a given time Wf / tr (Gill and Zylstra 2005). These definitions by themselves present us with some difficulty in applying the term flammability to a plant or forest. One forest may burn frequently and quickly due to its low grassy groundcover, but with only short flames. Another forest may be identical except for the presence of a shrubby understorey that facilitates the movement of flame into the tree tops and the development of crown fires. Ignitability could be the same in each, as could be the rate of burning from which we calculate combustibility and sustainability. All that differs between the two is the amount of fuel burning. To say that the two forests are equally flammable because all values of ignitability, combustibility and sustainability are equal is meaningless to a fire practitioner who is confronted by a small grassfire in one and a crown fire in the other. The same issue applies at the scale of a single plant. Etlinger and Beall (2004) found that the quantity of fuel consumed in a plant explained 89% of the rate of heat release for plants burned in laboratory conditions. Martin et al (1994) introduced a fourth characteristic of flammability, calling it ‘consumability’ or the quantity of fuel that the fire can consume. Consumability means that even if everything else in the two example forests described above is equal, the fact that the second forest has an understorey capable of conducting fire into the tree crowns means that the second forest can be considered more flammable. The additional fuel in the understorey causes the crown fuels to become available. 103 Flame length and duration can be used as measures of combustibility and sustainability respectively (Gill and Zylstra 2005). However, while sustainability relates directly to tr, the relationship of flame length to combustibility is not always straightforward. Although Gill et al (1978) found that flame length was proportional to the rate at which fuel beds burnt and could therefore be used as a measure of combustibility, this measure relates to the combustibility of the entire bed and does not necessarily describe that of the component fuel particles. The study found that drier fuel beds burnt more quickly; and as Gill and Moore (1996) established, moisture content directly affects the ignitability of leaves. The positive correlation between ignitability and combustibility is due to the fact that ignition of the complete fuel bed is not simultaneous, but faster fire spread (ignition of individual particles) caused by drier fuel results in more of the fuel burning at the same time and hence a reduction in the time taken for the entire fuel bed to burn. In this sense, increases in ignitability cause increases in consumability which in turn increase the combustibility of the whole bed. When the unit being considered is a combination of smaller units as in the case of a forest, a plant or a fuel bed, combustibility is probably influenced by a range of flammability properties of its components. In this context, the study of combustibility and sustainability at the scale of a single leaf can inform us about the flammability of larger arrays such as branches, plants and forests. Table 3.1 identified four aspects of the flammability of a leaf that are required for the model to operate. Corresponding to the three aspects of flammability which are ignitability, combustibility and sustainability, these are: temperature of ignition or P and ignition delay time as affected by different temperatures, flame length for a leaf, and flame residence time tr for a leaf. These aspects are examined below and mathematical models are constructed for each component. 4.2 Temperature of ignition Ignitability has two components – the minimum temperature of piloted ignition or endotherm P and the time taken to ignite or ignition delay time . The ignition isotherm T P is the point at which equation 3.1 is satisfied; that is where TλPα ≥ P and where t r ≥ ψ . The ability to 104 estimate the ignition isotherm requires knowledge of both of these parameters. These are discussed separately below. The temperature of ignition is the minimum temperature at which a fuel will ignite. The process of ignition can be seen in two steps. As a fuel particle is heated beyond a particular temperature threshold, compounds within the particle begin to turn to gas or volatilise – a process termed pyrolysis. Pyrolysis is an endothermic reaction, as the conversion of a compound into a gas requires a specific quantity of energy, and this point is referred to as the endotherm of the compound (Philpot 1970) and is represented by the symbol P. The second step of ignition is the point where these products begin to oxidise. In a piloted ignition, i.e. where a spark or flame is present, both steps of ignition occur at the point of pyrolysis. The second stage in an unpiloted ignition occurs when the products of pyrolysis are heated to the point where ignition occurs without a spark. As this point is marked by a release of energy it is referred to as the exotherm (Philpot 1970) and is represented by the symbol ε. Fire spread by definition requires an ignition source, so the temperature of ignition in this context refers to P. The temperature at which a leaf will ignite is not the same for all species, and Gill and Moore (1996) found that in 10% of the species they studied, not all leaves ignited at 400oC even when oven dried. Philpot (1970) found that the temperature at which leaves undergo volatilisation is related to the silica-free mineral content of the leaves. Using the techniques of thermogravimetric analysis and differential thermal analysis, he reported the temperature of the endotherm as a function of silica-free ash content according to: P = 354 – 13.9ln SFA - 2.91 (ln SFA) 2 Equation 4.1 Where P is the temperature of the endotherm (oC) and SFA is the silica-free ash content (% oven dry weight or ODW). King and Vines (1969) found that when leaves which were naturally slow to ignite were stored in an atmosphere of eucalypt vapour, they ignited far more vigorously. Although they 105 were not tested for ignitability at different temperatures, it is possible that what was being observed was a lowering of the endotherm for those leaves. It is possible therefore that the endotherm of a leaf could also be related to its content of volatile oil, although this has not been verified. 4.3 Ignition Delay Time 4.31 Introduction Xanthopoulos (1990) developed equations to determine the ‘time to ignition - temperature – moisture’ (TTM) for three species of North American conifers, later publishing these in Xanthopoulos and Wakimoto (1993). Ignition delay time was measured for the ignition of branches approximately 15cm long held at a constant temperature in the presence of a gas pilot flame. The three species Pinus ponderosa, Pseudotsuga menziesii and Pinus contorta were selected due to their wide variability in needle density and surface area: volume ratio. The equations for each species were: ψ = 291.917e (−0.00664T +0.00729 m ) Equation 4.2 for Ponderosa Pine Pinus ponderosa, ψ = 1408.922e (−0.00990T +0.00691m ) Equation 4.3 for Lodgepole Pine Pinus contorta, and ψ = 3607.697e (−0.01072T +0.00397 m ) Equation 4.4 for Douglas Fir Pseudotsuga menziesii. 106 The study of ignition delay time ψ across temperatures between 445 – 640oC and typical live moisture contents demonstrated that both temperature (T, oC) and moisture content (m, %ODW) were central to ψ for each species. The wide variability in all three coefficients for each species also demonstrated a strong variability between species as expected from the range in needle density and surface area: volume ratio. Gill and Moore (1996) examined the factors determining ignition delay time for 50 species from 19 families by placing leaves, phyllodes and cladodes of the plants in a muffle furnace heated to 400oC with a pilot source adjacent to the particle. The focus of this study was to identify those parameters that affected the ignition delay time for each species when temperature was held constant. Species were selected with a wide range of leaf dimensions, moisture, Sodium, Magnesium, Phosphorus, Sulphur and Chloride contents, although very small leaves or nanophylls typical of heathland were not measured. The study found that the only statistically significant variables used in the analysis were moisture content and the ratio of surface area to volume of the leaves, approximated in flat leaves or phyllodes by: θc = 2/d Equation 4.5 Where θc is the surface area to volume ratio of the particle and d is the particle thickness or diameter. And for terete leaves by: θc = 4/d Equation 4.6 The influence of moisture supported the findings of Xanthopoulos (1990) and Trabaud (1976), and the significance of surface area: volume ratio was also consistent with the implications of that study and the findings of Montgomery and Cheo (1971). The function 107 describing the relationship between leaf surface area to volume ratio, moisture content and ignition delay time at 400oC for green leaves is: ψ400 = (111.3 + m)(0.375θc -0.850) Equation 4.7 Where ψ400 is the ignition delay time at 400oC and m is the moisture content (% ODW) of the leaf. While this study produced a model capable of estimating ignition delay time across a wide range of species, the results were specific to a temperature of 400oC only. The models of Xanthopoulos and Wakimoto (1993) were applicable across a range of temperatures but were species-specific as they did not quantify the role of surface area to volume ration as Gill and Moore (1996) did later. These studies have established that the primary factors affecting ignition delay time are temperature, moisture content and surface area: volume ratio of the leaves. In order to produce a model of ignition delay time that utilises all of these factors, the methodology of Gill and Moore (1996) was extended as described below. 4.32 Methods As per the methodology of Gill and Moore, individual leaves were tested. The exception was the nanophyll Bossiaea foliosa, where small stems bearing 10 or more leaves were used to prevent the samples from falling through the wire platform or being covered by the tongs. Leaves from six species were studied, Eucalyptus stellulata, Bossiaea foliosa, Tasmannia xerophila, Olearia aglossa, Daviesia mimosoides, Eucalyptus pauciflora. These were chosen as they represented a wide range of surface area: volume ratios and moisture contents and were species common to the study area. For each species, 10 green and 10 oven dry replicates were tested at each temperature, and Whatman No. 54 hardened 70mm filter papers were used as controls. Ignition delay time was measured for the temperatures 220oC, 260oC, 300oC, 350oC, 400oC, 500oC, 600oC and 700oC. Altogether 960 leaves were tested. 108 Fresh leaves were collected from plants directly into sealed tins to retain live moisture content, and then tested within a day of picking. Surface area: volume ratio θc was calculated using equation 4.5 for flat leaves, and leaf thickness was measured with a micrometer at a point approximately 2/3 of the way to the midrib. In the case of Bossiaea foliosa, a leaf was randomly chosen from those present on the stem and its thickness used to represent that of the others on the stem. Leaf moisture was calculated as follows. Prior to testing for ignition delay time, the samples to be tested were taken randomly from the tins and placed on trays. The tins containing the remaining leaves were weighed and left open in the same room as the trays. After all samples in the run had been tested, the tins were weighed again then dried at 105oC for 24 hours and weighed again. The moisture of each sample was calculated as follows: w1 + m= S n (w2 − w1 ) − wt (n − 1) (wd − wt ) −1 Equation 4.8 Where m is the leaf moisture percent oven dry weight, w1 is the green weight when the tin was first opened, w2 is the green weight at the end of the run, Sn is the sample number for that sequence (e.g. Sn = 1 for the first sample), n is the total number of samples in the run, wt is the weight of the tin and wd is the oven dry weight. 4.321 Testing for ignition delay time Tests were carried out in a muffle furnace with the door open to provide adequate oxygen supply to samples. The muffle furnace was manufactured by Charles Moloney, Sydney and had a chamber size 15x10x23cm. This was the same furnace used in the experiments of Gill and Moore (1996). Using tongs, samples were placed apex first onto a wire cradle about 3.5cm above the floor of the furnace and held in place at the petiole. To provide an ignition source, the high frequency 109 electric spark gun built and described by Gill and Moore (1996) was held approximately 1 cm above each sample while it was being heated. All tests were recorded using a Sony Digital-8 DCR-TRV265E PAL Handycam and the ignition delay time measured from the video footage to reduce timing error. Due to the temperature gradient in the furnace as a result of the open door, a thermocouple (1mm diameter, k-type inconel sheathed) was fixed 1cm below the point where the leaves were placed on the wire cradle. The earlier experiments of Gill and Moore also utilised this technique, although they do not record the exact placement of the thermocouple so there may be discrepancy between the temperature readings of the two sets of experiments. Temperature readings were not taken from the furnace-designated temperature, but were based on the temperature at the sample as recorded from the thermocouple using a Lontek DT-8 Data logger measuring at 1-second intervals. Movement close to the furnace entrance or ignition of a sample caused wide fluctuations in temperature at the thermocouple, so samples were only introduced to the furnace when the temperature had stabilised to within +/- 10oC of the specified test temperature. The testing procedure occurred as follows. Sample tins were opened, samples removed, placed in order on trays and the tins weighed. As each sample was tested, its diameter was measured with the micrometer and the video camera and spark gun switched on. The sample was entered into the furnace with the spark gun and held in place until ignition was observed. After ignition, the spark gun was removed, the burning sample placed into a metal bucket and the video paused until the next sample. One run of samples consisted of all available species both green and oven dry, commencing with the filter paper control. Ten runs were conducted at each temperature (220oC, 260oC, 300oC, 350oC, 400oC, 500oC, 600oC, 700oC). In some cases, leaves of a particular species were temporarily unavailable so each run did not include all of the species and separate runs were conducted for those species at a later date. The ignition delay time was measured from the video footage using a digital stopwatch, and recorded to 0.1 second precision. At some lower temperatures (220oC and 260oC) some samples did not ignite. In these cases, the temperature datasets were examined with those 110 specimens removed; this meant that they had less samples and were skewed toward the responses of those species that did ignite. 4.322 Endotherm approximations As the ignition of leaves was tested at a range of temperatures, it was possible to collect information on the temperature of the endotherm for the species tested, and to provide limited generalisations based upon characteristics of the leaves. In order to test the effect of volatile oil content, leaves were divided into three groups based on an approximation of the volatile oil content. Volatile oil content was scored between 1 and 3 rating from ‘none’, to ‘very high concentration’ as determined using the methodology of Cornelissen et al (2003), relying on the smell of the crushed leaves. The number of leaves of each species that ignited in the relevant group was examined against the temperature of ignition, and P was estimated to be the point at which 75% or more of the samples ignited. If 75% or more samples ignited at the lowest temperatures, the endotherm was estimated as 220oC as this was the lowest temperature tested. If some or all of the samples did not ignite at the lower temperatures, the mean number of ignitions within that volatile oil score was plotted against temperature for the range of values from 220oC to the temperature at which all samples ignited. This data was examined by linear regression to find the point where 75% of samples ignited. 4.323 Statistical analysis For each temperature dataset, the recorded measurements were ignition delay time , leaf moisture m and surface area to volume ratio. Using the statistical package Minitab, a linear regression was carried out for each temperature, using the form: ψ =a m θc +b Equation 4.9 The values a and b were collected and plotted against temperature in Microsoft Excel, and curves were fitted to the data. These curves were then combined to produce a final model of ignition delay time as a function of temperature, moisture and surface area to volume ratio. 111 In order to identify any areas of systematic error in the model, the residuals from the model were plotted against the order of the data using Minitab. 4.33 Results 4.331 Ignition delay time For each temperature dataset, the relationship between ignition delay time and moisture / surface area: volume ration was well described by a linear equation (Figures 4.1 to 4.8). The R2 values and adjusted R2 for each regression are shown in table 4.1 along with the values of a and b. Raw data are given in appendix III. Table 4.1. Variables produced by linear analysis of the relationship between ignition delay time and moisture / surface area: volume ratio for each temperature dataset. a b R2 Adjusted R2 83 0.8431 12.72 0.835 0.835 260 110 0.7872 9.833 0.825 0.823 300 120 0.6297 7.817 0.824 0.823 350 120 0.4152 5.232 0.875 0.874 400 120 0.3466 3.235 0.927 0.926 500 120 0.1883 2.503 0.845 0.844 600 120 0.1492 1.287 0.926 0.925 700 120 0.09855 0.9547 0.875 0.874 Testing temp Successful (oC) ignitions 220 The coefficients a and b had an excellent fit to temperature data, displaying powerrelationships with R2 values of 0.99 (Figures 4.9 and 4.10). Values for the two lowest temperatures deviated slightly from the trend due to the smaller number of samples and species igniting. The combined model was: ψ = 97805.26T −2.10 m θc + 6452280.04T −2.40 Equation 4.10 112 This model has an R2 of 0.90; observed vs. expected values are shown at Figure 4.11. The plot of residuals against the order of the data is shown at fig. 4.12. The model overpredicts at 220oC (obs. 1 to 83), under-predicts for 260 - 300oC (obs. 84 to 313), and is increasingly accurate and precise over 300oC. 4.332 Temperature of the endotherm The species examined are tabled with volatile oil score, the number igniting at each temperature and their estimated endotherms in table 4.2. The three species with a score of 1 (little or no oily aroma when crushed) were the least likely to ignite at lower temperatures. The regression of these values against temperature is shown in Figure 4.13 and the equation produced was: Ignitions = 0.05T − 5.58 Equation 4.11 This equation has an R2 of 0.82. Rearranging and solving this equation for the value 7.5 gives a value for P of 264oC. For both oil scores two and three, greater than 75% of all samples ignited at 220oC, so in the absence of a more detailed study, the endotherm for species with a volatile oil score of either 2 or 3 may be estimated as 220oC. 4.34 Conclusion The present study of ignition delay time reinforced the findings of earlier work on the topic (Montgomery and Cheo 1971, Trabaud 1976, Xanthopoulos 1990, Xanthopoulos and Wakimoto 1993, Gill and Moore 1996), demonstrating that time to ignition is dependent upon the temperature to which the particle is being heated, and upon the moisture and the surface area to volume ratio of the particle. Although the study has focused upon sclerophyllous leaves and is not totally definitive, it has identified that these three factors can be used to generalise across species while retaining a high level of precision and providing a basis for a complete complex model. 113 A weak connection between volatile oil content and the temperature of the endotherm was established for these six species. Leaves having a strong aroma when crushed have an estimated endotherm of approximately 220oC, whereas those without such an aroma have an estimated endotherm of 260oC. Due to the small number of species studied in each category of oil content, it is recommended that these figures be used as an approximation only in the absence of better data or models. Table 4.2. Volatile oil and endotherm data for 240 samples from six species. *Value estimated using equation 4.11. Number of samples ignited Species Oil o 220 C 260 oC 280 oC 300 oC P score Bossiaea foliosa 1 0 2 2 10 300 oC Daviesia mimosoides 1 8 8 10 10 260 oC Olearia aglossa 1 9 10 10 10 220 oC 5.7 6.7 7.3 10 260oC* Mean Eucalyptus stellulata 2 10 10 10 10 220oC E. pauciflora 2 10 10 10 10 220oC 220oC Tasmannia xerophila 3 8 7 10 10 220oC 220oC 114 !" # $% & Figure 4.1. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 220oC !" # $% () Figure 4.2. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 260oC !" # $% & Figure 4.3. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 300oC !" # $% & Figure 4.4. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 350oC !" # $% & Figure 4.5. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 400oC !" # '% () Figure 4.6. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 500oC 115 Change in b w ith tem perature !" # $% & 12 10 y = 6452280.04x 2.40R20.99 = 8 6 4 2 0 200 300 400 500 600 700 800 900 1000 Tem p (C) Figure 4.7. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 600oC !" # $% & Figure 4.10. Change in the value of b with temperature *+ Figure 4.8. Ignition delay time as a function of leaf moisture/surface area: volume ratio at 700oC & & 120 Figure 4.11. Observed vs. expected results for all data . Change in a with temperature ,-. /0 ! 1.0 y = 97805.26x2.10R20.99 = 0.8 0.6 0.4 0.2 0.0 200 300 400 500 600 700 800 900 1000 Temp (C) Figure 4.9. Change in the value of a with temperature Figure 4.12. Residuals from the model vs. the order of the data. The temperature of the experiment increases toward the right with observation order. The model over-predicts at 220oC (obs. 1 to 83), under-predicts for 260 - 300oC (obs. 84 to 313), and is increasingly accurate and precise over 300oC. 116 4.4 Combustibility and Sustainability Gill and Moore (1996) defined combustibility as the rate at which fuels are consumed, as given by: C= w f − wi t f − ti Equation 4.12 Where C is combustibility, wf is the weight of fuels at the end of flaming combustion, wi is the weight of fuels at the point of ignition, tf is the time at which flaming combustion ceases, and ti is the time at which ignition occurs. The sustainability of combustion in a fuel can be defined using flaming or smouldering combustion (Gill and Zylstra 2005). However when considering the propagation of flame between fuels, smouldering combustion is probably irrelevant as it represents the period following the presence of flame. In this context then, sustainability will be used in this thesis to refer to flaming combustion unless otherwise specified. This is given by: tr = tf – ti Equation 4.12 Where tr is the sustainability of the fuel. 4.41 Introduction This study is an extension of a paper presented at the 5th International Conference on Forest Fire Research (Zylstra 2006). While it would be ideal to have an existing standardised test that measures particle ignition in a fire, no such measure yet exists. For widespread use, tests need to be simple and affordable while producing accurate and repeatable results. Dibble et al (2007) point out that a common 117 approach in quantifying flammability is to use measurements of physical or chemical properties as surrogates for “fire performance”, but unless such measurements enable the user to calculate physical fire characteristics, they do not meet the requirements of this model. Alternatively, engineering tests such as those in the ASTM F1870-05 guide (ASTM 2007) and others are typically meant for household furnishings, require a rather expensive laboratory set-up of calibrated and varying intensity radiators, and do not address the specific parameters of concern here. The test proposed can be standardised and repeated, but it is also possible to carry it out with rudimentary equipment. As the first step in a bid to understanding flames in forests and other natural vegetation, the length and duration of flames from individual burning leaves were studied in the laboratory. The aim was to define a standard methodology and to examine the effects of leaf characteristics on flame length and duration, linking flame dimensions to leaf morphology, moisture content or chemistry rather to individual species. As described earlier, flame length can be seen as a measure or approximation of combustibility for the array or element being measured (e.g. forest, plant or fuel bed), but is also a product of all the aspects of flammability of those elements that compose the unit. Combustibility and sustainability are inversely related (Gill and Zylstra 2005); therefore we can expect the same properties that cause an increase in one to also cause a decrease in the other. Although flame length is treated as a measure of combustibility in the following theory, this inverse relationship will be used to test the strength of flame length as a predictor at the scale of an individual leaf. 4.311 Factors affecting Flame Length and Duration The length and duration of flame from an isolated burning leaf is probably a product of the three primary factors: 1) Chemistry of the leaf 2) Dimensions of the leaf 3) Moisture of the leaf 118 It is not known how great the effect is from each of these factors; however they will be explored conceptually below and an attempt will be made to quantify each of them experimentally. Chemistry of the leaf Visible flame is produced by the oxidation of carbon particles (soot) and gases arising from the pyrolysis of a fuel (Lyons 1985). The exothermic oxidation reaction emits radiation; that part of the radiation within the visible spectrum forms the flame. When a fuel with high thermal energy content is burnt, the greater amount of energy produced facilitates faster pyrolysis and therefore a greater density of gases and carbon particles. Oxygen is limiting in the combustion of these materials at the centre of the flame as it must enter the flame through diffusion, producing a gradient of oxygen availability. The greater the densities of fuels above the stoichiometric fuel to oxygen ratio, the slower the rate of combustion and the longer it takes for all of the particles to be consumed. Higher temperatures produced by the flame also create increased gas buoyancy and therefore vertical velocity. The combination of the two factors means that soot particles transported by the gas travel a greater distance before they burn out and lose their incandescence; hence, a longer flame. Thermal energy content is a major explanatory variable that has been used in the modelling of combustibility in fuels. Byram (1952) used it as a plank in his definition of fire intensity, the product of heat yield, rate of spread and fuel load. Although heat yield is often considered a constant in the field, there is considerable variation in the standard value used with estimates ranging widely as discussed in chapter 2. As described earlier, higher levels of energy content produce faster rates of pyrolysis. The faster reaction time means that the leaf is consumed more quickly, therefore if we consider the earlier assertion stating that combustibility and sustainability are inversely proportional and if all other properties are kept constant, leaves with higher energy contents producing higher rates of combustion should also have lower values of sustainability. Rothermel (1972) introduced the idea of damping coefficients that reduce the quantity of energy released from its potential value, identifying both mineral and moisture damping. Sandberg et al (2007) added a third damping coefficient he termed ‘inefficient packing’, although Rothermel (1972) had already identified optimal packing ratios as a fundamental 119 issue. Of these, moisture will be examined separately, and packing is a property of the fuel bed rather than the particle, relating to heat conductivity and the movement of air. This does not relate to individual particles. Rothermel suggested that the percentage of silica free ash in a fuel would reduce the ‘reaction velocity’ , a term in his fire spread equations. There is no suggestion that two particles with the same energy content but different silica free ash contents will release different quantities of heat when ignited, rather the term is used only in connection with a fuel model representing a characteristic array of fuel strata. Philpot (1970) found that silica free ash content determined the temperature at which ignition occurred, a property that may be classified as part of ignitability. It is likely then that silica free ash content will have no significant effect on the burning properties of individual fuel particles once they are burning, but its influence on ignitability will affect the way a fuel array burns in a similar way to that in which changed ignitability properties affected the combustibility of fuel beds studied by Gill et al (1978). Although chemical tests were not carried out, the relative influences of thermal energy content and silica free ash content were inferred in this experiment from the amount of variability left in the flame parameters after examining the other variables. King and Vines (1969) found that the presence of flammable volatile oils in a leaf served to counter the effect of moisture to some extent, so that leaves with higher moisture content ignited more easily if they were treated first with Eucalyptus oil. Volatile oils have a heat of combustion roughly twice that of cellulose (King and Vines 1969), however the small quantities present in a leaf are more likely to affect the flame length by facilitating ignition of the leaf as described by King and Vines, rather than significantly affecting its energy content. Dimensions of the leaf Leaf dimensions affect flame length in two ways: by increasing the surface area of fuel burning and by broadening the flame base. The greater the surface area of leaf that is burning, the greater the area reacting and therefore the rate of heat release as given by Tewarson (1980): 120 Q = HM Equation 4.14 Where Q is the heat release rate per unit fuel surface area (kW m-2); H is the thermal energy content of the fuel (kW g-1), and M is the mass loss rate of the fuel in combustion per unit fuel surface area (g m-2 s-1). Because it allows more energy to be released at the same time, a greater rate of heat release affects the flame length in the same way as a higher thermal energy content, so that larger leaves produce longer flames. The longer flames may result from nothing more than an increase in the quantity of fuel burning, so that we do not see the expected decrease in sustainability as combustibility increases. Alternatively, the increased energy release may promote increased burning efficiency, thereby decreasing the time to burn per weight of fuel. In the first case we would say that the increased flame length is a result of increased fuel quantity, but in the second case the increased fuel quantity is accompanied by an increase in combustibility and its associated decrease in sustainability. In either case, the increased flame length from the leaves may cause the plant itself to be more flammable through improved heat conductivity and thereby an increase in consumability of the plant. The second way in which leaf dimensions affect flame length is by broadening the flame base. A broader flame base means that the burning particles and gases that form the centre of the flame have less oxygen available to them for combustion than those do at the outer edges. Consequently, much of the material that forms the visible flame takes longer to be consumed, forming a longer flame. The mechanism is much the same as that of cluster burning or flame merging (see Thomas 1963, Thomas et al 1965, Huffman et al 1967, Steward 1970, Chigier and Apak 1975, Tewarson 1980, Gill 1990, Heskestad 1998, Weng et al 2004, Liu et al 2009). In that case two or more flames in close proximity to each other block air entrainment to each other, causing each flame to tilt toward the centre of the group and merge into one longer flame with a broader base. As this effect of increasing leaf dimensions is to slow the rate of combustion within the flame by blocking air entrainment to its centre, the effect on sustainability is complex. By reducing the rate of combustion M is decreased and sustainability may be increased. Alternatively, the longer flame has a greater buoyant velocity and air entrainment at the lower edges of the flame will be increased as a conservation of energy. Increased air entrainment at the edges may in this case balance the reduction in rate 121 of combustion at the centre. The overall effect of leaf size on sustainability is not yet clear, it is intended that this study will better inform the matter. Other dimensions of a fuel particle are known to directly affect the period of flaming combustion. In various studies, Anderson (1969), Cheney et al (1990) and Burrows (2001) all found that flame residence time in burning round wood was a function of particle diameter. The residence time for the complete length of the wood will be dependent upon this value and the rate at which the flame spreads along it, unless the entire particle is ignited at the same time as is the intent in this study. In the context of flat leaves, the measure of diameter may relate to either the thickness of the leaf, its width or the area of its cross-section. Moisture of the leaf If a leaf is ignited at the tip, it may self extinguish depending upon the angle of the leaf and its moisture content (Burrows 1984). In this case, moisture content has a direct effect on flame length and duration, with higher moisture content reducing both values through a change in consumability. This is relevant to the flammability of a collection of leaves burning together such as in a plant or a fuel bed, and may be part of the damping effect alluded to earlier from the Rothermel model. King and Vines (1969) and later Bellamy (1993) examined flame residence in a similar context, defining flammability as the duration of flaming combustion for oven-dry leaves in the absence of an external heat source. Whether it is due to moisture or some other property of the leaf, this aspect will not be considered further here. While the ability of a leaf to maintain flaming combustion in the absence of an external heat source is an important aspect of flammability, this study only examines the context where a bushfire produces an ongoing, albeit variable external heat source. Moisture present in a leaf requires additional energy to evaporate it before the leaf can be heated to the point of ignition, directly affecting the rate of spread of a flame along a leaf (Weber 1990). A slower rate of spread means that less of a leaf will be burning at the same time, producing the same effect on flame length and duration as that of changing the size of the leaf. These circumstances are relevant when a leaf is ignited by a point ignition. In most field situations however, the size of the flaming front will be many times larger than the size of the leaf, so that the entire leaf will be ignited simultaneously. Once again, as this study 122 does not assume a point source but an area of ongoing external heat, this effect of moisture is not relevant here. The influence of water absorbing energy in its evaporation is to slow the rate of pyrolysis, releasing less carbon particles and gases to be consumed within the flame. This will affect the rate of combustion in the ongoing presence of an external heat source, so it is likely to have an affect on the results of this study. Leaves with higher moisture contents can be expected to combust more slowly so that the flame length will be shorter but the period of combustion will be longer. Summary If the conditions are limited to isolated leaves ignited as a whole in still conditions and exposed to an ongoing area heat source, it can be expected that: 1. Leaves with a higher thermal energy content will produce a longer flame, but for a shorter period of time. Silica free ash is not expected to affect flame length or duration in a single burning leaf, whereas volatile oil content may have a very small influence. 2. Larger leaves will produce longer flames, thicker or wider leaves will burn for longer 3. Moisture content will affect both flame length and duration, with wetter leaves producing smaller flames but burning for a longer period of time. These are qualitative expectations and the magnitude of the effect of each is unknown at the scale of the individual leaf, although some of these factors have been studied for whole plants (e.g. Etlinger and Beall 2004). The intention here is to produce a quantitative model which considers these points. 4.42 Methods Two hundred and ten simple (definition after Harden 1993) sclerophyllous leaves were picked from 10 species of shrubs and trees common to the Australian Alps (Table 4.3). The leaves were given three treatments; either sealed in tins to retain live fuel moisture levels, allowed to dry at ambient temperature for up to four days to approximate drought stressed or 123 dying green leaves, or oven dried. For each species, seven replicates were burnt in each of these three moisture categories. Table 4.3. Species examined in the study Species Growth Volatile Mean Mean Mean Max Form oils leaf burnt width moisture % (score) thickness length (mm) ODW (mm) (mm) Brachyloma daphnoides Shrub 1 0.22 4.3 2.1 130.3% Daviesia mimosoides Shrub 1 0.39 44.5 7.6 106.3% Eucalyptus niphophila Tree 2 0.54 60.5 16.0 110.5% E. pauciflora Tree 2 0.48 81.9 26.7 122.5% E. stellulata Tree 2 0.43 59.4 20.5 126.8% Helichrysum thyrsoideum Shrub 3 0.25 7.1 1.9 176.4% Orites lancifolia Shrub 1 0.41 31.5 11.9 125.7% Phebalium ovatifolium Shrub 2 0.42 9.8 7.0 130.8% Prostanthera cuneata Shrub 3 0.35 4.8 3.5 185.9% Tasmannia xerophila Shrub 3 0.36 40.9 9.0 140.4% Each leaf to be burnt was measured against a scale for length, and width at the widest point. Leaves were chosen randomly to provide a range of sizes, although new bright green growth was not tested. If a leaf was incompletely burnt during the experiment, the median remaining length of leaf was measured and this value subtracted from the total length to give length of leaf burnt Lb. Leaf area burnt was calculated from a formula that approximated the leaf shape as 2 connected triangles: A= WLb 2 Equation 4.15 Where A is the single-sided leaf area burnt (mm2), W is the leaf width (mm) and Lb is the length of leaf burnt (mm) (fig. 4.13). 124 W Lb Fig. 4.14. Plan view of leaf dimensions Leaf thickness (t) was measured to 0.1mm precision at a point two-thirds of the way to the midrib using a micrometer. Area of the cross section c was calculated as c = t w. Equation 4.16 As in the previous section, volatile oil content was scored between 1 and 3 rating from ‘none’, to ‘very high concentration’ as determined using the methodology of Cornelissen et al (2003), relying on the smell of the crushed leaves. 4.421 Burning methodology Maximum flame length and duration were found by igniting each leaf and holding the burning leaf over the 2-3cm flame of a candle. The candles used were plain “Candlelight Econopack” 20mm diameter white wax household candles. As the purpose of the study was not to study ignitability or consumability of the leaf in the absence of an external heat source, the entire leaf was preheated to facilitate simultaneous ignition of a greater part of each leaf; thereby more closely approximating the specified condition of an on-going area source of heat. The candle flame was chosen in preference to a gas flame such as that used by King and Vines (1969), Bellamy (1993) and Etlinger and Beall (2004), as the lower temperature flame and solid fuel (as opposed to a forced gas stream) would produce less vertical air movement with the potential to artificially lengthen the flame from a burning leaf. Liquid pool fires were also investigated, but the flame was found to be too turbulent and inconsistent to achieve repeatable results. The use of a candle which provides a point–source ignition and the wilting of vegetation samples as they are heated do present additional challenges if one wishes to achieve a consistent heating rate. However, our final method, as described next, seems to give a good compromise between ease of use and reproducibility. 125 To ignite the leaves, each sample was firmly held at the stalk end and starting at the leaf tip, moved back and forth over the tip of a candle flame at it’s visible extent, turning the leaf and heating it evenly until as much of the leaf as possible began to smoke. The larger leaves wilted as they were heated and were gently raised and angled as they wilted to ensure even heating and that no part of the leaf was inserted into the flame where the higher temperatures may cause it to ignite before the rest of the leaf was ready. Larger leaves were moved over the flame slightly faster to prevent parts of the leaf cooling before the rest of the leaf was heated. Once the entire sample was evenly heated, the leaf was held on a downward angle and the tip held at the tip of the flame until ignition. This process of gradual heating made it possible to ignite more of the leaf simultaneously. Once the leaf was burning, the advancing edge of the flame from the leaf was held over the tip of the candle flame to maintain a constant source of heat. Even so, some of the larger green samples only partially ignited, with the flame beginning to die after a large part of the leaf had been consumed. In these cases, the flame was allowed to die out rather than reigniting the leaf. The experiments were carried out in front of a scale marked with 10 mm increments for measurement of the flame dimensions, and each sample was recorded on video. The camera was situated at a point approximately level with the top of the candle flame at a distance of 1240mm, and the scale was 45 mm behind the flame. Leaves were burnt at three classes of moisture content with seven replicates of each class for every species. The classes were fresh (taken from the sealed tins), air dried or oven dried producing a range of moistures from oven-dry to those given in table 4.3. All experiments were conducted in still air conditions. After burning green or air-dried samples, the remainder in each tin was weighed and then oven dried at 105oC for 24 hours to determine moisture content. When burning oven dried samples, the tins containing the leaves were weighed before re-opening to determine any difference from their initial oven dry weight, and the current moisture content was then calculated from these values. 126 4.422 Measurements of flame length and duration The maximum flame length was found by close examination of the video record, and this length was measured by eye from the record to a precision of 5mm. This level was set to allow for the possibility that the electromagnetic range of the camera might differ slightly from that of the naked eye. The flame was measured from the point it broke contact with the angled leaf to its tip. Recorded maximum flame lengths were corrected for parallax error by subtracting from them a value of: E = S s tan l Sc Equation 4.17 Sc Ss Where E is the error in mm, Ss is the separation between the flame and the l Camera scale (mm), Sc is the separation between the camera and the flame (mm) and l is the recorded maximum flame length in mm (Figure 4.15). Figure 4.15. Dimensions for calculating parallax error Flame duration was measured from the video record in whole seconds, using a stopwatch. 4.423 Data analysis As the aim was to build flame length and duration equations that would be accurate across all species with characteristics within the ranges studied, differences in burning between species were not examined unless the analysis made it clear that the variables used were insufficient. As the results make clear, species-specific analysis was not necessary. An initial analysis was carried out by graphing flame length and flame duration tr as functions of each independent variable to determine what kind of regression was most appropriate. When it appeared that a linear relationship provided a good description of the dependence of the measured data upon the variables, this was examined thoroughly using a best subsets regression in the statistics package Minitab to find the strongest predictors. For each 127 parameter, the coefficient of correlation R was also calculated, and an ANOVA performed to find the level of significance. The R-value was only indicative for leaf dimensions, as the yintercept was set to 0 when constructing equations based on leaf dimensions to reflect the fact that leaves with a length, width or thickness of zero cannot produce any flame. This restriction is necessary if the equations are to be used in a modelling framework to avoid the possibility of modelling negative flame lengths for very small leaves. In cases where the relationship was clearly not linear, the data was examined several times after trying standard transformations (square root, cube root and logarithm) in order to discern the most appropriate non-linear relationship. Table 4.4. Correlation half-matrix for flame length, duration and all independent variables, showing significance for correlations with the 2 dependent variables. * Correlation significant at the 0.01 level, ** correlation significant at the 0.001 level. Thickness (mm) Thickness (mm) Leaf Width (mm) Burnt Length (mm) Volatile Oils (score) Moisture (% ODW) Burnt Leaf Area (mm2) Leaf Area ½ Leaf Width (mm) Burnt Length (mm) Volatile Oils (score) Moisture (% ODW) Burnt Leaf Area (mm2) Leaf Area Leaf Area 1/2 1/3 Log Leaf Area Crosssection Area (mm2) Leaf Volume (mm3) Duration (s) Corrected Flame length (mm) 1.00 0.54 1.00 0.48 0.86 1.00 -0.07 -0.11 -0.12 1.00 0.15 -0.11 -0.21 0.05 1.00 0.37 0.90 0.90 -0.05 -0.17 1.00 0.52 0.96 0.97 -0.11 -0.17 0.94 1.00 0.57 0.94 0.96 -0.14 -0.16 0.88 0.99 1.00 0.64 0.85 0.88 -0.15 -0.08 0.71 0.89 0.95 1.00 0.68 0.96 0.83 -0.11 -0.05 0.85 0.92 0.91 0.82 1.00 0.45 0.69** 0.89 0.77** 0.89 0.73** -0.05 -0.14 -0.13 0.09 0.98 0.67** 0.92 0.77** 0.87 0.78** 0.70 0.74** 0.89 0.85** 1.00 0.73** 1.00 0.48** 0.83** 0.89** -0.13 -0.39** 0.79** 0.89** 0.90** 0.83** 0.81** 0.79** 0.64** Leaf Area 1/3 Log Leaf Area Crosssection Area (mm2) Leaf Volume (mm3) Duration (s) Corrected Flame length (mm) Effort was made to produce an equation that both required minimal data input while maintaining maximum precision. Where this was not possible, more than one equation was produced. 128 1.00 Where more than one input variable was required, only predictors having a correlation with each other of less than 0.75 were used, and the model was chosen that had either the lowest standard error S or the value of Mallows C-P closest to n+1 where n was the number of independent variables. ! 42 $ 2 05 $$ " # /1+ 0 3 " For the purposes of simplicity, all # 4.424 Combustibility and flame length oven-dry density so that combustibility ! species were assumed to have an equal could be given in terms of volume consumed rather than weight. We therefore calculate combustibility C as leaf volume consumed per period of Figure 4.16. Flame length as a function of 3 A , showing the 95% confidence interval and the 95% prediction interval. The model was not forced through 0. time. ! 42 $ 2 05 $$ " # In order to find how best to use flame 8( # 3 " length as a measure of combustibility, 0 61 07 was graphed as a function of flame length and the strength of the correlation ! combustibility was 2 determined by comparing the R value $ for linear, exponential and power relationships. A predictive equation % & Figure 4.17. Flame length as a function of A , showing the 95% confidence interval and the 95% prediction interval. The model was not forced through 0. was produced for the strongest of these. ! The data were also analysed in Minitab * 3 4.43 Results ! " of the correlation. # using ANOVA to find the significance 4.431 Flame length The correlation half-matrix for all variables in the study is given as table ! ' ( )( $ Figure 4.18. Observed vs. expected results for equation 4.16 for flame length using A with the model forced through 0, showing the 95% confidence interval and the 95% prediction interval. 129 4.4 and all raw data are given in Appendix IV. All leaf dimensions had a highly significant (p < 0.01) relationship with the parallax-corrected maximum flame length . The maximum flame length in a burning leaf was most strongly correlated with 3 A (Fig. 4.16). However the best subsets regression (Table 4.5) found that when the y-intercept was set to 0, the standard error for a model based on A (Fig. 4.17) was lower and this was used to build the following model. λ = 5.25 A Equation 4.18 where A is the leaf area (mm) given by equation 4.15. The observed vs. expected results for equation 4.18 are given at fig. 4.18. The remaining parameters that were not strongly correlated with A were leaf thickness, moisture and oil score. The best subsets regression showed the strongest combination to be 3 A and leaf moisture. Leaf moisture had a significant correlation with the error from a model based on 3 A (fig 4.19). The equation based on 3 A and moisture content was: λ = 17.53 A − 0.277m − 0.27 Equation 4.19 where m is the moisture content (%ODW). The observed vs. expected results for equation 4.19 are given at Figure 4.20. The model with the greatest accuracy is equation 4.19, with a standard error (S) of 21.97 and an adjusted R2 of 0.87. This model is not suitable however for smaller leaves, as it attributes a negative flame length in these cases, causing significant skewing of the residuals in this area (Figure 4.20). The cluster of points on the left of the graph are better modelled by equation 130 4.18, with S = 27.52 and adjusted R2 of 79.9%. The moisture content above which equation 4.18 should be used was found by combining and rearranging equations 4.18 and 4.19 to find m: m= 17.53 A − 5.25 A − 0.27 0.277 Equation 4.20 Models using more than two variables were not considered as the best subsets regression indicated negligible improvement in precision while relying on variables that were correlated with each other. 4.432 Flame duration Flame duration tr was most strongly correlated with the area of the cross section (c) whether or not the model was forced through the origin (Tables 4.3 and 4.5). The relationship was linear; however there was considerable spread in the data in leaves with a larger cross-section area (Figures 4.22 and 4.23). 131 ' 2 ./ 0 " ' ! 01 * 9 1 0 " # / /0 $$ 3 " # 3 "+ ,# Figure 4.19. Relationship between the error from the 3 A model and leaf moisture ! " $ # Figure 4.22. Flame duration as a function of leaf cross-section area, showing the 95% confidence interval and the 95% prediction interval. The model was not forced through 0 ! * . 1 0 ! 3 ! ' /1 ( ./ 0 * 2 )11 ! * 2 )1 Figure 4.20. Observed vs. expected results for equation 4.19 using 3 A and m, showing the 95% confidence interval and the 95% prediction interval. The model was not forced through 0. Figure 4.23. Scatter-plot showing the spread of residuals in the larger fitted values when modelled using leaf cross-section area (equation 7) ! . 42 $ 2 " # ' 01 * 9 05 $$ 3 "+ ,# ! Figure 4.21. Scatter-plot showing the clustering and skewing of residuals in the lower fitted values for the model given at equation 4.19. Figure 4.24. Relationship between leaf moisture and the error from a model with cross-section area as the only predictor. 132 The most precise equation used both c and m as predictors, although the influence of m while significant (p<0.01) was considerably less than its influence on flame length (Figure 4.24). The equation for flame duration tr as a function of leaf cross-section area and moisture was: t r = 1.37c + 0.0161m − 0.027 Equation 4.21 The observed vs. expected results for equation 4.21 are shown at fig 4.25. The effect of decreasing precision with increasing leaf size is largely unchanged from a model using only c. The standard error S has been reduced from 3.59 to 3.47. ! 4.433 Combustibility and flame length * 3 The strength of flame length as a predictor of combustibility (C) is shown in Figure 4.26. The data had an R2 value of 73% for a linear relationship (Equation 4.22), and ANOVA results indicate that the correlation was highly significant (p < 0.01). Beyond a flame length of approximately 150mm, the relationship ' / 01 3 2 2/4/ * 2 2)1 Figure 4.25. Observed vs. expected results for equation 7 using c and m, showing the 95% confidence interval and the 95% prediction interval. The model was not forced through 0. shows increasing spread and rates of combustion are well beyond what the flame length trend indicates. Exponential and power curves however produced weaker R2 values (63% and 66%). C = 0.19 - 1.37 Equation 4.22 Figure 4.26. Flame length as a predictor of combustibility. Combustibility increases consistently with flame lengths up to a length of about 150mm. Beyond this point, the rates of combustibility are significantly higher than expected from the observed flame heights. 133 4.44 Discussion The high levels of significance and goodness-of-fit found using the methodology described suggest that the methods used are an effective means of determining flame length and duration experimentally. As it was more difficult to obtain consistent results with larger leaves, it may be that the procedure is best suited to leaves with an area less than about 1000mm2. 13.0 21.968 * 2 13.4 26.268 * 3 1.6 21.922 * 3 11.0 21.944 4 3.2 21.285 4 3.3 21.772 5 4.5 21.319 * 5 4.8 21.323 * Log Area 2 Cube root Area 27.647 28.651 Sqrt Area 140.5 166.2 Volume Standard error Cross-section Mallows C-p Leaf Area Moisture Oil Score Leaf Length Leaf Width Leaf Thickness Number of Variables 1 1 * * * * * * * * * * * * * * * * * * * * * * * * 134 * 81.1 3.6573 1 190.2 4.2922 2 53.7 3.4742 2 57.6 3.5005 3 23.3 3.2583 * * 3 25.9 3.2772 * * 4 9.7 3.1520 4 14.7 3.1890 5 3.5 3.0983 * 5 4.1 3.1028 * * Log Area Standard error Cube root Area Mallows C-p Sqrt Area Number of Variables 1 Volume Cross-section Leaf Area Moisture Oil Score Leaf Length Leaf Width Leaf Thickness Table 4.6. Best subsets regression for variables predictin * * * * * * * * * * * * * * * * * * * * * * * In the introduction, some qualitative expectations for the effects of leaf chemistry, dimensions and moisture on the length and duration of flames from burning leaves were outlined. Some of these have now been quantified, as described below in order of significance. 4.441 Leaf dimensions For both flame length and duration, leaf dimensions were the most significant predictors of those examined here. Leaf area explained 80.9% of the flame length variability, and leaf cross-section area explained 71.9% of the variability in flame duration. The variability in flame length and duration in larger leaves was possibly a reflection of the difficulty in producing simultaneous ignition of these samples. The flame from a burning leaf is often turbulent and therefore highly variable in length; burning leaves were commonly observed to produce small jets of flame in any direction, adding to the variability in flame length and burning consistency. If this is the case it is likely that improved experimental techniques may provide even stronger R2 values. Another possible explanation may relate to complexities around air entrainment and different efficiencies related to leaf shapes and the ratio of length to width. 135 Given the significance of leaf dimensions in explaining maximum flame length and duration from isolated burning leaves in still conditions, it appears that leaf dimensions are an important parameter for quantifying plant flammability. It may be the case that the effect of leaf size becomes insignificant in the context of other factors such as the structure of the plant or the ignitability of the leaves, but it is necessary to investigate whether this is actually the case and to quantify the effect on combustion of the whole plant. 4.442 Moisture content The influence of moisture content was evident, although minimal. As expected drier leaves produced longer flames and burnt more quickly. It is interesting that the effect of moisture was still significant after pre-heating, suggesting that enough moisture remains in a burning leaf to affect the rate of combustion. If we consider these findings along with those regarding ignition delay time, the effect of moisture on fire propagation through a live plant is to delay the ignition of the leaves, then once they are alight, shorten the flame length but make the period of combustion last slightly longer. For example, a species having thick, moist leaves can be expected to be difficult to ignite with leaves that produce short flames. 4.443 Leaf chemistry The scoring system for volatile oil content proposed by Cornelissen et al (2003) did not provide any statistically significant relationships, which may reflect that either volatile oil content has negligible influence on these parameters, or that the smell of crushed leaves is not a reliable methodology for estimating volatile oil content. It is likely that these factors affect the flammability of the plant more through leaf ignitability, perhaps by lowering the temperature of ignition (King and Vines 1969). It is proposed that although the energy and silica-free ash contents of the samples were not measured, the small amount of variability left in the results suggests that at the scale of individual leaves these factors play a lesser role in flame length and duration than do leaf dimensions and possibly leaf moisture. 4.444 Flame length as a measure of combustibility The strong and highly significant correlation between flame length and combustibility indicates that the increase in flame length with larger leaf size is a reliable measure of combustibility. The relationship does not suggest that the leaf tissue itself is somehow more 136 flammable in larger leaves of any species; the effect is related to the amount of fuel that is burning simultaneously. This effect is the same as that reported by Gill et al (1978) for fuel beds, where drier fuel beds allowed faster fire spread so that more fuel was burning at the same time and the overall effect was an increase in combustibility. In the same way, Etlinger and Beall (2004) found that drier leaves facilitated a greater amount of a plant to burn, producing an increase in the rate of heat release compared to plants with higher moisture contents. As indicated in fig. 4.26, increases in combustibility beyond about 40 mm3s-1 were not accompanied by increases in maximum flame length. A possible explanation is that although maximum flame length did not increase, mean flame length may have. This parameter was not measured in the study so the hypothesis cannot be tested. The reasons for the limit in maximum flame length may relate to the difficulties in simultaneous ignition of the largest leaves or limits to the distance to which air can be entrained into the one flame, but this also cannot be shown from these results. 4.5 Leaf Flammability As the results of both experiments demonstrated, the dominant factors affecting the flammability of leaves are their dimensions and moisture content, accounting for 90.0% of the variability in ignition delay time, 83.4% of the variability in flame length and 73.7% of the variability in flame duration. The temperature of the endotherm was affected by the chemistry of the leaves, with relationships demonstrated for silica-free ash content and volatile oil content. The interaction of these factors is complex; thicker or wetter leaves for instance take longer to ignite but also burn for a longer period in the presence of an external heat source. It is not possible to generalise the results to state that plants with leaves having a certain characteristic are more or less flammable as such conclusions can only be found by modelling the interactions of all of the processes as given in Figures 3.6 and 3.8. The models developed in this current chapter provide inputs for Table 3.1 and Figure 3.8. (Appendix II). 137 Chapter 5 Plant Flammability Having considered flammability at the scale of a single leaf, the next order of scale is that of an individual plant. A plant differs from a leaf in that it is a matrix of fuel particles separated by spaces, so this chapter is an application of the conceptual model presented in chapter three. Chapter four produced mathematical models of the aspects of flammability relevant to a leaf; chapter five now incorporates the geometry of the plant so that the role of the ignition isotherm can be modelled. When flammability is considered at the level of an individual plant, the variables to be quantified are 1. Whether the plant will ignite (ignitability), 2. The flame length to be expected from the plant (combustibility), and 3. The duration of the flame from the plant (sustainability). Section 3.23 presented a conceptual approach to answering these questions which required a number of sub-models to give quantitative answers and chapter four provided equations to describe the ignitability, combustibility and sustainability of individual leaves. The purpose of this chapter is to develop the conceptual model into a quantitative model by applying the equations from chapter four along with any others required. This chapter will be limited to 138 plant ignition from beneath as opposed to ignition by adjacent plants; the latter scenario will be examined in the following chapter. 5.1 Plant Architecture Figure 5.1 gives symbols for the various dimensions, w sides and angles of a simplified plant crown. Bold a A B b c C Hp capitals are used to name the sides and lower-case letters used to name the angles. The crown is symmetrical to D Ht d e f E He F allow the possibility of modelling fire introduction from any side without prior knowledge of the direction in which the fire is spreading. Hc Figure 5.1 Dimensions, sides and angles used to describe the leafy crown of a plant. The number of sides and angles can be varied. More sides can allow for more complex plant shapes, but will also complicate the modelling calculations. The number of sides shown accommodates basic variability in crown shapes, particularly allowing for those factors that affect the likelihood of a plant catching fire, and the depth of fuel bed that will be provided by that plant. These factors are primarily: 1) Height of the crown base, as it affects the likelihood of fire spreading upward into that stratum such as in the development of crown fires (Van Wagner 1976, Alexander, 1998, Cruz et al 2006, Morvan 2007). In the references cited this is given as a single value and this may be sufficient in many cases, however division of the crown base into two segments with heights Hc and He as shown allows more adequate description of species that do not have a level crown base, e.g. Figure 5.2. 2) Vertical depth of the crown at both the sides and the centre, and horizontal width. Unless the entire crown has been simultaneously ignited, both factors limit the quantity of foliage that is being burnt by a flame at a given angle. Examination of Figure 5.2 shows that as He is smaller than Hc while the top of the crown is relatively level; the vertical depth of crown fuel is slightly greater at the sides than at the centre. 139 Figure 5.2 Crown shape of Eucalyptus niphophila showing foliage closer to the ground at the edges than at the centre. The bold lines mark the sides E and F. 5.11 Angles and dimensions The angles of each of the vertices are found as follows: a = 360 − 2b Equation 5.1 b = 90 + tan −1 (H p − Ht ) 0.5w Equation 5.2 c=b Equation 5.3 d = 90 + tan −1 (H e − H c ) 0 .5 w Equation 5.4 140 e=d Equation 5.5 f = 360 − 2d Equation 5.6 The dimensions of the sides are: A= w 2. cos(b − 90) Equation 5.7 B=A Equation 5.8 C = Ht − He Equation 5.9 D=C Equation 5.10 E= w 2.Cos(d − 90) Equation 5.11 F=E Equation 5.12 141 5.2 Plant Ignitability As discussed in chapter three, ignition of the leaves of a plant will occur when the critical state is satisfied, that is, as given in equation 3.1, the temperature at the leaf is greater than P and the flame residence time is greater than . To satisfy equation 3.1, the value of P must be determined either experimentally, by using a model such as equation 4.1 or by using the estimates given in chapter four. The value of can be found from the moisture and surface area to volume ratio of the leaves using equation 4.10 if T P is known. If the flame heating the plant is burning in a plant, the flame residence time will be calculated using methods developed in this chapter. Where the flame is burning in a simple litter bed of dead fuels, the flame residence time can be found from the mean diameter of the surface fuel particles using a model such as those developed by Anderson (1969), Cheney et al (1990) or Burrows (2001). The flame duration model of Burrows (2001) for woody fuels gives Tr in seconds for particles of dmm diameter: Tr = 0.871d 1.875 Equation 5.13 Dead leaves were found to have a mean residence time approximately equivalent to sticks with a diameter of 4mm, so as surface fine fuels are defined as having diameters up to 6mm (McArthur 1967), the range of diameters for surface fine fuels should in general be limited to 4-6mm. Assuming that the heat source is a flame burning in a simple litter bed of dead fuel particles and the residence time can be calculated accordingly, what is still required is a model to estimate T P . 5.21 Temperature of a plume Weber et al (1995) described the temperature profile above a flame in 3 zones: 142 I – the zone of continuous flame II – the zone of flickering flame III – the zone of heated air above the visible flame The distance from the burning fuel to the top of region I refers to flame burning within a continuous fuel array, for instance the flame within a burning plant. As there is a constant fuel input dispersed throughout the zone, the temperature is approximately consistent and is given as a constant K. The outer edge of this zone is marked by the edge of the continuous fuels such as the outer leaves of the plant. The flame produced by single burning fuel particles such as leaves or a flat stratum such as a burning layer of forest litter does not have this zone as the fuel input does not come from a continuous array such as the leaves of a plant canopy. The distance to the outer edge of zone II is equal to the maximum flame length λ measured from the outer edge of the burning fuel unit. For a plant canopy this is the distance from the outer leaves of the plant to the maximum flame length. For a burning leaf or a flame arising from surface fuels, zone II is equal to the maximum flame length . The change in temperature from the ambient at any height Sp above the top of the leaf or canopy (f) is given by: 2 ∆T II = Ke (−α ( Pα − f ) ) Equation 5.14 Where α is a constant given by: α= 1 2λ (λ − f ) Equation 5.15 Beyond the point λ, temperature decreases according to: ∆T III = C Sp Equation 5.16 143 Where C is given by: 2 C = Kλ.e(−α (λ − f ) ) Equation 5.17 If we apply these equations to a flame produced by a burning particle or a layer of surface fuels that does not have an area of flame corresponding to zone I, we are able to simplify the equations by removing the term f. This term is found in equations 5.15 and 5.17 for α and C respectively. The simplified equations are given below as equations 5.18 and 5.19. α= 1 2λ2 Equation 5.18 C = Kλ.e (−αλ ) 2 Equation 5.19 If we combine these equations with equations 5.14 and 5.16 for Zones II and III respectively, the new equations are: − ∆T II = Ke S p2 2 λ2 Equation 5.20 ∆T III = Kλe −0.5 Sp Equation 5.21 The distance to an isotherm Sp for a given flame burning in a continuous fuel bed such as a plant can be found by making ∆T equal to the temperature required and solving equation 5.14 144 or 5.16 to find Sp, which is equal to (P -f). After incorporating equation 5.15, this equation within zone II is: S p = − 2λ (λ − f ) ln ∆T K Equation 5.22 The equation within zone III is rearranged from equation 5.16 to give: Sp = C ∆T Equation 5.23 The isotherm lt for a given flame without Zone I can be found by making ∆T equal to the temperature required and solving equation 5.20 or 5.21 to find Sp. Within zone II this equation is: [ ] S p = − 2λ2 . ln(K∆T ) Equation 5.24 Within zone III the equation is: Sp = Kλe −0.5 ∆T Equation 5.25 5.22 Plume angle In order to determine the temperature at the point where the heated plume enters a plant, the separation between the heat source and the point of entry (Ox, Oy) must be known. This distance P was shown in Figure 3.6 and is given in equation 3.7 and for backing flames in 145 equation 3.8. Progression of flame further into the plant after the ignition of the first leaves requires a different formulation of P , which is shown as equations 3.11 and 3.12 for ignition from the external heat source and as 3.13 for ignition by the flame burning within the plant where this is separate from the external flame. Solving equations 3.7 and 3.8, and 3.11 to 3.13 requires knowledge of the angle of the plume; the first three equations also require the value which can be found from the slope g as follows: δ = −O x tan (θ g ) Equation 5.26 The intention of this section is not to provide a full review of flame angle models, but to describe the main considerations in deriving flame angle as an input to the fire behaviour model, and present an acceptable flame angle model for this purpose. 