Report

Forecasting
hazards, averting
disasters
Implementing forecast-based
early action at scale
Emily Wilkinson, Lena Weingärtner, Richard Choularton,
Meghan Bailey, Martin Todd, Dominic Kniveton and
Courtenay Cabot Venton
March 2018
Overseas Development Institute
203 Blackfriars Road
London SE1 8NJ

Tel: +44 (0) 20 7922 0300

Fax: +44 (0) 20 7922 0399
Email: info@odi.org.uk

www.odi.org
www.odi.org/facebook
www.odi.org/twitter

Readers are encouraged to reproduce material from ODI publications for their own outputs, as long as they are not being sold commercially. As copyright
holder, ODI requests due acknowledgement and a copy of the publication. For online use, we ask readers to link to the original resource on the ODI website.
The views presented in this paper are those of the author(s) and do not necessarily represent the views of ODI.

© Overseas Development Institute 2018.

This work is licensed under a Creative Commons Attribution-NonCommercial Licence (CC BY-NC 4.0).

Cover photo: Men cover the windows of a car parts store in preparation for Hurricane Irma in San Juan, Puerto Rico ©2017 Alvin Baez / Reuters Pictures
Acknowledgements

The authors would like to thank Erin Coughlan (Climate Centre), Richard Ewbank (Christian Aid), Maggie Ibrahim
(World Vision), Kevin Huttly (VisionFund International), Debbie Hillier (Oxfam), Sheri Lim (CARE), Luke Caley
(Start Network), Sarah Barr (Start Network), Dunja Dujanovic (FAO), Benedikt Lucas Signer (World Bank), Hugh
MacLeman (OECD), Thorsten Klose-Zuber (German Federal Foreign Office), Steffen Lohrey (German Red Cross),
Stefanie Lux (German Red Cross), Rebecca Nadin (ODI), Nicola Ranger (DFID), Fergus McBean (DFID) and Henry
Donati (DFID) for reviewing this paper and/or providing additional inputs.
We gratefully acknowledge support from Rosalind West (DFID), Helen Ticehurst (UK Met Office) and Bill Leathes
(UK Met Office) in coordinating the review process, and Matthew Foley and Katherine Shaw (ODI) for copy-editing
and supporting the production of this paper.

3
Contents

Acknowledgements3

List of boxes, tables and figures 5

Acronyms6

1  Introduction 7
1.1  Methodology 8

2  Forecasting and decision-making 10

2.1  Characteristics of hazard information  10
2.2  Impact-based forecasting 12
2.3  Triggers for action  14

3  Timing and planning actions 16

4  Financing FbA 19
4.1  Dedicated FbA funds and funding windows 19

4.2  Insurance and contingent finance  19

4.3  Standard resource allocation processes 20

5  Mechanisms for delivering FbA 21

5.1  FbA linked to community development programmes  21
5.2  FbA with governments and through social protection and safety nets 21
5.3  International humanitarian response 22

6  The evidence base for forecast-based early action 23

6.1  The costs of anticipatory action 23
6.2  Other cost–benefit considerations 24

7  Taking forecast-based early action to scale 26

7.1  Approaches to scaling up 26
7.2  Embedding FbA in financing and delivery systems at scale 26
7.3  Challenges for taking FbA to scale 26

8  Conclusion 29

References30

Annexes32

4
List of boxes, tables and figures

Boxes

Box 1  Probabilistic forecasts and statistical methods for FbA 11

Box 2  Selecting hazard triggers 14

Box 3  Levels of automation in decision-making 15

Box 4  Examples of evidence on the costs and benefits of early action 25

Tables

Table 1 Methods for integrating hazard, vulnerability and exposure information to predict impact 13

Table 2 Examples of scaling up FbA 27

Table A1 List of FbA initiatives reviewed 32

Table A2 Selected FbA pilots 34

Table A3 List of key informants 36

Table A4 Glossary of FbA concepts 37

Figures

Figure 1  Map showing FbA initiatives and short descriptions of selected pilots 9

Figure 2  Triggering system for El Niño impacts across Peru 12

Figure 3  The interaction of climate-related hazards, vulnerability and exposure of human and natural systems 14

Figure 4  FbA, early response and late response in the case of droughts and cyclones 17

Figure 5  Illustration of possible outcomes of forecast-based early action 24

5
Acronyms

ACAPS Assessment Capacities Project

ARC African Risk Capacity
ARDIS African and Asian Resilience in Disaster Insurance Scheme
DFID UK Department for International Development
DREF Disaster Relief Emergency Fund
DRR disaster risk reduction
EAPs Early Action Protocols
ECMWF European Centre for Medium-Range Weather Forecasts
EENIP Extreme El Niño Insurance for Climate Change Prevention and Adaptation (Peru)
FAO Food and Agriculture Organization
FbA forecast-based early action
FbF forecast-based financing
FEWS NET Famine Early Warning System Network
FFP USAID Office of Food for Peace
HSNP Hunger Safety Net Programme (Kenya)
IASC Inter-Agency Standing Committee
INFORM Index for Risk Management
IPC Integrated Food Security Phase Classification
IPCC Intergovernmental Panel on Climate Change
IRI International Institute for Climate and Society
LEAP Livelihoods, Early Assessment and Protection (Ethiopia)
MFI microfinance institution
NDMA National Drought Management Authority (Kenya)
ROI return on investment
SST Sea Surface Temperature
SOPs Standard Operating Procedures
WFP World Food Programme

6
1  Introduction
Donors and humanitarian agencies are thinking carefully referred to here, forecast-based early action initiatives2
about how to use forecasts to provide earlier support are specialised mechanisms linking financing and early
to at-risk communities before a disaster occurs. While action to forecasts of hazards and disaster impacts.
this interest stems from a desire to reduce the growing No one definition of FbA has been agreed, but to help
humanitarian burden and reconsider how aid is spent on distinguish it from other risk financing arrangements and
humanitarian crises, forecast-based early action is also of humanitarian and disaster risk management practices we
interest to development professionals operating in social refer to the use of climate or other forecasts to trigger
protection, disaster risk management and risk financing: funding and action prior to a shock or before acute
preventive action should happen anyway, but in a impacts are felt, to reduce the impact on vulnerable
context of limited resources forecast-based early action people and their livelihoods, improve the effectiveness of
can help with decisions about how to best allocate funds emergency preparedness, response and recovery efforts,
in advance of an imminent impact. and reduce the humanitarian burden.
While practitioners agree on the importance of early While the paper draws on evidence from a wide
action, there is a wide interpretation of what this means range of FbA initiatives over the last five years, it is not
and when it can occur. Forecast-based early action (FbA) intended to provide a comprehensive review, but rather
initiatives are diverse, with very different approaches to draws out some of the commonalities and differences
the timing of decisions and actions, and to the types of between these initiatives within what is a disparate field
forecast, monitoring data and delivery mechanisms used. of practice. The paper situates FbA innovations within
They are similar in design to early warning systems: broader humanitarian, disaster risk management and
both are set up to minimise and prevent the impacts of development agendas and reform processes. The authors
imminent threats by providing information and support to examine the full chain of data use and decision-making:
at-risk communities.1 Forecasting and communication of from decisions about the forecast and monitoring data
early warnings have improved significantly in recent years, to be assessed to the selection of triggers and thresholds
but action based on those warnings has not kept pace (and methods for integrating bio-physical and socio-
due to a lack of readily available resources and internal economic impact data), protocols for action and the
inefficiencies in NGOs and UN and government agencies. financing mechanisms needed to deliver support to
FbA mechanisms respond directly to this challenge communities before a disaster happens.
by placing considerable emphasis on decision-making This has resulted in a typology of forecast-based early
protocols, so actors know what to do on the basis of a action. The typology includes questions around:
forecast; on ex ante financing of early action; and by using
cost–benefit analysis more rigorously to help promote 1. Forecasting and decision-making: FbA involves a
ex ante investment in disaster risk reduction (DRR). As range of forecasts, indicators and decision-making
such, FbA has the potential to revolutionise disaster risk mechanisms, from automated triggers to forecast-
management in a way that previous efforts to improve the informed decision-making.
links between early warning and early action have not. 2. Timing and planning early actions: FbA mechanisms
This paper identifies the core features of over 25 are designed to trigger and inform action across
FbA instruments designed to anticipate and reduce multiple time-scales before a disaster occurs, ranging
the impacts of natural and man-made hazards (see from several days (for a cyclone) to a year (in advance
Annex 1). It outlines how, by integrating forecast-based of an acute drought).
decision-making in existing national and international 3. Financing: Forecast-based action programmes
organisations and NGO delivery systems and in have applied a variety of financing tools, including
international humanitarian financing mechanisms, dedicated funds, specific windows in emergency
forecasts could play a more significant role in response funds, insurance and direct links to regular
humanitarian practice and disaster risk management. As resource allocation processes.

1 UNISDR (2017) defines an early warning system as: ‘An integrated system of hazard monitoring, forecasting and prediction, disaster risk
assessment, communication and preparedness activities, systems and processes that enables individuals, communities, governments, businesses and
others to take timely action to reduce disaster risks in advance of hazardous events’.

2 What is referred to here as FbA overlaps with other concepts such as early warning/early action and Forecast-based Financing (FbF).

7
4. Delivery: FbA mechanisms can be deployed through Humanitarian organisations that have developed
a range of delivery channels, including community- and piloted forecast-based early action mechanisms
based emergency preparedness processes and social understand and use the concepts of ‘forecast’ and ‘early
protection systems. action’ very differently (see Annex 4). Early action is
relative to the baseline of the implementing organisation
The paper concludes by considering the potential for FbA (the kind of action they were undertaking before) and
mechanisms to be adopted at scale in humanitarian and the type of hazard, whereby ‘early’ means something very
disaster risk management decision-making through the different for drought than it does for flooding. Similarly,
use of different sources of risk financing and national and the type of forecast used depends very much on the
international delivery mechanisms. There are significant context (what kinds of risks are faced), the relationship
challenges associated with using forecasts systematically with monitoring agencies and the mandate of the
to trigger the release of international and national organisation (whether it can support communities it does
humanitarian funds – and hence taking some control over not already operate in).
allocation away from donors, governments and NGOs – To date, most – but not all – FbA initiatives are
but this kind of step change is necessary if FbA is to have concerned with forecasting extreme weather-related
a significant impact on the lives of vulnerable populations. events that will have a negative impact on vulnerable
populations, so the analysis of FbA mechanisms in this
report focuses on these.
1.1  Methodology The report draws on project documentation and
grey literature, as well as 14 key informant interviews
This paper draws on data from a selection of operational with representatives of FbA initiatives (see Annex 3),
or piloted FbA mechanisms (see Annex 1). These were to extract relevant technical information on the nature
selected to ensure wide coverage of the various types of of the forecast and impact information used, the design
mechanisms, with different objectives, technical designs, of triggers and decision-making protocols and the
operational contexts, governance arrangements, scales of delivery systems through which support is provided to
operation, implementation location and hazard type. communities in advance of a disaster.

