Uncertainty Quantification of Climate Impact on Tobacco Yield: A Grid-Based Multimodal AI Framework for Dynamic Risk Stratification and Adaptive Management

<ns5:p>Increasing climate uncertainty poses a significant threat to the yield stability of high-value crops such as tobacco (Nicotiana tabacum L.). Existing methods have inherent limitations in quantifying these uncertainties at a fine-grained (sub-plot) level and in guiding dynamic, adaptive management, thus struggling to address complex and variable field conditions. To counter this challenge, this study proposes an innovative grid-based, multimodal AI framework designed to precisely quantify the impacts of climate change on tobacco yield and to enable dynamic risk stratification and adaptive management. The framework is centered on a Multi-modal Large Model (MLM) as its cognitive core, which models the complex interactions of the crop-environment system by deeply fusing multi-source, heterogeneous data from UAV remote sensing, meteorological time-series, and soil sensors. Results from large-scale field trials in the core production region of Guangxi demonstrate that the framework can achieve hourly dynamic risk assessment. Compared to the traditional mechanistic model (DSSAT), its disaster response efficiency is improved by nearly five-fold, and it significantly reduces the average yield reduction rate from 18.0% to 7.1% (a 60.6% decrease). Ablation studies prove that the MLM, as the core engine, increases the accuracy of risk assessment and decision-making from 78.2% to 94.5%. Furthermore, by introducing a human-in-the-loop mechanism, the success rate of critical interventions reached as high as 99.2%. Cross-regional back-testing on an independent dataset (91.7% accuracy) also validated the framework’s strong generalization capability. This study provides a powerful, interpretable, and quantitative decision-making tool for implementing precision agriculture management under uncertainty, showcasing the immense potential of advanced AI technology in ensuring the resilience and sustainability of key cash crop supply chains.</ns5:p>