Modeling Genotype X Environment Interaction and Yield Stability in Multi-Environment Maize (Zea Mays L.) Yield Trial
No Thumbnail Available
Date
2008-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Multi-environment trials (MET) play an important role to develop an understanding of
how genotypes of an agricultural crop perform under different growing conditions. [n a
MET, a number of genotypes are tested in a number of environments using designs that
involve several replications per environment. Plant breeders conduct multi-environment
trials primarily to make cultivar evaluation and recommendation for a target region.
However, this task is not generally easy due to the frequent presence of genotype x
environment interaction. Genotype x environment interaction (GxE) is differential
genotypic expression across envirorunents. A significant GxE for a quantitative trait such
as yield can seriously limit efforts in selecting superior genotypes for both new crop
introduction and improved cultivar development. A number of methods and models have
been proposed to cope with the presence of GxE in multi -environment trial s.
Traditional statistical analyses of multi -environment trials provide little or no insight into
the parti(;ular pattern or stru(;ture of the GxE. The additive main <;:[fects and multiplicative
interaction (AMMI) model incorporates both additive and multiplicative components of
the two-way data structure which account more effecti ve ly for the underlying interaction
patterns. The least squares estimates of the main effect parameters for an orthogonal
AMMI model are identical with the least squares estimates of the parameters for the
model reduced to its additive part whereas the estimates of the multiplicative parameters
are the leading terms of the singular value decomposition of the matrix residual to the
additive part. In practice the series of multiplicative terms is truncated at some point
beyond which further terms are believed to have little stati stical significance. Results
from AMMI analysis presented in biplots underpin better informed decisions on variety
selection and recommendation in plant breed ing research programs. A breeding trial for
J 5 genotypes of maize (Zea mays L.) grown in 8 location-year environments serves as an
example.In our study, standard ANOV A analysis showed a significant genotype x environment
interaction effect for grain yield, in addition to the main effects. The environment effect
accounted for more than 80% of the variabi lity in the observed yie ld response. The
magnitude of genotype x environment interaction sum of squares was larger than the
genotype sum of sq uares. From a breeding viewpoint, this result gives some indication of
the possible challenge in selecting superior genotypes. Generally, the larger the
magnitude of the interaction term relative to the genotype effect, the more complex the
problem of identifying broadly adapted genotypes.
In an attempt to explain the GxE of grain yield in the maize data, we subsequently fitted a
regression on the mean model and an AMMI model. Results suggest that the AMMI
model with two dimensions for the interaction (AMMI-2) was preferable to the more
familiar ANOV A and linear regression procedures. Based on the selected AMMlmodel,
genotypes G14, Gl and G2 were found to be relatively more stable varieties. Moreover,
genotype G 1 was identified as a more desirable variety. Interaction patterns revealed by
AMMI biplots indicated that these maize genotypes are narrowly adapted because no
genotype has superior performance (both high mean yield and high stability) in all
environments. Among the 8 location-year enviromnents Arsinegelle in 2004, Areka in
2004, Goffa in 2005 and Areka in 2005 exhibited larger interactions and were more
discriminating of genotypes, whereas Awassa in 2004 exhibited negligible interaction and
was the least discriminating of genotypes. In the current study, Goffa was identified as a
location that was highly predictable in year-to-year interaction with genotypes (making
specific cultivar recommendation more reliable). In contrast, Awassa, Arsinegelle and
Areka were less predictable. Finally, it is recommended that fllture data coll ection should
include information on external environmental and/or genotypic variables.
Description
Keywords
Modeling Genotype X Environment Interaction