Modelling sediment transfer in Malawi: Comparing backpropagation neural network solutions against a multiple linear regression benchmark using small data sets

The recent growth in neural network hydrological modelling has focused on t he provision of river flow estimates of one kind or another. Little or no s cientific research has been undertaken to assess the potential benefits for modelling sediment transfer. Some initial pathfinder experiments were ther efore conducted to assess the competence of a backpropagation network to pr oduce a combined model of sediment transfer occurring under different types of agriculture and land management conservation regimes. The results of th is investigation demonstrate that a neural network solution is able to exce ed the limitations of traditional multiple linear regression. The potential to create multiple solutions at different levels of generalisation and rob ust solutions that can be transferred to unknown catchment types is illustr ated. (C) 2000 Elsevier Science Ltd. All rights reserved.