Comparing lifetime and annual fitness measures reveals differences in selection outcomes

Selection analyses of long-term field data frequently use annual comparisons from long-lived species with overlapping generations to measure fitness differences with respect to phenotypic characteristics, such as annual phenological timing. An alternative approach applies lifetime estimates of fitness that encompass several annual events. We studied selection on emergence date from hibernation in male and female Columbian ground squirrels, Urocitellus columbianus (Ord 1915). From 32 years of records, we estimated lifetime fitness using either lifetime reproductive success (LRS) or matrix methods, and estimated annual fitness from individual yearly survival and reproduction. We also modified estimates to statistically control for changes in mean population fitness over the lives of individuals. We regressed lifetime fitness metrics on dates of emergence from hibernation, to quantify the strength of selection on emergence date (a heritable trait). All fitness metrics were highly correlated, but differences became apparent when estimating selection coefficients for emergence dates. The annual fitness metric and LRS produced lower effect sizes of selection coefficients than matrix-based lifetime fitness and a lifespan approach based on average annual fitness. Further, only these last two metrics revealed stabilizing selection. These results suggest that the choice of a fitness metric may influence our conclusions about natural selection.