Researchers from the Geospatial Analysis for Environmental Change Lab (UNSW) led by Mirela Tulbure and Mark Broich recently released a new paper on mapping surface water and flooding dynamics over the Murray-Darling Basin using three decades of time series of satellite data. Using Intersect Space storage and hpc.time, they mapped surface water and flooding dynamics (SWD) continuously through time (seasons and years over three decades) in the Murray-Darling Basin, a large (>1 million km2) semi-arid basin that experienced extreme hydroclimatic variability and competing water demands. [Download the paper]
They used the seasonally continuous Landsat TM and ETM + archives and generic random forest-based models, an ensemble classifier, necessary when mapping targets with high spectral variability such as floods. The combination of higher spatial resolution and using the entire Landsat archive allowed them to quantify changes in surface water and flooding dynamics with unprecedented detail (1986-2011).
Results revealed dramatic impacts of drought and wet extremes on SWD, with high inter- and intra-annual variability across time and low surface water/flooded area across the basin during the “Millennium Drought” (most severe drought on record in southeastern Australia, 1999–2009) followed by massive floods in 2010-11 (Figure below).
They validated the surface water and flooding time series using a statistically rigorous accuracy assessment. Accuracy of SWD was high and acceptable over time in wet and dry years. Providing an unbiased estimate of the accuracy of their SWD product and quantifying the uncertainty of the estimate are imperative steps needed when remotely sensed products are used for follow on applications.
Ongoing applications of the SWD product include vegetation response to flooding, connectivity and driver modelling.
Tulbure M.G., M. Broich, S.V. Stehman & A. Kommareddy (2016). Surface water extent dynamics from three decades of seasonally continuous Landsat time series at subcontinental scale in a semi-arid region. Remote Sensing of Environment 178: 142–157.[Download]