Combined ANOVA and genotype mean performance
Pooled analysis of variance (Table 2) was performed to elucidate the main effect and quantify the interactions among and within the source of variations. The mean squares exhibited significant differences among environment (except FC), genotypes and G×E interaction for all the assessed traits that might be due to environmental change or genetic makeup that vary along with changing environment. The subdivision of the G×E interaction percentage (% of GEI) is calculated from the total sum of the square which explains percentage of variation for all traits. For environment, all traits showed a substantial level of variation that spanned from 0.66 (fiber content) to 92.61% (plant height). With regard to genotype effect, the trait plant height (3.04%), days to last picking (5.38%) and fruits per plant (5.55%) had adjoining contribution while, fiber content (64.61%) had greater variation. The genotype by environment interaction (GEI) effect ranged from 4.35% (plant height) to 43.13% (average fruit girth). The significance in environment effect for fruit yield emphasized the necessity of multienvironment trials that will help in both ways viz., to find best hybrids having broad adaptability as well as hybrid selection with average adaptability to a particular agro-ecology.
Table 2
Conjoint ANOVA for evaluation of significant level for okra fruit yield and its components
Source of variation | Environment | Genotype | GEI | Residual |
df | 3 | 34 | 102 | 272 |
Trait | Mean Squares | % (G + E + GEI) | Mean Squares | % (G + E + GEI) | Mean Squares | % (G + E + GEI) | Mean Squares |
DF | 1109.24** | 58.67 | 26.94** | 16.15 | 14.00** | 25.18 | 4.77 |
DFP | 2014.68** | 70.31 | 26.78** | 10.59 | 16.10** | 19.10 | 6.11 |
DLP | 12607.96** | 86.67 | 69.03** | 5.38 | 34.00** | 7.95 | 16.04 |
AFL | 178.17** | 63.11 | 3.71** | 14.91 | 1.82* | 21.98 | 1.28 |
AFG | 2.59** | 12.16 | 0.84** | 44.71 | 0.27** | 43.13 | 0.02 |
AFW | 177.57** | 56.01 | 5.33** | 19.04 | 2.33** | 24.96 | 1.03 |
PH | 127848.10** | 92.61 | 370.40** | 3.04 | 176.49** | 4.35 | 86.86 |
NOB | 38.88** | 50.36 | 1.99** | 29.29 | 0.46** | 20.35 | 0.04 |
NON | 1497.35** | 88.10 | 10.62** | 7.08 | 2.41* | 4.82 | 1.92 |
IL | 124.69** | 61.89 | 3.31** | 18.61 | 1.16** | 19.49 | 0.53 |
NOF | 1661.81* | 88.47 | 9.20** | 5.55 | 3.30** | 5.98 | 1.85 |
FY | 320001.77 ** | 81.95 | 3318.24** | 9.63 | 967.28** | 8.42 | 370.04 |
FC | 0.06 | 0.66 | 0.50** | 64.61 | 0.09** | 34.73 | 0.05 |
DF: Days to 50% flowering, DFP: Days to first picking, DLP: Days to last picking, AFL: Average fruit length, AFG: Average fruit girth, AFW: Average fruit weight, PH: Plant height, NOB: Branches/plant, NON: Internodes/plant, IL: Internodal length, NOF: Fruits/plant, FY: Fruit yield/plant, FC: Fiber content, df: degrees of freedom, SS: Sum of squares, MS: Mean sum of squares, GEI: Genotype by environmental interaction, * significant at 5%, ** significant at 1% |
A significant level of divergence among all the three effects viz., environment effect, genotype effect and G×E interaction effect assured the existence of multi-environments with distinct genotypes as well as high yield potential (Yan and Kang, 2002). However, the specifics of genotype by environment interaction cannot be fully interpreted by analysis of variance. Subsequently, accessory statistical methods like multivariate techniques can aid in understanding the GEI (Oladosu, et al., 2017). The change in response of genotype to various environments is clearly shown by genotype × environment interaction effect, pinpointing the requirement of genotypes evaluation across varying environments. Similarly, GEI reveals the obstacles faced by plant breeders while selecting a superior genotype prior to its release as a variety for commercial cultivation (Yan and Kang, 2002).