8
 Tajikistan
Initiative: Start Network Anticipation Window
Kenya Period: July 2017 flooding and mudslides
Initiative: FAO Early Warning Early Action Numbers reached: 26,864 people Bangladesh
Period: December 2016, negative forecasts for Actions taken: Initiative: Red Crescent Forecast Based
rainy season Infrastructure for community-level Financing (Phase I)
Numbers reached: 60,000 people preparedness. Provision of tools and equipment Period: 2016 flood, 2017 flood, 2017 cyclone
Actions taken: for responding to flooding and landslides Numbers reached: 5,000 households
Livestock feed, supplements and health treatments, Emergency training, planning and simulation Actions taken:
destocking campaigns, training of local government Unconditional cash transfer
Figure 1  Map showing FbA initiatives and short descriptions of selected pilots

officials on livestock emergency procedures

Rehabilitation of boreholes and provision of water

9
Note: please see Annex 2 for full details on the selected pilots.
Peru
Initiative: Red Cross Forecast based Financing (Phase I)
Period: 2016 El Niño rains and flash floods
2016 cold wave
Numbers reached: 2,440 households Somalia
Actions taken: Initiative: SomReP
2016 El Niño: Period: 2014, late and inadequate Gu rains
Provision and purification of water, drainage Numbers reached: 6,438 households
Hygiene and first aid training and kits, actions Zimbabwe Actions taken:
against mosquitoes Initiative: WFP FoodSECuRE Livestock vaccination
Stabilisation of houses/roofs Period: 2015/2016 negative El Niño forecasts Conflict resolution (water and pasture)
Numbers reached: 4,500 people Vouchers and cash programming
2016 cold wave: Actions taken: Rehabilitation of boreholes and provision of water
Distribution of hay, veterinary kits Training farmers on ‘climate-smart’
First aid and winter gear kits agriculture and business practices,
provision of seeds and fertilisers
2  Forecasting and
decision-making
Understanding hazards and their potential impact is weather and climate variables (such as precipitation
central to promoting early action. This section describes and temperature), as well as river levels for flooding.
attempts to forecast hazards, and to use vulnerability For drought, many also use real-time monitoring of soil
and exposure information to predict disaster impacts and moisture and vegetation conditions from satellites, and
develop triggers for action. As FbA is born out of a desire assess food security status using in situ reports.
to more effectively translate early warning information Around half the systems reviewed in this research
into concrete action, these initiatives pay a great deal use probabilistic forecasts (including all the Red
of attention to the decision-making processes needed Cross pilot systems, the World Food Programme
to generate early action. Two principal approaches (WFP)’s FoodSECuRE programme and the Inter-
to decision-making dominate forecast-based action Agency Standing Committee (IASC) El Niño Standard
initiatives: automated triggers and forecast-informed Operating Procedures). Several mechanisms use real-time
decision-making. These are discussed below. monitoring data rather than forecasts, which can provide
advance warning of socio-economic and humanitarian
impacts of drought. These include social protection
2.1  Characteristics of hazard information systems (the Hunger Safety Net Programme (HSNP)
in Kenya and the Livelihoods, Early Assessment and
Forecast information comes in many forms, from raw Protection (LEAP) project developed by the government
data to qualitative statements such as bulletins from and WFP in Ethiopia); insurance-based systems such
national meteorological services.3 FbA mechanisms as African Risk Capacity (ARC) and the Extreme El
typically require quantitative information to define Niño Insurance for Climate Change Prevention and
objective triggers for decision-making. FbA mechanisms Adaptation in Peru (EENIP); and the Start Drought
must consider the hazards that need to be forecast and the Financing Facility. For slowly evolving drought hazards,
forecast data required (or monitoring data in some cases, real-time monitoring of impact precursors is clearly
for slow-onset events), the source of that information, favoured over forecast information. A few systems
whether the spatial/temporal scales and lead-times and the involve a hybrid of both real-time monitoring and
forecasted variable meet their requirements, whether the forecast information, including the USAID Food for
reliability of the forecasts (known as ‘forecasting skill’)4 is Peace’s use of the Famine Early Warning System Network
well established and whether deterministic or probabilistic (FEWS NET) Food Assistance Outlook Brief, the Food
forecasts are required. The Red Cross/Red Crescent and Agriculture Organization (FAO) Early Warning-
Forecast-based Financing (FbF) manual provides guidance Early Action system and the Start Fund’s Anticipation
on this process.5 Window.

2.1.1  Hazard forecasts and monitoring data 2.1.2  Sources of forecast information
Most of the FbA mechanisms reviewed in this report FbA uses a wide range of climate forecast information,
focus on the hazards posed by extreme weather and based on what is available and appropriate. Systems
climate, and so incorporate forecasts of extreme using probabilistic forecast information typically draw

3 Weather and climate forecasts vary in terms of their lead-times (e.g. forecasts for days, weeks, months and seasons ahead), their spatial coverage
and their detail. The more comprehensive and coordinated forecasts are obtained from numerical models of the atmosphere and/or the climate
system. These weather and climate forecasts are combined with bio-physical impacts, such as river flow, crop yields and fodder quality, to create a
hazard forecast. The Global Flood Awareness System (GloFAS), for example, couples weather forecasts with a hydrological model, with lead times
of up to 15 days to forecast river flow and flood risk globally.

4 A brief example of how to assess forecast skill is provided at https://www.metoffice.gov.uk/research/climate/seasonal-to-decadal/gpc-outlooks/user-

guide/interpret-roc

5 http://fbf.drk.de/fileadmin/Content/Manual_FbF/01_Manual/01_Manual_For_Forecast-Based-Financing.pdf

10
on products from international, regional and national 2.1.4  Forecast lead time
forecasting centres. Products from international and Forecast lead times in the systems surveyed vary from
regional forecasting centres are most common as these days through to seasons (and up to 12 months in
are freely available and considered reliable. Where advance of acute humanitarian impacts of drought). The
appropriate, these are complemented with products lead time depends in part on the hazard system under
from national hydrological and meteorological services. study, with short lead times (typically days) for pluvial
The Red Cross FbF pilots for flooding in Bangladesh flooding from heavy rain and heat/cold waves, days
and Togo, for example, use forecasts produced and used to weeks for fluvial flooding, depending on river basin
by national early warning systems in those countries. size, and months for drought and food security hazards.
Drought/food security hybrid systems typically use a range Systems using forecasts can generally provide longer
of information on food production, access and livelihood lead times than those based on monitoring information,
outcomes from national agencies and international although monitoring can provide usefully long lead times
assessments (e.g. FEWS NET, Integrated Food in more slowly evolving systems, for example in larger
Security Phase Classification (IPC)), and merge these into river basins (the Mono river in Togo is one example), and
an assessment of food security status and likely risk. for drought/food security.
A small number of systems operate over a range of
2.1.3  Spatial scale of operation forecast lead times, drawing on forecast information
The scales at which forecast information is assessed from seasons through to days. This can allow for the
range from districts to counties/provinces and the progressive staging of actions. A notable example is the
national level. Actions need to target specific places and Red Cross extreme rain/flood hazard system in Peru (see
people that are vulnerable to climate extremes. This can Figure 2), which involves preparedness actions triggered
present problems for the application of climate forecasts, automatically by forecasts at various lead times (days/
particularly seasonal ones, as these forecasts tend to be month/season).
made at a coarse scale with limited information for very
local decision-making.6 The type of forecast information 2.1.5  Degree of recognition and application of
required depends on the hazard; while coarse resolution forecast skill
forecasts may be suitable for droughts or heatwaves, Limited detail on the design of FbA systems makes
which are spatially extensive, forecasting flash floods it difficult to determine how much forecast skill is
requires very high-resolution modelling. There is a directly taken into account in agreeing triggers for
clear role for national meteorological and hydrological action, although the Red Cross FbF Manual does
services in providing such detailed information. encourage this. Other examples of initiatives that

Box 1  Probabilistic forecasts and statistical methods for FbA

Weather forecasts are usually accurate for hours or a few days ahead, but it might not be possible to predict
exact conditions at precise times. Nevertheless, it is possible to make forecasts of the statistics of atmospheric
conditions over an extended period of time (a month or a season), with a longer lead of up to many months,
e.g. a forecast of monthly or seasonal rainfall totals a few months ahead. These monthly/seasonal forecasts from
models are probabilistic, meaning that they typically come from multiple runs of the model (an ‘ensemble’).
Ensemble forecasting is now the standard approach used in major modelling centres and accounts for inherent
uncertainty in both the climate system and the models themselves. Probabilistic forecasts provide an estimate of
the likelihood of some event occurring, e.g. a 30% chance of rainfall greater than some value.
Monthly/seasonal forecasts can also be derived using statistical approaches such as regression equations,
predicting climate some months ahead, and Sea Surface Temperatures (SSTs), where a strong relationship exists
in historical data. This is a standard method used by many African national meteorological services. This type
of local approach is more appropriate for the context and can have greater skill (i.e. get it right more often)
than global models. The forecasts are often expressed as probabilities, reflecting uncertainty in the statistical
relationships. Statistical ‘calibration’ of numerical models can improve forecast skill: a good example is the
calibrated multi-model system used by WFP to derive drought forecast triggers, developed by the International
Research Institute for Climate and Society (IRI) at Columbia University. Multiple forecast products can also be
merged using an ‘expert judgement’ system, e.g. the consensus products of the Regional Climate Outlook Forums.

6 The resolution of numerical forecast models remains relatively coarse (at best ~40km grid cell). The most commonly used IRI multi-model
processed product is available on a 1-degree grid (~111km at the Equator), while some of the consensus products (e.g. from regional climate
outlook forums) provide only very broad regional distinctions. Tools are available to statistically downscale these forecasts (e.g. the Fact-Fit tool),
although these are not typically used in the systems we surveyed. For that reason, some smaller-scale systems use local hazard risk forecasts from
bespoke systems.

11
Figure 2  Triggering system for El Niño impacts across Peru

Note: the system uses observed information in addition to actual forecasts, specifically real-time sea surface temperatures, for which there is a
danger level threshold, but no probability threshold. For example, high-probability actions for long lead preparedness may be triggered from an
IRI seasonal forecast of (i) rainfall in the highest 10% of past events (the danger level) with (ii) a probability of 40% (i.e. 4x the normal likelihood).
Source: from Implementing forecast-based financing mechanism in Peru to enable preparedness for El Niño impacts, reproduced with
permission from the Red Cross Red Crescent Climate Centre.

explicitly take forecast skill into account include the 2.2  Impact-based forecasting
IASC, which evaluated forecast skill while developing
its Standard Operating Procedures for El Niño and FbA is most relevant for humanitarian actors if the hazard
La Niña, and WFP and IRI, which have built a series being forecast is likely to have a negative impact on a
of tools for evaluating forecast skill in their trigger vulnerable population, so where the hazard, vulnerability
design process. Overall, there seems to be widespread and exposure interact (see Figure 3). Understanding at what
awareness of the relationship between increasing lead point an action is relevant, where that action should be
times and the increasing uncertainty of forecasts, and focused and for whom requires linking the forecast hazard
the inherent trade-off in wanting to have a long lead to anticipated impacts. In order to prepare for and plan
time (which gives a greater range of action options) responses to forecast hazards, a clear understanding of the
and the risk of acting in vain (because the forecasting potential scale, severity and timing of impacts is needed.
skill is weaker for longer lead times). Identifying ‘low Many agencies have implemented or are in the process
regrets’ actions is a common approach to dealing of developing methods to integrate vulnerability and
with this trade-off. Other options include using exposure information with hazard forecasts to assess
observational data alongside forecasts, to reduce likely impact. Four major methods are currently being
uncertainty about the risk, and adding a mechanism used: the threshold method, the qualitative combination
that can stop implementation before large costs are method, the impact modelling method and the climate
incurred, if subsequent forecasts indicate that risk is sensitivity method (see Table 1).
below the threshold.

