A Statistical Approach to the Evaluation and Modelling Of Contamination in an Agro-Ecosystem

Abstract

Soil protection and preservation are crucial to continuing sustainable development of our society. Multidisciplinary research on bioindicators is under way to better understand the impact of chemical or physical stress on communities of soil organisms. In this context, our purpose is to develop a statistical discriminant approach to evaluate and analyse the stability of soil ecosystems in a farming context in northern France. Thirty-six physical, chemical, microbiological and biochemical parameters were measured, on two sites with conventional farming and pasture. In order to mimic the conditions of agricultural practices, soil microcosms were contaminated with minute quantities of copper and the same parameters were measured. The impact of agricultural practices and the analysis of relations between the microbiological, biochemical and physicochemical soil parameters indicated strong overlapping of the fungal, bacterial and functional soil descriptors. The results show that the Random Forests method is most appropriate for the available experimental data in this analysis. This statistical approach is scalable to a wider panel of variables and is applicable to situations with heavier contamination and more human presence.