The calculation and interpretation of climatic predictabilities

Descriptors of predictability for temperature and precipita- tion were calculated as sums of constancy (C) and contingency (M) components. These indices were compared with the proportions of vari- ability accounted for by models developed using Fourier and ARMA methods of time series analysis. Data from six sites in North America at latitudes ranging from arctic to sub-tropical were analysed. The predict- ability indices for both temperature and precipitation were similar across all sites, although climatic regimes differed widely. Similarity in predict- ability indices of temperature among these sites resulted from opposing latitudinal gradients in constancy and contingency; constancy decreased and contingency increased with increasing latitude. A model constructed using the single peak of Fourier spectra, a 52-week periodicity, indicated both a greater predictability for temperatures and greater differences among sites than did the indices. Patterns in the constancies or contingencies of precipitation were more complex. Predictability indices were lower than those for temperatures at each site, and latitudinal gradients were not evident in either component. Proportions of variation accounted for using models derived from Fourier and ARMA analyses were similar, but both indicated much lower predict- ability for precipitation than did the sum of constancy and contingency. Two factors affected the results and interpretation of indices of climate pattern: (1) the impact of rare climatic events when differing timespans are compared, and (2) effects of global versus local ranges in calculations of predictability and its components. The similarity in indices of the predict- ability of climates, while components of predictability differ widely, sug- gest that the influence of climate on life history characteristics of species may be due to differences in the impacts of constancy and contingency on life histories and predictability.