Advances in Biological Research 3 (1-2): 43-47, 2009

ISSN 1992-0067
© IDOSI Publications, 2009

Additive Main Effects and Multiplicative Interactions Analysis of

Yield Performances in Rice Genotypes under Coastal Saline Environments

A. Anandan, R. Eswaran, T. Sabesan and M. Prakash

Department of Agricultural Botany, Faculty of Agriculture,

Annamalai University, Chidambaram-608 002, Tamil Nadu, India

Abstract: Plant Breeders evaluate their germplasm in multi-environment to study the performance and
adaptation for specific or general environment. The objective of the present investigation was to analyse the
pattern of Genotype x Environment (G x E) interaction for grain yield of 46 genotypes by Additive Main effects
and Multiplicative Interaction (AMMI) model using the data generated from three saline stress environments
of east coastal region of Tamil Nadu, India. The results showed highly significant genotypic and G x E
interaction. The G x E interaction influenced the relative ranking of the genotypes across saline stress
environment condition. Among the genotypes tested, G45, G26, G27, G35 and G34 showed low interaction effect
with score nearest to zero with above average yield. While two genotypes, G36 and G19 exhibited high positive
interaction effect, gave the highest mean grain yield and is better suited to favourable environment. Similarly,
AMMI characterized the environments and identified E2 as a favourable environment for the better expression
of the trait studied.

Key words: Rice % Genotype x environment % AMMI % Yield

INTRODUCTION different environments to study genotype x environment

Rice is considered moderately sensitive to salinity.
Higher population growth rate and the conversion of
some highly productive rice cultivation lands for
industrial and residential purposes, has pushed rice
cultivation to less productive area such as saline,
drought and flood prone areas. However, cultivars
adaptable to these abiotic stress areas were originated or
selected for these situations were agronomically poor
(traditional cultivars). Salt tolerance cultivars such as
pokkali, cheriviruppu, nona bokra and SR26B are tall,
photosensitive, have poor grain quality and are low
yielding [1]. After green revolution, several breeding
programs were adapted to develop saline tolerant
genotypes at IRRI [2] and other rice growing countries.
The venture to develop high yielding genotype with very
good tolerance on par to traditional cultivars is still in
progress.
The developed cultivars should adapt to a wide range
of target environments, is the eventual goal of plant
breeders. Hence, pattern of response of genotypes is
studied by the plant breeders by testing genotypes in
(G x E) interaction. To estimate the level of interaction of
genotypes to environments and to eliminate as much as
possible the unexplainable and extraneous variability
contained in the data, several statistically techniques
have been developed to describe G x E and measure the
stability of genotypes. For further detailed study on
current statistics used to test and measure genotype
stability are given by Lin et al. [3] , Piepho [4 ] and Laxmi
and Gupta [5].
Since, G x E interaction is naturally multivariate, the
Additive Main effects and Multiplicative Interaction
(AMMI) offers on appropriate first statistical analysis of
yield trials that may have a G x E interaction [6]. The
objectives of this study were to assess the extent of G x E
interaction and to select the stable genotypes of rice in
saline stressed environment over season.
MATERIALS AND METHODS
Forty six rice genotypes (Table 1) from different parts
India were evaluated at Plant Breeding Research Farm,
Faculty of Agriculture, Annamalai University, Annamalai

Corresponding Author: A. Anandan, Department of Agricultural Botany, Faculty of Agriculture,

Annamalai University, Chidambaram-608 002, Tamil Nadu, India
43
 Advan. Biol. Res., 3 (1-2): 43-47, 2009

Table 1: Grain yield (kg/plot) of 46 rice genotypes grown in three environments and the first IPCA scores for the G x E interaction effect as derived from
AMMI analysis
Genotype Genotype code E1 E2 E3 Genotype mean IPCA 1
NDR 9730015 G1 1.78 2.50 1.83 2.04 -0.051
NDRK 50001 G2 2.33 3.20 2.77 2.77 0.202
NDRK 50002 G3 2.10 2.70 1.88 2.23 -0.240
NDRK 50004 G4 4.23 4.79 4.31 4.44 -0.070
NDRK 50005 G5 2.76 3.71 3.07 3.18 0.147
NDRK 50007 G6 1.34 2.14 1.63 1.70 0.102
CSAR 520 G7 5.19 6.12 5.55 5.62 0.175
CSAR 518 G8 3.48 4.37 3.62 3.82 0.033
TR 2003-25 G9 4.27 4.77 4.40 4.48 -0.055
TR 2003-82 G10 1.60 2.07 1.58 1.75 -0.152
TR 2003-3 G11 5.20 5.68 5.20 5.36 -0.136
Bio 5-1 G12 4.46 5.06 4.49 4.67 -0.090
BM 5 G13 2.01 2.71 2.36 2.36 0.122
GGV-SAT-05-1 G14 2.21 2.85 2.44 2.50 0.035
RP 4679-3-2-38-1-1 G15 4.49 4.99 4.28 4.59 -0.251
CR 2069-16-1 G16 3.95 4.50 4.37 4.27 0.129
CR 2094-155-4 G17 4.22 4.56 3.90 4.23 -0.351
CR 2071-245-3 G18 5.40 6.38 5.86 5.88 0.241
CR 2092-88-1 G19 5.81 6.64 6.09 6.18 0.108
RP 4658-IR 52707-20-7-1-3-5-P1-4 G20 1.76 2.42 2.06 2.08 0.084
RP 4659-IR 59418-20-2-1-5-7-P5-1 G21 1.39 1.93 1.32 1.55 -0.157
MTU 2716 G22 5.54 6.33 5.79 5.89 0.077
MTU 1061 G23 3.58 4.34 3.87 3.93 0.096
CN 1266-19-16 G24 3.50 4.22 3.63 3.78 -0.003
CN 1266-26-177 G25 3.94 4.64 3.92 4.17 -0.096
CN 1266-9-6 G26 6.00 6.77 6.15 6.31 0.016
CN 1266-3-4 G27 4.88 5.43 5.04 5.12 -0.022
CN 1912 G28 3.70 4.41 4.18 4.09 0.194
CN 1914 G29 3.80 4.25 3.61 3.89 -0.247
CN 1913 G30 3.73 4.02 3.45 3.73 -0.336
CSR 13 G31 7.24 7.71 7.20 7.38 -0.164
CSR 27 G32 2.92 3.58 3.09 3.20 0.003
CST 7-1 G33 5.64 6.46 5.87 5.99 0.071
Jaya G34 5.56 6.24 5.74 5.85 0.011
Pant Kalanamak 3131 G35 5.49 6.17 5.65 5.77 0.001
Pant Dhan 102 G36 6.03 6.95 6.41 6.46 0.176
RP 4677-6-16-1-1-2-B G37 4.44 5.04 4.58 4.69 -0.030
RP 46768-5-1-7-1-2-3 G38 4.49 5.22 4.67 4.80 0.024
MAUB 13 G39 1.60 2.39 1.68 1.89 -0.021
RP 4347-TSH-48-6-2-2-3-B G40 5.59 6.33 5.67 5.87 -0.030
RP 4339-MSC-55-1-4-2-3-8-B G41 4.18 5.13 4.44 4.58 0.115
RP 4676-23-6-1-2-5-B G42 4.66 5.52 5.03 5.07 0.162
RP 4343-4-2-2-1-3-1-B G43 4.40 5.13 4.67 4.73 0.078
RP 4677-16-6-1-1-2-1-2 G44 3.40 3.97 3.32 3.57 -0.163
RP 4343-18-2-1-1-1-1-B G45 4.06 4.84 4.22 4.37 0.023
TRY 1 G46 4.07 5.05 4.53 4.55 0.241
Environment mean 3.97 4.66 4.12 4.25
IPCA 1 -0.777 0.201 0.576

Nagar, located at East coastal region of Tamil Nadu, India
with soil pH of 8 to 8.5 and EC of 2.51 to 2.8 dSmG1. The
crop was evaluated in three seasons viz., E1 (Kharif, 2006),
E2 (Kharif, 2007) and E3 (Rabi, 2007). For all the trials,
the design used was RCBD with three replications.
The plot had 10 sq.m with spacing of 20 cm between
rows and 20 cm between plants. Management practices
were uniformly adapted to all seasons as per the
recommendation for rice in the irrigated condition. The
whole plot was used to determine the yield potential of
each genotype.
Statistical Analysis: AMMI is a unified approach that
fits the additive main effects of genotypes and
environments by the usual analysis of variance and
then describes the non-additive parts by principal
component analysis fitted to the AMMI model according
to the following equation.

44
 Advan. Biol. Res., 3 (1-2): 43-47, 2009

Yge = µ + "g + $e + E 8n ãgn *en + 2ge + 0ger Table 2: AMMI analysis of variance for grain yield of 46 rice genotypes
Source df SS MS F

Where, Yge is the yield of genotype g in environment e, Treatments 137 905.60 6.61 216.78**
Genotypes 45 865.30 19.23 630.56**
µ is the grand mean yield, "g is the genotype mean
Environments 2 36.50 18.24 5.43**
deviation, $e is the environment mean deviation, 8n is the
Interactions 90 3.90 0.04 1.42*
eigen value of the nth principal component, ãgn and *en are IPCA 1 46 2.90 0.06 2.04**
the genotype and environment scores for the nth principal Residuals 44 1.00 0.02 0.77
component; 2ge is the residual and 0ger is the random error Error 270 8.20 0.03
[6 , 7]. Total 413 934.00 2.26
The additive part of the AMMI model (µ, "g, $e) is *P<0.05 and **P<0.01
estimated from an analysis of variance and the
multiplicative part (8n, ã gn and * )en from a principal 25.64% with 51% and 49% of the degrees of freedom
component analysis. The interaction between any respectively for G x E interaction. The first principal
genotype and environment can be estimated by component factor has a high contribution to the
multiplying the score for the interaction principal interaction sum of squares while the residual is small.
component axis (IPCA) of a genotype (8n 0.5 x ãgn) by an This indicates that one fundamental factor affects G x E
environment IPCA score (8n0.5 x *en) [8]. Furthermore, interaction; this could be either genotypic or
Gauch and Zobel [9] proposed an approach to estimate environmental in nature [11].
the relevant portion of G x E for the trait studied to avoid The errors from uncontrolled variation were
spurious interpretation of statistical results. Partitioning calculated using the methods described by Gauch and
the errors from uncontrolled variation (noise) out of the Zobel [9]. The interaction of the grain yield contained
total G x E sum of squares (SS) is important because most 69% noise and 30.77% real structure, with the relevant
of the noise appears in the interaction. Noise SS, real (target) variation being 95.68% of the treatment sum of
structure SS and target relevant variation percentage were squares. Treatment sum of squares variation explained
calculated as described by Gauch and Zobel [9]. The above target percentages that would be attributed to
target percentage of the relevant variation explained by noise. The first IPCA component explained 95.87% of
IPCA in the AMMI analysis should equal SS (relevant the treatment sum of squares (2.9 + 865.3) / 905.6) which
variation) / SS (treatment). was very close to the target percentage sum of squares
explained (95.68%).
RESULTS AND DISCUSSION Data presented in Table 1 show that the mean yield
for 46 genotypes evaluated in three environments, plus
Bartlett’s test indicated homogenous error variance the genotype and environment means and first IPCA
for the trait yield in each of three environments and scores. The grain yield ranged from 1.32 kg/plot to
allowed to proceed further for pooled analysis across 7.71 kg/plot with an average of 4.25 kg/plot. The AMMI
environment. ANOVA across environments detected 1 estimated yield for any genotype in any location/season
significant variation among genotypes and for the G x E may be calculated from the biplot, as described by
for the trait grain yield (Data not shown). The analysis of Zobel et al. [6]. The biplot (Fig. 1) captured 99% of the
variance of AMMI (Table 2) showed that the mean sum of treatment sum of squares. Genotypes with IPCA 1 scores
squares due to treatments, genotypes, environments and near zero had little interaction across environments, while
genotype x environment interaction were significant, the genotypes with large IPCA 1 scores, either positive or
indicating broad range of diversity existed among negative direction were highly interactive. Similarly,
genotypes [10]. Significance of the environments locations/seasons with IPCA 1 scores near zero had
indicated distinctness of intrinsic factors in different little interaction across genotypes and low discrimination
environment. The significance exhibited by G x E among genotypes. Genotype and location combinations
interaction indicated that each of the genotype interacted with IPCA 1 scores of the same sign produced positive
differentially in various environment tested. The specific interaction effects, whereas combinations of
multiplicative variance of the treatment sum of squares opposite sign had negative specific interactions [12-14].
due to G x E interaction was partitioned into IPCA 1 and In Fig. 1 for any genotype-environment combination,
residual variation. About 74.3% of the interaction sum of the main effect equals the genotype mean plus the
squares was explained by IPCA 1 and residual explained environment mean minus the grand mean and the

45
 Advan. Biol. Res., 3 (1-2): 43-47, 2009
0.6 E3

0.4

G4 6 G18
0.2 G2 G28 E2 G7 G36
G6 G5 G4 2
G13 G16 G41
G2 0 G2 3 G43 G2 2 G33G19
G14 G8 G45 G38 G3 5 G3 4 G2 6
0.0 G3 9 G3 2 G2 4 G2 7
G9 G4 G37
IPCA 1
G1 G4 0
G2 1 G25 G12 G11
G10 G4 4 G3 1
-0.2
G3 G3 5 G15
G3 0 G17
-0.4

-0.6

-0.8 E1
1.5 2.0 2.5 3.0 3.5 4 .0 4.5 5.0 5 .5 6.0 6.5 7.0
MEA N

Fig. 1: AMMI – 1 biplot for grain yield of rice, showing both main and interaction effects for both genotypes and
environments

interaction is the genotype IPCA score times the
environment ICPA score [15]. For example, G5 grown in
environment E2 had a main effect of 3.18 + 4.66 - 4.25 =
3.59 kg/plot and an interaction effect of (0.147) x (0.201) =
0.029 kg/plot. Therefore, the AMMI model gives a grain
yield estimation of 3.59 kg/plot. This fits well the observed
value of 3.71 kg/plot.
From the Fig. 1, it is clear that, genotypes G17, G25
and G16 differ only in interaction effect and not in main
effect, while G39, G32, G24, G45, G27, G35, G34 and G26
showed little G x E interaction on IPCA 1 and grouped
together along the abscissa, although they differ
dramatically in main effect. The difference exhibited by
genotypes over interaction and main effect might have
been due to the factor salinity stress. The genotypes,
which had IPCA score nearest to zero, are G24, G26, G27,
G32 G34, G35, G39 and G45, the most stable across
environment studied. Among the above mentioned stable
genotypes, G45, G26, G27, G35 and G34 exhibited above-
average grain yield and indicated that these genotypes
were well adaptable to saline environment condition.
The biplot 1 (Fig. 1) indicated that, environments
were scattered without any grouping among them. This
indicated that, the test site also showed year to year
variation in mean grain yield and were evenly distributed
in all four quadrants indicating the importance of seasonal
climatic variation. This clearly suggested that, two or
more seasons of testing are better than a single year [16].
Among the three environments, E1 and E3 differed in
interaction but not main effect.
The genotype G36, followed by G19, could be
identified to the favourable environments with high
mean and high IPCA score. Similar signs of IPCA 1 score
for both genotype and environment imply positive
interaction and thus it attributes to higher yield of
genotype at that particular environment [13 , 17]. For
instance, the E2 is the most favourable for the genotypes,
G46, G41, G43, G42, G7, G22, G18, G33, G19 and G36. The
favourable response environment E2 might have been
due to evenly distribution of rainfall, this probably
reduced the salt toxicity by leaching.
The AMMI analysis appeared as a more useful
methodology in quantifying G x E and have greater insight
into the magnitude and nature of G x E interaction. This
model is effective in characterizing genotypes and
environments as well as interactive and discriminative
pattern of genotypes and environments respectively.
REFERENCES
1. Gregorio, G.B., D. Senadhira, R.D. Mendoza,
N.L. Manigbas, J.P. Roxas and C.Q. Guerta, 2002.
Progress in breeding for salinity tolerance and
associated abiotic stresses in rice. Field Crops Res.,
76: 91-101.
2. Gregorio, G.B., D. Senadhira and R.D. Mendoza, 1997.
Screening rice for salinity tolerance. IRRI Discussion
Paper Series No. 22, International Rice Research
Institute, Manila, Philippines.
3. Lin, C.S., M.R. Bains and L.P. Lefkovitvh, 1986.
Stability analysis: Where do we stand? Crop Sci.,
26: 894-900.
4. Piepho, H.P., 1999. Stability analysis using the SAS
system. Agron. J., 91: 154-160.

46
 Advan. Biol. Res., 3 (1-2): 43-47, 2009
5. Laxmi, R.R. and R. Gupta, 2000. Different approaches
for stability measures in relation to genotype-
environment interaction. Crop Res., 20(1): 118-125.
6. Zobel, R.W., M.J. Wright and H.G. Gauch, 1988.
Statistical analysis of yield trials. Agron. J.,
80: 338-393.
7. Guach, H.G., 1988. Model selection and validation for
yield trials with interaction. Biometrics, 88: 705-715.
8. Van Oosterom, E.J., D. Kleijn, S. Ceccarelli and
M.M. Nachit, 1993. Genotype-environment
interactions of barley in the Mediterranean region.
Crop Sci., 33: 669-74.
9. Gauch, J.H.G. and R.W. Zobel, 1997. Identifying
mega-environments and targeting genotypes. Crop
Sci., 37: 311- 326.
10. Vijayakumar, C.H.M., M.I. Ahmed, B.C. Viraktamath,
R. Balakrishnan and M.S. Ramesha, 2001. Genotype
x environment interaction effects on yield of rice
hybrids in India. Indian Journal of Genetics and
Plant Breeding, 61(2): 101-106.
11. Anandan, A. and R. Eswaran, 2006. Genotype by
environment interactions of rice (Oryza sativa L.)
hybrids in the east coast saline region of Tamil
Nadu. In the Proceedings of 2nd International Rice
Congress, pp: 226.
47
12. Crossa, J., 1990. Statistical analysis of multilocation
trials. Adv. Agron., 44: 55-85.
13. Crossa, J., P.N. Fox, W.H. Pfeiffer, S. Rajaram and
H.G. Gauch, 1991. AMMI adjustment for statistical
analysis of an international wheat yield trial.
Theoretical and Applied Genetics, 81: 27-37.
14. Asenjo, C.A., R. Bezus and H.A. Acciaresi, 2003.
genotype-environment interactions in rice (Oryza
sativa L.) in temperate region using the joint
regression analysis and AMMI methods. Cereal Res.
Communi., 31 (1-2): 97-104.
15. Gauch, H.G., 1992. Statistical analysis of regional
yield trials: AMMI analysis of factorial designs.
Elsevier Science publishers, Netherlands.
16. Debelo, D. S. Gelalcha, B. Yaie, B. Girma, B. Mamo
and D. Masresha, 2004. Grain yield stability of bread
wheat genotypes in favorable and stressed
environments. In 12th Regional Wheat Workshop for
Eastern, Central and Southern Africa, pp: 76-85.
17. Nimbalkar, C.A., A.P. Baviskar and P.A. Navale,
2004. Genotype × environment interaction effect
on seed yield of French bean. Indian J. Agril.
Sci., 74 (7): 366-369.