Researchers at Assam Agricultural University evaluated 31 rice genotypes across 21 environments, spanning seven locations over three consecutive growing seasons from 2022 to 2024. The environments captured the state’s striking agro-ecological diversity, including flood-prone areas where submergence-tolerant Sub1 varieties are especially valuable. Grain yield data were subjected to combined analysis of variance and a mixed-model framework that produced best linear unbiased predictions, or BLUPs, for each genotype in each environment.

The combined ANOVA revealed highly significant effects of genotype, environment, and their interaction, confirming that varietal performance shifts substantially depending on where and when the crop is grown. This genotype × environment interaction, often abbreviated GEI, is the central challenge of varietal recommendation in heterogeneous regions. A single average yield figure across all sites can obscure the fact that a variety ranks first in one location and near the bottom in another.

To address this, the team applied the WAASB index, which computes the weighted average of absolute scores from the BLUP-based GEI matrix. In practical terms, WAASB rewards genotypes that combine high mean yield with minimal fluctuation across environments, giving breeders a single stability metric grounded in mixed-model theory rather than simpler, older stability statistics.

WAASB-based selection identified BRRI Dhan 69 as the most stable genotype in the entire panel, meaning it maintained its performance level with the least deviation across all 21 test environments. Four additional varieties, Dholi, Surma Dhan, Ranjit Sub1, and CR Dhan 801, combined high yield with broad adaptation, marking them as strong candidates for widespread recommendation throughout Assam. Notably, several of these carry the Sub1 submergence-tolerance trait, making them particularly relevant for flood-prone districts where complete submergence during the growing season can wipe out an entire crop.

The researchers complemented the stability analysis with GGE biplot analysis, a graphical method that decomposes genotype-plus-genotype-by-environment variation into principal components and displays genotypes and environments together. Rather than seeking one variety that performs acceptably everywhere, GGE biplot analysis reveals mega-environments, clusters of locations where particular varieties hold a consistent advantage. The biplots exposed distinct mega-environments within Assam, and Labanya, the highest-yielding genotype overall, emerged as the winner across the largest of these clusters.

Other varieties showed clear specific adaptation. Ranjit Sub1, CR Dhan 802, and IR64 Sub1 excelled in particular environment groups rather than across the whole state, underscoring that the optimal deployment strategy for Assam is not a single universal variety but a portfolio matched to local conditions. This finding has direct implications for seed systems and extension services, which can now target specific Sub1 varieties to the districts where their advantages are greatest.

The study also evaluated the test environments themselves, identifying locations that combine high discriminative ability with representativeness of the broader target population of environments. Such locations are the most informative sites for future multi-environment trials, because they maximize the chance of detecting real genotypic differences while reflecting conditions farmers actually face. Optimizing the testing network in this way reduces cost and accelerates the release of improved varieties.

The analyses were conducted using the metan R package, which integrates BLUP-based stability indices and biplot tools within a single workflow, and the researchers emphasize that combining WAASB and GGE biplot approaches provides a more complete picture than either method alone. WAASB answers the question of which variety is most dependable on average, while GGE biplots answer which variety wins where. Together they support evidence-based varietal deployment tailored to the realities of climate-vulnerable rice landscapes.

For Assam’s rice farmers, the practical message is significant. Broadly adapted varieties such as BRRI Dhan 69, Dholi, Surma Dhan, Ranjit Sub1, and CR Dhan 801 offer a dependable foundation for the state’s rice production, while specifically adapted cultivars can be matched to their winning mega-environments. As flooding becomes more frequent and erratic under a changing climate, this kind of precision in varietal recommendation will be central to maintaining food security in flood-prone rice systems across Assam and comparable regions.

The distinction between broad and specific adaptation lies at the heart of modern plant breeding in heterogeneous regions. Broadly adapted cultivars deliver acceptable performance across a wide range of conditions, which simplifies seed multiplication, distribution, and extension messaging. Specifically adapted cultivars, by contrast, may achieve substantially higher yields than any broadly adapted competitor within a narrower band of environments, but they lose that advantage elsewhere. Neither strategy is inherently superior; the right choice depends on how predictable local growing conditions are and how finely a state’s seed system can segment its recommendations. The Assam study illustrates why both classes of cultivar deserve a place in a coherent deployment strategy.

