J F M A M J J Asondj F M A M

Months

FIGURE 5 The correlation function, calculated over the interval 1861-1988, showing the relationships between Santiago de Chile total monthly precipitation and tree-ring width variations at El Asiento in central Chile. Positive correlation indicates that above-average tree growth is associated with above-average values of precipitation. Correlation coefficients greater than 0.18 and 0.24 are significant at the 95 and 99% confidence level, respectively.

lected in 1974 by LaMarche (1975). For 1861-1988, an interval of time common to the tree-ring and instrumental records, a correlation function between tree-ring variations at El Asiento and Santiago precipitation shows that tree growth is significantly correlated with winter-spring precipitation (May-November; Fig. 5).

A gridded network of drought reconstructions over the continental United States has recently been derived from a large collection of climatically sensitive tree-ring chronologies (Cook et al., 1999). The reconstructions, based on the summer (June-August) Palmer Drought Severity Index (PDSI), were done on a 154-point, 2o latitude X 30 longitude regular grid. For the common time period 1700-1978, we correlated tree-ring variations from El Asiento in central Chile with the 154 PDSI reconstructions for across the United States. Not surprisingly, the strongest correlations between these records came from those PDSI reconstructions located in the Midwest and southern United States, two regions where climate variations are affected by SST changes in the equatorial Pacific.

It is well known that there is a strong teleconnection between SST changes in the tropical Pacific and precipitation anomalies in the southern United States and northern México (Ropelewski and Halpert, 1986; Ki-ladis and Diaz, 1989; Stahle and Cleaveland, 1993). Warmer SSTs in the tropical Pacific typically result in increased precipitation anomalies in this region. Spatial correlations between different indexes of tropical Pacific circulation and the 154-grid-point PDSI series show that the geographic location of the highest correlation field is the southwestern United States (Cook et al., 2000), a finding that is consistent with the patterns identified using instrumental records.

During the twentieth century, relationships between indexes of the tropical Pacific circulation and precipita tion records for the Midwest in the United States have been less obvious. Cook et al. (2000) showed that during the period 1897-1923 the teleconnections between the Southern Oscillation Index (SOI) and precipitation variations across the United States were radically different from those in 1924-1978. The significant SOI teleconnections with PDSI are much more extensive in the early period and penetrate northeastward into the upper Midwest-Great Lakes region. In contrast, the late-period teleconnections are highly contracted in the southwestern United States. Long-term teleconnec-tions across the United States were also investigated based on a winter SOI reconstruction from tree rings (1706-1977; Stahle et al., 1998). For the past three centuries, two contrasting patterns were suggested: one with expanded SOI teleconnections (1708-49 and 1850 - 99) and one with contracted teleconnections (1750-1999 and 1800-1949; Cook et al., 2000). Consequently, it is highly probable that the significant correlations between the central Chile record and the PDSI reconstructions for the Midwest reflect simultaneous changes in the teleconnections between the tropical Pacific and precipitation anomalies in these regions.

Hence, two PDSI reconstructions for the southwestern United States-northern México (370N) and two for the Midwest (3FN) were selected for comparison with the precipitation-sensitive records for South America. These four reconstructions explain, on average, more than 55% of the grid-point PDSI variance over the 1928-78 calibration period (Cook et al., 1999).

It is important to note that the strongest teleconnec-tions between precipitation in the southern United States-northern México and SSTs in the Pacific during the twentieth century have been identified for the winter months (Kiladis and Diaz, 1989), whereas the PDSI reconstructions that we used for comparison reflect mainly drought conditions during the summer. Despite this limitation, we found that the scores of the first PC from the four PDSI reconstructions in the United States are significantly correlated with tree-ring variations at El Asiento in central Chile (Fig. 6). For the common interval 1700-1978, the correlation coefficient between them is r = 0.32 (p <0.001), which could be considered as an indication of common modes of variations in these series.

