Water retention and availability in Brazilian Cerrado (neotropical savanna) soils under agricultural use: Pedotransfer functions and decision trees

Abstract

Soil water retention evaluation is critical for modeling water flow and solute transport in the vadose zone. The objectives of this study were to evaluate the performance of five data-fitting models for the soil water retention curve, to evaluate the performance of various pedotransfer functions for estimating model parameters, and to evaluate the use of decision trees for estimating water retention and availability in soil. Soil samples were collected from ten sites in the south-southwest central region of Mato Grosso State, in Brazil’s Cerrado biome (neotropical savanna). The sampling included five soil texture classes to determine the soil water retention curve (SWRC) and other soil properties. The mathematical description of SWRC was performed using the Brooks and Corey, Campbell, Libardi, Rogowski, and van Genuchten models. All models showed similar performance for matric potentials between −500 and −1500 kPa, while for higher potentials only the models of Brooks and Corey, Campbell and van Genuchten showed reasonable performance and were recommended for fitting the soil water retention curve from saturation to dryness. Among the best models, Campbell model showed lower performance for sandy soils, while van Genuchten model showed lower performance for sandy clay loam soils. Pedotransfer functions were created to estimate the parameters of the Brooks and Corey, Campbell, and van Genuchten models, but only the function fitted to the Brooks and Corey model showed adequate performance. Decision trees were created for estimating field capacity and permanent wilting point, thus providing a practical, easy-to-use tool that can be used as an alternative to pedotransfer functions in decision-making processes.