|
|
Author´s address: Ing.Agr.Julia Lúquez, M.Sc. Unidad Integrada Balcarce -
Ruta 226, km 73,5, cc 276, (7620) Balcarce, Argentina. -
e-mail:jluquez@balcarce.inta.gov.ar
2002
Abstract
Best-yielding
and most stables cultivars are identified by cultivating them in different environments.
Stability of quality grain traits has been less investigated than grain yield stability.
High oleic hybrids of sunflower have been available in the Argentinean seed market during
the last years. Research about stability of these genotypes is scarce. The objectives of
this work were i) to compare by using three different methods the stability and
adaptability of high oleic hybrids for grain yield and oil and oleic acid contents, ii) to
analyze the advantages and disadvantages of each method to select stables or adapted
genotypes with high grain yield and high quality. Stability and adaptability analysis were
made on results of grain yields and oil and oleic acid contents of 35 sunflower high oleic
hybrids from 17 Comparative Yield Trials conducted during two years in Argentina.
Stability was estimated by using two methods: the Fisher´s protected LSD test, which
compared hybrids with the best-yielding hybrid in each environment, and the test of
Relative Yield (RY), which uses standard deviation as the measure of stability.
Adaptability was estimated by using Piepho´s method of Multiple Comparisons with the
Best. The three methods can be applied for unbalanced data. Piepho´s method made little
discrimination of the hybrids. The LSD and RY coincided in classifying the hybrids as
stable and unstable in 85% of the cases for grain yield and 76% for oil content. It is
concluded that the more convenient method depends on characteristics of the experimental
design and of variability of the evaluated trait. Results from the LSD test depend on the
number of environments where the cultivar is tested. The RY method is valuable to classify
some cultivars as high yielding and stable, avoiding the bias that high-yielding
environments give to the general average. Using both methods together could be useful to
classify to hybrids when the number of environments is adequate.
Key words: stability, adaptability, grain yield, oil
content, oleic acid content, high oleic sunflower hybrids
Introduction
Using stable
cultivars for high grain yield and quality is important in sustainable agriculture.
Stability of grain yield for different crops has been evaluated through statistical models
that analyze genotype by environment interaction (GEI) in cultivar trials (Crossa, 1990;
Piepho, 1998). Stability of quality grain and oil traits has been less studied than grain
yield stability. Nevertheless, quality traits could highly influence the product price in
the market or it acceptation by consumers.
Stability has
been investigated by different procedures. Those included regression analysis (Yates &
Cochran, 1938; Finlay & Wilkinson, 1963; Eberhart & Russell, 1966), univariate and
multivariate analyses of variance (Mandel, 1971; Lin & Thompson, 1975; Ghaderi et
al., 1980; Brennan at al., 1981; Fox & Rossielle, 1982; Crossa et al.,
1993; Annicchiarico, 1997) and a multivariate analysis of the residuals from a main
effects additive model, using the AMMI approach (Additive Main effect and Multiplicable
Interaction) (Gauch, 1988; Gauch & Furnas, 1991). This last method allow the modeling
of GEI in more of one dimension (Vargas et al., 1998). Most of these methods
requires that all the genotypes should be present in all the environments. This condition
is not easy to fulfill in practice (i.e. due to the continuous replacement of the
cultivars in the seed market, losses of entries for climatic causes and pests attack).
Therefore, certain environments or cultivars are rendered useless, leaving only part of
the information obtained for analyses. In this context, it is possible to loss valuable
genotypes.
Changes in
cultivar ordering indicates GEI and lack of stability regarding the trait under study. GEI
reduce the correlation between the genotype and the phenotype hindering the evaluation of
the genetic potential of the cultivars (Kang & Gorman, 1989). Quite often, the
magnitude of the GEI is the reason to select genotypes adapted to different individual
locations through independent selection, in an attempt to reach the maximum yield
potential in a particular environment; all this increases costs substantially. The concept
of specific adaptability of genotypes explain that a genotype performed well in an
environment and not in other even when the differences between locations are consistent
from year to year (Fehr, 1987). In these cases, methods are used for evaluate cultivars
adaptability in each environment, as is the method proposed by Piepho (1995) assigning
cultivars to specific locations.
The
argentinian sunflower zone ( 33º and 38º Lat. S. Y 57º y 65º long. W, in 2.316.000
has.) extends on a wide range of environments. Studies about stability and adaptability in
sunflower in Argentina (Ludueña y Marta, 1979; Lorenzo y Lorenzo. 1987, Castaño et al.,
1987; de la Vega et al., 2000) were made with traditional genotypes in balanced designs
and no quality traits were studied.
High
oleic (HO) hybrids of sunflower have been available in the Argentinean seed market during
the last years. Research about stability of these genotypes is scarce. Studies involve in
most cases a few genotypes and environments (i.e. Uhart et al., 2000). The variability for
grain yield and oil and oleic acid content was recently assessed for numerous genotypes
cultivated under contrasting environments in the Argentine High Oleic Sunflower Trials
(Agüero et al., 1999). In the present paper, the same set of data used by Agüero
et al., (1999) was analyzed by three methods of stability and/or adaptability. The
methods are: i) the Fisher LSD protected test (Steel & Torrie, 1993), ii) the method
of relative yield (Yau & Hamblin, 1994), and iii) the Piepho´s method of Multiple
Comparison with the Best (1995). Methods i) and ii) are used to estimate stability and
method iii) to estimate adaptability. All these methods are easy to apply and can be used
when not all genotypes are present in all environments.
The
objectives of this work were i) to compare by using three different methods the stability
and adaptability of high oleic hybrids for grain yield and oil and oleic acid contents ii)
to analyze the advantages and disadvantages of each method to select stables or adapted
sunflower genotypes with high grain yield and high quality.
Materials
and Methods
Experiments
The
data set of grain yields (kg/ha) and oil and oleic acid contents (%) was obtained from 17
official Comparative Yield Trials conducted in two years. In 1995/96, 17 high oleic
experimental hybrids (HOEH) and 5 checks were evaluated in 7 locations. Checks were 4 high
oleic commercial hybrids (HOCH) (Aromo, Trisum 870, Sideral and P-6661) and 1 tradicional
hybrid (TH) (Contiflor 9). Locations were situated between latitude 34° - 38° south and
longitude 57° - 63° west. In 1996/97, 23 HOEH and 7 checks Aromo, Trisum 870, Sideral,
P-6661, Contiflor 9, ACA 884 and Dekasol 3881 (TH) were evaluated in 9 locations situated
between 33° - 38° south and longitude 57° - 65° west. All Trials were conducted using
a Randomized Complete Block design with 3 replications. The plots consisted of three 6-m
long furrows, 0.70 m apart. Sowing density was 71500 pl/ha.
Grain
yield was determined in the central furrow of each plot and expressed in kg/ha. Oil
content was determined by RMN (Robertson and Morrison, 1979) in grains harvested from
plants of free pollination from central furrow of each plot. Oleic acid content was
determined by CGL in all assays in 1995/96 and in 5 assays in 1996/97. Grains for oleic
acid determination were obtained from plants of self-pollination arising from central
furrow of each plot. Oleic acid content was only determined in 1 replicate of each
experiment.
Grain
yield and oil content were analyzed by analysis of variance procedures (SAS, 1992). in
each environment, the LSD mean tests at 5% of significance was made when statistical
differences were detected
Stability
estimates
Fisher protected LSD test (Steel & Torrie,
1993)
The means of all hybrids in each environment are
compared with the mean of the best-yielding hybrid in that environment according to the
LSD test of multiple comparisons at 5% of significance. The most stable hybrids and those
that do not differ significantly from them will be the best-yielding ones, in most
environments.
As this method estimates the stability of the hybrids
comparing them with the best-yielding one, it does not consider stable those genotypes
that differ significantly lower yielding than the best, and which for that reason would be
undesirable.
Stability estimations of oil content were performed as
described for yield. This method was not applied for oleic acid content.
- Method of the relative yield (RY, Yau &
Hamblin, 1994)
This method consists in expressing the yield of each
hybrid, in each environment, in a way relative to the average of the environment in which
it was determined, assigning the value 100 to the latter. The standard deviation of
Relative Yields of each hybrid across environments is used as a measure of stability. The
most stable hybrids will be those with smaller standard deviations. Those with values
higher than 100 will be the hybrids of specific adaptability to a particular environment.
This method has the advantage of considering each environment equally in the calculation
of the average through all of them, that is to say, it does not favor the best
environments, especially when the number of entries and/or locations is large.
Yau and Hamblin (1994) applied this method for grain
yield and named it Relative Yield. In this work, the method was also applied for oil and
oleic acid contents and was named Relative Oil (RO) and Relative Oleic Acid (ROA)
respectively. This method was the only one applied for oleic acid content (ROA) because
replications are not required for calculus.
Adaptability
estimate
Method of multiple comparisons with the best
(Piepho, 1995)
This procedure belongs to the category of multiple
comparisons of means with the best. Piepho (1995) used it to estimate the specific
adaptability of cultivars. According to this technique, cultivars are classified into
three categories: adapted, non-adapted, and unclassified. The existence of this last
category diminished the type-1 error, not classifying the cultivars into adapted or
non-adapted, when they are not so. The method consists in creating a value designated as d
(delta) that is compared with each one of the confidence intervals of each genotype for
the difference between all the cultivars and the best cultivar. The value of d is the
smallest difference among the cultivars that is considered significant. The election of d
is subjective and depends on what the breeder considers as a reasonable value, being
significantly different for each vegetable species in question. Piepho suggests taking a
value that is 5-10% of the mean of the character in question, of all the cultivars in a
particular environment. In a specific environment, a cultivar is considered adapted if its
difference from the best cultivar is significantly lower than the value of d , whereas it
is considered non-adapted if its difference with the best is significantly higher than d .
Cultivars whose differences with the best cultivar do not differ significantly from d are
considered as unclassified.
The use of this method implies to recognize that the use
of the mean square of the error, S2, for the construction of confidence
intervals in a given year and location, limits the inferences to that particular
environment, since these can change due to the interaction of the genotype with the
environment.
This method was not applied for oleic acid content.
General
features of the methods
In
the three methods, the limits to consider a hybrid stable are subjective and can be
established according to different approaches. In this work, using the LSD test, the
hybrids considered stable were i) the best hybrids of each environment ii) the hybrids
that did not have significant differences with the best hybrid of the environment in at
least half of the environments studied. Meanwhile, the method of the RY considered as
stable the hybrids that had a deviation lower than the half of the maximum value found for
the hybrids studied. With the method of multiple comparisons with the best (Piepho, 1995),
values for d of 15% of the mean yield of all hybrids in an environment were used to
classify the hybrids as adapted, non-adapted, or unclassified. The value of 15% was chosen
because with d = 10%, as Piepho (1995) recommends using, none of the hybrids was
classified as adapted.
Results
and Discussion
Table
1 shows mean yields, oil and oleic acid contents and coefficients of variation (CV, %) for
each environment (year x location). Table 1 here. The mean environmental yield
fluctuated between 1020,9 kg/ha for the environment Pergamino 96 and 2905,7 kg/ha for
Bragado 97. The mean environmental oil content fluctuated between 38,7% for the
environment Venado Tuerto 96 and 48,6% for Balcarce 97. Considering for calculus the high
oleic genotypes alone, the mean environmental oleic acid content fluctuated between 79,6%
for the environment Carlos Casares 96 and 84,7% for Venado Tuerto 97.
Stable
and adaptable hybrids for grain yield for the three methods are given in Table 2,
Table 2 here.
Contiflor 9 and 5 HOEH were judged as stable with both the LSD and RY methods.
These hybrids did not present significant differences with the best-yielding hybrid in
more than half of the environments, exceeding the general mean of the trials (1965 kg/ha).
Besides, hybrids had high absolute and relative yields and low standard deviations.
Ten
HOEH, P-6661, DK 3881 and ACA 884 were considered as stable by the LSD method and unstable
by the RY method, with high absolute and relative yields in high-yielding environments and
high standard deviations. However, hybrids with high RY mean were particularly adapted to
those environments. The HOEH 950306 was unstable by LSD method and stable by RY method,
with low grain yield and standard deviations lower than those of the general mean of the
trials. The rest of the hybrids (21) had an unstable performance according to the two
methods. It is not possible to recommend them for sowing on a broad area as the RY values
indicate specific adaptability to a particular environment. All the hybrids were judged to
be unclassified for grain yield according to Piepho’s method.
Table
3 shows results of hybrid stability and adaptability for oil content according to the
three methods (Table 3 here). Aromo, Sideral and 8 HOEH were judged as stable with both,
the LSD and RO methods. They had high oil contents and low standard deviations. P-6661, Dk
3881, Contiflor and 5 HOEH were stable according to the LSD method and unstable according
to the RO method. They had high standard deviations. Meanwhile, the HOEH 960102 and 960202
were considered as unstable by the LSD method and stable by the RY method. Twenty-two
hybrids had an unstable performance according to both stability methods. However, the
values of RO for each hybrid could be useful to identify its specific adaptability to a
particular environment (Table 3). The most of hybrids showed in at least one environment a
RO value higher than 100. Some hybrids were classified as adapted according to
Piepho’s method. In sunflower, variability for oil content in trials is typically
lower than variability for grain yield. The resulting lower Mean Square Error (used to
calculate the hybrid adaptability by this method) probably explain why some genotypes were
classified as adapted for oil content and none for grain yield. According to Piepho (1997,
personal communication), the fact that most hybrids appear as unclassified is a precaution
should prevent us from drawing erroneous conclusions. All the hybrids were classified as
adapted in at least one environment according to Piepho’s method, Table
3.
Nine
HOEH were stable for oleic acid content (Table 4), with oleic acid content values superior
to 80%. Table 4 here.
None
hybrid was classified as stable for the three characters according to Yau and Hamblin
method. The HOEH 950303 was classified as stable for this method for grain yield and oleic
acid content. The HOEH 950501, 950604 and 950305 were classified as stable for Yau and
Hamblin method for oleic acid and oil content .
LSD and Yau and
Hamblin methods coincided in classifying the hybrids as stable and unstable in most of
hybrids (77% of the cases for grain yield and 76% for oil content). Although these methods
use different parameters to estimate stability, they are complementary and they are useful
in sunflower breeding programs. Combining both LSD and Yau and Hamblin methods it could be
possible to minimize losses of valuable genotypes in sunflower breeding programs. However,
the approach used to consider a genotype as stable according to the LSD method depends
largely on the number of environments where the genotype was tested. When the number of
environments is low (which it is not the case in the present work), the method of the
Relative Yield is more valuable than LSD methods to classify some genotypes as
high-yielding and stable, avoiding the bias that high-yielding environments give to the
general average. In addition to the value of standard deviation as a measure of stability,
the values of RY, RO and ROA give an idea of specific adaptability to a particular
environment. Some hybrids with high RO values were classified as adapted according to
Piepho´s method, meanwhile none hybrid was classified as adapted according to this method
for grain yield. It can be claimed that this method did not discriminate genotypes as the
other methods did. Similar results were obtained with other crops like soybean (Giménez
et al., unpublished data) and maize (Lúquez et al., 2001).
Tabla 1: Means and variation
coefficients (VC,%) of grain yield and oil and oleic acid contents (%) for all
environments in 2 years.
Table 2: Stability and adaptability
results for grain yield according to Yau and Hamblin (1994), LSD and Piepho methods.
Table 3: Stability and adaptability
results for oil content according to Yau and Hamblin (1994), LSD and Piepho methods
Table 4: Stability results for oleic
acid content according to Yau and Hamblin method (1994)
Para descargar las tablas, hacer
click aquí (.doc 285 Kb)
Acknowledgement
This work was supported by Instituto Nacional de
Tecnología Agropecuaria (INTA) and Nidera S.A., Pioneer Argentina, Mycoyen S.A., Van der
Have S.A., Dekalb Argentina, Cargill Argentina, Sursem, Ciba Geigy y Eureka Seeds. L.
Aguirrezábal is member of the Consejo Nacional de Investigaciones Científicas y
Técnicas (CONICET, Argentina).
References
Agüero, ME, V. Pereyra, L.A.N. Aguirrezábal, and J.
Lúquez, 1999: Rendimiento de grano y porcentaje de aceite en híbridos de girasol
"alto oleico" en Argentina. AgriScientia XIV: 49-53 . Córdoba. Argentina.
Annicchiarico,P.,1997: Joint regression vs. AMMI
analysis of genotype environment interactions for cereals in Italy. Euphytica 94: 53-62.
Brennan, P.S., D.E. Byth, D.W.Drake, I.H. De Lacy, and D.G. Butler.
1981: Determination of the location and number of test environments for a wheat cultivar
evaluation program. Australian Journal Research 32: 189-201.
Castaño, F., J.A. Kesteloot, and M.Colabelli, 1987:
Evolución y comportamiento de cultivares de girasol de ciclo corto e intermedio en la
región girasolera argentina. V Reunión Técnica Nacional de Girasol. 207-212. Bahía
Blanca, Argentina.
Crossa, J., 1990: Statistical analyses of multilocation
trials. Adv. Agron. 44: 55-85.
Crossa, J., P.L.Cornelius, M. Seyedsadr, and P. Byrne, 1993: A shifted
multiplicative model analysis for grouping environments without genotypic rank change.
Theoretical and Applied Genetics 85: 577-586.
de la Vega, A., S.C.Chapman, and A.Hall, 2000: Genotype by environment
interaction and indirect selection in sunflower for Argentina. I. Multi-attribute two-mode
pattern analysis. XV Sunflower International Conference. Toulouse. Francia. June, 12- 16.
page A85-A90. Tome I.
Eberhart, S.A., and W.A. Russell, 1966: Stability parameters for
comparing varieties. Crop Science 6: 36-40.
Fehr, W., 1987: Genotype x environment Interactions. En Principles of
cultivar development. Collier McMillan Publishers. London. 636pp.
Fernández Martínez, J., J. Muñoz, and J.
Gómez-Arnau, 1993: Performance of near-isogenic high and low oleic acid hybrids of
sunflower. Crop Sci. 33: 1158-1163.
Finlay, K.E., and G.N. Wilkinson, 1963: The analysis of
adaptation in a plant breeding programme. Australian Journal of Agricultural Research 14:
742-754.
Fox, P.N., and A.A. Rossielle, 1982: Reference sets of
genotypes and selection for yield inunpredictable environments. Crop Sci. 22: 1171-1175.
Gauch, H.G. and R.E. Furnas, 1991: Statistical analysis
of yield trials with Matmodel. Agronomy Journal 83: 916-920.
Gauch, H, 1988: Model selection and validation for yield
trials with interaction. Biometrics 44: 705.
Ghaderi, A., E.H. Everson, and C.E. Cress, 1980:
Classification of environments and genotypes in wheat. Crop Science 20: 707-710.
Giménez, F., J.E. Lúquez y J.C.Suárez. 2001: Estabilidad y
adaptabilidad de cultivares de soja para rendimiento en el sudeste de la provincia de
Buenos Aires. Revista de la Facultad de Agronomía de La Plata. En prensa.
Hildrich, T.P. and P.N. Williams, 1964: The Chemical Constitution of
natural fats. 4th. Chapman & Hall, London, pp. 688.
Kang, M.S. , and D.P. Gorman, 1989: Genotype x environment interaction
in maize. Agronomy Journal 81: 662-664.
Lin. C.S., and B.Thompson, 1975: An empirical method of
grouping genotypes based on a linear function of the genotype-environment interaction.
Heredity 34: 255-263.
Lorenzo, M.L., and Lorenzo, A., 1987: Adaptación y estabilidad relativa
de variedades e híbridos de girasol en la República Argentina. V Reunión Técnica
Nacional de Girasol. 213-222. Bahía Blanca, Argentina.
Ludueña, P., y Marta, L., 1979: Adaptabilidad de
cultivares de girasol a distintas zonas ecológicas de la República Argentina. Informe
Técnico 154. INTA. 8 págs.
Lúquez, J., y L.Jurá. 2001: Stability of corn hybrids
(Zea mays L.) for grain yield in Argentina. Maydica 46: 69-74.
Mandel, J., 1971: A new analysis of variance model for
no additive data. Technometrics 13: 1-18.
Pavoni, J.C., 1994: El índice de adaptabilidad relativa
("IAR") como una herramienta de predicción del comportamiento de un cultivar.
III Congreso Nacional de Trigo y I Simposio Nacional de Cereales de Siembra
Otoño-Invernal. Bahía Blanca, Argentina.
Piepho, H.P., 1995: Assessing cultivar adaptability by multiple
comparison with the best. Agronomy Journal 87: 1225-1227.
Piepho, H.P., 1998: Methods for comparing the yield
stability of cropping systems- A review. Journal of Agronomy and Crop Sciences. 180:
193-213.
Robertson, J.A., and W.H. Morrison, 1979: Analysis of
oil content in sunflower seed by wide-line NMR. Journal of American Oil Chemistry Society 56:
961-964.
SAS Institute, 1992: SAS technical Report P-229,
SAS/STAT Software: changes and enhancements, release 6.07, Cary, NC.
Steel, R., and J. Torrie, 1993: Comparaciones
múltiples. En Bioestadística. Principios y Procedimientos. McGraw-Hill. Ed. Segunda
Edición. 622 pág.
Uhart, S., M. Frugone, G. Pozzi, R.Correa, and C.
Simonella, 2000: Rendimiento y estabilidad de rendimiento en híbridos de girasol
linoleicos y alto oleicos: II. Efecto de factores combinados o del ambiente. A 91. 15th
International Sunflower Conference. Tome 1. Toulouse. Francia.
Vargas, M. J. Crossa, K. Sayre, M. Reynolds, M.
Ramírez, and M. Talbot, M. 1998. Interpreting genotype x interaction in wheat by partial
least squares regression. Crop Science. 38: 679-689.
Yates, F., and W.G. Cochran, 1938: The analysis of
groups of experiments. J. Of Agric. Sci. 28: 556-580.
Yau, S.K, and J.Hamblin, 1994: Relative yield as a
measure of entry performance in variable environments. Crop Science 34:
831-817.Fehr, 1987.
|
|
