Genotype × environment interaction and stability analysis for seed yield and yield-attributing traits in soybean [Glycine max (L.) Merrill]

Introduction

Soybean [Glycine max (L.) Merrill, 2n = 40], the "miracle bean" (Kumar et al., 2005) [1] , is a key Indian oilseed crop valued for its high protein (38-42%) and oil (18-22%) content (Basavaraja et al., 2005; Dukariya et al., 2020) [2,3] . Global production reached an estimated 428.2 million metric tons in 2024-25, led by Brazil, the USA, and Argentina, with India among the top five producers (USDA, 2025) [4] ; Maharashtra, particularly Vidarbha, is a major domestic contributor, though productivity varies across seasons and locations. A key challenge in soybean improvement is genotype × environment (G × E) interaction, which shifts genotype rankings across environments and complicates cultivar selection (Comstock et al., 1963) [5] . Eberhart et al. (1966) [6] addressed this by defining a stable genotype as one with high mean performance, a regression coefficient (bᵢ) near unity, and a deviation from regression (S 2 dᵢ) near zero -partitioning G × E interaction into predictable and unpredictable components. This model, along with related approaches (AMMI, GGE biplot), has been widely applied in soybean, consistently revealing significant G × E interaction for yield and component traits (Deka and Talukdar, 1997 [7] ; Hossain et al., 2003 [8] ; Katna et al., 2017 [9] ; Pallavi et al., 2019 [10] ; Habtegebriel and Abebe, 2023 [11] ; Abebe et al., 2024) [12] , highlighting the need for multi-location testing. This study evaluated 30 soybean genotypes across three Maharashtra locations (Akola, Amravati, Washim) during Kharif 2025 to (i) assess seed yield potential across environments and (ii) characterize G × E interaction for yield and component traits using the Eberhart-Russell (1966) [6] model, identifying genotypes suited to broad or specific adaptation for future breeding.

Material an methods

The experimental material comprised 30 soybean genotypes (23 advanced test lines and 7 commercial checks: AMS 100-39, AMS-MB 5-18, JS 335, JS 93-05, JS 20-34, RVSM 2011-35, and MAUS 158) obtained from the All India Coordinated Research Project (AICRP) on Soybean, Regional Research Centre, Dr. Panjabrao Deshmukh Krishi Vidyapeeth (PDKV), Akola, Amravati, and evaluated during Kharif 2025 at three locations in Maharashtra treated as distinct environments E 1: Oilseeds Research Unit, Dr. PDKV, Akola (sown 04 July 2025); E 2: AICRP on Soybean (Dr. PDKV, Akola), Amravati (sown 25 June 2025); and E 3: Agriculture Research Station, Washim (sown 03 July 2025) using a Randomized Block Design (RBD) with three replications per location, each plot comprising five rows of 4 m length sown at 45 × 5 cm spacing with a basal fertilizer dose of 30:60:30 kg N:P 2 O 5 :K 2 O per hectare, and standard agronomic and plant-protection practices followed uniformly across locations. Observations were recorded on five randomly selected competitive plants per genotype per replication (and on a whole-plot basis for yield) for ten quantitative traits days to 50% flowering, days to maturity, plant height (cm), number of primary branches per plant, height of first pod insertion (cm), number of pods per plant, number of seeds per pod, 100-seed weight (g), seed yield per plot (kg), and seed yield per hectare (kg) with genotypic differences at each location tested following Panse et al. (1967) [13] and stability analysis performed using the joint linear regression model of Eberhart et al. (1966) [6] .

Result and discussion

Mean performance

Analysis of variance

Pooled analysis of variance based on the Eberhart et al. (1966) [6] model (Table 2) showed highly significant genotypic mean squares for plant height, branches per plant, height of first pod insertion, pods per plant, seeds per pod, 100-seed weight, seed yield per plot, and seed yield per hectare, but non-significant genotypic differences for days to 50% flowering and days to maturity when tested against the pooled G × E interaction. The Environment + (Genotype × Environment) component was highly significant for plant height, branches per plant, pods per plant, 100-seed weight, seed yield per plot, and seed yield per hectare, confirming substantial environmental influence.

