Some comments on the use of spatial interpolation techniques in studies of man-induced river channel changes

Abstract

Many studies of river channel changes induced by man have assumed that down-stream changes in channel size parameters relative to stream discharge or its surrogates behave as a simple power function over a wide range of spatial scales. This study and some previous empirical investigations, however, suggest that this assumption is doubtful. The pattern of variability of channel size along the River Elemi, Nigeria, and within the drainage basin, is briefly outlined and discussed in relation to catchment hydrology, geomorphology and channel perimeter cohesion. The relationship of channel capacity to stream discharge is scale dependent. Consequently, the extrapolation of observed upstream allometric relations to downstream reaches produces significant differences between actual field measured values and the estimated channel capacities of the downstream reaches, the latter being on average 210 per cent higher than the former. Extrapolation from the segmented regional relationship for the spatial scale of the downstream reaches, however, produces statistically valid estimates of channel capacities, the estimated values being on average only 21 per cent higher than the measured values. It is argued, therefore, that spatial interpolation for the purpose of quantifying man-induced river channel changes should entail the use of segmented regional relationships rather than the extrapolation of upstream relations to modified downstream reaches. It is necessary, therefore, to focus greater attention on detailed analysis of the spatial variability of equilibrium channel geometry through the use of such geostatistical techniques as fractal and spectral analyses.