The floodplain inundation history of the Murray-Darling Basin through two-monthly maximum water depth maps

Abstract

With growing concerns over water management in rivers worldwide, researchers are seeking innovative solutions to monitor and understand changing flood patterns. In a noteworthy advancement, stakeholders interested in the changing flood patterns of the Murray Darling Basin (MDB) in Australia, covering an area of 1 million km², can now access a consistent timeseries of water depth maps for the entire basin. The dataset covers the period from 1988 to 2022 at two-monthly timestep and was developed using remotely sensed imagery and a flood depth estimation model at a spatial resolution of ≈30 m, providing a comprehensive picture of maximum observed inundation depth across the MDB. Validation against 13 hydrodynamic model outputs for different parts of the MDB yielded a mean absolute error of 0.49 m, demonstrating reasonable accuracy and reliability of the dataset. The resulting dataset is best suited to system-wide analysis but might also be useful for those interested in the history of flooding at specific locations in the system. We provide the dataset, visualization tools, and examples to support ongoing research.

Fig. 1

Maximum floodwater depth. The maximum floodwater depth for the Murray Darling Basin calculated from the two-monthly floodwater depth dataset.

Fig. 2

Steps involved in building and distributing the flood water depth product for the Murray Darling Basin. The three panes show steps of the water depth product development. (a) Input data processing involves acquisition of two products: two-monthly maximum water surface extent (from Landsat) and a high-resolution Digital Elevation Model (combined from data sources), which we split into 23 regions for processing. (b) Floodwater Depth Estimation Tool (FwDET) algorithm v2 was used to identify the surface water elevation at the boundary (perimeter) of inundated areas. The perimeter water surface levels (elevations) were interpolated across inundated areas to provide continuous surface water levels. The depth was calculated by subtracting the Digital Elevation Model from the surface water levels and merging (recombining) across the Murray Darling Basin. (c) The resulting water depth rasters were archived in CSIRO’s Data Access Portal, and were also distributed through web services for machine access (i.e. Web Mapping Service) and presented through a geospatial visualisation platform for point-and-click visualisation of water depth across the floodplains of the Murray Darling Basin.

Fig. 3

Location of validation sites. These include three reaches of the River Murray in South Australia (A), one reach of River Murray in Victoria (B), and three reaches of the Balonne River (C), where the water depth data were compared to benchmarking hydrodynamic modelling results.

Fig. 4

Floodwater depth accuracy for each flood scene. The accuracy is expressed as benchmark minus prediction. The positive values represent underestimation of depth and negative values overestimation. Acronyms on axis labels are Lower Balonne System (LBS), Floodplain (FP), Gunbower Koondrook Perricoota Forest (GKPF) and South Australia (SA).