Performance Evaluation of GraphCast for Medium-Range Weather Forecasting over Brazil
A new study evaluated the performance of Google's GraphCast AI model for medium-range weather forecasting in Brazil. The research found that GraphCast's effectiveness varies depending on the season and specific climatic region. While it excels at capturing large-scale moisture transport during the summer wet season, it underperforms in predicting certain variables during the austral winter, particularly when complex weather systems are involved. AI
IMPACT Provides regional performance benchmarks for AI weather models, guiding future "tropicalization" efforts.