Printer Friendly

The GCTE validation of soil erosion models for global change.

Take a hammer! It may be used (or misused) to bang in nails and tacks, as a crowbar or lever, and for many other things. Now imagine that you are selecting a hammer for use in some odd job. You probably would not just grab whatever hammer happened to be closest to hand. Instead, you might first pick one or two from your toolbox and check that their heads are secure and their handles straight, and maybe even make a few experimental swings. Now, if you knew something about the work to be done, you would select a hammer of about the right size. OK, but what if you did not know in advance what the task might involve? Well, most probably you would take along a few hammers, big and small, light and heavy, long and short, and then - once you knew more about the work - use that which you judged to be the best suited.

An obvious strategy? In the development, exploration, and use of runoff and soil erosion models, the generally adopted methodology has not usually involved any such explicit consideration of 'experimental design' (as it has become known in other areas of science and engineering). Instead, a 'traditional' approach has evolved. This has been described by many researchers in a variety of ways, but all descriptions include steps akin to those in Table 1 (Haan et al. 1982). In particular, the reasons for choice of a particular model are rarely stated. Although this approach has been and continues to be used by numerous hydrologic and soil erosion modelers, and has proven to be extremely useful in helping us integrate our knowledge about processes and our databases, it would not seem to be the most appropriate experimental design for estimating soil erosion rates under the uncertain conditions of future global change (Ingram et al., this volume).

Estimates of future erosion rates will be required at a wide range of spatial and temporal scales; and at different scales, different processes dominate (Dickinson et al. 1990). Even at similar scales, erosional processes vary in importance from place to place and from time to time (Imeson and Kirkby, this volume). Thus to devise an appropriate balance between the various erosional processes at appropriate temporal and spatial scales is one of the main skills of the model-builder. Scientific ingenuity has not failed to rise to this challenge: in response, a wide diversity of erosion models has been and continues to be produced. However, the majority of erosion models have been produced for the (mostly humid-temperate) countries of the industrialized northern nations. Estimates of the effects of global change, however, will also be required for the developing countries of the south. An additional task is therefore to evaluate the performance of these models in the humid tropics and semi-arid areas. To do this, the Global Change and Terrestrial Ecosystems (GCTE) Soil Erosion Network has adopted the strategy in Table 2 for its validation of suitable modeling tools. This broadly resembles the hammer-selection analogy above. 'Model validation' as discussed here covers the middle three stages of Table 2, and is therefore used in a rather wide sense - paraphrasing the Oxford English Dictionary: 'to confirm that the model is sound and defensible for the conditions under which it will be used.'
Table 1. The 'traditional' approach to model-based studies of soil
erosion

Stage   Description                   Comments

A       Development/selection of a    Questions/issues to be
        model                         addressed may not be
clarified

B       Calibration of model with a   Data are often not quite what
        subset of available data      is required
        (the 'training set')

C       Verification/validation of    -
        model with a different
        subset of data (the
        'testing set')

D       Use model                     Perhaps with insufficient
                                      attention given to model
                                      assumptions, hypotheses, etc.


This discussion covers the experimental design of the GCTE model validation exercise, with a particular emphasis on the lessons learned from that part of it which has so far been completed.

Methodology

Selection of models. Given the large numbers and wide variety of available erosion models, a phased strategy is essential. One possible approach to categorizing erosion models is to group them by their spatial applicability; another is by the main erosive agent modeled (water or wind). For the GCTE model evaluation, a hybrid of the two is used, with models being validated in the following order:

* field-scale water erosion models;

* watershed-scale (catchment-scale in European usage) water erosion models;

* wind erosion models;

* models with a landscape-scale and larger focus.

Of these, field-scale water erosion models have been available for longest. Several studies have listed available models of this type (Knisel 1980b; Warner and Dysart 1980; Foster and Lane 1981; Foster 1982; Larson et al. 1982; Renard et al. 1982; Foster and Lane 1987; De Roo 1993). Data requirements of this class of models are generally less demanding than those of their distributed watershed-scale counterparts; in some cases (e.g., WEPP), field-scale models form the 'building-blocks' of watershed-scale models. Finally, because of their longer developmental history, representatives of this class of models tend to be more 'mature' and stable. Evaluation of the field-scale water erosion models first therefore offers some advantages.

