USING SATELLITE EARTH OBSERVATIONS TO UNDERSTAND THE FACTORS CONTROLLING THE TERRESTRIAL WATER CYCLE IN THE NORTHERN HIGH PLAINS, USA

Characterization of factors controlling the terrestrial water cycle (TWC) is imperative to better understand the dynamics, predict water storage and availability, and evaluate impacts on the water cycle due to global changes. This study utilizes and compares time-series data from satellite- and model-based products of the terrestrial water storage. Variables include precipitation (P) from PRISM and TRMM TMPA, soil moisture (SM) from AMSR-E and NLDAS, MODIS vegetation coverage (VC), MODIS and NLDAS land surface temperature (LST), snow water equivalent (SWE) from NLDAS, observed groundwater level (GWL), and TWS anomaly (dTWS/dt) from GRACE over the Northern High Plains (bounded by 38 – 44⁰ N Latitude and 106 – 96⁰W Longitude). Area-averaged time series (2003-2010) of P, SM, VC, LST, SWE, and dTWS/dt were extracted and processed using ENVI and GIS. The cross correlation function (CCF), which provides correlation with lags between the independent variables (e.g., P, LST) and dependent variables (e.g., SM, dTWS/dt) that are useful predictors of TWS anomaly, was applied to the multiple time-series products.

The results indicate there is a statistically significant correlation with time delays between the different TWC variables and dTWS/dt from GRACE. A relatively strong positive correlation (r = 0.77-0.83) exists between GWL and dTWS/dt with a time delay of 0 to 2 months. Soil moisture is positively correlated with dTWS/dt with little to no delays. However precipitation, vegetation coverage, and land surface temperature variables are negatively correlated (r = -0.4 – -0.6) with GRACE’s dTWS/dt with delays of 2 to 4 months. This could be associated with seasonality, where these variables are relatively higher during the summer season, while dTWS/dt decreases due to high evapotranspiration in this season. The reverse is also true during the winter season. There is consistency in correlation between the different products and GRACE with minor differences in time delays. For example, the correlation values of SM from NLDAS with dTWS/dt and SM from AMSRE with dTWS/dt were nearly equal in magnitude but different in time delay. Further improving this approach may help to predict TWS anomaly from different TWC variables and used to fill the gap between current and future GRACE mission.