Supplementary Fig. 1 portrays the variation for various traits in 35 okra genotypes in the form of box plots. The estimates of range, mean, and variance for the characteristic it reflects are portrayed by box plot.
Genotype by environment interaction
Table 3 represents the AMMI model-based analysis of variance for fruit yield of okra hybrids tested under four environments. The results showed major additive portion of the total sum of squares contributed by environment (81.95%) followed by genotype (9.63%) and GEI (8.42%). Thus, fruit yield is influenced by genotype, environment and genotype × environment interaction. The GEI effect was further partitioned into three multiplicative terms popularly known as interactive principal component (IPCA I, IPCA II and IPCA III) with the contribution of 67.60 percent, 27.70 percent and 4.60 percent of GEI sum of squares. Significant GEI effect and interactive components indicated the need to identify specific environment and genotype.
Table 3
AMMI based analysis of variance of okra fruit yield across four environments
Source | df | SS | MS | Total Explained variation (%) | GEI Contributed (%) |
Environment | 3 | 960005.32 | 320001.77** | 81.95 | - |
Genotype | 34 | 112820.10 | 3318.24** | 9.63 | - |
GEI | 102 | 98662.96 | 967.28** | 8.42 | - |
IPCA I | 36 | 66727.33 | 1853.54** | - | 67.60 |
IPCA II | 34 | 27374.47 | 805.13 ** | - | 27.70 |
IPCA III | 32 | 4561.16 | 142.54 | | 4.60 |
Residual | 272 | 100652.09 | 370.04 | - | - |
Total | 521 | 1374826.06 | 2638.82 | - | - |
df: degrees of freedom, SS: Sum of squares, MS: Mean sum of squares, GEI: Genotype by environmental interaction, IPCA: Interactive principal component analysis, * significant at 5%, ** significant at 1% |
AMMI analysis
The stability of genotype, fruit yield potential and association of test environments were visualized by use of different biplots. The additive main effects and multiplicative interaction I (AMMI I) biplot (fruit yield vs. IPCA I) depicted the relationship among okra genotypes and test environments. In AMMI I biplot, X-axis represent the first principal component while Y-axis represent significant interaction of fruit yield. The genotypic mean and its interaction with environment reflected a small degree of genotype similarity in this investigation. For fruit yield, environment four (E4) with PCA I score nearer to zero comparatively over other environments, depicts minor interaction effect and verified significant outcome of all hybrids. Placement of environments E3 and E4 on the right side of the biplot’s grand mean line indicated their potentiality for fruit yield while E1 and E2 placed on the left-hand side indicated poor environments. The PCA 1 axis estimates of hybrids H20, H3, H16, H24, H2, H6, H15 and H11 (Supplementary Fig. 2) was near to zero designating less influence by the environment on them as well as adaptability to all environments. Meanwhile, genotypes and environment with same sign and PCA I score but away from zero lines of biplot indicated that they show a positive correlation and those genotypes showed specific adaptation to that environment viz., genotypes H2 in E4, L2 in E1, T2 in E2 and H17, H21 in E3.
