Assessment of predicted and measured hydraulic parameters of different textured soils of Haryana

Abstract

The hydraulic properties i.e., the water retention curve and hydraulic conductivity are basic input parameters for water and solute flow simulation studies. Direct measurements of these properties are time-consuming and expensive due to spatial and temporal variations. Numbers of empirical methods have been developed to predict the soil hydraulic properties. To evaluate the applicability of these predictive models, an extensive comparison between measured and predicted observations results is needed for a wide range of soils. Therefore, soil samples were collected from twenty different locations over nine districts in the Haryana state belonging to six textural classes i.e., sand, loamy sand, sandy loam, loam, silty loam, and sandy clay loam. The study was carried out using the van Genuchten (VG) and two ANN models (ANN 1 and ANN 2) to estimate hydraulic parameters for water retention curves and comparing to the field measured retention curves data. Statistical analysis was carried out based on root mean square error (RMSE) and coefficient of determination (R2) to validate the performance of the three model approaches. The results showed that predicted values by VG model were closer to the measured values (R2 ranging between 0.963 and 0.995 and RMSE between 0.012 and 0.026 cm3 cm-3). The second approach ANN 1 (sand, silt, clay %, bulk density, soil moisture content at 0.3 bar and 15 bar) performed well having R2 ranging between 0.792–0.986 and RMSE ranging between 0.016-0.065 cm3 cm-3 as compared to ANN 2 (sand, silt, clay %, and bulk density) with R2 ranging between 0.477-0.788 and RMSE 0.040-0.135 cm3 cm-3. The present study concluded that the VG model provides a better simulation of soil water content than ANN model.