SBIR Phase I: Subseasonal Forecasting and Climate Risk Analytics Combining Physics and AI

The broader/commercial impact of this Small Business Innovation Research (SBIR) Phase I project lies in the development of a weather forecasting and climate prediction tool for subseasonal forecasting, extreme weather events, and long-term climatological changes. The proposed technology is expected to impact a significant number of industries, including agriculture, insurance, logistics/supply chains, and the public sector, with an initial focus and market entry in the energy sector. This market is financed by large banks, carries large insurance policies that are priced based on risk, and needs to allocate resources in both the short and long term to meet customer needs and prevent service interruptions. Without these forecasting capabilities, there is a risk of drastic economic and societal costs. For example, the 2022 Pacific Northwest heat wave resulted in $8.9 billion in damages and cost the lives of 1,400 people. With 4 weeks of advanced notice, energy companies could have adequately prepared, saving lives and minimizing the damage to physical assets. The suboptimal management of weather events costs the US an average of 839 lives and $161 B/year for the last five years (cumulative >$750B), a 2.5x increase from the previous five years. This Small Business Innovation Research (SBIR) Phase I project aims to establish the feasibility of utilizing physics-informed machine learning to create probabilistic models of crucial climatological parameters and extreme weather events. A proof-of-concept demonstration focused on a single forecast variable, temperature, capable of predicting temperature anomalies 2-4 weeks in advance with 30-50% higher accuracy than the leading physics-based forecast for North America. The climate prediction models operate by using unpublished, state-of-the-art physics-informed machine learning methods and data distillation to provide high-resolution subseasonal forecasts. This SBIR project aims to (1) increase the accuracy of the temperature predictions using cutting-edge transformer networks and AI-foundation models, (2) expand predictive capabilities to extreme weather such as severe convective storms, (3) and enhance the robustness of the product by leveraging improved Bayesian modeling to capture the uncertainty of forecasts. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.