Modelling Surface Water Dynamics (SWD) on Large River Basin Scale from Space: A Case Study for the Murray-Darling Basin (MDB) of Australia

Abstract:

Globally, increasing demands on water resources along with climate variability and change have led to alarming declines and deterioration of terrestrial surface water resources. The usage of earth-observation data and techniques for modeling SWD and its drivers represents a promising approach for sustainable management and restoration of surface water resources across broad geographic regions. The main objective of this research was to model SWD with a focus on floods, observed in 25 years of Landsat imagery (1986 - 2011), across a large and highly regulated river basin, the MDB in Australia. SWD were modelled as a function of river flow and spatially explicit time‑series data on soil moisture (Climate Change Initiative active passive microwave), evapotranspiration (Australian Water Resources Assessment land surface model) and rainfall (gauge-based). To enable a consistent modeling approach within the complex hydrological structure of the river basin, a unique spatial modeling framework was developed based on a fully directed and connected stream network, a categorization of the basin into floodplain and non-floodplain area and a regular grid of 10 by 10 km cells. Based on this framework, SWD on local floodplain units were successfully related to flow data from connected gauges by quantifying the lag time for each cell. Dynamic regression models of SWD were fitted locally for floodplains in each grid cell, with an average adjusted r² above 0.7. Validation against 10 years of test data that was left out for model fitting showed that the models can predict the test data with an average r² of 0.7, which makes them suitable for improving the ongoing management and allocation of environmental flows in the MDB. The models also revealed the relative importance of local climate conditions for SWD, with increased importance of soil moisture, evapotranspiration and rainfall in arid regions, in proximity to headwater catchments and on slow‑draining floodplains.