Strains of
Chlamydomonas
were cultured in different macroenvironments created by manipulating levels of nitrate, phosphate and bicarbonate in liquid growth media. Cell density, measured by optical transmittance, increased in a density-regulated manner, permitting the logistic parameters
r
and
K
to be estimated for each genotype–environment combination. The main empirical results of a factorial experiment were as follows. (i) A large proportion of the overall genotypic variance in fitness measures was attributable to genotype-by-environment (G × E) interaction: 65 % for
r
and 50 % for
K
. Variance components for
r
and
K
were uncorrelated, but components of the interaction variance may have been correlated with corresponding components of the environmental variance, such that the relative fitness of genotypes was most strongly affected by environmental factors that have the greatest effect on average fitness. Higher-order interactions were as large as lower-order interactions, so that relative fitness was sensitive to particular combinations of environmental factors as well as to their main effects. The covariance of
r
with
K
also showed strong G × E interaction, being negative in some macroenvironments and zero in others. (ii) An ‘environmental’ decomposition of the G × E interaction variance separates ‘inconsistency’, due to lack of complete correlation between genotypes over macroenvironments, from ‘responsiveness’, due to differences between environmental variances among genotypes. Inconsistency was much the larger component for both
r
and
K
, showing that the greater part of the interaction variance was created by changes in the ranking of genotypes with respect to fitness between macroenvironments. When reaction norms were defined as the linear regressions of genotypic value on mean environmental value, substantial variance among reaction norms was detected : nonlinear effects were also large. (ii) A ‘genetic’ decomposition of the G × E interaction variance separates a component due to lack of complete genetic correlation from one due to differences in genetic variance. Incomplete genetic correlation was much the larger effect, the mean correlation between genotypes in two macroenvironments being only about +0.23 for
r
and +0.45 for
K
. A very striking observation was that the genetic correlation decreased as the difference between environments increased. It declined from +0.31 (for
r
; + 0.58 for
K
) when one factor differed between macroenvironments to +0.18 ( + 0.40) when two factors differed, and to +0.13 ( + 0.24) when all three factors differed. Furthermore, the genetic correlation varied inversely with the difference between environmental values, approaching zero when this difference was maximal. A measure of environmental consistency was obtained by plotting the score of a genotype in a given macroenvironment on its mean score over all macroenvironments, to identify environments in which generally inferior genotypes performed relatively well and vice versa. This analysis revealed some differences between macroenvironments, but nonlinear effects were again large. (iv) The two major empirical results of this investigation were (
a
) that much of the variance in fitness among genotypes is due to G × E interaction caused by incomplete genetic correlation, and (
b
) that genetic correlation is smaller between environments that are less similar. Both the relevance and the limitations of these findings with respect to the interpretation of diversity are discussed.
Genetic Parameters and Genotype by Environment Interaction in Radiata Pine for Growth and Wood Quality Traits in Australia
