Processing of satellite images

Locations of riverbank erosion within the study area were identified based on exemplary historical satellite imagery with a cloud cover of less than 10%. Therefore, the modified normalised difference water index (MNDWI)⁵⁹ was applied within QGIS using openly available Landsat-7 and 8 images. First, the Semi-Automatic Classification Plugin⁶⁰ was used to calculate top-of-atmosphere (TOA) reflectance from raw digital numbers. Afterwards, the green and short-wave infrared bands were used to calculate the MNDWI as:

where ρGreen and ρSWIR1 are the TOA reflectance of the green and short-wave infrared bands. A threshold of 0.4 was used to delineate water from land mass. Some features (e.g. clouds) that did not reflect the actual bank lines were removed upon visual inspection. In order to identify areas of bank erosion and accretion based on historical satellite images, a centre line was defined using the river’s present-day geometry. Every 100 m along the centre line, orthogonal lines were drawn. The intersection points of these orthogonals with the shapes of the historical riverbanks were used to assess riverbank migration. These changes in riverbank location were plotted against their respective position along the RKM. Afterwards, high-resolution satellite images from Google Earth were used to check whether areas of land accretion corresponded to locations with aquaculture infrastructure. The identified areas were highlighted separately.