12
Table 1  Methods for integrating hazard, vulnerability and exposure information to predict impact

Methodology Data required Spatial scale Development Examples

Threshold Define a forecast At least one Defined for Defined in • Phase 1 of FbF implemented by the Red Cross
method threshold at which historical event, a specific advance. in Bangladesh, Peru, Uganda and elsewhere
people or infrastructure or simulations, location or • Heat health action plans set temperature
in a specific location are to identify the a specific thresholds for action based on historical
expected to be negatively magnitude of the infrastructure. relationships between temperature and
impacted, based on hazard impact. mortality/morbidity in a specific city. England’s
the vulnerability of that Heatwave Plan has a threshold for action
location/infrastructure. when maximum temperatures are forecast
to be 32° in London during the day and 18°
at night, with slightly modified thresholds for
other regions (Public Health England, 2014).
In India, the heatwave plan developed for
Ahmedabad has its lowest alert level starting
at 41.2° (Knowlton et al., 2014).
Qualitative Create a composite Vulnerability Large spatial Can be done • FAO’s early drought response in Kenya.
combination index that combines rankings of scale with in real-time • The Start Fund Anticipation Window, whose
method relative vulnerability locations or groups different discussions. rapid decision-making process uses inputs
with forecasted hazard within a larger vulnerability from forecasting partners such as IRI and the
magnitude to create a region. No historical groups. London School of Economics and Political
relative priority score, data is required. Science (LSE), a survey of its membership,
often a qualitative independent secondary data analysis from the
assessment by a group. Assessment Capacities Project (ACAPS) and
analysis from a technical advisory group called
FOREWARN.
• The UK Met Office brings together experts to
look at a weather forecast and assign colour
codes to different regions depending on a
combination of probability and impact, as part
of impact-based forecasting.
Impact Develop a model that Historical hazard Depends on Model • Dzud FbF, implemented by the Mongolia Red
modelling combines hazard and impact data the model. developed in Cross, is part of a second phase of Red Cross
method magnitude with as well as data on advance. projects that will build on the simpler threshold
vulnerability and exposure the relationships model and allow programmes to scale up
to predict a level of impact. between them to based on modelled impacts rather than a
improve the model specific threshold for a local area.
• Damage models are often run by the
insurance sector as an extreme event is
approaching and immediately after it hits.
• ARC runs crop models using satellite-derived
rainfall estimates, to estimate crop yields at
the end of the agricultural season. This is
combined with vulnerability data to trigger
insurance pay-outs.
• The Index for Risk Management (INFORM)
was used to calculate the potential impact of
El Niño. Forecasts were mapped onto risks
already quantified through INFORM.
Climate- Using a combination of Baseline socio- Large-scale, Developed • WFP and IRI’s approaches to food security,
sensitivity socio-economic baseline economic data, most often in advance which identify areas where food insecurity
method data and climate data, livelihood zones national level. to target FbA most closely correlates with climate risk,
identify areas where and climatology. and other and then develop and deploy tools based on
vulnerability is most climate risk the assumption that, in these places, efforts
closely correlated with management would have the greatest potential impact.
forecastable climate risks. tools. Studies using this method show that not all
food-insecure areas have high correlation with
climate risk, contrary to conventional wisdom.
This differs from the impact modelling method
in that it attempts to uncover the relationship
between climate risk and impacts, rather than
trying to quantify anticipated impacts.

13
Figure 3  The interaction of climate-related hazards, vulnerability and exposure of human and natural systems

IMPACTS

Vulnerability
CLIMATE SOCIOECONOMIC
PROCESSES
Socioeconomic
Natural variability
pathways
Hazards RISK
Anthropogenic Adaptation and
climate change mitigation actions

Governance

Exposure
EMISSIONS
and land-use change

Source: IPCC (2014).

Many methods for integrating impact-relevant

information overlap (for example, statistical modelling is Box 2  Selecting hazard triggers
just a more complicated version of the threshold method,
and the qualitative method still requires some sort of Hazard triggers are established and agreed in
advance, based on forecasts or measurements.
threshold in order to start a discussion). The method
For flood-related disasters, for example, 20mm of
selected is likely to be a function of several factors,
rainfall within a specific time period could be a
including data availability, how well we understand the
trigger to initiate a set of action(s). The threshold
hazard–impact relationship (and if it is too complicated
is referred to as the ‘danger level’ of interest. Often
to model), whether unexpected events can sway the result multiple increasing thresholds are selected to trigger
and the scale of the hazard itself. The lead time of the different levels of action (e.g. a forecast of 10mm
hazard is also a factor: it might not be possible to use a or 20mm or rain will trigger amber or red alerts).
complex qualitative method for a rapid-onset event like a However, this simple approach is complicated
flash flood. Finally, the characteristics of the infrastructure because forecasts are inherently uncertain, and so
or the population at risk will also determine the kind of are often expressed in a probabilistic form (for
assessment method that can be used: the threshold method example, there is a 30% probability of exceeding
might be best for a specific situation, such as whether a the threshold of 20mm of rainfall; see Section 2.1).
particular wind speed will cause a bridge to collapse or a In these systems, triggering actions requires defining
particular water level will cause a dam to burst. both the danger level threshold (e.g. 20mm of rain)
The field is rapidly evolving due to advances in and the probability of an occurrence of that danger
computing power and data availability. While all the level in the forecast (e.g. a 30% probability). Both
methods reviewed here are being used more frequently values have to be carefully selected so that actions
than in the past, quantitative modelling is growing are triggered with an acceptable level of frequency.
particularly rapidly. It is important to note, however, We can distinguish two types of trigger systems:
that the examples reviewed in this paper are all led or
1. Deterministic systems involving a single trigger
supported by international agencies. Low-income countries
(i.e. the danger level of some parameter), which
face significant cost and capacity limitations in developing
can be applied to either a deterministic forecast
impact models.
(which provides a single predicted outcome)
or, more usually, real-time monitoring of some
2.3  Triggers for action precursor to disaster combined with biophysical
information, e.g. upstream river flow or
FbA mechanisms are designed to release funds and vegetation condition.
initiate early actions when pre-established thresholds are 2. Probabilistic forecast systems, which require
met (see Box 2). Key to this is a predefined process for both a danger level and probability thresholds.
either directly initiating specific activities, or developing For climate extremes, this would be an ensemble
action plans in real time after the forecast has been forecast system.
issued. Throughout this process, effective coordination

14
and urgency are crucial in order not to miss the Box 3  Levels of automation in decision-making
opportunity for early action before a crisis escalates. The Red Cross/Red Crescent FbF projects remove
Start Network’s experience with its Crisis Anticipation as much real-time decision-making as possible
Window ahead of the arrival of Hurricane Irma in the through Standard Operating Procedures (SOPs),
Caribbean shows how, despite discussion starting a full or Early Action Protocols (EAPs) as they are now
week before the hurricane made landfall, uncertainty known. These are developed well in advance with
on where the storm would hit delayed action (Start decisions already made about triggers, actions
Network, 2017). While some practitioners have argued and targeting. Experts from national hydro-met
for and trialled automated triggers in FbA initiatives offices, disaster risk management authorities and
that require minimal real-time decision-making, several international scientific institutions are involved
key informants highlighted the importance of human in the definition of the forecast threshold used as
judgement in tracing evolving crises and deciding on a trigger. Upon submission of the EAP/SOP for
adequate responses on the ground. funding, the trigger is reviewed by expert bodies
to establish whether it qualifies for funding.
2.3.1  Automatic versus subjective triggers However, once the forecasts indicated in the
A key distinction exists between FbA systems that involve EAP reach the defined threshold, no more expert
judgment is used and the funding decision is taken
automatic triggers and subjective triggers based on expert
automatically. The use of pre-agreed automated
judgement (see Box 3). Institutions including the Red Cross
plans is a substantial change from the disaster-
and WFP have established automated triggers to release
specific decision-making processes that typically
funding to implement early action or contingency plans
characterise humanitarian operations within the
developed using expert opinion. These approaches front- Red Cross/Red Crescent movement.
load the decision-making process and directly link climate FAO uses a combined approach in its Early
forecasts to their potential consequences (see Box 2 for a Warning–Early Action activities. This means
more detailed description of hazard triggers). For example, cross-checking forecast triggers with human
the Uganda Red Cross uses GloFAS, jointly developed by judgement to validate triggers based on the
the European Commission and the European Centre for situation on the ground, to understand what
Medium-Range Weather Forecasts (ECMWF), to trigger is realistic in a given context and to assess the
plans developed through consultation with experts on quality and reliability of the system. The approach
water and sanitation programming, which in this case includes consultation with national and regional
involves the distribution of flood relief supplies such as experts and sub-national FAO offices to build a
water purification tablets and waterproof bags when flood common understanding of the situation before
risk reaches a predefined level. Other practitioners including implementing activities linked to a forecast.
those that form part of the Start Network, as well as FAO At the other end of the scale, the protocols
and WFP, use a forecast-informed decision-making process used by the Start Network determine how expert
to trigger early action (see Box 3). FEWS NET’s monthly opinion will be used in response to a forecast –
food assistance forecasts have been used by the USAID there is no automation. The Start Network uses
Office of Food for Peace (FFP) in its resource allocation and information from international forecasting centres,
procurement decisions. real-time monitoring by its members, independent
The threshold levels for both types of triggering secondary data analysis and inputs from a technical
systems are typically defined on the basis of experience advisory group to translate an alert into a funded
action plan in less than 72 hours. Based on this
of the hazard and impact – for example how much
information, a context-specific decision is taken
rainfall leads to damage. Some systems use specific event
about whether to trigger an action, what projects
‘return period’ values, such as a one in ten-year drought,
to fund and the amount of funding to disburse.
which provide an immediate and intuitive connection
WFP has tested both automatically triggered
with the frequency with which this might occur in the and non-automated actions through the forecast-
long term (see for example the START Drought Finance based finance window of FoodSECuRE and
Facility). Such information can be especially useful for its El Niño defensive procurement initiative.
planning. Unfortunately, the process and the rationale FoodSECuRE uses predefined climate forecast
for trigger value selection are not readily apparent for triggers to release funding from a central pool to
most systems. In most cases, the danger level and the implement contingency plans. During the 2014–
forecast probability thresholds are defined locally in 2016 El Niños, WFP analysed the impacts of El
conjunction with the system ‘users’, i.e. those who will Niño events on countries where the organisation
make decisions. Ideally this should involve an analysis purchased food and provided assistance, and
of the forecast skill such that the ‘false alarm’ and ‘miss’ used this information as part of its forward
rates are understood, but it is not clear that this is done procurement process to reduce the risks of price
in practice, despite being part of the FbF Manual ‘Menu increases in the supply chain.
of triggers’ process.

15
3  Timing and planning
actions
At the centre of FbA initiatives are efforts to provide second approach typically triggers funding and action
earlier support to at-risk communities. While right after and sometimes during a climate hazard, for
practitioners agree on the importance of early action, example after rains have failed but at the end of the
there is a wide interpretation of what this means and agricultural season when harvests fall short, or before
when it can occur. Three broad approaches are used: the worst impacts of that crop failure have unfolded
months later. The third approach is most often used
•• Before a hazard occurs. Practitioners including the Red in complex environments where multiple factors drive
Cross and WFP use forecasts of climate hazards linked humanitarian crises, and where FbA mechanisms need
to in-depth analysis of the impacts of these hazards to to prevent the humanitarian consequences of multiple
trigger action before and during the onset of climate evolving shocks. Figure 4 provides an illustration
or weather hazards. This means that action can be of the timelines for FbA for cyclones and droughts,
triggered days, weeks or months before the hazard illustrating these different approaches.
occurs. For example, WFP triggered funding for an Increasingly, early action is seen as a series of actions
anticipatory response in September 2015 based on a taken at different times, from an early point where there is
forecast of high drought risk due to El Niño conditions less certainty in a forecast to the point where a disaster is
during the main October to February agricultural about to happen. Especially early on there is a preference
season in Zimbabwe. The main humanitarian impacts for ‘low regrets’ actions, which provide benefits no matter if
of this drought were felt from mid-2016 to early 2017. a disaster actually happens or the forecast turns out to have
•• During and immediately following a climate hazard. been a false alarm. In WFP’s FbA initiatives, for example,
An increasing number of mechanisms use seasonal low-cost low-regret actions, such as checking and servicing
monitoring of climate and agro-climatic information weather gauges or updating and communicating emergency
to detect a shock. This analysis is linked to a forecast plans, are initiated when uncertainty is high; as the
of likely impacts, and used to create a trigger to weather deteriorates, high-cost actions such as evacuation
release finance for early action and early response. The become more acceptable to governments and affected
action may be triggered and implemented after shocks communities. The window for early action is much longer
such as rainfall deficits or changes in temperature for slower-onset events such as droughts, which allows for
have already occurred, but before they have unfolded more activities to be sequenced in the run-up to an event.
into fully fledged disasters. This approach entails As part of its forecast-based early action initiatives in East
often unclear overlaps between early action and early Africa, FAO organises training and awareness-raising
response, especially in the case of slow-onset events activities, scales up existing disaster risk reduction projects
such as droughts, where windows for both are longer. and provides livestock fodder and supplements to protect
Mechanisms such as ARC’s drought risk pooling pastoralist livelihoods. However, longer lead times can
facility and the HSNP are examples of this. also produce greater uncertainty around the correct timing
•• Across multiple time-scales (and for non-climate of interventions before a drought.
hazard-related shocks). Some practitioners use multiple Some preventative or mitigative forecast-based early
sources of early warning information, including actions, such as food distribution, fodder provision or
climate, market and conflict-related information, to cash transfers, can resemble emergency response activities
forecast the impact of a shock or series of shocks. This or shock-responsive social safety nets, but are delivered
approach blends data and uses impact forecasts before earlier, with the aim of allowing the population to take
impacts emerge or become acute. measures to protect themselves and their belongings,
reducing the need for people to use damaging coping
All three approaches link hazards or shocks to strategies, and to support health, education, food and
impacts, but the emphasis is slightly different. The first other expenditures, sustain household food security
approach is most often linked to automated trigger and protect livelihood assets such as breeding livestock
mechanisms that generate action before the hazard throughout a crisis. In its FoodSECuRE programme,
occurs, in order to reduce its impact (see Figure 4). The WFP is considering triggering supplementary nutrition

