Measurement of surface redistribution of rainfall and modelling its effect on water balance calculations for a millet field on sandy soil in Niger.

During rain there can be substantial redistribution of water at the surface of sandy soils in the Sudano-Sahelian zone, because of localised runoff and runon. This results in variable infiltration over a field. Measurements of spatial variability in infiltration and crop growth were made in a millet field at the southern supersite of the HAPEX-Sahel experiment in Niger. Infiltration was calculated from the change in soil water storage measured using a neutron probe at up to 33 locations, before and after rain storms exceeding 10 min. Data were obtained for five storms in 1993 and 1994. Infiltration varied from 0.3 to 3.4 times the recorded rainfall, though more than 80% of the locations had infiltration between 0.6 and 1.2 times the recorded rainfall. There was some consistency between storms, with locations at the extremes of infiltration having consistently high or low infiltration. The amount of infiltration had little discernable influence on crop growth, other than possibly at the very dry and very wet sites, where growth was reduced. The soil water balance model, SWIM, was used to assess the consequences of variable infiltration and crop growth on the partitioning of water losses between evaporation and drainage in 1992. Simulation of variable infiltration suggested that it has relatively little effect on evaporation, but considerable effect on point drainage. Once there was sufficient infiltration to cause drainage (which was achieved in all but extreme runoff areas), there was a linear relationship between any further cumulative infiltration and the annual loss through drainage, with typically more than 70% of any further input being lost as drainage. This linear relationship meant that on a field scale, variable infiltration had minimal effect on drainage, with increased drainage from runon areas tending to be at the expense of reduced drainage from runoff areas