Surface water dynamics of Africa: Analysing continental trends and identifying drivers for major lakes and reservoirs

Abstract

ABSTRACT Africa has the most dynamic demographic development worldwide. Current projections predict a population of > 3 billion people by the end of the century. Sub-Saharan Africa alone will likely see a 40% increase in population between 2020 and 2050. Although it is well known that large parts of Africa are in a constant state of water stress, its surface water resources remain understudied. This study analyses long-term trends of surface water in Africa. It identifies causal impacts on major lakes and reservoirs for the timeframe 2003–2020, as well as dynamic and causal similarities between the various lakes. For this, a set of daily time series based on Earth observation is employed. Global WaterPack data is used for a daily uninterrupted time series of the continent’s surface water area. Additionally, an array of relevant independent variables, namely precipitation, total evapotranspiration, groundwater, soil moisture, and gross primary productivity (GPP) in different land use areas is analysed. For causal identification, the Peter and Clark Momentary Conditional Independence algorithm is used. Findings show that > 42% of African countries and > 34% of African ecoregions experience shrinking surface water area. Over 80% of investigated surface water bodies are driven by the surface water in their upstream subbasins and GPP in agriculturally used areas. About 85% of investigated lakes are significantly driven by agricultural usage, often in the form of water abstraction, as referenced regional studies confirm. Our analysis demonstrates the feasibility of conducting causal analyses of surface water dynamics using Earth observation data. Dynamically similar lakes are often impacted by the same drivers, forming regional lake clusters. Considering the causes identified may greatly help adapt strategies for sustainable development. A causality analysis to identify drivers of surface water dynamics has, to our knowledge, never been performed before on this scale and at such high temporal resolution.