The statistical machinery behind these conclusions deserves some explanation. Mixed-model approaches allow researchers to treat environments, and sometimes genotype-by-environment interaction terms, as random effects, which improves the precision of genotype estimates by borrowing information across the entire trial network. The resulting best linear unbiased predictions are considered more reliable than simple arithmetic means because they adjust for the unequal quality of individual trials, accounting for factors such as imperfect recovery of genetic variance and heterogeneous error across sites. This is why BLUP-based indices have become standard practice in large multi-environment trials worldwide.

WAASB extends this framework by treating the interaction scores extracted from the BLUP matrix not as noise to be discarded but as information about instability to be quantified. Each genotype receives a weighted average of the absolute values of its interaction scores across the principal component axes retained in the model, with weights declining as each axis explains less of the interaction variation. A genotype that simultaneously shows a high BLUP for yield and a low WAASB value occupies the ideal corner of the performance-stability trade-off, and this joint interpretation is what allowed the study to separate dependably high performers from varieties whose average yields concealed erratic behavior.

The submergence-tolerance angle carries particular weight for Assam. The Sub1 trait, derived from an Indian landrace and introgressed into popular modern varieties through marker-assisted backcrossing, allows rice plants to survive complete submersion for roughly one to two weeks by suppressing growth during flooding, a quiescence strategy that conserves energy until floodwaters recede. Because flash floods in the Brahmaputra and Barak valleys can inundate fields suddenly during the growing season, varieties such as Ranjit Sub1 and IR64 Sub1 address a risk that yield trials alone historically ignored. Demonstrating that at least some Sub1 lines also rank well on stability metrics strengthens the case for their systematic deployment.

GGE biplot analysis approaches the same data from a complementary direction. Instead of summarizing each genotype with a single stability number, it partitions the genotype main effect plus genotype-by-environment interaction into principal components and plots both genotypes and environments in the same two-dimensional space. Environments that cluster together respond similarly to genetic differences, and the genotype positioned furthest in the direction of a cluster is its hypothetical winner. When the polygon connecting extreme genotypes is drawn over the biplot, the perpendicular sector boundaries delineate mega-environments, making it visually explicit where the winning variety changes. In Assam, this revealed that the state is not a single target population of environments but several.

The evaluation of test environments is an often-overlooked contribution of such studies. A testing location earns its place in a trial network only if it discriminates meaningfully among genotypes while remaining representative of the broader production zone. Locations that are highly discriminating but unrepresentative are useful for culling poor performers under stress, whereas locations that are representative but weakly discriminating waste resources by failing to reveal real differences. Sites combining both properties provide the greatest information per unit cost, and identifying them allows breeding programs to concentrate resources where they generate the most reliable selection decisions.

The wider literature on yield-trial analysis supports the dual-method approach adopted here. Decades of debate have compared AMMI and GGE models, with methodological work converging on the view that the choice of model matters less than fitting it to well-designed, adequately replicated trials and interpreting the results jointly with mean performance. Combining a BLUP-based stability index with biplot visualization has become a common recommendation precisely because the two answer different breeding questions, and software that implements both within one workflow has made this combined strategy accessible to national and regional programs, not only large international institutes.

For rice systems across eastern South Asia and similar flood-prone deltas elsewhere, the study’s framework offers a template. Multi-year, multi-location trials capture the temporal as well as spatial dimension of environmental variation, which single-season trials cannot. As climate variability increases the frequency of extreme hydrological events, the genotype-by-environment interaction term will likely grow rather than shrink, making stability analysis and mega-environment mapping increasingly central to varietal release decisions. The Assam results suggest that evidence-based deployment, pairing broadly stable cultivars with targeted specific adaptation and a rationalized testing network, is a practical pathway toward resilient rice production under uncertain conditions.

Subject of Research: Yield stability and genotype × environment interaction of rice varieties evaluated across multi-year, multi-location trials in Assam.

Article Title: Yield Stability and Adaptation of Rice Varieties Across Diverse Environments in Assam Using WAASB and GGE Biplot Analyses

Article References: Borgohain, R., Goswami, G., & Das, J. (2026). Yield Stability and Adaptation of Rice Varieties Across Diverse Environments in Assam Using WAASB and GGE Biplot Analyses. Indian Journal of Genetics and Plant Breeding. https://doi.org/10.1007/s44489-026-00040-1

Image Credits: AI Generated

DOI: 10.1007/s44489-026-00040-1

Keywords: rice, WAASB, GGE biplot, yield stability, genotype × environment interaction, multi-environment trials, Assam, Sub1 submergence tolerance, varietal recommendation, BLUP, climate resilience, plant breeding