Evaluation of Water Retention Functions and Computer Program “Rosetta” in Predicting Soil Water Characteristics of Seasonally Impounded Shrink–Swell Soils

Abstract

Soil water retention is a critical factor influencing irrigation decisions and hence agricultural crop yields. However, information on soil water retention characteristics (SWRC) is seldom available for irrigation planning, crop yield modeling, or hydrological simulations, especially for problematic soils, such as seasonally impounded shrink-swell soils. As large scale direct measurement of SWRC is not viable due to a number of reasons, researchers have developed pedotransfer functions (PTFs) to estimate SWRC from easily measured soil properties, such as texture, organic matter content, bulk density, etc. However, PTF applicability in locations other than those of data collection has been rarely reported. One of the most recent PTFs that has shown overall reasonable predictions in evaluation studies is Rosetta, a numerical code for estimating soil hydraulic parameters with hierarchical pedotransfer functions. Relatively, the development of large databases makes it one of the widely used PTFs. If validated for spatial application, it has immense use potential in countries like India, where data on soil hydraulic properties are seldom available, a deficiency that hampers better simulations in processes, like partitioning runoff and infiltration, assessing evapotranspiration, irrigation scheduling, etc. Rosetta is also relatively flexible allowing estimation of hydraulic properties from easily available minimum input of textural fractions. This study was conducted to evaluate (1) an applicability of four widely used soil water retention functions to describe SWRC; and (2) the computer program Rosetta for its validity. Statistical indices, i.e., root mean square error (RMSE), mean absolute error, maximum absolute error, and degree of agreement $\left(d\right)$ were computed to evaluate “goodness-of-fit” of the four functions to the measured SWRC data. These indices were also used to compare measured SWRC with estimates of SWRC by Rosetta. For soil samples collected from 41 profiles, 175 SWRC were measured in the laboratory. The van Genuchten function fitted relatively better $(\mathrm{RMSE}=0.052{\mathrm{m}}^3{\mathrm{m}}^{-3})$ to SWRC of clay soils, whereas the Brooks–Corey (BC) function was better in expressing SWRC of clay loam and sandy clay loam soils with $\mathrm{RMSE}=0.06$ and $0.07{\mathrm{m}}^3{\mathrm{m}}^{-3}$ , respectively. Campbell and Cass–Hutson (CH) functions were of intermediate value. Worst performing functions were BC (clay soils), Campbell (clay loam), and CH (sandy clay loam) with corresponding $\mathrm{RMSE}=0.059$ , 0.065, and $0.077{\mathrm{m}}^3{\mathrm{m}}^{-3}$ . Estimates of two important points on the SWRC curve, i.e., field capacity and permanent wilting point were predicted with relatively better accuracy for clay and sandy clay loam soils by all the four functions. RMSE and $d$ ranged from $0.027\text{to}0.043{\mathrm{m}}^3{\mathrm{m}}^{-3}$ and from 0.73 to 0.88 for clay soils. Corresponding values for sandy clay loam soils were $0.008–0.019{\mathrm{m}}^3{\mathrm{m}}^{-3}$ , and 0.92–0.98. However, in clay loam soils, only two functions were found suitable. Estimates of SWRC obtained by applying hierarchical rules in Rosetta were reliable $(\mathrm{RMSE}<0.05{\mathrm{m}}^3{\mathrm{m}}^{-3}$ ). Magnitude of average RMSE increased progressively in clay loam, clay and sandy clay loam soils $(0.028<0.035<0.042{\mathrm{m}}^3{\mathrm{m}}^{-3})$ . The study established that SWRC of the “Haveli” soils could be estimated using generic PTF and thus information that is prerequisite in simulating hydrological processes occurring in seasonally impounded soils could be acquired.