Satellite Gravimetry Improves Seasonal Streamflow Forecast Initialization in Africa

Abstract

West Africa is one of the poorest regions in the world and highly vulnerable to extreme hydrological events due to the lack of reliable monitoring and forecast systems. For the first time, we demonstrate that initial hydrological conditions informed by satellite‐based terrestrial water storage (TWS) estimates improve seasonal streamflow forecasts. TWS variability detected by the Gravity Recovery and Climate Experiment (GRACE) satellites is assimilated into a land surface model during 2003–2016 and used to initialize 6‐month hindcasts (i.e., forecasts of past events) during West Africa's wet seasons. We find that GRACE data assimilation (DA) generally increases groundwater and soil moisture storage in the region, resulting in increased evapotranspiration and reduced total runoff. Total runoff is particularly lower at the headwaters of the Niger River, positively impacting streamflow simulations and hindcast initializations. Compared to simulations without GRACE‐DA, hindcasts are notably improved at locations draining from large basin areas, in particular, over the Niger River basin, which is consistent with GRACE's coarse spatial resolution. The long memory of groundwater and deep soil moisture, two main TWS components updated by GRACE‐DA, is reflected in prolonged improvements in the streamflow hindcasts. Model accuracy at Niamey, Niger, the most populated city where streamflow observations are available, improved up to 33% during the flood season. These new findings directly contribute to ongoing developments in food security, flood potential forecast, and water‐related disaster warning systems for Africa.

Impacts of Fully Coupling Land Surface and Flood Models on the Simulation of Large Wetlands' Water Dynamics: The Case of the Inner Niger Delta

“…Models were driven with NASA's Modern‐Era Retrospective analysis for Research and Applications, version 2 (MERRA‐2; Reichle et al., 2017) meteorological data set, and precipitation from the Climate Hazards Group InfraRed Precipitation with Station data, version 2 (CHIRPS; Funk et al., 2015), which utilizes satellite‐based estimates and station‐based precipitation. CHIRPS station‐based component contributes to a superior spatial and temporal precipitation distribution in the continent, as demonstrated by several studies (e.g., Bichet & Diedhiou, 2018; Dembélé & Zwart, 2016; Dinku et al., 2018; Poméon et al., 2017) and, as a result, it has been widely used in monitoring water availability and forecast in Africa (Getirana, Jung, Arsenault,et al., 2020; Jung et al., 2017; McNally et al., 2019; Shukla et al., 2019). Model runs were first spun up for 60 years, allowing the models' water storage components to stabilize, followed by the 2002–2018 period experiments at a 15 min timestep.…”

“…Model experiments were evaluated with daily streamflow observations, satellite‐based altimetry, water extent, and evapotranspiration. Daily streamflow observations were made available at three gauging stations within or in the surroundings of the domain by the Comité permanent Inter état de Lutte contre la Sécheresse au Sahel (CILSS), as described in (Getirana, Jung, Arsenault, et al., 2020). Two of them are located upstream of the wetland at Koulikoro and Pankourou, on the Niger and Bagoé Rivers, respectively.…”

“…Constraining upstream boundary conditions is recommended in order to isolate errors in the wetland's physical processes evaluated in this study. Bichet & Diedhiou, 2018;Dembélé & Zwart, 2016;Dinku et al, 2018;Poméon et al, 2017) and, as a result, it has been widely used in monitoring water availability and forecast in Africa (Getirana, Jung, Arsenault,et al, 2020;Jung et al, 2017;McNally et al, 2019;Shukla et al, 2019). Model runs were first spun up for 60 years, allowing the models' water storage components to stabilize, followed by the 2002-2018 period experiments at a 15 min timestep.…”

“…The Niger inner delta model has constrained upstream the delta, over the Niger and Beni Rivers, streamflow simulations derived from an existing hydrological model for the whole Niger basin, as described in Getirana, Jung, Arsenault, et al (2020). The model run is composed of HyMAP one-way coupled with the Catchment Land Surface Model (CLSM; Koster et al, 2000) at 0.25° spatial resolution and 15 min time-step.…”

Developing a hydrological monitoring and sub-seasonal to seasonal forecasting system for South and Southeast Asian river basins

“…Future effort should be made to improve the accuracy of initial conditions estimated by the land surface model. This could include (1) assimilating ground or satellite-based observations into the land surface model (Getirana et al, 2020a, c;Wanders et al, 2014) and (2) better representation of anthropogenic influences, for example, irrigation (Nie et al, 2019) and reservoirs (Wanders and Wada, 2015;Getirana et al, 2020b), in the land surface model or hydrological model. Errors in the meteorological forecast are also a clear limitation on the forecast skill.…”