Principles of Plant Genetics and Breeding

Different models of ANOVA are used for partitioning
variance. The genotypes are never chosen at random
since they are deliberately selected by the breeder as
prospective cultivars. Similarly, locations are often not
randomly chosen as previously discussed. However,
they may be considered random if there are many of
them spread over a large region. The genotypes are
PERFORMANCE EVALUATION FOR CROP CULTIVAR RELEASE 423
evaluated over several years. The effects are random
since the environment is not controlled. Also, where the
genotypes are a random sample from a large population,
their effects are random.
For a two-factor mixed model (fixed genotypes
+ random environment) the ANOVA table is as
follows:
Source df MS EMS
Replications (r) ly(r − 1)
Years (y) y − 1
Location (l ) l − 1
y × l (y − 1)(l − 1)
Genotypes (G) g − 1 M 1 σ 2 e + rσ 2 gly + rlσ 2 gy + ryσ 2 gl + rlyσ 2 g
G × y (g − 1)(y − 1) M 2 σ 2 e + rσ 2 gly + ryσ 2 gl + rlyσ 2 g
G × l (g − 1)(l − 1) M 3 σ 2 e + rσ 2 gly + ryσ 2 gl
G × y × l (g − 1)(y − 1)(l − 1) M 4 σ 2 e + rσ 2 gly
Error (G × r) (g − 1)(r − 1)ly M 5 σ 2 e
Variances are calculated as follows:
σ 2 e = M 5
σ 2 gly = (M 4 − M 5 )/r
σ 2 gy = (M 2 − M 4 )/rl
σ 2 gl = (M 3 − M 4 )/ry
σ 2 g = (M 1 − M 2 − M 3 + M 4 )/rly
Breeding implications of ANOVA results
Lack of significant G × E interactions for genotype ×
location or location × years indicates that the breeder
may be able to select a superior genotype for release for
use throughout a specified production region, following
genotype evaluations in just 1 year. Crossover G × E
interactions are those that require careful interpretation
by breeders. In this instance decisions are based on the
practical significance of the results of the analysis. Breeders
need to include in their decision-making process factors
such as the magnitude of the change in rank. Consequently,
it is not uncommon for different conclusions
to be drawn by different breeders examining the same
results. General interpretations of G × E interactions
resulting from unpredictable causes are as follow:
1 If significant genotype × location effects are observed
and the rankings fluctuate by wide margins, the results
indicate that the breeder should consider establishing
separate breeding programs for the different locations
(i.e., develop different cultivars for different locations).
However, before making a decision, it is wise
to examine the data to see what specific factors are
responsible for the variation. If stable factors such
as soil are the source of variation, separate breeding
efforts may be warranted.
2 A significant genotype × year interaction is similar in
effect to genotype × location. However, because the
breeder cannot develop programs for different years, a
good decision would be to conduct tests over several
years and select the genotype with superior average
performance over the years for release. Because conducting
one trial per year for more years will prolong
the breeding program, the breeder may include more
locations and decrease the number of years.
3 The breeding implications for a complex interaction
like genotype × years × location is for the breeder to
select genotypes with superior average performance
across locations and over years, for release as new cultivars
for the production region. Farmers will benefit
from growing more than one cultivar each cropping
season. This strategy will reduce the effects of the
fluctuations attributed to genotype × year interactions.
4 The magnitude of a G × E interaction is influenced
by the genetic structure of the genotype. Genotypes
with less heterogeneity (e.g., pure lines, single-cross
hybrids, clones) or heterozygosity (e.g., pure lines)
generally interact more with the environment than
open-pollinated genotypes or mixtures, because of
lower amounts of adaptive genes.
5 Also, it is widely known that only G × location interactions
(rather than all the kinds of G × E interactions)
are useful for depicting adaptation patterns.
This is because they are the only interactions that
can be exploited by selecting for specific adaptation
or by growing specifically adapted genotypes. For