The Effect of Density Dependence and Weather on Population Size of a Polyvoltine Species

The identification of what factors determine the population dynamics of polyvoltine species has been a difficult problem in ecology because population dynamics can contain intra- and interannual variability, and because the time scale at which factors affect the population is often unknown. We created a comprehensive population model to determine how density dependence (linear, nonlinear, and time-delayed) and weather affected the rate of population growth of white-footed mice (Peromyscus leucopus) in an isolated woodlot. We studied this nonoutbreak, polyvoltine species using a 257-mo data set spanning 23 yr, which incorporated both detailed intra-annual and long-term dynamics, and we used this model to forecast future population size. We then evaluated whether 3-yr spans of monthly data or a 22-yr span of annual data were better able to identify the key determinants that drive population dynamics, and we identified which data type created more accurate forecasts. The 257-mo comprehensive model determined that the intra-annual cycle was caused by seasonally varying intrinsic growth rates and density dependence on a 1–2 mo scale and indicated that peak population size in one year did not affect the population in the subsequent year. Interannual variability in peak and trough density was caused by the effect of weather on monthly rate of growth with a 0–2 mo time delay, with the exception of two droughts. These droughts negatively affected the population for 9 mo; the effects were probably mediated through reduced seed crop. This model explained 81% of the variability in density. Because weather determined interannual variability in density, forecasts that did not use known weather data during the forecast period were poor. When weather data were used, forecasts were accurate within 1–3 animals (10%) of observed densities up to 8 mo in the future but were inaccurate beyond 8 mo. We found that short-term monthly data detected more factors affecting the population and created more accurate forecasts than long-term annual data, because all factors affecting the population (except droughts) occurred on a monthly scale. The annual model did not detect any weather effects except droughts and detected annual density dependence, which represents time-delayed density dependence in polyvoltine species. We argue that this annual relationship is spurious and caused by studying this polyvoltine species on an inappropriate time scale. Our work suggests that the time scale of the analysis may affect the conclusions drawn about which types of factors determine population size and with what time lag. It also suggests that, even when population fluctuations can be explained, accurately predicting future densities may be impossible when fluctuations are driven by weather.