©Haramaya University, 2022 ISSN 1993-8195 (Online), ISSN 1992-0407(Print) East African Journal of Sciences (2022) Volume 16(1): 77–92 Licensed under a Creative Commons *Corresponding author: atinkut947@gmail.com Attribution-NonCommercial 4.0 International License. Genotype x Environment Interaction and Grain Yield Stability of Improved Teff [Eragrostis tef (Zucc.) Trotter] Varieties in Northwestern Ethiopia Atinkut Fentahun1*, Tiegist Dejene2, and Kebebew Assefa3 1Adet Agricultural Research Center, Amhara Regional Agricultural Research Institute, P.O. Box 08, Bahir Dar, Ethiopia 2College of Agriculture and Environmental Science, Bahir Dar University, P.O. Box 79, Bahir Dar, Ethiopia 3Debre Zeit Agricultural Research Center, Ethiopian Institute of Agricultural Research, P.O. Box 32, Debre Zeit, Ethiopia Abstract Background: Teff is an important staple cereal crop in northwestern Ethiopia. However, the yield of the crop is very low due to, among others, lack of stable and high yielding varieties under varying environmental conditions because of genotype x environment interaction effect. Objective: The study was conducted to assess the effect of genotype by environment interaction, identify mega environments, and select high yielding and stable teff genotypes that interact less with the changing environment. Materials and Methods: Twenty improved teff varieties were evaluated using a randomized complete block design with three replications at Adet, Motta, Bichena, Debre-Tabor and Takussa districts for two consecutive years. Data were collected on days to heading and maturity, plant height, grain filling period, panicle and culm length, dry plant biomass and grain yield. The data were analyzed using a combined analysis of variance and genotype main effect plus genotype by environment interaction biplot. Results: The combined analysis of variance for grain yield revealed highly significant (P < 0.001) effects for genotype (16%), environment (54%) and genotype x environment interaction (23%). The effect of environment was three times higher than that of genotype, indicating significant and undesirable influence of the environment on genotype stability. The mean grain yield across the environments ranged from 1.65 to 2.77 tons ha–1 for Debre-Tabor and Takussa, respectively. The genotype mean yield ranged from 1.68 to 2.51 tons ha–1 for Simada and Hiber-1, respectively. Genotype by environment interaction biplot analysis grouped the ten test environments and twenty genotypes into three mega-environments and four genotype groups. Besides, Adet district and Bichena district had relatively the longest vector length and the smallest angles with the average environmental axis, thus being the most representative of all environments. Regarding genotypes, Hiber -1 followed by Kora, Etsub and Dukem were identified as the best yielding and relatively stable genotypes to increase teff productivity in the region. Conclusion: The biplot analysis of the genotype by environment interaction resulted in the identification of Adet and Bichena districts as the most favorable locations for teff production and as well as Hiber-1 as the most productive teff variety for cultivation in the study area. This implies that farmers in the two districts could be advised to take up this variety for enhancing yield of the crop and income from its production. Keywords: Combined analysis; Environmental axis; Genotypes; Genotype main effect; Genotype x environment interaction; Grain yield stability Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 78 1. Introduction Teff is an important staple cereal crop in Ethiopia as well as in Amhara National Regional State, particularly in northwestern Ethiopia. In Ethiopia, the crop covers 3.02 million hectares per year and accounts for 30% of the total land allotted to cereals. It is second in total production (5.28 million tons) accounting for 20% of grain production among all cereals grown, while its national average productivity is 1.75 tons ha–1 (Central Statistical Agency (CSA), 2018). In Amhara National Regional State (ANRS), teff productivity (1.8 tons ha–1) is slightly more than the national average (CSA, 2018). Amhara region (northwestern part of Ethiopia) is one of the major teff growing areas in the country. According to CSA (2018), the contribution of the region in terms of area coverage and total production is about 38% and 39%, respectively. In spite of the importance of the crop, both the national and regional average yields are very low as compared to other cereals grown in Ethiopia as well as the crop’s genetic potential. A study conducted under non-lodging condition has demonstrated that yield potential of the crop can further be increased up to 4.6 tons ha–1 (Yifru Teklu and Hailu Tefera, 2005). This shows there is higher difference between the potential of the crop and the actual yield, which is less than half. Some of the factors contributing to the low yield of teff include low soil fertility, lack of high yielding and widely adaptable varieties, weeds, erratic rainfall distribution in lower altitudes, lodging, waterlogging, and low moisture (Fufa Hundera, 1998). Among these factors, scarcity of stable and high yielding teff varieties under varying environmental conditions due to GEI effect is the main factor. The occurrence of genotype by environment interaction is the basic cause for differences among genotypes in terms of grain yield stability. Stability analysis can help to characterize the response of varieties to the changing environments and to determine the best and representative locations of the environmental diversity (Mohammed et al., 2008). Hence, conducting experiments in several locations and seasons are needed to determine stable and high yielding varieties of the crop. Multi-Environment Trials (MET) are carried out to evaluate grain yield stability performance of genetic materials under varying environmental conditions (Delacy et al., 1996; Farshadfar et al., 2012). Consequently, genotype x environment effects can be revealed by multi- environment trial experiments. The presence of a significant genotype x environment interaction for quantitative traits such as grain yield can lead to the failure of genotypes to achieve the same relative performance in different environments (Fekadu Gurmu et al., 2009). Information on genotype x environment interaction (GEI) leads to successful identification of stable genotypes, which could be used for wider cultivation. Yield is a complex quantitative character greatly influenced by environmental fluctuations; hence, the selection of superior genotypes based on grain yield performance at a single location in a year would not be very effective (Shrestha et al., 2012). Thus, evaluation of genotypes for stability of performance under varying environmental conditions for yield has become an essential part of any plant breeding program. Moreover, understanding of genotype x environment interaction enables us to effectively allocate resources and to characterize genotypic responses to diverse crop productivity levels (Tiruneh Kefyalew et al., 2000). Thus, it enables to eliminate unnecessary spatial and temporal replications of yield trails as well as to establish additional testing environments when the existing ones are under- represented (Basford and Cooper, 1998). Even though some studies have been conducted to elucidate the G x E interaction of teff in other parts of the country, there is little information on the G x E interaction of teff varieties in diverse environmental conditions of north western Ethiopia. The importance of conducting more studies across major teff growing environments has been suggested by Mathewos Ashamo and Getachew Belay (2012), so as to enable breeders to identify adaptable, stable, and high yielding teff genotypes. Therefore, the objectives of the present study were to assess the effect of GEI and identify mega- environments, and to select high yielding and stable teff genotypes that interact less with the changing environment in northwestern Ethiopia. Atinkut et al. Genotype x Environment Interaction and Yield Stability of Teff 79 2. Materials and Methods 2.1. Description of the Study Area A field experiment was conducted in five locations during the 2018 and 2019 main cropping seasons under rain-fed condition. The test locations were Adet (Yilmana-Densa), Motta (Hulet-Eju-enesie), Bichena (Enemay), Debre- Tabor, and Takusa districts (Figure 1). The climatic, edaphic and geographic descriptions of the locations are different and presented in Table 1. Figure 1. Map of Amhara National Regional State showing the testing sites. Table 1. Description of the locations and seasons. Test environments Altitude (m.a.s.l.) Geographical location Soil type Weather data Code Name Latitude Longitude Rainfall (mm) Temperature (°C) Max. Min. E1 Debre-Tabor-2018 2591 11051'N 38001'E Luvisol 1609.5 22.5 9.5 E2 Adet –2018 2240 11016' N 37029' E Nitosol 1432.0 25.6 10.8 E3 Motta–2018 2470 11020'N 37088' E Nitosol 1334.0 23.8 10.5 E4 Bichena–2018 2541 10046'N 38019'E Vertisol 1186.0 24.4 11.1 E5 Takussa–2018 1840 12010'N 37006' E Vertisol 870.0 15.0 28.0 E6 Debre–Tabor-2019 2591 11051'N 38001'E Luvisol 1926.1 22.7 10.0 E7 Adet–2019 2240 11016' N 37029' E Nitosol 1591.8 25.9 11.1 E8 Motta–2019 2470 11020'N 37088' E Nitosol 1457.5 24.1 11.4 E9 Bichena–2019 2541 10046'N 38019'E Vertisol 1126.4 23.5 10.7 E10 Takussa–2019 1840 12016'N 37006' E Vertisol Na Na Na Note: m.a.s.l. = meter above sea level; E = East; N = North; Max. = Maximum; Min. = Minimum; and Na = Not available for the specified periods. Weather data were collected from West Amhara Meteorological Cervices Center (2018 and 2019), and edaphic information’s were obtained from the respective research centers. 2.2. Plant Materials Twenty teff varieties released for cultivation in Ethiopia were used in this study (Table 2). The varieties were deliberately collected based on their phenology, adaptation, yield performance, and seed color, to select high yielding and stable teff genotypes that interact less with the changing environment in north western Ethiopia (Table 2). Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 80 2.3. Treatments and Experimental Management The treatments consisted of 20 genotypes (Table 2) and ten environments (locations) (Table 1). The experiment was laid out as a Randomized complete block design and replicated three times per treatment. Each plot consisted of six rows with 2 m length and 0.2 m spacing between rows. The spacing between blocks and plots were 1.5 m and 1 m, respectively. Land was prepared according to the conventional practice. It was ploughed five times using oxen before planting and the plot after the last ploughing was used for sowing. Planting was done from the beginning to end of July depending on the recommendations for the different test locations. Sowing was done by hand drilling using seeding rate of 15 kg ha– 1. A blanket recommendation rate of Urea and Di ammonium Phosphate (DAP) were applied to the plots at the rate of 40/60 N/P2O5 ha–1 for Nitisols and Luvisols and 60/60 N/P2O5 ha–1 for vertisols. All of the DAP was applied at planting but Urea was top dressed at tillering stage. All other pre-and post-planting management practices were made as per the recommendations for teff husbandry in all test locations. 2.4. Data Collection Data were collected on plant, row and plot basis. Five randomly taken plants were selected from the central four rows for plant-based parameters. The entire six rows in the plot were used for plot-based data scoring whereas four central rows were used for row-based data scoring. Plant height, panicle length, and culm length were recorded on plant basis. Shoot dry biomass and grain yield were recorded on row basis whereas days to heading and days to maturity, grain filling period were recorded on whole-plot basis. Table 2. Descriptions of teff genotypes used for the study. Genotype ID Year of release Suitable environment Seed color Breeding center Grain yield (tons ha–1) Research station Farmers’ field WellenKomi (DZ-01-787) G1 1978 HP Pale white Debre-Zeit 2.4-3.0 2.0-2.4 Tsedey (DZ-Cr-37) G2 1984 LW White Debre-Zeit 1.8-2.5 1.4-2.2 Dukem (DZ-01-974) G3 1995 HP Pale white Debre-Zeit 2.4-3.4 2.6-2.7 Ziquala (DZ-Cr-358) G4 1995 HP White Debre-Zeit 17-24 16-22 Ambo Toke (DZ-01-1278) G5 2000 HP White Holeta 2.4-3.5 Na Dega-Tef (DZ-01-2675) G7 2005 HP Pale white Debre-Zeit 1.5-2.2 1.6-2.0 Quncho (DZ-Cr-387RIL355) G8 2006 HP Very white Debre-Zeit 2.0-3.2 1.8-2.6 Etsub (DZ-Cr-3186) G9 2008 HP White Adet 1.9-2.7 1.6-2.2 Simada (DZ-Cr-285RIL295) G10 2009 LW White Debre-Zeit 1.9-2.8 1.6-2.5 Boset (DZ-Cr-40RIL-50d) G11 2012 LW Very white Debre-Zeit 1.8-2.0 1.4-1.8 Kora (DZ-Cr-438RIL-133B) G12 2014 HP Very white Debre-Zeit 2.5-2.8 1.8-2.2 Were-Kiyu (Acc.21476A) G13 2014 LW White Sirinka 2.22 Na Abola (DZ-Cr-438(RIL7) G14 2015 HP White Adet 2.0-2.8 1.46-1.68 Dagim (DZ-Cr-438-RIL91A) G15 2015 HP White Debre-Zeit 2.5-2.8 1.8-2.3 Areka-1(DZ-01-974*DZ-012788) G6 2017 LW White Areka 1.6-1.87 1.6-1.75 Negus (DZ-Cr-429RIL125) G16 2017 HP Very white Debre-Zeit 2.8 Na Felagot (DZ-Cr-442RIL77C) G17 2017 HP Brown Debre-Zeit 2.54 Na Tesfa (DZ-Cr-457-RIL181) G18 2017 HP Very white Debre-Zeit 2.5 Na Hiber-1 (DZ-Cr-419) G19 2017 LW White Adet 1.7-2.7 1.46-2.08 Abay (Acc#225931) G20 2018 HP White Adet 2.5-3.5 1.8-2.2 Note: ID = identification; HP = High potential; LW = Low moisture; and MoA = Ministry of Agriculture. Na = not available. Genotypes were obtained from MoA (2006–2018) and variety releasing centers. Atinkut et al. Genotype x Environment Interaction and Yield Stability of Teff 81 2.5. Data Analysis Analysis of variance for grain yield and yield related traits was carried out for each location separately using PROC GLM model of SAS computer program (SAS Institute, 2002). Duncan’s Multiple Range Tested was used for mean separation. The combined analysis of variance across locations was done using PROC GLM with MIXED procedure of SAS software which corresponds to the statistical model. Genotype effects were assumed to be fixed and environmental effects as random. To determine the validity of the combined analysis of variance, the homogeneity of error variance between environments were performed based on the ratio of the larger mean square of error (MSE) from the separate analysis of variance to the smallest mean square of error as: F– ratio = Large MSE Small MSE If the larger error mean square was not three-fold larger than the smaller error mean square, the error variance was considered homogeneous (Gomez and Gomez, 1984). Genotype by environment interaction was quantified using pooled analysis of variance, which partitions the total in to its component parts (genotype, environment, genotype x environment interaction and pooled error). The following statistical model was used for ANOVA of data of the individual environments: Yij = μ + Gi + Bj + Ɛij Where, Yij = observed value of genotype i in block j; µ = grand mean of the experiment; Gi = effect of genotype I; Bj = the effect of block j; and, Ɛij = error effect of genotype i in block j. In performing the combined analysis of variance, genotypes were assumed to be fixed while replications with in environments were assumed random. The following statistical model was used for combined analysis of variance over locations: Yijk = μ + Gi + Ej + GEij +Bkj+ Ɛijk Where, Yijk = observed value of genotype i in block k of environment (location) j, µ = grand mean, Gi = effect of genotype i, Ej= environment or location effect, GEij = the interaction effect of genotype i with environment j, Bkj = the effect of block k in location (environment) j, and Ɛijk = error (residual) effect of genotype i in block k of environment j. The combined analysis of variance was carried out to estimate the additive main effects of environment, genotype and GEI. Significance levels of these components were determined by using F-tests. Whenever the F-test was found significant, genotype x environment interaction was described using GGE biplot analysis (Yan et al., 2000) using GENSTAT analytical software version 18 (VSN International, 2015). Genotype main effect plus genotype x environment interaction biplot model (Yan et al., 2000) is the most commonly used and more efficient in determining the most stable and high yielding genotypes in multi-environment trials as compared to the earlier procedure (Eberhart and Russell, 1966; AMMI model, Guach and Zobel, 1988). The GGE biplot allows visual examination of the relationships among the test environments, genotypes and genotype x environment interactions. Thus, the first two principal components (PC1 and PC2) were used to graphically represent the GEI, and to identify the rank of the test genotypes and environments (Yan et al., 2000). GGE biplot analysis was based on the simplified model with two principal components (Yan et al., 2000). The model was: Yij–yij + l1xi1hj1 + l2xi2hj2 + εij In which, Yij is the productivity mean of cultivar i in environment j, yij is the general mean of the cultivars in environment j, l1 xi1 hj1is the first principal component (PCA1), l2 xi2 hj2 is the second principal component (PCA2), l1 and l2 are the eigenvalues associated with PCA1 and PCA2, respectively, xi1and xi2 are the values of the first and second principal components, respectively, for cultivar I, hj1 and hj2 are the values of the first and second principal components, respectively, for environment j, and εij is the error ij associated with the model. Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 82 3. Results and Discussion 3.1. Analysis of Variance Variance of homogeneity from results of the quick Bartlett test revealed that the mean squares of individual locations were homogenous for grain yield. Thus, the combined analysis of variance for gain yield of 20 improved teff varieties at ten environments showed highly significant (P < 0.001) effects of genotypes, environments and genotype x environment interaction (Table 3). Environments accounted for 54% of the total variation followed by the GEI (23%) whereas the genotype alone accounted 16% (Table 3). Habte Jifar et al. (2019) reported similar findings in teff multi-environment trials, where the largest proportion of total variation was attributed to locations and relatively smaller effects were noted due to genotype and genotype and environment interaction (GEI). Sewagegne Tariku et al. (2018) also found similar results in which environments contributed about 91% of the total variation in grain yield of teff, while genotypes and GEI accounted for about 0.87% and 3.63%, respectively. The high percentage of the environment sum squares is an indication that the major factor that influence yield performance of teff genotypes is the environment. Besides, the environmental effect was found to be highly significant. This may indicate presence of significant differences among testing locations due to variation in temperature, soil type, rainfall, and other environmental factors as also reported by Legesse Kassa et al. (2006). Table 3. Combined ANOVA for grain yield (ton ha–1) of genotypes tested at five environments during 2018 and 2019 cropping seasons. Source of variation Degrees of freedom Sum of squares Percent of total explained Mean squares Pr > F Blocks (Environments) 20 1.4323 0.925 0.072 <.0001 Environment 9 82.75 53.47 9.194 <.0001 Genotype 19 25.39 16.41 1.336 <.0001 Genotype × Environment 171 35.58 22.99 0.20807 <.0001 Residuals 380 9.61 6.21 0.02529 Total 599 154.76 100 Mean 2.16 Coefficient of determination (R2) 0.94 Coefficient of variation (%) 7.36 The large sums of squares associated with the environment in the present study indicate that the selected test environments were agro-ecologically diverse. This signifies the importance of site selection for teff cultivation. The mean grain yield across environments ranged from 1.65 tons ha–1 for Debre-Tabor to 2.77 tons ha–1 for Takussa. On the other hand, the grain yield means of the genotypes ranged from 1.68 to 2.51 tons ha–1 for Simada and Hiber-1, respectively. Takusa and Bichena were relatively high yielding environments compared to Debre-Tabor, Motta and Adet. The variety Hiber-1 performed best in most of the environments (locations) followed by the varieties Etsub and Kora. Apart from this, the teff varieties with higher grain yield at specific location respectively, were: Hiber-1, Dagim and Etsub at Adet; Kora, Hiber-1 and Dukem at Bichena; Etsub, Hiber-1 and Kora at Takussa; Worekiyu, Wellenkomi and Filagot at Motta; and Etsub, Dukem and Wellenkomi at Debre-Tabor (Table 4). The high variability in grain yield among the twenty teff varieties at the ten environments might be due to wide variability in climatic and soil conditions. Similarly, inconsistent grain yield performances of teff varieties have been found across locations (Solomon Chanyalew et al., 2009; Ayalneh Tilahun et al., 2012; Mathewos Ashamo and Getachew Belay, 2012; Wendwosen Shiferaw et al., 2012; Sewagegne Tariku et al., 2018; Habte Jifar et al., 2019). Atinkut et al. Genotype x Environment Interaction and Yield Stability of Teff 83 Table 4. Mean grain yield ton ha–1 of twenty teff genotypes for individual environments during the 2018 and 2019 main cropping seasons. Variety 2018 2019 Overall mean Rank D/T Adet Motta Bichena Takusa D/T Adet Motta Bichena Takusa Wellenkomi 2.18bc 2.38cde 2.12 a 2.93bc 2.36de 1.99bc 1.77fg 2.21ab 2.53abc 2.31defg 2.28 5 Tsedey 1.25h 1.95fg 1.98abc 2.35e 2.55cd 1.32j 2.03cde 1.94cdef 1.72f 2.29defg 1.94 17 Dukem 2.29b 2.29cde 1.78cde 3.13b 3.03ab 2.10ab 2.21abcd 1.78efg 2.85a 2.33def 2.38 4 Ziquala 1.62efg 2.24def 1.59e 2.53de 2.46cde 1.48fghi 2.00de 1.70g 1.92ef 2.28defg 1.98 15 Ambotoke 1.85de 2.59bc 1.94abcd 2.80bcd 3.05ab 1.53fgh 2.02de 2.01cbd 2.34bcd 2.22efg 2.24 7 Areka-1 1.72ef 2.12efg 1.67de 1.75f 2.90ab 1.89c 1.73gh 1.71g 1.86f 2.36cde 1.97 16 Dega-Tef 2.18bc 2.19ef 1.94abcd 2.82bcd 1.87f 1.93c 2.08cde 1.96cdef 2.72ab 1.89i 2.16 9 Quncho 1.96cd 2.53bcd 1.80bcde 2.96bc 3.03ab 1.60def 2.11bcd 1.89cdefg 2.61abc 2.10ghi 2.26 6 Etsub 2.52a 2.81ab 1.79bcde 2.94bc 3.20a 2.23a 2.20abcd 1.87cdef 2.60abc 2.64a 2.48 2 Simada 1.27h 1.57h 1.65e 1.96f 2.93ab 1.36ij 1.63gh 1.24h 1.08g 2.12fgh 1.68 19 Boset 1.49fgh 2.29cde 1.93abcd 2.88bc 2.88ab 1.39hij 2.00de 1.74fg 2.35bcd 2.57abc 2.15 10 Kora 1.99cd 2.56bcd 1.97abc 3.46a 3.03ab 1.70de 2.26abc 1.96cdef 2.91a 2.46abcd 2.43 3 Werekiyu 1.71ef 2.38cde 2.16a 2.77cd 2.13ef 1.56ef 2.00def 2.27a 2.36bcd 2.15efgh 2.15 11 Abola 1.68ef 2.29cde 1.97abc 2.88bc 2.37de 1.41ghij 1.99def 1.83defg 2.62abc 2.11fghi 2.12 13 Dagim 1.58fg 2.80ab 1.74cde 2.72cd 2.72bc 1.55fg 2.32ab 1.89cdefg 2.58abc 2.45abcd 2.24 7 Negus 1.52fg 2.82ab 2.00abc 2.91bc 2.39cde 1.52fgh 2.15bcd 2.02bcd 1.86f 2.17efgh 2.13 12 Felagot 1.39gh 2.12efg 2.06ab 2.93bc 3.12a 1.72d 2.20abcd 1.88cdefg 2.26cde 2.37bcde 2.21 8 Tesfa 1.27h 1.86g 1.82bcde 2.29e 3.08ab 1.34ij 1.51h 1.81defg 1.32g 1.99hi 1.83 18 Hibir-1 1.97cd 2.94a 1.99abc 3.15b 3.17a 1.97bc 2.39a 2.08abc 2.89a 2.58ab 2.51 1 Abay 1.41gh 2.11efg 1.98abc 2.86bcd 3.07ab 1.36ij 1.84efg 1.97cde 2.02def 2.21efgh 2.08 14 Mean 1.74 2.34 1.89 2.75 2.77 1.65 2.02 1.89 2.27 2.28 2.16 SEM (+) 0.076 0.098 0.084 0.105 0.108 0.0523 0.082 0.078 0.133 0.244 CV (%) 7.59 7.23 7.71 6.63 6.73 5.49 703 7.15 10.17 5.85 Note: Means in the same column followed by the same letters are not significantly different at P < 0.05 using Duncan’s Multiple Range Test. D/T = Debre-Tabor. CV = Coefficient of variation and SEM =Standard error of the mean. Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 84 3.3. Stability Analysis and Mega-Environment Classification Using GGE Biplot 3.2.1. The “Which-won-where” pattern The analysis of variance showed the presence of highly significant G x E mean squares for grain yield across the test environments. This result indicates that the use of the GGE biplot would be pertinent to decompose the G x GEI effects. The Principal Component axis1 (PC1) and axis 2 (PC2), are cumulatively, explained 68% of the total variation for grain yield (Figure 2). This result suggests that the biplot graphics explained most of the sums of squares for genotype by environment interaction. This outcome made it possible to have a safe genotype selection based on the multivariate analysis as per the suggestions of Yan (2001). The varieties and the environments found inside the polygon were less responsive to environment stimuli (Figure 2). Environments grouped inside the same polygon had similar influence on the genotypes. Environment groups deriving from the ten assessed environments revealed three mega-environments. The first one encompassed Takusa and Debre-Tabor areas with genotypes Etsub and Hiber-1 presented in the vertex. The second mega-environment contains Adet and Bichena areas and the third mega-environment contains only one location which is Motta with the vertex genotype Werkiyu (Figure 2). Related to this result, Karimizadeh et al. (2013), Yirga Belay (2016) and Sewagegne Tariku et al. (2018) identified different lentil, sesame and tef varieties, respectively, growing on mega environments. In the polygon view of the biplot analysis, the genotypes and test environments fell into four and three sectors, respectively. Varieties from the polygon vertex that did not group in any one of the environments were not fit varieties for the tested environment. The vertex variety Simada (G10) and Tesfa (G18) had no corresponding environment and hence have the lowest mean grain yield across environments (Figure 2). Habte Jifar et al. (2019) similarly reported non fit varieties for the tested environments in teff GGE biplot analysis. Figure 2. Polygon views of the GGE-biplot based on symmetrical scaling for the Which-won-where pattern analysis for varieties and environments (E1 = Debre-Tabor year-1, E2 = Adet year-1, E3 = Motta year-1, E4 = Bichena year-1, E5 = Takusa year-1, E6 = Debre-Tabor year-2, E7 = Adet year-2, E8 = Motta year-2, E9 = Bichena year-2, E10 = Takusa year-2, G1 = Wellenkomi, G2 = Tseday, G3 = Dukem, G4 = Ziquala, G5 = Ambotoke, G6 = Areka-1, G7 = Dega-Tef, G8 = Quncho, G9 = Etsub, G10 = Simada, G11 = Boset, G12 = Kora, G13 = Werekiyu, G14 = Abola, G15 = Dagim, G16 = Negus, G17 = Flagot, G18 = Tesfa, G19 = Hiber-1, and G20 = Abay). Atinkut et al. Genotype x Environment Interaction and Yield Stability of Teff 85 3.2.2. Evaluation of genotypes relative to ideal genotypes A genotype which is found at the center of the concentric circle is considered as an ideal genotype for teff grain production with its high mean yield and consequently stable characteristics, and genotypes that are close to the ideal genotype are considered as good genotypes (Figure 3). Accordingly, Hiber-1 (G19) being at the center of the concentric circle can be considered as an ideal genotype for teff grain production with high mean yield and stable characteristics. Likewise, Kora (G12), Etsub (G9) and Dukem (G3) that were close to the ideal genotypes are considered as good genotypes based on their yield performance as well as stability. On the other hand, Simada (G10), Tesfa (G18), Tseday (G2) and Areka- 1(G6) which are located farther from the first concentric circle are undesirable and low yielding genotypes (Figure 3). These results are confirmed by the mean separation test discussed earlier in Table 4. Similarly, Sewagegne Tariku et al. (2020) identified variety Hebir-1 is the most ideal genotype for teff grain production in teff variety verification trials. The relative contribution of stability and grain yield for identifying desirable genotype found in this study by the ideal genotype procedure of GGE biplot were also similar to Fan et al. (2007) maize hybrids stability studies. Figure 3. GGE-biplot based on genotype-focused scaling for comparison the genotypes with the ideal genotype (E1 = Debre-Tabor year-1, E2 = Adet year-1, E3 = Motta year-1, E4 = Bichena year-1, E5 = Takusa year-1, E6 = Debre-Tabor year-2, E7 = Adet year-2, E8 = Motta year-2, E9 = Bichena year-2, E10 = Takusa year-2, G1 = Wellenkomi, G2 = Tseday, G3 = Dukem, G4 = Ziquala, G5 = Ambotoke, G6 = Areka-1, G7 = Dega-Tef, G8 = Quncho, G9 = Etsub, G10 = Simada, G11 = Boset, G12 = Kora, G13 = Werekiyu. G14 = Abola, G15 = Dagim, G16 = Negus, G17 = Flagot, G18 = Tesfa, G19 = Hiber-1, and G20 = Abay). 3.2.3. Interrelationship among environment In GGE biplot, the cosine of the angle between any environment vectors stands for correlation intensity. Less than 90° indicates a positive correlation, more than 90° indicates a negative correlation and close to 90° indicates no correlation (Yan and Kang, 2003). The angle between the vectors of two environments has a meaningful relation with the correlation coefficient between them (Yan, 2002; Yan and Kang, 2003) and such a relationship is used to group the test environments. Thus, if two environments are positively correlated, the best yielding genotypes in one environment will perform best in the other environments. In contrast, if two environments are negatively correlated, the best yielding genotypes in one Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 86 environment perform the least in the other environment and vice versa (Yan, 2002; Yan and Kang, 2003). In the present study, as shown on Figure 4, Adet, Debretabor and Bichena with an angle less than 90o are positively correlated with each other. On the other hand, Takussa and Motta environments had greater than 90o angle and hence have negative correlations. Figure 4. GGE-biplot based on environment-focused scaling for environments (E1 = Debre-Tabor year-1, E2 = Adet year-1, E3 = Motta year-1, E4 = Bichena year-1, E5 = Takusa year-1, E6 = Debre-Tabor year-2, E7 = Adet year-2, E8 = Motta year-2, E9 = Bichena year-2, E10 = Takusa year-2, G1 = Wellenkomi, G2 = Tseday, G3 = Dukem, G4 = Ziquala, G5 = Ambotoke, G6 = Areka-1, G7 = Dega-Tef, G8 = Quncho, G9 = Etsub, G10 = Simada, G11 = Boset, G12 = Kora, G13 = Werekiyu. G14 = Abola, G15 = Dagim, G16 = Negus, G17 = Flagot, G18 = Tesfa, G19 = Hiber-1, and G20 = Abay). 3.2.4. Evaluation of environments relative to ideal environments An ideal environment should satisfy two conditions at the same time. These distinctly differentiate and discriminate the genotypes, and the representativeness for the target environments (Yan, 2010). Discriminating refers to an environment’s ability to maximize the variance among candidate genotypes in a study (Blanche and Myers, 2006). An ideal trial site can effectively screen genotypes that have high and stable yields. In GGE biplot graph, the small circle stands for an ideal environment, which depends on the mean coordinates of all test environments. There has been a positive correlation between the environment vector length and the environment discriminating ability while there has been negative correlation between the angle existing in environment vector with the ideal environment and the environment’s representativeness of the target environment (Yan, 2010). Accordingly, Figure 5 shows that the discriminating ability and the best representative environments for teff varieties was in declining order E2 (Adet year-1), followed by E7 (Adet year-2), E9 (Bichena year-2), E4 (Bichena year-1), E1 (Debre-Tabor year-1), E6 (Debre-Tabor year-2), E8 (Motta year-2), E3 (Motta year-2), E10 (Takusa year-2), and E5 (Takusa year-1). A test environment having a small angle with the average environmental axis is said to be more representative of other test environments (Yan and Tinker, 2006). In the present study, therefore, Adet (E2) which fell into the center of concentric circle and had the longest vector length and the smallest angle with the average environmental axis was identified to be the most Atinkut et al. Genotype x Environment Interaction and Yield Stability of Teff 87 representative of all test environments. Hence, Adet and Bichena are relatively ideal locations for teff cultivation among the test environments. In agreement with this finding, Habite Jifar et al. (2019) reported that Adet and Axum are relatively representative environments among test environments for teff production, but on the contrary, same researchers reported that they were not discriminative environments. Mahdieh et al. (2016) also reported that a testing environment has less power to discriminate genotypes when located far away from the center concentric circle or to an ideal environment. Hence, in connection to our result, Motta and Takusa testing locations, which are located far away from the center concentric circle, are considered as less powerful to discriminate genotypes. Figure 5. GGE-biplot based on environment-focused scaling for comparison of environment with the ideal environments (E1 = Debre-Tabor year-1, E2 = Adet year-1, E3 = Motta year-1, E4 = Bichena year-1, E5 = Takusa year-1, E6 = Debre- Tabor year-2, E7 = Adet year-2, E8 = Motta year-2, E9 = Bichena year-2, E10 = Takusa year-2, G1 = Wellenkomi, G2 = Tseday, G3 = Dukem, G4 = Ziquala, G5 = Ambotoke, G6 = Areka-1, G7 = Dega-Tef, G8 = Quncho, G9 = Etsub, G10 = Simada, G11 = Boset, G12 = Kora, G13 = Werekiyu. G14 = Abola, G15 = Dagim, G16 = Negus, G17 = Flagot, G18 = Tesfa, G19 = Hiber-1, and G20 = Abay). 