Multi-environment evaluation of forage maize genotypes for green forage and dry matter yield stability using AMMI and GGE Biplot analysis

Introduction

Maize (Zea mays L.) is one of the most important cereal crops globally and serves as an excellent forage crop owing to its high biomass production, palatability, and nutritional quality. In India, maize occupies a pivotal position in the livestock feed sector, contributing significantly to the total fodder availability. The crop is characterized by zero anti-nutritional content, higher crude protein, and superior in vitro dry matter digestibility (IVDMD) compared to other non-legume fodder crops, making it an ideal candidate for both green feeding and silage preparation (Mahanta et al., 2023) [9] . India is home to the largest and most diverse livestock population globally, with approximately 70% of households depending on the livestock and agriculture sectors for their livelihoods (Ghosh et al., 2016) [7] . Despite this, the country faces a significant deficit of 35.6% in green fodder, 10.5% in dry fodder, and 44% in concentrates (Anonymous, 2015) [1] . This perennial fodder scarcity has been identified as a major constraint limiting livestock productivity and, consequently, the livelihoods of millions of small and marginal farmers across the country. The development of high-yielding forage maize varieties adapted to diverse agro-climatic conditions offers a promising strategy to bridge this deficit. Maize, being a cross-pollinated crop, offers immense potential for harnessing heterosis, and both composite varieties and single-cross hybrids have been developed for forage purposes (Singh et al., 2021; Das et al., 2021) [19,3] . However, the presence of genotype × environment interaction (GEI) necessitates the measurement of both performance and stability for any genotypes being evaluated in breeding programs (Devi et al., 2019; Sah et al., 2016; Pour-Aboughadareh et al., 2022) [5,18,13] . This allows for a comprehensive assessment of the impact of GEI on phenotypic and genotypic values, avoiding potential biases in estimating heritability and predicting genetic advances.

Two commonly used biplots for comprehending GEI analysis are the AMMI biplot (Gauch, 1992) [6] and the genotype main effect plus genotype-environment interaction (GGE) biplot (Yan et al., 2000; Yan and Kang, 2003) [23,21] . The AMMI model partitions the GEI into multiplicative interaction principal component axes (IPCAs), while the GGE biplot eliminates the environment component and integrates the genotype with the GEI effect. The key distinction between the two lies in the fact that the GGE biplot focuses on the genotype and GE interaction, which are the most relevant for genotype evaluation, whereas the AMMI biplot separates the additive main effects from the multiplicative interaction effects (Yan et al., 2007) [24] . The All India Coordinated Research Project (AICRP) on Forage Crops and Utilization conducts systematic multilocation testing of forage maize genotypes across the country through a structured pipeline: Initial Varietal Trial in Maize (IVTM), followed by Advanced Varietal Trial in Maize-1 (AVTM-1), and then AVTM-2. These coordinated trials provide an excellent framework for evaluating genotypic stability across diverse environments. Consequently, this study examined the performance of forage maize genotypes tested under IVTM with the specified objectives: (i) to estimate the GEI of forage maize genotypes for green forage yield and dry matter yield, (ii) to identify high-yielding and stable genotypes, and (iii) to classify the testing locations into megaenvironments.

Materials and Methods

Plant materials and experimentation: The investigation involved the evaluation of 25 forage maize test entries along with three national checks, viz., J-1006, African Tall and COHM-8, across 20 locations during the designated cropping season under the AICRP on Forage Crops and Utilization. These 20 locations represented five distinct agro-climatic zones as defined by the AICRP framework: Hill Zone (Palampur, Srinagar, Rajouri and Almora), North West Zone (Ludhiana, Hisar and Pantnagar), North East Zone (Bhubaneswar, Ranchi, Pusa, Ayodhya and Imphal), Central Zone (Raipur, Jabalpur, Rahuri, Urulikanchan and Jhansi), and South Zone (Hyderabad, Coimbatore and Mandya). The multi-location testing represented a wide range of maize production ecosystems, distinguished by variations in latitude, altitude, soil type, and macro-climatic conditions. All locations adhered to the standard recommended practices for crop management as outlined by the AICRP.

Recording of observations:

The 28 forage maize genotypes were evaluated for two primary traits, viz., green forage yield (GFY; q/ha) and dry matter yield (DMY; q/ha). The green forage was harvested at 50% flowering stage, and the fresh weight was recorded for computing GFY.

