Testing Numerical Models in Geomorphology: How can we Ensure Critical Use of Model Predictions?

Abstract

Geomorphologists use numerical models over a wide range of spatial and temporal scales. How these models are validated and verified depends on the scales involved, not least because of the (non-) availability of some data critical for direct testing. Placing model testing within a formal framework of scale relations is suggested to increase the rigour of such testing. Two examples of numerical models are presented: the first concerns downstream sediment fining in river systems. The SEDROUT model has been tested in terms of its overall performance on three rivers, covering a c.40x range of discharge. At each site, internal components of the model have been validated against independent data. The testing programme for this model is considered to produce an overall evaluation that is sufficiently rigorous for the model to be more than a heuristic tool. The second example is of long-term landscape evolution models (surface process models, with or without coupled tectonic models) that have been used both to investigate specific sites and to study general principles of landscape evolution. Direct testing of such models is difficult on account of the long timescales and large spatial scales involved. A combination of imaginative formulation of model tests and the use of new technologies to develop novel model tests is advocated as providing potentially robust evaluation of such models. Both examples lead to the conclusion that field- and model-based studies need to be linked and that research programmes require greater integration than ever before.