Modelling the hydrological impacts of rural land use change: current state of the science and future challenges Neil McIntyre1*, Caroline Ballard1, Michael Bruen2, Nataliya Bulygina1, Wouter Buytaert1, Ian Cluckie3, Sarah Dunn4, Uwe Ehret5, John Ewen6, Alexander Gelfan7, Tim Hess8, Denis Hughes9, Bethanna Jackson10, Thomas Kjeldsen11, Ralf Merz12, Jong-Sook Park3, Enda O’Connell6, Greg O’Donnell6, Ludovic Oudin13, Ezio Todini14, Thorsten Wagener15 and Howard Wheater16 Imperial College London;2University College Dublin;3Swansea University;4The James Hutton Institute, Aberdeen;5Karlsruhe Institute of Technology; Newcastle University; 7Russian Academy of Sciences; 8Cranfield University; 9Rhodes University; 10University of Victoria Wellington; 11 Centre for Ecology and Hydrology; 12Helmholtz Centre for Environmental Research; 13Paris University; 14University of Bologna; 15 Pennsylvania State University;16University of Saskatchewan *Email: n.mcintyre@ic.ac.uk Abstract The potential links between rural land use and water-related hazards are well recognised, and there is increasing interest in managing the landscape to assist with flood risk mitigation and supply of good quality water However, our ability to quantify the impact of rural land use management on the hydrological cycle is limited and we are not yet able to provide consistently reliable evidence to support water-related planning and policy decisions Numerous projects in the UK and internationally have been attempting to produce new and better prediction methods and tools over the last few years A two-day international workshop, held at Imperial College London in June 2011, brought together 20 experts in hydrological impacts modelling with the objectives of sharing project outputs, knowledge and ideas on the topic of modelling the impacts of land use management change on hydrology This paper describes the main outputs from that workshop with the aim of summarising, for selected aspects of this field, the state of the science and priorities for future research The paper is structured into the following sections (1) Upscaling, including maximising the role of process knowledge using physics-based models and metamodelling approaches, and evaluation of how non-linear process dynamics can be maintained over scales (2) Model regionalisation and paired catchment analysis as a complement or potential alternative to physics-based modelling (3) Tools for quantifying and illustrating the integrating effect of the channel network and associated spatial sensitivities (4) Interactions between hydrological services and other aspects of ecosystem service provision and hence the need for holistic models The paper concludes with a list of research challenges, a summary of current initiatives that are addressing these challenges, and ideas for how they might be addressed in the future Introduction signals (Buytaert et al., 2007; McIntyre and Marshall, 2010), these cannot be safely extrapolated to different catchments and larger scales For larger catchments, seeking evidence of effects is even more of a challenge because of the important role of the channel network in dispersion of flow and chemistry signals (O’Donnell et al., 2011; Pattison and Lane, 2012) The lack of unambiguous evidence about the effects of land use presents a fundamental problem for the hydrological modellers, who rely on observed evidence to formulate, calibrate and validate models The wide perception that land use has a considerable role to play in solving our water management problems provides motivation for increasing our efforts to address this problem In the past few years, numerous research groups internationally have been conceiving, evolving, discussing and testing new approaches to doing so Some of the groups at the forefront of this research (represented by the authors of this paper) were invited to a workshop in London in summer 2011 sponsored Background Catchment management requires capacity for exploring hydrological impacts of rural land use scenarios For example, the interactions between land use and flooding, and land use and drought are of considerable practical interest Sediments, hydrochemistry and hydroecology are strongly linked to land use management, both directly and via the hydrological response, and integrated management would therefore require these links to be quantified (Wheater and Evans, 2009) Seeking evidence of the hydrological effects of rural land use change has been difficult: where signals of change may be expected, they tend to be obscured by other sources of variability and data uncertainty (Beven et al., 2008) For example, evidence-based methods of flood and low flow estimation (Kjeldsen, 2007; Holmes et al., 2005) cannot explicitly allow for rural land use changes Where inter-comparison of small catchments has revealed land use BHS Eleventh National Symposium, Hydrology for a Changing World, Dundee 2012 © British Hydrological Society by the UK Flood Risk Management Research Consortium (FRMRC, www.floodrisk.org.uk) to review progress and current capability, exchange ideas, and discuss future priorities This paper presents the outcomes of that workshop complex, introducing some of the limitations of physics-based modelling (prior specification