Info

°®®9Qoc1

)0¿¿¿6éÓÓÓÓÓ(SCD0

FIGURE 3 Eigenvalue spectrum of the tree-ring data employed in this study (O). For reference, the eigenvalues of red noise time series with similar autocovariance statistics (+) and of historical observations from the proxy sampling sites (X) are also shown.

both the proxy data and benchmark data calibration experiments. Not surprisingly, the proxy calibration captures the same pattern Villalba et al. (2000) observed. Resolution of this pattern is corroborated by a similar (but stronger) pattern from analysis of the benchmark instrumental data set (Table 1). However, the error in the resolution of this pattern [R, Eq. (7)] is large in both cases, due to the difficulty with which these sparse observational networks discern the climatic variability. Results described later show that observational error variances are, on average, about 80% of the signal variance.

4.3.3.2. Contribution of Individual Indicators to the Reconstructed Mode

Table 2 shows the correlations of individual tree-ring indicators with the single mode of reconstructed SST variability as a function of calibration period and time lag. We interpret these correlations as the contribution made by each indicator to the reconstructed pattern. Lag 0 or lag +1 correlations that are significant above the 90% confidence level in both calibration periods are in bold. Seven of the indicators (3, 4, 5, 7, 9, 12, and 15) meet these criteria for simultaneous correlations, and nine (2, 3, 4, 5, 7, 9, 11, 12, and 14) meet these criteria for lag + 1 correlations. Six indicators (3, 4, 5, 7, 9, and

FIGURE 3 Eigenvalue spectrum of the tree-ring data employed in this study (O). For reference, the eigenvalues of red noise time series with similar autocovariance statistics (+) and of historical observations from the proxy sampling sites (X) are also shown.

FIGURE 4 (a) Correlation of the first tree-ring principal component (PC) with the sea surface temperature (SST) field. (b) Correlation of the first PC derived from benchmark instrumental observations with the SST field.

12) are significantly correlated with the reconstructed mode at both lags and in both calibration periods. These subsets of indicators include precipitation- and temperature-sensitive tree-ring indicators from both hemispheres, in agreement with the hypothesis (Villalba et al., 2000) that the resolved SST pattern represents a common forcing of both Northern and Southern Hemisphere extratropical temperature and precipitation anomalies. Consequently, the reconstructed SST pattern produced independently by using the 1923-90 and 1856-1922 calibration intervals share almost 60% variance. However, the results also show that several indicators (1, 6, 8, and 13) did not contribute consistently within both calibration periods or for either lag relationship to the reconstruction, and two (13 and 14) changed the sign of their simultaneous correlations between calibration periods.

If the assumption is made that the relationships between SST and tree-ring-site climate represented statis-

TABLE 2 Contribution of Tree-Ring Indicators to Reconstruction

Correlation with time series of reconstructed pattern0

1923-1990 calibration 1856-1922 calibration

Indicator6

Lag 0

Lag +1

Lag 0

Lag +1

0 0

Post a comment