Modelling contaminant transport at catchment or regional scale

Most models for contaminant transport have been developed and tested at the scale of the field plot, the lysimeter or even the laboratory soil column. Modelling at the scale of the catchment or region therefore brings some interesting new challenges. The interaction between non-linearity in a model and variance in one of its parameters can cause errors in modelling even small areas of land; the reasons are explained and a test for non-linearity given. It is necessary to know whether this problem will intensify as the area modelled increases, and it is suggested that the variogram provides a means of answering this question. The intrinsic variability of the soil is only one of the factors that have to be considered at catchment or regional scale. At these scales, land use is likely to be the dominant source of variability in the loss of contaminants such as nitrate and phosphate. Simulating land use effects is essentially a data-handling exercise that has to involve the use of geographic information system with models for N turnover and leaching in what is essentially a decision-support system. Questions surrounding the development and use of such systems are discussed. Another extra dimension is introduced at catchment scale and above by the need to consider the subsequent fate of water after it has passed through the soil profile. It is necessary to model its partitioning between that which percolates into ground water and that which moves as surface runoff into streams and other surface waters, and possible approaches are described. The largest land use change in the UK this century has been the ploughing-up of old grassland. A brief account is given of the way the consequences were simulated using a simple model. The results suggest that the ploughing made a very substantial contribution to the nitrate problem in the UK. There is no fully satisfactory way of validating a model used at catchment or regional scale. All that can probably be achieved is to validate parts of the model for much smaller areas and to subject the whole model to a test of its general efficacy. (C) 1998 Elsevier Science B.V.