16
 FbA, early response and late response in the
case of droughts and cyclones
Figure 4  FbA, early response and late response in the case of droughts and cyclones

MONTHS

SEASONAL CLIMATE FORECAST -6

-5

No action Distribute drought

resistant seeds

Prepare -4 SEASONAL CLIMATE FORECAST

beneficiary lists

Tie down roofs

FAILING RAINS -3
Prepare No action
beneficiary lists

-2
CYCLONE WARNING
(72 HOURS)
Prepare Cash/in-kind
No action beneficiary lists distributions
Cash/in-kind Prepare
-1 distributions beneficiary lists
No action
Evacuate

HARVEST FAILS 0
CYCLONE HITS
NEGATIVE COPING STRATEGIES LOSS AND DAMAGE

Cash/in-kind Needs
1 distributions assessments

Cash/in-kind
distributions 2

Needs
assessments

5 Cash/in-kind
distributions

Cash/in-kind
distributions 7

Acute Medium reduced Less Less damage Medium damage Severe damage
malnutrition, food consumption, reduced food reduced loss reduced loss greater loss
slower recovery medium recovery consumption, of life, faster of life, medium of life, slower
faster recovery recovery recovery recovery

Note: the actions above represent a simplified chain of events in disaster preparedness and response, and are not meant to encompass all
The actions above represent a simplified chain of events in disaster preparedness and response, and are not meant to encompass all the actions
the actions that maybebenecessary
that may necessary to reduce
to reduce the impact
the impact of a or
of a drought drought
cyclone. or
Thecyclone. The authors
authors would also like towould
cautionalso like to caution
that although that, although
forecast-based early
action
forecast-based andaction
early early response can response
and early somewhat mitigate the effectsmitigate
can somewhat of a disaster,
the they will of
effects notaeliminate
disaster,thethey
needwill
for further response and
not eliminate the longer-term
need for further
risk reduction.
response and longer-term risk reduction.

17
programming 3–4 months prior to a drought for children Some development and risk reduction programmes
in high-risk areas. Other preventative measures are overlap with FbA initiatives, raising the question
geared towards scaling up risk reduction activities or whether early action activities would happen anyway if
adjusting livelihood practices in the run-up to droughts, no FbA system was in place. But as one key respondent
floods or other extreme weather conditions. Examples explained, although ideally many preventative and
include mapping flood risk areas and repairing dams and mitigative actions should happen anyway, FbA can
irrigation channels. In Kenya, livestock vaccinations are help with decisions about how to best allocate limited
among a set of preventative activities under the National resources, focusing on interventions that are risk-
Drought Management Authority (NDMA)’s Drought informed and in advance of an imminent impact. FbA
Contingency Fund (droughts are often accompanied can help reduce the impact in those areas, where longer-
by animal disease outbreaks). Finally, WFP’s defensive term preventative and mitigative measures have not yet
procurement is intended to increase the efficiency and been carried out. It can also support better management
reduce the costs of post-shock response through the of the ‘residual risks’ that remain despite DRR efforts, or
prepositioning of foodstuffs in the supply chain, ready for when DRR is not the most cost-effective option. The aim
distribution in case additional food and nutrition support is not for FbA to replace longer-term DRR programmes
is required after a drought has developed. but rather complement them and fill gaps as needed.

18
4  Financing FbA
All forecast-based early action mechanisms recognise uncertainty than the parent fund but based on a robust
the importance of being able to deploy funding and decision-making process. Scaling up social protection
other resources in a predictable and reliable way. FbA based on an early warning has been trialled with the
programmes have applied a variety of financing tools, Kenya Hunger Safety Net Programme. This involves
including dedicated funds, specific windows in emergency accessing a separate funding source, rather than funds
response funds, insurance or contingent finance and allocated for emergencies. The pilot was funded by the
direct links to regular resource allocation processes. UK Department for International Development (DFID),
Acting on the basis of forecasts does not necessarily with the expectation that, in the long term, the Kenyan
require new funding, but it certainly calls for the government will contribute financial support. The concept
more rational use of existing funds. In middle-income is gaining in popularity, and similar mechanisms are being
countries funding is usually available for preparedness, established across Africa and elsewhere to quickly scale
but is often not allocated consistently or quickly enough up public social protection systems based on forecasts or
to reduce disaster losses. Although many FbA initiatives early warnings. The Mongolian government has recently
do set aside resources to finance pre-determined actions, set aside 1.5% of all national government budgets to
infrastructure needs to be in place to deliver support, reducing disaster risk. Some of this will go to early action,
whether cash transfers or other types of assistance. for instance through the prepositioning of emergency
In terms of where the funding comes from, there hay reserves for herders, triggered by a seasonal climate
appears to be growing interest in combining different forecast. The International Federation of Red Cross
sources, triggered at different times and used to fund and Red Crescent Societies has established a Forecast-
different kinds of measures, from communicating based Financing Window as part of the Disaster Relief
information early on, when uncertainty is high, through Emergency Fund (DREF), to be used for forecast-based
to higher-cost activities as uncertainty is reduced. All action, and FAO has created an Early Action window
of this requires significant planning, clarification of within its SFERA emergency fund to support trigger-based
responsibilities, coordination and costing of anticipatory mitigation and prevention activities.
actions. As initiatives are scaled up, it will become
clearer to what extent FbA can help streamline, co-
benefit and increase the effectiveness of disaster risk 4.2  Insurance and contingent finance
management by donors, governments and humanitarian
and development organisations. Another option for FbA involves market-based
mechanisms (i.e. not public funds) such as insurance or
contingent finance. These could pay out before disaster
4.1  Dedicated FbA funds and funding impacts have resulted in losses for individual policy-
windows holders or businesses, or have led to credit defaults
for cooperatives, enterprises, banks or microfinance
Dedicated forecast-based and contingency funds have institutions (MFIs). There is as yet limited experience
been established under some initiatives. These can be with integrating hazard forecast-based triggers into such
stand-alone funds or part of emergency funds set aside risk financing and transfer mechanisms. One exception
specifically for FbA. The German Federal Foreign Office was an El Niño insurance product tested in Peru between
has established a fund that can be accessed by its partners 2012 and 2014, which released funds to MFIs and
using a standard operating procedure that includes a cooperatives based on forecasts, encouraging them to
forecast-based finance trigger. WFP’s FoodSECuRE take anticipatory action and to increase lending before,
programme includes a pilot fund with forecast-based during and after a disaster, when people were in need of
triggers linked to contingency plans. The risk financing additional resources to prepare and respond.
mechanism of Ethiopia’s Productive Safety Net Programme Other insurance and contingent finance initiatives that
allows the programme to be scaled up based on early are already using triggers to support early post-disaster
warning information evaluated by a national committee. response could be expanded to include more preventative
Windows in existing wider funds have been established and mitigative actions in the future. The African and
that can be triggered by forecast-based triggers and Asian Resilience in Disaster Insurance Scheme (ARDIS),
processes. One example is the Start Fund’s Anticipation launched in 2018 by VisionFund International, Global
Window, which functions with higher levels of Parametrics and the InsuResilience Investment fund, will

19
 institutions with access to credit and insurance-type
payouts. Both financing instruments are linked to either
a drought or a windstorm index that differs between
countries. Disaster risk financing initiatives targeted
at governments and NGOs, such as ARC’s sovereign
drought insurance and ARC Replica drought coverage,
might represent a further opportunity to expand into
even earlier action.

4.3  Standard resource allocation

processes
Forecast-informed decision-making processes are
often also resource allocation processes. For instance,
WFP’s El Niño Defensive Procurement exercise linked
forecasts of El Niño impacts on WFP’s supply chain to
procurement actions designed to limit those impacts. In
another example, local governments in the Philippines
are increasingly using a specific allocation for disaster
Scaling up social protection based on an early warning has been trialled with risk reduction, the Local Disaster Risk Reduction and
the Kenya Hunger Safety Net Programme. © Marisol Grandon/Department for
International Development Management Fund, before a disaster strikes. Funding is
at least 10% of the internal revenue allocation of each
local government unit, with 70% for DRR and 30% for
be implemented in Cambodia, Kenya, Malawi, Mali, rapid response. Local governments are also using parts of
Myanmar and Zambia in its first year. It will support this allocation for early action, and policy-level advocacy
post-disaster recovery lending by providing microfinance is ongoing to enable this shift.

20
5  Mechanisms for
delivering FbA
Forecast-based initiatives are being developed at many (as described in Section 4). These financial mechanisms
levels, and are being linked to a range of delivery are linked to planning tools known variously as Early
mechanisms. Expanding the use of FbA will require Action Protocols (EAPs), Standard Operating Procedures
continued technical innovation and financing, but perhaps (SOPs), early action plans and contingency plans. In the
no other factor is as important as the channels through case of the Red Cross societies, EAPs define ‘who takes
which forecast-based action is delivered. Initiatives that what action when, where, and with what funds’ (Cruz
use forecasts to trigger and deliver early action have Roja Peruana, German Red Cross and Red Cross Red
either provided support directly to communities (often in Crescent Climate Centre, 2016). These EAPs, which
coordination with local governments), or have worked are separate from government emergency plans, build
with national governments and partners to strengthen on existing Red Cross operational and programmatic
the development and delivery of FbA through state capacities. Despite using separate delivery channels, FbA
institutions. Whether or not actions triggered by forecasts initiatives are often implemented in close collaboration
are part of government early warning and action systems with national governments. Cash transfers using country-
or peripheral to them will depend on the capacity and wide public or private delivery systems are also growing
coordination of government actors, on the country in popularity, although there is a recognition that cash
context and on the mandates of agencies promoting FbA. is not always appropriate. For instance, if markets
Actions taken on the basis of forecasts can be carried out are unable to meet rising demand for goods before an
through existing delivery channels, such as public social event, in-kind distributions may be more suitable. The
protection and safety net programmes, or via stand-alone Bangladesh Red Crescent has distributed cash transfers
FbA mechanisms (Costella et al., 2017). based on a flood forecast; triggered by Dzud forecasting,
the Mongolia Red Cross has organised cash deliveries
to beneficiaries through local bank branches, and the
5.1  FbA linked to community START Network has disbursed cash based on food
development programmes insecurity forecasts in Somalia.

