ABSTRACT Accurate evaluation of maize performance across diverse environments is essential for improving yield stability and cultivar adaptation under a changing climate. However, conventional breeding approaches often struggle to address multi‐trait trade‐offs and complex genotype‐by‐environment (G × E) interactions while explicitly incorporating meteorological information. In this study, we developed an integrated framework that combines mixed‐model genetic evaluation, information‐theoretic multi‐trait indices, and machine‐learning‐based environmental prediction. Best linear unbiased prediction (BLUP) was used to estimate genotypic values and variance components from 146, 286 raw observations representing 34 maize genotypes evaluated across 53 environments during 2020–2024. Yield and disease resistance were summarized using a mutual‐information‐weighted yield performance index (YPIMI) and an entropy‐weighted disease resistance index (DRI), respectively. A random forest model was then used to predict YPIMI from meteorological and soil variables, yielding a cross‐validated R 2 of 0. 687 and an in‐sample R 2 of 0. 841. Variable‐importance analysis identified evapotranspiration, precipitation, BLUP‐derived genetic merit, shortwave radiation, and soil pH as the major drivers of performance variation. The proposed framework effectively distinguished genotypes that combined favorable yield potential with disease resistance and provided an interpretable basis for environment‐specific evaluation. Overall, this study offers a practical approach for integrating genetic merit, multi‐trait performance, and environmental information to support maize cultivar assessment and deployment under variable agroclimatic conditions.
Yue et al. (2026) studied this question.