Temporal instability of genotype adaptation limits environment targeting in dryland foxtail millet production systems

Introduction Genotype adaptation in dryland cropping systems is frequently challenged by environmental variability across both space and time. However, whether geographically defined testing environments provide repeatable information for genotype selection remains poorly understood. Methods We evaluated grain yield adaptation of foxtail millet (Setaria italica (L.) P. Beauv.) using replicated trials conducted at 20 locations across northern China during two growing seasons (2023–2024). A connected panel of 13 common genotypes evaluated across 40 site-year environments (1,560 plot observations) was used to quantify repeatable and non-static genotype-by-environment interactions (G×E). Heterogeneous-residual mixed models, reduced-rank covariance modelling, and environmental characterization were applied. Results Genotype-related variation was mainly attributable to genotype × location × year (50.3%), genotype × location (34.6%), and genotype × year (15.1%) interactions, with these percentages representing components of genotype-related rather than total phenotypic variance. The mean pairwise genotypic correlation across the 40 site-year environments was 0.037, and 48.7% of correlations were negative. The mean same-location cross-year genotypic correlation was only 0.027, with negative correlations at 12 of the 20 locations. Long-term climatic characteristics, soil properties, and seasonal anomalies showed limited correspondence with latent genotype-response dimensions. Genotypes differed in their environmental responsiveness despite relatively stable rankings for broad performance. Discussion These results indicate substantial season-specific genotype responses and limited evidence for consistently transferable location-specific selection information within the sampled two-year network. Multi-year testing and environmentally representative sampling are therefore important for improving genotype selection and deployment in variable dryland production systems.