AN INVESTIGATION INTO THE EFFECT OF SPATIAL SCALE ON THE PERFORMANCE OF A ONE-DIMENSIONAL WATER-BALANCE MODEL

Land surface parameterizations used in atmospheric general circulation models (AGCMs) are one dimensional representations of the transfer of energy and water vapour at the earth's surface over grid squares that are typically of the order of 10(4) km(2). It is known, however, that climate variables and land surface characteristics vary over scales o f orders of magnitude less than this. It is therefore important that t he effects of grid-scale aggregation are understood and appropriate me thodologies for incorporating the significant effects are developed. I n this paper, an established two-layer bucket scheme is used as a vehi cle to investigate the effects of aggregating spatially variable rainf all and soil moisture fields. The model is applied to two large UK cat chments, the rivers Severn and Thames, at a daily time-scale using a g rid scale of 40 km. A comparison with the observed river discharges fo r the period 1981 to 1990 shows that the model can simulate runoff to within 1 and 13% for the Severn and Thames catchments, respectively. D istribution functions were introduced into the model to represent sub- grid variability of rainfall and available soil water capacity and so remove the bias associated with spatial averaging. Distribution functi ons also proved to be a useful tool for analysing the key controls on runoff generation at the grid scale and so aid the development of larg e scale grid-based models. (C) 1997 John Wiley & Sons, Ltd.