Comparison of visible and multi-satellite global inundation datasets at high-spatial resolution

Several new satellite-derived and long-term surface water datasets at high-spatial resolution have recently become available at the global scale, showing different characteristics and abilities. They are either based on visible imagery from Landsat - the Global 3-second Water Body Map (G3WBM) and the Global Surface Water Explorer (GSWE) - or on the merging of passive/active microwave and visible observations - Global Inundation Extent from Multi-Satellite (GIEMS-D3) - that has been downscaled from a native resolution of 25 km x 25 km to the 90 m x 90 m resolution. The objective of this paper is to perform a thorough comparison of the different water surface estimates in order to identify the advantages and disadvantages of the two approaches and propose a strategy for future developments of high-resolution surface water databases. Results show that due to their very high spatial resolution (30 m) the Landsat-based datasets are well suited to retrieve open water surfaces, even at very small size. GIEMS-D3 has a better ability to detect water under vegetation and during the cloudy season, and it shows larger seasonal dynamics. However, its current version overestimates surface water extent on water-saturated soils, and due to its low original (i.e. before downscaling) spatial resolution, it is under-performing at detecting small water bodies. The permanent waters for G3WBM, GSWE, GIEMS-D3 and GLWD represent respectively: 2.76, 2.05, 3.28, and 3.04 million km(2). The transitory waters shows larger discrepancies: 0.48, 3.72, 10.39 and 8.81 million km(2). Synthetic Aperture Radar (SAR) data (from ENVIronment SATellite (ENVISAT), Sentinel and soon the Surface Water Ocean Topography (SWOT)) would be a good complementary information because they have a high nominal spatial resolution and are less sensitive to clouds than visible measurements. However, global SAR datasets are still not available due to difficulties in developing a retrieval scheme adequate at the global scale. In order to improve our estimates of global wetland extents at high resolution and over long-term records, three interim lines of action are proposed: (1) extend the temporal record of GIEMS-D3 to exploit the full time series of microwave observations (from 1978 to present), (2) develop an approach to fuse the GSWE and GIEMS-D3 datasets leveraging the strengths of both, and (3) prepare for the release of SAR global datasets.