and non-identifiability) and of data-based modelling (lack of sufficient signals of land use change in the data) The focus on uncertainty estimation by many conceptual modellers has also raised the complex question of characterising and modelling errors and how that may affect predictions (Mantovan and Todini, 2006) The choice between approaches in the context of modelling land use impacts not only depends on the scientific basis and potential errors associated with the methods, but also on practical criteria First, many land use interventions are made over areas that are remote from the much larger catchment scale at which flow predictions are needed This implies the need for a spatially distributed model Second, it is often the case that the hydrological model needs to support prediction of sediment yields or water quality modelling, which will dictate the modelled hydrological variables And thirdly, there may be data and computational limits to the degree of resolution of the model as previously discussed It may be that approaches should be integrated together, or each used for a different component of the catchment, or none used at all, instead reverting to qualitative methods A practical requirement, common across all hydrological modelling tasks, is the need for the model and its supporting documentation to be simple enough to use within reasonable constraints of expertise, data, human and computer resources, and to provide the outputs needed to solve the problem at hand Also common to all models is the good practice modelling procedure (Wagener et al., 2004), including conceptual design, calibration, testing and review Apart from these general requirements, the rural land use problem places particular requirements on the hydrological model and the way its parameters are estimated: (1) Extrapolation As introduced previously, the problem suggests a modelling approach that does not rely entirely on observed evidence because it is required to make predictions of responses under conditions never yet observed: some ability to integrate expert knowledge and theory is required (2) Scale The land use management changes of interest are likely to be at different scales from the hydrological responses of interest, for example flood risk management may involve interventions at the scale of ~100 m while often it is the catchment scale flow response that is important (Jackson et al., 2008) This suggests a multiscale, distributed approach (3) Uncertainty This is common issue in hydrological studies, but the difficulty of extrapolating to future land use scenarios, and the potential influence that a wrong prediction may have on planning and policy, provide a particular motivation for modellers to report uncertainty in predictions (4) Integrated analysis Managing the links between land use and water management cannot be restricted to plain hydrology — models ideally would have the capacity to also consider the roles of (at least) sediments and hydrochemistry General modelling approaches and their applicability The discussions at any workshop of this nature inevitably include debate on the relative merits of different fundamental approaches: top-down (metric, or data-based) and bottomup (mechanistic, or physics-based) modelling, or some combination of the two (conceptual modelling) Some introduction to these types of models and the surrounding issues is worthwhile The essential characteristic of top-down models is that they are based primarily on observations and seek to characterise system response from those data In principle, such models are limited to the range of observed data, and effects such as land use change cannot be represented with any confidence unless well-identified in the data As discussed above, such signals are rarely present (e.g Beven et al., 2008) Signals in hydrological extremes are especially difficult to detect, and data-based approaches such as standard flood frequency analysis are unlikely to detect land use influences (He et al., 2011) Intercomparison of small catchments has led in some cases to land use signals being identified from data — most notably the experiments conducted by the US soil conservation service in establishing the Curve Number method — but interpreting these data, applying them to new contexts, and upscaling to greater scales leads to results with high uncertainty (Bulygina et al., 2011) This is not to say that attempting to so is worthless, as discussed further below In contrast to empirical modelling, ‘bottomup’ or ‘physics-based’ modelling, uses models explicitly based on the best available understanding of the physical hydrological processes (Wheater et al., 1993;, O’Connell and Todini, 1996) They first became feasible in the 1970s when computing power became sufficient to solve the relevant coupled partial differential equations These models are characterised by parameters that have a direct physical significance; a theoretical advantage is that if the physical parameters can be determined a priori, the effects of land use change can be explicitly represented In practice, the underlying physics has been (necessarily) derived from small-scale, mainly laboratory-based, observations Hence the processes and parameter values may not apply