Forecast-based systems have been integrated into

community-based disaster risk reduction and development 5.2  FbA with governments and through
programmes, as well as linked with country-wide social protection and safety nets
contingency planning and response by humanitarian
agencies. For example, forecast-based financing has been Several countries have developed nationally led or
used by the Togolese Red Cross to preposition and then government-implemented FbA mechanisms. WFP is
distribute relief supplies as the risk of flooding increases. supporting government agencies to develop SOPs
World Vision builds on its longer-term community- with pre-agreed triggers and actions in five countries:
development programmes, which facilitate bottom-up Bangladesh, the Dominican Republic, Haiti, Nepal
generation of early warning information and early action. and the Philippines. In Zimbabwe, the Ministry of
Similarly, Christian Aid partners, for example through the Agriculture and WFP helped farmers switch to drought-
BRACED project, tend to use a participatory vulnerability tolerant crops when the risk of drought was high. At the
and capacity assessment approach. In combination with national level, forecast-based action has been integrated
improved access to climate information services, this aims into safety net programmes in Ethiopia, Kenya and
to increase coping capacity through community resilience Uganda, among others.
planning and motivate early action when a shock is forecast. Social protection systems can respond to shocks
FbA initiatives have occurred primarily through the through vertical or horizontal expansion, piggybacking
integration of finance and planning processes. Many on pre-established programmes, aligning social protection
organisations have their own financial mechanisms and and humanitarian systems or refocusing existing resources
dedicated funds to spend on early action, and deliver (O’Brien et al., 2018). By linking to established early
directly to high-risk communities in specific countries warning systems, regular cash transfers and other forms

21
of safety net can help reduce the impact of climate and responsibilities, taking into account the capacities
extremes on poor families. Additional features entail and resources of the actors involved.
climate- and disaster-sensitive targeting and planning; These delivery channels and governance processes
central coordination and registries for targeting and do not need to be separate. The Red Cross promotes
verifying disbursements; partnerships between public, government leadership in FbF, so that it can coordinate
private and non-state actors; pooling funds and smoothing and delegate responsibilities to the Red Cross and other
expenditures; financing and flexible systems for scaling partners. In practice, this could mean the Red Cross
up when a shock occurs; the provision of timely and carries out portions of government EAPs with government
predictable benefits; and interventions to support funding, or coordinates with the government to activate
livelihoods, all of which can increase the effectiveness of Red Cross EAPs using Red Cross funding. Coordination
social protection systems (Castello et al., 2017; O’Brien et between stakeholders operating at different scales is
al., 2018; Bastagli and Holmes, 2014; World Bank, 2013; key to ensuring that the right combination of actions is
Kuriakose et al., 2012). Social protection programmes that taken on time to minimise disaster impacts and reduce
already include shock-response systems, such as the HSNP, risks over the longer term. There are challenges with this.
are more suited to scaling up or integrating FbA. Existing FbA initiatives have been co-developed to varying
degrees with governments, NGOs, UN agencies and other
partners, and in some cases these initiatives have not been
5.3  International humanitarian response coordinated with government contingency plans and early
warning systems. Limited capacity or political will, lack
Finally, FbA has been integrated into international of data and weak planning systems and structures into
humanitarian response mechanisms. Within the UN which forecasts are to be integrated were all highlighted
system, the IASC has supported the development of as challenges by interviewees and in the literature on FbA.
global SOPs for El Niño for IASC partners and NGOs in While there is a general recognition that government
response to the devastating impacts of the 2015–2016 El ownership is desirable to ensure coordination and
Niño. The SOPs support the coordination of early action sustainability, it is less clear when a full handover from
at international, regional and country level, defining roles humanitarian organisations to governments is possible.

22
6  The evidence base for
forecast-based early action
Investment in FbA is expected to bring about several The costs and benefits of anticipatory action will
positive outcomes, and significant attention has been differ depending on whether the event is slow- or
paid to measuring the costs and benefits involved. In rapid-onset, and the degree of fragility/conflict. For
particular, studies have looked at outcomes related to: example, early action for a rapid-onset event can mean
the difference between life and death, and therefore in
•• an earlier response and reduced response time, so that this regard the benefits of early action can be obvious
aid gets to people faster, averting suffering and helping and have a high value. Slow-onset events give ample
to prevent more severe impacts; opportunity to respond months earlier, and hence offer
•• a decrease in the cost of humanitarian response through numerous benefits, although these opportunities are
greater prepositioning and early procurement; and often missed due to higher uncertainties associated with
•• better-quality programme design through pre-planning the forecasts. The costs and benefits of early action in
with more preventative measures, and potential co- fragile contexts can be hard to measure, as early action
benefits in non-crisis times. can be hampered by issues outside of a humanitarian
agency’s control, such as access to affected populations.
These outcomes would suggest that investing early through However, according to the Global Humanitarian
FbA is more cost-effective than waiting to provide a late Assistance Report (Development Initiatives, 2017) the
response. However, it would be wrong to assume that FbA confluence of climate and conflict dominates the majority
would be more cost-effective under all circumstances, and of crises, and therefore it is critical to understand the
there are many possible scenarios where FbA may not be relative costs and benefits of early action. The literature
cost-effective. These are described in greater detail below. is most limited with respect to examining early action in
Evidence on the costs and benefits of anticipatory fragile contexts.
action is very limited, and a meta-analysis of evaluations
of these initiatives was beyond the scope of this
report. Box 4 summarises key studies that have tried 6.1  The costs of anticipatory action
to quantify the costs and benefits of an early response.
Since empirical evidence around the impact of earlier 6.1.1  The cost of acting in vain
responses is scarce, most studies have relied on modelling Responding to a crisis before it has fully materialised
and estimations to assess the impact of alternative brings with it the increased risk of acting in vain. The
approaches. Protocols could usefully build in damage cost can be high. Many interventions that might be
and loss assessment, not only in areas where early action considered as part of an SOP will bring benefits regardless
was taken but also where it was not, in order to compare of whether the crisis occurs or not. From this perspective,
the differential outcomes. However, the benefits of early investment is highly likely to be cost-effective regardless
action can extend well beyond reducing loss and damage. of the accuracy of the forecast. Equally, some early action
By reducing damaging coping strategies, early action activities can have very high costs, for example evacuation
can have long-term effects on malnutrition, education of a large population. It is also important to consider the
and health that cannot easily be captured in a short opportunity cost of any investment, as triggering funding
timeframe. Furthermore, the impacts of crisis are multi- for an extreme event that does not occur will divert funds
dimensional, and teasing out attribution of outcomes away from another crisis, and thus limit the scope for
to specific activities can be difficult. It is therefore early response elsewhere. Nonetheless, it is generally felt
critical that any assessments of the costs and benefits of that a false early response is more than offset by the cost
early action through FbA use a mix of qualitative and of a late response. DFID’s Economics of Early Response
quantitative approaches to the full range of potential and Disaster Resilience study (see Box 4) found that, ‘for
impacts, including less quantifiable effects such as social every early response to a correctly forecast crisis, early
outcomes, as well as investigating the effectiveness of responses could be made 2–6 times to crises that do not
different activities for different hazards. materialise, before the cost of a single late response is met’.

23
Figure 5  Illustration of possible outcomes of forecast- in contexts with chronic and severe poverty. In many
based early action cases, the more expensive investment scenarios may be
offset by the avoided cost of humanitarian response,
but this is not a given and must be carefully assessed in
EXTREME EVENT NO EXTREME EVENT order to determine the most cost-effective package of
interventions in an SOP.

6.2  Other cost–benefit considerations

‘Act in Vain’
‘Worthy Action’
ACTION Action cost + additional 6.2.1  How early is early?
Action cost
costs/benefits In Box 4, the USAID and DFID models assume that ‘early’
means before negative coping strategies are employed.
However, in an assessment of the 2015 drought in Somali
region, Ethiopia, evidence suggested that people started to
be affected a full two years before the crisis was formally
recognised. This example raises the question how early is
‘Fail to Act’ ‘Worthy Inaction’ early enough? It also suggests that FbA would be more
INACTION Response cost No cost effective if longer-term development interventions were
more explicitly focused on building the resilience of
vulnerable populations to address chronic drought risk.
There is however little empirical evidence regarding the
Source: adapted from Suarez and Tall (2010). ‘right’ timing: assessments of the impact of intervening just
before the crisis, as compared to six months beforehand,
for example, are non-existent.
6.1.2  The cost of supplies
As shown in Box 4, early action can yield significant cost
savings through early procurement and prepositioning 6.2.2  Defining triggers
of humanitarian supplies. These savings will most likely In rapid-onset crises defining the trigger is more
be realised in slow-onset or predictable disaster contexts, straightforward than a slow-onset crisis, where the
where there is a good deal of lead time to be able to trigger may be a combination of rainfall forecasts and
respond. Not all items can be procured in advance at other weather-related indicators, as well as changes
a lower price, storing prepositioned supplies can be in prices of food and other commodities, politics, etc.
expensive and certain items will have shelf-lives that How the trigger is defined will affect how accurate the
may require disposal if a disaster does not occur in the forecast is, the likelihood of acting in vain and therefore
relevant timeframe. the overall cost-effectiveness of the response. Crises often
arise as a result of the interplay of changes in rainfall,
6.1.3  Cost of interventions at scale vegetation and prices, as well as other factors. As a result,
Investing in early action and wider resilience-building the cost-effectiveness of a response can be fundamentally
measures can be expensive. While many measures are affected by the composition of indicators used to trigger
‘no/low regret’, the types of actions that may mitigate it, and these will vary according to the context and types
an impending crisis can be costly at scale, for example of livelihoods affected by the disaster.

24
Box 4  Examples of evidence on the costs and benefits of early action
USAID Economics of Resilience (Cabot Venton, 2018). The study evaluated the economics of early response
and resilience across a population of 15 million people in Kenya, Ethiopia and Somalia. The study found that
investment in early response and resilience could have saved $4.3 billion over the previous 15 years, or an
average of $287 million per year. Every dollar spent on safety net and resilience programming results in net
benefits of between $2.30 and $3.30. The US government could have saved $1.6 billion over the last 15 years,
or 30%, on its humanitarian aid spend in these three countries. Incorporating the avoided losses to households,
the model estimates net savings of $4.2 billion to the US.
DFID Economics of Early Response (Cabot Venton et al., 2013). This study quantified the reduction in costs
as a result of procuring goods early in response to humanitarian crises in Kenya, Ethiopia and Niger, and
found that the cost of response decreased by between 11% and 45%. The study also modelled the impact of
commercial destocking and vet services in Kenya and Ethiopia, and found that these measures had a substantial
return on investment (ROI), as well as reducing food deficits by between 9% and 72%.
UNICEF/WFP Return on Investment for Emergency Preparedness (UNICEF/WFP, 2015). This study evaluated
the ROI to emergency preparedness, and found that prepositioning of emergency supplies can yield ROIs of
1.6–2.0, and generate significant time savings in response of between 14 and 21 days on average.
Evaluation of WFP Import Parity Approach (WFP, 2011). WFP’s import parity approach compares local and
international sourcing costs and delivery times for food. An analysis of the price differences between the lowest
and next-best quotes from suppliers for more than one-third of all WFP’s 2010 food procurement expenditure
suggests that the import parity approach led to savings of between 23% and 33% (at least $99 million) in the
cost of commodities.
African Risk Capacity Cost Benefit Analysis (Clarke and Vargas Hill, 2013). A cost–benefit analysis of the
African Risk Capacity facility looked at the impact of late response as compared to an early response. In
this analysis, the main costs to immediate and long-term welfare are assumed to come from reductions
in consumption, losses of productive assets (as a result of direct losses or distress sales) and investment
opportunities foregone. The study estimates late response losses at $1,294 per household, and early response
losses at $49 per household.
Cost Effectiveness of Early Warning (Hallegate, 2012). This study looked at the benefits that would arise
from investment in early warning by looking at the benefits of such systems in developed countries and then
extrapolating these to developing countries. It estimates returns of between $4 and $36 for every $1 spent on
investment in early warning.

