ISSN 2047-0371

Modelling Geomorphic Systems: Numerical Modelling

Christopher J. Hutton1
1
Centre for Water Systems, College of Engineering, Mathematics and Physical Sciences,
University of Exeter, UK (c.j.hutton@ex.ac.uk)

Numerical models seek to represent the interaction between landscape forms and processes
through mathematical equations. By integrating these equations over space and time, numerical
models have allowed geomorphologists to extend enquiry beyond observation alone, and explore
landscape dynamics over a range of temporal and spatial scales. Choosing the correct temporal
and spatial scale of investigation, the correct processes that control landscape form at these
scales, and then converting this conceptual model to a mathematical representation of these
process-form interactions is not straightforward. The decision requires careful consideration of
process dominance and scale, the ability of equations to parameterise these processes,
computational resources, and data availability to constrain model parameters and evaluate model
performance. These issues shall be considered in general terms, and illustrated mainly with
reference to catchment systems. Finally, numerical modelling of geomorphic systems is considered
from a Bayesian perspective to provide a conceptual grounding for the development and
application of numerical models, and therefore for their role in geomorphic enquiry.
KEYWORDS: Numerical modelling; Scale; Resolution; Evaluation; Data; Uncertainty.

Introduction
The interaction of landscape form and
process may be represented mathematically
in the form of a numerical model. Coupling
such models with observations provides a
formal framework to assemble scientific
understanding, and a powerful tool to
investigate landscape change. Numerical
models provide some liberation from the
temporal and spatial shackles imposed by
enquiry through observation alone; models
have allowed exploration of landscape
processes and evolution over spatial scales
ranging from particles to plate tectonics and
temporal scales ranging from milliseconds to
millennia (Figure 1; Bishop, 2007; Hardy,
2005). Numerical models have shown
potential as powerful tools for understanding
reductionist process-form interactions (e.g. Figure 1. Contrasting scales of model
Schmeeckle and Nelson, 2003; Wainwright et application: grain scale predictions of particle
al., 2008a), and also the relative importance paths (black lines) over water worked gravel
of autogenic versus allogenic controls for (modified from Hardy, 2005); convergent
larger scale system behaviour (Coulthard et orogen formation modelling, considering
al., 2005; Nicholas and Quine, 2007b; techtonic uplift and surface erosion (modified
Wainwright and Parsons, 2002). from Willett and Brandon, 2002).

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

 Numerical Modelling 2
In order to fully exploit the potential for driving conditions D (e.g. rainfall), a vector of
numerical models to elicit understanding of initial system states X0 (e.g. landscape
form-process interactions, and inform elevation), and vectors (with length t, the
landscape management, numerical models length of the simulation) representing future
must be developed and applied carefully system states X, and outputs Y (e.g.
considering three key questions: catchment runoff/ sediment flux):

1. What are the relevant form-process

interactions at the scale of enquiry?
2. What are the correct mathematical Geomorphic models are generally concerned
representations of these processes? with the action of a number of processes
which locally transport mass (e.g. sediment,
3. Are there appropriate data to constrain
including organics and nutrients) that lead to
model parameters and evaluate model changes in landscape form (X) at a specific
predictions? point over time (t). Though models generally
differ in the processes evoked to move
The main modelling issues that need to be sediment, a fundamental approach governing
considered in order to address these most geomorphic models is to divide the
questions are presented here to provide a landscape into units called control volumes
basis for more domain specific sections in (in 1, 2 or 3 dimensions). In most models a
Chapter 5, and illustrated mainly with quasi-2D conservation of mass is applied by
reference to catchment systems. The calculating the change in elevation in
strength of the assumptions made in response to sediment flux into and out of a
developing and applying a model will control, where dx indicates the size of the
determine the validity of model predictions, control in one dimension. The control may be
the strength of conclusions derived from divided into three stores, and M partly
model application, and therefore the ability of specified by (Figure 2; Tucker and Hancock,
models to inform us of real world process- 2010; Wainwright et al., 2008a):
form phenomena. A Bayesian approach
emphasising iterative dialogue between
model development and data collection is
recommended as a robust means to
appropriately develop numerical models, and
therefore geomorphic understanding.

Publications reviewing specific areas of

geomorphic modelling (e.g. Bishop, 2007;
Coulthard, 2001; Livingstone et al., 2007;
Merritt et al., 2003; Morgan and Nearing,
2010; Pelletier, 2011; Reinhardt et al., 2010;
Tucker and Hancock, 2010; Van de Wiel et
where Hr is depth of bedrock (m), Tu is
al., 2011; Wainwright et al., 2008a), and
tectonic uplift (m); Sc (ms-1) represents the
publications expanding on the more general
rate of conversion of rock to soil/surface
issues of model application to natural
regolith, Hs (m); d is sediment deposition rate
systems considered here (e.g. Beven, 2002;
(ms-1) and ɛ is sediment entrainment rate
Bloschl and Sivapalan, 1995; Brazier et al.,
(ms-1) from and into the equivalent depth of
2011; Church, 1996; Krueger et al., 2009;
sediment in transport Ht (m), and qs (m2s-1) is
Nicholas, 2005; Refsgaard et al., 2006; Van
sediment discharge across the surface, dx.
de Wiel et al., 2011; Wainwright and
Up to specifying the source terms, accounting
Mulligan, 2004; Wilcock and Iverson, 2003)
for density/particle size differences between
are additionally recommended.
the stores, and developing an appropriate
numerical solution, Equations 2-4 can
Model structure generally be used to simulate the evolution of
A model M contains equations with any point in the landscape, though some
associated parameters that represent the specific exceptions apply (Tucker and
functional relationship between a vector of Hancock, 2010).