under conditions and at scales of interest and, in general, calibration of scale-adjusted values is necessary The problem of limitations of observed data sets and non-identifiability of parameter sets then arises, leading to large uncertainty Another practical problem is that to resolve the small-scale non-linear processes with satisfactory numerical errors over a whole catchment may require many days or weeks of simulation on a personal computer Again, such problems not necessarily overcome the attractions of this approach, but some thought is required about how to manage them However, the most commonly applied hydrological models are the ‘conceptual’ class These models abandon the aspiration to either entirely specify the model using data, or entirely specify the model using prior knowledge Instead, prior knowledge is used to define the key components of the system, typically a series of storages and equations to describe the fluxes between storages, while calibration is used to define the parameter values Problems of parameter and non-identifiability have led to a general preference for parsimonious conceptual modelling and by extensive use of Monte Carlo methods to estimate parameter uncertainty These models not solve the land use impacts prediction problem — from one point of view they make it more The workshop Considerable advances in our land use impacts modelling techniques have been made over the past few years both nationally and internationally under numerous research projects The workshop was initiated and sponsored by the FRMRC in the recognition that there would be considerable benefit in sharing outcomes and ideas from these projects The workshop was two days, 12–14 June 2011 Some of the main outputs in terms of understanding the problems and tools and models, which have unique and lumped relationships between soil wetness and runoff, will produce inaccurate runoff predictions; and then showed how such hysteresis may be included to produce more satisfactory results Such attention to knowledge of physical processes is arguably essential to development of metamodelling (and conceptual modelling more generally), to allow for the challenge of prediction beyond the range of calibration data Another innovation towards managing the computational requirements of physics-based models, while including Monte Carlo-based uncertainty analysis for the purpose of flood risk assessments, is the specific censoring procedure of Gelfan (2010) This approach uses dynamicstochastic modelling (Kuchment and Gelfan, 1991), which couples a deterministic physics-based model with a stochastic weather generator A very large number of samples of weather inputs is generated, but these are censored so that only those that have the potential to pose a flood risk are run through the model Such censoring/screening of Monte Carlo samples may have valuable potential for managing land use impacts uncertainties as well as climate variability ideas for solving them are covered below For convenience this is broken down into four overlapping sub-topics: (1) Upscaling to maximise the use of process knowledge and physics-based models (2) Model regionalisation and paired catchment analysis as a complement or potential alternative to physics-based modelling (3) Tools for quantifying and illustrating the integrating effect of the channel network and associated spatial sensitivities (4) Interactions between hydrological services and other aspects of ecosystem service provision and hence the need for holistic models Upscaling to maximise the use of process knowledge and physics-based models A theoretically attractive approach to modelling land use impacts is physics-based modelling This involves discretising the catchment into many elemental units (e.g Ewen et al., 2006; Park et al., 2009; Gelfan, 2010) and integrating the elemental responses into a catchment scale response, explicitly accounting for both non-linearity and heterogeneity In design at least, this approach allows physical changes associated with land use scenarios to be represented as perturbations in model parameters values (Hashemi et al., 2000; Kuchment and Gelfan, 2002; Park and Cluckie, 2009) However, as noted above, the inherent problems with physics-based models means that data requirements are high, results can be highly uncertain, and catchment scale modelling expensive For example, in the recent work at Pontbren in Wales, UK, to resolve the physics acceptably, it was deemed necessary to use a cm vertical discretisation and m2 grid over a 20 000 m2 field (Jackson et al., 2008), resulting in (personal computer) run-times of several hours to cover a 1-month simulation period However, the required grid size depends on the non-linearity of the specific case and the accuracy criteria used: for example, applying the Topkapi model over larger scales, Martina and Todini (2008) concluded that 10-1000 m grid sizes can be sufficient The need to lower the expense but preserve the capability of physics-based models leads to the idea of metamodelling (Ballard, 2011) Here, metamodelling is taken to mean the substitution of a physics-based model by a conceptual model that both maintains the same basic hydrological principles as the physics-based model and