25
7  Taking forecast-based
early action to scale
As forecasts continue to improve, the use and geographical 2013). Scaling can also enable more comprehensive action
coverage of FbA mechanisms is growing. Learning from based on forecasts. The examples in Table 2 show how
successes and failures is contributing to the development experiences with scaling up FbA have involved much more
and institutionalisation of the approach. Organisations than simply covering larger populations.
and governments running FbA initiatives, as well as
donors investing in early action, are attempting to scale
up pilots in order to achieve greater impact in preventing 7.2  Embedding FbA in financing and
and dealing with disasters by covering more people, more delivery systems at scale
hazards and more countries. As these initiatives expand,
the community has gained experience and is continually Somewhat different from the examples in Table 2, the
growing the evidence base on effective processes and global ENSO SOPs (developed by the IASC) aim to
impacts. This experience and evidence provide guidance support the coordination of early action at international,
for further FbA development. Scaling up may also help regional and country levels. Although tailored to different
early action become more efficient through sharing country contexts, the SOPs are linked to a global analytical
information and resources and coordinating action. The cell of climatological, humanitarian and development
Red Cross in Peru and Ecuador, for instance, has been experts assessing forecasts, vulnerability and coping
covering heavy rains, volcanic ash, Amazon flooding, capacity to identify countries at high risk of being affected
snowfall and cold waves. Addressing several hazards, by an El Niño/La Niña event. In a first test run in 2016–
rather than just one, has made its presence, and efforts 2017, 19 high-risk countries were alerted and offered
to build and train local teams working on forecast-based support in early action planning and implementation. This
financing, more cost-effective. Scaling up FbA through indicates the potential for implementing FbA at scale to
institutionalisation may strengthen coherence between support coordination and anticipatory action through the
humanitarian and development approaches and between international humanitarian system, and in collaboration
different agencies and governments. Institutional change with national governments.
through FbA is expected to include more robust decision- At country level, operating at scale could mean
making and more effective allocation of aid, as well as coordinating FbA initiatives and stakeholders in contexts
improved long-term financial planning in government and where government capacity and an enabling environment
humanitarian systems. are in place. Close cooperation between humanitarian
and government agencies, as well as collaboration
between different sectors and ministries, were highlighted
7.1  Approaches to scaling up as crucial in the key informant interviews. In Mongolia
and Peru, for example, government agencies coordinate
Humanitarian agencies and governments are beginning forecasting and risk assessments with the Red Cross and
to scale up their FbA initiatives in a number of ways. FAO so that they can also coordinate interventions, with
Most have started by implementing pilots and gradually some actors focusing only on the highest-risk areas of the
expanding the reach and scope of these initial experiences. country, and others including areas that are less likely to
Scaling involves physical expansion (replicating be hit by a disaster.
approaches to new territory and addressing additional
hazards), greater social reach (increasing coverage in
number or scope of people targeted), extending political 7.3  Challenges for taking FbA to scale
engagement and institutional capacity (policy and budget
commitments and mainstreaming FbA within institutions 7.3.1  Scaling up pilot initiatives
by expanding early action to other programmes and Scaling up FbA initiatives from existing pilots can
institutional processes) and deepening the conceptual present challenges to governments and humanitarian
framework around FbA (transforming mindsets, agencies. In some cases, projects have been designed
administrative structures and power relations) (UNDP, to be pilots or catalytic, which means they are not

26
Table 2  Examples of scaling up FbA
Dimensions of scaling up Example
Physical The Togolese Red Cross has distributed non-food items and initiated evacuation plans in several
Replication in new geographic locations or for communities downstream of a hydropower dam before water is released that could cause flooding. It will
additional hazards now expand coverage to all potentially affected villages when notified that floodwaters will be released.

WFP is covering multiple hazards such as flooding and drought in very vulnerable areas that are likely to
need assistance.
Social The FbA system and risk mapping developed by the Mongolia Red Cross cover the entire country, but can
Increasing coverage in number or scope of only provide supplies to a limited number of people. The system will trigger action to support specifically
people targeted those forecasted to have the greatest risk of impact, no matter where they are in Mongolia. Rather than
expanding action to all areas with heightened risk, this means getting smarter about the households being
targeted in relation to their vulnerability.
Political Increasing the number and scope of its FbF pilots, the International Federation of Red Cross and Red
Policy and budget commitments Crescent Societies has established a Forecast-based Financing Window within the Disaster Relief
Emergency Fund (DREF), to be used specifically for forecast-based action.

The Start Network is adopting a layered approach to managing humanitarian financing for NGOs, with
different financial strategies for different scales of risk. For instance, where the Start Fund covers earlier
action for small- to medium-scale events, the Drought Financing Facility is designed to respond to drought
on a five-year return period.
Conceptual To anchor early warning/early action approaches more widely within the organisation, FAO is working with
Transforming mindsets, administrative operational staff to build capacity around early action. As standard technical and operational procedures, for
structures and power relations example for procurement, were not originally designed for early action, processes and mechanisms require
adaptation to match FbA and the timeliness required to make it work. An early warning/early action toolkit
currently under development is aimed at supporting capacity-building and embedding the concept more
widely within and beyond FAO operations.
Comprehensive In Somalia, SomReP is providing unconditional cash transfers, as well as information and advice tailored to
Expanding range of anticipatory actions taken the livelihoods of at-risk people.
to support beneficiaries based on forecasts,
enabling more comprehensive impact Many organisations are expanding their programmes and developing more comprehensive sets of actions.
This includes prepositioning stock, training staff and purchasing supplies, both to support forecast-triggered
distributions, and for post-event response where required.

embedded in a larger institutional context, can be too Greater clarity is required around who triggers action
narrow or are focused on a particular hazard context, for the ENSO SOPs, and how this is communicated to
making them unsuitable for direct replication elsewhere. relevant stakeholders. An Oxfam review of the Somalia
The need for longer-term support, joint programming Early–Warning, Early–Action dashboard in the 2016–17
and sufficient and predictable sources of finance is drought has highlighted the importance of building
another hurdle. There may also be a lack of clarity a common understanding around whether a system
around ownership, duty of care and sustainability once should facilitate early action, timely response or both. In
operating at scale. This may be linked to current M&E this instance, lack of clarity and diverging views of the
systems and the novelty of FbA approaches, which objective complicated implementation, created discontent
means that, while there is a general sense and initial with the mechanism and may make it more difficult to
evidence that early action pays, we do not yet know how scale up the system in the future (Oxfam, 2017).
FbA projects are changing attitudes or behaviours, or Within governments and humanitarian agencies,
the extent to which they are actually reducing disaster the expertise required for effective FbA implementation
impacts and helping build resilience in the longer term. at scale is often limited or absent, responsibilities for
Communication, coordination and timing is also leading within organisations or governments can be
challenging in some cases due to a lack of clarity on unclear and FbA is not a strategic objective for many
timeframes and on the benefits of different early actions, organisations. In Kenya, for example, the NDMA
and difficulties in collectively agreeing triggers and implements early action for drought, whereas flood
actions. As has become clear through the IASC’s ENSO preparedness and response sit with the National Disaster
SOPs and the Start Fund Anticipation Window, it is Operation Centre (NDOC), which does not integrate
essential to have pre-established strategies to deal with FbA. One key informant outlined that, while there is
the uncertainty inherent in FbA systems, to ensure that frequent exchange between the NDMA and the NDOC,
decisions on initiating action are taken early enough. their delivery systems, expertise and mandates are

27
different, and the complexity and shorter time windows media reaction to issuing a false alarm. Investing in FbA
for sudden-onset events such as flash floods further may mean foregoing or delaying other programmes
complicate the replicability of FbA mechanisms for these and investments. The potential to act in vain based
additional types of shocks. on forecasts and the lack of visibility of the benefits
Finally, limited or unclear forecasting skill, gaps in of early action are important political disincentives to
forecasting information and capacity limitations in early fully integrating FbA in international humanitarian
warning systems can restrict their use, reduce confidence financing and national and NGO delivery mechanisms.
in FbA and inhibit scaling up. As one interviewee noted, In addition, promoting inter-ministerial and sectoral
although this is not a ‘deal breaker’ for FbA – and can collaboration around early action is not straightforward.
be overcome through the use of other sources of forecast As one key informant pointed out, information-sharing
information and investment in forecasting capacity between ministries and links between emergency plans
– additional funding will be required to support the and centrally managed protocols are not always in place.
generation and provision of forecast information. Internal politics and competition between ministries over
funds get in the way.
7.3.2  The political economy of using forecasts How forecasts are being used is a key question for
at scale FbA initiatives. Interviewees highlighted that the gains
The political economy of using forecasts systematically from using FbA in a specific context depend on the
and at scale to trigger the release of funds and initiate ‘right’ interplay of risk profiles and hazards affecting a
action prior to a disaster is extremely complex. country and the institutional capacity and political will
Interviewees in this study pointed to a number of to forecast hazards or impacts and finance and deliver
issues, including capacity constraints (to produce and FbA. Haiti, for instance, is highly vulnerable to a range
interpret forecasts), lack of funding up-front (though of natural hazards, but competing priorities and limited
many governments have funds available for disaster capacity in government and meteorological agencies
response) and, critically, loss of political control over the challenge the development of effective FbA. Bangladesh,
allocation of resources. Governments and donors are, which is also frequently affected by natural hazards,
understandably, not keen on spending budgets on early has seen an influx of funding and proposals for FbA
action based on a forecast when levels of uncertainty and preparedness projects. This can be overwhelming
are high; even when uncertainty is low it is difficult for national agencies and demands strong coordination
to commit resources up-front. As one key informant at country level. Nepal was mentioned by one key
pointed out, the challenge remains that, even in high- informant as an example where FbA has made easier
income countries like the US and UK, there is insufficient and quicker advances due to a combination of available
political buy-in or confidence to automatically take resources, capacity improvements and political will.
decisions based on a trigger: decision-making power Strengthening forecasting capacities will also require
remains with technical or political institutions. In other longer-term investments, such as setting up higher
instances there may be a political desire to retain control education and training programmes.
over the parameters used to declare an emergency, Overall, there is a strong desire to institutionalise
limiting or blocking FbA initiatives from the outset. FbA in humanitarian and government risk-financing
In Ethiopia, the experimental forecast-based trigger mechanisms, and the process of developing the protocols
designed for the LEAP programme was not implemented with agreed actions and costs will certainly increase
partly because it removed subjectivity in decision-making confidence in these mechanisms. However, more
and the government’s control over communication of work needs to be done to identify and understand the
early warnings. In Kenya, on the other hand, concerns incentives and interests of all relevant stakeholders if
related more to a potentially hostile press and negative FbA is to become standard practice.

28
8  Conclusion
The field of early action is rapidly advancing and its management, some fundamental principles are needed.
proponents are identifying similar sets of challenges and FbA is gaining traction in countries where governments
modifying the mechanisms to deal with many of these. have some forecasting capacity and their own delivery
Although limitations persist in the forecasting skill and mechanisms for supporting vulnerable households, so
capacities needed to generate and interpret data, establish engaging with government agencies in the design and
triggers and target vulnerable populations, the evidence modification of these mechanisms will be critical. Care
seems clear that early action can reduce disaster losses must be taken, however, to ensure that FbA mechanisms
and has the potential to reduce the humanitarian burden. are free from political manipulation by donors or
More can be done in this regard with existing forecast national governments and retain their function as a
information and impact data. With the increasing risks robust, science-based and effective ex ante mechanism
of more severe and frequent weather events driven by for resource allocation.
climate change, FbA offers an important new tool to FbA requires a clear articulation of roles and
better manage these risks. responsibilities, based on comparative advantage,
As well as reducing loss and damage associated with capacity and access to resources. Careful monitoring of
known hazards and predictable disasters, FbA has the how funds are spent and regular evaluations can improve
potential to improve long-term financial planning in the the effectiveness of FbA and enhance accountability.
humanitarian sector and clarify responsibilities for early FbA should be seen as a more robust decision-making
action. It will only do so, however, if the use of forecasts approach to the allocation of humanitarian or
can be integrated into humanitarian and DRR systems, and contingency funding resources, and although new funds
decision-making protocols are established in advance to have been created in some cases to facilitate early action,
identify concrete actions, roles and responsibilities. To have in the future FbA will have to avoid creating parallel
a significant impact, FbA will need to be adopted at scale, funding systems and planning structures. Existing donor
building on existing delivery channels and strengthening and government funds and other financing mechanisms
these, and draw on a range of financing mechanisms. will need to be expanded to support FbA, linking
For these mechanisms to expand and become a core where possible to existing early warning systems and
component of humanitarian action and disaster risk contingency planning processes.