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

3 Christopher Hutton
In catchment systems over millennia climate-
qs induced fluctuations in sediment transport
and rock breakdown, alongside tectonic uplift,

Hill slope and Channel evolution

Soil Erosion and Sediment Transport

Tectonic-Climate Driven evolution
govern the evolution of plate tectonics.
Ht Therefore, changes in are important and
need to be included in the conceptual model
d ɛ of landscape change (Bishop, 2007). Over
decadal and centennial timescales,
catchments predominantly evolve in
Hs response to climate-induced fluctuations in
sediment transport, and therefore the stores
of sediment in the landscape evolve through
Sc transport between control volumes, and
may be considered fixed and the processes
that control them (Tu and Sc) relaxed. At this
scale the state vector (X) needs to consider
Hr not only the evolution of sediment and mass,
but also controlling and interacting factors
such as vegetation (Istanbulluoglu and Bras,
2005; Reinhardt et al., 2010) and potentially
Tu Fixed anthropogenic influence (Wainwright and
Datum Millington, 2010). Alongside exerting control
on sediment flux directly, these factors will
also respond independently to climatic
changes, creating potentially complex, and
non-linear landscape feedbacks (Corenblit
Figure 2. Key fluxes governing sediment and Steiger, 2009). At even smaller temporal
movement within a one-dimensional control scales, sediment flux is controlled by current
volume. The temporal scale of model weather conditions and the effect of previous
application determines the relevant equations events operating at all scales (Schumm and
and methods of parameterisation. Lichty, 1965), which manifest their effects
through the model initial conditions.
To completely specify M, first a conceptual Therefore many system states are fixed (e.g.
understanding of the relevant source terms Hr and vegetation cover) and only need to be
on the right hand side of equations 2-4 and specified in the initial conditions (X0), with no
the processes that control them are required additional equations required to simulate their
at the scale of enquiry. Second, the evolution.
conceptual model needs to be codified into a Similarly, over different spatial scales
set of equations, and an appropriate different processes will become important in
analytical/numerical solution sought. As with controlling landscape behaviour. At small
all geomorphic enquiry, an understanding of spatial scales on hill slopes instantaneous
scale underpins the specification of these fluxes of water (raindrops and overland flow)
source terms, and the answers to the control grain scale movements of sediment
questions posed in the introduction. (Brazier et al., 2011). As catchment size
increases overland flow concentrates to form
Process dominance and scale rills, gullies and channels which, alongside
mass movements, are increasingly important
The processes that govern changes in Hr in controlling catchment sediment flux
operate over larger timescales (e.g. are (Nichols, 2006). Therefore as the spatial and
relatively slower) than the processes temporal scales of interest reduce, the range
governing changes in Hs which in turn are of processes that must be considered also
slower than processes governing changes in reduces, and other, larger scale processes
Ht. Therefore as the timescale over which a are manifest through the model boundary
model needs to be applied tends to zero, so conditions (X0).
the number of relevant stores and processes
that control also reduce (Figure 2). Developing a model to address a specific
problem therefore requires a sound

