Data correlation structure controls pedotransfer function performance

Abstract

Soil hydraulic properties are essential for modeling and evaluating several soil–water processes. The use of pedotransfer functions (PTFs) is a viable alternative in the analysis of these processes due to the high costs or difficulty in measuring these properties. We aimed to (1) examine the performance of a physically-based PTF and a pool of empirical PTFs from both temperate and tropical climates to estimate soil water retention under subtropical climate and (2) performed correlation analysis between PTF inputs, outputs and estimation residuals to help gain understanding why some PTFs are more successful than others in a given study area. The study was carried out in the Pelotas River Watershed, Southern Brazil, where samples were taken at 100 locations in the 0–0.20-m soil layer along a 25-km-spatial transect. For each point, clay, silt, fine sand, total sand, and organic carbon contents were measured along with bulk and particle density, macro, micro, and total porosity, and the soil water retention curve. The PTFs were used to estimate field capacity, permanent wilting point, and the available water (AW) content. The performance of PTFs from different climate zones was mixed, and similarity in the data correlation structure between PTF development and application data sets appeared to be a good predictor of their predictive power. There was no clear grouping in such correlation structures within climate-zones, and we conclude that the often claimed geoclimatic difference or similarity between an empirical PTF’s origin and its application site is not, or at least not the sole driver of a PTF’s expected performance.