Genotype by Environment (G × E) Interaction and Yield Stability of Chickpea (Cicer arietinum L.) Varieties Across Agroecological Regions of Ethiopia

Genotype by environment (G × E) interaction obstructs breeding by persuading variations in genotype performance. The aim of the present study was to determine the stability and yield performance of Desi and Kabuli chickpea varieties at different agroecological regions of Ethiopia, using different stability parameters. The experiment was laid out in a randomized complete block design (RCBD) with three replications. The additive main effect and multiplicative interaction (AMMI) analysis of variance (ANOVA) indicated highly significant differences (p ≤ 0.01) for environments, genotypes, and importantly G × E interaction. AMMI and GGE biplot, AMMI's stability value (ASV) indicate that the Desi chickpea variety Teketay with mean yield of 2225.6 kg/ha (highest) and the variety Dimtu (1603.9 kg/ha) followed by Natoli with mean yield of 2004.9 kg/ha were found to be stable and adaptable to all environments. Similarly, from the Kabuli chickpea varieties, the variety Koka with mean grain yield of 2257.1 kg/ha (highest) and the variety Ejere with mean yield of 1997.6 kg/ha followed by Shasho (1798.59 kg/ha) were found to be stable and adaptable to all environments and should be promoted for production in chickpea‐growing areas of Ethiopia. In conclusion, identification of stable improved varieties for the different agroecological regions can assist the producers such as the farmers for the effective chickpea production. This leads to sustainable self‐sufficiency of food at the household and country level.

Ethiopian chickpea production shares 90% of the total production in sub-Saharan Africa and 3% of the world's production, and this makes Ethiopia the sixth among the world's major chickpea producers (Getahun et al. 2021).
Despite the crop being highly important, it is more sensitive to abiotic and biotic stresses particularly for the environmental variations such as edaphic and climatic factor in different areas of its production that leads to yield fluctuation (Biru et al. 2017).The presence of the genotype × environment (G × E) interaction contributes to the irregularity of crop yield over a wide range of environments.The occurrence of large G × E interaction during the multitrials makes the selection of superior genotypes difficult and affects the progress of the selection.In the absence of G × E interaction, the superior genotype in one environment may be regarded as the superior genotype in all, whereas in the presence of the G × E interaction, it confirms that particular genotypes are being superior in a particular environment (Beksisa 2021).Hence, the development of stable genotypes with improved yield has to be considered as one of the alternatives to mitigate the effects of G × E interaction and making the recommendation of genotypes with such attributes might be more reliable (Biru et al. 2017).According to Funga et al. (2017), evaluating genotypes of annual crops such as chickpea for yield and yield-related traits on a multilocation and multiyear trail often shows G × E interaction.In this context, developing genotypes that cope with G × E interaction by evaluating genotypes across year and location will be a crucial step in any breeding program.Therefore, this work was done with the objective of studying the suitability of chickpea varieties across the agroecological regions of Ethiopia.

| Description of Experimental Sites
The experiment was conducted at Adet, Debre Zeit, and Fogera agricultural research institutions (

| Experimental Materials
Twenty released chickpea (8 Desi and 12 Kabuli) varieties were used in the trial (Table 2).

| Experimental Design and Trial
The experiment was laid out in a randomized complete block design (RCBD) with three replications.That has been made all together in 20 experimental units.In both types of chickpea, the size of each plot was 1.5 m × 2 m.The distance between adjacent plots and replications was kept at 0.5 and 1 m, respectively.Each treatment was assigned randomly to the experimental plots within a replication.Each experimental unit consists of four rows of 2 m in length with spacing of 40 cm between rows and 10 cm between plants, so that 80 plants per plot were set.The total area was about 312 m 2 (for Desi, 8 × 15.5 = 124 m 2 , and for Kabuli, 12 × 23.5 = 188 m 2 ).Data were collected from the two central rows for the traits measured at plot bases, whereas for traits measured in plant bases, five randomly sampled plants were used.

| Phenological Parameters
Days to 50% flowering (DF): It was recorded as the number of days from planting to 50% of the plants producing flowers per plot through visual observation.
Days to 90% physiological maturity (DPM): It was recorded as the number of days from planting to when 90% of the plants shows yellowing pods and leaves through visual observation.

| Growth Parameters
Seedling stand count (STC): The total numbers of seedlings per plot were recorded in an early stage of plant growth.
Number of primary branches (NPB): It was determined from mean of five randomly selected plants with in the net plot area.
Number of secondary branches (NSB): It was determined from mean of secondary branches from the primary branches by selecting five plants randomly within the net plot area.
Plant height (PH): The average height in centimeters of five randomly selected plants in the net plot area was recorded from the soil surface to the top of the canopy of the plant.
Plant count at harvest (PCH): This represents the total number of plants in the net plot area at harvest.

| Yield and Yield-Related Traits
Number of pods per plant (NPP): The number of pods in a plant was determined mean of five randomly selected plants within the net plot area.
Number of seeds per pod (NSP): The NSP was determined mean of five pods randomly picked from five plants, which were previously tagged.
Biomass yield (BM): BM was recorded from a net plot area using the total above-ground biomass at harvest maturity.
Grain yield (GY): It was determined after harvesting all plants from each net plot area, threshing it and separating the straw from the grain and weigh in kilogram after adjusting the grain moisture content into 11% and it was change to per hectare.
Harvest index (HI): It was determined by dividing weight of actual yield (GY) for total biomass weight and multiply by hundred.

| Data Analysis
Shapiro's and Bartlett's tests were used to test the normality and homogeneity of variances of the data in that order.Yield stability of varieties was determined using different stability parameters as well as additive main effect and multiplicative interaction (AMMI) and genotype and G × E interaction biplot (GGE biplot) analyses were done using R statistical software version 4.2.0 (Team 2018).

| Stability Parameters
2.5.1.1 | Superiority Index (Pi).Lin, Binns, and Lefkovitch (1986) anticipated a superiority measure (Pi), which is defined as the distance mean square between the genotype response and the maximum response.They described that cultivar superiority measure involves calculations (across environments) of the mean square difference between the performance of a genotype and the best genotype within a given environment.It measures mean performance and stability simultaneously.Wricke (1962) describes the concept of ecovalence, to describe the stability of a genotype, as the contribution of each genotype to the G × E sum of squares.The ecovalence (W 2 ) or the stability of the ith genotype is its interaction with environments, squared and summed across environments.
The genotypes with the lower amount of W 2 could be considered as stable, while the genotype with the higher value of W 2 considered as unstable.The W 2 does not account for genotype performance.
where X ij is the GY of the ith genotype in the jth environment; X i is the mean GY of the ith genotype; X j is the mean GY of the jth environment; and X is the grand mean.).This statistic is describe as the predictability of response suggested by Pinthus (1973) as another stability parameter, in which a variation of mean yield was explained by genotype response across environments.This parametric stability statistic can be described with the following equation:

2.
where bi is slope of regression, Xij is the GY of the ith genotype in the jth environment, Xi. is the mean GY of the ith genotype, X.j is the mean GY of the jth environment, and X. is the grand mean.A genotype with the highest value is intended to be more stable.

| AMMI Analysis
AMMI analysis of variance summarizes most of the magnitude of G × E interaction into one or few interaction principal component analysis (IPCA).AMMI model was computed using the following formula suggested by Gauch (1992): where ε ij is the random error; ρij is the residual for the multiplicative components, and tjk is the jth element of the kth eigenvector; aik is the ith element of the kth eigenvector; λk is the singular value for the kth IPCA; τj is the deviation of the jth environment from the grand mean; αi is the deviation of the ith genotype from the grand mean; μ is the grand mean; Y ij is the value of the ith genotype in the jth environment.

| GGE Biplot Analysis
GGE biplot enables visual observation to understand which varieties are best performed in which environment/which varieties are stable and unstable.Moreover, GGE biplot is also used to visualize the discriminating ability and representativeness of the test environment (Yan WeiKai and Hunt 2002).
where ε ij = the error associated for the genotype i and environment j, hj 2 = eigen vectors of environment j for IPCA2; λ2 = the singular value for principal component 2 (IPCA2); σi 2 = eigen vectors of genotype I for IPCA2, hj1 = eigen vectors of environment j for IPCA1; σi 1 = eigen vectors of genotype i for IPCA1; λ1 = the singular value for principal component 1 (IPCA1); ej = the mean effect of the jth environment; μ = the grand mean; Yijr = the performance of the ith genotype in the jth environment and rth replication.
ASV as described by Purchase, Hatting, and Van Deventer (2000) was used to further investigate the stability of the genotypes.
Generally, the AMMI model does not make provision for a quantitative stability measure, such a measure is essential to quantify and rank genotypes according to their yield stability.In effect, the ASV is the distance from zero in a two-dimensional scatter of IPCA1 scores against IPCA2 score.Since the IPCA1 score contributes more to the G × E sum of squares, it has to be weighted by the proportional difference between IPCA2 score to compensate for the relative contribution of IPCA1 and IPCA2 to the total G × E sum of squares.
where ASV = AMMI's stability value, IPCA1 = interaction principal component analysis 1 scores for the specific genotype; IPCA2 = interaction principal component analysis 2 scores for the specific genotype; sum of square of the interaction principal component 1 and sum of square of the interaction principal component 2.

| Yield Stability Index (YSI)
The YSI was computed using Mahmodi, Yaghotipoor, and Farshadfar's (2011) formula as follows: where RASV = AMMI's stability value and RY = the mean yield rank of genotypes across environments.

| AMMI Analysis on Chickpea Varieties
In the Desi chickpea varieties, the total variation explained was 74.1% for environment, 13.4% for genotype, and 10.86% for G × E. Similarly, in Kabul chickpea varieties, the total variation explained was 77.8% for environment, 8.9% for genotype, and 12.9% for G × E (Table 3).Besides, to this, the three principal component analyses were significant (p ≤ 0.05) and highly significant (p ≤ 0.01).As a result, in the Desi-type chickpea varieties, 100% of the interaction sum of squares was cumulatively explained by the three principal components where the IPCA1, IPCA2, and IPCA3 explained the 50.6%, 34.0%, and 9.0% of the variations, respectively.In the Kabuli chickpea varieties, three of the principal component analyses were highly significant (p ≤ 0.01), and 100% of the interaction sum of squares was explained by IPCA1, IPCA2, and IPCA3 with 44.4%, 27.7%, and 19.0% of the G × E, respectively (Table 4).

| AMMI1 Biplot for Additive and Interaction Effects
To distinguish the G × E effect's contribution, an AMMI1 biplot was plotted using the environment and genotype mean yields versus their IPCA1 scores (Figure 1).The x-coordinate of the AMMI1 biplot indicates the average of the environments and genotypes, while the y-coordinate represents the IPC1.
The genotypes on the right side of the x-axis are high yielding (above the mean yielding), and the environments are favorable, while those on the left side are low-yielding genotypes and the environments are unfavorable (Yan and Tinker 2005).Therefore, in the Desi chickpea varieties, Adet 2022/23, Debre Zeit 2022/23, and Debre Zeit 2021/22 were classified as highyielding and favorable environments on the right side of the abscissa, whereas Adet 2021/22, Fogera 2021/22, and Fogera 2022/23 were classified as low-yielding and unfavorable environments.Likewise, the high-yielding Desi-type chickpea varieties that produce above average means were Teketay, Geletu, Natoli, Eshete, and Dalota, whereas the low-yielding genotypes were Mastewal and Fetenech suited in the left side of the x-coordinate (Figure 1).So, the varieties such as Teketay, Natoli, and Dalota were discovered to be relatively more stable and above average yielders; likewise, Fetenech was a stable and below average yielder, whereas Eshete and Geletu were varieties that are more unstable.On the other hand, environments  Likewise, the high-yielding genotypes that produce above-average means were Hora, Habru, Ejere, Dhera, Koka, and Shasho suited in the right side of the abscissa, whereas the low-yielding genotypes were Yelbie, Chefe, Kasech, DZ-10-4, and Akuri suited in the left side of the x-coordinate (Figure 1).The higher the IPCA scores from the origin (either positive or negative), the more specific a genotype is to certain environments with the highest contribution to the G × E, or the more the IPCA scores approach zero, the more stable the genotype is over all the environments sampled.Accordingly, Hora, Ejere, Dhera, and Habru were discovered to be relatively more stable and above average yielders; likewise, Arerti was a stable and below average yielder, whereas Akuri and DZ-10-4 were genotypes that are more unstable.

| AMMI2 Biplot for Demonstrating the Magnitude of G × E
The biplot was generated to illustrate environment and genotype effects simultaneously with the first two IPCAs (IPCA1 and IPCA2) (Figure 2).Therefore, in the Desi-type chickpea, the varieties Fetenech and Dimtu were located close to the origin, proved highly stable, and expressed a low G × E interaction (positive or negative), whereas the genotypes Geletu, Eshete, and Dalota located far away from the origin were highly unstable and expressed a higher G × E interaction (positive or negative) (Figure 2).Environments with short spokes (length of arrow lines) do not exert strong interactive forces, whereas those with long spokes exert strong interaction.Thus, Adet 2021/22, Fogera 2022/23, and Debre Zeit 2021/22 had long spokes (located far away from the biplot origin) and paid most to the G × E interactions, whereas Adet 2022/23, Fogera 2021/22, and Debre Zeit 2022/23 were located close to the origin and contributed most to the phenotypic stability of the genotypes (Figure 2).In the Kabuli-type chickpea, the varieties Yelbie, Shasho, Ejere, and Arerti were located close to the origin, proved highly stable, and expressed a low G × E interaction (positive or negative), whereas the genotypes Koka, Chefe, Dhera, Hora, and Akuri located far away from the origin were medium to highly unstable and expressed a higher G × E interaction (positive or negative) (Figure 2).

| Selection of Grand Varieties Using Mean GY and Stability Parameters
As indicated in   selection, the varieties such as Teketay (2225.6 kg/ha) followed by Geletu (2009.8kg/ha) and Natoli (2004.9kg/ha) expressed the highest mean GY across environments.Based on ASV stability parameter, Natoli, Dimtu, and Fetenech varieties with lower values of ASV were considered as stable.
The varieties Dimtu followed by Natoli and Fetenech were selected as the best materials due to their lowest values of Shukla's stability variance (σ 2 i).Additionally, Eberhart and Russel's joint regression model determined the stability of each variety through the slope of regression line (bi) and variance in regression deviation (s 2 di) stability parameter, and the varieties Fetenech, Mastewal, Dimtu, and Geletu had bi ≈ 1, of which Fetenech and Mastewal had lower GY than the average yield performance, which, therefore, were poorly adapted to all test environments.Therefore, the varieties Teketay, Natoli, Eshete, and Dalota showed bi > 1 and low rates of average stability; hence, they are suitable for high-yielding environments.The varieties Dimtu, Fetenech, Geletu, and Mastewal had bi < 1 of which Dimtu and Geletu produced higher GY compared to the average yield performance, and therefore, they were specificity adapted to low-yielding environments.Wricke's ecovalence (Wi 2 ) stability parameter also recognized that the varieties such as Dimtu, Natoli, and Fetenech are the most stable, while the WAASB stability parameter was used to better characterize ideal varieties based on both mean GY and stability so that the varieties Dimtu followed by Natoli and Dalota were stable relative to other varieties.Moreover, based on Pinhus's coefficients of determination (R 2 ), Natoli, Dimtu, and Fetenech were three top ranked stable varieties compared to other varieties.The best ranked varieties based on the ASV were Natoli, Dimtu, and Fetenech among the Desi-type chickpea varieties.In the context of the Kabuli-type chickpea varieties, mean GY has to be taken as the first criteria for variety selection; the varieties such as Koka (2257.1 kg/ ha) followed by Hora (2143.9kg/ha) and Dhera (2063.5 kg/ha) expressed the highest mean GY across environments.Similar stability trends were identified in Lin and Binns (Pi) and YSI stability parameters with mean GY.Using ASV stability parameter, varieties such as Yelbie, Ejere, and Arerti varieties with lower values of ASV were considered as stable.The varieties Dimtu followed by Ejere, Shasho, and Arerti were selected as the best materials due to their lowest values of Shukla's stability variance (σ 2 i).What is more, Eberhart and Russel's joint regression model determined the stability of each variety through the slope of regression line (bi) and variance in deviation from regression (s 2 di) stability parameter.
According to the model, the varieties Shasho, Koka, Habru, Ejere, and Geletu had bi ≈ 1, of which Shasho and Ejere had lower GY than the average yield performance; therefore, they were poorly adapted to all test environments.The varieties Hora, Kasech, DZ-10-4, Chefe, and Arerti showed bi > 1 and low rates of average stability; hence, they might be suitable for high-yielding environments.The varieties Yelbie, Shasho, Koka, Habru, Ejere, Dhera, and Akuri had bi < 1, of which Koka, Dhera, Ejere, and Habru produced higher GY as compared to the average yield performance, and therefore, they were specifically adapted to low-yielding environments.Wricke's ecovalence (Wi 2 ) stability parameter also recognized that the varieties such as Ejere, Shasho, Arerti, and Habru are the most stable, while the WAASB parameter was used to better characterize ideal varieties based on both mean GY and stability so that the varieties Ejere followed by Shasho, Arerti, and Yelbie were stable relative to other varieties.Moreover, based on Pinhus's coefficients of determination (R 2 ), Arerti, Kasech, and Ejere were three top ranked stable varieties compared to other varieties.The best ranked varieties based on the ASV were Yelbie, Ejere, and Arerti (Table 5).

| Correlation Between Stability Parameters
Table 6 shows each of the possible pair wise comparisons of the ranks of different stability parameters, which were determined by the Spearmen's rank correlation (Piepho 2012).Among the Desi-type chickpea varieties, mean GY was highly significant (p ≤ 0.01) and positively correlated with Lin and Binns's superiority index (Pi) but nonsignificant (p ≥ 0.05) and negatively correlated with all other parameters.All stability parameters showed a highly significant (p ≤ 0.01) and a positive correlation with Shukla's stability variance (σ 2 i) while nonsignificant (p ≥ 0.05) and negatively correlated with Lin and Binns's superiority index (Pi).Also, Wricke's ecovalence (Wi 2 ) was highly significant (p ≤ 0.01) and positively correlated with deviation from regression (s 2 di), ASV, and WAASB from the singular value decomposition of the matrix of BLUP but nonsignificant (p ≥ 0.05) and negatively correlated with Lin and Binns's superiority index (Pi).Furthermore, deviation from regression (s 2 di) was highly significant (p ≤ 0.01) and positively correlated with ASV and WAASB.However, it was nonsignificant (p ≥ 0.05) and negatively correlated with Lin and Binns's superiority index (Pi).On the same way, high significance and a positive correlation were observed between ASV and WAASB but nonsignificance (p ≥ 0.05) and a negative correlation with Lin and Binns's superiority index (Pi); besides that, coefficient of determination (R 2 ) was significant (p ≤ 0.05) and positively correlated with ASV and WAASB and highly significant and positively correlated with Shukla's stability variance (σ 2 i), deviation from regeration (s 2 di), and Wricke's ecovalence (Wi 2 ) but nonsignificant (p ≥ 0.05) and negatively correlated with mean GY and Lin and Binns's superiority index (Pi).In the context of the Kabuli-type chickpea varieties, mean GY was highly significant (p ≤ 0.01) and positively correlated with Lin and Binns's superiority index (Pi) but nonsignificant (p ≥ 0.05) and negatively correlated with all other parameters.All stability parameters showed high significance (p ≤ 0.01) and a positive correlation with Shukla's stability variance (σ 2 i), while nonsignificance (p ≥ 0.05) and a negative correlation were observed with Lin and Binns's superiority index (Pi).Wricke's ecovalence (Wi 2 ) showed high significance (p ≤ 0.01) and a positive correlation with deviation from regression (s 2 di), ASV, and WAASB from the singular value decomposition of the matrix of BLUP, but nonsignificance (p ≥ 0.05) and a negative correlation were observed with Lin and Binns's superiority index (Pi).Furthermore, deviation from regression (s 2 di) was highly significant (p ≤ 0.01) and positively correlated with ASV and WAASB; however; it was nonsignificant (p ≥ 0.05) and negatively correlated with Lin and Binns's superiority index (Pi).Similarly, high significance and a positive correlation were observed between ASV and WAASB but nonsignificance (p ≥ 0.05) and a negative correlation with Lin and Binns's superiority index (Pi).
In addition, coefficient of determination (R 2 ) was significant (p ≤ 0.05) and positively correlated with ASV and WAASB and highly significant and positively correlated with Shukla's stability variance (σ 2 i), deviation from regression (s 2 di), and Wricke's ecovalence (Wi 2 ) but nonsignificant (p ≥ 0.05) and negatively correlated with mean GY and Lin and Binns's superiority index (Pi).

| AMMI Analysis
AMMI model is primarily effective where the assumption of linearity of reactions of varieties to a variation in environment is not fully explained, which is important in stability analysis (Muricho et al. 2023).The high percentage of the variation explained by environment (ENV) implies that the environments were dissimilar, resulting in large differences among environments causing most of the variation in the GY.This is a very important clue for understanding environmental influence in which it is a major factor on yield performance of chickpea varieties in Ethiopia.Mekonnen et al. (2022) and Ojiewo et al. ( 2019) reported similar results.On the other hand, the response of genotypes varied considerably for GY due to the genetic makeup of the materials and the interaction between genetic constitution and environmental influences.This is in agreement with Gerrano et al. ( 2020) who reported that genotype and environment directly affect the yield potential of cowpea.Besides that, the higher yield variation contributed by the environment over genotype and interaction indicates that the test environments were highly variable and had a great impact on genotypes.So, the significant G × E necessitates the need to identify adaptable genotypes with consistent high GY (Mekonnen et al. 2022).

| AMMI1 and AMMI2 Biplot for Additive, Interaction, and Proving the Magnitude of G × E Effects
Identifying favorable environments for chickpea genotypes, environments that appear almost in a perpendicular line have similar means, and those that fall almost in a horizontal line have similar interaction pattern (Mulugeta and Girma 2014).Additionally, AMMI2 helps in the visual interpretation of the G × E pattern and identifies environments/genotypes that exhibit a high, medium/low interaction effects (Hailemariam and Tesfaye 2019).Points near the origin have small interaction effects, and points near each other have similar interaction effects (Gauch 1992).Likewise, according to Purchase, Hatting, and Van Deventer (2000), the genotypes that are positioned far away from the center are more responsive or unstable, while genotypes and environments that are closer to the center of the biplot have higher stability performance.
The higher the IPCA scores from the origin either positive/negative, the more specific a genotype is to certain environments with the highest contribution to the G × E/the more the IPCA Note: Numbers in parentheses show the ranking pattern of varieties based on each stability parameter.Abbreviations: ASV = AMMI's stability value, bi = regression coefficient, ns = nonsignficant at 0.05 and 0.01, Pi = Lin and Binns's superiority index, R 2 = coefficient of determination, s 2 di = deviation from regression, WAASB = weighted average of absolute scores from the singular value decomposition of the matrix of BLUP for GEI effects generated by a LMM, Wi 2 = Wricke's ecovalence, YSI = yield stability index, σ 2 i = Shukla's stability variance.*Significant at the 0.1 level, **Significant at the 0.05 level, and ***Significant at the 0.01 level.
scores approach zero, the more stable the genotype is over all the environments sampled (Gebreselassie et al. 2024).So, the varieties such as Teketay, Natoli, and Dalota were discovered to be relatively more stable and above average yielders likewise, Fetenech was a stable and below average yielder, whereas Eshete and Geletu were varieties that are more unstable.According to Mekonnen et al. (2022), the positive and negative IPCA scores of genotypes in AMMI analysis are the best indicators of stability/adaptation over environments.High positive interaction of the genotypes in an environment can exploit the agroecological conditions of the specific environment.Thus, it would be possible to identify adaptable and suitable genotype/s for the specific environment.A genotype performing high positive interaction in an environment clearly has the capacity to exploit the agroecological/agromanagement situations of the specific environment and therefore best suited to that environment (Mulugeta and Girma 2014).

| ASV for Measuring Quantitative Stability
ASV helps for the selection of comparatively stable and highyielding varieties.Thus, an ideal variety should have high mean GY and small ASV.Oladosu et al. (2017) reported that the higher the IPCA score, and ASV, the more specifically adapted a genotype is to a certain environment.On the other hand, a variety with the least ASV score is considered as the most stable; therefore, from the Desi-type chickpea varieties, Dimtu was the most stable followed by Natoli.Similarly, from the Kabuli-type chickpea varieties, the variety Yelbie was the most stable followed by Ejere and Arerti, rather than Dimtu and Yelbie.The most stable variety Dimtu was found to be the sixth interims on GY performance out of the eight Desi chickpea varieties, and Yelbie is the least in yield performance among the Kabuli chickpea varieties (Table 5).It shows that stability alone cannot be the basis for screening and selection of varieties for the higher GY since some varieties are stable for poor yields across environments and choosing them would lead to development of a variety, which is constantly low yielding (Ojiewo et al. 2019).Mekonnen et al. (2022) also added that both yield and stability performance of varieties should be considered simultaneously to exploit the useful effect of G × E and to recommend varieties for wide adaptations.the sectors in the polygon had no test environment and the best performing Kabuli chickpea variety without relaying in the test environment was Chefe (Figure 3).

| Discriminativeness and Representativeness
The circle indicated by the arrow represents the average environment.If the angle formed between the test environment and the line passing through the average environment is small, it means that this test environment is representative, and the larger the vector for each environment, the greater the discrimination ability (Yan and Tinker 2005) 4.

| Mean and Stability of Genotypes
The average environment coordination (AEC) method was used to evaluate the stability and average yield performance of the 20 chickpea varieties (8 Desi and 12 Kabuli) (Figure 5).The stability axis is the axis drawn as a double-headed arrow axis that passes at right angles to the AEC through the biplot origin.The genotype positioned further away from the AEC in the direction of either arrow is considered as less stable across environments, while the genotype nearest to the AEC in the direction of either arrow is considered as stable (Gebreselassie et al. 2024).Thus, in the Desi chickpea, varieties positioned nearest to the stability axis/relatively stable varieties were Natoli, Teketay, Dalota, Dimtu, Mastewal, and Fetenech whereas further varieties (unstable genotypes) were Eshete and Geletu (Figure 5).Similarly, in the Kabuli chickpea varieties, Ejere, Shasho, Dhera, Arerti, Chefe, and Yelbie were considered as relatively stable varieties while Koka, Hora, Akuri, Habru, Kasech, Akuri, and DZ-10-4 were considered as unstable (Figure 5).

| Ranking of Environments With the Ideal Environment
An ideal environment is representative and has the highest discriminating power (Yan and Tinker 2006).The most suitable environment is the one closest to the ideal environment, which is located in the first concentric circle of the environment focused GGE biplot (Gebreselassie et al. 2024).Thus, in the Desi chickpea among the environments, Debre Zeit 2022/23 and Adet 2022/23 were close to the ideal environment, and these environments were the most representative of the overall environments and the most powerful in discriminating genotypes; they were therefore identified as more desirable environments than the others were.Furthermore, Fogera 2021/22 and Adet 2021/22 were closer to the ideal environment and considered as the second most powerful to discriminate genotypes.Conversely, environments like Debre Ziet 2021/22 and Fogera 2022/22 were found to be less effective in differentiating between varieties, whereas in the Kabili chickpea varieties, environments like Adet 2021/22 and Fogera 2022/23 were found to be near the ideal environment.These environments were also found to be the most representative of all environments and had the greatest influence on differentiating between genotypes, making them more desirable than the others.Debre Zeit 2021/22 and Fogera 2021/22 were closer to the ideal environment and considered as the second most important to discriminate genotypes.On the other hand, environments Adet 2022/23 and Debre Zeit 2022/23 were far from the ideal environment and considered as less powerful to discriminate varieties (Figure 6).

| Ranking of Genotypes With the Ideal Genotype
According to Gebreselassie et al. ( 2024), the perpendicular line that passes through the origin to the AEC with double arrows represents the stability of varieties.The genotypes with the highest stability and the highest mean GY are the best candidates for selection.In the biplot, they are close to the origin and have a shorter vector from the AEC.Regardless of the direction, a longer projection to the AEC indicates a genotype's inclination toward a higher G × E, which implies less environmental stability.The ideal genotype, which is situated at the center of the concentric circles, can be used as a benchmark for selection.If a genotype is nearby to the ideal genotype, it is more desirable.Thus, to show the difference between each genotype and the ideal genotype, concentric circles were drawn around it.Genotypes located nearest to the "ideal genotype" are more desirable than those located farther away are. Figure 7 indicates the ranking of 20 chickpea varieties (8 Desi and 12 Kabuli) relative to the ideal genotypes using the GGE biplot model across six environments.The AEC of the GGE biplot view is used to rank varieties with the ideal variety.In addition, the AEC ordinate differentiates genotypes with lower than average means from those with higher than average means.More data closer to the concentric circles indicate a higher mean yield.As a result, the Desi chickpea varieties with the highest yielders were Teketay, Geletu, Natoli, Dalota, Eshete, Dimtu, and Mastewal while in the Kabuli chickpea, the highest yielders were Dhera, Koka, Hora, Habru, Ejere, Shasho, Arerti, Kasech, and Chefe.On the other hand, the variety Fetenech from the Desi type and the varieties such as Akuri, DZ-10-4, and Yelbie from the Kabuli type are out of the concentric circle and could be rejected in early breeding cycles (Figure 7).

| Relationship Among Environments
Interrelationships among the six test environments are shown in Figure 8. Lines connecting the biplot origin and environments markers are environment vectors while the angle between the vectors of two environments provides the correlation coefficient between them; the cosine of the angle between the vectors of two environments estimates the correlation coefficient between them (Otieno and Owuor 2019).Similarly, Yan and Tinker (2006) added that, when two angles between two environments vectors are less than 90°, then, these two environments are positively correlated, whereas they are independent if the angle is 90° and negatively correlated if the angle is greater than 90°.In view of that, in the Desi chickpea varieties, Debre Zeit 2021/22, Debre Zeit 2022/23, and Fogera 2021/22 environments were positively correlated to each other since all of the angles among their vectors were smaller than

| Conclusion
To maintain long-term food security in Ethiopia, improved Desi and Kabuli chickpea cultivars must be adjusted to a wide range of environmental conditions.Farmers are more interested in improved varieties, which produce consistent yield under their growing conditions.Likewise, chickpea breeders are intended to develop promising varieties, which are consistently high yielding with relatively broad adaptation and tolerant/resistant to both the biotic and abiotic challenges.Hence, plentiful evidence on G × E interaction and stability are principal importance for chickpea researchers and breeders and farmers engaged in chickpea improvement and production.
For that reason, multienvironment evaluation of chickpea varieties across locations and years/seasons were initiated to study the magnitude of G × E interaction and determine the GY performance and stability of the chickpea varieties.In this study, AMMI, GGE models, and other univariate and multivariate stability parameters were used to quantify the magnitude of genotype, environment, and G × E interaction to determine extent of stability and adaptability of varieties.Within this background, AMMI analysis of variance in this study point out that chickpea GY performances were highly influenced by environmental effects followed by G × E interaction and genotypes.The Desi chickpea varieties such as Teketay with mean GY of 2225.6 kg/ha (highest) and the variety Dimtu (1603.9kg/ha) followed by Natoli with mean GY of 2004.9 kg/ ha were found to be stable and adaptable to all environments.Similarly, from the Kabuli varieties, Koka with mean GY of 2257.1 kg/ha (highest) and variety Ejere with mean GY of 1997.6 kg/ha followed by Shasho (1798.59kg/ha) were found to be stable and adaptable to all environments.Finally based on the results of this study in GY and other traits, chickpea varieties in both types were recommended and should be promoted for production in chickpea-growing areas of Ethiopia.
In conclusion, identification of stable improved varieties for the different agroecological regions can assist the producers such as the farmers for the effective chickpea production.This leads to sustainable self-sufficiency of food at the household and country level.
Environments with scores near to zero have low G × E across genotypes and offer low discrimination among genotypes.As a result, Fogera 2021/22 and Adet 2021/22 showed low G × E. Only Debre Zeit 2021/22 was above the mean and a stable environment.In contrast, Debre Zeit 2022/23 and Adet 2022/23 with high interaction across genotypes provided the highest discrimination among genotypes with high-yielding performance tended to have the highest contribution to G × E (Figure 1).
Environments such as Debre Zeit 2022/23, Debre Zeit 2021/22, and Adet 2022/23 had long spokes (located far away from the biplot origin), contributed most to the G × E interactions, whereas Fogera 2021/22 and Adet 2021/22 located close to the origin, and contributed most to the GY stability of the varieties (Figure 2).
Note:Numbers in parentheses show the ranking pattern of varieties based on each stability parameter.Abbreviations: ASV = AMMI's stability value, bi = regression coefficient, Pi = Lin and Binns's superiority index, R 2 = coefficient of determination, s 2 di = deviation from regression, WAASB = weighted average of absolute scores from the singular value decomposition of the matrix of BLUP for GEI effects generated by a LMM, Wi 2 = Wricke's ecovalence, YSI = yield stability index, σ 2 i = Shukla's stability variance.
FIGURE3| The which-won-where display of varieties among the six environments.

FIGURE 6 |
FIGURE 6 | Ranks of the test environments.

TABLE 1 |
Descriptions of the experimental sites.

TABLE 2 |
Description of chickpea varieties used in the study.Y = the yield mean of the ith genotype in the jth environment.Y ij max = the average yield of the genotype with highest yield in the jth environment; and n = the number of environment.

TABLE 3 |
ANOVA for AMMI of G × E on grain yield of Desi and Kabuli chickpea varieties.
*Significant at the 0.1 level, **Significant at the 0.05 level, and ***Significant at the 0.01 level.FIGURE 1 | Plot of genotype and environment IPCA1 scores versus grain yield (GY).

TABLE 4 |
Principal component analyses (PCA)of genotype by environment interaction on grain yield of Desi-and Kabuli-type chickpea varieties.

Table 5
, based on the results of the stability parameters and mean GY as the first criteria for variety FIGURE 2 | Plot of IPCA1 versus IPCA2 scores.

TABLE 5 |
Mean grain yield and stability parameters of 20 chickpea varieties tested across six environments.
. Thus, in the Desi chickpea varieties, Adet 2021/22 and Debre Ziet 2021/22 were best at discriminating the varieties followed by Debre Ziet 2022/23 and Fogera 2022/23 in that order while in the Kabuli chickpea, Adet 2021/22 and Debre Ziet 2021/22 followed by Fogera 2022/23 and Debre Ziet 2022/23 are considered as discriminating environments.The average environment is at the point where the arrow is and the line being average environment axis (AEA).Within this context, a test environment that has a smaller angle with the AEA is more representative than the other test environments (Otieno and Owuor 2019).Thus, Adet 2022/23 and Debre Ziet 2022/23 (in the Desi-type chickpea varieties) and Adet 2021/22 and Debre Ziet 2022/23 (in the Kabuli type chickpea varieties) were considered as the most representative environments.Discriminating but nonrepresentative test environments are good for selecting specifically adapted varieties if they can divide under mega environments, because it could be useful for culling unstable varieties given that it was a single mega environment.So, Debre Ziet 2021/22 and a little bit Fogera 2022/23 (in the Desi chickpea varieties) and Fogera 2022/23 (in the Kabuli-type chickpea variety) could have been considered as discriminating but nonrepresentative test environments as shown in Figure