Field experiments were conducted at six different environments in the Rangpur Division, Northern part of Bangladesh, namely, Panchaghar (E1) (26˚20'N latitude and 88˚34'E longitude), Thakurgaon (E2) (26˚03'N latitude and 88˚46'E longitude), Nilphamari (E3) (25˚98'N latitude and 88˚91'E longitude), Lalmonirhat (E4) (25˚91'N latitude and 89˚45'E longitude), Dinajpur (E5) (25˚37'N latitude and 88˚39'E longitude), and Rangpur (E6) (25˚44'N latitude and 89˚15'5.98"E longitude) during Rabi season 2020 to 2021, respectively. The environments E1, E2, E3, E4, E5 and E6 are popular wheat growing regions that fall under different agro-ecological zones of Rangpur Divisions viz., Old Himalayan Piedmont Plain (AEZ 1), Active Tista Floodplain (AEZ 2), Tista Meander Floodplain (AEZ 3), High Barind Tract (AEZ 26) and North Eastern Barind Tract (AEZ 27) in Bangladesh (Figure 1). Details of the soil and meteorological conditions during the crop growth period are presented in Table 1.

Twelve wheat genotypes were used in this study (Table 2). Seeds of these wheat genotypes were obtained from the Bangladesh Agricultural Research Institute (BARI), Joydebpur, Gazipur, Bangladesh. The field trial was conducted using a randomized complete block design (RCBD) with three replications in each environment. Establishment and management of the experiment land were prepared by conventional ploughing three times through a power tiller for fine tilth and the basal fertilizer of 100 kg N + 80 Kg P₂O₅ + 20 Kg K₂O + 60 kg CaSO₄⋅2H₂O + 3 t/ha organic manure (FYM) was applied. Each genotype was planted in each environment had six rows of 2.5 m long spaced 20 cm apart with a plot area of 1.2 m × 2.5 m (3 m²). Initial germination and seed moisture content were 97% and 10%, respectively, and these were determined as defined by ISTA [41].

Figure 1. Spatial distribution of field experiment locations across Rangpur Division. Six districts were utilized as experimental sites: Panchaghar (E1), Thakurgaon (E2), Nilphamari (E3), Lalmonirhat (E4), Dinajpur (E5), and Rangpur (E6), representing the agro-ecological diversity of the region. Map coordinates range from 88.4˚E to 89.6˚E longitude with a 40 km scale reference.

Table 1.Weather parameters during the Rabi season 2020-2021.

Table 2.Description (background information) of twelve wheat genotypes used for the study.

^aNames in parenthesis represents the local names of the released variety; BAW, Bangladesh advanced wheat accession; d, day; no., number; t/ha, ton per hectare.

Agronomic practices and plant protection measures accomplished throughout the crop growth period as per the local farmer’s practices and irrigation (3 times) was applied. The morphological observations were recorded on 20 randomly selected plants in each genotype and each environment with three replications based on the descriptors of wheat [42]. The traits viz., plant height (cm), spike length (cm), number of tillers per plant, number of spikelets per spike, spike weight per plant, grains weight per spike (g), thousand seed weight (g) and grain yield (t/ha) were recorded. The mean data used in the analysis was the average value of each genotype from each environment with three replications.

The mean data of 20 randomly selected plants in each genotype for each trait were used for determining the range and overall mean of each environment. Correlation between grain yield and other yield attributing traits in each environment were performed using SPSS 16.0 version (SPSS. Inc USA, 2007). The G × E SPSS output consists of ready-to-use input files in R-package for multivariate analysis and univariate stability results. The morphological trait was subjected to analysis of variance (ANOVA) to determine phenotypic variations among the genotypes, locations, genotype by location using R-package. Simple correlation coefficients were determined using Spearman’s rank (a simplified version of R statistical software) developed by the R Core Team [43]. The Graphical user interface (GUI) package of R studio was used for GGE biplots, consisting of two concepts, the biplot concept [44] and the GGE concept [45]. The GGE biplots are a graphical picture to illustrate G × E interaction and genotype ranking based on mean and stability. The graph generated is based on mega environment evaluation (which-won-where pattern), Genotype evaluation (mean versus stability), and tested environment raking (discriminative versus representative). The GGE biplots were constructed using the first and second principal components (PC1 and PC2) that were derived by subjecting environment-centered means grain yield to singular value decomposition. The options used for data analysis were environment centering (Centering = 2), no standardization (Scale = 0) and no transformation (Transform = 0). The biplot was based on environment-focused singular-value partitioning (SVP = 2), which is suitable for picturing the relationships among the genotypes and locations.

