Crowdsourced ‘supercomputer’ enables more localised and accurate rainfall forecasts

Accurate forecasts of rainfall are crucial in Africa, where 95% of agriculture depends on highly localised rainfall. Currently, forecasts are based on satellite data and are not sufficiently accurate for small geographical areas. In order to improve local rainfall forecasts in the sub-Saharan region and enable farmers to plan their crop strategy more effectively, IBM has joined forces with TU Delft to set up the African Rainfall Project. Thanks to crowdsourced computing power pooled from volunteers in the IBM World Community Grid, forecasting models using ground observations and satellite data can now be tested.

A huge amount of computing power is needed to help small farms to predict rainfall patterns accurately. The Africa Rainfall Project is therefore encouraging thousands of volunteers worldwide to donate and pool unused computing power on their computers. Combined, they form one virtual supercomputer and are able to run highly detailed simulations of rainfall patterns. Ultimately, this will enable better forecasting and this in turn will allow farmers to plan more carefully and respond more effectively to climate change. The information could also prove useful for planning hydroelectric power plants and for bank loans and farm insurance policies.

Accuracy
This is the first time that rainfall in sub-Saharan Africa will be mapped for a whole rainy season with accuracy at a scale from 500 m to one kilometre. It is also the first time that this has happened based on crowdsourcing.

"This is a unique project, to which everyone can contribute”, says TU Delft's Professor Nick van de Giesen. “The results of the project will directly improve our understanding of rainfall distribution in Africa. Most of the rain there comes in the form of what we would call ‘summer rainstorms’, where one part of town is inundated while the other stays dry. Differences of this kind over short distances can only be simulated with a lot of computer power."

Anyone with an internet-connected computer can download a free and safe software program from IBM's World Community Grid. When the software senses that volunteers' devices are not otherwise in full, regular use, it automatically performs calculations for scientists. The results are then compared with information from The Weather Company, satellite data and ground observations. Scientists hope that this will improve local forecasting.

More information
Read more about the Africa Rainfall Project on the IBM World Community Grid website.
Read the Global Story 20,000 weather stations in Africa on the TU Delft Global Initiative website