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

 Numerical Modelling 4
conceptual understanding of the time and explicitly calculate an entrainment and
space scales over which processes operate deposition flux into, and out of transport Ht,
to control the landforms in question. The (Figure 3; Hairsine and Rose, 1992;
extent to which a process is represented in a Wainwright et al., 2008a). Developing better
model, however, depends on its parameterisations of equation 4 have been
mathematical formulation. limited by the difficulty of measuring sediment
in transport. Therefore, even at scales where
monitoring can take place to parameterise
Process representation system processes, competing process
Experimental work has been conducted, both representations may be derived reflecting
in the laboratory and in the field, to process uncertainty, and also different
investigate surface process, such as rain- experimental setups (Wainwright et al.,
splash and overland flow driven erosion 2000).
(Furbish et al., 2009; Wainwright et al., 2000).
Alongside fundamental, physical equations
(e.g. Navier-Stokes equations of fluid
motion), many such studies have provided us
with predictive equations from which process-
form interactions can be simulated (Tucker
and Hancock, 2010; Wilcock and Crowe,
2003). When combined with conservation of
mass equations and integrated over time,
such models represent the fundamental
mechanisms by which climatic fluctuations
manifest in catchment-scale landscape
evolution.
Given the temporal and spatial constraints on
observation, the majority of experimental
work has attempted to parameterise the
processes in equation 4, through what may
be determined process-based models
(Wainwright et al., 2008a; Wilcock and
Crowe, 2003). However, even at small spatial
and temporal (reductionist/observational)
scales, different process parameterisations
have been developed.
In catchment systems erosion, transport and
deposition of sediment by water is controlled
by both transport limited processes (TL; e.g.
presence and power/stress imparted by water Figure 3. Spatial pattern of at-a-point total
at the surface) and supply/detachment limited sediment movement (kg) predicted by
processes (SL; e.g. the resistive forces at the MAHLERAN soil erosion model when applied
sediment bed that impede sediment to an 18x35m runoff plot (Wainwright et
movement). The most widely applied al.2008b).
predictive equations have calculated a
sediment transport rate (qs) as a function of
TL and/or SL, which implicitly assume Specifying the source terms in equation 2-4
sediment transport is in equilibrium (and may also require separate models that
potentially at some capacity), and evolve Hs simulate the behaviour of phenomena that
according to sediment flux into and out of the control landscape change. In catchment
cell support (Wainwright et al., 2008a; systems transport limiting factors controlling
Wilcock and Crowe, 2003). Though such sediment transport are primarily derived from
methods may provide useful predictions, free surface flow. The fundamental equations
changes in sediment transport and soil depth that simulate free surface flow are the 3D
are inherently in disequilibrium. Alternative Navier-Stokes equations (Lane, 1998).
parameterisations have been developed that Although it has been argued that flow

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

5 Christopher Hutton
sediment interactions should be modelled movement, given the potential for non-linear
using these equations (Hardy, 2008), various error propagation.
simplifications are typically made because of
The increased availability of high resolution
computational limitations involved in deriving
topographic data has facilitated the
accurate numerical solutions – a class of
application of small scale model
models often termed “reduced complexity”
parameterisations over increasingly large
models.
domains. For computational reasons and
In the case of hydraulic/hydrologic modelling, availability of other distributed data, however,
such simplifications include depth averaging it is often necessary to model at coarser
to two dimensions (Lane et al., 1999) and the resolutions in both space and time. One
fully dynamic 1D St Venant equations with approach to deal with the problem of process
further simplifications thereof that neglect parameterisation at coarser scales is simply
potentially unimportant terms to derive the to apply the same equations developed at
diffusive- and kinematic-wave models (Hunter smaller scales. However, a key problem with
et al., 2007; Tucker and Hancock, 2010). this approach is that many if not all
Additional flow simplifications have led to a geomorphic laws are scale dependent. For
number of cellular approaches for flow example many geomorphic laws are slope
routing, which employ simplified rules to route dependent, thus increasing model cell size
flow in river channels and hillslopes (Favis- reduces slope, and can lead to inaccurate
Mortlock, 1998; Nicholas, 2009; Thomas and predictions of erosion (Kalin et al., 2003).
Nicholas, 2002). Furthermore, some Such inaccuracies occur because of how
approaches have sought to employ simpler changing scale of resolution affects both
cellular and physically based rules to geomorphic laws and the laws governing flow
simulate morphological change, and have (Brazier et al., 2011). Inferring the validity of a
simulated sand dune formation (Figure 4; model equation independently of the grid
Nield and Baas, 2008), meander migration within which it is applied may be difficult
(Coulthard and Van De Wiel, 2006), braided (Nicholas, 2005). Thus in model development
river evolution (Thomas et al., 2007), and changing model scale will affect the validity of
floodplain evolution (Figure 5; Karssenberg all processes included in the model.
and Bridge, 2008).
Although all spatial parameterisation is
lumped to some degree, critical scales in the
landscape that govern larger scale behaviour
(e.g. the scale of interest) should be
considered. For example, on hillslopes and in
channels coarser scale models will fail to
account for the spatial heterogeneity of flow.
Given the relationship between sediment
transport and flow is strong and non-linear,
neglecting this heterogeneity will lead to an
under-prediction of erosion and sediment
transport (Ferguson, 2003; Nicholas, 2000).
Figure 4. Nebkha dunes and vegetation (dark Relying on equilibrium concepts to model
bars) simulated with a cellular approach sub-grid scale channel features is a popular
(Modified from Nield and Baas, 2008). approach to deal with this problem in
landscape evolution models, but relies on
equilibrium concepts to model potentially
The specific conditions under which such non-equilibrium behaviour (Nicholas and
equations are ‘valid’ should be carefully Quine, 2007a; Tucker and Hancock, 2010).
considered (see Lane (1998) and Cao and Therefore the specific scale at which a model
Carling (2002) for a consideration of hydraulic equation is a valid representation of sub grid-
equations in a geomorphic context). The scale processes is an important consideration
scale and method by which transport limited when developing a numerical model.
conditions are modelled may be more
important than the equations that link At larger temporal scales of enquiry models
properties of flow to actual sediment have to deal with the disparity between the
timescales of individual events (e.g. rainfall
runoff) and the evolution of landscapes

