Yield Performance of Breeding Lines
The grain yield performance of 70 genotypes for every environment were separately demonstrated in Table 1. Mean yield performances were listed with respective genotypic code (G1 - G70) across the environments. In Cumilla, G61 (7.41 t/ha) was the highest yielder followed by G58, G38, G29, G17, G41 etc. G68 (3.3 t/ha) was the lowest yielder and followed by G33, G60, G70, G53 etc. In Gazipur, G41 (7.6 t/ha) was the highest yielder followed by G24, G42, G22, G32, G1 etc. G70 (3.4 t/ha) was the lowest yielder and followed by G2, G13, G17, G25 etc. In Rajshahi, G9 (7.08 t/ha) was the highest yielder followed by G9, G19, G50, G12, G26, G44, G24 etc. G47 (4.32 t/ha) was the lowest yielder genotype and followed by G11, G13, G14, G33, G45, G27 etc. (Table 1). These genotypes were suitable for specific environment by considering only the grain yield performance excluding G × E interaction.
Table 1. Yield (t/ha) performance of 70 genotypes across three breeding zones
Genotypes
|
C
|
G
|
R
|
M
|
Genotypes
|
C
|
G
|
R
|
M
|
IR 107971-B-B RGA-B RGA-20 (G1)
|
5.10
|
6.69
|
5.63
|
5.81
|
IR 107971-B-B RGA-B RGA-14 (G36)
|
6.19
|
5.82
|
5.56
|
5.86
|
IR 104006:4-15-B RGA-B RGA-30 (G2)
|
5.18
|
4.00
|
5.35
|
4.84
|
IR 103310-B-B RGA-B RGA-13 (G37)
|
5.87
|
5.53
|
6.08
|
5.83
|
IR 107982-B-B RGA-B RGA-164 (G3)
|
5.87
|
4.92
|
5.28
|
5.35
|
IR 107971-B-B RGA-B RGA-11 (G38)
|
7.13
|
4.74
|
5.10
|
5.66
|
IR 107982-B-B RGA-B RGA-105 (G4)
|
6.56
|
4.93
|
5.53
|
5.67
|
IR 108000-B-B-B-B-34 (G39)
|
6.29
|
5.38
|
6.22
|
5.96
|
IR 106457-B-B RGA-B RGA-335 (G5)
|
5.94
|
6.10
|
5.47
|
5.84
|
IR 108000-B-B-B-B-27 (G40)
|
6.06
|
5.01
|
5.12
|
5.39
|
IR 107995-B-B RGA-B RGA-441 (G6)
|
6.50
|
6.04
|
5.92
|
6.15
|
IR 103309-B-B RGA-B RGA-347 (G41)
|
6.95
|
7.63
|
6.18
|
6.92
|
IR 107976-B-B RGA-B RGA-175 (G7)
|
5.83
|
6.10
|
6.32
|
6.08
|
IR 106436-B-B RGA-B RGA-B RGA-275 (G42)
|
5.63
|
6.97
|
4.90
|
5.83
|
IR 107976-B-B RGA-B RGA-154 (G8)
|
5.42
|
6.31
|
5.99
|
5.90
|
IR 103757-B-B RGA-B RGA-B RGA-104 (G43)
|
5.32
|
5.97
|
5.53
|
5.61
|
IR 106449-B-B RGA-B RGA-B RGA-118 (G9)
|
5.80
|
6.45
|
7.08
|
6.44
|
IR 107989-B-B RGA-B RGA-153 (G44)
|
6.21
|
5.88
|
6.39
|
6.16
|
IR 100707-B-B RGA-B RGA-B RGA-224 (G10)
|
5.34
|
4.86
|
5.83
|
5.34
|
IR 107971-B-B RGA-B RGA-140 (G45)
|
6.92
|
5.21
|
4.67
|
5.60
|
IR 107995-B-B RGA-B RGA-487 (G11)
|
5.51
|
4.94
|
4.43
|
4.96
|
IR 103297-B-B-B-B-B-60 (G46)
|
5.86
|
5.63
|
5.40
|
5.63
|
IR 106449-B-B RGA-B RGA-B RGA-103 (G12)
|
6.49
|
5.81
|
6.50
|
6.27
|
IR 100707-B-B RGA-B RGA-B RGA-149 (G47)
|
5.56
|
5.57
|
4.32
|
5.15
|
IR 106449-B-B RGA-B RGA-B RGA-127 (G13)
|
5.45
|
4.24
|
4.56
|
4.75
|
IR 103314-B-B RGA-B RGA-171 (G48)
|
5.44
|
5.89
|
5.27
|
5.53
|
IR 107995-B-B RGA-B RGA-56 (G14)
|
5.64
|
6.38
|
4.64
|
5.55
|
IR 107982-B-B RGA-B RGA-108 (G49)
|
6.63
|
4.50
|
5.21
|
5.44
|
IR 100821-B-B RGA-B RGA-B RGA-32 (G15)
|
6.14
|
4.48
|
5.60
|
5.41
|
IR 107982-B-B RGA-B RGA-149 (G50)
|
5.71
|
5.52
|
6.51
|
5.91
|
IR 100707-B-B RGA-B RGA-B RGA-134 (G16)
|
6.01
|
6.54
|
5.26
|
5.94
|
IR 103757-B-B RGA-B RGA-B RGA-18 (G51)
|
6.61
|
5.44
|
5.91
|
5.98
|
IR 106449-B-B RGA-B RGA-B RGA-122 (G17)
|
7.01
|
4.32
|
5.41
|
5.58
|
IR 103314-B-B RGA-B RGA-182 (G52)
|
5.74
|
6.69
|
6.31
|
6.25
|
IR 107989-B-B RGA-B RGA-2 (G18)
|
6.33
|
5.89
|
5.42
|
5.88
|
IR 99129-B RGA-B RGA-213 (G53)
|
5.00
|
4.63
|
5.82
|
5.15
|
IR 107982-B-B RGA-B RGA-152 (G19)
|
6.06
|
6.25
|
6.83
|
6.38
|
IR 99129-B RGA-B RGA-138 (G54)
|
6.24
|
4.55
|
4.95
|
5.25
|
IR 103292-B-B-B-B-B-14 (G20)
|
5.68
|
6.46
|
5.38
|
5.84
|
IR 106436-B-B RGA-B RGA-B RGA-182 (G55)
|
6.51
|
5.81
|
5.19
|
5.84
|
IR 106457-B-B RGA-B RGA-336 (G21)
|
6.67
|
4.97
|
5.58
|
5.74
|
IR 106449-B-B RGA-B RGA-B RGA-288 (G56)
|
5.25
|
6.49
|
5.59
|
5.78
|
IR 100821-B-B RGA-B RGA-B RGA-50 (G22)
|
6.80
|
6.84
|
6.17
|
6.60
|
IR 107976-B-B RGA-B RGA-152 (G57)
|
6.47
|
5.79
|
5.85
|
6.04
|
IR 108000-B-B-B-B-16 (G23)
|
5.20
|
6.54
|
5.82
|
5.85
|
IR 106436-B-B RGA-B RGA-B RGA-132 (G58)
|
7.25
|
6.65
|
5.16
|
6.35
|
IR 107976-B-B RGA-B RGA-254 (G24)
|
5.59
|
7.07
|
6.35
|
6.34
|
IR 100707-B-B RGA-B RGA-B RGA-172 (G59)
|
5.66
|
6.67
|
5.65
|
5.99
|
IR 108000-B-B-B-B-31 (G25)
|
6.84
|
4.48
|
5.38
|
5.57
|
IR 107971-B-B RGA-B RGA-15 (G60)
|
4.50
|
5.71
|
5.87
|
5.36
|
IR 107995-B-B RGA-B RGA-14 (G26)
|
6.90
|
6.01
|
6.45
|
6.45
|
IR 107982-B-B RGA-B RGA-162 (G61)
|
7.41
|
6.27
|
5.22
|
6.30
|
IR 100707-B-B RGA-B RGA-B RGA-185 (G27)
|
5.57
|
5.89
|
4.88
|
5.45
|
IR 107989-B-B RGA-B RGA-141 (G62)
|
6.11
|
5.74
|
5.86
|
5.90
|
IR 103718-B-B RGA-B RGA-B RGA-B RGA-137 (G28)
|
6.10
|
5.12
|
5.70
|
5.64
|
IRRI 156 (G63)
|
5.68
|
5.41
|
5.63
|
5.57
|
IR 108000-B-B-B-B-33 (G29)
|
7.12
|
5.76
|
5.54
|
6.14
|
IRRI 174 (G64)
|
5.72
|
4.77
|
5.09
|
5.19
|
IR 107989-B-B RGA-B RGA-124 (G30)
|
6.45
|
5.74
|
5.07
|
5.75
|
IRRI 154 (G65)
|
5.11
|
6.24
|
5.61
|
5.65
|
IR 108000-B-B-B-B-30 (G31)
|
5.63
|
6.30
|
5.89
|
5.94
|
IRRI 104 (G66)
|
5.48
|
5.99
|
5.07
|
5.51
|
IR 103310-B-B RGA-B RGA-201 (G32)
|
5.43
|
6.70
|
5.88
|
6.00
|
IRRI 181 (G67)
|
6.19
|
5.32
|
6.28
|
5.93
|
IR 103757-B-B RGA-B RGA-B RGA-102 (G33)
|
3.57
|
6.05
|
4.66
|
4.76
|
BRRI dhan28 (G68)
|
3.30
|
4.53
|
6.33
|
4.72
|
IR 107995-B-B-B-B-4 (G34)
|
5.54
|
6.01
|
6.07
|
5.87
|
BRRI dhan67 (G69)
|
6.25
|
4.98
|
6.04
|
5.75
|
IR 107995-B-B RGA-B RGA-485 (G35)
|
5.29
|
6.09
|
5.79
|
5.72
|
BRRI dhan81 (G70)
|
4.92
|
3.44
|
5.51
|
4.62
|
Notes: Cumilla (C), Gazipur (G), Rajshahi (R), Mean (M).
AMMI Analysis of Variance (ANOVA)
AMMI analysis of variance of 70 genotypes was tested over three environments (Table 2). The higher value of sum of squares for genotype represented that the genotypes showed more variances in the level of mean performances. The lower value of sum of squares for environment represented that the genotypes showed less variances in the level of mean performances.
Table 2. AMMI analysis of variance for yield performance of 70 genotypes across three environments
Sources of variation
|
DF
|
Sum sq
|
Mean sq
|
F value
|
Pr (>F)
|
Environment
|
2
|
6.727
|
3.363
|
1.796
|
0.307
|
Replication
|
3
|
5.618
|
1.873
|
2.703
|
0.047 *
|
Genotype
|
69
|
86.550
|
1.254
|
1.810
|
0.0001 ***
|
Environment × Genotype
|
138
|
132.311
|
0.959
|
1.384
|
0.017 *
|
Residuals
|
207
|
143.448
|
0.693
|
|
|
Significant codes: ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05
The magnitude of sum of squares for environment × genotype was higher representing that the response of genotypes was different across environments. The variances of genotype were significant at 0.1 % level of significance. The variances of environment × genotype and replication were significant at 5 % level of significance.
AMMI 1 Biplot Analysis
The first primary component of interaction (IPC1) factor was responsible for 64.2 % variation due to G × E interaction. A biplot between mean yield performance and primary component of interaction (IPC1) of 70 genotypes (replicated) across three environments was shown in Fig. 1. High mean yield and lower IPCA1 scores (near zero) should be the characteristics of an ideal genotype. The similar characteristics are suggested for an ideal environment. The genotypes possessing lower AMMI1 value or IPCA1 scores are stable across environments. A genotype having both lower AMMI1 value and high yield potential is more desirable. These genotypes have the potential to give stable performance. G41 (IR 103309-B-B RGA-B RGA-347) was the highest yielder (6.92 t/ha) and followed by G22 (6.6 t/ha), G26 (6.45 t/ha), G9 (6.44 t/ha), G19 (6.38 t/ha), G58 (6.35 t/ha), G24 (6.34 t/ha) etc. On the other hand, G70 (BRRI dhan81) was the lowest yielder (4.62 t/ha) and followed by G68 (4.72 t/ha), G13 (4.75 t/ha), G33 (4.76 t/ha), G2 (4.84 t/ha), G11 (4.96 t/ha), G53 (5.15 t/ha), G47 (5.15 t/ha) etc. G22 (6.6 t/ha) was best considering AMMI1 value and yield performance. G5, G10, G14, G27, G30, G46, G47, G50, G53, G55, G63, G67 etc. all were more stable genotypes due to their lower IPCA1 scores (near 0) (Table 1, Fig. 1). The genotypes close to a particular environment are more suitable for that respective environment than others. G35, G65, G43, G27 etc. were more suitable for Rajshahi.
AMMI 2 Biplot Analysis
G × E interaction patterns can be demonstrated through biplot by using the AMMI method that displays the genotypes and environments concurrently. The genotypes location in the biplot far from the origin contribute comparatively more to the G × E interaction than those lying very close to origin or in the center of the origin (Mukherjee et al., 2013). The second primary component (PC2) factor accounted for 35.8% variation of the G × E interaction. These two-primary component (PC1 and PC2), all together accounted for 100% variation of the G × E interaction. The contribution of G68 was highest to the interaction. G70, G58, G42, G61, G45, G38, G14, G33, G60, G53, G9 had comparably more contribution than other genotypes to the interaction (Fig. 2).
On the other hand, the genotypes were in the center of the biplot or very near to the center representing stable performance across environments. Based on the positive or negative signs of the scores of the first two primary components (PC1 and PC2), the biplot represented four sectors. Positive PCA1 and PC2 scores by sector1, positive PCA1 and negative PCA2 scores by sector2, negative PCA1 and PCA2 scores by sector3 and negative PC1 and positive PC2 scores by sector4 were represented. The genotypes were distributed in these four sectors and environments were in three sectors. There was no environment in sector1, Cumilla in sector2, Gazipur in sector3 and Rajshahi in sector4. The ranking of environments based on mean yield performance was Rajshahi < Gazipur < Cumilla.
The genotypes located in the sector2 were relatively higher yielder and followed by sector1, sector 3, sector4. Not only the G × E interaction but also the ‘which win where’ pattern was displayed by AMMI2 biplot. The genotypes located far from the center were unstable across the environments. Lower value of PC1 scores indicated the lower interaction value and less variation. The genotypes having PC1 scores less than 0 reacted positively to the environments which had PC1 scores less than 0 but reacted negatively to the environments which had PC1 scores more than 0. G33 (IR 107995-B-B-B-B-4) having PC1 scores less than 0 reacted positively to Gazipur having PC1 scores less than 0 but reacted negatively to Cumilla. So, the interaction was positive for Gazipur and that genotype was more adaptive to the respective environment (Gazipur). The similar stability analysis was done through AMMI by Islam et al. (2015).
GGE Biplot Analysis
Environment can severely affect the genotypic performance (Islam et al., 2015). The length of the environment vectors can be visualized by the concentric circles of the GGE biplot. It is the measures of the ability of discrimination of the environment and proportional to the standard deviation within the respective environment (Negash et al., 2013). Ideal environments and stable genotypes locating on the concentric circles can be selected (Yan, 2002). Most desirable environment can be identified by observing the location, very near to the center circle or an ideal environment (Rad et al., 2013). GGE biplot analysis for grain yield of 70 genotypes over three breeding zones were conducted. It accounted for 81.74% variation due to G × E interaction comprising two principal components (PC1 and PC2) 45.62% and 36.12% respectively. The ranking of environments was done for selecting suitable environment for high yield potential and adaptability (Fig. 3). There was no environment in the center of the circle and so, there was no ideal environment. The distance from the center of circle was lower for Gazipur and followed by Rajshahi and Cumilla. Therefore, Gazipur was more suitable than others. By considering lower PC scoring, Rajshahi was more stable environment than others. These three environments were located on the two sectors. Gazipur and Rajshahi were in same sector whereas Cumilla was in another sector. So, Gazipur and Rajshahi showed similarity in performances. Highest mean yield performance and stability (PC scores close to 0) are the characteristics of an ideal genotype. The ideal genotype can perform with highest performance without the G × E interaction and so called most adaptable genotype (Akcura et al., 2011). It is expected that the best genotype should have high yield and stable performance across all environments. But such genotypes are hardly found. Hence, genotypes are evaluated by their high yielding performance with relatively stable over environments (Yan and Tinker, 2006). The ranking of 70 replicated genotype was done by using the phenomena that the genotype having highest mean yield performance and stability across environments. G41 locating in the center of the circle represented that highest mean yield performance and stability than the rest genotypes. It was the ideal genotype. G22, G58, G26, G61, G29, G24 etc. were more suitable genotype locating very near to the ideal genotype or around the center of the circle for yield and stability (Fig. 4).
Moreover, G70, G68, G33, G60, G2, G13, G53, G11 etc. were located far from the ideal genotype and so, they were low yielder and unstable genotypes. These are not suitable for further crop improvement program. The genotypes having less PC scores were stable across the environments. Most of the genotypes (e.g. G24, G7, G32 etc.) were more suitable for Rajshahi. No genotypes were found specifically suitable for Gazipur and Cumilla (Fig. 4). ‘Which-won-where’ and GEI (G × E interaction) pattern of grain yield trial of multiple environment were represented by the polygon derived from GGE-biplot. The polygon was made by joining the genotypes that were located far from the origin and the rest genotypes falling within the polygon. Erect lines were drawn from the origin to the side of polygon and extended beyond the polygon. Thus, seven lines divided the biplot into eight sectors and two environments (Rajshahi and Gazipur) fell into same sector. Cumilla was in another sector (Fig. 5). The vertex genotype of each sector represented best performer of the particular environment. From those eight sectors, two mega environments were selected. The first mega environment comprising two environments (Rajshahi and Gazipur) with G41 as the highest yielder genotype in those environment. Rajshahi was more stable whereas Gazipur was unstable environment. The performance of G22, G26, G58, G44 etc. were better in the specific environment than others.
The second mega environment, Cumilla was located in another sector with G61 as highest yielder genotype in that environment. G29 was also a good yielder than others in the particular environment. But Cumilla was an unstable environment. The rest sectors with the vertex genotype e.g. G38, G17, G70, G68, G33 had no environment. Those genotypes were not highest yielder at any environment. However, the genotypes within the polygon locating very near to the origin were less responsive to the environment than the vertex genotypes. The similar analysis was done for stability analysis through GGE biplot for ‘Which-won-where’ by Islam et al. (2015).