Automated Calibration Methods Improve Estimation of Soil Hydraulic Parameters for Soil Water Dynamics and Evaporation.

Estimation of soil hydraulic parameters is the key for predicting soil water dynamics, surface fluxes, and related biochemical processes in agricultural systems. However, high uncertainty in calibration of the parameters due to variations in space and time limit a model’s ability for prediction and application. In this study, a global search analysis method (Latin hypercube sampling, LHS) and a local optimization method (parameter estimation software, PEST) implemented with a gradient-based algorithm, Gasuss-Marquard-Levenberg (GML) were explored to calibrate soil hydraulic parameters in the Root Zone Water Quality Model (RZWQM2). Four methods of estimating Brook-Corey parameters for soil water retention and saturated hydraulic conductivity were evaluated to simulate measured soil water dynamics under fallow conditions in eastern Colorado, USA.  Soil evaporation and other water balance components were also computed and compared at four probe sites with different soils and topographic attributes.  Both LHS and PEST can improve the calibrated results compared with manual trial-and-error methods.  Based on paired t tests across four sites, PEST generally resulted in better calibration results for soil water (p<0.01 for root mean square error, RMSE) but worse validation results (p=0.03 for RMSE and p=0.01 for Nash-sutcliffe model efficiency, NSME) compared with the LHS results. By combining LHS and PEST, the calibration results were similar to the initial results from PEST, but the validation results were improved significantly (p<0.01 for RMSE and NSME).  The calibrated soil hydraulic parameters showed similar trends across the sites, but produced different simulation results due to parameter interactions. The simulated soil water balance suggested a high potential for increasing soil water storage under fallow by reducing soil evaporation and water drainage. Further investigations of soil water balance during crop growth periods will build upon these simulations for more efficient crop water use in a semiarid area.