You are on page 1of 14

Evaluation of genotype environment interaction and

stability of corn hybrids and relationship among


univariate parametric methods
Mahdi Changizi1, Rajab Choukan2, Eslam Majidi Heravan1,
Mohammad Reza Bihamta3, and Farrokh Darvish1
1
Department of Plant Breeding, Science and Research Branch, Islamic Azad University,
Tehran, Iran (e-mail: mahd.changizi@gmail.com); 2Seed and Plant Improvement Institute, Karaj, Iran; and
3
Department of Agronomy and Plant Breeding, University of Tehran, Iran.
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

Received 7 December 2013, accepted 5 May 2014. Published on the web 20 May 2014.
Changizi, M., Choukan, R., Heravan, E. M., Bihamta, M. R. and Darvish, F. 2014. Evaluation of genotypeenvironment
interaction and stability of corn hybrids and relationship among univariate parametric methods. Can. J. Plant Sci. 94:
12551267. There have been many approaches available in multi-location crop variety trial. However, the relationship
among these approaches is not understood. In this study, therefore, grain yields of 16 corn hybrids were measured in
12 locations in Iran in 2011 and 2012 in order to compare the 23 parametric methods and to assess stability and
adaptability of the hybrids. The combined ANOVA indicated that variances due to the genotypes, environments
and genotype environment interaction were substantially significant, which represents great variation among them.
Principal component analysis based on rank correlation matrix indicated that stability methods can be classified into
four groups. The group related to the dynamic concept and strongly associated with mean grain yield consisted of the
measures, superiority index (Pi), desirability index (DI), geometric adaptability index (GAI) and genotypic stability (Di2).
This group was more useful in agronomic goals in comparison with other methods. The second group also indicated
the dynamic concept contained slope of regression models. The third group reflected the static concept included, the
environmental variance (EV), the variance in regression deviation (S2di) and type IV stability concept (S2y(l)). The fourth
For personal use only.

group impressed concurrently by grain yield and stability included the measures coefficient of variability (CV), Wricks
ecovalence (W2), Shuklas stability variance (SH), Plaisted and Petersons parameter (pp59), Plaisteds parameter (p60),
yield reliability index (Ii), residual MS of regression models and coefficient of determination (R2). Based on both
concepts of stability (dynamic and static), hybrids (KLM76002/3MO17), (KLM77002/10-5-1 K19/1) and (K47/2 
MO17) were the most stable and (KSC704), (KSC720 (K74/1K19)) and (K48/3 K18) were found to be the most
adaptable to favorable environments. The methods of Pi, Di2, DI and GAI were more useful and more convenient than
other methods. S2y(l)a and S2y(l)b showed an acceptable static concept of stability methods whereas study S2y(l)b was more
efficient than S2y(l)a.

Key words: Genotype environment interaction, stability, univariate stability methods,


cluster analysis, Zea mays (L.)

Changizi, M., Choukan, R., Heravan, E. M., Bihamta, M. R. et Darvish, F. 2014. Evaluation de linteraction genotype 
environnement et de la stabilite des hybrides du mas, et lien entre les methodes danalyse unidimensionnelles. Can. J. Plant Sci.
94: 12551267. Les essais de varietes realises a divers endroits recourent a maintes approches. Cependant, on saisit mal les
liens qui existent entre ces approches. Dans la presente etude, les auteurs ont mesure le rendement grainier de 16 varietes
hybrides de mas a 12 endroits, en Iran, en 2011 et 2012, avec pour objectifs de comparer les vingt-trois methodes
parametriques ainsi que devaluer la stabilite et la capacite dadaptation des cultivars. Lanalyse combinee de la variance
indique que la variance attribuable aux interactions du genotype, de lenvironnement et genotypeenvironnement est
nettement significative, ce qui signifie que les resultats varient beaucoup de lune a lautre. Lanalyse en composantes
principales, qui repose sur une grille de correlation des rangs revele quon peut repartir les methodes de stabilite en quatre
groupes. Le groupe de la stabilite dynamique, fortement associe au rendement grainier moyen, comprend les mesures,
lindice de superiorite (Pi), lindice de caractere desirable (DI), lindice geometrique de la capacite dadaptation (GAI) et la
stabilite du genotype (Di2). Ce groupe savere plus utile que les autres pour les objectifs agronomiques. Le deuxieme groupe
integre aussi le concept de stabilite dynamique exprime par la pente des modeles de regression. Le troisieme reflete
davantage le concept de stabilite statique et englobe la variance environnementale (EV), la variance de lecart de regression
(S2di) et le concept de stabilite de type IV (S2y(l)). Enfin, le quatrieme groupe, affecte a la fois par le rendement grainier et la
stabilite, comprend le coefficient de variabilite (CV), lecovalence de Wrick (W2), la variance de la stabilite de Shukla (SH),
le parametre de Plaisted et Peterson (pp59), le parametre de Plaisted (p60), lindice de fiabilite du rendement (Ii), la

Abbreviations: EV, environmental variance; GEI, genotype 


environment interaction; PCA, principal component analysis

Can. J. Plant Sci. (2014) 94: 12551267 doi:10.4141/CJPS2013-386 1255


1256 CANADIAN JOURNAL OF PLANT SCIENCE

variance residuelle des modeles de regression et le coefficient de determination (R2). Dapres les deux concepts de stabilite
(dynamique et statique), les hybrides (KLM76002/3MO17), (KLM77002/10-5-1 K19/1) et (K47/2MO17) saverent
les plus stables, tandis que les hybrides (KSC704), (KSC720 (K74/1K19)) et (K48/3K18) sont ceux qui sadaptent le mieux
a des conditions favorables. Les methodes Pi, Di2, DI et GAI sont plus utiles et plus commodes que les autres. S2y(l)a et S2y(l)b
offrent un concept statique acceptable pour les methodes de stabilite, la seconde etant plus efficace que la premiere.

Mots cles: Interaction genotype environnement, stabilite, methodes de stabilite unidimensionnelles,


analyse par grappe, Zea mays (L.)

Corn (Zea mays L.) production and trading mainly and biological limitations in the regression approach
aim for livestock feeding; however, corn is an important (Lin et al. 1986; Westcott 1986; Crossa 1990; Flores 1998).
and integral part of the human food basket as well. The other univariate stability parameters are Hansons
Given the importance of corn and its planting potential (1970) genotypic stability index (Di2), Hernandezs (1993)
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

throughout the world and in Iran, the sustainability desirability index (DI), Kataokas (1963) yield reliability
of its yield has been seriously taken into account in index (Ii), Pinthuss (1973) coefficients of determination
previous research. The features of stable genotype are (R2), geometric adaptability index (GAI) (Mohammadi
complex due to genotype environment interactions and Amri 2008), Lin and Binns (1988) superiority index
(GEI) (Alwala 2010; Moghaddam et al. 2013). As a (Pi), and type IV stability concept (S2y(l)a) and (Sy(l)
2
b).
result, plant breeders target GEI and undertake multi In the present study, the experiments were carried out
environment trials in appropriate and inappropriate during 2 yr in 12 locations under the macro-environment
environments (Alwala 2010). In this way, different res- condition and the aims of the study were to assess the
ponses of genotypes for changing environmental condi- stability and adaptability of desired hybrids in order to
tions are used to estimate the mean yield and to identify introduce stable genotypes and also the evaluation of
high yield and stable genotypes (Yates and Cochran stability assessment methods by comparing their relative
1938; Moghadam et al. 2012; Tsegaye et al. 2012). It efficiency and relationships.
can be considered that the GEI are due to predictable
For personal use only.

and unpredictable effects (Allard and Bradshow 1964).


