Genotype × environment interaction, phenotypic stability and multivariate divergence of rice (Oryza sativa L.) genotypes

Introduction

Rice (Oryza sativa L.) is a major staple crop and a cornerstone of food security in Asia. Its productivity is strongly influenced by environmental heterogeneity, while grain yield itself is a complex, quantitatively inherited trait that integrates the effects of multiple yield components and environmental factors. In India, substantial variation in productivity among rice-growing regions underscores the importance of developing genotypes that combine desirable mean performance with predictable responses across target environments. (FAO, 2023) A central challenge in rice improvement is that grain yield is a complex, polygenically controlled trait that responds differentially to environmental variation. This differential response, expressed as genotype × environment (G×E) interaction, can cause genotypes selected under one set of conditions to perform inconsistently when exposed to different agro-ecological conditions (Comstock and Moll, 1963; Lin et al., 1986; Becker and Leon, 1988; Crossa, 1990) [3,8,9] . Consequently, identifying genotypes that combine high mean performance with predictable responses across environments remains a key objective of multi-environment testing. Phenotypic stability analysis provides a quantitative framework for characterising differential genotypic responses to environmental Variation. The joint regression model of Eberhart and Russell (1966) describes stability using the regression coefficient (bi), which reflects the responsiveness of a genotype to environmental quality, and the deviation from regression (S²di), which represents the extent of non-linear or unpredictable response. A genotype with high mean performance, bi close to unity and a small, nonsignificant S²di is generally considered broadly adapted, whereas bi > 1 indicates greater responsiveness to favourable environments and bi < 1 indicates relatively greater performance under less favourable environments when the deviation from regression is small. (Eberhart and Russell, 1966). Stability statistics used in multi-environment trials include parametric approaches such as the Eberhart-Russell regression parameters, Wricke's ecovalence and Shukla's stability variance, as well as non-parametric ranking procedures. Multivariate methods, including the Additive Main Effects and Multiplicative Interaction (AMMI) model and Genotype plus Genotype-by-Environment (GGE) biplot, provide complementary approaches for describing complex G×E patterns and visualising genotypeenvironment relationships (Yan and Tinker, 2006; Somanagoudra and Manonmani, 2020) [34,36] . The Eberhart-Russell model was selected here because its parameters provide readily interpretable measures of environmental responsiveness and predictability, while cluster analysis and PCA were used to characterise multivariate phenotypic differentiation among the genotypes. Despite extensive research on G×E interaction and stability in rice (Arumugam et al., 2007 [2,5,6,13,20,24,31] , comparatively limited information is available on the stability behaviour and multivariate phenotypic diversity of newly developed rice germplasm under contrasting environments of the Indo-Gangetic Plain. Such information is particularly relevant for eastern Uttar Pradesh, where spatial and temporal environmental heterogeneity can complicate the identification of consistently performing genotypes. Integrating stability analysis with multivariate characterisation can therefore help distinguish genotypes with desirable mean performance from those exhibiting environmentspecific responses. The objectives of the present study were therefore to: (i) quantify the nature and magnitude of G×E interaction across three contrasting environments; (ii) estimate Eberhart-Russell stability parameters for eleven yield and yieldrelated traits; (iii) characterise multivariate phenotypic divergence among the 24 genotypes and identify potentially divergent parental combinations; and (iv) identify genotypes warranting further multi-environment evaluation for stable and superior performance in the target region.

Materials and methods

Plant Material and Experimental Design

Twenty-four rice genotypes predominantly newly developed salttolerant lines, together with two check varieties (PUSA 44 and CSR 43) were evaluated during Kharif 2019. The experiment was conducted simultaneously across three environments representing contrasting agronomic conditions at Acharya Narendra Deva University of Agriculture and Technology, Kumarganj, Ayodhya: (i) the Genetics and Plant Breeding Research Farm, (ii) the Student Instructional Farm, and (iii) the MES Farm. Each environment was treated as an independent experimental location. At each location, the experiment followed a Randomised Complete Block Design (RCBD) with three replications. Each genotype was grown in a plot of 2.4 m × 2.0 m, with a spacing of 20 cm × 15 cm (row × plant). Recommended agronomic practices for the Kharif season were followed uniformly across all three environments.

