Abstract

Patterns of correlation between tree rings and local temperature or precipitation are investigated using 762 International Tree-Ring Data Bank standardized ring width site chronology time series, and a gridded dataset of temperature and precipitation. Coherent regional- and, in some cases, hemispheric-scale patterns of correlation are found in the extratropical Northern Hemisphere for both the summer prior to and the summer concurrent with ring width formation across different species and over large distances.

Among those chronologies that are primarily linked to temperature, thicker ring widths are generally associated with anomalously cool prior summer temperature and anomalously warm concurrent summer temperature. Reconstructions of local summer temperature using prior, concurrent, and/or subsequent year ring widths as predictors demonstrate that useful climate–growth information generally exists in ring widths that are both concurrent with and subsequent to the summer temperature anomaly. Consistent prior summer temperature–ring width relationships have received relatively little previous attention. Among those chronologies that are primarily linked to precipitation, thicker ring widths are generally associated with high summer precipitation in both the year prior to and the year concurrent with ring formation. The magnitude and spatial consistency of temperature correlations are greater than those for precipitation, at least on the hemispheric scale. These results support and serve to generalize the conclusions of prior regionally restricted and/or species-specific studies relating ring width to energy and/or water limitations.

Regional- and hemispheric-scale patterns of ring width–temperature or ring width–precipitation correlations show up more clearly in species-specific and frequency-dependent analyses. Different species respond differently to temperature and precipitation anomalies. Consistent with the hemispheric patterns described above, most standardized ring width time series more faithfully record the high frequency component of the temperature signal than the low frequency component. The potential for enhanced coherence in regionally restricted, species-specific, and frequency-dependent analyses is independently verified by examining the correlation between ring width time series over geographical distance. This broader characterization of relationships between tree-ring widths and local climate provides an objective basis for selecting tree ring or other similarly high-resolution proxy data for regional-, hemispheric-, or global-scale paleoclimate reconstructions.

1. Introduction

Many paleoclimate reconstructions rely on tree-ring width measurements or multiproxy data assemblages including ring widths. Proxy-based reconstructions incorporating ring widths have been performed for climate variables such as regional, hemispheric, or global average surface air temperature (Jones et al. 1998; Mann et al. 1999; Crowley and Lowery 2000; Briffa et al. 2001; Esper et al. 2002; Mann and Jones 2003; Jones and Mann 2004; Huang 2004; Moberg et al. 2005; Jones et al. 2009), spatially varying fields of surface temperature and sea level pressure (Briffa et al. 2002a), and the time history of indices of leading patterns of climate variability such as the North Atlantic–Arctic Oscillation (Woodhouse 1997; Cook et al. 1998; Cullen et al. 2001; Glueck and Stockton 2001; Cook et al. 2002; D’Arrigo et al. 2003), the Southern Annular Mode (Jones and Widmann 2003), the El Niño–Southern Oscillation (Stahle et al. 1998; Mann et al. 2000), and the Pacific decadal oscillation (Gedalof and Smith 2001; Biondi et al. 2001; D’Arrigo et al. 2001; MacDonald and Case 2005). The various reconstructions based on different data and different reconstruction approaches tend to be qualitatively similar but quantitatively different (e.g., Esper et al. 2004). Reconstructions of average Northern Hemisphere surface temperature have received considerable scrutiny.

There are at least two reasons for such differences. First, the use of different reconstruction techniques applied to different proxy (predictor) and instrumental (predictand) time series can result in dissimilarities between reconstructions. Second, reconstructions may capture different proportions of the total and/or frequency-dependent variance because different proxies have potentially different frequency responses for reasons having to do with their particular ecophysiological, chemical, and/or physical characteristics and environments. Differences in total and frequency-dependent calibration skill can translate into varying representations of the reconstructed past. There is therefore a need to better understand the region-, species-, and/or frequency-dependent differences in local tree-ring climate sensitivity so that paleoproxy reconstructions using tree-ring predictors can be as biologically and physically based as possible. The species and frequency dependence of tree-ring reconstructions has been described in papers too numerous to cite here but usually in the context of specific regions and/or species.

