Lake Tanganyika basin water storage variations from 2003–2021 for water balance and flood monitoring

Lake Tanganyika in East Africa contains 17% of the free freshwater on the Earth's surface and provides important ecosystem services to ∼13 million people in the region. It is one of the great lakes in East Africa for which a significant rise in water level between 2019 and 2020 led to flooding, with major environmental consequences and social impacts. This study focused on the Lake Tanganyika basin water balance between 2003 and 2021 to assess the influence of recent climate variability on lake water level variations (due in particular to the floods of 2020 and 2021) and to explore early warnings of flooding in the lake's surrounding lowlands. This process is performed using remote sensing data. For the computation of the basin's water balance, we compared variations in the watershed total water storage (TWS) with the basin water flux calculated using rainfall, evaporation (E), evapotranspiration (ET) and discharges data. The space–time variations in rainfall, E and ET were analyzed by decomposing their time series into trend and seasonal signals and applying (only for rainfall) multivariate statistical analysis to the decomposed signals. For flood mapping, we calculated the MNDWI spectral water index from Sentinel–2 images acquired between 2017 and 2022. Our study showed that the basin water balance is closed when rainfall from Era5 is combined with E and ET from GLEV and MOD16A2, respectively. During the 2003–2021 period, over the entire watershed, water losses of ∼70 km³ due to lake E were offset by an increase in water inflows of ∼100 km³ in the rest of the watershed. During the period from 2003 to 2021, the E rate from the lake was stable overall, while the ET and rainfall mainly in the Malagarasi basin increased significantly. The surface water storage (SWS), which represents the variation in lake water volume derived from altimetry measurements, corresponds to 41.8% of the TWS, groundwater storage corresponds to 57.7% of the TWS, and the soil moisture is less than 0.5%. The TWS strongly correlated with the SWS (∼91%), with a one-month lag in the SWS variations in response to the TWS fluctuations. Therefore, the SWS in May, when the flood risk is the highest, was estimated using TWS in February, March and April with accuracies of 85%, 94% and 95%, respectively. This valuable information could be integrated into flood management tools, particularly for areas such as Gatumba city and the Ruzizi Delta Nature Reserve, which were heavily affected by the M