The cross-spectral analyses of tree-ring records for extratropical regions of North and South America suggest the existence of common decadal modes of climate variations along the western Americas. Over the length of common record, independent tree-ring records for North and South America are significantly correlated in a positive sense (Figs. 4 and 6), implying that similar trends in temperature and precipitation along the western coasts of the Americas have prevailed for the past three to four centuries. However, when correlation coefficients between the tree-ring series from North and South America are calculated for different subintervals, the magnitude of the relationships between these records changes over time, being typically weaker during the twentieth century.
Although local or regional climatic changes may contribute to the lack of correlation during some sub-periods, a change in the strength of the tropical Pacific teleconnections is also a valid explanation. We used singular spectral analysis (SSA), a data-adaptive method for extracting signals from noise in a time series (Vautard and Gill, 1989; Vautard, 1995), to determine changes in the oscillatory modes that most contribute to the positive relationships between the independent temperature- and precipitation-sensitive records for North and South America. Similar to other spectral techniques, the choice of lags in SSA is a compromise between the amount of information to be retained (resolution) and statistical significance (stability). We experimented with lags ranging from 5 to 15% of the series length, and we found that a lag equal to 10% of the series length (39 and 29 for the temperature- and precipitation-sensitive records, respectively) adequately resolved most decadal-scale oscillatory modes.
Three major waveforms, representing decadal modes of common variance at 9, 12.8, and >50 years, were isolated from the original temperature-sensitive records (Fig. 11). In general, the temporal evolution of these components shows similar fluctuations in amplitude and intensity from 1600 to ca. 1850. After 1850, relationships between waveforms break up, particularly for the 9- and 12.8-year components. This change is clearly revealed by changes in the correlation coefficients between the northern Patagonia and Alaska waveforms before and after 1850 (Fig. 11). For instance, the correlation coefficient between the components centered on 12.8 years changes from r = 0.73 to r = 0.02 for the intervals 1592-1849 and 1850-1988, respectively.
As a way of validating the previous results, we split both the northern Patagonia and Alaska records into two independent series (1592-1849 and 1850-1988) and proceeded to spectrally characterize each individual segment using BT spectral analysis. As expected, most of the peaks identified for the interval 1592-1849 agree with those previously reported for the entire series (Fig. 12). In contrast, the spectral estimates for the most recent portion of the series look somewhat different. A significant peak at 10-12 years is observed for the northern Patagonia record, whereas much of the variability for the 1850-1988 part of the Alaska records is confined to a period of 6.6-10 years (Fig. 12). In addition, cross-spectral analyses between the northern Patagonia and Alaska records show large differences in coherence between the early (1592-1849) and late (1850-1988) parts of the series. Tree-ring oscillations for northern Patagonia and Alaska were highly coherent from 1600-1850, particularly at decadal-scale oscillatory modes of 6.7, 8.4-16.5, and >50 years. Yet, except for oscillations longer than 50 years, there is no significant
Was this article helpful?