Representative samples were oven-dried at 65°C until constant weight for determination of dry matter content and subsequent calculation of DMY.

Data Analysis: Analysis of variance (ANOVA) was performed to analyse genotype (G), environment (E), and their interactions (GEI) effects. The proportion of variation attributable to G, E, and GEI was computed from the respective sum of squares. Additive main effects and multiplicative interaction (AMMI) analysis was conducted following Gauch (1992) [6] , and genotype main effect plus genotype-environment interaction (GGE) biplot analysis was performed following Yan et al. (2000) [23] and Yan and Tinker (2006) [22] . In the analysis, no scaling was applied (scaling = 0) to generate a tester-centered (centering = 2) GGE biplot. For genotype evaluation, genotype-focused singular value partitioning (SVP = 1) was applied using the Mean versus Stability option, while environment-focused singular value partitioning (SVP=2) was employed for environmental evaluation. The Which-Won-Where option was utilized to identify the winning genotype within a given set of environments. The AMMI Stability Value (ASV) was computed following Purchase et al. (2000) [14] , and the Genotype Selection Index (GSI) was calculated as the sum of ranks for mean yield and ASV. All analyses were performed using the metan package (Olivoto and Lúcio, 2020) [12] in R statistical software (R Studio, 2020) [17] .

Results and Discussion

Analysis of Variance: The multi-location testing selected for the present experiment reflected diverse forage maize production ecosystems across India, characterized by differences in latitude, altitude, and macro-climatic conditions spanning five agro-climatic zones. The analysis of variance (ANOVA) indicated the significant influence of G, E, and GEI for both traits studied in the experiment (Table 2). The analysis revealed that the environmental factor accounted for a substantial proportion of the observed variation in the dataset: 67.47% for GFY and 73.63% for DMY, while the contributions of genotype (9.35% and 7.34%) and GEI (23.18% and 19.03%) were relatively lower (Table 2). Notably, the contribution of G outweighed the contribution of GEI for DMY, while for GFY the GEI component was substantially higher than the genotypic component, suggesting that genotype performance for forage yield was considerably influenced by environmental interaction. The significant GEI component underscored the necessity for stability analysis to identify genotypes with consistent performance across environments. The predominance of environmental effects is consistent with earlier reports in maize (Mohammadi et [10,8,2,26] and sorghum (Rakshit et al., 2012) [16] . However, the higher GEI proportion observed in the present study (19-23%) compared to Yathish et al. (2024) [26] who reported 1-3% GEI in baby corn, could be attributed to the wider environmental spread (20 locations across 5 zones vs. 4 locations across 3 zones) and the greater genetic diversity (28 genotypes vs. 13) tested in the present investigation.

When considering location-wise means, the highest GFY was observed in Coimbatore (510.28 q/ha) followed by Pantnagar (484.12 q/ha), while the lowest was recorded in Srinagar (139.18 q/ha). For DMY, Urulikanchan (109.89 q/ha) recorded the highest mean, while Pantnagar recorded the lowest (Table 2).

Mean performance and stability of the genotypes across locations:

The intricate interaction between genotypes and the environment was effectively simplified and visualized using GGE biplot analysis, which represents the data graphically through principal components (Yan and Tinker, 2006) [22] . The first two PCs explained 55.57% (PC1 = 42.43%, PC2 = 13.14%) and 58.51% (PC1 = 41.87%, PC2 = 16.64%) of the total variation for GFY and DMY, respectively, validating the GGE biplots as reliable graphical representations of the multi-location data. A cumulative variation of >60% (PC1 and PC2) was considered to provide a reliable approximation of the G × E variation (Yang et al., 2009) [25] . In the mean vs. stability graph (Fig. 1a-b), the line characterized by a single arrowhead represented the AEC abscissa, which intersected the biplot origin and served as a marker for the average environment, progressing towards higher mean values. The AEC ordinate, perpendicular to the AEC abscissa, reflected genotype stability, with longer projections indicating lesser stability. It was evident from the mean vs. stability graph that genotype DFH 25-2 outperformed all check genotypes, displaying the highest mean values for both GFY (414.24 q/ha) and DMY (86.73 q/ha). This genotype exhibited 10.00% superiority over the best check (African Tall) for GFY and 12.54% for DMY. JHFM 25-2 (GFY: 380.64 q/ha) and PMC-18 (GFY: 380.62 q/ha) were the next best performers. Among the checks, African Tall (376.59 q/ha) outperformed COHM-8 (350.63 q/ha) and J-1006 (288.26 q/ha).

