The Lake Chad hydrology under current climate change

Lake Chad, in the Sahelian zone of west-central Africa, provides food and water to ~50 million people and supports unique ecosystems and biodiversity. In the past decades, it became a symbol of current climate change, held up by its dramatic shrinkage in the 1980s. Despites a partial recovery in response to increased Sahelian precipitation in the 1990s, Lake Chad is still facing major threats and its contemporary variability under climate change remains highly uncertain. Here, using a new multi-satellite approach, we show that Lake Chad extent has remained stable during the last two decades, despite a slight decrease of its northern pool. Moreover, since the 2000s, groundwater, which contributes to ~70% of Lake Chad’s annual water storage change, is increasing due to water supply provided by its two main tributaries. Our results indicate that in tandem with groundwater and tropical origin of water supply, over the last two decades, Lake Chad is not shrinking and recovers seasonally its surface water extent and volume. This study provides a robust regional understanding of current hydrology and changes in the Lake Chad region, giving a basis for developing future climate adaptation strategies.

with MODIS surface water maps, we use atmospherically corrected surface reflectance Lansat-8  to increase the accuracy of the GRACE total water storage estimates 5 . Although GRACE spatial 48 resolution is ∼300 km, the product we use is distributed on a 0.25°×0.25°pixel-size grid. 49 We use GLEAM 3.3 version dataset 6, 7 to estimate monthly variation of root-zone soil mois-  Figure S2 shows all satellite datasets used in this study, and some available satellite datasets 53 could be used in future studies. 54 Figure S2: List of satellite datasets used in this study.

55
Examples of monthly surface water extent maps derived from MODIS imagery over Lake Chad 56 for January, April, July and October 2003 is shown in Figure S3. Vegetation cover over the Lake is 57 also presented. It is clear that vegetation cover is limited at the beginning of the year (in January), Land Surface Water Extent Mapping with Landsat-8 and Sentinel-2 Imagery.

61
The variations of surface water extent at both regional and global scales have been monitored using higher spatial resolution images from Landsat 8-11 , and recently Sentinel-2 12, 13 observations. The main principle of these studies is to distinguish water and non-water bodies based on the application of water indices 13 , such as the Normalized Difference Vegetation Index (NDVI) 14 , the Normalized Difference Water Index (NDWI) 15 , the Modified Normalized Difference Water Index (MNDWI) 16 , or the Automated Water Extraction Index (AWEI) 17 . Some authors compared the performance of these indices in different regions 12, 16, 18-20 , and concluded that the MNDWI normally gives the best result among all the indices. Over Lake Chad basin, the MNDWI was also reported to work better than other indices in detecting surface water bodies 20, 21 . In this study, we apply a threshold on the MNDWI to separate water and non-water bodies within Lake Chad region.
By definition, the MNDWI is the ratio between the green band and the middle infrared band 16 (see Equation (1)). Landsat-8 and Sentinel-2 images, but results were not satisfied because the MNDWI histogram 73 distributions were not bi-modal as the number of non-water pixels is much larger than the number 74 of of water pixel (see Figure S4). After many careful tests on different thresholds between 0 and the 75 Otsu threshold (0.3176), the MNDWI threshold for Lake Chad was set to 0.2. This is consistence  from Landsat and Sentinel-2 imagery. Figure S5 shows inundation frequency of Lake Chad de-  Figure S6).

103
The southern pool appears clearly in both MODIS and Sentinel-2 surface water extent maps, and compared to Sentinel-2. This is expected because the spatial resolution of Sentinel-2 is 250 times 111 higher than MODIS, therefore, Sentinel-2 can detect much better small water bodies than MODIS.

112
Total surface water area detected by Sentinel-2 is 14%-15% higher than that detected by MODIS.

113
The difference in the rainy seasons is lower (9%-10%) than in the dry seasons (18%-20%). The dif-114 ference mostly comes from the river, and especially from the Archipelagos where the environment Annual minimum (red) and maximum (red, light and dark blue) of surface water extent of 118 Lake Chad over the last 18 years are shown in Figure S8. It clearly shows that the southern pool of 119 Lake Chad is very stable during almost the last 20 years, at both minimum and maximum states.

120
In the northern pool, surface water extent variation is more dynamic.

121
13 Figure S9 shows the increase of permanent vegetation cover between the two pools as shown 122 in Figure S8. An increase of ∼30% of permanent vegetation is evidenced between 2001 (∼3800 123 km 2 ) and 2018 (∼5200 km 2 ). station. Figure S11 clearly shows the decreasing trend of discharge from 1950s to the end of 1980s, 131 but from the beginning of 1990s until present time, discharge is slowly increasing again.