Impact of rainfall distribution on the parameterisation of a soil-moisture balance model of groundwater recharge in equatorial Africa

SummaryRobust calibration of hydrological models, driven by gridded precipitation data derived from either Regional Climate Models or statistical downscaling of General Circulation Models, is essential to the quantitative analysis of the impacts of climate change on catchment hydrology and freshwater resources. Predicted warming in equatorial Africa, accompanied by greater evaporation and more frequent heavy precipitation events, may have substantial but uncertain impacts on terrestrial hydrology. In this study, we examine how the spatial representation of precipitation influences the parameterisation and calibration of a soil-moisture balance model (SMBM) in the humid tropics of equatorial Uganda. SMBMs explicitly account for changes in soil-moisture and partition effective precipitation into groundwater recharge and runoff. The semi-distributed SMBM, calibrated with daily station data over a 15 year period (1965–1979), estimates a mean annual recharge of 104 mm a−1 and mean annual surface runoff of 144 mm a−1. Interpolation of station precipitation by inverse distance weighting produces a more uniform distribution, and a 7% increase, in mean annual catchment precipitation relative to point-based station data. Application of interpolated (gridded), uncorrected precipitation to the SMBM results in an underestimation of runoff and overestimation of recharge by 57% and 52%; respectively whereas use of corrected, gridded precipitation results in an underestimation of recharge and runoff by 10% and 64%; respectively. Recalibration of the SMBM using gridded precipitation data requires a 3% reduction in potential evapotranspiration, a 12% increase in the runoff-coefficient, and an 18% reduction in the rainfall threshold. These values are inconsistent with local, point-based observations of these parameters. Although current efforts seek to improve the distribution and duration of key hydrological measurements (e.g. soil-moisture, groundwater levels) in data-poor regions, the parameterisation of gridded hydrological models remains largely empirical due to the discrepancy between gridded and locally observed hydrological parameters.