This discussion mainly concerns the ongoing validation of the field-scale water erosion models. Much of the methodology will be appropriate to the subsequent validation of the other classes of model, however. The field-scale models being evaluated by GCTE are listed in Table 3. Note that these include both event and continuous (i.e., multiple event) types.

Overall design of the validation exercise. While the model-evaluation strategy of Table 2 is essentially pragmatic, being driven by the needs of GCTE, it does broadly follow the form of a theoretical framework for model evaluation outlined by Klemes (1986). Klemes' widely used strategy is hierarchical, based on setting the model progressively harder tests, which, as each is passed, increase the confidence of the user in the model. The first stage is a split-sample test, with one half of the data record being used for calibration and the other for validation. If the model passes this first test it is then subjected to a proxy-location test (calibrated using data from one location and tested with data from another), to assess its geographic transferability. The next stage is a differential split-sample test, which tests the model's ability to simulate conditions different to those of the data used for calibration (e.g., the model may be calibrated for one land use or set of climatic conditions but validated on another). The final stage is a proxy-basin differential split-sample test. If the model passes this, then it can be considered transferable in terms of geographical location, climate, and land use. While elegant, this methodology does not address the need to corroborate the theoretical basis of many models, nor does it address the problems of applying models without calibration. These are discussed below.

Datasets: common and user-supplied. A central pillar of the GCTE model evaluation is the use of common - i.e., shared - datasets. These were distributed to the model users in a standard format developed for the GCTE erosion model evaluation, based on that proposed by Hunt et al. (1994). This standard is in turn based upon the widely used IBSNAT (1989) format for crop models. Details of the datasets used in the first stage of the field-scale evaluation - three U.S., two Portuguese, and two Canadian - are given in Table 4.

If several models are run using the same common datasets, then this provides an objective way of comparing their output. There are disadvantages to this widely-used methodology, however, including the following:

* Each dataset will probably not include all the variables needed by a particular model; consequently default or 'best-guess' values will have to be used. Since model users may not be familiar with the datasets, the values used for these may not be optimum. In this sense, model results may tend to be 'worst-case'.

* For the GCTE evaluations, 'model users' are commonly the same people as 'model developers' (cf. Table 3). This means that they will possess greater insight into the model's workings than will the 'typical' user; which in turn is likely to affect the quality of model results (Botterweg 1995); from this point of view, those obtained in the GCTE evaluation might be expected to be 'best-case'.

* 'Constrained calibration' (see below) is not practically possible if different modelers are running the models.

As a partial response to these problems, a parallel set of model evaluations were carried out for some of the field-scale models. Instead of common datasets, these employed datasets supplied by the model users (Table 3). These additional evaluations gave the following advantages:

* Because the model users supply their own data, they are familiar with it. Values supplied to the models are therefore likely to be near optimum for that dataset, representing 'best-case' conditions.

* Model users in this instance are not model developers, but more nearly represent the 'typical' user.

* Constrained calibration (below) is possible if one user runs several models.

* The extra results provide a crude check on the results from the common-dataset evaluations - if a model does well in one type of evaluation, it might reasonably be expected to do well in the other.

* The extra datasets extend the range of conditions covered.
Table 2. The experimental design adopted for the GCTE evaluation of
soil erosion models

Stage   Description                  Comments

1       Select models                -

2       Verify models                Use common datasets to ensure
                                     that models work as expected
                                     under the conditions for which
                                     they were designed (their
'home
                                     areas'). This may involve
                                     calibration, and is then
                                     equivalent to stages B and C
in
                                     Table 1.

3       Determine applicability      Determine the range of
                                     conditions (geographical
                                     location, climate, land use
                                     etc.) under which each model
                                     may reasonably be applied,
also
                                     by use of common datasets (but
                                     not from 'home areas').

4       Global change sensitivity    Ensure that models are
        analysis                     realistically sensitive to
                                     to those elements of climate
                                     and use which are expected to
                                     change in the future (provided
                                     that these are within the
range
                                     of stage 3) by the use of
                                     perturbed common datasets.