Genotypic and environmental evaluation:

One major advantage of the graphical presentation of GEI is the facile identification of genotypes that are closer to the ideal genotype (Yan and Tinker, 2006) [22] . The ideal genotype, characterized by high yield and greater stability, is represented by the longest vector length of the highestyielding genotype with zero GEI, located at the center of the concentric circles (Fig. 2a-b). Based on this study, it could be concluded that genotype DFH 25-2 most closely approached the ideal genotype for both GFY and DMY traits. JHFM 25-2, PMC-18, and JHFM 25-1 were the next closest to the ideal, all outperforming the national checks in terms of combined mean performance and stability. The graphical representation of biplots also facilitated the interpretation of the interaction between testing environments through the angle between their vectors (Fig. 3a-b). The cosine of the angle between two environment vectors indicates their correlation (Yan and Tinker, 2006) [22] . Analysis of both GFY and DMY traits revealed that environments within the same agro-climatic zone generally displayed acute vector angles, indicating positive correlations. For instance, the Hill Zone locations (Palampur, Srinagar, Rajouri, Almora) formed a tight cluster, as did the Central Zone locations (Raipur, Jabalpur, Urulikanchan). In contrast, nearly right or obtuse angles were observed between locations from different zones, such as between Hill and South Zone environments, revealing that genotypes performing well in one zone might exhibit different rankings in another. This crossover GEI justified the classification of testing locations into multiple megaenvironments. Which won where and mega environment identification A Which-Won-Where graph was constructed by connecting the most distant genotypes to form a polygon in the biplot, with perpendicular equality lines drawn from the origin to each polygon edge (Yan, 2001) [20] . The genotypes at the vertices of the polygon represented the best performers in the environments falling within their respective sectors (Fig. 4a-b).

For the GFY trait, the Which-Won-Where plot identified DFH 25-2 as the winning genotype across the largest number of environments, including Palampur, Rajouri (Hill Zone), Ludhiana, Hisar (NW Zone), and Raipur, Jabalpur (Central Zone), confirming its broad adaptability. JHFM 25-2 was the winning genotype in Pantnagar, Ayodhya, and Mandya, while JHFM 25-1 won in Imphal, Jhansi, and Hyderabad. For the DMY trait, a similar pattern emerged with DFH 25-2 winning across five locations spanning the NE and Central Zones. This suggested that although testing was conducted at 20 locations across 5 zones, the environments could be grouped into distinct megaenvironments based on winning genotype patterns. However, it is essential to verify this mega-environment pattern through multi-year and multi-environment trials, as recorded in studies involving wheat (Yan et al., 2000) [23] and peanut (Putto et al., 2008; Daemo et al., 2024) [15,4] .

AMMI Analysis:

The AMMI analysis complemented and confirmed the GGE biplot findings. The first two AMMI interaction principal component axes (IPCA [1] and IPCA [2] ) accounted for 42.34% (IPCA [1]: 24.26%, IPCA [2]: 18.08%) and 46.82% (IPCA [1]: 27.96%, IPCA [2]: 18.86%) of the GEI sum of squares for GFY and DMY, respectively. The significant proportion of the sum of squares attributed to environmental factors implied notable variations in means across environments, which predominantly explained the observed variability in both traits, consistent with the GGE biplot results. In the AMMI [1] biplot (Fig. 5a-b), the abscissa and ordinate axes represent the trait main effect and IPCA [1] , respectively. Genotypes with IPCA [1] scores close to zero indicated general adaptation across environments, while larger IPCA [1] scores indicated specific adaptation (Murphy et al., 2009; Li et al., 2023) [11,8] . For GFY, genotypes IFH-11-250 (ASV = 1.21), HPFM-14 (ASV = 1.27), and PH-421266 (ASV = 1.69) displayed the highest stability, as evidenced by their proximity to the zero line. However, the high-yielding genotype DFH 25-2 also exhibited moderate stability (ASV = 4.12), positioning it as a genotype with both high yield and acceptable stability. For DMY, a similar pattern was observed with IFH-11-250 showing the lowest ASV (0.61), followed by BAIF Maize-9 (0.93) and COHM-8 (1.07).