also, closely replicates its flood responses under a range of relevant climate and land use scenarios Previous applications of this general idea in hydrology are few, the closest being the UP framework of Ewen (1997) and the emulation framework of Young (2010) The method includes uncertainty analysis which allows uncertainties associated with data, models and the upscaling procedure to be propagated through to predictions After several years of research under the FRMRC, this upscaling method was considered to have considerable merits for land use change impacts analysis However, uncertainty was high, especially where the physicsbased models were developed without supporting small-scale measurements In the Hodder catchment case study, the errors in predicted changes in peak flows due to the metamodelling step had magnitudes between and 20% at small scale and and 10% at 260 km2 catchment scale (Ballard, 2011) In many cases these errors were larger than the predicted change — the degree of error should ideally be assessed for each individual case The metamodelling approach requires that the conceptual model retains as much of the relevant hydrological processes as possible Martina et al (2011), using the Topkapi model (Todini and Ciarapica, 2001), illustrate the strong hysteresis that can exist in the soil moisture-saturated area relationship for both hill-slopes and large catchments, and the importance of saturation excess processes after the end of rainfall They also illustrated that typical conceptual Model regionalisation and paired catchment analysis Model regionalisation is a general term used to describe the process of generalising an empirical or conceptual model over a whole region (or country, or even continent), rather than just a particular catchment This is relevant because if a model can be successfully generalised over space, capturing relevant spatial signals in land use, then there is some basis for generalising the model over future land use scenarios (Merz and Bloschl, 2009; Oudin et al.; 2010; Buytaert and Beven, 2011; Wagener and Montanari, 2011) Various problems arise however: the well-known difficulty of, and uncertainty in, generalising conceptual model parameters over space (McIntyre et al., 2005, Oudin et al., 2010); signals of rural land use are weak or non-existent within most regionalisation studies (Merz and Bloschl, 2009, 2011) or not cover all scenarios of interest (McIntyre and Marshall, 2010); and there is an incompatibility of scale between regionalisation studies (usually catchment scales of at least several km2) and the scales at which land use change is implemented, for example field scale Nevertheless, where the physics-based upscaling route is considered too expensive and/or uncertain due to lack of the necessary data, regionalisation may be the best applicable source of information Several groups have taken this view and proposed approaches which attempt to address the three main problems listed above (Yadav et al., 2007; Merz and Bloschl, 2009; Bulygina et al., 2011; Singh et al., 2011) The method of Bulygina et al (2011) encompasses ideas from all these groups They condition a conceptual field-scale model using regionalised values of flow response indices: the baseflow index from the HOST soils classification system (Boorman et al., 1995) and the Curve Number from the USDA soil classification system (USDA, 1986) The indices are used to condition parameters of the model using Bayes’ equations, giving posterior parameter distributions for a given land use scenario The posterior likelihood of a sampled parameter set is proportional to the consistency of the simulated hydrological response with the response indices predicted for the same catchment by the HOST/ USDA systems, taking into account, not just the expected value of the index, but also the probability distribution of the index derived from these sources Trading space for time, the posterior distribution of parameter values is propagated to uncertainty in predictions under land use scenarios In validation tests over different land uses, the regionalisation gave hydrograph outputs broadly consistent with observed differences (Bulygina et al., 2009, 2011) although with in an extremely accurate and efficient way The resulting pixel-scale sensitivities are illustrated on maps These can be used to answer questions such as: Which areas of the Hodder are most sensitive to change? What is the importance of soil type on land use impacts? And how does impact change with the rainfall pattern? The main assumption behind this method is ‘spatial linearity’ whereby it is assumed that the total downstream impact is the sum of the impacts from individual pixels When tested on the Hodder, the error introduced by this assumption was small for winter storms and larger for summer storms due to increased non-linearity Further exploration of this method is recommended high uncertainty A major limitation of this approach is the incomplete and uncertain information about field-scale land use management effects contained in the USDA and HOST databases, and it was recommended that new informative indices are sought, and carefully selected information from the metamodelling procedure described previously is also being