The African Rainfall Project (ARP) of the World Community Grid is running again! We are excited to share an update on our progress since Fall 2024.
Introduction and Updates
95% of farms in Sub-Saharan Africa depend on rainfall to grow their crops successfully. However, due to changes in weather patterns resulting from climate change, significant decreases in rainfall are expected across the continent. By leveraging a high-resolution (1km) application of the Weather Research and Forecasting Model (WRF), we are building simulations that will significantly improve rainfall predictions in the region. Once complete, ARP predictive models will help scientists build more accurate weather forecasts, which will ultimately support farmers in timing their crops to maximize their yield.
The large-scale simulations that are being built by the ARP require an extraordinary quantity of computing power, which would not be possible without WCG volunteers. Rainfall simulations are now about two thirds completed, and are expected to be fully complete within 12 months. The simulations uniquely cover the continent at a very fine resolution (1km), allowing simulation of the intense convective rainstorms that bring about 80% of the rain. As of July 7, 2025, volunteers have helped to generate over 10.5M results for this project. We greatly appreciate your ongoing support!
Where AI Falls Short: The Ongoing Need for the African Rainfall Project
As artificial intelligence (AI) capabilities rapidly advance, a common question arises: could AI-driven weather predictions eventually replace the complex and costly calculations of physical weather models?
To understand the context, consider ERA5: a reanalysis project within the European Centre for Mid-range Weather Forecasts (ECMWF). It boasts a spatial resolution accurate to 31 km and global coverage since 1940, including numerical weather models, all ground observations, and satellite data. ERA5 is an extraordinary scientific achievement and currently represents the most accurate estimate of global atmospheric conditions over the past 85 years, making it a cornerstone for climate research.
Despite its success, ERA5 came at a high cost: about twenty billion dollars. To save money without losing progress, the next logical step would be to use AI to mimic ERA5 and predict weather patterns across the globe. This is exactly what is being done by most large AI companies who have found that running such AI models costs millions instead of billions.
While AI presents a promising solution, unfortunately, ERA5-based rainfall predictions do not perform well for Africa specifically. When compared against ground data from the TAHMO network over Kenya, even at monthly time scales, ERA5 does not capture the anomalies that make the difference between drought and flood. Therefore, if all forecasts are AI-based and trained on ERA5, African rainfall will be poorly covered by the network.
With the African Rainfall Project, we have built an “ERA5” that is relevant to Africa. Though its time span may be limited, the right processes will be captured at the right scale for the population. A logical next step would be to see if AI can be used to mimic the physical model that was generated thanks to the World Community Grid and its thousands of volunteers around the world. If this succeeds, then weather forecasting in Africa will have taken a huge step forward.