SPATIAL INTERPOLATION OF SOIL-MOISTURE RETENTION CURVES

Methods for interpolating soil moisture curves at unrecorded sites fro m measurements of soil water content at fixed pressure heads at sample points are investigated. A spatial prediction procedure was devised w ith two steps: interpolation of the retention data followed by fitting the Van Genuchten function to the interpolated values. For the interp olation, four different methods were considered. Two are based on a te ssellation into areas of equal values, namely the nearest neighbour an d the soil classification. The other two were autokriging and cokrigin g. The performance of these methods were compared in an area of 92 ha of alluvial soil where soil changes were gradual. Mean deviations and mean square deviations between predicted and observed retention data a t validation sites were compared for two sampling densities. Kriging p erformed best in both cases. Cokriging was only slightly better than a utokriging, and the latter is recommended for use. The performance of the nearest neighbour method diminished markedly with decreasing sampl ing density. Conversely, the performance of soil classification dimini shed only slightly, and this technique appeared as a reasonable altern ative to kriging for sparse samples.