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

 Numerical Modelling 6
(Tucker and Hancock, 2010). As a result of uncertain to what extent fine scale processes
computational limitations, many landscape need to be resolved explicitly in geomorphic
evolution models applied over larger temporal models, or whether simpler treatments,
scales relate at-a-point discharge to upslope relying on for example regime theory, are
contributing area, which implies runoff is in applicable (Nicholas and Quine, 2007).
equilibrium with uniform rainfall. Such Choosing a specific process representation
approximations subsequently used to therefore reflects a specific modelling
simulate sediment transport fail to account for hypothesis regarding the relevant processes
spatial and temporal variability. Sub-scale governing a different problem. Multiple
events may be important in controlling runoff representations, and therefore hypotheses of
production and therefore the sediment the same processes may require
transport that actually governs landscape investigation (Krueger et al., 2009). To help
behaviour. Simple averaging to an effective overcome this issue, data are required to
event may miss that some events are more constrain model parameters and evaluate
important in controlling erosion and sediment model hypothesis (Kleinhans et al., 2012).
transport than others (Nichols, 2006).
Furthermore, modification of the landscape
by continued operation of smaller events may Data and model evaluation
be (more) important in controlling Data availability is an essential factor
morphological change (Goodrich et al., 2008; governing model development, as data
Sambrook-Smith et al., 2010). provides the modeller with the ability to
The issue of what scales to consider and constrain model parameters and evaluate the
therefore what processes to resolve explicitly quality of model predictions. The evaluation
in a given model structure points towards a of model process representation is an
fundamental issue for geomorphologists: essential step as it often occurs prior to
given natural systems often display non- application of models to investigate so called
linear, threshold responses, it is uncertain - “what if” questions (Michaelides and Wilson,
and debated in the literature - to what extent 2007; Nicholas and Quine, 2010). Such
small scale processes (in both space and model application is often at space and time
time) control larger scale system behaviour scales over which data are insufficient to
(Lane and Richards, 1997). It is therefore differentiate between competing model

Figure 5. Simulated channel belt and floodplain evolution (e.g. by bifurcation, avulsion and
aggradation) in response to base level rise (Karssenberg and Bridge, 2008).
British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)
7 Christopher Hutton
hypotheses (Pelletier, 2011). and also to results derived from physical
model experiments (Nicholas et al., 2009).
In the ideal case, all process
parameterisations of equations 2-4 will be A number of implicit assumptions made in
known and data will be available at the scale model calibration may potentially undermine
of the model cell size to constrain model model application; First, it is often assumed
parameters and the initial model conditions. that parameter uncertainty is the only form of
For example in soil erosion modelling modelling uncertainty; Second, that the
distributed information on particle size may model is equal to reality; Third, the initial
be required to parameterise both roughness states are the true initial states (e.g. the DEM
for overland flow modelling and the supply is error-free); and Fourth, that the
limiting factors controlling sediment input/driving conditions and output data to
entrainment (Wainwright et al., 2008b). evaluate model performance are true. In
However, often such data are unavailable, or most, if not all situations these conditions do
inconsistent with the scale of model not hold due to the problems of modelling an
application (Brazier et al., 2011). open system, where true model validation
and verification is impossible (Oreskes et al.,
1994).
As a result of these assumptions incorrect
model parameters can be identified that
reproduce catchment outlet conditions (i.e.
larger scale measurements) with insufficient
consideration of how well they reproduce the
internal spatial patterns of process-form
interaction that ultimately control larger (and
longer) scale response. Furthermore, model
parameters may be identified that are highly
unique to specific settings (Nearing, 1999;
Nearing, 2000), and inapplicable elsewhere
because of the non-linear open nature of
natural systems. Worse still, a number of
parameter combinations within a specific
model may provide equally good predictions -
a form of model equifinality (Brazier et al.,
2000). Similarly, another form of model
equifinality may occur if the data are
Figure 6. Comparison of measured (a) and
insufficient to differentiate between
modelled (b) elevation change in the braided
competing models. Equifinality has arisen at
Avoca River, New Zealand (Nicholas and
a range of scales, from using metrics of
Quine, 2007a).
landscape form to differentiate between
transport and supply limited models of
landscape evolution (Pelletier, 2011),
As a consequence, models may be calibrated
evaluating alluvial fan evolution (Nicholas and
by comparing model outputs to observations.
Quine, 2010), and at smaller scales when
In the case of catchment modelling, models
applying models with complex, and ill
are typically calibrated by adjusting internal
constrained parameters (Brazier et al., 2000).
model parameters to derive the best fit
However, equifinality is not all bad if it avoids
between model outputs (e.g. sediment/water
over confidence in the information content of
flux at a catchment outlet) and the equivalent
data and therefore the potential rejection of
observations at specific locations (Canfield
good model structures. In the face of
and Goodrich, 2006; Nearing, 2000). In
equifinality, simpler models may be preferred
morphological modelling, distributed model
that are justified by the data: A model is only
predictions may be compared to:
as good as the data available to constrain
observations or morphological change
model structure, parameters, and therefore
(Figure 6; Nicholas and Quine, 2007a); to
predictions.
results derived from models with a stronger
physical basis for prediction (Nicholas, 2009);

