The Bezos Earth Fund has awarded a $50,000 (about R925,000) grant to Wits University’s Climate Risk Modelling project to develop AI-powered weather forecasting tools for Africa.
Led by Prof. Rendani Mbuvha of the Wits School of Statistics and Actuarial Science, the project uses advanced machine learning models—such as graph neural networks and diffusion models—to improve climate predictions in areas with limited data.
“We’re thrilled to democratize skillful AI-driven weather forecasts across the Global South,” Mbuvha said. “This support inspires us to accelerate inclusive, community-informed climate solutions.”
Africa has just one-eighth of the weather station coverage recommended by the World Meteorological Organization, hindering accurate early-warning systems. The Wits initiative will deploy automatic weather stations in underserved areas and integrate satellite, ground, and historical data to build open-source, continent-wide forecasting models.
The project also emphasizes ethical AI use, prioritizing data privacy, sustainability, and accessibility.
“This recognition strengthens our school’s research in this critical area,” said Prof. Stephen Jurisich, head of the Wits statistics school.
The grant is part of the Bezos Earth Fund’s $100 million AI Grand Challenge for Climate and Nature. Wits is now eligible for a second-phase grant of up to $2 million (about R37 million) to scale the project over the next two years.
