Statistical analysis of the effects on overland flow of spatial variability in soil hydraulic conductivity

The effects of spatial variation of the saturated hydraulic conductivity (K(s)) of the soil on the variation of overland flow were tested by analysing 2000 synthetic rainfall-runoff events, all generated from real, observed rainfall events but with runoff modelled by a two-dimensional distributed model using different spatially variable K(s) fields in a small (12 ha) agricultural catchment. The purpose is to determine the influence of spatial variation in K(s) on runoff generation. The statistical measures used to describe the variation in the generated K(s) were its coefficient of variation and correlation length. Both of these had two levels of typical values obtained from field measurements in other studies. The storms were analysed at a general event level, first using simple graphical and statistical methods and then using analysis of variance (ANOVA). The observed scale of the spatial variation of K(s) does cause statistically significant variation in overland flow. The graphical analysis showed that the first flow peak in a multi-event storm had the largest variation and that differences were greater in the rising part of the hydrograph than in its recession. The greatest variation in overland flow was produced by the combination of the greater coefficient of variation and the longer correlation lengths. The smallest variation in overland flow was produced by the combination of the smaller coefficient of variation and the shorter correlation lengths. ANOVA showed that the coefficient of variation and correlation length alone did not explain all the variation of the total flow. ANOVA was not very useful due to the many restrictive assumptions that were not satisfied by the nature of the data and therefore analysis methods with less restrictive assumptions need to be tested.