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

 Numerical Modelling 8
Accounting for uncertainty when In many applications only a single model
structure is considered, and associated
developing numerical models parameters are either derived from previous
studies or are optimised to the specific data
The preceding sections have discussed the available. As a result the final two terms in
main issues to consider in developing and Equation 5 collapse to a single set of
applying a numerical model to address a structural assumptions. In such cases
geomorphic problem. Even if we consider our overconfidence in the data for the reasons
conceptual model of the system and the discussed above – both in its accuracy and
dominant processes to be accurate, its general applicability to a wide range of
computational resources and data availability settings – may lead to inappropriate model
will limit our ability to apply the preferred rejection and narrowing of the posterior
model at the desired scale. Furthermore, probability of all possible models. This may
uncertainty surrounding the relationship lead to an entrenchment of modelling
between process dominance and scale is a concepts that may prevent wider exploration
fundamental geomorphic question governing of and therefore potentially more
model development. Therefore there is no appropriate models for particular
single answer to the three questions posed in circumstances (Nicholas and Quine, 2007a;
the introduction, and nor can each be Wainwright et al., 2008a).
answered independently. As a result of these
factors a model can and will only remain as a A better position to develop models and
(working) hypotheses of how processes and therefore understanding of process-form
landforms interact. In order to develop, interactions is to consider the uncertainty in
evaluate and use models in geomorphic different model parameters, and therefore
enquiry, uncertainty in both data and models consider a range of possible combinations of
needs to be dealt with in a robust manner. parameters that may reproduce the
data, according to a likelihood function that
Considering numerical modelling from a considers potential uncertainty in the data.
Bayesian perspective provides a suitable Significant advances have been made in
framework for robust model development of developing appropriate likelihood functions in
non-linear open systems. As a result of the related discipline of hydrology that
modelling uncertainties we should be consider different forms of model uncertainty
interested in obtaining the probability that (Beven, 2006; Schoups and Vrugt, 2010).
the model and its parameters are a Such statistical treatments of model
correct representation of reality, given the uncertainty require further adaptation to
available data initial model conditions geomorphic problems, including potential
and driving conditions . In order to do uncertainty in model boundary conditions and
this we combine our prior beliefs about the elevation data (Hutton and Brazier, 2012;
model structure, and associated Nicholas and Quine, 2010; Wheaton et al.,
parameters, (which are dependent on 2010). Calibration and the related sensitivity
the specific model structure) with some data analysis conducted by exploring adequately
using a Likelihood Function (e.g. a measure different parameter combinations that
of model performance based on a given constitute can guide the modeller as
dataset) , to obtain our to which parameters are most important in
posterior belief from Bayes’ equation (Draper, controlling system response (Hutton et al.,
1995): 2012; Saltelli, 1999). Such information can be
then used to guide further data collection
targeted at constraining the most important
parameters.
Furthermore, when different models
and therefore different hypotheses of form-
process interactions are confronted with the
Thus, our confidence in the model is same observations (Hancock et al., 2011;
specifically dependent on the data to Krueger et al., 2009; Pelletier, 2011; Tatard
constrain initial conditions, driving conditions et al., 2008) the strengths and weaknesses of
and that used to evaluate model different model structures may be identified.
performance. Such information can be used to guide further

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

9 Christopher Hutton
data collection and understand the conditions overconfidence in the results of a single
(e.g. ) under which different process model prediction.
parameterisations (e.g. ) are valid.
Conclusion
Numerical Models have, especially over
recent years, become a central tool in
Model
geomorphic enquiry, and have allowed
Data Parameters exploration of a range of system dynamics at
a range of spatial and temporal scales.
Appropriate use of numerical models should
consider the scale of model application, the
Likelihood potential processes controlling landscape
form at the scale of application,
computational resources, data availability,
Posterior and the validity of modelling concepts derived
from previous modelling applications.
Data Collection Given the many uncertainties governing
model application, not least uncertainty
Figure 7. Schematic illustration of the iterative regarding fundamental issues concerning
use of a Bayesian Approach for model process dominance and scale, an approach
development to model development considering such
uncertainties from a Bayesian perspective is
recommended. Such an approach provides a
Therefore posterior understanding derived robust framework for model development,
from comparison to data can guide further model rejection and therefore hypothesis
data collection (Figure 7). Depending on the testing that considers uncertainty in both data
similarities between the conditions used to and models. Advances in data collection and
derive the posterior (e.g. ) and the specification of errors in available data will
newly collected data, the posterior in facilitate robust model development.
equation 5 (e.g. the left hand side) can then
become the prior (e.g. move to the right hand
side of Equation 5) for comparison to the References
newly collected data when combined with a Beven K. 2002. Towards a coherent
likelihood function. When using model philosophy for modelling the environment
structures and parameters derived from Proceedings of the Royal Society of London
previous applications it is up to the modeller A 458: (2026): 2465-2484.
to consider how appropriate such models are
to the situation under current consideration, Beven K. 2006. A manifesto for the
and whether such model application is equifinality thesis Journal of Hydrology 320:
supported by available data. 18-36.