Traits Assessed

Data were recorded for eleven quantitative traits: days to 50% flowering, plant height (cm), productive tillers per plant, panicle length (cm), leaf area index, number of fertile spikelets per panicle, spikelet fertility (%), biological yield per plant (g), harvest index (%), grain length-to-breadth (L/B) ratio of grain, and grain yield per plant (g). Measurements followed the Standard Evaluation System for Rice (IRRI, 2013).

Statistical Methods

All data were subjected to pooled analysis of variance (ANOVA) under the RCBD framework after assessment of homogeneity of error variances across environments using Bartlett's test. Statistical significance was evaluated at the 5% and 1% probability levels (p ≤ 0.05 and p ≤ 0.01, respectively). Phenotypic stability was assessed using the joint linear regression model of Eberhart and Russell (1966) [11] . For each genotype, mean performance, the regression coefficient (bi), and the mean squared deviation from regression (S²di) were estimated. The regression coefficient quantified responsiveness to environmental quality, with bi ≈ 1 denoting average responsiveness, bi > 1 denoting greater responsiveness to favourable environments, and bi < 1 denoting relatively greater performance under less favourable environments. A small, non-significant S²di was interpreted as greater predictability of response. The environmental index (Ij) was calculated as the deviation of each environmental mean from the grand mean. Multivariate phenotypic divergence among the 24 genotypes was assessed using hierarchical cluster analysis based on standardised values of the eleven traits and Ward's minimum-variance linkage method. The reported intra-and inter-cluster distance measures were used to identify divergent groups and potential parental combinations for subsequent breeding programmes. Because the analysis was based on phenotypic traits rather than molecular markers, the resulting groups are interpreted as phenotypically divergent clusters rather than genetically distinct populations. To complement the cluster analysis, Principal Component Analysis (PCA) was performed on standardised cluster-mean values for five key traits (days to 50% flowering, plant height, grain yield per plant, harvest index, and spikelet fertility) using the correlation matrix. The first two principal components were visualised as a biplot to identify major trait contributions to multivariate differentiation among clusters.

RESULTS

Pooled Analysis of Variance

Pooled analysis of variance across the three environments revealed highly significant (p ≤ 0.01) mean squares for genotype for all eleven characters examined, indicating substantial phenotypic variation among the test genotypes (Table 1). Environments were also highly significant for all characters, reflecting pronounced differences among the three test environments. Most importantly for stability assessment, genotype × environment (G×E) interaction mean squares were significant (p ≤ 0.05 or p ≤ 0.01) for all traits, demonstrating differential genotypic responses to environmental variation. The linear component of the G×E interaction was significant or highly significant for the majority of traits, indicating that a substantial proportion of the interaction was associated with differential genotypic regression on the environmental index. This supports the use of the Eberhart-Russell (1966) [11] framework for characterising environmental responsiveness.

However, significant pooled deviations for several traits, particularly days to 50% flowering and spikelet fertility, also indicate that a component of the response remained non-linear and environment-specific. Partitioning of the pooled sum of squares for grain yield per plant showed that environment accounted for the largest share of the variation (approximately 46%), followed by G×E interaction (approximately 39%) and genotype (approximately 15%). Thus, within the present trial, environmental heterogeneity was the dominant contributor to variation in grain yield per plant. This result reinforces the importance of multi-environment evaluation when assessing the consistency of genotype performance (Arumugam et al., 2007; Biswas et al., 2012) [2,6] .