In the AMMI [2] biplot (Fig. 6a-b), which employs IPCA [1] and IPCA [2] as axes, environments closer to the biplot origin indicate less interaction and are considered favorable for selecting genotypes with broad adaptation (Murphy et al., 2009) [11] . The angle between genotype and environment vectors determines the nature of GEI: acute angles indicate positive interaction, right angles negligible interaction, and obtuse angles negative interaction. Through the AMMI [2] analysis, genotypes near the origin, particularly BAIF Maize-9, COHM-8, and NT-009, were identified as the most stable across environments for both traits. 3). The GSI, calculated as the sum of ranks for mean yield and ASV, provides a balanced criterion for selecting genotypes that are both high-yielding and stable. Genotypes DFH 25-2 and JHFM 25-1 had the lowest GSI values (13 each), indicating the best combination of high yield and stability for GFY. DFH 25-2 ranked 2nd for mean GFY (414.24 q/ha) with an ASV of 4.12 (rank 11), while JHFM 25-1 ranked 5th for mean (379.90 q/ha) with an ASV of 3.18 (rank 8). For DMY, DFH 25-2 (GSI = 11), BAIF Maize-9 (GSI = 11), and COHM-8 (GSI = 11) showed the best combination of yield and stability. The Finlay-Wilkinson regression coefficient (bi) provides additional insight into genotype responsiveness. A bi close to 1.0 indicates average responsiveness, bi > 1.0 indicates above-average response to favourable environments, and bi < 1.0 indicates better adaptation to unfavourable environments. DFH 25-2 displayed a bi of 1.15 for GFY, indicating slightly above-average responsiveness to favourable environments while maintaining high yield across all environments. In contrast, PH-421266 (bi = 0.82) showed better adaptation to stress environments, making it a candidate for low-input farming systems.

Conclusion

According to the integrated GGE biplot and AMMI analysis, it was inferred that the environmental assessment could categorize the 20 testing locations across five agroclimatic zones into distinct mega-environments. Genotype DFH 25-2 exhibited superior performance and stability across multiple locations for both GFY (414.24 q/ha, 10% superiority over best check) and DMY (86.73 q/ha, 12.54% superiority) traits, followed by JHFM 25-2, PMC-18, and JHFM 25-1. The substantial GEI observed (19 to 23 percent of total variation) emphasized the importance of multi-location testing for reliable genotype evaluation. These promising genotypes possess the capability to address the scarcity of green fodder in the nation and can be advanced to AVTM-1 trials for further evaluation and potential release. Furthermore, the parental lines of these superior genotypes can be employed in breeding programs, utilizing different mating designs with other elite germplasm to develop improved forage maize cultivars.

