Adaptive hedging under climate and market asymmetries: a comparative deep learning and AutoGluon forecasting approach to U.S. and Chinese agricultural futures

This study examines the hedging effectiveness of agricultural commodity futures in the United States and China by integrating climate-aware probabilistic forecasting and deep reinforcement learning into dynamic risk-management frameworks. Using daily data from 2015 to 2025, covering major disruptions such as the U.S.–China trade war, COVID-19 pandemic and the Russia–Ukraine conflict, AutoGluon is employed to generate quantile forecasts that inform adaptive hedge ratios, while TD3 and SAC algorithms are used to model sequential hedging decisions. The results show that AI-based strategies improve variance reduction, tail-risk control and portfolio preservation, although their effectiveness differs across market structures. Chinese futures benefit from persistent nonlinear patterns and wider uncertainty bands, whereas U.S. markets require more conservative tail-sensitive hedges. The findings highlight the value of AI-driven hedging tools for strengthening resilience in climate-exposed and institutionally heterogeneous agricultural futures markets.