Data handling and pattern recognition for metal contaminated soils

In the laboratory sciences good experimental design minimises the effects of any disturbing variables so that hypotheses are amenable to to relatively unambiguous testing. But in the field sciences such variables cannot be controlled and data are inherently variable. Subsequent hypothesis testing must rely on a careful statistical interpretation of noisy data. This paper describes one systematic approach to interpreting the results from surveys of metal contaminated soils. Since contaminating metals are also present naturally in soil, anthropogenic excesses are recognised through statistical tests on the data. The nature of pollution processes also leads to the generation of distinct spatial patterns which may be evaluated through appropriate computergraphic techniques.