A BSC team wins second prize in a challenge to improve temperature and precipitation forecasts using AI
WMO launched this initiative with the aim of improving sub-seasonal temperature and precipitation forecasts with machine learning and artificial intelligence.
Researchers from the Department of Earth Sciences of the Barcelona Supercomputing Center (BSC) have won the second prize in the Challenge to improve sub-seasonal to seasonal predictions using artificial intelligence (S2S AI challenge). The prize consisted of a contribution of CHF 10,000.
The World Meteorological Organization (WMO) has launched this initiative that aims to improve, through artificial intelligence and/or machine learning, the current precipitation and temperature forecasts for 3 and 6 weeks into the future from the best computational fluid dynamics models available today. The challenge began last June and ended on October 31, although the official results have not been made public until now. In total, 9 international research teams have participated.
The BSC team has been led by Lluís Palma and Llorenç Lledó, along with Sergi Bech, Andrea Manrique and Carlos Gómez, all from the Earth System Services group. This group already has several projects in this line, but, as Lledó assures us, “until now we had worked with statistical methods and now we have taken the leap using machine learning methods. Specifically, we have used classic machine learning methods trained with the forecasts generated by the European Meteorology Center. This has allowed us to understand what biases this dynamic model has and how they can be corrected to obtain better forecasts.”
Challenge to improve S2S Predictions
Improved sub-seasonal to seasonal (S2S) forecast skill would benefit multiple user sectors immensely, including water, energy, health, agriculture and disaster risk reduction. The creation of an extensive database of S2S model forecasts has provided a new opportunity to apply the latest developments in machine learning to improve S2S prediction of temperature and total precipitation forecasts up to 6 weeks ahead, with focus on biweekly averaged conditions around the globe.
This challenge is part of the Subseasonal-to-Seasonal Prediction Project (S2S Project), coordinated by the World Weather Research Programme (WWRP)/World Climate Research Programme (WCRP), in collaboration with the Swiss Data Science Center and the European Centre for Medium-Range Weather Forecasts (ECMWF).
- All codes and forecasts of the method used by the BSC team are available here: https://renkulab.io/gitlab/lluis.palma/s2s-ai-challenge-bsc
- Check results here: https://s2s-ai-challenge.github.io/#leaderboard