The decrease of water resources can enhance poverty and increase insecurity in dry regions, at the same time leading to loss of biological diversity. For these reasons, the information about surface perennial and well-known water sources in the arid and semi-arid regions of Africa has been mapped. However, seasonal water can be missed in mapping due to their short and erratic appearance, while the mapping of any aquatic resources represents a foremost priority for protecting social, economic and biological values in the e.g. Sahara–Sahel transition zone. Therefore, Remote Sensing becomes crucial to monitor a variety of wetland systems in these regions. This work evaluates the performance of three Normalized Difference Water Indexes [Gao’s NDWI (NDWINIR/MIR), McFeeters’ NDWI (NDWIG/NIR) and Xu’s NDWI (NDWIG/MIR)] in mapping of water systems across Mauritania. Maps with seasonal and permanent water were derived, using a multi-temporal series of Landsat 5 TM and Landsat 7 ETM+ images. The performance of indexes was compared based on 551 control points collected during five fieldwork missions to Mauritania between 2007 and 2011. Control points were separated in three classes of water availability (permanent, seasonal and non-water points) and then randomly assigned into two data sets: one for selecting the water availability thresholds for index reclassification and another for threshold validation. NDWIG/MIR and NDWINIR/MIR had good performances in detecting permanent and seasonal water, respectively, while NDWIG/NIR failed to detect most of the water bodies. The threshold selection generated water maps with seasonal and permanent features that might be missing in simple mapping of aquatic systems. The extensive data collection provides novel information about NDWI performances for water delineation in arid and semi-arid regions and for a future management of aquatic environments of the Sahara–Sahel transition zone.
Journal: Journal of Hydrology