Abstract
The objective of this study was to test the practicability of defining hydrologic response units as combinations of soil, land use and topography for modelling infiltration at the hillslope and catchment scales. In an experimental catchment in the East African Highlands (Kwalei, Tanzania), three methods of measuring infiltration were compared for their ability to capture the spatial variability of effective hydraulic conductivity: the constant head (CH) method; the tension infiltration (TI) method; and the miniโrainfall simulation (RS) method. The three methods yielded different probability distributions of effective hydraulic conductivity and suggested different types of hydrologic response units. Independently from these measurements, the occurrence of infiltrationโexcess overland flow was monitored over an area of 6 ha by means of overland flow detectors. The observed pattern of overland flow occurrence did not match any of the patterns suggested by the infiltration measurements. Instead, clusters of spots with overland flow were practically independent from field borders. Geostatistical analysis of the overland flow confirmed the absence of spatial correlation for distances over 40 m. The RS method yielded the pattern closest to the observations, probably because the method simulated better the processes that trigger infiltrationโexcess overland flow, i.e. soil sealing and infiltration through macroporosity. The RS hydrologic response unit correlated significantly with observed overland flow frequency. However, the location of clusters and โhot spotsโ of overland flow remained largely unexplained by land use, soil and topographic variables. It is concluded that using such landscape variables to define hydrologic units may create artificial boundaries that do no correspond to physical realities, especially if the stochastic component within hydrologic units is neglected. Copyright ยฉ 2005 John Wiley & Sons, Ltd.