The AMMI stability value (ASV) was calculated by the method formulated by Purchase et al.[46].

where SS represents the sum of squares of first (PC₁) and second (PC₂) interaction principal component axes; and PC₁ and PC₂ are the genotypic scores obtained from the AMMI model.

Genotype selection index (GSI) was obtained by following the method devised by Farshadfar and Sutka [47].

where GSI_i denotes the genotype selection index for i^th genotype, RY_i is rank of mean grain yield for i^th genotype, RASV_i represents rank for the AMMI stability value for the i^th genotype.

GGE analysis was performed using the software GEA-R version 4.1 Angela et al.[48] with the model equation:

where Y_ij is the yield of i^th genotype in the j^th environment; G_i and E_j represent the genotype and environment deviations from the grand mean, respectively; µ denotes the grand mean λ_k is the eigenvalue of the PCA axis k; α_ik and γ_jk indicate the genotype and environment PC scores, respectively, for the axis k and e_ij denotes the error term.

Analysis of Variance AMMI analysis of variance for pooled mean grain yield of 12 genotypes from six environments (E1, E2, E3, E4, E5 and E6) explained that the major portion of the total sum of squares contributed by genotypic effects (58.34%) followed by GEI effects (24.55%) and environmental effects (17.10%) (Table 5). AMMI ANOVA revealed significant differences among 12 wheat genotypes and six environments. This depicts that the grain yield of wheat influenced by genotype (G), environment (E) and the interaction between genotype and environment (GEI). AMMI analysis further partitioned the GEI into the first two multiplicative terms namely PC1 and PC2 with a contribution of 40.80% and 29.70% of GEI sum of squares. The presence of a significant proportion of GEI necessitates the analysis of the stability of wheat genotypes over environments.

Table 5.AMMI analysis of variance for mean grain yield of 12 wheat genotypes from six different locations during Rabi season 2020-2021.

Grain yield analysis of wheat revealed that the average environment means ranged from 3.75 t/ha (E5) to 3.81 t/ha (E3). Seven genotypes showed above-average yield in the E1 environment. Six genotypes in E2, ten genotypes in E3, and nine genotypes in the E4, E5 and E6 environment outperformed the average yield of the corresponding genotypes in a particular environment (Table 6). The present study showed that the environment E1 was found with the longest vector having more discriminating power compared to the other five environments. The Average Environment Axis (AEA) view compared the environments in relation to an ideal environment. The environment E1 had the smallest angle with the AEA, hence the E1 environment is highly representative. The Average Environment Coordinate (AEC) axis projected the stability of the accessions. The highly stable genotypes indicated by a small perpendicular line to the AEC axis whereas the increase in the length of the perpendicular line denoted the decrease in stability of the genotypes. Genotypes (G5, G2, G12 and G11) were highly stable with low to good yielding ability. The genotype G4 was highly unstable because it was far away from the AEC axis followed by G10, G8 and G6. These genotypes expressed good yielding ability (Figure 2).

AMMI analysis showed the interaction component PC1 with 40.8% of the total interaction sum of squares, whereas PC2 attributed for 29.7%. The performance of genotypes in a specific environment, as well as the overall performance across all the test environments, can be effectively analyzed in the presence of two or more PCA axes. PC1 scores for grain yield over six locations were plotted against genotype and environment scores along with the environments (E1, E2, E3, E4, E5 and E6) (Figure 2). The genotypes plotted on right side of the central axis formed based on grand mean, exhibited high yield compared to those on the left side of the axis. Four genotypes (E4) exhibited the above-average performance with positive interaction effect were present in the quadrant I. Three genotypes in the quadrant IV also showed the above-average performance but having a negative interaction effect. About two genotypes along with the environment E5 falling under quadrant II showed below-average performance with a positive interaction effect. The environment E3 is between quadrant (I and II), E1, E2 and E6 is between quadrant III and IV. About three genotypes present in quadrant III also exhibited below-average performance with negative interaction. About five genotypes having PC1 score nearer to zero as it exhibited between −0.13 to 0.30 (Table 6) which were stable across all the test environments (E1, E2, E3, E4, E5 and E6). A set of four genotypes found to be less stable as it explained moderately larger scores.

Table 6.Mean grain yield, AMMI stability value (ASV) and genotype selection index (GSI) of wheat accessions.

^aNames in parenthesis represents the local names of the released variety; E1, Panchaghar; E2, Thakurgaon; E3, Nilphamari; E4, Lalmonirhat; E5, Dinajpur; E6, Rangpur.

