Exploring and Anticipating Extreme East African Short Rains

During the 16 rainy seasons since October-November-December (OND) of 2016, the eastern Horn of Africa (eHorn) has experienced an exceptional sequence of extreme rainy seasons, with 8 dry seasons, 6 wet seasons, and just 2 normal rainy seasons (Figure 1). In 2016/17 and 2020/22  climate change-enhanced west Pacific sea surface temperatures (SST) amplified the influence of La Nina, leading to hazard two-season and five-season drought sequences that forced millions of people into starvation as crops failed and millions of livestock perished. Drought conditions during 2020-22 were exceptionally intense, persistent, extensive and hot, devastating livelihoods and producing repetitive, debilitating and cumulative shocks to herds, crops, water availability, and household incomes. More than eight million livestock died and millions of people faced the threat of starvation, and emergency humanitarian relief efforts required more than $2 billion USD. Extreme rains in March-April-May (MAM) of 2018, due to a Madden-Julien Oscillation brought flooding and displacement, while positive Indian Ocean Dipole (IOD) conditions in 2019 and 2023 contributed to excessive rains, flooding and displacement. These extremes provide potential opportunities for prediction, proactive risk management, and improved agricultural and water management outcomes. Here, focusing on OND rains, we explore the use of a new Indo-Pacific Heating Gradient indicator to understand and predict extreme eHorn rains.

Methods
This data set draws from five widely used sources: the Climate Hazard Center Infrared Precipitation with Stations archive (CHIRPS), the Centennial Trends Gridded Rainfall archive, the NOAA Extended Reconstruction sea surface temperature data set (version 5), ERA5 Reanalysis atmospheric heating values, and seasonal SST forecasts from the North American Multi-Model Ensemble (NMME). While all of these data are publicly available, we pull together in this dataset all the salient time series supporting the basic results in our paper. All data are for October-November-December (OND).