Phenotypic stability for grain yield per plant

Grain yield per plant ranged from 27.96 g (CSR 2711-103) to 40.28 g (CSR 2016-IR-18-12), with an overall mean of 29.50 g across the three environments (Table 2). Eleven genotypes recorded mean grain yield above the general mean. The Eberhart-Russell analysis revealed substantial variation among genotypes in both environmental responsiveness and predictability. Genotypes combining above-average yield with relatively small deviation estimates included CSR 43 (32. [25,27,28,29,32] . CR 2851-S-B-1-2B-1 (30.69 g; bi = 0.917), CSR-2748-4441-193 (29.76 g; bi = 0.578), and CR3437-1-5-200-83-1 (30.36 g; bi = 0.134) exhibited regression coefficients below unity. When associated with small deviation from regression, such responses indicate relatively greater performance under less favourable environments. These genotypes therefore merit consideration for further evaluation under resource-limited or stress-prone conditions (Subudhi et al., 2008; Kesh et al., 2023) [18,35] . CSR 2016-IR-18-12 recorded the highest mean grain yield (40.28 g plant⁻¹) but an exceptionally large positive S²di value (152.66), indicating substantial environmental unpredictability. Despite its superior mean performance, the large deviation estimates limits confidence in its consistency across environments and supports its prioritisation for further targeted evaluation rather than immediate recommendation for broad adaptation (Bhakta and Das, 2008; Subudhi et al., 2008; Kumar et al., 2020) [4,10,19,35] . 2. Vertical and horizontal dashed lines mark bi = 1 and the grand mean yield (30.15 g), respectively, delineating quadrants corresponding to the Eberhart and Russell (1966) [11] adaptability classes. CSR 2016-IR-18-12 (bi = -8.92) lies outside the plotted range and is indicated separately owing to its markedly high S²di (152.66), reflecting unpredictable performance.

Stability Analysis for Other Yield Components

Stability analysis of the remaining ten traits revealed substantial variation in the magnitude and direction of environmental responsiveness. For several traits, the linear component of G×E interaction predominated, whereas significant deviations from regression for selected traits indicated additional non-linear, environment-specific responses. The trait-wise stability estimates are presented in Supplementary Tables S1

Multivariate Phenotypic Divergence Analysis

Hierarchical cluster analysis using standardised values of the eleven traits and Ward's minimum-variance linkage method partitioned the 24 rice genotypes into five phenotypically divergent clusters (Table 3). Cluster I was the most populous (nine genotypes), followed by Cluster II (seven genotypes) and Clusters III and IV (three genotypes each), while Cluster V comprised two genotypes (CR 2851-S-B-1-2B-1 and CSR 2016-IR-18-12). Because the clustering was based on agronomic and phenotypic traits, these groups represent multivariate phenotypic differentiation rather than molecularly defined genetic structure. Intra-cluster distances were modest (range: 0.227-1.746), whereas the maximum reported inter-cluster distance occurred between Clusters II and V (D² = 5.526), followed by Clusters I and II (D² = 5.273) and Clusters II and III (D² = 4.709) (Figure 1). These differences indicate phenotypically divergent groups that may be considered when selecting parents for subsequent crossing and recombination.

Principal Component Analysis

Principal Component Analysis of the five cluster-mean traits reduced the original trait space to two dominant axes: PC1 accounted for 52.2% of total variance and PC2 accounted for a further 27.7%, together explaining 79.8% of the variation among clusters (Figure 2). PC1 was most strongly and positively associated with harvest index, grain yield per plant and spikelet fertility, and negatively associated with days to 50% flowering, broadly representing a productivity-earliness gradient. PC2 was dominated by plant height and, to a lesser extent, days to 50% flowering, representing a plant-stature dimension that was comparatively independent of the principal productivity gradient. Cluster V occupied the positive end of PC1, consistent with its higher mean grain yield, harvest index and spikelet fertility, whereas Cluster III occupied the opposite region. Cluster IV separated primarily along PC2, reflecting differences in plant stature and flowering time, while Clusters I and II were relatively close along PC2 but differed along PC1. The agreement between the cluster structure and PCA ordination provides complementary evidence of multivariate phenotypic differentiation; however, it should not be interpreted as molecular evidence of genetic divergence.

