Can population growth rates vary with the spatial scale at which they are measured?

Abstract

1. The ratio of successive population censuses is often assumed to reflect population growth rates. We identify three simple potential sources of bias in the estimation of population growth rates that relate to either the total number of censused individuals or the spatial areas over which censuses are conducted. 2. The commonly used method of adding a constant to time series data to avoid problems caused by division by zero can lead to underestimation of growth rates at low densities in increasing populations. 3. Variances associated with density estimates can lead to positive bias in estimation of growth rates when populations are distributed in ephemeral patches. The spatial variance and spatio‐temporal covariance in bank vole census data suggest that this bias could be severe when small trapping grids are used. Use of logged estimators of growth rate avoids this problem. 4. Using census data from non‐randomly placed trapping grids that are smaller than twice the maximum range of natal dispersal to estimate population growth rates can lead to negatively biased estimates, particularly at low population densities. 5. These three sources of bias are evaluated as explanations for scale‐dependent changes in the estimates of growth rates identified in populations of snowshoe hare (Lepus americanus), bank voles (Clethrionomys glareolus) and lemmings (Lemmus lemmus).