GSA Annual Meeting in Indianapolis, Indiana, USA - 2018

Paper No. 161-2
Presentation Time: 8:15 AM

ASSESSMENT OF THE SPATIO-TEMPORAL VARIATIONS IN GROUND WATER STORAGE USING DOWN-SCALED GRACE DATA


SAHOUR, Hossein1, SULTAN, Mohamed1, ABDELMOHSEN, Karem2, KARKI, Sita1, GEBREMICHAEL, Esayas1 and ELBAYUMI, Tamer3, (1)Geological and Environmental Sciences, Western Michigan University, 1903 W Michigan Ave, Kalamazoo, MI 49008-5421, (2)Department of Geological and Environmental Sciences, Western Michigan University, 1903 W. Michigan, Kalamazoo, MI 49008, (3)Department of Statistics, Indiana University Bloomington, Bloomington, 47408; Department of Statistics, Mansoura University, Mansoura, 35516, Egypt

The Gravity Recovery and Climate Experiment (GRACE) has been successfully used for the assessment of terrestrial water storage (TWS) and ground water storage (GWS) for hydrologic systems worldwide, yet such applications are hindered by the low spatial resolution inherited in GRACE data. To address this shortcoming, we applied stepwise multivariate regression models to downscale (from 1º x 1º to 0.25º x 0.25º) the GRACE RL05M Mascon data acquired (2002 to 2016) over Michigan’s Lower Peninsula (LP). We first selected the variables that contribute and/or correlate with GRACE to be used as model inputs, then constructed and tested two types of models, a unified model and a pixel-based approach. For the former, we built one regression model to establish the relationship between the variables and GRACE terrestrial water storage (GRACETWS) over the entire investigated area and for the pixel-based approach, an individual regression model was constructed for each of the investigated pixels. The accuracy of the two approaches was assessed using statistical coefficients (R-square, Normalized RMSE and Nash-Sutcliffe efficiency [NSE]) and the pixel-based approach was found to produce a better performance (R-square ranging from 67 to 86, RMSE ranging from 3.8 to 5.5, and NSE ranging from 0.66 to 0.88). We then, applied the pixel-based models to extract GRACETWS at the spatial resolution of the variables (0.25º x 0.25º). Using the downscaled GRACETWS and outputs of GLDAS land surface model we estimated trends for TWS and GWS on the county scale. Three groups of counties were identified: (1) counties showing a modest increase (5 to 7mm/yr) in GRACEGWS in the central and southern parts of the LP(e.g., Cass, St. Joseph) (2) counties showing a high increase (8 to 11 mm/yr) in GRACEGWS in the central parts of LP (e.g., Montcalm and Gratiot) and (3) counties experiencing an excessive increase (12 to 17 mm/yr) in GRACEGWS in the northern parts of LP (e.g., Gladwin and Clare). The observed increase in GRACETWS during the period 2002 to 2016 is here attributed to the increase in snow water equivalent (southern LP: 0 to 14 mm/yr; northern LP: 14 to 30 mm/yr), but not in rainfall that showed negligible change during the same period.