Soil Depth Map for Sri Lanka
Description
Although harmonised profile data containing depth information are available, no attempt has been made to comprehensively map this important soil property across Sri Lanka. Hereby we provide the depth map resulting from the kriging map (OK), kriging variance map (OK_var) and random forest map (RFR). The resulting map can be used to provide essential information for environmental models that are sensitive to soils with variable depth.
Files
Steps to reproduce
We compared the ability of kriging and machine learning models to assess and reproduce the spatial variability of soil depth across the country. The result of 10-fold cross-validation showed that both methods could map the variability to a limited degree. Kriging methods tend to reduce the variability to produce smoother patterns, while random forest methods could capture the range of all possible values for soil depth without the need for any spatial random process modelling. Random forest method has the advantage of less pre-processing requirements if good number of covariates are available. Kriging in contrast does not perform well with the presence of covariates.