DISCUSSION

The significant genotype × environment (G×E) interaction observed in the present study demonstrates that the performance of the evaluated rice genotypes was not consistent across the three test environments. Such differential genotypic responses are a well-established phenomenon in rice and have been reported across diverse agro-ecological conditions (Mahapatra, 1993 [2,5,6,10,14,16,20,22,33] . The predominance of the linear component of G×E interaction for most traits indicates that a substantial proportion of the observed interaction could be explained by differential genotypic responses to environmental quality. This finding supports the utility of the Eberhart-Russell (1966) [11] model for characterising environmental responsiveness and predictability. Nevertheless, the significant non-linear component observed for traits such as days to 50% flowering and spikelet fertility indicates that a proportion of the genotypic response remained environmentspecific and could not be adequately explained by linear regression alone. Similar complexity in G×E interaction has been documented in rice stability studies (Arumugam et al., 2007; Bhatt et al., 2015; Lee et al., 2023) [2,5,20] . Environmental variation accounted for the largest proportion of the variation in grain yield per plant, followed by G×E interaction and genotype effects. This indicates that environmental heterogeneity among the three test locations exerted a stronger influence on grain-yield expression than genotype effects alone during the study period. Comparable patterns, in which environmental effects contributed substantially to yield variation, have been reported in multi-environment rice trials (Habib et al., 2024; Islam et al., 2024) [14,16] . From a breeding perspective, this finding emphasises the importance of evaluating candidate genotypes across multiple environments and seasons rather than relying on performance under a single test environment. Multienvironment testing provides a more reliable basis for distinguishing genotypes with consistent performance from those exhibiting strong environment-specific responses (Crossa, 1990) [9] . CSR 43 combined an above-average grain yield (32.07 g plant⁻¹) with a low regression coefficient (bi = 0.239) and a small adjusted deviation estimate (S²di = -7.36). According to the Eberhart-Russell framework, a regression coefficient substantially below unity indicates relatively low responsiveness to improvement in environmental quality, whereas a small deviation from regression indicates comparatively predictable performance (Eberhart and Russell, 1966) [11] . Therefore, the performance of CSR 43 should not be interpreted as evidence of average responsiveness or broad adaptation solely on the basis of its low bi value. Rather, its response suggests relatively limited environmental responsiveness and potential suitability under less favourable conditions, provided that this pattern is confirmed across additional environments. The concept of stability as a combination of desirable mean performance and predictable response has also been emphasised by Becker and Leon (1988) and Purchase et al. (2000) [3,30] . Several other genotypes displayed regression coefficients close to or moderately above unity. CSR-C27 SM-117 (bi = 0.936) and RP 5706-120-7-2-2 (bi ≈ 1) exhibited responses close to the average environmental trend, suggesting comparatively broad environmental responsiveness when considered together with their relatively small deviation estimates. Similar interpretations of near-unity regression coefficients and small deviations have been used for identifying rice genotypes with relatively stable performance across environments ( [23,25,27,28,29,32] . However, because the present evaluation involved only three environments in a single season, these genotypes should be considered as candidates for further multi-environment evaluation rather than as immediately suitable for formal varietal release. In contrast, CSR 2711-103 (bi = 3.88), CR 3890- [25,27,28,29,32] . These genotypes may therefore be useful for production environments characterised by favourable resource availability, although their specific adaptation should be validated through targeted multi-environment testing [4,10,17,19,35] . Therefore, although CSR 2016-IR-18-12 represents an important highyielding genotype, its performance should be examined across a larger number of environments before considering it for broad adaptation. Its contrasting response may also make it useful as a breeding resource for investigating yield plasticity and identifying progeny capable of maintaining high productivity under variable environmental conditions. The multivariate analysis provided additional information on the phenotypic structure of the evaluated germplasm. Hierarchical cluster analysis grouped the 24 genotypes into five phenotypically divergent clusters, indicating substantial variation in the joint expression of the measured agronomic traits. The lack of a clear correspondence between geographic origin and cluster membership suggests that geographic origin alone may not adequately reflect phenotypic similarity among rice genotypes. Similar observations have been reported in multivariate analyses of rice germplasm, where clustering based on agronomic traits has revealed substantial divergence among genotypes irrespective of their geographic background (Chandra et al., 2007) [7] . Therefore, the observed cluster structure may provide useful information for selecting parents with contrasting phenotypic attributes.

The maximum reported inter-cluster distance occurred between Clusters II and V (D² = 5.526), indicating substantial multivariate divergence between these groups. In breeding programmes, crosses between phenotypically divergent parents may provide opportunities for generating segregating populations with complementary trait combinations. In the present study, Cluster V was characterised by comparatively high grain yield, harvest index, biological yield and panicle length, whereas Cluster I showed relatively lower plant height and higher spikelet fertility. Such contrasting trait profiles may provide useful parental combinations for subsequent breeding programmes. However, cluster distance alone does not guarantee superior combining ability or transgressive segregation; therefore, parental selection should subsequently be validated through crossing, progeny evaluation and combining-ability studies.

