Envirotyping
From the environmental covariates we performed the similarity analysis considering 14 enviromic kernels (Fig. 01), where we can visualize the formation of four groups (mega-environments), which we worked on individually. In the similarity analysis, we visualize that environments E6 (São Gotardo* - dry) and E8 (Sete Lagoas) are contrasting in relation to the other environments that did not group with any other environment, where such environments represent a mega-environment. On the other hand, environments E3 (Viçosa irrigated 2019), E4 (Viçosa irrigated 2020) and E5 (Viçosa dry 2020) have greater similarity for the environmental covariates analyzed, indicating that the relative performance of genotypes in a same city did not change as a function of the water regime. This indicates that the variance components of the G × E interaction between these environments are not significant.
The environment E1 (Rio Paranaíba - irrigated 2018), E2 (Rio Paranaíba - Irrigated 2019) and E7 (São Gotardo irrigated 2020) have greater similarity, these environments formed another subgroup, possibly explained by the proximity between the three environments.
Figure 02 shows the biplot for the Principal Component Analysis between environmental variables, grain yield, and genotypic variation in the environments. The two principal component axes explained 88.6% of the total variation. There was a positive correlation between TMED, TMAX, SPV, and ETP, and a negative correlation of these climate variables with RH.
The PCA scores suggest that E3, E4, E5, and E8 were mainly characterized as having a higher TMD, TMAX, TMIN, a lower RH, and consequently a higher VPD, which is the difference (deficit) between the volume of moisture in the air and how much moisture the air can hold when it is saturated. The VPD can be then used as an accurate indicator of current air evaporative capacity. The combination of higher temperatures with high VPD resulted in higher rates of ETP observed for E4, E5, and E8. Therefore, it can be concluded that in warmer environments, the loss of water by evapotranspiration is greater than in cooler environments. This is supported by the positive association between ETP and environment temperatures (mean, minimum, and maximum).
The positive association of TMAX, TRANGE, and VPD and the negative association of RH with WAASB suggest that these environmental variables play a key role in the manifestation of genotypic variation in each environment. Therefore, good prospects for the selection of climate-resilient wheat genotypes for the Cerrado are expected.
Figure 03 shows the distribution of the maximum air temperature within each mega environment. It is observed that maximum temperature (TMAX ºC d− 1) showed a variable distribution, especially in mega environment four, which showed low average grain yield values (see the distribution for other variables in Supplementary material S1).
Investigating the maximum air temperature in each of the five development stages (tillering, boosting, final flowering, kernel milk stage and physiological maturity) it was observed that the higher temperatures of the E8 (Fig. 03) were mainly observed during boosting, heading/flowering, and kernel milk stage with temperatures between 29 and 39ºC during more than 75% of the days in these stages. On the other hand, environment E1 (Rio Paranaíba -irrigated), in which the wheat genotypes showed the highest BLUP for grain yield, we were able to analyze in the workflow that the relative frequency of maximum air temperature in the development stages was reduced, which contributed to the excellent results obtained in this environment. Similar results for grain yield and maximum air temperature, we can observe for environments E3 (Viçosa irrigated 2019), E4 (Viçosa irrigated 2020) and E7 (São Gotardo irrigated).
Wheat is cereal of the most important food crops worldwide in terms of production and human nutrition. Brazil is a major consumer of wheat, with an approximate consumption of 12 million tons (Conab 2021), it is not self-sufficient and produces only 50% of the demand. Brazil has a high potential for expanding the cultivation of wheat, especially into regions of lower latitude, such as the Brazilian Cerrado (Casagrande et al. 2020). However, in the Brazilian Cerrado we need to develop and release cultivars with tolerance to heat, once those high temperatures same that for short periods have the potential to significantly reduce grain production (Nuttal et al. 2018). In the research by Hofmann et al. (2021) the importance and potential of the Brazilian Cerrado is highlighted with a prediction of maximum air temperature for 2050. The authors highlighted an increase in maximum temperature of 4.0, 3.4, 4.4, and 6.0 ºC in the months of July, August, September and October, respectively. During these months, the wheat crop will be in full development in the field in this region. In parallel, the increase in maximum air temperature, the study also indicates a ~ 15% decrease in relative humidity.
Of the eight environments evaluated, in three (E6, E8 and E5) the wheat genotypes showed grain yields below the general average of the other environments, where E8 and E5 were environments with maximum air temperatures greater than 29°C at critical development stages, such as heading/flowering. The high temperature at anthesis or after anthesis decreased chlorophyll content, individual grain weight and grain yield of wheat (Pradham et al. 2012). High temperature stress decreases chlorophyll content as a result of damaged thylakoid membranes or lipid peroxidation of chloroplast membranes (Ristic et al. 2007; Djanaguiraman et al. 2010).
With these data, we understand that the development cycle of the genotypes was accelerated in this environment, which reduces the time for the accumulation of photo-assimilated reserves in the stalk and leaves that would be translocated to the grain later. High temperature, especially above 34°C at flowering, as we observed, increases flower sterility in the spikelets, reducing the number of grains per spike and thus grain yield. Once, that reduced grain number may only be partially compensated by increased grain filling due to proportionally greater allocation of assimilate to the remaining kernels (Jenner et al. 1991; Nuttall et al. 2018). At growth stage (double ridge), high temperature adversely affects spikelet formation (Shpiler and Blum 1986), at meiosis decreases grain number per spike by inducing ovule and pollen sterility and anther indehiscence (Prasad et al 2008a), at anthesis, stress decreases the grain number by adversely affecting ovarian development, pollen germination and pollen tube growth (Prasad et al. 2008b) At grain filling period, high temperature decreases leaf chlorophyll content and accelerates senescence (Zhao et al. 2007), leading to a shorter grain filling duration with an ultimate decrease in individual grain weight and yield (Pradhan et al 2012).
Based on these results we can also highlight the huge importance of considering the application of envirotyping in understanding the GxE interaction (Heinemann et al. 2022), for deciphering environmental impacts of covariable climates on wheat genotypes. Wheat genotypes vary significantly in their sensitivity to high temperature, where many studies have showed significant differences in the response in heat tolerance (Pradhan et al. 2012; Farooq et al. 2011; Stone and Nicolas 1998). Thus, considering the impact that maximum air temperature has on the wheat crop and that the expansion area in Brazil will have an increase in maximum temperature, we need to select wheat genotypes that are resilient to these environmental conditions. Currently, a first step can be taken by analyzing the genetic variability of germplasm and recommending wheat genotypes that are more resilient to higher temperature environments.
Genotypic mean performance
The BLUPs represented in Table 04 refer to the mean genotypic value in the various environments and capitalizes the average interaction with all evaluated environments. There is a high environmental variability according to the estimates of the genotypes evaluated. The environment with the highest average was E1, in Rio Paranaíba 2018, followed by the environments E3 and E7, Viçosa 2019 and São Gotardo 2020, with mean yields per environment of 5,029.03, 4,292.80, and 4,265.80 kg ha− 1, respectively. These environments, classified as favorable environments or with high technology, if compared to the national average of wheat productivity in Brazil in 2020 (2,663 kg ha− 1), are 1.89 (E1), 1.61 (E3), and 1.60 (E7) times superior. It is important to point out that the Brazilian average of wheat productivity mainly reflects the South region, which historically cultivates wheat. In view of this, it is possible to verify a high potential of these cultivation environments for the cultivation of wheat in warmer regions.
Table 04
Best linear unbiased predictor (BLUP) for 36 tropical wheat genotypes in eight environments considering G x E interaction as evaluated in the 2018, 2019, and 2020 crop seasons in the Brazilian Cerrado.
Genotypes
|
\(\mu +{g}_{i}+{\left(ge\right)}_{ij}\)
|
Mean\(\mu +{g}_{i}\)
|
Environments
|
E1a*
|
E2bd
|
E3b
|
E4b
|
E5b
|
E6c
|
E7cd
|
E8e
|
BRS 264
|
5282.23
|
4375.40
|
4594.23
|
4309.59
|
4281.84
|
1388.56
|
4780.20
|
3097.16
|
3918.94
|
CD 151
|
5133.16
|
3981.86
|
3926.71
|
3794.05
|
3484.88
|
1319.71
|
4620.75
|
2623.31
|
3654.93
|
ORS 1403
|
4766.96
|
3736.36
|
4220.72
|
3889.71
|
3180.30
|
1257.97
|
4654.84
|
2338.53
|
3586.24
|
TBIO ATON
|
5245.11
|
4194.03
|
4931.53
|
4342.03
|
3907.79
|
1403.87
|
4381.30
|
3334.58
|
3888.73
|
TBIO DUQUE
|
4721.24
|
4002.78
|
4320.12
|
4029.86
|
3515.71
|
1194.67
|
4173.36
|
2472.32
|
3617.74
|
VI 09004
|
5405.13
|
3823.78
|
4219.48
|
4673.11
|
3945.29
|
1567.34
|
4064.07
|
4068.09
|
3890.86
|
VI 09007
|
5306.46
|
4077.72
|
4036.95
|
4251.37
|
3699.00
|
1348.74
|
3712.14
|
2862.33
|
3688.52
|
VI 09031
|
5113.06
|
3957.37
|
4254.31
|
4028.08
|
3583.42
|
1515.56
|
3624.16
|
2929.14
|
3664.81
|
VI 09037
|
5349.56
|
4144.17
|
3984.18
|
3818.95
|
3395.13
|
1322.17
|
4082.31
|
3068.60
|
3677.91
|
VI 130679
|
4898.96
|
3482.30
|
4240.45
|
4341.23
|
3503.98
|
1290.43
|
4090.40
|
2681.78
|
3625.88
|
VI 130755
|
4621.50
|
3607.51
|
4299.76
|
3979.09
|
3228.69
|
1536.86
|
4279.23
|
2956.01
|
3624.17
|
VI 130758
|
4420.23
|
3883.09
|
4110.79
|
3996.55
|
3162.77
|
1465.33
|
4025.40
|
2914.98
|
3580.82
|
VI 131313
|
4638.85
|
3365.61
|
4406.71
|
3928.31
|
3836.29
|
1027.85
|
4130.87
|
3607.03
|
3659.61
|
VI 14001
|
4767.34
|
3882.66
|
4113.74
|
4005.18
|
3987.86
|
1274.39
|
4414.19
|
3260.62
|
3722.19
|
VI 14022
|
5795.32
|
3893.18
|
3657.15
|
3906.81
|
3444.77
|
1409.05
|
4351.73
|
2821.19
|
3687.25
|
VI 14026
|
5587.80
|
4356.68
|
4478.56
|
4073.46
|
3590.93
|
1644.69
|
4746.96
|
3500.30
|
3908.31
|
VI 14050
|
5138.20
|
3740.85
|
4237.25
|
4146.22
|
3727.61
|
1120.71
|
3846.74
|
2737.49
|
3639.43
|
VI 14055
|
5317.19
|
3807.28
|
4461.51
|
4550.30
|
3900.65
|
1305.89
|
4654.96
|
3781.99
|
3891.97
|
VI 14088
|
4424.98
|
3853.66
|
4527.73
|
4318.06
|
3484.28
|
1447.72
|
4055.07
|
3289.19
|
3697.20
|
VI 14118
|
5104.55
|
4268.61
|
4110.80
|
4387.65
|
3901.30
|
1460.54
|
4242.79
|
3212.28
|
3802.63
|
VI 14127
|
5706.27
|
4364.26
|
4603.74
|
4747.49
|
3915.44
|
1740.92
|
4482.76
|
3393.77
|
3988.15
|
VI 14194
|
5676.27
|
4162.07
|
4961.24
|
4398.18
|
3427.17
|
1629.55
|
4427.03
|
2851.64
|
3871.78
|
VI 14197
|
5884.07
|
4282.66
|
4379.20
|
4424.87
|
3995.84
|
1551.62
|
4602.84
|
3402.94
|
3952.90
|
VI 14204
|
5459.91
|
3703.28
|
4151.58
|
4351.95
|
3572.02
|
1646.64
|
4380.79
|
2555.84
|
3731.69
|
VI 14208
|
5005.01
|
3947.98
|
4307.44
|
4315.33
|
3684.80
|
1439.66
|
4344.18
|
2242.44
|
3687.88
|
VI 14214
|
5060.30
|
4182.44
|
4368.34
|
4813.63
|
3932.94
|
1467.10
|
4521.61
|
2750.58
|
3836.07
|
VI 14239
|
4509.86
|
3706.07
|
3816.21
|
3738.57
|
3646.55
|
1090.03
|
3869.91
|
2648.90
|
3502.80
|
VI 14327
|
5206.94
|
3838.80
|
4342.12
|
4467.27
|
3680.39
|
1304.04
|
4322.40
|
3075.91
|
3765.74
|
VI 14426
|
4814.14
|
3970.33
|
3994.47
|
3881.57
|
3657.52
|
1469.28
|
4270.01
|
2844.01
|
3656.32
|
VI 14668
|
5227.38
|
3898.25
|
4148.27
|
4154.12
|
3817.16
|
1671.45
|
4182.97
|
3525.37
|
3797.43
|
VI 14708
|
4962.56
|
3870.11
|
4414.74
|
4199.28
|
3464.51
|
1191.21
|
4292.40
|
3077.76
|
3703.08
|
VI 14774
|
4384.20
|
4211.41
|
4495.34
|
4339.31
|
3882.48
|
1745.95
|
4368.15
|
3800.89
|
3846.77
|
VI 14867
|
4420.44
|
4165.90
|
4756.99
|
4198.06
|
3345.16
|
1515.56
|
4374.17
|
3551.35
|
3773.08
|
VI 14881
|
4626.01
|
3964.26
|
4359.08
|
4332.43
|
3526.44
|
1232.24
|
3906.15
|
2712.93
|
3636.52
|
VI 14950
|
4482.00
|
3748.62
|
3625.09
|
3651.65
|
3269.52
|
1178.56
|
3753.46
|
3359.90
|
3506.29
|
VI 14980
|
4581.94
|
4252.46
|
4684.44
|
4616.68
|
3972.25
|
1448.69
|
4538.37
|
4093.31
|
3925.40
|
Mean
|
5029.03
|
3964.00
|
4292.80
|
4205.56
|
3654.24
|
1406.24
|
4265.80
|
3095.68
|
3739.17
|
* Environments followed by the same letter show non-significant G x E interaction components at 5% probability by F test. |
The least productive environment was the E6, São Gotardo 2020, a rainfed environment with sowing before the end of the rainy season and with pressure by disease (blast inoculation). Grain yield in this environment was 1,406.24 kg ha− 1. The second lowest productive average was in Sete Lagoas 2020 (E8), followed by Viçosa 2020 (E5). Among the three most unfavorable environments, two are environments with rainfed cultivation (E5 and E6). These results corroborate those Pereira et al. (2019) described. The authors concluded that there is a need for a greater drought tolerance in wheat to unlock the production potential of the Brazilian Cerrado.
Another unfavorable environment was E8, characterized as the environment with the highest average temperature during the cultivation period (21.8 ºC) and the lowest relative humidity (58.9%) among all environments evaluated (Table 01). Leaf respiration is highly associated with higher air temperature in wheat and inversely proportional to yield, especially during the grain filling phase, with negative correlation estimates between 0.50 and 0.85 (Pinto et al., 2017). Heat tolerance in wheat is a polygenic trait difficult to quantify, and there are still no effective methods for selecting parents that are heat-tolerant and few molecular markers are used for the selection of this trait (Ni et al., 2018; Zhongfu et al. 2018). However, this characteristic, coupled with tolerance to water deficit, are essential factors for the advancement of wheat in Brazil, especially for lower regions in the Southeast and Midwest regions and in the North and Northeast States in general.
The environmental stratification grouped the environments E2, E3, E4, and E5. This indicates that the variance components of the G × E interaction between these environments are not significant. The three environments in Viçosa, together with the Rio Paranaíba 2019 environment, established a mega-environment (b). The environments E6 and E7, rainfed and irrigated, of São Gotardo established another subgroup (c), indicating that the relative performance of genotypes in a same city did not change as a function of the water regime. The environments irrigated Rio Paranaíba (2020) and irrigated São Gotardo (2019) formed another subgroup, possibly explained by the proximity between the two environments. The environments E1 and E8, with lower and higher average air temperature, did not group with any other with non-significant G × E interaction. They are thus independent environments.
Among the ten most productive genotypes in the E1 environment, the most favorable environment in this study, six are among the group of the ten most productive genotypes in the E8, with artificial inoculation of blast, namely: VI 14127, VI 14204, VI 14026, VI 14194, VI 09004, and VI 14197. These results suggest the classification of these strains as an ideal genotype due to the high performance of these materials in favorable as well as in unfavorable environments. The VI 14127 genotype was present in the selection of the ten most productive genotypes in all evaluated environments. Other published studies have also reported a high performance of this genotype (Machado e Silva et al. 2021). Figure 02 shows the relationship between climate variables, environments, and grain yield.
Table 05 shows the results of stability (HMGV), adaptability (RPGV), stability, and adaptability. The HMGV results indicate that the most productive and stable genotypes were VI 14127, VI 14774, VI 14026, VI 14197, and VI 09004. The HMGV analysis simultaneously brings together a selection based on two concepts: productivity and stability. Productivity is the result of an ordering of genotypes based on their genotypic values (BLUP), and stability is calculated through the standard deviation of the genotype behavior in the environments evaluated. The smaller the standard deviation, the greater the harmonic mean of genotypes. The genotypes VI 14980 and BRS 264, the 3rd and 4th most productive genotypes based on environment means (Table 04), are not among the five genotypes with the highest estimate of HMGV. These results indicate that, despite being productive when the average across environments is analyzed, they present little predictable behaviors.
Table 05
Genotypic stability (HMGV), genotypic adaptability (RPGV), simultaneous genotypic adaptability and stability (HMRPGV), genotypic value multiplying adaptability (RPGV*µ), and genotypic value penalized by instability and multiplied by adaptability (HMRPGV*µ) for 36 tropical wheat genotypes as for grain yield (GY) in eight environments evaluated in the 2018, 2019, and 2020 crop seasons in the Brazilian Cerrado.
Line
|
HMGV
|
Line
|
RPGV
|
RPGV*µ
|
Line
|
HMRPGV
|
HMRPGV*µ
|
VI 14127
|
3670.10
|
VI 14127
|
1.1117
|
4156.81
|
VI 14127
|
1.1092
|
4147.30
|
VI 14774
|
3570.17
|
VI 14197
|
1.0872
|
4065.32
|
VI 14197
|
1.0857
|
4059.75
|
VI 14026
|
3539.64
|
VI 14980
|
1.0845
|
4055.24
|
VI 14980
|
1.0747
|
4018.48
|
VI 14197
|
3534.09
|
VI 14026
|
1.0772
|
4027.91
|
VI 14026
|
1.0729
|
4011.71
|
VI 09004
|
3507.41
|
VI 09004
|
1.0743
|
4017.05
|
VI 09004
|
1.0640
|
3978.65
|
VI 14980
|
3493.21
|
VI 14774
|
1.0711
|
4005.12
|
BRS 264
|
1.0629
|
3974.48
|
VI 14668
|
3459.32
|
BRS 264
|
1.0662
|
3986.57
|
VI 14774
|
1.0596
|
3961.97
|
VI 14194
|
3437.34
|
TBIO ATON
|
1.0568
|
3951.45
|
TBIO ATON
|
1.0552
|
3945.41
|
BRS 264
|
3399.40
|
VI 14055
|
1.0560
|
3948.58
|
VI 14055
|
1.0496
|
3924.61
|
TBIO ATON
|
3391.96
|
VI 14194
|
1.0545
|
3942.86
|
VI 14194
|
1.0474
|
3916.27
|
VI 14867
|
3356.20
|
VI 14668
|
1.0412
|
3893.19
|
VI 14668
|
1.0359
|
3873.43
|
VI 14118
|
3348.19
|
VI 14214
|
1.0364
|
3875.43
|
VI 14214
|
1.0317
|
3857.54
|
VI 14214
|
3341.72
|
VI 14118
|
1.0289
|
3847.26
|
VI 14118
|
1.0276
|
3842.32
|
VI 14055
|
3338.90
|
VI 14867
|
1.0253
|
3833.60
|
VI 14867
|
1.0177
|
3805.48
|
VI 14204
|
3291.34
|
VI 14204
|
1.0029
|
3749.85
|
VI 14327
|
1.0009
|
3742.40
|
VI 14088
|
3243.72
|
VI 14327
|
1.0024
|
3747.99
|
VI 14204
|
0.9933
|
3714.16
|
VI 09031
|
3215.70
|
VI 14088
|
0.9912
|
3706.27
|
VI 14088
|
0.9876
|
3692.79
|
VI 14327
|
3197.80
|
VI 14001
|
0.9905
|
3703.50
|
VI 14001
|
0.9870
|
3690.73
|
VI 14426
|
3190.58
|
VI 09031
|
0.9773
|
3654.11
|
VI 09031
|
0.9733
|
3639.15
|
VI 130755
|
3187.44
|
VI 09007
|
0.9752
|
3646.32
|
VI 09007
|
0.9716
|
3633.07
|
VI 14001
|
3163.61
|
VI 14208
|
0.9744
|
3643.62
|
VI 14426
|
0.9703
|
3628.11
|
VI 14022
|
3145.82
|
VI 14022
|
0.9739
|
3641.74
|
VI 14708
|
0.9701
|
3627.43
|
VI 09007
|
3140.78
|
VI 14708
|
0.9732
|
3638.97
|
VI 09037
|
0.9675
|
3617.54
|
VI 14208
|
3126.08
|
VI 14426
|
0.9722
|
3635.29
|
VI 14022
|
0.9670
|
3615.95
|
VI 09037
|
3121.60
|
VI 09037
|
0.9704
|
3628.36
|
VI 14208
|
0.9626
|
3599.26
|
VI 130758
|
3116.07
|
VI 130755
|
0.9639
|
3604.21
|
VI 130755
|
0.9600
|
3589.65
|
CD 151
|
3069.25
|
CD 151
|
0.9581
|
3582.51
|
CD 151
|
0.9530
|
3563.51
|
VI 14708
|
3064.48
|
VI 131313
|
0.9558
|
3573.88
|
VI 14881
|
0.9464
|
3538.71
|
VI 130679
|
3035.88
|
VI 14881
|
0.9499
|
3551.69
|
VI 130679
|
0.9439
|
3529.26
|
VI 14881
|
3020.73
|
VI 130679
|
0.9468
|
3540.24
|
VI 130758
|
0.9421
|
3522.56
|
TBIO DUQUE
|
2955.48
|
VI 130758
|
0.9449
|
3533.15
|
VI 131313
|
0.9392
|
3511.69
|
ORS 1403
|
2937.66
|
VI 14050
|
0.9427
|
3524.85
|
VI 14050
|
0.9367
|
3502.45
|
VI 14050
|
2935.84
|
TBIO DUQUE
|
0.9377
|
3506.30
|
TBIO DUQUE
|
0.9320
|
3484.96
|
VI 131313
|
2912.19
|
ORS 1403
|
0.9263
|
3463.43
|
ORS 1403
|
0.9173
|
3429.93
|
VI 14950
|
2900.67
|
VI 14950
|
0.9060
|
3387.55
|
VI 14950
|
0.9005
|
3367.07
|
VI 14239
|
2808.35
|
VI 14239
|
0.8932
|
3339.80
|
VI 14239
|
0.8891
|
3324.56
|
Genotypic adaptability (RPGV) is expressed as the mean value of the proportion of predicted genotypic values in relation to the general mean of each environment. In this sense, the most adapted genotypes were those with the highest estimate of HMGV, plus the genotype VI 14980. The genotype VI 14980 had previously the 6th highest value of HMGV. The HMRPGV, which combines the concepts of productivity, stability and adaptability, classifies the genotypes exactly in the same way as the RPGV statistic informs. According to this method, the most productive, stable, and adapted genotypes are VI 14127, followed by the genotypes VI 14197, VI 14980, VI 14026, and VI 09004. No commercial genotype was among the top five for each parameter evaluated. The best commercial genotype was the BRS 264, with HMRPVG *µ 3,974.48 kg ha− 1. This method is very similar to the classical method proposed by Linn and Binns (1988), but in a genotypic and not in a phenotypic context.
Other authors have already reported the use of this method for soybean (Gonçalves et al. 2020), cotton (Peixoto et al. 2020), and other oilseed species that produce biodiesel (Alves et al. 2018). For wheat, there is only one study in the literature (Machado e Silva et al. 2021); however, only three environments were used in it.
The first two principal components (PC1 and PC2) in the GGE Biplot analysis encompassed approximately 70% of the total variation (64.75%) present in the environments (Fig. 05A). Other authors, in a similar study with 50 wheat genotypes and nine environments, observed an explanation between PC1 and PC2 of 50%. According to Yan et al. (2000), PC1 indicates the degree of adaptability of the genotypes; it is correlated with the performance per se in each environment. PC2 indicates the degree of stability of each genotype. In this type of analysis, the cosine of the angle between two environments corresponds to the genetic correlation between them. There is a high negative genetic correlation between the environments E1 and E8. The high discordance between the performance of genotypes in these environments can be explained by the agroclimatic differences existing between them (Table 01).
The comparison between mean and stability (Fig. 05B) considers the continuous green line with the arrow, called “average-environment axis” (AEA), classifying the genotypes with the highest average performance across the environments; the line perpendicular to the AEA indicates a greater environmental productivity variability (less stability) in any direction, such that the longer the dotted green line, the less stable the genotype (Yan and Tinker 2006). Based on these concepts, the sister lines G21 (VI 14127) and G23 (VI 14197) present genetic superiority in relation to the others. In addition, other genotypes show high stability and productivity: G1 (BRS 264), G16 (VI 14026), G18 (VI 14055), and G4 (TBIO ATON). The closer to the X axis (green line with arrow to the left), the greater the stability. In this case, the most stable genotypes are G27 (VI 14239), G34 (VI 14881), and G31 (VI 14708). However, these genotypes have a low grain yield in most environments. In general, the less productive a genotype is, the more stable it tends to be, as it consistently underperforms in many environments. The G36 (VI 14980) presented the greatest distance, evidencing a low estimate of prediction of productive behavior, which is different from the previous HMRPGV analysis.
The genotypes derived from cultivar BRS 264 (G21, G22 and G23) in its pedigree are among the best results for stability. Other studies have already reported the high productivity, adaptability, and stability of this genotype under Brazilian Cerrado conditions (Albrecht et al., 2007). This cultivar occupies 70% of the cultivated area in the Cerrado of Brazil and the world record for daily productivity: 9,630 kg ha− 1, that is, 80.9 kg ha− 1 day− 1. Based on this, it is suggested that stability and adaptability are complex traits and are confused with grain yield. Their inheritance should be better investigated to optimize the selection of this trait in wheat breeding programs.
Based on the graphic representation Which-won-Where (Fig. 05C), the genotypes G21, G23, G22, G15, G3, G27, G35, G32, and G36 are the furthest from the Biplot origin and have a better performance in one or more environments. Therefore, these genotypes delimit the area of the polygon. The blue dotted lines leaving the center of the Biplot (0,0) delimit the diagram in nine different sectors, with the formation of three distinct mega-environments (ME). The first mega-environment (ME1) is composed of only E1, the second mega-environment (ME2) is composed of the environments E2, E3, E4, E5, E6 and E7, and the third mega-environment (ME3) is composed of the E8 environment. Each mega-environment can be defined as a group of environments where one or more genotypes show a high adaptability, similar to what occurs in environmental stratification analyses. The three mega environments formed by the GGE Biplot analysis corroborate the results obtained in the environmental stratification analysis due to the formation of subgroups with a non-significant G × E interaction. G22 showed high adaptability to ME1, G21 and G23 showed better performance in ME2, while the genotypes G36 and G32 showed high adaptability to ME3. The genotypes present in sectors where there is no environment were not responsive to any environment studied.
ME1 and ME3 were more discriminative, while ME2 was more representative (Fig. 05D), especially the E6, a rainfed environment with blast inoculation. Mushayi et al. (2020) report that ideal environments for selection must be discriminatory and representative; however, no ME in this study was classified as ideal (Fig. 05E). Discriminatory but not representative environments can be used to select genotypes adapted to specific environments. Representative and homogeneous environments are ideal for the selection of widely adapted lines (Bányai et al., 2020). A genotype is considered ideal when high productive performance is linked to high stability. The genotypes closer to the center of the concentric circles are the most desirable and present a behavior close to that of the ideal genotype ideotype (Fig. 05F). In this study, the genotypes G21, G23, G16, G1, G4, and G22 are the closest to a hypothetical ideal genotype according to the GGE Biplot methodology. These results corroborate those found through the WAASBY index (Fig. 06).