29
References

Action Against Hunger, ADRA, CARE, COOPI, DRC, OXFAM and World Vision (2014) ‘Becoming early action agencies:
the SomReP community based early action system in Somalia, from conception to practice’ (www.wvi.org/sites/default/
files/SomReP%20EWEA%20System%20October%202014.pdf)
American Meteorological Society (2012) Glossary of meteorology: skill. (http://glossary.ametsoc.org/wiki/Skill)
Bastagli, F. and Holmes, R. (2014) Delivering social protection in the aftermath of a shock: lessons from Bangladesh,
Kenya, Pakistan and Viet Nam. London: ODI (www.odi.org/publications/8468-delivering-social-protection-
aftermath-shock-lessons-bangladesh-kenya-pakistan-and-viet-nam)
Cabot Venton, C., Fitzgibbon, C., Shitarek, T., Coulter, L. and Dooley, O. (2013) The economics of early response and
resilience: lessons from Kenya and Ethiopia. London: DFID (www.gov.uk/government/uploads/system/uploads/
attachment_data/file/67330/Econ-Ear-Rec-Res-Full-Report_20.pdf)
Cabot Venton, C. (2018) The economics of resilience to drought. Washington DC: USAID (www.usaid.gov/sites/default/
files/documents/1867/Kenya_Economics_of_Resilience_Final_Jan_4_2018_-_BRANDED.pdf)
Clarke, D. and Vargas Hill, R. (2013) Cost–benefit analysis of the African Risk Capacity Facility, IFPRI Discussion
Paper 01292. Washington DC: IFPRI (http://ebrary.ifpri.org/cdm/ref/collection/p15738coll2/id/127813)
Coughlan de Perez, E., van den Hurk, B.J.J.M., Van Aalst, M.K., Jongman, B., Klose, T. and Suarez, P. (2015) ‘Forecast-
based financing: an approach for catalyzing humanitarian action based on extreme weather and climate forecasts’,
Natural Hazards and Earth System Sciences 15(4)
Costella, C., Jaime, C., Arrighi, J., Coughlan de Perez, E., Suarez, P. and van Aalst, M. (2017) ‘Scalable and sustainable:
how to build anticipatory capacity into social protection systems’, IDS Bulletin 48(4) (http://bulletin.ids.ac.uk/
idsbo/article/view/2885/ONLINE%20ARTICLE)
Cruz Roja Peruana, German Red Cross and Red Cross Red Crescent Climate Centre (2016) ‘Forecast based Financing
(FbF)’ (http://www.climatecentre.org/downloads/files/BrochureFbF%20-%20Vertical%20-%20Ingles%20
20sep2016.pdf)
Development Initiatives (2017) Global humanitarian assistance report. Bristol: Development Initiatives
(http://devinit.org/wp-content/uploads/2017/06/GHA-Report-2017-Full-report.pdf)
FAO (2018) Saving livelihoods saves lives. Rome: FAO (www.fao.org/3/i8463en/I8463EN.pdf)
Giuffrida, V. (2017) Exploring projected outcomes of the FoodSECuRE Small Grains project: evidence from Zimbabwe.
Rome: World Food Programme (www.wfp.org/content/zimbabwe-exploring-projected-outcomes-food-secure-small-
grains-april-2017)
Hallegatte, S. (2012) ‘A cost effective solution to reduce disaster losses in developing countries and evacuation’, World
Bank Policy Research Working Paper. Washington DC: World Bank
Ibrahim, M. and Kruczkiewicz, A. (2016) ‘Learning from experience: a review of early warning systems’ Columbia
University and World Vision UK (www.wvi.org/sites/default/files/WV_EWEA_Doc_FINAL_Web.pdf)
IPCC (2014) Climate change 2014: impacts, adaptation, and vulnerability. Cambridge: Cambridge University Press.
Knowlton, K. et al. (2014) ‘Development and implementation of South Asia’s first heat-health action plan in Ahmedabad
(Gujarat, India)’, International Journal of Environmental Research and Public Health 11(4) (www.ncbi.nlm.nih.gov/
pmc/articles/PMC4024996/)
Kuriakose, A., Heltberg, R., Wiseman, W., Costella, C., Cipryk, R. and Cornelius, S. (2012) Climate-responsive
social protection, Discussion Paper 1210. Washington DC: World Bank (https://openknowledge.worldbank.org/
handle/10986/13555)
Machenda, T. (2015) ‘WFP helps farmers facing drought in Zimbabwe’ WFP (www.wfp.org/stories/
wfp-helps-farmers-facing-drought-zimbabwe)
O’Brien, C., Scott, Z., Smith, G., Barca, V., Kardan, A., Holmes, R., Watson, C. and Congrave, J. (2018) Shock responsive
social protection systems research. Synthesis report. Oxford: Oxford Policy Management (www.opml.co.uk/sites/
default/files/OPM_Synthesis_Report_Shock_Responsive_SP.pdf)
Oxfam (2017) From early warning to early action in Somalia. What can we learn to support early action to
mitigate humanitarian crisis? Oxfam Discussion Papers (https://policy-practice.oxfam.org.uk/publications/
from-early-warning-to-early-action-in-somalia-what-can-we-learn-to-support-earl-620345)
Public Health England (2014) Heatwave plan for England 2014. (www.gov.uk/government/uploads/system/uploads/
attachment_data/file/429384/Heatwave_Main_Plan_2015.pdf)
Red Cross Red Crescent Climate Centre (n.d.) ‘Forecast-based financing’ of preparedness: developing an operational
protocol. The Hague: Red Cross Red Crescent Climate Centre

30
Red Cross Red Crescent Climate Centre (2016) ‘German-supported forecast-based action in Bangladesh will be all-cash’
(www.climatecentre.org/news/716/german-supported-forecast-based-action-in-bangladesh-will-be-all-cash)
Start Network (2017) Start Fund Crisis Anticipation Window annual report 2017 (https://start-network.app.box.com/s/5
dr4zhh0ng0sze2pdlvjfya4b594f1qf)
Suarez, P. and Tall, A. (2010) ‘Towards forecast-based humanitarian decisions: climate science to get from early
warning to early action’, Humanitarian Futures Programme. London: King’s College
UNDP (2013) Scaling up development programmes, Guidance Note (www.undp.org/content/undp/en/home/librarypage/
poverty-reduction/participatory_localdevelopment/guidance-note--scaling-up-development-programmes.html)
UNICEF/WFP (2015) UNICEF/WFP Return on Investment for Emergency Preparedness study: final report, Boston
Consulting Group (www.unicef.org/publications/files/UNICEF_WFP_Return_on_Investment_for_Emergency_
Preparedness_Study.pdf)
UNISDR (2017) Terminology (www.unisdr.org/we/inform/terminology)
Western Water Assessment (2018) ‘Weather and Climate Forecasts’ (www.colorado.edu/climate/forecasts.html)
World Bank (2013) Building resilience to disaster and climate change through social protection. Washington DC:
World Bank (http://documents.worldbank.org/curated/en/187211468349778714/pdf/796210WP0Build0Box
0377381B00PUBLIC0.pdf)
World Food Programme (2011) ‘Efficiency at WFP’, Executive Board Second Regular Session.
World Food Programme (2016) ‘FoodSECuRE: Food Security Climate Resilience Facility supporting community
resilience-building before and after climatic shocks’, World Food Programme (http://documents.wfp.org/stellent/
groups/public/documents/communications/wfp279583.pdf?_ga=2.68272209.1783856577.1518546169-
1309923270.1510071402)
World Food Programme (2017) ‘Food Security Climate Resilience Facility’, World Food Programme
(http://climateinitiativesplatform.org/index.php/Food_Security_Climate_Resilience_Facility)
World Food Programme (2018) ‘Food Security Climate Resilience Facility (FoodSECuRE)’, World Food Programme
(http://documents.wfp.org/stellent/groups/public/documents/communications/wfp279583.pdf?_ga=2.68272209.
1783856577.1518546169-1309923270.1510071402)
World Meteorological Organization (2015) WMO guidelines on multi-hazard impact-based forecast and warning
services. Geneva: WMO (www.wmo.int/pages/prog/www/DPFS/Meetings/ET-OWFPS_Montreal2016/documents/
WMOGuidelinesonMulti-hazardImpact-basedForecastandWarningServices.pdf)
World Meteorological Organization (2017) Public Weather Services (PWS). (https://public.wmo.int/en/programmes/
public-weather-services-programme)

31
Annexes

Annex 1

Table A1 List of FbA initiatives reviewed

FbA system Countries Space/timescale Hazard type
WFP FoodSECuRE Guatemala, Philippines, Zimbabwe Country-level project but forecast Drought, flood, storms
information provided at district level
WFP Early Warning Unit Bangladesh, Dominican Republic, National but trying to focus on Flood
Haiti, Nepal, Philippines populations that are vulnerable
and exposed
Red Cross Red Crescent FbF Uganda River basin scale; days Flood
Red Cross Red Crescent FbF Togo River basin scale; days Flood
Red Cross Red Crescent FbF Peru Local scale; days–season Heavy rain (and associated flood risk);
cold waves
Red Cross Red Crescent FbF Ecuador Regional – river basin/national Floods/volcanos (ash fall, wind speed)
(volcanos)
Red Cross Red Crescent FbF Bangladesh Regional – river basin/coastal zones Floods, cyclones
Red Cross Red Crescent FbF Solomon Islands National Droughts
Red Cross Red Crescent FbF Mongolia National Dzud
Red Cross Red Crescent FbF Kenya Regional – river basin Floods and drought with new project
at national level
Red Cross Red Crescent FbF Mali Regional – river basin Floods
Red Cross Red Crescent FbF Niger Regional – river basin Floods (droughts to be explored)
Red Cross Red Crescent FbF Zambia Regional – river basin Floods
Red Cross Red Crescent FbF Malawi Regional – river basin Floods
Red Cross Red Crescent FbF Tanzania Local – city level Floods – urban
Red Cross Red Crescent FbF Ethiopia Regional – river basin Floods (drought to be defined)
Red Cross Red Crescent FbF Philippines National – river basin Typhoons/floods
Red Cross Red Crescent FbF Vietnam City level Heatwaves (city)
Red Cross Red Crescent FbF Indonesia River basin Floods (new project at national level –
not yet started)
IFRC FbF window DREF Global (190 National Societies)
Start Network – Collective Zimbabwe Country level; seasonal Various
Crisis Modifier
WFP El Niño Defensive Procurement Global Drought, floods, storms
LEAP – EUPORIAS pilot Ethiopia National system – woreda information Drought
Kenya HSNP and livestock asset Kenya Sub-county level Drought
protection policy

32
FbA system Countries Space/timescale Hazard type
Kenya drought EWS and DCF Kenya County level Drought
Kenya FSSG Kenya County Drought and food security
FAO EW/EA pilots Kenya, Madagascar, Pacific Country level Various
Islands, Paraguay, Sudan; with
Guatemala, Philippines planned
for late 2017
Somalia Resilience Program Somalia Community-based, piloted in 3 regions Drought, floods, conflict,
(SomReP) of Somalia where SomReP partners climate change
are operating
GlobalAgRisk Extreme El Niño Peru District El Niño-related hazards
Insurance Products (EENIP)
Darfur Rain Timeline Sudan Precipitation affecting logistics
Start Fund Crisis Global Country level All hazards
Anticipation Window
Start Network Drought Pakistan, Zimbabwe Country level Drought
Financing Facility
Welthungerhilfe (WHH) Drought Madagascar Drought
Forecast-based Financing
Inter-Agency Standing Global Global–country Multiple: all ENSO-related hazards
Committee (IASC) ENSO
Standard Operating Procedures
African Risk Capacity (ARC) Africa Insurance Drought
ARC Replica Coverage (Start Mali, Mauritania, Senegal Insurance Drought
Network and WFP)
FEWS Food Assistance Outlook Briefing Central America, Central Asia, Multiple Food security
and monthly procurement cycle Sub-Saharan Africa, Haiti, Yemen
Improved Early Warning Early Action Ethiopia Woreda level Multi-hazard
(ACCRA and Oxfam)
Urban Early Action Early Warning Kenya City (Nairobi) Multi-hazard
Start Fund Bangladesh Bangladesh Multi-scale sub-national, time Multi-hazard
depending on hazard
IIED/Christian Aid Kenya County level Drought, food security
Christian Aid/RWAN Philippines Municipality Cyclones, ENSO/drought
Christian Aid BRACED Burkina Faso, Ethiopia District/woreda Drought, food security
Christian Aid/partners Malawi District Drought, flood, food security
Christian Aid/Centro Humboldt Nicaragua National Drought and long-term
climate scenarios
Christian Aid/GEAG India State Drought, flood, food security

