Process controls of water balance variability in a large semi-arid catchment: downward approach to hydrological model development

The process controls on water balance are examined at the annual, monthly and daily scales. A systematic 'downward' approach for the formulation of models of appropriate complexity is presented based on an investigation of the climate, soil and vegetation controls on water balance. Starting with a simple model, complexity is added in steps, with the models tested progressively against signatures of runoff variability at each time scale. The inter-annual variability of runoff is the first signature considered, followed by the intra-annual (mean monthly) variation of runoff. The flow duration curve is the third key signature and is used to test predictions of the daily water balance model. These analyses are carried out using observed data from the Collie River Basin in Western Australia. At the annual time scale, a simple water balance model including saturation excess overland flow and evaporation is found adequate, provided spatial variability of soil depths and rainfall are introduced through multiple buckets. At the monthly time scale, additional processes are required-the key process is subsurface runoff, but in our case we also separated total evapotranspiration into bare soil evaporation and transpiration to represent the heterogeneous vegetation cover. At the daily time scale, inclusion of non-linearity in the storage-discharge relationship for subsurface runoff generation was important, and more crucially, the inclusion of a deeper groundwater store to capture prolonged low flows was important. Model predictions were very sensitive to the assumed distribution of soil depths, both within each subcatchment, and between subcatchments on a regional basis. Streamflow routing was important for large catchments to capture high flows. The overall conclusion is that in this semi-arid catchment. spatial variability of soil depths appear to be the most important control on runoff variability at all time and space scales, followed by the spatial variability of climate and vegetation cover. (C) 2001 Elsevier Science B.V. All rights reserved.