Density of soil observations in digital soil mapping: A study in the Mayenne region, France

• Published in: Geoderma Regional Vol. 24; p. e00358
• Main Authors: Loiseau, Thomas, Arrouays, Dominique, Richer-de-Forges, Anne C., Lagacherie, Philippe, Ducommun, Christophe, Minasny, Budiman
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2021; Elsevier
• Subjects: Agricultural sciences; calibration; data collection; Digital soil mapping; France; kriging; Life Sciences; Multiple soil classes; particle size; particle size distribution; prediction; Prediction performance; Sampling density; Sampling strategy; silt; Soil study; topsoil; Topsoil particle-size distribution; uncertainty; France; Sampling strategy; Prediction performance; Digital soil mapping; Sampling density; Multiple soil classes; France; Topsoil particle-size distribution
• DOI: 10.1016/j.geodrs.2021.e00358
• ISSN: 2352-0094

The density of soil observations is a major determinant of digital soil mapping (DSM) prediction accuracy. In this study, we investigated the effect of soil sampling density on the performance of DSM to predict topsoil particle-size distribution in the Mayenne region of France. We tested two prediction algorithms, namely ordinary kriging (OK) and quantile random forest (QRF). The study area is a region of ~5000 km2 with the highest density of field soil observations in France (1 profile per 0.64 km2). The number of training sites was progressively reduced (from n = 7500 to n = 400, corresponding to 1 profile per 0.7 km2 to 1 profile per 13 km2) to simulate the different density of observations. For OK and QRF, we tested random subsampling for splitting the data into training and testing datasets using k-fold cross validation. For QRF we also tested conditioned Latin hypercube sampling based on the point coordinates or the covariates. The results indicated that, with increasing density of observations, OK performed as well or even better than QRF, depending on the particle-size fraction. For silt prediction, OK was systematically better than QRF. However, the prediction intervals were much larger for OK than for QRF, and OK did not seem to estimate uncertainty correctly. Overall, the performance indicators increased with the density of observations with a threshold at about 1 profile per 2 km2 which suggests that the main limitation of DSM prediction accuracy using QRF is the amount of data collected in the field, not the type of calibration sampling strategy. Future DSM activities should focus on gathering more field observations. •Topsoil particle-size distribution was predicted using digital soil mapping in a French region.•The density of soil observations was the main determinant of prediction accuracy.•The predictions intervals were larger for ordinary kriging than for quantile random forest.•The performance increased with the training density with a threshold of ca 1sample/2 km2.