5.221 Wind effects A range of flame angle models exist, however only a number of these have been validated. Weise and Biging (1996) tested 2 flame angle models in a Paper Birch (Betula papyrifa Marsh.) and Quaking Aspen (Populus tremuloides Michx.) vertical stick array. The models tested were Putnam (1965), Nelson and Adkins (1986) and an adjusted version of Nelson and Adkins (1986). The authors found that the models were adequate on flat ground but overestimated the flame angle to horizontal on slopes. Anderson et al (2006) provide validation for models provided by Taylor (1961), Nelson (1980) and Albini (1981a) using five different datasets; one of which has a shrub or herb component and the rest of which are composed of litter, wood shavings or heartwood. Their findings did not identify a consistent trend as did the previous study, but identified Albini’s (1981a) model as the best of those studied; although it required knowledge of flame height as an input. Alexander (1998) notes that no explicit model exists to describe the angle of a plume, so in the formulation of his crown fire model he focuses on examining and validating Van Wagner’s (1973) model for flame angle as an approximation of the plume angle. As this model has been used to estimate the plume angle in the same context as that of this thesis, the same approach will be taken here. 146 The angle of a flame to the horizontal is a balance of the intensity of the flame and the strength of the wind. The energy produced by the flame causes a vertical flux of buoyant air, and this vector is balanced by the horizontal vector produced by the wind. Van Wagner (1973) expressed flame angle f as: f = tan-1 (K(b IB/u3)0.5) Equation 5.27 Where K is an empirical correction added by Alexander (1998), b is a temperature dependent buoyancy term, u is the wind speed (m/sec) at mid flame height and IB is Byram’s fire line intensity (kW/m). As the temperature of the plume decreases with distance from the heat source and u increases with height in the forest profile (Cionco et al 1963); it is reasonable to assume that the plume angle decreases with height. However, for the purpose of simplicity at this time, assumption 5.1 will be made, consistent with the approach of Alexander (1998) and others. Plume angle remains equal to flame angle. Assumption 5.1 The value K was assumed by Van Wagner (1973) to be equal to one; however based on 54 experimental burn results collected by Fendell et al (1990), Alexander (1998) approximated K as 0.345. Due to the small effect of temperature, Alexander also suggested that the buoyancy term b can be approximated to a standard value of 0.0256. The metric form of Byram’s (1959) fire line intensity is: IB = 258 2.17 Equation 5.28 Where is the flame length (m). Substituting this equation and the values for K and b into equation 5.27 and simplifying gives flame angle as a function of flame length and wind speed. 147 θ f = tan −1 66.048λ 2.17 0.345 u3 0.5 Equation 5.29 5.222 Slope effects The angle of the flame to the slope is equal to the angle of the flame from the horizontal minus the angle of the slope. When considering fire spread into a single plant where the canopy is disconnected from the ground by a stem however, the slope is only considered in reference to the flame beneath the plant, as flame propagation through the plant occurs above the ground. For flames that do originate from the ground and therefore do not allow for air entrainment from directly beneath, slope also has an impact on flame angle by blocking air entrainment from the side. Rothermel (1984) described the phenomenon of “flame attachment” to a slope, where flames burning on slopes were observed to lean in toward the slope and exhibit a smaller angle to horizontal than expected from the fire intensity and wind conditions. Above a certain level of slope which Rothermel gave as 45o, flames were observed to align themselves with the slope regardless of wind conditions. Rothermel explained the phenomenon as the in-draft to the flame being blocked on the uphill side. The effect of slope in blocking air entrainment can be estimated geometrically for line and point fires using the process described below and by referring to Figure 5.3. The following assumptions apply: The slope is flat and constant for that point and can be approximated by a tilted plane. Assumption 5.2 148 The effect of slope on flame angle is relevant at low wind speeds only. At higher wind speeds the blocking of air entrainment by the slope is insignificant in comparison to the overall air movement. In this way, the slope-determined flame angle represents a maximum flame angle possible for that slope; or for fires burning down-slope and against the wind, a minimum. Assumption 5.3 f g Figure 5.3. Angles used for estimating slope effects on flame attachment. On flat ground, air entrainment is available from 180o for a line fire. If a slope is added, it removes a percentage of the air entrainment available from the uphill side equal to the percent slope, so that the flame angle becomes fl: θ fl = 90 1 − Direction of slope θg 90 Equation 5.30 n For a point fire, air entrainment is available from a 360o radius so a slope does not block as much air entrainment to the flame as it does to a point fire. Figure 5.4. The angle of the line b relative to a line a normal to the slope is denoted by n. 149 At an angle normal to the slope, the slope has 0% of its possible blocking effect. At an angle varying from a line normal to the slope by slope blocking air entrainment ge n degrees (Figure 5.4), the effective amount of becomes: θ ge = sin −1 [cos(θ n )sin (θ g )] Equation 5.31 The percentage of the potential blocking effect B is the average value of the curve produced when ge/ g is plotted against n. If this is calculated in one degree increments for n from - 90o to 90o, the equation is: θ n = 90 B= θ n = −90 θ ge 181 Equation 5.32 Equation 5.30 can then be adjusted for point fires with the inclusion of B to give θ fp = 90 1 − fp: Bθ g 90 Equation 5.33 The main limitation to the flame angle equations at this stage is that there is no mechanism for determining at what point a downslope wind can overwhelm the effects of equations 5.30 and 5.33 for fires burning downhill. Consequently, the model is still at this stage not recommended for fires in this situation (downslope but burning with the wind behind them). 150 5.23 The Critical State Test The potential for ignition of leaves at point Ox, Oy where a heated plume from a surface fire intersects a plant is shown in Figures 5.5 and 5.6. These constitute the “critical state test” from Figures 3.6 and 3.8. Figures 5.5 and 5.6 show the general process for calculating ignition used throughout a plant. To reduce the number of calculations overall, ignition of the first leaves can be calculated without setting a value for t, so that equation 3.1 is substituted for equation 3.15 in Figure 5.6. 5.231 Depth of ignition The depth of ignition is determined by the distance to which the flame can extend the ignition isotherm into the plant. The temperature of the ignition isotherm (equation 3.15) is limited by P and for the leaves. The effect of is to determine for the leaf dimensions and moisture the temperature that is required to ignite the leaf within the calculation time frame. It is not possible to rearrange equation 4.10 to find temperature, but the distance along the plume into the plant until the temperature of the endotherm is reached given the ambient air temperature can be found from equation 5.23: − (λ − f S pp Kλe 2λ = P −T ) − Pα Equation 5.34 This equation ignores any blocking effects due to the porosity of the fuel. The potential depth of ignition is the minimum of Spp and the distance along the plume, either from where it enters the plant canopy or from the flame base in the plant to where the plume leaves the plant canopy (available plume pathway Ppa). This distance is greater than or equal to Sl, the mean spacing between leaves. As it should not be assumed that the temperature of the endotherm will be high enough for leaves to ignite due to constraints imposed by , an alternative is to divide the minimum of Ppa and Spp (P ) into n steps. Ignition can then be calculated using the model at Figure 5.6 for each step in addition to P ; the depth of ignition is equal to the sum of the steps where ignition was successful. The following section presents the methodology for finding the distance along the plume through a plant canopy. 151 Slope Fireline length Wind speed Surface flame length Line fire T θ f = tan −1 0.345 0 .5 66.048λ2.17 u3 θ fl = 90 1 − Equ. 5.29 F θg θ fp = 90 1 − 90 Equ. 5.30 Bθ g 90 Equ. 5.33 Ox, Oy Flame in plant disconnected from ground flame δ = −O x tan (θ g ) T Pα = (O F − O y ) + (O xn − O x ) 2 yn 2 Equ. 3.13 Equ. 5.26 Fire backing against the wind T f = min{Equ. 5.29, Equ. 5.30/5.33) F f = min{Equ. 5.29, Equ. 5.30/5.33) Pα = (O + δ )sin (90 + θ g ) yn sin (θ f − θ g ) Equ. 3.11 P Pα = (O yn [ + δ )sin (90 − θ g ) sin (180 − θ f ) + θ g ] Equ. 3.12 Figure 5.5. Determining P 152 P t Slope Fireline length Wind speed Surface flame length T θ f = tan −1 0. 345 0.5 66.048λ2.17 u3 Mean surface fuel particle diameter Line fire θ fl = 90 1 − Equ. 5.29 F θg θ fp = 90 1 − 90 Equ. 5.30 Bθ g 90 Tr = 0.871d 1.875 Equ. 5.33 Equation 5.13 Ox, Oy Flame in plant disconnected from ground flame δ = −O x tan (θ g ) T Pα = (O F − O y ) + (O xn − O x ) 2 yn 2 Equ. 3.13 Fire backing against the wind T f Surface >P Equ. 5.26 = min{Equ. 5.29, Equ. 5.30/5.33) T F f F = min{Equ. 5.29, Equ. 5.30/5.33) Pα = (O yn ∆T III = + δ )sin (90 + θ g ) sin (θ f − θ g ) Equ. 5.16 Equ. 3.11 Pα = P (O yn [ C Sp + δ )sin (90 − θ g ) sin (180 − θ f ) + θ g ] 2 ∆T II = Ke (−α ( Pα − f ) ) Equ. 3.12 Equ. 5.14 Leaf surface area to volume ratio Leaf moisture ψ = 97805.26T 2.10 m θc + 6452280.04T − 2.40 Equ. 4.10 Leaf silica-free ash content TλPα ≥ P, (Tr − t ) ≥ ψ , (Tr − t m ) ≤ t Equ. 3.15 P = 354 – 13.9ln SFA - 2.91 (ln SFA) 2 Equ. 4.1 T IGNITION F NO IGNITION Figure 5.6. The critical state test 153 5.24 Dimensions of the plume pathway In the situation where a surface fire is heating a plant or a burning fuel stratum is heating the plants above it, the part of a plant above the heat source exposed to the highest temperatures will be the base, shown on Figure 5.1 as sides E and F. The straight pathway that the plume follows through the plant from the base is denoted as Px, where the subscript x refers to the number of the pathway as described below. The origin of Px is given by the coordinates Ox and Oy measured from the plant stem base as in Figure 5.7. The coordinate Oy is always positive, but if the origin lies to the left of vertex f, Ox will have a negative value. While sides C and D of the plant crown are always vertical, A, B, E and F can connect to sides C and D with either acute or obtuse angles. Given these limits, there are six different straight pathways for a heated plume to penetrate a plant from below. For flames where f>90o , the sides of the plant can be reversed. The six different pathways are: P1 Plume enters through E and exits through A P2 Plume enters through E and exits through B P3 Plume enters through E and exits through D P4 Plume enters through E and exits through F P5 Plume enters through F and exits through B P6 Plume enters through F and exits through D The six scenarios describe the path of a heated plume as a straight, one-dimensional line. In doing so, the following simplifications are made: i) Flame is assumed to originate from a point source rather than an area ii) Lateral force exerted by the wind is assumed to be constant both over time and over the vertical profile of the plant. Wind speed is not adjusted for height in the plant. 154 iii) The plume itself is treated as a non-turbulent straight line. In reality, heated plumes on this scale result from turbulent diffusion flames and the motion of the plume although approximately linear has been shown to be subject to the action of vortices caused by shear stresses imposed by the interaction of the wind and the surrounding obstructions, including the foliage of the plant (Morvan 2007). Although models exist that can calculate this turbulence (e.g. Croba et al. 1994, Linn 1997, Dupuy and Morvan 2005), these require calculations in three dimensions and are therefore very slow to operate and cannot provide answers within a useful operational time frame. By designating the plume as a one-dimensional straight line, we are in fact referring only to an approximation of its central axis. It is unknown to what extent the three factors listed would modify the results of this model, however due to the complexity of modelling these factors they will not be considered here. This is an area where future research may improve on the model. The heated plume of the flame shown in Figure 5.7 below is represented by the bold arrow extending diagonally upward from the slope (bold line), entering the plant through side E and leaving the top of the plant through side A. • The base of the flame heating the plant is a diagonal distance Sx cm along the slope to the left of where the plant stem connects to the ground. • The slope raises the plant in relation to the flame base by a distance of 1 cm, so that the plant is raised higher above the flame than it would be on level ground. The diagonal distance along the plume from the flame base to a line level with the base of the plant stem is p cm. The ground directly below the Ox, Oy is cm above the base of the plant stem; is therefore negative on an uphill slope and positive on a downhill slope. • The point at which the plume enters the plant is denoted by the Greek letter alpha , having the coordinates Wx cm to the right of vertex d and Wy cm above vertex d; and 155 Ox to the right of the stem and Oy above the stem base. As in Figure 5.7 is below vertex d and to the left of the stem, both Wy and Ox are negative. • The total distance through the plant along this pathway is Px cm, where the x subscript refers to the pathway number as described below for pathways P1 to P6. • The distance from the point where the plume leaves the plant to the point where it reaches a horizontal line level with vertex a of the plant is Px cm, where the subscript x refers to the side of the plant that is being crossed, in this case side A; and the subscript • indicates that the plume is leaving the plant. The distance along the plume from to a horizontal line marking vertex a is Pp cm. It should be noted that this represents the initial plume pathway, and that this pathway is altered as parts of the plant ignite and change the buoyancy of the flame. Px Px Pp Wy Oy P 1 p Sx Wx Ox Figure 5.7 Plume pathway nomenclature 156 5.241 Finding the coordinates of the plant ignition point The coordinates of the point at which the plant ignites are needed before the dimensions of the plume pathway can be calculated. These coordinates can be pre-determined by calculating fire spread through the plant at different points such as at each of the vertices d, e and f, and at points midway between d and e, and between e and f. The time from exposure of the vertex d to the flame until the time of ignition equal to p in seconds is plus the time taken for the point of intersection between the plume and the plant to progress laterally to . As Wx is measured in cm and R in km/h, the equation is: ψ p =ψ + 0.036W x R Equation 5.35 The earliest ignition will occur at the minimum value of p and using the methodology described in the previous section. Other ignition points however may produce greater flame lengths and therefore be more important to understanding fire behaviour in the fuel complex. The coordinates Wx and Wy can be converted to coordinates Ox and Oy measured from the base of the plant stem as below: Ox = −w + Wx 2 Equation 5.36 Oy = H e + Wy Equation 5.37 Ox and Oy describe the origin of the plant flame. After the initial leaves that caught fire have burnt out to a distance P into the plant along the line of the plume, the values Ox and Oy are then adjusted to Oxn and Oyn by finding the horizontal and vertical dimensions of P as given in equations 3.9 and 3.10. 157 5.242 Dimensions of the plume pathway To find Sx, measure a right angle triangle with the vertices marked by the flame base, the intersection of a vertical line extending down from and with a horizontal line extending across from the flame base. The horizontal separation between these two points can be found by the relationship x = lcos( ), where l is equal to P and equal to this value minus Ox (as Ox is negative when is equal to f. The value Sx is is left of the plant centre) divided by the cosine of the slope: Sx = Pα cos(θ f ) − O x cos(θ g ) Equation 5.38 Pp is found from the relationship l = y , where y is the vertical distance from sin(θ ) to the peak of the plant and is the flame angle f. Pp = H p − Oy sin(θ f ) Equation 5.39 The value of was given in equation 5.26. 1 is equal to: δ 1 = S x sin (θ g ) Equation 5.40 The distance along the plume from the point where it leaves the relevant side to a horizontal line extending from vertex a at the top of the plant is equal to P and can be found by creating a triangle with vertex one at the point where the plume leaves the plant, vertex two at a in the plant and vertex three where a horizontal line from a intersects the plume. The same process can be used for all three measures of P . 158 − Ox − PAΩ = H p − Oy tan (θ f ) a 2 sin 90 − [ ] sin θ f − (b − 90) ) Equation 5.41 And for backing flames: w − Ox + PAΩ = w 2 H p − Oy − tan (180 − θ f ) sin 90 − [ a 2 ] sin (180 − θ f ) − (b − 90) ) Equation 5.42 H p + δ1 PBΩ = tan (θ f ) − S x cos(θ g ) sin 90 − sin (270 − b − θ f a 2 ) Equation 5.43 And for backing flames: H p +δ PBΩ = tan (180 − θ f ) + S x cos(θ g ) sin 90 − sin (90 − b + (180 − θ f PDΩ = )) w − Ox tan(θ f 2 sin(θ f ) H p − Oy − a 2 ) Equation 5.44 159 And for backing flames: w + O x tan (180 − θ f 2 sin (180 − θ f ) H p − Oy − PDΩ = ) Equation 5.45 With the exception of P4, The various pathways within the plant can be found using the equation: Px = Pp − PxΩ Equation 5.46 P1 was shown in Figure 5.7 and the remaining pathways are shown in Figures 5.8 to 5.12; the equation for each relevant pathway is provided with each Figure. Pathway 1. Pathway 1 is shown in Figure 5.7 P1 = Pp − PAΩ Equation 5.47 160 Pathway 2. PB B P2 E Figure 5.8 Flame pathway 2. P2 = Pp − PBΩ Equation 5.48 Pathway 3. PD P3 D E Figure 5.9 Flame pathway 3. P3 = Pp − PDΩ Equation 5.49 161 Pathway 4. P4 2 E F 1 3 Figure 5.10. Flame pathway 4. P4 is not calculated from Pp, so equation 5.46 cannot be used. In this case, the length is found by producing a triangle with vertex one at the flame base, vertex two where the plume exits the plant and vertex three where a line following side F extends downward until it meets a horizontal line extending from the flame base. The vertices are marked as numbers 1 – 3 on Figure 5.10. The Sine Rule gives the length from vertices one to two; P4 is this length minus P. H c + S x sin (θ g ) f sin 90 + f 2 tan 2 − Pα f sin 90 − − θ f 2 S x cos(θ g ) − P4 = Equation 5.50 162 Pathway 5. B PB P5 F Figure 5.11. Flame pathway 5. P5 = Pp − PBΩ Equation 5.51 Pathway 6. PD P6 D F Figure 5.12 Flame pathway 6. 163 P6 = Pp − PDΩ Equation 5.52 5.243 Identifying which flame pathway will be followed The angle of the plume, the dimensions of the plant and the values Ox and Oy determine which pathway the plume will take through the plant. To identify the initial pathway and thereby the maximum depth of foliage that can be instantaneously ignited, we first find the angles from the point Ox, Oy to the various vertices of the plant as given in fig. 4.14. The following equations were derived using the basic trigonometric ratios. The angle to vertex a is given by: θ1 = tan −1 H p − Oy − Ox Equation 5.53 And for backing flames: θ1 = 180 − tan −1 H p − Oy Ox Equation 5.54 The angle to vertex c is given by: θ 2 = tan −1 H t − Oy 0.5w − O x Equation 5.55 164 And for backing flames: θ 2 = tan −1 H t − Oy 0.5w + O x Equation 5.56 The angle to vertex e is given by: θ 3 = tan −1 H e − Oy 0.5w − Ox Equation 5.57 The angle to vertex f is given by: θ 4 = tan −1 H c − Oy − Ox Equation 5.58 And for backing flames: θ 4 = 180 − tan −1 H c − Oy Ox Equation 5.59 Given these limits, the flame pathway can be identified using the decision tree in Figure 5.13. 5.25 Preheating Preheating of fuels occurs when a flame heats the fuel and drives off some of the moisture, but equation 3.1 is not satisfied and ignition is yet to occur. In the model framework 165 proposed, preheating can be calculated at each point for each time-step preceding the current moment using assumption 5.4 below. The proportion of moisture removed from a leaf during a single time-step is equal to the proportion of the ignition delay time taken up by the time step. Assumption 5.4 Mathematically, this is: Ph = t ψ Equation 5.60 Inherent in this assumption is a secondary assumption that the direction of the plume remains constant during this time step. 166 f <90 T F Ox<0 Ox>0 T F T F f< 2 T f< 1 P6 T f< 2 T T T No ignition T P6 F T F P1 f> 2 P1 P5 T T F P4 P5 P2 F f> 4 P3 F F f> 3 P2 F f< 4 F f> 1 F f< 3 f> 2 T No ignition P3 F P4 Figure 5.13. Decision tree for identifying the plume pathway through a plant crown heated from the base 167 As each successive time-step removes more moisture from the leaf, the amount of preheating by any given time is equal to the sum of all the values of Ph that have occurred before that moment. This value directly reduces the ignition delay time of the leaf according to: ψ n = ψ (1 − {Ph1 + Ph 2 ...Phn −1 }) Equation 5.61 Where n is the ignition delay time adjusted for the amount of preheating that has occurred by time n, and Ph1 to Phn-1 are the amounts of preheating of the location that have occurred in each preceding time step. Equation 3.15 is then altered to incorporate equation 5.61 to become: TP P, (tr – t) n, (tr – tm) t Equation 5.62 The model presented at Figures 5.5 and 5.6 can then be adjusted as shown in Figure 5.14. 168 P t Slope Fireline length Wind speed Surface flame length T θ f = tan −1 0.345 0.5 66.048λ 2.17 u3 Mean surface fuel particle diameter Line fire θ fl = 90 1 − Equ. 5.29 F θg θ fp = 90 1 − 90 Equ. 5.30 Bθ g 90 Tr = 0.871d 1.875 Equ. 5.33 Equation 5.13 Ox, Oy Flame in plant disconnected from ground flame δ = −O x tan (θ g ) T Pα = (O F − O y ) + (O xn − O x ) 2 yn 2 Equ. 3.13 Fire backing against the wind T f Surface >P Equ. 5.26 = min{Equ. 5.29, Equ. 5.30/5.33) T F f F = min{Equ. 5.29, Equ. 5.30/5.33) Pα = (O ∆T III = + δ )sin (90 + θ g ) yn sin (θ f − θ g ) Equ. 3.11 Pα = P (O yn [ C Sp Equ. 5.16 + δ )sin (90 − θ g ) sin (180 − θ f ) + θ g ] 2 ∆T II = Ke (−α ( Pα − f ) ) Equ. 3.12 Equ. 5.14 Leaf surface area to volume ratio Leaf moisture ψ = 97805.26T 2.10 m θc + 6452280.04T −2.40 Equ. 4.10 Leaf silica-free ash content ψ n = ψ (1 − {Ph1 + Ph 2 ...Phn −1 }) Equ. 5.62 P = 354 – 13.9ln SFA - 2.91 (ln SFA) 2 Equ. 4.1 TλPα ≥ P, (Tr − t ) ≥ ψ n , (Tr − t m ) ≤ t Equ. 5.63 T IGNITION F NO IGNITION Figure 5.14. The critical state test adjusted for preheating 169 5.26 Finding Depth of Ignition The factors examined so far are sufficient to find the depth of ignition within a plant from a flame with known dimensions and coordinates. The process can be described in the following steps: 1. Identify the coordinates for the base of the existing flame 2. Find the length of the flame, then the flame angle using equations 5.29, 5.30 and 5.33. Consider the slope if the flame origin is external to the plant crown. 3. Find Spp using equation 5.34 and Ppa using equations 5.38-5.59 and Figures 5.5, 5.6, 5.13 and 5.14. Take the minimum of these to find P , then divide by n steps. Larger values of n will give greater precision but will also increase the calculation time. The model developed for this study uses 10 steps. 4. For the initial ignition, find the temperature at each step using equations 5.14 or 5.16, then use this to find the ignition delay time with equation 4.10. Ignition occurs if the flame duration is greater than the flame duration time (Figures 5.5 and 5.6). Flame duration for dead surface fuels can be found using equation 5.13; the model for flame duration in plants has yet to be developed. 5. Following ignition, an additional calculation must be carried out for every time step preceding the moment examined. Steps 3 and 4 above are repeated to find the amount of preheating the point has received, and the ignition delay time for the current moment. The ignition delay time is then reduced by the preheating according to equation 5.61 and Figure 5.14. If the ignition delay time is less than or equal to t, the depth of ignition is increased by P /n. 6. This process is repeated for each of the n divisions to find the proportion of P that has been ignited within each successive time-step. 5.3 Plant Combustibility Flame length has been recognised as a measure of combustibility, and as section 4.33 established, the relationship between flame length and combustibility at a particle scale is highly significant. Sections 3.23 and 4.31 propose that flame length in a plant is a product of 170 the flammability of the leaves, where individual flames coalesce to form a larger flame as described by flame merging theory. On this basis, the length of a flame produced by a burning plant is determined by the factors given in Table 5.1 below; these factors are discussed in the following sections. Table 5.1. Factors and equations determining flame length from a burning plant. ‘TBD’ indicates that the equation is yet to be developed. Factor Components Model/Equations Flame length from a burning Flame length model Equations 4.18, 4.19, leaf determined by 4.21 Number of leaves burning Depth to which a flame will Equations 5.34, 5.38-5.59, simultaneously ignite Figures 5.5, 5.6, 5.13, 5.14 Maximum plume pathway Equations 5.38 – 5.59 Actual length of plume pathway TBD Number of leaves per depth of TBD plant ignited Flame length from the leaves Flame merging model TBD burning 5.31 Branch architecture As the shape of the plume has been simplified to a simple line, it becomes necessary to introduce two further assumptions to find the number of leaves burning; these are: The depth to which an ignition isotherm penetrates a branch or clump of leaves represents the proportion of the clump that will ignite. Assumption 5.5 The convective plume will only ignite those branches immediately in its path, branches on either side may ignite afterward but will not contribute to the maximum flame height. Assumption 5.6 171 The number of leaves that will burn if ignition penetrates a plant crown to depth Pi is therefore determined by: 1) Branch or clump diameter Wb and spacing Sb 2) Number of leaves within each branch The number of branches that will ignite Nbe is approximated by: N be = Pi Wb + S b Equation 5.63 As it is impossible to know the direction that fire will come from before the event, Wb is an average of three dimensions for a natural clump of leaves. The clump does not need to be produced by a single stem, but can be approximated by eye. Sb is the distance from the edge of one clump to the closest edge of its neighbour. This approximation has the characteristic that it will underestimate the number of branches ignited when Pi is small, especially when Sb is large. If Pi is equal to Wb, an entire branch will ignite, but it will be calculated as a proportion of the branch Wbi according to: Wbi = Wb Wb + S b Equation 5.64 To correct for this, the approximation can be corrected to find the actual number of branches Nb by rounding Nbe down to the nearest integer Nbi and subtracting the difference. The difference can then be compared to Wb to find the actual proportion of the branch that has been ignited. 172 N b = min ( N be − N bi ) Wb + S b ,1 + N bi Wb Equation 5.65 5.311 Number of leaves in a branch In order to find the number of leaves that will burn if one branch is ignited, it is necessary to estimate the number of leaves in a branch. This is affected by the size of the clump, the density of stems within the clump and the spacing between leaves on each stem and can be expressed as: Number of leaves = Clump size * branch density / leaf spacing Based on this concept, a simple empirical model was constructed to calculate the number of leaves in a branch. Methods Clump size, branch density and branch spacing were approximated by mean clump diameter, stem order and mean space between leaves respectively. Clump diameter was taken as the average of three measurements for each clump. Measurements were taken in centimetres from the outside edge of the leaves ignoring individuals that extended well past the edge, and following the longitudinal axis of the branch, the width at the widest point and the width at a point when the branch is rotated 90o along its axis. Stem order Os is based on the concept of stream order (Strahler 1952), and is calculated beginning at the outermost stems. An individual stem bearing leaves or clusters of leaves is a first order stem; two first-order stems connecting produce a second order stem and so on. Figure 5.15 shows a fourth order network. Stem order was used due to its simplicity and Figure 5.15. A fourth-order branch network. repeatability. 173 The mean space between leaves Sl was calculated by dividing the length of a stem segment (cm) by the number of leaves on the segment. This was measured by selecting three stem segments for each clump based on a visual assessment of less than average leaf density for the clump, average leaf density and greater than average leaf density. Measurements were taken from eight species of trees and shrubs native to the study area and representing a broad range of leaf density and sizes. Six branches were measured from each species giving a total sample of 48. For each branch, Wb, Sl and Os were measured, and the total number of leaves in the clump Nl was counted manually. Data analysis The measurements Wb, Sl and Os were combined as described into an index Nli relating to leaf number: N li = Wb Os Sl Equation 5.66 Nl was compared to Nli with non-linear regression, the R2 calculated and an equation produced to predict Nl. The percent error of the model for each sample was estimated as: % Error = Observed − Expected Observed × 100 1 Equation 5.67 Where Observed − Expected is the absolute value of the difference between the observed value and the expected value. The model error was compared to the input parameters and a Coefficient of Correlation calculated for each to determine whether any one of these provided a bias in the model. 174 Results Table 5.2 shows the range of measurements between the different species. The average branch diameter measured was 24.7 cm and clump sizes are skewed toward smaller branches. Table 5.2. Statistics of species examined Species x Wb x Sl x Os x Nl Max Nl Grevillea lanigera 8.7 0.16 2.3 359 1077 Eucalyptus pauciflora 48.8 2.89 2.8 125 375 Eucalyptus mannifera 37.0 2.17 2.5 116 348 Bursaria spinosa 16.8 0.30 3.3 462 1387 Eucalyptus niphophila 40.7 2.05 4.4 193 579 Tasmannia xerophila 22.2 0.48 2.5 254 616 Olearia aglossa 16.3 0.59 2.7 135 281 Bossiaea foliosa 7.0 0.20 4.0 597 1422 The regression of Nl against Nli is shown at Figure 5.16 for the 48 points. The fit of the data was good (R2 = 0.88) using a power function, with the level of error increasing for nanophylls such as Bursaria spinosa and Bossiaea foliosa. Linear regression produced an R2 of 0.84. The equation produced by the regression was: N l = 0.88 N li 1.18 Equation 5.68 When combined with equation 5.66, the equation for estimating the number of leaves in a branch is: WO N l = 0.88 b s Sl 1.18 Equation 5.69 175 The error of the model did not correlate with any of the input measurements (Table 5.3), indicating that none of the parameters caused significant bias to the model. TABLE 5.3. Correlation of independent variables with the model error. Figure 5.16. Linear regression of Nl against Nli Parameter Wb Sl Os Coefficient of Correlation 0.0077 -0.0037 -0.0128 Discussion The ratio of clump size x branch density to leaf spacing provided a robust predictive index for estimating the number of leaves in a branch while utilising techniques of measurement that are simple, repeatable and require no specialist equipment. While Wb and Sl are direct measurements of the clump characteristics, Os is a proxy measure. It is likely that this accounts for a significant part of the error in the model; the amount is dependent upon the strength of the correlation. It is likely that a better model to calculate the number of leaves in a clump could be developed if stem density could be better approximated. 5.32 The number of leaves burning simultaneously The actual length of the plume pathway at any time step is equal to the depth of ignition so far minus the length of pathway that has already extinguished. The depth of extinction occurring during one time-step Edt is determined by the time for which the leaves in each time-step have been burning and the flame duration in those leaves. Flame duration is found using Equation 4.21 when the burning leaves are being heated by an external source; if the external source has extinguished or an independent crown fire has 176 burnt ahead of the ground fire, the flame residence time is greatly reduced in green leaves. In the absence of a relevant equation, the model developed in this study assumes that the flame duration without an external heat source is one second. As described in section 5.26, a depth of foliage up to P cm is ignited in each time-step; this depth was referred to as Ppi in equation 3.15. Values of Ppi can be assigned to each time-step for the fuels ignited at that time, where the subscript n refers to the number of the time-step. Let {t1, t2, t3…tn} refer to time-steps 1 through to n. The total length of the plume pathway alight at any moment (Ppi) is then described by a nested “IF…” statement such as: [ ] [ ] [ ] Ppi = {IF t (t n − t1 ) < Tr , Ppi1 ,0 } + {IF t (t n − t 2 ) < Tr , Ppi 2 ,0 } + ... + {IF t (t n − t n −1 ) < Tr , Ppin −1 ,0 } Equation 5.70 As described in equation 3.16 and 3.18, the depth of extinction for each step changes the coordinates Oxn, Oyn for the flame origin in the plant. Using time-specific terms, the coordinates for the flame origin at any time are described by nested “IF…” statements: [ ] [ ] O xn = O x + {IF t (t n − t1 ) < Tr ,0, Ppi1 }cos θ f + {IF t (t n − t 2 ) < Tr ,0, Ppi 2 }cos θ f + ... [ ] + {IF t (t n − t n −1 ) < Tr ,0, Ppin −1 }cos θ f Equation 5.71 [ ] [ ] O yn = O y + {IF t (t n − t1 ) < Tr ,0, Ppi1 }sin θ f + {IF t (t n − t 2 ) < Tr ,0, Ppi 2 }sin θ f + ... [ ] + {IF t (t n − t n−1 ) < Tr ,0, Ppin−1 }sin θ f Equation 5.72 The number of leaves that are burning concurrently can be estimated using equation 5.70 to find the depth of penetration into the crown, equations 5.63 and 5.65 to find the number of branches that this entails if assumptions 5.5 and 5.6 are accepted, and equation 5.69 to find the number of leaves in each branch. 177 5.33 Flame merging 5.331 Area effects The concept of flame merging was raised under “flame length” in section 3.23 and has been studied by Thomas 1963, Thomas et al 1965, Huffman et al 1967, Steward 1970, Chigier and Apak 1975, Tewarson 1980, Gill 1990, Heskestad 1998, Weng et al 2004 and Liu et al 2009. One of the simpler models for flame merging is described below and used to find flame length. Thomas et al (1965) described the relationship between the flame to fuel dimension ratio and the maximum distance between burning fuel elements at which flame merging will occur as: λ D =9 σ3 DW 2 Equation 5.73 Where cm, is the length of the flame in cm, D is the width or diameter of the burning zone in is the distance between burning fuel elements in cm, and W is the length of the long side of the fuel bed in cm. Gill (1990) rearranged equation 5.73 and simplified it to describe the common field situation where the burning areas are circular such that D = W, or when considering a shrub, w. Flame merging can then be said to happen when fuel elements with distance of separation burning simultaneously, where are is given by: σ =w λ 2 3 w Equation 5.74 Based upon measurements from the work of Thomas et al (1965), Gill estimated the merged flame length λm from N fuel elements, each with an individual flame length λ: λm = λ . N0.4 Equation 5.75 178 The calculation difficulty in the model is reduced by scaling the model outward as flame propagates. At the scale of a branch or clump, the value of λ used is for a single burning leaf. At the scale of a plant, the value of λ is that of a burning branch or clump and at the scale of a stratum, the value of λ is that of an individual burning plant. Equation 5.74 can be dispensed with in the context of a plant by making two further assumptions. These are: Where one burning particle or clump ignites others, the flames of all burning particles or clumps will merge. Assumption 5.7 All of the flames from leaves burning in a single clump will merge. Assumption 5.8 Assumption 5.7 is supported by flame merging theory, as the mechanism for increasing flame length is the limitation of oxygen by blocking its entrainment into the flame. If the plume from one flame ignites further fuels, then the entrainment of air into the new flame is partially blocked by the pilot flame as the oxygen content in that flame is in the process of being consumed. Whether this affects flame length in the same way as lateral blocking is unknown, and it is suggested that further work be carried out to test this assumption. Assumption 5.8 is based on assumption 5.7, and on the fact that all lateral neighbours of leaves are relatively closely spaced to each other. This assumption is a simplification intended to minimise the number of calculations. 5.332 Longitudinal effects An additional process occurs along the path of the actual plume, where the merged flames accelerate the movement of hot gases and thereby extend the length of the flame. This area has been less studied, however Mitler and Steckler (1995) describe that for a vertical wall, the 179 area effect of the flame is insufficient to describe the full flame length and must be amplified proportional to the length of surface ignited. Although there are differences between this scenario and that of a flame burning through vegetation, some fundamental similarities suggest that the process is comparable and that the model should be applied in this situation as well: 1. The flame merging theory described so far addresses the effect that blocked air entrainment and heat feedback have on flame lengths, but does not address the effect that increased buoyancy due to heating from below has on flame length. This effect is common to both flames on a vertical wall and to flames burning along the pathway of a heated plume at an angle as the direction of movement of gases is the operative issue. 2. In both the wall scenario and in the plant, it is possible to achieve a distance of ignition where Ppi > m, which is not a physically plausible scenario and indicates that an additional mechanism is affecting the flame. The value m is assumed to represent zone II of the flame and therefore does not include the distance Ppi which represents the portion of foliage already ignited. The model given by Mitler and Steckler (1995) is: [ λml = (λm + aPpi ) + Ppi 4 ] 4 0.25 Equation 5.76 Where ml is the merged flame including longitudinal effects and a is a proportionality constant given as unity by Mitler and Steckler. 5.333 Combining the effects Given these formulae and assumptions, the length of a flame from a burning plant can be calculated as follows by building on the summary in section 5.32: 1. Find the number of leaves in each branch (equation 5.69) 2. Find the flame length from a single burning leaf (equations 4.17, 4.18) 3. Find Ppi (equation 5.70) 180 4. Find the number of leaves that this entails (equations 5.63 and 5.65) 5. Find m (equation 5.75), with N = the number of burning leaves and λ =the flame length from one leaf. 6. Find ml from m and Ppi These calculations suggest that the combustibility of a plant as measured by flame length is a variable value. A different flame length is derived depending upon the depth of foliage ignited, so the length of flame varies over time and any representative value of combustibility for a plant is a generalisation. 5.4 Plant Flame Sustainability The sustainability of flame in a plant is determined by both the sustainability of flame in the component leaves and by the rate at which fire spreads through the plant. Flame sustainability in the leaves is related to the cross-section area of the leaf and its moisture (Equation 4.21) if the leaves have an external heat source to assist the ongoing combustion. As discussed in section 3.23 however, removal of an external heat source reduces the sustainability and in the absence of a reliable study, the default value used here is one second. In this way, the sustainability of flame in a plant is both an inherent property and dependent up external conditions. The rate at which fire spreads through the plant affects the sustainability because faster spread occurs through the ignition of greater parts of the plant in each time step. Equations 5.71 and 5.72 show that the origin of the flame moves through the plant as the ignited section reaches extinction. Plant flame extinction occurs if either the flame is unable to continue spread through the plant or if the flame origin reaches the edge or the top of the plant. To find the duration of flame in a plant, it is necessary to model the spread of flame through the plant to identify when the flame either self-extinguishes or consumes the plant. The flame origin will move laterally through the plant as described in equation 5.71 so that Oxn increases up to a maximum value of w/2 and Oyn increases to a maximum value of [max(Hp, Ht) min(Hc, He)]. As the subscript n refers to the number of the time-step, n increases by a value 181 of one with each step. The value of n where the flame origin reaches the right hand side of the plant, i.e. where O xn = w may be termed xn and the value where the flame origin reaches the 2 top of the plant i.e. O yn = max (H p , H t ) − min (H c , H e ) may be termed yn. The total flame residence time in the plant trp is therefore equal to: [ ] t rp = t min (O xn , O yn ) Equation 5.77 5.5 Summary The spread of fire through a plant is governed primarily by the quantity of the plant that can be ignited at any given time. Greater ignitability whether resulting from easier-to-ignite leaves or from a larger contributing flame produces greater depths of fuel alight at any given time, which produces greater flame lengths and therefore combustibility. By igniting larger parts of the plant at a time, the fuel is consumed more quickly so that the sustainability of the plant is reduced. Factors that produce larger flames or closer flames to the plant such as the combustibility of lower fuel strata or the density of plants in the array act to increase the ignitability and combustibility of the plant but reduce the sustainability. Factors inherent to the plant that affect this include the dimensions and moisture of the leaves and the density of the leaves in the plant. Thicker, moister leaves both reduce the ignitability (and therefore combustibility) and increase the sustainability of the plant flame. Longer leaves increase the combustibility, wider leaves increase the sustainability and denser packing of leaves increases the number of leaves ignited with a given flame penetration, thereby increasing the combustibility of the plant. The Leaf Area Index (LAI) is the one-sided area of leaves in a plant divided by the area of the plant on the ground (Campbell 1987). This value is therefore larger in plants with denser leaf packing and larger leaves, which raises the possibility that there may be a correlation between this and the flammability of the plant. LAI does not however allow for the thickness of the leaves or their moisture content, so there will be an unknown level of noise introduced 182 by these factors. It also should not be assumed that the flammability of a plant correlates with forest flammability, as other factors are also involved in this case. These factors will be explored in the next chapter. 183 Chapter 6 Forest Flammability Context As established in chapter one, the relationship between fuels and fire is not adequately described by the existing fire behaviour models. Chapter two explored the reasons for this, identifying that existing models depend upon the assumption that fuel quantity related directly to the flammability of a fuel array. This assumption was examined in chapter two and shown to be invalid, identifying the need for a model that could accurately quantify the relationship. Chapter three described a conceptual model capable of addressing this by treating fire spread as a complex system where flame propagation depends on both the physical conditions imposed by the effect of the flammability of fuel particles on the ignition of new fuels, and by the spatial relationships of the fuels. Chapter four described experimental work to quantify the flammability of leaves as the basic particles that compose plants; and chapter five described the mathematical framework whereby this is linked to plant geometry so that the flammability of a plant can be quantified. Chapter six concludes the modelling process by extending the mathematical framework introduced in chapter five to describe the structure of an entire fuel array up to the complexity of a forest. A number of other factors relevant to the scale of the forest are also sourced from the literature, explored mathematically or determined experimentally; so that models are provided to describe the influence of forest structure on wind speed, dead and live fuel moisture content and soil moisture content. 184 The chapter concludes by describing the way all components combine into a complete fire behaviour model, a copy of which is attached as a spreadsheet on the CD included. A worked example demonstrates how the calculations are performed. 6.1 Forest structure and geometry As the scale of calculations increases to the level of the forest, it becomes possible to reduce the detail in the geometry used to describe the fuels. The primary considerations have moved from the shape of the plant and its impact on the length of the plume pathway to the separation between fuels and the vertical dimension of each stratum. To achieve this, each fuel stratum can be visualised as a broad, flat plant. The horizontal dimension can be considered infinite and the vertical dimensions Hb and Hu representing the base and top of each stratum are the minimum of He and Hc and the maximum of Ht and Hp respectively. Branch dimensions are the same as for the component plants, but clump separation is a weighted average of the separation between clumps Sb and the separation between plants Sf. In practice, Sf is easier to measure or estimate from the centre of one plant to the next than it is to measure from the closest edges of each plant. As a result Sb and Sf do not compare directly, and the distance from the edge of one plant to the next is equal to Sf – w. Clump separation at the level of the stratum Sbs can be found as follows: S bs = S f − w + Sb w 1+ Wb + S b Equation 6.1 185 6.2 Interactions of forest structure with wind and moisture 6.21 Wind speed The velocity of wind acting on a flame in a forest is reduced by friction and turbulence created by the vegetation, so that wind speeds on the forest floor are less than those above the canopy. The effect of this on fire behaviour is that the horizontal flux on flames burning in lower strata is less than in the upper strata, so that flames burning close to the ground will be more upright than flames of an equal length burning in the canopy. 6.211 Modelling the wind profile above a surface The velocity profile of the atmospheric boundary layer above a surface such as a forest or a fuel bed is reasonably well understood. Under stable atmospheric conditions, the widely accepted form (e.g. Schlichting 1979, Ch20) is: u 1 z = ln uτ k z 0 Equation 6.2 Where u is the horizontal wind velocity at height z, k is von Kárman’s constant ( k ≈ 0.4 ), z0 is the intercept on the z-axis and u is the “friction velocity” defined by: uτ = τw ρ Equation 6.3 Where w is the shear stress and is the density of the air. A simpler estimation of the logarithmic profile (Whiteman 2000) is: u 2 = u1 ln(z 2 / z 0 ) ln(z1 / z 0 ) Equation 6.4 186 Where u1 and u2 are the horizontal wind velocities at two different heights z1 and z2 above the surface (height z0 = 0.1m). 6.212 Modelling the wind profile within a forest or porous fuel bed While such profiles facilitate the calculation of wind speed above a layer and are therefore useful for situations involving flames above a fuel bed under stable conditions (Viegas and Neto 1991), they do not explain the variation in wind speed within or beneath a porous fuel array such as winds experienced beneath a forest canopy. This topic has been studied to some degree in Australia, however the majority of the literature is confined to non peer-reviewed sources and the quality of these studies varies widely. McArthur (1966b) reported wind speeds at 10m and 1.5m (V10 and V1.5) heights in two different forest structures, 13-19m and 32-48m; the second of these had been earlier printed as a graph in McArthur (1962). Linear regression of the seven data points for each forest produced equations 6.5 and 6.6 below, both with R2 values of 0.99. No details were given as to methodology or the total number of measurements taken. The average ratio of V10 : V1.5 (wind reduction factor) was 2.7 (S.E. = 0.4) for low forest and 3.8 (S.E. = 0.6) for tall forest. V1.5(13−19 m ) = 0.32V10 + 0.74 Equation 6.5 V1.5(32− 48m ) = 0.22V10 + 0.59 Equation 6.6 McArthur (1967, p.8) produced three graphs of wind reduction without equations or stated wind reduction factors and it is not clear which of these is assumed in his rate of spread model. No data is provided to support the lines and no information is provided as to their sources. 187 Visual approximations from the hand-drawn graphs give wind reduction factors of approximately 2.5 for short open forest or woodland, 3.3 for medium height (13-19m) forest and 5 for tall forest (32m or taller). Tran and Pyrke (1999) produced a more thorough study of wind reduction factors in a woodland environment, recording specific details of methodology, site descriptions, data obtained and statistical analysis. They conclude that the wind reduction factor in this woodland community is between 3 and 4, and that a denser canopy produces a larger factor. Gould et al (2007a) recorded wind speeds in E. marginata forest at 30m (above the canopy), 10m, 5m and 1.5m at two different sites. The Dee Vee site had a denser tree canopy but thinner midstorey and lower understorey than the McCorkhill site. Site descriptions and methodology were described but data was only displayed for selected instances and no statistical analysis or reporting of important statistics such as R2 was given. The wind profile at Dee Vee demonstrated a greater drop in wind speed between 30m to 10m than the more open McCorkhill canopy, and a lesser drop in the lower measurements amongst the lighter understorey (pp. 43-45). No V10 : V1.5 wind reduction factors or equations were produced, however the use of additional anemometers at other heights demonstrated the change in shear stress between the different fuel strata. The peer-reviewed literature has focused more on models of wind speed with height rather than wind reduction profiles relevant only to a specific height. Albini (1981b) produced an exponential model using conservation of energy principles that gave the wind speed profile as a function of shear stress and an empirical constant related to wind speed above the canopy. While such a model is informative, like the wind reduction profiles it is only of use in forest structures where empirical measurements have provided measures of shear stress and the empirical constant. It also works on the assumption that the drag area density is uniform across the profile and is therefore of more limited use in forests where this is not the case. Cionco et al (1963) produced an exponential model of wind speed with height that was dependent upon an attenuation coefficient termed the Canopy Flow Index determined empirically from shear stress profiles. The model has the form: 188 u = u H exp γ z −1 H Equation 6.7 Where u is the horizontal wind speed at height z within the canopy, uH is the horizontal wind speed at the top of the canopy and H is the average height of the canopy. The empirical nature of again places restraints on the application of the model across different forest types where field studies have not been carried out, and attempts have been made to use standard values for generalised forest descriptions (e.g. Greene and Johnson 1996) which perhaps generalise the situation too far for use in fire behaviour modelling. In contrast to empirical values required by Albini (1981b), has the property that it is related to the physical components of forest structure vegetation density and flexibility (Cionco 1972). Vegetation flexibility was described in three categories: 1. Rigid (e.g. wooden peg arrays), 2. Semi-rigid (will bend with the wind but spring back and forth) and 3. Flexible (will streamline with the flow, Cionco 1972). Higher values of were often associated with the more flexible arrangements (Wang and Cionco 2007). Vegetation density has two components, the leaf density of the actual plants and the density of plants on the ground. Both of these elements can be combined into the measure of Leaf Area Index (LAI – the one-sided area of leaf divided by the area of ground taken up by the plants) for a vegetation community, and Wang and Cionco (2007) have suggested (based on three data points) that positively correlates with LAI. As most forest situations fall into the semi-rigid classification of vegetation flexibility, it may be reasonable to give this some credence and approximate by disregarding flexibility and using LAI as an independent measure. Certainly this will allow a more objective approximation than that suggested by Greene and Johnson (1996) while making the model useable in field situations with the more accessible LAI measures. 189 To better estimate an approximation, LAI values were collected from the literature and fitted to the values of given in Table 3 of Wang and Cionco (2007) where they were missing. The forests described as ‘gum-maple’, ‘maple-fir and ‘oak-gum lacked a precise description, so values were collected for forests that had at least one component of these present. An additional value of for ‘jungle’ was sourced from Pinker and Moses (1981). All values are given in table 6.1, and the data is shown in Figure 6.1. Table 6.1. Published data for Canopy Flow and Leaf Area Indices. All values are taken from Wang and Cionco (2007) except for the second value for ‘jungle’, which was taken from Pinker and Moses (1981). Canopy LAI Flexibility LAI Source Corn 2.40 3.00 Flexible Wang & Cionco 2007 Corn 4.10 2.90 Flexible Wang & Cionco 2007 Wheat 2.50 4.35 Flexible Bavec et al 2007 Forest 1.70 1.00 Semi-rigid Wang & Cionco 2007 Gum-Maple 4.42 3.89 Semi-rigid Sampson et al 2001 Gum-Maple 4.42 7.10 Semi-rigid Gower et al 1999 Jungle 3.84 6.00 Semi-rigid Clark et al 2008 Jungle 4.04 6.00 Semi-rigid Pinker and Moses 1981, Clark et al 2009 Maple-fir 4.03 8.60 Semi-rigid Thomas and Winner 2000 Maple-fir 4.03 7.10 Semi-rigid Gower et al 1999 Maple-fir 4.03 7.80 Semi-rigid Chason et al 1991 Oak-gum 2.68 4.89 Semi-rigid Gower et al 1999 Oak-gum 2.68 4.20 Semi-rigid Gower et al 1999 Oak-gum 2.68 3.87 Semi-rigid Sampson et al 2001 Spruce 2.74 5.95 Semi-rigid Homolová et al 2007 Spruce 2.74 4.80 Semi-rigid Gower et al 1997 Spruce 2.74 3.46 Semi-rigid Sampson et al 2001 Spruce 2.74 5.60 Semi-rigid Stenberg et al 1994 190 5 Canopy Flow Index 4 3 2 y = 1.031Ln(x) + 1.681 2 R = 0.371 1 0 0 1 2 3 4 5 6 7 8 9 10 Leaf Area Index Figure 6.1. Canopy Flow Index as a function of Leaf Area Index Data were compared using non-linear regression, producing equation 6.8 with an R2 of 0.37, and adjusted R2 of 0.37 and a P-value of 0.008. This equation provides a useful estimate of and negates the need for conjecture and approximation which was characteristic of the earlier studies quoted. Further work on the topic is warranted to improve the precision of the model. γ = 1.031 ln( LAI ) + 1.681 Equation 6.8 Linear regression of the data produces a slightly stronger fit (R2 = 0.38, adjusted R2 = 0.35 and P = 0.006). This model (equation 6.9) is useful for LAI values greater than or equal to one, but cannot be used in more open environments as it will produce a canopy flow index of 1.9 in an environment devoid of vegetation. γ = 0.269 LAI + 1 .90 Equation 6.9 191 6.213 Modelling the wind profile under complex canopies Cionco (1978) demonstrated the difference in the profiles of sub-canopy wind in complex canopies from that in simple canopies (Figure 6.2). Complex canopies are distinguished by horizontal layering in the different strata, producing a heterogeneous vertical distribution of vegetation density. Simple canopies by contrast have a relatively homogeneous vertical density distribution and are typified by monocultures and single-strata forest arrangements. In homogeneous canopies, wind speed decreases along a reasonably predictable pattern which is easily modelled by a line. Examination of the profiles in Figure 6.2 however shows that when the fuel array is heterogeneous due to spaces between or beneath strata or to strata of differing densities, the variation in wind speed does not follow a predictable course. Three of the profiles (jungle forest, larch trees and soybeans) all exhibit a similar drop in wind speed in the canopy. The jungle profile continues with this trend until about three quarters of the way to the forest floor, where the air is still. In contrast, both Larch trees and Soybeans demonstrate very little change in wind velocity below a point half to three quarters of the way up into the canopy. In this context, does not have to be treated as a constant throughout the forest profile but can be adjusted to different heights in the canopy (e.g. Wright 1965, Pinker and Moses 1981). This approach has merit when attempting to understand the effect of different elements of the forest array as it can identify the effect of individual strata on the wind speed profile. The process of calculating for each stratum is the same as it is for an entire array; except that the LAI is calculated only for that stratum and then converted to Figure 6.2. Figure 2 of Cionco (1978), showing measured wind profiles from four different complex canopies. using equation 6.8 or 6.9. To calculate the entire forest wind profile, the method proposed is to calculate for each stratum, then assume either no loss in velocity 192 between strata or alternatively assume a logarithmic loss such as that described by equations 6.2 or 6.4. Figure 6.2 demonstrates that for the three profiles given for citrus, larch and soybeans the assumption of no loss of wind speed is reasonable. Leaf Area Index can readily be calculated from the fuel descriptors described earlier by combining equations 4.15 and 5.69 to find the one-sided area of leaves in a branch, then calculating the number of branches in a plant and the area of ground occupied by the plants The area of leaves in a branch Alb is: WL WO Alb = l l × 0.88 b s Sl 2 1.18 Equation 6.10 Where Wl is the width of the leaf, Ll is the length of the leaf, Wb is the diameter of the branch, Os is the stem order within the branch and Sl is the separation between leaves. The number of branches in a plant can be found by dividing the volume of the plant by the volume of a branch and accounting for the spacing between branches. Plants can be visualised from chapter 5 as cylinders with cones on either end, and the volume of a plant Vp is: Vp = w π 2 2 × (H p − H t ) 3 w + π 2 2 × (H t − H e ) + w π 2 2 × (H e − H c ) 3 Equation 6.11 All other nomenclature is given in chapter 5 at Figure 5.1. The volume of a branch and associated spacing between branches can be calculated by considering each branch as a sphere with diameter equal to Wb plus half of the gap between the branches Sb. The LAI for a plant then is then equal to the volume of a plant crown divided 193 by the volume of a branch, times the number of leaves in a branch and all divided by the horizontal area of the crown: Vp LAI p = 4 Wb + 0.5S b π 3 2 3 × Alb π 0 .5 w 2 Equation 6.12 The LAI for an entire stratum LAIs is equal to the sum of the LAIp for each species multiplied by its percentage composition and by the percent cover of that stratum. Wind speed is calculated in steps progressing downward from the top of the canopy using equation 6.7 and the relevant value of for each stratum. The value used for UH in the canopy is the wind speed at 10m in the open and H is the height of the top of the canopy. UH and H are then adjusted in the lower strata, so that UH is the wind speed at the bottom of the stratum above, and H is the height of the stratum being measured. 6.22 Dead fuel moisture The moisture content of dead fine fuels (DFMC) has been addressed with a wide range of models (e.g. McArthur 1962, 1966a, 1967, Simard 1968, Fosberg and Deeming 1971, Van Wagner 1972, Anderson et al 1978, Nelson 1984, Rothermel 1983, Sneeuwjagt and Peet 1985, Rothermel et al 1986, Péch 1989, Matthews 2006, Gould et al 2007a, Sharples et al 2009, Matthews et al 2010). These range in sophistication from book-keeping approaches such as Sneeuwjagt and Peet and empirical models such as those of Gould et al and McArthur through to detailed physical models. If the role of forest structure and composition in affecting fire behaviour is to be understood, it is necessary to quantify its effects on dead fuel moisture as one of the sub-models. 194 6.223 Effects of forest structure on dead fine fuel moisture Physically, dead fuel moisture content is a product of the effects of latent heat, vapour exchange and precipitation (Viney 1991). Water is added to fuels through precipitation and the absorption of humidity or water in surrounding soils, and lost as heat converts it into vapour form. Fuels become drier or wetter as they approach equilibrium with the surrounding energy and water environment, exhibiting a delay depending primarily upon the surface area to volume ratio of the particle (Viney and Catchpole 1991). This process is examined in detail by Viney (1992) and Matthews (2006). While the dominant factors affecting DFMC are rainfall, air temperature, relative humidity and amount of incident sunlight, the microclimate in which the fuels are located affects the energy and water balance so that forest structure has a role in determining DFMC. While a number of such parameters exist, two dominant local influences have been recognised by a range of authors (e.g. Byram and Jemison 1943, Van Wagner 1969, Wotton 2009); these are shading and reduced wind speed. Shading by vegetation and terrain reduces the incidence of solar radiation at the soil surface and thereby, direct heating of the fuels. As discussed in the previous section, greater density of vegetation also has the effect of reducing wind speed at the ground surface, so that a humid microclimate can be maintained and heat loss via convection reduced. The effect of these factors together can maintain the air at the ground surface at a different temperature and humidity than it would be outside of the forest environment, directly impacting the moisture of the dead fuels. Van Wagner (1969) gave the temperature of the fuel Tf in relationship to the ambient temperature Ta, incident radiation and wind speed as: T f = Ta + aIe− ku Equation 6.13 Where a and k are constants specific to the fuel type. Van Wagner reported values of a as 0.035 K m2 W-1 for jack pine (Pinus banksiana) litter and 0.028 K m2 W-1 for trembling aspen (Populus tremuloides) litter. Viney (1991) used a value of 0.224 for k. 195 The saturation vapour pressure is a function of temperature defined by the Clausius- Clapeyron equation (e.g. Whiteman 2000) as: ς = ς 0e L 1 1 − Rv T0 T Equation 6.14 Where 0 is 6.11 mb the saturation vapour pressure at T0 = 273K, Rv is the gas constant for water vapour = 461 J K-1 kg-1 and T is measured in Kelvin. If is calculated for both Ta and Tf, the relative humidity at the fuel is (Wotton 2009): RH f = RH ς a × 100 ς f Equation 6.15 In order to adequately account for the effect of forest structure on fire behaviour, it is necessary to consider in some way the influence of shading by the forest canopy, and ideally also the effects of reduced wind speed on surface fuel moisture. Although dead fuel moisture affects the behaviour of surface fires in a range of models, this may have little bearing on fire behaviour as a whole. It is possible that the most significant effect of shading and wind speed reduction is in the way these factors slow the evaporation of water following rainfall, although the effect of this is difficult to quantify without the use of dynamic models such as Matthews (2006) or Matthews et al (2010) which cannot be used for the typical point modelling of fire behaviour used in incident management as they require historical parameters. 196 6.23 Live fuel moisture Chapter 4 demonstrated the role of leaf moisture m in all aspects of the flammability of leaves, and chapter 5 provides models that quantify the way this affects the flammability of a plant. As with dead fuel moisture, m can be measured directly for different species or modelled with empirical or deterministic models. A third option is the use of satellite images to calculate m from spectral reflectance characteristics (e.g. Serrano et al 2000, Dasgupta et al 2007, Yebra et al 2008, Yebra and Chuvieco 2009), which is of particular value in heath and grasslands where the fuel structure is relatively simple and uniform. Such methodology derives a representative value for the species present within a pixel, but may be of limited value in a complex forest structure. In such environments, modelling m becomes an important tool. Modelling live fuel moisture contents Unlike dead leaves, live fuels have the characteristic that they are capable of responding to environmental stimuli via opening or closing of the stomata on the leaves so that drying is not necessarily a simple, predictable pattern. The stomata themselves are each governed by a pair of ‘guard cells’ that swell when they are filled with water, opening the stoma. Open stomata freely exchange water with the atmosphere, whereas closed stomata reduce transpiration and loss of water from the plant (Starr and Taggart 1995). Species-specific traits which determine the surface area of plant exchanging water with the atmosphere include the degree of sclerophylly, the surface area to volume ratio of the leaves and the LAI of the plant. Another factor complicating the modelling of m is the capacity of the plant to actively source water via the root system. Just as different species have leaves of differing thicknesses, density of stomata and density on the plant, so root systems also vary widely and their effectiveness in collecting water may be influenced by mutualistic relationships with mycorrhizal fungi (Augé 2001), which themselves are frequently dependent for distribution upon fire/climatic influences and a range of fauna (e.g. Claridge and May 1994). Despite these complicating biotic factors, it is possible to determine a number of trends in the moisture content of live leaves, and in some cases to build predictive models. Castro et al (2006), Pellizzaro et al (2007) and others have noted that there are a range of different mechanisms in plants which determine their moisture response to certain stimuli. While some 197 species respond strongly to soil moisture and weather conditions, others maintain consistently low moisture with very little variation and others are seasonally driven by their phenology. The moisture content of the first group of species can potentially be modelled using meteorological values, the second group can be characterised by a representative moisture value and the third group by values representative of the season. Because of this enormous potential variability, adequate models representing m across species have not been developed, although some generalisations have been made and a number of species-specific models developed (table 6.2). Table 6.2. Fuel moisture models developed for a range of species Species/community Study Chaparral Olsen 1980 Acacia dealbata Viegas et al 2001 Arbutus unedo Calluna vulgaris Chamaespartium Tridentatum Cistus monspeliensis Cistus albidus Erica arborea Eucalyptus globulus Quercus coccifera Rosmarinus officinalis Pinus halapensis Ulex europeus Cistus monspeliensis Castro et al 2003 Avena sterillis Dimitrakopoulos and Bemmerzouk 2003 Parietaria diffusa Piptatherum miliaceum Chaparral threshold for large fires Dennison et al 2008 In order to test and use the model against a range of local conditions, moisture models were developed for six species which were quantitatively important to the areas studied. A slightly 198 different approach was taken to the approaches listed above by investigating the impact of longer-term weather conditions on leaf moisture. These are captured to some degree with drought indices such as the KBDI and the Canadian Drought Code, however they are then generalised to describe the moisture of the soil or some other parameter. Consequently, such models do not address the impact of meteorological anomalies such as prolonged heatwaves which may occur with or without dry soils. Details of the study are described below. 6.231 Methods Temperature and humidity data loggers (Tinytag TGP-1500) were placed within a 50m radius of stands containing the plants of interest. Data loggers were suspended under tree canopies approximately 1.5m above the ground beneath covers providing both shade and free air movement. Leaves were collected from seven different sites all located in Kosciuszko National Park, as shown in table 6.3. The species E. pauciflora and E. niphophila were combined due to their taxonomic and morphological similarities. Table 6.3. Data collection sites for live fuel moisture Station Aspect Elevation (deg.) (m.a.s.l.) EU1 80 1398 7 1224 E. pauciflora, Daviesia mimosoides EU2 150 1186 21 943 Daviesia mimosoides EU4 320 1149 20 887 Daviesia mimosoides EU5 340 1173 12 922 Daviesia mimosoides Brooks Ridge 200 1490 4 1393 E. pauciflora, Olearia aglossa, Slope (deg.) Mean annual Species studied Rainfall (mm) Tasmannia xerophila, E. stellulata, Bossiaea foliosa Sawpit Ck 320 1470 8 1307 E. pauciflora, Bossiaea foliosa, Daviesia mimosoides Diggers Ck 280 1500 13 1344 E. pauciflora, Olearia aglossa, 199 Tasmannia xerophila, Bossiaea foliosa Cottage 16 280 920 5 642 E. pauciflora, E. stellulata Fuel sampling occurred at irregular intervals. An average of 11.2g of leaves (excluding lightgreen new growth) and stems <1mm diameter were picked and sealed in tins which were weighed within an hour of picking. Soil samples were taken at each site from a depth of approximately 20cm and sealed in tins, also to be weighed within one hour. After initial weighing of leaves and soil samples, the lids were removed and the tins dried in an oven at 105oC for 24 hours, then weighed again with their lids to find the moisture of the samples as percent oven dry weight (%ODW). The moisture of the leaves was collated with soil moisture ms, temperature t and humidity RH as well as with a range of derived values including dew point DP, dew point depression DPD (the difference between temperature and dew point) and averages of all meteorological values for up to one week before picking. Independent variables were divided into two classes – accessible values and difficult to access values. The accessible values were soil moisture, temperature, relative humidity, dew point and dew-point depression; all other variables were deemed difficult to access as they require historical records of temperature and humidity rather than the operationally-accessible ambient values. The raw data for the accessible variables are given in appendix V. The coefficient of correlation was calculated for each time-based average compared to live fuel moisture, and the results graphed to determine if there were any significant trends. All variables were then entered into the statistics package ‘Minitab’, where they were analysed with the aim of building a model with a maximum of four independent variables. The steps were: 1) As the best-subsets regression is limited to 31 predictors and there were 49 available, stepwise regression was used to find the strongest five of these. 