Additive main effects and multiplicative interaction II (AMMI II) is a graphical representation of PC I vs PC II scores that identifies genotypes with broad or narrow adaptation and have some benefits than combined regression-based analysis. Figure 1 depicts AMMI II biplot with the top two PCs accounting 95.30% of the fruit yield variance due to G + G×E interaction. This implicit those first two principal components chiefly presented the interaction of 35 okra genotypes evaluated under four environments. This was in accordance with Gauch and Zobel, 1996 who affirmed the first two PCs as adequate for estimation of the AMMI model. The two-line which passes vertically and horizontally divides the biplot (0, 0) in four sectors (Fig. 1). The distance between biplot origin and genotype or environment will decide the degree of interaction by the environment over the genotype and vice versa. In this study, we found that for fruit yield, environment E2 with a shorter vector is less interactive and contributes much to the stability of genotypes comparatively, over other environments. Also the genotypes that are placed far from the origin of biplot and near to any environment are considered as winning over the environments in which they fall. Genotypes H5, H3 and T1 found close to environment E1, depicted their top performance and adaptation in this environment than the other three. Whereas, testers T3 and T4 performed better in environment E2, hybrids H13, H12, H23, H22, H9 and H21 performed better in environment E3, while hybrids H2, H24, H6 and H1 performed better in environment E4. Similar results of okra hybrids with high yield under specific environments have been reported by Sanwal et al. 2021, Nwangburuka et al. 2011. The presence of G×E interaction was reflected owing to dissimilarity in comparative performance of genotypes in diverse environments. Therefore, okra hybrids with environment-specificity should be chosen for various agro-ecologies and environmental conditions.
In general, most of the genotypes clustered together on the biplot origin implying their parallel response in every evaluated environment than other genotypes situated at distant from each other, which cofirms the report of Akter et al. 2014. Also genotypes positioned away from biplot origin portrayed their sensitivity to environmental interaction. AMMI II model is one the best model to figure out performance and stability of genotypes with genetic variation among them and their relationship to test environments (Miranda et al., 2009).
Genotypes with low AMMI stability value (ASV) indicate low G×E interaction and high stability. In the present experiment, hybrids H2, H16 and H15 having lowest ASV (0.90, 0.97 and 1.09) and mean fruit yield of 164.06, 155.50, and 166.41 g, respectively, showed high stability (Table 4). But they did not posses high yield and therefore they might not be selected where higher economic gains are required.
Table 4
Mean fruit yield (g), IPCA I, IPCA II and IPCA III score, AMMI stability value (ASV), rank of ASV, Yield stability index (YSI) and rank of YSI of okra hybrids tested across four environments
Genotype | Mean | IPCA I | IPCA II | IPCA III | ASV | rASV | YSI | rYSI |
H1 | 167.711 | 0.642 | 1.542 | –0.436 | 2.196549 | 9 | 19 | 10 |
H2 | 164.062 | 0.303 | 0.526 | 0.451 | 0.906719 | 1 | 14 | 13 |
H3 | 146.596 | 0.443 | –0.801 | 0.04 | 1.344754 | 4 | 31 | 27 |
H4 | 156.314 | 2.572 | 1.805 | 0.002 | 6.523803 | 28 | 50 | 22 |
H5 | 154.303 | 1.631 | –1.675 | 0.342 | 4.313445 | 19 | 43 | 24 |
H6 | 162.91 | –0.071 | 1.591 | 0.188 | 1.600243 | 6 | 21 | 15 |
H7 | 152.607 | 1.571 | –3.407 | 0.413 | 5.126832 | 23 | 48 | 25 |
H8 | 163.162 | –1.470 | 0.981 | 1.119 | 3.715415 | 16 | 30 | 14 |
H9 | 196.301 | –3.006 | –0.893 | –0.346 | 7.381608 | 30 | 32 | 2 |
H10 | 174.847 | –1.071 | 0.389 | 0.424 | 2.640173 | 12 | 18 | 6 |
H11 | 171.762 | 0.344 | –1.884 | 1.28 | 2.062459 | 8 | 15 | 7 |
H12 | 193.148 | –1.533 | 0.053 | 1.881 | 3.737426 | 17 | 20 | 3 |
H13 | 175.583 | –1.471 | –0.198 | –0.732 | 3.592319 | 15 | 20 | 5 |
H14 | 168.532 | –3.529 | 0.615 | 1.435 | 8.624699 | 33 | 42 | 9 |
H15 | 166.417 | –0.342 | –0.715 | 1.44 | 1.09803 | 3 | 14 | 11 |
