Soil thickness prediction models: Types, accuracy and influencing factors

Abstract

Soil thickness is a key soil property that critically influences geomorphological processes, hydrological dynamics, and related domains. It is affected by many factors and has strong spatial heterogeneity; thus, accurate soil thickness prediction is typically difficult to achieve. In this paper, the types, accuracies and influencing factors of soil thickness prediction models were reviewed. We collected 123 publications related to soil thickness prediction worldwide and performed statistical analysis on 732 observations. The results showed that the spatial interpolation method had the highest prediction accuracy among the various models. Furthermore, the highest prediction accuracy among the different study area sizes was achieved for areas $\le$ 25,000 km². The sample density had a significant positive effect (P < 0.05), whereas the observed range of soil thickness and the number of predictors had significant negative effects (P < 0.05). However, under different methods and area sizes, the effects of the number of predictors, observed range of soil thickness, sample density, and DEM resolution on prediction accuracy needs to be specifically analyzed. Future research on soil thickness prediction should prioritize the integration of methods, uncertainty assessment of predictive results, and interpretability of predictive results.