5       Use models for global        -
        change studies


Calibration. Clearly, calibration is not possible when erosion models are run using future datasets; models that require calibration are therefore likely to be of little use for global change studies. However, for present conditions some models perform better (in the sense of matching measured values) when calibrated. What is the best way of handling this?

If more than one model is being validated, three approaches are possible: no calibration, 'independent' calibration, or 'constrained' calibration (Favis-Mortlock 1994). With independent calibration, modelers may freely adjust values of selected input variables for each model (e.g., those pertaining to infiltration such as SCS curve number or saturated conductivity, or to surface flow conditions such as Manning's) to achieve the best fit to measured data. This is likely to result in different values of the same variable being used for different models, or by different users for the same model (Botterweg 1995). With constrained calibration, the values of input variables are constrained so that the same values are used for all models. Constrained calibration therefore has the advantage of ensuring consistency between models. However, it is more complex, and is only practically possible when the same person is carrying out all simulations. Thus for model evaluations which involve several modelers - such as that part of the GCTE exercise which makes use of common datasets - only two approaches to calibration are possible: either calibration is disallowed (which would be a disadvantage [TABULAR DATA FOR TABLE 3 OMITTED] to those models for which calibration is essential or desirable), or independent calibration must be permitted. The latter option was chosen for the common-dataset part of the field-scale model evaluation. Modelers were requested, however, to specify which variables (if any) were used for calibration. Extensive use of calibration would strongly disadvantage a model for global change studies.

Training sets and testing sets. If calibration is permitted, a potential danger is 'getting the right answer for the wrong reason,' where goodness of fit between measured and simulated values results merely from calibration. A standard method of reducing this risk was adopted for the GCTE evaluation. Measured time series data for runoff and erosion was partitioned into 'training' and 'testing' subsets (Table 1). The training subset was supplied to each modeler, and could be used by modelers to check (and/or calibrate) their model's performance. The testing subset was withheld for later comparison with model output.

Simple evaluation and consideration of process representations. In broad terms, two styles of validation are possible: the first is a simple evaluation of the model's ability to predict certain occurrences, e.g., runoff and soil loss; the second involves evaluating the adequacy of the model's treatment of processes producing erosion. While less straightforward, this is necessary to further reduce the possibility of 'the right answer for the wrong reason.' When model outputs do not match measured values, a legitimate conclusion is that the model does not adequately represent reality. However, even when model outputs more or less coincide with measured values, it is still not possible to be sure that the model is a good representation of reality. For example, it may be that two (or perhaps several) components of the model are yielding significant but compensating errors for the range of input data explored (Fleming and Shoemaker 1994; Oreskes et al. 1994). Even if processes are being described accurately, there are likely to be parameterization errors. These may lead to different parameter sets giving the same model output as noted by Beven (1993) and Quinton (1994a, b). An approach therefore adopted by both these authors was to use Monte Carlo simulation techniques to produce probability distributions of model outputs. However, this approach is extremely time-consuming and hence beyond the scope of the GCTE exercise. Another possible check is to split the models into smaller sub-models (Bunnel 1989), such as a hydrologic model and an erosion model, or even the splash erosion and flow erosion components. Outputs from these may then be evaluated separately, either against measured data or by comparing them with relationships reported in the literature; this method was adopted by Quinton (1994a).

Although both simple evaluation and evaluation of process representation may be used either separately or in combination, the relative emphasis placed upon each will depend in part upon the nature of the model used.

Model categories. Three categories of model are currently in use within the soil erosion research community: those based entirely on empirical relationships (often derived from field plot experiments); e.g., USLE (Wischmeier and Smith 1978), SLEMSA (Elwell 1981), or RUSLE (Renard et al. 1991); and those that are largely based on mathematical descriptions of physical processes; e.g., WEPP (Nearing et al. 1989) or EUROSEM (Morgan et al. 1992). The third class is intermediate, incorporating some process descriptions but retaining a substantial empirical base; e.g., CREAMS/GLEAMS (Knisel 1980a; Leonard et al. 1987) and EPIC (Williams 1985). Completely empirical models embody no scientific theories (i.e., they describe only the relationships between model inputs and outputs in the classic 'black box' sense) and so can only be evaluated, whereas process-based models can also have their scientific rigour examined through corroboration and refutation of their constituent theoretical base (Medawar 1968).

