With the aim of improving the seasonal prediction of rainfall in Sub-Saharan Africa, with a particular focus on the regions where the CRD hubs are located, a group of ALBATROSS researchers, led by Paolo Ruggieri, from the University of Bologna, has tested a novel methodology for the seasonal forecast of precipitation in sub-Saharan Africa.

Having access to skilful seasonal rainfall predictions is instrumental to support the development of climate services, inform decision-making and accelerate climate adaptation in Sub-Saharan Africa. The aim of this study was to test and develop a model combining traditional weather prediction models with artificial intelligence (AI). The focus is on using AI to better predict large-scale modes of climate variability, known as “teleconnections”, such as El Niño Southern Oscillation (ENSO), the Indian Ocean Dipole (IOD), and the Atlantic Niño, which usually have a strong influence on rainfall in specific regions of Africa.

This piece of work builds on the baseline established in a previous analysis made in the project, which provided a comprehensive assessment of seasonal rainfall forecasting skill over Sub-Saharan Africa. Based on those results, the present study focuses on regions and seasons characterised by relatively high predictive skill, with a particular focus on precipitation. These methods include approaches that combine European Centre for Medium-Range Weather Forecasts (ECMWF) systems with AI-driven models, leveraging machine learning techniques to enhance the prediction of region-specific precipitation drivers. Such hybrid frameworks were expected to improve the accuracy of forecasts, thereby increasing their utility for agricultural planning.

The methodology developed produces daily and monthly data for temperature and precipitation, highlighting the role of teleconnections as key sources of seasonal predictability was then tested in three case studies to measure their accuracy: one for the East Africa region, one for the West Africa, and the other for South Africa. In short, while the AI-enhanced approach did not always outperform traditional methods, especially in South Africa during the austral summer, it showed clear benefits in some regions, particularly East Africa during the short rains (October-December). The researchers suggest that further research is needed to expand the approach to other climate patterns and regions.