AMMI BIPLOT ANALYSIS OF
GENOTYPE × ENVIRONMENT INTERACTION
IN WHEAT IN EGYPT
M. B. Ali1, A. N. El-Sadek2, M. A. Sayed1 and M. A. Hassaan2
1. Agronomy Dept, College of Agri., Assiut Univ., Assiut, Egypt
2. Ecology and Dry Land Agri. Division, Desert Research Center, El-Matarya, Cairo, Egypt
ABSTRACT
The genotype by environment (G×E) interaction has an impact on the selection of genotypes. The yield productivity of 49 wheat (Triticum aestivum L.) genotypes was evaluated using additive main effects and multiplicative Interaction (AMMI) model. Experiments were implemented in four locations (Barrani, Matrouh, Assiut and El-Kharga) across Egypt using a randomized complete block design with three replications. Analysis of variance for grain yield (g/m2) showed that the sum of squares of locations accounted for 86% of total sum of squares. The AMMI analysis of variance showed that two interaction principal components (IPCs) out of three were highly significant (P < 0.01). The IPC1 and IPC2 accounted for 55.4% and 31.4% of the G×E interaction, respectively. Genotypes 40 and 48 were well adapted to Assiut location whereas genotype 18 was adapted to Barrani location. The AMMI model identified the best combinations of genotypes and environments for yield.
Key words: Interaction principal components (IPCs), Grain yield, Which-won-where, CIMMYT, General linear-bilinear model (GLBM), Drought stress, Favorable environment.
INTRODUCTION
Multi-environmental trials (METs) are a crucial step in breeding programs of all major crops around the world. The presence of G × E interactions is universal in METs. Some plant breeders aim to identify genotypes that show consistent performance across a wide range of diverse environments. However, the changes in genotypes’ relative ranking across environments (crossover interaction) reflect significant G × E interactions (Allard and Bradshaw 1964). Therefore, other plant breeders select specifically adapted genotypes. The gain from selection and the correlation between phenotype and genotype are reduced because of the G×E interaction (Comstock and Moll 1963). An adaptive genotype possesses a high mean yield with minimal fluctuation in performance across diversified environments. Plant breeders carry out METs to identify superior genotypes for a target region; the target region may be subdivided into mega-environments (Yan et al 2000).
The yielding ability of any genotype in an environment is a function of environment main effect (E), a genotype main effect (G), and the G×E interaction (Yan and Kang 2003).To detect G×E interaction using statistical procedures, there must be at least two diverse genotypes evaluated in at least two distinct environments(Yan and Kang 2003). Usually, 80% or higher of the total yield variation is attributed to E; nevertheless, both G and G×E interaction are important for genotype evaluation (Yan 2002). Biplot analysis is a multivariate analytical procedure that graphically shows two-way data and allows visualization of the relationship among environments, genotypes, and between genotypes and environments (Dehghani et al 2006). There are two biplot approaches that have been utilized to graphically display and summarize two-way G×E data. These are the additive main effect and multiplicative interaction (AMMI) biplot (Gauch 1988; Zobel et al 1988) and the genotype main effect + G×E interaction (GGE) biplot (GGE biplot) (Yan et al 2000).
The most commonly used models of the biplot analysis are AMMI1 (the AMMI model with one PC), GGE2 (the GGE model with two PCs) and AMMI2 (the AMMI model with two PCs) (Yang et al 2009). Both AMMI and GGE are two cases of GLBM (Yang et al 2009). Gauch and Zobel (1997) implemented the first use of AMMI1 graph to address which-won-where pattern. Ebdon and Gauch (2002b) claimed that mega-environment classification based on AMMI1 method should be virtually the same as that based on a GGE biplot. The AMMI1 biplot enables a coinciding visualization of the mean performance and the stability of genotypes (Yang et al 2009). The AMMI2 and AMMI1 biplots can be used for genotype and environment identifications (which-won-where pattern) (Yan et al 2007; Gauch et al 2008); however, the AMMI1 has simple and straightforward geometry for genotype and environments identification. Yang et al 2009, summarized description and interpretation of the AMMI2. Briefly, the scores of genotypes and environments are displayed as vectors in a two-dimensional space. These vectors are drawn from the origin (0.0) to the ends of their scores. The angle “∠” between vectors of genotypes and environments determine the genotypic response at a specific environment; e.g. 90º > “∠” > 270º shows that the genotype has a positive response at that environment; on the other hand, 90º ≤ “∠”≤ 270º shows a negative genotypic response. Moreover, the cosine of this angle estimates the phenotypic correlation between two genotypes or environments.
Egypt imported about 18.1 million tons of cereals in 2014/2015 and around 55% of its cereal imports represented wheat; therefore, Egypt is the world’s largest wheat importer with 10 million tons (FAO 2015). Egypt is working towards increasing its agricultural efficiency. To do that, the agricultural sector of Egypt is attempting to improve wheat production through identification and introduction of adaptive genotypes. These genotypes can be used in wheat breeding programs to improve yielding ability of the Egyptian wheats. Therefore, our objectives were to: (1) use the AMMI biplot procedures to investigate the possible presence of mega environments in wheat-growing regions across Egypt, and (2) determine the best genotype for each mega-environment.
MATERIALS AND METHODS
Genotypes and field experiments
A set of 49 breeding materials (genotypes) of advanced lines resulted from CIMMYT’s breeding program (CIMMYT=International Maize and Wheat Improvement Center) (Table 1) were sown at four locations across Egypt (Figure 1) during 2014/2015 season. These locations represented a wide range of wheat-production environments in Egypt (Table 2). The 49 genotypes were sown in a randomized complete-block design. Each genotype was sown in three replications in 3×3.5 m2 plots. Each plot consisted of 10 rows, each 3 m long with 35 cm between rows.
Traits studied
Heading date (HD) for each genotype was recorded as the number of days from the sowing date until 50% of tillers had emergence a half of spikes from the flag leaf sheath (Zadoks et al 1974). Then the number of days to 50% heading was calculated by subtracting the heading date from the planting date. Anthesis date (AD) for each genotype was recorded as the number of days from the sowing date till spikes of 50% of tillers showed at least one anther on the spikelets (Zadoks et al 1974). At anthesis time, chlorophyll content (Chl.) of the flag leaf was measured using a self-calibrating SPAD chlorophyll meter (Model 502, Spectrum Technologies, Plainfield, IL). This measurement directly estimated the chlorophyll concentration of the flag leaf.
At maturity time, 10 individual plants were randomly chosen to measure plant height (PH; cm) and spike length (SL; cm). At harvest, a guarded (plantshaving another plant on each side of the harvested area) square meter was harvested to measure the following traits: number of grains per spike (NGPS), 1000-kernel weight (1000-KW; g), grain yield (GY; g/m2), biological yield (BY; g/ m2) and harvest index (HI). The HI was calculated as: (GY/BY) ×100.
Statistical analyses
The ANOVA, other statistics of all studied traits for separate environments, the AMMI and GGE analyses were carried out using the CropStat7.2 software package developed by IRRI (IRRI, 2009).This software has been extensively used for AMMI analyses in many studies, e.g., De Vita et al (2010), Annicchiarico et al (2013) and Abebe et al (2015).
The AMMI analysis was carried out using ordinary F-test (Gollob 1968) which combined both analysis of variance (ANOVA) and singular value decomposition that is also known as principal component analysis as a single model with additive and multiplicative parameters. The analysis was implemented using general linear-bilinear model (GLBM) (e.g., Cornelius and Seyedsadr 1997; Cornelius et al 2001).
Table 1. List of CIMMYT genotypes along with their pedigree
1 / KACHU #1
2 / QUAIU #1
3 / BAJ #1
4 / FRANCOLIN #1
5 / KACHU/BECARD//WBLL1*2/BRAMBLING
6 / QUAIU #1/SUP152
7 / QUAIU #1/SUP152
8 / KACHU//KIRITATI/2*TRCH
9 / KIRITATI//HUW234+LR34/PRINIA/3/BAJ #1
10 / ND643/2*WBLL1//VILLA JUAREZ F2009
11 / SUP152/FRNCLN
12 / BAJ #1/SUP152
13 / WHEAR/KUKUNA/3/C80.1/3*BATAVIA//2*WBLL1/5/PRL/2*PASTOR/4/CHOIX/STAR/3/HE1/3*CNO79//2*SERI
14 / CROC_1/AE.SQUARROSA (205)//BORL95/3/PRL/SARA//TSI/VEE#5/4/FRET2/5/2*DANPHE #1
15 / FRET2*2/4/SNI/TRAP#1/3/KAUZ*2/TRAP//KAUZ/5/2*FRNCLN
16 / BAJ #1/3/2*HUW234+LR34/PRINIA//PFAU/WEAVER
17 / KISKADEE #1*2//KIRITATI/2*TRCH
18 / MUTUS*2/HARIL #1
19 / BAJ #1*2/TINKIO #1
20 / BAJ #1*2//ND643/2*WBLL1
21 / WBLL1*2/BRAMBLING*2//BAVIS
22 / PRL/2*PASTOR//WHEAR/SOKOLL
23 / WHEAR/KUKUNA/3/C80.1/3*BATAVIA//2*WBLL1/4/WAXWING*2/KRONSTAD F2004
24 / WHEAR/KIRITATI/3/C80.1/3*BATAVIA//2*WBLL1/4/BECARD
25 / FRET2*2/4/SNI/TRAP#1/3/KAUZ*2/TRAP//KAUZ/5/KIRITATI/2*TRCH/6/BAJ #1
26 / FRET2*2/BRAMBLING//KIRITATI/2*TRCH/3/FRET2/TUKURU//FRET2
27 / KACHU*2/SUP152
28 / DANPHE/PAURAQUE #1//MUNAL #1
29 / KIRITATI//2*PRL/2*PASTOR/3/CHONTE/5/PRL/2*PASTOR/4/CHOIX/STAR/3/HE1/3*CNO79//2*SERI
30 / KIRITATI//HUW234+LR34/PRINIA/3/CHONTE/5/PRL/2*PASTOR/4/CHOIX/STAR/3/HE1/3*CNO79//2*SERI
31 / KIRITATI//HUW234+LR34/PRINIA/3/FRANCOLIN #1/4/BAJ #1
32 / MUTUS//KIRITATI/2*TRCH/3/WHEAR/KRONSTAD F2004
33 / ND643/2*WBLL1//2*KACHU
34 / PAURAQ/5/KIRITATI/4/2*SERI.1B*2/3/KAUZ*2/BOW//KAUZ/6/PAURAQUE #1
35 / PAURAQ/4/WHEAR/KUKUNA/3/C80.1/3*BATAVIA//2*WBLL1/5/PAURAQUE #1
36 / FRANCOLIN #1*2//ND643/2*WBLL1
37 / FRANCOLIN #1/CHONTE//FRNCLN
38 / BAJ #1*2/KISKADEE #1
39 / WHEAR/KUKUNA/3/C80.1/3*BATAVIA//2*WBLL1*2/4/KIRITATI/2*TRCH
40 / TAM200/PASTOR//TOBA97/3/FRNCLN/4/WHEAR//2*PRL/2*PASTOR
41 / TOB/ERA//TOB/CNO67/3/PLO/4/VEE#5/5/KAUZ/6/FRET2/7/VORB/8/MILAN/KAUZ//DHARWAR DRY/3/BAV92
42 / FALCIN/AE.SQUARROSA (312)/3/THB/CEP7780//SHA4/LIRA/4/FRET2/5/DANPHE #1/11/CROC_1/AE.SQUARROSA (213)//PGO/10/ATTILA*2/9/KT/BAGE//FN/U/3/BZA/4/TRM/5/ALDAN/6/SERI/7/VEE#10/8/OPATA
43 / BAVIS/NAVJ07
44 / CROC_1/AE.SQUARROSA (213)//PGO/10/ATTILA*2/9/KT/BAGE//FN/U/3/BZA/4/TRM/5/ALDAN/6/SERI/7/VEE#10/8/OPATA/11/ATTILA*2/PBW65
45 / W15.92/4/PASTOR//HXL7573/2*BAU/3/WBLL1/5/DANPHE #1
46 / BAVIS/3/ATTILA/BAV92//PASTOR/5/CROC_1/AE.SQUARROSA (205)//BORL95/3/PRL/SARA//TSI/VEE#5/4/FRET2
47 / BABAX/LR42//BABAX/3/ER2000/4/PAURAQUE #1
48 / VEE/MJI//2*TUI/3/PASTOR/4/BERKUT/5/BAVIS
49 / SOKOLL/3/PASTOR//HXL7573/2*BAU/5/CROC_1/AE.SQUARROSA (205)//BORL95/3/PRL/SARA//TSI/VEE#5/4/FRET2
Table 2. Seasonal rainfall† and soil type of the test locations.
Location / Latitude / Longitude / Rainfall† (mm) / Altitude (m) / Temperatureǂ (ºC) / Soil typeMin. / Max.
Barrani (Ba) / 31° 34′
19 ″ / 25° 59′ 16″ / 30.75 / 33 / 10.46 / 21.06 / Sandy Clay Loam
Matrouh (Ma) / 31° 21′
12″ / 27° 11′ 14″ / 81.60 / 10 / 11.18 / 21.72 / Sandy Clay Loam
El-Kharga (Kh) / 25° 31′
36″ / 30° 37′ 27″ / 0.00 / 32 / 13.05 / 28.73 / Sandy Clay Loam
Assiut (As) / 27° 11′
18″ / 31° 09′ 47″ / 0.00 / 70 / 11.33 / 26.42 / Clay
†Rainfall data were obtained from the Global Summary of the Day (GSOD) dataset of the National Climatic Data Center NNDC (ftp://ftp.ncdc.noaa.gov/pub/data/gsod/) for the period from November 2014 to May 2015. ǂ Average temperature was calculated from the daily temperature data over the growing season
Figure 1. Map of Egypt showing the four locations.
RESULTS AND DISCUSSION
Genotype performance
The genotypes showed highly significant differences for all studied traits in Barrani, Matrouh and ElKharga. However, in Assiut, genotypes exhibited highly significant differences for all studied traits, except plant height, spike length, number of grains per spike andchlorophyll content (Table 3).
Table 3. Mean squares and summary statistics for studied traits in separate locations
PH† / SL† / NGPS† / 1000-GW† / GYPM† / BYPM† / HI† / Chl. † / HD† / AD†
Barrani / Rep. / 0.004 / 0.124 / 1.778 / 0.198 / 47.336 / 273.078 / 0.007 / 2.420 / 0.092 / 1.235
Gen. / 46.389
** / 1.845
** / 55.903
** / 48.575
** / 35.536
** / 439.502
** / 0.005
** / 39.869
** / 10.967
** / 6.793
**
Error / 1.520 / 0.396 / 3.617 / 1.201 / 2.390 / 19.743 / 0.001 / 1.323 / 0.217 / 0.214
CV% / 3.980 / 12.640 / 12.140 / 4.820 / 20.130 / 12.510 / 13.590 / 12.450 / 1.010 / 0.950
Mean / 30.950 / 4.983 / 15.660 / 22.710 / 7.681 / 35.520 / 0.209 / 9.239 / 45.970 / 48.720
Min. / 22.950 / 3.000 / 8.000 / 11.400 / 2.080 / 13.050 / 0.103 / 1.300 / 40.500 / 45.000
Max. / 41.150 / 7.450 / 32.000 / 38.050 / 18.940 / 74.870 / 0.330 / 18.250 / 50.000 / 52.500
S.E. / 0.872 / 0.445 / 1.345 / 0.775 / 1.093 / 3.142 / 0.020 / 0.813 / 0.329 / 0.327
Matrouh / Rep. / 30.8898 / 0.060 / 1.062 / 1.77796 / 1973.72 / 895.553 / 0.008 / 5.53969 / 0.010 / 0.041
Gen. / 186.54
** / 4.249
** / 227.235
** / 389.499
** / 13004
** / 38677.2
** / 0.043
** / 33.240
** / 8.577
** / 8.309
**
Error / 17.838 / 0.332 / 10.980 / 0.50796 / 114.061 / 578.133 / 0.001 / 2.642 / 0.177 / 0.270
CV% / 5.000 / 6.710 / 12.370 / 2.150 / 8.020 / 4.950 / 8.780 / 3.460 / 0.480 / 0.570
Mean / 84.550 / 8.587 / 26.800 / 33.070 / 133.200 / 486.100 / 0.274 / 46.930 / 88.090 / 90.960
Min. / 59.210 / 5.100 / 10.200 / 8.500 / 29.820 / 246.000 / 0.051 / 34.100 / 83.000 / 86.000
Max. / 102.400 / 12.220 / 68.500 / 85.800 / 326.600 / 954.00 / 0.578 / 53.950 / 94.000 / 96.000
S.E. / 2.986 / 0.407 / 2.343 / 0.504 / 7.552 / 17.000 / 0.017 / 1.149 / 0.297 / 0.367
Assiut / Rep. / 13.224 / 2.452 / 9.184 / 10.449 / 4083.13 / 142585 / 0.002 / 0.132 / 5.398 / 5.398
Gen. / 48.035 / 2.800 / 14.391 / 28.930
** / 19079.5
** / 70829.6
** / 0.008
** / 4.912 / 14.189
** / 9.367
**
Error / 77.230 / 1.946 / 12.184 / 4.087 / 690.041 / 19801.8 / 0.002 / 5.251 / 1.065 / 1.065
CV% / 9.71 / 10.99 / 8.44 / 4.35 / 4.83 / 10.29 / 9.720 / 4.36 / 1.26 / 1.21
Mean / 90.53 / 12.69 / 41.37 / 46.45 / 544.4 / 1368 / 0.4028 / 52.5 / 81.72 / 85.44
Min. / 81.5 / 10.25 / 35 / 37.9 / 357.1 / 933.3 / 0.2924 / 48.7 / 77 / 82
Max. / 106.2 / 16.75 / 48 / 56.7 / 902.4 / 1724 / 0.5653 / 55.2 / 87.5 / 89.5
S.E. / 6.214 / 0.987 / 2.468 / 1.429 / 18.57 / 99.5 / 0.028 / 1.62 / 0.7296 / 0.730
El-Kharga / Rep. / 3.197 / 0.155 / 190137 / 0.083 / 59.606 / 85322900 / 0.021 / 1.349 / 0.500 / 0.255
Gen. / 79.741
** / 2.162
** / 192024
** / 80.248
** / 19903.2
** / 35421900
** / 0.044
** / 138.99
** / 15.586
** / 16.623
**
Error / 24.942 / 0.257 / 191124 / 0.470 / 109.731 / 30694400 / 0.003 / 2.0547 / 0.292 / 0.338
CV% / 6.79 / 5.79 / 2.62 / 2.070 / 4.25 / 248.360 / 20.950 / 4.250 / 0.750 / 0.770
Mean / 73.55 / 8.758 / 39.65 / 33.19 / 246.6 / 2231 / 0.263 / 33.71 / 72.26 / 75.7
Min. / 63.25 / 7.15 / 29 / 21.53 / 105 / 195.6 / 0.055 / 17.5 / 67 / 71
Max. / 87.5 / 11.1 / 50.9 / 45.59 / 484.5 / 26190 / 0.574 / 51.4 / 76 / 80
S.E. / 3.531 / 0.359 / 0.735 / 0.485 / 7.407 / 3918 / 0.039 / 1.014 / 0.382 / 0.411
∗∗Significant at the 0.01 probability level.
†PH = plant height measured in cm; SL = spike length measured in cm; NGPS = number of grains per spike; 1000-GW = 1000 grain weigh measured in grams; GYPM = grain yield per square meter measured in grams; BYPM = biological yield per square meter measured in grams; HI =harvest index calculated as a ratio by dividing GYPM/BYPM; Chl. = chlorophyll content taken at anthesis; HD = heading date and AD = anthesis date.