[TABULAR DATA FOR TABLE 4 OMITTED]

[TABULAR DATA FOR TABLE 5 OMITTED]
Table 6. Approaches to validation in the GCTE experimental design

Stage   Description                Approach to validation

1       Select models              -

2       Verify models              Mainly simple evaluation of
                                   modelled and measured results.
                                   Assume that in general, model
                                   will adequately represent
                                   processes in its 'home area'

3       Determine applicability    Simple evaluation of modelled
                                   and measured results using
                                   datasets from outside each
                                   model's 'home area', plus
                                   consideration of the validity
                                   of the modelling approach used

4       Global change              Simple evaluation by comparing
        sensitivity analysis       modelled results, also
                                   consideration of the
                                   applicability of the modelling
                                   approach to those attributes or
                                   processes affected by global
                                   change

5       Use models for global      -
        change studies
Table 7. Example global change perturbations

Variable                        Perturbation

Land use                        natural vegetation to cropland
                                rotated crops to monoculture
                                dryland to irrigated crops

Mean temperature                present + 1.5 [degrees] C
                                present + 3.0 [degrees] C

Rainfall amount                 present [+ or -] 10%
                                present [+ or -] 20%

Number of rain-days             present [+ or -] 5%
                                present [+ or -] 10%

Atmospheric C[O.sub.2]          to 490 ppmv


In the cautious extrapolations from current conditions which are characteristic of global change erosion studies (Table 5), purely empirical approaches are problematic. For example, the protection of the soil surface given by a growing crop will depend in a complex way on both future weather and atmospheric C[O.sub.2] concentration (Favis-Mortlock and Boardman 1995). Wholly empirical models such as the USLE cannot easily capture such interactions. More process-based modelling approaches are therefore demanded (Williams et al. this volume). Note though, that against these advantages of process-based models must be set their greater (frequently much greater) data requirement. Global change studies, therefore, are likely to use only models which are at least partly process-based: thus both simple evaluation and evaluation of each model's theoretical base can take place at different stages of the validation (Table 6).

Expression of model results. The traditional way of expressing long-term soil loss - whether measured, or as results from continuous simulation erosion model - is as a mean annual erosion rate. While attractively simple, this is unsatisfactory in at least two respects.

* Frequency distributions constructed from time series data of measured erosion events are usually highly skewed (e.g., Hallsworth 1987; Boardman 1988; Dickinson et al. 1990; Govers 1991; Evans 1993). The use of a measure such as the mean is therefore statistically inappropriate. Such skewness enables a relatively few high values in the distribution to greatly inflate the mean for the whole sample. The median is a more statistically suitable measure of central tendency for such distributions (Evans 1990). However, the use of the mean is well entrenched in current practice.

* A similar annual rate could be the result of many small events, or a few large ones.

Therefore, where possible in the GCTE exercise, each continuous simulation model must produce estimates of the following:

* long-term mean rates (up to the length of erosion measurements)

* the number of events

* the distribution of events.

For the event models, comparison of measured and simulated hydrographs and sedigraphs is preferred.

It should be noted that failure to match measured values does not mean that a model is useless. The aims of any modeling exercise may include both prediction and improved understanding of the system modeled (Kirkby et al. 1992; Parsons and Abrahams 1992). An aberrant result may be very valuable in highlighting inadequate system description (Quinton 1994b), or a need for improved data, to the benefit of future modelers.

Construction of global change datasets. The fourth stage in the GCTE validation (Table 2) is a comparison of model sensitivity to global change. Variants of the common datasets used in earlier stages of the exercise will be perturbed within the range of expected future climate and land use change (Table 7). For changes in climate, this sensitivity-based approach is preferred to the use of downscaled GCM output, since this can currently give little information regarding future rainfall intensities (Favis-Mortlock and Boardman, 1995). Note that the use of generated weather can bring problems of its own - care should be taken to ensure that low-frequency, high-magnitude events are reproduced sufficiently, otherwise erosion rates may be radically under-simulated (Favis-Mortlock 1995).

