Calibration and evaluation of pedotransfer functions to estimate available water capacity of seasonally impounded shrink-swell soils of central India

Abstract

Pedotransfer functions (PTF) to estimate available water capacity of seasonally impounded shrink-swell soils of central India are presented. Performance of the calibrated PTFs is compared with that of ‘Rosetta’ a widely used general PTF. Available information on soil properties contained nine point soil water retention data for 175 samples measured at varied potentials, textural composition, bulk density and organic carbon content. Nine widely used water retention functions proposed by different researchers were fitted to the measured data and evaluated for efficacy to describe water retention characteristics (WRC). Of the nine functions evaluated, Brooks-Corey, van Genuchten, and Campbell functions were recommended for describing WRC of these soils. We present point PTFs to estimate available water capacity (AWC) using two approaches-regression and artificial neural networks (ANN). Point estimation PTFs were calibrated for water contents at −33 and −1500 kPa and consequently AWC. Performance evaluation with root mean square error (RMSE) criteria suggested that ANN based PTFs were better than regression PTFs. Performance evaluation of ‘Rosetta’ suggested its limited applicability for the study area. Region-specific PTFs to predict AWC were recommended. Increasing the number of predictor variables improved performance of neural PTFs and ‘Rosetta’.

Keywords:

• regressions
• soil
• water
• moisture