VALIDATION OF AN EXISTING GROUND-WATER FLOW MODEL OF THE MISSISSIPPI RIVER VALLEY ALLUVIAL AQUIFER USING WATER-LEVEL AND WATER-USE DATA FOR 1998-2005

Ground-water flow models typically are developed and calibrated using available water-level and water-use data. The calibrated model can then be used to evaluate the response of the ground-water system based on hypothetical future pumping conditions. Validation of a ground-water flow model after its development, using post-calibration period water-level observations and water-use data, is less common. In 2003 a ground-water flow model of the Mississippi River Valley alluvial aquifer in eastern Arkansas calibrated to simulate ground-water flow for the period 1918-1998, using water-level observations from 1972, 1982, 1992, and 1998 and water-use data through 1997. The calibrated model was subsequently used to simulate water-level response from 1999 to 2049, and showed that continued 1997-pumping rates led to unsustainable ground-water conditions.

The calibrated model was validated with the addition of water-level and water-use data that extended the end of the observation period to 2005. Water-level observations for each year from 1998 to 2005 were used for model validation, totaling 3,616 water-level observations from 698 locations. Water-use data were compiled and used for the same period, totaling 290,005 discrete water-use values from 43,440 wells, with as many as 39,169 wells pumping in any one year. Total pumping (which is primarily agricultural) for this period was about 2.3 trillion cubic feet of water, distributed over approximately 10,340 square miles within the model area.

Using the calibrated model stressed with the 1998-2005 water-use data and comparing the simulated water levels with observed water levels provides for validation of the ground-water flow model, as well as a quality check of the water-level and water-use data compiled. The 1998-2005 water-level observations ranged from 74 to 293 feet above National Geodetic Vertical Datum across the model area. The maximum water-level residual (observed minus simulated value) for the 3,616 water-level observations was 52 feet, the minimum water-level residual was 60 feet, the average annual root mean squared error was 8.2 feet, and the annual average absolute residual was 6.0 feet. An R² value of 0.96 was calculated for the line of best fit for observed to simulated water-levels for the combined 1998-2005 dataset, indicating a good fit to the data.