H16 | 155.502 | 0.006 | –0.977 | 0.985 | 0.976755 | 2 | 25 | 23 |
H17 | 169.982 | –1.094 | 1.953 | 0.04 | 3.306421 | 13 | 21 | 8 |
H18 | 162.492 | –0.977 | 0.936 | –1.574 | 2.558812 | 10 | 26 | 16 |
H19 | 159.556 | –0.447 | 2.403 | –0.598 | 2.638387 | 11 | 29 | 18 |
H20 | 136.506 | 0.264 | –1.684 | 1.467 | 1.803123 | 7 | 41 | 34 |
H21 | 201.322 | –5.470 | –0.695 | –1.273 | 13.35069 | 35 | 36 | 1 |
H22 | 178.597 | –3.095 | –0.562 | –0.221 | 7.564532 | 32 | 36 | 4 |
H23 | 165.521 | –2.062 | 0.034 | –0.310 | 5.025222 | 21 | 33 | 12 |
H24 | 158.308 | 0.028 | 1.463 | 0.525 | 1.464755 | 5 | 24 | 19 |
L1 | 162.466 | 1.282 | 2.097 | 1.471 | 3.763921 | 18 | 35 | 17 |
L2 | 144.371 | 2.302 | 2.649 | –1.823 | 6.205585 | 27 | 56 | 29 |
L3 | 157.602 | –2.538 | –2.933 | –1.510 | 6.846434 | 29 | 49 | 20 |
L4 | 157.246 | –0.634 | 3.033 | –0.411 | 3.403793 | 14 | 35 | 21 |
L5 | 138.829 | 2.1 | 0.133 | –0.726 | 5.119642 | 22 | 53 | 31 |
L6 | 144.477 | 1.784 | 0.662 | –2.395 | 4.397645 | 20 | 48 | 28 |
T1 | 141.096 | 1.961 | –1.941 | –1.135 | 5.159161 | 24 | 54 | 30 |
T2 | 137.3 | 2.092 | –3.435 | –0.974 | 6.149053 | 26 | 59 | 33 |
T3 | 134.269 | 3.078 | –0.528 | 0.012 | 7.52095 | 31 | 66 | 35 |
T4 | 147.82 | 2.373 | –1.447 | –0.521 | 5.961573 | 25 | 51 | 26 |
SC | 137.76 | 4.034 | 0.906 | 1.471 | 9.874241 | 34 | 66 | 32 |
E1 | 126.788 | 4.713 | –5.291 | –3.466 | | | | |
E2 | 111.631 | 4.098 | –1.073 | 4.938 | | | | |
E3 | 235.175 | –10.377 | –1.621 | 0.133 | | | | |
E4 | 167.01 | 1.566 | 7.985 | –1.606 | | | | |
YSI integrates fruit yield and stability across environments. Lower YSI index represents genotypes with higher stability and greater productivity. The hybrids H2, H15, H11, H10 and H1 were recognized based on their YSI values (14, 14, 15, 18 and 19) as highly stable for fruit yield (Table 4). Although hybrids H21 and H9 showed good average fruit yield but their distance from biplot origin indicated their unstable performance. However they exhibited specific adaptation to environment E3. The hybrid H12 had high mean yield as well as located close to the origin suggesting it to be stable.
GGE biplot analysis
Which-won-where pattern analysis
Samonte et al. (2005) and Yan et al. (2001) followed the SREG model to construct two dimensional biplot named ‘which-won-where polygon’. The pattern analysis helps to identify the most suitable genotype for a given environment. Figure 2 represents the polygon view formed by joining the outermost genotypes for trait fruit yield. The variation for G + G×E for fruit yield was recorded 84.70%. The polygon is divided into eight different sectors using rays (dotted lines) that were starting from origin of plot and passing perpendicular to the sides of the polygon which identifies suitable genotype for a particular environment. As shown in the biplot, the four environments viz., E1: summer 2019, E2: late summer 2019, E3: monsoon 2019 and E4: late monsoon 2019 were positioned under three mega- environments where different genotypes excelled in each mega-environment. This result provided the confirmation on distinct interaction between genotype and environment for fruit yield. Biplot showed that hybrids H9, H12 and H21 placed at polygon vertex were most stable and yielded maximum in environment E1 and E2. This clearly depicts specific adaptations by genotypes to mega-environments (MGE) and thus positive exploitation of the G×E interaction (Mohammadi et al. 2010). Genotypes L4, L2, H20, T2 and H27 placed at vertex of the polygon but none of the environment found in that section showed their low yielding capacity under all tested environments. According to Yan and Tinker 2005, the ideal way to fully use the favourable effects of GE interaction positively would be to cluster the test environments into distinct MGEs and selection of various genotypes under each MGE. Genotypes located close to the origin of biplot were most adapted to low yielding environments while the genotypes positioned in the polygonal area were low performing to that environment as compared to vertex genotypes. Similar findings in okra were also reported by Sanwal et al. (2021), Nwangburuka et al. (2011) and Olayiwola and Ariyo (2013).
