Global Rice Array: from Environmental Diversity to climate-smart breeding strategies

Rice breeding for climate adaptation requires a robust framework to evaluate genotype performance across diverse and interacting production environments. However, most breeding pipelines remain constrained by limited environmental representation and insufficient understanding of genotype by environment (G x E) interactions. This study introduces the Global Rice Array (GRA), a coordinated multi-environment trial network specifically designed to capture major sources of environmental heterogeneity in lowland rice systems and to support environment-targeted breeding decisions. Unlike conventional location-based testing, the GRA integrates standardized trials across Asia, Africa, and Latin America, enabling a functional characterization of environments and genotypes at global scale. A total of 58 diverse rice genotypes nominated by breeders across Asia, Africa, and Latin America were evaluated. The set includes genotypes tolerant to heat, drought, cold, salinity, and major biotic stress. Fifty-one field trials were conducted across 12 countries in three continents from 2018 to 2020. Grain yield, days to flowering, and plant height were measured across genotypes, sites and seasons. Using the additive main effect and multiplicative interaction (AMMI) model, environment explained 56% of grain yield variation, followed by G x E interaction (26%) and genotype (18%), confirming the limited predictability of performance from single-environment trials. Importantly, environment grouping revealed seven functional environment groups (EGs) structured by climate, water regime, and biotic pressures rather than geographic origin. This demonstrates that lowland rice production environments can be more effectively defined as integrated stress environments rather than administrative or regional units. Low-yielding EGs achieved only 58% of the yield of high-yielding EGs and reflected climate-water and biotic stress interaction constrains (higher night temperatures >24 °C, lower solar radiation, higher pest and disease pressure). Cluster analysis further identified four genotype groups with contrasting adaptation strategies, including broadly adapted high-yielding genotypes and specifically adapted stress-resilient genotypes. Broadly adapted genotypes contributed to yield stability in favorable environments, whereas specific adaptation was associated with stress tolerance, including escape strategies associated with variation in flowering time. Phenology analyses further showed that maturity groups were stable across environments, but that stress consistently delayed development and reduced yield across genotypes. Overall, our results demonstrate that climate-smart rice breeding should integrate environment characterization, genotype × environment analysis, and complementary use of broadly and specifically adapted germplasm. Breeding pipelines must explicitly target functional environmental groups and combine stress-resilient traits with appropriate maturity types to ensure productivity and stability under evolving climate conditions.