33
Annex 2

Table A2 Selected FbA pilots

Somalia Bangladesh Kenya Peru Tajikistan Zimbabwe

Initiative SomReP Red Crescent FbF FAO Early Warning Red Cross FbF Start Network WFP FoodSECuRE
(Phase I) Early Action (Phase I) Anticipation Window
Hazard(s) Drought, floods, Floods, cyclones Drought El Niño, floods, Multi-hazard Drought
conflict, climate snowfall, cold waves
change
Fund Donor pooled EW/ German Federal Early Action window German Federal Start Fund Multilateral
EA fund Foreign Office FbF in the Special Fund Foreign Office FbF Anticipation Window contribution from
fund for Emergency fund Norway
and Rehabilitation
(SFERA)
Information Community-based Weather forecasts Kenyan government Risk analysis World Bank Climate WFP’s forecast
used in early warning of the Bangladesh drought information (DesInventar, Investment Fund, analysis based on
forecast/early indicators Meteorological early warning national statistical Tajikistan country climate models
warning Data on food department and the system (short-range and sectoral information, district- (climatology/
security and Bangladesh Water weather forecasts, agencies, level meteorology precipitation),
livelihoods, health Development Board hydrological data, meteorological stations within Zimbabwe
and nutrition and market and trade offices) and Tajikistan, Weather Meteorological
conflict information, socio- climatological/ Online, for analysis department and
FSNAU/FEWS NET economic indicators, meteorological of number of the Southern
livestock movement forecasts days experiencing Africa Regional
indicators); IRI (meteorological precipitation in Climate Outlook
seasonal rainfall offices, geophysical 2016–17 winter, Forecast (SARCOF),
forecasts; Global institutes, NOAA, IRI, verbal reports from long-term trends in
Agriculture Stress European Centre, district-level staff food insecurity and
Index; FAO’s etc.) for a range of and community on vulnerability through
Predictive Livestock seasonal to daily emergency situation the Zimbabwe
Early Warning forecasts Integrated Context
System (forage Analysis, Zimbabwe
coverage forecast); Vulnerability
Kenyan government Assessment
Long and Short rain Committee
assessments (crop (ZimVAC), WRSI
production, livestock
prices, food security)
Type of trigger Phased approach Short-term hydro- Phased approach Seasonal and short Combination Combination
based on meteorological based on a term climatological of forecasting of forecasting
combination of forecast using combination of and meteorological information and information and
information: normal, threshold method indicators (normal, forecasts using expert analysis of expert analysis of
alert, alarm and providing a lead alert, alarm), each threshold method risk that a crisis risk that a crisis
emergency, with time of 48 hours for corresponding to a (El Niño: 3-month, occurs (qualitative occurs (qualitative
each deterioration cyclones and 7–10 different set of early 1-month, 7-day; information method) information method)
triggering a range days for floods actions. Combination floods: 1-month, 9-
of early actions of quantitative day, 2-day; snowfall
using qualitative (rainfall, hydrological and cold waves:
information method thresholds) as well 5-day)
as qualitative (expert
analysis of where the
drought is likely to hit
hardest)
Implementing ACF, ADRA, CARE, Bangladesh Red FAO Peruvian Red Cross Start Network WFP
agency/ies COOPI, Danish Crescent Society Technical support members
Refugee Council, (BRCS) from German Red (Welthungerhilfe,
Oxfam, World Vision Technical support Cross and Red Mercy Corps and
from German Red Cross Red Crescent ACTED involved in
Cross and Red Climate Centre previous activation)
Cross Red Crescent
Climate Centre

34
 Somalia Bangladesh Kenya Peru Tajikistan Zimbabwe
Funding $777,791 funding Approximate funds $400,000 Approximate funds £145,704 spent $100,000
released to gap approved spent in direct cash for relief goods and
date payments: services activated:
Flood 2016: El Niño 2015–16
€92,000 rains: €240,000
Flood 2017:
€54,000 Cold wave 2016:
Cyclone 2017: €60,000
€124,000

Sources: Start Network (2017); Action Against Hunger, ADRA, CARE, COOPI, DRC, Oxfam and World Vision (2014); Red Cross Red
Crescent Climate Centre (2016); FAO (2018); Machenda (2015); World Food Programme (2016); World Food Programme (2017); World
Food Programme (2018); Giuffrida (2017); Cruz Roja Peruana, German Red Cross and Red Cross Red Crescent Climate Centre (2016);
Ibrahim and Kruczkiewicz (2016).

35
Annex 3

Table A3 List of key informants

Name Affiliation
Davaajargal Batdorj Mongolian Red Cross

Lorenzo Bosi WFP

Mathieu Destrooper German Red Cross
Dunja Dujanovic FAO
Brenden Jongman WB GFDRR
Georgina Jordan World Vision
Michael Kühn Welthungerhilfe
Romain Lare Togolese Red Cross
Jesse Mason WFP
Emily Montier Start Network
Sunya Orre Kenyan National Drought Management Authority
Greg Puley OCHA
Sanna Salmela-Eckstein IFRC
Jerry Skees Global Parametrics

36
Annex 4

Table A4 Glossary of FbA concepts

Forecast-based financing Forecast-based Financing (FbF) is a mechanism first developed by the Red Cross to release humanitarian funding
based on forecast information for planned activities which reduce risks, enhance preparedness and response and
make disaster risk management overall more effective. The Red Cross, meteorological services and communities at
risk agree on selected actions to be taken once a forecast reaches a certain threshold of probability. Each action is
then allocated a budget to be activated when a forecast is received (Red Cross Red Crescent Climate Centre, n.d.).
Forecast-based early action Action taken in the short term after the issuance of a science-based early warning, but before a potential disaster
materialises (Coughlan de Perez et al., 2015). In this study we found multiple interpretations of the word ‘forecast’;
some organisations focused on forecasting climate hazards based on an analysis of their possible impacts, while others
focused on forecasting the impacts themselves, with differing levels of complexity. As a result, there is little consensus
on what counts as forecast-based action and when it is taken. Some adopt a broad interpretation including actions to
reduce vulnerability, training and prepositioning relief (Coughlan de Perez et al., 2015). Others include early response
after a climate hazard has already had an impact, or before multiple shocks and stressors have worsened an existing
humanitarian crisis.
Early response There is often confusion over whether early action refers to action taken ahead of an impending shock to reduce its
impact, based on forecasts/predicted needs – or simply a faster, more timely humanitarian response (Oxfam, 2017).
For clarity, we use early response to refer to the latter.
Late response Late response most often refers to a humanitarian response implemented when severe impacts of a hazard have
already begun to occur. In the case of drought this may be as much as six months after a failed agricultural season. In
cost–benefit studies this is often formalised as a humanitarian response that arrives after negative coping strategies
have been employed and after prices of food and other items have begun to destabilise (e.g. Cabot Venton, 2018).
Impact (humanitarian and disaster) Disaster impact is the total effect, including negative effects (e.g. economic losses) and positive effects (e.g. economic
gains), of a hazardous event or disaster. The term includes economic, human and environmental impacts, and may include
death, injuries, disease and other negative effects on human physical, mental and social wellbeing (UNISDR, 2017).
Impact-based forecasting Forecasting the impact of a hazard, or multi-hazards, on individuals or communities at risk. Examples include
forecasting the possible impact of rainfall on road users during rush hour, or the impact on passengers of closing
an airport due to strong winds. These could be done in a subjective way working alongside transport customers,
or in an objective way through developing an impact model using vulnerability and exposure datasets as well as
meteorological information (World Meteorological Organization, 2015).
Vulnerability The conditions determined by physical, social, economic and environmental factors or processes which increase the
susceptibility of an individual, a community, assets or systems to the impacts of hazards (UNISDR, 2017).
Exposure The situation of people, infrastructure, housing, production capacities and other tangible human assets located in
hazard-prone areas (UNISDR, 2017).
Risk Risk is defined as the probability and magnitude of harm attendant on human beings and their livelihoods and
assets because of their exposure and vulnerability to a hazard. The magnitude of harm may change due to response
actions to either reduce exposure during the course of the event or reduce vulnerability to relevant hazard types in
general (World Meteorological Organization, 2015).
Disaster risk reduction Disaster risk reduction is the concept and practice of reducing disaster risks through systematic efforts to analyse
and reduce the causal factors of disasters. Reducing exposure to hazards, lessening the vulnerability of people and
property, wise management of land and the environment and improving preparedness and early warning for adverse
events are all examples of disaster risk reduction (UNISDR, 2017).
Disaster risk management Disaster risk management is the application of disaster risk reduction policies and strategies to prevent new disaster
risk, reduce existing disaster risk and manage residual risk, contributing to the strengthening of resilience and
reduction of disaster losses (UNISDR, 2017). DRM systems therefore include early action based on hazard forecasts.
Early warning system An integrated system of hazard monitoring, forecasting and prediction, disaster risk assessment, communication
and preparedness activities, systems and processes that enables individuals, communities, governments,
businesses and others to take timely action to reduce disaster risks in advance of hazardous events (UNISDR, 2017).

FbA systems are in many ways a subset of early warning systems focused on better translation of forecasts into
anticipatory action. They also allow for action to be taken based on probabilistic information, and therefore for
responses to be triggered that may not be followed by a disaster event.

37
Contingency (emergency) planning A management process that analyses disaster risks and establishes arrangements in advance to enable timely,
effective and appropriate responses. Contingency planning results in organised and coordinated courses of action with
clearly identified institutional roles and resources, information processes and operational arrangements for specific
actors at times of need. Based on scenarios of possible emergency conditions or hazardous events, contingency
planning allows key actors to envision, anticipate and solve problems that can arise during disasters. Contingency
planning is an important part of overall preparedness. Contingency plans need to be regularly updated and exercised
(UNISDR, 2017). FbA systems all link to some kind of contingency planning, SOP, EAP or decision-making process.
Forecast A forecast is a prediction or estimate of future events, especially coming weather or a financial trend. In this
study, most initiatives focused on climate and weather forecasts. Weather forecasts provide information about
the expected state of the weather up to 10–14 days in advance, while climate forecasts and outlooks provide
information about the expected state of regional climate beyond the timeframe of long-range weather forecasts
(~10–14 days) (Western Water Assessment, 2018).
Forecast skill A statistical evaluation of the accuracy of forecasts or the effectiveness of detection techniques. Forecast skill is
determined by comparison of the disseminated forecast with a reference forecast, such as persistence, climatology
or objective guidance; it shows what ‘value’ the forecast adds to simple schemes (American Meteorological Society,
2012). Forecast accuracy is determined by comparison of the disseminated forecast with actual observations (World
Meteorological Organization, 2017).

38
ODI is the UK’s leading independent
think tank on international development
and humanitarian issues.

Readers are encouraged to reproduce

material for their own publications,
as long as they are not being sold
commercially. As copyright holder, ODI
requests due acknowledgement and
a copy of the publication. For online use,
we ask readers to link to the original
resource on the ODI website. The views
presented in this paper are those of the
author(s) and do not necessarily represent
the views of ODI or our partners.

© Overseas Development Institute 2018.

This work is licensed under a Creative
Commons Attribution-NonCommercial
Licence (CC BY-NC 4.0).

All ODI publications are available from

www.odi.org

Cover photo: Men cover the windows

of a car parts store in preparation for
Hurricane Irma in San Juan, Puerto Rico
©2017 Alvin Baez / Reuters Pictures

Overseas Development Institute

203 Blackfriars Road
London SE1 8NJ

Tel: +44 (0) 20 7922 0300

Fax: +44 (0) 20 7922 0399
Email: info@odi.org.uk

Shaping policy for development odi.org | @ODIdev