Multivariate Random Forest for Digital Soil Mapping

• DOI: 10.2139/ssrn.4240513
• Publication Date: 2022-10-08T03:10:43Z
• Authors: Cornelius Stephan...; Gerard B. M. Heuv...; David P. Hofmeyr
• Keywords: Digital Soil Mapping

In digital soil mapping (DSM) maps are usually produced in a univariate manner, that is, each map is independently and therefore, when multiple properties mapped the underlying dependence structure between these ignored. This may lead to inconsistent predictions simulations. For example, organic carbon total nitrogen show unrealistic carbon-nitrogen ratios. last decade production of with machine learning models has become increasingly popular as able capture complex non-linear relationships environmental covariates. However, producing multivariate still lacking requires much investigation DSM. this paper we present combined random forest model. We applied model nitrogen, compared it results two separate models. The comparison was done by means stochastic simulations determined sampling from conditional distributions properties, given covariates, estimated quantile regression forest. superior terms maintaining consequently, also produce more realistic were on basis prediction accuracy. found accuracy comparable performed similar co-kriging kriging