Bayes’ equation therefore provides both a Bishop P. 2007. Long-term landscape

conceptual and probabilistic framework for evolution: linking tectonics and surface
model development that can explicitly processes Earth Surface Processes And
consider uncertainty in models and data, and Landforms 32(3): 329-365.
therefore appropriately frame the use of Bloschl G, Sivapalan M. 1995. Scale Issues
models in developing understanding of In Hydrological Modeling - A Review.
process-form interactions. Process-form Hydrological Processes 9(3-4): 251-290.
understanding is developed iteratively
through continuous dialogue between models Brazier RE, Beven KJ, Freer J, Rowan JS.
and data. Furthermore such development 2000. Equifinality and uncertainty in
provides a more robust grounding for physically based soil erosion models:
subsequent model application to investigate Application of the glue methodology to
“what if” type questions by considering a WEPP-the water erosion prediction project-
range of possible model structures supported for sites in the UK and USA. Earth Surface
by available data, which will prevent Processes And Landforms 25(8): 825-845.

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

 Numerical Modelling 10
Brazier RE, Hutton CJ, Parsons AJ, and land-surface evolution. Journal Of
Wainwright J. 2011. Scaling Soil Erosion Geophysical Research-Earth Surface 114:
Models in Space and Time. In Handbook of F00A03. doi:10.1029/2009JF001265.
Erosion Modelling, RPC Morgan RPC and
Goodrich DC, Unkrich CL, Keefer TO,
Nearing MA (eds). Blackwell: 98-117.
Nichols MH, Stone JJ, Levick LR, Scott RL.
Canfield HE, Goodrich DC. 2006. The impact 2008. Event to multidecadal persistence in
of parameter lumping and geometric rainfall and runoff in southeast Arizona.
simplification in modelling runoff and erosion Water Resources Research 44: W05S14.
in the shrublands of southeast Arizona. doi:10.1029/2007WR006222.
Hydrological Processes 20(1): 17-35.
Hairsine PB, Rose CW. 1992. Modeling Soil
Cao Z, Carling PA. 2002. Mathematical Erosion due to overland flow using physical
modelling of alluvial rivers: reality and myth principles: 2 Sheet Flow. Water Resource
Part I: General review. Proceedings of the Research 28: 237-243.
Institution of Civil Engineers-Water and
Hancock GR, Coulthard TJ, Martinez C,
Maritime Engineering 154(3): 207-219.
Kalma JD. 2011. An evaluation of landscape
Church M. 1996. Space time and the evolution models to simulate decadal and
mountain - how do we order what we see? In centennial scale soil erosion in grassland
The scientific nature of geomorphology, catchments. Journal Of Hydrology 398(3-4):
Rhoads BL and Thorn CE (eds). Wiley: 171-183.
Chichester; 147-170.
Hardy RJ. 2005. Modelling granular sediment
Corenblit D, and Steiger J. 2009. Vegetation transport over water-worked gravels. Earth
as a major conductor of geomorphic changes Surface Processes And Landforms 30: 1069-
on the Earth surface: toward evolutionary 1076.
geomorphology. Earth Surface Processes
Hardy RJ. 2008. Geomorphology Fluid flow
And Landforms 34(6): 891-896.
Modelling: Can Fluvial Flow Only Be
Coulthard TJ. 2001. Landscape evolution Modelled Using a Three-Dimensional
models: a software review Hydrological Approach? Geography Compass 2(1): 215-
Processes 15: 165-173. 234.
Coulthard TJ, Lewin J, and Macklin MG. Hunter NM, Bates PD, Horritt MS, Wilson
2005. Modelling differential catchment MD. 2007. Simple spatially-distributed
response to environmental change models for predicting flood inundation: A
Geomorphology 69: 222-241. review. Geomorphology 90(3-4): 208-225.
Coulthard TJ, Van De Wiel MJ. 2006. A Hutton C, Brazier R. 2012. Quantifying
cellular model of river meandering. Earth riparian zone structure from airborne LiDAR:
Surface Processes And Landforms 31(1): Vegetation filtering anisotropic interpolation
123-132. and uncertainty propagation. Journal Of
Hydrology 442: 36-45
Draper D. 1995. Assessment and
Propagation of Model Uncertainty. Journal of Hutton C, Brazier RE, Nicholas AP, Nearing
the Royal Statistical Society Series B- MA. 2012. On the Effects of Improved Cross-
Methodological 57(1): 45-97. section Representation in One-dimensional
Flow Routing Models Applied to Ephemeral
Favis-Mortlock D. 1998. A self-organizing
Rivers. Water Resources Research 48:
dynamic systems approach to the simulation
W04509. doi:10.1029/2011WR011298.
of rill initiation and development on hillslopes.
Computers & Geosciences 24(4): 353-372. Istanbulluoglu E, Bras RL. 2005. Vegetation-
modulated landscape evolution: Effects of
Ferguson RI. 2003. The missing dimension:
vegetation on landscape processes drainage
effects of lateral variation on 1-D calculations
density and topography Journal Of
of fluvial bedload transport. Geomorphology
Geophysical Research-Earth Surface 110:
56: 1-14.
F02012. doi:10.1029/2004JF000249.
Furbish DJ, Childs EM, Haff PK, Schmeeckle
Kalin L, Govindarju RS, Hantush MM. 2003.
MW. 2009. Rain splash of soil grains as a
Effect of geomorphic resolution on modeling
stochastic advection-dispersion process with
of runoff hydrograph and sedimentograph
implications for desert plant-soil interactions