References

1. Anonymous. IGFRI Vision 2050. Jhansi: Indian Grassland and Fodder Research Institute; 2015.
2. Bocianowski J, Nowosad K, Rejek D. Genotypeenvironment interaction for grain yield in maize (Zea mays L.) using the additive main effects and multiplicative interaction (AMMI) model. Journal of Applied Genetics. 2024;65:653-664.
3. Das AK, Choudhary M, Kumar P, Karjagi CG, Yathish KR, Kumar R, et al. Heterosis in genomic era: advances in the molecular understanding and techniques for rapid exploitation. Critical Reviews in Plant Sciences. 2021;40:218-242.
4. Daemo BB, Wolancho GB, Arke ZA, Wakalto DD, Onu MH. Performance evaluation and stability of maize (Zea mays L.) genotypes for grain yield using AMMI and GGE biplot. International Journal of Agronomy. 2024;2024:8801999.
5. Devi R, Sood VK, Chaudhary HK, Kumari A, Sharma A. Identification of promising and stable genotypes of oat (Avena sativa L.) for green fodder yield under varied climatic conditions of north-western Himalayas. Range Management and Agroforestry. 2019;40:67-76.
6. Gauch HG. AMMI analysis of yield trials. In: Kang MS, Gauch HG, editors. Genotype-by-Environment Interaction. Boca Raton (FL): CRC Press; 1992. p.1-40.
7. Ghosh PK, Palsaniya DR, Srinivasan R. Forage research in India: issues and strategies. Agricultural Research Journal. 2016;53:1-12.
8. Li Y, Bao H, Xu Z, Hu S, Sun J, Wang Z, Yu X, et al. AMMI and GGE biplot analysis of grain yield for drought-tolerant maize hybrid selection in Inner Mongolia. Scientific Reports. 2023;13:18800.
9. Mahanta SK, Kumar S, Singh P, Meena BS. Forage production in India: bottlenecks and opportunities. Asian Journal of Dairy and Food Research. 2023;42:275-285.
10. Mohammadi R, Aghaee M, Haghparast R, Pourdad SS, Rostaii M, Ansari Y, et al. Association among nonparametric measures of phenotypic stability in four annual crops. Middle Eastern and Russian Journal of Plant Science and Biotechnology. 2009;(Special Issue 1):20-24.
11. Murphy SE, Lee EA, Woodrow L, Seguin P, Kumar J, Rajcan I, et al. Genotype × environment interaction and stability for isoflavone content in soybean. Crop Science. 2009;49:1313-1321.
12. Olivoto T, Lúcio AD. metan: an R package for multienvironment trial analysis. Methods in Ecology and Evolution. 2020;11:783-789.
13. Pour-Aboughadareh A, Khalili M, Poczai P, Olivoto T. Stability indices to deciphering the genotype-byenvironment interaction (GEI) effect: an applicable review for use in plant breeding programs. Plants. 2022;11:414.
14. Purchase JL, Hatting H, Van Deventer CS. Genotype × environment interaction of winter wheat in South Africa: II. Stability analysis of yield performance. South African Journal of Plant and Soil. 2000;17:101107.
15. Putto W, Patanothai A, Jogloy S, Hoogenboom G. Determination of mega-environments for peanut breeding using the CSM-CROPGRO-Peanut model. Crop Science. 2008;48:973-982.
16. Rakshit S, Ganapathy KN, Gomashe SS, Rathore A, Ghorade RB, Nagesh MVK, et al. GGE biplot analysis to evaluate genotype, environment and their interactions in sorghum multi-location data. Euphytica. 2012;185:465-479.
17. RStudio Team. RStudio: Integrated Development for R. Boston (MA): RStudio, PBC; 2020.
18. Sah RP, Ahmed S, Malaviya DR, Saxena P. Identification of consistence performing dual purpose maize (Zea mays L.) genotypes under semi-arid condition. Range Management and Agroforestry. 2016;37:162-166.
19. Singh SB, Kumar P, Kasana RK, Choudhary M, Kumar S, Kumar R, et al. Unveiling combining ability and heterotic grouping of newly developed winter maize (Zea mays L.) inbreds. Indian Journal of Agricultural Sciences. 2021;91:1586-1591.
20. Yan W. GGE biplot: a Windows application for graphical analysis of multi-environment trial data and other types of two-way data. Agronomy Journal. 2001;93:1111-1118.
21. Yan W, Kang MS. GGE Biplot Analysis: A Graphical Tool for Breeders, Geneticists, and Agronomists. Boca Raton (FL): CRC Press; 2003.
22. Yan W, Tinker NA. Biplot analysis of multienvironment trial data: principles and applications. Canadian Journal of Plant Science. 2006;86:623-645.
23. Yan W, Hunt LA, Sheng Q, Szlavnics Z. Cultivar evaluation and mega-environment investigation based on GGE biplot. Crop Science. 2000;40:597-605.
24. Yan W, Kang MS, Ma BL, Woods S, Cornelius PL. GGE biplot vs. AMMI analysis of genotype-byenvironment data. Crop Science. 2007;47:643-653.
25. Yang RC, Crossa J, Cornelius PL, Burgueno J. Biplot analysis of genotype × environment interaction: proceed with caution. Crop Science. 2009;49:15641576.
26. Yathish KR, Kumar S, Rao TV, Kumar P, Karthik M, Das AK, et al. GGE biplot and AMMI analysis for stability and adaptability of dual-purpose maize hybrids tested across multi-environments for baby corn and fodder yield. Range Management and Agroforestry. 2024;45:49-56.