LATEST DATA CHALLENGE
SOLVING FOR FROST RISK
OVERVIEW OF THE CHALLENGE
Frost is one of the most damaging weather risks to agriculture in California and across the U.S., causing losses that exceed those from other weather-related events. Because traditional frost protection relies heavily on manual measurements and local experience, this challenge examined how advanced analytics and high-performance computing can deliver earlier, more accurate frost forecasts to help growers take timely action. The Frost Risk Forecasting Data Challenge investigated how data-driven methods could provide earlier and more reliable frost predictions for growers. University teams developed machine learning and spatial models using localized microclimate weather data from California, with approaches designed to be transferable to other regions. The goal was to improve frost risk detection and support proactive decision-making in agricultural operations.
CHALLENGE OVERVIEW
Launched in November 2025, the Frost Risk Data Challenge invited interdisciplinary university teams - including undergraduates, graduate students, faculty, and staff - to develop predictive models using real-world datasets and supercomputing resources. The challenge was supported by the San Diego Supercomputing Center and the National Data Platform. Projects were evaluated based on accuracy, creativity, and solution effectiveness.
Winning teams
All teams that submitted valid and reproducible entries received a Certificate of Participation, while the top three teams earned an official F3i Digital Credential, verifiable through an open badging system for professional and academic use. Top teams were also invited to present their work to F3i’s industry and grower partners and may be invited to participate in future sprint design teams and commercialization cohorts. Cash prizes will be awarded to the top three teams: $1,500 for 1st place, $750 for 2nd place, and $400 for 3rd place.
1
FrostByte, San Jose State University
Probabilistic forecasting system built on XGBoost, with spatial validation and an interactive decision-support dashboard
2
TylerAziz, University of California, Davis and California Polytechnic State University, San Luis Obispo
Physics-informed tree-based machine learning using Random Forest and XGBoost with strict station-grouped cross-validation to ensure spatial generalization.
3
AgriFrost AI, Independent Contributor
Systematic feature engineering with class-balanced gradient boosting (LightGBM) for probabilistic frost risk forecasting.