Partitioning of the environmental component revealed that Environment (linear) was highly significant for the same major traits, indicating that a large share of environmental variance followed a predictable linear trend. The G × E (linear) component was highly significant for plant height, seeds per pod, seed yield per plot, and seed yield per hectare traits for which genotypic responses to environ mental change can therefore be reliably predicted by linear regression. Pooled deviation (the non-linear component) was also highly significant for plant height, branches per plant, pods per plant, seed yield per plot, and seed yield per hectare, indicating that unpre-dictable, non-linear responses contributed meaningfully to G × E interaction for these traits as well consistent with reports by Katna et al.

(2017) [9] and Pallavi et al. (2019) [10] in soybean.

Enviornmental indices

The environmental index quantified the relative favorability of each test location for each trait (

Stability analysis

Because significant G × E interaction was detected for most traits, stability parameters (mean, bᵢ, S 2 dᵢ) were estimated for all ten characters following Eberhart et al. (1966) [6] . Full genotype-wise results for the primary economic trait, seed yield per hectare, are given in Table 4; a synthesis of stability classification across all ten traits is given in Table 5. dᵢ = +6,569.595, non-significant), which yielded above and near the grand mean respectively, showed the closest approach to the ideal stability combination among genotypes with non-significant S 2 dᵢ. High-yielding genotypes AMS 2022-1, AMS 25-3, AMS 26-1, AMS 25-6, AMS 25-2, and AMS 25-22 combined mean yields well above the grand mean with regression coefficients above unity (bᵢ > 1) and non-significant S 2 dᵢ, marking them as below-average in stability but specifically and strongly responsive to favorable, high-input environments such as Amravati. Conversely, AMS 26-2, AMS 25-1, and AMS-MB 5-18 showed bᵢ < 1 with non-significant S 2 dᵢ, denoting above-average stability and buffering capacity suited to resource-limited or stress-prone environments such as Washim. Genotypes with significant S 2 dᵢ (e.g., AMS 22-14, AMS 24-7, AMS 25-21, AMS 22-2, AMS 25-9, AMS 25-7) showed unpredictable, non-linear responses across environments and are not recommended for either broad or targeted release without further evaluation. Table 5. Summary of stability classification of soybean genotypes across ten traits (Eberhart et al., 1966) [6] [14] and Ramteke et al. (2012), who similarly reported that high-yielding soybean genotypes with regression coefficients above unity performed best under favorable, high-input management, while genotypes with bᵢ near or below unity were more dependable under variable or resource-limited conditions.

Character

Conclusions and reccommendatins

This multi-location evaluation of 30 soybean genotypes under the Eberhart et al. (1966) [6] stability model confirmed significant genotype × environment interaction for seed yield and its major component traits, with the linear component predominating for plant height, seeds per pod, seed yield per plot, and seed yield per hectare. Amravati was identified as the most environmentally favorable test site, Washim as the most stress-prone, and Akola as intermediate. Based on combined mean performance and stability parameters: AMS 26-1, AMS 25-6, and AMS 2022-1 are recommended for cultivation in favorable, high-input environments (e.g., Amravati), owing to their high mean seed yields (2,743.33, 2,700.00, and 2,635.00 kg/ha, respectively) and above-unity regression coefficients (bᵢ > 1), and are recommended as prime candidates for multi-location release trials and as donor parents for high-yield breeding.

1. AMS 100-39 (C) and RVSM 2011-35 (C) displayed the closest approach to general (average) stability for seed yield per hectare among genotypes with non-significant S 2 dᵢ, supporting their continued use as broadly dependable check cultivars.

2. AMS 26-2, AMS 25-1, and AMS-MB 5-18, with regression coefficients below unity, are better suited to low-input or stressprone environments such as Washim and merit inclusion in breeding programs targeting drought-or stress-resilient soybean cultivars.

