Author:
Chapman S. C.,Cooper M.,Butler D. G.,Henzell R. G.
Abstract
Past sorghum hybrid trials in north-eastern Australia have detected
substantial genotype by environment (G×E) interactions for yield in
sampling a variable target population of environments (TPE) that is affected
by spatial and seasonal differences in crop water supply. Three datasets,
comprising yields of commercial and final stage experimental hybrids and
covering 9–17 years (Y) and up to 30 locations (L), were analysed to
quantify variance components for trial error, genotypic
(σ2g), and G×E
(σ2gl,
σ;2gy, and
σ2gly) interaction effects.
Whereas trial means varied 2–3-fold across seasons, a greater range was
estimated for variance components of trial error (range of 0.05–0.5), G
(0–>0.3), and G×L interaction (0.05–>1.0). There was
substantial seasonal variation in the ratio of
σ2g to
(σ2g
+σ2gl), and in two
datasets, 73% of the seasonal
σ2gl was due to poor genetic
correlations among locations. This implies that any given set of hybrids in a
random set of locations would be ranked differently from season to season.
Analysis of locations over years detected 90% of the total G×E
interaction as G×L×Y, rather than G×L or G×Y, although
this was reduced by accounting for genotype maturity. To achieve
repeatabilities of >80%, trials would need to be conducted over at
least 5 years and 20 locations per year.
The variable and unpredictable nature of much of the G×E interaction in
the region implies that broad adaptation to different water regimes is
required, unless prior knowledge of the seasonal weather can be used to choose
‘narrowly adapted’ cultivars. With current approaches, a large
sample of environments is needed to identify such hybrids, and testing across
locations and years is equally important. Alternative breeding strategies
based on classifying environment types are discussed.
Subject
General Agricultural and Biological Sciences