Drylands occupy one third of the Earth’s surface and are home to around 400 million people, yet the water resources of these regions are often poorly understood because of a lack of fundamental hydrological data. Rivers are often low-gradient with multiple channels across large floodplains, adapted to transmitting episodic and slow-moving flood pulses with flows diminishing downstream due to high transmission losses. At a coarse-scale, it is known that these losses sustain the biologically rich and diverse floodplain-waterhole “boom-and-bust” ecosystems, yet a detailed understanding of the way these losses are partitioned and other fundamental questions of (eco)hydrological function of these river systems cannot be understood at a detailed scale in dryland environments.
This thesis aims to develop remotely-sensed data approaches to support hydrodynamic modelling in order to improve our understanding of hydrological processes in data-sparse dryland landscapes. Four objectives, which constitute the main chapters of this thesis and which are presented as 4 peer-reviewed papers, were investigated, they are:
(i) to evaluate the accuracy and effectiveness of satellite derived altimetry data for estimating flood water depths in low-gradient, multi-channel rivers in central Australia;
(ii) to detect and map flood extents and optimise the trade-off between image frequency and spatial resolution using Landsat and MODIS imagery;
(iii) to determine the optimum Digital Elevation Models (DEMs) for hydrodynamic modelling in low gradient dryland environments in relation to accuracy, DEM preparation and trade-offs in model grid size; and
(iv) to use a hydrodynamic model supported by altimetry, optical and optimal DEM data to investigate the partitioning of transmission loss within a multi-channel river in central Australia.
Regarding objective (i), based on an assessment of six altimetry satellites at two lakes (in Victoria and Western Australia) and six sites along the Diamantina River and Cooper Creek, ICESat (mean = 0.00m, RMSE = 0.04m) was found to have the highest accuracy, while Jason-2 (mean = -0.04m, RMSE = 0.28m) offered potential for ongoing applications where water elevation time series are of use.
Regarding objective (ii), two advanced blending algorithms were applied to MODIS and Landsat imagery; STARFM (Spatial and Temporal Adaptive Reflectance Fusion Model) and ESTARFM (Enhanced STARFM) and evaluated for nine common indices used in vegetation studies, environmental moisture assessment and standing water identification across three sites with different flood extents and spatio-temporal variability. The investigation of these factors, including the order that index calculation and blending should occur (i.e. (i) ‘Index-then-Blend’ (IB); and (ii) ‘Blend-then-Index’ (BI)), found that the IB approach was more accurate for both blending algorithms as it was; computationally less intensive due to blending single indices rather than multiple bands; more accurate due to reduced error propagation; and less sensitive to the choice of algorithm.
Near-global coverage Digital Elevation Models (DEMs) such as Shuttle Radar Topography Mission (SRTM) and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) are required for hydrological–hydrodynamic modelling in remote areas. For objective (iii), point accuracy and geometric co-registration error; effects of DEM preparation (vegetation smoothed and hydrologically corrected); and effects of grid size on hydrodynamic model performance were investigated. SRTM outperformed ASTER after applying geometric correction followed by vegetation smoothing and hydrological correction and a grid size of around 120m offered the optimal balance between hydrodynamic model and computational performance.
Combining modelling, remotely-sensed data and limited field measurements for objective (iv), a calibrated hydrodynamic model (TUFLOW) was used to investigate the overall water balance , with a focus on transmission losses, across seven flood pulses along a 180 km reach of the Diamantina River. Results showed that transmission losses were up to 68% (mean = 46%) of the total inflow to the system, with evaporation the most significant component (21.6%), then infiltration (13.2%) and terminal water storage (11.2%).
This research concluded that it is now possible to realistically constrain water balances in data-sparse dryland rivers using hydrodynamic models in combination with remote sensing and simple field measurements to address limitations in the availability of conventional hydrological datasets. This research has implications for the opportunities, limitations, and future directions of using remotely-sensed data to better understand water balance and hydrodynamics of data-sparse regions. This knowledge is imperative for improved management of the limited water resources in dryland, low-gradient, and multi-channel river systems both in Australia and around the world.