Identifying river geodiversity in Lower Mekong tributaries using supervised classification of Sentinel-2 imagery

Abstract

This project aims to improve the environmental data available for river basin managers when deciding Freshwater Protected Area (FPA) location. Due to the absence of information on critical habitat location and difficulty in obtaining local ecological knowledge at large scales, FPA establishment is based on socioeconomic indicators rather than environmental indicators.

This study investigates the methods and extent to which supervised classification of Sentinel-2imagery could be used to identify areas with high river geodiversity, acting as a proxy for river biodiversity. Land use and geomorphological river features were classified for the Xe Banghieng, Sekong and Sesan rivers in Laos and Cambodia.

All analyses were carried out using open-source data in QGIS, Google Earth Engine, and Microsoft Excel. Existing FPAs were found to feature a high occurrence of deep pools and sandy/rocky geomorphological river features (GMRF). A methodology for GMRF detection and identification through supervised classification was developed on Xe Banghieng river and extrapolated for application to the Sekong and the Lower Sesan rivers. This was then analysed using a processing model to find hydrobasins with the highest percentage of in-channel GMRF area, the chosen proxy for river geodiversity.

The Xe Banghieng has the most detected GMRF from all the rivers studied, and downstream hydrobasins had a higher percentage of GMRF area within the river channel. In the Sekong river, upstream has a GMRF deficit, and the top five hydrobasins are located either side of the Laos-Cambodia border. The Sesan river has a GMRF deficit below the Lower Sesan Dam II.