200 2) Using the strongest five predictors, best-subsets regression was performed to identify the model with a value of Mallows Cp closest to n + 1, thereby minimising the risk of multicollinearity. 3) Linear regression was used to create a model with the best identified predictors, then the best-subsets regression repeated with the accessible parameters so that a second model could be constructed on the best of these. 4) If any model had a significant (P ≤ 0.05) correlation, the model was considered acceptable. Model value for validation and general use was ranked based on the R2 as a measure of precision using table 6.4. Where the available data could not explain the trends in plant moisture, generalisations were made based around the observations of Castro et al (2006) and Pellizzaro et al (2007). R2 Model strength 0 – 33% Weak 33.1 – 66% Moderate 66.1 – 100% Strong Table 6.4. Criteria for defining the strength of each moisture model. 6.232 Results for six species The size of samples required due to the weight of the tins meant that the sampling procedure was highly time consuming and as a result, only a limited number of samples was collected. It is proposed that a sampling procedure requiring smaller samples such as that employed by Viegas et al (2001) may be more practical for future extensions of this study. Snowgum (Eucalyptus pauciflora / niphophila) 100 samples were collected for these species. The correlation of m with all time-based averages is shown in Figure 6.3. The three strongest predictors temperature, dew point and dew point depression showed a minor trend with time, with a minor drop in correlation with data averaging up to 12 hours, then a gradual, albeit fluctuating climb in strength up to 168 hours. Based on visual analysis, the strength of the trend was not considered strong enough to project the average back further than one week. 201 The best-subsets regression identified the 100% five strongest predictors as soil moisture, 90% 80% dew point depression (DPD), dew point relative humidity averaged over six hours (RH6) and relative humidity averaged over 168 hours regression (RH168). of m The against best-subsets these factors produced the best model with three predictors (Mallows Cp = 3.8). The regression of this is equation 6.16 (R2 = 66.9%, adjusted R2 = 65.9%, P = 0.000). % Correlation depression averaged over two hours (DPD2), 70% 60% 50% 40% 30% 20% 10% 0% 0 24 48 72 96 120 Period of average (hrs) 144 168 Temperature Relative Humidity Dew P o int Dew P o int Depressio n Figure 6.3. Correlation between the leaf moisture of E. pauciflora / E. niphophila and four meteorological indices averaged over periods up to one week prior to the measurement. Best-subsets regression using only accessible data produced equation 6.17 (predictors = 3, Mallows Cp = 3.0, R2 = 66.1%, adjusted R2 = 65.0%, P = 0.000). m = 1.19m s − 1.41DPD − 0.230 RH 6 + 130 Equation 6.16 m = 1.28ms + 0.272 DP − 0.687 DPD + 104 Equation 6.17 A further simplified model was developed with little loss of precision (R2 = 65.9%, adjusted R2 = 65.2%, P = 0.000). m = 1.74ms − 0.765DPD+107 Equation 6.18 202 Olearia aglossa 100% 49 samples of this species were collected. 90% 80% averages is shown in Figure 6.4. The three strongest predictors relative humidity, dew % Correlation 70% The correlation of m with all time-based 60% 50% 40% 30% point and dew point depression showed an 20% increase for 12 to 24 hours followed by a 10% 0% decline to around 48 hours then a gentle increase to around one week. Based on visual analysis, the strength of the trend was not considered strong enough to project the average back further than one week. 0 24 48 72 96 120 Period of average (hrs) 144 168 Temperature Relative Humidity Dew P oint Dew Po int Depressio n Figure 6.4. Correlation between the leaf moisture of Olearia aglossa and four meteorological indices averaged over periods up to one week prior to the measurement. The best-subsets regression identified the five strongest predictors as 12-hour RH, two hour temperature, six-hour RH, 24-hour temperature and 168-hour temperature. The best-subsets regression of m against these factors produced the best model with four predictors (Mallows Cp = 4.0). The regression of this is equation 6.19 (R2 = 69.9%, adjusted R2 = 67.2%, P = 0.000). Best-subsets regression using only accessible data produced equation 6.20 (predictors = 3, Mallows Cp = 3.1, R2 = 51.6%, adjusted R2 = 41.4%, P = 0.000). m = 5.91T2 + 1.64RH 6 − 3.55T24 + 2.33T168 − 59.9 Equation 6.19 m = 1.34RH + 0.905DP + 3.91DPD − 17.5 Equation 6.20 203 Tasmannia xerophila 43 samples of this species were collected. 100% The correlation of m with all time-based 90% averages is shown in Figure 6.5. The three 80% strongest predictors relative humidity, dew % Correlation 70% 60% point depression and temperature showed a 50% decrease for 12 to 24 hours. All three 40% increased to varying degrees after this point 30% then declined after 96 hours. The size of the 20% 10% oscillations in the trends may be a result of 0% 0 24 48 72 96 120 Period of average (hrs) 144 168 Temperature Relative Humidity Dew Point Dew Point Depression Figure 6.5. Correlation between the leaf moisture of Tasmannia xerophila and four meteorological indices averaged over periods up to one week prior to the measurement. the small sample size, but it appears that a one-week average is sufficient information and that longer-term studies are unlikely to reveal stronger correlations. The best-subsets regression identified the five strongest predictors as 12-hour DPD, 48-hour RH, 96-hour RH, 72-hour DPD and 24-hour DPD. The best-subsets regression of m against these factors produced the best model with four predictors (Mallows Cp = 7.9). The regression of this is equation 6.21 (R2 = 88.2%, adjusted R2 = 87.0%, P = 0.000). Best-subsets regression using only accessible data produced equation 6.22 (predictors = 3, Mallows Cp = 3.1, R2 = 24.1%, adjusted R2 = 18.3%, P = 0.189). This model is not statistically valid and will not be used. m = 186 − 22.4 DPD36 − 6.67 RH 48 + 6.67 RH 96 + 21.2 DPD72 Equation 6.21 m = 218 − 0.196RH − 1.06DP − 2.24DPD Equation 6.22 204 Eucalyptus stellulata 44 samples of this species were collected. 100% 90% 80% averages is shown in Figure 6.6. No 70% correlations stood out as stronger than others, and all except dew point displayed the pattern of increasing strength up to six hours % Correlation The correlation of m with all time-based 60% 50% 40% 30% 20% followed by decline. Dew point increased in strength to 48 hours then decreased before 10% 0% 0 24 increasing slightly up to 168 hours. The best-subsets regression identified only four strong predictors; these were six hour dew point depression, 72-hour DPD, 12-hour 48 72 96 120 Period of average (hrs) 144 168 Temperature Relative Humidity Dew P oint Dew P oint Depression Figure 6.6. Correlation between the leaf moisture of Eucalyptus stellulata and four meteorological indices averaged over periods up to one week prior to the measurement. RH, 96-hour RH and 24-hour DPD. The best-subsets regression of m against these factors produced the best model with three predictors (Mallows Cp = 6.5). The regression of this is equation 6.23 (R2 = 83.7%, adjusted R2 = 82.5%, P = 0.000). Best-subsets regression using only accessible data produced equation 6.24 (predictors = 3, Mallows Cp = 3.5, R2 = 31.2%, adjusted R2 = 26.1%, P = 0.002). m = 56.8 − 0.912DPD6 + 3.88DPD72 + 0.551RH12 Equation 6.23 m = 110 − 0.566ms + 0.316RH − 0.190DP Equation 6.24 205 Bossiaea foliosa 85 samples of this species were collected. The correlation of m with all time-based averages is shown in Figure 6.7. The correlation with dew point was slightly weaker than the others, which all decreased in strength up to 12 hours, increased again to 24 hours then fluctuated up to 168 hours. Both temperature and dew point were increasing in strength as the period approached one week; dew point in particular increased significantly from 144 to 168 hours, so that further investigation of the influence of averages lasting longer than a week may be warranted. The best-subsets regression identified the five strongest predictors as two, six and 12-hour dew point depression, 24-hour RH and soil moisture. The best-subsets regression of m against these factors produced two possible models, one with three predictors (Mallows Cp = 4.9, R2 = 32.5%) and another with four predictors (Mallows Cp = 4.0, R2 = 34.8%). The increase in R2 with the second equation is too small to justify the use of another variable, so the first model is preferable. The regression of this is equation 6.25 (adjusted R2 = 30.0%, P = 0.000). 100% Best-subsets regression using only accessible 90% data produced equation 6.26 (predictors = 3, 80% 2 Mallows Cp = 3.1, R = 6.3%, adjusted R = 2.8%, P = 0.154). This model is not statistically valid so it will not be used. 70% % Correlation 2 60% 50% 40% 30% 20% m = 83.9 + 0.895DP2 − 3.67 DP6 + 2.94DP12 Equation 6.25 10% 0% 0 24 48 72 96 120 Period of average (hrs) 144 168 Temperature Relative Humidity Dew P oint Dew P oint Depression Figure 6.7. Correlation between the leaf moisture of Bossiaea foliosa and four meteorological indices averaged over periods up to one week prior to the measurement. m = 85.9 + 0.0819ms + 0.033T − 0.0580RH Equation 6.26 206 Daviesia mimosoides 78 samples of this species were collected. 100% 90% The correlation of m with all time-based 80% averages 70% shown in Figure 6.8. Temperature and dew point were the strongest predictors after six hours, maximising at one to two days before very % Correlation is 60% 50% 40% 30% 20% gradually declining. It appears that the majority of influence occurs in the day or 10% 0% 0 24 two before measurement. The best-subsets regression identified the five strongest predictors as 36-hour DP, soil moisture, 24-hour DPD, 24-hour RH, 48 72 96 120 Period of average (hrs) 144 168 Temperature Relative Humidity Dew P oint Dew Po int Depressio n Figure 6.8. Correlation between the leaf moisture of Daviesia mimosoides and four meteorological indices averaged over periods up to one week prior to the measurement. and 12-hour DPD. The best-subsets regression of m against these factors produced the best model with four predictors (Mallows Cp = 9.4). The regression of this is equation 6.27 (R2 = 40.3%, adjusted R2 = 37.0%, P = 0.000). Best-subsets regression using only accessible data produced equation 6.28 (predictors = 3, Mallows Cp = 4.5, R2 = 24.7%, adjusted R2 = 21.7%, P = 0.000). m = 0.446ms + 1.36 RH 24 + 1.41DP36 + 6.34 DPD24 − 45.1 Equation 6.27 m = 81.2 + 0.487ms + 1.38DP + 0.657DPD Equation 6.28 6.233 Discussion The leaf moisture of all six species was successfully described, although the equations for the two leguminous species (Bossiaea foliosa and Daviesia mimosoides) were not as strong as 207 those for the other species. Operational models were produced for four of the species, with the strongest models developed for Snowgum and Olearia aglossa. Of the six species studied, only Snowgum and Daviesia mimosoides were found to have a significant dependence upon soil moisture. E. pauciflora has been regarded as possibly the most drought sensitive of the eucalypt species (Körner and Cochrane 1985). This model may provide some explanation of that observation as another study has found that the moisture of many tree species has little dependence on topsoil moisture (Dimitrakopoulos and Bemmerzouk 2003). Three species – Olearia aglossa, Tasmannia xerophila and E. stellulata were found to be influenced by temperatures and atmospheric moisture during the three to seven preceding days and yet were not affected by soil moisture. This suggests that events such as heat waves may cause moisture stress in these species in the absence of drought. The two shrubs Olearia aglossa and Tasmannia xerophila were the most dependent upon temperature and dew point for the preceding days, and as both of these were observed during the course of the study to resprout following fire, it is possible that this reflects a more robust root system. This presents another hypothesis worthy of future investigation – whether the leaf moisture of resprouting shrubs is in general more dependent upon long-term atmospheric conditions than on soil moisture. The lesser predictability of the two leguminous species presents a second area for future investigation – whether leguminous shrubs are less responsive to environmental stimuli than other species because their greater access to the Nitrogen required for green leafy growth equips them with alternative adaptive mechanisms. Throughout the study period for instance, plants of both species were observed to shed foliage during dry periods and regrow during wetter seasons. These observations were ad hoc, however they do present a plausible explanation worthy of further investigation. The models are summarised in table 6.5, and the observed vs. expected results for each are shown in Figures 6.9 to 6.18. 208 Table 6.5. Models of LFMC developed for six species Species Model type Equation Strength Snowgum Main 6.16 Strong Operational 6.17 Strong Operational 6.18 Moderate Main 6.19 Strong Operational 6.20 Moderate Tasmannia xerophila Main 6.21 Strong E. stellulata Main 6.23 Strong Operational 6.24 Weak Bossiaea foliosa Main 6.25 Moderate Daviesia mimosoides Main 6.27 Moderate Operational 6.28 Weak Olearia aglossa - 5 * - 5 * 5 / 3 5/ 3 Figure 6.9. Observed vs. expected results for equation Figure 6.10. Observed vs. expected results for 6.16 describing Snowgum LFMC equation 6.17 describing Snowgum LFMC with 1st operational model - 5 * 5 ) * 3 5) 3 Figure 6.11. Observed vs. expected results for equation 6.18 describing Snowgum LFMC with 2 Figure 6.12. Observed vs. expected results for nd equation 6.19 describing Olearia aglossa LFMC operational model 209 * 5 6 5 * 3 5 3 Figure 6.13. Observed vs. expected results for Figure 6.14. Observed vs. expected results for equation 6.20 describing Olearia aglossa LFMC equation 6.21 describing Tasmannia xerophila LFMC - - * * 5 3 5 3 Figure 6.15. Observed vs. expected results for Figure 6.16. Observed vs. expected results for equation 6.23 describing E. stellulata LFMC equation 6.24 describing E. stellulata LFMC % * * 3 3 Figure 6.17. Observed vs. expected results for Figure 6.18. Observed vs. expected results for equation 6.25 describing Bossiaea foliosa LFMC equation 6.27 describing Daviesia mimosoides LFMC 210 * 5 6.24 Soil moisture 5 3 As noted in the previous section, soil moisture affects the live moisture of many plant species, so the ability to estimate it is an important component of a complete system Figure 6.19. Observed vs. expected results for equation 6.28 describing Daviesia mimosoides LFMC model. This section does not study the link between forest structure and soil moisture, but examines an existing method for approximating soil moisture so that this variable can be made available for modelling LFMC. Both water movement within the soil and evaporation have been well studied and can be estimated using a range of models (e.g. Penman 1948, Jury et al 1991), but many such models developed require difficult to obtain measurements such as evapotranspiration. Where such measurements are not available as is commonly the case in remote forest areas or in broad scale management of fire-prone lands, the only models available are drought indices such as the Keetch–Byram Drought Index or KBDI (Keetch and Byram 1968) and the Soil Dryness Index or SDI (Mount 1972). Such indices take a broader empirical approach by encapsulating all aspects of water movement to and from the topsoil profile in one equation. The output of each index is a measure of the water required to saturate a given soil to its field capacity. As the KBDI is used operationally in Australia and internationally (e.g. Burgan 1988, Finkele et al 2006), it presents a measure of soil moisture that is readily accessible or easily calculated. The objective of this section is to explain the relationship between measures of the soil moisture deficit and percent soil moisture. The previous studies of LFMC were consequently related to soil moisture rather than a drought index. Due to its widespread use, the KBDI will be adopted for this purpose and the present study will provide some validation. Converting soil moisture deficit to percent moisture content The drought indices KBDI and SDI refer to the soil moisture deficit or the amount of rain required to raise soil to the point of field capacity fc from the plant wilting point (Finkele et 211 al 2006) to an unspecified depth. The KBDI considers soil to be at field capacity when it holds 8 inches (203.2mm) of rain, calculating evapotranspiration (ET) according to: (203.2 − KBDIt −1 )(0.968e 0.0875T max +1.5552 KBDI t = KBDI t −1 + 1 + 10.88e −0.00173 Pa )10 − 8.3 −3 − Pt Equation 6.29 (Keetch and Byram 1968, Finkele et al 2006); t refers to the time interval of one day, Pt is the amount of precipitation for the day (mm) Tmax is the maximum temperature (oC) for the day and Pa is the mean annual precipitation in mm. Pt is adjusted for runoff and interception by vegetation by subtracting 5mm from the first 24 hours of rainfall in any consecutive period. Although the rate at which water is lost from the soil via evapotranspiration is exponential in the KBDI calculation, water additions are additive. Soil moisture content is assumed to display a linear relationship with KBDI, where an index of 0 is equal to the field capacity of the soil and an index of 203.2 is equal to the lower limit of soil dryness, which is the wilting point moisture (Keetch and Byram 1968). The field capacity of a soil is the percent water content of the soil after saturation and following 2-3 days for gravitational water to drain (Veihmeyer and Hendrickson 1931). This value varies depending on the texture of the soil as it is determined by the number of small pores (less than 0.05mm diameter) which are produced by the particle sizes in the soil (Craze and Hamilton 1991). The depth of soil imposes limits on its water holding capacity, however the field capacity is unaffected as this is a measure per volume or weight of soil. The lower limit of soil moisture expected at the maximum value of the index is the wilting point moisture mwp, defined by Powers (1922) as “the moisture content of the soil at which the plant wilts permanently or at which it cannot maintain its turgidity”. This value is assumed to vary little between species, but significantly between soil textures so that soils with a fine texture and therefore capable of holding more water also have higher wilting points (Powers 1922). 212 If a uniform distribution of water is assumed throughout the volume of soil that can be brought to field capacity with 203.2 mm rain, then the soil moisture for any value of the KBDI is: ms = (m fc − mwp )(203.2 − KBDI ) 203.2 + mwp Equation 6.30 Where mfc and mwp are the field capacity and wilting points of the soil expressed as percentages of the soil weight. As the SDI is based upon 177.8 mm of water (Mount 1972), the equivalent conversion for this index is: ms = (m fc − m wp )(177.8 − KBDI ) 177.8 + m wp Equation 6.31 Determining field capacity and wilting point Field capacity is usually determined in the laboratory by saturating a soil core sample then applying a water tension of 10 kPa, however it can be determined in the field by saturating a soil and allowing it to drain for 48 hours (Craze and Hamilton 1991). Published values of field capacity based upon the soil texture (e.g. Salter and Williams 1967, 1969) are useful for modelling soil hydrology with a minimum of field data collection. Table 6.6 shows standard values of soil field capacity and wilting point as fraction weight of the soil. Permanent wilting point is determined in a laboratory by applying a soil water tension of 1500 kPa to small soil cores, or by using plants as part of a bioassay. Field capacity and permanent wilting points are usually expressed in volume of water / volume of soil. These values can be converted to percent Oven Dry Weight (% ODW) as follows, substituting mwp for mfc to calculate this value where necessary: 213 %m fc = m fc ρ w ρs Equation 6.32 Where w is the density of water (g/cm-3) equal to 1.0 and s is the density of the soil. TABLE 6.31 Soil field capacity and wilting point (Craze and Hamilton 1991). g/cm3 Soil texture Field capacity W/W Wilting point Sand 0.14 0.03 – 0.1 1.55 – 1.80 Loamy sand 0.18 0.03 – 0.1 1.60 Sandy Loam 0.26 0.06 – 0.12 1.40 – 1.60 Loam 0.30 0.11 – 0.17 1.35 – 1.50 Silt Loam 0.39 0.09 – 0.21 1.30 Clay Loam 0.34 0.15 – 0.2 1.30 – 1.40 Silt Clay Loam 0.43 0.17 – 0.24 1.35 Silty Clay 0.47 0.17 – 0.22 1.25 – 1.35 Clay 0.42 0.19 – 0.24 1.20 – 1.30 6.241 Validating the model Methods To validate this model, weather stations were constructed at five sites along a soil catena between a high rainfall ridge and a low rainfall ridge in the Eucumbene Cove area of Kosciuszko National Park. The two ridge top stations, two mid slope, and one gully station represented a range of rainfall, slope, aspect and soil conditions (table 6.32) to provide broader validation. As detailed in table 6.33, each station recorded soil moisture, with three stations recording temperature and relative humidity and two recording wind speed and rainfall. All measurements were 30 minute mean values, with the exception of the rain gauge which recorded total rainfall for each 30 minute interval. Soil moisture was measured using Tinytag s/n 2409/04 resistance soil probes inserted at 45 degrees into the slope so that the electrodes were situated approximately 20 cm beneath the soil surface. Temperature and Relative Humidity were measured using Tinytag TGX-3580 sensor / data-loggers affixed to a three sided wooden shelter at 1.4m above ground to provide 214 protection from direct sunlight. Wind speed was measured using three-cup anemometers at 1.8m, and rainfall was measured using a tipping-bucket rain gauge recording 0.2mm increments and placed at 1.6m above the ground surface. Each station was visited approximately every 2 months for maintenance. Table 6.32. Topographic details of the weather stations calculated from NPWS (2005) using ArcView GIS. Rainfall data taken from NPWS (1998). Station Aspect (deg.) Elevation (m.a.s.l.) Slope (deg.) Mean annual Rainfall (mm) EU1 80 1398 7 1224 EU2 150 1186 21 943 EU3 130 1131 3 877 EU4 320 1149 20 887 EU5 340 1173 12 922 Table 6.33. Locations and sensors at the weather stations Station Grid Reference* Landform Sensors Soil Temp/RH EU1 643813E, 5999025N Ridge * EU2 644568E, 5998320N Lower slope * EU3 644647E, 5998188N Valley * EU4 644679E, 5998124N Mid slope * EU5 644746E, 5998081N Ridge * Wind Rain * * * * * * * *All grid references are in zone 55 using AGD66 datum Calibrating Soil Moisture Probes As the soil moisture probes calculated water content indirectly by measuring electrical resistance, these were calibrated to soil conditions for well-drained soil and for poorly drained soil. All sites were classified as well-drained except for EU3, which was situated on a creek floodplain and was poorly drained. The probes were calibrated by comparing their measurements with measurements of the same core using oven-drying techniques, and a corrective equation was developed. Soil moisture was measured with the probe after removing loose organic material from the soil surface. After inserting and removing the probe, a hand auger was used to remove a core of the same 215 soil that had been measured by the probe. The probes sample a section of soil with a diameter of 25mm and a depth of 60mm, while the soil auger samples a core of diameter 50mm and a depth of 60mm. After each sample, another was taken by inserting the probe into the hole left by the auger, then taking a core of that sample. In all, 30 samples were taken in the 2 drainage categories – 19 samples of the well-drained soil and 11 samples of the poorly drained soil. Core samples were weighed, dried in an oven at 105o for 24 hours then weighed again to calculate the % oven dry weight. In order to find a calibrating function for the soil probes, the measured soil moisture msm was plotted as a function of the moisture value given by the probe msp, and an equation fitted to the data. Determining KBDI KBDI values were calculated for the area of the stations using equation 6.29. Temperatures taken from EU1, EU3 and EU5 were used for those stations and interpolated for EU2 and EU4 using the equation: Ti = (Eu − El )(Ei − El ) (Tu − Tl ) Equation 6.33 Where E is the elevation of each station and T is the temperature. The station is indicated by the subscript u for the station above (either EU1 or EU5), i for the station being studied, and l for the station below (EU3). The average annual rainfall for each site was taken from an ESOCLIM model dataset for the area (NPWS 1998). 216 Rainfall values were calculated from EU3 for the first 111 days, but due to storm damage to the gauge, the remaining 163 days were measured at the EU1 site. Values were adjusted according to the average annual rainfall for each site using the equation: Rad = Rr R As R Ar Equation 6.34 Where R is the rainfall and the subscripts ad, As, Ar and r respectively refer to the adjusted value for the site, the annual value for the site, the annual rainfall for the recorded site (either EU3 or EU1) and the recorded values. The starting value for the KBDI record was estimated from Bureau of Meteorology data (www.bom.gov.au), using an average between the values for the two closest stations Cabramurra in the higher rainfall mountains to the west and Cooma in the drier tablelands to the east. Method of validation As the KBDI relies on both the field capacity of the soil and the permanent wilting point, two options exist for validating the model. Soil moisture can be calculated using measured values or by using approximated values based on field texture measurements. As the second option relies on further estimations, the error estimate from the results would contain the error of the field soil texture estimates and tabulated values as well as that of the KBDI model. Consequently, the first option was chosen to best examine the accuracy of the model in question. The soil field capacity at each site was found by plotting measured soil moisture as a function of KBDI at each site. Data obtained within 48 hours of soil saturation were removed by disregarding those dates where the KBDI was equal to 0 and the day following. A linear function was fitted to each station’s data and the field capacity was calculated from the function as the y-intercept. In the absence of laboratory studies of soil samples, it was not possible to measure the wilting point and a standard value of zero was applied to all soil textures. Such a simplification will introduce a greater level of error as the KBDI increases, but within the range examined it may be a reasonable simplification. 217 Using these values, soil moisture was modelled with equation 6.30. Data was again filtered to remove any values greater than 99% of the field capacity as gravitational water is irrelevant to the plant water balance (Craze and Hamilton 1991). Modelled values were compared with measured values using ANOVA to find the significance of the correlation. Statistical analysis was carried out in Minitab and the percent error was calculated in MS Excel. Results Soil Probe Calibration Soil moisture probes were found to vary between the two drainage groups. In dry conditions the probes were accurate for well-drained soils but overestimated the moisture content of poorly-drained soils. In wet conditions, soil moisture was underestimated in welldrained soils but still over-estimated in poorly-drained soils. The results for the measurements are given in Figures 6.20 and 6.21. Figure 6.20. Calibration curve for well-drained soils The equation produced using data for well-drained soils was: ms = 4.03e0.09sps Equation 6.35 Where ms is soil moisture (% ODW), and sps is the soil moisture indicated by the probe in well-drained soil. The equation fit the data with an R2 of 0.96. Figure 6.21. Calibration curve for poorly-drained soils The equation found for poorly-drained soils was: 218 ms = 3.65e0.05spc Equation 6.36 Where spc is the soil probe moisture in poorly drained soil. The equation fit the data with an R2 of 0.97. KBDI calculations The KBDI was calculated for each individual weather station from the 16th February 2005 to the 9th November; a total of 267 days or 1335 station-days. During this period, the KBDI showed a marked increase during autumn which is typical of the region, commencing with a value of 19 0n February 16 and reaching a maximum of 108 on June 11 at EU5 before falling quickly to 0 between the 10 July and the 4th August depending on the station. EU1 had the lowest indices, peaking at 84 on the same day. EU2, EU3 and EU4 all displayed similar values to each other midway between EU1 and EU5. Following the onset of the winter rains, the KBDI oscillated between 0 and the high teens for the remaining 3 months. KBDI for each of the stations is shown at Figure 6.22. Model validation Data for 286 station-days represented saturated soils with a significant gravitational water content and was therefore removed. Irreparable damage by wildlife to EU2 soil moisture station also occurred, causing the loss of 160 day’s data so that the database was reduced to 889 values. These were plotted for each station with soil moisture as a function of KBDI (Figure 6.23), and the y-intercept was found to provide an estimate of field capacity for individual sites. This measured field capacity is shown in table 6.34 along with the value calculated from a field texture measurement based on Northcote (1979). 219 KBDI Values 120 110 100 90 80 KBDI 70 60 50 40 30 20 10 0 16-Feb 13-M ar 7-Apr 2-M ay 27-M ay 21-Jun 16-Jul 10-Aug 4-Sep 29-Sep 24-Oct Date EU1 EU2 EU3 EU4 EU5 Figure 6.22. KBDI values at the five Eucumbene weather stations Table 6.34. Measured and tabled values for field capacity, tabled wilting points (Craze and Hamilton 1991) Station Soil texture Measured Tabled mfc (% ODW) Tabled mwp (% ODW) mfc (% ODW) EU1 Silty clay 38% 47% 25% EU2 Sandy clay loam 26% 26% 15% EU3 Light clay loam 19% Not tabled Not tabled EU4 Clayey sand 6% Not tabled Not tabled EU5 Loam 14% 30% 13% KBDI values were converted to soil moisture using equation 6.30, and these were compared to the measured soil moisture values after removing 220 values greater than 99% of the measured field capacity. Measured and modelled soil moisture stats for the 669 station-days are shown in Figure 6.24, and the values are given in Appendix VI. 220 Figure 6.23. KBDI values used to calculate the soil field capacity at each site The overall performance of KBDI as a predictor of soil moisture across the five sites was highly significant (P = 0.000) and was ranked as ‘strong’ using table 6.4 (R2 = 73.7%, adjusted R2 = 73.6%). The results of the ANOVA analysis are shown in Table 6.35. When the data were ranked from lowest moisture to highest and the percent error calculated (Figure 6.25), the error was greatest in dry soils and decreased steadily as soil moisture increased. The mean error was 3.1% moisture. Regression of the mean error for each site against the measured field moisture content (Figure 6.26) demonstrated that error mse increased exponentially with field capacity according to equation 6.37 (R2 = 0.81). m se = 0.68e 0.07 m fc Equation 6.37 221 Table 6.35. ANOVA statistics for soil moisture calculated from the KBDI against measured soil moisture values. Source DF SS MS F P Regression 1 45041.7 45041.7 1868.10 0.000 Error 667 16082.0 24.1 Total 668 61123.7 -65 -65 3 Figure 6.24. Soil moisture values calculated from the KBDI shown as a function of measured soil moisture values. Model % error Mean error and field capacity 160% 12 140% 10 120% 8 Percent error % Error 100% 80% 60% 6 4 40% 2 20% 0% 0 0 5 10 15 20 25 30 35 40 Soil moisture (%ODW) Figure 6.25. Percent error of soil moisture values calculated from the KBDI shown as a function of measured soil moisture values. 0 10 20 30 40 M e as ur e d fie ld capacity Figure 6.26. Percent error of soil moisture values as a function of measured field moisture. 222 Conclusion The results demonstrate that across a broad range of rainfall, slope, aspect and soil field capacities, KBDI can be used to estimate soil moisture to a high level of accuracy. Accuracy was highest in soils with a low field capacity, which is to be expected as this restricts the range of moisture values possible. As expected due to the absence of information on the permanent wilting point of each soil, the accuracy of the model decreased with drier soils. However as Figure 6.24 shows, the error was predominantly one of over-prediction rather than the under-prediction that might be expected from using a value of zero as the minimum. This suggests that the KBDI is less effective for modelling soil moisture at high index values. The over-prediction in this study suggests that a precautionary approach may be to assume a value of zero for all wilting points, and to calculate the error using equation 6.37. Conversion of KBDI or other indices of soil moisture deficit to actual values of soil moisture requires information on the field capacity by weight of the soils; where this is available only on a volume basis it can be converted to a weight basis with additional information on the soil density. 6.3 Fire spread in surface fuels As discussed in section 3.23, the level of complexity is low when a fire is burning in the surface fuel stratum alone. For this reason, no new work will be carried out in this area; rather a simple existing empirical model will be used. Although the models of McArthur (1962, 1967) and Gould et al (2007a) use surface fuels as inputs, neither the experimental design with which they were constructed or the details in the publications themselves adequately demonstrate that the fire behaviour attributed to surface fuels is indeed produced solely by these fuels. Consequently, the surface spread model of Burrows (1999a) is used as it provides a defensible model limited to surface fuels free of the influence of other higher fuels. The equations for rate of spread and flame length then are: 223 (0.0245u 2.22 s ) (0.003 + 0.1000922m + 0.071 Rs = d ) 1000 Equation 6.38 Where Rs is the rate of spread (km/h), us is the wind speed at mid-flame height (km/h), md is the moisture content of dead fuels and g is the slope of the ground in degrees. The equation for flame length in metres is: λ = 2.4 Rs + 0.036ws Equation 6.39 Where ws is the weight of surface fuels in t/Ha. 6.4 Forest fire behaviour 6.41 Fire spread within a stratum The same process is used to model fire spread in a stratum as in a plant, with adjustments for the differences in geometry as described earlier. A stratum is ignited when one plant is ignited as described in chapter five. Fire spread occurs through the stratum if the flame from one plant is able to ignite its neighbour. To achieve this, the critical state test must be satisfied and the flame angle must allow the plume to pass through adjacent plants according to equation 3.3. The test is the same process as described in section 5.2, although there are differences in the way P is calculated. When fire is burning from one plant to the next within a stratum (flame connection), P is initially: Pα = S f − w − Ox cos(θ f − θ g ) Equation 6.40 224 After spread has commenced in the stratum, P is simplified to the maximum value of flame depth and (Sf – w). If the flame angle is greater than 90 degrees (backing fire), connection between plants is not possible according to equation 3.3 so there is no equation for backing flames. The plume pathway is also simplified as a result of the simpler geometry, so that equation 3.4 applies to flat terrain. This negates the need for equations 5.39 – 5.60, and for flames where f ≤ 90, the plume pathway can be found using equation 6.41: Pp = hu − hb sin(θ f − θ g ) Equation 6.41 For backing fires, this becomes: Pp = hu − hb sin ((180 − θ f ) − θ g ) Equation 6.42 To assess the critical state, it is necessary to know the flame length and duration from the contributing plant. Contributing flame length within the stratum is the merged flame length of the burning plants, calculated at each time step as in section 5.3 and merged as described by equations 5.75 and 5.76. The simplest way to provide an input flame length is to use the merged flame length updated at each time step. This approach is valid if only one species is considered for the stratum, however when two or more are considered it presents some over-simplifications due to the different rate of burning between species. An alternative is to standardise the flame lengths of the different species by ranking them from longest to shortest, so that merging of flames will be synchronised. This approach may be described with the following assumption: 225 As plants burning in close proximity to one another affect the rate at which their neighbours burn through radiative heat transfer and their influence on air movement, plant combustion will become synchronised so that flames from individual plants reach their maximums at approximately the same time. Assumption 6.1 When calculating equation 6.40, the origin of the contributing flame (Ox) can be taken from the origin of the plant flame for the relevant time step; however this will not be valid if flame lengths have been standardised between species as the flame lengths themselves are linked to specific origins. A simplification that will overcome this is to use the coordinates for the maximum flame length in the contributing plant. The issue of flame origin also applies when determining flame connection with equation 3.3, as Oy is an input to the equation. 6.42 Fire spread between strata Ignition of a new stratum is described in section 5.2, however to find the contributing flame length from a burning plant stratum requires a level of simplification which increases as the number of fuel strata increase. Figure 3.9 addressed this, giving the combined flame length as the combination of all component flames minus the overlap and spaces between them. This conclusion requires the following assumption: If the plume from one burning plant stratum aligns with the flame from the stratum above it, the two flames may be treated as one. Assumption 6.2 The length of overlap and/or spacing ( neg) between each contributing flame can be found from the lower flame height and the base height hb of the higher flame according to: λ neg = abs (O xu − f hl ) sin θ fl Equation 6.43 226 Where fhl is the lower flame height and fl is the angle of the lower flame. By treating neg as an absolute value, both gaps or overlaps between flames are subtracted from the total flame length in the same way. Once the contributing flame length has been determined, ignition of the new stratum can be calculated as given in section 5.2. The flame angle required for calculating P requires an estimate of mid-flame wind speed for the existing flame; this can be calculated from the wind speed at mid stratum for each burning stratum beneath the canopy, weighted by its contribution to the total flame length. This is shown in equation 6.44. u = u1 λ1 (λ1 + λ2 + ... + λn ) + u2 λ2 (λ1 + λ2 + ... + λn ) + ... + u n λn (λ1 + λ 2 + ... + λn ) Equation 6.44 Where u with no subscript denotes the wind speed acting on the contributing flame, and the subscripts 1 to n refer to all strata beneath the stratum being ignited. Preheating can be simplified by generalising the fire behaviour in the strata heating the new stratum, so that the temperature of preheating is based on the mean flame length prior to ignition of a new stratum, and the time of preheating from that flame length is the time taken till ignition of the new stratum. For example, ignition of the canopy is based on the combined flame length of all strata burning below. In an instance where surface, near-surface, elevated and midstorey fuels ignite before the canopy does, pre-heating is received from the surface fuels for ns seconds before near surface fuels ignite; from combined surface and near surface flame length for nns seconds before the elevated fuels ignite, and from combined surface, near-surface and elevated fuels for ne seconds before the midstorey ignites. Due to stability issues with MS Excel which was used as the model platform for the study, preheating from lower strata was only estimated for the first five time-steps. This was considered sufficient to determine the effective ignition of the new stratum as model trials demonstrated that the first few time-steps allowed the development of a self-sufficient flame where this was likely. With a more stable platform it is desirable to continue calculating preheating from lower strata for all time-steps to ensure that there are no other unforseen effects. 227 6.421 Overlapping strata effects The discussion so far has focused on the situation where the plants from one stratum overlap or grow directly over the top of the plants in the stratum below. In this case, the convective plume from a vertical flame in one stratum will pass through the plants in the stratum above it. In some cases however, factors may prevent plants from growing beneath the canopy of the next stratum, in which case a vertical plume from the lower plants will not heat the higher stratum. In such cases, the contributing flame to the higher stratum is calculated using Figure 3.9 and equation 6.43 using only those strata that occur directly beneath the upper stratum. 6.43 Flame dimensions As in the case of burning plants, flame dimensions from a burning fuel stratum are determined by the factors given in table 5.1. The contributing flame is the merged length of burning plants ms, found by averaging the flame length from all species and weighting this by their relative abundance to provide a value of for equation 5.75; i.e.: λms = {η sp1λsp1 + η sp 2 λ sp 2 + ... + η spn λ spn }N 0.4 Equation 6.45 Where spn, sp1, sp2, spn are the percent compositions of species one to n species, and sp1, sp2, are the plant flame lengths from species 1 to n. The longitudinal merging is then found by using this value in equation 5.76. To find the flame length at a given time step, find separately for each species: 1) Number of leaves per branch (equation 5.70) 2) Flame length from a single burning leaf (equations 4.17, 4.18) 3) Depth of ignition into the stratum for each time step (section 5.26), substituting equations 6.41 and 6.42 for equations 5.38-5.59, and using sp as the contributing flame. 4) Total depth of penetration (equation 5.70) 228 5) The number of branches ignited (equations 5.63 and 5.65 with equation 6.1 substituted for Sb) and therefore the number of leaves 6) The flame length is found using equation 5.75, with N = the number of burning branches and = the merged flame length from an individual branch, and equation 5.76 The maximum flame length of the entire forest array is equal to the sum of flame lengths for all strata burning minus neg (equation 6.43) for each stratum. The mean flame length can be estimated from the average across the array. As flame moves through the array, there are areas where plants from all strata are present and the fuel ladder facilitates flame growth but there may also be other areas where a space in one of the strata breaks the fuel ladder and flame length is potentially altered. To find the average flame length, the potential combination of presence/absence of the different strata needs to be examined so that an average can be generated, weighted by the percent occurrence of each scenario. Due to the influence of the forest canopy on the wind profile, scenarios without the canopy should not be examined as the sub-canopy wind speed is an average across the area rather than a value applying only to sites directly beneath an individual canopy. For n strata including the canopy then, the number of presence/absence scenarios ns is: n s = 2 n −1 Equation 6.46 The weighting of flame length determined from each scenario is found from the percent cover of each stratum, where percent cover can be found by describing each crown as a square with an average width wm, according to: C= wm (w + (S m 2 − wm )) 2 f Equation 6.47 229 Where flame merging or connection occurs in a stratum, the gaps between plants become obsolete and the percent cover is estimated as 100% cover. Percent occurrence for each stratum is dependent upon whether the stratum is present or absent in that scenario, for example if a stratum has 70% cover then the percent occurrence in a scenario with that stratum absent is 1- 70%, or 30%. The weighting of a scenario is the product of the percent occurrence for each stratum. The maximum flame height fh for one stratum is derived from the flame length and relevant angles plus the height of the flame origin Oy. f h = λ sin (θ f − θ g ) + O y Equation 6.48 Flame origin varies throughout the time for which fire spread is calculated so there is no set Figure that can be used. A conservative estimate is to substitute Hb as the minimum value; a more realistic approach is to use the value of Oy at the time of maximum flame length. The maximum flame height for the entire fuel array is the maximum of all flame heights for each stratum, and the average flame height is calculated across the array using scenarios as for flame length. 6.44 Rate of spread The rate of spread for a stratum is calculated using equation 3.16, with Ppi equal to Pi from equation 5.70. As fire moves forward however, the wind speed relative to the flame motion decreases up to the point that where a flame is moving at the same speed as the wind, the effective wind speed is zero. Because the wind speed influences the angle of the flame and therefore the potential depth of penetration and overall rate of spread, the rate of spread and effective wind speed cannot be calculated in the one step. This issue is addressed in the use of time steps, so that the effective wind speed at time step n (uen) is equal to the overall wind speed at that height in the array minus the rate of spread in the previous time step Rn-1. 230 u en = u − Rn−1 Equation 6.49 Because of this, the rate of spread should not be taken from a single time step but from a minimum of two steps, to account for the wind correction. As discussed in section 3.23, when fire in one stratum out-paces the fire burning in the strata beneath it, the flame residence time is reduced and Equation 4.21 no longer applies. This is particularly relevant to the rate of spread in the forest canopy as it is often well-elevated above the lower strata. A further assumption will be introduced here, that: Where rates of spread in near-surface, elevated or midstorey fuels are greater than those in the strata below them, radiation and falling burning debris from the higher strata will maintain the surface fire beneath them so that the heat source remains relatively constant. Assumption 6.3 The effect of this assumption is that the rate of spread for a fuel array can be equal to the maximum of all strata burning beneath the canopy, except where specific circumstances allow a crown fire to be maintained at an even faster rate. The rate of spread for crown fires will be examined below. 6.441 Crown fires Three types of crown fires have been described (Van Wagner 1976, Rothermel 1991), passive, active and independent. Passive crown fires In a passive fire, individual tree crowns are ignited and although some flame merging may occur across the fire front, the flame angle relative to the slope does not allow flame connection within the stratum and fire cannot spread from crown to crown. Consequently, a passive crown fire increases the flame dimensions but does not increase the rate of spread. In some cases it may even reduce the rate of spread as the greater thermal updraft caused by the 231 larger flames may cause the lower flames to be more upright and reduce the incidence of flame connection (flame angle damping). Active crown fires In an active crown fire, equation 3.3 is satisfied and fire is able to spread from crown to crown. Due to the higher wind conditions at the height of the crown, it is often the case that an active crown fire will spread faster than fire burning in the strata beneath, however as already discussed, the residence time of flames in the leaves is sharply reduced when the crown fire burns ahead of the plume from the ground fire. Section 5.4 gives a default flame residence of one second in the absence of a heat source; a heat source here is defined as a temperature greater than or equal to 100oC. It is unknown whether these Figures are valid and this is an area in need of further study. This assumption is stated below. Where leaves are burning in the absence of an external heat source producing a temperature of 100oC or greater, the residence time of flame in the leaves will be reduced to one second. Assumption 6.4 The temperature at the crown is calculated using equation 3.5 to find P for each stratum, and equations 5.14 and 5.16 for the temperature. The horizontal distance to which the canopy is heated Dh can be found using equation 6.50 below for each stratum burning below the canopy; where the subscript l refers to the lower stratum and the values He and Hc apply to the canopy base. He + Hc − O yl 2 Dh = O xl + tan (θ fl − θ g ) Equation 6.50 A second factor affecting limiting the spread of active crown fires is the fact that the angle of the plume directs the ignition path out through the top of the canopy rather than parallel to it – another form of flame angle damping. Once the flame origin has reached the top of the 232 canopy, the crown fire is extinguished. Radiative heat transfer is not considered in this analysis; it is possible that this also plays a role in crown fires although experimental analysis of fire spread in vertical arrays suggests that convective transfer may be the main factor (Weber 1990a, Weber and deMestre 1991). Future work may involve analysis of the role of radiation in this area. Although the rate of spread in active crown fires is not set by the rate of spread through the tree crowns, two mechanisms are proposed whereby the crown fire influences the overall rate of spread, these are crown defoliation and crown pulsing. In the first instance, the faster moving crown fire removes the foliage of the forest canopy ahead of the ground fire, allowing higher wind speeds at lower levels in the fuel array as described by equations 6.7 and 6.8. Gill et al (1996) noted that wind profiles in their study exhibited significantly higher wind speeds at lower levels when the forest had been burnt at high intensity, removing the canopy. Calculating the effect of canopy defoliation can be achieved by the following steps: 1) Calculate the rate of spread for all strata 2) Where the rate of spread in the canopy exceeds that of the lower strata, remove the canopy from the wind profile calculations 3) Repeat step one with the altered wind profile Crown pulsing occurs when the crown fire burns some distance ahead of the ground fire before extinguishing itself. This is Figure 6.26. Abrupt variability in fire impact indicative of crown pulsing evidenced in burn patterns observed after high intensity fire spread where fire impact on the vegetation changes abruptly (Figure 6.26). Direct observation by fireground personnel also testifies to its occurrence (Pers. Comms. A. Grant, February 2010). 233 It is proposed that the effect of radiative heat transfer downward from the crown along with the shower of burning debris and drawing effect of the fire itself may cause the continual ignition of new surface fires beneath the temporarily independent crown front. The development of these fires is too slow to maintain the crown fire, but when it selfextinguishes at cp metres from where the fire first started growth after burning in the crown for cr seconds, the surface fire once again develops into the crown, taking cd seconds before the crown fire is again burning. The overall rate of spread for the fire is equal to the distance covered divided by the time taken. In km/h, this is: R= 3.6c p cd + cr Equation 6.51 The distance cp shown below and cr is the time in seconds from when the crown is first ignited to when it self-extinguishes. The value cp is a product of the distance fire travels through the crowns and the horizontal distance from ground ignition to crowning ignition. c p = O x max + I NSx + I Ex + I Mx Equation 6.52 Where Oxmax is the horizontal distance that fire has progressed through the crown stratum, INSx, IEx and IMx are the horizontal depths ignited in the near surface, elevated and midstorey strata respectively. The time cd is the sum of the ignition delay times of each stratum below and including the crown, plus the time taken from ignition to maximum flame length for each stratum ( ); i.e.: cd = {ψ ns + β ns + ψ e + β e + ψ m + β m + ψ c } Equation 6.53 Equation 6.53 relies on a further assumption, that: 234 New strata do not ignite before the stratum beneath produces its maximum flame length. Assumption 6.5 This is an abstraction of reality that may be removed with further work. Although assumption 6.4 treats the flame duration in canopy fuels differently from the lower strata, fire spread in lower strata is still subject to the second limitation of flame angle damping. Where this occurs, spread may still be accelerated by pulsing within that stratum, using the same process as that of crown pulsing. Independent crown fires The third type of crown fire is the independent crown fire. An independent crown fire can theoretically occur if the limitations described under active crown fires are overcome; that is, that the rate at which the fire is spreading is faster than the rate of its extinction, and the angle of the plume is approximately parallel to the slope. The first limitation is overcome more easily in some species and forest structures than in others. Factors that act to decrease the ignition delay time (larger surface area to volume ratio, lower leaf moisture) facilitate faster ignition. Factors that cause longer flames extend the ignition isotherm further into the unburnt fuels; these include longer leaves, closer plant spacing and greater bulk density (a product of leaf separation, branch orders and branch dimensions) as such factors increase the individual flame length of particles as well as increasing the number of particles burning together. Just as these factors can increase the rate of ignition through the stratum, doing so produces a greater flame length (and consequently greater flame angle) so that the likelihood of overcoming the second limitation actually decreases. A more upright flame also means that the length of crown pulses cp is decreased so that the rate of spread in active crown fires may be slower in more flammable vegetation under certain conditions. The two factors required to overcome the second limitation are slope and wind; sufficient wind is required to tilt the flame parallel to the slope and therefore steeper slopes reduce the quantity of wind required. A standard value can be used to approximate turbulent diffusion in the heated plume, so that 235 crown fires can be classified as independent when the plume angle is within a certain range of the slope angle. A value of 10 degrees is used in the model developed here. 6.5 Discussion 6.51 Worked example The following example provides a hypothetical illustration of fire spread through a simple forest environment consisting of surface litter, a shrub layer of one species and a single species tree canopy. The example demonstrates the steps required to calculate each increment of fire spread, and the resulting fire behaviour. Weather and terrain conditions are shown in table 6.34, and fuels are described in tables 6.35 to 6.36. The calculations are based on a line fire burning upslope with the wind behind it (assumed parallel to the slope), using t = one second. Table 6.34 Weather and terrain details Temperature o 28 C Relative humidity Wind speed (km.h-1, 10m) Slope 20% 15 10o Table 6.35 Structural details Parameter Value Mean surface fuel particle 5mm diameter Surface fuel load 9t/ha Elevated plant spacing 0.8m Canopy plant spacing 5m 236 Table 6.36 Plant details Parameter Shrubs Trees m (%ODW) 100% 120% P ( C) 260 220 % Dead 10 0 Leaf form Flat Flat d (mm) 0.40 0.6 Wl (mm) 11.0 18 Ll (mm) 58.0 68 Sl (cm) 1.3 1.1 Os 2.2 5 Sb (cm) 12 85 Wb (cm) 26 120 He (cm) 5 380 Ht (cm) 50 555 Hc (cm) 1 400 Hp (cm) 60 700 w (cm) 60 370 o The plant community is shown in Figure 6.27 along with the resulting modelled flame. Figure 6.27. FFM output of community structure and flame dimensions for the worked example. 237 6.511 Steps in calculating fire behaviour The list below shows the calculations and steps involved in modelling the spread of flame into the fuel array, using an assumed surface flame with given length and rate of spread. Additional equations are provided at times to clarify the process. Wind profile 1. Using Equ. 6.11 find the Leaf Area LAI for trees is 0.9, for shrubs is 0.7 Index for the shrubs and trees 2. Multiply this by the percent cover to Percent cover of the canopy is 55% and of find the LAI for each stratum. Percent the understorey is 56%, therefore the stratum cover Cp can be calculated from plant LAI values are 0.47 for the canopy and 0.39 separation Sf and crown width w using a for the shrub layer square to represent the plant crown area to give equation 6.53: Cp = w2 [w + (S f ] − w) 2 Equation 6.53 3. Use Equ. 6.8 to convert this to The values of are 0.91 for canopy and 0.71 for the shrub layer. 4. Use Equ. 6.7 to find the wind profile for Table 6.37 shows wind speeds calculated for the top, middle and base of the two heights in the fuel array (z) starting from the strata. top of the canopy (7.0m) where the 10m wind speed is assumed. The wind speed of 6.37m is reduced for the height of 5.4m Figure 6.28. Wind profile for the example forest 238 of 5.4m using equation 6.7 to 16.4km/h. Table 6.37. Wind speeds at different heights in the forest profile z (m) 0.0 0.3 0.6 3.8 5.4 7.0 Wind (km/h) 3.1 4.4 6.3 9.9 12.2 15.0 Surface fire behaviour 5. Surface rate of spread (Equation 6.38): 0.07km/h 6. Surface flame length (Equation 6.39): 0.49m 7. Flame angle from surface wind - Equ. 16.4o 5.29 8. Maximum angle from slope - Equ. 5.30 80o. Flame angle is minimum of wind and slope effects Plant ignition 9. Flame residence time in surface fuels – 17.8s Equ. 5.13 10. Identify potential ignition points in shrub base. Model on CD uses 5 points Ox = -30cm Oy = 5cm – left edge, halfway to centre, centre, halfway to right, right edge. All pathways are examined and the one producing the longest flames chosen. In this example, this is the downhill edge. 11. Find - Equ. 5.26 5cm 239 12. Find P – Equ. 3.11 89cm 13. Identify in which flame zone Ox, Oy lies Zone III by comparing P to the flame length 14. Find f (distance from flame base to top of burning plant) Flame is only burning in surface fuels so f is 0cm 15. Find Spp. Equ. 5.34 29cm 16. Find the temperature at Ox, Oy + one 325oC tenth of the penetration depth – Equ. 5.16 17. Ignition delay time of shrub leaves at 15.1s that temperature – Equ. 4.10 18. Check for ignition Eq.3.15 Ignition occurs Flame propagation through the plant for one time step 19. Find 1 to 4, Equations. 5.53 – 5.59 1 = 1.07 radians, radians, 4 20. Identify the plume pathway – Fig. 5.13 Pathway 3 21. Find Pp – Equ. 5.39 195cm 22. Find PD – Equ. 5.44 133cm 23. Find Ppa for P3 – Equ. 5.49 63cm 2 = 0.64 radians, 3 = 0.00 = -0.13 radians 240 24. Find P (minimum of Ppa and Spp) 29cm 25. Increment length (P / 10) 2.9cm 26. Find T for P + one increment, 325oC Equations. 5.14 or 5.16 27. – Equ. 4.10 28. Critical state test – Equ. 3.15 for t + initial 29. Repeat steps 26 to 28 for all ten 15.1s 325oC >260oC, 15.1+0 ≥ 15.1. Ignition occurs. Ignition occurs for one of ten steps increments 30. Depth of ignition = Increment length * 2.9cm ignited in first time step. number of increments ignited 31. Number of branches ignited – Equ. 5.65 0.1 branches 32. Number of leaves in a branch – Equ. 75 leaves 5.69 33. Flame length from one leaf – Equations. 9.4cm 4.18, 4.20 34. Merged flame length from merged 0.21m flame – Equ. 5.75 35. Flame length produced in 1st time step 0.25m (merged flame with longitudinal effect – Eq. 5.76) 241 36. Angle of plant flame from mid-shrub 8.1o wind speed – Equ. 5.29 37. Find horizontal distances (xi, yi) ignited; xi1 = 2.7cm x = cos( f)*penetration yi1 = 0.8cm y = sin( f)*penetration Maximum distance is edge of plant Flame propagation through plant for following time-steps 38. Determine whether external flame still burns Time elapsed (15.1 + t * 1) < 17.8s, so surface flame unchanged 39. Flame angle from understorey wind for 16.4o external flame - Equ. 5.29 40. Find 1 to 4, Equations. 5.53 – 5.59 1 = 1.07 radians, = 0.00 radians, 41. Identify the plume pathway for the combined 4 2 = 0.64 radians, 3 = -0.13 radians, Pathway 3 flame – Fig. 5.13 42. Find Pp – Equ. 5.39 195cm 43. Find PD – Equ. 5.44 132cm 44. Find Ppa for P3 – Equ. 5.49 - depth of 60cm penetration 45. Find Spp. Equ. 5.34 29cm 46. Find P (minimum of Ppa and Spp) 29cm 47. Identify the plume pathway for plant flame – Pathway 3 242 Fig. 5.13 48. Find Pp – Equ. 5.39 391cm 49. Find PD – Equ. 5.44 330cm 50. Ppa = pathway 3 – depth of penetration 58cm 51. Find Spp. Equ. 5.34 62cm 52. Find P (minimum of Ppa and Spp) 58cm 53. Increment length (Max of 2 P values – plant 6cm penetration) / 10 54. Temperature from new flame in plant, using 949oC P = penetration from last second + step number * increment length 55. Find the temperature from the combined 316oC, flame 56. Preheating from surface fuels occurring No NS fuels, so no preheating before NS fuels ignited 322oC 57. Temperature at this point from before ignition (preheating T1) 58. Preheating1 - from [27] divided by for 1.0 preheating T1 59. from maximum temp out of plant/external 0.0s flame multiplied by 1-preheating 1 243 949oC >260oC, 1 > 0.0s. Ignition 60. Critical state test – Equ. 3.15 for t occurs. 61. Repeat steps 54 to 60 for all ten increments Ignition occurs for three of ten steps 62. Depth of ignition = Increment length * 17.3cm ignited in second time step. number of increments ignited 63. Find tr for leaves – Equ. 4.21 Surface flame still burning, so tr = 7.6s 64. Find P – is the time that has elapsed since 1s elapsed, less than 7.6s so P is 0. If the first leaves were ignited greater than tr? elapsed time was greater than 7.6s, then (Section 3.23) P would be equal to previous plant penetration, i.e. 1.7cm 65. Find Ox2, Oy2 – Equ. 3.9, 3.10 Ox2 = -30.0cm, Oy2 = 5.0cm 66. Find horizontal/vertical distances (xi, yi) xi2 = 19.9cm ignited; yi2 = 3.2cm xin = cos( f)*penetration + xi(n-1) yin = sin( f)*penetration + yi(n-1) 20.2cm 67. Find total penetration Pt: Pt = [y − (O i2 ] − O y ) + [xi 2 − (O x 2 − O x )] 2 y2 2 Equation 6.54 68. Number of branches ignited from total 0.8 branches penetration – Equ. 5.65 69. Flame length produced in 2nd time step 0.68m 244 70. Angle of plant flame from mid-shrub wind 23.1o speed – Equ. 5.29 This process is repeated for each consecutive second, with the following effects accumulating with each new step: 1. Steps 57 & 58 are calculated for every preceding time step as well as for preheating due to the new flame in the plant. For example, at time-step five, preheating is calculated from time-steps one to four for the external flame, and two to four for the plant flame (which was not present at step one) 2. Step 64 must be calculated for each preceding time-step so that P is equal to all leaves that have extinguished at the current time-step The progression of the flame through the plant is shown in table 6.38. The maximum flame length was 1.50m, occurring six to eight seconds after ignition. During these three seconds, the flame height relative to the slope was 1.12m. The flame origin remains stationary until the 9th second as the initial leaves ignited are still burning, but after this point the flame takes five seconds before it has burnt across the plant and six seconds to burn to the top of the plant. In this case, the entire plant has been consumed. Table 6.38. Flame propagation through an example shrub Step Ox (cm) Oy (cm) 1 -30.0 5.0 0.25 8.1 0.08 2 -30.0 5.0 0.68 23.1 0.32 3 -30.0 5.0 0.76 25.7 0.38 4 -30.0 5.0 1.34 41.7 0.94 5 -30.0 5.0 1.43 43.8 1.04 6 -30.0 5.0 1.50 45.2 1.12 7 -30.0 5.0 1.50 45.2 1.12 8 -30.0 5.0 1.50 45.2 1.12 (m) f (degrees) Fh (m) 245 9 -27.3 5.8 1.47 44.5 1.09 10 -10.1 8.2 1.30 40.7 0.93 11 -6.1 9.9 1.24 39.2 0.88 12 29.2 27.0 0.80 26.9 0.63 13 30.0 43.4 0.46 15.7 0.56 14 30.0 53.5 0.00 0.0 0.53 15 30.0 53.5 0.00 0.0 0.53 These results refer to an isolated plant as covered in chapter five; the model now moves on to model the spread of fire through the stratum. Initial flame spread through stratum 71. Rank all flame lengths from longest to shortest so that different species’ flame lengths can be synchronised. 72. Calculate (cm) for each flame length – Equ. 5.74, and merged flame length m – Equ. 5.75 73. Find representative values of Ox, Oy for plant flame from those of maximum m 111 111 111 109 107 103 100 97 72 70 65 50 33 14 14 14 1.71 1.71 1.71 1.66 1.61 1.48 1.42 1.34 0.80 0.76 0.68 0.46 0.25 0.00 0.00 0.00 Ox = -30.0cm Oy = 5.0cm flame length 74. Find P for stratum – Equ. 6.39 103cm 246 75. Find Spp – Equ. 5.34 403cm 76. Find T from P + (Spp / 10) 696o 77. Find 2.8s 78. Test flame connection – Equ. 3.3 Connection does not occur The rate of spread (km/h) within a stratum can be calculated from the values of Ox and the slope according to: R= 0.036 xin cos(θ g )t n Equation 6.55 Where xni is the value of xi at time n and tn is the number of seconds passed since ignition. In this instance however, the rate of spread at this point is still determined by the surface fuels as connection did not occur within the shrub stratum. The next step is to determine whether this flame is able to ignite the next stratum, which in this case is the crown. . Crown ignition 79. Identify potential ignition points in tree Ox = -185cm base. The left-hand edge is used in this Oy = 380cm example. 80. Find - Equ. 5.26 33cm 81. Find P – Equ. 3.11 4677cm 82. Identify in which flame zone Ox, Oy lies Zone III by comparing P to the flame length 247 83. Find f (distance from flame base to top of burning shrubs based on Oy and 0.44m f values during the longest flame) 84. Find Spp. Equ. 5.34 0cm 85. Find the temperature at Ox, Oy + one 37oC tenth of the penetration depth – Equ. 5.16 86. Ignition delay time of leaves at that 2746s temperature – Equ. 4.10 87. Critical state test – Equ. 3.15 37oC < 220oC, Ignition does not occur. 6.512 Resulting fire behaviour The process examined above produces a fire that spreads through the surface fuels igniting shrubs with flames large enough to produce lateral merging, but with insufficient wind or slope to allow flame connection. The resulting flame is 1.12m tall but only spreads at the speed of the surface fire which is 0.07km/h. Despite the 1.5m flame length, the 45o angle of the flame means that most of the heat has dissipated by the time it reaches the tree crowns. Consequently, the temperature in the crown is too low and crown fire does not initiate. 6.52 Summary The conceptual model of fire behaviour presented in chapter three consisted of two central aspects – the spatial arrangement of fuels in an array and the flammability properties that determine whether new fuels will be ignited. Chapter four provided mathematical models quantifying the flammability of individual fuel particles. The influence of the geometric relationships between these particles on fire spread was studied in the next two chapters, along with the influence of fuel structure on other factors such as wind and moisture. Chapter six produced a completed model, which has been built into a spreadsheet and attached on a 248 separate CD. The following chapters provide some validation of the model, and then demonstrate the ways in which it contrasts with some of the older models in the direction it provides for fuel and fire management. 249 Chapter 7 Model Validation 7.1 Overview 7.11 Validating models Validation is considered by some to be non-essential to the development of a model (eg Mankin et al 1977), but important for gaining the confidence of users and demonstrating its legitimacy in the field of operations (Rykiel 1996). Others point to the inability to ever adequately validate a model, suggesting that a model is a hypothesis and as such can only ever be falsified (Holling 1978). In this sense, validation can be seen more as a process of identifying the limits of or the confidence around a model’s usefulness, the areas where it does appear to reflect reality. Validation is a distinct process from verification, which is concerned with the logic of the model formulation and its correct transfer into the mathematical computing framework (Fishman and Kiviat 1968). Verification of this model can be achieved by examining the model description in the preceding chapters, the worked example provided at the end of chapter six and the attached CD containing the model rendered as an Excel spreadsheet. 250 One process often referred to as validation is that of ‘predictive validation’ (Rykiel 1996) which involves a comparison of modelled fire behaviour with measured values. By itself it cannot answer the question as to whether a model is “good enough”; all it can provide is a relative measure of its performance across a range of variables. Because it is relative, the results only have meaning if presented in the context of other standards. Some other standards to compare the accuracy of the model with include random results within a certain range, estimates of experienced people or results from another model. A more informative part of model validation is that of analysing its credibility. Credibility is related to the usefulness of the model for decision making (Holling 1978, Sargent 1984, Rykiel 1996) and is therefore an important part of the assessment for a model such as this which is concerned with providing guidance for management decisions. Incident management frequently involves binary decision processes such as identifying whether the conditions are safe to conduct a backburn, to place suppression crews in front of a fire or severe enough to evacuate a town. Such decisions are made on broadly recognised criteria which provide cut-off points based on aspects of fire behaviour, so it is irrelevant whether the model is “nearly right” if the predicted behaviour does not fall into the range that would guide the management effectively. A third aspect of validation is what has been referred to as an ‘extreme condition test’ (Sargent 1984). Rykiel (1996) states that the purpose of an extreme condition test is to “reveal if behaviour outside of normal operating conditions is bounded in a reasonable manner”. By itself this test does not validate or invalidate a model, rather it identifies reasonable limits. 