SPLINTEX: A physically-based pedotransfer function for modeling soil hydraulic functions

The determination of soil hydraulic properties is laborious and expensive, especially in large-scale applications. One often used substitute for measured hydraulic properties are pedotransfer function (PTFs) estimates. Most PTFs, however, are statistical models that tend to produce biased results for data outside their –often limited– calibration databases. In addition, most PTFs have been established on data derived from temperate regions causing the question whether such models are applicable to soils in tropical regions. This work aimed to evaluate the performance of the Splintex PTF to predict the hydraulic functions for sandy and clayey soils from several tropical and subtropical Brazilian datasets. Splintex is somewhat unique in that it is based on physical principles using a modification of the Arya-Paris method while allowing the estimation of van Genuchten parameters from limited data. In addition, Splintex has an option to include measured soil water retention points, in principle allowing it to produce accurate estimates for a variety of soils. Estimates by Splintex were compared with the empirical Rosetta PTF, which also has an option to use one (or two) retention points. Estimates by both PTFs were compared to observed retention data and field capacity, available water capacity, hydraulic conductivity, and diffusivity using metrics such as Pearson correlation (r), mean absolute error (MAE) and root mean square error (RMSE). Both Splintex and Rosetta yielded similar results and sometimes produced significant biases in estimated quantities. In the majority of cases it appears that Splintex produced somewhat better estimates than the 2001 version of Rosetta, indicating that Splintex is a viable, physically-based, alternative to estimating hydraulic properties.