introduced (Bulygina et al., 2012) In using regionalisation for land use impacts analysis, errors are also introduced in translating evidence of regional variations into speculations of future time variations Bulygina et al (2011) and Holman et al (2011) discuss this in the context of how to apply the US Curve Number system to UK impacts studies and Buytaert and Beven (2009) propose including an additional term in the Bayes’ equation to formally include this evident translation error The problem of course changes depending on the availability of data to support regionalisation In cases where hundreds of well-gauged catchments are available to support empirical regionalisation (McIntyre et al., 2005; Merz and Bloschl, 2009; Oudin et al., 2011; Sawicz et al., 2011) the role of expert knowledge may be relatively small However, in more typical cases, the specification of the model and its parameter uncertainty may rely heavily on expert knowledge (Kapangaziwiri et al., 2009) These authors describe the challenge in South Africa of predicting under environmental change, where the collection and formalisation of expert knowledge is crucial; and they describe a procedure for comparing and combining this knowledge with regionalised indices Interactions between hydrological services and other aspects of ecosystem service provision The preceding discussion highlights some of the difficulties in demonstrating the benefits of various land use practices for managing water hazards This may make it difficult to justify expenditure on changes in land use because the cost effectiveness cannot be quantified In practice, land management activities may have multiple effects above and beyond those that are being targeted, through modifications to sediment or pollutant transport, ecological responses and biodiversity Some of these other responses may be easier to quantify and the multiple benefits that the land use changes deliver may become easier to justify on economic grounds This calls for implementation of an ecosystem services approach that takes a holistic overview of all of the processes occurring and their various roles in providing services (Millenium Ecosystem Assessment, 2005; Morris et al., 2005; Jackson et al., 2012) Examples of land use activities that might have multiple benefits include temporary flood storage zones (Environment Agency, 2012) As well as attenuating a flood peak through slowing the movement of water through the catchment, flood storage ponds often provide suitable conditions for deposition of sediment and associated particulate pollutants, such as phosphorus The wetland environment created can also be valuable for decreasing nitrate concentrations through enhanced de-nitrification Finally, wetland environments often bring biodiversity benefits and can be visually appealing Set against this, consideration of potential negative factors also needs to be made For example, enhancement of the de-nitrification process will generate increased greenhouse gas production, and if sediments are allowed to accumulate in the storage pond, then they may be remobilised, causing enhanced pollutant transport at different times Many other land management activities such as riparian management, tree planting, minimum tillage and drain blocking will also affect more than one aspect of the water environment, pointing to the need for consideration of multiple responses Another issue is that many land use interventions (such as those that might be implemented by a land owner) are typically quite small, meaning that on their own they would be ineffective at achieving any measurable changes in catchment responses However, where multiple activities are undertaken, their combined effectiveness could become valuable (Balana et al., 2012) This calls for the need of a catchment approach to address land use and a holistic overview to identify synergistic activities or potential conflicts Furthermore, it highlights the need for buy-in from land managers if land use management techniques are to become effective in mitigating against environmental hazards Consideration of all of these aspects clearly makes the modelling challenge much more complex, but ultimately is more likely to deliver sustainable management practices that are both efficient and cost-effective in their outcomes Quantifying and illustrating the integrating effect of the channel network Runoff from the land surface travels to sites of interest via open-channel flow in networks of ditches, streams and river channels The impact of a land use change on the shape of a hydrograph can depend strongly on the properties of the channel network, and hence the accurate representation of channel routing processes can be critical to the impacts assessment The aim when estimating impact at a downstream site is to estimate the outcome when the effects of the runoff impact hydrographs from the various subcatchments or pixels interact and accumulate as their runoff moves through the network This can be approached by running a catchment hydrological model, with an explicit channel routing model, before and after suitable adjustments are made to the parameters describing land use at selected pixel(s) However, achieving a sufficiently accurate model is a