Sampling, Estimating and Understanding soil Pollution

Abstract

Soil pollution and environmental protection have replaced soil fertility and agricultural production as the principle concerns of soil surveys in Europe in recent years. Is the soil polluted by specified substances? If so by how much? Where is the soil polluted? Is it getting worse (or better)? What is or was the source of pollution? The first four questions can be answered only on a statistical basis, and in the absence of written records even the last can depend on the interpretation of statistical evidence. The first two will usually be in the context of a defined region, no particular form of spatial variation is implied, the answers will be global ones for the region as a whole, and classical (design-based) sampling and estimation should serve well. The third question demands local answers requiring spatial prediction, and in most instances it is best answered geostatistically, based on a model of random variation. Monitoring the soil (the fourth question) may be based on either design or a model. It must be sensitive to detect change in a short time, a large sample is likely to be necessary, but bulking and pairing samples should increase sensitivity and bring economies. Suitable design can lead to inferential tests. The last question brings into play an unlimited statistical repertoire, including geostatistics and other forms of spatial analysis, many forms of multivariate analysis, and the testing of specific factors.