Assessing long-term changes in annual monsoon inundations in the Mekong Delta (Cambodia): Testing an innovative approach linking remote sensing and in-situ measurements to overcome data scarcity

The annual monsoon inundations are vital in maintaining the fertility and productivity of the delta of the Mekong, Southeast Asia’s largest river. During the inundations, which traditionally last from July until November, nutrient-rich sediments are deposited on the floodplains, groundwater is recharged, and fish populations regenerate in the shallow waters. Consequently, local agriculture and fisheries are keyed to the timing of flood arrival and recession and reliant on overall flood duration. However, in recent years, the hydrological dynamics of the region have shifted. The Mekong’s hydrological regime has been impacted by shifts in land cover, the construction of hydropower infrastructure, and climate change. 

Yet the effects of these changes on the spatio-temporal patterns of inundations in the Mekong Delta remain largely unstudied, especially at local scales. Part of the reason for this is data sparsity: there is a lack of consistent long-term data on spatial inundation dynamics. No concerted in-situ monitoring efforts of flood extents existed until recently, while optical earth observation satellite missions such as Landsat often fail to provide data during the wet season due to cloud cover. Hydrological modelling approaches struggle with insufficiently precise elevation data - due to the flat topography of the Mekong Delta, even high-resolution Digital Elevation Models (DEMs) fail to capture small-scale dykes that determine whether large swaths of land become flooded. 

To cope with this data-scarce environment, we propose an innovative methodology harnessing recent satellite missions and long-term in-situ river water level measurements. This approach uses remote sensing data from the Sentinel-1 and 2 missions operated by the European Space Agency. Since 2017, these satellites provide optical and synthetic aperture radar (SAR) data at a spatial resolution of 10 m and a return frequency of 5-6 days. Furthermore, SAR provides data independent of cloud cover, which makes it particularly well-suited for operational flood monitoring purposes. After deriving inundation maps from available Sentinel images, we link these maps to water levels measured at a local hydrological station through a correlative approach to create a water-level flood link (WAFL). Using this link, we can describe the evolution of inundation patterns in the Mekong Delta since the 1990s. To quantify uncertainties, comparisons with historical inundation maps derived from available Landsat images,  and with a high- resolution DEM were carried out.  

The approach was tested in two study areas in the Cambodian Mekong Delta.  The results indicate that the accuracy of the WAFL for quantifying inundations on a per-pixel basis lies at 87%, reaching up to 93%. The spatio-temporal analysis shows that inundation incidence in the early wet season has declined by 21% since 1991 and that the average duration of inundations has decreased by 19 days. This illustrates that annual monsoon inundations have become an increasingly volatile resource, with significant impacts on agriculture, fisheries, and ecosystems.