7.12 The approach adopted The model presented on the attached Excel spreadsheet (referred to hereafter as the Forest Flammability model or FFM) was tested against a number of experimental and unplanned fires. Equations and symbology used in the model are given in appendices II and III, and instructions for use of the spreadsheet are given in appendix VII. The model used a maximum flame temperature of 950o as an approximation of temperatures reported by Gould et al 251 (2007a), the equations of Gould et al (2007a) to find dead fuel moisture, and estimated available surface fuel using McArthur (1962) as given by Tolhurst (2010). The number of fires used is limited due to the lack of resources required for adequate validation, however as approximately only one-third of fire behaviour models receive any form of validation (see table 1.1), this is an important step and it is anticipated that a more complete treatment will be carried out in the future. Validation was carried out in three steps. Firstly, a Monte Carlo analysis was performed for two different forest structures as an ‘extreme condition test’ (Sargent 1984) to identify the potential limits to average flame length and rate of spread for 1000 different scenarios in each array. The scenarios tested fire behaviour for each array for random values of slope from -50 degrees to +50 degrees, temperature from one to 50 degrees, relative humidity from one to 100% and wind speeds from zero to 100km/h. The model used for dead fuel moisture in all validations was Gould et al (2007b), as rendered into equation form by Tolhurst (2010). This model is a compilation of three sub-models for use at different times and different light conditions and was adjusted to accommodate canopy shading by introducing an extra criterion, where canopy cover of greater than 50% is treated in the same way as cloud cover of greater than 50%. The Monte Carlo analysis varied the fuel moisture model for all days of the year and all times of day or night. The FFM was tested alongside three Australian forest fire models in popular use – Leaflet 80 (McArthur 1962), the McArthur Meter (McArthur 1967), and Project Vesta (Gould et al 2007a). These were all compared to recorded maximum rates of spread to assess whether extension of each model to extreme conditions was realistic. While the most extreme values were used for the empirical models, this was not sufficient for the FFM as the steepest slope or strongest wind speed etc may not produce the fastest spread or tallest flames due to complexity and feedback mechanisms discussed in chapter three. The second step was a predictive validation that involved a case by case examination of a range of fires, both experimental and unplanned. The modelled results of these were compared with the observed results, and also with modelled results from the three Australian empirical models. The experimental fires were designed to compare the value of the model as a decision support tool for prescribed burning and management of low intensity unplanned 252 fires (bushfires), and to provide direct observation of flame behaviour when compared to modelled behaviour so that aspects of the FFM requiring further work could be identified. The bushfires provided a wider range of fire behaviour to asses the value of the FFM as a decision support tool against the other models under higher fire danger conditions. Some of the test conditions involved high intensity fire which is almost certainly outside of the domain of the empirical models. This was intentional, as none of the three empirical model sources actually state the domain of their models, and in the absence of an alternative these models are in practice used for managing high intensity fire incidents (e.g. Tolhurst 2009) or marketed for such use (e.g. CSIRO 2008). The credibility analysis was conducted alongside the validation, and produced a ranking of the usefulness of each model for decision making against the widespread criteria employed in fire management. The scores produced were weighted by the magnitude of the error. Details of each analysis are provided below. 7.2 Extreme Condition Test 7.21 Methods Two communities were examined in Monte Carlo analyses to identify the extremes of predicted fire behaviour. The Monte Carlo analysis was not intended to identify the statistical distribution of fires, but to identify where the limits lay. The communities examined were the regrowth subalpine Eucalyptus niphophila forest, and old growth E. regnans forest classified by Ashton (2000) as “Type A”. These two were chosen as they represent opposite extremes of forest structure ranging from a ‘young’ (6 years since fire, Figure 7.1) low heath-like structure to the tallest hardwood forests on earth (112 years since fire, Figure 7.2), and both have good data available to describe the fuel arrays. The relevant fuel parameters for the empirical models are shown in table 7.1. These were found by using the most extreme of the weather and terrain variables for the range specified 253 (temperature = 50oC, relative humidity = 1%, wind = 100km/h, KBDI = 203, drought factor = 10, time since rain = 30 days for 1mm rain, slope = 30o). The limitation to 30o is a response to Gould et al, who specify that slope values greater than 30o should not be used. Fuel moisture was also given a minimum of 1% where the relevant models predicted less. All strata are assumed to overlap in both communities. Tolhurst (2010) allows for a changed wind reduction factor in McArthur’s 1967 model, assuming a default factor of three. Neither of the McArthur models however state that this is the case, although McArthur (1962) does provide a graph for tall forest with a reduction factor of approximately 1:4.5 as the default calculation for the earlier model (equation 6.6). Due to this disparity and the fact that the McArthur meter was not designed with the ability to alter the wind reduction factor, the factor was changed for McArthur (1962 – regrowth Snowgum = 1, Mountain Ash = 6) but not for McArthur (1967) calculations. The dead fuel moisture content for this model was also set to a value of one, as a negative value was predicted in the spreadsheet. Figure 7.1. Regrowth E. niphophila forest 254 Fuel parameters for the E. niphophila regrowth site were collected in a field survey of a site near Guthega, in Kosciuszko National Park at an altitude of around 1620m and a north-westerly aspect. Surface fuel loads were modelled from Zylstra (2007) with the parameters given in table 2.2. All fuel variables used in the Forest Flammability model are given along with their derivation in tables 7.3 to 7.8. The phrase “accurate to x%, y% confidence” defines the sample as accurate to within +/- x% of the mean within a y% Figure 7.2. Mature E. regnans forest, community A from Ashton 2000 confidence interval as determined with a ztest (Freund and Simon 1997, p320). The terms “estimate” and “observation” indicate that the variable was not measured but in the first case it was approximated and in the second case it was a simple observation such as the shape of a leaf. The term “unknown” indicates that there is no information yet available on the variable and an uninformed estimate has been used. Where the terms “unknown” or “estimate” occur, further work is desirable. Fuel parameters for the E. regnans old growth forest were measured from botanical structural diagrams and field survey results provided by Ashton (2000), with other species-specific values measured from herbarium samples or collected in field surveys. Remaining values were estimated as stated. Surface fuel loads were modelled from Ashton (1975) with the parameters given in table 2.2. Ashton compares the average density of woody species for two surveys in the old growth forest, showing the change in density between 1950 and 1995 (Ashton 2000, table 7). In order to estimate the 2009 density, an assumption of linear change was used from these two years to give the species density used for fire behaviour calculations in 2009. The parameters used in the Ash study do not enable a full random data set as the live fuel moisture values were set to moisture-stressed values rather than modelled on the conditions. 255 All fuel variables used in the Forest Flammability model are given along with their derivation in tables 7.4 to 7.14. Plant spacing Where plant spacing was not measured directly, it was calculated from quadrat surveys by assuming an even distribution. In an even, gridded distribution, plants have four neighbours on the same x and y coordinates and four neighbours diagonally adjacent to them. For the neighbours on the same axes, the spacing Sf1 is equal to the square root of the quadrat area Aq divided by the number of plants np: Aq s f1 = np Equation 7.1 For the diagonal neighbours, the spacing sf2 can be found using Pythagoras’ theorem as follows: 2 s f 2 = s f1 + s f1 2 Equation 7.2 By combining and simplifying the two equations, the average separation between plants is: sf = 2 Aq np Equation 7.3 256 Table 7.1. Fuel parameters used in the empirical models for the E. niphophila community FUEL PARAMETER Model Surface Near surface Near surface Elevated height -1 McArthur 10t.Ha (1962) (Zylstra 2007) McArthur 10t.Ha-1 (1967) (Zylstra 2007) Gould et al Score = 2 Elevated height - - - - - - - - Score = 4 11cm - 44cm (2007a) Table 7.2. Fuel parameters used in the empirical models for the E. regnans community FUEL PARAMETER Model Surface Near surface Near surface Elevated height -1 McArthur 22.25t.Ha (1962) (Ashton 1975) McArthur 22.25t.Ha-1 (1967) (Ashton 1975) Gould et al Score = 4 Elevated height - - - - - - - - Score = 4 50cm - 2m (Maximum (2007a) from Gould et al 2007b) Table 7.3 Details for E. niphophila community Parameter Surface fuel Value 10.0t.Ha Details -1 Mean 6 year old fuel load (Zylstra 2007) load Near surface Poa costiniana species Near surface Clear dominant from observation, only minor occurrences of other species. Closed cover Observation Elevated Olearia phlogopappa var. flavescens 30m transect survey result species (10%), Bossiaea foliosa (33%) plant spacing 257 Elevated plant 0.431m Measured from plant centre to nearest neighbours for spacing 42 measurements. Canopy species E. niphophila Clear dominant from observation Canopy plant 4.1m 66 mature trees in 1250m2 and 26 new saplings in 400m2 quadrats spacing Soil texture Clay loam Soil texture typical for granite soils in the area Soil moisture Modelled using equation 6.29 Table 7.4 Poa costiniana details Parameter Value Details m (%ODW) 80 Estimate only P (oC) 260 Standard value for volatile oil content of one (Table 4.2) % Dead 50 Estimated level of curing representative of this altitude for drought conditions in mid summer Leaf form Round Observation d (mm) 0.25 Estimate Wl (mm) 0.25 Estimate Ll (mm) 110 Estimate Sl (cm) 0.03 Standard value for fine grasses Os 1 Observation Sb (cm) 0 Observation Wb (cm) 20 Visual estimate/ actual measure makes little difference in closed cover He (cm) 0 Observation Ht (cm) 11 Mean of 13 measurements, accurate to 25%, 99.5% confidence Hc (cm) 0 Observation Hp (cm) 11 Mean of 13 measurements, accurate to 25%, 99.5% confidence w (cm) 20 Visual estimate/ actual measure makes little difference in closed cover Table 7.5 Olearia phlogopappa var. flavescens details Parameter Value m (%ODW) Details Modelled using equation 6.19 P (oC) 260 Standard value for volatile oil content of one (Table 4.2) % Dead 20 Estimate 258 Leaf form Flat Observation d (mm) 0.32 Mean of 10 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 4.5 Mean of 15 measurements, accurate to 25%, 99.5% confidence Ll (mm) 14.9 Mean of 15 measurements, accurate to 25%, 99.5% confidence Sl (cm) 0.09 Counted for 9 leaves Os 4 Mean of 5 measurements Sb (cm) 12.9 Mean of 15 measurements, accurate to 25%, 95.0% confidence Wb (cm) 26.8 Mean of 12 measurements, accurate to 25%, 90.0% confidence He (cm) 29 Mean of 16 measurements, accurate to 25%, 99.5% confidence Ht (cm) 57 Mean of 16 measurements, accurate to 25%, 99.5% confidence Hc (cm) 25 Mean of 10 measurements, accurate to 25%, 99.5% confidence Hp (cm) 74 Mean of 10 measurements, accurate to 25%, 99.5% confidence w (cm) 60 Mean of 15 measurements, accurate to 25%, 97.5% confidence Table 7.6 Bossiaea foliosa details Parameter Value Details m (%ODW) 85 Mean value of measurements (Appendix V) P (oC) 300 Experimentally derived value (Table 4.2) % Dead 0 Observation Leaf form Flat Observation d (mm) 0.17 Mean of 152 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 5 Little variability, single measurement Ll (mm) 5 Little variability, single measurement Sl (cm) 0.20 Mean of 20 leaves Os 5 Estimate Sb (cm) 0 Observation Wb (cm) 57 Mean of plant diameter and Hp-Hc He (cm) 18.4 Mean of 10 measurements, accurate to 25%, 99.0% confidence Ht (cm) 42.0 Mean of 10 measurements, accurate to 25%, 99.5% confidence Hc (cm) 16.3 Mean of 10 measurements, accurate to 25%, 97.5% confidence Hp (cm) 56.9 Mean of 10 measurements, accurate to 25%, 99.5% confidence w (cm) 72.8 Mean of 10 measurements, accurate to 25%, 90.0% confidence Table 7.7 E. niphophila details 259 Parameter Value m (%ODW) o Details Modelled using equation 6.16 P ( C) 220 Experimentally derived value (Table 4.2) % Dead 0 Observation – dead leaves not retained on live branches Leaf form Flat Observation d (mm) 0.54 Mean of 10 samples, accurate to 0.05mm, 99.5% confidence Wl (mm) 20.5 Mean of 10 samples, accurate to 25%, 99.5% confidence Ll (mm) 65.2 Mean of 11 samples. Accurate to 25%, 99.5%confidence Sl (cm) 2.3 Mean of 18 leaves Os 5.5 Mean of 5 branches Sb (cm) 0 Observation Wb (cm) 110 Mean of crown width & height He (cm) 78 Mean of 10 samples, accurate to 25%, 97.5% confidence Ht (cm) 150 Mean of 10 samples, accurate to 25%, 99.0% confidence Hc (cm) 111 Mean of 10 samples, accurate to 25%, 95.0% confidence Hp (cm) 174 Mean of 10 samples, accurate to 25%, 97.5% confidence w (cm) 158 Mean of 10 samples, accurate to 25%, 95.0% confidence Table 7.8 Details for E. regnans Community A, 112 years since low intensity fire Parameter Surface fuel Value 22.2t.Ha -1 Details Reference Equilibrium fuel load for E. regnans Ashton (1975) Ashton (2000) load Near surface Polystichum Dominant fern and graminoid used species proliferum 73%, to represent ferns and graminoids Lepidosperma based on the ration in elatius 27% presence/absence plots 1.15m Polystichum proliferum projected to Near surface plant spacing Ashton (2000) 96% cover by the year 2009. Elevated Olearia argophylla Projected value; clear dominant species 48.8% species. Three other species have Ashton (2000) similar lesser density but no fuel dimensions were available. Elevated plant Closed spacing Midstorey Calculated from projected number Ashton (2000) of plants per Ha using equation 7.3 Acacia dealbata 1%, Species of different heights so Ashton (2000) 260 species Pomaderris aspera Acacia ideally forms taller stratum, 99% but heights overlap and Pomaderris is strongly dominant. Midstorey plant 3.98m spacing Calculated from projected number Ashton (2000) of 1260.8 plants per Ha using equation 7.3 Canopy species E. regnans 100% Study describes forest as a pure Ashton (2000) stand Canopy plant 20.6m spacing Calculated from projected number Ashton (2000) of plants per Ha using equation 7.3 (47.1) Soil texture Clay loam Typical soil type for E. regnans on E.g. Robinson et al granite soils (2003), Ringrose and Nielsen (2005), Table 7.9 Polystichum proliferum details Parameter Value Details m (%ODW) 239 Single measurement 347 Calculated from Silica Free Ash content of Pteridium esculentum o P ( C) (Dickinson and Kirkpatrick 1985) using equation 4.1 % Dead 50 Estimated value due to drought and heat wave conditions Leaf form Flat Field observations d (mm) 0.24 Mean of 5 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 5 Single measurement, little variability Ll (mm) 11.3 Mean of 7 measurements, accurate to 25%, 99.5% confidence Sl (cm) 0.33 Counted for 32 leaves Os 2 Observation, no variability between plants Sb (cm) 10 Visual estimate Wb (cm) 35 Mean of 3 measurements, little variability in plant He (cm) 0 Single plant Ht (cm) 62 Mean of 4 measurements from 1 plant Hc (cm) 0 Single plant Hp (cm) 100 Single plant w (cm) 160 Mean of 2 measurements from 1 plant 261 Table 7.10 Lepidosperma elatius details Parameter Value Details m (%ODW) 100 Unknown P (oC) 260 Estimate for VOC1 % Dead 65 Low end of curing range reported for February 2009 (McCaw et al 2009) Leaf form Flat Observed from photograph d (mm) 0.40 Default value for sedges Wl (mm) 13 Average published width (http://plantnet.rbgsyd.nsw.gov.au/cgibin/NSWfl.pl?page=nswfl&lvl=sp&name=Lepidosperma~elatius ) Ll (mm) 140 Model maximum Sl (cm) 0.6 Estimate for coarse graminoid Os 1 Default value for graminoids Sb (cm) 0 Default value for graminoids Wb (cm) 150 Mean published plant diameter (http://www.yarraranges.vic.gov.au/Directory/S2_Item.asp?Mkey=77 6&S3Key=38 ) He (cm) 0 Estimated from photograph Ht (cm) 100 Estimated proportionally from photograph Hc (cm) 0 Default value for graminoids Hp (cm) 125 Average published height (http://plantnet.rbgsyd.nsw.gov.au/cgibin/NSWfl.pl?page=nswfl&lvl=sp&name=Lepidosperma~elatius ) w (cm) 150 Mean published plant diameter (http://www.yarraranges.vic.gov.au/Directory/S2_Item.asp?Mkey=77 6&S3Key=38) Table 7.11 Olearia argophylla details Parameter Value Details m (%ODW) 70 Estimate from driest sample of Olearia aglossa 328 Calculated from Silica Free Ash content using equation 4.1 o P ( C) (Dickinson and Kirkpatrick 1985) % Dead 0 Assumption from common observation Leaf form Flat Observation 262 d (mm) 0.27 Mean of 10 measurements accurate to 0.05mm, 99.5% confidence Wl (mm) 33.2 Mean of 14 measurements accurate to 25%, 99.5% confidence Ll (mm) 78.6 Mean of 14 measurements accurate to 25%, 99.5% confidence Sl (cm) 1.16 From 21 leaves on herbarium specimen (http://plantnet.rbgsyd.nsw.gov.au) Os 4 Unknown Sb (cm) 20 Unknown Wb (cm) 40 Unknown He (cm) 116 Mean of 9 plants in diagram (Ashton 2000) Ht (cm) 290 Mean of 9 plants in diagram (Ashton 2000) Hc (cm) 116 Mean of 9 plants in diagram (Ashton 2000) Hp (cm) 464 Mean of 9 plants in diagram (Ashton 2000) w (cm) 464 Mean of 9 plants in diagram (Ashton 2000) Table 7.12 Pomaderris aspera details Parameter Value Details m (%ODW) 100 Value unknown P (oC) 260 Estimate for VOC1 % Dead 0 Assumption Leaf form Flat Observation d (mm) 0.31 Mean of 9 samples, accurate to 0.05mm, 99.5% confidence Wl (mm) 13.1 Mean of 12 samples on herbarium specimen, accurate to 25%, 99.5% confidence (http://plantnet.rbgsyd.nsw.gov.au) Ll (mm) 42.2 Mean of 11 samples on herbarium specimen, accurate to 25%, 99.5% confidence (http://plantnet.rbgsyd.nsw.gov.au) Sl (cm) 1.67 From 26 leaves on herbarium specimen (http://plantnet.rbgsyd.nsw.gov.au) Os 4 Unknown Sb (cm) 20 Unknown Wb (cm) 40 Unknown He (cm) 406 Mean of 17 plants in diagram (Ashton 2000) Ht (cm) 884 Mean of 17 plants in diagram (Ashton 2000) Hc (cm) 478 Mean of 17 plants in diagram (Ashton 2000) Hp (cm) 1072 Mean of 17 plants in diagram (Ashton 2000) w (cm) 435 Mean of 17 plants in diagram (Ashton 2000) 263 Table 7.13 Acacia dealbata details Parameter Value Details m (%ODW) 135 Single indicative measure 333 Calculated from Silica Free Ash content (Dickinson and Kirkpatrick o P ( C) 1985) using equation 4.1 % Dead 0 Assumption Leaf form Flat Observation d (mm) 0.22 Mean of 10 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 20 Mean of 10 measurements, accurate to 25%, 99.5% confidence. Halved to account for bipinnate structure Ll (mm) 28 Mean of 10 measurements, accurate to 25%, 99.5% confidence. Halved to account for bipinnate structure Sl (cm) 1.05 Mean of 21 leaves Os 5.2 Mean of 5 branches Sb (cm) 30 Visual estimate Wb (cm) 40 Visual estimate He (cm) 1424 Mean of 7 trees in diagram (Ashton 2000) Ht (cm) 1996 Mean of 7 trees in diagram (Ashton 2000) Hc (cm) 1294 Mean of 7 trees in diagram (Ashton 2000) Hp (cm) 2000 Mean of 7 trees in diagram (Ashton 2000) w (cm) 706 Mean of 7 trees in diagram (Ashton 2000) Table 7.14 Eucalyptus regnans details Parameter Value Details m (%ODW) 102 Equation 6.16 P (oC) 220 Estimate for VOC2/3 % Dead 0 Assumed Leaf form Flat Observation d (mm) 0.39 Mean of 15 mature E. delegatensis leaves Wl (mm) 28 Mean of 15 mature E. delegatensis leaves Ll (mm) 129 Mean leaf length from (Ashton 1975) Sl (cm) 2.5 Mean from 31 mature E. delegatensis leaves Os 5.4 Mean from 5 mature E. delegatensis branches Sb (cm) 208 Mean of 25 measurements from diagram (Ashton 2000) 264 Wb (cm) 486 Mean of 25 measurements from diagram (Ashton 2000) He (cm) 3409 Mean of 5 trees in diagram (Ashton 2000) Ht (cm) 6678 Mean of 5 trees in diagram (Ashton 2000) Hc (cm) 3130 Mean of 5 trees in diagram (Ashton 2000) Hp (cm) 6933 Mean of 5 trees in diagram (Ashton 2000) w (cm) 1577 Mean of 5 trees in diagram (Ashton 2000) 7.22 Results The behaviour described by each model is given in table 7.15. By comparison, maximum reported rates of spread for forest fires have an approximate upper limit of 10 - 11km.h-1 (e.g. Billing 1983, McCaw et al 2009), with some reports of spread up to 12km.h-1 (Tolhurst 2009). All three empirical models predicted well beyond this range, while the FFM predicted beyond the range for one community. The excessive values produced by the empirical models reflect the fact that these models are predicting well beyond the domains within which they were developed. By itself, the data in table 7.15 does not invalidate the models within their domains, but it does demonstrate that the application of these models to conditions beyond their domains is inadvisable. Unfortunately, the conditions commonly experienced during bushfire events are frequently beyond the domains within which the models were developed, suggesting that these models are not necessarily applicable as incident management tools except under more moderate conditions. Table 7.15. Maximum rates of spread and flame heights as modelled by three Australian empirical forest fire behaviour models and the FFM using the fuel parameters given in tables 7.1 to 7.14 with 30o slope, 100km/h wind, temperature of 50oC, relative humidity of 1% and drought factor/fuel availability at maximum. LFMC values for the E. regnans forest were set to moisture – stressed values in the absence of available fuel moisture models. Model McArthur Max ROS (km/h) Max Flame height Max ROS (km/h) Max Flame height E. niphophila (m) E. niphophila E. regnans (m) E. regnans 17,015 4,831 37,859 10,749 64 832 142 1,854 166 152 481 890 (1962) McArthur (1967) Gould et al 265 (2007a) FFM 38.7 21 9.8 125 Peak fire behaviour modelled by the FFM was less intense than that produced by the other models. The fastest rates of spread were 38.7km/h and 9.8km/h and the greatest flame heights were 21m and 125m for each fuel array respectively (Figures 7.3 and 7.4). The very high figures achieved in the regrowth Snowgum community represent outliers, where 97% of results were values of less 12km/h with flame heights less than 5m tall. Such results represent conditions that have not been recorded to date; the extreme rates of spread for example were modelled as independent crown fire spread advancing for about 200m ahead of the surface fire before extinguishing. These values represent approximations of the possible maxima that the model will attain for each community, but do not give the maximum attainable by the model itself as other communities or untested conditions may produce faster spreading fires or taller flames. Figure 7.3. Results of the Monte Carlo analysis for six year-old E. niphophila regrowth Figure 7.4. Results of the Monte Carlo analysis for E. regnans Community A with growth projected to 112 years post-fire 7.23 Discussion This exercise examined fire behaviour under conditions more severe than any yet on record, although the weather of February 7 2009 in Victoria did approach the limits albeit in gentler terrain. Because of the extremity of the conditions, it is to be expected that the models will predict outside of the historic range, and even possible that new fire phenomena such as 266 independent crown fire will be encountered that predict significantly beyond the range of past observations. What the exercise demonstrates is that unlike the empirical models which are limited by a finite domain, the FFM is feasible for broad general usage and can therefore be applied to new and extreme conditions because the output from the FFM contains natural limits to spread. For the conditions examined, fires were unable to spread faster than they were modelled because there was no physical mechanism available for them to do so. This creates the possibility of modelling potential extreme scenarios under changed climate conditions for particular fuel arrays. In the context of the goal for this test stated by Rykiel (1996) which was to “reveal if behaviour outside of normal operating conditions is bounded in a reasonable manner”, this analysis suggests that in contrast to the empirical models, fire behaviour modelled by the FFM is bounded in a reasonable manner. The Monte Carlo analysis reveals another interesting difference to the two major empirical models by McArthur (1967) and Gould et al (2007a) when they are examined in the same way using random slope and weather parameters within the bounds specified for the FFM and slope limited to 30o (Figures 7.5 and 7.6). The assumption that flame height and rate of spread are directly correlated is not made in the FFM, and as a result the two communities show significantly different patterns in these areas. Fire behaviour in the E. niphophila community showed very little trend in flame heights except that, excluding the outliers, the tallest flames occurred in fires spreading less than 5km/h. Above this speed, flames were typically less than 2.5m tall. Due to the range of weather conditions applied, low intensity fire behaviour was not observed in the E. regnans community, however three distinct groups of fires are visible based on flame height. At the slowest rates of spread, flames are typically between 50 and 90m tall, representing passive crown fires and slow active crown fires. At rates of spread approaching 1km/h, a second and very distinct group appears with flame heights of 15-20m, representing ignition of the midstorey but not the canopy. Examination of the data shows that these occurred only on slopes of 17o or greater but at all wind speeds. Flames greater than 100m 267 representing well developed active crown fires occurred at rates of spread of 2km/h or greater. Thse groupings demonstrate the effect of vegetation stratification on fire behaviour and the limits that are imposed by the fire growth patterns described in section 3.22. Figure 7.5. Results of the Monte Carlo analysis for six year-old E. niphophila regrowth using McArthur 1967 and Gould et al 2007a. Figure 7.6. Results of the Monte Carlo analysis for E. regnans Community A with growth projected to 112 years post-fire, using McArthur 1967 and Gould et al 2007a. 7.3 Predictive Validation and Credibility Analysis 7.31 Methods of validation 7.311 Predictive validation The validation provided is limited to a small number of experimental fires and three unplanned fires. Options were limited due to the time and resource constraints of a PhD project; however the adequacy of the sample set is confirmed via statistical analysis. The experimental fires were all conducted in one forest community with a simple and relatively consistent understorey. Although the FFM uses data from the two dominant species (Daviesia mimosoides and E. pauciflora) along with others to create equations for leaf flammability in chapter four, the model is not in any way calibrated to this community 268 through use of any empirical constants and its predictions cannot be expected to be any better for this community than for any other. The three unplanned fires studied were chosen because there was good data available to describe the necessary parameters of fuel, terrain and weather, and because they represented three completely different scenarios. The Guthega fire occurred on January 26th 2003 and represented what may be described as a “typical” bushfire under adverse conditions; that is, the forest was mature and the weather conditions were within a normal summer range. The Tooma Dam fire (13th October 2006) provided an example of a fire burning in a similar environment under quite similar weather conditions, but with a young regrowth forest. The Kilmore fire (8th February 2009) contrasted with both of these by providing a scenario of the most extreme fuel and weather conditions, burning in a fuel array composed of species that had not contributed at any of the leaf flammability equations in chapter four. Fuel and weather parameters for the various fires were collected from sources described separately in each instance. Field data collection varied as it was often achieved opportunistically, such as during the course of fire suppression operations. In some cases data was collected directly from the site; in others published studies or equations developed in this thesis were used, or measurements made from herbarium samples. Where there were no other options available, estimates of some plant measurements were made based on other similar species or default values inserted in the absence of any information. For each fuel complex, the way the data was sourced is described for each parameter. In all cases, all strata were assumed to overlap. Predictions from McArthur (1962), McArthur (1967), and Gould et al (2007a) were generated to find the strength of the model in the context of the other options available. Predictions for these models were developed using the spreadsheet of Tolhurst (2010), modified so that the relevant wind reduction factor was applied to McArthur 1962 but not to McArthur 1967, consistent with those reports. The strength of each model was compared by examining the mean error of the collected results. Where the model prediction was larger than the observed parameter, the mean error was calculated as: 269 E= e−o o Equation 7.4 Where E is the error, e is the expected or modelled value and o is the observed value. When the prediction was less than the observed value, the calculation was: E= o−e e Equation 7.5 For each model, the mean error was calculated, and mean of the FFM was compared with those of the other models using a paired t-test with the Null hypothesis that the means were equal. 7.312 Credibility analysis No published credibility analysis was found that was suitable for use with a fire behaviour model considering the number of parameters to be examined, so the following methodology was developed for this thesis. Value in prescribed burning support As part of the credibility analysis, the value of each model as a decision support tool for prescribed burning was assessed against the criteria given by McArthur (1962) for suitable fire prescriptions, which was that flame heights were to be between one and three feet (approximately 30cm to 1m). Model predictions in each instance were classified according to table 7.16 as destructive/dangerous (D), useful (G) or conservative (C). D, G and C categories were then tallied; results with a rating of “D2” or C2” were given a weight of two, and all others had a weight of one. Each model was given a percent accuracy based on equation 7.6, a destructive/dangerous rating as given by equation 7.7 and a conservativeness rating as given by equation 7.8. A= G G+ D1 + 2 D2 + C1 + 2 C2 Equation 7.6 270 D= D1 + 2 G+ D1 + 2 D2 D2 + C1 + 2 C2 Equation 7.7 C= C1 + 2 G+ D1 + 2 D2 + C2 C1 + 2 C2 Equation 7.8 Models with a destructive/dangerous component under-predict fire behaviour and give prescriptions that would produce destructive or dangerous fire behaviour. Conversely, models with a conservative ranking over-predict flame height and will therefore produce unworkably low-intensity prescriptions or conditions where fire will not propagate. Table 7.16. Rating of model predictions Modelled behaviour Below Range Within range Above Range Observed Above range D2 D1 G behaviour Within range D1 G C1 Below Range G C1 C2 Value in incident management support The second part of the credibility analysis tested the value of each model as a support tool in incident management against standard criteria used in decision making based on flame height, rate of spread and spotting distance. As with prescribed burning, decision making in incident management requires categorised answers to strategic questions; these include: • What is the safest and most effective mode of attack for the fire? • Are initial attack methods likely to be successful or should additional resources be sourced to manage a longer term campaign? • Will the rate of spread allow time to complete a backburn or will it endanger fire crews? To what distance will the fire impact in a given time span? 271 To assess the value of each model in these decision making processes, three criteria were designated. The first of these was the “attack method criterion”. The standard and widely used (Australasian Fire Authorities Council 1996) flame height values for determining the form of attack are: 1. Flames 0 to 1.5m, direct attack 2. Flames 1.5 to 3m, Parallel attack 3. Flames 3m to 10m, Indirect attack 4. Flames greater than 10m, property protection Model results were rated using the principles of table 7.16, so that if observed fire behaviour fell within a given category, the model results were rated as conservative, good or dangerous depending on whether they fell within the stated range for that form of attack or not. In addition, the weighting system was extended due to the larger number of behaviour categories, so that modelled results were weighted by the number of categories that they differed from the observed behaviour. For example, if the observed behaviour could be addressed by direct attack but the model suggested that it required property protection, the prediction has a weighting of three because it is three categories removed from the correct one. The second criterion was the “initial attack criterion”. The values for this are taken from a guideline given in Gould et al (2007a) and attributed to McCarthy and Tolhurst (1997). The guideline states that the estimated maximum fire behaviour allowing first attack success is where fires do not spread faster than 600m per hour or spot further than 50m. Spotting distances were calculated using equation 7.9, which is a regression of the data given in table 10.1 of Gould et al (2007a) based on flame height and wind speed. The regression had an R2 of 0.97 based solely on flame height; wind speed was not fitted but as the R2 indicates, it has little bearing on the outcome. This model follows a geometric progression and has no upper asymptote to limit the distance by factors such as the burnout time of firebrands, so the spotting distances reported beyond an approximate maximum of 20km were reported as “maximum”. 272 S d = 11.98 f h 2.19 Equation 7.9 Although the data for this equation are taken from Gould et al (2007a), this is not the model used in the report, as discussed in chapter one. Predictions were rated as given in table 7.17, and the percent accuracy, destructive/dangerous and conservativeness ratings calculated using equations 7.6 to 7.8. Table 7.17. Rating of the initial attack criterion Modelled behaviour Within range Observed behaviour Beyond range Within range G C1 Beyond range D1 G The third analysis was the forward planning criterion. There are no standardised or accepted guidelines for this, although it is perhaps the most important of the assessments in that it guides decision making for the majority of campaign fires and determines essential decisions such as the evacuation of residential areas. Consequently, the performance of each model is examined by weighting the number of conservative and dangerous directions given by the model based on the magnitude of the error. A further weighting is applied based on the flame height, as it is potentially less dangerous to miscalculate the arrival of a low intensity fire than a high intensity event. This weighting was based on the flame height categories used in the attack method criterion, so that where the actual flame height called for direct attack, the error weighting was divided by four. Where the actual flame height called for parallel attack, the error weighting was halved; the error weighting for flame heights requiring indirect attack were not altered and flame heights calling for property protection were doubled. The divisions for rate of spread errors are given in table 7.18. Unlike the other scores, those of the forward planning criterion are added together rather than averaged. The addition of scores provides an indication of the spread, so that larger values 273 indicate more severe or more frequent bad consequences, whether these are dangerous decisions or missed opportunities. Table 7.18. Rating of the forward planning criterion Under prediction Rating Over prediction Rating 0-50% G 0-50% G 50-100% D1 50-100% C1 100-200% D2 100-200% C2 200-400% D3 200-400% C3 400-800% D4 400-800% C4 800-1600% D5 800-1600% D5 >1600% D6 >1600% C6 The value of each model was given in terms of all four values for prescribed burning, attack method, initial attack and forward planning. As well as examining the two traditional variables rate of spread and flame height, the FFM was queried to identify any further fire behaviour phenomena, and these were described along with the way in which they were derived. Fire behaviour was then related back to the emergent behaviours described in section 3.22 to ascertain whether any of these had been described by the model. 7.32 Experimental studies The experimental studies were carried out under moderate to high fire danger (FFDI 8-14), within the range of weather conditions often utilised for prescribed burning or for direct attack of bushfires. A typical prescription limits flame dimensions by the amount of crown scorch that they will produce 7.321 Methods Six experimental burns were conducted near Eucumbene Cove in a forest environment dominated by Snowgum Eucalyptus pauciflora with a small component of Mountain Gum E. dalrympleana, an understorey composed of a near monoculture of Hop Scrub Daviesia 274 mimosoides with no grassy near-surface fuels. The burns were conducted in 2005 before the model had been constructed, so that some of the fuel variables were not measured at the time as they had not yet been specified. These burns serve to provide validation for low intensity fires in this fuel array, with a maximum rate of spread and flame height of 0.08km/h and 1.1m. The burns were conducted in plots measuring 10m by 4m, with half of the block designated for ignition and fire build-up (Figure 7.7). A mineral earth containment line was Figure 7.7. Layout for the experimental burn plots, scale is in metres. cleared for approximately 1.5m on the tail and flanks of the fire, and for approximately 3m at the head and at least one Category nine fire fighting “striker unit” was situated adjacent to each fire for quick containment should any escapes occur. The study area of the plot was 5m long with a width of 4m to allow for adequate head fire width for a low intensity fire based on the findings of Wotton et al (1999), while being narrow enough to capture the flame dimensions close to the scale poles. Gould et al (2007a) have suggested that the head fire width should be significantly larger, however it appears that this assertion was based on higher intensity fires and may relate to atmospheric interactions other than the flame merging phenomenon which appears to have driven the findings of Wotton. As the experimental fires studied here were of a low intensity, the width of 4m is considered sufficient. At each 1m interval along the axis of spread, a star picket was hammered into the ground, painted white to the height of 1m then marked with 20cm increments above this point. These served the purpose of marking horizontal and vertical distance for measuring spread rates and flame dimensions. Each fire was ignited using a drip torch with a mixture of three parts diesel to one part petrol, igniting a line initially then augmenting this where necessary throughout the ignition area. 275 Fire behaviour was captured on video using three Sony Digital-8 DCR-TRV265E PAL Handycam cameras located at the tail and flanks of each fire. Measurements of rate of spread and flame dimensions were made from the video footage by recording the time at which the fire front crossed each 1m marker, and the length and height of the flames at this time. The average flame dimensions are therefore a spatial average rather than a time-based average. 7.322 Weather Wind speed, temperature and relative humidity were measured at 30-second intervals using a Kestrel 4000 weather meter with in-built data-logger mounted on a purpose-built wind vane at 1.5m above the ground surface and placed in the shade of a tree. The drought factor was calculated using a McArthur Meter (McArthur 1967) from the current conditions for the area. The recorded wind speed was adjusted to give an estimate of above-canopy wind (10m) using a wind reduction factor of 1.5 found using equation 6.7 for the area of relatively open canopy (10m tree spacing) where it was placed; recorded weather conditions averaged over the time of each burn are given in table 7.19. The wind reduction factor in the plots was 3.3. Table 7.19. Weather and slope conditions recorded for the experimental burns Date Time Temperature Relative Wind humidity speed Wind speed -1, th o Drought FFDI Slope factor -1 (km.h (km.h , 1.5m) 10m) 7 Jan 05 1130 14 C 36% 3 4.5 7 6 15o 25th Feb 05 1200 23oC 34% 8 12.0 10 19 22o 13thApril 05 1310 20.8 oC 37% 2.3 3.5 10 9 7o 13thApril 05 1400 20.8 oC 37% 2.3 3.5 10 9 15o 28th April 05 1330 20.3 oC 33% 5.1 7.7 10 21 20o 28th April 05 1415 19.8 oC 33% 7.9 11.9 10 28 12o 7.323 Fuels Surface fuel loads were measured from the area immediately adjacent to each plot before it was cleared to form a fire break. Two quadrat samples of 0.35 by 0.35m were taken randomly from this area, giving a total sample area of 0.25m2 or 1.2% of the study area. Samples were dried in a furnace at 105oC for 24hours then weighed and the surface fuel load calculated. Other parameters used in the model of Gould et al were measured on site as described in 276 Gould et al (2007b), except for elevated fuel height, which was estimated from the average height of the shrubs multiplied by their percent cover. Near surface fuels in all blocks were composed of suspended leaf litter, with only very sparse, scattered grasses. All parameters for the empirical models are given in tables 7.20 to 7.25. Parameters for the live fuels were measured directly before the burns, through separate field trips to characterise species and from video footage as indicated in the following tables. Standard values are used for Daviesia mimosoides as measured from across the area, except that the density of D. mimosoides were measured separately for most plots as these varied. This was not measured for the burns carried out on the 25th of February or at 14:15 on the 28th of April. In the first case, the width and plant spacing used was the taken from the adjacent plot burnt on April 13th at 13:10; in the second case values were taken from the plot adjacent to that burn, burnt on April 28th at 13:30. Values used for E. pauciflora crown dimensions were estimated from field observations; however the density of trees and the leaf dimensions were measured as indicated in the relevant tables. All fuel parameters for the FFM are shown in tables 7.26 to 7.28. Table 7.20. Fuel parameters used in the empirical models for 7th January burn Model McArthur Surface Near surface Near surface Elevated height (cm) height (m) 15.8t.Ha-1 - - - 15.8t.Ha-1 - - - 3.5 2 10 0.12 (1962) McArthur (1967) Gould et al (2007a) Table 7.21. Fuel parameters used in the empirical models for 25th February burn Model McArthur Surface 11t.Ha -1 Near surface - Near surface Elevated height (cm) height (m) - - 277 (1962) McArthur 11t.Ha-1 - - - 3 2 10 0.12 (1967) Gould et al (2007a) Table 7.22. Fuel parameters used in the empirical models for 13th April 13:10 burn Model McArthur Surface Near surface Near surface Elevated height (cm) height (m) -1 - - - 9.2t.Ha-1 - - - 2 2 10 0.15 9.2t.Ha (1962) McArthur (1967) Gould et al (2007a) Table 7.23. Fuel parameters used in the empirical models for 13th April 14:00 burn Model McArthur Surface Near surface Near surface Elevated height (cm) height (m) -1 - - - 9.2t.Ha-1 - - - 2 2 10 0.12 9.2t.Ha (1962) McArthur (1967) Gould et al (2007a) Table 7.24. Fuel parameters used in the empirical models for 28th April 13:30 burn Model McArthur Surface Near surface Near surface Elevated height (cm) height (m) -1 - - - 9.0t.Ha-1 - - - 2 2 10 0.25 9.0t.Ha (1962) McArthur (1967) Gould et al (2007a) 278 Table 7.25. Fuel parameters used in the empirical models for 28th April 14:15 burn Model McArthur Surface Near surface Near surface Elevated height (cm) height (m) 9.0t.Ha-1 - - - 9.0t.Ha-1 - - - 2 2 10 0.25 (1962) McArthur (1967) Gould et al (2007a) Table 7.26 Details for experimental burn sites Parameter Value Details DFMC (%ODW) 7/1/5 – 10.0% Measured on-site using a Wiltronics moisture meter 25/2/5 – 10.2% 13/4/5 (13:10) – 7.4% 13/4/5 (14:00) – 8.7% 28/4/5 – 9.5% Near surface None Observation – very isolated grass tussocks and forbs only species Near surface plant - spacing Elevated species Daviesia mimosoides Monoculture within blocks Elevated plant 7/1/5 – 1.35, 0.50m 20m quadrat results, calculated using equation 7.3. 2nd measure spacing 25/2/5 – 1.07m for 7/1/5 from 6 on-ground nearest-neighbour measurements 13/4/5 (13:10) – 0.97m along central axis of burn, testing only distances within clumps. 13/4/5 (14:00) – 1.07m 28/4/5 – 0.74m Canopy species Canopy plant E. pauciflora 89%, E. Only E. pauciflora present within burn plots; data determined dalrympleana 11% from 250m2 quadrat 5.1m 19 trees in 250m2, calculated using equation 7.3 spacing Soil texture Silty clay 279 Table 7.27 Daviesia mimosoides details Parameter Value Details m (%ODW) 7/1/5 – 96% Measured on site using Wiltronics TH Moisture Meter 25/2/5 – 123.2% 13/4/5 – 113.2% 28/4/5 – 101.8% P (oC) 260oC Laboratory measurement (Table 4.2) % Dead 0% Observation Leaf form Flat Observation d (mm) 0.38 Mean of 153 samples. Accurate to 0.05mm, 99.5%confidence Wl (mm) 10.8 Mean of 25 samples. Accurate to 25%, 99.5%confidence Ll (mm) 58.2 Mean of 25 samples. Accurate to 25%, 99.5%confidence Sl (cm) 1.26 Mean of 178 leaves Os 2.2 Mean of 5 branches Sb (cm) 12.4 Mean of 20 samples. Accurate to 25%, 99.5%confidence Wb (cm) 26.4 Mean of 12 samples. Accurate to 25%, 99.5%confidence He (cm) 33 (7/1/5) 7/1/5 measured from photograph, mean of 8 samples, 25% error, 90.0% confidence. Ht (cm) 54 (7/1/5) 7/1/5 measured from photograph, mean of 9 samples, 25% 36 error, 95.0% confidence. Others estimated from video, mean of 8 samples Hc (cm) 12 (7/1/5) 7/1/5 measured from photograph, mean of 7 samples, 25% error, 99.5% confidence. Hp (cm) 71 (7/1/5) 7/1/5 measured from photograph, mean of 6 samples, 25% 46 error, 99.5% confidence. Others estimated from video, mean of 5 samples w (cm) 55 7/1/5 measured on ground, mean of 17 samples, 25% error, 99.5% confidence. Others estimated from 13/4/5 video, mean of 6 samples 280 Table 7.28 E. pauciflora details Parameter Value m (%ODW) o Details Modelled using equation 6.16 P ( C) 220 Laboratory measurement (Table 4.2) % Dead 0 Observation Leaf form Flat Observation d (mm) 0.49 Mean of 160 samples. Accurate to 0.05mm, 99.5%confidence Wl (mm) 26.67 Mean of 21 samples. Accurate to 25%, 99.5%confidence Ll (mm) 67.9 Mean of 11 samples. Accurate to 25%, 99.5%confidence Sl (cm) 1.06 Mean of 18 leaves Os 5 Mean of 5 branches Sb (cm) 84 Taken from E. Niphophila table 7.7 Wb (cm) 121 Taken from E. Niphophila table 7.7 He (cm) 400 Estimate Ht (cm) 700 Estimate Hc (cm) 400 Estimate Hp (cm) 900 Estimate w (cm) 400 Estimate 7.324 Modelled fire behaviour The observed and modelled fire behaviour for each plot is shown in tables 7.29 to 7.34. For most of the time, the observed fire spread was in the surface fuels, with individual shrubs igniting but not contributing to the rate of spread because equation 3.3 was not satisfied and connection did not occur. McArthur (1962) predicted rates of spread with considerably greater accuracy and consistency than either Gould et al or McArthur (1967), both of which displayed high levels of error. McArthur (1967) displayed both poor accuracy and very high variability in both rate of spread and flame height predictions, but while rate of spread predictions by Gould et al showed poor accuracy and very similar levels of variability, the flame height predictions by this model were nearly as accurate as those of McArthur (1962) and the variability in accuracy was even less. 281 Rate of spread and flame height predictions were 1.3 and 1.7 times better than those of McArthur (1962), 8.2 and 13.0 times better than those of McArthur (1967) and 5.2 and 2.4 times better than those of Gould et al, respectively. The standard deviation for rate of spread errors was slightly less than the FFM for McArthur (1962), with 69% compared to 81%. In all other cases the FFM had more consistent results. FLAME HEIGHT Mean absolute error RATE OF SPREAD Mean absolute error 1200% 1400% 1000% 1200% 1000% 800% 792% 600% Error Error 651% 416% 800% 600% 400% 400% 104% 200% 146% 79% 200% 0% 106% 61% 0% McArthur 1962 McArthur 1967 Gould et al 2007 Zylstra 2011 McArthur 1962 McArthur 1967 Gould et al 2007 Zylstra 2011 Figure 7.8. Mean absolute error of each model for the experimental burns. Table 7.29. Modelled and observed fire behaviour for 7/1/5 Model Mean ROS (km/h) % Error Mean Flame height (m) % Error McArthur (1962) 0.01 71% 0.16 394% McArthur (1967) 0.24 1096% 4.90 513% Gould et al (2007a) 0.06 185% 0.39 105% FFM 0.07 226% 1.13 41% Observed 0.02 0.8 Table 7.30. Modelled and observed fire behaviour for 25/2/5 Model Mean ROS (km/h) % Error Mean Flame height (m) % Error McArthur (1962) 0.13 150% 0.68 29% McArthur (1967) 0.66 1229% 9.28 1651% Gould et al (2007a) 0.67 1244% 2.31 336% FFM 0.11 112% 0.47 13% Observed 0.05 0.5 282 Table 7.31. Modelled and observed fire behaviour for 13/4/5 13:10 Model Mean ROS (km/h) % Error Mean Flame height (m) % Error McArthur (1962) 0.02 209% 0.34 102% McArthur (1967) 0.13 123% 1.95 186% Gould et al (2007a) 0.033 82% 0.27 153% FFM 0.04 58% 0.91 33% Observed 0.06 0.7 Table 7.32. Modelled and observed fire behaviour for 13/4/5 14:00 Model Mean ROS (km/h) % Error Mean Flame height (m) % Error McArthur (1962) 0.03 84% 0.34 13% McArthur (1967) 0.23 276% 3.24 978% Gould et al (2007a) 0.058 7% 0.39 30% FFM 0.07 6% 0.86 187% Observed 0.06 0.3 Table 7.233. Modelled and observed fire behaviour for 28/4/5, 13:30 Model Mean ROS (km/h) % Error Mean Flame height (m) % Error McArthur (1962) 0.11 46% 0.70 58% McArthur (1967) 0.63 725% 8.31 656% Gould et al (2007a) 0.40 429% 1.73 57% FFM 0.09 21% 0.69 60% Observed 0.08 1.1 Table 7.34. Modelled and observed fire behaviour for 28/4/5, 14:15 Model Mean ROS (km/h) % Error Mean Flame height (m) % Error McArthur (1962) 0.08 10% 0.82 34% McArthur (1967) 0.40 469% 5.34 769% Gould et al (2007a) 0.44 521% 1.83 198% FFM 0.11 53% 0.47 30% Observed 0.07 0.6 283 The observed fire behaviour could be described in categories or stages of growth. These are: Stage one: Fire burns only the surface and suspended litter but does not ignite the shrubs. Flame heights are low and the rate of spread is slow and predictable. Stage two: Fire ignites the shrubs but does not spread between them. Flame heights are larger but short lived. Flame lengths are larger but flames are frequently upright and will not usually breach fire breaks. Rates of spread are slow but subject to sudden change if conditions allow a shift to stage three. Stage three: Fire spreads from one shrub to the next. Flame heights are consistently greater, flame lengths are much larger and the low angle of the flames allows them to breach fire breaks. Rates of spread are much larger (e.g. three to four times faster in these fires) and continue to increase as the length of pulses increases (spread acceleration). Stage three only occurs when the combination of wind and slope with plant height and spacing facilitates a low enough flame angle for connection to occur. The observed fire behaviour supports the claim of chapter three that fire spread increases in threshold changes interspersed between more steady periods of growth. The FFM was able to effectively capture most of these threshold changes for the experimental fires and thereby performed with a smaller mean error than the empirical models (Figure 7.8). 7.33 Guthega, 26th January 2003 On the 26th January 2003, strong winds producing very high fire danger conditions caused fire to escape from a control line to the north west of Guthega Village in Kosciuszko National Park, NSW. The fire front crossed the Snowy River then climbed the north western slopes of Mount Blue Cow before largely self-extinguishing along the summit ridge. 7.331 Fire behaviour A Line scan (RFS 2003) at 1530 hrs EDT was taken showing the fire run completed, however rate of spread cannot be estimated as there are no other air or ground observations available from which to estimate a starting time. Photographs taken by fire crews (Gibson 2003, Figure 284 7.9) display flames of approximately 1.5m height and other anecdotal reports describe 10m flames (Border Mail 2003). A survey of the site in 2004 revealed that along the centre of the main fire run where the fuels were later sampled, all fine branches in some trees were removed by the fire to the top of the canopy. The direct impact of flame was distinguished from leaf loss due to flame scorch by the removal of actual branch material as opposed to the loss of leaves alone. This allowed an estimation of flame height for 10 trees in the study area based on the highest point in the tree where material had been burnt from the fine branches. The studies indicated point values of 3.5 - 6m corresponding with the top branches of the trees. In some cases these top branches had been consumed to a branch tip diameter of 4mm, suggesting flames which may have been considerably taller. These indicate maximum potential heights occurring at points where fuels permitted; occurring in all cases where the understorey was dominated by Orites lanceolata. As there was no recorded rate of spread for the fire and the flame heights were highly variable, the models will be tested against the range of flame heights (1.5 – 10m) and the error calculated from the flame height closest to that estimated by the model. 7.332 Weather The closest weather station to the site was located on the top of the range at Thredbo 15km to the south. Weather conditions from this site are limited as fire impacted on the station so that the last weather recording for the day was at 1300hrs (table 7.35), approximately 2.5 hours earlier. As this time was close to the commencement of the run, most data from this recording will be used and fire behaviour assessed for the lower parts of the run. The photographed fire behaviour (figure 7.9)adds to this data as it shows little to no bending of the upper tree branches due to wind, giving a Beaufort scale of 3-4 and corresponding with a minimum 12km/h for the time at which the photograph was taken. It is possible that wind speeds had lessened after the 1300hr record, so a range of wind speeds will also be used and the most accurate prediction from each model used. 285 Table 7.35. Slope and weather conditions recorded for 1300hrs at Thredbo upper weather station Temperature Relative Wind humidity speed Drought FFDI Slope 11-29 17o factor -1 (km.h ) o 21.4 C 27.0% 12 - 55 8 7.333 Fuels An area in the lower part of the fire run was surveyed post-fire and measurements were taken of the plant skeletons to find their dimensions and spacing. To find the species composition and characteristics, vegetation was also surveyed in the same forest type, altitude and aspect approximately 1km away within an area protected from fire by the ground crews. Surface fuel loads are taken from the equilibrium value for the community (Zylstra 2007). Survey results and parameters for the four models are given in tables 7.36 to 7.41. Table 7.36. Fuel parameters used in the empirical models for 26th January 2003 Model McArthur Surface Near surface Near surface Elevated height height -1 - - - 12.1t.Ha-1 - - - 3 4 15 0.20m 12.1t.Ha (1962) McArthur (1967) Gould et al (2007a) Table 7.37. Details for Guthega site Parameter Value Details DFMC 7.2 Modelled from Gould et al (2007b) using Tolhurst (2010) Poa costiniana Clear dominant under shrub cover with occasional minor forbs Closed Observation Olearia phlogopappa Calculated from 50m line transect (%ODW) Near surface species Near surface plant spacing Elevated 286 species 80%, Orites lanceolata 20% Elevated plant 1.68m Average of 29 nearest neighbour measurements Canopy species E. niphophila Pure stand Canopy plant 6.2m 66 trees counted in a 1250m2 quadrat Clay loam Typical texture for granite soils (e.g. Robinson et al 2003, spacing spacing Soil texture Ringrose and Nielsen 2005) Table 7.38. Poa costiniana details Parameter Value Details m (%ODW) 80 Estimate only P ( C) 260 Standard value for volatile oil content of one (Table 4.2) % Dead 50 Estimated level of curing representative of this altitude for drought o conditions in mid summer Leaf form Round Observation d (mm) 0.25 Estimate Wl (mm) 0.25 Estimate Ll (mm) 110 Estimate Sl (cm) 0.03 Standard value for fine grasses Os 1 Observation Sb (cm) 0 Observation Wb (cm) 20 Visual estimate/ actual measure makes little difference in closed cover He (cm) 0 Observation Ht (cm) 11 Mean of 13 measurements, accurate to 25%, 99.5% confidence Hc (cm) 0 Observation Hp (cm) 11 Mean of 13 measurements, accurate to 25%, 99.5% confidence w (cm) 20 Visual estimate/ actual measure makes little difference in closed cover 287 Table 7.39. Olearia phlogopappa var. flavescens details Parameter Value Details m (%ODW) 70 Modelled using equation 6.19 P (oC) 260 Standard value for volatile oil content of one (Table 4.2) % Dead 20 Estimate Leaf form Flat Observation d (mm) 0.32 Mean of 10 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 4.5 Mean of 15 measurements, accurate to 25%, 99.5% confidence Ll (mm) 14.9 Mean of 15 measurements, accurate to 25%, 99.5% confidence Sl (cm) 0.09 Counted for 9 leaves Os 4 Mean of 5 measurements Sb (cm) 12.9 Mean of 15 measurements, accurate to 25%, 95.0% confidence Wb (cm) 26.8 Mean of 12 measurements, accurate to 25%, 90.0% confidence He (cm) 22.6 Mean of 10 measurements, accurate to 25%, 90.0% confidence Ht (cm) 41.7 Mean of 10 measurements, accurate to 25%, 97.5% confidence Hc (cm) 26.2 Mean of 10 measurements, accurate to 25%, 99.5% confidence Hp (cm) 63.1 Mean of 10 measurements, accurate to 25%, 99.5% confidence w (cm) 57.2 Mean of 15 measurements, not significant Table 7.40. Orites lanceolata details Parameter Value Details m (%ODW) 80 Estimate P ( C) 260 Standard value for volatile oil content of one (Table 4.2) % Dead 0 Observation – dead leaves not retained by plant Leaf form Flat Observation d (mm) 0.49 Mean of 10 samples, accurate to 0.05mm, 99.5% significance Wl (mm) 7.3 Mean of 10 samples, accurate to 25%, 99.5% significance Ll (mm) 21.1 Mean of 10 samples, accurate to 25%, 99.5% significance Sl (cm) 0.29 Mean of 31 leaves Os 4.8 Mean of 5 measurements Sb (cm) 0 Observation Wb (cm) 137 Mean of plant width and canopy height He (cm) 67 Mean of 10 measurements, not significant Ht (cm) 128 Mean of 10 measurements, not significant o 288 Hc (cm) 43 Mean of 3 measurements, not significant Hp (cm) 120 Mean of 3 measurements, not significant w (cm) 198 Mean of 6 measurements, accurate to 25%, 99.5% significance Table 7.41. E. niphophila details Parameter Value Details m (%ODW) 126 Equation 6.16 P (oC) 220 Measured value (table 4.2) % Dead 0 Observation – dead leaves rarely retained on trees Leaf form Flat Observation d (mm) 0.60 Mean of 10 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 18.1 Mean of 10 measurements, accurate to 25%, 99.5% confidence Ll (mm) 67.9 Mean of 11 measurements, accurate to 25%, 99.5% confidence Sl (cm) 1.06 Mean of 18 leaves Os 5 Estimate from observations Sb (cm) 84 Mean of 10 measurements, not significant Wb (cm) 121 Mean of 11 measurements, accurate to 25%, 90.0% confidence He (cm) 380 Mean of 10 measurements, accurate to 25%, 95.0% confidence Ht (cm) 554 Mean of 10 measurements, accurate to 25%, 95.0% confidence Hc (cm) 398 Mean of 10 measurements, accurate to 25%, 90.0% confidence Hp (cm) 697 Mean of 10 measurements, accurate to 25%, 99.0% confidence w (cm) 366 Mean of 10 measurements, accurate to 25%, 95.0% confidence 7.334 Modelled fire behaviour All predictions except that of the FFM were within the specified range if the lower wind speed was used (table 7.42). The higher wind speeds caused all empirical models to predict well beyond the range, and the FFM to produce lower flame heights. The FFM predicted flame heights within the specified range when wind speeds were less than 10km/h or when the understorey was dominated by Orites lanceolata. Where this was the case, crown fire was expected to produce flame heights approaching 10m for wind speeds of up to 20km/h (figure 7.10). 289 Table 7.42. Model results and error for each model Model Mean flame % ROS (km/h) height (m) Error McArthur (1962) 8.0 0% 0.10-0.90 McArthur (1967) 7.4 0% 0.50-1.37 Gould et al (2007a) 5.5 0% 2.07 – 11.57 FFM 1.02 47% 0.58-1.57 Observed 1.5 – 10 12 Flame height (m) 10 8 6 4 2 0 0 20 40 60 80 100 Wind speed (km/h) Figure 7.9. Observed fire behaviour and wind at Guthega th Figure 7.10. Flame height and rate of spread as on 26 January 2003. Flame heights are approximately functions of wind speed for the given conditions 1.5m tall, the smoke column has aligned with the hillside at Guthega under different fuel structures. The y- but there is no evidence of tree branches bending. Used by axis gives flame height (m), and the x-axis shows permission, NSWFB wind speed from 0 to 100km/h. The black line indicates fire behaviour for the average fuel conditions, the grey area beneath the line indicates an understorey free of Orites and the grey area above the line has an understorey dominated by Orites. 290 Figure 7.11. FFM modelling of flame dimensions under a 12km/h wind scenario. 7.34 Tooma Dam, 13th October 2006 The Tooma Dam fire was ignited from an abandoned campfire on October 9th 2006, following a dry winter. The area that the fire was burning had been burnt three years and nine months earlier, so the combination of young fuels, high altitude cool spring conditions and a low KBDI suggested that fire behaviour would be benign. 7.341 Fire behaviour After escaping containment on the 12th October, the fire burnt at a low intensity through the Alpine Ash forest under high fire danger conditions until early afternoon on the 13th of October. The intensity remained low until the fire reached midway up the gully running up the northwest slopes of Mount Toolong, where the forest transitioned toward subalpine Snowgum E. niphophila. At approximately 1500hrs, an air observer mapped the edge of the fire as shown in Figure 7.12. By the time the observer had returned to the front an hour later, the fire had advanced over the top of Mt Toolong and extended in a narrow arm to the eastsoutheast, well beyond the fallback containment lines that were being prepared on the Round Mountain fire trail (Figure 7.13). The total spread in the hour was 6.6km, although the unusual shape of the front consisting of a broad front for approximately two thirds of this distance and a narrow finger extending from this suggests some unusual external conditions, perhaps related to wind channelling through the gully on the north-western slopes of Mt 291 Toolong, and spotting. The large spot fire to the north of the narrow run suggests that this was the characteristic spotting distance, which in turn suggests that the narrow burnt area may not be a contiguous area but a narrow ember shower resulting from the fire run up the north-west gully of Mt Toolong. If this is the case, the actual spread of the fire front can be measured up to the end of the main burnt area before the narrow neck. This equates to a mean rate of spread of approximately 3.5km/h. The spread in the narrow finger beyond that point was predominantly through grasslands (Pers. Comms. D. Corcoran, Air Attack Supervisor Feb. 2010). Flame height in the forest was not measured at the time, however post-fire observation of burnt Snowgum forest shows significant consumption of both live and dead wood to the top of the trees (Figure 7.14), indicating that flames were considerably taller than the 2003 regrowth vegetation (1.5m). Estimates based on minimum branch tip diameter from the taller deadwood are not valid as dead wood continues to burn after the front has passed, so no firm estimates can be made except to say that the flames were taller than the 1.5m cut-off point for direct attack. The smoke column over the narrow finger of fire was described as black in the lower parts, bent very low to the ground and light grey on top with a smooth rippled upper surface (Pers. Comms. D. Corcoran, Air Attack Supervisor Feb. 2010). Black smoke is an indicator of incomplete combustion typical of crown fire events; however the low angle of the column suggests that the flame height cannot have been of the same scale as a crown fire in a mature forest. The isolated spotfire to the north of the main run appears to have started at least 3km ahead of the front. Solving equation 7.9 to find the necessary flame height to achieve such a distance gives an estimated flame height of 12m. Gould et al (2007a) specify that long distance spotting is produced by the tallest flames rather than the average, so this value represents an upper limit, most likely achieved in the steep mid-upper slope part of the climb up Mt Toolong. A reasonable range for estimations then is that flames in the forest were over 1.5, with maximums of approximately 12m in the mid-upper slope climb. Flame heights in the grasslands were 1-2m high with a very shallow flame angle, and were very fast spreading (Pers. Comms. D. Corcoran, Air Attack Supervisor Feb. 2010). 292 Figure 7.12. The Tooma Dam fire front at 1500hrs Figure 7.13. The Tooma Dam fire front at 1600hrs. The narrow finger of fire appears to be one or more spot fires that have since linked up with the main front coming from behind. 293 The behaviour of the Tooma Dam fire is important from an incident management point of view, as the increase in fire spread was unexpected and placed the operators preparing the fallback lines at considerable risk as it crossed the Round Mountain Fire Trail. Incident planners were aware that the weather on the 13th October was very similar to the preceding day, although the wind speed was slightly lower and the temperatures slightly higher. They were also aware that the area had been burnt less than four years previously, and that the fire was leaving the higher surface fuel loads of the Alpine Ash forest for the lower fuel loads of the Snowgums. 7.342 Weather and terrain The closest weather station to the site was Figure 7.14. Burn severity of the Tooma Dam fire. The green vegetation is two year-old regrowth from the fire, the taller dead sticks are the regrowth from the 2003 fire that was burnt in 2006, the large dead wood is regrowth from the 1988 Round Mtn/Ogilvie’s Ck fire and the large trees in the background are the original vegetation. at Cabramurra, which is 16km to the NNE at approximately the same altitude and in similar terrain. The weather at 1500hrs produced an FFDI of 19, with wind speeds averaging 35km/h from the WNW and gusting to 57km/h. The influence of the gully on the NW slopes of Mt Toolong however should be factored in as the shape of the fire front strongly suggests forced channelling. In the absence of the capacity to model the magnitude of the channelling, an estimate can be gained from empirical evidence elsewhere; for example radiosonde measurements in the Dead Sea area have demonstrated that wind speeds in confined mountain valleys were on average 138% higher than those in surrounding areas (Shafir et al 2008). Based on this Figure, wind speeds and slopes will be calculated for the length of the main fire run, with winds increased by 138% on the north western face of Mt Toolong. The fire run is divided into natural divisions based on changes in slope, and a wind reduction factor of 2 was used for calculations with McArthur (1962), derived using equations 6.7 and 6.8. 