On the effect of uncertainty in spatial distribution of rainfall on catchment modelling

The development of distributed modelling technology has been hampered by incomplete understanding of spatially-variable processes and the lack of adequate data to characterize spatially varying inputs and state variables. This seems to be particularly true with respect to rainfall fields. This paper examines the effect of uncertainty in spatial estimates of rainfall on predictions of runoff volume, peak runoff and sediment yield. This is achieved by applying a distributed runoff-erosion model to a 6.73 km2 catchment with a dense recording raingauge network in an area dominated by convective air-mass storm rainfall and infiltration-excess overland flow. Results indicate that the density of the raingauge network, the spatial arrangement of individual raingauges within the catchment, and the spatio-temporal characteristics of storm events all greatly influence catchment response predictions. The implication of this is that first, the utility of a spatially distributed approach to model catchment response is compromised by large uncertainties in the spatial distribution of rainfall input. Second, application of distributed catchment models may require knowledge of the spatial variability of rainfall at the same scale used to represent the catchment area and geometry. Third, estimates of areal rainfall using traditional techniques like the Thiessen method may be accurate at the catchment scale but they do not produce accurate estimates of rainfall for distributed catchment modelling.