Reconstructions also rely on proxy measurements with a limited geographic distribution, though tree-based measurements are relatively widespread compared to other proxy measurements. Variability in different tree-ring measurements (ring width, ring density, and isotopic concentrations) is driven by growth limiting factors including, but not limited to, temperature, precipitation, or other site-specific factors such as disturbance or forest density. Species and site selection are therefore critical components of climate reconstruction (Fritts 1976), but the possibility for regional or species assemblages to exhibit characteristic simultaneous and lagged relationships with local climate may be as important to the reconstruction skill as site selection.

Trees are sensitive to their most limiting factor. Drought reconstructions (e.g., Woodhouse and Overpeck 1998; Cook et al. 1999; Woodhouse 2003; Cook et al. 2004) have used ring width time series selected based on established a priori expectations that the combination of selected species and sites will indicate precipitation and/or evapotranspiration within a chosen precipitation-limited environment. Tree-ring species and sites for temperature-based reconstructions are likewise selected based on established a priori expectations that temperature is a growth limiting factor.

Briffa et al. (1990) demonstrated that maximum latewood density and total ring width time series could be used in combination to reconstruct Fennoscandian summer temperatures. Briffa et al. (1998) described a frequency-dependent climate signal between regional averages of summer temperature and both regionally averaged maximum latewood density and regionally averaged ring width along with a divergence in the relationship between wood density and summer temperatures at decadal time scales during the latter half of the twentieth century. Briffa et al. (2002b) explored hemispheric-scale temperature and precipitation relationships with both maximum latewood density and total ring width. They found the strongest temperature associations occurred in the summer concurrent with the tree-ring measurements and that the concurrent summer temperature signal was generally stronger in maximum latewood density than in total ring width, consistent with the results of Schweingruber et al. (1987). They also found precipitation relationships that were generally weaker than those for concurrent summer temperature.

Frequency-dependent variance (Fritts 1976, 268–275 and 295–300) and coherence (e.g., Briffa et al. 2002b) have been difficult to verify in some paleoclimate reconstructions, which has led to critiques (e.g., Schmutz et al. 2000; Esper et al. 2004). Frequency fidelity is generally dependent on proxy type (Jones et al. 1998) as well as on the particular measurement (e.g., D’Arrigo et al. 1992; Briffa et al. 2002b). Erlandsson (1936) proposed a method of preserving regionally coherent low frequency variability in tree-ring assemblages while removing artifacts from sources such as age or species dependence. This approach has been altered and refined over the years, for example, by Meko (1981), Briffa et al. (1992, 2001), Cook et al. (1995), and Esper et al. (2003, 2008). To preserve relevant information at the desired frequencies (and, ultimately, to generate the best possible reconstruction) it is useful to verify that the desired frequencies are captured in the underlying proxy–local climate relationship. A number of existing reconstructions deal explicitly with the frequency fidelity of the predictors in relation to the targeted predictand (e.g., Guiot 1985; Osborn and Briffa 2000) but not all do. The combined influence of geographical-, biophysical-, and frequency-dependent effects on the ring width–climate relationship has received relatively little attention.

Knowing the frequency fidelity of ring width–climate relationships on very broad spatial scales and for all species along with the same information from particular regions and/or species could contribute practical information on the heavily utilized ring width site chronology time series within the International Tree-Ring Data Bank (ITRDB). This paper characterizes the influence of both intrinsic and climate-related autocorrelation on the temperature and precipitation growth responses in tree-ring widths throughout the extratropical Northern Hemisphere. An illustrative subset of species-specific relationships is also shown. The generally increased coherence in region, species, and frequency-dependent ring width–climate relationships that result are independently supported by a nested analysis of correlation between ring width time series over distance. While we recognize that this paper has many features in common with existing studies, most notably Briffa et al. (1998, 2002b), we believe that our focus on autocorrelation and the out-of-phase correlations between ring width and summer temperature in the years prior to and concurrent with ring formation is unique. We also believe that the general characterization of latitudinal, regional, species-, and frequency-dependent correlations between ring width site chronologies and local summer temperature or precipitation are noteworthy, especially considering the widespread use of ring width time series in reconstructions.

