Yield Stability Analysis in An Underutilized Legume ‘Winged Bean’ (Psophocarpus tetragonolobus L.)

Psophocarpus tetragonolobusis commonly known as winged bean, it belongs to legume tribe with chromosome number 2n=2x=18. It is native of Papua New Guinea and thrives well under tropical and sub tropical regions and loamy soils with adequate drainage. In India its cultivation spread in Assam, Chhattisgarh, Tripura, Meghalaya, Orissa, West Bengal and other southern States. Plant parts viz., flowers, green pods, seeds and tuberous roots are all edible and nutritious (Garcia and Palmar, 1980). The pods are rich in protein (29.8-39.0%), carbohydrate (23.9-42.0%), fat (15.0-18.0%) and vitamin A (300 to 900 IU). This multipurpose crop has not yet been fully exploited in view of lack of genotypes suited for different purposes (Mohamadali and Madalageri, 2007).
        
Three agro-climatic zones (Northern Hill, Central Plain and Bastar Plateau) prevailing in Chhattisgarh state of India and there is a large scope for cultivation of this potential underutilized legume crop. High yielding an ideal genotype should be perform well under any environment, but as expression of quantitative traits are dependent on environment, due to that most of genotypes do not perform well in all environments (Carvalho et al., 1983). An interaction of genotype and environment provide potent evaluation of genotypes towards stability and stable genotype could be used for wider cultivation. Quantitative character like yield is greatly influenced by different environmental condition; hence, superior genotypes may be selected based on yield performance at a single location in a year and it may be very effective. The AMMI (additive main effects and multiplicative interaction) and genotype and genotype × environment (GGE) biplot models can be robust tools for effective estimation and interpretation of multi environment data structure in breeding programs (Yan et al., 2000; Ebdon and Gauch, 2002; Samonte et al., 2005). AMMI and GGE biplot methods have simultaneously been used by many researchers to confirm high yielding stable genotypes in different crop species (Bhartiya et al., 2017; Mortazavian et al., 2014; Rad et al., 2013 and Yan et al., 2007). Therefore, the present investigation applied AMMI and GGE biplot techniques to study the patterns of GEI in 05 winged bean genotypes, to find high yielding and most stable genotype(s) suited to Chhattisgarh state of India.

A total of 05 winged bean genotypes (Table 1) including two checks viz., AKWB-1 (National Check) and RMDWB-1(Local Check) were grown at three locations across Chhattisgarh region of India representing three agro-climatic zones i.e., Ambikapur (E1, E4; Northern hill zone), Jagdalpur (E2, E5; Bastar plateau zone) and Raipur (E3, E6; Central plain zone) for two subsequent crop years (2015 and 2016; Table 1). The experiment was conducted under randomized block design with four replications. Each test genotype was grown in a 4 m long row length containing 6 rows with 30 × 60 cm spacing. Good crop were harvested in each test environment by adopting recommended agronomic practices. Meteorological data of two crop years were recorded over the entire environment (Table 1). Analysis of variance along with combined analysis of variance was performed for different environments. The seed yield data were subjected to AMMI and GGE biplots analysis. All statistical analyses were conducted by using PBTools version 1.4. IRRI, Philippines.
 

Significant variation among test genotype over environments was observed for seed yield, suggesting broad range of variation among genotypes. Environment and G × E mean sum of square were also highly significant for seed yield (Table 2).
 

 
AMMI biplot analysis
 
Additive mean effect and multiplicative interaction analysis for seed yield showed high significant differences among genotypes, environments and gene and environment interactions (Table 1 and Table 2). The gene× environment components was further divided and explained by two IPCA (interaction principal components axes) namely, IPCA I and IPCA II. First two IPCA axes explained more than 84.11% (PC 1= 58.3%; PC 2= 25.8%) of total variation and thus this model was effective in explaining gene× environment components and interaction in the present study (Fig 1 and 2).
 

 

        
Graphical analysis of IPCA I with average seed yield revealed that genotype G₄ ‘AKWB-1’ had the high value for yield but genotype G₂ ‘AKWB13-5’ had the highest positive AMMI1 score (Fig 1). Among environments, E₄ was most favorable for seed yield but high negative interaction with genotypes (-13.22) followed by E₅ shows low yield and negative interaction (-2.92) with test genotypes were observed. Positive interaction with test genotypes was observed for environments E₁, E₂, E₃ and E₆ even though mean value was less than E₄. In AMMI model, if genotype having high value for trait and it is greater than grand mean and near to zero IPCA score are considered under general adaptability across environments. Thus, G₄ was having general adaptability. However, genotypes with high value for trait and IPCA scores towards larger value are considered under specific adaptability to the environments. Genotypes, G₅ (RMDWB-1) and G₂ (AKWB13-5) were considered under specific adaptation due to high seed yield and large IPCA score.
        
