Prediction of the soil-water characteristic curve of Brazilian soils using general-purpose pedotransfer functions

Abstract

Several predictive models have been proposed in the literature to estimate soil-water characteristic curves (SWCCs) using basic soil properties. However, most of these models were not originally calibrated for tropical soils, such as bimodal materials commonly found in Brazil. To assess the applicability of some of these models to Brazilian tropical soils, which often present fine-particle cementation and lateritic behavior, this study evaluates four physico-empirical approaches: Arya & Paris (1981), Arya & Dierolf (1989), Aubertin et al. (2003), and Wang et al. (2017). A database containing information on 95 Brazilian soils was compiled and used as the basis for the analysis. The results showed that, in their original form, none of the evaluated models consistently represented the SWCCs of these soils with sufficient accuracy. Due to the models’ poor performance, specific geotechnical concepts not incorporated in the original formulations were examined and found to significantly influence the predictions. This analysis revealed that the Arya & Paris (1981) model holds promising potential for estimating the SWCC, provided that particle aggregation is addressed by combining grain-size distributions obtained under different dispersion states. Accordingly, this article proposes guidelines for the enhanced application of the Arya & Paris (1981) model in Brazilian geotechnical contexts, with the aim of improving predictive accuracy and broadening its applicability.