Mapping soil hydraulic properties using random forest based pedotransfer functions and geostatistics

Abstract

Abstract. Spatial 3D information on soil hydraulic properties for areas larger than plot scale are usually derived with indirect methods due to lacking measured information on those. Soil hydraulic properties are calculated with applying pedotransfer functions (PTFs) – which describe the relationship between the desired soil hydraulic parameter and easily available soil properties determined on a soil hydraulic point dataset – on available soil maps. Our aim was to analyse difference in performance and spatial patterns between soil hydraulic maps derived with indirect (using PTFs) and direct (geostatistical) mapping methods. We performed the study on Balaton catchment in Hungary, where density of measured soil hydraulic data fulfils the requirements of geostatistical methods. Maps of saturated water content (THS), field capacity (FC) and wilting point (WP) for 0–30, 30–60 and 60–90 cm soil depth were prepared. PTFs were derived with random forest method on the whole Hungarian soil hydraulic dataset (MARTHA: soil chemical, physical, taxonomical and hydraulic information of some 12000 samples) complemented with information on topography, climate, parent material, vegetation and land use. As a direct method random forest combined with kriging (RFK) was applied on 359 MARTHA soil profiles located in the Balaton catchment. There was no significant difference between the direct and indirect methods in case of six out of nine maps having root mean squared error values between 0.052 and 0.074 cm3 cm−3, which is in accordance with the internationally accepted performance of hydraulic PTFs. The PTFs based mapping method performed significantly better than the RFK for the THS at 30–60 and 60–90 cm soil depth, in case of WP the RFK outperformed the PTFs at 60–90 cm depth. Difference between the PTF based and RFK mapped values are less than 0.025 cm3 cm−3 for 65–86 % of the catchment. In RFK uncertainty of input environmental covariate layers is less influential on the mapped values which is preferable. In the PTFs based method the uncertainty mapping of the soil hydraulic properties is less computational intensive. Detailed comparison of the maps derived by the PTF based and the RFK is presented in the paper.