Pedotransfer Functions to Predict Soil Moisture Constants in Shrink–Swell Soils of Haveli Tract in Jabalpur District of Madhya Pradesh

Knowledge of soil moisture constants namely field capacity (FC) and permanent wilting point (PWP) is required in many irrigation-related applications. However, data on these moisture constants is seldom available. Pedotransfer functions (PTF) bridge the gap between available and unavailable information. Research on PTFs for estimating soil hydraulic properties has been extensive because measurement of these properties is a time consuming, manpower intensive, tedious process, which is also expensive. PTFs could be either point-specific like estimating water retention (φ) at a specified suction pressure (h) or function-specific like estimating parameters of an h-φ relationship. In this study, point PTFs to predict FC (φ at –33 kPa) and PWP (φ at –1500 kPa) for ‘Haveli’ tract (Jabalpur district, Madhya Pradesh, India) were developed and evaluated for their performance. ‘Haveli’ soils are the agricultural soils that are impounded by rainwater during the monsoon season. Samples were collected and analysed for basic and water retention properties. Regression and artificial neural network (ANN) methods were used for building PTFs. Neural networks proved superior to regression as a PTF building technique. In neural networks use of sigmoidal function as an activation function emerged better in comparison to hyperbolic tangent function. The calibrated PTFs were shown to predict FC and PWP with acceptable reliability. Root mean square error (RMSE) in prediction of FC and PWP was lower than 0.05 m³ m⁻³–an average reported in the literature.