Estimation of temporal variability of survival in animal populations

Temporal variation of demographic characteristics for animal populations is of interest to both ecologists and biological modelers. The standard deviation of a series of estimated parameter values (e.g., estimated population size) or some function thereof (e.g., log of the estimated parameters) is commonly used as a measure of temporal variability. These measures of temporal variation overestimate the true temporal variation by not accounting for sampling variability inherent to the estimation of unknown population parameters. Using a variance-components approach to partitioning the total variability of an estimated parameter, we demonstrate the ease with which sampling variation can be removed from the observed total variation of parameter estimates. Estimates of temporal variability of survival are given after removal of sampling variation for three bird species: the federally listed Roseate Tern (Sterna dougallii), Black-capped Chickadees (Parus atricapillus), and Mallard ducks (Anas platyrhynchos). Sampling variation accounted for the majority of the total variation in the survival estimates for nearly all of the populations studied. Substantial differences in observed significance levels were observed when testing for demographic differences in temporal variation using temporal variance estimates adjusted and unadjusted for sampling variance.