Surrogate estimation of soil water retention characteristics of seasonally impounded soils

Abstract

Measurement of soil hydraulic properties is an arduous task, especially to meet the large scale data requirements. This has prompted numerous studies on surrogate estimation of hydraulic properties. Objective of this study was to build regional hydraulic database and develop pedotransfer functions (PTF) for estimation of water retention characteristics (WRC) of seasonally impounded shrink-swell soils. Soil samples from 41 soil profiles were analyzed in the laboratory for soil-water retention at nine pre-determined suction pressure points and basic soil properties. These data were used for calibrating neural PTFs to predict Van Genuchten (VG) function parameters for prediction of WRC from basic soil properties data. Routine method of deriving and relating retention function parameters to basic soil properties and Neuro-m method resulted in unacceptable PTFs (RMSE > 0.05 m3 m-3) . In place of parameters derived from SWRC, we used ‘Rosetta’ estimates of VG parameters to relate basic soil properties. Resultant hierarchical parametric PTFs with prediction accuracy of RMSE