The Oxford meeting and beyond

The second stage of the GCTE field-scale erosion model evaluation, verification of models in their 'home' conditions (Table 2), has now been completed. This took place at a NATO-funded Advanced Research Workshop held in Oxford, UK, on September 11-14, 1995.

Results from this evaluation will be presented in Boardman and Favis-Mortlock (in press); results from the 'user-supplied data' studies (Table 5) will appear in Favis-Mortlock (in press) and Wilcox and Simanton (in press). Plans have been made for the third stage (Table 2); results from this are to be compared and discussed at a further GCTE Soil Degradation meeting planned for April 1997 in Utrecht, The Netherlands. Planning is under way for the initial evaluation of watershed-scale models at the same meeting.

Conclusions

The GCTE model validation exercise, while it has a relatively straightforward experimental design, is fairly complex in terms of the detailed methodology. Results from this evaluation will be invaluable in the development of suitable tools for estimating the effects of global change upon soil erosion.

As well as contributing to the migration of erosion models from the experimenter to the policy-maker, the GCTE model evaluation will assist in the development of the next generation of models, by initiating or strengthening the dialogue between model builders and field workers, between agricultural engineers and geomorphologists, and between countries. For this, the GCTE common datasets are likely to continue to be of value in evaluating progress. Additionally, many modelers will also be involved in the subsequent catchment-scale exercise. The resultant cross-fertilization of lessons learned can only improve our understanding of the art and science of erosion modeling, and continue to extend the boundaries of erosion model application.

In a changing world, trustworthy tools are needed more than ever.

REFERENCES CITED

Abrahart, R.J., M.J. Kirkby, and M.L. McMahon. 1994. MEDRUSH - a combined Geographical Information System and large scale distributed process model. Proceedings 2nd National Conference on GIS Research, Leicester, UK. pp 67-76.

Beven, K. 1993. Prophecy, reality, and uncertainty in distributed hydrological modeling. Advances in Water Resources 16, 41-51.

Boardman, J. 1988. Severe erosion on agricultural land in East Sussex, UK, October 1987. Soil Technology 1, 333-348.

Boardman, J., R. Evans, D.T. Favis-Mortlock, and T.M. Harris. 1990. Climate change and soil erosion on agricultural land in England and Wales. Land Degradation and Rehabilitation 2(2), 95-106.

Boardman, J., and D.T. Favis-Mortlock. 1993. Climate change and soil erosion in Britain. Geographical Journal 159(2), 179-183.

Boardman, J., and D.T. Favis-Mortlock (eds). (in press). Modeling Soil Erosion. Springer-Verlag NATO-ASI Global Change series.

Botterweg, P. 1994. Modelling the effects of climate change on runoff and erosion in central southern Norway. In: R.J. Rickson (ed), Conserving Soil Resources - European Perspectives. CAB International, Wallingford, UK. pp. 273-285.

Botterweg, P. 1995. The user's influence on model calibration results: an example of the model SOIL, independently calibrated by two users. Ecological Modeling 81, 71-81.

Bunnel, F.R. 1989. Alchemy and Uncertainty: What Good are Models? USDA Forest Service, Pacific Northwest Research Station. pp. 27.

De Roo, A.P.J. 1993. Modeling Surface Runoff in Catchments using Geographical Information Systems. Faculteit Ruimtelijke Wetenschappen Universiteit Utrecht, Utrecht, Netherlands. 295 pp.

Dickinson, W.T., G.J. Wall, and R.P. Rudra. 1990. Model building for predicting and managing soil erosion and transport. In: J. Boardman, I.D.L. Foster, and J.A. Dearing (eds), Soil Erosion on Agricultural Land. Wiley, Chichester, UK. pp. 415-428.

Elwell, H.A. 1981. A soil loss estimation technique for southern Africa. In: R.P.C. Morgan (ed), Soil Conservation - Problems and Prospects. Wiley, Chichester, UK. pp. 281-292.

Evans, R. 1990. Water erosion in British farmers' fields - some causes, impacts, predictions. Progress in Physical Geography 14(2), 199-219.

