Important drivers of East African monsoon variability and improving rainy season onset prediction

Abstract

Monsoon rain and its year-to-year variability have a profound influence on Africa’s socio-economic structure by heavily impacting sectors such as agricultural and energy. This study focuses on major drivers of the east African monsoon during October-November-December (OND) which is the standard time window for the onset of the rainy season, be it unimodal or bimodal. Two drivers viz. Indian Ocean Dipole (IOD) and El Niño Southern Oscillation (ENSO) both separately indicate very strong positive connections with monsoon (OND) rain not only in the OND season with zero seasonal lag, but the signal is also present even taking IOD and ENSO a season ahead. A compositing approach is applied that can additionally identify strong signals when different combinations of ENSO and IOD phases act as confounding factors. Results of precipitation anomaly suggest that when IOD and ENSO are both on the same phase in July-August-September (JAS), a significant OND rainfall anomaly occurs around the east African sector: A deficit (excess) of OND monsoon rain occurs when both drivers are in a negative (positive) phase during JAS. A location Kibaha in Tanzania, for which station data are available, is considered for a more in-depth analysis. The uncertainty range in cumulative OND rainfall is also reduced to a large degree when IOD and ENSO phases are both negative in JAS. These results can be used for prediction purposes and interestingly, that criterion of IOD and ENSO being of same phase in JAS was again matched in 2022 (both negative) and hence it was possible to deliver early warnings for a deficit in rainfall a season ahead. Techniques to compute the monsoon onset as determined by meteorological services such as the Tanzania Meteorological Authority rely on various thresholds, which may also vary by country. To overcome some of the issues with thresholds-based techniques, other definitions of ‘onset’ take into account cumulative rainfall amount and such technique has also been tested and compared. In both approaches, late (early) onsets dominate in years when ENSO and IOD are both negative (positive) during JAS. In these cases, it is therefore possible to provide an estimation of cumulative rainfall and onset for OND in terms of average, median value, range and distribution of rainfall one season in advance. Such results have implications for optimizing agricultural, water and energy management, also mitigating possible severe production losses, which would impact the livelihoods of millions of Africans.