Analyzing soil enzymes to assess soil quality parameters in long-term copper accumulation through a machine learning approach

Soil contamination by agrochemicals is a big concern for soil health and ecosystem functioning. This is especially true for non-degradable substances like heavy metals (HM) that, because of their long-term use, are reaching significant values today due to soil accumulation. Among agrochemicals, copper (Cu) has been important in fighting fungal diseases on perennial crops for centuries. Laboratory experiments can be useful to understand the highest potential toxic effect of Cu but need to reflect what happens with long-term application in a dynamic and living agroecosystem. This study uses multivariate data analysis and machine learning to investigate long-term Cu accumulation on soil quality parameters, especially soil extracellular enzymatic activities. We collected soil samples from 21 apple orchards in South Tyrol, Italy. The orchards had different concentrations of Cu. We took 315 samples in total and analyzed them for various soil properties. We also measured the concentrations of elements in apple leaves and the activities of soil extracellular enzymes. We depicted the effect of Cu on several enzymatic activities, shedding light on the effect of Cu on the soil microbial communities functionality. Our results show that Cu concentrations in the study area affect only phosphatase activity, showing effects above 60 mg kg⁻¹ of available Cu. Protease activity was positively correlated with Cu, while soil organic matter and management mainly influenced the carbon (C) cycle enzymes. Phosphatase decrease could be of concern for the potential disruption of the Phosphorus (P) cycle in the soil and plays a role in plant nutrition, as seen by P concentration in apple trees' leaves. We demonstrated how machine learning can help interpret complex and multivariate environmental data and overcome some downsides of traditional statistical models.