Evans, R. 1993. Extent, frequency and rates of tilling of arable land in localities in England and Wales. In: S. Wicherek (ed), Farmland Erosion in Temperate Plains, Environments and Hills. Elsevier, Amsterdam, The Netherlands. pp. 177-190.

Favis-Mortlock, D.T. 1994. Use and Abuse of Soil Erosion Models in Southern England, Unpublished PhD Thesis. University of Brighton, Brighton, UK. 310 pp.

Favis-Mortlock, D.T. 1995. The use of synthetic weather for soil erosion modeling. In: D.F.M. McGregor and D.A. Thompson (eds.), Geomorphology and Land Management in a Changing Environment. Wiley, Chichester, UK. pp. 265-282.

Favis-Mortlock, D.T. (in press). Evaluation of field-scale erosion models on the UK South Downs. In: J. Boardman and D.T. Favis-Mortlock (eds), Modeling Soil Erosion. Springer-Verlag NATO-ASI Global Change series.

Favis-Mortlock, D.T., R. Evans, J. Boardman, and T.M. Harris. 1991. Climate change, winter wheat yield, and soil erosion on the English South Downs. Agricultural Systems 37(4), 415-433.

Favis-Mortlock, D.T., and J. Boardman. 1995. Nonlinear responses of soil erosion to climate change: a modeling study on the UK South Downs. Catena 25(1-4), 365-387.

Favis-Mortlock, D.T., and M.R. Savabi (in press). Shifts in rates and spatial distributions of soil erosion and deposition under climate change. In: M.G. Anderson and S. Brooks (eds), Advances in Hillslope Processes. Wiley, Chichester, UK.

Fleming, R.A., and C.A. Shoemaker. 1994. Validation of large scale process-oriented models for managing natural resource populations: a case study. In: J. Grasman and G. van Staten (eds.), Predictability and Nonlinear Modelling in Natural Sciences and Economics, Kluwer, Dordrecht, Holland. pp. 138-148.

Foster, G.R. 1982. Modeling the erosion process. In: C.T. Haan, H.P. Johnson, and D.C. Brakensiek (eds), Hydrologic Modeling of Small Watersheds, ASAE Monograph 5. American Society of Agricultural Engineers, St. Joseph, Michigan. pp. 297-380.

Foster, G.R., and L.J. Lane. 1981. Simulation of erosion and sediment yield from field-sized areas. In: R. Lal and E.W. Russell (eds.), Tropical Agricultural Hydrology, Wiley, Chichester, UK. pp. 375-394.

Foster, G.R., and L.J. Lane. 1987. Beyond the USLE: advancements in soil erosion prediction. In: L.L. Boersma et al. (eds), Future Developments in Soil Science Research, Soil Science Society of America, Madison, Wisconsin. pp. 315-326.

Govers, G. 1991. Rill erosion on arable land in central Belgium: rates, controls, and predictability. Catena 18,133-155.

Haan, C.T., H.P. Johnson, and D.L. Brakensiek. 1982. Hydrologic Modeling of Small Watersheds, ASAE Monograph 5. American Society of Agricultural Engineers, St. Joseph, Michigan.

Hallsworth, E.G. 1987. Anatomy, Physiology and Psychology of Soil Erosion. Wiley-Interscience, Chichester, UK.

Hunt, L.A., J.W. Jones, G. Hoogenboom, D.C. Godwin, U. Singh, N. Pickering, P.K. Thornton, K.J. Boote, and J.T. Ritchie. 1994. General input and output file structures for crop simulation models. In: P.F. Uhlir and G.C. Carter (eds), Crop Modeling and Related Environmental Data. CODATA, Paris, France.

Imeson, A.C., and M.J. Kirkby. 1996. Scaling up soil erosion models for global change studies. Journal of Soil and Water Conservation 51 (5) pp. 391-396.

Ingram, J.S.I., J.J. Lee, and C. Valentin. 1996. The GCTE Soil Erosion Network: a multi-participatory research programme. Journal of Soil and Water Conservation, 51(5). pp. 377-380.

International Benchmark Sites Network for Agrotechnology Transfer. 1989. Technical Report 5: Documentation for the IBSNAT Crop Model Input and Output Files, Version 1.1. Department of Agronomy and Soil Science, College of Tropical Agriculture and Human Resources, University of Hawaii.