demanding requirement because it is known that even models that are calibrated accurately against observed discharge can have false sensitivities and give inaccurate estimates for impact (Ewen et al., 2006) Such perturbation analysis can also be computationally demanding if we require extensive explorations of sensitivity, for example if we want to explore spatial variability of sensitivity, which may involve perturbations of land use at hundreds or thousands of pixels A new method of ‘information tracking’ (O’Donnell et al., 2011) helps solve these problems To explain this method, consider the example of a model of the 260 km2 Hodder catchment, where there are 2634 pixels (average area 0.0925 km2) Runoff from each pixel is modelled using an 8-parameter conceptual model In total there are 21 072 (i.e 2634 × 8) parameters that can change as a result of a change in land use If 21 072 simulations are run in each of which only one parameter is perturbed, this gives the basic data required to calculate the sensitivity to a change in each parameter This is, in effect, what is done in information tracking, except that reverse algorithmic differentiation (Griewank, 2000) is used to calculate the 21, 072 sensitivities Conclusion References The problem of modelling the water-related impacts of rural land use change is closely related to that of prediction in ungauged catchments, which have been extensively studied over the last decade and more But in some respects the land use problem is more difficult It usually necessitates the ability to upscale local (e.g field scale) land use changes; it usually requires consideration of non-stationary of vegetation, soils and geomorphology; water managers, under public and political pressure, want modellers to produce nearimmediate and definitive answers while, of course, not getting it wrong; and the importance of integrated land use management requires models of hydrology-driven variables such as nutrients and sediments The June 2011 workshop at Imperial College London brought together representatives of international research groups taking a lead in addressing these problems The need to upscale the impacts of small-scale change and make best use of process knowledge means that there is significant interest in physics-based modelling The workshop discussed methods of addressing the problem using physics-based models within data and computational cost constraints (Jackson et al., 2008; Gelfan, 2010), including the estimation of errors in upscaling procedures (Martina and Todini, 2008; Ballard, 2011) There is clearly much work to on providing guidance for upscaling and trade-offs between errors and computational expense The workshop also covered the provision of tools and data to support spatially distributed models The need for sufficient quality and resolution of rainfall data is clear, and work on remote sensing (Zhu et al 2008) and stochastic modelling (Gelfan, 2010) has improved its provision To understand and communicate the spatial aspect of land use change impacts, the workshop discussed tools that assess how catchment scale impacts of a given land use change will vary with the location of that change (O’Donnell et al., 2011) Looking at more empirical and conceptual modelling methods, the workshop explored concepts for classifying catchments and dynamic signatures, to help us identify and categorise changes and to predict possible change (Wagener et al., 2007; Sawicz et al., 2011) The need for spatially distributed models also requires us to use all possible sources of information about these signatures for model identification, including exploring merging different information types (Kapangaziwiri et al., 2009; Bulygina et al., 2012) There is also continued interest in managing errors and uncertainty (Buytaert and Beven, 2009; Bulygina et al., 2011) Clearly many of the problems and proposed solutions discussed at the workshop are not new, and lessons learnt in the past should not be forgotten Nevertheless, it was clear that the problem of land use impacts modelling is a relatively young and dynamic research field, and there are substantial challenges yet to be met in all the sub-topics discussed Perhaps the greatest challenge, encompassing many theoretical and practical hurdles, will be moving on from pure hydrology to provide multi-constituent models needed to support a more integrated approach to land use impacts management Balana, B.B., Lago, M., Baggaley, N., Castellazzi, M., Sample, J., Stutter, M., Slee, B and Vinten, A.J.A 2012 Integrating economic and biophysical data in assessing cost-effectiveness of buffer strip placement J Environ Qual., 41, 380–388 Ballard, C 2011 The role of physics based models for simulating runoff responses to rural land management scenarios PhD Thesis, Imperial College London, 340pp Beven, K., Young, P., Romanowicz, R., O’Connell, E., Ewen, J., O’Donnell, G., Homan, I., Posthumus, H., Morris, J., Hollis, J., Rose, S., Lamb, R and Archer, D 2008 Analysis of historical data sets to look for impacts of land use and management change on flood generation Defra R&D Final Report FD2120, Defra, London Boorman, D., Hollis, J and Lilly, A 1995 Hydrology of soil types: a hydrologically-based classification of the soils of