The PCA further complemented the cluster analysis by summarising the major sources of multivariate variation. The first two principal components of the cluster-mean dataset accounted for 79.8% of the total variation, with harvest index, grain yield and plant height contributing substantially to genotype differentiation. The identification of grain yield and harvest index among the major contributors is consistent with the importance of productivity-related traits in differentiating rice germplasm. PCA and biplot approaches have been widely used to identify major contributors to phenotypic variation and to visualise relationships among rice genotypes (Yan and Tinker, 2006; Yang et al., 2025) [36, 37] . Importantly, the agreement between cluster analysis and PCA should be interpreted as complementary evidence of phenotypic structure rather than independent confirmation of molecular genetic divergence, because both analyses were derived from the same phenotypic dataset. The integration of stability analysis with multivariate phenotypic characterisation provides a more comprehensive framework for genotype prioritisation than reliance on a single statistical criterion. Stability parameters describe how genotypes respond to environmental variation, whereas cluster analysis and PCA provide information on multivariate phenotypic relationships among genotypes. Such complementary approaches can assist breeders in identifying genotypes with desirable performance as well as potential parental contrasts for subsequent breeding cycles (Crossa, 1990; Nayak et al., 2022; Kesh et al., 2023; Abebe et al., 2023; Yuriyah et al., 2026) [1,9,18,26,38] . In the present study, CSR 43, CSR-C27 SM-117 and CSR-2748-4441-193 were particularly noteworthy because of their relatively favourable grain-yield performance and distinct environmental response patterns, while the divergent clusters provided additional opportunities for selecting contrasting parental material. Overall, the findings indicate that genotype performance was strongly influenced by environmental heterogeneity and that substantial variation existed among genotypes in both environmental responsiveness and predictability. The results support the use of multi-environment stability analysis in combination with multivariate phenotypic characterisation for prioritising rice genotypes. However, the present conclusions are based on three environments evaluated during a single season; therefore, additional multi-season and multi-location testing is necessary to establish the repeatability of the observed stability patterns and to support subsequent varietal advancement or breeding decisions.

Conclusions

The present study demonstrated significant genotype × environment interaction for the eleven traits evaluated across three contrasting environments, with the linear component accounting for a substantial part of the interaction for most traits. Eberhart-Russell analysis revealed marked differences among genotypes in environmental responsiveness and predictability. CSR 43 recorded an above-average grain yield (32.07 g plant⁻¹) together with a low regression coefficient (bi = 0.239) and a small adjusted deviation estimate (S²di = -7.36), indicating relatively limited responsiveness and comparatively stable performance under less favourable conditions. CSR-C27 SM-117 and RP 5706-120-7-2-2 showed responses closer to the average environmental trend, whereas CSR 2711-103, CR 3890-35-1-3-4 and CSRRIL-01-IR 165 exhibited strong responsiveness to favourable environments. CSR 2016-IR-18-12 recorded the highest mean yield but a large deviation estimate, indicating substantial environmental unpredictability. These genotypes should therefore be considered for further multi-environment and multi-season evaluation before varietal advancement decisions are made. Multivariate analysis grouped the 24 genotypes into five phenotypically divergent clusters, with the largest reported inter-cluster distance between Clusters II and V (D² = 5.526). PCA of cluster means showed that the first two components accounted for 79.8% of the total variation, with harvest index, grain yield and plant height contributing strongly to multivariate differentiation. The combined use of stability parameters and multivariate phenotypic characterisation provides a useful framework for prioritising genotypes and potential parental contrasts. However, because the study was based on three environments in a single season, the identified stability patterns should be validated across additional environments and seasons before formal varietal recommendation. The findings therefore provide a quantitative foundation for continued breeding evaluation of rice germplasm under heterogeneous production conditions.

Acknowledgement

Acknowledgements: The authors gratefully acknowledge the facilities and support provided by Acharya Narendra Deva University of Agriculture and Technology, Kumarganj, Ayodhya, and the guidance of the Student Advisory Committee.

