Comparing complex fitness surfaces: among-population variation in mutual sexual selection in Drosophila serrata

Despite a dramatic increase in empirical estimates of phenotypic selection over the past two decades, we remain remarkably ignorant about variation in the multivariate fitness surfaces that shape the adaptive landscape. We develop a novel approach for quantifying patterns of spatial and/or temporal variation in multivariate selection that directly compares vectors of linear selection gradients (beta) and matrices of nonlinear selection gradients (gamma) that describe the multivariate fitness surface in each population. We apply this approach to estimates of sexual selection on a suite of cuticular hydrocarbons (CHCs) in males and females from nine geographic populations of Drosophila serrata. In males, variation in linear sexual selection was associated with the presence of the related species Drosophila birchii, suggesting that female mate preferences for male CHCs differ between sympatry and allopatry. This is consistent with previous experimental results suggesting that reproductive character displacement of male CHCs has resulted from selection caused by the presence of D. birchii. No significant associations were found for nonlinear sexual selection in males. In females, large-scale variation in both linear and nonlinear sexual selection was negatively associated with assumed-neutral population genetic structure, suggesting a key role for chance events in male mate preference divergence.