Evaluation of modelled spatially distributed predictions of soil erosion by water versus field-based assessments

Policy makers concerned about soil erosion and its impacts need good quality information on which to base their decisions. There is a trend toward using erosion models to aid such decision making. Such models are based on data obtained from experimental plots. The theoretical results need to be compared with information gained from monitoring erosion in the field to assess if theory accords with reality. Data from the Minimum Information Requirement version of the Water Erosion Prediction Project model (MIRSED) are compared to information gained from field monitoring over a 5-year period (1982-1986) in 11 localities widely spread throughout England and Wales. Two of the localities, Gwent and Shropshire, are examined in detail. The model seriously over predicts erosion, both in amount and extent. Also, the statistical distributions of the data values are different. The model predicts erosion will happen where it does not. The reasons why the two assessments of erosion differ greatly are explored. This comparison shows there is an urgent need to develop models which incorporate information gained from field-based observations. Until better models are devised, policy makers and decision takers should treat the results of modelling exercises with great caution. (c) 2005 Elsevier Ltd. All rights reserved.