3.3.5. Mean grain yield and stability performance of genotypes Ranking of twenty teff varieties based on mean yield performance and stability is presented in Figure 6. The single arrow line passing through the biplot origin and the average environment indicated by the small circle is the average environments coordinate (AEC) axis, which is defined by the average PC1 and PC2 scores of all environments (Yan and Kang, 2003). This line points towards higher mean yield across environments. Hence, in the present biplot, G19 gave the highest mean yield followed by G9, G12, G3, G1, G8, G5, G15 and G17. The remaining genotypes had bellow grand mean yield (Figure 6). In lien with this result Sewagegne Tariku et al. (2020) reported that Hiber-1 is the highest mean grain yield and relatively stable variety in teff variety verification trial. The line which passes through the biplot origin and perpendicular to the AEC axis shows the measure of stability. Either direction away from the biplot origin, on this axis, indicates greater Genotype x environment interaction and poor stability or vice versa (Kaya et al., 2006). Thus, in terms of stability the genotypes ranked as G17 > G12 > G18 > G8 > G5 > G19 > G11 > G20 > Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 88 G2 > G15 > G4 > G14 > G3 > G16 > G1 > G7 > G9 > G6 > G10 > G13 (Figure 6). Stability was reported to have lower heritability than mean performance (Eskridge, 1996) hence; it is useful only when considered jointly with mean performance. Yan and Tinker (2006) also noted that stability refers to the relative performance of a genotype and it is meaningful only when associated with mean performance. Figure 6. GGE-biplot based on environment-focused scaling for comparison of environment with the ideal environments (E1 = Debre-Tabor year-1, E2 = Adet year-1, E3 = Motta year-1, E4 = Bichena year-1, E5 = Takusa year-1, E6 = Debre- Tabor year-2, E7 = Adet year-2, E8 = Motta year-2, E9 = Bichena year-2, E10 = Takusa year-2, G1 = Wellenkomi, G2 = Tseday, G3 = Dukem, G4 = Ziquala, G5 = Ambotoke, G6 = Areka-1, G7 = Dega-Tef, G8 = Quncho, G9 = Etsub, G10 = Simada, G11 = Boset, G12 = Kora, G13 = Werekiyu. G14 = Abola, G15 = Dagim, G16 = Negus, G17 = Flagot, G18 = Tesfa, G19 = Hiber-1, and G20 = Abay). Atinkut et al. Genotype x Environment Interaction and Yield Stability of Teff 89 4. Conclusions Analysis of GEI is necessary to determine the stability and performance of varieties across different environments. The results of the combined analysis of variance in this study have demonstrated that teff grain yield and plant height, panicle length, dry biomas, days to heading and maturity were significantly affected by environment (E), followed by G x E interaction and genotype (G) effects, respectively. The results of the research revealed that the varieties Hiber-1 (G19), Kora (G12), Etsub (G9) and Dukem (G3) were found to be good genotypes based on their yield performance as well as stability. On the other hand, Simada (G10), Tesfa (G18), Tseday (G2) and Areka-1(G6) were found as unstable and low yielding genotypes. Thus, Hiber-1 could be recommended for wide cultivation across the areas of north western Ethiopia because of its high yield potential and yield stability. The results have also demonstrated that the variety Kora for Bichena, Etsub for Debre-Tabor and Takusa and Werekiyu for Motta could be potentially productive for specific adaptation to boost grain production of the crop. Furthermore, the present study revealed the existence of three mega- environments and four teff genotype groups in north western parts of Ethiopia. Environment, Adet, and Bichena have the longest vector length and the smallest angle with average environmental axis was the most discriminating and representative of all test environments, respectively. It could, thus, be concluded that Adet and Bichena districts are the best locations for teff production and Hiber-1 is the best teff variety to be produced in the region. 5. Acknowledgements The authors are grateful to Adet Agricultural Research Center for providing financial supports and teff breeding staff of Adet and Debre-Tabor Research sub-center for technical supports. 6. References Ayalneh Tilahun, Habtamu Zeleke and Amsalu Ayana. 2012. Genetic variability, heritability and genetic advance in tef [Eragrostis tef (Zucc.) Trotter] lines in Sinana and Adaba. International Journal of Plant Breeding and Genetics, 6(1): 40–60. Basford, K.E. and Cooper, M. 1998. Genotype x environment interactions and some considerations of their implications for Wheat breeding in Australia. Australia Journal of Agricultural Research, 49(2): 153–174. Blanche, S.B. and Myers, G.O. 2006. Identifying discriminating locations for cultivar selection in Louisiana. Crop Science, 46 (2): 946–949. CSA (Central Statistical Agency). 2018. Central Statistical Agency, Agricultural Sample Survey 2017/2018 (2010 E.C.). Report on Area and Production of Major Crops (Private Peasant Holdings, Meher Season), Volume I. Statistical Bulletin 586. Addis Ababa, Ethiopia. Pp. 14–18 Delacy, I.H., Basford, K.E., Cooper, M. and Bull, J.K. 1996. Analysis of multi environment trials and historical perspective. In: Cooper, M. and Hammer, G.L. eds.). Plant Adaptation and Crop Improvement. CAB International. Eberhart, S.A. and Russell, W.A. 1966. Stability parameters for comparing varieties. Crop Science, 6 (1): 36–40. Eskridge, K.M. 1996. Analysis of multiple environment trials using the probability of outperforming a check. Pp. 273–307. In: Kang, M.S. and Gauch Jr. H.G. (eds.). Genotype x environment interaction. CRC Press, Boca Raton, FLorida. Fan, X.M., Kang, M.S., Chen, H., Zhang, Y., Tan, J. and Xu, C. 2007. Yield stability of maize hybrids evaluated in multi-environment trials in Yunnan, China. Agronomy Journal, 99(1): 220–228. Farshadfar, E., Mohammadi, R., Aghaee, M. and Vaisi, Z. 2012. GGE biplot analysis of genotype x environment interaction in Wheat-Barley disomic addition lines. Australian Journal of Crop Science, 6(6): 1074–1079. Fekadu Gurmu, Hussein Mohammed and Getinet Alemaw. 2009. Genotype x environment interaction and stability of Soybean for grain yield and nutritional quality. African Crop Science Journal, 17(2): 87–99. Fufa Hundera. 1998. Variations of morph-agronomic characters and grain chemical composition of released varieties of tef [Eragrostis tef (Zucc.) Trotter]. Journal of Genetics and Breeding, 52(4): 307– 311. Gauch, H.G. and Zobel, R.W. 1988. “Predictive and postdictive success of statistical analysis of yield trials”. Theoretical and Applied Genetics, 76(1): 1–10. Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 90 Gomez, K.A. and Gomez, A.A. 1984. Statistical Procedures for Agricultural Research. 2nd edition. John Willey and Sons, Toronto, Canada. Pp. 680. Habte Jifar, Kebebew Assefa, Kassahun Tesfaye, Kifle Dagne and Zerihun Tadele. 2019. Genotype x environment interaction and stability analysis in grain yield of improved Tef (Eragrostis tef) varieties evaluated in Ethiopia. American Journal of Experimental Agriculture, 35(5): 1–13. Karimizadeh, R., Mohammadi, M., Sabaghni, N., Mahmoodi, A.A., Roustami, B., Seyyedi, F. and Akbari, F. 2013. GGE biplot analysis of yield stability in multi-environment trials of lentil genotypes under rainfed condition. Notulae Scientia Biologicae, 5(2): 256–262. Kaya, Y., Akcura, M. and Taner, S. 2006. GGE-biplot analysis of multi-environment yield trials in bread wheat. Turkey Journal of Agriculture, 30: 325–337. Legesse Kassa, Marie, F. and Fufa, Hundra. 2006. Stability analysis of grain yield of tef (Eragrostis tef) using the mixed model approach. South African Journal of Plant and Soil, 23(1): 38–42. Mahdieh, R., Alireza, P., Hamid Reza, B. and Fariborz, S. 2016. Grain yield stability analysis of soya bean genotypes by AMMI method. Azarian Journal of Agriculture, 3(6): 119–128. Mathewos Ashamo and Getachew Belay. 2012. Genotype x environment interaction analysis of tef grown in southern Ethiopia using additive main effects and multiplicative interaction model. Journal of Biology, Agriculture and Healthcare, 2(1): 66–72. MOA (Ministry of Agriculture). 2018. Crop Variety Register No.21. Ministry of Agriculture, Plant Variety Release, Protection, and Seed Quality Control Directorate, Addis Ababa, Ethiopia. Mohammadi, R., Pourdad, S.S. and Amri, A. 2008. Grain yield stability of Spring Safflower (Carthamus tinctorius L.). Australian Journal of Agricultural Research, 59: 546–553. SAS (System Analysis Software) Institute. 2002. Proprietary Software version 9.00, Cary, North Carolina, USA. Sewagegne Tariku, Atinkut Fentahun and Ataly Fentahun. 2018. Participatory evaluation and selection of improved tef varieties in a researcher - managed trial in western Ethiopia. International Journal of Emerging Technology and Advanced Engineering, 8(5): 28–42. Sewagegne Tariku, Atinkut Fentahun, Gedefaw Misganaw, Atalay Fentahun, Tesfaye Alemayehu, Abiy Legesse and Solomon Mitiku. 2020. GGE biplot analysis and agronomic performance of tef genotypes in moisture deficit stress areas of eastern Amhara Regional State. Blue Nile Journal of Agricultural Research, 1(2): 1–17. Shrestha, S.P., Asch, F., Dusserre, J., Ramanantsoanirine, A. and Brueck, H. 2012. Climate effects on yield components as affected by genotypic responses to variable environmental conditions in Upland rice systems at different altitudes. Field Crops Research, 134: 216–228. Tiruneh Kefyalew, Hailu Tefera, Kebebew Assefa and Mulu Ayele. 2000. Phenotypic diversity for qualitative and phenologic characters in germplasms collections of tef (Eragrostis tef). Genetic Resource and Crop Evolution, 47: 73–80. VSN International. 2015. Genstat Software Reference Manual (Release 18), Part 1 Summary. VSN International, Hemel Hempstead, UK. Wondewosen Shiferaw, Alemayehu Balcha and Husien Mohammed. 2012. Genetic variation for grain yield and yield related traits in tef [Eragrostis tef (Zucc.) Trotter] under moisture stress and non- stress environments. American Journal of Plant Sciences, 3(8): 1041–1046. Yan, W., Hunt, L.A., Sheng Q. and Szlavnics, Z. 2000. Cultivar evaluation and mega environment investigation based on the GGE biplot, Crop Science, 40(3): 597–605. Yan, W. 2001. GGE biplot a Windows application for graphical analysis of multi-environment trial data and other types of two way data. Agronomy Journal, 93(5): 1–11. Yan, W. 2002. Singular-value partitioning in biplot analysis of multi-environment trial data. Agronomy Journal, 94(5):990–996. Yan, W. and Kan, M.S., editors. 2003. GGE biplot analysis: A graphical tool for breeders, geneticists and agronomist. CRC Press, Boca Raton, Florida. Yan, W. and Tinker, N.A. 2006. Biplot analysis of multi- environment trial data: Principles and application. Canadian Journal of Plant Science, 86(3): 623–645. Atinkut et al. Genotype x Environment Interaction and Yield Stability of Teff 91 Yan,W.K. 2010. Optimal use of biplot in analysis of multi-location variety test data. Acta Agronomy Sinica, 36(11): 1805–1819. Yifru Teklu and Hailu Tefera, 2005. Genetic improvement in grain yield potential and associated agronomic traits of tef (Eragrostis tef). Euphytica, 141: 247–254. Yirga Belay. 2016. Genotype x environment interaction and yield stability of white seeded sesame (Sesamum indicum L.) genotypes in Northern Ethiopia. M.Sc. Thesis, Haramaya University, Ethiopia. Atinkut et al. East African Journal of Sciences Volume 16(1): 77–92 92