LINKING SOIL WATER STORAGE TO LONG-TERM OUTLOOKS OF WATER SUPPLY ANOMALIES IN TEXAS

The 2011 drought in Texas resulted in total water storage (TWS) deficit of 62 km³ with soil water storage (SWS) accounting for 20-100%. Although we can reasonably monitor precipitation, stream flow, reservoir capacity, and groundwater levels, soil moisture is more enigmatic. As drought continues in Texas, water managers are becoming aware of the linkage between soil moisture deficit and water resources. Existing forecasting models for reservoir storage do not account for such anomalous departures, yielding poor stream and reservoir forecasts. For this study, we develop a multiple linear, time regressive model based on the cross-correlation time series between monthly soil moisture products to forecast reservoir levels at 30, 60, and 90 day in Texas and within 6 of its largest watersheds. A total of 8 parameters were evaluated by backward step-wise regression including monthly precipitation and temperature from the NCEP North American Regional Reanalysis (NARR); TWS from the NASA Gravity Recovery and Climate Experiment (GRACE); SWS from 0-100 cm from the North American Land Data Assimilation System (NLDAS) multi-model including Noah, Mosaic, VIC, and their ensemble mean; and SWS from 0-5 cm from the European Space Agency’s Soil Moisture and Ocean Salinity (SMOS) missions. Our results show that state-wide reservoir capacity is remarkably predictable to 90 day lags, with correlations coefficients in excess of 0.7 between reservoir storage and SWS anomalies from all products; GRACE data produced the highest correlation at 60 days (R = 0.81). Basin-level correlation was significant for three basins (Brazos, Colorado, and Trinity) and more dependent on time series length, indicating management influences primarily related to increasing capacity over the past 20 years. The multiple time regressive model showed successful hindcast predictions despite several compounding factors, including: (1) Texas and its watersheds transition between wet and dry climates; (2) large variability of spatial scales from 14 km cells in NLDAS to 200 km cells in GRACE; (3) karstic underlying geology in numerous reservoir locations; and (4) tremendous anthropogenic pressure and intensive management of each watershed. This model can in turn be used to optimize water allocation decisions at the state and basin levels.