Abstract: Early warning of flood events is critical to disaster risk reduction systems in South Sudan, one of the most food insecure countries globally. Seasonal flooding of the Sudd Wetlands, formed by the White Nile in central South Sudan, has increased exponentially since its expansion in 2019. In November 2024, the United Nations Office for the Coordination of Humanitarian Affairs (OCHA) reported 1.4 million people affected by the 2024 flood event, including 379,000 people displaced, as well as increased incidences of malaria and cholera outbreaks in flooded areas (https://www.unocha.org/publications/report/south-sudan/south-sudan-floods-snapshot-29-november-2024).

Sudd water levels are largely controlled by outflows from Lake Victoria and regional precipitation. Skillful flood forecasts are challenging due to the scarcity of in-situ data in the Sudd region and the difficulty of simulating complex, dam-controlled surface water systems via continental-scale forecasting systems. However, satellite datasets such as the NOAA VIIRS Flood Mapping (VFM) product provide useful estimates of flooded area in the Sudd. This talk will present a least absolute shrinkage and selection operator (LASSO) model for skillful forecasts of monthly VFM inundation area in the Sudd Wetlands at seasonal and subseasonal timescales. The model leverages lag correlation between VFM area and satellite and Earth system model datasets. The forecasts produced by the model during 2024 and 2025 were used by decision support organizations (e.g., FEWS NET and OCHA) and the South Sudan government to prepare for and respond to major flood events in the Sudd Wetlands.