Regional estimation of soil properties from local observations

This paper presents four diierent approaches for estimating the average value of a soil property over a region, say a province, from point data: 1) arithmetic average, 2) declustered mean, 3) block kriging, and 4) stochastic simulation. The last two techniques are based on geostatistics which allows one to account for the pattern of spatial dependence of observations while providing a measure of the uncertainty attached to the estimate. The diierent techniques are illustrated using an environmental data set related to heavy metal contamination of a 14.5 km 2 area in the Swiss Jura