Mean yield vs stability pattern analysis
For fruit yield, the GGE biplot's "mean vs. stability" pattern captured 90.83% of the variation in G + G×E. The AEC abscissa in the vertical dimension and the AEC ordinate in the horizontal dimension are the two lines that make up the biplot graph. Line one (Fig. 3) representing single arrowhead and passing through the origin depicted greater mean performance of the genotype. Genotypes T4, H3, H5, H2, H18, H8, H10 and H12 produced greater fruit yield.
Further, the distance between genotype position and AEA (dotted line in the graph) in ranking biplot will decide the stability of genotypes. It will be represented by genotypes that are distant from the origin yet on the AEA or nearby. Accordingly, genotypes H10, H13, H14, H12, H5, H3 and T4 exhibited high stability.
However, high yielding hybrids such as H2, H18 and H8 exhibited G×E interaction effect which rendered them less stable. Earlier reports on fruit yield in okra also showed that the high yielding genotypes are not necessarily the most stable (Sanwal et al., 2021; Nwangburuka et al., 2011). Also hybrids H13 and H14, identified as stable manifested poor yield performance (Fig. 3). They might be utilized to enhance crop stability in the future breeding strategies.
GGE biplot graph helps in identifying genotypes which posses’ favourable mean performance and stability. The genotypes that had greater protrusion on AEC depicting higher mean together with small scale projection on AEA representing higher stability would be regarded as best genotypes (Yan et al. 2001; Yan and Kang, 2002; Farshadfar et al. 2010). Accordingly, genotypes T4, H3, H5, H10 and H12 were identified combining both, high and stable fruit yield. Thus, GGE biplot analysis helps in identification of supreme genotype. Utilizing this method, stable and high yielding okra genotypes have been identified by many authors (Sanwal et al. 2021; Nwangburuka et al. 2011).
The study evidently explained that it is not necessary that any good yielding genotype will also exhibit good stability. This is because yield is a complex quantitative trait whose performance being governed by polygenes is highly affected by genotype × environment interaction. Different environment situations including temperature variation or water stress causes differential expression of polygenes as a response to which difference in performance of genotype with environment occurs.
Discriminativeness vs representiveness pattern analysis
The discriminating ability viz., environment’s ability to differentiate genotype is measured by the distance of the vector from the biplot's origin and representativeness viz., the environment’s capability to represent all the tested environments is measured by angle between vector and average environment axis (AEC) (Oladosu, et al. 2017; Yan and Kang 2003). Higher discrimination capacity is shown by a longer length of the vector, and more environment representation is indicated by a smaller angle (Yan et al., 2007). In present study, as shown in supplementary Fig. 3, E1 environment with longest vector is most suitable to distinguish the genotypes and constitutes the most discriminative environment while, E2 with smallest angle is highly representative environment. This result indicated that generally adopted genotypes were selected from E2 while specifically adopted genotypes selected from the E1, E3 and E4 environments. However, none of the environment was found both discriminating as well as representative. Non-discriminating environments viz., E2, E3 and E4 does not appreciably distinguish between genotypes in a significant way and are thus not recommended (Yan and Tinker 2006). Similar results were also found by Sanwal et al. 2021 and Nwangburuka et al. 2011.
