INVERSELY ESTIMATING SOIL HYDRAULIC FUNCTIONS USING EVAPOTRANSPIRATION FLUXES

Abstract

Numerical modelling of the transient moisture fluxes helps in the development of appropriate water management practices, but requires site-specific information on soil hydraulic properties. Forward simulations with a given set of Van Genuchten (VG) model parameters indicate that different soils act differently in their temporal variable evapotranspiration (ET) response under deep-water table/free drainage conditions. Since actual ET fluxes can be recovered from remote sensing measurements, a new possibility is established to derive soil hydraulic functions under actual field conditions for a range of spatial scales. Thus, inverse modelling of ETa fluxes is a promising way to estimate the so-called effective soil hydraulic functions. We numerically explore the possibility of inverse ET modelling to derive VG model parameters for semi-arid regions. The utility of this technique is evaluated using a simulation model to create a 'no error data set' of ET fluxes for cotton crop. Objective functions using actual ET and actual transpiration (T) are defined. Backward simulations are carried out to re-assess selected VG model parameters for the three-soil types, i.e. sand, loamy sand and sandy clay loam. A realistic ET data set is created from the no error data set by incorporating different levels of random error. Seasonal simulated water balance components (ET, deep percolation and change in profile storage) are compared to study the hydrological performance of the inversely estimated soil hydraulic functions. Results indicate that the moisture stress period under fully developed crops is most appropriate for sampling ET fluxes to solve the proposed inverse problem. It is also observed that frequent observations of ET fluxes are desired to reduce undesirable correlation between different fitting parameters. It is observed that when ET fluxes are very accurate, the VG model parameters α, n and θs (θr, ks and λ fixed at actual values) are optimised precisely with 12ET data points. Inverse fitting of these parameters (α, n and θs) utilising perturbed data on ET fluxes results in effective soil hydraulic functions, which reliably predict different water balance components for sand and loamy sand soils.