3. Genotypes with significant deviation from regression (S 2 dᵢ > 0), such as AMS 22-14, AMS 24-7, AMS 25-21, AMS 22-2, and AMS 25-9, exhibited unpredictable performance and require further multi-season testing before recommendation. 4. Given its superior seed mass (15.94 g/100 seeds) and pod load, AMS 26-1 is additionally recommended as a valuable donor parent for improving seed size and sink capacity in future hybridization programs, while genotypes AMS 25-2, AMS 2022-1, and AMS 2021-3, with first pod insertion heights of 17.47, 16.13, and 15.78 cm respectively (compared with 8.84 cm in JS 20-34), are recommended for breeding mechanical-harvest-compatible cultivars.

5. Genotypes with significant deviation from regression (S 2 dᵢ > 0), such as AMS 22-14, AMS 24-7, AMS 25-21, AMS 22-2, and AMS 25-9, exhibited unpredictable performance and require further multi-season testing before recommendation.

6. Given its superior seed mass (15.94

References

1. Kumar J, Badiyala D. Effect of seed rate, row spacing and sowing time on yield and yield attributes of soybean. Legume Research. 2005;28(4):288-290.
2. Basavaraja GT, Naidu GK, Salimath PM. Evaluation of vegetable soybean genotypes for yield and component traits. Karnataka Journal of Agricultural Sciences. 2005;18(1):27-31.
3. Dukariya G, Shah S, Singh G, Kumar A. Soybean and its products: Nutritional and health benefits. Journal of Nutritional Science and Healthy Diet. 2020;1(2):22-29.
4. United States Department of Agriculture, Foreign Agricultural Service. Oilseeds: World Markets and Trade. USDA FAS; 2025.
5. Comstock RE, Moll RH. Genotype-environment interactions. Statistical Genetics and Plant Breeding. 1963;982:164-196.
6. Eberhart SA, Russell WA. Stability parameters for comparing varieties. Crop Science. 1966;6(1):36-40.
7. Deka SD, Talukdar P. Stability behaviour of some soybean (Glycine max [L.] Merrill) genotypes under environmental variability. Indian Journal of Genetics and Plant Breeding. 1997;57(1):36-39.
8. Hossain MA, Rahman L, Shamsuddin AKM. Genotype-environment interaction and stability analysis in soybean. Journal of Biological Sciences. 2003;3(11):1026-1031.
9. Katna G, Kumari V, Dev J, Singh A, Bala A. Stability analysis for seed yield and related traits in soybean (Glycine max [L.] Merrill). Agricultural Research Journal. 2017;54(2):165-169.
10. Pallavi M, Kumar GP, Kishore NS. Yield stability and association of its component characters in soybean (Glycine max [L.] Merrill) genotypes. Journal of Oilseeds Research. 2019;217.
11. Habtegebriel MH, Abebe AT. Grain yield stability of soybean (Glycine max [L.] Merrill) for different stability models across diverse environments of Ethiopia. Agrosystems, Geosciences & Environment. 2023;6(3):e20396.
12. Abebe AT, Adewumi AS, Adebayo MA, Shaahu A, Mushoriwa H, Alabi T, Derera J, Agbona A, Chigeza G. Genotype x environment interaction and yield stability of soybean (Glycine max L.) genotypes in multi-environment trials (METs) in Nigeria. Heliyon. 2024;10(19).
13. Panse VG, Sukhatme PV. Statistical Methods for Agricultural Workers. ICAR, New Delhi; 1967.
14. Tyagi SD, Khan MH, Tyagi V. Genotype x environment interaction for yield, protein and maturity duration in soybean. Crop Improvement. 2013;39(1):70-73.
15. Basavaraja GT, Naidu GK, Salimath PM. Pattern of genotype x environment interaction in soybean. Annals of Arid Zone. 2007;46(1):43.
16. How to cite this article: Hiwarkar AV, Nichal S, Patil P, Burghate S, Jadhav P, Pradhan S, et al. Genotype × environment interaction and stability analysis for seed yield and yield-attributing traits in soybean [Glycine max (L.) Merrill]. International Journal of Agriculture and Food Science. 2026;8(9):710-715. doi: https://www.doi.org/10.33545/2664844X.2026.v8.i9j.1941