Geostatistical and multivariate statistical methods for the assessment of polluted soils—merits and limitations

The complexity and the large variance of environmental data sets limit the use of common statistical methods for the assessment of the state of pollution. Therefore, the application of geostatistical and multivariate statistical methods is recommended. In principle, both types of statistics are able to detect spatial or temporal structures in data sets. The merits and limitations of these statistical methods shall be demonstrated for the investigation of three very different examples of soil pollution. The first case study is characterized by a distinct spatial structure and a relatively large number of samples. Both geostatistical and multivariate statistical methods are well suited for the characterization of the state of pollution. The second example is typical for a case study under practical and economic limitations. In this case it is possible to describe the polluted area semiquantitatively by means of multivariate statistical methods. The third data set is characterized by a relatively diffuse distribution of the contaminants in an old uranium mining waste dump. Methods of homogeneity testing can be used as an alternative to geostatistical and multivariate statistical methods.