A taxonomy for Random Forest in the spatial regression framework

Random Forest (RF) is a widely recognized algorithm adopted in various fields due to its flexibility in modelling the response-predictors relationship, even in the presence of strong non-linearities. However, in environmental applications, the phenomenon of interest may exhibit spatial and/or temporal dependence that cannot be explicitly considered in the standard RF. In this regard, we propose a taxonomy to categorize strategies proposed in the scientific literature that attempt to incorporate spatial information into regression RF. These strategies are classified as Pre-, In-, and/or Post-processing, depending on the stage of analysis at which the spatial information is included. This work falls inside the Agriculture Impact On Italian Air (AgrImOnIA) project.