10.5. SPATIAL CORRELATION PATTERNS BETWEEN TREE-RING RECORDS AND PACIFIC SSTs

Precisely dated tree-ring records for North and South America suggest that similar temporal modes of climate variations have affected the extratropical coasts

Years

FIGURE 6 Comparison of precipitation-sensitive records for central Chile (thick line) and the Midwest-southern United States (thin line) during the past 300 years. The El Asiento chronology is considered an estimate of past precipitation fluctuations in central Chile. The United States record is the leading principal component (PC) resulting from a PC analysis of the four Palmer Drought Severity Index (PDSI) reconstructions listed in Table 1 (Cook et al. 1996). The central Chile and United States records are significantly correlated at the 99.99% confidence level.

Years

FIGURE 6 Comparison of precipitation-sensitive records for central Chile (thick line) and the Midwest-southern United States (thin line) during the past 300 years. The El Asiento chronology is considered an estimate of past precipitation fluctuations in central Chile. The United States record is the leading principal component (PC) resulting from a PC analysis of the four Palmer Drought Severity Index (PDSI) reconstructions listed in Table 1 (Cook et al. 1996). The central Chile and United States records are significantly correlated at the 99.99% confidence level.

of the western Americas during the past three to four centuries. Assuming this is to some degree correct, the next step in our analysis was to identify the existence of a common forcing mechanism for these climatic oscillations. For comparison across the Americas, tree-ring records were selected for regions where climate variations are, to some degree, associated with SSTs in the tropical Pacific. Then, we proceeded to establish the spatial correlation fields between tree rings and SSTs across the Pacific. If the correlation patterns between the North American tree-ring records and Pacific SSTs are similar to those that result from using the South American chronologies, then they provide an indication of the role of SSTs across the Pacific as a common forcing mechanism of climate variations in the extra-tropics.

Correlation fields between tree-ring records and SSTs across the Pacific were calculated for the interval 1857-1988. SST records on a 5° latitude X 5° longitude grid were obtained from Kaplan et al. (1997). The SST data set, which consists of 583 points, covers the entire North Pacific from 60°N to the equator. Coverage across the South Pacific is more limited, particularly at higher latitudes. Except for the grids located along the South American coast, there is no information on SST for the southeastern Pacific south of 25°S. At 55°S, the SST coverage is reduced to four points: two south of New Zealand and two off the South American coast. There are no grid points south of 60°S.

Although trees respond primarily to seasonal rather than annual variations in climate, we used annual SSTs for calculating the correlation fields. The use of seasonal averages of SST may increase the magnitude of the correlation fields between tree rings and SSTs. However, because tree growth responses to climate vary between regions along the Americas, the comparison of spatial patterns based on different seasons could be problematic.

The spatial patterns that result from correlating the Alaska and northern Patagonia tree rings with Pacific SSTs are qualitatively similar (Fig. 7 [see color insert]). In both fields, positive correlations in the tropical Pacific and at mid- to high latitudes along the western coasts of the Americas contrast with negative correlations in the extratropical central and western Pacific. Significant positive correlations penetrate westward into the tropical Pacific region. Interestingly, both sets of temperature-sensitive records are not significantly connected with strong SST anomalies in the eastern equatorial Pacific. Similar spatial patterns were obtained when the tree-ring series were replaced in the calculations by annual SST anomalies from those grid cells located near the chronology sites (Fig. 7), indicating that the tree-ring estimates of temperature are capturing the large-scale spatial pattern of SST.

Correlation fields between SSTs across the Pacific Ocean and the precipitation-sensitive records for the Midwest-southern United States and central Chile were also calculated. In both patterns, significant positive correlations in the central Pacific near the date line extend eastward to the subtropical coasts of North and South America near 30°N and 30°S latitude, respective ly. Negative correlations in the extratropical North and South Pacific are also prominent (Fig. 8 [see color insert]).

The spatial patterns obtained by correlating the tree rings with the SST records closely resemble those observed for the decadal mode of Pacific SST variability identified by Latif et al. (1997) and Zhang et al. (1997). In contrast to the interannual mode of El Niño/Southern Oscillation (ENSO) variability, the decadal mode is characterized by less pronounced anomalies in the eastern Pacific (the classic key ENSO region) and is not narrowly confined along the equator. Similar to the tree-ring spatial patterns, the documented decadal oscillatory mode of Pacific SST shows anomalies in the western Pacific that extend to the northeast and southeast into the American subtropics.

0 0

Post a comment