抄録
Genotype-by-environment (G×E) interactions represent a fundamental constraint in stabilizing wheat yield under increasing climate variability, yet their quantitative characterization and translation into adaptive variety deployment strategies remain insufficiently resolved. Artificial intelligence (AI) has emerged as a powerful paradigm for integrating heterogeneous agricultural data and capturing nonlinear G×E relationships, enabling a shift from empirical assessment toward predictive and decision-oriented crop management. This review synthesizes recent advances in AI-driven wheat adaptability evaluation and precision planting strategies, with emphasis on the integration of multi-source datasets, including meteorological, soil, remote sensing, and phenotypic information, through machine learning and deep learning frameworks. We further examine how AI models are employed to quantify genotype performance, predict yield stability, and assess stress tolerance across diverse environments, and how these outputs are translated into variety recommendation and site-specific planting optimization systems. Despite substantial progress, key limitations persist in data standardization, cross-regional model transferability, and the interpretability of black-box models, which hinder large-scale operational deployment. Emerging directions including multimodal data fusion, transfer learning across environments, and explainable AI (XAI) are discussed as critical enablers for improving model robustness and transparency. By consolidating current methodologies into a unified analytical framework linking G×E modeling with adaptive decision-making, this review provides a conceptual basis for advancing predictive breeding systems and precision wheat production under climate uncertainty.
