A Comparison of Geostatistical Simulation Approaches for Estimating the Spatial Uncertainty of Soil Texture

Abstract

Soil texture is often used in environmental modelling approaches for predicting the variation of more difficult-to-measure input properties, like soil hydraulic properties. This paper investigates on a case study sequential Gaussian and indicator simulation algorithms for estimating the spatial uncertainty of texture. It also evaluates whether the texture obtained on purposively chosen soil profiles during soil survey are adequate for conditioning the texture simulations. The analysis of the reproduction of model statistics by the simulated realizations and the accuracy of the predicted local probability distributions shows that the performance of the Gaussian algorithm is the most stable, and that randomly sampled data should be preferred for conditioning the simulations instead of the data measured on the soil surveyor's soil profiles, which underestimate the actual spatial variability of soil properties.