Sunflower yield modeling with explainable artificial intelligence: Historical weather impacts across half a century of American production

Abstract This study applies explainable artificial intelligence (XAI) to analyze the impact of inter‐year variation in weather conditions on yields of oilseed sunflower ( Helianthus annuus L.) across the United States. By integrating historical county‐level yield data from 1976 to 2022 with monthly meteorological data over the same period, we identified key weather predictors influencing sunflower yields at national and state levels along with critical yield‐sensitive threshold temperature and precipitation values that predict reduced yield. Across the sunflower production range, the most critical climate variables identified are July and August maximum temperatures and total precipitation, reflecting yield vulnerability to summer heat waves and drought during budding and flowering. Secondarily, overly cool temperatures during spring planting and establishment (May–June) reduce yields, as do overly cool end‐of‐season temperatures during seed maturation and harvest (September–October), indicating risk of frost or insufficient growing degree days to support plant development. Winter precipitation and temperatures were also detected as important to overall yield dynamics, in particular where wetter winters benefitted yields. Specific temperature and precipitation tipping points vary across the geographic extent of production, but align with existing agronomic knowledge. Our XAI approach enhances model transparency, offering valuable insights for farmers and policymakers to develop adaptive strategies for sunflower cultivation under climate change. Future research incorporating additional factors like soil characteristics and agricultural practices can further refine yield predictions.