Multi-trait stability index (MTSI) analysis
Y × WAASB biplot is constructed using fruit yield on the abscissa and WAASB value on the ordinate (Fig. 4). This plot divides placed genotypes and environments in four quadrants which allow simultaneous selection of genotype with high yield and stability. First quadrant, with their high contribution to GEI, included unstable genotypes of low productivity and discriminative environments. According to biplot, 9 genotypes along with environment E1 and E2 fall in first quadrant. Those genotypes were unstable due to high WAASB value while, environment E1 had high discrimination ability due to high WAASB value. Second quadrant included unstable genotypes possessing higher yield above grand mean. In this quarter, 4 genotypes were placed along with environment E3 and E4. The genotypes were unstable while, environment E3 had high discrimination ability due to its highest WAASB index. It also suggested giving special attention to environment to get more yield. Third quadrant contains genotypes with greater stability (due to low WAASB value) but lower productivity. This quadrant covers 9 genotypes. The fourth quadrant indicated the genotypes with high stability and productivity due to low WAASB value and greater yield performance (Huang et al. 2021). Remaining 13 genotypes fall in the fourth quadrant of biplot among which hybrids H12, H13, H10 and H11 were highly productive and stable. Since all of the significant principal components are used in WAASB index calculation, it shows a better stability interpretation and the genotypes selected have more reliable stability. Thus, it can be used for broad adaptation and identification of stable genotypes for all environments.
The WAASBY values were calculated that weight between mean performance and stability for individual traits using Pearson's correlation matrix. With the purpose of grouping the traits with correlation into factors, those values were utilized to calculate an exploratory factor analysis. The factor analysis clustered thirteen traits into four factors which explained 78.39% of the total variation (Supplementary table 2). The communality ranged from 0.44 (Branches/plant) to 0.93 (Days to first picking) with an average of 0.78 after varimax rotation which means that a large proportion variance of each variable can be explained by the factors (FA). Plant height, internodes/plant, internodal length and fruits/plant were grouped in FA1 while days to 50% flowering, days to first picking and days to last picking in FA2, average fruit length, average fruit weight and fruit yield in the FA3 and FA4 included average fruit girth, branches/plant and fiber content (Table 5). Thus, factor analysis helped in reduction of data dimension.
Table 5
Selection differential, heritability and selection gains for mean and selection differential for WAASBY index obtained using MTSI for 13 traits of okra hybrids across four environments
Trait* | Factor | Mean performance | WAASBY |
Overall (X0) | Selected genotype (Xs) | SD (%) | h2 | SG (%) | Overall (X0) | Selected genotype (Xs) | SD (%) |
PH | FA 1 | 91.76 | 95.62 | 4.212 | 0.524 | 2.205 | 64.85 | 75.71 | 10.87 |
NON | FA 1 | 14.79 | 15.18 | 2.647 | 0.773 | 2.046 | 57.83 | 60.81 | 2.98 |
IL | FA 1 | 5.51 | 5.89 | 6.768 | 0.651 | 4.405 | 64.04 | 79.38 | 15.34 |
NOF | FA 1 | 13.37 | 13.91 | 3.983 | 0.641 | 2.553 | 51.35 | 62.71 | 11.36 |
DF | FA 2 | 42.63 | 43.55 | 2.150 | 0.480 | 1.033 | 33.53 | 46.89 | 13.36 |
DFP | FA 2 | 48.52 | 49.45 | 1.914 | 0.399 | 0.763 | 38.64 | 51.75 | 13.12 |
DLP | FA 2 | 79.33 | 81.57 | 2.815 | 0.507 | 1.429 | 55.77 | 76.94 | 21.17 |
AFL | FA 3 | 9.90 | 10.34 | 4.433 | 0.509 | 2.255 | 48.96 | 68.23 | 19.28 |
AFW | FA 3 | 10.18 | 10.63 | 4.454 | 0.563 | 2.508 | 51.13 | 67.27 | 16.14 |
FY | FA 3 | 160.20 | 172.10 | 7.491 | 0.708 | 5.308 | 48.79 | 64.78 | 15.99 |
AFG | FA 4 | 1.56 | 1.59 | 2.155 | 0.678 | 1.462 | 44.69 | 45.38 | 0.69 |
