Using data mining to predict soil quality after application of biosolids in agriculture

The amount of biosolids recycled in agriculture has steadily increased during the last decades. However, few models are available to predict the accompanying risks, mainly due to the presence of trace element and organic contaminants, and benefits for soil fertility of their application. This paper deals with using data mining to assess the benefits and risks of biosolids application in agriculture. The analyzed data come from a 10-yr field experiment in northeast France focusing on the effects of biosolid application and mineral fertilization on soil fertility and contamination. Biosolids were applied at agriculturally recommended rates. Biosolids had a significant effect on soil fertility, causing in particular a persistent increase in plant-available phosphorus (P) relative to plots receiving mineral fertilizer. However, soil fertility at seeding and crop management method had greater effects than biosolid application on soil fertility at harvest, especially soil nitrogen (N) content. Levels of trace elements and organic contaminants in soils remained below legal threshold values. Levels of extractable metals correlated more strongly than total metal levels with other factors. Levels of organic contaminants, particularly polycyclic aromatic hydrocarbons, were linked to total metal levels in biosolids and treated soil. This study confirmed that biosolid application at rates recommended for agriculture is a safe option for increasing soil fertility. However, the quality of the biosolids selected has to be taken into account. The results also indicate the power of data mining in examining links between parameters in complex data sets.

Keywords: