Variation in selection and its evolutionary consequences

Abstract

It is widely appreciated that many aspects of selection change temporally when populations experience stochastic environments. However, the long term consequences for adaptive evolution has received less consideration. The availability of long term data, including phenotypic and life history information, along with high quality pedigrees, allows old approaches and theoretical ideas to be re-considered, and applied, empirically.

To investigate viability selection in a wild population of Soay sheep (Ovis aries) I estimated the sensitivity of selection to changes in population size. Using environmentally structured fitness functions I investigated pathways through which changes in population size drive changes in selection estimates. Much of the observed variation in selection was driven by changes in mean fitness. If generally true in other populations this may have important consequences regarding the occurrence of fluctuating selection. I then developed a method to estimate both individual, and cohort, success using theory based around Fisher’s reproductive value. This approach overcomes some complications associated with overlapping generations and age structure common in natural populations. Using pedigree information I tracked relative contributions to the future population through time and additionally used alternative theory from demographic literature to calculate stochastic total cohort reproductive values. The different estimates corresponded closely. Finally, I weighted selection differentials by measures of cohort reproductive success. When generations overlap, correctly considering the contribution a cohort makes to the future population can greatly reduce the expected response, as cohorts that experience stronger selection typically contribute less. This might provide one explanation for the paradox of stasis.

By linking approaches from multiple different fields with high quality, long term, data, I demonstrate that ecological pathways that influence selection can be identified, providing greater understanding of longer term evolutionary dynamics. Further applications of the methods developed could generate opportunities to investigate generalities in temporal evolutionary patterns not previously considered.