A global drought monitoring framework using GRACE/-FO data assimilation

Abstract

Extreme climate phenomena like droughts can lead to a shortage of available water resources that in turn can cause serious consequences such as famines. Monitoring of extremes is extremely important, and past research mainly focused on fluxes like precipitation or streamflows and surface waters because of easy access to measurements. More complicated is the use of subsurface water storages like groundwater for observing water shortages, mostly monitored via in-situ observations that are spatially irregularly distributed. The stations require maintenance and access can be restricted because, e.g., of political reasons. Hydrological models provide simulations of surface and subsurface water as well but they use strong process simplifications, the forcing data are error-prone, and thus the models imperfectly represent reality. <br /> Another possibility to derive information about water storages is via satellite observations. Since 2002, the gravity missions Gravity Recovery And Climate Experiment (GRACE) and its successor GRACE Follow-On (GRACE-FO) provide measurements of surface and subsurface water globally from space that can be represented as Total Water Storage Anomalies (TWSA). Concretely, TWSA is the change in the aggregated volume of water stored in different compartments, among others, rivers, soil moisture, and groundwater. However, due to the orbit constellation, only a spatial resolution of about 300 km is given, which is too coarse for most drought applications. A possibility to simultaneously improve the spatial resolution, disaggregate TWSA into the single water compartments, and improve the models’ realism can be achieved by data assimilation. With data assimilation, the TWSA observations are integrated into a hydrological model and simulated model output is pushed towards the observations. <br /> So far, no global framework exists on how to optimally use water storage information from GRACE/-FO within assimilation for monitoring drought and its water propagation through the hydrological water cycle, i.e. how precipitation deficits lead to a decrease in water storage that in turn affects vegetation growth. Therefore this thesis has three major goals: (1) developing a framework that allows assimilating GRACE/-FO TWSA into the Water Global Assessment and Prognosis (WaterGAP) hydrological model for the first time globally, (2) analyzing dominant hydrological signatures and identifying the signatures of precipitation - water storage - vegetation seasonal maxima and non-seasonal events, and, (3) developing a drought monitoring framework and an unprecedented prototype warning alert system under the consideration of GRACE/-FO and assimilation outputs. <br /> The global GRACE/-FO assimilation into the WaterGAP model is set up with ensemble-based filters, tuned for global application, e.g. to include spatial observation error correlations via localization techniques and provided as Global Land Water Storage (GLWS) release 3.0 – the update of GLWS release 2.0. The analysis of dominant signatures (e.g., linear trends) shows that TWSA from assimilation outputs inherit properties from both, the GRACE/-FO observations and the simulations, thus, present a smooth transition between them. The timing of seasonal or episodic high or low precipitation leading to an increase or decrease in water storage is for most vegetation regimes, as expected, found to be shortest for soil moisture, longer for surface water, and longest for groundwater. Validation with respect to independent data sets shows an overall improvement of GLWS compared to WaterGAP simulations for TWSA and groundwater but the assimilation does not have a major impact on surface water storage and soil moisture as these storages do not significantly improve. <br /> In drought monitoring it is common to define indices that would classify actual drought conditions as “severe”, “moderate”, etc., to facilitate an easy means of communication to decision makers. Therefore, in this thesis, the performance of existing GRACE drought indices is studied. As it is difficult to evaluate such indices, a synthetic study is set up that reveals how trends, accelerations, and noise in GRACE/-FO time series are biasing drought detection. Hereinafter, droughts are analyzed globally and in selected focus regions for TWSA, and subsequently, I investigate subsequent drought events from soil moisture to surface water and groundwater. An approach is developed for determining drought hazard risk maps from assimilation-derived soil moisture and the combination of surface water and groundwater. Finally, a prototype of a warning alert system is set up and analyzed to identify the spatiotemporal dynamics of droughts for hydrological basins. This system might pave the way for future implementations of TWSA or water storages from GRACE/-FO assimilation into existing or future operational early warning systems.