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

11 Christopher Hutton
over small watersheds. Journal Of Hydrology Nearing MA. 1999. Variability in soil erosion
276: 89-111. data from replicated plots. Soil science
society of America Journal 63(6): 1829-1835.
Karssenberg D, Bridge, JS. 2008. A three-
dimensional numerical model of sediment Nearing MA. 2000. Evaluating soil erosion
transport, erosion and deposition within a models using measured plot data: Accounting
network of channel belts, floodplain and hill for variability in the data. Earth Surface
slope: extrinsic and intrinsic controls on Processes And Landforms 25(9): 1035-1043.
floodplain dynamics and alluvial architecture.
Nicholas AP. 2000. Modelling bedload yield
Sedimentology 55(6): 1717-1745.
in braided gravel bed rivers. Geomorphology
Kleinhans MG. de Haas T, Lavooi E, 36: 89-106.
Makaske B. 2012. Evaluating competing
Nicholas AP. 2005. Cellular Modelling in
hypotheses for the origin and dynamics of
Fluvial Geomorphology Earth Surface
river anastomosis Earth Surface Processes
Processes And Landforms 30: 645-649
And Landforms doi: 101002/esp3282.
Nicholas AP. 2009. Reduced-complexity flow
Krueger T, Quniton JN, Freer J, Macleod CJ,
routing models for sinuous single-thread
Bilotta GS, Brazier RE, Butler P, Haygarth
channels: intercomparison with a physically-
PM. 2009. Uncertainties in Data and Models
based shallow-water equation model. Earth
to Describe Dynamics of Agricultural
Surface Processes And Landforms 34: 641-
Sediment and Phosphorus Transfer. Journal
653
of Environmental Quality 38: 1137-1148.
Nicholas AP, Clarke L, Quine TA. 2009. A
Lane SN. 1998. Hydraulic modelling in
numerical modelling and experimental study
hydrology and geomorphology: A review of
of flow width dynamics on alluvial fans. Earth
high resolution approaches Hydrological
Surface Processes And Landforms 34(15):
Processes 12(8): 1131-1150.
1985-1993
Lane SN, Bradbrook KF, Richards KS, Biron
Nicholas AP and Quine TA 2007a Crossing
PA, Roy AG. 1999. The application of
the divide: Representation of channels and
computational fluid dynamics to natural river
processes in reduced-complexity river
channels: three-dimensional versus two-
models at reach and landscape scales.
dimensional approaches. Geomorphology
Geomorphology 90(3-4): 318-339
29(1-2): 1-20.
Nicholas AP, Quine TA. 2007b. Modeling
Lane SN, Richards KS. 1997. Linking river
alluvial landform change in the absense of
channel form and process: Time space and
external environmental forcing Geology
causality revisited. Earth Surface Processes
35(6): 527-530
And Landforms 22(3): 249-260.
Nicholas AP, Quine TA. 2010. Quantitative
Livingstone I, Wiggs GFS, Weaver CM. 2007.
assessment of landform equifinality and
Geomorphology of desert sand dunes: A
palaeoenvironmental reconstruction using
review of recent progress Earth-Science
geomorphic models. Geomorphology 121(3-
Reviews 80(3-4): 239-257.
4): 167-183.
Merritt WS, Letcher RA, Jakeman AJ. 2003.
Nichols MH. 2006. Measured sediment yield
A review of erosion and sediment transport
rates from semiarid rangeland watersheds
models. Environmental Modelling and
Rangeland Ecology & Management 59(1):
Software 18: 761-799
55-62
Michaelides K, Wilson MD. 2007. Uncertainty
Nield JM, Baas ACW. 2008. Investigating
in predicted runoff due to patterns of spatially
parabolic and nebkha dune formation using
variable infiltration. Water Resources
cellular automation modelling approach.
Research 43(2): W02415.
Earth Surface Processes And Landforms
doi:10.1029/2006WR005039.
33(724-740).
Morgan RPC, Nearing MA (eds) 2010
Oreskes N, Shraderfrechette K, Belitz K.
Handbook of Erosion Modelling. Wiley-
1994. Verification Validation And
Blackwell 416 pp.
Confirmation Of Numerical-Models In The
Earth-Sciences. Science 263(5147): 641-646

