Journal of Experimental Agriculture International

Climate-resilient food production requires decisions that are meteorologically informed, agronomically plausible and adaptive to rapidly changing conditions. Agrometeorology provides observations and forecasts of weather and climate risk; agronomy translates those risks into crop, soil and management responses; and artificial intelligence can fuse high-volume, heterogeneous data, learn complex relationships and accelerate prediction. These domains are increasingly connected, yet their evidence bases remain partly fragmented. This critical narrative review evaluates how their integration can support sustainable crop production, with particular attention to the chain from environmental observation and prediction to management action and measurable resilience outcomes. Literature published from 1 January 2000 to 8 August 2026 was examined through relevant agricultural, multidisciplinary and biomedical scholarly sources, with supplementary citation searching and verification of bibliographic metadata. The synthesis shows that the strongest case for integration is not a replacement of process knowledge by data-driven prediction, but a division of labour: meteorological models define time-varying hazard, agronomic models constrain biological plausibility and management options, and artificial intelligence improves data fusion, bias correction, pattern recognition and computational efficiency. Evidence is comparatively mature for remote sensing, yield prediction and process-based crop modelling, and is rapidly advancing for machine-learned weather prediction. By contrast, end-to-end demonstrations that connect forecast skill to farm decisions, yield stability, resource-use efficiency, profitability and environmental outcomes remain limited. Hybrid crop-model and machine-learning approaches are promising, but uncertainty calibration, domain shift, sparse ground truth, interoperability, digital inequality and data governance remain substantial constraints. Digital twins offer a useful integration architecture, although agricultural implementations are still less mature than the concept implies. Future research should prioritise multi-environment prospective validation, decision-centred forecast verification, causal evaluation of management recommendations, uncertainty-aware hybrid modelling and equitable deployment. The most defensible conclusion is that integration can improve climate resilience only when predictive accuracy is coupled to agronomic actionability, transparent uncertainty and outcome-based evaluation.