Autocorrelated rates of change in animal populations and their relationship to precipitation

I examined the prevalence of autocorrelation in mammalian, avian, and precipitation time series, how well autocorrelation in the environment translates into autocorrelation in animal populations, and length of the time series needed to accurately characterize the degree of autocorrelation. These are important questions because more-complex population models are incorporating autocorrelation terms in life-history characteristics and the intrinsic rate of increase. Including inaccurate or nonsignificant autocorrelation can alter the conclusions reached, providing either an unduly rosy or bleak picture of the likelihood of population viability and persistence. Using autocorrelation analysis in 175 vertebrate and 88 precipitation data sets, I found that 17.8% of the mammalian time series, 61.5% of the avian time series, and 97.7% of the precipitation data sets were autocorrelated. Carnivore populations were more likely than herbivore populations to show significant autocorrelation at lags of 2 or more years. I found only two cases of significant cross correlation between rate of population increase and local precipitation. This indicates that, although some environmental variables may be highly autocorrelated, it does not translate into autocorrelation in the resident animal populations. Based on subsampling of the precipitation and vertebrate data, I found that 15 years of data is sufficient to produce an autocorrelation not significantly different from one based on 100 years of data, although the variance continues to decrease with the length of the time series, as expected. My results suggest that, although some populations show temporal autocorrelation, it is not ubiquitous, and that environmental autocorrelation may not be a good predictor of autocorrelation in rates of increase. Population modelers should determine if autocorrelation exists in populations of interest prior to modeling their viability or probability of persistence because not all populations are equally influenced by autocorrelation.