Prolonged hydrologic drought disturbs the natural state of ecosystems, stresses regional water supplies, and can adversely affect agricultural production. Determining the severity of hydrologic drought traditionally depended on evaluations of historical rainfall, stream flow, and soil moisture; yet, a comprehensive measure of the magnitude of a drought's impact on all components of the terrestrial hydrologic system, including surface, soil, and groundwater storage, remains lacking in standard drought analyses. NASA's Gravity Recovery and Climate Experiment (GRACE) satellite mission fills a gap by providing monthly measures of terrestrial water storage anomalies (TWSA) based on time-variable gravitational fields. This dissertation details an investigation of regional hydrological extremes (e.g., drought and flood) using both satellite remote sensing data and outputs from NASA's Catchment Land Surface Model (CLSM).
The first project presented in this thesis involves discussion of a novel quantitative, GRACE-based framework for measuring the severity of hydrologic drought. GRACE observations are used to quantify drought by calculating the deviation of monthly-average terrestrial water storage anomalies from the regional climatological reference, where negative deviations represent storage deficits. Each deficit conveys the volume of water that would be required to recover from a drought. Moreover, this finite deficit observation allows for the calculation of a likely time for recovery based on statistical percentiles of storage change distributions, for every month through the end of the event.
The second portion of work evaluates and compares the characteristics of subsurface terrestrial water storage variables from the CLSM, assimilated with GRACE satellite observations (CLSM-DA), for the purposes of: acquiring near-real time analysis, downscaling GRACE's spatial resolution, and vertically disaggregating GRACE column-integrated water storage anomalies. Several zones throughout the United States were selected to quantify differences between hydrologic extremes identified by CLSM-DA and those measured by GRACE. Results establish that CLSM-DA TWSA outputs improved those from CLSM Open-loop runs in all regions with R2 increases from 5-14%. We also compared CLSM surface soil moisture content with independent surface moisture observations from the AMSR-E satellite to assess improvements after data assimilation. Results established that assimilation produced modest improvements in correlations between CLSM and AMSR-E in all regions.
CLSM-DA hydrologic extremes are comparable to GRACE, however the data-assimilated model continues to struggle with matching the some of the amplitudes of extreme events, in part due to model structure and parameters that do not possess enough information about the hydrologic system to accurately depict changes in TWSA as observed by GRACE. Since CLSM continues to run through the near-present month (April 2014), beyond the current, publically available GRACE month (January 2014), an assessment of the CLSM's performance between assimilation updates is also provided.
The final project details the development of a linear multivariate, multi-frequency regression model to estimate monthly water storage change and extremes before and beyond the currently available GRACE observation period (April 2002-April 2014). The regression model provides coefficients that can then be used with any precipitation and evapotranspiration dataset, to calculate the associated amount of water storage change for our study region, California's Central Valley (e.g., Sacramento, San Joaquin, and Tulare river basins). Model results show that 82% of GRACE's TWSA signal can be explained with a combination of precipitation and evapotranspiration. The June 2014 storage estimate from the regression model revealed that water storage deficits persisted in the Central Valley with a monthly value of -28.8 km3 (±1.22 km3).
This work concludes that GRACE satellite data can successfully be utilized for regional scale drought analysis and has implications for improving drought early warning lead times together with drought preparation and management efforts. The storage deficit method demonstrates the added benefits of explicitly recognizing the beginning and end of storage deficit periods and of providing additional information about the effects of meteorological drought on regional water storage. Data assimilation increases the usability of GRACE for near-present monitoring, while implementation of the linear multi-frequency regression model allows for the extension of water storage anomalies.