2. Data and methods

Annually resolved tree-ring index data from the Northern Hemisphere were obtained from the International Tree-Ring Data Bank (information available online at http://www.ncdc.noaa.gov/paleo/paleo.html) described by Grissino-Mayer and Fritts (1997).1 The 762 standard ring width site chronology time series used throughout this study represent all ITRDB species with sites poleward of roughly 30° to 40° latitude. We started by selecting all sites poleward of 40° outside of arid regions that have been the focus of many prior studies on drought sensitivity (e.g., Woodhouse and Overpeck 1998; Cook et al. 1999; Woodhouse 2003) and subsequently extended the study to include the whole geographic range of the species so selected; hence, the irregular southern boundary of our study domain.

Alternate ITRDB tree-ring measurement types such as earlywood width, latewood width, and wood density measurements were also considered but generally do not provide either the spatial coverage or species diversity available for total ring width. For example, at the time of this paper’s submission, there were many times more ring width site chronology time series than latewood density time series available through the ITRDB. Furthermore, even if previous studies (e.g., Schweingruber et al. 1987; Briffa et al. 2002b) have noted the generally superior summer temperature correlations with maximum latewood density, many researchers use ITRDB ring widths in a variety of reconstructions that emphasize different seasons, species, and frequencies. The standard ring width site chronology time series are unaltered from their downloadable form, except where either a low-order 5-year high-pass or a 5-year low-pass digital Butterworth filter has been applied to investigate frequency dependence. By construction, adding the high-pass- and low-pass filtered data together produces the raw (unfiltered) data so that raw signals can be viewed as the sum of separate frequency-dependent signals.

The many ITRDB contributors individually standardized their ring width site chronologies prior to submitting them. Though methods of standardization vary, it is reasonable to speculate that many used regression-based detrending methods and/or autoregressive approaches such as Auto-Regressive Standardization (ARSTAN) (Cook 1985) and that standardization was generally conducted to combine multiple ring width time series into a single chronology representing the coherent signal in a single species at the site. ARSTAN and some alternate standardization protocols are described in Cook et al. (1990). The likelihood that alternate standardization procedures could materially affect the general sense of our results and conclusions is discussed later in the paper.

Relationships between ring width and temperature or precipitation were inferred from linear regression coefficients and Pearson correlation coefficients. The 762 ring width site chronologies were compared with the 427 closest grid point time series in a 1°- resolution version of the 1901–2000 gridded surface temperature and precipitation dataset developed by Mitchell et al. (2004). Ring width time series were required to contain at least 51 years of data in the twentieth century such that all results are based on at least 50 years of mutual record.

3. Results

Figure 1 shows maps of the one-year lagged autocorrelation for the 762 tree-ring width and the 427 summer temperature and precipitation time series. Given space limitations and the generally greater strength of summer correlations within our study domain [not shown, but in general agreement with Briffa et al. (2002a)], only summer results are shown. The standard ring width time series used in this study have a median one-year lagged autocorrelation of roughly 0.5 and median lagged autocorrelation values that are slightly positive even at lags of greater than five years. Prewhitened (residual) site chronology time series are also available through the ITRDB and would undoubtedly have substantially lower autocorrelation, but we do not include them here for fear of discarding potentially useful climate information, an issue we will revisit later in the discussion of our results.

Fig. 1.

Autocorrelation in 762 annually resolved tree-ring width and 427 summer- (JJA) averaged temperature and precipitation time series from grid points nearest the ring width sites. One-year lagged autocorrelations are shown in maps, where circle diameter scales linearly with autocorrelation magnitude, and smaller autocorrelations (circles) are plotted over larger ones. Medium blue, light blue, yellow, orange, light red, and deep red correspond to autocorrelations ranging from −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, 0.5 to 0.75, and 0.75 to 1, respectively. Histograms of the one-year lag autocorrelations are shown below each map with the 50th, 10th, and 90th percentile values indicated. (top right) Histograms of the zero–10-year lagged autocorrelation in ring widths are shown with the median, interquartile range, spread (1.5 times interquartile range), and outliers indicated.

