Evaluation of SWAT Soil Water Estimation Accuracy Using Data from Indiana, Colorado, and Texas

. Highlights SWAT soil water assessment using soil water measurements Dryland SWAT model soil water content was greater than the irrigated SWAT model Using SWAT soil water estimates for real-time (daily) irrigation management purposes with the existing SWAT soil water subroutines and available soils data considered risky The surface had showed the greatest soil water variability compared to deeper layers Abstract Soil Water Content (SWC) is a challenging measurement at the field, watershed, and regional scales. The Soil and Water Assessment Tool (SWAT) soil water were evaluated at three locations: 1) St. Joseph River Watershed (SJRW) located in northeast Indiana, 2) the United States Department of Agriculture- Agricultural Research Service (USDA-ARS) Conservation and Production Research Laboratory (CPRL) at Bushland, Texas, and 3) the USDA-ARS Limited Irrigation Research Farm (LIFR) at Greeley, Colorado. The soil water estimates were evaluated under two scenarios: 1) for defined soil profile, and 2) by individual layer. Each site‘s soil water assessment was performed based on the existing management conditions during each experiment, whether dryland or irrigated, and for various periods depending on soil water content measurement availability at each site. The SWAT soil water was evaluated as follows: Indiana site under dryland conditions using daily soil water observations for one year. The Texas site was evaluated for a ten-year period under irrigated and dryland conditions using weekly soil water observations for four lysimeters. The Colorado site was evaluated under irrigated conditions for a four-year period. The simulated soil water was evaluated by comparing the model simulates with observed daily and weekly soil water measurements at the three sites. Based on the results, even though all the SWAT models were considered to perform as good models following calibration (streamflow, ET, etc), the soil water simulations were unacceptable for the defined soil profile and for individual layers for the three sites. Deeper soil layers had observations greater than field capacity values, indicating poor soil parameterization. The dryland model had greater water content than the irrigated model, contradicting the soil water measurements. This greater soil water simulation with the dryland model is a result of SWAT model uncertainties with ET reduction under dryland conditions due to water stress. This paper indicated that soil water estimation using the default SWAT soil water equations have many sources of uncertainties. Two apparent sources result in the SWAT model poor performance: (i) SWAT soil water routines that do not fully represent soil water moving between layers to meet plant demand and (ii) uncertainty in soil parameterization.