A hybrid data–model approach to map soil thickness in mountain hillslopes

Abstract. Soil thickness plays a central role in the interactions between vegetation, soils, and topography, where it controls retention release of water, carbon, nitrogen, metals. However, mapping soil thickness, here defined as mobile regolith layer, at high spatial resolution remains challenging. Here, we develop hybrid model that combines process-based empirical relationships to estimate heterogeneity with fine (0.5 m). We apply this two aspects hillslopes (southwest- northeast-facing, respectively) East River watershed Colorado. Two independent measurement methods – auger cone penetrometer are used sample 78 locations calibrate local value unconstrained parameters within model. Sensitivity analysis using reveals diffusion coefficient hillslope modeling has largest sensitivity among all input parameters. In addition, our results from both sampling show that, general, northeast-facing deeper layer than southwest-facing hillslope. By comparing estimated machine-learning approach model, provides higher accuracy requires less data. Modeling further reveal slightly faster surface erosion rate production hillslope, which suggests relatively dense vegetation cover drier soils on slopes influence properties. With seven total for calibration, can provide realistic map small amount dataset approach. Integrating statistical not only thorough understanding fundamental mechanisms prediction but also integrates strengths approaches approaches.