Retro- and forecasting inundation extents for trend analysis in the Cambodian Mekong Delta - a new method combining Sentinel-1 and 2 and local water level measurements

Annual monsoon inundations are integral to the hydrological, ecological, and economic processes of the Mekong Delta. However, the dynamics of seasonal flooding shifted over the past decades due to upstream influences like hydropower infrastructure, land use change, and modified precipitaiton patterns. Characterizing these changes has been a major challenge in these ungauged areas with scarce environmental data, imposing constraints on modelling approaches.

 

Here, hydrological remote sensing through Sentinel-1 and -2 provides valuable insights into inundation processes due to its enhanced revisit frequencies and SAR capabilities. However, the relatively short duration of the Sentinel mission makes it impossible to analyze long-term trends. To address this, we developed a methodology harnessing Sentinel-derived inundation maps and local water level measurements available for the past 30 years. By linking inundation maps to water levels through a correlation model taking into account flood propagation delays we constructed a water level-flood link to retro- and forecast inundation extents. Its performance was assessed using historical Landsat imagery and the TanDEM elevation model. Subsequently, it was used to analyse inundation dynamics since 1991 and to forecast future developments based on streamflow projections.

 

Preliminary results yield an accuracy of up to 93% when compared to historical Landsat inundation maps. The analysis indicates a marked decrease in inundation incidence in the first half of the flood season (-24%) and a decrease in inundation durations by 19 days.