Kirkby, M.J., P.S. Naden, T.P. Burt, and D.P. Butcher. 1992. Computer Simulation in Physical Geography, (2nd edition). Wiley, Chichester, UK. 180 pp.

Kirkby, M.J., and N.J. Cox. 1995. A climatic index for soil erosion potential (CSEP), including seasonal and vegetation factors. Catena 25, 333-352.

Kirkby, M.J., R. Abrahart, M.D. McMahon, J. Shao, and J.B. Thornes. (in press). MEDALUS Soil Erosion Models for Global Change. Paper presented at 3rd International Geomorphology Conference, Hamilton, Ontario (for publication in an edited volume).

Kirkby, M.J., A.J. Baird, J.G. Lockwood, M.D. McMahon, P. Mitchell, J. Shao, J.E. Sheehy, J.B. Thornes, and F.I. Woodward. (in press b). The MEDALUS slope catena model: a physically based process model for hydrology, ecology and land degradation interactions in MEDALUS I. In: J.B. Thornes (ed), Mediterranean Desertification and Land Use, Wiley, Chichester, UK.

Klemes, V. 1986. Operational testing of hydrological simulation models. Hydrological Sciences Journal 31, 13-24.

Knisel, W.G. (ed.). 1980a. CREAMS - a Field Scale Model for Chemicals, Runoff and Erosion from Agricultural Management Systems. U.S. Department of Agriculture Research Report No. 26. 643 pp.

Knisel, W.G. 1980b. Erosion and sediment yield models - an overview. In: Symposium on Watershed Management, Volume I. American Society of Civil Engineers, pp. 141-150.

Larson, C.L., C.A. Onstad, H.H. Richardson, G. Dale, and K.N. Brooks. 1982. Some particular watershed models. In: C.T. Haan, H.P. Johnson, and D.C. Brakensiek (eds.), Hydrologic Modeling of Small Watersheds, ASAE Monograph 5. ASAE, St. Joseph, Michigan. pp. 409-434.

Lee, J.J., D.L. Phillips, and R.F. Dodson. (in press). Sensitivity of the U.S. corn belt to climate change and elevated C[O.sub.2]: II. Soil erosion and organic carbon. Agricultural Systems.

Leonard, R.A., W.G. Knisel, and D.A. Still. 1987. GLEAMS: Groundwater Loading Effects of Agricultural Management Systems. Transactions of the American Society of Agricultural Engineers 30(5), 1403-1418.

Medawar, P.B. 1968. Induction and Intuition in Scientific Thought. American Philosophical Society, Philadelphia.

Misra, R.K., and C.W. Rose. 1990. Manual for Use of Program GUEST. Division of Australian Environmental Studies Report, Griffith University, Brisbane, Australia 4111.

Morgan, R.P.C., J.N. Quinton, and R.J. Rickson. 1992. EUROSEM Documentation Manual. Silsoe College, Silsoe, Bedford, UK.

Nearing, M.A., G.R. Foster, L.J. Lane, and S.C. Finckner. 1989. A process-based soil erosion model for USDA Water Erosion Prediction Project Technology. Transactions of the American Society of Agricultural Engineers 32, 1587-1593.

Nicks, A.D. 1991. Climate data inputs to models. In: D.B. Beasley, W.G. Knisel, and A.P. Rice (eds), Proceedings of the CREAMS/GLEAMS Symposium, Publication 4, Agricultural Engineering Dept., University of Georgia, Athens pp. 97-107.

Nicks, A.D. 1993. Modeling hydrologic impacts of global change using stochastically generated climate change scenarios. In: Y. Eckstein and A. Zaporozec (eds), Industrial and Agricultural Impacts on the Hydrologic Environment. Proceedings of the Second USA/CIS Joint Conference on Environmental Hydrology and Hydrogeology.

Nicks, A.D., and R.D. Williams. 1993. Estimating water resources of a mixed land use basin under changing climate and environment. In: Proceedings of International Conference on Environmentally Sound Water Resources Utilization, Bangkok, Thailand. pp. II-8 - II-17.