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

 Numerical Modelling 12
Pelletier JD. 2011. Fluvial and slope-wash historic braided river evolution
erosion of soil-mantled landscapes: Geomorphology 90: 302-217.
detachment- or transport-linited? Earth
Tucker GE, Hancock G. 2010. Modelling
Surface Processes And Landforms 37(1): 37-
landscape evolution. Earth Surface
51.
Processes and Landforms 35: 28-50
Refsgaard JC, van der Sluijs JP, Brown J,
Van de Wiel MJ, Coulthard TJ, Macklin MG,
van der Keur P. 2006. A framework for
Lewin J. 2011. Modelling the response of
dealing with uncertainty due to model
river systems to environmental change:
structure error. Avdances in Water
Progress problems and prospects for palaeo-
Resources 29: 1586-15977.
environmental reconstructions. Earth-Science
Reinhardt L, Jerolmack D, Cardinale BJ, Reviews 104(1-3): 167-185.
Vanacker V, Wright J. 2010. Dynamic
Wainwright J, Millington JDA. 2010. Mind, the
interactions of life and its landscape:
gap in landscape-evolution modelling. Earth
feedbacks at the interface of geomorphology
Surface Processes And Landforms 35(7):
and ecology. Earth Surface Processes And
842-855.
Landforms 35(1): 78-101.
Wainwright J, Mulligan M (eds). 2004.
Saltelli A. 1999. The role of sensitivity
Environmental Modelling: Finding Simplicity
analysis Transparency relevance robustness
in Complexity. John Wiley and Sons
and parsimony in modelling Safety and
Chichester 430 pp.
Reliability Vols 1 & 2: 1181-1186.
Wainwright J, Parsons AJ. 2002. The effect
Sambrook-Smith GHS, Best JL, Ashworth PJ,
of temporal variations in rainfall on scale
Lane SN, Parker NO, Lunt IA, Thomas RE,
dependency in runoff coefficients. Water
Simpson CJ. 2010. Can we distinguish flood
Resources Research 38(12): 1271,
frequency and magnitude in the
doi:10.1029/2000WR000188.
sedimentological record of rivers? Geology
38(7): 579-582. Wainwright J, Parsons AJ, Abrahams AD.
2000. Plot-scale studies of vegetation
Schmeeckle MW, Nelson JM. 2003. Direct
overland flow and erosion interactions: case
numerical simulation of bedload transport
studies from Arizona and New Mexico.
using a local dynamic boundary condition.
Hydrological Processes 14(16-17): 2921-
Sedimentology 50: 279-301.
2943.
Schoups G, Vrugt JA. 2010. A formal
Wainwright J, Parsons AJ, Muller EN, Brazier
likelihood function for parameter and
RE, Powell DM, Fenti, B. 2008a. A transport-
predictive inference of hydrologic models with
distance approach to scaling erosion rates: I
correlated heteroscedastic and non-Gaussian
Background and model development. Earth
errors. Water Resources Research 46:
Surface Processes And Landforms 33(5):
W10531. doi:10.1029/2009WR008933.
813-826.
Schumm SA, Lichty RW. 1965. Time space
Wainwright J, Parsons AJ, Muller EN, Brazier
and causality in geomorphology American.
RE, Powell DM, Fenti, B. 2008b. A transport-
Journal of Science 263: 110-119.
distance approach to scaling erosion rates: 2
Tatard L, Planchon O, Wainwright J, Nord G, Sensitivity and evaluation of MAHLERAN.
Favis-Mortlock D, Silvera N, Ribolzi O, Earth Surface Processes And Landforms
Esteves M, Hua Huang C. 2008. 33(6): 962-984.
Measurement and modelling of high-
Wheaton JM, Brasington J, Darby SE, Sear
resolution flow-velocity data under simulated
DA. 2010. Accounting for uncertainty in
rainfall on a low-slope sandy soil Journal Of
DEMs from repeat topographic surveys:
Hydrology 348(1-2): 1-12
improved sediment budgets. Earth Surface
Thomas R, Nicholas AP. 2002. Simulation of Processes And Landforms 35(2): 136-156.
braided river flow using a new cellular routing
Wilcock P, Crowe JC. 2003. Surface-based
scheme. Geomorphology 43(3-4): 179-195.
transport model for mixed sized sediment.
Thomas R, Nicholas AP, Quine TA. 2007. Journal of Hydraulic Engineering 129(2): 120-
Cellular modelling as a tool for interpreting 128.

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)

13 Christopher Hutton
Wilcock PR, Iverson RM (eds). 2003.
Prediction in Geomorphology. Geophysical
Monograph Series Vol 135, Washington DC
256 doi:101029/GM135 pp
Willett SD, Brandon MT. 2002. On steady
states in mountain belts. Geology 30(2): 175-
178

British Society for Geomorphology Geomorphological Techniques, Chap. 5, Sec. 2 (2012)