Mapping nitrate concentrations in upper groundwater using Random Forest

The aim of this study was to create a map of nitrate concentrations leaching from the root zone of Dutch agricultural soils. We used nitrate data from a national monitoring network and the Random Forest algorithm as prediction and interpolation method.

A large set of spatial auxiliary data, like soil types, groundwater levels and crop types, was used as dependent variables. The explained variance and statistical errors indicate that the interpolation and map visualisation is suitable for interpretation of the spatial variability of the nitrate concentrations in the Netherlands.