Using Machine Learning to Predict the Impact of Agricultural Factors on Communities of Soil Microarthropods

With the newly arisen ecological awareness in the agriculture the sustainable use and development of the land is getting more important. With the sustainable use of soil in mind, we are developing a decision support system that helps making decisions on managing agricultural systems and is able to handle both conventional and genetically modified crops as a part of the ECOGEN project. The decision support system considers economical and agricultural factors and actions including crop selection, crop sequence, pest and weed control actions etc. For such decision support system to work, it needs modules that predict results of different agricultural actions. One of the most important factors for sustainable use and fertility of soil is soil flora and fauna. Any change of that community can influence the short or long term soil fertility and soil usability. With soil fauna being one of the most important factors we first need to model it. However, since the function of the individual species is not known, the only action we have is to try and model the community of soil fauna. We start by modelling the community soil microarthropods. For that goal we used machine learning methods regression trees, model trees and linear equations. We identified previous crops and time since different kinds of tillage as the most important factors for the community of soil microarthropods.