NOB | FA 4 | 1.31 | 1.44 | 9.929 | 0.768 | 7.629 | 53.29 | 61.84 | 8.55 |
FC | FA 4 | 4.60 | 4.59 | –0.291 | 0.821 | –0.239 | 54.07 | 49.10 | -4.97 |
FA1: Factor 1, FA2: Factor 2, FA3: Factor 3, FA4: Factor 4, DF: Days to 50% flowering, DFP: Days to first picking, DLP: Days to last picking, AFL: Average fruit length, AFG: Average fruit girth, AFW: Average fruit weight, PH: Plant height, NOB: Branches/plant, NON: Internodes/plant, IL: Internodal length, NOF: Fruits/plant, FY: Fruit yield/plant, FC: Fiber content X0: Mean of the original population, Xs: Mean of the selected genotypes, SD: Selection differential, h2: heritability, SG: Selection gain |
Genotype selection has so far been researched primarily on the average performance and stability of the single characteristic viz., fruit yield, whereas the selection based on above average stability and high mean performance for various traits is what a breeder needs because selection for multiple traits altogether is a daunting task. This study uses MTSI (Olivoto et al., 2019) to select okra genotypes where majority of the traits exhibit the preferred stability and positive gains. MTSI analysis allocates rank to each of the 35 okra accessions depending on desired trait value (Fig. 5). The selection pressure of ~ 15% identified the top five genotypes with lowest MTSI, which included H23 (MTSI = 4.46), H10 (MTSI = 4.66), H2 (MTSI = 4.90), H8 (MTSI = 5.12) and H12 (MTSI = 5.33). These genotypes were selected with maximum stability and high mean performance of analyzed traits through MTSI. Red circle in Fig. 7 indicated the cutoff point with MTSI value of 5.33 of H12. Genotype T2 had higher MTSI value (MTSI = 8.51) followed by H16 (MTSI = 8.05) and SC (MTSI = 7.83), these genotypes were recognized as unstable genotypes with poor performance of traits. Selected genotypes can be used to advance generations and identify suitable lines or even as parents in future breeding programs aiming superior hybrid development that performs better under various environmental conditions.
MTSI provided selection differential for all the traits. The Table 5 shows selection differential, broad-sense heritability, selection gain for mean performance and selection differential for WAASBY index of all traits. Selection differential for mean performance has range of -0.30 (Fiber content) to 7.50 (Fruit yield), while heritability ranged from 40% (Days to first picking) to 82% (Fiber content) with 70% for fruit yield. As a result, most of the yield components exhibited anticipated selection gains. The fruit yield (5.31%) and fiber content (-0.23%) recorded desired gains in positive and negative direction, respectively. The high selection gains explained that character’s variation is mainly owing to genetic makeup and hence probably to be incorporated in potential future fillial generations through breeding techniques. The selection differential for WAASBY index is found in the range of -4.97 (Fiber content) to 21.17 (Days to last picking). The positive estimates of selection differential for WAASBY index of all the traits except FC, recommended the proficiency of the technique in selecting ideal hybrids under diverse environments and the likelihood for obtaining gain with selection on all traits examined. Use of MTSI for selection of superior genotypes have also been reported by Hussain et al. (2021) in chickpea and Benakanahalli et al. (2021) in guar.
In the mean vs stability' biplot, the majority of genotypes identified by MTSI were shown to be closer to the AEC (Fig. 3). Among these selected genotypes, H23 and H10 had lowest MTSI value and they were fallen in the fourth quadrant of Y x WAASB biplot (Fig. 4), near to origin in which-won where biplot (Fig. 2) and also found near to origin in AMMI II biplot (Fig. 1). These results support use of hybrids H23 and H10 for cultivation over diverse environments. They can even be incorporated as a genitor for future breeding programmes to develop new cultivar.