All the six environments showed different mean for yield and because of that AMMI2 biplot does not show the additive main effects, interaction as suggested by AMMI1 biplot, but interaction component is very informative (Fig 2). This graph is useful when IPCA2 is sizeable and significant. In AMMI2 biplots, if a genotype is located near to the bioplot centre it will considered more stable than those located far from the centre. Genotype G₅ (RMDWB-1) followed by G₂ (AKWB13-5) were found stable genotypes. Most stable was environment E₂ followed by E₃ and E₂ as observed in AMMI2 score. Genotypes G2 and G4 are having positive interaction with E₃, whereas G₇ and G₅ had high positive interaction with E₂.
        
According to the IPCA I vs IPCA II scores of genotypes and environments, when a genotype is near to an environment, it indicates that the genotype is specifically adapted to that environment (Shafii et al., 1992; Kumar et al., 2016). Thus, genotypes G2 and G5 were recognized as superior and stable genotypes for environment E2 (Fig 2). In order to select appropriate environment with high ability for distinguishing genotypes, environments should have a high IPCA I and low IPCA II (Mohammadi et al., 2008). According to IPCA I, E1 and E2 environments had the most stability and the least contribution of interaction, whereas E3 and E6 with the least IPCA I had the most contribution to produce GEI. Most ideal environment was found to be E2 (based on the high IPCA I and the low IPCA II). AMMI stability parameters for environments have been used by several researchers in order to analyze GEI and found stable and compatible genotypes to such environments (Yan et al., 2000; Yan, 1999; Yan and Rajcan, 2002; Mohammadi et al., 2008).
 
GGE biplot analysis
 
Graphical virtualization for identification and evaluation of genotypes, environments and their interactions is facilitates by GGE biplot (Yan et al., 2000). Genotype × genotype  environment (GGE) biplot analysis revealed that the first two principal components PC1 and PC2 explained 94% of the total variation comprising PC1 = 77.5% and PC2 = 16.5% (Fig 3). Genotypes with have high PC1 scores and low PC2 scores were considered under ideal. Environments should be considered as ideal it has high PC1 scores and low PC2 scores (Yan and Rajcan, 2002; Yan et al., 2000). Accordingly, the genotypes G4 and G5are high yielder and G1 and G2 with large negative PC1 scores were comes under low yielder genotypes (Fig 3). Genotypes with low PC2 scores such as G3 can be considered as stable. Large PC1 scores of environment are those environments that better differentiate the genotypes and PC2 scores near zero are represent an average suitable environment (Yan, 2001; Yan et al., 2000). Projection to the y-axis (AEA line) produces measure for the stability of the genotypes. This signifies that, greater the absolute length of the projection of a genotype, the less stable it is and vice-versa (Yan, 2001).
 

        
The AEA line partitioned genotypes which yield below and above the mean yield (Fig 3). The genotypes to the right of this line are high yielders while left side is low yielders. Therefore, the genotype ranking according to this interpretation is in the order of G4, G5, G3, G2 and G1 (Fig 4). G1 is the poorest genotype for grain yield, whereas genotype ‘G4’ was identified as the ideal genotype as shown by the concentric circles around it (Fig 4). Further, genotype ‘G4’ (AKWB-1) had a projection on the y-axis that is zero and therefore it has absolute stability i.e., wider adaptation to all the test environments and it would be recommended uniformly for cultivation in all the three agro-climatic zones of the Chhattisgarh state, India. The local check genotype ‘G5’ (RMDWB-1) is also among the high yielding and relatively stable genotype. Using E3 as an ideal environment, environments in closer concentric circles e.g., E5 and E2 were considered as ideal environments while E1 and E4 were poor environments (Fig 5). Assessment of genotypes under different environment is essential to evaluate quantitative characters, to measures stability and adoptability. A complex trait like yield is highly influenced by environment. Further, to evaluate multi-environment data in effective way use of both the models are recommended (Gauch and Zobel, 1988).
 

 

The present study revealed a better understanding of the G×E interaction through AMMI model and results revealed that winged bean yield performance was significantly influenced by G×E interaction followed by genotypic (G) and environment (E) effects, respectively. GGE biplot facilitated identification of genotypes possessing stable yields as well as discriminating environments and specificity in adaptability of the genotypes to specific environments in a ‘which won where’ pattern. As per the GGE biplot the check ‘G4’ (AKWB-1) was identified as ideal cultivar as it had longest vector length among high yielding genotypes as well as high stability. ‘G4’ i.e., AKWB-1 would be recommended for commercial cultivation in all the three prevailing agro-climatic zones of the Chhattisgarh state, India. Further, other high yielding and stable genotypes i.e., G5 (RMDWB-1) and G3 (Ambika 11-3) can also be used as parents for winged bean improvement program.

Authors are thankful to Director Research Services, Indira Gandhi Krishi Vishwavidyalaya (IGKV), Raipur, India for financial support and allocation of yield evaluation trials at different research centers of Chhattisgarh state, India.