Recent advances in multimodal AI, including deep learning, foundation models, and vision–language models, combined with rapid progress in remote sensing, are transforming how agricultural systems are monitored and managed under climate stress. These approaches integrate data across scales, from satellites and UAVs to in-situ sensors, enabling more precise and timely decision-making with increasing relevance to practical, farm-level outcomes. However, major challenges remain. Agricultural data are often heterogeneous and poorly standardized, limiting interoperability, scalability, and model generalization across environments. Issues such as data quality, uncertainty, missing data, and transferability hinder the transition from proof-of-concept models to robust, field-deployable solutions. This session invites contributions addressing these gaps through data-centric and model-centric approaches, including multimodal data fusion, standardized frameworks, foundation models for agriculture, and high-quality reference datasets. We encourage work that bridges sensing and decision-making while supporting actionable, economically viable solutions, enabling scalable, climate-resilient agricultural systems and improving real-world impact.