Using remote sensing and geostatistics to map 137Cs-derived net soil flux in south-west Niger

A small (760 m x 740 m) study area in south-west Niger was sparsely sampled for (137)Cs which was modelled to provide measurements of net soil flux. A variogram of net soil flux was computed and a model fitted using weighted least squares. The parameters of the variogram were used to solve kriging equations to provide net soil flux estimates and estimation variances over 20 m X 20 m blocks throughout the study area. At the sample locations net soil flux was substituted for the block-averaged net soil flux estimates in order to compute additional auto- and cross-variograms of intensively sampled SPOT satellite data (20 m x 20 m). These variograms were modelled as before and the parameters used to solve co-kriging equations to provide net soil flux estimates and estimation variances. The co-kriging estimation variances were considerably smaller than those for ordinary kriging, suggesting that problems of accurately mapping net soil flux can be overcome by making use of the additional sampling intensity of the spatially interdependent data in SPOT band XS3. This methodology has considerable potential for accurate and cheap net soil flux mapping over very large areas with limited (137)Cs ground measurements. (C)1998 Academic Press Limited.