Subseasonal-to-seasonal forecasting is critical for mitigating risks, enhancing decision-making, and optimizing outcomes in various sectors. For example, in agriculture, knowing the likelihood of a drought or heavy rainfall two weeks in advance can allow farmers to adapt irrigation and planting plans to mitigate the impact of adverse weather conditions on their crops. In the energy sector, anticipating extreme temperature or wind patterns can enable decision-makers to optimize resource allocations.

Subseasonal-to-seasonal forecasts also play a vital role in disaster preparedness and response. Predicting the likelihood of extreme weather events, such as hurricanes, heatwaves, or heavy precipitation weeks in advance enables authorities to issue early warnings, evacuate vulnerable areas, and implement measures to reduce the impact of disasters on communities and infrastructure. This proactive approach can save lives, minimize property damage, and improve overall resilience to natural hazards.

Among climate data researchers, subseasonal-to-seasonal forecasting contributes to improving climate modeling. By analyzing atmospheric, oceanic, land, and glacial conditions over intermediate time scales, scientists can gain insights into the mechanisms driving climate variability and change. This, in turn, informs the development of more accurate long-term climate models and predictions, which enables policymakers and other stakeholders to develop effective climate change mitigation and adaptation strategies that support community resilience.

In recent years, advancements in technology, such as improved numerical weather prediction models and enhanced computing power, have enhanced the skill and reliability of subseasonal forecasts. Machine learning algorithms and data assimilation techniques have also been employed to better incorporate observational data and improve forecast accuracy. Yet, accurate climate predictions remain a major challenge for the climate community because of inherent uncertainties in atmospheric, oceanic, land, and glacial processes; model biases; and the complex interactions between different components of the Earth system. Addressing these challenges requires ongoing research and collaboration among scientists and stakeholders.

Given these challenges, LEAP’s January 2025 Hackathon will:

• focus on improving the prediction of subseasonal-to-seasonal climate time scales (from a few days to a few weeks) using machine learning;
• provide historical data that will be used to train the machine learning algorithms and for validation of the strategy on unseen data (years left out from the training); and
• include a “green” computing component where the simplest models with high skill will be rewarded, as it aligns with our mission to reduce greenhouse gas emissions. 

For those with climate and data science research backgrounds, the following resources provide additional background:

• Subseasonal-to-Seasonal Prediction Project
• “Outcomes of the WMO Prize Challenge to Improve Subseasonal to Seasonal Predictions Using Artificial Intelligence”

For those coming from other disciplines and areas of expertise, the following link provides additional definitions and background:

• To learn more about climate modeling at LEAP, check out “LEAP: Our Story”