Aurora Hunter: A Two-Stage Framework for Probabilistic Visibility Forecasting
Researchers have developed "Aurora Hunter," a two-stage framework designed to improve the forecasting of aurora borealis visibility. The system first predicts the likelihood of an aurora occurring using physics-based features and then forecasts the probability of clear observation conditions, considering cloud cover and lunar illumination. This decoupled approach achieved a high ROC-AUC of 0.937 on test data, outperforming single-stage baselines and demonstrating strong generalization across different sites. AI
IMPACT Enhances predictive accuracy for space weather phenomena, benefiting scientific research and tourism.