References

1. Abebe D, Mohammed W, Tesfaye A. Genotype × environment interaction and stability analysis in upland rice (Oryza sativa L.) varieties in Ethiopia. J Crop Sci Biotechnol. 2023;26(1):51-62.
2. Arumugam M, Rajanna MP, Vidyachandra B. Stability of rice genotypes for yield and yield components over extended dates of sowing under Cauvery command area in Karnataka. ORYZA Int J Rice. 2007;44(2):104-107.
3. Becker HC, Leon J. Stability analysis in plant breeding. Plant Breed. 1988;101(1):1-23.
4. Bhakta N, Das SR. Phenotypic stability for grain yield in rice. ORYZA Int J Rice. 2008;45(2):115-119.
5. Bhatt JP, Nanavati JI, Mehta AM. Stability analysis and G × E interactions in lowland rice promising genotypes. ORYZA Int J Rice. 2015;52(3):217-221.
6. Biswas PL, Nath UK, Ghosal S, Patwary AK. Genotypeenvironment interaction and stability analysis of four fine rice varieties. J Bangladesh Agric Univ. 2012;10(1):1-7.
7. Chandra R, Pradhan SK, Singh S, Bose LK, Singh ON. Multivariate analysis in lowland rice. J Cent Eur Agric. 2007;8(1):65-74.
8. Comstock RE, Moll RH. Genotype-environment interactions. In: Statistical Genetics and Plant Breeding. 1963. p. 164-196.
9. Crossa J. Statistical analyses of multilocation trials. Adv Agron. 1990;44:55-85.
10. Das SR, Choudhury PKD. Stability analysis for some growth and yield characters in rice (Oryza sativa). ORYZA Int J Rice. 1996;33:264-267.
11. Eberhart SA, Russell WA. Stability parameters for comparing varieties. Crop Sci. 1966;6(1):36-40.
12. Food and Agriculture Organization of the United Nations. FAOSTAT statistical database. Rome: Food and Agriculture Organization of the United Nations; 2023. Available from: https://www.fao.org/faostat
13. Ghazy MI, Abdelrahman M, El-Agoury RY, El-Hefnawy TM, El-Naem SA, Daher EM, et al. Exploring genetics by environment interactions in some rice genotypes across varied environmental conditions. Plants. 2024;13(1):74. doi:10.3390/plants13010074.
14. Habib MA, Azam MG, Haque MA, Hassan L, Khatun MS, Nayak S, et al. Climate-smart rice (Oryza sativa L.) genotypes identification using stability analysis, multi-trait selection index, and genotype-environment interaction at different irrigation regimes with adaptation to universal warming. Sci Rep. 2024;14:13836. doi:10.1038/s41598-02464808-9.
15. International Rice Research Institute. Standard evaluation system for rice (SES). Los Baños, Philippines: International Rice Research Institute; 2013.
16. Islam SS, Sarker MBU, Rana MM, Hasan AK, Karim MR, Khomphet T. Comprehensive assessment of the genotypeenvironment interaction and yield stability of Boro rice genotypes under four environments in Bangladesh using AMMI analysis. Scientifica. 2024;2024:7800747. doi:10.1155/2024/7800747.
17. Jain BT, Sarial AK, Kaushik P. Understanding G × E interaction of elite Basmati rice (Oryza sativa L.) genotypes under north Indian conditions using stability models. Appl Ecol Environ Res. 2019;17(6):13249-13263. doi:10.15666/aeer/1706_1324913263.
18. Kesh H, Battan RK, Kumar R. Genotype × environment interaction and stability analysis in Basmati rice (Oryza sativa L.) genotypes. Indian J Agric Res. 2023;57(1):16-22. doi:10.18805/IJARe.A-5876.
19. Kumar DBM, Purushottam AP, Raghavendra P, Vittal T, Shubha KN, Madhuri R. Genotype-environment interaction and stability for yield and its components in advanced breeding lines of red rice (Oryza sativa L.). Bangladesh J Bot. 2020;49(3):425-435.
20. Lee SM, Kwon Y, Jeung JU, Mo Y. Multi-environment trials and stability analysis for yield-related traits of commercial rice cultivars. Agriculture. 2023;13(2):256. doi:10.3390/agriculture13020256.