the United Kingdom Institute of Hydrology, Wallingford Bulygina, N., McIntyre, N and Wheater, H 2009 Conditioning rainfall-runoff model parameters for ungauged catchments and land management impacts analysis Hydrol Earth Syst Sci., 13, 893–904 Bulygina, N., McIntyre, N and Wheater, H 2011 Bayesian conditioning of a rainfall-runoff model for predicting flows in ungauged catchments and under land use changes Water Resour Res., 47, 2, W02503 Bulygina, N., Ballard, C., McIntyre, N., O’Donnell, G And Wheater, H 2012 Integrating different types of information into hydrological model parameter estimation: application to ungauged catchments and land use scenario analysis Water Resour Res, doi:10.1029/2011WR011207, in press Buytaert W., Iñiguez, V., De Bièvre B., 2007 The effects of Pinus patula forestation on water yield in the Andean páramo Forest Ecol Manage., 251, 22–30 Buytaert, W and Beven, K 2011 Models as multiple working hypotheses: Hydrological simulation of tropical alpine wetlands Hydrol Proc., 25, 11, 1784–1799 Buytaert, W and Beven, K 2009 Regionalisation as a learning process Water Resour Res., 45, W11419 Environment Agency 2012 Greater working with natural processes in flood and coastal erosion risk management EA Report., Available online at http://publications.environmentagency.gov.uk/PDF/GEHO0811BUCI-E-E.pdf Ewen, J 1997 ‘Blueprint’ for the UP modelling system for large scale hydrology Hydrol Earth Syst Sci., 1, 1, 55–69 Ewen, J., O’Donnell, G., Burton, A and O’Connell, E 2006 Errors and uncertainty in physically-based rainfall-runoff modelling of catchment change effects J Hydrol., 330, 641–650 Gelfan, A.N 2010 Extreme snowmelt floods: frequency assessment and analysis of genesis on the basis of the dynamic-stochastic approach J Hydrol., 388, 85–99 Griewank, A 2000 Evaluating Derivatives: Principles and Techniques of Algorithmic Differerentiation Society for Industrial and Applied Mathematics, Philadelphia, USA, 369 pp Hashemi, A.M., Franchini, M and O’Connell, P.E 2000 Climatic and basin factors affecting the flood frequency curve: Part I – A simple sensitivity analysis based on the continuous simulation approach Hydrol Earth Syst Sci., 4, 463–482 He, Y., Bárdossy, A and Zehe, E 2011 A review of regionalization for continuous streamflow simulation Hydrol Earth Syst Sci., 15, 3539–3553 Hess, T., Holman, I., Rose, S., Rosolova, Z and Parrott, A 2010 Estimating the impact of rural land management changes on catchment runoff generation in England and Wales Hydrol Proc., 24, 1357–1368 Acknowledgements The workshop was supported by the Flood Risk Management Research Consortium Phase (www.floosdrisk.org.uk), Engineering and Physical Sciences Research Council grant EP/F020511/1 A wide variety of funders supported the research presented at the workshop Holman, I.P., Hess, T.M and Rose, S.C 2011 A broad-scale assessment of the effect of improved soil management on catchment Baseflow Index Hydrol Proc., 25, 16, 2563– 2572 Holmes, M.G.R., Young, A.R., Goodwin, T.H and Grew, R 2005 A catchment-based water resource decision-support tool for the United Kingdom Environ Model Software, 20, 2, 197–202 Jackson, B.M., Wheater, H.S., McIntyre, N.R., Chell, J., Francis, O.J., Frogbrook, Z., Marshall, M., Reynolds, B and Solloway, I 2008 The impact of upland land management on flooding: insights from a multiscale experimental and modelling programme J Flood Risk Manage., 1, 2, 71–80 Jackson, B., Pagella, T., Sinclair, F., Orellana, B., Henshaw, A., Reynolds, B., McIntyre, N., Wheater, H and Eycott, A 2012 Polyscape: a GIS mapping toolbox providing efficient and spatially explicit landscape-scale valuation of multiple ecosystem services Urban and Landscape Planning, in review Jakeman, A.J., Letcher, R.A., Norton, J.P 2006 Ten iterative steps in development and evaluation of environmental models Environ Model Software, 21, 602–614 Kapangaziwiri, E., Hughes, D and Wagener, T 2009 Towards the development of a consistent uncertainty framework for hydrological predictions in South Africa Proc of IAHS, Hyderabad, 6–12 Sept 2009, 84–93 Kjeldsen, T 2007 The revitalised FSR/FEH rainfall-runoff method – a user handbook Flood Estimation Handbook Supplementary Report No 1, Centre for Ecology and Hydrology, Wallingford, UK www.ceh.ac.uk/refh Kuchment, L.S and Gelfan, A.N 1991 Dynamic- stochastic models of rainfall and snowmelt runoff Hydrol Sci J., 36, 153–169 Kuchment, L.S and Gelfan, A.N 2002 Estimation of extreme flood characteristics using physically based models of runoff generation and stochastic meteorological inputs Water International, 27, 77–86 Mantovan, P and Todini, E 2006 Hydrological forecasting uncertainty assessment: Incoherence of the GLUE methodology J Hydrol., 330, 1-2, 368–381 Martina, M.L.V., Todini, E and Liu, Z 2011 Preserving the dominant physical processes in a lumped hydrological model J Hydrol, 399, 1-2, 121–131 Martina, M.L.V and Todini, E 2008 Watershed hydrological modelling: towardphysically meaningful processes representation, vol 63 In: Hydrological Modelling and the Water Cycle, Springer, pp 229–241 McIntyre, N., Lee, H., Wheater, H., Young, A and Wagener, T 2005 Ensemble predictions of runoff in ungaged catchments Water Resour Res., 41, W12434, doi:10.1029/2005WR004289 