Global intercomparison of satellite-derived variability in reservoir storage

Abstract Understanding surface water reservoir storage variability enables vital insights into water management techniques, the impact of humans on global water storage, and climate-driven influences on water availability. In the past few years, recent improvements in satellite data availability and cloud computing have led to rapid growth in satellite-derived global observation of reservoir storage fluctuations, yet little research has directly compared the growing number of published storage datasets. Here we perform a first global intercomparison between five global satellite-derived reservoir storage datasets, namely GLWS, GRS, GloLakes, GRDL-Y and GRDL-L. Overall, we find generally good agreement in relative storage time series (median RMSE between datasets = 8.7% of capacity), with little substantial difference in agreement between datasets. Agreement in absolute storage is much worse (median = 19.4%), and for absolute storage, GloLakes has higher error than the other datasets. We find that agreement is worse in highly variable reservoirs, new reservoirs, and, notably, in developing countries. All datasets agree that, globally, construction of new reservoirs is driving a net increase in reservoir storage over 1999-2018, yet there is disagreement in the magnitude of these trends. Overall, our results lend confidence to the utility of satellite-derived global reservoir storage datasets for water management applications. We suggest that future research should involve reducing errors in water area observations, increasing consistency in which reservoirs are observed, and improving storage algorithm performance especially in developing areas.