Print PageVariability and autocorrelation in climate factors have been theoretically and empirically implicated as strong predictors of extinction risk, invasive spread and population dynamics. In particular, and quite apart from mean conditions, variance and autocorrelation act together to determine the probability that non-native species will successfully invade a new location, and the rate of spread. Previous data have suggested that coastal locations will experience greater autocorrelation than inland locations, and thus might be more susceptible to invasion. However, climate change models may suggest increasing variability in environmental signals in many locations. Interpreting these predictions is sometimes difficult, however, since these models are generally at very large spatial scales. I concentrate on characterizing the variability and autocorrelation of temperature data in locations restricted to the northeastern half of North America.
Results/Conclusions
An analysis of minimum temperature data for 1932 to 2002 from weather stations east of the 100th meridian suggests that coastal (< 40 km from the coast), midcontinental (560-1600 km inland), and continental (>1600 km inland) locations have experienced different impacts of climate change. As expected, continental conditions were colder and more variable than midcontinental and coastal locations. Somewhat unexpectedly, there were no detectable differences in the autocorrelation of variance estimated using both Fourier analysis and detrended fluctuation analysis. Strong positive autocorrelation was found for all locations. However, a comparison of 1932 to 1970 and 1971 to 2002, indicated that climate change has altered the structure of variation differently in these locations. In continental and midcontinental locations, daily minimum temperature has become less variable and more strongly autocorrelated over time. This trend has serious implications for invasive spread and community dynamics in these locations.
