Convolutional Neural Networks for Digital Soil Mapping

A framework, built on 'torch', for mapping a soil property from point observations and a stack of aligned rasters with a convolutional network that reads the neighbourhood of each point. It extracts the patches; resamples by spatial blocks or by k-fold nearest neighbour distance matching, after Linnenbrink et al. (2024) ; tunes and selects under a noise floor and the one-standard-error rule; refits the selected configuration under several seeds; and predicts median, mean and interval maps over rasters of any extent. The map's centre is the ensemble median over seeds; the mean corrects its back-transformation by smearing, after Duan (1983) . The intervals are conformal, calibrated three ways and each checked on the test set: on the cross-validated residuals; on a calibration set held out of every fold, as split conformal, after Lei et al. (2018) ; or with every fold's model at each pixel, as CV+, after Barber et al. (2021) -- with points, or whole groups of them, weighing alike, after Dunn et al. (2023) . The area of applicability of Meyer and Pebesma (2021) is mapped beside them. What a model learned is read by permutation, by SHAP values from expected gradients, after Erion et al. (2021) , by SAGE, after Covert et al. (2020) , by accumulated local effects, after Apley and Zhu (2020) , and by refits without each variable. Random forest and multilayer perceptron baselines run under the same folds.