Oreskes, N., K. Shrader-Frechette, and K. Belitz. 1994. Verification, validation, and confirmation of numerical models in the Earth Sciences. Science 263, 641-646.

Parsons, A.J., and A.D. Abrahams. 1992. Preface. In: A.J. Parsons and A.D. Abrahams (eds.), Overland Flow. UCL Press, London, UK.

Phillips, D.L., J.L. Lee, and R.F. Dodson. (in press). Sensitivity of the U.S. corn belt to climate change and elevated C[O.sub.2]: I. Corn and soybean yields. Agricultural Systems.

Quinton, J.N. 1994a. The Validation of Physically-Based Erosion Models - with Particular Reference to EUROSEM. Unpublished PhD thesis, Cranfield University, Silsoe, Bedford, UK.

Quinton, J.N. 1994b. Validation of physically based erosion models, with particular reference to EUROSEM. In: R.J. Rickson (ed), Conserving Soil Resources - European Perspectives. CAB International, Wallingford, UK. pp. 300-313.

Renard, K.G., W.J. Rawls, and M.M. Fogel. 1982. Currently available models. In: C.T. Haan, H.P. Johnson, and D.C. Brakensiek (eds), Hydrologic Modeling of Small Watersheds, ASAE Monograph 5. American Society of Agricultural Engineers, St. Joseph, Michigan. pp. 507-522.

Renard, K.G., G.R. Foster, G.A. Weesies, and J.P. Porter, J.P. 1991. RUSLE - Revised Universal Soil Loss Equation. Journal of Soil and Water Conservation 46(1), 30-33.

Warner, R.E., and B.C. Dysart III. 1980. Erosion modeling approaches for construction sites. In: Symposium on Watershed Management. American Society of Civil Engineers, pp. 465-484.

Wilcox, B.P., and J.R. Simanton. (in press.) Evaluation of the WEPP hillslope model on stable and eroding semiarid woodlands. In: J. Boardman and D.T. Favis-Mortlock (eds), Modeling Soil Erosion. Springer-Verlag NATO-ASI Global Change series.

Williams, J.R. 1985. The physical components of the EPIC model. In: S.A. El-Swaify, W.C. Moldenhauer, and A. Lo. (eds), Soil Erosion and Conservation. Soil Conservation Society of America, Ankeny, Iowa. pp. 272-284.

Williams, J.R., A.R. Nicks, M.A. Nearing, E. Skidmore, and C. Valentin. 1996. Using soil erosion models for global change studies. Journal of Soil and Water Conservation 51 (5).

Wischmeier, W.H., and D.D. Smith. 1978. Predicting Rainfall Erosion Losses. U.S. Department of Agriculture, Agricultural Research Service, Handbook no. 537, 58 pp.

David T. Favis-Mortlock is a research scientist, Environmental Change Unit, University of Oxford, 1a Mansfield Road, Oxford OX1 3TB, UK; John N. Quinton is a lecturer in Soil and Water Management, Department of Water Management, Silsoe College, Cranfield University, Silsoe, Bedford MK45 4DT, UK; and W. Trevor Dickinson is a professor, School of Engineering, University of Guelph, Guelph, Ontario N1G 2W1, Canada. We would like to acknowledge funding from the NATO Office for Scientific Affairs for the NATO Advanced Research Workshop 'Global Change: Modelling Soil Erosion by Water.' Also, many thanks to our GCTE colleagues for their stimulating inputs. In particular, our thanks to those who commented on an early draft of this paper. This paper is a contribution to the Soil Erosion Network of the GCTE, which is a core research project of the International Geosphere-Biosphere Programme.
COPYRIGHT 1996 Soil & Water Conservation Society
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1996 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Special Issue: Global Change & Terrestrial Ecosystems; Global Climate and Terrestrial Ecosystems Core Project
Author:Favis-Mortlock, David T.; Quinton, John N.; Dickinson, W. Trevor
Publication:Journal of Soil and Water Conservation
Date:Sep 1, 1996
Words:5501
Previous Article:Scaling up processes and models from the field plot to the watershed and regional areas.
Next Article:Wetlands potentially exempted and converted under proposed delineation changes.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters