AMMI analysis of variance revealed highly significant effects of environment, genotype and genotype × environment interaction (GEI) on pod yield and kernel yield (Table 1). The environment accounted for the largest proportion of the total variation, contributing 78.58% for pod yield and 79.73% for kernel yield. In comparison, genotype effects explained only 7.68% and 6.75% of the variation for pod yield and kernel yield, respectively, while GEI contributed 13.73% and 13.50%. Similarly,
Mahamane et al. (2026) reported the environment accounted up to 84.7% of the total yield variation. These results indicate that environmental and GEI effects had a much greater influence on yield performance than genotype alone, highlighting the strong sensitivity of groundnut productivity to agro-ecological variability. Consequently, genotype performance varied considerably across environments. Similar observations have been reported in earlier studies, highlighting the critical role of GEI in yield expression and adaptability
(Agrawal et al., 2025; Sharma et al., 2025 and
Choudhary et al., 2026). Hence, multivariate analytical approaches such as AMMI and GGE biplot analysis are essential for partitioning interaction effects, understanding genotypes stability and identifying high-yielding, widely adaptable genotypes across diverse environments
(Shilpa et al., 2025).
Genotype performance and AMMI stability analysis
The mean performance, IPCA scores and stability parameters of all genotypes are presented in Table 2. Genotype stability for pod yield and kernel yield was assessed using the Eberhart and Russell model, which evaluates stability based on mean performance, regression coefficient (bi) and deviation from regression (S
2di) (
Agrawal et al. 2025). For both pod and kernel yield, TCGS 2410 (2672.42, 1754.29), Kadiri Lepakshi (2587.52, 1741.27), TCGS 2306 (2399.19, 1588.38), TCGS 2409 (2306.81, 1547.10) and Visishta (2249.09, 1512.55) recorded comparatively higher mean performance. Among them, Visishta (bi = 1.14, 1.16), TCGS 2410 (bi = 1.34, 1.27) and Kadiri Lepakshi (bi = 1.55, 1.52) showed regression coefficients greater than unity coupled with non-significant deviations from regression, indicating responsiveness to favourable environments and predictable performance. In contrast, TCGS 2409 (bi=0.97, 0.91) and TCGS 2306 (bi=1.10, 1.08) showed significant deviations from regression, indicating instability despite relatively high mean yield.
The AMMI Stability Value (ASV) identified genotypes with stable performance across environments, where lower ASV values indicate reduced GEI effects and greater stability (
Reddy et al. 2025). For pod yield, TCGS 2306 (8.4), TCGS 2020 (9.0), TCGS 2347 (13.3), Kadiri Lepakshi (13.3), Kadiri 6 (14.4) and Visishta (15.8) exhibited relatively low ASV values. Similarly, for kernel yield, TCGS 2347 (3.6), TCGS 2020 (5.5), TCGS 2306 (13.0), Kadiri Lepakshi (14.2), K 1774 (14.6) and TCGS 2410 (14.6) recorded lower ASV values, suggesting consistent performance across environments. Interestingly, TCGS 2306 and Kadiri Lepakshi recorded pod and kernel yields above the overall mean, highlighting their superior and stable performance.
The Genotype Stability Index (GSI), which integrates yield and stability, identified TCGS 2306 (11.4) and Kadiri Lepakshi (15.3) as desirable genotypes for pod yield, while TCGS 2347 (12.6), TCGS 2410 (15.6), TCGS 2306 (16.0) and Kadiri Lepakshi (16.2) showed lower GSI values for kernel yield. Similarly, the Yield Stability Index (YSI) indicated superior stability and yield performance for TCGS 2306 (4), Kadiri Lepakshi (6), TCGS 2410 (10) and Visishta (11) for pod yield and for Kadiri Lepakshi (6), TCGS 2306 (6), TCGS 2410 (7) and TCGS 2347 (10) for kernel yield. Overall, the consistent performance of TCGS 2306 and Kadiri Lepakshi across GSI and YSI analyses highlights their potential as stable and high-yielding genotypes suitable for diverse agro-ecological conditions.
AMMI biplot analysis
The AMMI2 biplot, constructed using the first two interaction principal component (PC1 and PC2), provides an effective graphical depiction for interpreting GEI and identifying genotypes with wide or specific adaptability
(Khan et al., 2021). For both pod yield and kernel yield (Fig 1), the first two PCs explained a substantial proportion of the GEI variance. In pod yield, PC1 (41.3%) and PC2 (25.9%) together accounted for 67.2% of the total GEI, similarly for kernel yield, PC1 (46.3%) and PC2 (21.2%) jointly explained 67.5% of the interaction variance. These results indicates that the first two components sufficiently captured the major interaction patterns, enabling reliable interpretation of the AMMI model. Similar findings were reported by
Khan et al. (2021) and
Ntsoli et al. (2026). The biplot further revealed that environments located near the origin exhibited minimal interaction effects and greater stability, whereas those farther from the origin showed stronger interaction effects and greater discriminating ability
(Esan et al., 2023). Accordingly, Utukuru, Ananthapuramu and Yellamanchili had shorter vectors, indicating lower interaction effects for both traits. Conversely, Kadiri, Peddapuram and Darsi displayed longer vectors, reflecting greater discriminatory power and suitability for identifying specifically adapted genotypes. For pod yield, genotypes Kadiri 6, Visishta, K 1774 and TCGS 2020 were positioned near the origin, indicating wider adaptability and lower interaction effects. Conversely, TCGS 2019, TCGS 2438, TCGS 2349, TCGS 2409, Kadiri Lepakshi and TCGS 2306 were located farther from the origin, suggesting specific adaptation to particular environments. A similar trend was observed for kernel yield, where TCGS 2410, K 1774 and TCGS 2020 showed relatively stable performance across environments, while TCGS 2019, TCGS 2438, TCGS 2306 and TCGS 2409 exhibited stronger interaction effects and specific adaptation. Furthermore, Kadiri Lepakshi and TCGS 2410 were closely associated with the Kadiri environment, suggesting better performance under those conditions, whereas Kadiri 6 and TCGS 2432 appeared better adapted to the Peddapuram environment.
GGE biplot
The AMMI model, through biplot analysis, effectively identifies GEI across diverse environmental conditions. However, the GGE biplot provides a more comprehensive assessment by simultaneously considering genotype main effects and GEI effects, enables clearer and more intuitive interpretation of genotype behaviour in METs, particularly for evaluating newly developed varieties and hybrids
(Khan et al., 2021).
Unlike the AMMI2 model, which applies principal component analysis (PCA) solely to the GEI component, the GGE biplot incorporates both genotype main effects and GEI effects. Consequently, it is considered more effective for identifying specifically adapted genotypes as well as those combining high yield and stability across environments
(Khan et al., 2021 and
Ntsoli et al. 2026). In the GGE approach, PC1 represents the mean performance of genotypes, while PC2 reflects stability and the magnitude of interaction effects. In this study, PC1 accounted 56.54% of variation for pod yield and 52.37% for kernel yield, whereas PC2 explained 17.11% and 23.91% of variation, respectively. Together, these components captured a substantial proportion of the total variation, enabling effective interpretation of genotype performance across environments.
Mean vs stability
The mean versus stability GGE biplot, constructed using the first two principal components, was used to evaluate the average performance and stability of 14 groundnut genotypes for pod yield (Fig 2A) and kernel yield (Fig 3A). For pod yield, PC1 and PC2 together explained 73.65% of the total G + GE variation. In this approach, the Average Environment Coordination (AEC) abscissa represents the direction of increasing mean performance, with genotypes located toward the positive end exhibiting higher mean yield. Accordingly, Kadiri Lepakshi and TCGS 2410 were positioned on the positive side of the AEC abscissa, reflecting superior performance for both pod yield and kernel yield across environments. Genotype stability is determined by the projection onto the AEC ordinate, where shorter projections indicate greater stability and longer projections reflect higher instability
(Esan et al., 2023). In this context, Visishta, TCGS 2409 and TCGS 2020 exhibited better stability coupled with moderate to good kernel and pod yield. Furthermore, Kadiri Lepakshi and TCGS 2410 showed high mean performance with minimal deviation from the AEC axis, highlighting their desirable stability and wide adaptability. In contrast, TCGS 2347, TCGS 2349, TCGS 2306 and TCGS 2432 exhibited larger deviations from the AEC ordinate, suggesting lower stability and greater sensitivity to environmental variation, thereby indicating specific rather than broad adaptation.
Relationships among the environments
The relationship among test environments for the evaluated traits are illustrated in Fig 2B for pod yield and Fig 3B for kernel yield. For both pod yield and kernel yield Kadiri, Peddapuram, Utukuru, Ananthapuramu and Tirupati were positioned along the positive side of PC1, indicating similar environmental responses and positive correlations among these locations. Among them, Kadiri and Tirupathi were located farther from the origin, suggesting stronger discriminating ability and greater contribution to GEI. In contrast, Darsi was also positioned far from the origin but in a different direction, indicating pronounced interaction effects and a distinct environmental response compared to the other locations. Conversely, Ananthapuramu and Yellamanchili were situated close to the origin, reflecting lower discriminating ability and weaker interaction effects.
