ORIGINAL PAPER

EVALUATION OF GRAIN YIELD STABILITY, RELIABILITY AND CULTIVAR

RECOMMENDATIONS IN SPRING WHEAT (TRITICUM AESTIVUM L.) FROM
KAZAKHSTAN AND SIBERIA
Gómez-Becerra HUGO FERNEY1*, Morgounov ALEXEI2, Abugalieva AIGUL1,3
1
Research and Production Center of Farming and Crop Science, Karasaisky rayon, 483133, Almalibak, Kazakhstan.
2
CIMMYT Turkey, P.K. 39 Emek, 06511 Ankara, Turkey
3
Kazakhstan-Siberia Network for Spring Wheat Improvement (KASIB), Astana, Kazakhstan.
*Corresponding author: Tel: + 57 3272 983608, Email: hugoferney2004@yahoo.com

Manuscript received: September 2, 2006; Reviewed: January 29, 2007; Accepted for publication: January 29, 2007

ABSTRACT
The investigation was carried out to determine the stability and adaptability patterns of a set of 40 promising spring
wheat genotypes from Kazakhstan and Siberia evaluated in a multievironment yield trial across 22 environments. Some
of the most widely known parametric stability parameters were used as well as the less frequently cited reliability index
(I). Grain yield correlated significantly and positively with the stability parameters b and S2 and the reliability index
(I); but did not correlate with AMMI ASV. However, the stability parameters failed in detecting adaptability patterns.
In contrast, the reliability index (I) was probed to be more useful in supporting practical decisions. With regard to the
genotypes, cultivars Lutescens 54, Lutescens 30-94, Lutescens 29-94, Tertsia, Omskaya 35, and Shortandynskaya 95
showed to be the widest adapted and the most reliable cultivars.
KEYWORDS: spring bread wheat (Triticum aestivum L.), Kazakhstan, Siberia, parametric stability analysis, reliability
index

Volume 7 (2006) No. 4 (649-660) 649

Gómez-Becerra HUGO FERNEY, Morgounov ALEXEI, Abugalieva AIGUL

1. INTRODUCTION Forty spring wheat genotypes from Kazakhstan and

Among the objectives of multienvironment yield Siberian breeding programs (Table 1) were evaluated
trials are the establishment of adaptation strategies at the 4th and 5th Kazakhstan Siberia Network Spring
for breeding programs and definition of domains for Bread Wheat Breeding Yield Trials (4th and 5th KASIB-
cultivar recommendations. The adaptation strategy SBWBYT) across 11 locations in Kazakhstan and the
objectives focuses on responses of a set of genotypes Siberian region of Russia. The locations in Kazakhstan
to obtain indications and generate predictions relative were: Karabalyk, Karaganda, Pavlodar, Aktube,
to future breeding material that may be produced from Shortandy, Petropavlovsk and Almaty; and in Siberia:
the genetic bases of which the tested genotypes are Omsk, Kurgan, Barnaul (Altay region of Russia) and
assumed to be a representative sample, while, for cultivar Chelyabinsk. The nursery was evaluated in 2003 and
recommendations the most important information 2004 as a yield trial in a randomized complete block
concerns the response of, and comparison between, high- design with two replications.
yielding genotypes [2]. 2.2 Yield stability and reliability methods
High yield stability usually refers to a genotype’s ability
2.2.1 The Finlay and Wilkinson’ b regression
to perform consistently, whether at high or low yield
coefficient
levels, across a wide range of environments [2]. Several
biometrical methods including univariate and multivariate According to Finlay and Wilkinson [4], the modeled
ones have been developed to assess stability [1]. Between genotype response is represented by:
them the most widely used are the regression coefficient Rij = ai + bimj (1)
[4], the environmental variance [12], the Shukla’s [16] where Rij = modeled genotype yield response in
stability variance and Wricke’s ecovalence [17]. More environment j, ai = intercept value, and mj = the
recently, Purchase [14] developed the AMMI Stability environmental mean yield. They showed that a
Value (ASV) based on the AMMI (Additive Main Effects regression coefficient approximating 1.0 indicated an
and Multiplicative Interaction) model’s PCA1 and PCA2 average stability, and in association with high yield,
(Principal Components Axis 1 and 2 respectively) scores the entry possesses general adaptability. However,
for each cultivar. This AVS is in effect the distance from entries with a low yield would be poorly adapted to the
the coordinate point to the origin in a two dimensional environment. Regression coefficient values increasing
scattergram of PCA1 scores against PCA2 scores [11]. above 1.0 describe genotypes with increasing sensitivity
The practical interest of combining high levels of mean to environmental change, thus below average stability.
yield and yield stability has led to the development of Regression coefficients decreasing below 1.0 provide a
the yield reliability concept [3] [6], where a reliable measure of greater resistance to environmental change,
genotype is characterized by consistently high yield thus above average stability [15], where the greatest
across environments [2]. The use of a yield reliability stability is b = 0.
index facilitates genotype selection or recommendation, 2.2.2 The environmental variance (S2)
as the mean yield and the yield stability are combined
The environmental variance (S2) is one of the major
into a unique measure of genotype merit [2].
stability measures for the static stability concept (type
A few number of studies of genotype by environment 1 stability) [12], i.e. the variance of genotype yields
interactions (GxE) and stability have been reported on recorded across test or selection environments (i.e.
wheat in Kazakhstan [8] [9] [10]. However, no stability and individual trials). For the genotype i:
reliability studies have been performed for spring bread
wheat developed by Kazakhstan and Siberian breeding
Si2 = ∑ (Rij - mi)2/(e - 1) (2)
programs and tested together in a multienvironment yield
trial. The objectives of the present study were to evaluate
the grain yield of promising spring wheat genotypes where Rij = observed genotype yield response in
in different environments in Kazakhstan-Siberia and the environment j, mi = genotype mean yield across
to determine their stability and reliability for cultivar environments, and e = number of environments. Greatest
recommendations. stability is S2 = 0.

