Satellite gravimetry in water-limited environments: applications and spatial enhancement

Large-scale hydrology is often insufficiently represented by ground monitoring networks and land surface models. The launch of the Gravity Recovery and Climate Experiment (GRACE) satellite mission provides an independent source to quantify water cycle dynamics over large basins. The main objective of this PhD thesis is to improve our understanding of water resources in water-limited environments through the use of spatially-improved GRACE observations.\ud \ud Firstly, the benefits and challenges from readily available GRACE global solutions at a 4° resolution were examined. GRACE global solutions from CSR and GRGS were applied to investigate the recent decadal groundwater depletion trend over the Hai River Basin in China. Results show an agreement between GRACE and in-situ groundwater observations: the Hai River Basin, especially the North China Plain, exhibited a constant declining trend in groundwater storage between 2003 and 2012. Continuous groundwater depletion is mainly attributed to anthropogenic over-extraction rather than climatic variability. This study highlights the value of using GRACE observations to provide rapid and independent estimates of regional groundwater storage change, which can supplement the scarcity of ground measurements. However, the coarse spatial resolution associated with GRACE global solutions prevents the analysis of spatial variability in groundwater depletions within the basin.\ud \ud Secondly, the benefits of applying newly developed the regional GRACE solutions at a 2° resolution for the estimation of large-scale evapotranspiration (ET) in two semi-arid and arid Australian basins were assessed. The GRACE ET estimates from these regional solutions were compared with three continental ET products; one derived from a land surface model and two from energy-based satellite models. The results demonstrate that the satellite energy balance ET models poorly represent water availability at different temporal and spatial scales; ET was overestimated during water stressed conditions and underestimated when it was wet. Integrated with optical and thermal satellite retrievals, however, these energy-based models are capable of detecting ET heterogeneities over land surfaces. In comparison, water balance constrained GRACE (P-△S) estimates are able to capture dynamic ET variations over large spatial domains.\ud \ud As a result of these findings, a wavelet fusion approach was proposed to further refine GRACE ET estimates, as well as to better constrain energy-based ET estimates. The improvements were achieved by combining GRACE (P-△S) and an ET product forced by optical and thermal remote sensing retrievals. A wavelet-based multi-scale fusion approach embodied with a flexible coefficient weighting scheme highlights the large-scale ET patterns from GRACE but also the small-scale ET features from the energy balance constrained model. Compared with original ET sources, our fused ET results display a significant improvement in spatial accuracy and a better sensibility to rainfall variability over temporal scales.\ud \ud Overall, this research confirms GRACE's capability and potential contributions in understanding the hydrology of arid and semi-arid environments. In these regions, water fluxes are limited by water availability rather than energy, hence the water balance framework used in GRACE studies is ideal. A prospective analysis highlights that the GRACE fusion method developed in this thesis has potential applications for other components of the water cycle for the purpose of enhancing estimation accuracy in time and/or space