The present investigation was carried out in Research Area of Wheat and Barley Section, Department of Genetics and Plant Breeding, CCS Haryana Agricultural University. The experiment was conducted during rabi 2015-16. For this experiment randomly selected seven durum genotypes namely; WHD 896, WHD 912, WHD 943, WHD 948, PDW 291, PDW 233, PDW 314 were used.

The genotypes were evaluated under four environmental conditions i.e. timely sown irrigated condition (E1), late sown irrigated condition (E2), timely sown rainfed condition (E3), and late sown rainfed condition (E4). After restricting these conditions other recommended packages and practices were followed throughout the cropping season of wheat. The experiment was conducted in Randomized Block Design (RBD) with three replications.

In the present experiment data was collected from randomly selected five plants from each genotype. Plants were tagged and harvested separately at the time of maturity. Threshing of these plants was done and grain yield of each plant was weighted. The data was then analyzed for the stability of genotypes and grouping of environments based on GGE biplot analysis given by Yan et al. (2000). For the data analysis, R software was used. GGE biplot was generated using statistical package ‘GGEbiplotGUI’. This package can be downloaded in R by using the command ‘install.package(GGEbiplotGUI) and can be loaded for analysis by the command ‘library (GGEbiplotGUI)’. Grain yield of seven genotypes from the above four environments was saved in a comma separated file (.csv) and imported into R studio. After this biplot was generated by a simple command ‘GGEbiplot(data_file)’. Different types of biplots can be generated by using the option ‘biplot analysis’.

Grain yield in wheat is a complex trait as it has both genotypic and environmental origin (Farshadfar et al. 2013). In the present study it was found that yield of seven durum wheat genotypes varied across environments which proved that environment influenced this trait (Table 1).

E1: Timely sown irrigated environment; E2: Late sown irrigated environment; E3: Timely sown rainfed environment; E4: Late sown rainfed environment.

Pooled analysis of yield data for seven durum wheat genotypes from four environments was carried out to partition the total variability into its component parts. Analysis of variance (ANOVA) showed that the studied genotypes were significantly different from each other for grain yield (Table 2). Further highly significant variations were also noticed for environment and genotype × environment interaction (GE). Only 28.10% of variability is caused by genotype itself. It is clear from the table that about 50.48% of variability was caused only by environment itself whereas its interaction with genotype contributed for about 7.38% of variability. It is clear from earlier studies that durum wheat has enough genetic variability for grain yield (Rathwa et al. 2018). Grain yield in wheat is affected by significant influence of environment (Kaya and Akcura 2014). Earlier researchers have also noticed significant interaction of genotypes with environment Solonechnyi et al. (2015). The significant interaction of genotype and environment is the major cause of change in ranking of genotypes under different environments (Matus-Cadiz 2003). This interaction also creates problems while selecting a genotype on the basis of grain yield (Crespo et al. 2017).

**Significant at p>0.01, DF: Degree of freedom; SS: Sum of squares.

Environmental effect can make the evaluation more confusing. In the present case about 50% of variability was caused due to environment. Under such situation the data was further analyzed for genotype and genotype plus environment interaction (GGE) using the standard procedure of GGE biplot analysis suggested by Yan et al. (2000). In GGE biplot analysis the irrelevant environmental variance is separated from total variance and evaluation of genotypes are carried out only using two parameters viz. genotype and GEI (Yan and Kang 2003). Using this method two major principle components; PC1 and PC2 were calculated which accounted for 71.30% and 17.34% of variability respectively shown on x and y axis of GGE biplot (Fig. 1). The first two principle components accounted for a total of 88.64% of genetic variability for grain yield among these genotypes caused by G+GE interaction. In a similar study first two PCs were used to represent 88.50% of variability caused by GEI (Solonechnyi et al. 2015). When first two PCs can measure maximum variability exploration of other PCs become wasteful (Verma et al. 2020).

In our present GGE biplot, green color was used to indicate positions of genotypes on biplot whereas environments were represented by blue color. To get the maximum information from the biplot, black dotted lines were used to represnt origin of biplot and blue dotted lines as environment vectors i.e. line joining the environments to the origin of biplot. The environment vectors were formed by calculating standard deviations of genotypes in a particular environment. Thus, the length of environmental vectors can be utilized in identification of a suitable environment. In the present study, environment E3 had maximum length of environment vector (Fig. 2). Grouping of environments based on GGE biplot is more easy and valuable (Yan 2000).

Thus, environment E3 had more discriminating power as genotypes perform differentially in this environment. All other environments had similar environment vectors and thus had same discriminating powers. The other information we get from environment vectors is the correlation between the environments. This correlation is a measure of cosine angular distance between two environment vectors. Larger the angle, lesser will be correlation between the environments and vice versa. In the present case, environment E1 and E2 were highly correlated while others with high angles were less correlated. Thus instead of E1 and E2 we can use either of these environment for future studies as both environments are alike and can give similar results. Another feature of GGE biplot is that environments can be grouped based on average or ideal environment. In Fig. 2, the bold and arrowed blue line was used to represent average environment. This line is called average environment coordinate (AEC) and the angle between this line and a test environment is used to check the reliability power of that test environment (Braun et al. 1996). In present case, environment E4 and E1 had almost equal angle with AEC and had higher discriminating power than other environments. Thus out of these four environments only three environments i.e. E1, E2 and E3 are best and one can drop the fourth environment for genotypic assessment.

The genotypes WHD 948 was best for three environments viz. E1, E2 and E3 whereas WHD 896 was most suitable for E4 (Fig. 3a). Variability in a genotype and its yield stability can be accessed by locating the genotype on AEC. From Fig. 3, it is clear that genotype WHD 948 is more stable and higher yielder than other genotypes. Genotypes WHD 943, WHD 912 and PDW 233 were situated away from ATC lines and thus varied from one environment to other. In other way we can say that performance of these three genotypes was changing across different environments (Fig. 3b). As GGE biplot can classify the genotypes according to their stability this technique is very useful for plant breeders (Hagos and Abay 2013; Sagar et al. 2014). GGE biplot has been utilized by Solonechnyi et al. (2015) to identify stable genotypes.