The former effects are due to macro-environment and MATERIALS AND METHODS
the latter ones are attributable to temperature and micro- Sixteen corn hybrids (Table 1) were tested during 2011
environment (Abo-Hegazy 2013). In addition, the adapt- 2012 at 12 locations including Karaj, Shiraz, Gharakhil-
Ghaemshahr, Moghan, Khorramabad, Miandoab,
ability of a cultivar results from its stability under
Mashhad, Gorgan, Dezfoul, Isfahan, Jiroft and Ilam
different conditions (De Souza Goncalves et al. 2003;
in Iran, The environmental characteristics of these loca-
Rasul et al. 2005; Sharma et al. 2010; Moghaddam
tions are presented in Table 2. A randomized com-
et al. 2011). There are various methods for describing
plete block design with four replicates was used at
the effects of GEI along with identifying and recom-
each location. The plot size was as follows: four rows
mending stable genotypes in breeding programs. The con-
ventional methods are grouped in parametric (univariate
and multivariate) and nonparametric approaches based Table 1. Code, hybrids, mean yield (t ha 1) and grouping means by
on different strategies. The univariate parametric strategy Duncans test (at the 5% level)
includes the methods which are based on variance com-
ponents and joint regression. The combined variance Code Hybrids Grain mean yield (t ha 1)
analysis is the most common method used for the G1 KLM 76002/3MO17 10.002a
identification of GEI in repeated multi-environment G2 KLM 77002/10-5-1K19/1 10.003a
trials. Environmental variance (EV) [Roemer (1917), G3 K48/3K19/1 9.378cde
as reported in Becker and Leon (1988)], Francis and G4 K3612/2K19 8.523g
G5 K3651/2K19 8.726gf
Kanenbergs (1978) coefficient of variability (CV), G6 K3651/1K19/1 9.261de
Wrickes (1962) ecovalence (W2), Shulkas (1972) stabi- G7 KLM 77002/10-2-3K19/1 8.965efg
lity variance (SH), Plaisted and Petersons (1959) mean G8 K47/2MO17 9.933ab
variance component for genotype environment interac- G9 KLM 77007/7K3615/2 9.253de
G10 KLM 77002/10-6-1K19/1 9.09def
tion (GEI) (PP59), Plaisteds (1960) variance compo- G11 K47/22A679 9.495bcd
nent for GEI (P60), regression analysis as detailed by G12 K48/3K18 9.897ab
Finlay and Wilkinson (1963) (FW), Eberhart and G13 KLM 78027/2MO17 8.748gf
Russells (1966) deviation from the regression line G14 KLM 78018/6MO17 8.719gf
G15 KSC 720 (K74/1K19) 9.921ab
(S2di) and regression coefficient (bi), Perkins and Jinks G16 KSC 704 9.779abc
(1968) (PJ), and Freeman and Perkins (1971) (FP)
methods have been widely used as important measures ag Grouping mean yield of hybrids by Duncans test at the 0.05 level;
of stability. Though there are well-recognized statistical similar letters indicate similar grouping.
CHANGIZI ET AL. * GE INTERACTIONS AND STABILITY OF CORN HYBRIDS 1257

Table 2. Geographical positions and information of where the experiments were conducted

Code Growing season Environment Longitude Latitude Rainfall mm Mean yield (t ha1)

E1 20102011 Karaj 51.00E 35.49N 320 12.134


E2 20102011 Shiraz 52.43E 46.25N 286.6 12.021
E3 20102011 G-Ghaemshahr 52.25E 31.28N 808.4 8.152
E4 20102011 Moghan 47.32E 39.41N 376.7 9.417
E5 20102011 Khoramabad 48.2E 33.29N 287.9 5.33
E6 20102011 Miandoab 32.24E 46.06N 287.5 16.476
E7 20102011 Mashhad 59.38E 36.16N 250 13.593
E8 20102011 Gorgan 54.25E 36.45N 309 7.323
E9 20102011 Dezful 32.24E 48.28N 290 8.138
E10 20102011 Isfahan 51.51E 32.51N 164 5.512
E11 20102011 Jiroft 56.17E 26.43N 68.8 7.434
E12 20102011 Ilam 46.16E 33.5N 298 8.194
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

E13 20112012 Karaj 51.00E 35.49N 274 9.375


E14 20112012 Shiraz 52.43E 46.25N 226.5 10.179
E15 20112012 G-Ghaemshahr 52.25E 31.28N 585 6.285
E16 20112012 Moghan 47.32E 39.41N 275.7 9.35
E17 20112012 Khoramabad 48.2E 33.29N 165.3 5.67
E18 20112012 Miandoab 32.24E 46.06N 200 14.607
E19 20112012 Mashhad 59.38E 36.16N 288 15.347
E20 20112012 Gorgan 54.25E 36.45N 400 6.093
E21 20112012 Dezful 32.24E 48.28N 254 4.579
E22 20112012 Isfehan 51.51E 32.51N 126.3 10.909
E23 20112012 Jiroft 56.17E 26.43N 175.6 9.964
E24 20112012 Ilam 46.16E 33.5N 142.5 8.455

(furrows), 6 m length, 0.75 m apart and 0.35 m between The stability analysis was used according to 23 uni-
For personal use only.

hills and plant density fixed at 76 000 plants ha 1. All of variate stability methods as follows: Environmental
the common practices were followed throughout the variance [Roemer (1917), as reported in Becker and Leon
growing season. The inner two rows of each plot were (1988)] indicates the variance of each genotype across
harvested and grain production was measured based on all of the tested environments as phenotype stability.
14% of grain moisture. Francis and Kanenbergs (1978) coefficient of variability
indicates the general CV% of each genotype as a stability
Statistical Methods measure (Lin and Binns 1986). Wrickes (1962) equiva-
At each location, the statistical design analysis was lence (W2) reveals the GE interaction for genotype (i)
carried out for determining the mean yield. Bartletts across all environments squared and summed as a geno-
test was employed to test the homogeneity of error mean type stability measure (Lin and Binns 1986; Mohammadi
squares. The yield data were subjected to ANOVA by et al. 2012). Lin and Binns (1988) introduced the super-
using the following model: iority index (Pi) as the general superiority of genotype
and defined it as the mean square distance between the
Yijkr  mYk Lj LYjk Rr (Lj Yk )Gi GLij genotypes response and the maximum response at all
GYik GLYijk eijkr locations (Mohammadi et al. 2012).
Plaisted and Peterson (1959) proposed the variance
where Yijkr is the response variable of ith genotype at jth mean for each pair of genotypes in order to estimate
location, kth year, and rth replication, m is the overall the interaction variance (pp59). Plaisted (1960) intro-
mean, y is the main effect of year k, L is the main effect duced the variance-based contribution of each genotype
of location j, LY is the interaction effect of location-by- in forming the genotype environment interaction as
year jk, R is the effect of replication (r) within location the stability parameter (p60). Hansons (1970) genotypic
and year (Lj Yk), G is the main effect of genotype i, GL is stability parameter (Di2) is derived from regression ana-
the interaction effect of genotype-by-location ij, GY is lysis, since it uses the minimum slope from the Finlay
the interaction effect of genotype-by-year ik, GLY is the and Wilkinson (1963) method. Hernandezs (1993) pro-
interaction effect of genotype-by-location-by-year ijk posed a desirability index (DI) that would combine both
and e is the random error ijkr. yield and regression coefficient (Dehghani et al. 2008).
In the above-mentioned model, genotypes were se- Geometric Adaptability Index (GAI) (Mohammadi and
lected by Seed and Plant Improvement Institute, Karaj, Amri 2008) is used to evaluate the adaptability of
Iran, and locations were considered as fixed factors. genotypes and it is calculated as follows:
Because the weather conditions can change from year
to year, the effect of year was considered as a random p
E

factor. GAI  (X 1:)(X 2:) . . . (X i:)
1258 CANADIAN JOURNAL OF PLANT SCIENCE

Where, X 1:; X 2: and X i: are the mean yields of the first, interaction with the environment employing joint re-
second and ith genotypes across all environments and E gression; after that Finlay and Wilkinson (1963) and
is the number of environments (E: locationyear). then Eberhart and Russell (1966) modified the method
The genotypes with the higher GAI will be acceptable (Tiawari et al. 2011). Then, it was further developed by
(Mohammadi and Amri 2008; Pourdad 2010). Based on Perkins and Jinks (1968), and also by Freeman and
Kataoka index, Eskridge (1990) calculated the lowest Perkins (1971). In the Finlay and Wilkinson (1963)
yield expected for a specified probability (say 0.95). regression model (FW), the difference between the grand
In this way, for each genotype i, the index is given as mean of the environments and the general mean is
follows: shown by the regression of observations and environ-
mental indices. Eberhart and Russell (1966) further
(Ii mi Z(p) Si )
developed FWs regression model of stability and pro-
where, mi is the mean yield, Si is the square root of posed the use of two stability measures (S2di and bi)
stability variance, and Z(p) is the percentage of normal when determining the performance of one cultivar across
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

distribution at the probability level [Food and Agricul- various environments. The Perkins and Jinks (1968)
ture Organization of the United Nations (FAO) 2002] regression coefficient (PJ) is similar to the FW method,
for which the cumulative distribution function reaches but the observations are adjusted for site effects. Tai
the value P. Z(p) can assume the following values depe- (1971) proposed two measures of stability (ai and li).
nding on the chosen P level: 0.675 for P0.75; 0.840 for These two stability parameters are very similar to the
P 0.80; 1.040 for P 0.85; 1.280 for P 0.90; and regression coefficient and the deviation from regression
1.645 for P 0.95 (FAO 2002). of Eberhart and Russell, but are attained by a method
This reliability index, Ii, gives the lowest yield ex- that is a continuance of the analysis of variance. They
pected in l00P% cases. The selection based on Ii gives are obtained using the principle of structural relation-
an confidence that in 100P% cases the yield would be ships (Dehghani et al. 2008). Pinthus (1973) (R2)
greater than Ii and selection of genotype can be based on proposed that the ability of the linear regression model
the greatest Ii (Shulka 2004). for definition of the observed variation could be
The residual MS of Perkins and Jinks (1968) (MSPJ) characterized using coefficients of determination which
For personal use only.

and Freeman and Perkin (1971) (MSFP) regression are computed by an individual linear regression analysis
models were used as the stability parameter. Lin and (Mohammadi et al. 2012).
Binns (1988) proposed type IV stability concept on the
basis of predictable and unpredictable component non- Rank Correlation Coefficient and Principal
genetic variation; predictable component and unpredict- Component Analysis
able components are related to the location and year, The stability methods were evaluated using Spearmans
respectively (FAO 2002). In fact, the estimation of this rank. In this method, the linear relationship between the
variation as provided by My(L) is inflated by the ex- considered methods is evaluated based on the computed
perimental error variance. The actual variance of this parameters ranking. The other approach is based on
effect (S2y(l)) could be estimated as follows: principal component analysis (PCA) (Piepho and Lotto
S2y(l)  My(l) (Merr =r) 1992). For this purpose, the biplot analysis was carried
out with the rank correlation matrix to determine the
Where, Merr is pooled error (i.e., average experimental relationship between the stability parameters (Lipkovich
error for the genotypes) in the combined ANOVA, and r and Smith 2002; Kilic et al. 2010). Most of the analyses
is the number of experiment replicates (in the present were undertaken by employing SAS software V.9.1.3
paper, it is shown as (S2y(l)a)). (SAS Institute, Inc.) as well as SPSS V.16.0 (SPSS Inc.,
S2y(l) values can also be estimated through a hierarch- 2007) and Minitab V.16.2.1 (MINITAB Inc., 2010).
ical ANOVA performed on plot values of each geno-
type, which includes the MS value for the replicate RESULTS
within years as a source of variation (Mr(y)). In this case,
it is shown as follows: ANOVA Results and Joint Regression Analysis
The results of variance analysis are presented in (Table 3).
S2y(l)  (My(l) Mr(y) )=r The effect of year indicated no significant difference
2 while the main effect of location and genotype were
(In this paper it is represented as (Sy(l) b)). significant (PB0.01) indicating the differences in geno-
2
The current estimates of Sy(l) values may differ slightly types and locations. The highest difference is determined
from the estimate obtained from the previous formula for the main effect of location (65.58%), while the
(FAO 2002). differences related to G, G L, G Y, and G Y L
totally reach (8.579%), which demonstrates the diversity
Joint Regression of the tested environments. A comparable percent has
Yates and Cochran (1938) explained, for the first time, already been reported by Alwala (2010) and Namorato
the combined evaluation of phenotypic value and its et al. (2009). As is indicated in Table 1, the highest mean
CHANGIZI ET AL. * GE INTERACTIONS AND STABILITY OF CORN HYBRIDS 1259

Table 3. Combined analysis of variance for yield data of 16 corn between these parameters were very high and equal
genotypes tested across 12 locations in two years to 1 (Table 6). Karimizadeh et al. (2012) proposed
that the above parameters emphasize the gen-
Total variation
SOV DF SS MS explained (%)
otype environment interaction and establish the
stable genotypes with lower contribution in the
Year (Y) 1 25.034 25.034 NS 0.12 interaction. In addition, Mohebidini et al. (2006)
Location (L) 11 13659.639 1241.785** 65.586 and Kilic (2010) reported a strong relationship
LY 11 2264.632 205.876** 10.874
R (LY) 72 551.409 7.658 2.648
between the above parameters using the correlation
Genotype (G) 15 394.345 26.29** 1.893 method. Based on the superiority index (Pi) (Lin
GL 165 772.427 4.681** 3.709 and Binns 1998), a genotype is stable if it has a low
GY 15 78.351 5.223** 0.376 Pi (Kilic 2010). In this term, G8, G15 and G2 were
GLY 165 541.787 3.284** 2.601 the most stable genotypes respectively. And, G10,
Error 1080 2539.427 2.351 0.154
Corrected total 1535 20827.051 G7, G14, G5, G13 and G4 were not acceptable
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

R2 0.878071 because of their high Pi. In desirability index (DI)


(Hernandez et al. 1993) G12, G15 and G8 had the
** Significant at the 1% level; NS, not significant. highest DI values as well as average yield, and were
stable, but G4, G13, G14, and G5 were unstable.
Based on Hansons (1970) genotypic stability para-
yields are attributed to G2, G1, G12, G8, G15 and G16, meter (D2i ), G1, G8 and G2 had the lowest Di2
respectively; their mean yields are classified in a statistic values and thus were stable, while G14, G4, G5,
group using Duncan method. The computed para- and G10 had the highest Di2 values and were
meters, values and the rankings are illustrated in Tables unstable. The GAI is based on geometric mean of
4 and 5, respectively. Based on the EV (Roemer 1917) genotype; thus, genotypes with high GAI are accept-
G13, G4, and G1 were more stable and exhibited bio- able (Pourdad 2011). In this respect, G1, G2 and
logical stability. The results of the CV stability para- G8 were the most acceptable genotypes and G4,
meter showed little difference and G1, G2, and G13 were G14, G5, and G13 had the lowest GAI. Kataokas
For personal use only.

more stable and also based on Fig. 1 (CV vs. MEAN) (1963) yield reliability index (Ii) is obtained from
a combination of yield and stability. A genotype
the genotypes can be categorized in four groups:
is stable if it has a high Ii value. In this way, G16
and G8 were classified in the stable group and G11
1. Genotypes with relatively high mean yield and low
and G15 were unstable. Type IV stability (Lin and
coefficient of variability that are proposed as stable
Binns 1988) (S2y(l)) has been estimated with two
genotypes (group 1).
different formulas (FAO 2002), which are similar in
2. Genotypes with the mean yield higher than total
results. The lower the parameter, the more stable
average and high coefficient of variability (indicat-
the genotype. Accordingly, G4 was the most stable,
ing the specific stability (group 2)). Among them G3 while G3, G16 and G12 were the most unstable;
and G13 indicated the lower mean yield than others however, G16 and G12 had a relatively accept-
in this group; thus, they are not recommended. able mean yield but low stability. The most unstable
3. Genotypes with relatively low mean yield and also genotype was G3 and the most stable one was G4,
low coefficient of variability, which are indicated but these were not acceptable due to theri unac-
as stable genotypes due to their low coefficient of ceptable mean yields; these results were similar to
variability, but they were not acceptable because of environmental variance.
their low mean yield (group 3). Among them G4,
G13 and G7 indicated the lowest mean yield; thus, The results of the different linear regression stability
they are not recommended at all. parameters are given in Table 4. The deviations from
4. Genotypes with low mean yield and high coefficient linear regression of Eberhart and Russells (1966) (S2di)
of variability, which are not recommended due showed that G13 and G4 were stable genotypes and
to their low yield and high variability (group 4). based on slope (bi) they had specific adaptability to poor
Genotypes G6 and G8 revealed the best stability environments (b 0.852 for G13 and b0.865 for G4),
according to their SH values (Table 5) similar to but G16 had specific adaptability to favorable environ-
W2 results, but G6 was not acceptable because of ments (b 1.149). According to Eberhart and Russells
its low mean yield; where G3 and G11 had the (1966) method, the use of diagram (S2di vs. bi) is useful
lowest stability as well as average yield. Results of for a better understanding of genotypes stability [Haufe
Plaisteds (1960), and Plaisted and Petersons (1959) and Geidel (1978) as cited by Becker and Leon (1988)].
methods were very similar to Wrikes (1962) and Genotypes G13, G4, G1, G7, G2, and G6 are classified
Shulkas (1972) methods. Table 5 shows that the in the high yield stability group and were adapted
genotype rank, based on these four stability para- to low-yielding environment; they are classified in this
meters were similar and the correlation coefficients group due to their coefficient, which is less than 1
1260 CANADIAN JOURNAL OF PLANT SCIENCE
Table 4. Mean grain yield values (t ha) and 23 stability parameters of 16 corn genotypes across 24 environmentsz

MEAN EV CV W2 pp59 p60 SH FW PJ MSPJ bi S2di R2 FP MSFP Pi Di2 DI ai li Ii S2y(l)a S2y(l)b GAI

G1 10 10.02 31.67 21.72 1.01 1.01 1.01 0.92 0.081 0.91 0.919 10.4 0.913 0.92 0.91 1.7 59.5 11.08 0.0821 1.241 4.91 13.62 3.2 9.51
G2 10 10.45 32.29 24.19 1.07 1 1.13 0.93 0.066 1.05 0.934 10.87 0.9039 0.93 1.05 1.5 63.4 11.1 0.067 1.429 5.22 13.53 3 9.48
G3 9.38 13.37 38.93 34.16 1.3 0.97 1.63 1.05 0.048 1.53 1.048 13.95 0.891 1.05 1.52 2.8 89 10.61 0.048 2.08** 4 26.11 5.84 8.68
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

G4 8.52 8.91 34.97 19.1 0.95 1.02 0.88 0.87* 0.127* 0.68 0.873* 9.13 0.9265 0.87* 0.68 4.4 105 9.55 0.128* 0.931 8.43 5.79 1.19 8.06
G5 8.73 12.06 39.81 16.33 0.88 1.03 0.74 1.02 0.023 0.74 1.024 12.6 0.9415 1.02 0.74 4 100 9.93 0.023 1.001 9 16.11 3.57 8.12
G6 9.26 10.57 35.2 10.28 0.74 1.05 0.44 0.97 0.032 0.46 0.967 11.04 0.9587 0.97 0.46 2.6 63.7 10.4 0.033 0.619 5.46 12.19 2.14 8.72
G7 8.96 10.18 35.59 18.26 0.93 1.02 0.84 0.93 0.066 0.78 0.934 10.6 0.9268 0.93 0.78 3.5 82.6 10.06 0.067 1.060 4.18 16.18 3.45 8.44
G8 9.93 11.92 34.81 13.35 0.81 1.04 0.59 1.02 0.023 0.6 1.023 12.46 0.9518 1.02 0.6 1.3 59.7 11.14 0.023 0.818 9.35 12.36 2.11 9.35
G9 9.25 12.47 38.21 23.86 1.06 1 1.11 1.03 0.027 1.08 1.028 13.03 0.9176 1.03 1.08 2.7 83.2 10.46 0.028 1.464 7.13 9.59 2.33 8.65
G10 9.09 12.5 39 23.48 1.05 1 1.1 1.03 0.029 1.06 1.030 13.06 0.919 1.03 1.06 3.1 93.3 10.3 0.03 1.438 7.64 11.05 2.68 8.49
G11 9.49 14.09 39.57 31.62 1.24 0.98 1.5 1.09 0.086 1.35 1.087 14.65 0.908 1.09 1.35 2.3 88 10.77 0.087 1.84* 1.24 18.37 4.49 8.81
G12 9.9 13.81 37.45 25.14 1.09 1 1.18 1.09 0.086 1.06 1.087 14.35 0.9268 1.09 1.06 1.8 75 11.18 0.087 1.438 6 23.92 5.39 9.22
G13 8.75 8.47 33.22 14.98 0.85 1.03 0.67 0.86* 0.139** 0.46 0.86** 8.63 0.9479 0.86* 0.46 4 86.1 9.76 0.14* 0.627 6.13 13.11 1.87 8.31
G14 8.72 11.93 39.81 25.57 1.1 1 1.2 1 0.0007 1.16 0.999 12.47 0.907 1 1.16 4 109 9.9 0.0007 1.581 3.74 10.4 2.39 8.07
G15 9.92 12.73 36.05 20.74 0.99 1.01 0.96 1.05 0.05 0.92 1.046 13.29 0.931 1.05 0.92 1.4 68.5 11.15 0.046 1.25 1.61 18.61 4.62 9.3
G16 9.78 15.02 39.72 25.36 1.09 1 1.19 1.14* 0.14* 0.92 1.142* 15.48 0.9412 1.14* 0.92 1.7 81.8 11.12 0.143* 1.256 9.24 24.97 5.24 9.01

z
MEANgrain mean yield (t ha1); biregression coefficient of Eberhart and Russell; FWregression coefficient of Finlay and Wilkinson; PJregression coefficient of Perkins and
Jinks; FPregression coefficient of Freeman and Perkins; MSFPresidual MS of Freeman and Perkins model; MSPJresidual MS of Perkins and Jinks model; S2dideviation from
regression Eberhart and Russell; R2 Pinthuss coefficient of determination; W2 Wricks equivalence; EV environmental variance; CV coefficient of variability; SHShuklas stability
For personal use only.

variance; pp59Plaisted and Petersons stability parameter; p60 Plaisted stability parameter; DIHernandezs desirability index; a and l Tais stability parameters; Pi Lin and Binns
superiority index; Di2 Hansons genotypic stability parameter; Iireliability index; GAIgeometric adaptability index; Ii yield reliability index; (Sy(l)
2 2
a and Sy(l)b) Lin and Binns type
IV stability. *, ** Significant at the 0.05 and 0.01 probability levels, respectively.

Table 5. Rank of the 16 corn genotypes using mean grain yield and 23 univariate methods of stability in 24 environmentsz

MEAN EV CV W2 pp59 p60 SH FW PJ MSPJ bi S2di R2 FP MSFP Pi Di2 DI ai li Ii S2y(l)a S2y(l)b GAI

G1 2 3 1 8 8 8 8 14 14 7 14 3 12 14 7 5 1 6 14 7 11 9 9 1
G2 1 5 2 11 11 11 11 12 12 10 13 5 15 13 10 3 3 5 12 10 10 8 8 2
G3 8 13 11 16 16 16 16 4 4 16 6 13 16 4 16 10 12 8 4 16 13 16 16 9
G4 16 2 5 6 6 6 6 15 15 4 15 2 9 15 4 16 15 16 15 4 4 1 1 16
G5 14 9 16 4 4 4 4 8 8 5 9 9 4 8 5 15 14 13 8 5 3 10 11 14
G6 9 6 6 1 1 1 1 11 11 1 11 6 1 11 1 8 4 10 11 1 9 5 4 8
G7 12 4 7 5 5 5 5 13 13 6 12 4 7 12 6 12 8 12 13 6 12 11 10 12
G8 3 7 4 2 2 2 2 9 9 3 8 7 2 9 3 1 2 3 9 3 1 6 3 3
G9 10 10 10 10 10 10 10 7 7 13 7 10 11 7 13 9 9 9 7 13 6 2 5 10
G10 11 11 12 9 9 9 9 6 6 11 5 11 10 6 12 11 13 11 6 11 5 4 7 11
G11 7 15 13 15 15 15 15 2 2 15 2 15 13 3 15 7 11 7 2 15 16 12 12 7
G12 5 14 9 12 12 12 12 2 2 11 3 14 7 2 11 6 6 1 2 11 8 14 15 5
G13 13 1 3 3 3 3 3 16 16 2 16 1 3 16 2 14 10 15 16 2 7 7 2 13
G14 15 8 15 14 14 14 14 10 10 14 10 8 14 10 14 13 16 14 10 14 14 3 6 15
G15 4 12 8 7 7 7 7 5 5 8 4 12 6 5 8 2 5 2 5 8 15 13 13 4
G16 6 16 14 13 13 13 13 1 1 9 1 16 5 1 9 4 7 4 1 9 2 15 14 6

z
MEAN grain mean yield (t ha 1); biregression coefficient of Eberhart and Russell; FWregression coefficient of Finlay and Wilkinson; PJregression coefficient of Perkins and
Jinks; FPregression coefficient of Freeman and Perkins; MSFP residual MS of Freeman and Perkins model; MSPJ residual MS of Perkins and Jinks model; S2dideviation from
regression Eberhart and Russell; R2 Pinthuss coefficient of determination; W2 Wricks ecovalence; EV environmental variance; CV coefficient of variability; SHShuklas stability;
pp59Plaisted and Petersons stability parameter; p60 Plaisted stability parameter; DIHernandezs desirability index; a and l Tais stability parameters; PiLin and Binns
superiority index; Di2 Hansons genotypic stability parameter; Iireliability index; GAIgeometric adaptability index, Iiyield reliability indexl (S2y(l)a and S2y(l)b) Lin and Binns type
IV stability.
CHANGIZI ET AL. * GE INTERACTIONS AND STABILITY OF CORN HYBRIDS 1261
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

Fig. 1. Grain yield means versus CV% from 16 corn genotypes in 24 environments.
For personal use only.

(Fig. 2). Regarding the mean yield, G1 and G2 with high as stable (Table 4). However, G6, G8 and G13 were res-
yield were more acceptable than others. Genotype G14 pectively the genotypes with higher determination coef-
showed a general adaptability and is placed in the ficient. Genotypes G3 and G2 had a lower R2 (Table 5).
middle of the chart. Other genotypes were placed in the
low yield stability region, and were adapted to high- Methods Comparison and their Association
yielding environments (their coefficient was more than The rank correlations between the mean yield and
1). In terms of their mean yield, G8, G15, G12, and G16 stability methods are given in Table 6. To find out the
were more acceptable than other genotypes and, accord- relationships among the rank-based statistics, the PCA
ing to Duncans test, they are classified in the high mean was carried out on the rank correlation matrix. Figure 3
yield group, respectively. Coefficients of regression indicates the loading of the first two principal comp-
based on FW, PJ models and the ai parameter marked onents (PCs) of ranks in the stability methods. These
G16 as stable. Based on the PJ regression model, geno- first two PCs accounted for 77.5% of the variation of
type G16 gave a significant positive coefficient but G4 the original variance (55.4 and 22% by PC1 and PC2,
and G13 gave a significant negative coefficient. As a respectively). Based on biplot of both PC axes of the
stability measures, the area of biplot can be divided into
result, G16 was stable because of the high regression
four separate groups. The place of mean yield (MEAN),
coefficient and had specific adaptability to favorable
Pi, GAI, Di2 and DI were in the first group, indicating
environments. The results of independent linear regres-
that the selection of the most stable genotypes based
sion coefficient (Freeman and Perkins 1971) including re- on these measures caused high yield genotypes to be
gression coefficients and deviation mean square (Table 4) presented as most stable genotypes. In the second group,
imparted that G16 was stable. The assessments of the there were coefficients of FW, FP, PJ, bi and ai re-
parameters ai and li for genotypes are given in Table 4. gression models. The EV, S2y(l)a, S2y(l)b and S2di pro-
G16 was marked as stable and G3 and G11 regarded as cedures were in group 3. Group four consisted of 10
unstable based on the ai and li parameters, respectively. stability parameters including the W2, SH, pp59, p60, li,
Based on Pinthuss (1973) coefficients of determination MSPJ, MSFP, Ii, CV, and R2 procedures.
(R2), the stability parameter values are the predictability
of response estimates (R2 1) (Kilic 2010), and range DISCUSSION
from 0.891 to 0.958; the variation of mean yield is The stability analysis should be carried out in two steps:
explained by the genotype response across different first to distinguish high-yielding genotypes via ANOVA;
environments. None of the values of coefficient of then use of the stability measures (Elsahookie 1996).
determinations was significantly different from 1.0. In Analysis of variance can measure the main effects of
terms of this parameter, all genotypes could be regarded years, locations, genotypes and their interaction. In fact,
1262 CANADIAN JOURNAL OF PLANT SCIENCE
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

Table 6. Spearmans rank correlation coefcients between yields and stability parametric measures for 16 corn genotypes across 24 environmentsz
MEAN EV CV W2 SH pp59 p60 bi FW PJ FP ai S2di MSFP MSPJ li R2 Pi Di2 DI Ii S2y(l)a S2y(l)b GAI

MEAN 1
EV 0.256 1
CV 0.406 0.712** 1
W2 0.162 0.641** 0.474 1
SH 0.162 0.641** 0.474 1.000** 1
pp59 0.162 0.641** 0.474 1.000** 1.000** 1
p60 0.162 0.641** 0.474 1.000** 1.000** 1.000** 1
bi 0.294 0.979**0.653**0.535* 0.535* 0.535* 0.535* 1
FW 0.294 0.996**0.674**0.602* 0.602* 0.602* 0.602* 0.985** 1
For personal use only.

PJ 0.294 0.996**0.674**0.602* 0.602* 0.602* 0.602* 0.985** 1.000** 1


FP 0.265 0.991**0.682**0.579* 0.579* 0.579* 0.579* 0.985** 0.996** 0.996** 1
ai 0.294 0.996**0.674**0.602* 0.602* 0.602* 0.602* 0.985** 1.000** 1.000** 0.996** 1
S2di 0.256 1.000** 0.712** 0.641** 0.641** 0.641** 0.641**0.979**0.996**0.996**0.991**0.996** 1
MSFP 0.106 0.653** 0.532* 0.932** 0.932** 0.932** 0.932**0.568* 0.615* 0.615* 0.597* 0.615* 0.653** 1
MSPJ 0.115 0.658** 0.528* 0.937** 0.937** 0.937** 0.937**0.571* 0.621* 0.621* 0.603* 0.621* 0.658** 0.999** 1
li 0.115 0.658** 0.528* 0.937** 0.937** 0.937** 0.937**0.571* 0.621* 0.621* 0.603* 0.621* 0.658** 0.999** 1.000** 1
R2 0.096 0.159 0.121 0.796** 0.796** 0.796** 0.796**0.049 0.116 0.116 0.085 0.116 0.159 0.817** 0.814** 0.814** 1
Pi .944** 0.353 0.285 0.121 0.121 0.121 0.121 0.424 0.39 0.39 0.365 0.39 0.353 0.079 0.087 0.087 0.041 1
Di2 .856** 0.106 0.650** 0.279 0.279 0.279 0.279 0.026 0.057 0.057 0.065 0.057 0.106 0.329 0.319 0.319 0.259 0.826** 1
DI .906** 0.547* 0.121 0.259 0.259 0.259 0.259 0.597* 0.589* 0.589* 0.576* 0.589* 0.547* 0.215 0.23 0.23 0.004 0.929** 0.718** 1
Ii 0.135 0.112 0.021 0.429 0.429 0.429 0.429 0.056 0.062 0.062 0.056 0.062 0.112 0.465 0.469 0.469 0.542* 0.082 0.009 0.082 1
S2y(l)a 0.456 0.571* 0.209 0.385 0.385 0.385 0.385 0.532* 0.583* 0.583* 0.594* 0.583* 0.571* 0.262 0.277 0.277 0.046 0.382 0.241 0.574* 0.294 1
S2y(l)b 0.379 0.741** 0.45 0.606* 0.606* 0.606* 0.606* 0.679**0.737**0.737**0.747**0.737** 0.741** 0.550* 0.562* 0.562* 0.287 0.315 0.068 0.556* 0.368 0.903** 1
GAI .994** 0.229 0.424 0.109 0.109 0.109 0.109 0.276 0.268 0.268 0.241 0.268 0.229 0.053 0.062 0.062 0.043 0.944** 0.885** 0.897** 0.126 0.426 0.347 1

z
MEANgrain mean yield (t ha1); biregression coefficient of Eberhart and Russell; FWregression coefficient of Finlay and Wilkinson; PJregression coefficient of Perkins and
Jinks; FPregression coefficient of Freeman and Perkins; MSFPresidual MS of Freeman and Perkins model; MSPJ residual MS of Perkins and Jinks model; S2dideviation from
regression Eberhart and Russell; R2  Pinthuss coefficient of determination; W2 Wricks equivalence; EVenvironmental variance; CV coefficient of variability; SHShuklas stability
variance; pp59Plaisted and Petersons stability parameter; p60 Plaisted stability parameter; DIHernandezs desirability index; a and lTais stability parameters; PiLin and Binns
superiority index; Di2 Hansons genotypic stability parameter; Iireliability index; GAIgeometric adaptability index; Iiyield reliability index; (S2y(l)a and Sy(l)
2
b)Lin and Binns type
IV stability.
*, ** Significant at the 0.05 and 0.01 probability levels, respectively.
CHANGIZI ET AL. * GE INTERACTIONS AND STABILITY OF CORN HYBRIDS 1263
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

Fig. 2. Interpretation of parameters bi and S2di for the regression approach for 16 corn genotypes in 24 environments. [Adapted
from Haufe and Geidel (1978) as cited by Becker and Leon (1988).]
For personal use only.

the ANOVA is not convenient for the description of and Mohammadi et al. (2010) in wheat reported high
GE interactions, and the other statistical procedures are positive rank correlation between the Pi, GAI and mean
necessary to understand and describe GE interactions. yield, and they can be helpful for identifying high yield
Almost all breeders have used the term stability to and stable genotypes; also, Karimzadeh et al. (2012)
characterize a genotype which ever indicated a constant reported that mean yield and the Di2 and DI methods
yield under variable environmental conditions (Dehghani were in the same group of stability measures in durum
et al. 2008). This suggestion of stability concurs with wheat and can also be used for genotype selections
the concept of homeostasis broadly used in quantitative based on high performance and stability. The coeffi-
genetics, and might be considered as a biological (static) cients of FW, FP, PJ, bi and ai regression models
concept of stability (Becker and Leon 1988). Also, a geno- were in the second group (Group II); they were in the
type that combines both high mean yield and stability dynamic region and also placed near the mean yield area
performance together is favorable, and so it is suitable of biplot. Mohebodini et al. (2006), Dehghani et al.
over variable environmental conditions (Allard and (2008), Karimzadeh et al. (2012) and Mohammadi et al.
Bradshaw 1964). This suggestion for identifying favor- (2012) reported that the coefficients of regression models
able genotypes reveals the agronomic (dynamic) concept benefit from the dynamic concept of stability.
of stability. The agronomic concept of stability is pre- Group III consisted of the static stability concept as
ferable for plant breeders, because in this concept of EV, S2di, S2y(l)a and S2y(l)b methods. Due to the high
2 2
stability, it is not necessary for the genotypic response to correlation of (Sy(l) a) and (Sy(l) b) with environmental
environmental conditions to be equal for all genotypes. variance, these two methods could be appropriate
The relationships among the parametric stability mea- alternatives for static concept of stability methods and
2
sures are graphically displayed in a biplot of PC1 versus also because of more consistent results, Sy(l) b measure
PC2 in Fig. 3, where both axes were considered con- with static concept of stability methods was more
currently. Four groups in Fig. 3 can be classified: Group efficient than S2y(l)a. Mohammadi et al. (2012) in wheat
1 contained mean yield and the parametric measures and Karimzadeh et al. (2012) in durum wheat reported
Pi, GAI, Di2 and DI. These measures were positively high positive rank correlation between the EV and S2di.
correlated with each other and with mean yield (Table 6). Lin et al. (1986) stated that the types I and IV stability
2 2
This result indicates that the selection of stable geno- (Sy(l) a and Sy(l) b) are grouping in static concept of
types, based on these measures, caused high yield geno- stability. The stability measures of MSPJ, MSFP, W2,
types to be introduced as stable genotypes. In this SH, pp59, p60, li, R2, CV, and Ii were in group IV. The
assay, the four above-mentioned measures had the high correlation between W2, SH, pp59 and p60, and
highest efficiency for introducing high yield and stable the introduction of similar stable genotypes shows that
hybrids simultaneously. Pourdad (2010) in safflower using only one of them is sufficient. For this reason,
1264 CANADIAN JOURNAL OF PLANT SCIENCE
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

Fig. 3. Biplot of PC1 versus PC2 for 23 parametric methods of 16 corn hybrids stability.

EV and S2di could be considered as appropriate alter- stability methods, plant breeders have found some
For personal use only.

natives for each other. inconsistencies between the results, especially when the
Kang and Pham (1991) indicated that W2 showed a stability is classified in four groups (Lin and Binns
high correlation with SH. Flores et al. (1998) found that 1994). Numerous stability measures that were used in
the SH, S2di and li methods were related to each other. this study quantified genotype stability with considering
Lin et al. (1986) and Kang et al. (1987) suggested that the yield. Mohammadi and Amri (2008) and Dehghani
Wrickes ecovalence (W2) and stability variance (SH) et al. (2008) stated that the stability values do not
were the same; these two methods should not be used as provide adequate information for attaining definitive
separate methods, because the stability variance is a conclusion, and both stability and yield should be
coded value of ecovalence. There is also a high correla- considered concurrently. Accordingly, among genotypes
tion between these methods and the pp59 and p60 with high mean yield, G1 by the EV method and G1 and
measures. Mohammadi et al. (2012) stated that R2 G2 by the CV method were introduced as the most
showed a positive correlation with SH, W2, MSPJ and stable; G8 was the most stable based on W2, SH, pp59,
li, in bread wheat. Although the CV should be located
p60, li, Ii and the R2 procedures; and, based on the
in group III, but placed in group IV because of its high
slope of the regression models, G16 was the most stable;
correlation with MSPJ, MSFP and li; this group also
according to the evaluation methods of stability dy-
showed the static concept of stability. Dehghani et al.
(2008) reported that parameters MSPJ, MSFP, W2, SH, namic concept that strongly correlated with the yield
pp59, p60 and li were grouped in static concept of mean, the genotypes of G2, G1, G8, G12, G15 and G16
stability. Yield reliability index (Ii) was located in the were the most stable ones with a little difference. This
fourth group and showed the static concept of stability. little difference can be seen in Figs. 4 and 5. These
Pourdad (2010) in safflower and Mohammadi et al. genotypes had the highest grain yield among the corn
(2010) in barley reported that Ii was grouped in the genotypes studied (and also among them G1, G2
static concept by its correlation with EV and CV. and G8 were introduced as stable genotypes by static
However, in this study the Ii was placed in group IV methods). Among the high-yielding genotypes (Fig. 1)
by its correlation with R2. In other words, the 14 mea- those with lower CV, indicate greater stability and those
sures above (groups III and IV) follow the biological with greater CV indicate lower stability, but by using
stability concept, and the selection of stable genotypes, the plot of (S2di vs. bi) (Fig. 2) it is indicated that those
based on these measures, caused the introduction of genotypes with higher average yield and high CV could
stable genotypes that show static stability. be related to specific stability. Becker and Leon (1988)
One of the complex issues about stability concepts stated that definition of Eberhart and Russell (1966)
is that in reviewing rank correlation across different about regression model is in accordance with the
CHANGIZI ET AL. * GE INTERACTIONS AND STABILITY OF CORN HYBRIDS 1265
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16
For personal use only.

Fig. 4. Superiority index versus grain yield means of 16 corn genotypes in 24 environments and position of stable genotypes.

dynamic concept, but it is preferable to use the eco- In summary, according to Figs. 1 and 2 and pre-
valence instead, which combines bi and S2di, into one vious results, we can state that among high-yielding
parameter and this approach is schematically shown in genotypes, G1, G2 and G8, introduced as stable
(S2di vs. bi) plot. genotypes by both dynamic and static methods, were

Fig. 5. Desirability index versus grain yield means of 16 corn genotypes in 24 environments and proximity of stable genotypes.
1266 CANADIAN JOURNAL OF PLANT SCIENCE

the most stable, and G16, G15 and G12, with specific Eskridge, K. M. 1990. Selection of stable cultivars using a
stability, had a high mean yield and can be recom- safety-rst rule. Crop Sci. 30: 369374.
mended with regard to both stability and yield. Finlay, K. W. and Wilkinson, G. N. 1963. The analysis of
adaptation in a plant breeding programme. Aust. J. Agric.
Res. 14: 742754.
CONCLUSIONS Flores, F., Moreno, M. T. and Cubero, J. I. 1998. A comp-
In this study we found: (1) Since both performance and arison of univariate and multivariate methods to analyze envi-
stability must be considered simultaneously to exploit ronments, Field Crop. Res. 56: 271286.
the beneficial effects of GE interaction and to have more Food and Agriculture Organization of the United Nations. 2002.
accurate selection, Pi, Di2, DI and GAI methods were Genotype environment interactions: Challenges and oppor-
more useful and more convenient than other methods. tunities for plant breeding and cultivar recommendations.
(2) Most of the parametric stability measures in each FAO Plant Production and Protection Paper 174. FAO,
concept can be alternated by another; it is not necessary Rome, Italy.
to use all of them. (3) Using the plots of (CV vs. MEAN) Francis, T. R. and Kannenberg, L. W. 1978. Yield stability
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

and (S2di vs. bi) simultaneously can help improve the studies in short-season maize: I. A descriptive method for
interpreting quality of stability of genotypes based on grouping genotypes. Can. J. Plant Sci. 58: 10291034.
Freeman, G. H. and Perkins, J. M. 1971. Environmental and
both concepts. (4) Type IV stability parameter (S2y(l)a and
genotype-environmental components of variability VIII. Rela-
S2y(l)b), which previous research did not take seriously
tions between genotypes grown in different environments and
into account, was used, and its comparison with other measures of these environments. Heredity 27: 1523.
parameters showed that it could be an acceptable Hanson, W. D. 1970. Genotypic stability. Theor. Appl. Genet.
parameter for the determination of stability as well as 40: 226231.
2
static concept of stability methods and Sy(l) b measure Haufe, W. and Geidel, H. 1978. Zur beurteilung der ertagssi-
2
showed a better efficiency than Sy(l)a in this study. cherheit von sorten und zuchstammen.I. Denitionen. Stabi-
litatsparameter und deren nterpretationsmoglichkeiten. Z.
Panzenuchtg 80: 2437.
ACKNOWLEDGEMENTS Hernandez, C. M., Crossa, J. and Castillo, A. 1993. The area
The authors extend their appreciations to the Seed and under the function:an index for selecting desirable genotypes.
For personal use only.

Plant Improvement Institute, Karaj-Iran for technical Theor. Appl. Genet. 87: 409415.
supports in all eld and laboratory activities. Kang, M. S. and Pham, H. N. 1991. Simultaneous selection for
high yielding and stable crop genotypes. Agron. J. 83: 161165.
Abo-Hegazy, S. R. E., Selim, T. and Ashrie, A. A. M. 2013. Kang, M. S., Miller, J. D. and Darrah, L. L. 1987. A note
Genotype environment interaction and stability analysis for on relationship between stability variance and ecovalence.
yield and its components in lentil. J. Plant Breed. Crop Sci. 5: Heredity 78: 107112.
8590. Karimizadeh, R., Mohammadi, M., Sabaghnia, N., Shefazadeh,
Allard, R. W. and Bradshaw, A. D. 1964. Implications of M. K. and Pouralhosaini, J. 2012. Univariate stability analysis
genotype-environmental interactions in applied plant breeding. methods for determining genotype environment interaction
Crop Sci. 4: 503508. of durum wheat grain yield, Afr. J. Biotechnol. 11: 25632573.
Alwala, S., Kwolek, T., McPherson, M., Pellow, J. and Meyer, Kataoka, S. 1963. A stochastic programming model. Econo-
D. 2010. A comprehensive comparison between Eberhart and metrika 31: 181196.
Russel joint regression and GGE biplot analyses to identify Kilic, H., Akura, M. and Aktas, H. 2010. Assessment of
stable and high yielding maize hybrids. Field Crop. Res. 119: parametric and non-parametric methods for selecting stable
225230.
and adapted durum wheat genotypes in multi-environments.
Becker, H. C. and Leon, J. 1988. Stability analysis in plant
Not. Bot. Hortic. Agrobot. Cluj. 38: 271279.
breeding. Plant Breed. 101: 123.
Lin, C. S. and Binns, M. R. 1988. A superiority measure of
Crossa, J. 1990. Statistical analyses of multi-location trials.
Adv. Agron. 44: 5585. cultivar performance for cultivar location data. Can. J. Plant
Dehghani, H., Sabaghpour, S. H. and Ebadi, A. 2010. Study of Sci. 68: 193198.
genotype environment interaction for chickpea yield in Iran. Lin, C. S. and Binns, M. R. 1994. Concepts and methods for
Agron. J. 102: 18. analyzingregional yield trial data for cultivar and location
Dehghani, H., Sabaghpour, S. H. and Sabaghnia, N. 2008. selection. Plant Breed. Rev. 12: 271297.
Genotype environment interaction for grain yield of some Lin, C. S., Binns, M. R. and Lefkovitch, L. P. 1986. Stability
lentil genotypes and relationship among univariate stability analysis: where do we stand? Crop Sci. 26: 894900.
statistics. Span. J. Agric. Res. 6: 385394. Lipkovitch, I. and Smith, E. P. 2002. Biplot and SVD macros
De Souza Gonzales, P., Bortoletto, N., Lucio Mello Martins, for EXCEL. J. Stat. Softw. 7: 5.
A., Brito Da Costa, R. and Boller Gallo, P. 2003. Genotype- MINITAB Inc. 2010. Ver. 16. Minitab, Inc., State College, PA.
environment interaction and phenotypic stability for girth Moghadam, A., Pietsch, G., Ardakani, M. R., Raza, A.,
growth and rubber yield of Hevea clones in Sao Paulo State, Vollmann, J. and Friedel, J. K. 2011. Genetic diversity and
Brazil. Genet. Mol. Biol. 26: 441448. distance among Iranian and European alfalfa (Medicago sativa
Eberhart, S. A. and Russell, W. A. 1966. Stability parameters L.) genotypes. Crop Breed. J. 1: 1328.
for comparing varieties. Crop Sci. 6: 3640. Moghaddam, A., Raza, A., Vollmann, J., Ardakani, M. R.,
Elsahookie, M. M. 1996. Applications on stability analysis of Wanek, W., Gollner, G. and Friedel, J. K. 2013. Carbon isotope
genotypes. Iraqi J. Agric. Sci. 27: 1120. discrimination and water use efciency relationships of alfalfa
CHANGIZI ET AL. * GE INTERACTIONS AND STABILITY OF CORN HYBRIDS 1267

genotypes under irrigated and rain-fed organic farming. Eur. J. Pourdad, S. S. 2011. Repeatability and relationships among
Agron. 50: 8289. parametric and non-parametric yield stability measures in
Moghaddam, A., Vollmann, J., Wanek, W., Ardakani, M. R., safower (Carthamus tinctorius L.) genotypes. Crop Breed. J.
Raza, A., Pietsch, G. and Friedel, J. K. 2012. Suitability of 1: 109118.
drought tolerance indices for selecting alfalfa (Medicago sativa Rasul, S., Khan, M. I., Javed, M. M. and Haq, I. 2005. Stability
L.) genotypes under organic farming in Austria. Crop Breed. J. and adoptability of Maize Genotypes in Pakistan. J. Appl. Sci.
2: 7989. Res. 1: 307312.
Mohammadi, R. and Amri, A. 2008. Comparison of parametric Roemer, J. 1917. Sinde die ertagdreichen Sorten ertagissi-
and non-parametric methods for selecting stable and adapted cherer? Mitt DLG 32: 8789.
durum wheat genotypes in variable environments. Euphytica Sharma, C. S., Morgounov, A. I., Braun, H. J., Akin, B., Keser,
159: 419432. M., Bedoshvil, D., Bagci, A., Martius, C. and Van Ginkel, M.
Mohammadi, M., Karimzadeh, R., Sabaghnia, N. and 2010. Identifying high yielding stable winter wheat genotypes
Shefazadeh, M. K. 2012. Genotype environment interaction for irrigated environments in Central and West Asia. Euphy-
and yield stability analysis of new improved bread wheat
tica 171: 5364.
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16

genotypes. Turk. J. Field Crops. 17: 6773.


Shulka, G. K. 1972. Some aspects of partitionity genotype-
Mohammadi, R., Roostaei, M., Ansari, Y., Aghaee, H. and
Amri, A. 2010. Relationships of phenotypic stability measures environmental components of variability. Heredity 28: 237
for genotypes of three cereal crops. Can. J. Plant Sci. 90: 245.
819830. Shulka, G. K. 2004. On the choice of genotype using location
Mohebodini, M., Dehghani, H. and Sabaghpour, S. H. 2006. specic information when stability variances are heteroge-
Stability of performance in lentil (Lens culinaris Medik) neous. J. Ind. Soc. Ag. Stat. 57: 94102.
genotypes in Iran. Euphytica 149: 343352. SPSS Inc. 2007. SPSS for Windows, Version 16.0. SPSS Inc.,
Namorato, H., Miranda, G. V., Souza, L. V., Oliviera, L. R., Chicago, IL.
Delima, R. O. and Mantovani, E. E. 2009. Comparing biplot Tai, G. C. C. 1971. Genotypic stability analysis and applica-
multivariate analysis with Eberhart and Russell method tion to potato regional trials. Crop Sci. 11: 184190.
for genotype environment interaction. Crop Breed. Appl. Tiawari, D. K., Pandey, P., Singh, R. K., Singh, S. P. and
Biotechnol. 9: 299307. Singh, S. B. 2011. Genotype environment interaction and
Perkins, J. M. and Jinks, J. L. 1968. Environmental and stability analysis in elite clones of sugarcane (Saccharum
genotype-environmental components of variability. Heredity ofcinarum L.). Int. J. Plant Breed. Genet. 5: 9398.
For personal use only.

23: 339356. Tsegaye, D., Wuletaw, T. and Muluken, B. 2012. Genotype 


Piepho, H. P. and Lotto, S. 1992. Rank correlation among environment interactions and grain yield stability of haricot
parametric and nonparametric measures of phenotypic stability. bean varieties in Northwest Ethiopia. Sci. Res. Essays 7: 3487
Euphytica 64: 221225. 3493.
Pinthus, J. M. 1973. Estimate of genotype value: a proposed Westcott, B. 1987. A method of assessing the yield stability of
method. Euphytica 22: 121123. crop genotypes. J. Agric. Sci. 108: 267274.
Plaisted, R. I. 1960. A shorter methods for evaluating the Wricke, G. 1962. Uber eine methodezur erfassung der ekolo-
ability of selections to yield consistently over locations. Am. gischen streubreite in feldversuchen. Z.F. Panzenzuecht. 47:
Potato J. 37: 166172. 9296.
Plaisted, R. I. and Peterson, L. C. 1959. A technique for Yates, F. and Cochran, W. G. 1938. The analysis of groups of
evaluating the ability of selection to yield consistently in experiments. J. Agric. Sci. 28: 556580.
different location or season. Am. Potato J. 36: 381385.
This article has been cited by:

1. Ali Moghaddam, Amir Raza, Johann Vollmann, M. Reza Ardakani, Wolfgang Wanek, Gabrielle Gollner, Juergen K. Friedel. 2015.
Biological nitrogen fixation and biomass production stability in alfalfa ( Medicago sativa L.) genotypes under organic management
conditions. Biological Agriculture & Horticulture 31, 177-192. [CrossRef]
Can. J. Plant Sci. Downloaded from www.nrcresearchpress.com by 186.46.90.98 on 06/07/16
For personal use only.

You might also like