Spatial variation in soil and the role of kriging

Abstract

Introduction Soil varies from place to place in such a way that anyone of i ts properties is best considered mathematically as a random function. The semi-variogram of regionalized variable theory can describe such variation in a region. I t can be estimated from data, and simple mathematical models can usually be found to f i t the sample values well. In these circumstances local estimates of soil properties can be made from sample data by kriging. These estimates are optimal in the sense that they are unbiased and have minimum variance. The theory has been applied recently in several case studies to characterize variation in soil in the lateral plane, to perform local estimation and to map soil properties from sample data. I t has also been used to make regional estimates with substantial gains in precision over traditional practice, Experience suggests that the intrinsic hypothesis of stationary mean and variance of differences of terms is adequate in practice in most instances, and that block krlging is l ikely to be the most profitable form of kriging. The procedures are il lustrated using data from a survey of cover loam in Eastern England.