Cokriging point data on moisture deficit

Existing computer calculations by simulation of the moisture deficit (MD) were used for 500 point observations in an area of 404 ha with sandy soils in the Netherlands. The statistical prediction techniques kriging and cokriging were used to predict MD-values in 100 points selected at random from the 500 available points. The MD-data at the remaining 400 points were used for the predictions. The mean variance of prediction error (MVPE) and the mean squared error of prediction (MSEP) decreased only slightly when kriging is compared with cokriging using the mean highest water-table level as a covariable. The number of MD-values used in cokriging could, however, be reduced from 400 to 160 with only a small loss of accuracy in using 400 observations on the mean highest water-table (MHW) as a covariable. As the MD-variables being considered are four times as expensive to determine as the covariable, which is routinely estimated during soil survey, this also represents a considerable reduction of costs. Cokriging can thus be a useful technique to more effectively utilize available soil survey information.