Harvesting spatially dense legacy soil datasets for digital soil mapping of available water capacity in Southern France

摘要: Abstract Although considerable work has been conducted in recent decades to build soil databases, the legacy data from a lot of former survey campaigns still remain unused. The objective this study was determine interest harvesting such for mapping available water capacities (SAWCs) at different rooting depths (30 cm, 60 cm, 100 cm) and maximal observation depth, over commune Bouillargues (16 km2, Occitanie region, southern France). An increasing number auger hole observations with SAWC estimations – 0 2781 were added existing profiles calibrate quantile regression forests (QRFs) using Euclidean buffer distances sites as covariates. first mapped separately layers, outputs pooled estimate required SAWC. uncertainty prediction estimated uncertainties individual layers by an error propagation model first-order Taylor analysis. performances predictions their evaluated 10-fold cross validation that iterated 20 times. results showed use forest fed used covariates considerably augmented (percentages explained variance 0.39 0.70) compared performance classical DSM approach, i.e., QRF solely only environmental 0.04 0.51). analysis revealed also dependent on spatial patterns examined SAWCs limited observational determined holes. best tended provide view overestimation uncertainty. Despite these gains performance, cost-efficiency augmentation not cost efficient because highly time-consuming manual protocol. However, result did account observed gain map details. Furthermore, efficiency could be further improved automation.