Assessing climate risks to crops requires understanding how climatic thresholds and interactions vary across environments. However, most climate-risk assessments rely on generalized thresholds despite evidence that crop responses differ among regions, growth stages, and climatic conditions. Using wheat-growing systems across India, we applied an explainable machine-learning framework to analyze climate–yield relationships during 2010–2020 and identify region- and growth-stage-specific critical climatic thresholds, post-threshold yield responses, and compound climate interactions. Results revealed pronounced spatial heterogeneity in wheat responses to climate. Minimum (Night-time) temperature (February) during reproductive development (booting, heading, flowering) stages emerged as the dominant climatic constraint across most agro-climatic regions. Corresponding climatic thresholds ranged from approximately 10.5 °C in cool semi-arid systems to 16 °C in tropical environments, whereas the highland region exhibited positive yield responses above approximately 8 °C, indicating substantial regional differences in thermal adaptation under current climatic conditions. Yield sensitivity beyond these thresholds also varied significantly, from approximately 6% yield loss per °C in the cool semi-arid region to approximately 1% per °C in the humid plains. Interaction analyses further demonstrated that wheat responses cannot be explained by individual climatic drivers alone. Despite increasingly negative impacts of individual climatic variables beyond their critical thresholds, many compound interactions remained positive, indicating that climatic continuity across consecutive crop developmental periods can partially buffer thermal stress. Conversely, larger inter-stage thermal transitions across climatic thresholds were associated with stronger negative interactions and greater yield penalties, emphasizing the importance of thermal continuity for wheat productivity. Overall, wheat productivity is governed by region-specific climatic thresholds, post-threshold yield sensitivities, and compound climate interactions. Incorporating these spatially explicit climatic thresholds and their responses into climate-risk assessments can improve adaptation planning for increasingly variable and warming climates.
No takes yet. Share an insight, caveat, or question.
Goswami et al. (2026) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: