Genotype by Environment Interaction by AMMI and GGE Bi-Plot Analysis for Maize (Zea Mays L.) for Transitional High Land Agroecology of Ethiopia.


 The current research examined the magnitude of genotype by environment interaction (G x E) and evaluated the adaptability and stability of maize genotypes for grain yield in Ethiopia's transitional highland agroecology using an additive main effects and multiplicative interaction (AMMI) model. The study's goals were to first assess the yield output and stability of maize genotypes in Ethiopia's transitional highlands, and then to investigate the effect of genotype- environment interaction on genotype yield. During the main season of 2017/2018, thirteen advanced maize genotypes which was selected from different observation trials with two commercial check hybrids were evaluated at five representative locations for agroecology. The experiment was set up using an alpha lattice (3*5) with three replications and two rows per plot. AMMI showed highly significant(P < 0.001) variation of grain yield was observed due to the effect of genotype (G), Environment(E) and their interaction (G x E). In fact, all genotypes evaluated in representative locations for this agroecology had higher grain yield advantages than the best commercial check except one genotype. Overall, this study discovered the possibility of fast releasing and overtake of new maize hybrids for transitional high land agroecology of Ethiopia to exploits availability maize germplasm to maximize production. The best candidate genotype, MABK181261 is a stable and high-yielding product. It is recommended for release as a commercial hybrid alternative after national variety verification trial in a high land transitional agroecology of Ethiopia. In addition, the parental lines of this genotypes can be used to enhance germplasm of opposite heterotic group in maize breeding for East Africa.


Abstract
The current research examined the magnitude of genotype by environment interaction (G x E) and evaluated the adaptability and stability of maize genotypes for grain yield in Ethiopia's transitional highland agroecology using an additive main effects and multiplicative interaction (AMMI) model. The study's goals were to rst assess the yield output and stability of maize genotypes in Ethiopia's transitional highlands, and then to investigate the effect of genotype-environment interaction on genotype yield. During the main season of 2017/2018, thirteen advanced maize genotypes which was selected from different observation trials with two commercial check hybrids were evaluated at ve representative locations for agroecology. The experiment was set up using an alpha lattice (3*5) with three replications and two rows per plot. AMMI showed highly signi cant(P < 0.001) variation of grain yield was observed due to the effect of genotype (G), Environment(E) and their interaction (G x E). In fact, all genotypes evaluated in representative locations for this agroecology had higher grain yield advantages than the best commercial check except one genotype. Overall, this study discovered the possibility of fast releasing and overtake of new maize hybrids for transitional high land agroecology of Ethiopia to exploits availability maize germplasm to maximize production. The best candidate genotype, MABK181261 is a stable and high-yielding product. It is recommended for release as a commercial hybrid alternative after national variety veri cation trial in a high land transitional agroecology of Ethiopia. In addition, the parental lines of this genotypes can be used to enhance germplasm of opposite heterotic group in maize breeding for East Africa.

Background
Maize is one of the stable crops that millions of Ethiopians depend on for their daily calories. It grows from the lowlands to the highlands in the agroecology of the country. Ethiopian farmers have used several commercial corn hybrids developed and released by National Maize Breeding Research, such as BH540, BH546 and BH547 for low to mid-altitudes and BH660 and BH661 for transitional highlands.
Selecting superior and stable hybrids from advanced variety trials planted in multiple locations is critical to the agroecology of the transitional highlands of Ethiopia. The performance of genotypes varies in all environments, especially in tropical regions with high seasonal and spatial variability [1][2][3], is referred to as genotype by environment interaction. Experiments in single environment (location) do not cause general conclusions to be drawn about the genotypes being studied. As a result, multi-environment trials are required to identify a dependable and superior hybrid. Therefore, the general strategy of the testing trial in year-by-locations representing agroecological conditions rather than multiple tests combined year after year [4,5]. When genotypes rank differently in different environments (locations and/or years), G x E interaction is important from the standpoint of breeders. The stability of suitable genetic characters is critical for producing improved varieties that can be used commercially under a wide range of agroclimatic conditions. The evaluation of genotypes for the G x E provides useful information on their behavior in speci c environments [6,8]. Therefore, each genotype responds differently to changing climate and soil conditions; some have higher G x E interactions, while others have lower interactions. The main task of maize breeders is to select superior genotypes for narrow or broad adaptation, and to determine the best experimental sites that can be used to select excellent and stable genotypes [5,7]. As a result, yield stability and superiority of maize genotypes evaluated in different environments can help breeders develop superior hybrid. The additive main effects and multiplicative interaction models are frequently used for data analysis to sort out statically signi cant different within genotype, Environment, and their interaction (G x E) to select superior and stable genotype [4,9]. The primary goals of this study were to (1) evaluate the yield performance and stability of maize genotypes for Ethiopia's transitional high land, and (2) investigate the in uence of genotype by environment interaction on genotype yield.

Germplasm used and Description of locations
Thirteen advanced maize genotypes which was selected from different observation trials with two commercial check hybrids were used.
Seeds for 13 genotypes and 2 commercial checks maize hybrids were produced at nursery of breeding blocking through hand pollination at Bako National Maize Research Center ( Table 1). The experiment was evaluated at ve environments vs Bako, Arsi negele, Haramaya, Pawe and F/Selam (Table 2 and Figure 1). Environments are representative for maize growing area of transitional highland altitude agroecology of Ethiopia.  Genotypes were planted in 5.1m long two-row plots with 0.75m between rows and 0.25m between plants. At the planting and owing stages, the recommended NPS was applied. Also, pre-emergency herbicide (Grama Gold) was applied to control weeds. When it's appropriate, hand weeding rehearsals are conducted. The experiment was carried out during the rainy season.

Experimental design and statically analysis
The experimental design used was alpha lattice (3 x 5) with three replications with two rows per plot for each location. Software available online http://www.pbstat.com/ ''PBSTAT-GE'' was used for data analysis. Environments were treated as a random effect in the combined analysis, whereas genotypes were treated as a xed effect. Least square means were simultaneously generated for AMMI model. Following single site ANOVA for grain yield, the combined ANOVA and AMMI analysis of variance were done for G, G x E and G x E interaction [10][11]. The following formula used to estimate yield of genotype, environment and G X E interaction through stability analysis using AMMI analysis model [12] as follow Whereas Yge=GY of genotype, e =environment, µ= the grand mean; ag= genotype mean deviation; βe= environment mean deviation; kn = Eigen value of the principal component (PCA) axis, cgn and gen = the genotype and environment PCA scores for the PCA axis, n N = number of PCA axes retained in the model and hge = residual.

Data collected
According to Getachew [13], data on grain yield and other important agronomic traits are collected as follows: (1) Grain weight (t/ha (2) Actual moisture content (
However, BH660 (check1) produced the lowest yielder for ve individual locations: Bako (5.52 t-ha -1 ), A/Nagel (4.41 t-ha-1), Pawe (3.51 tha -1 ), F/Selam (4.87 t-ha -1 ) and Haromaya (3.6 t-ha -1 ). The average grain yield for all genotypes evaluated in multi-environment was higher than the best commercial check (BH661), with the exception of MABK181254. This implies possibility of fast releasing and overtake of new maize hybrids for transitional high land agroecology of Ethiopia to exploits available of maize inbred line germplasm to maximize production and productivity. Wender [21] also reported similar result that grain yield of some hybrids evaluated over location was higher than grain yield of commercial check. From total studied genotypes, seven genotypes had more than grand mean (6.55 t-ha -1 ) of grain yield. This study approves Legesse [14] which

GGE Biplot analysis
The GGE biplot can be thought of as a graphical representation of matrix multiplication. As a result, fteen hybrids, including two commercial checks, were evaluated at the ve sites shown in Figures 2, 3, 4, and 5. GGE biplot analysis gathered a lot of useful information from various biplot graphs [4, 5 and 15]. The x-axis was the rst principal component (PC1) scores, and the y-axis was the second principal component (PC2) scores. Biplot's total variation in relation to G and GEI explained 80.1 percent of the variance ( Figure   2). PC1 explained 65.4 percent of genotype by GGE from total variation, while PC2 explained 14.7 percent. Two principal components (PC1 & PC2) explained more than half of the G x E interpretation. Therefore, the rst two PCs can be used to predict the best AMMI model to explain the two interpretable modes which supported nding of Yan & Tinker [20] and Yan [21]. The GGE biplot is shown in Figure 4, in which the environments are connected to the biplot origin vs the vectors. This perspective aids in comprehending the environment's interrelationships. According to Yan [22], the vector length close to the standard deviation in each test environment is a measure of the environment's ability to distinguish genotypes. As a result, the F/Selam's vector is abnormally short in comparison to the plot size. Therefore, predicted genotype differences based on vectors from such environments may not be reliable and only re ect noise. The polygon view of the GGE biplot is useful for visualizing the multi-environment trial dataset's "which-won-where" pattern. PC1 accounted for 62.7 percent of variation, while PC2 accounted for 17.5 percent ( Figure 5). Demisew [15], Hortense [23] and Legesse [14] also reported the variation of PC1 was much greater than PC2 in their studies of maize hybrids. Grain yield was explained by 80.4 percent of total variation on both axes. Draw vertical lines on each side of the polygon starting at the biplot origin. As shown in Figure 5, the rays of the two-line graphs divide the graph into six parts, with ve environments appearing in one sector and the remaining environments appearing in another. The sectors had different high yielding vertex genotypes, such as MABK181261, MABK181255, MABK181253, MABK181254, these genotypes were among the most responsive to the environments in their respective directions. When a test environment is divided into mega Environments, it means that different environments have different high yielding genotypes for those sectors, indicating crossover G x E [24]. Based on this, the polygon views divided the test environment into three mega environments, with Bako, Pawe and Haromaya in best genotype MABK181262 group I, F/Selam MABK181259 group II and A/Negele in best genotype 6 (group III). Eliyas [16] also used GGE-biplot models and classi ed eight environments into two mega-environments for sugar cane genotypes evaluation for yield. To avoid random GEI, genotype evaluation in mega environment should be based on both mean performance and stability [25]. As a result, the genotypes MABK181261 and MABK181262 were identi ed as high yielders in the AMMI analysis. However, MABK181261 was identi ed as stable. only MABK181261 was chosen as the ideal genotype because it met both of the criteria. Demissew [15] also selected three maize hybrids for high land mid altitude of Ethiopia using the same criteria. Genotype MABK181262 is the highest yielding vertex hybrid in all the test environments that share the sector with it. Whereas MABK181259 was the highest yielding found in separate sector at F/selam site.
The environment -vector view of the GGE biplot with the average -environment axis was shown in Figure 3. A test environment with a smaller angle with average environmental axis is the most representative of the others. Thus, Pawe's vector is very short, ensuring that all genotypes perform similarly. Consequently, it no information or gave little information about genotype difference [4 and 14]. According to Yan [22] the biplot's concentric circles aid in visualizing the length of the environment vectors. This is a measure of the environments' discriminating ability and representativeness that is proportional to the standard deviation of each environment Bako has long vectors and small angles, and when combined with absicca, which is more discriminating genotypes and representative of test environments, the result is ideal for genotype selection. Through discriminating and representativeness by GGE biplot analysis, testing environments were used to distinguish environments that select high yielder genotypes in sets of environments. A/NEGEL has a long vector with a large angle as the result it cannot be used to select genotype.
performing genotypes, but it can provide information on genotypes that are unstable [24]. The discriminating test environments of A/NEGEL and F/SELAM are not representative, but they are useful for selecting speci c adapted genotypes and discarding unstable genotypes. The performance of genotypes can be easily identi ed in a test environment with high representativeness and discriminating ability [21]. Only one test environment (Bako) is both representative and discriminating, making it a good place to nd genotypes that are generally adapted. Similar results as many authors Y Matana [4]; SALEEM [26], Dagnechew [27], Ndlala [28], Dao [29] and Moges [30], who used GGE biplot to identify high representativeness and discriminating for different crops.

Stability of genotype
The stability of maize genotypes and average grain yield in all environments should be assessed within a single mega environment. Figure 4 shows the stability and average performance of the evaluation genotype. According to Yan [21], the AEC abscissa is a single arrowed line that points to higher mean yield cross environments. Thus, the maximum grain yield was MABK181262, followed by MABK181261, MABK181259, and so on. The acute angle between MABK181262 and MABK181259 indicates that these two genotypes respond similarly in all environments, and the difference between them is proportional. Whereas, an obsute angle (e.g., MABK181261 vs BH660) indicates that the genotypes reacted inversely, with the MABK181261 outperforming the BH660 and vice versa. The difference in performance between genotypes MABK181261 and MABK181262, MABK181253 and MABK181251, and MABK181254 and BH660 in the respond environment was due to G x E, where the genotypes were presented at right angles. AEC ordinate is the double arrowed line. It indicates a higher level of variability (poor stability in both directions). As a result, G1 was extremely stable, whereas MABK181259 was extremely unstable.

Conclusion
Identifying stable, high-yielding maize hybrids in multi-environment trials is critical to the success of commercial hybrids in Ethiopian highland transitional agroecology. In fact, all genotypes evaluated in representative of locations for this agroecology had higher grain yield advantages than the best commercial check (BH661) except one genotype. According to the GGE biplot model, only the Bako environment was chosen for both representative and discriminating genotype selection, indicating that it is an appropriate environment for selecting generally adapted genotypes. The test environment was divided into three mega environments by polygon views: Bako, Pawe, and Haromaya. The biplot's rays divided the plot into six sections, with four environments appearing in one sector and one appearing in a different sector. Seven genotypes which had different high yielding vertex with the longest vectors are located at the polygon's corner of vertex sector. In comparison to other hybrids, these genotypes were among the most responsive to the environments in their respective directions. In fact, all genotypes evaluated in representative locations for this agroecology had higher grain yield advantages than the best commercial check except one genotype. Overall, this study discovered the possibility of fast releasing and overtake of new maize hybrids for transitional high land agroecology of Ethiopia to exploits availability maize germplasm to maximize production. The best candidate genotype, MABK181261 is a stable and high-yielding product. It is recommended for release as a commercial hybrid alternative after national variety veri cation trial in a high land transitional agroecology of Ethiopia. In addition, the parental lines of this genotypes can be used to enhance germplasm of opposite heterotic group in maize breeding for East Africa.