Treering Records

Along the western coasts of North and South America, there is a gradual environmental gradient from the dry and warm subtropics to the wet and cold high latitudes. Tree-ring records for subtropical regions, such as southern California and central Chile, are remarkably sensitive to precipitation variations (Haston and Michaelsen, 1994; Boninsegna, 1988). In the transitional zones to higher latitudes, tree-ring responses to climate are largely determined by site conditions. Depending on elevation, aspect, slope, and soil characteristics, tree growth can be influenced by temperature, precipitation, or, more commonly, by a combination of both. At the extreme wet and cold environments in high-latitude upper tree lines, temperature is the limiting factor controlling tree growth (Wiles et al., 1996, 1998; Villalba et al., 1997). These changes in tree response with latitude along the western Americas were instrumental in our setting the strategies for selecting tree-ring records sensitive to temperature and precipitation variations. Upper-elevation chronologies for coastal sites around the Gulf of Alaska and northern Patagonia were selected as proxy records of temperature for North and South America, respectively. Tree-ring records for dry environments in the south-central United States and central Chile were used for the in-terhemispheric comparison of precipitation-sensitive records (Fig. 1).

FIGURE 1 Geographical locations of tree-ring chronologies (triangles) and Palmer Drought Severity Index (PDSI) reconstructions (circles) considered in this study. See Table 1 for site names and identification of the PDSI reconstructions.

FIGURE 1 Geographical locations of tree-ring chronologies (triangles) and Palmer Drought Severity Index (PDSI) reconstructions (circles) considered in this study. See Table 1 for site names and identification of the PDSI reconstructions.

To facilitate the comparison between the North and South American records, ring-width measurements for all trees at each site were standardized by using similar techniques. Standardization involves fitting an expected growth curve to the observed ring-width series and computing an index of the observed ring widths in relation to the expected values. This procedure reduces the variance among cores and transforms ring widths into dimensionless index values. Thus, standardization permits computation of average tree-ring chronologies without the average being dominated by the faster growing trees with larger ring widths. Residuals from the expected growth curve were calculated. Potential biases introduced by standardizing tree-ring series as ratios between actual and expected ring widths were avoided by taking residuals from the growth curve after appropriate power transformation of tree-ring series (Cook and Peters, 1997).

Tree-ring chronologies were produced by the ARSTAN program (Cook, 1985). ARSTAN generates chronologies by combining standardized tree-ring series with biweighted robust estimation. To preserve the long-term fluctuations in the series, we used a conservative method of detrending. Only negative exponential curves or linear regressions were used.

0 0

Post a comment