Figure 2.(a) GGE biplot based on environment focused scaling for the comparison of six environments in relation to average environment axis (AEA). (b) Average Environment Coordination (AEC) view based on environment focused scaling for the pooled mean performance and the stability of 12 wheat genotypes.

Figure 3.(a) AMMI1 biplot showing main effects and PC1 interaction effects of 12 wheat genotypes and six environments on grain yield (GY) of wheat. (b) AMMI2 biplot of first two principal components (PC1 vs PC2) of interaction effects.

For the maximum exploration of GEI, the PC1 scores visualized against PC2 in the form of a graph (Figure 3). The three genotypes with PC2 value ranging between −0.07 to 0.05 located nearer to the center of the biplot, produced highly stable grain yield across all the test environments. Three genotypes fall under medium stable genotypes category which surrounded the highly stable genotypes in the graph. The highly unstable six genotypes that were far away from the center, had also been identified from the biplot. Furthermore, the biplot showed a highly interactive environment as E6 and E2 as well as the genotypes (G2 and G4) that contributed largely to the GEI. The genotypes G8, G1 and G11 particularly performed well in E1 and E3. The genotypes G7 and G10 were particularly adapted to E4 and E5, whereas the genotype G2 (E6 and E2) and G4 had more interaction with E4 and E5. According to ASV, the genotypes close to 0 (G11, G10 and G9) were highly stable, whereas the genotypes having high value (G7, G6, G3, and G8) were found to be highly unstable (Table 6). GSI integrates both yield and stability across environments. Genotypes with lower GSI (G11, G9, and G4) were desirable since they combine high mean yield performance with stability.

The GGE biplot analyses (which-won-where, genotype ranking, discriminativeness, representativeness and relationship among environments) for grain yield (Figures 4(a)-(d)). Figure 4(a) shows the which-won-where grain yield performance of the wheat genotypes evaluated in the six environments. Locating the most suitable accession for each environment can be done by which-won-where pattern analysis. In this, a polygon is produced by joining the genotypes far away from the origin consisted of all the other genotypes inside the polygon. The polygon is further portioned into five different sectors using the rays (dot line) that were starting from biplot origin and passing perpendicular to the sides of the polygon. The genotypes in a sector are similar in performance compared to the genotypes in other sectors. E3 and E5 environment is more suitable for the following genotypes viz., G11 and G5 which are located in the different sector whereas the environment E2 and E4 is highly desirable for G4 and G10. The sector that consisted of the E1 environment contained G8, G6, and G1 genotypes thus portraying that the E1 environment is highly suitable for the expression of most of the genotypes. Few of those genotypes that highly performed in E6 were G2, and G12. The genotypes plotted at each vertex of the polygon were the best performing genotype to the environment nearer to the vertex. In this case, G2, G4, G10, G11, and G8 were the peak genotypes that are highly suitable for the environments E1, E2, E3, E4, E5, and E6 respectively.

An ideal genotype should have both high mean performance and high stability across environments. The center of the concentric circle (Figure 4(b)) is the location for the ideal genotype. Among the test genotypes, the one closest to the point is the best. However, G11 had the highest grain yield among the 12 genotypes; G1 that possessed both high mean grain yield and high stability is closest to the ideal genotype for grain yield with consistency of performance across environments. The ‘discrimination and representativeness’ view (Figure 4(c)) have the graphical ability to unravel the discriminating ability and representativeness present among the six test environments. The plot is more relevant for genotypes evaluation as it was constructed by genotype-metric preserving (SVP = 2). Thus, genotypes G7 (BARI Gom 25) had maximum yield across environments followed by G4 (BARI Gom 30), G11 (BARI Gom 21), G6 (BARI Gom 27) and G5 (BARI Gom 28) while G10 (BARI Gom 24) followed by G9 (BARI Gom 23), and G1 (BARI Gom 31) were recorded lowest yields. Thus, genotypes G2 (BARI Gom 32) and G4 (BARI Gom 30) were showed more variability i.e. highly unstable whereas entry G12 (BARI Gom 19) and G8 (BARI Gom 26) were more stable. The GGE-biplot approach, which was based on environment focused scaling, was used to estimate the relationships between the environments (Figure 4(d)). This GGE biplot approach suggested that E2 and E6 were the most closely correlated environments with E6 and E5 closely behind. However, the largest correlation coefficients were between E3 between E1. Some contradictions between the figures and actual correlations were predictable because the biplot did not estimate 100% of the GGE variation.

Figure 4.GGE Biplot for root yield indicating (a) superior genotype in a particular environment (b) ranking of genotypes (c) discriminativeness vs. representativeness (d) relationship among environments.