294 Table 7.43. Slope, and weather conditions recorded for 1500hrs at Cabramurra with wind speed adjusted for forced channelling. Segment Distance Temperature (km) o Relative 6hr Mean Wind speed DF FFDI Slope -1 humidity RH (km.h ) Foot-slope 0.8 23.1 C 24% 30.5% 48 7 26 7o Mid-slope 0.4 23.1oC 24% 30.5% 48 7 26 14o Upper slope 0.6 23.1oC 24% 30.5% 48 7 26 6o Lee side 0.7 23.1oC 24% 30.5% 35 7 19 -9o Toolong Range 1 23.1oC 24% 30.5% 35 7 19 8o 7.343 Fuels The fireground fuels were observed and measured both during the course of the fire and during subsequent visits. Grass fuels were unusually high and cured for the time of the year due to the poor snow cover during the previous winter. Winter snowfall at this altitude (14001600m) typically compresses the grasses, causing a large amount of stem death but flattening this into a dense thatch between tussocks. The winter of 2006 experienced uncharacteristically low snowfalls, so that although much stem death had occurred, the grass had not been flattened fully and early thaw had allowed quick recovery. In addition, the fire in 2003 had encouraged the growth of dense, tall flowering stems so that the amount of dead material was considerably higher than usual. Live fuels were measured both at the mid-slopes of Mt Toolong and at the point where the fire first crossed the Round Mountain Fire Trail. The Snowgum community varies between these points from a montane formation with tall parent trees killed in the 2003 fire interspersed with occasional Alpine Ash (E. delegatensis) to a subalpine community consisting of dense Snowgum regrowth broken by areas of grassland along drainage lines and higher flat areas. The starting point of the run represents the transition from montane to a subalpine community, so the fuels were represented by two groups. The foot and mid-slopes of Mt Toolong were represented by Snowgum regrowth with a shrub understorey, and those of the forested areas in the subalpine for the remainder of the fire run were represented by Snowgum regrowth with a grassy understorey. Surface fuel loads were visual estimates taken at the time of the fire. As the 2003 fire had caused complete stem death in the trees, surface fuel accumulation proceeded at a much slower rate than the typical rate for Snowgum forest 295 and the litter on the ground consisted only of sparse, scattered leaves. Survey results and parameters for the four models are given in tables 7.44 to 7.49. Table 7.44. Fuel parameters used in the empirical models for 13th October 2006 Model McArthur Surface Near surface Near surface Elevated height (cm) height (m) -1 - - - 3t.Ha-1 - - - 1 4 11 0.97, 0 on lee 3t.Ha (1962) McArthur (1967) Gould et al (2007a) side and Toolong Range Table 7.45. Details for the montane forest Parameter Value Details DFMC 6.6 Modelled from Gould et al (2007b) using Tolhurst (2010) Poa costiniana Clear dominant under shrub cover with occasional minor forbs Closed Observation Olearia phlogopappa Observed clear dominant 1.02m Average of 11 nearest neighbour measurements. Accurate to 25%, (%ODW) Near surface species Near surface plant spacing Elevated species Elevated plant spacing 95.0% significance Canopy species E. niphophila Observed clear dominant Canopy plant 1.1m Two quadrats measured. Coppice from mature trees in 225m2 – 16 trees; New saplings in 60m2 – 102 trees spacing Soil texture Clay loam Typical texture for granite soils (e.g. Robinson et al 2003, Ringrose and Nielsen 2005) 296 Table 7.46. Details for the subalpine forest Parameter Value Details DFMC 6.6 Modelled from Gould et al (2007b) using Tolhurst (2010) Poa costiniana Clear dominant under shrub cover with occasional minor forbs Closed Observation Canopy species E. niphophila Observed clear dominant Canopy plant 1.1m Two quadrats measured. Coppice from mature trees in 225m2 – 16 (%ODW) Near surface species Near surface plant spacing trees; New saplings in 60m2 – 102 trees spacing Soil texture Clay loam Typical texture for granite soils (e.g. Robinson et al 2003, Ringrose and Nielsen 2005) Table 7.47. Poa costiniana details Parameter Value Details m (%ODW) 80 Estimate only P ( C) 260 Standard value for volatile oil content of one (Table 4.2) % Dead 70 Ground observation during the fire Leaf form Round Observation d (mm) 0.25 Estimate Wl (mm) 0.25 Estimate Ll (mm) 110 Estimate Sl (cm) 0.03 Standard value for fine grasses Os 1 Observation Sb (cm) 0 Observation Wb (cm) 20 Visual estimate/ actual measure makes little difference in closed cover He (cm) 0 Observation Ht (cm) 11 Mean of 13 measurements, accurate to 25%, 99.5% confidence Hc (cm) 0 Observation Hp (cm) 11 Mean of 13 measurements, accurate to 25%, 99.5% confidence w (cm) 20 Visual estimate/ actual measure makes little difference in closed cover o 297 Table 7.48. Olearia phlogopappa var. flavescens details Parameter Value Details m (%ODW) 100 Modelled using equation 6.19 P (oC) 260 Standard value for volatile oil content of one (Table 4.2) % Dead 20 Estimate based on common condition Leaf form Flat Observation d (mm) 0.32 Mean of 10 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 4.5 Mean of 15 measurements, accurate to 25%, 99.5% confidence Ll (mm) 14.9 Mean of 15 measurements, accurate to 25%, 99.5% confidence Sl (cm) 0.09 Counted for 9 leaves Os 4 Mean of 5 measurements Sb (cm) 12.9 Mean of 15 measurements, accurate to 25%, 95.0% confidence Wb (cm) 26.8 Mean of 12 measurements, accurate to 25%, 90.0% confidence He (cm) 24 Mean of 4 measurements, not significant Ht (cm) 45 Mean of 4 measurements, not significant Hc (cm) 15 Single measure Hp (cm) 95 Single measure w (cm) 75 Mean of 2 measurements Table 7.49. E. niphophila details Parameter m (%ODW) o P ( C) % Dead Value Details 129 Equation 6.15 220 Measured value (table 4.2) 0 Observation – dead leaves rarely retained on trees Leaf form Flat Observation d (mm) 0.54 Mean of 10 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 20.5 Mean of 10 measurements, accurate to 25%, 99.5% confidence Ll (mm) 65.2 Mean of 10 measurements, accurate to 25%, 99.5% confidence Sl (cm) 2.3 Mean of 23 leaves Os 5.5 Estimate from observations Sb (cm) 0 Observation Wb (cm) 54 Mean of plant height and width He (cm) 116 Mean of 10 measurements, accurate to 25%, 99.5% confidence Ht (cm) 151 Mean of 10 measurements, accurate to 25%, 99.5% confidence 298 Hc (cm) 85 Mean of 10 measurements, accurate to 25%, 99.5% confidence Hp (cm) 142 Mean of 10 measurements, accurate to 25%, 99.5% confidence w (cm) 101 Mean of 12 measurements, accurate to 25%, 97.5% confidence 7.344 Modelled fire behaviour The Tooma Dam fire illustrates an important point due to its young fuels and comparatively mild conditions (fire danger classified as “high”). The surface fuel load for the Tooma Dam fire was exceptionally low as the regenerating canopy had not yet matured sufficiently to begin shedding many leaves, bark or sticks, so the two models based on surface fuel load (McArthur 1962, 1967) predicted very low rates of spread (table 7.50). All models predicted flame heights within the range except for Gould et al, which was slightly over the probable maximum (table 7.51). Gould et al and the FFM produced identical rate of spread predictions, both slightly under the expected rate. The primary reason for the underprediction by the FFM was the slow lee-slope rate of spread. The required maximum flame height to account for spotting can be explained by small pockets of taller vegetation permitting greater flame heights. Table 7.50. Observed and modelled rates of spread (km/h) for the Tooma Dam fire run. Gould et al FFM 0.149 2.94 3.05 0.494 0.241 4.77 4.34 0.6 0.284 0.139 2.75 2.83 Lee slope 0.7 0.037 0.036 0.71 0.04 Toolong range 1.0 0.120 0.118 2.29 3.10 0.22 0.12 2.57 2.57 1491% 2817% 36% 36% Segment Length McArthur McArthur (km) (1962) (1967) Foot-slope 0.8 0.305 Mid-slope 0.4 Upper slope Average Observed/error 3.5 Table 7.51. Observed and modelled flame heights (m) for the Tooma Dam fire run. Segment McArthur McArthur Gould et Length (km) (1962) (1967) al Foot-slope 0.8 4.30 0.7 11.56 2.90 Mid-slope 0.4 4.30 1.9 16.40 3.13 Segment FFM 299 Upper slope 0.6 4.30 0.5 11.00 2.58 Lee slope 0.7 1.62 -0.8 2.22 0.89 Toolong 1.0 1.62 0.3 5.19 2.11 Average 3.00 0.36 8.33 2.17 Maximum 4.30 1.9 16.4 2.98 0% 0% 37% 0% range Observed 1.5 - 12 range/ error Figures 7.15 and 7.16 are outputs from the FFM showing the modelled changes in rate of spread and flame height with increases in wind speed for the montane and subalpine communities at set slopes. In both cases, fire spread increases suddenly at a threshold representing flame connection where flame angle allows propagation from one plant to the next. In both cases, the threshold wind speed is only a little lower than that experienced at the site, suggesting that the observed rate of spread only occurred because the wind had reached this velocity at this time of the day. With this knowledge, it appears that the sudden and unexpected change in fire behaviour occurred for the following reasons: 1) The fire front burnt from the Alpine Ash forest into the montane Snowgum community, where the lack of a tall forest canopy and the forced channelling effect of the Mt Toolong gully exposed the front to higher wind speeds. 2) The post-fire regrowth of the Snowgum communities produced a fuel array that could support a very fast spreading fire in these conditions. 300 Figure 7.15. Modelled rate of spread and flame height for the Montane Snowgum community in the mid-slopes of Mt Toolong. The x-axis is wind speed from still to 100km/h, and the y-axis shows rate of spread in km/h and flame height in m. Note the sudden increase in flame height for winds above 20km/h. Figure 7.16. Modelled rate of spread and flame height for the Subalpine Snowgum community on the Toolong Range. The x-axis is wind speed from still to 100km/h, and the y-axis shows rate of spread in km/h and flame height in m. Note the sudden increase in rate of spread for wind speeds above 16km/h. Analysis of the model shows that fire spread as a passive crown fire, with spread in the grasses dominating the overall fire velocity and igniting the canopy as it progressed. 7.35 Kilmore East Mountain Ash, 7th February 2009 7.351 Observed fire behaviour It is difficult to reconstruct the fire behaviour from the 7th February 2009 as the available sources provide little information on this. The four sources available at the time of writing are the report by the Bushfire Cooperative Research Centre on fire behaviour (McCaw et al 2009), reports by Taylor (2009) and Tolhurst (2009) who worked as Fire Behaviour Analyst during the incident, and the submissions to the Royal Commission (VBRC 2009a,b, accessed 10/2/10). The information they provide on rates of spread and flame dimensions is summarised below. The report on fire behaviour by the Bushfire CRC (McCaw et al 2009) provides a map showing the direction of fire spread and states that the fire covered 50km in 4.75 hrs, giving a rate of spread of 10.5km/h. No explanation is given as to the details of this spread and where it occurred, and no indications are given of flame heights. 301 Taylor (2009) provides more specific information about fire behaviour, making use of material submitted to the Royal Commission. Key details given are: • Fire ignition occurred at 11:49 on Saunders Road, Kilmore East. • Fire spread through grazed pasture and a pine plantation to reach the Hume Highway at Heathcote Junction at 13:58 (10.6km in 2 hours and 9 minutes; ROS on average 4.9km/h). • A generalised statement is made that the fire travelled a further 6km by 17:00, indicating a mean ROS of 2km/h. • Reached the base of Mt Disappointment at 15:30 travelling at an estimated 8km/h. VBRC (2009a) makes it clear that the 8km/h rate of spread occurred during the climb up Mt Disappointment, although it is not clear how this number was derived. • The fire progressed more quickly along the southern escarpment of Mount Disappointment than it did through the Wallaby Creek area on the plateau, also known as the “Big Ash” forest (Ashton 2000). • Following a south-west wind change, the fire front reached Flowerdale at 19:40 – an indirect distance of 10 to 15km in 4hrs 10 minutes, giving a rate of spread of 2.4 to 3.6km/h. Tolhurst (2009) reported that the Kilmore fire averaged 12km/h for an hour, although he does not specify where this occurred. Table two in his report gives an average rate of spread of 8km/h while under the influence of the northerly wind, and 5km/h under the influence of the southerly. Tolhurst also states that flames in excess of 100m high were observed, and detailed evidence given by Fire Investigator Mr Fabian Crowe described flames in excess of 50m amongst the E. regnans on Mt Disappointment (VBRC 2009c). Table 2 states that the maximum intensity of the fire was 150,000 kW/m; however he also reports that there was an average fine fuel load across the fire ground of 25t.Ha-1, reaching 45t.Ha-1 in mountain forests. This stated intensity suggests either a maximum rate of spread in the forest of approximately 6.5km/h, or that the estimate of fuel load was an overestimate. The Royal Commission also provides three maps of the fire edge at given times based on linescan imagery (VBRC 2009b). The map derived from the linescan taken at 2145hrs places the fire front 1km to the north-west of Strathewen, however brigades had listed the town as burnt by 19:00 when they first arrived, a map of ground observations places the front at the 302 same point by 19:25 and the town of Kinglake which lies further to the east recorded the arrival of the fire front at 17:45 (VBRC 2009a). The discrepancy may be due to spotting ahead of the main front, or to part of the image having been obscured. The original of the linescan is not provided in the Royal Commission documents. Fire tower observations for the area reported major spot fires establishing quickly approximately 10 to 15 km ahead of the main front, so two scenarios are possible: 1. The main front crossed the Hume Highway at 13:58 and arrived at Strathewen 30.5 km distant around 19:00hrs, and the areas burnt further to the east were due to new fronts ignited by spot fires. This gives a main front rate of spread of 6.1km/h. 2. The main fire front crossed the Hume Highway at 13:58 and arrived at Kinglake 33.5km distant at 17:45, giving a mean rate of spread of 8.9km/h. Without further information to clarify which scenario is the case, predictions will be validated against the range 6.1 - 8.9km/h, and for flames between 50 to 100m tall. It is unclear where the upper value of 12 km/h reported by Tolhurst (2009) may have occurred. The value of 10.5km/h provided by McCaw et al (2009) is problematic as it does not appear to fit with the timings given by the Royal Commission, but as stated in the report, not all such records were available at the time of writing. 7.352 Weather The weather conditions used in each model were taken from the Bureau of Meteorology 3pm observations for Kilmore Gap weather station (BOM 2009), and are given in table 7.52. The Drought factor was calculated using a KBDI of 100, as McCaw et al (2009) reported that KBDIs for the area met or exceeded this level. A wind reduction factor of 6 was used for calculations with McArthur (1962). Table 7.52. Weather conditions recorded for 1500hrs at Kilmore Gap Temperature Relative Wind Drought humidity speed factor FFDI (km.h-1) 41.4oC 10% 63 10 155.1 303 7.353 Fuels Approximately 4km of the fire run from Heathcote Junction to Strathewen burnt through Mountain Ash (E. regnans) forest, mapped in detail by Ashton (2000). The structure of the forest varies with fire history; however Ashton’s work makes it possible to quantify the fuels in this area with sufficient precision for validation. Values for fuels have been supplied by a number of sources. VBRC (2009a) quotes as their sole source CFA volunteers stating that the fuels in this area were in the range of 40 - 50t.Ha1 , although it is not stated how these were measured. Such values conflict with published equilibrium values of surface fuels (22.25t.Ha) measured with appropriate scientific rigor by Ashton (1975) and should be viewed with caution in the absence of evidence to demonstrate that they are more reliable. Tolhurst (2009) states that due to drought and heat-wave conditions, live fuels were available and he estimates the fuel load as 45t.Ha-1 in forest areas. Tolhurst is clear that this value includes fuels made available by drought conditions and it may be that this is the case, however the use of such values with the equations of Noble et al (1980) derived from McArthur (1967) conflicts with McArthur’s principle of using a ‘drought factor’ to approximate the addition of fuels due to drought stress. In a sense, these fuels are being added twice to the equation. McCaw et al (2009) measured fuel loads at two dry open eucalypt forest sites and provided fuel parameters for both McArthur (1967) and Gould et al (2007); these were surface fuel load of 20t.Ha-1, surface fuel hazard score of 3.5, near-surface score of 3 and near-surface fuel height of 20cm. These values are applicable for the low dry, open forest where they were obtained but cannot be used for the tall wet Mountain Ash forests. McCaw et al (2009) gave the lowest reliability score they had to the fuel values they provided as these were only “typical of equilibrium level in a dry sclerophyll forest”. 304 By focusing instead on a well-known section of the forest such as the “Big Ash” forest studied by Ashton, it is possible to refine the fuel parameters to the highest reliability score given in the report. Ashton (2000) also provided detailed structural diagrams and community descriptions of the Big Ash forest, allowing for measurement of the necessary dimensions required in the FFM. Further details were taken from Dickinson and Kirkpatrick (1985) for leaf chemistry, and where necessary some values for leaf dimensions were taken from field measures, in some instances using closely related species where access to the actual species was not possible. The major limitation to the study is that there is only information on the rate of spread averaged across a large distance containing different slopes and several other Figure 7.17. 27 year-old E. regnans Community C (Ashton 2000). The forest canopy dates from an earlier fire. plant communities, so the value of 7.5 km/h may not actually be accurate for the forest in question and can only be used as a best estimate. The flame height quoted by Tolhurst (2009) is also a generalised value referring to maximum flame heights. Spread is calculated for two communities of Ash making up the majority of the area – Type A and Type C. Although the area occupied by each community is not given by Ashton (2000), the map given in the paper shows an approximate ratio of two to one for the mixture of Type A and C. The parameters used in the empirical models are those given in table 7.2 for Community A, and the parameters for the FFM are given in tables 7.8 to 7.14 for Community A and tables 7.53 to 7.60 for Community C along with the source and methods by which they were derived. The rate of spread modelled for the area based on these parameters is the weighted average of the two communities, and the maximum flame height (the only dimension quoted) is the maximum found in either community. Spread was calculated in this way for positive slopes of five degrees and for flat terrain as the Big Ash forest is located on a gently undulating plateau, and the final quoted rate of spread is the average for both slopes. 305 Table 7.53. Details for E. regnans Community C Parameter Surface fuel Value 22.2t.Ha Details -1 Equilibrium fuel load for E. regnans (Ashton 1975) load Near surface Lepidosperma species elatius 50%, Equal dominance in presence/absence quadrats. (Ashton 2000) Polystichum proliferum 50% Near surface 1.70m Calculated from number of plants per Ha in 1953 using equation 7.3; plant spacing L. elatius assumed to have equal density to P. esculentum. (Ashton 2000) Elevated Olearia argophylla Taken from 1953 survey data as no more recent data or trends were species 53%, Bedfordia identified. B. arborescens understood to now be a dominant from arborescens 47% text. (Ashton 2000) Closed (0m) Assumed from text of study (Ashton 2000) Midstorey Acacia dealbata 6%, Taken from 1953 survey data as no more recent data or trends were species Pomaderris aspera identified. (Ashton 2000) Elevated plant spacing 94% Midstorey plant 9.3m Calculated from 230.2 plants per Ha using equation 7.3. 1953 density spacing assumed to be current (Ashton 2000). Canopy species E. regnans Study describes forest as a pure stand (Ashton 2000) Canopy plant 20.6m Calculated from 47 trees per Ha using equation 7.3. 1953 density spacing Soil texture assumed to be current (Ashton 2000). Clay loam Typical soil type for E. regnans on granite soils (e.g. Robinson et al 2003, Ringrose and Nielsen 2005) Table 7.54. Polystichum proliferum details Parameter Value Details m (%ODW) 239 Single measurement P (oC) 347 Calculated from Silica Free Ash content (Dickinson and Kirkpatrick 1985) of Pteridium esculentum using equation 4.1 % Dead 30 Estimated value due to drought and heat wave conditions Leaf form Flat Field observations d (mm) 0.24 Mean of 5 measurements, accurate to 0.05mm, 99.5% confidence Wl (mm) 5 Single measurement, little variability 306 Ll (mm) 11.3 Mean of 7 measurements, accurate to 25%, 99.5% confidence Sl (cm) 0.33 Counted for 32 leaves Os 2 Observation, no variability between plants Sb (cm) 10 Visual estimate Wb (cm) 35 Mean of 3 measurements, little variability in plant He (cm) 0 Single plant Ht (cm) 62 Mean of 4 measurements from 1 plant Hc (cm) 0 Single plant Hp (cm) 100 Single plant w (cm) 160 Mean of 2 measurements from 1 plant Table 7.55. Lepidosperma elatius details Parameter Value Details m (%ODW) 100 Unknown P ( C) 260 Estimate for VOC1 % Dead 65 Low end of curing range reported for February 2009 (McCaw et al 2009) Leaf form Flat Observed from photograph d (mm) 0.40 Default value for sedges Wl (mm) 13 Average published width (http://plantnet.rbgsyd.nsw.gov.au/cgi- o bin/NSWfl.pl?page=nswfl&lvl=sp&name=Lepidosperma~elatius) Ll (mm) 140 Model maximum Sl (cm) 0.1 Default value for graminoids Os 1 Default value for graminoids Sb (cm) 0 Default value for graminoids Wb (cm) 150 Mean published plant diameter (http://www.yarraranges.vic.gov.au/Directory/S2_Item.asp?Mkey=776&S 3Key=38) He (cm) 0 Estimated from photograph Ht (cm) 100 Estimated proportionally from photograph Hc (cm) 0 Default value for graminoids Hp (cm) 125 Average published height (http://plantnet.rbgsyd.nsw.gov.au/cgibin/NSWfl.pl?page=nswfl&lvl=sp&name=Lepidosperma~elatius) w (cm) 150 Mean published plant diameter (http://www.yarraranges.vic.gov.au/Directory/S2_Item.asp?Mkey=776&S 3Key=38) 307 Table 7.56. Bedfordia arborescens details Parameter Value Details m (%ODW) 250 Mean of 2 measurements of heat stressed foliage, rounded to nearest 50% P (oC) 260 Estimate for VOC1 % Dead 20 Approximation from observation Leaf form Flat Observation d (mm) 0.26 Mean of 6 samples, accurate to 0.05mm, 99.0% confidence Wl (mm) 33 Median published value (Costermans 1994). Ll (mm) 200 Median published value (Costermans 1994). Sl (cm) 0.78 From 16 leaves on herbarium specimen (http://plantnet.rbgsyd.nsw.gov.au) Os 3 Estimate Sb (cm) 60 Estimate Wb (cm) 30 Estimate He (cm) 116 Unknown O. argophylla value used Ht (cm) 290 Unknown O. argophylla value used Hc (cm) 200 Estimate Hp (cm) 464 Unknown O. argophylla value used w (cm) 464 Unknown O. argophylla value used Table 7.57. Olearia argophylla details Parameter Value Details m (%ODW) 70 Estimate from driest sample of Olearia aglossa P (oC) 328 Calculated from Silica Free Ash content using equation 4.1 (Dickinson and Kirkpatrick 1985). % Dead 0 Assumption from common observation Leaf form Flat Observation d (mm) 0.27 Mean of 10 measurements accurate to 0.05mm, 99.5% confidence Wl (mm) 33.2 Mean of 14 measurements accurate to 25%, 99.5% confidence Ll (mm) 78.6 Mean of 14 measurements accurate to 25%, 99.5% confidence Sl (cm) 1.16 From 21 leaves on herbarium specimen (http://plantnet.rbgsyd.nsw.gov.au) Os 4 Unknown 308 Sb (cm) 20 Unknown Wb (cm) 40 Unknown He (cm) 116 Size taken from Community A to reflect growth from 1951 (Ashton 2000). Ht (cm) 290 Size taken from Community A to reflect growth from 1951 (Ashton 2000). Hc (cm) 116 Size taken from Community A to reflect growth from 1951 (Ashton 2000). Hp (cm) 464 Size taken from Community A to reflect growth from 1951 (Ashton 2000). w (cm) 464 Size taken from Community A to reflect growth from 1951 (Ashton 2000). Table 7.58. Acacia dealbata details Parameter Value Details m (%ODW) 135 Single indicative measure 333 Calculated from Silica Free Ash content using equation 4.1 o P ( C) (Dickinson and Kirkpatrick 1985). % Dead 0 Assumption Leaf form Flat Observation d (mm) 0.22 Mean of 10 measurements Wl (mm) 20 Mean of 10 measurements halved to account for bipinnate structure Ll (mm) 28 Mean of 10 measurements halved to account for bipinnate structure Sl (cm) 1.05 Mean of 21 leaves Os 5.2 Mean of 5 branches Sb (cm) 30 Visual estimate Wb (cm) 40 Visual estimate He (cm) 1424 Size taken from Community A to reflect growth from 1951 (Ashton 2000) Ht (cm) 1996 Size taken from Community A to reflect growth from 1951 (Ashton 2000) Hc (cm) 1294 Size taken from Community A to reflect growth from 1951 (Ashton 2000) Hp (cm) 2000 Size taken from Community A to reflect growth from 1951 (Ashton 2000) w (cm) 706 Size taken from Community A to reflect growth from 1951 (Ashton 309 2000) Table 7.59. Pomaderris aspera details Parameter Value Details m (%ODW) 100 Value unknown P ( C) 260 Estimate for VOC1 % Dead 0 Assumption Leaf form Flat Observation d (mm) 0.31 Mean of 9 samples, accurate to 0.05mm, 99.5% confidence Wl (mm) 13.1 Mean of 12 samples on herbarium specimen, accurate to 25%, 99.5% o confidence (http://plantnet.rbgsyd.nsw.gov.au) Ll (mm) 42.2 Mean of 11 samples on herbarium specimen, accurate to 25%, 99.5% confidence (http://plantnet.rbgsyd.nsw.gov.au) Sl (cm) 1.67 From 26 leaves on herbarium specimen (http://plantnet.rbgsyd.nsw.gov.au) Os 4 Unknown Sb (cm) 20 Unknown Wb (cm) 40 Unknown He (cm) 406 Mean of 17 plants in diagram (Ashton 2000) Ht (cm) 884 Mean of 17 plants in diagram (Ashton 2000) Hc (cm) 478 Mean of 17 plants in diagram (Ashton 2000) Hp (cm) 1072 Mean of 17 plants in diagram (Ashton 2000) w (cm) 435 Mean of 17 plants in diagram (Ashton 2000) Table 7.60. Eucalyptus regnans details Parameter Value Details m (%ODW) 102 Equation 6.16 P ( C) 220 Estimate for VOC2/3 % Dead 0 Assumed Leaf form Flat Observation d (mm) 0.39 Mean of 15 mature E. delegatensis leaves Wl (mm) 28 Mean of 15 mature E. delegatensis leaves Ll (mm) 129 Mean leaf length calculated from Ashton (1975) Sl (cm) 2.5 Mean from 31 mature E. delegatensis leaves o 310 Os 5.4 Mean from 5 mature E. delegatensis branches Sb (cm) 318 Mean of 25 measurements from diagram (Ashton 2000) Wb (cm) 434 Mean of 25 measurements from diagram (Ashton 2000) He (cm) 2224 Mean of 7 trees in diagram (Ashton 2000) Ht (cm) 3768 Mean of 7 trees in diagram (Ashton 2000) Hc (cm) 2155 Mean of 7 trees in diagram (Ashton 2000) Hp (cm) 4448 Mean of 7 trees in diagram (Ashton 2000) w (cm) 1138 Mean of 7 trees in diagram (Ashton 2000) 7.354 Modelled fire behaviour The modelled fire behaviour is compared with the estimated behaviour in tables 7.61 to 7.63. Of the empirical models, McArthur (1967) was most accurate at estimating both the rate of spread and flame height; although Gould et al performed acceptably on the flame height. Gould et al significantly over estimated the rate of spread (258% error), and McArthur (1962) showed over-estimates of 178% and 362% on both factors respectively. Output from the FFM demonstrated a faster rate of spread in Community A compared to Community C on flat ground, but faster fire in C on sloping ground. The average rate of spread weighted across both communities and slopes was 4.31km//h, which was slower than the estimated spread and the prediction by McArthur (1967) but significantly closer than the estimates from the other two models. The maximum rate of spread and flame height occurred in Community C on a 5o slope. In this case, spread occurred by crown pulses which spread at 24km/h for 22m ahead of the surface fire before expiring and reigniting from the surface. As the weather readings were taken from a site below the plateau it is possible that the wind speeds on-site were actually higher than those reported in table 7.52, which has implications for Figure 7.18. Rate of spread related to wind speed in Community C on a 5o slope. The maximum rate of spread is 9.2km/h, occurring in winds of 100km/h. 311 modelled rates of spread. Figure 7.18 shows the range of spread rates possible for Community C on a 5o slope for winds between 0-100km/h. These results suggest that if wind speeds were higher on the plateau as could be expected, the modelled rate of spread would also be higher. Table 7.61. Rates of spread and flame heights as estimated above and modelled by three Australian empirical forest fire behaviour models and the FFM using the fuel parameters given in tables 7.2, 7.8 to 7.14 and 7.53 to 7.60 with 0o slope. Model McArthur ROS (km/h) Max Flame height ROS (km/h) Max Flame height Community A (m) Community A Community C (m) Community C 20.5 462.0 20.5 462.0 4.1 57.2 4.1 57.2 26.4 109.3 26.4 109.3 4.07 68.9 2.39 72.7 6.1–8.9km/h 50-100m 6.1–8.9km/h 50-100m (1962) McArthur (1967) Gould et al (2007a) FFM Observed estimate Table 7.62. Rates of spread and flame heights as estimated above and modelled by three Australian empirical forest fire behaviour models and the FFM using the fuel parameters given in tables 7.2, 7.8 to 7.14 and 7.53 to 7.60 with 5o slope. Model McArthur ROS (km/h) Max Flame height ROS (km/h) Max Flame height Community A (m) Community A Community C (m) Community C 29.0 462.0 29.0 462.0 5.85 79.4 5.85 79.4 37.33 140.3 37.33 140.3 4.46 69.5 6.43 74.0 6.1–8.9km/h 50-100m 6.1–8.9km/h 50-100m (1962) McArthur (1967) Gould et al (2007a) FFM Observed estimate 312 Table 7.63. Final reported rates of spread and flame heights based on a weighted average between both communities [(2A+C)/3] and the average of both slopes. ROS (km.h-1) Model % Error Max Flame height (m) % Error McArthur (1962) 24.8 178% 462.0 362% McArthur (1967) 5.0 22% 79.4 0% Gould et al 31.9 258% 140.3 40% 4.31 42% 74.0 0% (2007a) FFM 7.4 Model strengths and weaknesses 7.41 Predictive validation The observed vs. expected results for each model are shown in tables 7.64 and 7.65, and the results are graphed in Figures 7.19 to 7.30. The spotting distance modelled from the predicted flame height using equation 7.9 is shown in table 7.66, and all observed vs. expected results are shown in Figures 7.19 to 7.30. The values of mean absolute error for flame height and rate of spread are shown in Figures 7.31 and 7.32. Table 7.64 Summary of observed and modelled flame heights (m) Date Time Observed McArthur McArthur Gould et al 1962 1967 2007a FFM 7/1/5 1130 0.8 0.16 4.90 0.39 1.13 25/2/5 1200 0.5 0.68 9.28 2.31 0.47 13/4/5 1310 0.7 0.34 1.95 0.27 0.91 13/4/5 1400 0.3 0.34 3.24 0.39 0.86 28/4/5 1330 1.1 0.70 8.31 1.73 0.69 28/4/5 1415 0.6 0.82 5.34 1.83 0.47 26/1/3 1300 1.5-10 8.00 7.4 5.5 1.02 13/10/6 1500 1.5-12 4.30 1.90 16.40 2.98 7/2/9 1500 50-100 462.0 79.40 140.30 73.97 313 Table 7.65 Summary of observed and modelled rates of spread (km.h-1) Date Time Observed McArthur McArthur Gould et al 1962 1967 2007a FFM 7/1/5 1130 0.02 0.01 0.24 0.06 0.07 25/2/5 1200 0.05 0.13 0.66 0.67 0.11 13/4/5 1310 0.06 0.02 0.13 0.03 0.04 13/4/5 1400 0.06 0.03 0.23 0.06 0.07 28/4/5 1330 0.08 0.11 0.63 0.40 0.09 28/4/5 1415 0.07 0.08 0.40 0.44 0.11 26/1/3 1300 - 0.10-0.90 0.50-1.37 2.07-11.57 0.58-1.57 13/10/6 1500 3.5 0.22 0.12 2.57 2.57 7/2/9 1500 6.1-8.9 24.8 5.0 31.9 4.31 McArthur 1967 Gould et al 2007a FFM Table 7.66 Spotting distances (m), modelled from equation 7.9 Date Time From McArthur observed fh 1962 7/1/5 1130 7 0 390 2 16 25/2/5 1200 3 5 1580 75 2 13/4/5 1310 6 1 52 1 10 13/4/5 1400 1 1 157 2 9 28/4/5 1330 15 5 1241 40 5 28/4/5 1415 4 8 471 45 2 26/1/3 1300 29 - 2294 1142 962 502 13 13/10/6 1500 29 - 3000 293 49 5504 131 7/2/9 1500 Max distance Max distance Max distance Max distance Max distance 314 20 15 10 5 Modelled flame height (m) 500 Modelled flame height (m) Modelled ROS (km/h) 25 450 400 350 300 250 200 150 100 50 0 0 5 10 15 20 0 25 observed rates of spread Figure for observed 7 6 5 4 3 2 1 7.20. 0 4 5 6 7 observed 7.22. rates 8 McArthur (1967) Modelled of 0.5 and Figure heights for observed flame 9 10 spread 1.0 1.5 7.21. Modelled and heights for flame McArthur (1962), flames <1.5m. 10 90 80 70 60 50 40 30 20 9 8 7 6 5 4 3 2 1 0 0 0 Observed ROS (km/h) Figure 0.2 Modelled 10 3 0.4 Observed flame height (m) Modelled flame height (m) Modelled flame height (m) 8 2 0.6 0.0 100 9 1 0.8 100 150 200 250 300 350 400 450 500 McArthur (1962) 10 Modelled ROS (km/h) 50 and McArthur (1962) 0 1.0 Observed flame height (m) Modelled 7.19. 1.2 0.0 0 Observed ROS (km/h) Figure 1.4 10 20 30 40 50 60 70 80 90 100 0 Observed flame height (m) and Figure for observed 7.23. flame McArthur (1967) 1 2 3 4 5 6 7 8 9 10 Observed flame height (m) Modelled and Figure heights for observed 7.24. flame Modelled and heights for McArthur (1967), flames <1.5m. 315 160 30 25 20 15 10 5 3.0 Modelled flame height (m) Modelled flame height (m) Modelled ROS (km/h) 35 140 120 100 80 60 40 2.5 2.0 1.5 1.0 0.5 20 0 0 10 20 0 30 0.0 0 Observed ROS (km/h) Figure 40 60 80 100 120 140 160 0.0 Observed flame height (m) Modelled 7.25. 20 and Figure Modelled 7.26. 0.5 1.0 1.5 2.0 2.5 3.0 Observed flame height (m) and Figure 7.27. Modelled and observed rates of spread for Gould observed flame heights for Gould observed flame heights for Gould et al et al et al, flames <1.5m. 10 Modelled flame height (m) Modelled ROS (km/h) 8 7 6 5 4 3 2 1 2.0 Modelled flame height (m) 100 9 90 80 70 60 50 40 30 20 10 0 0 2 4 6 8 10 7.28. Modelled 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0 0.0 0 Observed ROS (km/h) Figure 1.8 10 20 30 40 50 60 70 80 90 100 0.0 Observed flame height (m) and Figure Modelled 7.29. 0.5 1.0 1.5 2.0 Observed flame height (m) and Figure 7.30. Modelled and observed rates of spread for the observed flame heights for the observed flame heights for the FFM FFM FFM, flames <1.5m. The mean percent error of each model tested against the fires is shown in Figures 7.31 and 7.32, and the results of the Paired t-test are shown in table 7.67. Table 7.67. Results of the paired t-test comparison of the models RATE OF SPREAD MODEL FLAME HEIGHT Mean error Significance Mean error Significance McArthur (1962) 287% N.S. 111% N.S. McArthur (1967) 843% 95.0% 528% 95.0% Gould et al (2007) 349% 90.0% 106% N.S. FFM 69% 46% 316 RATE OF SPREAD Mean absolute error FLAME HEIGHT Mean absolute error 1800% 1200% 1600% 1000% 1400% 800% 1000% Error Error 1200% 843% 800% 600% 400% 400% 349% 287% 69% 200% 528% 600% 0% 200% 106% 111% 46% 0% McArthur 1962 McArthur 1967 Gould et al 2007 Zylstra 2011 McArthur 1962 McArthur 1967 Gould et al 2007 Zylstra 2011 Figure 7.31. Mean absolute percent error in rate of Figure 7.32. Mean absolute percent error in flame spread predictions calculated for all models, based on height predictions calculated for all models, based on eight fires ranging from low to extreme intensity ten scenarios ranging from low to extreme intensity For the limited although diverse dataset examined and based on the mean absolute percent error, the Forest Flammability Model was between 4 and 12 times more accurate than the empirical models for estimating rate of spread, and between 2 and 12 times more accurate in estimating flame height. The difference in model accuracy was statistically significant against both McArthur (1967) and Gould et al for rate of spread but only significant against McArthur (1967) for flame height; although the significance was high (95.0%). The flame height models of both McArthur (1962) and Gould et al (2007) was comparably accurate to the FFM. The most consistently accurate of the empirical models was McArthur (1962), although Gould et al was slightly better for predicting flame heights. The accuracy of McArthur (1962) was largely dependent upon the wind reduction factor. Predictions by McArthur (1967) had very little relationship to the recorded values. Although this dataset is small, it demonstrates that although the experimental basis of McArthur (1967) was the largest of all of the models, its accuracy was much lower than those other models which included fuel structure, the separation of the influence of wind from other weather parameters and/or the use of a wind reduction factor. 317 In McArthur (1962) the treatment of wind was improved by its separation from other weather parameters and use of a wind reduction factor, but fuels were all combined into a single value. This model was quite accurate for milder weather conditions where surface fuels dominated fire behaviour, but less accurate for higher intensity fires where other fuels dominated. The FFM did not however produce the same level of under-prediction in the Tooma Dam fire or over-prediction in the Kilmore fire because it was capable of calculating the roles of these higher fuels. Gould et al separated wind from other weather parameters and provided some further information on fuel structure, but did not allow for a wind reduction factor. The reduced wind reduction in the Tooma Dam fire would have facilitated faster fire spread and perhaps been adequate to reproduce the most accurate rate of spread, but would have caused significant over-preediction in flame heights. By raising the wind reduction factor for the Kilmore fire to a realistic value, rate of spread predictions may have been significantly improved in this case and flame heights reduced from their excessive value. There is potential for this model to be improved in this way. Rates of spread for the experimental fires were significantly overpredicted when wind speeds greater than 5km/h were present. This may be an artefact of the small plot size preventing fires from reaching their potential spread rate as stated in the report for this model, or it may result from the use of near surface fuels for determining rate of spread rather than elevated fuels. As elevated fuels produce greater flame heights, greater wind speeds are required to produce flame connection. Further experimentation in similar fuels using larger plot sizes may be of benefit in answering these questions. McArthur (1967) consistently over-predicted both rates of spread and flame heights in all experimental fires, reflecting the exaggerated value given to surface fuels by the model just as in the under-prediction for the Tooma Dam fire. Inclusion of a wind reduction factor is unlikely to produce any improvement in the model as wind is not separated from the other weather parameters. 318 7.42 Credibility analysis 7.421 Value for planning prescribed burns The performance of the four models as decision making tools for planning prescribed fire in high fire danger conditions or less is summarised in table 7.68 using the criteria from table 7.64. McArthur (1962) and th FFM performed equally well for determining valuable prescriptions, While error from the FFM was equally distributed between conservative and destructive, errors from McArthur were only destructive, suggesting that prescriptions were safe when flames were in fact too tall. Gould et al performed slightly less well, with a larger emphasis on conservative predictions and McArthur (1967) produced only one useful prediction, with all others suggesting that flames were too tall when they were not. Table 7.68. Performance of each model as a decision support tool in prescribed burning for fires examined in Low to High fire danger conditions Date th Time Observed behaviour McArthur McArthur Gould et al (1962) (1967) (2007a) FFM 7 Jan 05 1130 Within range D1 C1 G C1 25th Feb 05 1200 Within range G C1 C1 G 13 April 05 1310 Within range G C1 D1 G 13thApril 05 1400 Within range G C1 G G 28 April 05 1330 Above range D1 G G D1 28th April 05 1415 Within range G C1 C1 G 1500 Above range G C1 G G Accuracy 0.71 0.14 0.57 0.71 Destructive/dangerous 0.28 0.00 0.14 0.14 Conservative 0.00 0.86 0.29 0.14 th th th 13 Oct. 06 7.422 Value for incident management As with the prescribed burn assessment, the attack method criterion is based on flame height, so the models that predicted flame height effectively (Figures 7.27, 7.30, 7.32) provided the 319 most useful predictions. Rate of spread accuracy affected the initial attack criterion as much as flame height, which determined spotting distance. The forward planning criterion produces a series of scores weighted by the accuracy of rate of spread predictions and the consequences of error as determined by the flame height. Attack method The best model for determining attack method was McArthur (1962, table 7.69), which advised the best method in every instance. The FFM perfomed next best, with one failure where direct attack had been advised under conditions where parallel or more conservative attack methods were needed. The analysis of this fire (26th Jan 2003) makes it clear that this problem could be avoided by modelling for variability in the vegetation composition. Gould et al performed next best, making a number of low-consequence conservative errors. McArthur (1967) performed least reliably, producing a number of large conservative errors. In four instances, the model advised indirect methods (backburning) for fires which could have been addressed through direct attack. Table 7.69. Performance of each model against the attack method criterion Date th 7 Jan 05 Time Observed behaviour McArthur McArthur Gould et al (1962) (1967) (2007a) FFM 1130 Direct attack G C2 G G th 1200 Direct attack G C2 C1 G th 1310 Direct attack G C1 G G th 1400 Direct attack G C2 G G th 1330 Direct attack G C2 C1 G th 1415 Direct attack G C2 C1 G th 1300 Indirect attack to G G G D1 G G G G Property protection G G G G Accuracy 1.00 0.21 0.67 0.89 Dangerous 0.00 0.00 0.00 0.11 Conservative 0.00 0.78 0.33 0.00 25 Feb 05 13 April 05 13 April 05 28 April 05 28 April 05 26 Jan 03 property protection th 13 Oct. 06 1500 Indirect attack to property protection th 7 Feb 09 1500 320 Initial attack Both McArthur (1962) and the FFM advised correctly as to initial attack success in every instance (table 7.70), whereas Gould et al had only a single and minor conservative error. McArthur (1967) in contrast advised falsely on all but two fires, in the majority of these suggesting that initial attack efforts would fail when they were in fact more likely to succeed. Table 7.70 Performance of each model in determining the initial attack criterion Date Time From observed McArthur behaviour 1962 McArthur 1967 Gould et al 2007a FFM 7/1/5 1130 Success G C1 G G 25/2/5 1200 Success G C1 C1 G 13/4/5 1310 Success G C1 G G 13/4/5 1400 Success G C1 G G 28/4/5 1330 Success G C1 G G 28/4/5 1415 Success G C1 G G 26/1/3 1300 Fail G G G G 13/10/6 1500 Fail G D1 G G 7/2/9 1500 Fail G G G G Accuracy 1.00 0.22 0.89 1.00 Dangerous 0.00 0.11 0.00 0.00 Conservative 0.00 0.67 0.11 0.00 Forward planning The FFM significantly out-performed all of the empirical models in forward planning, as the consequence of errors was approximately half of that for the next best model (Table 7.71 and figure 7.33) and the consequence for good results was nearly twice as high as the next best performer. The model gave good advice more than twice as often as it gave poor advice, and the poor advice tended toward conservative decisions. Both McArthur (1962) and Gould et al had similar levels of consequence for errors, although Gould et al produced good predictions more often than the other. Gould et al characteristically produced conservative decisions so that the consequence of overly-cautious actions was 8 times greater than that of good actions. McArthur (1962) produced the least good consequences; however the error types were relatively even. McArthur (1967) predicted well for the Kilmore fire which had the highest consequence, however all other predictions had moderate to extreme error levels and consequences, equally divided between conservative and dangerous. 321 Table 7.71 Performance of each model in determining the forward planning criterion, final weightings are given in brackets. McArthur 1967 Gould et al 2007a D1 (0.25) C5 (1.25) C2 (0.5) C3 (0.75) 1200 C2 (0.5) C5 (1.25) C5 (1.25) C2 (0.5) 13/4/5 1310 D2 (0.5) C2 (0.5) D1 (0.25) D1 (0.25) 13/4/5 1400 D2 (0.5) C3 (0.75) G (0.25) G (0.25) 28/4/5 1330 G (0.25) C4 (1) C5 (1.25) G (0.25) 28/4/5 1415 G (0.25) C4 (1.25) C4 (1) C1 (0.25) 13/10/6 1500 D5 (5) D6 (6) G (1) G (1) 7/2/9 1500 C2 (4) G (2) C3 (6) G (2) Accuracy 0.5 2 1.25 3.5 Dangerous 6.25 6 0.25 0.25 Conservative 4.5 6 10 1.5 TOTAL 11.25 14 11.5 5.25 Date Time 7/1/5 11:30 25/2/5 McArthur 1962 FFM Forward planning reliability McArthur (1962) McArthur (1967) Overly cautious Gould et al (2007) Safe FFM Dangerous Figure 7.33. Reliability of the models for forward planning during incidents. The total length of the bars represents the spread in the consequence of the predictions whereas the different shadings represent the types of error. 322 7.43 Discussion The predictive validation was sufficient for analysing rate of spread predictions and provided statistically valid results against two of the three models. The flame height predictions were less conclusive, with statistical significance established for the comparison with McArthur (1967) but not for either of the other models. Further analysis is suggested to determine whether significance is possible against the other two models. The credibility analysis demonstrated that if used as a decision making tool, the FFM is equivalent in value to McArthur (1962) and superior to the other two models for developing prescriptions in the conditions examined. For managing incidents, the FFM is slightly less effective than McArthur (1962) and more effective than the other two models for determining the mode of attack, is equally effective with McArthur (1962) and more effective than the other two models for determining the likelihood of initial attack success, and is significantly more effective than any of the other models for forward planning. The value of the FFM could be improved in incident planning by incorporating variability in fuel structure and weather conditions at the site. Of the empirical models, McArthur (1962) was the most accurate and credible in general, although Gould et al was slightly better in flame height prediction due to a better prediction for the Kilmore fire. Gould et al performed only slightly less well than McArthur (1962) in the credibility analysis, although the very strong tendency of the model to over-predict is likely to result in overly-cautious incident management decisions. McArthur (1967) performed poorly in all tests. The range of error for predictions by this model is very large and direction provided by the model is highly unreliable. Outcomes for the FFM The FFM is very sensitive to thresholds marking flame connection and the ignition of plants, and this can introduce significant error to predictions. Also, as mentioned in chapter five the downhill rate of spread is not necessarily reliable when the fire is burning with the wind as it is unknown whether a ground-based flame will back toward the slope or lean with the wind. This may have reduced the overall modelled rate of spread for the Tooma Dam fire due to the lee-slope section of fire spread. The following changes are suggested: 323 1. Fire behaviour during incidents should be modelled accounting for variability in the fuel array and wind speeds. 2. Minimise the complexity resulting from optimal angle affecting plant flame lengths (Figure 7.8) until or unless this is proven to be legitimate. This can be achieved by ascribing a single value to plant base height and possibly the height at the top of the plant. 3. Produce specific predictions from smoothed data (e.g. Figure 7.9) to allow for error in the calculation of flame connections. 4. Investigate solutions for downhill rates of spread. 7.431 Observed fire growth patterns In addition to the standard validation covered here, the fires examined demonstrated examples of the growth patterns covered in chapter three. In particular, abrupt changes in fire behaviour were demonstrated in the Monte Carlo analysis of the E. regnans forest as well as in the plots of fire behaviour against wind speed for a number of the fires. These changes were not predetermined with assigned values but were emergent behaviours caused by progression of the flame through the fuel ladder (section 3.221) and optimal angle effects (section 3.227). Sudden and unexpected drops in fire behaviour also occurred where optimal angle and flame angle damping caused the crown fire to extinguish and only allowed a surface fire. 324 Chapter 8 Implementation and Applications 8.1 Implementation of the model as an operational tool The model is ready for use in its current spreadsheet form and available in the accompanying CD. However due to the large file size the spreadsheet is somewhat unweildy and a purposebuilt software package would provide greater speed and ease of use. The model would also have much greater practical functionality if it was implemented within a spatial platform such as a Geographic Information System. The primary barrier to implementation of the model however is the collection of fuels data. Such data can be time-consuming to collect even for simple models, however in this case the number of variables required is quite large. Because of this, the model will have limited use as a tool that can be utilised without earlier preparatory work. It may be possible to reduce the amount of work required by investigating and where appropriate implementing the following steps: 1. Conduct a full sensitivity analysis of all fuel parameters to identify those which can be generalised. It appears at this stage that the most sensitive parameters are the plant spacing and crown dimensions. 325 2. The level of complexity modelled by differences between He and Hc, and between Ht and Hp should be examined in greater detail to determine its importance. If this detail is not sensitive, the parameters can be generalised into single measures of plant top and base height. 3. Plant characteristics such as leaf dimensions that do not vary greatly within the species or express variability that can be modelled from growth stages or climatic variability can be collected and characterised for different species in a centralised database to negate the need for repeated collection. 4. Vegetation mapping can be used to identify the dominant species in an area, and the basic characteristics of those species taken from the database. 5. Plant base height, top height and width can be measured using remote sensing platforms such as LIDAR (e.g. Lee et al 2009, Loudermilk et al 2009). 6. Plant moisture can be either modelled with species-specific models such as those presented in this thesis, or measured directly via remote sensing (e.g. Dasgupta et al 2007, Yebra et al 2008, Yebra and Chuvieco 2009). The steps outlined here have the potential to enable mapping of fuels across large areas with the use of remote sensing, enabling a transition from the traditional approach of point samples to landscape-scale mapping. Unless these steps are shown not to be viable, the collection of fuels data may not prove to be an obstacle to implementation. As the model is not calibrated to any particular fuel array via empirical constants but can be set to any environment using repeatable data collection methods, it has potential for application in any fuel array. There are however some current limitations in that the models for leaf flammability have been derived from sclerophyllous vegetation, and that some plant forms such as vines are difficult to accommodate in the present structure. Priority research in this area then will include: 1. Repetition of the leaf flammability studies in other leaf forms such as grasses, coniferous and deciduous vegetation. 2. Development and trialling of fuel parameters for vines and any other identified different forms. 326 A further area of development to enable the application of the model to wet forests or rainforest environments is: 3. Upgrade of the dead fuel moisture model to a model that can incorporate the effects of incident radiation and wind speed on dead fuel moisture. 8.2 Application of the model 8.21 Fuel management As discussed in chapter two, the response of plants to fire can vary widely across forest areas; the Forest Flammability Model has particular value in determining bushfire risk and treatment because it is able to quantify the effects of the local fire ecology on potential fire behaviour. Prioritising management objectives Effective risk management requires the logical and strategic application of fire behaviour science to fuel management. There is, for example, no strategic value in managing fuels to enable direct attack of flames when the location of the site is too steep or remote and indirect attack is a more likely approach during an incident. Consequently, fuels may be managed for a range of different objectives such as: 1. Enabling direct attack, parallel or indirect attack under a wider range of conditions (Australasian Fire Authorities Council 1996) 2. Reducing flame height to minimise spotting distance below an identified maximum (Ellis 2000) 3. Reducing fibrous bark to minimise the number of embers available for short distance spotting (Gould et al 2007a) 4. Maintaining a good cover of surface litter with minimal shrub cover to facilitate down-slope back-burns (Burrows 1999 a, b) 5. Reducing rate of spread to allow for evacuation or to minimise fire perimeter prior to initial attack (Weber and Sidhu 2006). 6. Increasing the likelihood of direct attack success (McCarthy and Tolhurst 1997) 327 7. Increasing the predictability of low intensity fires so that remote fire crews are safer and less likely to be caught out by sudden changes. 8. Reducing the length of flames and their variability to maximise containment probability (Mees et al 1993) It will be the exception if all or even most of these goals can be achieved simultaneously. Section 7.22 demonstrated that flame height and rate of spread are not necessarily correlated for instance, so fuel management that reduces one may have no effect on or may even increase the other. Mechanical thinning of trees in forestry coupes may reduce the chance of active crown fire (Stephens and Moghaddas 2005, Sabo et al 2009, Stephens et al 2009), but by reducing the LAI of the canopy it may also allow greater wind speeds and light access at ground level so that ground fires are both faster and more frequent. The critical question is which outcome is most desirable for the site in question. Effective fuel management then requires a rational prioritising of objectives based upon what is most important and what can realistically be achieved. Given the enormous range of ecological responses to fire discussed in chapter two and the effect of these on fire behaviour demonstrated by the Forest Flammability Model, it is no longer defensible to simply assume Zedler and Sieger’s (2000) fuel-age paradigm (Fernandes and Botelho 2003). Chapter seven demonstrated that the empirical models examined rely on fuels which play little or no role in fire behaviour, although those used by Gould et al (2007a) appear to be more representative than those of either of McArthur’s models. Because these fuels increase with time however, the models report on what is effectively a pre-determined and incorrect answer. In effect, the fuel-age paradigm is inbuilt in these models. 8.211 Case study To illustrate this point, the following case study uses data collected by the author from the subalpine Snowgum community, which was raised in chapter two as a long-standing problem that has not yet been addressed. Observations that the fuel-age paradigm did not hold in this environment were first recorded in 1893 (Helms 1893) and have since been raised by graziers and examined in detail in the peer-reviewed literature. This study uses the FFM along with McArthur (1967) and Gould et al (2007a) to demonstrate the differences in model output that occur between those that are directed by the paradigm and the FFM which is not. 328 Indicators of fire behaviour were modelled for 35 different age classes ranging from directly after the fire through to 50years post-fire. A site at Guthega was surveyed at six years since fire (tables 7.3 to 7.7) and in a long-unburnt state, possibly about 50 years since fire (tables 7.37 to 7.41). As only two age classes had been surveyed, the fuel parameters were estimated by assuming a power growth rate for the plants up to 6 years (Equation 8.1) followed by logarithmic growth (Equation 8.2), and by allowing for two separate seral stages based on the regrowth and mature understoreys. The change in plant spacing was modelled with linear growth. The change to mature understorey is placed at 20 years as other studies have shown a decline in Bossiaea foliosa at this time (Leigh et al 1987). Fire is assumed not to spread at zero years. The constants for the growth formulae are given in tables 8.1 and 8.2. The basic formulae are: V = ay b Equation 8.1 V = a ln( y ) + b Equation 8.2 Where V is the variable being considered, y is the time since fire in years and a and b are the constants from the table below. Four flammability indicators were calculated by modelling fire behaviour for every day of one summer, and either taking an average for the summer or calculating the percentage of days in the summer for which the criterion was satisfied. The indicators were flame height, rate of spread, spotting distance and direct/parallel attack failure. The weather used for the analysis was taken from the Australian Bureau of Meteorology records for 3pm weather over the summer of 2008/09 for the Thredbo upper weather station (Crackenback) and the Cabramurra weather station. Each parameter was then recalculated for the altitude of the survey site at 1617m, assuming a linear change in the parameter between the upper Crackenback altitude of 1957m and the lower Cabramurra altitude of 1475m. Wind was assumed to be up-slope at all times for simplicity. 329 Table 8.1. Constants for Equation 8.1 in young (0-6 years) E. niphophila forest CONSTANTS YOUNG Percent dead Thickness (mm) Width (mm) Length (mm) Separation (cm) Branch orders Clump separation Clump diameter (cm) He (cm) Ht (cm) Hc (cm) Hp (cm) w (cm) NEAR SURFACE Poa costiniana a b 0.50 0.00 0.17 0.22 0.17 0.22 14.06 1.15 0.03 0.00 0.37 0.56 0.00 0.00 1.97 1.29 0.00 0.00 3.87 0.59 0.00 0.00 3.87 0.59 1.97 1.29 ELEVATED O.Phlogopappa a 0.15 0.19 2.32 4.57 0.03 2.18 1.54 2.32 2.43 3.55 2.23 4.11 3.65 b 0.17 0.28 0.37 0.66 0.64 0.34 1.19 1.37 1.38 1.66 1.36 1.61 1.68 Orites lanceolata a b 0.00 0.00 0.13 0.13 2.47 0.39 2.47 0.39 0.15 0.17 2.47 0.39 0.00 0.00 3.55 1.66 1.85 1.27 2.99 1.48 1.74 1.24 3.55 1.66 4.08 1.61 CANOPY Eucalyptus niphophila a b 0.00 0.00 0.26 0.41 20.28 0.01 62.87 0.02 2.16 0.03 5.28 0.02 0.00 0.00 5.13 1.71 4.23 1.63 6.11 1.79 5.16 1.71 6.64 1.82 6.29 1.80 Table 8.2. Constants for Equation 8.2 in older (>6 years) E. niphophila forest CONSTANTS OLD Percent dead Thickness (mm) Width (mm) Length (mm) Separation (cm) Branch orders Clump separation Clump diameter (cm) He (cm) Ht (cm) Hc (cm) Hp (cm) w (cm) NEAR SURFACE a b -0.05 0.59 0.00 0.25 0.00 0.25 0.00 110.00 0.00 0.03 0.00 1.00 0.00 0.00 0.00 20.00 0.00 0.00 0.00 11.00 0.00 0.00 0.00 11.00 0.00 25.00 a 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -3.30 -7.07 0.47 -5.19 -1.41 ELEVATED b a 0.20 0.00 0.32 0.00 4.47 0.00 14.93 0.00 0.09 0.00 4.00 0.00 12.90 0.00 26.80 0.00 34.92 0.00 69.68 0.00 24.15 0.00 83.30 0.00 62.54 0.00 CANOPY b 0.00 0.49 7.30 21.10 0.29 4.80 0.00 137.00 67.00 128.00 43.00 120.00 198.00 a 0.00 0.03 -1.13 1.27 -0.59 -0.24 39.62 5.19 142.44 190.54 135.36 246.67 98.10 b 0.00 0.49 22.53 62.92 3.35 5.92 -70.99 100.70 -177.21 -191.41 -131.53 -267.97 -17.77 The value for each of the four parameters is given in Figures 8.1 to 8.4. Three broad stages are apparent – a primary inhibition period of reduced flame height, rate of spread and spotting distance, which all rapidly increase as the vegetation regrows. After a sharp peak in rate of spread and a broader plateau in flame heiught and spotting distance, the parameters again decline into a more permanent secondary inhibition period. This secondary decline is directly opposite to the expectations of the fuel-age paradigm, which assumes that all such parameters will be greater in old fuels compared to youger fuels. 330 These stages can be understood in terms of the principles described in chapter three. The fuel ladder in young, regrowing fuels is vertically contiguous, but low to the ground. As a result, the full fuel profile ignites easily so that no overhead vegetation is left to protect the ground fuels from wind and nearly all fires spread at the maximum rate of spread for the conditions. As fuels age and grow in height but maintain vertical continuity, flame heights also increase. After three years post-fire, sufficient space develops between the shrub layer and the recovering Snowgum coppice to prevent ignition of this canopy layer in some instances. Where crown fires do occur, flame heights continue to increase as the forest grows and the trees increase in height and cover, but at the same time they become less frequent. As the flames from the crown fires increase in size, flame angle damping also slows the rate of spread and prevents head development, so that even crown fires are generally slower. This is not always the case – the strongest winds are still capable of producing fully developed active crown fires, but on average the reduction in crown fire incidence and the change from active toward passive crown fires causes a significant and sudden reduction in rates of spread. The timing of this is related to the growth rates of the plants, and to the slope of the terrain. On gentler slopes, the flame angles will be more upright in relation to the vegetation and at lower altitudes the vegetation will grow more quickly, so as a generalisation flatter slopes and faster plant growth will produce earlier peaks. Although the rate of spread reduces quickly after stratification starts to develop, the average flame height continues to increase because even though crown fires are less frequent, they produce significantly larger flames. This is evident when mean rate of spread is compared with the values for spotting distance and attack failure. Mean flame height begins to reach equilibrium at around 10 years, yet spotting distance still increases significantly after this point because the increasingly rare crown fires spot for sufficient distance to outweigh the minor changes in the ground fires. Attack failures however peak sooner, as these are driven by the frequency at which flames exceed a given threshold. So after 10 years, the frequency of flames greater than 1.5m in height drops rapidly, as do 3m flames after 12 years. Despite this, the mean flame height does not peak until 16 years and the tallest flames probably still continue to become taller as the trees grow, facilitating longer distance spotting; but the frequency of crown fires becomes so rare after this point that mean flame height begins to decrease. 331 Mean summer Flame height Mean summer ROS 3.5 2.5 2.0 2.5 ROS (km/h) Flame height (m) 3.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 0 10 20 30 40 0 50 10 20 30 40 50 Time since fire (years) Time since fire (years) Figure 8.1. Mean summer flame height modelled for Figure 8.2. Mean summer rate of spread modelled one summer in Snowgum forest on a 17 degree slope for one summer in Snowgum forest on a 17 degree slope Mean summer attack failures 100% 0.40 90% Direct attack 0.35 80% Parallel attack 0.30 70% Failure rate Distance (km) Mean summer spotting distance 0.25 0.20 0.15 60% 50% 40% 30% 0.10 20% 0.05 10% 0% 0.00 0 10 20 30 40 0 50 10 20 30 40 50 Time since fire (years) Time since fire (years) Figure 8.3. Mean summer spotting distance modelled Figure 8.4. Mean summer attack failures modelled for one summer in Snowgum forest on a 17 degree for one summer in Snowgum forest on a 17 degree slope slope The graph of mean rate of spread modelled by both McArthur (1967) and Gould et al (2007a) is shown in Figure 8.5. In both cases, these models predicted the classical Olsen curve (Figure 2.1), where the mean rate of spread increased initially until reaching an equilibrium point, after which it maintains stasis. The implication of this is that rate of spread can always be reduced by reducing the fuel age; that is, the fuel age paradigm is supported. By examining the more complete picture presented by the FFM however, it can be seen that the fuel age paradigm does not hold true for each of the indices. 332 Applying the findings to management Once the behaviour-age curves of interest have been developed for specific sites as above, the next step is to analyse the costs and benefits of each potential course of action. To illustrate the process, a standard approach to fuel management based on the fuel-age paradigm is to assign blocks a regular burning return interval such as Figure 8.5. Mean of the summer rate of spread for Snowgum forest on a 17 degree slope modelled with Gould et al (2007) and McArthur (1967) every ten years. If this was applied to the site in question for a 50year-old forest, the effect on first attack success for instance (Figure 8.4) would be distinctly counterproductive. In its long-unburnt state, first attack is likely to fail only 17% of the summer. Once the forest has been burnt however, first attack can be expected to succeed for 100% of the next summer, fail 59% of the summer by the 7th year, and maintain this rate until the fuels are 10 years old. The average fail rate for the 10 year period is 33%, which means that first attacks will fail twice as often than they would if nothing was done. By the same process, parallel attacks would fail nearly three times more often. In addition, the next prescribed burn would be scheduled for a point when flame heights are reaching their peak, so that containing the fire will be very difficult. Also, such use of frequent fire on the steep erodible granite soils will cause soil loss, which along with the degradation of the catchment will encourage even more dense growth of shrubs such as Bossiaea foliosa and Helichrysum species (Newman 1954). Greater shrub density and height extends the ability of fire crossing the gap into the canopy, so that the peak is delayed and peak rates of spread and flame heights are greater. Examination of the other indicators produces additional considerations. Where long distance spotting is a consideration, it may be valuable to burn the forest more frequently, although this would only be the case in forests with long bark streamers rather than in the smoothbarked E. niphophila forest. As stated, the empirical models, while giving widely differing results both upheld the fuel age paradigm when the rate of spread was examined. By examining the role of all other fuels and processes in the forest however, the FFM was able to identify that the fuel-age paradigm was 333 at best an over-simplification in this instance. Frequent fire did reduce the incidence of problem spotting, but it is not true to say that fire behaviour was less severe in young fuels. Overall, fires in old forest spread more slowly and were easier to control; so if these are priorities for fuel management, then the most effective course of action at this site is to exclude fire. This is not a “do nothing” approach; if the site has strategic significance, the fuel management plan should identify the need to protect it from fire where possible. The longunburnt nature of the site has become an asset in itself. It should be noted that the study is an analysis of regeneration following moderate to high intensity fire, and the experience of some is that low intensity fire in similar communities does not promote the same level of regrowth (Pers. Comms. B. Aitchison, 2008), but may potentially be used to reduce the cover of shrubs. While implementation of low intensity fire for a small isolated site may be possible under ideal conditions, application to a whole block requires that the necessary intensity is achieved in all treated areas across the full duration of the burn. Effective fuel management requires that these possibilities be investigated thoroughly along with the feasibility of their implementation. 8.22 Weather and climate interactions The model provides a framework for assessing the connection between any potential changes to a fuel array or its environment and the flammability of the array. Of particular interest is the potential impact of climate change on fire behaviour and fire regimes. Palaeoenvironmental studies have demonstrated that warming climates produce increased fire presence in parts of the landscape (e.g. Clark 1988, Sharp 1992), and modelling results have argued that the effects of severe fire weather such as drought or strong winds overwhelm any effects of fuel age in some environments (e.g. Bessie and Johnson 1995, Littell et al 2009). These results suggest that fire will have an increasing impact under future scenarios of climate change; however the specifics of this are vague. 8.221 Basic measures of fire weather Australian studies have focused on the broad impact of warming on the FFDI (Beer and Williams 1995, Hennessy et al 2005, Lucas et al 2007), demonstrating a clear projected increase in this which can be expected to translate into more severe fire conditions. The next step in this work should be the assignment of specific effects to different areas, such as the 334 impact of rising temperatures, changed rainfall timing or increased wind speeds. The FFDI combines all of these effects together; however as has been demonstrated, this is an oversimplification. Wind speed has an effect on dead fuel moisture and therefore fire occurrence, but the dominant effect is the tilting of the convection column and the resulting capacity of fire to propagate horizontally and vertically. Together with slope, wind is probably the most dominant force determining fire rates of spread and flame heights. The previous chapter demonstrated that when wind is not separated from the other weather parameters, the results are confused. As discussed in chapter three the relationship is complex, so understanding the effect of increased wind speeds needs to be examined in the context of the localised terrain and vegetation The effect of changed temperature and rainfall is significantly different to the effect of changed wind speeds. Both have impacts on fuel moisture, with temperature being a primary determinant of dead fuel moisture and therefore the ability for fires to spread. Because of these different responses shown by different communities, the FFM has the potential for both understanding the specific ways that forecast changes in the weather may impact fire behaviour, and for developing unique fire danger indices relevant to different areas. Two examples of this follow. Snowgum regrowth The data generated by the Monte Carlo analysis for six year-old Snowgum regrowth shown in Figure 7.3 of the previous chapter are shown graphed against the independent variables of FFDI, wind and dead fuel moisture content for rate of spread and flame height in Figures 8.6 to 8.11. These modelled results suggest that the FFDI is a poor predictor of fire danger for this community, with fast spreading fires and/or fires with large flame heights common under the lower fire danger indices. The effect of wind speed was far more pronounced; in particular, rate of spread was heavily influenced by wind speed (R2 = 0.47) . Dead Fuel Moisture content however had almost no effect on either spread rate or flame height, as its main influence is to permit the spread of fire through the surface fuels rather than to encourage faster or larger fires. As temperature 335 and humidity primarily affect dead fuel moisture, the combination of this with wind speed has rendered the FFDI a poor tool for predicting fire danger in this community. Modelling of the effects of changing climate that utilises the FFDI is unlikely to be very reliable in this instance. ROS vs FFDI Fh vs FFDI 35 25 20 25 20 Fh (m) ROS (km/h) 30 15 15 10 10 5 5 0 0 0 100 200 FFDI 300 0 400 100 200 FFDI 300 400 Figure 8.6. Results of a Monte Carlo analysis of six Figure 8.7. Results of a Monte Carlo analysis of six year-old Snowgum regrowth showing rate of spread year-old Snowgum regrowth showing flame height as as a function of the Forest Fire Danger Index for a function of the Forest Fire Danger Index for modelled results up to an FFDI of 400 modelled results up to an FFDI of 400 ROS vs Wind Fh vs Wind 25 45 40 20 30 Fh (m) ROS (km/h) 35 25 20 15 10 15 10 5 5 0 0 0 50 Wind speed (Km /h) 100 0 20 40 60 80 Wind speed (Km /h) 100 Figure 8.8. Results of a Monte Carlo analysis of six Figure 8.9. Results of a Monte Carlo analysis of six year-old Snowgum regrowth showing rate of spread year-old Snowgum regrowth showing flame height as as a function of wind speed for 1000 model results a function of wind speed for 1000 model results 336 Fh vs DFMC ROS vs DFMC 45 25 40 20 30 Fh (m) ROS (km/h) 35 25 20 15 10 15 10 5 5 0 0% 5% 10% 15% 0 0% DFMC % 5% 10% 15% DFMC % Figure 8.10. Results of a Monte Carlo analysis of six Figure 8.11. Results of a Monte Carlo analysis of six year-old Snowgum regrowth showing rate of spread year-old Snowgum regrowth showing flame height as as a function of dead fuel moisture content for 1000 a function of dead fuel moisture content for 1000 model results model results Mountain Ash forest The data generated by the Monte Carlo analysis for 112 year old Mountain Ash Community A shown in Figure 7.4 of the previous chapter are shown graphed against the independent variables of FFDI, wind and dead fuel moisture content for rate of spread and flame height in Figures 8.12 to 8.17. Unlike the Snowgum regrowth, the FFDI had a strong relationship to rate of spread (R2 = 0.49), however flame heights in general were unrelated and tall, passive crown fires occurred at low values. The FFDI did set some limits, so that sub-canopy fires were rare above values of 200. This is essentially a hypothetical concern as such values have not yet been recorded in Mountain Ash forests. The effect of wind on rate of spread was even stronger than for the Snowgum (R2 = 0.62), however it demonstrated a bifurcation at around 25km/h. Fires burning in conditions beyond this wind speed divided into two possible types, representing crown and sub-crown fires. While wind speed had little effect on sub-crown fires, crown fires steadily increased in speed as the wind increased. In the same way, sub-crown flame heights appeared to be almost 337 unaffected by wind speed, whereas the tallest flames in crown fires occurred between 20 and 60km/h where wind speeds facilitated longer plume pathways. Above this point, the additional buoyancy due to growth in flame length was less than the horizontal force of the wind, so flame heights decreased. As in the Snowgum community, dead fuel moisture content had no visible relationship to fire behaviour in this range of conditions. This raises the question as to why FFDI had such a strong relationship to rate of spread. Part of the answer lies in the strong relationship with wind, however temperature, humidity and KBDI also affect live fuel moisture contents, which in turn affect the ignitability of plants and therefore the capacity for flame to progress upward through the fuel strata to where the faster spread occurs. This overcomes the problem of bifurcation evident from wind speed alone, however the fit is weaker as the values have been spread by the artificial combination of indepent variables. These examples suggest that although FFDI may have value for predicting fire risk in some environments, it is not a universally applicable index and it there is value in separating the different weather components. ROS vs FFDI Fh vs FFDI 140 8 7 120 100 5 Fh (m) ROS (km/h) 6 4 3 80 60 2 40 1 20 0 0 200 400 600 FFDI 0 0 200 400 600 FFDI Figure 8.12. Results of a Monte Carlo analysis of Figure 8.13. Results of a Monte Carlo analysis of Mountain Ash Community A at 112 years since fire Mountain Ash Community A at 112 years since fire showing rate of spread as a function of the Forest Fire showing flame height as a function of the Forest Fire Danger Index for modelled results. Danger Index for modelled results. 