Using a rainfall simulator and a physically based hydrological model to investigate runoff processes in a hillslope

A large rainfall simulator (LRS, area 970 m(2)) is being used to measure flow, sediment and contaminant runoff from a pasture hillslope in the North Island of New Zealand. Results from a winter experiment corresponding to a I in 8 year storm were used to calibrate a complex, physically based hydrological model (SHETRAN, Ewen et al., 2002). The ability of SHETRAN to reproduce firstly the observed runoff hydrograph and secondly the soil moisture response, was assessed. Surface runoff was widely distributed over the simulator plot but was not uniform because of uneven topography and possibly because of spatial variations in soil properties. Most model parameters were determined from field measurements, but two (vertical hydraulic conductivity and surface friction) were 'calibrated' by matching the observed and predicted runoff hydrograph. The calibrated model had a coefficient of determination (R-2) of 98%, and reproduced accurately the shape of the hydrograph and the total volume of runoff. The model predicted soil moisture that matched TDR probe measurements within the calibration uncertainties, at one location, and underestimated the reduction in soil moisture following the end of rain, possibly indicating that the calibrated value of vertical hydraulic conductivity was too low. This study assumed no-flux boundary conditions under and around the LRS, and while this approximation may be acceptable when modelling a single storm, vertical and downslope drainage will need to be quantified in future longer-term applications. From both field data and model simulations, it was concluded that surface runoff occurred principally as a result of the infiltration excess (IE) mechanism. Saturation excess (SE) and variable source area (VSA) were rejected as the principal generation mechanisms. Copyright (c) 2005 John Wiley & Sons, Ltd.