Space-based data-driven solutions for quantifying dynamics, drivers and quality of surface water resources across space and time

Integrated management of river systems requires synoptic and continuous information about the state of surface- and groundwater resources and recent advances in Earth-observation technology hold great, yet often not fully explored, potential for providing this type of systematic observation. The overall goal of this thesis was to explore some of this potential by developing data-driven methods for space-based quantification of dynamics, drivers and quality of surface water (SW) resources at the large river basin scale. To achieve this objective, this thesis employs semi-parametric time series regression techniques to model empirical relationships between spatio-temporal satellite data sets and in-situ data, using Australia’s Murray-Darling Basin (MDB) as a case study. Chapter 2 describes the development and validation of a framework for modeling SW extent dynamics from a time series of Landsat-based SW maps (1986-2011) as a function of in-situ river flow, rainfall, evapotranspiration and soil moisture. The incorporation of flow travel times and hydroclimatic drivers represent major advances compared to existing approaches. In Chapter 3, this framework is applied across the entire MDB using 18.521 individual modeling units, a spatial river network and 68 river gauges. Goodness of fit was consistently-high (r2 > 0.6) for floodplain areas and hydroclimatic conditions held most explanatory power in unregulated sub-catchments with high levels of climatic variability. Chapter 4 evaluates the ability of the statistical models and Landsat-MODIS blending for overcoming the sparse and irregular sampling frequency of Landsat for 30 m resolution SW mapping (>= 16 days). Both approaches provided a temporally-consistent mapping of floodplain inundation, but image blending is limited when changes in inundation levels become exceedingly rapid (i.e. <= 8 days). Chapter 5 uses generalized additive mixed models to show that changes in terrestrial water storage levels as measured by the Gravity Recovery and Climate Experiment (GRACE) partially explain long-term trends in SW salinity across the MDB. In summary, this thesis contributes a number of novel and peer-reviewed methodologies that provide improved understanding of continental-scale SW inundation and salinity dynamics along with new modeling capabilities for integrated management of water resources at the large river basin scale