338 ROS vs Wind Fh vs Wind 140 12 120 10 Fh (m) ROS (km/h) 100 8 6 80 60 4 40 2 20 0 0 0 20 40 60 80 Wind speed (Km /h) 100 0 20 40 60 80 Wind speed (Km /h) 100 Figure 8.14. Results of a Monte Carlo analysis of Figure 8.15. Results of a Monte Carlo analysis of Mountain Ash Community A at 112 years since fire Mountain Ash Community A at 112 years since fire showing rate of spread as a function of wind speed for showing flame height as a function of wind speed for 1000 model results 1000 model results ROS vs DFMC Fh vs DFMC 140 12 120 10 Fh (m) ROS (km/h) 100 8 6 80 60 4 40 2 20 0 0% 5% 10% 15% DFMC % 0 0% 5% 10% 15% DFMC % Figure 8.16. Results of a Monte Carlo analysis of Figure 8.17. Results of a Monte Carlo analysis of Mountain Ash Community A at 112 years since fire Mountain Ash Community A at 112 years since fire showing rate of spread as a function of dead fuel showing flame height as a function of dead fuel moisture content for 1000 model results moisture content for 1000 model results 339 8.222 Modelling the effects of climate change on fire While the effects of temporal weather conditions on fire behaviour are considered by other fire models to some degree, attention has been given to indicative measures such as the FFDI (Beer and Williams 1995, Hennessy et al 2005, Lucas et al 2007) and the individual impacts of specific weather changes such as those examined in the previous section have not been addressed. The FFM introduces the capacity to model the effects of specific changes in local weather conditions as well as considering more complex interactions and feedbacks such as those introduced by plant moisture dynamics or differential responses of species to stimuli (e.g. Freckleton 2004). The fuel moisture studies demonstrated that some species of plants can become moisture stressed following protracted periods of high temperatures or low humidity even when soils are moist, whereas others will dry as the soil dries so that even in cool moist air, dry soils can cause moisture stress and thereby increased flammability of trees such as E. pauciflora (Körner and Cochrane 1985). It follows that changes in weather that do not necessarily alter broad indices such as the FFDI may have profound effects on forest flammability; for instance lighter, more frequent rainfall may result in drier soils but the same final drought factor due to the effect of reduced time since rain on McArthur’s drought factor calculation (Noble et al 1980). The drier soils in turn may promote moisture stress in some species. In the same way while one day of extreme temperatures will affect the FFDI, the effect of a week of extreme temperatures cannot be captured by that index but will nevertheless severely affect the moisture of some species and thereby the flammability of the fuel array. Temperature and water balance variables can also have less obvious, long-term effects on forest flammability that may turn out to be significant feedback mechanisms in the carbon cycle. The distribution of plant species and communities in the Australian Alps for instance has been shown to be strongly affected by temperature components such as the occurrence of extreme frosts (e.g. Newman 1954, Fallon 2008). Ashton (1975) showed that the period of active growth of E. regnans and associated understorey species was limited by daily mean temperatures, so that a warming climate will produce a longer growing period. Leaf size also varied throughout the growing period, suggesting that a different distribution of temperatures will potentially mean a different mean leaf size, changed leaf flammability and altered shade and wind profiles due to the LAI. Ashton and Turner (1979) also related the light compensation point (the maximum amount of shade where plants will still grow) of E. 340 regnans to both temperature and atmospheric CO2 concentration, showing that plants are less shade tolerant in warmer temperatures but that this is mitigated by increased CO2 levels. The net effect of these factors needs to be examined in Mountain Ash forests to model the changes that can be expected in midstorey density of E. regnans saplings following low intensity fires, and its consequent impact on potential fire behaviour via shading, wind speed modification or a role as ladder fuels. Where changed weather conditions alter the flammability of the forest, the new conditions will favour different species. Whether these are introduced species or current minor components of the system, the rise to dominance of a new species can markedly change the flammability of the forest so that potentially, a new threshold may be crossed and a new community created. Higuera et al (2009) for instance found from palaeoenvironmental studies in Alaska that the establishment of Picea mariana in the region ca. 5500 yr BP increased the flammability of the entire landscape so that the reduction in fire frequency expected from the cooling trend of the time was overwhelmed by the effect of the plant’s dominance. The Forest Flammability Model provides a transparent and adaptable structure with which to examine such scenarios, so that informed decisions can be made where necessary for targeted mitigation programs. 341 Conclusions Fulfilment of objectives The introduction of this thesis outlined three objectives, these are addressed below. 1. To identify the factors governing the flammability (ignitability, combustibility and sustainability) of leaves and the way in which these affect the flammability of plants The flammability of leaves was quantified in detail in chapter four, and equations produced that accurately modelled ignition delay time, flame length and flame duration from individual leaves as measures of ignitability, combustibility and sustainability. The temperature of the endotherm was not examined in detail as models already exist to explain this, but a simple methodology was identified whereby this value can be estimated from the volatile oil content. The way in which the flammability of leaves affects that of the plants was discussed in chapters four and five, and the model framework was constructed in chapter five 2. To develop a fire behaviour model capable of examining the way that all fuels in an array affect wildland fire behaviour through flammability of plants and the geometry of the fuel array. The priority for this is the forest environment. 342 The models for leaf and plant flammability were built into a larger framework to model fire behaviour in chapter six. This was called the Forest Flammability Model (FFM), and incorporates sub-models for wind speed, live and dead fuel moisture, and soil moisture. The model was validated in chapter seven, and the indications so far are strong that this model provides significantly greater accuracy than three empirical Australian models which it was tested against. Errors in the modelling of the FFM were readily identifiable and attributable to specific causes. This transparency in the modelling process is a clear advantage over the empirical form, which provides an answer with no explanation and no means of improving the model. In addition to the standard rate of spread and flame height predictions expected of models, the FFM provides the following information which can be used for strategic planning and to identify weaknesses in the model: • Rate of spread in each stratum • Details of pulsing behaviour, i.e. the distance of pulses within each stratum, their velocity and the time taken for recovery after the pulse dies • The growth stage of the fire as described in section 7.31 • Type of crown fire • Flame height and length in each stratum • Average and maximum flame height and length in the full fuel array • Flame dimensions and rate of spread over time (fire growth) • Flame angles in each stratum as well as the full array. The FFM was tested against a range of operational standards for both prescribed burn planning and incident management, and performed better than the three empirical models provided for comparison. 3. To demonstrate the validity of this model as a tool in fuel and land management. Chapter eight provided by way of discussion and demonstration of examples evidence that the FFM is able to give specific, targeted insights into the flammability dynamics of a given 343 fuel array and environment. By providing this information, the model is able to equip fire managers to create more effective prescriptions, fire management plans and measures of fire risk. This can also reduce wasted and counter-productive effort and expenditure as well as minimising adverse and unnecessary environmental and ecosystem function damage. In addition, examples were provided where the FFM is uniquely able to address other aspects of bushfire risk management related to climate change. Specific outcomes Context and necessity of a new model Examination of the assertion that scientific knowledge of fuel dynamics and management is unnecessary in Australia due to an unbroken retention of effective traditional knowledge on the subject found this to be unsupported and highly questionable, that is that “the popular European-Australian understanding of Aboriginal fire does not accurately translate the oral tradition into practical management”. A second assumption that fire behaviour in Australia is sufficiently well understood by the current science revealed that while high-quality peerreviewed science did exist in the Australian literature, work of this standard examining the priority fuel arrays for the country is non-existent, and the material used in its place is not peer-reviewed or adequately validated. Further to this, the fuel-age paradigm that is grounded in this work has been shown to be an unworkable over-simplification that excludes the changes to a fuel array dictated by the ecology of the site. Consequently, it was established that a peer-reviewed model was needed that was capable of accommodating these differences and complexities. Development of a conceptual model Fire behaviour was described from physical principles and observations as a complex system that could not be effectively modelled with a simple empirical approach. A range of feedback mechanisms and non-linear behaviours were identified that were dependent upon the history of the flame progression, so that fire spread through an array was dependent upon repeated assessments of a critical state determined by the geometry of the fuels and their ignitability, combustibility and sustainability. A conceptual model of fire behaviour as a dynamical process was produced; the critical state test defined and the necessary parameters for utilising the model were identified. 344 Leaf flammability The following three chapters of the thesis (chapters four to six) were occupied with identifying models to fill these parameters. Leaf ignitability was described as having two parts – the minimum temperature of ignition which was dependent upon the chemistry of the leaf, and the ignition delay time which was determined by the temperature of exposure, the moisture content and the surface area to volume ratio of the leaf. These three factors were used to create a model for ignition delay time with an R2 of 0.90. The minimum temperature of ignition has been quantified in other work, but a simple index was created whereby this value may be estimated based on a field volatile oil content test. Leaf combustibility and sustainability were shown to have a very strong relationship to the dimensions and moisture of the leaf, so that other factors such as the energy content of the leaves may actually have only a very small influence on the matter. The combination of onesided leaf area with moisture was sufficient to produce a model for the flame length from a single burning leaf with an adjusted R2 of 0.87, and the combination of cross-section area with moisture produced a model for flame duration from a burning leaf supported by an outside heat source, with an adjusted R2 of 0.73.6%. Plant flammability In order to model fire spread through a plant, these models for leaf flammability were incorporated with geometric data to describe the plant crown, the leaf arrangements and the dimensions in relation to a flame. Existing models to describe flame angle using flame length and wind speed were utilised, and a theoretical model developed to explain the effect of slope on flame angle. Flame propagation through the plant was then described in a step-wise, history-dependent process which utilised a series of equations to describe the pathway of the plume through a plant with an idealised hexagonal crown shape. Ignition of new fuels was described by a series of equations and flow-charts based around the critical state test, and a model to incorporate preheating was introduced. Flame dimensions from a burning plant were related to leaf combustibility, flame merging theory and an empirical model developed to estimate the number of leaves within a given cross-section of plant. Flame length was varied with each time-step based upon these factors as affected by the number of new ignitions less the number of leaves that had extinguished. 345 Forest flammability Chapter six addressed the issue of translating plant flammability into the behaviour and spread of fire through an entire forest array. The forest arrangement and environment were found to have an external influence on fire behaviour beyond the direct flammability and geometry of the fuels, and the effects of these factors on wind speed, dead and live fuel moisture and soil moisture were examined. Wind speed was examined using an existing model supported by a new empirical model for estimating the Canopy Flow Index from the Leaf Area Index, which although having a significant relationship had a low level of precision. The influence of canopy shading and wind reduction on dead fuel moisture were discussed. Empirical models were produced to describe the moisture of six plant species based upon soil moisture and atmospheric conditions, with varying levels of success. Plant moisture in the species examined was found to be dominated by one of three processes – soil moisture, longterm atmospheric moisture and processes internal to the plant such as the shedding of leaves during periods of moisture stress. Implications of these three different mechanisms were discussed. In order to better model plant moisture, the value of the KBDI as a measure of soil moisture was examined and validated experimentally. The results had high statistical correlation and provided a strong model. With these external factors considered, the spread of fire through a forest was examined within surface fuels using an existing empirical model, and within the rest of the forest with an extension of the conceptual model developed in chapter three. Fire behaviour was modelled with rate of spread, flame length, flame angle and height and a number of other dimensions. Mathematical techniques were introduced to differentiate between the different mechanisms of fire spread, including the three forms of crown fire. Model validation The model was validated with six low intensity experimental fires and three high intensity unplanned fires. Validation included the measurement of the mean error as well as an assessment of the way the FFM and three other Australian empirical models performed as decision making tools for the development of burn prescriptions and in incident management. The comparison of models was necessary to determine the performance of the FFM against what has been accepted as the Australian standard to date. 346 Validation encompassed three components: 1. Extreme condition test 2. Predictive validation 3. Credibility analysis The extreme condition test demonstrated that the FFM did effectively model fire behaviour outside of the normal working conditions with reasonable boundaries, which was an advantage over the empirical models. Based on the mean error, the predictive analysis showed that the FFM was between 4 and 12 times more accurate than the other models when estimating rate of spread, and up to 12 times more accurate when estimating flame height. The improvements in rate of spread were statistically significant against two of the three empirical models (significance McArthur 1967 95.0%, Gould et al 2007a 90.0%). The flame height improvements were significant against McArthur (1967, 95.0%), but not against the other two models. The credibility analysis demonstrated that as a decision making tool, the FFM was better than Gould et al (2007a) and much better than McArthur (1967) for creating prescriptions, was very slightly less effective than McArthur (1962) for predicting the appropriate attack method for unplanned fires but better than Gould et al (2007a) and much better than McArthur (1967). Both the FFM and McArthur (1962) achived perfect scores when predicting initial attack success, which was better than Gould et al (2007a) and much better than McArthur (1962); but when used for forward planning the FFM far outperformed all other models by providing good predictions when they were most needed and by giving much less damaging errors when these did occur. Overall, the FFM was demonstrated to be more accurate than the other three models for this dataset, as well as providing a transparent platform that can be scrutinised to identify the submodels that may be responsible for any observed errors. This was a very satisfactory outcome, as the objective of developing a fire behaviour model was firstly to demonstrate the validity of the model structure and approach, identifying sub-models that needed further 347 work. That the model performed so well with only this first generation of sub-models was surprising. Implementation and applications Implementation of the FFM will require extensive study of fuels using the new parameters identified. Work was proposed to identify any parameters that could potentially be discarded or simplified and a works program for further research and application was outlined that would facilitate application of the model on a landscape scale. The value of the model as a tool for effective fuel management was established conceptually and as a case study. The FFM is not grounded in the flawed reasoning of the fuel-age paradigm, so it is capable of identifying where prescribed burning may be of assistance in fuel management and where it will be ineffective or counter-productive. As it does not employ a priori assumptions such as a correlation between flame height and rate of spread, the model is able to produce specific fire plans valid for strategic fuel management, including precise prescriptions. The model was used to examine trends in the influence of a small number of weather parameters on fire behaviour for two different forest communities. The demonstration outlined the different role of the parameters in different forest structures, and introduced the possibility of using the model to predict the impact of specific forecast changes due to climate change, and for developing site-specific fire danger indices that can more effectively capture the local risks. Further discussion outlined extensions of this work that would provide insights into more complex changes such as the impact of accumulated days of high temperature or increased CO2 levels on the flammability of the community. The model was shown to have the potential for capturing complex feedbacks between climate and flammability that may have important implications for fire impact and the carbon cycle. 348 REFERENCES Albini FA, (1981a) A model for the wind-blown flame from a line fire. Combustion and Flame 43, 155-174. Albini FA (1981b) A phenomenological model for wind speed and shear stress profiles in vegetation cover layers. Journal of Applied Meteorology 20(11), 1325-1335 Albini FA (1985) A model for fire spread in wildland fuels by radiation. Combustion Science and Technology 42, 229–258. doi:10.1080/00102208508960381 Albini FA (1986)Wildland fire spread by radiation – a model including fuel cooling by natural convection. Combustion Science and Technology 45, 101–113. doi:10.1080/00102208608923844 Albini FA (1996) Iterative solution of the radiation transport equations governing spread of fire in wildland fuel. Combustion, Explosion, and Shock Waves 32(5), 534–543. doi:10.1007/BF01998575 Alexander ME (1998) Crown fire thresholds in exotic pine plantations of Australasia. PhD Thesis, Australian National University, Canberra, ACT Anderson HE (1969) Heat transfer and spread. USDA Forest Service, Intermountain Research Station Research Paper INT-69. (Ogden, UT) Anderson HE (1970) Forest fuel ignitability. Fire Technology 6, 312-322 Anderson HE, Schuette RD, Mutch RW (1978) Timelag and equilibrium moisture content of Ponderosa Pine needles. USDA Forest Service, Intermountain Research Station, Research Paper INT-202. (Ogden, UT) Anderson W, Pastor E, Butler B, Catchpole E, Dupuy J-L, Fernandes P, Guijarro, M, Mendes-Lopes J-M, Ventura J (2006) Evaluating models to estimate flame characteristics for free-burning fires using laboratory and field data. In ‘Proceedings of the V International Conference on Forest Fire Research’ (Ed. Viegas DX) (Figueira Da Foz: Portugal) Andrews PL, Bevins CD, Seli RC (2008) ‘BehavePlus fire modeling system. Version 4.0 Users guide.’ USDA Forest Service Rocky Mountain Research Station, Research Paper RMRS-GTR-106WWW. (Ogden, UT) Ashton DH (1975) The seasonal growth of E. regnans F. muell. Australian Journal of Botany 23, 239-252 349 Ashton DH (1981) Fire in tall open forests (wet sclerophyll forests). In ‘Fire and the Australian biota’. (Eds AM Gill, GH Groves, IR Noble) pp. 339-366 (Australian Academy of Science: Canberra) Ashton DH (2000) The Big Ash Forest, Wallaby Creek, Victoria – changes during one lifetime. Australian Journal of Botany 48, 1-26 Ashton DH, Turner JS (1979) Studies on the light compensation point of Eucalyptus regnans F. Muell. Australian Journal of Botany 27, 589-607 ASTM (2007). http://www.astm.org/cgibin/SoftCart.exe/DATABASE.CART/REDLINE_PAGES/F1870.htm?E+mystore http://www.astm.org/cgibin/SoftCart.exe/DATABASE.CART/REDLINE_PAGES/F1870.htm?E+mystore Attiwill P, Packham D, Barker T, Hamilton I (2009) ‘The people’s review of bushfires 20022007 – Final Report 2009.’ (Vega: Vic) Augé RM (2001) Water relations, drought and vesicular-arbuscular mycorrhizal symbiosis. Mycorrhiza 11(1), 3-42 Australian Alps Liaison Committee (2006). Australian Alps Fire Response Database. Australasian Fire Authorities Council (1996) ‘Wildfire suppression 2, 2.29 Learning manual’, AFAC National Fire Curriculum. (Longman: Melbourne) Baeza M, DeLuís M, Raventós J, Escarré A (2002) Factors influencing fire behaviour in shrublands of different stand ages and the implications for using prescribed burning to reduce wildfire risk. Journal of Environmental Management 65(2), 199-208. doi: 10.1006/JEMA.2002.0545 Bak P, Tang C, Weisenfeld K (1988) Self organised criticality. Physical Review A 38(1), 364-374 Banks JCG (1982) The use of Dendrochronology in the interpretation of the dynamics of the snow gum forest. PhD Thesis, ANU: Canberra Banks JCG (1986) Fire and stand histories in subalpine forests on the Thredbo ski slopes, Kosciussko National Park, N.S.W. In: Ed.GC Jacoby, Procedings of the international symposium on ecological aspects of tree rings. Columbia University: Palisades Banks JCG (1988) A history of forest fire in the Australian Alps. In Ed. RG Good: The scientific significance of the Australian Alps. Australian Academy of Science and AALC: Canberra 350 Baum HR, Ezekoye OA, McGrattan KB, Rehm RG (1994) Mathematical modelling and computer simulation of fire phenomena. Theoretical and computational fluid dynamics 6(2-3), 125-139 Bavec M, Vukovi K, Grobelnik Mlakar S, Rozman , Bavek F (2007) Leaf area index in winter wheat: response on seed rate and nitrogen application by different varieties. Journal of Central European Agriculture 8(3), 337-342 Beer T , Williams A (1995) Estimating Australian forest fire danger under conditions of doubled carbon dioxide concentrations. Climatic Change 29, 169-188 Behm AL, Duryea ML, Long AJ, Zipperer WC (2004) Flammability of native understorey species in pine flatwood and hardwood hammock ecosystems and implications for the wildland-urban interface. International Journal of Wildland Fire 13, 355-365 Bellamy C (1993) The use of plants to minimise risk of bushfire damage to buildings. In: The Burning question: Fire management in NSW. Conference Proceedings, UNE Armidale: NSW Benson J (1994) The native grasslands of the Monaro region: Southern Tablelands of NSW. Cunninghamia 3(B), 609-650 Berry SL, Roderick ML (2002) CO2 and land-use effects on Australian vegetation over the last two centuries. Australian Journal of Botany 50, 511-531 Bessie WC, Johnson EA (1995) The relative importance of fuels and weather on fire behaviour in subalpine forests. Ecology 76(3), 747-762 Bilgili E, Saglam B (2003) Fire behaviour in Maquis fuels in Turkey. Forest Ecology and Management 184(1-3), 201-207. doi:10.1016/S0378-1127(03)00208-1 Billing P (1983) ‘Otways fire no. 22 – 1982/83’. NRE Fire Research Branch Report No. 20. NRE: Vic Boer MM, Sadler RJ, Wittkuhn RS, McCaw L, Grierson PF (2009) Long-term impacts of prescribed burning on regional extent and incidence of wildfires – evidence from 50 years of active fire management in SW Australian forests. Forest Ecology and Management 259, 132-142 Boer M, Stafford Smith M (2003) A plant functional approach to the prediction of changes in Australian rangeland vegetation under grazing and fire. Journal of Vegetation Science 14, 333-344 BOM (2009) ‘Kilmore Gap, Victoria. February 2009 daily weather observations’ www.bom.gov.au/climate/dwo/IDCJDW0000.pdf Border Mail (2003) ‘Team of 11 save resort town.’ Border Mail, 28 Jan 2003, P8. 351 Bowman DMJS, Kirkpatrick JB (1986) Establishment, suppression and growth of Eucalyptus delegatensis RT Baker in multiaged forests. I. The effects of fire on mortality and seedling establishment. Australian Journal of Botany 34, 63-72 Bowman DMJS, Wilson BA (1988) Fuel characteristics of coastal monsoon forests, Northern Territory, Australia. Journal of Biogeography. 15, 807-817 Bradstock RA, Auld TD (1995) Soil temperatures during experimental bushfires in relation to fire intensity: consequences for legume germination and fire management in south-eastern Australia. Journal of Applied Ecology 32, 76-84 Bradstock RA, Gill AM (1993) Fire in semi-arid, Mallee shrublands: size of flames from discrete fuel arrays and their role in the spread of fire. International Journal of Wildland Fire 3(1), 3-12 Bradstock RA, Kenny BJ (2003) An application of plant functional types to fire management in a conservation reserve in southeastern Australia. . Journal of Vegetation Science 14, 345-354 Bradstock RA, Tozer MG, Keith DA (1997) Effects of high frequency fire on floristic composition and abundance in a fire-prone heathland near Sydney. Australian Journal of Botany 45, 641-655 Brown MJ, Podger FD (1982) Floristics and fire regimes of a vegetation sequence from sedgeland-heath to rainforest at Bathurst Harbour, Tasmania. Australian Journal of Botany 30, 659-676 Buchanan M (2000) ‘Ubiquity: why catastrophes happen’. Three river press: NY Buckley A (1988) Fire behaviour and fuel reduction burning. Bemm River wildfire, October 1988. Victorian Department of Conservation and Environment Research Report No. 28. (Victoria, Australia) Burgan RE (1988) 1988 revisions to the 1978 National Fire Danger Rating System. USDA SE Forest Experiment Station Research Paper SE-273 (Asheville: CA) Burgan RE, Rothermel RC (1984) ‘BEHAVE: Fire behaviour prediction and fuel modelling system – FUEL subsystem.’ USDA Forest Service, Intermountain Forest and Range Experiment Station Research Paper INT-167. (Ogden, UT) Burrows ND (1984) Describing forest fires in Western Australia: a guide for fire managers. Forest Department of Western Australia: Perth. Tech. Pap. No. 9. Burrows ND (1994) Experimental development of a fire management model for Jarrah (Eucalyptus marginata Donn ex Sm.) forest. PhD thesis, Australian National University, Canberra 352 Burrows ND (1999a) Fire behaviour in jarrah forest fuels: 1. Laboratory experiments. CALMScience 3(1), 31-56 Burrows ND (1999b) Fire behaviour in jarrah forest fuels: 2. Field experiments. CALMScience 3(1), 57-84 Burrows ND (2001) Flame residence times and rates of weight loss of eucalypt forest fuel particles. International Journal of Wildland Fire 10, 137-143 Burrows ND, McCaw WL (1990) Fuel characteristics and bushfire control in Banksia low woodlands in Western Australia. Journal of Environmental Management 31, 229-236 Burrows ND, van Didden G (1991) Patch-burning desert nature reserves in Western Australia using aircraft. International Journal of Wildland Fire 1(1), 49-55 Burrows N, Ward B, Robinson A (1991) Fire behaviour in spinifex fuels on the Gibson Desert Nature Reserve, Western Australia. Journal of Arid Environments 20, 189-204 Bush Fire Front (2009) Avoiding Megafires in Australia. Occasional paper No. 5. (www.bushfirefront.com.au) Butler BW, Bartlette RA, Bradshaw LS, Cohen JD, Andrews PL, Putnam T, Mangan RJ (1998) Fire behaviour associated with the 1994 South Canyon Fire on Storm King Mountain, Colorado. USDA Forest Service, Rocky Mountain Research Station Research paper RMRS-RP-9. (Ogden, UT) Byram GM (1952) Thermal properties of forest fuels. USDA Division of Fire Research Interim Tech. Rep. AFSWP – 404. Byram GM (1959) Combustion of forest fuels. In: ‘Forest fire control and use.’ (Ed. KP Davis) (McGraw-Hill: NY) Byram GM and Jemison GM (1943) Solar radiation and forest fuel moisture. Journal of Agricultural Research 67, 149-176 Campbell JB (1987) ‘Introduction to remote sensing’. (Guidford: USA) Cary, G.J. and Golding, C.K. (2002). Fuel modelling for fire management. A review for the Northern Directorate of the NSW NPWS. ANU: Canberra Castro FX, Tudela A, Gabriel E, Montserrat D, Canyameres E, Segarra M (2006) Evolution of live fuel moisture in mediterranean forest. In ‘Proceedings of the V International Conference on Forest Fire Research’. 27-30 November 2006, Figuira da Foz, Portugal. (CD-ROM) (Elsevier BV: Amsterdam) Castro FX, Tudela A, Sebastià MT (2003) Modeling moisture content in shrubs to predict fire risk in Catalonia (Spain). Agricultural and Forest Meteorology 116, 49-59 353 Catchpole WR (1987) Heathland fuel and fire modelling. PhD Thesis, University of New South Wales, Canberra: Australia Catchpole WR, Bradstock RA, Choate J, Fogarty LG, Gellie N, McCarthy GJ, McCaw WL, Marsden-Smedley JB, Pearce G (1998) Co-operative development of equations for heathland fire behaviour, In ‘Proceedings of the 3rd International Conference on Forest Fire Research and the14th Conference on Fire and Forest Meteorology.’ (Ed. DX Viegas) pp631-645 (Coimbra: Portugal) Catchpole WR, Bradstock RA, Choate J, Fogarty LG, Gellie N, McCarthy GJ, McCaw WL, Marsden-Smedley JB, Pearce G (1999) Co-operative development of prediction equations for fire behaviour in heathlands and shrublands. Australian Bushfire Conference 1999. (Albury: NSW) Chason JW, Baldocchi DD, Huston MA (1991) A comparison of direct and indirect methods for estimating forest canopy leaf area. Agricultural Forest Meteorology 57,107–128 Cheney NP (1990) Fuel load or fuel structure: discussion paper for RWG 6. Unpublished paper presented to Australia Foresrty Council Research Working Group 6, Fire Management Meeting, Adelaide Cheney NP (1993) The changing fire environment. In ‘The burning question: fire management in NSW’ (UNE: Armidale) Cheney NP, Gould JS (1997) Letter to the editor – fire growth and acceleration. International Journal of Wildland Fire 7(1), 1-5 Cheney NP, Gould JS, Catchpole WR (1993) The influence of fuel, weather and fire shape variables on fire-spread in grasslands. International Journal of Wildland Fire 3(1) 31-44 Cheney NP, Gould JS, Catchpole WR (1998) Prediction of fire spread in grasslands. International Journal of Wildland Fire 8(1) 1-13 Cheney N P, Gould J S, Hutchings PT (1990) Sampling available fuel and damage to trees retained after thinning and burning. Management of Eucalypt re-growth in East Gippsland. Department of Conservation and Environment Victoria, Technical Report 9 Chigier NA, Apak G (1975) Interaction of multiple turbulent diffusion flames. Combustion Science and Technology 10(5-6) 219-231 Cionco RM (1972) A wind-profile index for canopy flow. Boundary-layer Meteorology 3, 255-263 Cionco RM (1978) Analysis of canopy index values for various canopy densities. Boundarylayer meteorology 15, 81-93 354 Cionco RM, Ohmsteade WD, Appleby JF (1963) A model for airflow in an idealized vegetative canopy. USAERDAA, Fort Huachuca Meteorology Research Notes 5. (Arizona: USA) Claridge AW, May TW (1994) Mycophagy among Australian mammals. Australian Journal of Ecology 19, 251-275 Clark DB, Olivas PC, Oberbauer SF, Clark DA, Ryan MG (2008) First direct landscape-scale measurement of tropical rain forest Leaf Area Index, a key driver of global primary productivity. Ecology letters 11, 163-172 Clark JS (1988) Effect of climate change on fire regimes in northwestern Minnesota. Nature 334, 233-235 Clarke MF (2008) Catering for the needs of fauna in fire management: science or just wishful thinking? Wildlife Research. 35, 385-394 Clarke WB (1860) ‘Researches in the southern goldfields of NSW’. (Reading and Wellbank: Sydney) Cochrane MA (2003) Fire science for rainforests. Nature 421, 913-119 Cochrane MA, Alencar A, Schulze MD, Souza CM, Nepstad DC, Lefebvre, P, Davidson EA (1999) Positive feedbacks in the fire dynamic of closed canopy tropical forests. Science 284(5421), 1832-1835 Cochrane MA, Messina JP (2006) Fire dynamics in tropical moist and wet forests. In ‘V International Conference on Forest Fire Research’. (Ed. DX Viegas). (Figuera da Foz: Portugal) Cocke AE Fulé PZ, Crouse JE (2005) Forest change on a steep mountain gradient after extended fire exclusion: San Francisco Peaks, Arizona, USA. Journal of Applied Ecology 42, 814-823 Constantin P, Foias C (1988) ‘Navier-Stokes Equations. Chicago lectures in mathematics.’ (University of Chicago Press: Chicago) Cornelissen JHC, Lavorel S, Garnier E, Díaz S, Buchmann N, Gurvich DE, Reich PB, ter Steege H, Morgan HD, van der Heijden MGA, Pausas JG, Poorter H (2003) A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Australian Journal of Botany 51, 335-380 Costermans LF (1994) ‘Native trees and shrubs of south-eastern Australia’. (Landsowne: Sydney) Costin AB (1954) ‘A study of the ecosystems of the Monaro region of New South Wales with special reference to soil erosion’. (Pettifer: Sydney) 355 Craze B, Hamilton GJ (1991) Soil physical properties. In ‘Soils – their properties and management’. (Eds PEV Charman, BW Murphy) pp. 147-164. (Sydney Uni. Press: Sydney) Croba D, Lalas D, Papadopoulos C, Tryfonopoulos D (1994) Numerical simulation of forest fire propagation in complex terrain. In ‘Proceedings of the 2nd International Conference on Forest Fire Research, Vol. 1’, November 1994, Coimbra, Portugal. (Ed. DX Viegas) pp. 491–500. (DX Viegas: Coimbra, Portugal) Cruz MG, Alexander ME, Wakimoto RH (2002) Predicting crown fire behaviour to support forest fire management decision-making. (Ed DX Viegas) ‘Forest fire research and wildland fire safety. Proceedings of the IV International Conference on Forest Fire Research, 2002 Wildland Fire Safety Summit, November 2002, Luso and Coimbra, Portugal’ pp. 18-23. (Millpress: Rotterdam) Cruz MG, Alexander ME, Wakimoto RH (2005). Development and testing of models for predicting crown fire rate of spread in conifer forest stands. Canadian Journal of Forest Research 35, 1626-1639 Cruz MG, Butler BW, Alexander ME, Forthofer, JM, Wakimoto RH (2006) Predicting the ignition of crown fuels above a spreading surface fire. Part I. Model idealisation. International Journal of Wildland Fire 15, 47-60 Cruz MG, Gould J (2009) Field-based fire behaviour research: past and future roles. In ‘18th World IMACS / MODSIM Congress’, (Cairns: Australia) CSIRO (2008) www.publish.csiro.au/pid/5993.htm Dasgupta S, Qu JJ, Hao X, Bhoi S (2008) Evaluating remotely sensed live fuel moisture estimations for fire behaviour predictions in Georgia, USA. Remote Sensing of Environment 108, 138-150 Delitti W, Ferran A, Trabaud L, Vallejo VR (2005) Effects of fire recurrence in Quercus coccifera L. shrublands of the Valencia region (Spain): I. plant composition and productivity. Plant Ecology 177, 57-70 de Mestre NJ, Rothermel RC, Wilson RA, Albini FA (1985) Radiation screened fire propagation. ADFA Maths Department Report No. 2, 85. University of NSW Australian Defence Force Academy: Canberra Dennison, PE, Moritz MA, Taylor RS (2008) Evaluating predictive models of critical live fuel moisture in the Santa Monica Mountains, California. International Journal of Wildland Fire 17, 18-27 356 Dibble AC, White RH, LeBow PK (2007) Combustion characteristics of north-eastern USA vegetation tested in the cone calorimeter: invasive versus non-invasive plants. Int.J.Wild.Fire 16, 426-443 Dickinson KJM, Kirkpatrick JB (1985) The flammability and energy content of some important plant species and fuel components in the forests of southeastern Tasmania. Journal of Biogeography 12, 121-134 Dimitrakopoulos AP (2001) A statistical classification of Mediterranean species based on their flammability components. International Journal of Wildland Fire 10, 113-118 Dimitrakopoulos AP, Bemmerzouk AM (2003) Predicting live herbaceous moisture content from a seasonal drought index. International Journal of Biometeorology 47, 73-79 Dodge JM (1972) Forest fuel accumulation – a growing problem. Science 177, 139-142 Dodson JR, De Salis T, Myers CA, Sharp AJ (1994) A thousand years of environmental change and human impact in the alpine zone at Mt Kosciuszko, New South Wales. Australian Geographer 25(1), 77-87 Dodson JR, Mooney SD (2002) An assessment of historic human impact on south-eastern Australian environmental systems, using late Holocene rates of environmental change. Australian Journal of Botany 50, 455-464 DSE (2009) Major bushfires in Victoria. www.dse.vic.gov.au Dunn AT (1989) The effects of prescribed burning on fire hazard in the chaparral: toward a new conceptual synthesis. In ‘Proceedings of the symposium on fire and watershed management’. pp. 23-29. USDA Forest Service General Technical Report PSW-109 Dupuy JL, Morvan D (2005) Numerical study of a crown fire spreading toward a fuel break using a multiphase physical model. International Journal of Wildland Fire 14(2), 141– 151. doi:10.1071/WF04028 Elkin AP (1938) ‘The Australian Aborigines’. (Angus and Robertson: Sydney) Ellis PF (2000) The aerodynamic and combustion characteristics of eucalypt bark – a firebrand study. PhD Thesis, Australian National University, Australia. Ellis RC (1985) The relationships among eucalypt forest, grassland and rainforest in a highland area in north-eastern Tasmania. Australian Journal of Ecology 10, 297-314 Ellis PF (2000) The aerodynamic and combustion characteristics of eucalypt bark – a firebrand study. PhD Thesis, Australian National University, Australia. England JR, Attiwill PM (2008) Patterns of growth and nutrient accumulation in expanding leaves of Eucalyptus regnans (Myrtaceae). Australian Journal of Botany 56, 44-50 357 Espelta JM, Verkaik I, Eugenio M, Lloret F (2008) Recurrent wildfires constrain long term reproduction ability in Pinus halapensis Mill. International Journal of Wildland Fire 17, 579-585 Etlinger MG, Beall FC (2004) Development of a laboratory protocol for fire performance of landscape plants. International Journal of Wildland Fire 13, 479-488 Fallon M (2008) Application of aerial photography for assessing vegetation change in northern Kosciuszko National Park. Hons. Thesis University of New South Wales. Sydney: NSW Fendell FE, Carrier GF, Wolff MF (1990) Wind-aided fire spread across arrays of discrete fuel elements. Department of Defence, Defence Nuclear Agency Technical Report DNATR-89-193. (Washington, DC) Fernandes PAM (2001) Fire spread prediction in shrub fuels in Portugal. Forest Ecology and Management 144(1-3), 67-74. doi:10.1016/S0378-1127(00)00363-7 Fernandes PM, Botelho HS (2003) A review of prescribed burning effectiveness in fire hazard reduction. International Journal of Wildland Fire 12, 117-128 Fernandes PM, Botelho HS, Loureiro C (2002) Models for the sustained ignition and behaviour of low-to-moderately intense fires in maritime pine stands. In ‘Proceedings of the IV International Conference on Forest Fire Research 2002 Wildland Fire Safety Summit’, 18-23 November 2002, Luso, Portugal. (Ed. DX Viegas) p. 98. (Millpress Science Publishers: Rotterdam, the Netherlands) Finkele K, Mills GA, Beard G, Jones DA (2006) National gridded drought factors and comparison of two soil moisture deficit formulations used in prediction of Forest Fire Danger Index in Australia. Australian Meteorological Magazine 55, 183-197 Finlayson HH (1936) ‘The red centre. Man and beast in the heart of Australia’. (Angus and Robertson: Sydney) Fishman GS, Kiviat PJ (1968) The statistics of discrete event simulation. Simulation 10, 185191 Flood J (1980) ‘The moth hunters. Aboriginal pre-history of the Australian Alps’. (Australian Institute of Aboriginal Studies: Canberra) Floyd AG (1966) Effect of fire upon weed seeds in the wet sclerophyll forests of northern New South Wales. Australian Journal of Botany 14, 243-256 Forestry Canada Fire Danger Group (1992) Development and structure of the Canadian Forest Fire Behaviour Prediction System. Forestry Canada Science and Sustainable Development Directorate, Information Report ST-X-3. (Ottowa, ON) 358 Fosberg MA, Deeming JE (1971) Derivation of the 1 and 10-hour timelag fuel moisture calculations for fire-danger rating. United States Department of Agriculture, Forest Service, Research Note RM-207. Rocky Mountain Forest Experimental Station, Fort Collins, Colorado. Frandsen WH, Andrews PL (1979) Fire behaviour in non-uniform fuel. United States Department of Agriculture, Forest Service, Research Paper INT-232. Intermountain Research Statioan, Ogden: UT Freckleton RP (2004) The problems of prediction and scale in applied ecology: the example of fire as a management tool. Journal of Applied Ecology 41, 599-603 Freund JE, Simon GA (1997) ‘Modern elementary statistics’. (Prentice Hall: NJ) Galton F (1864-5) Hereditary character and talent. MacMillan’s Magazine 11, 157-166, 318324 Galton F (1869) ‘Hereditary genius: an enquiry into its laws and consequences’. (Cosimo: NY) Gibbs L, Reid N, Whalley, RDB (1999) Relationships between tree cover and grass dominance in a grazed temperate Stringybark (Eucalyptus laevopinea) open-forest. Australian Journal of Botany 47, 49-60 Gibson M (2003) ‘Fighting bushfires at Guthega Village 26th January 2003.’ NSWFB Fire Investigation and Research Unit Task Force Commander report to NSW Fire Brigades. Lakemba: Australia Gill AM (1990) Fire behaviour in discontinuous fuels: a contract report to the New South Wales National Parks and Wildlife Service. CSIRO Plant Industry: Canberra Gill AM (2008) ‘Underpinnings of fire management for biodiversity conservation in reserves.’ Department of Sustainability and Environment Report No. 73. DSE: Vic Gill AM, Moore PHR (1996) Ignitability of leaves of Australian plants. Contract report to the Australian flora foundation. CSIRO: Canberra Gill AM, Moore PHR, Williams RJ (1996) Fire weather in the wet-dry tropics World Heritage Kakadu National Park, Australia. Australian Journal of Ecology 21, 302-308. Gill AM, Trollope WSW, MacArthur DA (1978) Role of moisture in the flammability of natural fuels in the laboratory. Australian Journal of Forest Research 8, 199-208 Gill AM, Zylstra P (2005) Flammability of Australian forests. Australian forestry 68(2), 8793 359 Gómez-González S, Cavieres LA (2009) Litter burning does not equally affect seedling emergence of native and alien species of the Mediterranean-type Chilean matorral. International Journal of Wildland Fire 18, 213-221 Gordon, N.D., McMahon, T.A., Finlayson, B.L. (1992). Stream hydrology: an introduction for ecologists. Wiley: England Gould JS, McCaw WL, Cheney NP, Ellis PF, Knight IK, Sullivan AL (2007a) Project Vesta – fire in dry Eucalypt forest: fuel structure, fuel dynamics and fire behaviour. EnsisCSIRO, Canberra ACT, and Department of Environment and Conservation, Perth WA Gould JS, McCaw WL, Cheney NP, Ellis PF, Matthews S. (2007b). Field Guide – Fuel assessment and fire behaviour prediction in dry eucalypt forest. Ensis-CSIRO, Canberra ACT, and Department of Environment and Conservation, Perth WA Gould RA (1971) Uses and effects of fire among the western desert Aborigines of Australia. Mankind 8, 14-24 Gower ST, Kucharik CJ, Norman JM (1999) Direct and indirect estimation of Leaf Area Index, fAPAR, and net primary production of terrestrial ecosystems. Remote Sensing Environment 70, 29-51 Gower ST, Vogel J, Norman JM, Kucharik CJ, Steele SJ, Stow TK (1997) Carbon distribution and aboveground net primary production in aspen, jack pine, and black spruce stands in Saskatchewan and Manitoba, Canada. Journal of Geophysical Research 102(D24), 29,029 – 29,041 Green DG (1989) Simulated effects of fire, dispersal and spatial pattern on competition within vegetation mosaics. Vegetation 82, 139-153 Greene DF, Johnson EA (1996) Wind dispersal of seeds from a forest into a clearing. Ecology 77(2), 595-609 Griffiths T (2001) ‘Forests of Ash – an environmental history’. (Cambridge: UK) Griffiths T (2002) How many trees make a forest? Cultural debates about vegetation change in Australia. Australian Journal of Botany 50, 375-389 Halsey RW (2004) ‘Fire, chaparral, and survival in southern California’. (Sunbelt Publications: San Diego) Hamilton CD (1964) The effect of Monterey Pine (Pinus radiate D. Don) on properties of natural Eucalyptus forest and woodland soils. M.Sc Thesis, Australian National University, Canberra, ACT Harden GJ (Ed.) (1990) ‘Flora of New South Wales, Volume 1’. Royal Botanic Gardens and Domain Trust. (NSWU: Sydney) 360 Harris J (1990) ‘One blood: two hundred years of Aboriginal encounter with Christianity’. P. 89 (Albatross: NSW) Hart SC, Classen AT, Wright RJ (2005) Long-term interval burning alters fine root and mycorrhizal dynamics in a ponderosa pine forest. Journal of Applied Ecology 42, 752-761 Hartland ES (1898) The High Gods of Australia. Folklore. 9(4). 290-329 Haynes K, Tibbits A, Coates L, Ganewatta G, Handmer J, McAneney J (2008) 100 years of Australian bushfire fatalities: exploring the trends in relation to the ‘stay or go’ policy. Report for the Bushfire CRC. Risk Frontiers, Bushfire CRC, RMIT University: Australia Hedger R (2003) Unpublished letter to the “Boys from Snowy River” gathering in Dalgety. Helms R (1893) Report on the grazing leases of the Mount Kosciusko plateau. Agricultural Gazette of NSW 4, 530-531 Helms R (1895) Anthropological notes. Proceedings of the Linnaean Society of NSW 20, 387-408 Hennessy K, Lucas C, Nicholls N, Bathols J, Suppiah R, Rickets J (2005) ‘Climate change impacts on fire-weather in south-east Australia’. CSIRO Marine and Atmospheric Research and Australian Bureau of Meteorology report, Australia Heskestad G (1998) Dynamics of the fire plume. Philosophical Transactions, Royal Society of London 356, 2815-2833 Higuera PE, Brubaker LB, Anderson PM, Hu FS, Brown TA (2009) Vegetation mediated the impacts of postglacial climate change on fire regimes in the south-central Brooks Range, Alaska. Ecological onographs. 79(2), 201-219 Hill RS (1982) Rainforest fire in western Tasmania. Australian Journal of Botany. 30, 583589 Hill RS, Read J (1984) Post-fire regeneration of rainforest and mixed forest in western Tasmania. Australian Journal of Botany 32, 481-493 Hill SJ, French K (2004) Potential impacts of fire and grazing in an endangered ecological community: plant composition and shrub and eucalypt regeneration in Cumberland Plain Woodland. Australian Journal of Botany 52, 23-29 Hobbs RJ, Atkins L (1990) Fire related dynamics of a Banksia woodland in south-western Western Australia. Australian Journal of Botany 38, 97-110 Hodgkinson KC, Oxley RE (1990) Influence of fire and edaphic factors on germination of the arid zone shrubs Acacia aneura, Cassia nemophila and Dodonaea viscosa. Australian Journal of Botany 38, 269-279 361 Hogenbirk, JC, Sarrazin-Delay CL (1995) Using fuel characteristics to estimate plant ignitability for fire hazard reduction. Water, air and soil Pollution 82, 161-170 Holling CS (1978) ‘Adaptive environmental assessment and management.’ (John Wiley and Sons: NY.) Homolová L, Malenovský Z, Hanuš J, Tomášková I, Dvo áková M, Pokorný R (2007) Comparison of different ground techniques to map leaf area index of Norway Spruce forest canopy. In Proceedings 10th International Symposium on Physical Measurements and spectral signatures in remote sensing (Eds Schaepman ME, Liang S, Groot NE, Kneubühler M), Davos, Switzerland - Davos (CH): ISPRS, 2007 (part 7/C50) Howitt AW (1904) ‘The native tribes of South-East Australia’. (Australian Institute of Aboriginal Studies: Canberra) Huffman KG, Welker JR, Sliepcevich CM (1967) Wind and interaction effects on freeburning fires. Technical Report, Defense Technical Information Centre: USA Hull HM (1875) ‘Lecture on the Aborigines of Australia’. (Mercury Steam Press: Hobart) Hutchings PT, Oswald KM (1975) Litter fall and litter accumulation in eucalypt forests of the Australian Capital Territory. In ‘Proceedings of the 3rd Australian Specialist Conference in Soil Biology’. (Adelaide: SA) IPCC (2007). ‘Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change’ (Eds. S Solomon, D Qin, M Manning, Z Chen, M Marquis, KB Averyt, M Tignor, HL Miller) 996 pp. (Cambridge University Press, Cambridge, United Kingdom and New York, NY) Jones R (1975) The Neolithic Palaeolithic and hunting gardeners: man and land in the antipodes. In ‘’Quaternary studies’. (Eds. RP Suggate, MM Cresswell) pp, 21-34. (Royal Society of New Zealand: Wellington). Cited in Nicholson PH (1981) Fire and the Australian Aborigine – an enigma. In ‘Fire and the Australian biota’. (Eds. AM Gill, RH Groves, IR Noble) pp. 55-76. (Australian Academy of Science: Canberra) Jury WA, Gardner WR, Gardner WH (1991) Water flow in saturated soils. In ‘Soil physics’ 5th edn. (Wiley: New York) Karafyllidis I, Thanailakis A (1996) A model for predicting forest fire spread using cellular automata. Ecological Modelling 99, 87-97 Keeley JE (1986) Seed germination patterns of Salvia mellifera in fire-prone environments. Oecologia (Berl.) 71, 1-5 362 Keeley JE (1987) Role of fire in seed germination of woody taxa in California chaparral. Ecology 68, 434-443 Keeley JE, Fotheringham CJ (2001) Historic fire regime in Southern California shrublands. Conservation Biology 15, 1536-1548 Keeley JE, Fotheringham CJ, Morais M (1999) Reexamining fire suppression impacts on brushland fire regimes. Science 284, 1829-1832 Keetch JJ, Byram GM (1968) A drought index for forest fire control. U.S.D.A Forest Service, Southeastern Forest Experiment Station Research Paper SE-38. (Asheville, NC) Keith DA, Holman L, Rodoreda S, Lemmon J, Bedward M (2007) Plant functional types can predict decade-scale changes in fire-prone vegetation. Journal of Ecology 95, 1324-1337 Kershaw AP, Clark JS, Gill AM, D’Costa DM (2002) A history of fire in Australia. In ‘Flammable Australia: the fire regimes and biodiversity of a continent’. (Eds. RA Bradstock, JE Williams, AM Gill) (Cambridge: UK) King NK, Vines RG (1969) Variation in the flammability of the leaves of some Australian forest species. CSIRO Div. App. Chem. Report Kirkpatrick JB, Bridle KL, Wild AS (2002) Succession after fire in alpine vegetation on Mount Wellington, Tasmania. Australian Journal of Botany 50, 145-154 Knight IK, Sullivan AL (2003) A semi-transparent model of bushfire flames to predict radiant heat flux. International Journal of Wildland Fire 13(20), 201-207 Körner Ch, Cochrane PM (1985) Stomatal responses and water relations of Eucalyptus pauciflora in summer along an elevational gradient. Oecologia 66, 443-455 Ku-ring-gai City Council (2008) Bushfire management policy. (Ku-ring-gai, NSW) Langton M (1999) Fire! The Australian experience. ‘Proceedings of the National Academics Forum seminar, University of Adelaide’. (Snap: SA) Lavarch M (1997) Bringing them home: Report of the National Inquiry into the separation of Aboriginal and Torres Strait Islander children from their families. Attorney General’s report. (Commonwealth: Canberra) Lee H, Slatton KC, Roth B, Cropper WP (2009) Prediction of forest canopy light interception using three-dimensional airborne LiDAR data. International Journal of Remote Sensing 30, 189-207 Leigh JH, Wimbush DJ, Wood DH, Holgate MD, Slee AV, Stanger MG, Forrester RI (1987) Effects of rabbit grazing and fire on a subalpine environment. I. Herbaceous and shrubby vegetation. Australian Journal of Botany 35, 433-464 363 Leonard JE, Bowditch PA (2003) Findings of studies of houses damaged by bushfire in Australia. 3rd International Wildland Fire Conference, 3-6 October. Sydney: Australia Linn RR (1997) A transport model for prediction of wildfire behaviour. Los Alamos National Laboratory, Science Report LA-13334-T. (Los Alamos, NM) Littell JS, McKenzie D, Peterson DL, Westerling AL (2009) Climate and wildfire area burned in western U.S. ecoprovinces 1916-2003. Ecological Applications 19(4). 10031021 Liu N, Liu Q, Lozano JS, Shu L, Zhang L, Zhu J, Deng Z, Satoh K (2009) Global burning rate of square fire arrays: experimental correlation and interpretation. Proceedings of the Combustion Institute, 32(2), 2519-2526 Loudermilk EL, Hiers JK, O’Brien JJ, Mitchell RJ, Singhania A, Fernandez JC, Cropper WP Jr., Slatton KC (2009) Ground-based LIDAR: a novel approach to quantify fine-scale fuelbed characteristics. International Journal of Wildland Fire 18, 676-685 Lucas C, Hennessy K, Mills G, Bathols J (2007) ‘Bushfire weather in southeast Australia: recent trends and projected climate change impacts.’ Bushfire CRC and Australian Bureau of Meteorology consultancy report prepared for the Climate Institute of Australia. Bushfire CRC: Melbourne Lunt ID (2002) Grazed, burnt and cleared: how ecologists have studied century-scale vegetation changes in Australia. Australian Journal of Botany 50, 391-407 Lyons JW (1985) ‘Fire’. (Scientific American Library: USA) Malamud BD, Morein G, Turcotte, DL (1998) Forest fires, an example of self-organized critical behaviour. Science 281, 1840-1842 Mankin JB, O’Neill RV, Shugart HH, Rust BW (1977) The importance of validation in ecosystem analysis. (Ed. G Innis ‘New directions in the analysis of ecological systems. Part I: The society for computer simulation.’ (LaJolla: CA), pp. 63-71 Marsden-Smedley JB, Catchpole WR (1995a) Fire behaviour modelling in Tasmanian Buttongrass moorlands I. Fuel characteristics. International Journal of Wildland Fire 5(4), 203-214 Marsden-Smedley JB, Catchpole WR (1995b) Fire behaviour modelling in Tasmanian Buttongrass moorlands II. Fire Behaviour. International Journal of Wildland Fire 5(4), 215-228 Martin RE, Gordon DA, Gutierrez MA, Lee DS, Molina DM, Schroeder RA, Sapsis DB, Stephens SL, Chambers M (1994) Assessing the flammability of domestic and wildland 364 vegetation. In ‘Proc. 12th Conf. on Fire and Forest Meteorology Jekyll Island, GA’ pp. 130-137. (Society of American Foresters: Bethesda, MD) Mason R (2003) Aboriginal people and fire. ‘Kosciuszko Today’. (NPWS: Jindabyne) Matthews S (2006) A process-based model of fine fuel moisture. International Journal of Wildland Fire 15, 155-168 Matthews S, Gould J, McCaw L (2010) Simple models for predicting dead fuel moisture in eucalyptus forests. International Journal of Wildland Fire 19, 459-467 McAlpine RS (1995) Testing the effect of fuel consumption on fire spread rate. International Journal of Wildland Fire 5(3), 143-152 McArthur AG (1962) ‘Control burning in eucalypt forests.’ Commonwealth Australia Forestry Timber Bureau Leaflet 80, Canberra McArthur AG (1966a) Weather and grassland fire behaviour. Commonwealth Australia Forestry Timber Bureau Leaflet 100, 23p McArthur AG (1966b) ‘Notes on fire behaviour in eucalypt forests.’ Commonwealth Australia Forestry Timber Bureau Report produced for NSW NPWS, Canberra McArthur AG (1967) Fire behaviour in Eucalypt forests. Forestry and Timber Bureau Leaflet 107, 9th Commonwealth Forestry Conference¸ India. McArthur AG, Douglas DR, Mitchell LR (1966) ‘The Wandiloo fire, 5 April 1958: fire behaviour and associated meteorological and fuel conditions.’ Forest and Timber Bureau, Forest Research Institute, Leaflet No. 98, Canberra McCarthy GJ, Tolhurst KG (1997) Effectiveness of firefighting first attack operations. Victorian Department of Conservation and Natural Resources Research Report no. 45, Fire research CFTT, Department of Natural Resources and Environment, Victoria McCarthy GJ, Tolhurst KG (2001) Effectiveness of broadscale fuel reduction burning in assisting with wildfire control in parks and forests in Victoria. Victorian Department of Natural Resources and Environment Research Research report n. 51, Fire Research Science Centre, Department of Natural Resources and Environment, Victoria McCarthy GJ, Tolhurst KG, Chatto K (1999) ‘Overall fuel hazard guide’ (NRE: Victoria) McCarthy MA, Gill AM, Bradstock RA (2001) Theoretical fire-interval distributions. International Journal of Wildland Fire 10, 73-77 McCaw LM (1997) Predicting fire spread in Western Australian mallee-heath shrubland . PhD thesis, University of New South Wales, Canberra. 365 McCaw WL, Gould JS, Cheney NP (2008) Quantifying the effectiveness of fuel management in modifying wildfire behaviour. Proceedings ‘Fire Environment and Society’15th AFAC Conference”. (Adelaide,SA) McCaw WL, Mills G, Sullivan A, Hurley R, Ellis P, Matthew S, Plucinski M, Pippin B, Boura J (2009) Research results from February 7th 2009 Victorian fires findings on: Fire behaviour investigation. Bushfire Cooperative Research Centre Final Report. (Melbourne: VIC) McColl JG (1966) Accession and decomposition of litter in spotted gum forests. Australian Forestry 30, 191-198 Mees R, Strauss D, Chase R (1993) Modeling wildland fire containment with uncertain flame length and fireline width. International Journal of Wildland Fire. 3(3), 179-185 Melville City Council (2008) City of Melville Bushfire Management Strategy. (Melville, WA) Mitler HE, Steckler KD (1995) Spread – a model of flame spread on vertical surfaces. NISTIR 5619. Building and Fire Research Laboratory, U.S. Department of Commerce, Gaithersburg: MD Montgomery KR, Cheo PC (1971) Moisture and salt effects on fire retardance in plants. American Journal of Botany 56, 1028-1032 Mooney SD (2004) ‘Looking back as a way forward… pre-historic fire in the high altitude ecosystems of mainland south-eastern Australia’. University of NSW report prepared for the NSW Department of Environment and Conservation. (DEC: Sydney) Mooney SD, Harrison SP, Bertlein PJ, Daniau A-L, Stevenson J, Brownlie KC, Buckman S, Cupper M, Luly J, Black M, Colhoun E, D’Costa D, Dodson J, Haberle S, Hope GS, Kershaw P, Kenyon C, McKenzie M, Milliams N (2010) Late Quaternary fire regimes of Australasia. Quaternary Science Reviews doi:10.1016/j.quascirev.2010.10.010 Moriarty, O. M. (1993). Conservation or destruction of natural resources of alpine regions of South Eastern Australia. Publisher Unknown Morvan D (2007) A numerical study of flame geometry and potential for crown fire initiation for a wildfire propagating through shrub fuel. International Journal of Wildland Fire 16, 511-518 Mount AB (1972) The derivation and testing of a soil dryness index using runoff data. Tasmania Forestry Commission Bulletin No. 4, (Hobart:Tas) Muñoz MR, Fuentez ER (1989) Does fire induce shrub germination in the Chilean Matorral? Oikos 56, 177-181 366 Neal L (1988) “It doesn’t snow like it used to – memories of Monaro and the Snowy Mountains”. (Stateprint: Ultimo) Nelson RM Jr. (1980) Flame characteristics for fires in southern fuels. USDA Forest Service, Southeast Forests Experimental Station Research Paper SE-205. (Asheville, NC) Nelson RM (1984) A method for describing equilibrium moisture content of forest fuels. Canadian Journal of Forest Research 14, 597-600 Nelson RM Jr., Adkins CW (1986) Flame characteristics of wind-driven surface fires. Canadian Journal of Forest Research 16, 1293-1300 Newman JC (1954) Burning on subalpine pastures. Journal of Soil Conservation Service NSW. 10, 135-40 Noble, JC. (1991). Behaviour of a very fast grassland wildfire on the riverine plain of southeastern Australia. International Journal of Wildland Fire 13, 189-196 Noble IR, Bary GAV, Gill AM (1980) McArthur’s fire danger meters expressed as equations. Australian Journal of Ecology 5, 201-203 Northcote KH (1979) ‘A factual key for the recognition of Australian soils.’ (Rellim: Glenside) NPWS (No date). Incident Control System. Fire management – strategies – tactics – tasks matrix. CJB: ICS/PLAN V10. NPWS: Hurstville NPWS (1998). Bioregional Strategy for NSW Rainfall Layer. ESOCLIM annual rainfall model output. NPWS (2004). Snowy Mountains Region fire history database. Unpublished GIS database NPWS (2005). 25m Digital Elevation Model. GIS database NPWS (2007). Snowy Mountains Region Forest Ecosystems Database. NSW Health (2009) Coronary heart disease deaths by health. www.nsw.health NSW RFS (2003) ‘Linescan Run 11, Kosciuszko South Complex, 1530 hrs’ NSW RFS (2007). Draft bush fire risk management planning guidelines for bush fire management committees. RFS: Sydney Olsen CM (1980) An evaluation of the Keetch-Byram Drought Index as a predictor of foliar moisture content in a chaparral community. In ‘Proceedings of the 6th Conference on forest and fire meteorology’,22-24 April 1980 (Seattle) Olsen JS (1963) Energy storage and the balance of producers and decomposers in ecological systems. Ecology 44, 322-331 Ong WJ (2002) ‘Orality and Literacy: The technologizing of the word’. (Routledge: London and New York) 367 Ottino JM (2003) Complex systems. AIChE Journal 49(2), 292-299 Parker WJ, LeVan SL (1989) Kinetic properties of the components of Douglas-fir and the heat of combustion of their volatile pyrolysis products. Wood and Fibre Science 21(3), 289-305 Péch G (1989) A model to predict the moisture content of Reindeer Lichen. Forest Science 35, 1014-1028 Peet GB, Williamson AJ (1968) An assessment of forest damage from the Dwellingup fires in Western Australia. In ‘5th conference of the Institute of Foresters of Australia’ (Perth) Pellizzaro G, Duce P, Ventura A, Zara P (2007) Seasonal variations of live fuel moisture content and ignitability in shrubs of the Mediterranean Basin. International Journal of Wildland Fire 16, 633-641 Penman H (1948) Natural evaporation for open water, bare soil and grass. Proceedings of the Royal Society of London. A193, 120-146 Pepper P, De Araugo T (1985) ‘The Kurnai of Gippsland’. P. 11 (Hyland: Aust) Philpot CW (1970) Influence of mineral content on the pyrolysis of plant materials. Forest Science 16(4), 461-471 Pinker RT, Moses JF (1981) On the Canopy Flow Index of a tropical forest. Boundary Layer Meteorology 22, 313-324 Plucinski MP (2003) The investigation of factors governing ignition and development of fires in heathland vegetation. PhD Thesis, University of New South Wales, Canberra Possingham HP, Comins HN, Noble IR (1995) The fire and flammability niches in plant communities. Journal of Theoretical Biology. 174, 97-108 Powers WL (1922) Field moisture capacity and wilting point of soils. Soil Science. 14(2), 159-166 Price O, Bradstock RA (2010) The effect of fuel age on the spread of fire in sclerophyll forest in the Sydney region of Australia. International Journal of Wildland Fire 19. 35-45 Prober SM, Lunt ID, Thiele KR (2008) Effects of fire frequency and mowing on a temperate, derived grassland soil in south-eastern Australia. International Journal of Wildland Fire 17, 586-594 Pryor LD (1939) The bushfire problem in the Australian Capital Territory. Australian Forestry 4(1), 33-38 Pulsford IF (1991) History of disturbances in the White Cypress Pine (Callitris glaucophylla) forests of the lower Snowy River valley. Masters Thesis, ANU: Canberra 368 Putnam AA 1965 A model study of wind-blown free-burning fires. In ‘Proceedings of the 10th Symposium (International) on Combustion.’ pp. 1039-1046, Combustion Institute, Pittsburgh, Pa., 17-21 Aug. 1964, (University of Cambridge, Cambridge: UK) Raeder-Roitzsch JE, Phillips ME (1958) An erosion and siltation survey in the upper Tooma catchment. Snowy Mountains Authority Report No. SS13. Snowy Mountains Authority: Cooma Raison RJ, Woods PV, Khanna PK (1986) Decomposition and accumulation of litter after fire in sub-alpine Eucalypt forests. Australian Journal of Ecology 11, 9-19 Resnick M (1994) ‘Turtles, termites and traffic jams’. MIT Press (Cambridge: Massachusetts) Richards R, Cary G, Bradstock R (2001) The sensitivity of snowgum to fire scarring in relation to Aboriginal landscape burning. Bushfire 2001: Australasian bushfire conference: Christchurch Ringrose C, Neilsen WA (2005) Growth response of Eucalyptus regnans and soil changes following periodic fertilisation. Soil Science Society of America Journal 69, 1806-1812 Robinson N, Rees D, Reynard K, MacEwan R, Dahlhaus P, Imhof M, Boyle G, Baxter N (2003) A Land Resource Assessment of the Corangamite Region. Department of Primary Industries Rothermel RC (1972) A mathematical model for predicting fire spread in wildland fuels. USDA Forest Service, Intermountain Forest and Range Experiment Station Research Paper INT-115. (Ogden, UT) Rothermel RC (1983) How to predict the spread rate and intensity of forest and range fires. USDA Forest Service, Intermountain Research Station Paper INT-143. (Ogden, UT) Rothermel RC (1984) Fire behaviour considerations of aerial ignition. In ‘Proceedings of prescribed fire by aerial ignition’. (Missoula: Montana) Rothermel RC (1991) Predicting behaviour and size of crown fires in the northern Rocky Mountains. USDA Forest Service, Intermountain Forest and Range Experiment Station Research Paper INT-438. (Ogden: UT) Rothermel RC, Deeming JE (1980) Measuring and interpreting fire behaviour for correlation with fire effects. USDA Forest Service, Intermountain Forest and Range Experiment Station Research Paper INT-93. (Ogden: UT) Rothermel RC, Wilson RA, Morris GA, Sackett SS (1986) Modelling moisture content of fine dead wildland fuels: input to the BEHAVE fire prediction system. USDA Forest Service, Intermountain Research Station, Research Paper INT-359. (Ogden, UT) 369 Rudolph LR (1993) Vegetation and bushfire. Part II. A system for the assessment of the behaviour of vegetation as applied to the landscaping around buildings in bushfire areas. Landscape Australia 15, 115-123 Ryan DG, Ryan JE, Starr BJ (1995) ‘The Australian Landscape – observations of explorers and early settlers’. (Murrumbidgee Catchment Management Committee: NSW) Rykiel EJ (1996) Testing ecological models: the meaning of validation. Ecological Modelling 90, 229-244. Sabo KE, Sieg CH, Hart SC, Bailey JD (2009) The role of disturbance severity and canopy closure on standing crop of understorey planyt species in ponderosa pine stands in nortrhern Arizona, USA. Forest Ecology and Management 257, 1656-1662 Salter PJ, Williams JB (1967) The influence of texture on the moisture characteristics of soils. IV. A method for estimating the available water capacities of profiles in the field. Journal of Soil Science 18, 174-181 Salter PJ, Williams JB (1969) The influence of texture on the moisture characteristics of soils. V. Relationships between particle size composition and moisture contents at the upper and lower limits of available-water. Journal of Soil Science 20, 126-131 Sampson DA, Vose JM, Allen HL (2001) A conceptual approach to stand management using Leaf Area Index as the integral of site structure, physiological function and resource supply. http://coweeta.uga.edu/publications/1233.pdf Sandberg DV, Riccardi CL, Schaaf MD (2007) Reformulation of Rothermel’s wildland fire behaviour model for heterogeneous fuelbeds. Canadian Journal of Forest Research 37, 2438-2455 Sargent RG (1984) ‘A tutorial on verification and validation of simulation models’. Procedings of the 1984 winter simulation conference IEEE 84CH2098-2 (Eds. S Sheppard, U Pooch, D Pegden), pp. 115-122 Schlichting H (1979) ‘Boundary layer theory’. (McGraw-Hill: NY) Schmoldt DL, Peterson DL, Keane RE, Lenihan JM, McKenzie D, Weise DR, Sandberg DV (1999) Assessing the effects of fire disturbance on ecosystems: a scientific agenda for research and management. USDA Forest Service General Technical Report PNW-GTR455. Portland, OR. Schwilk DW, Keeley JE, Knapp EE, McIver J, Bailey JD, Fettig CJ, Fielder CE, Harrod RJ, Moghaddas JJ, Outcalt KW, Skinner CN, Stephen SL, Waldrop TA, Yaussy DA, 370 Youngblood A (2009) The national Fire and Fire Surrogate study: effects of fuel reduction methods on forest vegetation structure and fuels. Ecological Applications 19(2), 285-304 Serrano L, Ustin SL, Roberts DA, Gamon GA, Penuelas J (2000) Deriving water content of chaparral vegetation from AVIRIS data. Remote Sensing of Environment 74, 570-581 Seydack AHW, Bekker SJ, Marshall AH (2007) Shrubland fire regime scenarios in the Swartberg mountain range, South Africa: implications for fire management. International Journal of Wildland Fire 16, 81-95 Shafir H, Jin FJ, Lati Y, Cohen M, Alpert P (2008) Wind channelling by the Dead Sea Wadis. The Open Atmospheric Science Journal. 