McIntyre, N., and Marshall, M 2010 Identification of rural land management signals in runoff response Hydrol Proc., 24, 3521–3534 Merz, R and Blöschl, G 2009 A regional analysis of event runoff coefficients with respect to climate and catchment characteristics in Austria, Water Resources Research, 45, W01405, doi:10.1029/2008WR007163 Merz, R., Parajka, J and Blöschl, G 2011 Time stability of catchment model parameters: Implications for climate impact analyses Water Resour Res., 47, W02531, doi:10.1029/2010WR009505 Millennium Ecosystem Assessment 2005 Ecosystems and Human Well-being: Synthesis Island Press, Washington DC Morris, J., Hess, T.M., Gowing, D.G., Leeds-Harrison, P.B., Bannister, N., Vivash, R.M.N and Wade, M 2005 A framework for integrating flood defence and biodiversity in Washlands in England Int J River Basin Management, IAHR &INBO, 3, 2, 1–11 O’Connell, P.E and Todini, E 1996 Modelling of rainfall, flow and mass transport in hydrological systems: an overview J Hydrol., 175, 3–16 O’Connell, P.E., Beven, K., Carney, J.N., Clements, R.O., Ewen, J., Fowler, H., Harris, G., Hollis, J., Morris, J., O’Donnell, G.M.O., Packman, J.C., Parkin, A., Quinn, P.F., Rose, S.C., Shepher, M and Tellier, S 2004 Review of Impacts of Rural Land Use and Management on Flood Generation, R&D Technical Report FD2114/TR, DEFRA, London, 152pp O’Donnell, G., Ewen, J and O’Connell, P.E 2011 Sensitivity maps for impacts of land management on an extreme flood in the Hodder catchment, UK Phys Chem Earth, 36, 630–637 Oudin, L., Kay, A., Andréassian, V and Perrin, C 2010 Are seemingly physically similar catchments truly hydrologically similar? Water Resour Res., 46, W11558, doi:10.1029/2009WR008887 Oudin, L., Andreassian, V., Lerat, J., Michel, C 2008 Has land cover a significant impact on mean annual streamflow? An international assessment using 1508 catchments J Hydrol., 357:303–316 doi: 10.1016/j.jhydrol.2008.05.021 Park, J.S and Cluckie, I.D 2009 Modelling the impact of land use change in a strategic wetland with a fully distributed model Proc.f 8th International Conference on Hydroinformatics, Concepción, Chile, 12–16 January 2009, vol 2, pp1368–1377 Park, J.S., Ren, Q., Chen, Y., Cluckie, I.D., Butts, M and Graham, D 2009 Effectiveness of complex physics and DTM based distributed models for flood risk management of the River Tone, UK IAHS Red Book, 331, 114–121 Pattison, I and Lane, S.N 2012 The link between land-use management and fluvial flood risk: A chaotic conception? Prog Phys Geogr., 36, 1, 72–92 Sawicz, K., Wagener, T., Sivapalan, M., Troch, P.A and Carrillo, G 2011 Catchment classification: empirical analysis of hydrologic similarity based on catchment function in the eastern USA Hydrol Earth Syst Sci Discussion, 8, 4495–4534 Singh, R., Wagener, T., van Werkhoven, K., Mann, M., and Crane, R 2011 A trading-space-for-time approach to probabilistic continuous streamflow predictions in a changing climate Hydrol Earth Syst Sci Discuss., 8, 6385–6417 Todini, E., Ciarapica, L 2001 The TOPKAPI model In Singh, V.P et al (Eds.) Mathematical Models of Large Watershed Hydrology, Water Resources Publications, Littleton, Colorado USDA 1986 Urban hydrology for small watersheds, United States Department of Agriculture, TR 55, 1–164 Wagener, T., Wheater, H and Gupta, H.V 2004 Rainfallrunoff modelling in gauged and ungauged catchments Imperial College Press, 306 pp Wagener, T., Sivapalan, M., Troch, P and Woods, R 2007 Catchment classification and hydrologic similarity Geogr Compass, 1, 4, 901–931 Wagener, T and Montanari, A 2011 Convergence of approaches toward reducing uncertainty in predictions in ungauged basins Water Resour Res., 47, W06301, doi:10.1029/2010WR009469 Wheater, H.S., Jakeman, A.J and Beven, K.J 1993 Progress and directions in rainfall-runoff modelling In: Modelling Change in Environmental Systems, Ed A.J Jakeman, M.B Beck and M.J McAleer, Wiley, 101–132 Wheater, H and Evans E 2009 Land use, water management and future flood risk Land Use Policy, 26S, S251–S264 Yadav, M, Wagener, T and Gupta, H 2007 Regionalization of constraints on expected watershed response behavior for improved predictions in ungaged basins Adv Water Resour., 30, 1756–1774 Young, P.C 2010 The data-based mechanistic approach to the emulation of large dynamic hydrological process models, Proc BHS 2010 International Conference, Newcastle Zhu, D., Park, J-S., Rico-Ramirez, M.A and Cluckie, I.D 2008 Sensitivity Analysis of a Distributed Hydrological Model for the Upper Medway Catchment using Point and Radar-based Rainfall Data, In: Sustainable Hydrology for the 21st Century, Proc 10th BHS National Hydrology Symposium, Exeter, 153–158 ... range of relevant climate and land use scenarios Previous applications of this general idea in hydrology are few, the closest being the UP framework of Ewen (1997) and the emulation framework of. .. of the difficulties in demonstrating the benefits of various land use practices for managing water hazards This may make it difficult to justify expenditure on changes in land use because the. .. illustrating the integrating effect of the channel network Runoff from the land surface travels to sites of interest via open-channel flow in networks of ditches, streams and river channels The impact of