2.2.3 AMMI model and the AMMI Stability Value

2. MATERIALS AND METHODS
(ASV)
2.1 Plant material and locations The AMMI model postulates additive components for

650 Journal of Central European Agriculture Vol 7 (2006) No 4

 EVALUATION OF GRAIN YIELD STABILITY, RELIABILITY AND CULTIVAR RECOMMENDATIONS IN SPRING WHEAT
(TRITICUM AESTIVUM L.) FROM KAZAKHSTAN AND SIBERIA

Table 1. Mean yield and origin of wheat genotypes.

Code Name of genotypes Yield (t/ha) Origin
1 Omskaya 34 2.52 Siberian Research Institute
2 Omskaya 35 2.86 Siberian Research Institute
3 Lutescens 148-97-16 2.38 Siberian Research Institute
4 Chernyava 13 2.60 Omsk State Agrarian University
5 Sonata 2.85 Omsk State Agrarian University
6 Niva 2 2.61 Omsk State Agrarian University
7 Golubkovskaya 2.71 Omsk State Agrarian University
8 Iren 2.34 Krasnoufimsk Breeding Station
9 Irgina 2.01 Krasnoufimsk Breeding Station
10 Krasnoufimskaya 90 2.26 Krasnoufimsk Breeding Station
11 Sibirskaya 12 2.64 Siberian Research Institute
12 Sibirskaya 123 2.57 Siberian Research Institute
13 Novosibirskaya 15 2.25 Siberian Research Institute
14 Novosibirskaya 29 2.44 Siberian Research Institute
15 Udacha 2.79 Siberian Research Institute
17 Lutescens 574 2.81 Altay Research Institute
18 Lutescens 424 2.65 Altay Research Institute
19 Altaiskaya 50 2.11 Altay Research Institute
20 Aria 2.85 Kurgan Agric Res Institute
21 Tertsia 2.97 Kurgan Agric Res Institute
22 Fora 2.04 Kurgan Agric Res Institute
23 Chelyaba 2.58 Chelyabinsk Agric Res Institute
24 Chebarkulskaya 2.68 Chelyabinsk Agric Res Institute
25 Astana 2.70 Scientific Prod Center of Grain Farming
26 Bayterek 2.73 Scientific Prod Center of Grain Farming
27 Shortandynskaya 95 2.85 Scientific Prod Center of Grain Farming
28 Lutescens 13 2.92 Karabalyk Experimental Res Station
29 Lutescens 54 3.16 Karabalyk Experimental Res Station
31 Nadezhda 2.18 Kazakh Res. Center of Farming
32 No. 18 2.15 Kazakh Res. Center of Farming
33 Eritrospermum 727 2.55 Kazakh Res.Center of Farming
34 Lutescens 29-94 3.01 Pavlodar Agric Research Institute
35 Lutescens 30-94 3.11 Pavlodar Agric Research Institute
36 Lutescens 53-95 2.78 Pavlodar Agric Research Institute
37 Stepnaya 1 2.82 Aktube Experimental Res Station
38 Aktubinka 2.62 Aktube Experimental Res Station
39 Aktube 32 2.76 Aktube Experimental Res Station
40 GVK 1860/8 2.58 East Kazakhstan Agric Res Institute
41 GVK 1369/2 2.85 East Kazakhstan Agric Res Institute
42 GVK 1857/9 2.77 East Kazakhstan Agric Res Institute
max 3.16
min 2.01
mean 2.62

J. Cent. Eur. Agric. (2006) 7:4, 649-660 651

Gómez-Becerra HUGO FERNEY, Morgounov ALEXEI, Abugalieva AIGUL

the main effects of genotypes (αi) and environments in unfavorable cropping regions) to 0.70 for modern
(βj) and multiplicative components for the effect of the agriculture in most favorable regions [2].
interaction (фij). Thus, the mean response of genotype i 2.3 Computer program
in an environment j is modeled by:
Calculations were performed by IRRISTAT 4.3 software
[5] using the cross site analysis procedure, which gives
Ŷ= μ + αi + βj + λkγikδjk + ρij + εij (3)
outputs of AMMI and joint regression models including
in which фij is represented by:
analysis of variance, regression coefficients, as well as
λkγikδjk genotypes and environments means and Biplots graphics.
For the rest of calculations as stability and reliability
index an ordinary spread sheet program was used.
where λk is the size, γik is the normalized genotype
vector of the genotype scores or sensitivities, δjk is the
normalized environmental vector of the scores describing 3. RESULTS AND DISCUSSION
environments, ρij are the AMMI residuals , and εij is the
error term.
3.1 Mean yield
As mentioned above, the AVS is the distance from the
The mean yield of the 40 genotypes across 22
coordinate point to the origin in a two dimensional
environments (location x year) ranged from 2.01 t/
scattergram of PCA1 scores against PCA2 scores.
ha to 3.16 t/ha (Table 1). The difference in the rank of
Because the PCA1 score contributes more to the GxE
the genotypes in the various environments indicated
sum of squares, a weighted value is needed. This weight
the presence of GxE interactions (Table 2), that was
is calculated according to the relative contribution of
confirmed by the significant effect of the genotype x
PCA1 to PCA2 to the interaction SS [11].
environment interaction (explaining 15.76% of the G + E
+ GE) in the AMMI analysis of variance (Table 3). From
ASV = {[(SSPCA1 / SSPCA2) (GPCA1score)]2 + (GPCA2s Table 2 and Figure 1 it is possible to see that genotype
core)}1/2 (4) 29 (Lutescens 54) was present in the top 5 rank in 13
out of 22 environments (being identified as dominant
cultivar in 6 environments); followed by genotype 34
where SSPCA1 / SSPCA2 is the weight given to the PCA1
(Lutescens 29-94) that appeared in the top 5 rank in 8
value by dividing the PCA1 sum of squares by the PCA2
of 22 environments, being the dominant cultivar in 3
sum of squares, GPCA1score is the PCA1 score for that
environments; genotype 35 (Lutescens 30-94) was the
specific genotype, and GPCA2score is the PCA2 score
best in 2 environments and appeared in the top 5 rank in
for that specific genotype.
7 of 22 environments; genotype 36 (Lutescens 53-95),
was the dominant cultivar in 3 environments and was
2.2.4 The reliability index (I) inside the top 5 rank in 6 environments; genotype 21
The reliability index as proposed by Kataoka [7] for (Tertsia), was the dominant cultivar in 2 environments
economic analysis can be used for estimating, on the and in 6 environments appeared within the top 5 rank;
basis of the distribution of yield values observed across genotype 15 (Udacha), was winner in 3 environments
test environments, the lowest yield expected for a given and was within the top 5 rank in 5 environments; and
genotype and a specified probability of negative events genotype 39 (Aktube 32), that was the dominant cultivar
[3]. It can be calculated by the following expression: in 3 environments. Other genotypes that, although were
not dominant cultivars in more than one environment,
Ii = mi - Z(P) Si (5) but appeared consistently in the top 5 rank across 22
environments were: genotype 27 (Shortandynskaya 95),
genotype 2 (Omskaya 35), and genotype 5 (Sonata), (7, 7
where mi = mean yield, Si = square root of the
and 5 times inside the top 5 rank respectively).
environmental variance (S2), and Z(P) = percentile from
Table 4 serves to illustrate the importance of recommending
the standard normal distribution for which the cumulative
the right genotype for each environment. In our case, an
distribution function reaches the value P. Z(P) can assume
improvement of 0.893 t/ha could be achieved across the
the following values depending on the chosen P level:
22 environments if only the dominant cultivar for each
0.675 for P = 0.75; 0.840 for P = 0.80; 1.040 for P =
environment had been sown.
0.85; 1.280 for P = 0.90; and 1.645 for P = 0.95. P values
may vary between 0.95 (for subsistence agriculture

652 Journal of Central European Agriculture Vol 7 (2006) No 4

 EVALUATION OF GRAIN YIELD STABILITY, RELIABILITY AND CULTIVAR RECOMMENDATIONS IN SPRING WHEAT
(TRITICUM AESTIVUM L.) FROM KAZAKHSTAN AND SIBERIA

Table 2. Environments grouped by their winning genotypes, including the first 5 recommended
cultivars for each environment, based on the AMMI2 estimates.
Environments Dominant cultivar AMMI2 cultivar recommendations
yield yield
Codea Code 1st 2nd 3rd 4th 5th
(t/ha) (t/ha)
B 4.63 36 5.58 36 29 34 41 37
H 2.74 36 3.7 36 28 42 35 34
R 2.14 36 2.9 36 42 21 2 29
A 4.46 35 5.73 35 23 29 34 27
L 4.40 35 5.8 35 20 5 21 2
O 3.02 29 4.57 29 35 25 2 34
Q 2.67 29 4.19 29 17 2 24 25
G 2.54 29 3.42 26 29 34 11 36
U 1.48 29 2.29 29 18 21 27 28
T 0.75 29 1.03 28 7 29 27 6
V 1.99 29 2.47 11 5 27 29 2
S 2.99 15 4.5 15 5 21 8 9
J 1.85 15 2.4 27 26 17 15 25
N 2.29 15 3.12 15 35 2 42 36
M 2.43 21 3.2 21 27 2 29 15
P 2.07 21 3.15 21 5 24 15 29
F 1.87 39 2.27 39 29 37 38 7
K 1.86 39 2.29 39 25 18 35 27
E 1.85 39 2.32 24 39 18 37 17
C 3.86 34 4.59 37 38 20 34 41
D 3.46 34 5.22 41 34 35 40 29
I 2.44 34 2.69 8 31 36 5 34

a
The full name of the environments are: A = Barnaul 2004, B = Chelyabinsk 2004, C = Kurgan 2004, D = Omsk 2004, E
= Aktube 2004, F = Karabalyk 2004, G = Shortandy 2004, H = Almaty 2004, I = Pavlodar 2004, J = Petropavlovsk 2004,
K = Karaganda 2004, L = Barnaul 2003, M = Chelyabinsk 2003, N = Kurgan 2003, O = Omsk 2003, P = Aktube 2003, Q
= Karabalyk 2003, R = Shortandy 2003, S = Almaty 2003, T = Pavlodar 2003, U = Petropavlovsk 2003, V = Karaganda
2003.

Table 3. AMMI analysis of variance for the yield (t/ha) of 40 spring wheat genotypes in 22 environments.
Explained
Source Df SS MS
(%)
Environment 21 854.539 40.6923 77.76
Genotype 39 71.0508 1.82182 6.46
Genotype X Environment 819 173.217 0.211498 15.76
AMMI component 1 59 43.7750 0.741948 25.27
AMMI component 2 57 28.3835 0.497956 16.38
Total 879 1098.81

J. Cent. Eur. Agric. (2006) 7:4, 649-660 653

Gómez-Becerra HUGO FERNEY, Morgounov ALEXEI, Abugalieva AIGUL

Table 4. Yield improvement on the trial if only the first AMMI2 recommendation
was planted at each environment.
Environment Dominant Cultivar
Environmenta improvement
yield (t/ha) cultivar yield (t/ha)
B 4.63 36 5.58 0.953
H 2.74 36 3.7 0.960
R 2.14 36 2.9 0.759
A 4.46 35 5.73 1.272
L 4.40 35 5.8 1.397
O 3.02 29 4.57 1.551
Q 2.67 29 4.19 1.525
G 2.54 29 3.42 0.884
U 1.48 29 2.29 0.807
T 0.75 29 1.03 0.279
V 1.99 29 2.47 0.478
S 2.99 15 4.5 1.512
J 1.85 15 2.4 0.549
N 2.29 15 3.12 0.834
M 2.43 21 3.2 0.773
P 2.07 21 3.15 1.079
F 1.87 39 2.27 0.405
K 1.86 39 2.29 0.427
E 1.85 39 2.32 0.465
C 3.86 34 4.59 0.726
D 3.46 34 5.22 1.764
I 2.44 34 2.69 0.251
Average 2.63 3.52 0.893
a
The full name of the environments are: A = Barnaul 2004, B = Chelyabinsk 2004, C = Kurgan
2004, D = Omsk 2004, E = Aktube 2004, F = Karabalyk 2004, G = Shortandy 2004, H =
Almaty 2004, I = Pavlodar 2004, J = Petropavlovsk 2004, K = Karaganda 2004, L = Barnaul
2003, M = Chelyabinsk 2003, N = Kurgan 2003, O = Omsk 2003, P = Aktube 2003, Q =
Karabalyk 2003, R = Shortandy 2003, S = Almaty 2003, T = Pavlodar 2003, U = Petropavlovsk
2003, V = Karaganda 2003.

3.2 The Finlay and Wilkinson’ (b) regression Genotypes 29 (Lutescens 54), 35 (Lutescens 30-94) and
coefficient 34 (Lutescens 29-94) ranked as the three best yielding
Table 5 showed genotypes 9 (Irgina), 31 (Nadezhda), 32 cultivars (the only three with mean yield over 3 t/ha), but
(No. 18), 19 (Altaiskaya 50), and 13 (Novosibirskaya genotype 29 (Lutescens 54) adapted better by having the
15) as the cultivars with the greatest stability because of best mean yield (rank 1) and lower regression coefficient
their b value significantly lower than 1.0. However, as all (rank 33 versus 39 and 40 respectively).
of them were within the cultivars with the lowest yield
mean, the conclusion is that they were poorly adapted to 3.3 The environmental variance (S2)
the test environments. The environmental variance (S2) gave almost the same
Genotypes 17 (Lutescens 574), 4 (Chernyava 13), 27 rank of the cultivars as the regression coefficient. This
(Shortandynskaya 95), and 2 (Omskaya 35) possesses near perfect correlation is confirmed by the Spearman
average stability due to their regression coefficient near correlation coefficient r = 0.97*** (Table 6). Genotypes
to 1.0, and can be consider as well adapted cultivars 31(Nadezhda), 9 (Irgina), and 32 (No. 18) with the lowest
across the environments because their good mean yield, (S2) values (the most stable cultivars), were also within the
except for cultivar 4 (Chernyava 13). lowest yielding, and thus performed as the least adapted

654 Journal of Central European Agriculture Vol 7 (2006) No 4

 EVALUATION OF GRAIN YIELD STABILITY, RELIABILITY AND CULTIVAR RECOMMENDATIONS IN SPRING WHEAT
(TRITICUM AESTIVUM L.) FROM KAZAKHSTAN AND SIBERIA

Table 5. Rank and estimation of stability and reliability measures, and grain yield for
40genotypes across environments.
Regression
Rank ASV stability S2 stability coefficient Reliability index Mean yield
Genotype ASV Genotype S2 Genotype b genotype I Genotype Yield
1 17 0.112 31 0.617 9 0.669* 29 2.115 29 3.16
2 21 0.189 9 0.669 31 0.685* 28 1.988 35 3.11
3 38 0.205 32 0.715 32 0.716* 35 1.967 34 3.01
4 20 0.232 13 0.756 19 0.740* 17 1.939 21 2.97
5 7 0.263 14 0.783 13 0.755* 21 1.935 28 2.92
6 3 0.289 22 0.818 8 0.785 27 1.931 2 2.86
7 28 0.294 19 0.848 24 0.805 25 1.888 41 2.85
8 1 0.308 10 0.919 14 0.808* 15 1.884 27 2.85
9 27 0.328 25 0.940 22 0.827* 24 1.862 20 2.85
10 14 0.380 8 0.949 10 0.862 5 1.854 5 2.85
11 33 0.388 24 0.950 25 0.893 34 1.847 37 2.82
12 25 0.399 1 0.988 11 0.948 26 1.843 17 2.81
13 39 0.405 11 1.033 18 0.948 2 1.842 15 2.79
14 26 0.409 18 1.049 15 0.954 20 1.809 36 2.78
15 11 0.450 6 1.065 6 0.956 11 1.788 42 2.77
16 15 0.464 17 1.086 1 0.970 18 1.785 39 2.76
17 6 0.490 26 1.119 26 0.976 39 1.784 26 2.73
18 18 0.494 15 1.175 23 0.983 6 1.739 7 2.71
19 23 0.533 23 1.198 17 0.996 37 1.730 25 2.70
20 12 0.585 33 1.199 4 1.006 36 1.729 24 2.68
21 22 0.597 27 1.201 27 1.007 41 1.725 18 2.65
22 4 0.622 28 1.223 2 1.010 14 1.692 11 2.64
23 37 0.724 4 1.240 33 1.028 42 1.686 38 2.62
24 2 0.738 3 1.274 3 1.037 1 1.682 6 2.61
25 13 0.767 38 1.318 28 1.050 4 1.666 4 2.60
26 32 0.791 12 1.328 5 1.051 23 1.664 23 2.58
27 40 0.841 39 1.346 38 1.082 7 1.657 40 2.58
28 36 0.842 5 1.404 12 1.104 38 1.654 12 2.57
29 29 0.853 2 1.466 39 1.120 33 1.633 33 2.55
30 31 0.891 21 1.529 20 1.129 12 1.602 1 2.52
31 5 0.910 20 1.535 36 1.137 31 1.525 14 2.44
32 10 0.910 29 1.545 21 1.142 8 1.519 3 2.38
33 35 0.978 7 1.561 29 1.145 13 1.517 8 2.34
34 24 0.997 36 1.568 7 1.164 40 1.505 10 2.26
35 8 1.040 40 1.623 42 1.174 10 1.453 13 2.25
36 19 1.113 42 1.659 40 1.181 32 1.440 31 2.18
37 42 1.188 37 1.694 41 1.248* 3 1.433 32 2.15
38 41 1.215 41 1.804 37 1.257* 19 1.332 19 2.11
39 34 1.262 35 1.858 35 1.291* 9 1.324 22 2.04
40 9 1.346 34 1.913 34 1.313* 22 1.275 9 2.01
*indicates coefficients significantly different from the overall regression coefficient which is 1.

J. Cent. Eur. Agric. (2006) 7:4, 649-660 655

Gómez-Becerra HUGO FERNEY, Morgounov ALEXEI, Abugalieva AIGUL

cultivars across the test environments. Between the top 3.4 AMMI model and the AMMI Stability Value
yielding genotypes, cultivar 29 (Lutescens 54) performed (ASV)
as the widest adapted by having a lower (S2) value, The PCA scores of a genotype in the AMMI analysis
and being the top one yielding genotype. Conversely, are an indicator of the stability of a genotype over
genotypes 35 (Lutescens 30-94) and 34 (Lutescens 29- environments. The greater the PCA scores, either negative
94) were the two least stable cultivars and although or positive, the more specifically adapted a genotype
wide adapted, they showed some specific adaptation to is to certain environments. The more the PCA scores
locations Barnaul and Omsk respectively (Table 2). approximate zero (0), the more stable the genotype is

Table 6. Spearman correlation between yield, S2, b, ASV and I ranks.

S2 b ASV I
Yield 0.75*** 0.73*** -0.12 0.87***
2
S 0.97*** 0.01 0.40**
b -0.08 0.38*
ASV -0.26
* Indicates significance at P = 0.05
** Indicates significance at P = 0.01
*** Indicates significance at P = 0.001

D
A= Barnaul 2004,
B= Chelyabinsk 2004,
C= Kurgan 2004,
D= Omsk 2004,
E= Aktube 2004,
F= Karabalyk 2004,
G= Shortandy 2004,
H= Almaty 2004,
I= Pavlodar 2004,
J= Petropavlovsk 2004, A
K= Karaganda 2004,
L=Barnaul 2003, B
M= Chelyabinsk 2003,
N= Kurgan 2003, 35
O= Omsk 2003,
P= Aktube 2003,
Q= Karabalyk 2003,
R= Shortandy 2003,
S= Almaty 2003, G
T= Pavlodar 2003,
U= Petropavlovsk 2003,
V= Karaganda 2003
29 36
R H L
C

Q
U O
F N
K V
39 E
T I 15
P 21 M

Figure 1. Grouping of environments according to the superior cultivar as determined by theAMMI2 estimates at
each environment. Note that according to the information provided in Table 2, environments I, C, and D and can join
the genotype’s 34 group; environment J, the genotype’s 15 group; and environment V, the genotype’s 29 group.

656 Journal of Central European Agriculture Vol 7 (2006) No 4

 EVALUATION OF GRAIN YIELD STABILITY, RELIABILITY AND CULTIVAR RECOMMENDATIONS IN SPRING WHEAT
(TRITICUM AESTIVUM L.) FROM KAZAKHSTAN AND SIBERIA

Figure 2. AMMI2 genotype x environment biplot for 40 genotypes and 22 environments. Genotypes and
environments names correspond to those in Tables 1 and 2.

over all environments sampled [15]. and 21 (Tertsia) appeared as the most stable genotypes
From Figure 2, it is possible to visualize some interesting as they were located very close to the origin point (zero
patterns. For genotypes, the GVK cultivars (genotypes PCA scores), and genotype 19 as the most instable (and
40, 41, and 42) were very in close relation to each also the least adapted due to its low yield mean across
other indicating similar germplasm. The same situation the environments). However, when analyzed together the
happens to genotypes 11 and 12 (Sibirskaya 12 and PCA scores with the mean yield, genotype 21 (Tertsia)
Sibirskaya 123) and some of the Lutescens cultivars emerges not only as very stable cultivar, but also as a
(genotypes 29, 34, 35 and 36). This group of Lutescens cultivar with wide adaptation pattern. The AMMI ASV
cultivars (Lutescens 54, Lutescens 29-94, Lutescens 30- stability value confirms what it can be seen graphically
94, and Lutescens 53-95) fell into the upper-right side in the biplot. However, the ASV ranked genotype 34
quadrant of the AMMI2 biplot, indicating their good (Lutescens 29-94) as the second most instable cultivar,
yield potential, especially in high yielding environments and genotype 9 (Irgina) as the most instable one (and one
as A, B and L (Barnaul 2004, Chelyabinsk 2004, and of the least adapted to the testing environments because
Barnaul 2003 respectively). Cultivars 17 (Lutescens 574) of its low mean yield). The fact that one high yielding

J. Cent. Eur. Agric. (2006) 7:4, 649-660 657

Gómez-Becerra HUGO FERNEY, Morgounov ALEXEI, Abugalieva AIGUL

cultivar (genotype 34), and one low yielding cultivar In regard to the genotypes, cultivars 29, 35, 34, 21, 2,
(genotype 9) were the two most instable cultivars serves and 27 (Lutescens 54, Lutescens 30-94, Lutescens 29-
to demonstrate the importance of analyzing both stability 94, Tertsia, Omskaya 35, and Shortandynskaya 95
and yield performance to determine also the adaptability respectively) showed to be the widest adapted and the
patterns for cultivar recommendations. most reliable cultivars. Many of the Lutescens cultivars
(genotypes 29, 35, 34, 36) and cultivar Omskaya 35
3.5 The reliability index (I) (genotype 2) dominated the reliability and mean yield
Making assumption that the technological level of ranks. It can be explained by the fact that they possess
agriculture in Kazakhstan and Siberia falls between in their pedigrees blood from the landrace Poltavka, a
the subsistence agriculture and modern agriculture, we very broad adapted genotype due to its ability to tolerate
took (P) = 0.8, which corresponds to a Z(p) = 0.84 to be abiotic stresses and present in more than 76% of the
inserted in equation 5 (see section 2.2.4). The reliability modern spring wheat varieties released in Kazakhstan
index (I) ranked the genotypes according to the lowest [13].
average yield one can expect for each cultivar across In summary, the following cultivars were identified
the test environments. Table 5 shows that the top 5 as promising for practical recommendations at the
reliable genotypes were: genotype 29 (Lutescens 54), 28 different spring wheat producing regions in Kazakhstan
(Lutescens 13), 35 (Lutescens 30-94), 17 (Lutescens 574) and Siberia, according to their performance across 22
and 21 (Tertsia), which agrees in grand extent to the top 5 environments:
yielding genotypes (Spearman correlation coefficient r = • Cultivar 29 (Lutescens 54): Chelyabinsk,
0.87***). It is not surprising that genotype 17 (Lutescens Shortandy, Barnaul, Omsk, Karabalyk, Petropavlosk,
574) appeared as one of the most reliable genotypes Pavlodar and Karaganda.
because from the regression coefficient analysis it was • Cultivar 35 (Lutescens 30-94): Barnaul,
identified as one with average adaptation (b coefficient Karaganda, Omsk and Kurgan.
close to 1.0), and from the biplot graphic and AMMI • Cultivar 27(Shortandynskaya 95): Petropavlosk,
ASV value, as the most stable cultivar; all this plus its Karaganda and Almaty.
good yield potential (although not being the best) placed • Cultivar 34 (Lutescens 29-94): Pavlodar, Omsk
it as 4th in the reliability rank. and Chelyabinsk
• Cultivar 36 (Lutescens 53-95): Pavlodar and
3.6 Comparisons and concluding remarks Shortandy.
No one of the three stability parameters (b, S2, and • Cultivar 21 (Tertsia): Almaty and Aktube.
AMMI ASV) was sufficiently accurate in its own to be • Cultivar 24(Chebarkulskaya): Aktube.
considered of practical interest for genotype selection and • Cultivar 15 (Udacha): Almaty.
recommendation as it was manifested by the Spearman
correlation coefficient between them and the mean yield
rank (Table 6). For instance, b and S2 values, although ACKNOWLEDGEMENTS
with a middle-high correlation coefficients (0.73*** The authors are grateful with all institutions, scientist
and 0.74*** respectively) with respect to yield, failed and technicians involved in the Kazakhstan-Siberian
in detecting more yielding and wider adapted genotypes Network on Spring Wheat Improvement (KASIB) that in
as Lutescens 29-94 (genotype 34), and Lutescens 30- excellent manner conducted the trials and collected the
94 (genotype 35). Similarly, the AMMI ASV value did data used in this paper.
not correlate with the yield of the genotypes (r = -0.12).
Only when they were combined with the yield data and
assisted by a table of mean yield of environments and REFERENCES
genotypes and cultivars ranks for each environment (as [1] Akçura M., Kaya Y., Taner S., Genotype-
Table 2 in the present study), the use of these stability environment interaction and phenotypic stability analysis
parameters take a more practical application for selection for grain yield of durum wheat in the Central Anatolian
and recommendation. Conversely, the reliability index region, Turk. J. Agric. For. (2005) 29: 369-375.
(I), due to its nature of combining a derived stability [2] Annicchiarico P., Genotype X environment
measure (the Si value) and yield data, gave more useful interactions. Challenges and opportunities for plant
information for selection and recommendation. It was breeding and cultivar recommendations, FAO, Rome,
confirmed by its correlation with yield (r = 0.87***). 2002.

658 Journal of Central European Agriculture Vol 7 (2006) No 4

 EVALUATION OF GRAIN YIELD STABILITY, RELIABILITY AND CULTIVAR RECOMMENDATIONS IN SPRING WHEAT
(TRITICUM AESTIVUM L.) FROM KAZAKHSTAN AND SIBERIA

[3] Eskridge K.M., Selection of stable cultivars using [12] Lin C.S., Binns M.R., Lefkovitch L.P., Stability
a safety-first rule, Crop Sci. (1990) 30: 369-374. analysis: where do we stand?, Crop Sci. (1986) 26: 894-
[4] Finlay K.W., Wilkinson G.N., The analysis of 900.
adaptation in a plant-breeding programme, Aust. J. Agric. [13] Martynov S.P., Dobrotvorskaya T.V.,
Res. (1963) 14: 742-754. Morgounov A.I., Urazaliev R.A., Absattarova A.S.,
[5] International Rice Research Institute, IRRISTAT Genealogical analysis of diversity of spring bread wheat
4.0 for windows. Tutorial manual, IRRI, Manila, 2003. cultivars released in Kazakhstan from 1929-2004, Acta
Agronomica Hungarica (2005) 53(3): 261-272.
[6] Kang M.S., Pham H.N., Simultaneous selection
for high yielding and stable crop genotypes, Agron. J. [14] Purchase J.L., Parametric analysis to describe
(1991) 83: 161-165. genotype x environment interaction and yield stability in
winter wheat, (Ph.D. Thesis), University of Free State,
[7] Kataoka S., A stochastic programming model,
Bloemfontein, 1997.
Econometrika (1963) 31: 181-196.
[15] Schoeman L.J., Genotype x environment
[8] Kokhmetova A., Analysis of genotype-
interaction in sunflower (Helianthus annuus) in South
environment interaction in wheat breeding for drought
Africa, (M.Sc. Thesis), University of Free State,
resistance, Annual Wheat Newsletter (2000) 46: 69-69.
Bloemfontein, 2003.
[9] Kokhmetova A., Genotype-environment
[16] Shukla G.K., Some statistical aspects of
interaction and yield stability in wheat varieties, Annual
partitioning genotype-environmental components of
Wheat Newsletter (2001) 47: 91-93.
variability, Heredity (1972) 29: 237-245.
[10] Kokhmetova A., Genotypic structure of winter
[17] Wricke G. Über eine Methode zur Erfassung
wheat variety Progress, Agromeridian (2005) 1: 59-66.
der ökologischen Streubreite in Feldversuchen, Z.
[11] Leeuvner D.V., Genotype x environment Pflanzenzüchtg (1962) 47: 92-96.
interaction for sunflower hybrids in South Africa, (M.Sc.
Thesis), University of Pretoria, Pretoria, 2005.

J. Cent. Eur. Agric. (2006) 7:4, 649-660 659