2, 139-152 Shalders G (2007). Fuel accumulation for reserves affected by the 2003 fires within the South West Slopes Region. Enfireonment report to National Parks and Wildlife South West Slopes Region. Enfireonment, Moruya Sharp AJ (1992) A Palaeoenvironmental reconstruction and assessment of human impact in the alpine region of Kosciusko National Park. New South Wales, Australia. Honours Thesis. University of New South Wales, :Kensington:NSW Sharples JJ, McRae RHD, Weber RO, Gill AM, (2009) A simple method for assessing fuel moisture content and fire danger rating. ‘Proceedings of the 18th World IMACS / MODSIM Congress, Cairns, Australia 13-17 July 2009’. http://mssanz.org.au/modsim09 Simard AJ (1968) The moisture content of forest fuels – 1. A review of the basic concepts. Canadian Department of Forest and Rural Development, Forest Fire Research Institute, Information Report FF-X-14, Ottawa, Ontario Smith W (1994) The ecology and taxonomy of the Southern Lined Earless Dragon (Tympanocryptis lineata pinguicolla). Honours Thesis, Australian National University, Canberra, ACT Sneeuwjagt RJ, Peet GB (1985) Forest fire behaviour tables for Western Australia. Department of Conservation and Land Management, (Perth, WA) Starr C, Taggart R (1995) ‘Biology – the unity and diversity of life’. 7th edition. (Wadsworth: USA) Stenberg P, Linder S, Smolander H, Flower-Ellis J (1994) Performance of the LAI-2000 plant canopy analyser in estimating leaf area index of some Scots pinestands. Tree Physiology 14, 981–995. Stephens SL, Moghaddas JJ (2005) Experimental fuel treatment impacts on forest structure, potential fire behaviour, and predicted tree mortality in a mixed conifer forest. Forest Ecology and Management 215, 21-36 371 Stephens SL, Moghaddas JJ, Edminster C, Fiedler CE, Haase S, Harrington M, Keeley JE, Knapp EE, McIver JD, Metlen K, Skinner CN, Youngblood A (2009) Fire treatment effects on vegetation structure, fuels and potential fire severity in western U.S. forests. Ecological Applications 19(2), 305-320 Stephenson NL (1999) Reference conditions for giant sequoia forest restoration: Structure, process and precision. Ecological Applications 9, 1253-1265 Stevenson J (2001) Inquest into the deaths of Mark Douglas Cupit, Claire Wynne Dean, George Allan Fitzsimmons and Eric Furlan and inquiry into fire at Mt Kuring-gai National Park. Coroner’s Report, (Westmead: NSW) Steward FR (1970) Prediction of the height of turbulent diffusion buoyant flames. Combustion Science and Technology 2(4), 203-212 Steward FR (1971) A mechanistic fire spread model. Combustion Science and Technology 4, 177-186 Stocks BJ (1989) Fire behaviour in immature jack pine. Canadian Journal of Forest Research 17:.80-86. Stott P (1986) The spatial pattern of dry season fires in the savanna forests of Thailand Strahler AN (1952) Hypsometric (area-altitude) analysis of erosional topography. Bulletin of the Geological Society of America 63, 1117-1142 Stretton LEB (1939) ‘Report of the Royal Commission to enquire into the causes and measures taken to prevent the bushfires of January, 1939, and to protect life and property.’ (Government Printer: Melbourne) Sturtevant BR, Zollner PA, Gustafson EJ, Cleland DT (2004) Human influence on the abundance and connectivity of high-risk fuels in mixed forests of northern Wisconsin, USA. Landscape Ecology 19, 235 – 253 Sullivan AL (2009a) Wildland surface fire spread modelling, 1990-2007. 1. Physical and quasi-physical models. International Journal of Wildland Fire 18, 349-368 Sullivan AL (2009b) Wildland surface fire spread modelling, 1990-2007. 2. Empirical and quasi-empirical models. International Journal of Wildland Fire 18, 369-386 Sullivan AL, Ellis PF, Knight IK (2003) A review of radiant heat flux models used in bushfire applications. International Journal of Wildland Fire 12, 101-110 Susott RA (1982) Characterization of the thermal properties of forest fuels by combustible gas analysis. Forest Science 2, 404-420 Tan B, Midgley S, Douglas G, Short L (2007) ‘A methodology for assessing bushfire attack.’ Report by the NSW RFS. (RFS: Sydney) 372 Tanskanen H, Granstrom A, Larjavaara M, Puttonen P (2007) Experimental fire behaviour in managed Pinus sylvestris and Picea abies stands of Finland. International Journal of Wildland Fire 16(4), 414-425. doi:10.1071/WF05087 Taylor C (2009) Victorian February fires: a report on driving influences and land tenures affected. Report to the Black Saturday Royal Commission by the Victorian National Parks Association, Australian Conservation Foundation and the Wilderness Society. Taylor GI (1961) Fire under influence of natural convection. In ‘International Symposium on the use of models in fire research’ pp.10-31. (Ed. Berl WG) (National Academy of Sciences National Research Council, Washington DC) Tewarson A (1980) Heat release rate in fires. Fire and materials 4, 185-191 Thomas PH (1963) The size of flames from natural fires. ‘Ninth symposium (International) on combustion’. pp 844-859. (Academic: NY) Thomas PH, Baldwin R, Heselden AJM (1965) Buoyant diffusion flames: some measurements of air entrainment, heat transfer and flame merging. ‘Tenth symposium (International) on combustion’. pp 983-996. The combustion institute Thomas SC, Winner WE (2000) Leaf Area Index of an old-growth Douglas-fir forest estimated from direct structural measurements in the canopy. Canadian Journal of Forest Research 30(12), 1922-1930 Thompson DF (1949) ‘Economic structure and the ceremonial exchange cycle in Arnhem Land’. (Macmillan: Melbourne). Cited in Nicholson PH (1981) Fire and the Australian Aborigine – an enigma. In ‘Fire and the Australian biota’. (Eds. AM Gill, RH Groves, IR Noble) pp. 55-76. (Australian Academy of Science: Canberra) Tolhurst KG (2010) ‘FireBehaviourCalcs_SouthernAustralia_20100403’. Excel spreadsheet Tolhurst KG, Chong DM (2009) Incorporating the effect of spotting into fire behaviour spread prediction using Phoenix. ‘Proceedings of the AFAC conference, 2009’. Gold Coast: Queensland. Trabaud L (1976) Inflammabilite et combustibilite des principales especes des garrigques de la region mediterraneenne. Oecologia Plantarum 11, 117-136 Tran C, Pyrke A (1999) ‘Wind speed reduction factors in a heathy woodland.’ Australian Bushfire Conference, Albury Tran C, Wild C, Zubrinich T (2001) Testing Mutch's hypothesis in SE Qld: plant flammability revisited. ‘Bushfire 2001: Australasian bushfire conference’, (Christchurch: NZ) 373 Turner JS, Crocker RL, Evans JW, Costin AB (1957) ‘A report on the condition of the high mountain catchments of New South Wales and Victoria.’ (Australian Academy of Science: Canberra) Vaillant NM, Fites-Kaufman JA, Stephens SL (2009) Effectiveness of prescribed fire as a fuel treatment in Californian coniferous forests. International Journal of Wildland Fire 18, 165-175 Van Wagner CE (1967) Calculations of forest fire spread by radiation. Canadian Department for Rural Development Publication No. 1185 Van Wagner CE (1969) Combined effect of sun and wind on surface temperature of litter. Canadian Forest Service Information Report PS-X-10. Petawawa Forest Experimental Station, (Chalk River, Ontario) Van Wagner CE (1972) Equilibrium moisture contents of some fine forest fuels in eastern Canada. Canadian Forestry Service, Information Report PS-X-36. Petawawa Forest Experimental Station, (Chalk River, Ontario) Van Wagner CE (1973) Height of crown scorch in forest fires. Canadian Journal of Forest Research 3: 373-378 Van Wagner CE (1976) Conditions for the start and spread of crown fire. Canadian Journal of Forest Research 7, 23-34 Van Wagner CE, Pickett TL (1985) Equations and FORTRAN program for the Canadian Forest Fire Weather Index System. Canadian Forestry Service Technical Report 33. 18 pages. Vaz GC, Andre JCS, Viegas DX (2004) Fire spread model for a linear front in a horizontal solid porous fuel bed in still air. Combustion Science and Technology 176(2), 135-182. doi:10.1080/00102200490255343 Vega JA, Fernández C, Pérez-Gorostiaga P, Fonturbel T (2008) The influence of fire severity, serotiny and post-fire management on Pinus pinaster Ait. Recruitment in three burnt areas in Galicia (NW Spain). Forest Ecology and Management. 256, 1596-1603 Veihmeyer FJ, Hendrickson AH (1931) The moisture equivalent as a measure of the field capacity of soils. Soil Science. 32(3), 181 Victorian Bushfires Royal Commission (2009a) Interim Report. http://www.royalcommission.vic.gov.au/getdoc/208e8bcb-3927-41ec-8c232cdc86a7cec7/Interim-Report Victorian Bushfires Royal Commission (2009b) Submissions: http://www.royalcommission.vic.gov.au/Hearing-Schedule/SPAusnet-Submission 374 Victorian Bushfires Royal Commission (2009c) Attachment 1: Evidence of fire behaviour. RESP.3000.005.0240 Viegas DX (2006) Analysis of eruptive fire behaviour. Proceedings of the Fifth International Conference on Forest Fire Research, 27-30th November 2006, Figueira da Foz, Portugal Viegas DX (2009) (Ed.) ‘Recent forest fire related accidents in Europe.’ JCR Scientific and Technical Reports. European Commission Joint Research Centre Institute for Environment and Sustainability. (Ispra: Italy) Viegas DX, Neto LPC (1991) Wall shear-stress as a parameter to correlate the rate of spread of a wind induced forest fire. International Journal of Wildland Fire 1(3), 177-188 Viegas DX, Piñol J, Viegas MT, Ogaya R (2001) Estimating live fine fuels moisture content using meteorologically-based indices. International Journal of Wildland Fire 10, 223-240 Viegas DX, Pita LP, Ribeiro L, Palheiro P (2005) Eruptive fire behaviour in past fatal accidents. ‘VIII. International Wildland Fire Safety Summit’ (Missoula: MT) Viney NR (1991) A review of fine fuel moisture modelling. International Journal of Wildland Fire 1(4), 215-234 Viney NR (1992) Modelling fine fuel moisture content. PhD Thesis, University of New South Wales, Canberra Viney NR, Catchpole EA (1991) Estimating fuel moisture response times from field observations. International Journal of Wildland Fire 1(4), 211-214 Wahren C-HA, Papst WA Williams RJ (1994) Long-term vegetation change in relation to cattle grazing in subalpine grassland and heathland on the Bogong High Plains: An analysis of vegetation records from 1945 to 1994. Australian Journal of Botany 42, 607639 Walker J (1981) Fuel dynamics in Australian vegetation. In ‘Fire and the Australian biota’. (Eds AM Gill, RH Groves, IR Noble) pp. 101-127 Wang Y, Cionco R (2007) ‘Wind profiles in gentle terrains and vegetative canopies for a three-dimensional wind field (3DWF) model.’ Army Research Laboratory, Report ARLTR-4178. Adelphi: Maryland Washington State University (2009). Soil texture. http://www.bsyse.wsu.edu/cropsyst/soil_editor/manual/texture.htm Weber RO (1989) Analytical models for fire spread due to radiation. Combustion and Flame. 78, 398-408 Weber RO (1990a) Modelling fire spread through fuel beds. Prog. Energy Combust. Sci. 17, 67-82 375 Weber RO (1990b) A model for fire propagation in fuel arrays. Mathematical and Computer Modelling, 13(12), 95-102 Weber RO (1990c) The moist strand and the burning bed. Canadian Journal of Forest Research, 20, 845-848 Weber RO, de Mestre NJ (1990) Flame spread measurements on single Ponderosa Pine needles: Effect of sample orientation and concurrent external flow. Combustion Science and Technology, 70, 17-32. Weber RO, de Mestre NJ (1991) Buoyant convection: A physical process as the basis for fire modelling. In ‘Proceedings of the Conference on Bushfire Modelling and Fire Danger rating systems’. (Eds NP Cheney, AM Gill) pp. 43-50 (CSIRO Div. Forestry: Australia) Weber RO, Gill AM, Lyons RRA, Moore PHR, Bradstock RA, Mercer GN (1995) Modelling wildland fire temperatures. CALM Science Supplement 4, 23-26 Weber RO, Sidhu HS (2006) A Dynamical Systems Model for Fireline Growth with Suppression ANZIAM J 47, C462-C474 Weick KE (1993) Tool retention and fatalities in wildland fire settings: conceptualising the naturalistic. In ‘Linking expertise and naturalistic decision making’. (Eds E Salas, G Klein) pp. 321-336. (Lawrence Erlbaum and Associates: NJ) Weick KE (2001) The collapse of sensemaking in organisations: The Mann Gulch disaster. Administrative Science Quarterly. 38, 628-652 Weise DR, Biging GS (1996) Effects of wind velocity and slope on flame properties. Canadian Journal of Forest Research 26(10), 1849–1858. Weng W, Kamikawa D, Fukuda Y, Hasemi Y, Kagiya K (2004) Study on flame height of merged flame from multiple fire sources. Combustion Science and Technology 176(14), 2105-2123 WFLC. (2009). The true cost of wildfire in the western U.S. Western Forestry Leadership Coalition: US Whiteman CD (2000) ‘Mountain meteorology – fundamentals and applications’ (Oxford: New York) Williams RJ, Ashton, DH (1987) Effects of disturbance and grazing by cattle on the dynamics of heathland and grassland communities on the Bogong High Plains, Victoria. Australian Journal of Botany 35, 413-31 Wilson AAG (1993) Elevated fuel guide. CNR Fire Management Branch Research Report No. 35. (Melbourne: Vic) 376 Wimbush DJ, Costin AB (1979a) Trends in vegetation at Kosciuszko. I. Grazing trials in the subalpine zone, 1957 – 1971. Australian Journal of Botany 27, 741-787 Wimbush DJ, Costin AB (1979b) Trends in vegetation at Kosciuszko. II. Subalpine range transects, 1959 – 1978. Australian Journal of Botany 27, 789-831 Wotton MB (2009) A grass moisture model for the Canadian Forest Fire Danger Rating System. In ‘Proceedings of the 8th symposium on forest and fire meteorology, 13th October 2009’. US Wotton BM, McAlpine RS, Hobbs MW (1999) The effect of fire front width on surface fire behaviour. International Journal of Wildland Fire 9(4), 247-253 Wright JL (1965) Evaluating turbulent transfer aero-dynamically within the microclimate of a cornfield. Ph.D. thesis, Cornell University Ithaca, New York, 174 pp. Xanthopoulos G (1990) Development of a wildland crown fire initiation model. PhD thesis, University of Montana, Missoula, MO Xanthopoulos G, Wakimoto RH (1993) A time to ignition – temperature – moisture relationship for branches of three western conifers. Canadian Journal of Forest Research 23(2), 253-258 Yebra M, Chuvieco E (2009) Linking ecological information and radiative transfer models to estimate fuel moisture content in the Mediterranean region of Spain: Solving the ill-posed inverse problem. Remote Sensing of Environment 113, 2403-2411 Yebra M, Chuvieco E, Riaño D (2008) Estimation of live fuel moisture content from MODIS images for fire risk assessment. Agricultural and Forest Meteorology 148, 523-536 Young M (2005) ‘The Aboriginal people of the Monaro – a documentary history.” (DEC:NSW) Zalba SM, Cuevas YA, Boó RM (2008) Invasion of Pinus halapensis Mill. Following a wildfire in an Argentine grassland nature reserve. Journal of Environmental Management 88, 539-546 Zedler PH, Seiger LA (2000) Age mosaics and fire size in chaparral: a simulation study. In ‘2nd interface between ecology and land development in California’. (Eds JE Keeley, M Baer-Keeley and CJ Fotheringham), pp. 9-18, US Geological Survey Open-File Report 0062 Zylstra P (2006a) ‘Fire History of the Australian Alps’. (AALC: Canberra) Zylstra P (2006b) Physical properties determining flammability in sclerophyllous leaves and flame propagation within shrubs and trees in the Australian Alps. Forest Ecology and Management. 234S, S81 377 Zylstra P (2007) ‘Surface fuel accumulation and implications for bushfire risk management in Kosciuszko National Park’. Kosciuszko National Park Snowy Mountains Region Fire Unit report to the Kosciuszko National Park Fire Management Strategy Team. (Jindabyne, NSW) Zylstra P (2008) Assessing bushfire risk at Cabramurra. NSW Department of Environment and Climate Change, Kosciuszko National Park Fire Management Strategy Team report to Snowy Hydro Ltd. (Jindabyne, NSW) 378 Appendices 379 Appendix I: Symbol A a C c cd Cp d Df Dh DP DPD f fh Hb Hc He Hp Ht Hu INSx IEx IMx K KBDI LAI Lb m mfc ms N Nb Nl Ox Oxn Oy Oyn m P P P P Ph Pp Symbols used in the model Definition Units One-sided surface area of a leaf Proportionality constant Combustibility Cross section area of a leaf Time taken for an active crown fire to re-initiate after a pulse has expired Percent cover of stratum Diameter of a fuel particle Flame depth Horizontal distance to which a crown will receive sufficient heat from a ground fire to maintain combustion Dew point Dew point depression Distance along a plume from the flame base to the edge of the plant being burnt Flame height Height of the base of a plant or stratum above the stem base Height of the bottom centre of a plant crown Height of the bottom edge of a plant crown Height of the top centre of a plant crown Height of the top edge of a plant crown Height of the top of a plant or stratum above the stem base Horizontal depth ignited in the near surface stratum Horizontal depth ignited in the elevated stratum Horizontal depth ignited in the midstorey stratum Maximum temperature within a flame Keetch Byram Drought Index Leaf area index Burnt length of a leaf Moisture Field capacity moisture of a soil Soil moisture Number of fuel units burning Number of branches that will ignite within a given depth of ignition Number of leaves in a branch or clump Flame origin measured horizontally to the right from the plant base Ox at time step n Flame origin measured vertically above the plant stem base Oy at time step n Moisture content Temperature of the endotherm Distance from the flame base to the point being heated Minimum of Ppa and Spp Distance into a plant to which burning leaves have extinguished in one time-step Proportion of moisture removed from a leaf via preheating Length of the plume pathway or from to a point horizontally equal with Hp mm2 mm3s-1 mm2 s % mm m m o C C cm m cm cm cm cm cm cm cm cm cm o C mm o mm % ODW % ODW % ODW cm cm cm cm % ODW o C cm cm cm % cm 380 Ppa Ppi Px Px Pt R Rs RH Sb Sbs Sf SFA Sp Spp Sx T t te tm tr trp TP u uen uH W w Wb ws Wx Wy xi yi Available plume pathway Length of the plume pathway ignited Actual plume pathway through a plant crown, one of 6 possibilities The difference between Pp and Px in a plant crown Total penetration Rate of spread Rate of spread in surface fuels Relative humidity Distance from the edge of one branch or clump to the edge of a neighbouring branch Clump separation averaged for a stratum Distance from one plant centre to the next Silica-free ash content Distance along a plume from the edge of the burning plant to the point being heated Potential penetration: distance along a plume to the point where T = P Distance along the slope between the base of a plant and the base of the flame heating the plant Temperature Length of the time-step (default is 1) Time elapsed since ignition Time taken for a measurement Flame residence time Flame residence time in a plant Temperature from flame length at separation P Wind speed Wind speed adjusted for rate of spread Wind speed at 10m or the top of a forest canopy Width of a leaf Width of a plant crown Width of a branch or clump Surface fuel load Distance of a flame origin from the left hand side of a plant crown Distance of a flame origin above the lower left corner of a plant crown Horizontal distance ignited Vertical distance ignited cm cm cm cm cm km/h km/h % cm cm m % ODW cm cm cm o C s s s s s o C km/h km/h km/h mm cm cm t/Ha cm cm cm cm Greek Symbols 1 p TII TIII 1 Entry point of a flame to a plant, having the coordinates Ox, Oy Canopy flow index Vertical distance from the ground surface to the flame origin minus Oy Vertical distance from the base of a surface flame to the base of a plant stem Distance along a plume from the base of a surface flame to a point equal in height with the base of a plant stem Change in temperature in zone II Change in temperature in zone III Saturation vapour pressure Angle of a plume to vertex a (fig. 5.1) in a plant cm cm cm o C C o degrees 381 2 3 4 c f fl fp g m ml ms neg Angle of a plume to vertex c (fig. 5.1) in a plant Angle of a plume to vertex e (fig. 5.1) in a plant Angle of a plume to vertex f (fig. 5.1) in a plant Surface area to volume ratio Flame angle Flame angle of a line fire as affected by slope Flame angle of a point fire as affected by slope Slope Flame length Area merged flame length Longitudinal merged flame length Merged flame length weighted by species Length of overlap or space to be subtracted when combining flame lengths from more than one stratum Ignition delay time degrees degrees degrees Degrees Degrees Degrees Degrees mm or m m m m m s 382 Appendix II: Main equations and assumptions used in the Forest Flammability Model Description Surface flame length Flame angle due to wind Equation λ = 2.4 Rs + 0.036ws θ f = tan θ fl = 90 1 − Point flame angle due to slope θ fp = 90 1 − Separation between backing surface heat source and potential fuel Flame residence time in surface fuels Plume temperature in zone III Plume temperature in zone II Distance from the edge of a burning fuel to the point beyond a flame where the temperature is ToC above the ambient Pα = (O Equ. 5.30 θg 90 Equ. 5.33 Bθ g 90 Pα = Pα = Equ. 5.26 (O Equ. 3.13 − O y ) + (O xn − O x ) 2 yn 2 + δ )sin (90 + θ g ) yn Source Equ. 6.39 from Burrows (1999a) Equ. 5.29, From Van Wagner (1973) & Byram (1959) 0.5 δ = −O x tan (θ g ) Vertical difference due to slope Separation between surface heat source and potential fuel 66.048λ 2.17 0.345 u3 −1 Line flame angle due to slope Separation between heat source and potential fuel within plant Notes Equ. 3.11 sin (θ f − θ g ) (O yn [ + δ )sin (90 − θ g ) sin (180 − θ f ) + θ g Equ. 3.12 ] Equ. 5.13 from Burrows (2001) Tr = 0.871d 1.875 ∆T III = C Sp 2 ∆T II = Ke (−α ( Pα − f ) ) Sp = C ∆T Developed on a scale of metres Equ. 5.16 from Weber et al (1995) Developed on a scale of metres Equ. 5.14 from Weber et al (1995) Developed on a scale of metres Equ. 5.23, rearranged from Weber et al (1995) 383 Ignition delay time Temperature of the endotherm Flame duration in burning leaves with an outside heat source Preheating ψ= 97805.26T 2.10 m θc Equ. 4.10 + 6452280.04T − 2.40 P = 354 – 13.9ln SFA - 2.91 (ln SFA) 2 Equ. 4.1 from Philpot 1970 Tr = 1.37c + 0.0161m − 0.027 Equ. 4.21 ψ n = ψ (1 − {Ph1 + Ph 2 ...Phn −1 }) Equ. 5.61 Depth of ignition S pp = Equ. 5.34 C − Pα P −T Critical state test TλPα ≥ P, (Tr − t ) ≥ ψ , (Tr − t e ) ≤ t Equ. 3.15 Critical state test with preheating TλPα ≥ P, (Tr − t ) ≥ ψ n , (Tr − t e ) ≤ t Equ. 5.62 Number of branches ignited Number of leaves per branch Equ. 5.65 W + Sb ,1 + N bi N b = min ( N be − N bi ) b Wb WO N l = 0.88 b s Sl Equ. 5.69 1.18 Pi = {IF[t (t n − t1 ) < Tr , Pi1 ,0]} + {IF[t (t n − t 2 ) < Tr , Pi 2 ,0]} + ... + {IF[t (t n − t n−1 ) < Tr , Pin−1 ,0]} Equ. 5.70 Flame length from one burning leaf λ = 17.5.3 A − 0.277m − 0.27 Equ. 4.19 Flame length from one burning leaf below critical moisture λ = 5.25 A Equ. 4.18 Critical moisture for single flame length model 17.53 A − 5.25 A − 0.27 m= 0.277 Equ. 4.20 Length of a plume pathway alight λm = λ . N0.4 Area merged flame length Maximum separation for flame merging Longitudinal merged flame length Plant flame residence time σ =w [ 2 3 λ w λml = (λm + aPpi ) + Ppi 4 [ ] 4 0.25 ] Trp = t min (O xn , O yn ) Approximate model, developed from scanty data Approximate model, developed from scanty data Equ. 5.75, from Gill (1990) Mitler and Steckler (1995) Equ. 5.76 Equ. 5.74, from Gill (1990) Equ. 5.77 384 Limits for flame connection tan −1 H u − Oy ≥ (θ f − θ g ) ≥ tan S f − Ox Calculation of clump separation in a stratum S bs = Wind speed adjustment under plant canopy −1 H b − Oy Equ. 3.3 S − Ox Equ. 6.1 S f − w + Sb w 1+ Wb + S b Eq. 6.7, from Cionco et al (1963) z −1 u = u H exp γ H Canopy Flow Index γ = 1.031 ln( LAI ) + 1.681 Equ. 6.8 Canopy Flow Index – linear model γ = 0 .269 LAI + 1 .90 Equ. 6.9 Plant Leaf Area Index Equ. 6.12 Vp LAI p = Percent cover of stratum [w + (S f ] − w) 2 Equ. 6.40 S f − w − Ox cos(θ f − θ g ) Equ. 6.41 hu − hb Pp = sin(θ f − θ g ) Plume pathway in stratum for backing fires Pp = Subtraction factor for combining flames between strata Mid-flame wind speed for contributing flame Equ. 6.54 w2 Pα = Plume pathway in stratum for head fires × Alb π 0 .5 w 2 Cp = Separation between heat source and potential fuel within stratum 3 4 Wb + 0.5S b π 3 2 λ neg = u = u1 λ1 (λ1 + λ2 + ... + λn ) Equ. 6.42 hu − hb sin ((180 − θ f ) − θ g ) + u2 abs (O xu − f hl ) sin θ fl λ2 (λ1 + λ2 + ... + λn ) + ... + u n Equ. 6.43 λn Equ. 6.44 (λ1 + λ2 + ... + λn ) Length of flame from combined species λms = {η sp1λsp1 + η sp 2 λ sp 2 + ... + η spn λ spn }N 0.4 Equ. 6.45 Flame height from a burning f h = λ sin (θ f − θ g ) + O y Equ. 6.48 385 stratum Wind speed adjusted for ROS u en = u − Rn−1 Equ. 6.49 Horizontal distance to which a forest canopy is heated ahead of a ground fire He + Hc − O yl 2 Dh = O xl + tan (θ fl − θ g ) Equ. 6.50 Total penetration Pt = Surface fire ROS [y − (O i2 (0.0245u 2.22 s R= Time taken for a crown fire to redevelop ) (0.003 + 0.1000922m Equ. 6.38 from Burrows (1999a) d ) 1000 Rate of spread Rate of spread due to pulsing Equ. 6.55 + 0.071 Rs = Rate of spread from x ignited 2 ] − O y ) + [xi 2 − (O x 2 − O x )] 2 y2 Ppi cos(θ f − θ g ) Equ. 3.16 Tm R= 0.036 xin cos(θ g )t n Equ. 6.56 3.6c p Equ. 6.51 R= cd + cr Equ. 6.52 c p = O x max + I NSx + I Ex + I Mx ASSUMPTIONS Fire spread through discontinuous fuels occurs via convective heat transfer Assumption 3.1 Flames and the resulting plumes can be simplified to a central vector without significant loss of accuracy. The effects of turbulent diffusion and flame depth can be approximated using simple means rather than using complex fluid dynamics. Assumption 3.2 Plume angle remains equal to flame angle. Assumption 5.1 386 The slope is flat and constant for that point and can be approximated by a tilted plane. Assumption 5.2 The effect of slope on flame angle is relevant at low wind speeds only. At higher wind speeds the blocking of air entrainment by the slope is insignificant in comparison to the overall air movement. In this way, the slope-determined flame angle represents a maximum flame angle possible for that slope; or for fires burning down-slope and against the wind, a minimum. Assumption 5.3 The proportion of moisture removed from a leaf during a single time-step is equal to the proportion of the ignition delay time taken up by the time step. Assumption 5.4 The depth to which an ignition isotherm penetrates a branch or clump of leaves represents the proportion of the clump that will ignite. Assumption 5.5 The convective plume will only ignite those branches immediately in its path, branches on either side may ignite afterward but will not contribute to the maximum flame height. Assumption 5.6 Where one burning particle or clump ignites others, the flames of all burning particles or clumps will merge. Assumption 5.7 All of the flames from leaves burning in a single clump will merge. Assumption 5.8 387 As plants burning in close proximity to one another affect the rate at which their neighbours burn through radiative heat transfer and their influence on air movement, plant combustion will become synchronised so that flames from individual plants reach their maximums at approximately the same time. Assumption 6.1 If the plume from one burning plant stratum aligns with the flame from the stratum above it, the two flames may be treated as one. Assumption 6.2 Where rates of spread in near-surface, elevated or midstorey fuels are greater than those in the strata below them, radiation and falling burning debris from the higher strata will maintain the surface fire beneath them so that the heat source remains relatively constant. Assumption 6.3 Where leaves are burning in the absence of an external heat source producing a temperature of 100oC or greater, the residence time of flame in the leaves will be reduced to one second. Assumption 6.4 New strata do not ignite before the stratum beneath produces its maximum flame length. Assumption 6.5 388 Appendix III: Ignition delay times of leaves Tests at 220oC Dry weight 3 Tin %moist M/SAV 13.33 Wet weight 3.1 0 3.3% 0.3 13.33 3.1 3 0 3.3% 0.3 8.9 0.14 14.29 3.1 3 0 3.3% 0.2 10.1 Control 0.13 15.38 3.1 3 0 3.3% 0.2 13.0 Control 0.15 13.33 3.1 3 0 3.3% 0.3 13.8 Control 0.13 15.38 3.1 3 0 3.3% 0.2 12.9 Control 0.15 13.33 3.1 3 0 3.3% 0.3 13.1 Control 0.13 15.38 3.1 3 0 3.3% 0.2 11 Control 0.12 16.67 3.1 3 0 3.3% 0.2 10.4 Control 0.18 11.11 2.1 2 0 5.0% 0.5 6.8 Control 0.18 11.11 2.1 2 0 5.0% 0.5 9.7 Control 0.18 11.11 4.1 3.8 0 7.9% 0.7 12 Control 0.17 11.76 4.1 3.8 0 7.9% 0.7 17.5 Control 0.17 11.76 4.1 3.8 0 7.9% 0.7 38.7 Species Thickness SA/V Control 0.15 Control 0.15 Control 11.1 E. stellulata 0.4 5.00 100.1 99.9 79.5 1.0% 0.2 8.6 E. stellulata 0.33 6.06 100.11 99.9 79.5 1.0% 0.2 8.4 E. stellulata 0.37 5.41 100.12 99.9 79.5 1.1% 0.2 10.2 E. stellulata 0.22 9.09 100.13 99.9 79.5 1.1% 0.1 7.2 E. stellulata 0.3 6.67 100.14 99.9 79.5 1.2% 0.2 10.2 E. stellulata 0.35 5.71 100.16 99.9 79.5 1.3% 0.2 11.3 E. stellulata 0.25 8.00 100.17 99.9 79.5 1.3% 0.2 8.5 E. stellulata 0.28 7.14 100.18 99.9 79.5 1.4% 0.2 9 E. stellulata 0.22 9.09 100.19 99.9 79.5 1.4% 0.2 7 E. stellulata 0.24 8.33 100.2 99.9 79.5 1.5% 0.2 9 E. stellulata 0.38 5.26 106.2 91.8 80 122.0% 23.2 34.1 E. stellulata 0.38 5.26 106.13 91.8 80 121.5% 23.1 29.6 E. stellulata 0.4 5.00 106.07 91.8 80 120.9% 24.2 42.1 E. stellulata 0.33 6.06 106.00 91.8 80 120.3% 19.9 29.3 E. stellulata 0.32 6.25 105.93 91.8 80 119.8% 19.2 31.8 E. stellulata 0.3 6.67 105.87 91.8 80 119.2% 17.9 31.2 E. stellulata 0.34 5.88 105.80 91.8 80 118.6% 20.2 32.6 E. stellulata 0.38 5.26 105.73 91.8 80 118.1% 22.4 43 E. stellulata 0.36 5.56 105.67 91.8 80 117.5% 21.2 34.5 E. stellulata 0.33 6.06 105.6 91.8 80 116.9% 19.3 31.6 Tasmannia xerophila 0.4 5.00 95.4 85.7 80.1 173.2% 34.6 35 Tasmannia xerophila 0.42 4.76 95.33 85.7 80.1 172.0% 36.1 34 Tasmannia xerophila 0.39 5.13 95.27 85.7 80.1 170.8% 33.3 37.6 Tasmannia xerophila 0.37 5.41 95.13 85.7 80.1 168.5% 31.2 41.3 Tasmannia xerophila 0.38 5.26 95.00 85.7 80.1 166.1% 31.6 45 Tasmannia xerophila 0.38 5.26 94.93 85.7 80.1 164.9% 31.3 42.5 Tasmannia xerophila 0.35 5.71 94.87 85.7 80.1 163.7% 28.6 39.8 Tasmannia xerophila 0.36 5.56 94.8 85.7 80.1 162.5% 29.2 39.1 Tasmannia xerophila 0.55 3.64 105.00 104.5 94 4.8% 1.3 16.7 Tasmannia xerophila 0.58 3.45 105.00 104.5 94 4.8% 1.4 8.9 Tasmannia xerophila 0.5 4.00 105.00 104.5 94 4.8% 1.2 14.1 Tasmannia xerophila 0.41 4.88 105.00 104.5 94 4.8% 1.0 14.1 Tasmannia xerophila 0.45 4.44 105.00 104.5 94 4.8% 1.1 10.19 Tasmannia xerophila 0.51 3.92 105 104.5 94 4.8% 1.2 12.2 389 Olearia aglossa 0.49 4.08 96.2 87.7 79.7 106.3% 26.0 27 Olearia aglossa 0.39 5.13 96.07 87.7 79.7 104.6% 20.4 27.2 Olearia aglossa 0.32 6.25 96.00 87.7 79.7 103.8% 16.6 25.2 Olearia aglossa 0.38 5.26 95.93 87.7 79.7 102.9% 19.6 26.1 Olearia aglossa 0.41 4.88 95.87 87.7 79.7 102.1% 20.9 35.8 Olearia aglossa 0.35 5.71 95.80 87.7 79.7 101.3% 17.7 29.4 Olearia aglossa 0.42 4.76 95.73 87.7 79.7 100.4% 21.1 38.1 Olearia aglossa 0.41 4.88 95.67 87.7 79.7 99.6% 20.4 35.7 Olearia aglossa 0.41 4.88 95.6 87.7 79.7 98.7% 20.2 28.8 Olearia aglossa 0.5 4.00 111.4 111.1 93.7 1.7% 0.4 17.9 Olearia aglossa 0.51 3.92 111.42 111.1 93.7 1.9% 0.5 24.6 Olearia aglossa 0.54 3.70 111.47 111.1 93.7 2.1% 0.6 24.3 Olearia aglossa 0.52 3.85 111.49 111.1 93.7 2.2% 0.6 25.3 Olearia aglossa 0.48 4.17 111.51 111.1 93.7 2.4% 0.6 13.4 Olearia aglossa 0.52 3.85 111.53 111.1 93.7 2.5% 0.6 16 Olearia aglossa 0.45 4.44 111.6 111.1 93.7 2.9% 0.6 16.4 Daviesia mimosoides 0.5 4.00 121.2 107.9 94.2 97.1% 24.3 38.3 Daviesia mimosoides 0.49 4.08 121.17 107.9 94.2 96.8% 23.7 38.5 Daviesia mimosoides 0.44 4.55 121.13 107.9 94.2 96.6% 21.3 26.8 Daviesia mimosoides 0.51 3.92 121.10 107.9 94.2 96.4% 24.6 34.1 Daviesia mimosoides 0.49 4.08 121.07 107.9 94.2 96.1% 23.5 36.4 Daviesia mimosoides 0.41 4.88 121.03 107.9 94.2 95.9% 19.7 27.3 Daviesia mimosoides 0.37 5.41 121.00 107.9 94.2 95.6% 17.7 19.8 Daviesia mimosoides 0.48 4.17 120.93 107.9 94.2 95.1% 22.8 25.4 Daviesia mimosoides 0.29 6.90 108.8 108.5 93.4 2.0% 0.3 11.4 Daviesia mimosoides 0.33 6.06 108.83 108.5 93.4 2.2% 0.4 9.5 Daviesia mimosoides 0.39 5.13 108.87 108.5 93.4 2.4% 0.5 9.5 Daviesia mimosoides 0.33 6.06 108.88 108.5 93.4 2.5% 0.4 17.4 Daviesia mimosoides 0.34 5.88 108.89 108.5 93.4 2.6% 0.4 17.5 E. pauciflora 0.5 4.00 96.2 87.1 79.5 119.7% 29.9 32.5 E. pauciflora 0.48 4.17 96.13 87.1 79.5 118.9% 28.5 32.9 E. pauciflora 0.68 2.94 96.07 87.1 79.5 118.0% 40.1 36.4 E. pauciflora 0.43 4.65 96.00 87.1 79.5 117.1% 25.2 31.7 E. pauciflora 0.59 3.39 95.93 87.1 79.5 116.2% 34.3 38.8 E. pauciflora 0.43 4.65 95.87 87.1 79.5 115.4% 24.8 39.9 E. pauciflora 0.42 4.76 95.80 87.1 79.5 114.5% 24.0 37.8 E. pauciflora 0.48 4.17 95.73 87.1 79.5 113.6% 27.3 34.9 E. pauciflora 0.46 4.35 95.67 87.1 79.5 112.7% 25.9 36.4 E. pauciflora 0.51 3.92 95.6 87.1 79.5 111.8% 28.5 34.8 E. pauciflora 0.53 3.77 116.1 115.7 93.6 1.8% 0.5 12.9 E. pauciflora 0.33 6.06 116.13 115.7 93.6 2.0% 0.3 9.6 E. pauciflora 0.35 5.71 116.17 115.7 93.6 2.1% 0.4 8.6 E. pauciflora 0.57 3.51 116.20 115.7 93.6 2.3% 0.6 13.2 E. pauciflora 0.37 5.41 116.23 115.7 93.6 2.4% 0.4 11.1 E. pauciflora 0.49 4.08 116.27 115.7 93.6 2.6% 0.6 10.7 E. pauciflora 0.51 3.92 116.30 115.7 93.6 2.7% 0.7 11.7 E. pauciflora 0.4 5.00 116.33 115.7 93.6 2.9% 0.6 9.6 E. pauciflora 0.6 3.33 116.37 115.7 93.6 3.0% 0.9 14.7 E. pauciflora 0.49 4.08 116.4 115.7 93.6 3.2% 0.8 13 390 Tests at 260oC Species Thickness SA/V Wet weight Dry weight Tin %moist Control 0.17 11.76 2.1 2 0 5.0% 0.4 5 Control 0.175 11.43 2.1 2 0 5.0% 0.4 6.8 Control 0.18 11.11 2.1 2 0 5.0% 0.5 9.3 Control 0.17 11.76 2.1 2 0 5.0% 0.4 9.3 Control 0.18 11.11 2.1 2 0 5.0% 0.5 11.6 Control 0.18 11.11 2 2 0 0.0% 0.0 9.3 Control 0.18 11.11 2.03 2 0 1.5% 0.1 13 Control 0.17 11.76 2.05 2 0 2.5% 0.2 9.6 Control 0.18 11.11 2.08 2 0 4.0% 0.4 8.5 Control 0.17 11.76 4.1 3.8 0 7.9% 0.7 6.6 M/SAV Control 0.17 11.76 4.2 3.8 0 10.5% 0.9 11.5 E. stellulata 0.39 5.13 116.3 105.7 94.2 92.2% 18.0 26.6 E. stellulata 0.38 5.26 116.27 105.7 94.2 91.9% 17.5 26.2 E. stellulata 0.44 4.55 116.23 105.7 94.2 91.6% 20.2 31.9 E. stellulata 0.41 4.88 116.20 105.7 94.2 91.3% 18.7 28.9 E. stellulata 0.39 5.13 116.17 105.7 94.2 91.0% 17.7 26.7 E. stellulata 0.36 5.56 116.13 105.7 94.2 90.7% 16.3 26.2 E. stellulata 0.41 4.88 116.10 105.7 94.2 90.4% 18.5 30.4 E. stellulata 0.36 5.56 116.07 105.7 94.2 90.1% 16.2 27.3 E. stellulata 0.36 5.56 116.03 105.7 94.2 89.9% 16.2 22.9 E. stellulata 0.39 5.13 116 105.7 94.2 89.6% 17.5 24.5 E. stellulata 0.44 4.55 112.8 112.4 92.7 2.0% 0.4 9.3 E. stellulata 0.34 5.88 112.80 112.4 92.7 2.0% 0.3 9 E. stellulata 0.38 5.26 112.80 112.4 92.7 2.0% 0.4 7.2 E. stellulata 0.46 4.35 112.80 112.4 92.7 2.0% 0.5 8.7 E. stellulata 0.44 4.55 112.80 112.4 92.7 2.0% 0.4 7.3 E. stellulata 0.39 5.13 112.80 112.4 92.7 2.0% 0.4 6.9 E. stellulata 0.29 6.90 112.80 112.4 92.7 2.0% 0.3 6.6 E. stellulata 0.36 5.56 112.80 112.4 92.7 2.0% 0.4 5.8 E. stellulata 0.46 4.35 112.80 112.4 92.7 2.0% 0.5 6.7 E. stellulata 0.45 4.44 112.8 112.4 92.7 2.0% 0.5 9.8 Bossiaea foliosa 0.16 12.50 110.9 110.5 93.5 2.4% 0.2 14.4 Bossiaea foliosa 0.16 12.50 110.91 110.5 93.5 2.4% 0.2 17.5 Bossiaea foliosa 0.18 11.11 110.93 110.5 93.5 2.5% 0.2 16.9 Bossiaea foliosa 0.15 13.33 110.95 110.5 93.5 2.6% 0.2 9.6 Bossiaea foliosa 0.17 11.76 110.96 110.5 93.5 2.7% 0.2 14.8 Bossiaea foliosa 0.16 12.50 110.97 110.5 93.5 2.8% 0.2 11.1 Bossiaea foliosa 0.15 13.33 110.98 110.5 93.5 2.8% 0.2 23.2 Bossiaea foliosa 0.16 12.50 111 110.5 93.5 2.9% 0.2 15.1 Tasmannia xerophila 0.53 3.77 122.1 104.4 93.4 160.9% 42.6 34.9 Tasmannia xerophila 0.5 4.00 122.01 104.4 93.4 160.1% 40.0 31.6 Tasmannia xerophila 0.48 4.17 121.83 104.4 93.4 158.5% 38.0 35.1 Tasmannia xerophila 0.52 3.85 121.66 104.4 93.4 156.9% 40.8 48.2 Tasmannia xerophila 0.53 3.77 121.57 104.4 93.4 156.1% 41.4 36.7 Tasmannia xerophila 0.48 4.17 121.48 104.4 93.4 155.3% 37.3 38.8 Tasmannia xerophila 0.41 4.88 121.39 104.4 93.4 154.4% 31.7 37.4 Tasmannia xerophila 0.37 5.41 103.6 103.5 93.5 1.0% 0.2 10.5 Tasmannia xerophila 0.35 5.71 103.62 103.5 93.5 1.2% 0.2 8.6 Tasmannia xerophila 0.38 5.26 103.67 103.5 93.5 1.7% 0.3 9.9 Tasmannia xerophila 0.39 5.13 103.69 103.5 93.5 1.9% 0.4 9.6 Tasmannia xerophila 0.39 5.13 103.73 103.5 93.5 2.3% 0.5 8.6 391 Tasmannia xerophila 0.33 6.06 103.78 103.5 93.5 2.8% 0.5 6.2 Tasmannia xerophila 0.41 4.88 103.8 103.5 93.5 3.0% 0.6 5.2 Olearia aglossa 0.48 4.17 119.6 106.9 94.2 100.0% 24.0 17.9 Olearia aglossa 0.46 4.35 119.49 106.9 94.2 99.1% 22.8 19.2 Olearia aglossa 0.55 3.64 119.38 106.9 94.2 98.3% 27.0 24.7 Olearia aglossa 0.48 4.17 119.27 106.9 94.2 97.4% 23.4 21.1 Olearia aglossa 0.51 3.92 119.16 106.9 94.2 96.5% 24.6 27.1 Olearia aglossa 0.5 4.00 119.04 106.9 94.2 95.6% 23.9 21 Olearia aglossa 0.46 4.35 118.93 106.9 94.2 94.8% 21.8 20 Olearia aglossa 0.52 3.85 118.82 106.9 94.2 93.9% 24.4 25.7 Olearia aglossa 0.48 4.17 118.71 106.9 94.2 93.0% 22.3 21.1 Olearia aglossa 0.47 4.26 118.6 106.9 94.2 92.1% 21.6 20.1 Olearia aglossa 0.42 4.76 106.4 106.3 93.7 0.8% 0.2 5.1 Olearia aglossa 0.43 4.65 106.43 106.3 93.7 1.1% 0.2 7 Olearia aglossa 0.42 4.76 106.47 106.3 93.7 1.3% 0.3 7.5 Olearia aglossa 0.45 4.44 106.50 106.3 93.7 1.6% 0.4 7.3 Olearia aglossa 0.45 4.44 106.53 106.3 93.7 1.9% 0.4 5.1 Olearia aglossa 0.48 4.17 106.57 106.3 93.7 2.1% 0.5 6.8 Olearia aglossa 0.47 4.26 106.60 106.3 93.7 2.4% 0.6 7.8 Olearia aglossa 0.44 4.55 106.63 106.3 93.7 2.6% 0.6 11.4 Olearia aglossa 0.47 4.26 106.67 106.3 93.7 2.9% 0.7 6.8 Olearia aglossa 0.42 4.76 106.7 106.3 93.7 3.2% 0.7 9.5 Daviesia mimosoides 0.44 4.55 129.29 111.3 93.8 102.8% 22.6 34 Daviesia mimosoides 0.4 5.00 129.18 111.3 93.8 102.2% 20.4 25.4 Daviesia mimosoides 0.42 4.76 129.07 111.3 93.8 101.5% 21.3 27.5 Daviesia mimosoides 0.4 5.00 128.96 111.3 93.8 100.9% 20.2 23 Daviesia mimosoides 0.42 4.76 128.84 111.3 93.8 100.3% 21.1 27.8 Daviesia mimosoides 0.41 4.88 128.62 111.3 93.8 99.0% 20.3 24.8 Daviesia mimosoides 0.34 5.88 128.51 111.3 93.8 98.3% 16.7 23.2 Daviesia mimosoides 0.42 4.76 128.4 111.3 93.8 97.7% 20.5 27.1 Daviesia mimosoides 0.38 5.26 108.9 108.8 94.1 0.7% 0.1 10.1 Daviesia mimosoides 0.44 4.55 108.91 108.8 94.1 0.8% 0.2 9.5 Daviesia mimosoides 0.48 4.17 108.92 108.8 94.1 0.8% 0.2 9.7 Daviesia mimosoides 0.44 4.55 108.93 108.8 94.1 0.9% 0.2 19.6 Daviesia mimosoides 0.37 5.41 108.94 108.8 94.1 1.0% 0.2 9.4 Daviesia mimosoides 0.41 4.88 108.96 108.8 94.1 1.1% 0.2 23 Daviesia mimosoides 0.46 4.35 108.97 108.8 94.1 1.1% 0.3 7.5 Daviesia mimosoides 0.39 5.13 108.98 108.8 94.1 1.2% 0.2 12.4 Daviesia mimosoides 0.4 5.00 108.99 108.8 94.1 1.3% 0.3 8.1 Daviesia mimosoides 0.46 4.35 109 108.8 94.1 1.4% 0.3 12 E. pauciflora 0.52 3.85 160.8 123.9 93.8 122.6% 31.9 33.8 E. pauciflora 0.58 3.45 160.68 123.9 93.8 122.2% 35.4 40.4 E. pauciflora 0.5 4.00 160.56 123.9 93.8 121.8% 30.4 48.1 E. pauciflora 0.52 3.85 160.43 123.9 93.8 121.4% 31.6 36.5 E. pauciflora 0.61 3.28 160.31 123.9 93.8 121.0% 36.9 42.5 E. pauciflora 0.52 3.85 160.19 123.9 93.8 120.6% 31.3 35.1 E. pauciflora 0.59 3.39 160.07 123.9 93.8 120.2% 35.4 40.8 E. pauciflora 0.49 4.08 159.94 123.9 93.8 119.7% 29.3 33.5 E. pauciflora 0.55 3.64 159.82 123.9 93.8 119.3% 32.8 45.5 E. pauciflora 0.55 3.64 159.7 123.9 93.8 118.9% 32.7 46 E. pauciflora 0.39 5.13 114.1 113.7 94.1 2.0% 0.4 7.2 E. pauciflora 0.39 5.13 114.12 113.7 94.1 2.2% 0.4 9.7 E. pauciflora 0.42 4.76 114.14 113.7 94.1 2.3% 0.5 8.8 E. pauciflora 0.35 5.71 114.17 113.7 94.1 2.4% 0.4 7.1 392 E. pauciflora 0.4 5.00 114.19 113.7 94.1 2.5% 0.5 8 E. pauciflora 0.35 5.71 114.21 113.7 94.1 2.6% 0.5 7.3 E. pauciflora 0.36 5.56 114.23 113.7 94.1 2.7% 0.5 7 E. pauciflora 0.35 5.71 114.26 113.7 94.1 2.8% 0.5 11.6 E. pauciflora 0.41 4.88 114.28 113.7 94.1 2.9% 0.6 7.7 E. pauciflora 0.39 5.13 114.3 113.7 94.1 3.1% 0.6 8.1 393 Tests at 300oC Species Thickness SA/V Wet weight Dry weight Tin %moist M/SAV Control 0.18 11.11 2 2 0 0.0% 0.0 Control 0.17 11.76 2 2 0 0.0% 0.0 5 Control 0.18 11.11 3.8 3.6 0 5.6% 0.5 5.3 Control 0.17 11.76 3.8 3.6 0 5.6% 0.5 4.2 Control 0.18 11.11 3.8 3.6 0 5.6% 0.5 5.6 Control 0.17 11.76 3.8 3.6 0 5.6% 0.5 4.8 Control 0.18 11.11 3.8 3.6 0 5.6% 0.5 5.7 Control 0.17 11.76 56.3 55.8 0 0.9% 0.1 4.3 Control 0.19 10.53 56.38 55.8 0 1.0% 0.1 4.9 Control 0.17 11.76 56.45 55.8 0 1.2% 0.1 3.9 Control 0.17 11.76 56.53 55.8 0 1.3% 0.1 4.1 Control 0.18 11.11 56.6 55.8 0 1.4% 0.1 4.3 E. stellulata 0.44 4.55 115.7 105.7 94.2 87.0% 19.1 16.8 E. stellulata 0.34 5.88 115.69 105.7 94.2 86.9% 14.8 15.5 E. stellulata 0.45 4.44 115.68 105.7 94.2 86.8% 19.5 17.9 E. stellulata 0.37 5.41 115.67 105.7 94.2 86.7% 16.0 15 E. stellulata 0.33 6.06 115.66 105.7 94.2 86.6% 14.3 14.8 E. stellulata 0.39 5.13 115.64 105.7 94.2 86.5% 16.9 19.7 E. stellulata 0.39 5.13 115.63 105.7 94.2 86.4% 16.8 18.9 E. stellulata 0.49 4.08 115.62 105.7 94.2 86.3% 21.1 25.1 E. stellulata 0.37 5.41 115.61 105.7 94.2 86.2% 15.9 16.9 E. stellulata 0.41 4.88 115.6 105.7 94.2 86.1% 17.6 21 E. stellulata 0.32 6.25 148 147.9 93.7 0.2% 0.0 7.4 4.6 5.9 E. stellulata 0.3 6.67 148.02 147.9 93.7 0.2% 0.0 E. stellulata 0.31 6.45 148.04 147.9 93.7 0.3% 0.0 4 E. stellulata 0.4 5.00 148.07 147.9 93.7 0.3% 0.1 7.2 E. stellulata 0.38 5.26 148.09 147.9 93.7 0.3% 0.1 4.2 E. stellulata 0.43 4.65 148.11 147.9 93.7 0.4% 0.1 5.6 E. stellulata 0.29 6.90 148.13 147.9 93.7 0.4% 0.1 4.2 E. stellulata 0.37 5.41 148.16 147.9 93.7 0.5% 0.1 4.4 E. stellulata 0.33 6.06 148.18 147.9 93.7 0.5% 0.1 4.9 E. stellulata 0.3 6.67 148.2 147.9 93.7 0.6% 0.1 6.9 Bossiaea foliosa 0.22 9.09 125.3 113.2 93.5 61.4% 6.8 21.3 Bossiaea foliosa 0.22 9.09 125.24 113.2 93.5 61.1% 6.7 21.6 Bossiaea foliosa 0.18 11.11 125.19 113.2 93.5 60.9% 5.5 20.6 Bossiaea foliosa 0.19 10.53 125.13 113.2 93.5 60.6% 5.8 26.5 Bossiaea foliosa 0.28 7.14 125.08 113.2 93.5 60.3% 8.4 23.6 Bossiaea foliosa 0.24 8.33 125.02 113.2 93.5 60.0% 7.2 18.8 Bossiaea foliosa 0.18 11.11 124.97 113.2 93.5 59.7% 5.4 18.2 Bossiaea foliosa 0.21 9.52 124.91 113.2 93.5 59.4% 6.2 24.6 Bossiaea foliosa 0.21 9.52 124.86 113.2 93.5 59.2% 6.2 25.8 Bossiaea foliosa 0.21 9.52 124.8 113.2 93.5 58.9% 6.2 23.9 Bossiaea foliosa 0.16 12.50 127.4 127.3 93.9 0.3% 0.0 9.2 Bossiaea foliosa 0.18 11.11 127.42 127.3 93.9 0.4% 0.0 10.9 Bossiaea foliosa 0.15 13.33 127.44 127.3 93.9 0.4% 0.0 8.1 Bossiaea foliosa 0.15 13.33 127.47 127.3 93.9 0.5% 0.0 5.2 Bossiaea foliosa 0.21 9.52 127.49 127.3 93.9 0.6% 0.1 5.4 Bossiaea foliosa 0.17 11.76 127.51 127.3 93.9 0.6% 0.1 6.5 Bossiaea foliosa 0.18 11.11 127.53 127.3 93.9 0.7% 0.1 6 Bossiaea foliosa 0.17 11.76 127.56 127.3 93.9 0.8% 0.1 6.2 Bossiaea foliosa 0.16 12.50 127.58 127.3 93.9 0.8% 0.1 4.7 394 Bossiaea foliosa 0.18 11.11 127.6 127.3 93.9 0.9% 0.1 8.5 Tasmannia xerophila 0.6 3.33 124.6 106.1 94 152.9% 45.9 33.9 Tasmannia xerophila 0.59 3.39 124.57 106.1 94 152.6% 45.0 33.6 Tasmannia xerophila 0.6 3.33 124.53 106.1 94 152.3% 45.7 32.1 Tasmannia xerophila 0.53 3.77 124.50 106.1 94 152.1% 40.3 33 Tasmannia xerophila 0.59 3.39 124.47 106.1 94 151.8% 44.8 34 Tasmannia xerophila 0.58 3.45 124.43 106.1 94 151.5% 43.9 33.7 Tasmannia xerophila 0.56 3.57 124.40 106.1 94 151.2% 42.3 31.4 Tasmannia xerophila 0.53 3.77 124.37 106.1 94 151.0% 40.0 39.9 Tasmannia xerophila 0.59 3.39 124.33 106.1 94 150.7% 44.5 30.6 Tasmannia xerophila 0.53 3.77 124.3 106.1 94 150.4% 39.9 33.3 Tasmannia xerophila 0.42 4.76 126.3 126.2 93.5 0.3% 0.1 7.5 Tasmannia xerophila 0.42 4.76 126.31 126.2 93.5 0.3% 0.1 8.3 Tasmannia xerophila 0.42 4.76 126.32 126.2 93.5 0.4% 0.1 11.5 Tasmannia xerophila 0.42 4.76 126.33 126.2 93.5 0.4% 0.1 6.6 Tasmannia xerophila 0.39 5.13 126.34 126.2 93.5 0.4% 0.1 7.3 Tasmannia xerophila 0.39 5.13 126.36 126.2 93.5 0.5% 0.1 3.3 Tasmannia xerophila 0.44 4.55 126.37 126.2 93.5 0.5% 0.1 5 Tasmannia xerophila 0.54 3.70 126.38 126.2 93.5 0.5% 0.1 5.4 Tasmannia xerophila 0.36 5.56 126.39 126.2 93.5 0.6% 0.1 7 Tasmannia xerophila 0.44 4.55 126.4 126.2 93.5 0.6% 0.1 8.5 Olearia aglossa 0.59 3.39 127.4 112.8 93.7 76.4% 22.5 17.9 Olearia aglossa 0.56 3.57 127.36 112.8 93.7 76.2% 21.3 23.6 Olearia aglossa 0.52 3.85 127.31 112.8 93.7 76.0% 19.8 23 Olearia aglossa 0.55 3.64 127.27 112.8 93.7 75.7% 20.8 22.7 Olearia aglossa 0.58 3.45 127.22 112.8 93.7 75.5% 21.9 22.6 Olearia aglossa 0.57 3.51 127.18 112.8 93.7 75.3% 21.5 24.1 Olearia aglossa 0.5 4.00 127.13 112.8 93.7 75.0% 18.8 21.4 Olearia aglossa 0.58 3.45 127.09 112.8 93.7 74.8% 21.7 19.9 Olearia aglossa 0.54 3.70 127.04 112.8 93.7 74.6% 20.1 22.1 Olearia aglossa 0.54 3.70 127 112.8 93.7 74.3% 20.1 19.8 Olearia aglossa 0.64 3.13 130 129.8 92.7 0.5% 0.2 7.9 Olearia aglossa 0.57 3.51 130.01 129.8 92.7 0.6% 0.2 5.7 Olearia aglossa 0.67 2.99 130.02 129.8 92.7 0.6% 0.2 5.5 Olearia aglossa 0.58 3.45 130.03 129.8 92.7 0.6% 0.2 4.4 Olearia aglossa 0.73 2.74 130.04 129.8 92.7 0.7% 0.2 6.6 Olearia aglossa 0.58 3.45 130.06 129.8 92.7 0.7% 0.2 5 Olearia aglossa 0.59 3.39 130.07 129.8 92.7 0.7% 0.2 7.2 Olearia aglossa 0.6 3.33 130.08 129.8 92.7 0.7% 0.2 5.8 Olearia aglossa 0.64 3.13 130.09 129.8 92.7 0.8% 0.2 8.2 Olearia aglossa 0.61 3.28 130.1 129.8 92.7 0.8% 0.2 6.7 Daviesia mimosoides 0.45 4.44 124.5 107.9 93.5 115.3% 25.9 23.3 Daviesia mimosoides 0.48 4.17 124.47 107.9 93.5 115.0% 27.6 23.4 Daviesia mimosoides 0.47 4.26 124.43 107.9 93.5 114.8% 27.0 30.9 Daviesia mimosoides 0.49 4.08 124.40 107.9 93.5 114.6% 28.1 24 Daviesia mimosoides 0.43 4.65 124.37 107.9 93.5 114.4% 24.6 21.3 Daviesia mimosoides 0.44 4.55 124.33 107.9 93.5 114.1% 25.1 20.1 Daviesia mimosoides 0.46 4.35 124.30 107.9 93.5 113.9% 26.2 21.9 Daviesia mimosoides 0.44 4.55 124.27 107.9 93.5 113.7% 25.0 18.8 Daviesia mimosoides 0.48 4.17 124.23 107.9 93.5 113.4% 27.2 22.3 Daviesia mimosoides 0.43 4.65 124.2 107.9 93.5 113.2% 24.3 25.9 Daviesia mimosoides 0.38 5.26 118.5 118.5 93.7 0.0% 0.0 7.9 Daviesia mimosoides 0.35 5.71 118.52 118.5 93.7 0.1% 0.0 8.8 Daviesia mimosoides 0.36 5.56 118.54 118.5 93.7 0.2% 0.0 9.3 395 Daviesia mimosoides 0.34 5.88 118.57 118.5 93.7 0.3% 0.0 8.9 Daviesia mimosoides 0.35 5.71 118.59 118.5 93.7 0.4% 0.1 8.9 Daviesia mimosoides 0.35 5.71 118.61 118.5 93.7 0.4% 0.1 4.2 Daviesia mimosoides 0.34 5.88 118.63 118.5 93.7 0.5% 0.1 3.9 Daviesia mimosoides 0.3 6.67 118.66 118.5 93.7 0.6% 0.1 6.5 Daviesia mimosoides 0.38 5.26 118.68 118.5 93.7 0.7% 0.1 4.6 Daviesia mimosoides 0.3 6.67 118.7 118.5 93.7 0.8% 0.1 3 E. pauciflora 0.49 4.08 148.6 121.8 93.6 95.0% 23.3 22.3 E. pauciflora 0.48 4.17 148.53 121.8 93.6 94.8% 22.8 25.5 E. pauciflora 0.48 4.17 148.47 121.8 93.6 94.6% 22.7 24.6 E. pauciflora 0.5 4.00 148.40 121.8 93.6 94.3% 23.6 23.4 E. pauciflora 0.51 3.92 148.33 121.8 93.6 94.1% 24.0 26.5 E. pauciflora 0.47 4.26 148.27 121.8 93.6 93.9% 22.1 23.8 E. pauciflora 0.51 3.92 148.20 121.8 93.6 93.6% 23.9 20.3 E. pauciflora 0.51 3.92 148.13 121.8 93.6 93.4% 23.8 24.6 E. pauciflora 0.53 3.77 148.07 121.8 93.6 93.1% 24.7 22.1 E. pauciflora 0.48 4.17 148 121.8 93.6 92.9% 22.3 22.1 E. pauciflora 0.33 6.06 114.3 113.7 94.1 3.1% 0.5 4.3 E. pauciflora 0.32 6.25 114.31 113.7 94.1 3.1% 0.5 4 E. pauciflora 0.38 5.26 114.32 113.7 94.1 3.2% 0.6 4.8 E. pauciflora 0.39 5.13 114.33 113.7 94.1 3.2% 0.6 4.7 E. pauciflora 0.38 5.26 114.34 113.7 94.1 3.3% 0.6 7.7 E. pauciflora 0.44 4.55 114.36 113.7 94.1 3.3% 0.7 6.7 E. pauciflora 0.4 5.00 114.37 113.7 94.1 3.4% 0.7 6.1 E. pauciflora 0.39 5.13 114.38 113.7 94.1 3.5% 0.7 6.2 E. pauciflora 0.48 4.17 114.39 113.7 94.1 3.5% 0.8 7.6 E. pauciflora 0.45 4.44 114.4 113.7 94.1 3.6% 0.8 7 396 Tests at 350oC Species Thickness SA/V Wet weight Dry weight Tin %moist M/SAV Control 0.17 11.76 3.8 3.6 0 5.6% 0.5 2.7 Control 0.18 11.11 3.8 3.6 0 5.6% 0.5 3.2 Control 0.18 11.11 3.8 3.6 0 5.6% 0.5 3 Control 0.17 11.76 3.8 3.6 0 5.6% 0.5 3.1 Control 0.17 11.76 3.8 3.6 0 5.6% 0.5 2.4 Control 0.18 11.11 4.2 3.9 0 7.7% 0.7 2.7 Control 0.18 11.11 4.15 3.9 0 6.4% 0.6 3.5 Control 0.17 11.76 4.1 3.9 0 5.1% 0.4 3 Control 0.18 11.11 57.6 55.8 0 3.2% 0.3 4 Control 0.16 12.50 57.65 55.8 0 3.3% 0.3 3.4 Control 0.17 11.76 57.70 55.8 0 3.4% 0.3 3.6 Control 0.17 11.76 57.75 55.8 0 3.5% 0.3 3.6 Control 0.18 11.11 57.8 55.8 0 3.6% 0.3 3.7 E. stellulata 0.35 5.71 133.3 114.4 92.7 87.1% 15.2 11 E. stellulata 0.37 5.41 133.27 114.4 92.7 86.9% 16.1 10.2 E. stellulata 0.4 5.00 133.23 114.4 92.7 86.8% 17.4 10.9 E. stellulata 0.34 5.88 133.20 114.4 92.7 86.6% 14.7 9.8 E. stellulata 0.35 5.71 133.17 114.4 92.7 86.5% 15.1 11.1 E. stellulata 0.32 6.25 133.13 114.4 92.7 86.3% 13.8 10.4 E. stellulata 0.37 5.41 133.10 114.4 92.7 86.2% 15.9 11.8 E. stellulata 0.35 5.71 133.07 114.4 92.7 86.0% 15.1 9.6 E. stellulata 0.34 5.88 133.03 114.4 92.7 85.9% 14.6 10.4 E. stellulata 0.33 6.06 133 114.4 92.7 85.7% 14.1 9.4 E. stellulata 0.32 6.25 148.3 147.9 93.7 0.7% 0.1 3.9 E. stellulata 0.4 5.00 148.30 147.9 93.7 0.7% 0.1 7 E. stellulata 0.29 6.90 148.30 147.9 93.7 0.7% 0.1 3.3 E. stellulata 0.35 5.71 148.30 147.9 93.7 0.7% 0.1 3.4 E. stellulata 0.4 5.00 148.30 147.9 93.7 0.7% 0.1 4 E. stellulata 0.37 5.41 148.30 147.9 93.7 0.7% 0.1 3.9 E. stellulata 0.37 5.41 148.30 147.9 93.7 0.7% 0.1 7.8 E. stellulata 0.33 6.06 148.30 147.9 93.7 0.7% 0.1 3.2 E. stellulata 0.34 5.88 148.30 147.9 93.7 0.7% 0.1 3.8 E. stellulata 0.37 5.41 148.3 147.9 93.7 0.7% 0.1 3.8 Bossiaea foliosa 0.15 13.33 124.5 113.2 93.5 57.4% 4.3 7.8 Bossiaea foliosa 0.18 11.11 124.47 113.2 93.5 57.2% 5.1 9.1 Bossiaea foliosa 0.11 18.18 124.43 113.2 93.5 57.0% 3.1 9.4 Bossiaea foliosa 0.15 13.33 124.40 113.2 93.5 56.9% 4.3 8.3 Bossiaea foliosa 0.22 9.09 124.37 113.2 93.5 56.7% 6.2 8.9 Bossiaea foliosa 0.22 9.09 124.33 113.2 93.5 56.5% 6.2 6.5 Bossiaea foliosa 0.16 12.50 124.30 113.2 93.5 56.3% 4.5 8.2 Bossiaea foliosa 0.21 9.52 124.27 113.2 93.5 56.2% 5.9 7.4 Bossiaea foliosa 0.17 11.76 124.23 113.2 93.5 56.0% 4.8 8.9 Bossiaea foliosa 0.14 14.29 124.2 113.2 93.5 55.8% 3.9 8.5 Bossiaea foliosa 0.18 11.11 127.7 127.3 93.9 1.2% 0.1 6 Bossiaea foliosa 0.18 11.11 127.71 127.3 93.9 1.2% 0.1 5.6 Bossiaea foliosa 0.19 10.53 127.72 127.3 93.9 1.3% 0.1 6.5 Bossiaea foliosa 0.2 10.00 127.73 127.3 93.9 1.3% 0.1 5.7 Bossiaea foliosa 0.21 9.52 127.74 127.3 93.9 1.3% 0.1 5.6 Bossiaea foliosa 0.2 10.00 127.76 127.3 93.9 1.4% 0.1 5 Bossiaea foliosa 0.2 10.00 127.77 127.3 93.9 1.4% 0.1 5.5 Bossiaea foliosa 0.2 10.00 127.78 127.3 93.9 1.4% 0.1 4.9 397 Bossiaea foliosa 0.18 11.11 127.79 127.3 93.9 1.5% 0.1 4.8 Bossiaea foliosa 0.2 10.00 127.8 127.3 93.9 1.5% 0.1 4.5 Tasmannia xerophila 0.49 4.08 182.2 129.3 93.5 147.8% 36.2 20.5 Tasmannia xerophila 0.56 3.57 182.17 129.3 93.5 147.7% 41.3 20.5 Tasmannia xerophila 0.58 3.45 182.13 129.3 93.5 147.6% 42.8 19.2 Tasmannia xerophila 0.64 3.13 182.10 129.3 93.5 147.5% 47.2 23.9 Tasmannia xerophila 0.58 3.45 182.07 129.3 93.5 147.4% 42.7 19.3 Tasmannia xerophila 0.58 3.45 182.03 129.3 93.5 147.3% 42.7 22.4 Tasmannia xerophila 0.57 3.51 182.00 129.3 93.5 147.2% 42.0 22.3 Tasmannia xerophila 0.57 3.51 181.97 129.3 93.5 147.1% 41.9 21 Tasmannia xerophila 0.59 3.39 181.93 129.3 93.5 147.0% 43.4 20.5 Tasmannia xerophila 0.59 3.39 181.9 129.3 93.5 146.9% 43.3 22.8 Tasmannia xerophila 0.44 4.55 126.4 126.2 93.5 0.6% 0.1 3.5 Tasmannia xerophila 0.36 5.56 126.41 126.2 93.5 0.6% 0.1 6.2 Tasmannia xerophila 0.39 5.13 126.42 126.2 93.5 0.7% 0.1 4.7 Tasmannia xerophila 0.4 5.00 126.43 126.2 93.5 0.7% 0.1 8.5 Tasmannia xerophila 0.45 4.44 126.44 126.2 93.5 0.7% 0.2 5.1 Tasmannia xerophila 0.46 4.35 126.46 126.2 93.5 0.8% 0.2 5.9 Tasmannia xerophila 0.34 5.88 126.47 126.2 93.5 0.8% 0.1 5.4 Tasmannia xerophila 0.5 4.00 126.48 126.2 93.5 0.8% 0.2 8.4 Tasmannia xerophila 0.48 4.17 126.49 126.2 93.5 0.9% 0.2 4.2 Tasmannia xerophila 0.36 5.56 126.5 126.2 93.5 0.9% 0.2 6.6 Olearia aglossa 0.58 3.45 168.4 134.5 92.7 81.1% 23.5 11.5 Olearia aglossa 0.55 3.64 168.38 134.5 92.7 81.0% 22.3 12.6 Olearia aglossa 0.59 3.39 168.36 134.5 92.7 81.0% 23.9 11.6 Olearia aglossa 0.55 3.64 168.33 134.5 92.7 80.9% 22.3 11.6 Olearia aglossa 0.56 3.57 168.31 134.5 92.7 80.9% 22.6 12 Olearia aglossa 0.54 3.70 168.29 134.5 92.7 80.8% 21.8 11.5 Olearia aglossa 0.54 3.70 168.27 134.5 92.7 80.8% 21.8 12.1 Olearia aglossa 0.57 3.51 168.24 134.5 92.7 80.7% 23.0 12.2 Olearia aglossa 0.59 3.39 168.22 134.5 92.7 80.7% 23.8 13.1 Olearia aglossa 0.55 3.64 168.2 134.5 92.7 80.6% 22.2 11.8 Olearia aglossa 0.66 3.03 130.2 129.8 92.7 1.1% 0.4 4.9 Olearia aglossa 0.65 3.08 130.21 129.8 92.7 1.1% 0.4 5.6 Olearia aglossa 0.59 3.39 130.22 129.8 92.7 1.1% 0.3 4.7 Olearia aglossa 0.66 3.03 130.23 129.8 92.7 1.2% 0.4 5.6 Olearia aglossa 0.62 3.23 130.24 129.8 92.7 1.2% 0.4 4.7 Olearia aglossa 0.62 3.23 130.26 129.8 92.7 1.2% 0.4 4.2 Olearia aglossa 0.58 3.45 130.27 129.8 92.7 1.3% 0.4 4 Olearia aglossa 0.65 3.08 130.28 129.8 92.7 1.3% 0.4 4.4 Olearia aglossa 0.63 3.17 130.29 129.8 92.7 1.3% 0.4 4.8 Olearia aglossa 0.58 3.45 130.3 129.8 92.7 1.3% 0.4 5.1 Daviesia mimosoides 0.48 4.17 144.6 120.7 93.7 88.5% 21.2 21.9 Daviesia mimosoides 0.35 5.71 144.54 120.7 93.7 88.3% 15.5 13.2 Daviesia mimosoides 0.47 4.26 144.49 120.7 93.7 88.1% 20.7 18.5 Daviesia mimosoides 0.38 5.26 144.43 120.7 93.7 87.9% 16.7 10.8 Daviesia mimosoides 0.32 6.25 144.38 120.7 93.7 87.7% 14.0 12.9 Daviesia mimosoides 0.32 6.25 144.32 120.7 93.7 87.5% 14.0 12 Daviesia mimosoides 0.5 4.00 144.27 120.7 93.7 87.3% 21.8 24 Daviesia mimosoides 0.36 5.56 144.21 120.7 93.7 87.1% 15.7 11.7 Daviesia mimosoides 0.35 5.71 144.16 120.7 93.7 86.9% 15.2 11.5 Daviesia mimosoides 0.33 6.06 144.1 120.7 93.7 86.7% 14.3 11.6 Daviesia mimosoides 0.33 6.06 118.7 118.5 93.7 0.8% 0.1 10.4 Daviesia mimosoides 0.35 5.71 118.71 118.5 93.7 0.9% 0.1 6.5 398 Daviesia mimosoides 0.32 6.25 118.72 118.5 93.7 0.9% 0.1 4 Daviesia mimosoides 0.35 5.71 118.73 118.5 93.7 0.9% 0.2 7 Daviesia mimosoides 0.34 5.88 118.74 118.5 93.7 1.0% 0.2 7.4 Daviesia mimosoides 3 0.67 118.76 118.5 93.7 1.0% 1.5 8.8 Daviesia mimosoides 0.37 5.41 118.77 118.5 93.7 1.1% 0.2 8.2 Daviesia mimosoides 0.3 6.67 118.78 118.5 93.7 1.1% 0.2 4.2 Daviesia mimosoides 0.32 6.25 118.79 118.5 93.7 1.2% 0.2 6.5 Daviesia mimosoides 0.32 6.25 118.8 118.5 93.7 1.2% 0.2 3.2 E. pauciflora 0.49 4.08 107.6 107.1 94.1 3.8% 0.9 3.9 E. pauciflora 0.39 5.13 107.60 107.1 94.1 3.8% 0.8 4.6 E. pauciflora 0.42 4.76 107.60 107.1 94.1 3.8% 0.8 4.5 E. pauciflora 0.39 5.13 107.60 107.1 94.1 3.8% 0.8 3.7 E. pauciflora 0.39 5.13 107.60 107.1 94.1 3.8% 0.8 3.2 E. pauciflora 0.35 5.71 107.60 107.1 94.1 3.8% 0.7 3.4 E. pauciflora 0.42 4.76 107.60 107.1 94.1 3.8% 0.8 4.2 E. pauciflora 0.35 5.71 107.60 107.1 94.1 3.8% 0.7 4.1 E. pauciflora 0.38 5.26 107.60 107.1 94.1 3.8% 0.7 4.6 E. pauciflora 0.37 5.41 107.6 107.1 94.1 3.8% 0.7 3.7 E. pauciflora 0.58 3.45 229.1 160.5 93.5 102.4% 29.7 19.6 E. pauciflora 0.58 3.45 229.03 160.5 93.5 102.3% 29.7 20.5 E. pauciflora 0.64 3.13 228.97 160.5 93.5 102.2% 32.7 21.5 E. pauciflora 0.65 3.08 228.90 160.5 93.5 102.1% 33.2 21.9 E. pauciflora 0.59 3.39 228.83 160.5 93.5 102.0% 30.1 21 E. pauciflora 0.64 3.13 228.77 160.5 93.5 101.9% 32.6 22.7 E. pauciflora 0.58 3.45 228.70 160.5 93.5 101.8% 29.5 19 E. pauciflora 0.6 3.33 228.63 160.5 93.5 101.7% 30.5 20.1 E. pauciflora 0.72 2.78 228.57 160.5 93.5 101.6% 36.6 23.7 E. pauciflora 0.65 3.08 228.5 160.5 93.5 101.5% 33.0 22 399 Tests at 400oC Species Thickness SA/V Wet weight Dry weight Tin %moist Control 0.18 11.11 59.1 55.8 0 5.9% 0.5 2.4 Control 0.18 11.11 59.09 55.8 0 5.9% 0.5 2.2 Control 0.17 11.76 59.08 55.8 0 5.9% 0.5 2.2 Control 0.16 12.50 59.07 55.8 0 5.9% 0.5 2.4 Control 0.18 11.11 59.06 55.8 0 5.8% 0.5 2.5 Control 0.18 11.11 59.04 55.8 0 5.8% 0.5 2.4 Control 0.18 11.11 59.03 55.8 0 5.8% 0.5 2.3 Control 0.16 12.50 59.02 55.8 0 5.8% 0.5 2.1 Control 0.17 11.76 59.01 55.8 0 5.8% 0.5 2.6 Control 0.17 11.76 59 55.8 0 5.7% 0.5 2.6 E. stellulata 0.35 5.71 111.3 110.5 93.7 4.8% 0.8 1.9 E. stellulata 0.39 5.13 110.92 110.7 92.7 1.2% 0.2 3.5 E. stellulata 0.45 4.44 110.94 110.7 92.7 1.4% 0.3 3.3 E. stellulata 0.45 4.44 110.97 110.7 92.7 1.5% 0.3 2.8 E. stellulata 0.36 5.56 110.99 110.7 92.7 1.6% 0.3 2.5 E. stellulata 0.37 5.41 111.01 110.7 92.7 1.7% 0.3 3.1 E. stellulata 0.34 5.88 111.03 110.7 92.7 1.9% 0.3 2.2 E. stellulata 0.45 4.44 111.06 110.7 92.7 2.0% 0.4 2.7 E. stellulata 0.34 5.88 111.08 110.7 92.7 2.1% 0.4 2.3 E. stellulata 0.5 4.00 111.1 110.7 92.7 2.2% 0.6 3.4 E. stellulata 0.41 4.88 183.2 140.9 93.7 89.6% 18.4 10.6 E. stellulata 0.42 4.76 183.14 140.9 93.7 89.5% 18.8 7.6 E. stellulata 0.34 5.88 183.09 140.9 93.7 89.4% 15.2 7.5 E. stellulata 0.43 4.65 183.03 140.9 93.7 89.3% 19.2 10.2 E. stellulata 0.41 4.88 182.98 140.9 93.7 89.1% 18.3 10.7 E. stellulata 0.38 5.26 182.92 140.9 93.7 89.0% 16.9 8.1 E. stellulata 0.47 4.26 182.87 140.9 93.7 88.9% 20.9 9.4 E. stellulata 0.37 5.41 182.81 140.9 93.7 88.8% 16.4 9.3 E. stellulata 0.32 6.25 182.76 140.9 93.7 88.7% 14.2 6.9 E. stellulata 0.41 4.88 182.7 140.9 93.7 88.6% 18.2 8.7 Bossiaea foliosa 0.15 13.33 127.6 127.4 93.9 0.6% 0.0 2.9 Bossiaea foliosa 0.16 12.50 127.62 127.4 93.9 0.7% 0.1 3 Bossiaea foliosa 0.16 12.50 127.64 127.4 93.9 0.7% 0.1 2.5 Bossiaea foliosa 0.17 11.76 127.67 127.4 93.9 0.8% 0.1 4.3 Bossiaea foliosa 0.16 12.50 127.69 127.4 93.9 0.9% 0.1 3.1 Bossiaea foliosa 0.15 13.33 127.71 127.4 93.9 0.9% 0.1 2.6 Bossiaea foliosa 0.17 11.76 127.73 127.4 93.9 1.0% 0.1 2.7 Bossiaea foliosa 0.17 11.76 127.76 127.4 93.9 1.1% 0.1 3.2 Bossiaea foliosa 0.17 11.76 127.78 127.4 93.9 1.1% 0.1 2.9 Bossiaea foliosa 0.14 14.29 127.8 127.4 93.9 1.2% 0.1 2.9 Bossiaea foliosa 0.13 15.38 114.3 108 93.8 44.4% 2.9 4.4 Bossiaea foliosa 0.18 11.11 114.27 108 93.8 44.1% 4.0 4.3 Bossiaea foliosa 0.15 13.33 114.23 108 93.8 43.9% 3.3 4.1 Bossiaea foliosa 0.18 11.11 114.20 108 93.8 43.7% 3.9 4.4 Bossiaea foliosa 0.18 11.11 114.17 108 93.8 43.4% 3.9 3.6 Bossiaea foliosa 0.15 13.33 114.13 108 93.8 43.2% 3.2 3.7 Bossiaea foliosa 0.14 14.29 114.10 108 93.8 43.0% 3.0 4 Bossiaea foliosa 0.15 13.33 114.07 108 93.8 42.7% 3.2 4.1 Bossiaea foliosa 0.1 20.00 114.03 108 93.8 42.5% 2.1 2.9 Bossiaea foliosa 0.2 10.00 114 108 93.8 42.3% 4.2 4.2 Tasmannia xerophila 0.49 4.08 127 126.4 93.5 1.8% 0.4 2.9 M/SAV 400 Tasmannia xerophila 0.46 4.35 127.01 126.4 93.5 1.9% 0.4 3.2 Tasmannia xerophila 0.45 4.44 127.02 126.4 93.5 1.9% 0.4 5.1 Tasmannia xerophila 0.3 6.67 127.03 126.4 93.5 1.9% 0.3 4.3 Tasmannia xerophila 0.48 4.17 127.04 126.4 93.5 2.0% 0.5 2.2 Tasmannia xerophila 0.6 3.33 127.06 126.4 93.5 2.0% 0.6 3.7 Tasmannia xerophila 0.31 6.45 127.07 126.4 93.5 2.0% 0.3 3.3 Tasmannia xerophila 0.44 4.55 127.08 126.4 93.5 2.1% 0.5 5.4 Tasmannia xerophila 0.4 5.00 127.09 126.4 93.5 2.1% 0.4 2.1 Tasmannia xerophila 0.45 4.44 127.1 126.4 93.5 2.1% 0.5 2.4 Tasmannia xerophila 0.58 3.45 127.4 108.5 94.1 131.3% 38.1 17.9 Tasmannia xerophila 0.53 3.77 127.36 108.5 94.1 130.9% 34.7 13.8 Tasmannia xerophila 0.53 3.77 127.31 108.5 94.1 130.6% 34.6 13.4 Tasmannia xerophila 0.54 3.70 127.27 108.5 94.1 130.3% 35.2 14.9 Tasmannia xerophila 0.5 4.00 127.22 108.5 94.1 130.0% 32.5 14.8 Tasmannia xerophila 0.49 4.08 127.18 108.5 94.1 129.7% 31.8 14.2 Tasmannia xerophila 0.52 3.85 127.13 108.5 94.1 129.4% 33.6 13.9 Tasmannia xerophila 0.5 4.00 127.09 108.5 94.1 129.1% 32.3 14.7 Tasmannia xerophila 0.54 3.70 127.04 108.5 94.1 128.8% 34.8 15.6 Tasmannia xerophila 0.41 4.88 127 108.5 94.1 128.5% 26.3 10.2 Olearia aglossa 0.53 3.77 130.1 130 92.7 0.3% 0.1 2.5 Olearia aglossa 0.51 3.92 130.12 130 92.7 0.3% 0.1 3 Olearia aglossa 0.55 3.64 130.14 130 92.7 0.4% 0.1 3.5 Olearia aglossa 0.55 3.64 130.17 130 92.7 0.4% 0.1 2.8 Olearia aglossa 0.5 4.00 130.19 130 92.7 0.5% 0.1 3.1 Olearia aglossa 0.54 3.70 130.21 130 92.7 0.6% 0.2 3 Olearia aglossa 0.5 4.00 130.23 130 92.7 0.6% 0.2 3.2 Olearia aglossa 0.53 3.77 130.26 130 92.7 0.7% 0.2 2.4 Olearia aglossa 0.5 4.00 130.28 130 92.7 0.7% 0.2 2.3 Olearia aglossa 0.53 3.77 130.3 130 92.7 0.8% 0.2 2.7 Olearia aglossa 0.64 3.13 117.4 107.3 93.6 73.7% 23.6 10 Olearia aglossa 0.68 2.94 117.36 107.3 93.6 73.4% 25.0 11.1 Olearia aglossa 0.6 3.33 117.31 107.3 93.6 73.1% 21.9 8.8 Olearia aglossa 0.59 3.39 117.27 107.3 93.6 72.7% 21.5 8.8 Olearia aglossa 0.66 3.03 117.22 107.3 93.6 72.4% 23.9 9.1 Olearia aglossa 0.68 2.94 117.18 107.3 93.6 72.1% 24.5 13 Olearia aglossa 0.58 3.45 117.13 107.3 93.6 71.8% 20.8 9.5 Olearia aglossa 0.63 3.17 117.09 107.3 93.6 71.5% 22.5 9.9 Olearia aglossa 0.62 3.23 117.04 107.3 93.6 71.1% 22.0 11.4 Olearia aglossa 0.62 3.23 117 107.3 93.6 70.8% 21.9 11.3 Daviesia mimosoides 0.4 5.00 118.8 118.6 93.7 0.8% 0.2 3.6 Daviesia mimosoides 0.35 5.71 118.82 118.6 93.7 0.9% 0.2 5.3 Daviesia mimosoides 0.37 5.41 118.84 118.6 93.7 1.0% 0.2 3.1 Daviesia mimosoides 0.31 6.45 118.87 118.6 93.7 1.1% 0.2 4.1 Daviesia mimosoides 0.36 5.56 118.89 118.6 93.7 1.2% 0.2 6.1 Daviesia mimosoides 0.27 7.41 118.91 118.6 93.7 1.2% 0.2 2 Daviesia mimosoides 0.4 5.00 118.93 118.6 93.7 1.3% 0.3 2.6 Daviesia mimosoides 0.39 5.13 118.96 118.6 93.7 1.4% 0.3 5.8 Daviesia mimosoides 0.35 5.71 118.98 118.6 93.7 1.5% 0.3 3.5 Daviesia mimosoides 0.4 5.00 119 118.6 93.7 1.6% 0.3 3.1 Daviesia mimosoides 0.4 5.00 115.9 105.5 94 90.4% 18.1 11.5 Daviesia mimosoides 0.33 6.06 115.84 105.5 94 90.0% 14.8 8.3 Daviesia mimosoides 0.33 6.06 115.79 105.5 94 89.5% 14.8 9.7 Daviesia mimosoides 0.37 5.41 115.73 105.5 94 89.0% 16.5 11.8 Daviesia mimosoides 0.33 6.06 115.68 105.5 94 88.5% 14.6 8.2 401 Daviesia mimosoides 0.38 5.26 115.62 105.5 94 88.0% 16.7 11.3 Daviesia mimosoides 0.35 5.71 115.57 105.5 94 87.5% 15.3 10.7 Daviesia mimosoides 0.37 5.41 115.51 105.5 94 87.1% 16.1 10.1 Daviesia mimosoides 0.33 6.06 115.46 105.5 94 86.6% 14.3 7.2 Daviesia mimosoides 0.37 5.41 115.4 105.5 94 86.1% 15.9 10.8 E. pauciflora 0.53 3.77 144.1 143.2 93.5 1.8% 0.5 4.5 E. pauciflora 0.5 4.00 144.12 143.2 93.5 1.9% 0.5 3.7 E. pauciflora 0.56 3.57 144.14 143.2 93.5 1.9% 0.5 4.3 E. pauciflora 0.48 4.17 144.17 143.2 93.5 1.9% 0.5 4 E. pauciflora 0.55 3.64 144.19 143.2 93.5 2.0% 0.5 4.2 E. pauciflora 0.49 4.08 144.21 143.2 93.5 2.0% 0.5 3.5 E. pauciflora 0.48 4.17 144.23 143.2 93.5 2.1% 0.5 3.9 E. pauciflora 0.45 4.44 144.26 143.2 93.5 2.1% 0.5 3.7 E. pauciflora 0.53 3.77 144.28 143.2 93.5 2.2% 0.6 3.7 E. pauciflora 0.57 3.51 144.3 143.2 93.5 2.2% 0.6 3.6 E. pauciflora 0.53 3.77 160 128.9 93.6 88.1% 23.3 12.5 E. pauciflora 0.52 3.85 159.91 128.9 93.6 87.9% 22.8 13.4 E. pauciflora 0.56 3.57 159.82 128.9 93.6 87.6% 24.5 12.8 E. pauciflora 0.57 3.51 159.73 128.9 93.6 87.3% 24.9 10.2 E. pauciflora 0.52 3.85 159.64 128.9 93.6 87.1% 22.6 11.8 E. pauciflora 0.47 4.26 159.56 128.9 93.6 86.8% 20.4 9.8 E. pauciflora 0.57 3.51 159.47 128.9 93.6 86.6% 24.7 14.3 E. pauciflora 0.61 3.28 159.38 128.9 93.6 86.3% 26.3 14.6 E. pauciflora 0.52 3.85 159.29 128.9 93.6 86.1% 22.4 11.7 E. pauciflora 0.41 4.88 159.2 128.9 93.6 85.8% 17.6 9.4 402 Tests at 500oC Species Thickness SA/V Wet weight Dry weight Tin %moist Control 0.2 10.00 49.5 49.3 0 0.4% 0.0 1.5 Control 0.2 10.00 49.52 49.3 0 0.5% 0.0 1.4 Control 0.18 11.11 49.57 49.3 0 0.5% 0.0 1.7 Control 0.2 10.00 49.59 49.3 0 0.6% 0.1 1.8 Control 0.2 10.00 49.61 49.3 0 0.6% 0.1 1.6 Control 0.2 10.00 49.63 49.3 0 0.7% 0.1 1.8 Control 0.2 10.00 49.66 49.3 0 0.7% 0.1 1.5 Control 0.2 10.00 49.68 49.3 0 0.8% 0.1 1.5 Control 0.19 10.53 49.7 49.3 0 0.8% 0.1 1.7 Control 0.2 10.00 46.7 46.5 0 0.4% 0.1 2 Control 0.2 10.00 46.83 46.5 0 0.7% 0.0 2 Control 0.2 10.00 46.95 46.5 0 1.0% 0.1 1.8 Control 0.2 10.00 47.08 46.5 0 1.2% 0.1 1.7 Control 0.21 9.52 47.2 46.5 0 1.5% 0.1 1.7 E. stellulata 0.34 5.88 145.2 145.1 93.7 0.2% 0.0 2.5 E. stellulata 0.35 5.71 145.21 145.1 93.7 0.2% 0.0 1.9 E. stellulata 0.38 5.26 145.22 145.1 93.7 0.2% 0.0 2.1 E. stellulata 0.35 5.71 145.23 145.1 93.7 0.3% 0.0 2 E. stellulata 0.36 5.56 145.24 145.1 93.7 0.3% 0.1 2.1 E. stellulata 0.39 5.13 145.26 145.1 93.7 0.3% 0.1 2.5 E. stellulata 0.37 5.41 145.27 145.1 93.7 0.3% 0.1 2 E. stellulata 0.38 5.26 145.28 145.1 93.7 0.3% 0.1 1.9 E. stellulata 0.38 5.26 145.29 145.1 93.7 0.4% 0.1 1.8 E. stellulata 0.34 5.88 145.3 145.1 93.7 0.4% 0.1 2.4 E. stellulata 0.3 6.67 132.3 111.3 93.8 120.0% 18.0 6.4 E. stellulata 0.36 5.56 132.26 111.3 93.8 119.7% 21.6 6.9 E. stellulata 0.33 6.06 132.21 111.3 93.8 119.5% 19.7 6.8 E. stellulata 0.32 6.25 132.17 111.3 93.8 119.2% 19.1 6.3 E. stellulata 0.29 6.90 132.12 111.3 93.8 119.0% 17.3 5 E. stellulata 0.3 6.67 132.08 111.3 93.8 118.7% 17.8 5.9 E. stellulata 0.36 5.56 132.03 111.3 93.8 118.5% 21.3 7 E. stellulata 0.33 6.06 131.99 111.3 93.8 118.2% 19.5 6.2 E. stellulata 0.32 6.25 131.94 111.3 93.8 118.0% 18.9 6.4 E. stellulata 0.33 6.06 131.9 111.3 93.8 117.7% 19.4 6.2 Bossiaea foliosa 0.15 13.33 127.2 127.1 93.9 0.3% 0.0 3.1 Bossiaea foliosa 0.18 11.11 127.20 127.1 93.9 0.3% 0.0 2.5 Bossiaea foliosa 0.18 11.11 127.20 127.1 93.9 0.3% 0.0 2.6 Bossiaea foliosa 0.17 11.76 127.20 127.1 93.9 0.3% 0.0 2.1 Bossiaea foliosa 0.16 12.50 127.20 127.1 93.9 0.3% 0.0 2.7 Bossiaea foliosa 0.15 13.33 127.20 127.1 93.9 0.3% 0.0 2.5 Bossiaea foliosa 0.16 12.50 127.20 127.1 93.9 0.3% 0.0 2 Bossiaea foliosa 0.23 8.70 127.20 127.1 93.9 0.3% 0.0 3 Bossiaea foliosa 0.17 11.76 127.20 127.1 93.9 0.3% 0.0 2.7 Bossiaea foliosa 0.15 13.33 127.2 127.1 93.9 0.3% 0.0 2 Bossiaea foliosa 0.21 9.52 125 111.1 94 81.3% 8.5 5.4 Bossiaea foliosa 0.22 9.09 124.94 111.1 94 81.0% 8.9 5 Bossiaea foliosa 0.17 11.76 124.89 111.1 94 80.6% 6.9 4.3 Bossiaea foliosa 0.16 12.50 124.83 111.1 94 80.3% 6.4 5.2 Bossiaea foliosa 0.19 10.53 124.78 111.1 94 80.0% 7.6 5.1 Bossiaea foliosa 0.18 11.11 124.72 111.1 94 79.7% 7.2 4.7 Bossiaea foliosa 0.19 10.53 124.67 111.1 94 79.3% 7.5 4.3 M/SAV 403 Bossiaea foliosa 0.19 10.53 124.61 111.1 94 79.0% 7.5 4.9 Bossiaea foliosa 0.15 13.33 124.56 111.1 94 78.7% 5.9 3.4 Bossiaea foliosa 0.18 11.11 124.5 111.1 94 78.4% 7.1 5.2 Tasmannia xerophila 0.41 4.88 125.3 125.2 93.5 0.3% 0.1 2.7 Tasmannia xerophila 0.42 4.76 125.31 125.2 93.5 0.4% 0.1 1.7 Tasmannia xerophila 0.43 4.65 125.32 125.2 93.5 0.4% 0.1 3.8 Tasmannia xerophila 0.46 4.35 125.33 125.2 93.5 0.4% 0.1 2.4 Tasmannia xerophila 0.37 5.41 125.34 125.2 93.5 0.5% 0.1 2.1 Tasmannia xerophila 0.33 6.06 125.36 125.2 93.5 0.5% 0.1 1.2 Tasmannia xerophila 0.42 4.76 125.37 125.2 93.5 0.5% 0.1 3.8 Tasmannia xerophila 0.49 4.08 125.38 125.2 93.5 0.6% 0.1 3 Tasmannia xerophila 0.34 5.88 125.39 125.2 93.5 0.6% 0.1 2.4 Tasmannia xerophila 0.36 5.56 125.4 125.2 93.5 0.6% 0.1 1.4 Tasmannia xerophila 0.4 5.00 137.1 108.2 93.8 200.7% 40.1 7.7 Tasmannia xerophila 0.39 5.13 137.04 108.2 93.8 200.3% 39.1 9.1 Tasmannia xerophila 0.46 4.35 136.99 108.2 93.8 199.9% 46.0 9 Tasmannia xerophila 0.45 4.44 136.93 108.2 93.8 199.5% 44.9 11.1 Tasmannia xerophila 0.42 4.76 136.88 108.2 93.8 199.2% 41.8 8.9 Tasmannia xerophila 0.44 4.55 136.82 108.2 93.8 198.8% 43.7 9.4 Tasmannia xerophila 0.4 5.00 136.77 108.2 93.8 198.4% 39.7 9.6 Tasmannia xerophila 0.36 5.56 136.71 108.2 93.8 198.0% 35.6 7.6 Tasmannia xerophila 0.44 4.55 136.66 108.2 93.8 197.6% 43.5 10.5 Tasmannia xerophila 0.49 4.08 136.6 108.2 93.8 197.2% 48.3 10.8 Olearia aglossa 0.5 4.00 128.2 128.2 92.7 0.0% 0.0 1.9 Olearia aglossa 0.52 3.85 128.21 128.2 92.7 0.0% 0.0 2.3 Olearia aglossa 0.53 3.77 128.22 128.2 92.7 0.1% 0.0 2.1 Olearia aglossa 0.49 4.08 128.23 128.2 92.7 0.1% 0.0 2.2 Olearia aglossa 0.56 3.57 128.24 128.2 92.7 0.1% 0.0 2.3 Olearia aglossa 0.53 3.77 128.26 128.2 92.7 0.2% 0.0 2.7 Olearia aglossa 0.53 3.77 128.27 128.2 92.7 0.2% 0.0 2 Olearia aglossa 0.51 3.92 128.28 128.2 92.7 0.2% 0.1 1.7 Olearia aglossa 0.49 4.08 128.29 128.2 92.7 0.3% 0.1 2 Olearia aglossa 0.49 4.08 128.3 128.2 92.7 0.3% 0.1 1.7 Olearia aglossa 0.45 4.44 116.2 103.5 93.9 132.3% 29.8 5.4 Olearia aglossa 0.43 4.65 116.14 103.5 93.9 131.7% 28.3 6.2 Olearia aglossa 0.41 4.88 116.09 103.5 93.9 131.1% 26.9 6.1 Olearia aglossa 0.4 5.00 116.03 103.5 93.9 130.6% 26.1 5 Olearia aglossa 0.48 4.17 115.98 103.5 93.9 130.0% 31.2 7 Olearia aglossa 0.37 5.41 115.92 103.5 93.9 129.4% 23.9 4.7 Olearia aglossa 0.42 4.76 115.87 103.5 93.9 128.8% 27.1 6 Olearia aglossa 0.42 4.76 115.81 103.5 93.9 128.2% 26.9 6 Olearia aglossa 0.4 5.00 115.76 103.5 93.9 127.7% 25.5 5.3 5.1 Olearia aglossa 0.37 5.41 115.7 103.5 93.9 127.1% 23.5 Daviesia mimosoides 0.24 8.33 117.8 117.7 93.7 0.4% 0.0 3.9 Daviesia mimosoides 0.32 6.25 117.81 117.7 93.7 0.5% 0.1 3.5 Daviesia mimosoides 0.36 5.56 117.82 117.7 93.7 0.5% 0.1 3.2 Daviesia mimosoides 0.39 5.13 117.83 117.7 93.7 0.6% 0.1 3.4 Daviesia mimosoides 0.27 7.41 117.84 117.7 93.7 0.6% 0.1 3.5 Daviesia mimosoides 0.24 8.33 117.86 117.7 93.7 0.6% 0.1 3.3 Daviesia mimosoides 0.25 8.00 117.87 117.7 93.7 0.7% 0.1 3 Daviesia mimosoides 0.27 7.41 117.88 117.7 93.7 0.7% 0.1 3.1 Daviesia mimosoides 0.36 5.56 117.89 117.7 93.7 0.8% 0.1 3.3 Daviesia mimosoides 0.25 8.00 117.9 117.7 93.7 0.8% 0.1 3.2 Daviesia mimosoides 0.24 8.33 125.3 106.9 93 132.4% 15.9 4.2 404 Daviesia mimosoides 0.33 6.06 125.24 106.9 93 132.0% 21.8 5.7 Daviesia mimosoides 0.25 8.00 125.19 106.9 93 131.6% 16.4 5 Daviesia mimosoides 0.26 7.69 125.13 106.9 93 131.2% 17.1 4.7 Daviesia mimosoides 0.23 8.70 125.08 106.9 93 130.8% 15.0 3.8 Daviesia mimosoides 0.23 8.70 125.02 106.9 93 130.4% 15.0 5.4 Daviesia mimosoides 0.25 8.00 124.97 106.9 93 130.0% 16.2 4.7 Daviesia mimosoides 0.25 8.00 124.91 106.9 93 129.6% 16.2 4.6 Daviesia mimosoides 0.31 6.45 124.86 106.9 93 129.2% 20.0 6.5 Daviesia mimosoides 0.29 6.90 124.8 106.9 93 128.8% 18.7 6.3 E. pauciflora 0.48 4.17 138.3 138.3 93.5 0.0% 0.0 3 E. pauciflora 0.53 3.77 138.31 138.3 93.5 0.0% 0.0 2.3 E. pauciflora 0.43 4.65 138.32 138.3 93.5 0.0% 0.0 2.4 E. pauciflora 0.45 4.44 138.33 138.3 93.5 0.1% 0.0 2.2 E. pauciflora 0.48 4.17 138.34 138.3 93.5 0.1% 0.0 2.6 E. pauciflora 0.5 4.00 138.36 138.3 93.5 0.1% 0.0 2.7 E. pauciflora 0.51 3.92 138.37 138.3 93.5 0.1% 0.0 2.8 E. pauciflora 0.47 4.26 138.38 138.3 93.5 0.2% 0.0 2.5 E. pauciflora 0.5 4.00 138.39 138.3 93.5 0.2% 0.0 1.9 E. pauciflora 0.51 3.92 138.4 138.3 93.5 0.2% 0.1 2.5 E. pauciflora 0.61 