<p>Rice cultivation faces increasing challenges from climate variability, threatening global food security. This study presents a climate-aware, IoT-enabled framework that integrates explainable artificial intelligence for real-time assessment of rice-growing suitability using a normalized favourability score. The proposed approach uses a primary dataset from eight Northern Bangladesh districts covering the Aman, Aus, and Boro varieties, with feature engineering informed by Bangladesh Rice Research Institute (BRRI) expertise to generate comprehensive favourability metrics. The core contribution is the development of optimized tree-based ensemble methods, Stacked Ensemble, Optimized Random Forest, and Optimized Extra Trees, designed to mitigate overfitting while achieving high predictive accuracy. Experimental results demonstrate robust performance, with the Stacked Ensemble achieving a root mean square error of 0.0159, an R² of 0.9365, and a mean absolute error of 0.0075; the Optimized Random Forest attaining a root mean square error of 0.0160, an R² of 0.9358, and a mean absolute error of 0.0072; and the Optimized Extra Trees yielding a root mean square error of 0.0162, an R² of 0.9338, and a mean absolute error of 0.0072. The deployed IoT infrastructure enables real-time data acquisition, while explainable AI ensures transparency and user trust. This framework advances precision agriculture by providing an effective decision-support system for climate-resilient rice production and sustainable food security. Code publicly available at: <a href="https://github.com/abidhasanrafi/rice-yield-predictor">https://github.com/abidhasanrafi/rice-yield-predictor</a>.</p>