Discriminativeness and representativeness
The identification of an ideal test environment involves assessing both representativeness and discriminating ability, which can be effectively achieved using the GGE biplot approach
(Khan et al., 2021). In the present study, the environment vector view of the GGE biplot for pod yield (Fig 2C) and for kernel yield (Fig 3C) illustrates the Average Environment Axis (AEA), which passes through the biplot origin and the mean position of all environments. This visualization enables interpretation of relationships among test environments in terms of correlation, representativeness and discriminate ability.
The angle between the vectors of any two environments indicates their degree of correlation. Tirupati and Kadiri were oriented in a similar direction with relatively longer vectors, indicating strong positive association and high discriminating ability for genotypic performance. Peddapuram also exhibited a comparatively long vector, suggesting good discriminatory capacity. In contrast, Utukuru, Yellamanchili and Ananthapuramu were positioned closer to the origin, showed weaker GEI effects and reduced discriminating capacity.
Environments forming smaller angles with the AEA are considered more representative of the target environments. An ideal test environment should therefore combine high discriminating ability with strong representativeness for effective selection of widely adapted genotypes. In this study, Kadiri and Darsi showed high discriminating ability, with Darsi exhibiting the longest vector. However, Kadiri, Tirupati, Peddapuram and Utukuru formed smaller angles with the AEA, indicating better representativeness across testing conditions. In contrast, Darsi showed a larger angle with the AEA, suggesting lower representativeness despite its strong discriminating capacity. Overall, Kadiri emerged as the most ideal test environment, combining high discriminating ability with strong representativeness for effective evaluation and selection of groundnut genotypes in METs.
Which-won-where
The “which-won-where” view of the GGE biplot is a powerful graphical tool for visualizing crossover GEI, identifying mega-environment differentiation and determining genotype-specific adaptation for pod yield (Fig 2D) and kernel yield (Fig 3D). In the polygon view, genotypes farthest from the origin form the convex hull, while perpendicular rays divide the biplot into sectors representing potential mega-environments (
Kona et al. 2024). The vertex genotypes of each sector are considered the highest-performing genotypes within the respective environments. For pod yield, the polygon vertices included TCGS 2306, TCGS 2432, Kadiri 6, K 1774, K 2349, TCGS 2410 and Kadiri Lepakshi, representing the best performers in their respective environments. The biplot formed six sectors, of which two contained test environments, indicating two major mega-environments. The first mega-environment comprised Kadiri, Tirupati, Utukuru and Peddapuram, where Kadiri Lepakshi and TCGS 2410 were identified as the winning genotypes. The second mega-environment included Darsi and Yellamanchili, where TCGS 2019, TCGS 2349, K 1774 and Kadiri 6 showed superior performance. Genotypes positioned near the origin exhibited greater stability but lower responsiveness to specific environments. While for kernel yield, the winning genotypes were TCGS 2306, Kadiri 6, K 1774, K 2349, TCGS 2019 and Kadiri Lepakshi and the biplot was divided into 4 sectors. Peddapuram, Utukuru, Kadiri, Tirupati and Ananthapuramu constituted the first mega-environment where Kadiri Lepakshi showed superior performance, while Darsi and Yellamanchili formed the second mega-environment, where TCGS 2019, TCGS 2349, K 1774 and Kadiri 6 performance best. Similar results were reported by
Rao et al. (2023);
Soni et al. (2026) and
Mahamane et al. (2026), highlighting the effectiveness of the GGE biplot analysis in identifying mega-environments and specifically adapted high-performing genotypes.
Multi trait stability index
The Multi-Trait Stability Index (MTSI) analysis identified superior genotypes among the 14 evaluated groundnut entries, as presented in Fig 4. The selected genotypes exhibited higher mean performance (Xs) compared with the overall population mean (Xo), along with predominantly positive selection differentials for most traits, indicating considerable scope for genetic improvement through MTSI-based selection. The negative selection differential for days to 50% flowering reflects a desirable shift toward early-maturing genotypes. Factor analysis showed that pod yield and kernel yield (Factor 1) were the primary contributors to overall productivity, while test weight and sound mature kernels (Factor 2) were associated with kernel quality attributes (Table 3). Additionally, days to 50% flowering and maturity (Factor 3) captured variation related to phenological adaptation. Based on the lowest MTSI scores, TCGS 2438 and Kadiri Lepakshi were identified as superior genotypes, combining high yield potential with stable performance across environments, thereby demonstrating their suitability for broad adaptation and multi-trait improvement
(Zuffo et al., 2020 and
Ntsoli et al., 2026).