3.28 115 102.5 92.5 125.0% 38.1 12.5 E. pauciflora 0.69 2.90 114.92 102.5 92.5 124.2% 42.9 14.6 E. pauciflora 0.58 3.45 114.84 102.5 92.5 123.4% 35.8 11.9 E. pauciflora 0.58 3.45 114.77 102.5 92.5 122.7% 35.6 10.9 E. pauciflora 0.59 3.39 114.69 102.5 92.5 121.9% 36.0 12.2 E. pauciflora 0.56 3.57 114.61 102.5 92.5 121.1% 33.9 12.8 E. pauciflora 0.56 3.57 114.53 102.5 92.5 120.3% 33.7 9.8 E. pauciflora 0.57 3.51 114.46 102.5 92.5 119.6% 34.1 11.3 E. pauciflora 0.61 3.28 114.38 102.5 92.5 118.8% 36.2 12 E. pauciflora 0.56 3.57 114.3 102.5 92.5 118.0% 33.0 11.3 405 Tests at 600oC Species Thickness SA/V Wet weight Dry weight Tin %moist Control 0.16 12.50 59.1 55.8 0 5.9% 0.5 1.1 Control 0.17 11.76 59.09 55.8 0 5.9% 0.5 0.7 Control 0.17 11.76 59.08 55.8 0 5.9% 0.5 1 Control 0.18 11.11 59.07 55.8 0 5.9% 0.5 1.1 Control 0.18 11.11 59.06 55.8 0 5.8% 0.5 0.9 Control 0.18 11.11 59.04 55.8 0 5.8% 0.5 0.8 Control 0.17 11.76 59.03 55.8 0 5.8% 0.5 0.7 Control 0.17 11.76 59.02 55.8 0 5.8% 0.5 0.9 Control 0.18 11.11 59.01 55.8 0 5.8% 0.5 1 Control 0.18 11.11 59 55.8 0 5.7% 0.5 0.8 Control 0.18 11.11 58.4 55.8 0 4.7% 0.4 0.7 Control 0.17 11.76 58.4 55.8 0 4.7% 0.4 1 E. stellulata 0.38 5.26 148.1 147.9 93.7 0.4% 0.1 1.2 E. stellulata 0.39 5.13 148.10 147.9 93.7 0.4% 0.1 1.1 E. stellulata 0.37 5.41 148.10 147.9 93.7 0.4% 0.1 1.3 E. stellulata 0.33 6.06 148.10 147.9 93.7 0.4% 0.1 1.3 E. stellulata 0.41 4.88 148.10 147.9 93.7 0.4% 0.1 1.5 E. stellulata 0.32 6.25 148.10 147.9 93.7 0.4% 0.1 1.3 E. stellulata 0.42 4.76 148.10 147.9 93.7 0.4% 0.1 0.9 E. stellulata 0.36 5.56 148.10 147.9 93.7 0.4% 0.1 1.2 E. stellulata 0.35 5.71 148.10 147.9 93.7 0.4% 0.1 1.1 E. stellulata 0.32 6.25 148.1 147.9 93.7 0.4% 0.1 1.2 E. stellulata 0.36 5.56 182.5 140.9 93.7 88.1% 15.9 3.3 E. stellulata 0.39 5.13 182.47 140.9 93.7 88.1% 17.2 4 E. stellulata 0.42 4.76 182.43 140.9 93.7 88.0% 18.5 4.6 E. stellulata 0.43 4.65 182.40 140.9 93.7 87.9% 18.9 3.5 E. stellulata 0.36 5.56 182.37 140.9 93.7 87.9% 15.8 3.1 E. stellulata 0.38 5.26 182.33 140.9 93.7 87.8% 16.7 3.8 E. stellulata 0.37 5.41 182.30 140.9 93.7 87.7% 16.2 3.7 E. stellulata 0.35 5.71 182.27 140.9 93.7 87.6% 15.3 4.1 E. stellulata 0.4 5.00 182.23 140.9 93.7 87.6% 17.5 4.7 M/SAV E. stellulata 0.39 5.13 182.2 140.9 93.7 87.5% 17.1 4.8 Bossiaea foliosa 0.15 13.33 127.9 127.4 93.9 1.5% 0.1 1.5 Bossiaea foliosa 0.15 13.33 127.92 127.4 93.9 1.6% 0.1 1.2 Bossiaea foliosa 0.16 12.50 127.94 127.4 93.9 1.6% 0.1 1 Bossiaea foliosa 0.15 13.33 127.97 127.4 93.9 1.7% 0.1 1.3 Bossiaea foliosa 0.16 12.50 127.99 127.4 93.9 1.8% 0.1 1.2 Bossiaea foliosa 0.18 11.11 128.01 127.4 93.9 1.8% 0.2 1.5 Bossiaea foliosa 0.18 11.11 128.03 127.4 93.9 1.9% 0.2 1.2 Bossiaea foliosa 0.16 12.50 128.06 127.4 93.9 2.0% 0.2 1.1 Bossiaea foliosa 0.17 11.76 128.08 127.4 93.9 2.0% 0.2 1.2 Bossiaea foliosa 0.18 11.11 128.1 127.4 93.9 2.1% 0.2 1.6 Bossiaea foliosa 0.16 12.50 113.7 108 93.8 40.1% 3.2 1.4 Bossiaea foliosa 0.14 14.29 113.68 108 93.8 40.0% 2.8 1.2 Bossiaea foliosa 0.14 14.29 113.66 108 93.8 39.8% 2.8 1.6 Bossiaea foliosa 0.14 14.29 113.63 108 93.8 39.7% 2.8 1.6 Bossiaea foliosa 0.13 15.38 113.61 108 93.8 39.5% 2.6 1.3 Bossiaea foliosa 0.17 11.76 113.59 108 93.8 39.4% 3.3 1.7 Bossiaea foliosa 0.18 11.11 113.57 108 93.8 39.2% 3.5 1.4 Bossiaea foliosa 0.16 12.50 113.54 108 93.8 39.0% 3.1 1.6 Bossiaea foliosa 0.18 11.11 113.52 108 93.8 38.9% 3.5 1.8 406 Bossiaea foliosa 0.15 13.33 113.5 108 93.8 38.7% 2.9 1 Tasmannia xerophila 0.49 4.08 127.2 126.4 93.5 2.4% 0.6 1.8 Tasmannia xerophila 0.42 4.76 127.21 126.4 93.5 2.5% 0.5 1.6 Tasmannia xerophila 0.3 6.67 127.22 126.4 93.5 2.5% 0.4 1.3 Tasmannia xerophila 0.43 4.65 127.23 126.4 93.5 2.5% 0.5 1.3 Tasmannia xerophila 0.45 4.44 127.24 126.4 93.5 2.6% 0.6 1 Tasmannia xerophila 0.4 5.00 127.26 126.4 93.5 2.6% 0.5 1 Tasmannia xerophila 0.41 4.88 127.27 126.4 93.5 2.6% 0.5 1.2 Tasmannia xerophila 0.49 4.08 127.28 126.4 93.5 2.7% 0.7 1.3 Tasmannia xerophila 0.45 4.44 127.29 126.4 93.5 2.7% 0.6 1.1 Tasmannia xerophila 0.33 6.06 127.3 126.4 93.5 2.7% 0.5 1.6 Tasmannia xerophila 0.49 4.08 126.8 108.5 94.1 127.1% 31.1 4.4 Tasmannia xerophila 0.44 4.55 126.78 108.5 94.1 126.9% 27.9 5.2 Tasmannia xerophila 0.47 4.26 126.76 108.5 94.1 126.8% 29.8 5.6 Tasmannia xerophila 0.45 4.44 126.73 108.5 94.1 126.6% 28.5 6.2 Tasmannia xerophila 0.53 3.77 126.71 108.5 94.1 126.5% 33.5 6.5 Tasmannia xerophila 0.52 3.85 126.69 108.5 94.1 126.3% 32.8 6 Tasmannia xerophila 0.58 3.45 126.67 108.5 94.1 126.2% 36.6 5.9 Tasmannia xerophila 0.5 4.00 126.64 108.5 94.1 126.0% 31.5 7.1 Tasmannia xerophila 0.43 4.65 126.62 108.5 94.1 125.8% 27.1 5.1 Tasmannia xerophila 0.54 3.70 126.6 108.5 94.1 125.7% 33.9 6.5 Olearia aglossa 0.52 3.85 130.4 130 92.7 1.1% 0.3 1.3 Olearia aglossa 0.48 4.17 130.41 130 92.7 1.1% 0.3 1.2 Olearia aglossa 0.49 4.08 130.42 130 92.7 1.1% 0.3 1.4 Olearia aglossa 0.51 3.92 130.43 130 92.7 1.2% 0.3 1.4 Olearia aglossa 0.49 4.08 130.44 130 92.7 1.2% 0.3 1.2 Olearia aglossa 0.47 4.26 130.46 130 92.7 1.2% 0.3 1.1 Olearia aglossa 0.5 4.00 130.47 130 92.7 1.3% 0.3 1.2 Olearia aglossa 0.58 3.45 130.48 130 92.7 1.3% 0.4 0.9 Olearia aglossa 0.47 4.26 130.49 130 92.7 1.3% 0.3 1.2 Olearia aglossa 0.49 4.08 130.5 130 92.7 1.3% 0.3 1 Olearia aglossa 0.6 3.33 116.9 107.3 93.6 70.1% 21.0 3.7 Olearia aglossa 0.62 3.23 116.88 107.3 93.6 69.9% 21.7 4.3 Olearia aglossa 0.62 3.23 116.86 107.3 93.6 69.7% 21.6 3.4 Olearia aglossa 0.66 3.03 116.83 107.3 93.6 69.6% 23.0 3.5 Olearia aglossa 0.65 3.08 116.81 107.3 93.6 69.4% 22.6 3.4 Olearia aglossa 0.69 2.90 116.79 107.3 93.6 69.3% 23.9 4.1 Olearia aglossa 0.64 3.13 116.77 107.3 93.6 69.1% 22.1 3.7 Olearia aglossa 0.64 3.13 116.74 107.3 93.6 68.9% 22.1 4.4 Olearia aglossa 0.58 3.45 116.72 107.3 93.6 68.8% 19.9 4.3 Olearia aglossa 0.64 3.13 116.7 107.3 93.6 68.6% 22.0 4.8 Daviesia mimosoides 0.39 5.13 119.1 118.6 93.7 2.0% 0.4 1.7 Daviesia mimosoides 0.36 5.56 119.10 118.6 93.7 2.0% 0.4 1.6 Daviesia mimosoides 0.3 6.67 119.10 118.6 93.7 2.0% 0.3 1.3 Daviesia mimosoides 0.4 5.00 119.10 118.6 93.7 2.0% 0.4 1.3 Daviesia mimosoides 0.4 5.00 119.10 118.6 93.7 2.0% 0.4 1.5 Daviesia mimosoides 0.32 6.25 119.10 118.6 93.7 2.0% 0.3 1.3 Daviesia mimosoides 0.37 5.41 119.10 118.6 93.7 2.0% 0.4 1.5 Daviesia mimosoides 0.4 5.00 119.10 118.6 93.7 2.0% 0.4 1.2 Daviesia mimosoides 0.43 4.65 119.10 118.6 93.7 2.0% 0.4 1.4 Daviesia mimosoides 0.32 6.25 119.1 118.6 93.7 2.0% 0.3 1.4 Daviesia mimosoides 0.31 6.45 115.2 105.5 94 84.3% 13.1 2.7 Daviesia mimosoides 0.31 6.45 115.17 105.5 94 84.1% 13.0 3.3 Daviesia mimosoides 0.34 5.88 115.13 105.5 94 83.8% 14.2 4.2 407 Daviesia mimosoides 0.29 6.90 115.10 105.5 94 83.5% 12.1 3.7 Daviesia mimosoides 0.34 5.88 115.07 105.5 94 83.2% 14.1 4 Daviesia mimosoides 0.35 5.71 115.03 105.5 94 82.9% 14.5 3.7 Daviesia mimosoides 0.34 5.88 115.00 105.5 94 82.6% 14.0 4 Daviesia mimosoides 0.3 6.67 114.97 105.5 94 82.3% 12.3 3.3 Daviesia mimosoides 0.34 5.88 114.93 105.5 94 82.0% 13.9 4 Daviesia mimosoides 0.42 4.76 114.9 105.5 94 81.7% 17.2 4.6 E. pauciflora 0.54 3.70 144.5 143.2 93.5 2.6% 0.7 1.5 E. pauciflora 0.45 4.44 144.51 143.2 93.5 2.6% 0.6 1.8 E. pauciflora 0.46 4.35 144.52 143.2 93.5 2.7% 0.6 1.5 E. pauciflora 0.52 3.85 144.53 143.2 93.5 2.7% 0.7 1.8 E. pauciflora 0.5 4.00 144.54 143.2 93.5 2.7% 0.7 1.6 E. pauciflora 0.48 4.17 144.56 143.2 93.5 2.7% 0.7 1.3 E. pauciflora 0.55 3.64 144.57 143.2 93.5 2.7% 0.8 1.7 E. pauciflora 0.46 4.35 144.58 143.2 93.5 2.8% 0.6 1.6 E. pauciflora 0.41 4.88 144.59 143.2 93.5 2.8% 0.6 1.4 E. pauciflora 0.48 4.17 144.6 143.2 93.5 2.8% 0.7 2 E. pauciflora 0.56 3.57 158.9 128.9 93.6 85.0% 23.8 4.9 E. pauciflora 0.52 3.85 158.86 128.9 93.6 84.9% 22.1 5.1 E. pauciflora 0.52 3.85 158.81 128.9 93.6 84.7% 22.0 5 E. pauciflora 0.55 3.64 158.77 128.9 93.6 84.6% 23.3 5.1 E. pauciflora 0.48 4.17 158.72 128.9 93.6 84.5% 20.3 4.9 E. pauciflora 0.5 4.00 158.68 128.9 93.6 84.4% 21.1 4.9 E. pauciflora 0.46 4.35 158.63 128.9 93.6 84.2% 19.4 3.6 E. pauciflora 0.53 3.77 158.59 128.9 93.6 84.1% 22.3 5.3 E. pauciflora 0.48 4.17 158.54 128.9 93.6 84.0% 20.2 5 E. pauciflora 0.52 3.85 158.5 128.9 93.6 83.9% 21.8 5.6 408 Tests at 700oC Species Thickness SA/V Wet weight Dry weight Tin %moist Control 0.17 11.76 58.4 55.8 0 4.7% 0.4 0.7 Control 0.18 11.11 58.38 55.8 0 4.6% 0.4 0.6 Control 0.17 11.76 58.36 55.8 0 4.6% 0.4 0.6 Control 0.19 10.53 58.33 55.8 0 4.5% 0.4 0.7 Control 0.18 11.11 58.31 55.8 0 4.5% 0.4 0.5 Control 0.17 11.76 58.29 55.8 0 4.5% 0.4 0.6 Control 0.17 11.76 58.27 55.8 0 4.4% 0.4 0.6 Control 0.18 11.11 58.24 55.8 0 4.4% 0.4 0.6 Control 0.16 12.50 58.22 55.8 0 4.3% 0.3 0.6 Control 0.17 11.76 58.2 55.8 0 4.3% 0.4 0.6 E. stellulata 0.34 5.88 148.2 147.9 93.7 0.6% 0.1 0.8 E. stellulata 0.36 5.56 148.22 147.9 93.7 0.6% 0.1 0.8 E. stellulata 0.29 6.90 148.24 147.9 93.7 0.6% 0.1 0.6 E. stellulata 0.36 5.56 148.27 147.9 93.7 0.7% 0.1 1 E. stellulata 0.37 5.41 148.29 147.9 93.7 0.7% 0.1 1 E. stellulata 0.29 6.90 148.31 147.9 93.7 0.8% 0.1 1.2 E. stellulata 0.32 6.25 148.33 147.9 93.7 0.8% 0.1 1 E. stellulata 0.38 5.26 148.36 147.9 93.7 0.8% 0.2 1.2 E. stellulata 0.37 5.41 148.38 147.9 93.7 0.9% 0.2 1.2 E. stellulata 0.44 4.55 148.4 147.9 93.7 0.9% 0.2 0.9 E. stellulata 0.28 7.14 113.7 104.6 94 85.8% 12.0 1.9 E. stellulata 0.39 5.13 113.62 104.6 94 85.1% 16.6 3 E. stellulata 0.4 5.00 113.54 104.6 94 84.4% 16.9 2.6 E. stellulata 0.28 7.14 113.47 104.6 94 83.6% 11.7 1.9 E. stellulata 0.41 4.88 113.39 104.6 94 82.9% 17.0 3.1 E. stellulata 0.28 7.14 113.31 104.6 94 82.2% 11.5 1.7 E. stellulata 0.44 4.55 113.23 104.6 94 81.4% 17.9 3.1 E. stellulata 0.26 7.69 113.16 104.6 94 80.7% 10.5 1.7 E. stellulata 0.35 5.71 113.08 104.6 94 80.0% 14.0 1.6 E. stellulata 0.32 6.25 113 104.6 94 79.2% 12.7 2.5 Bossiaea foliosa 0.21 9.52 127.4 127.3 93.9 0.3% 0.0 0.9 Bossiaea foliosa 0.17 11.76 127.44 127.3 93.9 0.4% 0.0 0.7 Bossiaea foliosa 0.23 8.70 127.49 127.3 93.9 0.6% 0.1 1.1 Bossiaea foliosa 0.19 10.53 127.53 127.3 93.9 0.7% 0.1 0.6 Bossiaea foliosa 0.16 12.50 127.58 127.3 93.9 0.8% 0.1 1.1 Bossiaea foliosa 0.19 10.53 127.62 127.3 93.9 1.0% 0.1 1 Bossiaea foliosa 0.14 14.29 127.67 127.3 93.9 1.1% 0.1 0.8 M/SAV Bossiaea foliosa 0.2 10.00 127.71 127.3 93.9 1.2% 0.1 1 Bossiaea foliosa 0.17 11.76 127.76 127.3 93.9 1.4% 0.1 0.7 Bossiaea foliosa 0.21 9.52 127.8 127.3 93.9 1.5% 0.2 1.1 Bossiaea foliosa 0.13 15.38 109.7 105.8 93.4 31.5% 2.0 1.6 Bossiaea foliosa 0.17 11.76 109.61 105.8 93.4 30.7% 2.6 1.9 Bossiaea foliosa 0.2 10.00 109.52 105.8 93.4 30.0% 3.0 1.3 Bossiaea foliosa 0.17 11.76 109.43 105.8 93.4 29.3% 2.5 1.3 Bossiaea foliosa 0.18 11.11 109.34 105.8 93.4 28.6% 2.6 1.1 Bossiaea foliosa 0.16 12.50 109.26 105.8 93.4 27.9% 2.2 1.5 Bossiaea foliosa 0.19 10.53 109.17 105.8 93.4 27.2% 2.6 1.7 Bossiaea foliosa 0.17 11.76 109.08 105.8 93.4 26.4% 2.2 1.2 Bossiaea foliosa 0.15 13.33 108.99 105.8 93.4 25.7% 1.9 1.2 Bossiaea foliosa 0.17 11.76 108.9 105.8 93.4 25.0% 2.1 1 Tasmannia xerophila 0.33 6.06 126.4 126.2 93.5 0.6% 0.1 0.7 409 Tasmannia xerophila 0.43 4.65 126.42 126.2 93.5 0.7% 0.1 0.9 Tasmannia xerophila 0.43 4.65 126.44 126.2 93.5 0.7% 0.2 1 Tasmannia xerophila 0.49 4.08 126.47 126.2 93.5 0.8% 0.2 0.6 Tasmannia xerophila 0.48 4.17 126.49 126.2 93.5 0.9% 0.2 1.3 Tasmannia xerophila 0.37 5.41 126.51 126.2 93.5 1.0% 0.2 0.8 Tasmannia xerophila 0.49 4.08 126.53 126.2 93.5 1.0% 0.2 0.6 Tasmannia xerophila 0.4 5.00 126.56 126.2 93.5 1.1% 0.2 0.6 Tasmannia xerophila 0.48 4.17 126.58 126.2 93.5 1.2% 0.3 0.7 Tasmannia xerophila 0.42 4.76 126.6 126.2 93.5 1.2% 0.3 0.7 Tasmannia xerophila 0.53 3.77 129.4 108.8 93.3 132.9% 35.2 3.7 Tasmannia xerophila 0.47 4.26 129.32 108.8 93.3 132.4% 31.1 4.4 Tasmannia xerophila 0.5 4.00 129.24 108.8 93.3 131.9% 33.0 3.9 Tasmannia xerophila 0.51 3.92 129.17 108.8 93.3 131.4% 33.5 4.5 Tasmannia xerophila 0.51 3.92 129.09 108.8 93.3 130.9% 33.4 3.8 Tasmannia xerophila 0.51 3.92 129.01 108.8 93.3 130.4% 33.3 4 Tasmannia xerophila 0.46 4.35 128.93 108.8 93.3 129.9% 29.9 3.4 Tasmannia xerophila 0.48 4.17 128.86 108.8 93.3 129.4% 31.1 3.6 Tasmannia xerophila 0.46 4.35 128.78 108.8 93.3 128.9% 29.6 4.2 Tasmannia xerophila 0.45 4.44 128.7 108.8 93.3 128.4% 28.9 3.7 Olearia aglossa 0.59 3.39 130 129.8 92.7 0.5% 0.2 0.8 Olearia aglossa 0.65 3.08 130.02 129.8 92.7 0.6% 0.2 0.8 Olearia aglossa 0.67 2.99 130.04 129.8 92.7 0.7% 0.2 1 Olearia aglossa 0.58 3.45 130.07 129.8 92.7 0.7% 0.2 0.9 Olearia aglossa 0.67 2.99 130.09 129.8 92.7 0.8% 0.3 0.6 Olearia aglossa 0.62 3.23 130.11 129.8 92.7 0.8% 0.3 1 Olearia aglossa 0.65 3.08 130.13 129.8 92.7 0.9% 0.3 0.7 Olearia aglossa 0.62 3.23 130.16 129.8 92.7 1.0% 0.3 0.7 Olearia aglossa 0.62 3.23 130.18 129.8 92.7 1.0% 0.3 1.2 Olearia aglossa 0.63 3.17 130.2 129.8 92.7 1.1% 0.3 0.8 Olearia aglossa 0.58 3.45 110.7 102.9 93.6 83.9% 24.3 3.6 Olearia aglossa 0.54 3.70 110.64 102.9 93.6 83.3% 22.5 2.9 Olearia aglossa 0.5 4.00 110.59 102.9 93.6 82.7% 20.7 2.4 Olearia aglossa 0.55 3.64 110.53 102.9 93.6 82.1% 22.6 3 Olearia aglossa 0.59 3.39 110.48 102.9 93.6 81.5% 24.0 1.9 Olearia aglossa 0.68 2.94 110.42 102.9 93.6 80.9% 27.5 3.1 Olearia aglossa 0.57 3.51 110.37 102.9 93.6 80.3% 22.9 3.3 Olearia aglossa 0.52 3.85 110.31 102.9 93.6 79.7% 20.7 2.6 Olearia aglossa 0.55 3.64 110.26 102.9 93.6 79.1% 21.8 2.6 Olearia aglossa 0.55 3.64 110.2 102.9 93.6 78.5% 21.6 3.1 Daviesia mimosoides 0.35 5.71 118.6 118.5 93.7 0.4% 0.1 0.8 Daviesia mimosoides 0.37 5.41 118.62 118.5 93.7 0.5% 0.1 1.1 Daviesia mimosoides 0.31 6.45 118.64 118.5 93.7 0.6% 0.1 1.1 Daviesia mimosoides 0.33 6.06 118.67 118.5 93.7 0.7% 0.1 1 Daviesia mimosoides 0.33 6.06 118.69 118.5 93.7 0.8% 0.1 0.8 Daviesia mimosoides 0.32 6.25 118.71 118.5 93.7 0.9% 0.1 1 Daviesia mimosoides 0.32 6.25 118.73 118.5 93.7 0.9% 0.2 0.8 Daviesia mimosoides 0.31 6.45 118.76 118.5 93.7 1.0% 0.2 0.9 Daviesia mimosoides 0.32 6.25 118.78 118.5 93.7 1.1% 0.2 0.8 Daviesia mimosoides 0.34 5.88 118.8 118.5 93.7 1.2% 0.2 0.8 Daviesia mimosoides 0.4 5.00 107.3 101.1 93.3 79.5% 15.9 3 Daviesia mimosoides 0.42 4.76 107.23 101.1 93.3 78.6% 16.5 2.9 Daviesia mimosoides 0.35 5.71 107.17 101.1 93.3 77.8% 13.6 2.7 Daviesia mimosoides 0.34 5.88 107.10 101.1 93.3 76.9% 13.1 2.2 Daviesia mimosoides 0.34 5.88 107.03 101.1 93.3 76.1% 12.9 2.8 410 Daviesia mimosoides 0.33 6.06 106.97 101.1 93.3 75.2% 12.4 2.3 Daviesia mimosoides 0.44 4.55 106.90 101.1 93.3 74.4% 16.4 3.6 Daviesia mimosoides 0.37 5.41 106.83 101.1 93.3 73.5% 13.6 1.8 Daviesia mimosoides 0.38 5.26 106.77 101.1 93.3 72.6% 13.8 3.1 Daviesia mimosoides 0.33 6.06 106.7 101.1 93.3 71.8% 11.8 2 E. pauciflora 0.45 4.44 143.2 143 93.5 0.4% 0.1 0.6 E. pauciflora 0.41 4.88 143.23 143 93.5 0.5% 0.1 0.4 E. pauciflora 0.5 4.00 143.27 143 93.5 0.5% 0.1 0.9 E. pauciflora 0.48 4.17 143.30 143 93.5 0.6% 0.1 1.1 E. pauciflora 0.42 4.76 143.33 143 93.5 0.7% 0.1 1 E. pauciflora 0.44 4.55 143.37 143 93.5 0.7% 0.2 0.8 E. pauciflora 0.5 4.00 143.40 143 93.5 0.8% 0.2 0.7 E. pauciflora 0.46 4.35 143.43 143 93.5 0.9% 0.2 0.8 E. pauciflora 0.48 4.17 143.47 143 93.5 0.9% 0.2 1.2 E. pauciflora 0.47 4.26 143.5 143 93.5 1.0% 0.2 1 E. pauciflora 0.48 4.17 112.4 105.2 94.1 64.9% 15.6 3.3 E. pauciflora 0.44 4.55 112.30 105.2 94.1 64.0% 14.1 3.3 E. pauciflora 0.46 4.35 112.20 105.2 94.1 63.1% 14.5 2.4 E. pauciflora 0.47 4.26 112.10 105.2 94.1 62.2% 14.6 2.6 E. pauciflora 0.45 4.44 112.00 105.2 94.1 61.3% 13.8 2.7 E. pauciflora 0.5 4.00 111.90 105.2 94.1 60.4% 15.1 2.7 E. pauciflora 0.42 4.76 111.80 105.2 94.1 59.5% 12.5 3 E. pauciflora 0.47 4.26 111.70 105.2 94.1 58.6% 13.8 3.2 E. pauciflora 0.58 3.45 111.60 105.2 94.1 57.7% 16.7 3.9 E. pauciflora 0.44 4.55 111.5 105.2 94.1 56.8% 12.5 2.8 411 Appendix IV: Length and duration of flames from different leaf species Species Daviesia mimosoides G Daviesia mimosoides G Daviesia mimosoides G Daviesia mimosoides G Daviesia mimosoides G Daviesia mimosoides G Daviesia mimosoides G Daviesia mimosoides S Daviesia mimosoides S Daviesia mimosoides S Daviesia mimosoides S Daviesia mimosoides S Daviesia mimosoides S Daviesia mimosoides S Daviesia mimosoides D Daviesia mimosoides D Daviesia mimosoides D Daviesia mimosoides D Daviesia mimosoides D Daviesia mimosoides D Daviesia mimosoides D Brachyloma daphnoides G Brachyloma daphnoides G Brachyloma daphnoides G Brachyloma daphnoides G Brachyloma daphnoides G Brachyloma daphnoides G Brachyloma daphnoides G Brachyloma daphnoides S Brachyloma daphnoides S Brachyloma daphnoides S Brachyloma daphnoides S Brachyloma daphnoides S Brachyloma daphnoides S Brachyloma daphnoides S Brachyloma daphnoides D Brachyloma daphnoides D Brachyloma daphnoides D Brachyloma daphnoides D Brachyloma daphnoides D Brachyloma daphnoides D Brachyloma daphnoides D Eucalyptus pauciflora G Eucalyptus pauciflora G Eucalyptus pauciflora G Eucalyptus pauciflora G Eucalyptus pauciflora G Eucalyptus pauciflora G Eucalyptus pauciflora G Eucalyptus pauciflora S Eucalyptus pauciflora S Eucalyptus pauciflora S Eucalyptus pauciflora S Eucalyptus pauciflora S Leaf thickness (mm) Leaf width (mm) Burnt length (mm) Volatile oils (score) Leaf Moisture (% ODW) Max flame length (mm) Parallax corrected (mm) Flame duration (s) 0.5 0.5 0.4 0.4 0.4 0.4 0.4 0.3 0.4 0.4 0.4 0.6 0.4 0.3 0.3 0.4 0.3 0.3 0.3 0.4 0.4 0.2 0.3 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.3 0.3 0.2 0.3 0.2 0.3 0.2 0.1 0.2 0.2 0.2 0.5 0.6 0.6 0.5 0.5 0.5 0.5 0.5 0.4 0.5 0.4 0.5 7 7 7 7 7 7 7 7 7 7 7 7 10 7 6 8 9 8 6 11 11 1.5 1.5 1.5 1.5 1.5 2 3 3 2 3 4 3 3 3 1 1 2 1 1.5 2 2 20 20 20 20 20 25 28 24 40 33 36 26 35 36 30 32 34 40 45 25 37 25 27 37 60 31 50 60 70 60 64 70 66 6 7 6 8 7 4 5 5 4 5 6 5 6 6 0.5 0.5 0.5 0.7 0.6 4 4 40 50 13 12 15 32 96 90 135 125 140 95 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 104.5 104.5 104.5 104.5 104.5 106.3 106.3 91.3 91.3 91.3 91.3 91.3 53.3 53.3 1.0 1.0 1.0 1.0 1.0 1.8 1.8 128.3 128.3 128.3 128.3 128.3 129.3 130.3 90.4 90.4 90.4 90.4 90.4 90.4 90.4 1.7 1.7 1.7 1.7 1.7 2.7 3.7 122.5 122.5 122.5 122.5 122.5 103.2 103.2 62.8 62.8 62.8 62.8 62.8 20 60 60 30 50 25 25 30 50 80 20 40 95 100 85 100 110 105 110 170 180 10 10 10 10 10 10 15 15 10 10 10 10 10 10 15 15 10 10 10 10 15 100 70 70 70 70 80 120 140 170 210 240 230 19.3 57.8 57.8 28.9 48.2 24.1 24.1 28.9 48.2 77.1 19.3 38.5 91.5 96.4 81.9 96.4 106.0 101.2 106.0 163.8 173.4 9.6 9.6 9.6 9.6 9.6 9.6 14.5 14.5 9.6 9.6 9.6 9.6 9.6 9.6 14.5 14.5 9.6 9.6 9.6 9.6 14.5 96.4 67.5 67.5 67.5 67.5 77.1 115.6 134.9 163.8 202.3 231.2 221.6 11 13 13 9 12 6 7 7 8 4 4 6 8 4 5 7 7 7 7 6.5 8 1 1 1 2 1.5 1.5 2 2 1.5 2 2.5 2 2 2 2 1 1 1 1 1 1 27 13 16 21 13 19 25 20 25 22 15 20 412 Eucalyptus pauciflora S Eucalyptus pauciflora S Eucalyptus pauciflora D Eucalyptus pauciflora D Eucalyptus pauciflora D Eucalyptus pauciflora D Eucalyptus pauciflora D Eucalyptus pauciflora D Eucalyptus pauciflora D Tasmannia xerophila G Tasmannia xerophila G Tasmannia xerophila G Tasmannia xerophila G Tasmannia xerophila G Tasmannia xerophila G Tasmannia xerophila G Tasmannia xerophila S Tasmannia xerophila S Tasmannia xerophila S Tasmannia xerophila S Tasmannia xerophila S Tasmannia xerophila S Tasmannia xerophila S Tasmannia xerophila D Tasmannia xerophila D Tasmannia xerophila D Tasmannia xerophila D Tasmannia xerophila D Tasmannia xerophila D Tasmannia xerophila D Orites lancifolia G Orites lancifolia G Orites lancifolia G Orites lancifolia G Orites lancifolia G Orites lancifolia G Orites lancifolia G Orites lancifolia S Orites lancifolia S Orites lancifolia S Orites lancifolia S Orites lancifolia S Orites lancifolia S Orites lancifolia S Orites lancifolia D Orites lancifolia D Orites lancifolia D Orites lancifolia D Orites lancifolia D Orites lancifolia D Orites lancifolia D Prostanthera cuneata G Prostanthera cuneata G Prostanthera cuneata G Prostanthera cuneata G Prostanthera cuneata G Prostanthera cuneata G Prostanthera cuneata G Prostanthera cuneata S Prostanthera cuneata S Prostanthera cuneata S Prostanthera cuneata S Prostanthera cuneata S Prostanthera cuneata S 0.5 0.6 0.4 0.3 0.4 0.4 0.5 0.5 0.5 0.5 0.5 0.3 0.4 0.3 0.3 0.4 0.5 0.3 0.3 0.3 0.3 0.3 0.3 0.4 0.4 0.4 0.4 0.4 0.3 0.3 0.5 0.4 0.3 0.4 0.5 0.5 0.5 0.4 0.4 0.4 0.5 0.3 0.4 0.3 0.4 0.4 0.4 0.4 0.4 0.4 0.5 0.5 0.4 0.5 0.4 0.4 0.4 0.4 0.2 0.2 0.3 0.2 0.2 0.2 35 23 28 28 20 25 32 29 28 8 8 7 10 8 6 10 8 11 10 13 10 11 13 8 7 8 9 8 8 8 10 15 13 11 9 12 7 14 15 12 12 18 7 7 14 9 13 16 12 12.5 12 5 4 4 3 3 3 3 4 4 4 4 4 2 127 85 96 78 82 92 108 108 101 18 20 36 51 35 29 56 19 45 46 56 45 62 63 30 40 42 43 42 40 40 32 44 37 26 35 32 25 23 26 29 40 27 13 14 38 36 36 42 34 38 35 6 6 6 5 5 5 5 4 4 5 5 6 3 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 3 3 3 3 3 3 3 3 3 3 3 3 62.8 62.8 2.4 2.4 2.4 2.4 0.7 0.7 0.7 140.4 140.4 110.3 110.3 110.3 110.3 110.3 122.7 83.9 83.9 83.9 83.9 83.9 83.9 2.1 2.1 2.1 2.1 2.1 2.2 2.2 124.7 124.7 124.7 124.7 124.7 124.7 125.7 81.1 81.1 81.1 81.1 81.1 81.1 81.1 2.2 2.2 2.2 2.2 2.2 2.2 2.2 185.9 185.9 185.9 185.9 185.9 185.9 185.9 63.5 63.5 63.5 63.5 63.5 63.5 240 180 180 190 200 200 250 210 230 45 70 40 70 75 35 60 50 80 70 90 50 70 70 90 90 80 120 130 150 90 65 70 40 50 65 65 30 85 55 55 70 100 60 25 110 120 120 110 135 130 145 10 5 10 10 5 5 10 10 5 10 10 30 5 231.2 173.4 173.4 183.1 192.7 192.7 240.8 202.3 221.6 43.4 67.5 38.5 67.5 72.3 33.7 57.8 48.2 77.1 67.5 86.7 48.2 67.5 67.5 86.7 86.7 77.1 115.6 125.3 144.5 86.7 62.6 67.5 38.5 48.2 62.6 62.6 28.9 81.9 53.0 53.0 67.5 96.4 57.8 24.1 106.0 115.6 115.6 106.0 130.1 125.3 139.7 9.6 4.8 9.6 9.6 4.8 4.8 9.6 9.6 4.8 9.6 9.6 28.9 4.8 17 25 16 10 10 13 16 17 14 7 10 3 7 3 4 7 8 7 6 6 5.5 6 5.5 5 5 6 5 5 3 4.5 7 13 11 5 7 5 5.5 9 6 7 6 8 3.5 5 4 6 6 7 7 6 5 2 1 2 1.5 1 1 1.5 2 1 1 2 1 0.5 413 Prostanthera cuneata S Prostanthera cuneata D Prostanthera cuneata D Prostanthera cuneata D Prostanthera cuneata D Prostanthera cuneata D Prostanthera cuneata D Prostanthera cuneata D Phebalium ovatifolium G Phebalium ovatifolium G Phebalium ovatifolium G Phebalium ovatifolium G Phebalium ovatifolium G Phebalium ovatifolium G Phebalium ovatifolium G Phebalium ovatifolium S Phebalium ovatifolium S Phebalium ovatifolium S Phebalium ovatifolium S Phebalium ovatifolium S Phebalium ovatifolium S Phebalium ovatifolium S Phebalium ovatifolium D Phebalium ovatifolium D Phebalium ovatifolium D Phebalium ovatifolium D Phebalium ovatifolium D Phebalium ovatifolium D Phebalium ovatifolium D Helichrysum thyrsoideum G Helichrysum thyrsoideum G Helichrysum thyrsoideum G Helichrysum thyrsoideum G Helichrysum thyrsoideum G Helichrysum thyrsoideum G Helichrysum thyrsoideum G Helichrysum thyrsoideum S Helichrysum thyrsoideum S Helichrysum thyrsoideum S Helichrysum thyrsoideum S Helichrysum thyrsoideum S Helichrysum thyrsoideum S Helichrysum thyrsoideum S Helichrysum thyrsoideum D Helichrysum thyrsoideum D Helichrysum thyrsoideum D Helichrysum thyrsoideum D Helichrysum thyrsoideum D Helichrysum thyrsoideum D Helichrysum thyrsoideum D Eucalyptus niphophila G Eucalyptus niphophila G Eucalyptus niphophila G Eucalyptus niphophila G Eucalyptus niphophila G Eucalyptus niphophila G Eucalyptus niphophila G Eucalyptus niphophila S Eucalyptus niphophila S Eucalyptus niphophila S Eucalyptus niphophila S Eucalyptus niphophila S Eucalyptus niphophila S Eucalyptus niphophila S 0.2 0.4 0.4 0.3 0.4 0.4 0.4 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.3 0.4 0.3 0.3 0.3 0.2 0.3 0.2 0.2 0.3 0.3 0.3 0.3 0.2 0.2 0.3 0.1 0.3 0.2 0.3 0.2 0.3 0.3 0.2 0.6 0.7 0.7 0.6 0.7 0.6 0.7 0.5 0.6 0.4 0.5 0.5 0.5 0.4 3 3 3 2 4 3 3.5 4 9 8 10 9 9 7 6 7 6 7 6 7 7 7 7 6 5 6 5 8 5 3 2 3 2 2 2 3 2 1.5 2 1 2 1.5 1 1.5 1.5 2 1.5 2 1.5 1 10 11 14 11 12 14 15.5 16 21 21 24 14 14 16 5 5 4 4 5 5 4 4 12 10 11 12 10 9 9 11 10 11 10 11 10 10 10 9 9 7 9 10 6 10 7 9 5 9 10 12 5 4 5 4 5 5 10 8 6 6 6 6 7 9 58 62 72 60 64 56 75 47 60 65 75 46 55 47 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 63.5 2.0 2.0 2.0 2.0 2.0 2.0 2.0 129.8 129.8 129.8 129.8 129.8 129.8 130.8 67.0 67.0 67.0 67.0 67.0 67.0 67.0 2.4 2.4 2.4 2.4 2.4 2.4 2.4 176.4 176.4 176.4 176.4 176.4 176.4 176.4 42.9 42.9 42.9 42.9 42.9 42.9 37.9 2.8 2.8 2.8 2.8 2.8 2.8 2.4 110.5 110.5 110.5 110.5 110.5 110.5 110.5 51.3 51.3 51.3 51.3 51.3 51.3 51.3 15 5 10 5 15 15 5 15 25 20 20 25 20 15 10 25 25 25 15 40 30 10 30 65 70 30 70 70 25 5 5 5 0 15 5 20 15 10 10 15 15 15 20 25 20 20 15 30 30 60 90 110 140 80 120 150 130 100 190 130 170 110 150 125 14.5 4.8 9.6 4.8 14.5 14.5 4.8 14.5 24.1 19.3 19.3 24.1 19.3 14.5 9.6 24.1 24.1 24.1 14.5 38.5 28.9 9.6 28.9 62.6 67.5 28.9 67.5 67.5 24.1 4.8 4.8 4.8 0.0 14.5 4.8 19.3 14.5 9.6 9.6 14.5 14.5 14.5 19.3 24.1 19.3 19.3 14.5 28.9 28.9 57.8 86.7 106.0 134.9 77.1 115.6 144.5 125.3 96.4 183.1 125.3 163.8 106.0 144.5 120.4 1.5 1 2 1 1.5 2 1 1.5 6 5 6 6 5 4 5 7 6 6 4 5 5 4 3 3 3 3 1.5 3 3 1 3 0.5 0 2 2 2.5 0.5 1 1.5 0.5 1 1 1.5 1 1 1 1 1 1 1 19 26 19 20 16 16 36 12 13 20 18 18 10 9.5 414 Eucalyptus niphophila D Eucalyptus niphophila D Eucalyptus niphophila D Eucalyptus niphophila D Eucalyptus niphophila D Eucalyptus niphophila D Eucalyptus niphophila D Eucalyptus stellulata G Eucalyptus stellulata G Eucalyptus stellulata G Eucalyptus stellulata G Eucalyptus stellulata G Eucalyptus stellulata G Eucalyptus stellulata G Eucalyptus stellulata S Eucalyptus stellulata S Eucalyptus stellulata S Eucalyptus stellulata S Eucalyptus stellulata S Eucalyptus stellulata S Eucalyptus stellulata S Eucalyptus stellulata D Eucalyptus stellulata D Eucalyptus stellulata D Eucalyptus stellulata D Eucalyptus stellulata D Eucalyptus stellulata D Eucalyptus stellulata D 0.5 0.5 0.4 0.4 0.6 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.4 0.3 0.4 0.3 0.3 0.4 0.4 0.4 0.5 0.5 0.4 0.4 0.4 0.4 20 23 14 11 21 17 17 32 24 20 16 25 23.5 20 30 14 28 15 24 22 28 19 19 16 13 13 16 14 54 80 65 55 67 53 54 63 62 59 45 55 65 55 59 40 88 37 74 60 50 75 74 65 65 55 50 52 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 8.5 8.5 8.5 8.5 8.5 8.5 8.5 126.8 126.8 126.8 126.8 126.8 126.8 126.8 29.5 29.5 29.5 29.5 29.5 29.5 29.5 1.6 1.6 1.6 1.6 1.6 1.6 1.6 150 180 150 160 210 150 150 140 100 100 130 130 70 70 180 80 130 40 130 160 160 170 180 150 160 170 150 170 144.5 173.4 144.5 154.2 202.3 144.5 144.5 134.9 96.4 96.4 125.3 125.3 67.5 67.5 173.4 77.1 125.3 38.5 125.3 154.2 154.2 163.8 173.4 144.5 154.2 163.8 144.5 163.8 14 13 12 12 16 11 13 24 29 15 16 21 19 15 8 6 19 5 7 5 4 6 8 10 7.5 4.5 6 7 415 Appendix V: Moisture content (%ODW) and corresponding soil/weather conditions 1. Eucalyptus pauciflora / E. niphophila leaves Year/month/day/time 604041100 604041100 604041100 604041100 604111130 604111130 604111130 604111130 605181600 605181600 605190900 605190930 605191000 605191100 605191100 605191100 605301120 605301150 605301220 606061000 606061125 606061150 609191130 609191220 609191250 609191320 610231400 610231415 610231420 610231425 610251345 610251400 610251405 610251415 611011100 611011135 611011155 611011215 612181500 612181525 612181545 612190950 612191010 612191030 612191050 Plant Moisture 138.8 149.2 147.4 126.7 158.1 151.7 125.4 116.5 135.6 135.7 152.6 145.1 143.3 119.0 114.6 120.2 143.8 148.1 135.9 136.7 140.7 139.2 129.8 136.8 141.7 144.0 108.1 104.2 120.3 113.7 108.7 106.2 96.9 102.3 116.8 120.3 118.8 120.0 117.9 118.4 115.5 104.1 99.6 98.3 105.6 Soil 28.3 28.3 28.3 28.3 29.1 29.1 29.1 29.1 30.5 30.5 30.3 30.3 30.3 20.0 20.0 20.0 36.8 36.8 36.8 24.9 24.9 24.9 35.8 35.8 35.8 35.8 24.6 24.6 24.6 24.6 6.0 6.0 6.0 6.0 10.8 10.8 10.8 10.8 25.1 25.1 25.1 7.3 7.3 7.3 7.3 Day 94 94 94 94 101 101 101 101 139 139 139 139 139 140 140 140 151 151 151 157 157 157 262 263 263 263 297 297 297 297 299 299 299 299 305 305 305 306 353 353 353 353 353 353 353 Temp deg C 10.4 10.4 10.4 10.4 13.4 13.4 13.4 13.4 5.7 5.7 6.7 6.7 9.3 10.9 10.9 10.9 3.1 3.1 3.6 1.6 1.8 1.8 17.1 16.2 16.2 18.3 18.2 18.2 18.9 18.9 20.6 20.6 20.6 20.6 15.9 16.1 17.1 18.1 26.9 24.5 24.8 18.9 18.9 18.4 17.5 RH 66.8 66.8 66.8 66.8 45.2 45.2 45.2 45.2 94.3 94.3 91.5 91.5 83.1 72.7 72.7 72.7 100.0 100.0 100.0 100.0 100.0 100.0 41.6 42.5 42.5 40.2 24.1 24.1 23.7 23.7 31.4 31.4 31.4 31.4 30.4 31.7 30.2 28.2 21.5 22.5 22.1 47.7 47.7 48.9 50.5 Dew Point deg C 4.5 4.5 4.5 4.5 1.9 1.9 1.9 1.9 4.9 4.9 5.4 5.4 6.6 6.2 6.2 6.2 3.1 3.1 3.6 1.6 1.8 1.8 4.1 3.6 3.6 4.7 -2.5 -2.5 -2.1 -2.1 3.3 3.3 3.3 3.3 -1.4 -0.6 -0.4 -0.5 3.3 2.0 2.0 7.7 7.7 7.6 7.2 Dew Point Depression deg C 5.9 5.9 5.9 5.9 11.5 11.5 11.5 11.5 0.8 0.8 1.3 1.3 2.7 4.7 4.7 4.7 0.0 0.0 0.0 0.0 0.0 0.0 13.0 12.6 12.6 13.6 20.7 20.7 21.1 21.1 17.3 17.3 17.3 17.3 17.3 16.7 17.5 18.5 23.6 22.5 22.8 11.2 11.2 10.8 10.3 416 701041110 701041145 701041205 701041225 701041238 701041130 701041155 701041214 701041231 701041244 701041258 701041250 701041525 701041525 701041535 701041535 701101415 701101435 701101446 701101455 701101505 701101516 701101521 701101524 701111524 701111525 701111530 701111530 701111537 701111540 701111555 701111555 702281130 702281137 703010950 703011010 703011020 703011031 703011041 703011055 703011125 703011130 703011130 703051455 703051520 703051534 703051545 703051558 703051628 703051644 703051656 703061128 703061130 703061132 105.1 106.3 107.4 103.0 107.8 113.7 107.6 115.6 110.6 106.9 107.0 107.7 94.1 93.9 92.9 90.4 110.7 113.6 107.7 105.9 120.4 112.0 110.4 108.9 94.6 103.6 102.5 100.0 94.0 98.9 94.6 90.7 153.3 144.6 114.1 146.2 194.9 152.5 165.9 139.7 93.8 95.4 95.7 142.7 134.2 143.6 144.0 162.5 140.0 128.7 141.2 111.0 115.7 119.8 6.7 6.7 6.7 6.7 6.7 10.6 10.6 10.6 10.6 10.6 10.6 6.7 5.8 5.8 5.8 5.8 19.9 19.9 19.9 19.9 19.9 19.9 19.9 19.9 8.2 8.2 8.2 8.2 8.2 8.2 8.2 8.2 23 23 25.9 25.9 25.9 25.9 25.9 25.9 17.5 17.5 17.5 25.5 25.5 25.5 25.5 25.5 25.5 25.5 25.5 12.3 12.3 12.3 4 4 5 5 5 4 5 5 5 5 5 5 5 5 5 5 11 11 11 11 11 11 11 11 12 12 12 12 12 12 12 12 59 59 60 60 60 60 60 60 60 60 60 64 64 64 64 64 64 64 64 65 65 65 20.4 20.9 20.7 20.9 20.9 20.9 20.7 20.7 20.9 20.9 22.5 22.5 30.0 30.0 30.0 30.0 24.1 24.4 24.8 24.8 24.8 25.6 25.6 25.6 35.9 35.9 37.4 37.4 37.4 37.4 37.4 37.4 13.7 13.7 15.8 15.8 14.1 14.1 14.1 13.3 17.9 17.9 17.9 17.4 17.4 17.4 17.4 17.4 16.2 16.2 16.2 10.8 10.8 10.8 41.6 40.7 41.6 40.5 40.5 40.7 41.6 41.6 40.5 40.5 37.3 37.3 23.2 23.2 23.2 23.2 18.1 17.3 19.0 19.0 19.0 18.5 18.5 18.5 14.4 14.4 14.0 14.0 14.0 14.0 14.0 14.0 92.0 92.0 71.0 71.0 83.7 83.7 83.7 93.1 80.9 80.9 80.9 66.8 66.8 66.8 66.8 66.8 67.3 67.3 67.3 63.9 63.9 63.9 7.1 7.2 7.3 7.1 7.1 7.2 7.3 7.3 7.1 7.1 7.4 7.4 7.1 7.1 7.1 7.1 -1.4 -1.8 -0.1 -0.1 -0.1 0.2 0.2 0.2 5.2 5.2 6.0 6.0 6.0 6.0 6.0 6.0 12.4 12.4 10.7 10.7 11.4 11.4 11.4 12.2 14.7 14.7 14.7 11.3 11.3 11.3 11.3 11.3 10.3 10.3 10.3 4.3 4.3 4.3 13.3 13.7 13.3 13.8 13.8 13.7 13.3 13.3 13.8 13.8 15.1 15.1 22.9 22.9 22.9 22.9 25.5 26.2 24.9 24.9 24.9 25.4 25.4 25.4 30.7 30.7 31.4 31.4 31.4 31.4 31.4 31.4 1.3 1.3 5.2 5.2 2.7 2.7 2.7 1.1 3.3 3.3 3.3 6.2 6.2 6.2 6.2 6.2 6.0 6.0 6.0 6.5 6.5 6.5 417 703061134 114.1 12.3 65 10.8 Day 157 157 158 158 158 158 234 234 263 263 263 263 263 297 297 297 297 313 313 313 313 313 313 313 313 353 353 353 11 11 11 11 11 11 11 11 64 64 64 64 64 64 64 64 148 148 Temp deg C 1.6 1.8 2.3 2.3 2.3 2.8 9.0 7.7 16.2 16.2 18.3 18.3 18.3 17.1 17.1 17.6 17.6 12.8 12.9 12.9 12.9 12.9 13.9 13.9 13.9 22.8 24.2 24.2 24.4 24.4 24.8 24.8 24.8 25.6 25.6 25.6 16.5 17.4 17.4 17.4 16.2 16.2 16.2 16.2 8.4 8.7 63.9 4.3 RH 100.0 100.0 100.0 100.0 100.0 100.0 56.0 62.3 42.5 42.5 40.2 40.2 40.2 28.3 28.3 27.6 27.6 46.4 48.7 48.7 48.7 48.7 48.5 48.5 48.5 27.2 22.2 22.2 17.3 17.3 19.0 19.0 19.0 18.5 18.5 18.5 72.5 66.8 66.8 66.8 67.3 67.3 67.3 67.3 72.3 74.8 Dew Point deg C 1.6 1.8 2.3 2.3 2.3 2.8 0.7 1.0 3.6 3.6 4.7 4.7 4.7 -1.3 -1.3 -1.1 -1.1 1.6 2.4 2.4 2.4 2.4 3.3 3.3 3.3 3.1 1.5 1.5 -1.8 -1.8 -0.1 -0.1 -0.1 0.2 0.2 0.2 11.6 11.3 11.3 11.3 10.3 10.3 10.3 10.3 3.8 4.5 6.5 2. Olearia aglossa leaves Year/month/day/time 606061030 606061140 606061300 606061320 606061345 606061400 608211430 608211510 609191205 609191230 609191305 609191340 609191340 610231350 610231405 610231415 610231420 611081426 611081445 611081455 611081505 611081515 611081525 611081530 611081535 612181450 612181520 612181540 701101425 701101440 701101447 701101458 701101507 701101515 701101531 701101535 703051455 703051525 703051535 703051546 703051601 703051629 703051646 703051659 703281035 703281100 Moisture 113.5 116.9 112.6 127.3 118.9 121.8 100.9 89.9 88.6 93.5 89.7 100.8 86.7 78.9 88.2 88.2 86.2 97.2 89.2 89.9 102.2 100.0 102.5 95.8 98.9 82.2 79.8 72.4 108.1 116.7 112.1 117.8 110.5 111.8 117.7 115.1 107.7 104.6 101.3 117.3 102.1 102.4 111.0 117.3 97.2 100.9 Soil 30.1 30.1 30.1 28.1 28.1 28.1 31.7 31.7 33.5 33.5 33.5 33.5 33.5 24.6 24.6 24.6 24.6 25.6 25.6 25.6 25.6 23.1 23.1 23.1 23.1 19.5 19.5 19.5 23.5 23.5 23.5 23.5 23.5 23.5 23.5 23.5 25.5 25.5 25.5 25.5 25.5 25.5 25.5 25.5 31.2 31.2 Dew Point Depression deg C 0.0 0.0 0.0 0.0 0.0 0.0 8.3 6.7 12.6 12.6 13.6 13.6 13.6 18.4 18.4 18.8 18.8 11.1 10.5 10.5 10.5 10.5 10.6 10.6 10.6 19.7 22.7 22.7 26.2 26.2 24.9 24.9 24.9 25.4 25.4 25.4 4.9 6.2 6.2 6.2 6.0 6.0 6.0 6.0 4.7 4.2 418 703281125 703281145 703281205 97.7 97.3 103.0 31.2 31.2 31.2 148 148 148 8.6 8.6 8.6 Day 95 95 95 95 102 102 102 102 139 139 139 139 139 139 151 151 151 157 157 158 234 234 262 263 263 263 313 313 313 313 353 353 353 60 60 60 60 60 148 148 148 148 148 Temp deg C 9.0 9.0 9.0 9.0 13.4 13.4 13.4 13.4 5.7 5.7 5.7 6.7 6.7 9.3 3.1 3.6 3.6 1.6 1.8 2.3 9.0 7.7 17.1 16.2 18.3 18.3 13.9 15.9 15.9 15.9 26.2 26.5 26.5 15.8 15.8 14.1 14.1 14.1 8.4 8.7 8.7 8.6 8.6 76.7 76.7 77.5 4.7 4.7 4.9 3.8 3.8 3.7 RH 69.9 69.9 69.9 69.9 38.9 38.9 38.9 38.9 94.3 94.3 94.3 91.5 91.5 83.1 100.0 100.0 100.0 100.0 100.0 100.0 56.0 62.3 41.6 42.5 40.2 40.2 48.5 38.7 38.7 38.7 20.2 19.6 19.6 71.0 71.0 83.7 83.7 83.7 72.3 74.8 74.8 76.7 77.5 Dew Point deg C 3.8 3.8 3.8 3.8 -0.2 -0.2 -0.2 -0.2 4.9 4.9 4.9 5.4 5.4 6.6 3.1 3.6 3.6 1.6 1.8 2.3 0.7 1.0 4.1 3.6 4.7 4.7 3.3 2.0 2.0 2.0 2.0 1.8 1.8 10.7 10.7 11.4 11.4 11.4 3.8 4.5 4.5 4.7 4.9 Dew Point Depression deg C 5.2 5.2 5.2 5.2 13.6 13.6 13.6 13.6 0.8 0.8 0.8 1.3 1.3 2.7 0.0 0.0 0.0 0.0 0.0 0.0 8.3 6.7 13.0 12.6 13.6 13.6 10.6 13.9 13.9 13.9 24.3 24.8 24.8 5.2 5.2 2.7 2.7 2.7 4.7 4.2 4.2 3.8 3.7 3. Tasmannia xerophila leaves Year/month/day/time 604041100 604041100 604041100 604041100 604111130 604111130 604111130 604111130 605181600 605181600 605181600 605190900 605190915 605191000 605301130 605301200 605301230 606061015 606061135 606061250 608211440 608211520 609191140 609191225 609191305 609191330 611081440 611081450 611081500 611081510 612181450 612181520 612181540 703011000 703011015 703011025 703011035 703011045 705281030 705281055 705281115 705281135 705281200 Moisture 203.8 204.8 220.4 191.9 202.2 192.1 190.7 182.2 210.1 194.8 203.0 212.9 219.7 202.8 195.8 196.3 190.9 193.4 195.7 193.2 192.2 186.4 188.5 192.7 194.3 194.2 158.2 163.0 162.0 165.7 147.6 149.5 157.8 182.4 180.6 178.9 180.9 189.6 154.1 148.4 157.2 155.6 154.1 Soil 39.3 39.3 39.3 39.3 40.2 40.2 40.2 40.2 41.6 41.6 41.6 41.4 41.4 41.4 36.8 36.8 36.8 27.0 27.0 27.0 36.6 36.6 42.0 42.0 42.0 42.0 23.1 23.1 23.1 23.1 22.4 22.4 22.4 24.4 24.4 24.4 24.4 24.4 51.8 51.8 51.8 51.8 51.8 419 4. E. stellulata leaves Year/month/day/time 606061330 606061350 606061410 608211500 608211500 609191200 609191220 609191300 609191325 609191345 609201530 609201530 609201530 609201530 610251345 610251400 610251405 610251415 610251430 610251435 610251445 610251450 610251450 610251455 701041520 701041530 701041530 701041540 701041546 701111527 701111535 701111543 701111556 703061217 703061220 703061224 703061226 703061230 703061231 703061234 703281155 703281158 703281202 703281202 Moisture 112.4 109.9 111.8 104.1 102.5 104.8 107.1 101.8 104.7 101.7 111.9 112.9 106.0 116.9 122.4 121.1 117.7 122.9 114.1 113.9 114.0 104.2 114.1 110.5 100.0 104.9 102.2 100.0 101.9 106.0 106.0 101.2 103.5 123.5 126.1 122.2 117.6 123.9 126.9 123.9 108.9 110.1 113.0 116.0 Soil 35.9 35.9 35.9 38.3 38.3 31.5 31.5 31.5 31.5 31.5 28.3 28.3 28.3 28.3 6.0 6.0 6.0 6.0 6.8 6.8 6.8 6.8 6.8 6.8 9.4 9.4 9.4 9.4 9.4 6.0 6 6 6 15.9 15.9 15.9 15.9 15.9 15.9 15.9 14.4 14.4 14.4 14.4 Day 158 158 158 234 234 262 262 263 263 263 264 264 264 264 299 299 299 299 299 299 299 299 299 299 5 5 5 5 5 12 12 12 12 65 65 65 65 65 65 65 88 88 88 88 Temp deg C 1.8 1.8 2.3 9.0 9.0 17.1 17.1 16.2 16.2 16.2 14.1 14.1 14.1 14.1 20.6 20.6 20.6 20.6 22.4 22.4 22.4 22.4 22.4 22.4 30.0 30.0 30.0 30.0 30.0 35.9 37.4 37.4 37.4 17.9 17.9 17.9 17.9 17.9 17.9 17.9 23.9 23.9 23.9 23.9 RH 100.0 100.0 100.0 56.0 56.0 41.6 41.6 42.5 42.5 42.5 40.3 40.3 40.3 40.3 31.4 31.4 31.4 31.4 27.7 27.7 27.7 27.7 27.7 27.7 23.2 23.2 23.2 23.2 23.2 14.4 14.0 14.0 14.0 47.5 47.5 47.5 47.5 47.5 47.5 47.5 38.7 38.7 38.7 38.7 Dew Point deg C 1.8 1.8 2.3 0.7 0.7 4.1 4.1 3.6 3.6 3.6 0.9 0.9 0.9 0.9 3.3 3.3 3.3 3.3 3.1 3.1 3.1 3.1 3.1 3.1 7.1 7.1 7.1 7.1 7.1 5.2 6.0 6.0 6.0 6.7 6.7 6.7 6.7 6.7 6.7 6.7 9.2 9.2 9.2 9.2 Dew Point Depression deg C 0.0 0.0 0.0 8.3 8.3 13.0 13.0 12.6 12.6 12.6 13.2 13.2 13.2 13.2 17.3 17.3 17.3 17.3 19.3 19.3 19.3 19.3 19.3 19.3 22.9 22.9 22.9 22.9 22.9 30.7 31.4 31.4 31.4 11.2 11.2 11.2 11.2 11.2 11.2 11.2 14.7 14.7 14.7 14.7 420 5. Bossiaea foliosa leaves Year/month/day/time 605181600 605181600 605190850 605190855 605190955 605301200 605301210 605301230 606060945 606061120 606061145 606061240 606061310 606061340 606061355 608211450 608211520 609191145 609191215 609191240 609191245 609191315 609191315 611011050 611011130 611011145 611011210 612181430 612181515 612181535 612190930 612191005 612191025 612191040 701041120 701041150 701041210 701041230 701041240 701041255 701041135 701041200 701041216 701041235 701041246 701041300 701101430 701101443 701101450 Moisture 78.8 85.4 84.2 85.5 74.6 79.4 84.7 85.7 87.4 81.4 80.1 80.5 79.9 82.0 83.7 83.0 79.7 79.8 78.0 77.4 83.3 80.6 84.5 94.7 94.6 92.6 97.2 93.9 100.0 103.7 86.4 81.1 86.3 84.9 85.7 76.7 91.1 79.5 83.3 82.1 88.1 82.9 82.9 72.0 88.3 82.4 79.6 81.1 88.1 Soil 41.6 41.6 41.4 41.4 41.4 36.8 36.8 36.8 19.3 19.3 19.3 19.3 29.9 29.9 29.9 18.0 18.0 16.3 36.4 16.3 36.4 16.3 36.4 13.2 13.2 13.2 13.2 21.4 21.4 21.4 9.2 9.2 9.2 9.2 7.1 7.1 7.1 7.1 7.1 7.1 8.7 8.7 8.7 8.7 8.7 8.7 17.7 17.7 17.7 Day 139 139 139 139 139 151 151 151 157 157 157 157 158 158 158 234 234 262 263 263 263 263 263 305 305 305 305 353 353 353 353 353 353 353 4 4 5 5 5 5 4 5 5 5 5 5 11 11 11 Temp deg C 5.7 5.7 5.7 6.7 9.3 3.6 3.6 3.6 1.1 1.8 1.8 1.8 2.3 2.3 2.8 9.0 7.7 17.1 16.2 16.2 16.2 18.3 18.3 15.9 16.1 17.1 17.1 22.3 22.8 24.2 15.9 18.9 18.4 18.4 20.9 20.9 20.7 20.9 20.9 22.5 20.9 20.7 20.9 20.9 22.5 22.5 24.4 24.4 24.8 RH 94.3 94.3 95.5 91.5 83.1 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 56.0 62.3 41.6 42.5 42.5 42.5 40.2 40.2 30.4 31.7 30.2 30.2 23.9 27.2 22.2 56.1 47.7 48.9 48.9 40.7 40.7 41.6 40.5 40.5 37.3 40.7 41.6 40.5 40.5 37.3 37.3 17.3 17.3 19.0 Dew Point deg C 4.9 4.9 5.0 5.4 6.6 3.6 3.6 3.6 1.1 1.8 1.8 1.8 2.3 2.3 2.8 0.7 1.0 4.1 3.6 3.6 3.6 4.7 4.7 -1.4 -0.6 -0.4 -0.4 0.9 3.1 1.5 7.2 7.7 7.6 7.6 7.2 7.2 7.3 7.1 7.1 7.4 7.2 7.3 7.1 7.1 7.4 7.4 -1.8 -1.8 -0.1 Dew Point Depression deg C 0.8 0.8 0.7 1.3 2.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 8.3 6.7 13.0 12.6 12.6 12.6 13.6 13.6 17.3 16.7 17.5 17.5 21.4 19.7 22.7 8.6 11.2 10.8 10.8 13.7 13.7 13.3 13.8 13.8 15.1 13.7 13.3 13.8 13.8 15.1 15.1 26.2 26.2 24.9 421 701101502 701101510 701101520 701101525 701101528 702281130 702281137 703010945 703011005 703011020 703011028 703011038 703011050 703051505 703051525 703051535 703051550 703051612 703051634 703051650 703051702 703071535 703071546 703071600 703280940 703281015 703281036 703281102 703281005 703281030 703281053 703281124 705281045 705281110 705281130 705281150 82.1 81.4 82.6 76.9 84.3 81.1 81.9 84.5 87.5 83.6 88.2 89.0 87.2 93.5 83.3 91.5 93.5 79.2 86.1 81.0 86.5 85.0 89.1 84.1 81.3 83.5 82.8 84.1 88.0 87.0 102.5 89.1 84.7 87.9 78.0 79.7 17.7 17.7 17.7 17.7 17.7 12.2 12.2 33.9 33.9 33.9 33.9 33.9 33.9 24.3 24.3 24.3 24.3 24.3 24.3 24.3 24.3 16.6 16.6 16.6 11.4 11.4 11.4 11.4 19 19 19 19 29 29 29 29 11 11 11 11 11 59 59 60 60 60 60 60 60 64 64 64 64 64 64 64 64 68 68 68 87 87 87 87 87 87 87 87 148 148 148 148 24.8 24.8 25.6 25.6 25.6 13.7 13.7 15.8 14.1 14.1 14.1 14.1 13.3 17.4 17.4 17.4 17.4 16.2 16.2 16.2 16.2 20.1 19.0 19.0 14.3 15.7 16.9 18.2 15.7 16.9 18.2 18.9 8.4 8.7 8.6 8.6 Day 66 66 66 74 74 74 74 102 102 146 146 146 Temp deg C 18.0 19.7 19.5 26.5 27.7 28.5 28.3 20.6 22.7 11.8 11.8 14.0 19.0 19.0 18.5 18.5 18.5 92.0 92.0 71.0 83.7 83.7 83.7 83.7 93.1 66.8 66.8 66.8 66.8 67.3 67.3 67.3 67.3 53.1 56.8 56.8 39.1 32.6 25.3 20.8 32.6 25.3 20.8 29.8 72.3 74.8 76.7 77.5 -0.1 -0.1 0.2 0.2 0.2 12.4 12.4 10.7 11.4 11.4 11.4 11.4 12.2 11.3 11.3 11.3 11.3 10.3 10.3 10.3 10.3 10.3 10.3 10.3 0.7 -0.6 -2.9 -4.5 -0.6 -2.9 -4.5 1.0 3.8 4.5 4.7 4.9 24.9 24.9 25.4 25.4 25.4 1.3 1.3 5.2 2.7 2.7 2.7 2.7 1.1 6.2 6.2 6.2 6.2 6.0 6.0 6.0 6.0 9.7 8.6 8.6 13.6 16.3 19.9 22.7 16.3 19.9 22.7 17.9 4.7 4.2 3.8 3.7 RH 32.6 29.7 32.9 34.0 31.8 30.5 28.5 51.2 45.2 47.5 50.8 38.0 Dew Point deg C 1.5 1.7 2.9 9.6 9.7 9.8 8.6 10.3 10.4 1.1 2.0 0.0 Dew Point Depression deg C 16.5 18.1 16.6 16.9 18.0 18.7 19.7 10.3 12.3 10.8 9.8 14.0 6. Daviesia mimosoides leaves Year/month/day/time 503071300 503071330 503071500 503151330 503151430 503151430 503151500 504121500 504121530 505261430 505261445 505261500 Moisture 118.6 116.7 111.0 124.2 119.0 106.6 101.6 94.0 93.6 97.3 87.1 79.7 Soil 34.3 28.2 5.6 19.8 22.1 5.5 8.2 5.3 7.7 4.9 4.9 7.0 422 505261500 505261620 505261630 507141600 507141600 508021130 508021200 508161430 508161500 508161500 509221540 509221540 510061300 510061300 604041200 604041200 604041300 604041300 604041240 604041240 604041230 604041230 604111300 604111300 604111300 604111300 604111330 604111330 604111330 604111330 604111330 604111330 604111330 604111330 604121200 604121200 604121200 605191100 605191100 611011115 611011140 611011200 611011220 611141420 611141430 611141440 611141455 611141505 611141515 611141520 611141530 612190915 612190955 612191015 612191030 703071525 703071540 703071553 88.7 98.8 100.0 113.9 114.4 92.6 96.8 105.0 95.2 111.6 111.4 111.7 102.9 109.9 95.0 92.9 101.3 98.9 94.6 86.4 83.1 83.1 113.6 102.5 110.9 128.9 85.6 84.1 87.0 90.0 88.6 91.5 82.7 95.8 91.4 92.2 91.7 95.3 109.6 97.3 101.6 103.4 103.8 90.2 97.3 109.4 94.7 93.6 98.6 92.2 95.7 122.4 134.5 125.8 127.6 99.2 98.8 98.2 7.1 9.8 9.6 43.8 43.8 35.6 36.2 42.2 41.5 41.5 13.2 13.2 38.6 38.6 23.4 23.4 11.2 11.2 5.0 5.0 6.3 6.3 11.7 11.7 11.7 11.7 6.2 6.2 6.2 6.2 5.2 5.2 5.2 5.2 15.9 15.9 15.9 20.0 20.0 7.4 7.4 7.4 7.4 13.1 13.1 13.1 13.1 13.1 13.1 13.1 13.1 6.4 6.4 6.4 6.4 10.2 10.2 10.2 146 146 146 165 165 214 214 228 228 228 265 265 279 279 94 94 94 94 94 94 94 94 101 101 101 101 101 101 101 101 101 101 101 101 102 102 102 139 139 305 305 305 306 319 319 319 319 318 318 318 318 353 353 353 353 68 68 68 14.0 9.6 9.6 3.6 3.6 8.5 9.3 10.1 10.1 10.1 22.3 22.3 12.3 12.3 14.0 14.0 14.6 14.6 14.0 14.0 13.9 13.9 19.8 19.8 19.8 19.8 18.9 18.9 18.9 18.9 18.9 18.9 18.9 18.9 17.7 17.7 17.7 10.9 10.9 16.1 16.1 17.1 18.1 15.3 15.3 15.3 15.1 15.1 13.2 13.2 13.2 15.9 18.9 18.4 18.4 20.1 20.1 19.0 38.0 48.3 48.3 90.5 90.5 49.8 43.1 45.0 46.4 46.4 40.0 40.0 31.7 31.7 50.7 50.7 63.3 63.3 53.7 53.7 53.5 53.5 29.9 29.9 29.9 29.9 30.6 30.6 30.6 30.6 30.8 30.8 30.8 30.8 39.4 39.4 39.4 72.7 72.7 31.7 31.7 30.2 28.2 28.9 28.9 28.9 26.0 26.0 33.5 33.5 33.5 56.1 47.7 48.9 48.9 53.1 53.1 56.8 0.0 -0.7 -0.7 2.2 2.2 -1.4 -2.6 -1.3 -0.8 -0.8 8.2 8.2 -4.0 -4.0 4.0 4.0 7.8 7.8 4.8 4.8 4.7 4.7 1.9 1.9 1.9 1.9 1.4 1.4 1.4 1.4 1.5 1.5 1.5 1.5 3.9 3.9 3.9 6.2 6.2 -0.6 -0.6 -0.4 -0.5 -2.6 -2.6 -2.6 -4.2 -4.2 -2.4 -2.4 -2.4 7.2 7.7 7.6 7.6 10.3 10.3 10.3 14.0 10.3 10.3 1.4 1.4 9.9 11.9 11.4 10.9 10.9 14.1 14.1 16.3 16.3 10.0 10.0 6.8 6.8 9.2 9.2 9.2 9.2 17.9 17.9 17.9 17.9 17.5 17.5 17.5 17.5 17.4 17.4 17.4 17.4 13.8 13.8 13.8 4.7 4.7 16.7 16.7 17.5 18.5 17.9 17.9 17.9 19.3 19.3 15.6 15.6 15.6 8.6 11.2 10.8 10.8 9.7 9.7 8.6 423 703280949 703281022 703281046 703281112 705281230 705281240 705281250 705281300 95.1 95.7 93.9 94.8 95.5 95.9 95.5 101.9 8.9 8.9 8.9 8.9 7.6 7.6 7.6 7.6 87 87 87 87 149 149 149 149 15.7 16.9 18.2 18.2 10.0 10.0 10.3 10.3 32.6 25.3 20.8 20.8 63.7 63.7 64.2 64.2 -0.6 -2.9 -4.5 -4.5 3.5 3.5 3.8 3.8 16.3 19.9 22.7 22.7 6.5 6.5 6.4 6.4 424 Appendix VI: KBDI and measured/modelled soil moisture values KBDI Measured soil moisture Date EU1 EU2 EU3 EU4 EU5 EU1 16-Feb 18.0 18.0 18.0 18.0 18.0 17-Feb 19.5 19.4 19.5 19.6 18-Feb 21.3 21.0 21.1 21.3 19-Feb 24.0 23.0 23.0 20-Feb 26.8 25.1 21-Feb 3.0 22-Feb EU2 Modelled soil moisture EU3 EU4 EU5 EU1 EU2 EU3 EU4 EU5 21.1 15.3 5.5 9.4 34.6 23.7 17.3 5.5 12.8 20.0 19.8 15.2 5.5 9.3 34.3 23.5 17.2 5.4 12.6 21.9 19.5 15.2 5.4 9.1 34.0 23.3 17.0 5.4 12.5 23.2 24.1 18.3 14.9 5.5 8.9 33.5 23.1 16.8 5.3 12.3 25.0 25.4 26.8 18.3 14.7 5.5 8.7 33.0 22.8 16.7 5.3 12.2 8.1 9.5 9.6 10.3 47.9 16.8 5.5 9.1 37.4 25.0 18.1 13.3 0.0 5.6 7.3 7.4 8.3 43.8 16.3 5.5 11.0 25.3 18.3 13.4 23-Feb 1.5 7.0 8.7 8.9 10.1 5.6 11.4 25.1 18.2 13.3 4.2 9.6 11.3 11.8 13.6 41.5 38.6 15.6 24-Feb 15.3 5.6 11.4 37.2 24.8 17.9 13.1 25-Feb 7.9 12.5 14.0 15.1 17.7 36.2 15.0 5.6 10.8 36.5 24.4 17.7 12.8 26-Feb 12.3 16.4 17.8 18.9 22.0 33.4 14.9 5.5 10.8 35.7 23.9 17.3 5.4 12.5 27-Feb 14.5 18.5 19.9 21.1 24.6 31.1 14.9 5.6 10.4 35.3 23.6 17.1 5.4 12.3 28-Feb 16.5 20.4 21.8 23.3 27.4 29.2 14.6 5.5 10.2 34.9 23.4 17.0 5.3 12.1 1-Mar 19.4 23.0 24.3 26.2 30.8 28.7 14.7 5.5 9.8 34.4 23.1 16.7 5.2 11.9 2-Mar 22.3 26.0 27.1 29.1 34.1 30.3 14.9 5.6 9.6 33.8 22.7 16.5 5.1 11.6 3-Mar 23.8 27.6 28.7 30.9 36.4 29.7 14.7 5.6 9.4 33.5 22.5 16.3 5.1 11.5 4-Mar 26.3 29.6 30.7 32.8 38.7 27.4 14.6 5.4 9.3 33.1 22.2 16.1 5.0 11.3 5-Mar 24.6 29.5 31.0 33.1 38.8 34.9 14.7 9.1 9.1 33.4 22.2 16.1 5.0 11.3 6-Mar 21.2 27.0 28.7 30.7 36.4 37.9 15.0 5.7 9.1 34.0 22.5 16.3 5.1 11.5 7-Mar 22.0 27.7 29.4 31.5 37.3 34.9 14.9 5.6 9.1 33.9 22.5 16.3 5.1 11.4 8-Mar 23.0 28.7 30.4 32.5 38.4 32.2 14.7 5.6 9.1 33.7 22.3 16.2 5.0 11.4 9-Mar 24.4 29.9 31.6 33.7 39.7 29.7 14.6 5.5 8.9 33.4 22.2 16.0 5.0 11.3 10-Mar 26.0 31.4 33.1 35.2 41.3 28.2 14.4 5.5 8.9 33.1 22.0 15.9 5.0 11.2 11-Mar 27.5 32.7 34.4 36.6 42.9 26.0 14.4 5.5 8.7 32.9 21.8 15.8 4.9 11.0 12-Mar 29.7 34.6 36.2 38.4 44.9 24.8 14.4 5.5 8.7 32.4 21.6 15.6 4.9 10.9 13-Mar 32.4 37.2 38.8 40.9 47.6 23.5 14.3 5.5 8.7 31.9 21.2 15.4 4.8 10.7 14-Mar 35.4 39.7 41.3 43.3 50.2 21.7 14.2 5.5 8.4 31.4 20.9 15.1 4.7 10.5 15-Mar 38.4 42.2 43.7 45.8 52.9 19.8 13.9 5.4 8.4 30.8 20.6 14.9 4.6 10.4 16-Mar 41.2 44.7 46.1 48.3 55.4 23.5 13.6 5.4 8.4 30.3 20.3 14.7 4.6 10.2 17-Mar 39.8 44.7 46.5 48.6 55.6 22.9 13.9 5.3 8.4 30.6 20.3 14.6 4.6 10.2 18-Mar 39.8 44.9 46.7 48.9 55.9 22.1 13.7 5.4 8.4 30.6 20.3 14.6 4.6 10.2 19-Mar 39.9 45.1 46.9 49.1 56.1 21.1 13.9 5.4 8.2 30.5 20.2 14.6 4.6 10.1 20-Mar 40.5 45.8 47.7 49.8 57.0 19.5 13.7 5.4 8.2 30.4 20.1 14.5 4.5 10.1 21-Mar 41.2 46.7 48.6 50.9 58.1 18.3 13.6 5.3 8.2 30.3 20.0 14.5 4.5 10.0 22-Mar 41.4 46.9 48.9 51.1 58.3 20.4 13.7 5.3 8.2 30.3 20.0 14.4 4.5 10.0 23-Mar 41.5 47.1 49.1 51.3 58.6 24.4 13.9 5.4 8.2 30.2 20.0 14.4 4.5 10.0 24-Mar 41.9 47.5 49.5 51.7 59.1 22.5 13.7 5.4 8.1 30.2 19.9 14.4 4.5 9.9 25-Mar 42.4 48.1 50.1 52.4 59.9 21.7 13.7 5.4 8.2 30.1 19.8 14.3 4.5 9.9 26-Mar 42.6 48.6 50.6 52.9 60.3 20.7 13.6 5.4 8.1 30.0 19.8 14.3 4.4 9.8 27-Mar 43.1 49.1 51.3 53.6 61.0 19.8 13.6 5.4 8.2 29.9 19.7 14.2 4.4 9.8 28-Mar 44.0 50.2 52.4 54.7 62.1 18.6 13.5 5.3 7.9 29.8 19.6 14.1 4.4 9.7 29-Mar 44.5 50.8 53.0 55.4 62.9 18.0 13.5 5.4 8.1 29.7 19.5 14.0 4.4 9.7 30-Mar 45.0 51.3 53.5 56.0 63.5 16.9 13.5 5.3 8.1 29.6 19.4 14.0 4.3 9.6 31-Mar 45.5 51.8 54.0 56.4 64.1 16.9 13.5 5.3 8.1 29.5 19.4 14.0 4.3 9.6 1-Apr 46.4 52.9 55.1 57.6 65.3 16.1 13.5 5.2 8.1 29.3 19.2 13.8 4.3 9.5 2-Apr 47.8 54.3 56.6 59.1 67.0 15.1 13.2 5.3 7.9 29.1 19.1 13.7 4.3 9.4 3-Apr 50.0 56.3 58.6 61.1 69.0 14.3 13.2 5.3 7.9 28.6 18.8 13.5 4.2 9.2 4-Apr 51.5 57.7 59.9 62.4 70.3 14.1 13.1 5.3 7.9 28.4 18.6 13.4 4.2 9.2 5-Apr 51.8 58.0 60.3 62.8 70.8 13.6 13.2 5.3 7.9 28.3 18.6 13.4 4.1 9.1 6-Apr 52.7 58.9 61.1 63.7 71.8 13.3 13.2 5.2 7.9 28.1 18.5 13.3 4.1 9.1 7-Apr 54.0 60.3 62.6 65.1 73.3 12.8 13.1 5.3 7.7 27.9 18.3 13.1 4.1 8.9 425 8-Apr 54.8 60.9 63.2 65.8 74.1 12.5 13.1 5.3 7.7 27.8 18.2 13.1 4.1 8.9 9-Apr 55.8 61.7 64.0 66.7 75.1 12.1 12.9 5.2 7.7 27.6 18.1 13.0 4.0 8.8 10-Apr 57.6 63.3 65.6 68.1 76.5 12.1 12.9 5.3 7.6 27.2 17.9 12.9 4.0 8.7 11-Apr 58.5 64.8 67.1 69.5 77.9 12.3 13.2 5.3 7.7 27.1 17.7 12.7 3.9 8.6 12-Apr 59.3 65.6 67.9 70.4 79.0 13.3 12.9 5.2 7.7 26.9 17.6 12.7 3.9 8.6 13-Apr 60.4 66.8 69.0 71.5 80.0 13.1 12.8 5.3 7.7 26.7 17.5 12.5 3.9 8.5 14-Apr 61.5 68.0 70.2 72.6 81.1 12.8 12.7 5.2 7.7 26.5 17.3 12.4 3.9 8.4 15-Apr 60.1 68.1 70.8 73.0 81.3 14.9 13.7 5.2 7.7 26.8 17.3 12.4 3.8 8.4 16-Apr 60.1 68.2 70.8 73.1 81.3 16.6 13.2 5.2 7.7 26.8 17.3 12.4 3.8 8.4 17-Apr 60.6 68.7 71.3 73.6 81.9 16.1 13.1 5.2 7.6 26.7 17.2 12.3 3.8 8.4 18-Apr 61.4 69.5 72.1 74.3 82.6 15.5 12.8 5.2 7.6 26.5 17.1 12.3 3.8 8.3 19-Apr 62.2 70.2 72.8 75.0 83.3 14.9 12.7 5.2 7.6 26.4 17.0 12.2 3.8 8.3 20-Apr 62.8 70.9 73.5 75.6 83.9 14.1 12.5 5.2 7.6 26.2 16.9 12.1 3.8 8.2 21-Apr 63.3 71.5 74.1 76.2 84.4 13.3 12.5 5.2 7.6 26.2 16.9 12.1 3.8 8.2 22-Apr 64.3 72.2 74.9 76.9 85.2 13.1 12.5 5.2 7.6 26.0 16.8 12.0 3.7 8.1 23-Apr 65.3 73.4 76.2 77.9 86.1 12.5 12.5 5.2 7.6 25.8 16.6 11.9 3.7 8.1 24-Apr 66.0 74.4 77.1 78.8 87.0 12.3 12.5 5.2 7.6 25.7 16.5 11.8 3.7 8.0 25-Apr 67.0 75.3 78.2 79.8 87.9 12.1 12.2 5.2 7.4 25.5 16.4 11.7 3.6 7.9 26-Apr 67.9 76.1 78.9 80.5 88.6 11.9 12.4 5.2 7.4 25.3 16.3 11.6 3.6 7.9 27-Apr 69.2 77.3 80.1 81.6 89.5 11.4 12.2 5.1 7.6 25.1 16.1 11.5 3.6 7.8 28-Apr 70.3 78.4 81.2 82.5 90.3 11.0 12.2 5.0 7.4 24.9 16.0 11.4 3.6 7.8 29-Apr 71.1 79.6 82.4 83.6 91.2 11.0 12.1 5.1 7.4 24.7 15.8 11.3 3.5 7.7 30-Apr 71.6 80.1 83.0 84.1 91.6 10.8 12.1 5.1 7.4 24.6 15.8 11.2 3.5 7.7 1-May 72.0 80.7 83.7 84.7 92.1 10.8 12.1 5.1 7.4 24.5 15.7 11.2 3.5 7.7 2-May 72.2 81.0 83.9 84.9 92.4 10.8 12.1 5.1 7.3 24.5 15.6 11.2 3.5 7.6 3-May 72.7 81.6 84.6 85.5 92.9 10.6 12.1 5.1 7.4 24.4 15.6 11.1 3.5 7.6 4-May 73.6 82.7 85.7 86.5 93.8 10.6 12.1 5.1 7.3 24.2 15.4 11.0 3.4 7.5 5-May 74.4 83.3 86.3 87.1 94.6 10.4 12.0 5.0 7.1 24.1 15.3 10.9 3.4 7.5 6-May 75.4 84.0 87.1 87.9 95.5 10.4 11.8 5.0 7.3 23.9 15.2 10.9 3.4 7.4 7-May 76.1 84.6 87.6 88.5 96.2 10.4 11.7 5.1 7.3 23.8 15.2 10.8 3.4 7.4 8-May 76.3 84.8 87.8 88.8 96.6 10.2 11.8 5.1 7.3 23.7 15.2 10.8 3.4 7.3 9-May 76.7 85.3 88.3 89.2 97.2 10.2 11.7 5.1 7.3 23.6 15.1 10.7 3.4 7.3 10-May 77.1 85.7 88.8 89.7 97.6 10.0 11.7 5.1 7.3 23.6 15.0 10.7 3.4 7.3 11-May 77.6 86.3 89.4 90.2 98.0 10.2 11.7 5.0 7.3 23.5 15.0 10.6 3.3 7.2 12-May 78.2 87.2 90.3 91.0 98.7 10.2 11.7 4.9 7.1 23.4 14.8 10.6 3.3 7.2 13-May 78.6 87.6 90.8 91.5 99.2 10.0 11.6 5.0 7.1 23.3 14.8 10.5 3.3 7.2 14-May 79.1 88.2 91.4 92.0 99.7 10.0 11.5 5.0 7.1 23.2 14.7 10.5 3.3 7.1 15-May 79.3 88.6 91.8 92.5 100.2 10.0 11.6 5.0 7.1 23.2 14.7 10.4 3.3 7.1 16-May 79.4 88.8 92.1 92.7 100.4 10.0 11.6 4.9 7.1 23.1 14.6 10.4 3.3 7.1 17-May 79.4 88.9 92.1 92.8 100.5 9.8 11.6 5.0 7.1 23.1 14.6 10.4 3.3 7.1 18-May 79.4 89.0 92.2 92.9 100.6 10.0 11.5 4.9 7.0 23.1 14.6 10.4 3.3 7.1 19-May 79.4 89.0 92.3 92.9 100.7 10.0 11.6 5.0 7.1 23.2 14.6 10.4 3.3 7.1 20-May 79.4 89.0 92.3 93.0 100.8 10.0 11.6 5.0 7.1 23.2 14.6 10.4 3.3 7.1 21-May 79.6 89.4 92.7 93.3 101.2 10.0 11.6 4.9 7.1 23.1 14.6 10.3 3.2 7.0 22-May 79.9 89.7 93.0 93.6 101.6 10.0 11.6 4.9 7.1 23.1 14.5 10.3 3.2 7.0 23-May 80.2 90.0 93.4 94.0 102.0 9.8 11.5 5.0 7.1 23.0 14.5 10.3 3.2 7.0 24-May 80.6 90.5 93.9 94.4 102.4 9.6 11.6 5.0 7.0 22.9 14.4 10.2 3.2 6.9 25-May 80.8 90.6 94.0 94.6 102.7 9.8 11.6 4.9 7.1 22.9 14.4 10.2 3.2 6.9 26-May 80.9 90.9 94.3 94.9 102.9 9.6 11.5 5.0 7.0 22.9 14.4 10.2 3.2 6.9 27-May 81.0 91.0 94.4 95.0 103.2 9.8 11.6 4.8 7.0 22.8 14.4 10.2 3.2 6.9 28-May 81.0 91.0 94.5 95.1 103.3 9.8 11.5 4.9 7.0 22.9 14.4 10.2 3.2 6.9 29-May 81.0 91.1 94.5 95.2 103.4 9.8 11.5 4.9 7.0 22.9 14.3 10.2 3.2 6.9 30-May 81.0 91.1 94.6 95.2 103.6 9.6 11.6 4.9 7.0 22.9 14.3 10.2 3.2 6.9 31-May 81.2 91.3 94.8 95.5 103.9 9.6 9.3 11.6 4.8 7.0 22.8 14.3 10.1 3.2 6.8 1-Jun 81.4 91.5 94.9 95.7 104.3 9.8 9.1 11.6 4.9 7.0 22.8 14.3 10.1 3.2 6.8 2-Jun 81.5 91.6 95.0 95.9 104.6 9.4 8.9 11.6 4.9 6.9 22.8 14.3 10.1 3.2 6.8 426 3-Jun 81.9 91.7 95.1 96.1 105.0 9.6 9.1 11.5 4.9 7.0 22.7 14.3 10.1 3.2 6.8 4-Jun 82.3 91.9 95.3 96.4 105.5 9.4 9.1 11.6 4.9 6.9 22.6 14.2 10.1 3.2 6.7 5-Jun 82.6 92.1 95.5 96.6 105.8 9.6 9.1 11.6 4.9 7.0 22.6 14.2 10.1 3.1 6.7 6-Jun 82.8 92.3 95.7 96.8 106.1 9.6 9.1 11.6 4.9 6.9 22.5 14.2 10.1 3.1 6.7 7-Jun 83.1 92.5 95.9 97.1 106.4 9.6 9.1 11.6 4.9 7.0 22.5 14.2 10.0 3.1 6.7 8-Jun 83.5 92.9 96.2 97.4 106.8 9.6 9.1 11.6 4.9 7.0 22.4 14.1 10.0 3.1 6.6 9-Jun 83.8 93.1 96.4 97.7 107.1 9.3 9.1 11.6 4.8 6.9 22.3 14.1 10.0 3.1 6.6 10-Jun 84.0 93.3 96.6 97.9 107.3 9.6 9.1 11.6 4.9 7.0 22.3 14.1 10.0 3.1 6.6 11-Jun 84.3 93.5 96.8 98.1 107.6 9.6 9.1 11.6 4.9 6.9 22.2 14.0 9.9 3.1 6.6 12-Jun 78.8 90.4 94.3 95.5 104.8 15.1 18.0 11.8 4.9 6.9 23.3 14.4 10.2 3.2 6.8 13-Jun 78.8 90.5 94.4 95.6 104.9 14.6 13.6 11.8 4.9 7.0 23.3 14.4 10.2 3.2 6.8 14-Jun 76.1 89.5 93.8 95.0 104.1 16.6 21.7 12.0 5.0 7.0 23.8 14.5 10.2 3.2 6.8 15-Jun 76.1 89.6 93.9 95.1 104.2 16.1 16.6 12.0 5.0 7.0 23.8 14.5 10.2 3.2 6.8 16-Jun 76.0 89.6 93.9 95.1 104.3 15.5 23.5 12.1 5.0 6.9 23.8 14.5 10.2 3.2 6.8 17-Jun 73.9 89.1 93.8 95.0 104.0 16.6 24.8 13.3 4.9 7.0 24.2 14.6 10.2 3.2 6.8 18-Jun 69.0 85.4 90.4 91.6 100.4 22.1 22.1 13.1 5.0 7.0 25.1 15.1 10.5 3.3 7.1 19-Jun 68.1 85.4 90.4 91.5 100.4 21.7 19.1 12.9 5.0 7.0 25.3 15.1 10.5 3.3 7.1 20-Jun 56.8 77.9 83.7 84.8 93.3 34.9 26.4 15.2 5.0 7.0 27.4 16.0 11.2 3.5 7.6 21-Jun 46.4 69.9 76.4 77.3 85.5 36.2 28.2 15.3 5.0 7.0 29.3 17.1 11.9 3.7 8.1 22-Jun 45.5 69.2 75.7 76.7 84.9 34.9 28.2 15.0 5.0 7.1 29.5 17.1 11.9 3.7 8.2 23-Jun 45.3 69.1 75.7 76.7 84.9 32.8 43.8 14.6 5.0 7.1 29.5 17.2 11.9 3.7 8.2 24-Jun 31.4 59.7 67.2 68.1 75.7 38.6 32.8 15.3 5.0 7.3 32.1 18.4 12.7 4.0 8.8 25-Jun 31.2 59.6 67.2 68.1 75.8 36.2 30.3 14.9 5.0 7.7 32.2 18.4 12.7 4.0 8.8 26-Jun 31.2 59.6 67.2 68.1 76.0 35.6 30.3 14.6 5.0 8.4 32.2 18.4 12.7 4.0 8.8 27-Jun 31.1 59.6 67.2 68.1 76.0 33.4 28.7 14.6 5.0 8.7 32.2 18.4 12.7 4.0 8.8 28-Jun 31.1 59.7 67.3 68.2 76.1 33.4 26.0 14.4 5.0 9.1 32.2 18.4 12.7 4.0 8.8 29-Jun 31.2 59.7 67.3 68.3 76.3 32.8 23.5 14.4 5.0 8.9 32.2 18.4 12.7 4.0 8.7 30-Jun 23.0 54.6 63.0 63.9 71.5 37.9 34.3 14.9 5.0 9.4 33.7 19.0 13.1 4.1 9.1 1-Jul 0.0 29.1 39.2 39.8 46.5 43.8 35.6 16.8 5.2 11.4 22.3 15.3 4.8 10.8 2-Jul 0.0 27.4 37.6 38.2 44.9 40.7 35.6 16.1 5.1 11.9 22.5 15.5 4.9 10.9 3-Jul 0.0 27.3 37.6 38.2 44.9 39.3 32.8 15.8 5.2 11.9 22.5 15.5 4.9 10.9 4-Jul 0.2 27.4 37.6 38.4 45.2 38.6 29.2 15.5 5.2 12.1 22.5 15.5 4.9 10.9 5-Jul 0.4 27.4 37.7 38.5 45.6 36.9 26.4 15.3 5.3 12.1 22.5 15.5 4.9 10.9 6-Jul 0.6 27.6 37.8 38.7 45.9 36.2 24.4 15.2 5.2 11.7 22.5 15.5 4.9 10.8 7-Jul 0.9 27.8 38.0 39.0 46.2 35.6 21.7 15.0 5.2 11.9 22.4 15.4 4.8 10.8 8-Jul 0.9 27.9 38.1 39.1 46.5 34.9 21.1 15.0 5.1 11.7 22.4 15.4 4.8 10.8 9-Jul 0.7 27.8 38.1 39.0 46.4 36.2 31.7 16.4 5.3 11.9 22.4 15.4 4.8 10.8 10-Jul 0.0 26.2 36.9 37.9 45.2 41.5 34.3 18.3 5.2 12.5 22.6 15.5 4.9 10.9 11-Jul 0.0 13.2 25.2 26.0 32.7 43.0 35.6 17.5 5.3 13.1 24.3 16.6 5.2 11.7 12-Jul 0.0 3.2 15.9 16.7 23.2 42.2 35.6 16.9 5.4 13.1 17.5 12.4 13-Jul 0.0 0.0 8.9 9.5 15.9 43.0 36.2 17.7 5.6 13.6 18.2 12.9 14-Jul 0.0 0.0 2.8 3.4 9.6 43.8 36.9 17.9 5.6 13.6 15-Jul 0.0 0.0 0.0 1.0 7.0 43.0 36.9 16.9 5.6 13.6 16-Jul 0.0 0.0 0.0 0.0 7.1 42.2 36.2 16.4 5.5 13.6 17-Jul 0.0 0.0 0.0 0.0 7.2 42.2 37.9 16.9 5.6 13.3 18-Jul 0.0 0.0 0.0 0.0 7.1 41.5 35.6 16.4 5.6 13.8 19-Jul 0.0 0.0 0.0 0.0 7.3 40.0 33.4 16.1 5.6 13.6 13.5 20-Jul 0.0 0.0 0.0 0.1 7.6 40.0 31.7 15.8 5.6 13.6 13.5 21-Jul 0.1 0.0 0.0 0.2 8.0 39.3 29.7 16.0 5.5 13.3 13.5 22-Jul 0.4 0.1 0.0 0.4 8.5 42.2 28.7 15.8 5.5 13.3 13.4 23-Jul 0.5 0.2 0.2 0.7 9.0 42.2 26.4 15.8 5.6 13.6 13.4 24-Jul 0.7 0.3 0.3 0.9 9.3 42.2 24.8 15.8 5.5 13.3 13.4 25-Jul 0.7 0.4 0.4 1.0 9.5 43.0 23.9 15.8 5.6 13.3 13.3 26-Jul 0.7 0.5 0.5 1.2 9.7 41.5 23.9 15.8 5.5 13.3 13.3 27-Jul 0.7 0.6 0.7 1.3 10.0 40.0 22.9 15.6 5.5 13.3 13.3 28-Jul 0.9 0.9 1.0 1.7 10.3 39.3 22.1 15.6 5.5 13.1 13.3 13.3 427 29-Jul 0.9 1.1 1.2 1.9 10.5 38.6 20.4 15.6 5.4 13.1 13.3 30-Jul 0.9 1.1 1.3 2.0 10.7 37.9 19.5 15.5 5.4 13.1 13.3 31-Jul 1.1 1.3 1.5 2.2 10.9 36.9 18.3 15.6 5.5 13.1 13.2 1-Aug 1.3 1.7 1.9 2.6 11.4 36.2 17.2 15.6 5.4 13.1 13.2 2-Aug 1.6 2.1 2.4 3.0 11.8 36.2 16.6 15.6 5.4 12.5 13.2 3-Aug 2.2 2.7 2.9 3.7 12.6 34.9 15.5 15.5 5.4 12.5 13.1 4-Aug 0.0 0.0 0.0 0.0 0.0 51.5 28.2 151.4 5.5 14.1 5-Aug 0.0 0.0 0.0 0.0 0.0 44.6 30.3 18.1 5.5 14.1 6-Aug 0.0 0.0 0.0 0.0 0.0 43.8 29.2 17.9 5.5 14.3 7-Aug 0.0 0.0 0.0 0.0 0.0 43.0 27.4 17.7 5.6 14.1 8-Aug 0.0 0.0 0.0 0.0 0.0 42.2 32.8 17.3 5.6 13.8 9-Aug 0.1 0.1 0.1 0.1 0.1 40.7 31.1 16.8 5.5 13.8 10-Aug 0.0 0.0 0.0 0.0 0.0 47.0 36.2 63.7 5.6 14.6 11-Aug 0.0 0.0 0.0 0.0 0.0 44.6 35.6 18.3 5.5 14.1 12-Aug 0.0 0.0 0.0 0.0 0.0 43.0 33.4 17.9 5.6 14.3 13-Aug 0.0 0.0 0.0 0.0 0.0 42.2 32.2 17.3 5.6 14.3 14-Aug 0.0 0.0 0.0 0.0 0.0 44.6 32.2 17.3 5.6 14.1 15-Aug 0.3 0.3 0.3 0.3 0.4 43.0 34.9 17.1 5.6 14.1 16-Aug 0.4 0.5 0.5 0.5 0.6 42.2 34.9 16.9 5.5 14.1 17-Aug 0.7 0.7 0.7 0.8 1.1 41.5 32.2 16.8 5.6 13.8 18-Aug 1.3 1.3 1.3 1.4 1.8 40.0 30.3 16.8 5.6 14.1 19-Aug 2.0 2.0 2.0 2.1 2.5 39.3 28.7 16.8 5.6 13.8 20-Aug 2.4 2.3 2.4 2.5 2.9 36.9 26.9 16.8 5.6 13.8 21-Aug 2.4 2.4 2.5 2.6 3.1 36.9 26.0 16.6 5.5 13.8 22-Aug 2.5 2.6 2.6 2.8 3.3 36.9 23.9 16.4 5.6 13.8 23-Aug 2.3 2.6 2.6 2.8 3.4 38.6 26.4 16.9 5.6 13.6 24-Aug 2.3 2.6 2.7 2.9 3.6 37.9 26.0 16.8 5.6 14.1 25-Aug 2.6 3.0 3.1 3.4 4.3 36.9 24.4 16.4 5.5 14.1 26-Aug 3.4 3.7 3.7 4.2 5.3 36.2 23.5 16.4 5.6 13.8 37.4 27-Aug 3.9 4.1 4.2 4.8 6.1 34.9 22.1 16.4 5.5 13.8 37.3 25.5 28-Aug 4.9 4.9 5.0 5.7 7.4 34.9 20.7 16.3 5.6 13.6 37.1 25.4 29-Aug 6.6 6.2 6.2 7.2 9.1 34.3 19.5 16.4 5.6 13.6 36.8 25.2 18.4 13.4 30-Aug 7.4 6.9 6.8 7.8 9.8 32.8 18.0 16.4 5.5 13.6 36.6 25.1 18.4 13.3 31-Aug 1.3 3.4 4.0 4.8 6.5 37.9 19.1 22.1 5.4 13.8 1-Sep 0.0 0.0 0.0 0.0 2.7 41.5 26.9 17.3 5.5 14.6 2-Sep 0.0 0.0 0.0 0.0 0.0 43.8 25.5 16.9 5.5 14.3 3-Sep 0.0 0.0 0.0 0.0 0.0 44.6 29.7 17.3 5.6 14.6 4-Sep 0.1 0.2 0.2 0.3 0.3 41.5 28.2 16.9 5.6 14.6 5-Sep 0.1 0.4 0.4 0.5 0.6 40.7 26.0 16.8 5.6 14.3 6-Sep 0.4 0.8 0.9 1.0 1.2 39.3 24.4 16.8 5.5 14.3 7-Sep 0.9 1.3 1.4 1.5 1.9 38.6 22.9 16.6 5.5 14.3 8-Sep 1.7 2.0 2.0 2.3 2.8 36.9 20.7 16.6 5.5 14.1 9-Sep 3.6 3.6 3.6 3.9 4.7 36.2 18.6 16.4 5.6 14.3 10-Sep 0.0 0.0 0.0 0.0 0.0 47.9 28.7 19.2 5.6 14.9 11-Sep 0.0 0.0 0.0 0.0 0.0 48.8 34.9 151.4 5.7 15.8 12-Sep 0.0 0.0 0.0 0.0 0.0 44.6 33.4 151.4 5.7 15.1 13-Sep 0.0 0.0 0.0 0.0 0.0 43.8 33.4 137.7 5.7 15.1 14-Sep 0.0 0.1 0.1 0.2 0.3 43.0 31.7 26.7 5.7 14.9 15-Sep 0.1 0.2 0.3 0.3 0.4 41.5 20.5 5.7 14.9 16-Sep 0.3 0.4 0.5 0.5 0.6 41.5 19.7 5.7 14.9 17-Sep 0.4 0.6 0.7 0.8 1.0 40.7 19.0 5.7 14.9 18-Sep 0.0 0.0 0.0 0.0 0.0 43.0 19.0 5.9 15.1 19-Sep 0.0 0.0 0.0 0.0 0.1 42.2 18.4 5.7 14.9 20-Sep 0.2 0.2 0.2 0.2 0.6 40.7 18.1 5.9 15.1 21-Sep 0.4 0.5 0.5 0.6 1.1 40.7 17.9 5.7 14.9 22-Sep 1.3 1.6 1.7 1.8 2.4 39.3 17.7 5.9 14.6 13.5 37.3 428 23-Sep 2.9 3.1 3.3 3.2 4.1 38.6 17.3 5.9 14.9 37.5 24-Sep 3.9 4.0 4.2 4.1 5.3 37.9 17.5 5.7 14.9 37.3 25.5 25-Sep 5.2 5.2 5.3 5.3 6.8 36.2 17.3 5.9 14.6 37.0 25.3 18.5 26-Sep 6.0 6.0 6.2 6.2 7.8 36.2 17.1 5.7 14.6 36.9 25.2 18.4 27-Sep 1.6 3.8 4.5 4.5 5.8 39.3 17.3 5.9 14.9 28-Sep 2.3 4.6 5.3 5.3 6.8 39.3 16.9 5.7 14.6 29-Sep 0.0 0.0 0.0 0.0 0.0 50.5 57.4 5.7 16.1 30-Sep 0.0 0.0 0.0 0.0 0.0 46.2 20.1 6.0 15.8 1-Oct 0.6 0.4 0.5 0.4 0.5 43.0 18.6 6.0 15.5 2-Oct 1.8 1.3 1.4 1.4 1.8 40.7 18.1 5.9 15.5 3-Oct 3.0 2.3 2.3 2.3 2.8 39.3 17.7 6.0 14.9 37.4 4-Oct 4.7 3.7 3.6 3.7 4.3 37.9 17.5 6.0 15.1 37.1 5-Oct 5.9 4.9 4.8 5.0 5.8 36.2 17.3 6.0 14.9 36.9 6-Oct 1.8 3.1 3.5 3.6 4.3 36.9 17.9 6.0 15.1 7-Oct 2.3 3.6 4.2 4.1 4.9 37.9 17.5 6.0 15.1 8-Oct 0.0 0.0 0.0 0.0 0.0 50.5 26.7 6.0 16.1 9-Oct 0.0 0.0 0.0 0.0 0.0 45.4 18.8 6.1 15.8 10-Oct 0.1 0.2 0.2 0.2 0.2 43.8 18.4 6.1 15.8 11-Oct 0.9 0.9 0.9 0.8 0.9 41.5 17.9 6.1 15.5 12-Oct 1.9 1.6 1.6 1.6 1.7 40.0 17.9 6.2 15.1 13-Oct 2.8 2.7 2.7 2.5 2.7 38.6 18.1 6.2 15.5 37.5 14-Oct 4.0 3.8 3.7 3.6 3.9 37.9 18.1 6.2 15.1 37.2 15-Oct 5.4 4.9 4.9 4.8 5.1 36.2 18.3 6.2 15.1 37.0 25.4 16-Oct 6.7 6.3 6.4 6.2 6.6 34.9 18.6 6.2 14.9 36.8 25.2 18.4 17-Oct 7.2 6.8 6.9 6.8 7.3 34.3 18.4 6.3 14.6 36.7 25.1 18.4 13.5 18-Oct 8.1 7.8 7.8 7.8 8.5 32.8 18.4 6.2 14.6 36.5 25.0 18.3 13.4 19-Oct 9.0 8.8 8.8 8.8 9.8 32.2 18.4 6.2 14.6 36.3 24.9 18.2 13.3 20-Oct 10.2 10.0 10.1 10.0 11.1 31.1 18.4 6.2 14.3 36.1 24.7 18.1 13.2 21-Oct 11.4 10.9 10.9 10.8 12.1 30.3 18.3 6.2 14.3 35.9 24.6 18.0 13.2 22-Oct 12.8 12.0 12.0 11.9 13.3 29.7 18.3 6.1 14.1 35.6 24.5 17.9 13.1 23-Oct 13.6 12.6 12.7 12.6 14.1 28.7 18.1 6.2 14.1 35.5 24.4 17.8 13.0 24-Oct 15.5 14.0 14.1 14.0 15.7 27.4 18.3 6.1 13.6 35.1 24.2 17.7 12.9 25-Oct 11.0 11.9 12.5 12.4 13.7 34.9 18.4 6.2 13.6 35.9 24.5 17.8 13.1 26-Oct 11.6 12.4 13.0 12.9 14.4 32.2 18.3 6.2 13.6 35.8 24.4 17.8 13.0 27-Oct 13.0 13.4 13.9 13.8 15.6 32.2 18.3 6.2 13.6 35.6 24.3 17.7 12.9 28-Oct 13.7 14.0 14.6 14.5 16.4 30.3 18.3 6.1 13.8 35.4 24.2 17.6 12.9 29-Oct 15.2 15.6 16.2 16.0 17.9 29.2 18.1 6.2 13.6 35.2 24.0 17.5 12.8 30-Oct 0.0 0.0 0.0 0.0 7.3 48.8 23.2 6.1 13.6 31-Oct 0.0 0.0 0.0 0.0 0.0 51.5 115.0 6.3 16.6 1-Nov 0.0 0.0 0.0 0.0 0.0 44.6 115.0 6.5 16.6 2-Nov 0.0 0.0 0.0 0.0 0.0 46.2 115.0 24.4 16.9 3-Nov 2.7 2.3 2.4 2.3 2.5 42.2 115.0 16.6 16.6 4-Nov 5.2 4.5 4.6 4.5 4.8 40.7 115.0 14.9 15.8 37.0 25.4 5-Nov 6.8 5.9 6.0 5.9 6.4 37.9 72.9 13.6 16.1 36.7 25.2 6-Nov 9.0 8.0 8.2 8.1 8.7 36.2 21.8 12.8 15.8 36.3 25.0 18.2 13.4 7-Nov 11.1 10.1 10.2 10.2 11.0 35.6 19.9 12.1 15.5 35.9 24.7 18.0 13.2 8-Nov 0.0 0.0 0.0 0.0 0.0 46.2 115.0 19.5 16.6 9-Nov 1.7 1.5 1.5 1.5 1.6 43.0 28.5 15.5 16.6 13.5 25.4 25.4 18.4 429 Appendix VII: Operating the Forest Flammability Model Spreadsheet A copy of the FFM in the format of an Excel spreadsheet has been provided with this thesis so that any model output can be checked, and to demonstrate the way that the model can be structured. The model will run on Excel 2003 but does not appear to run on Open Office. Please Note: The attached spreadsheet is the 19th July 2011 version of the model and the following directions apply to this version. Launching the model Due to its size, the model takes a few minutes to open, so please be patient. All macros must be enabled to use the model. Model worksheets There are 14 worksheets in the model, 9 of which have been hidden as they are not used by the operator and should not be changed. Some parts of the visible worksheets are also hidden for aesthetic purposes, to prevent accidental alterations and to make the model run more efficiently. Each worksheet is described in table 1. Table 1. Worksheets in the model Worksheet Description Register Lists all changes to the spreadsheet Settings Table of constants that can be altered with improved data Datasets Contains a collection of macros that will fill fuel or weather variables for a number of scenarios a Main interface for entry of weather and environmental variables along with broad fuel descriptors, and to describe the fire behaviour Fuels Entry point for descriptive fuel variables Monte Carlo Lists flame height and rate of spread data from Monte Carlo analyses Report1 Provides a summary of fire behaviour for series Study data Collects weather and fire behaviour data for series Dynamics Working sheet for fire growth and calculating pulsing behaviour S Working sheet for surface fire behaviour and some other variables NS Working sheet for near surface fire behaviour E Working sheet for elevated fire behaviour M Working sheet for midstorey fire behaviour C Working sheet for canopy fire behaviour Graph Working sheet for fuel and flame structure graph on sheet a Status Hidden Hidden Visible Visible Visible Visible Visible Visible Visible Hidden Hidden Hidden Hidden Hidden Hidden Main interface Sheet a (Figure 1) has five main areas, these are: 1. 2. 3. 3. 4. Weather/terrain conditions Forest structure Fire behaviour Graphs Analyses 430 The first two of these are for data entry, the next two provide information about the modelled fire behaviour and the ‘Analyses’ are a series of macros for analysing the fire behaviour. There are two graphs – the series graph on the left and the fuel/flame structure graph on the right. Figure 1. Main interface for the FFM Entering data Data need only be entered into the sheets ‘a’ and ‘fuels’. On each worksheet, coloured cells should not be changed and all data can be entered into the white cells. Weather and terrain data Weather and terrain variables are entered into sheet a as indicated by the cell names. The cell ‘Mean’ can be left blank, this is used only for long-term means such as the dew point averaged over a few days, or for other unique inputs. The cell “soil moisture” is optional, as soil moisture will be calculated from the KBDI and soil texture. If a number is entered into soil moisture, this will override the automatic calculation. Soil texture should be chosen from the drop-down menu associated with the cell. “Altitude” may be left blank unless weather is to be interpolated for a specific point. This will be described later. The two unnamed cells below this refer to atmospheric pressure (used for dead fuel moisture calculations) and soil texture, which is chosen from a drop-down menu and is used in some live fuel moisture models. The cell K3 may also be left blank; this provides a title for reports when series are being studied. 431 Fuel data Fuel data is predominantly entered into the ‘Fuels’ sheet (fig. 2), although some values can be entered into sheet ‘a’. The surface fuel load can be either entered directly into the cell O5 on ‘Fuels’, or calculated automatically by leaving this cell blank and providing the litter depth and percent cover in the lower 2 cells. Fuel load is calculated as 5t.Ha-1 for every cm depth of fuel. Mean fuel diameter should be set to 5 in most circumstances, but may be lowered to 4 in young regrowth with no bark or sticks or raised to 6 in areas of heavy woody litter. Dead fuel moisture will be calculated from the weather parameters using Gould et al (2007a) as rendered in Tolhurst (2010) and adjusted for modelled fuel temperature and humidity, however a measured value can be used to override the automatic calculation if it is entered into cell O9 on the sheet ‘Fuels’. Live fuel moisture for each species can be entered directly into the white cells Q24 to Q34 on sheet a if they have been measured, or fuel models can be entered into the white cells at B6 to L6 on ‘Fuels’ if models exist. Fuel models referring to dew point should reference cell S!B11, and those referring to dew point depression should reference cell S!B10. The remaining fuel information has four parts – plant spacing, plant count, vertical relationships and detailed fuel data. The plant spacing data is entered into the white cells in column O on sheet a and the plant count into column P rows 24, 25, 27, 28, 30, 31 and 33, 34. Above the “count” and “moisture” cells of the near-surface and understorey strata and of the “count”cell for the midstorey are cells with drop-down menus that describe whether the stratum grows under the strata above it. The default value is ‘T’, indicating that they do. Changing the cell to ‘F’indicates that they do not, and changing it to ‘auto’ allows the model to decide based on the comparative heights of the strata. If the lower stratum is taller than the upper one, the model decides the answer is ‘F’. The remaining data is entered into the white cells on Fuels. One limitation of the spreadsheet is that it reports an error if some of the fuel columns are left empty, so if there are no species recorded these should be populated with dummy values and the species composition on ‘a’ left blank. Care should also be taken to ensure that plant dimensions are logical as an error will result if canopy base height is greater than top height for example.. The structure of the fuel array will be shown on the fuel/flame structure graph on sheet a. 432 Figure 2. The Fuels datasheet Using stored data A small number of fuel and weather datasets have been stored in macros and can be accessed from the ‘Datasets’ page by clicking on any of the symbols. Three categories of data are stored, weather scenarios record the weather over a short time period such as during a fire run, seasons record three months worth of weather for a site and forest profiles record the relevant fuel parameters for a given community. Operating the model The model can be run from the commands under the ‘Analyses’ heading, and the fuel/flame structure graph can be zoomed in or out by clicking on the words ‘Near Surface’ through to ‘Zoom to forest’ in the ‘Forest Structure’ area. The analyses buttons (fig. 3) are described below. Clear This command will clear the records from the Study data sheet that are shown in the series graph Calculate This command will run the model over a series of iterations to find the relevant mean values and to calculate changes in wind structure with the onset of active crown fire. Wind This will run the model keeping all factors equal but changing wind speed from 0 to 100km/h at 2km/h intervals. The results will be shown in the series graph. Figure 3. The Analyses menu Slope This will do the same as the wind command, but will alter slope from minus 50o to 50o. 433 Scenario This is intended for incident management, and analyses a short series of weather from the ‘Study data’ sheet (try one of the stored macros for an example). The model will be then run for each of the weather records and the output shown on the series graph. Season This is used for flammability management studies, and analyses the fuel and terrain conditions against three months worth of weather data from the ‘Study data’ page. This data can either be entered manually or one of the stored data sets used. The results are shown in the series graph as well as in ‘Report1’. The report summarises the data and provides indices such as the percentage of the season for which initial attack would have failed. Smooth This smooths the data composing the series graph by averaging it out over the three values on either side of each point. This is useful for wind and slope studies. Monte Carlo This analysis maintains the fuels information, but runs 1000 scenarios of randomly derived variables for the weather and terrain data within the bounds described in chapter seven of this thesis. The analysis takes many hours, so the number of the run being examined is shown in the cell below the word ‘Monte Carlo’ and the completion time is shown in the cell below that. The final data produced are given on the ‘Monte Carlo’ sheet. Studying forest flammability As discussed in chapter eight, managing forest flammability is not a matter of simply “reducing the fuels”; it is a matter of managing the whole forest. The FFM is able to assess the structure of a forest against strategic fire management objectives using risk statistics on the ‘Report1’ sheet (fig. 4). To study the flammability of a forest against identified objectives such as maximising the percentage of time for which direct attack is possible or minimising the number of days where long-distance spotting will exceed a designated maximum; the following steps should be followed: 1. 2. 3. 4. Enter the relevant fuel and slope parameters for the site Enter a season’s worth of weather into the ‘Study data’ page Enter the maximum desired spotting distance to cell Q31 on the report Click ‘Season’ on sheet a. The 4th graph on the report (Forest Structure) can be tailored to the dimensions of the forest using the four buttons in the ‘structure zoom’ box on the top right of the worksheet. 434 Figure 4. The flammability report 435