Hydrogeologic Aspects of GRACE Modeling: A Case Study of the Upper Mississippi Basin

Thursday, 18 December 2014
Alla Skaskevych1, Jejung Lee1, Frederick S Policelli2, John D Bolten2 and John L David2,3, (1)University of Missouri Kansas City, Geosciences, Kansas City, MO, United States, (2)NASA Goddard Space Flight Center, Greenbelt, MD, United States, (3)Science Systems and Applications, Inc., Lanham, MD, United States
Gravity Recovery and Climate Experiment (GRACE) modeling is an emerging field in hydrology. Investigation of groundwater change using remote sensing data helps overcome data limitation at a global and regional scale. However, its application to regional or local scale hydrogeologic settings has been limited, especially for the use of the publically available Level 3 data. We conducted a study to estimate the change of groundwater using remotely sensed GRACE and ground truth data for the Upper Mississippi Basin in the US. The modeling conditions that affect the model accuracy are soil moisture models, groundwater fluctuations in the monitoring wells, and the hydrogeologic conditions of the aquifer. We adopted three different land surface models for soil moisture: CLM (Common Land Model), Noah, and Mosaic. The ground truth data from monitoring wells were obtained from the USGS National Water Information System. The results showed that the best-fit soil moisture model is CLM. The correlation coefficient is 86.1%, which signifies strong correlation between remote sensed and ground truth data. As for the effect of aquifer, the best selection of well observations is when the groundwater data is collected from the sand and gravel aquifer. Correlation with well observations in sand and gravel aquifers were 73.4%. The best-fit condition is therefore when the CLM was adopted for soil moisture in the GRACE calculation as well as 11 well observations from the sand and gravel aquifer used for the ground-truth calculation. Under the best-fit conditions, the correlation coefficient between the GRACE and the ground truth is 91.8%.