21. Lin CS, Binns MR, Lefkovitch LP. Stability analysis: where do we stand? Crop Sci. 1986;26(5):894-900.
22. Mahapatra KC. Phenotypic stability of some promising varieties of rice. ORYZA Int J Rice. 1993;30:176-180.
23. Makwana DD, Patel JS, Mehta AM. Interpreting genotype × environment interaction in rice (Oryza sativa L.) using Eberhart and Russell model. Int J Bio-resour Stress Manag. 2025;16(2):01-09. doi:10.23910/1.2025.5822.
24. Mamidi A, Lavuri K, Badri J, Sai Prasad SV, Sundaram RM. A multi-model genotype × environment interaction analysis discerning phenotypic plasticity of the strong culm trait in rice. Front Plant Sci. 2026;17:1727579. doi:10.3389/fpls.2026.1727579.
25. Manjunatha B, Malleshappa C, Niranjana KB. Stability analysis for yield and yield attributing traits in rice (Oryza sativa L.). Int J Curr Microbiol Appl Sci. 2018;7(6):16291638. doi:10.20546/ijcmas.2018.706.196.
26. Nayak GP, Sujatha M, Chandra MY, Saida ND, Kiran BT. Stability analysis for grain yield and yield contributing traits in hybrid rice. Biol Forum Int J. 2022;14(1):1230-1234.
27. Pandey V, Singh SK, Korada M, Singh DK, Khaire AR, Habde S, et al. Stability analysis in rice (Oryza sativa L.) genotypes with high grain zinc. Indian J Agric Res. 2020;54(6):689-698. doi:10.18805/IJARe.A-5353.
28. Panwar LL. Stability analysis in rice (Oryza sativa L.). ORYZA Int J Rice. 2008;45(3):194-197.
29. Prajapati KN, Kathiria KB. Genotype × environment interactions and stability for grain yield and its components in rice (Oryza sativa L.). Int J Bio-resour Stress Manag. 2018;9(1):137-141. doi:10.23910/IJBSM/2018.9.1.1829.
30. Purchase JL, Hatting H, van Deventer CS. Genotype × environment interaction of winter wheat (Triticum aestivum) in South Africa: II. Stability analysis of yield performance. S Afr J Plant Soil. 2000;17(3):101-107.
31. Rahim Soroush H, Hoseini Chaleshtori M, Haghighi Hasanalideh A. Investigating the genotype × environment interaction and estimating the grain yield stability of rice promising genotypes. Cereal Biotechnol Biochem. 2024;3(3):409-424. doi:10.22126/cbb.2024.11230.1088.
32. Ramalingam J, Nadarajan N, Vanniarajan C, Rangasamy P. Stability analysis in rice. Madras Agric J. 1990;77(2):73-78.
33. Singh D, Prakash V, Verma OP, Singh PK, Chauhan VS. Genotype × environment interaction and stability analysis in rice (Oryza sativa L.) genotypes evaluated under multienvironment trials in the Indo-Gangetic Plains. ORYZA Int J Rice. 2023;60(1):32-41.
34. Somanagoudra C, Manonmani S. Stability analysis of twoline rice hybrids (Oryza sativa) in diverse environments utilizing AMMI model. Int J Curr Microbiol Appl Sci. 2020;9(6):2896-2904. doi:10.20546/ijcmas.2020.906.352.
35. Subudhi HN, Meher J, Bose LK. Stability analysis for yield and quality traits in rice. J Cent Eur Agric. 2008;9(4):647655.
36. Yan W, Tinker NA. Biplot analysis of multi-environment trial data: principles and applications. Can J Plant Sci. 2006;86(3):623-645. doi:10.4141/P05-169.
37. Yang H, Yang J, Wang L, et al. Phenotypic diversity evaluation of medium indica rice germplasm resources based on principal component analysis and cluster analysis. J South China Agric Univ. 2025;46(4):471-479. doi:10.7671/j.issn.1001-411X.202412004.
38. Yuriyah S, Herlina L, Afza H, Prasetiyono J, Dadang A, Andarini YN, et al. Development of stable and high-yielding rice (Oryza sativa L.) genotypes for suboptimal agroecosystems. SABRAO J Breed Genet. 2026;58(1):1222. doi:10.54910/sabrao2026.58.1.2.
39. Average/Wide RP 5706-120-7-2-2 178.33 1.209 -11.18 Favourable Adaptability class assigned as per Eberhart and Russell (1966): bi ≈ 1 (0.8-1.2) = Average/Wide; bi > 1.2 = Favourable environments; bi < 0.8 = Unfavourable environments; S²di > 10 = Unstable.