ABSTRACT
Depth to Bedrock (DTB) is required by the specification of Globalsoilmap. This paper presents a global estimation of DTB. About 130,000 Soil profiles data and 1.6 million boreholes were combined as observations. Additional pseudo-observations were generated by expert knowledge for deserts and steep area. These observations were overlaid with 155 soil-related covariates. The predictions were generated at the spatial resolution of 250 m as an ensemble prediction of random forest and Gradient Boosting Tree. The cross-validation shows that the models explain 59% for absolute DTB and 34% for censored DTB (depths deep than 200 cm are predicted as 200 cm). Visual comparisons of predictions with regional detailed maps show that the spatial patterns match well in general. Limitation of the data set and extrapolation in data spare areas should not be ignored in applications. More borehole drilling logs are needed in under-represented areas.
