Evaluation of field-scale and catchment-scale soil erosion models

One of the tasks of the International Geosphere-Biosphere Programme-Global Change and Terrestrial Ecosystems (IGBP-GCTE) Soil Erosion Network is to de termine the suitability of modelling approaches for the estimation of soil erosion under global change. To achieve this, current erosion models are be ing evaluated in a series of GCTE meetings. This paper presents a synthesis of results from the first two meetings, which focused upon models for soil erosion by water at field and catchment scales, respectively. Apart from t his comparison, discussions were held on model use and quality. The main co nclusions from these discussions are included here as well. For both sets o f evaluations, common datasets which had been split into a 'training set' a nd a 'testing set' were prepared and distributed to the participating model lers. For the 'testing set' data, measured values for runoff and sediment y ield (field-scale models) or for erosion only (catchment-scale models) were withheld from the modellers. The data used for the field-scale evaluation represented 73 site-years from seven sites in three countries: six field-sc ale erosion models took part in the evaluation. For the catchment-scale eva luation, data for 10 events on a 40 ha catchment in the Netherlands was use d to evaluate seven catchment models. Conclusions from both field-scale and catchment-scale exercises include the following: calibration is desirable for many models, and necessary for some. Calibrati on is most effective if the event(s) to be estimated lie inside the range o f calibration events; total discharge is generally better predicted than peak discharge and both are better predicted than sediment discharge; for continuous-simulation models, long-term average results are better simu lated than results for individual time periods. In general, results are les s good for shorter time periods, although there are exceptions; while for certain events models may not perform. well (absolute results), c orrelation coefficients between observed and predicted values are acceptabl e (relative results); at a catchment scale, the predicted spatial runoff pattern is as important as a correct prediction of the net output; from the discussions it was clear that additional 'soft' information, in pa rticular regarding the change in soil structure as a result of agricultural activities and/or climate, greatly improves the quality of input data and model results. (C) 1999 Elsevier Science B.V. All rights reserved.