Fig. 1.

Autocorrelation in 762 annually resolved tree-ring width and 427 summer- (JJA) averaged temperature and precipitation time series from grid points nearest the ring width sites. One-year lagged autocorrelations are shown in maps, where circle diameter scales linearly with autocorrelation magnitude, and smaller autocorrelations (circles) are plotted over larger ones. Medium blue, light blue, yellow, orange, light red, and deep red correspond to autocorrelations ranging from −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, 0.5 to 0.75, and 0.75 to 1, respectively. Histograms of the one-year lag autocorrelations are shown below each map with the 50th, 10th, and 90th percentile values indicated. (top right) Histograms of the zero–10-year lagged autocorrelation in ring widths are shown with the median, interquartile range, spread (1.5 times interquartile range), and outliers indicated.

Based on Fig. 1, it is evident that ring width time series exhibit substantially more year-to-year autocorrelation than summer temperature or precipitation time series. The larger interannual autocorrelation in standard ring widths than in meteorological data is a result that has been well known in the tree-ring community for many years (e.g., Guiot 1986) but is perhaps less well known outside that community. A few large regions (e.g., the Mediterranean coast, central and western North America, and south-central Russia) exhibit one-year-lagged temperature correlations consistently on the order of 0.25 or higher. Regions exhibiting consistent precipitation autocorrelation are more limited, at least within the domain of this study. The fact that the autocorrelation in the tree rings is so different from that of the temperature and precipitation time series suggests a potential challenge in using ring width time series as climate indicators, at least insofar as variance is apparently concentrated at different frequencies in the tree-ring and meteorological data. This challenge has also been well known for decades in the dendroclimatological community and much work has gone into methodologies to usefully extract climate signal (Erlandsson 1936; Meko 1981; Briffa et al. 1992, 2001; Cook et al. 1995; Esper et al. 2003, 2008).

Correlations between annually resolved ring width and seasonally averaged local temperature or precipitation time series (Fig. 2) show consistent regional- and, in some cases, hemispheric-scale patterns, in general agreement with Briffa et al. (2002b). Temperature correlation magnitudes and the spatial consistency of those correlations are greater for the prior and concurrent summer (Fig. 2), compared with the prior winter (not shown). Correlations between ring width and prior summer temperature are mostly negative, so cooler summer temperatures tend to be associated with larger ring widths the following year. Conversely, positive correlations between ring width and concurrent summer temperature are more prevalent, so increased ring widths are also associated with anomalously warm concurrent summer temperatures at most sites. The negative correlation between ring width and prior summer temperature is particularly noteworthy as it would not necessarily have been anticipated based on the results described in Briffa et al. (2002b). The sign as well as the magnitude of the ring width correlations with concurrent summer temperature depend on latitude. High-latitude ring width time series are more likely to have a relatively strong positive concurrent summer temperature correlation, whereas low-latitude sites are more likely to have a strong negative correlation.

Fig. 2.

Correlation coefficients between the 762 ring width time series from Fig. 1 and seasonally averaged (top) local temperature and (bottom) precipitation time series from (left) the summer prior to (JJA−) and (right) the summer concurrent with (JJA) ring width formation. Circle color and diameter indicate the sign and magnitude of the correlations—deep blue, medium blue, light blue, yellow, orange, and red circles indicate correlation coefficients ranging from −1 to −0.5, −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, and 0.5 to 1, respectively. Circle diameter scales linearly with correlation coefficient magnitude, and smaller correlations (circles) are plotted over larger ones. A histogram of the correlation coefficients with the 50th, 10th, and 90th percentile values indicated is shown below each map.

Fig. 2.

Correlation coefficients between the 762 ring width time series from Fig. 1 and seasonally averaged (top) local temperature and (bottom) precipitation time series from (left) the summer prior to (JJA−) and (right) the summer concurrent with (JJA) ring width formation. Circle color and diameter indicate the sign and magnitude of the correlations—deep blue, medium blue, light blue, yellow, orange, and red circles indicate correlation coefficients ranging from −1 to −0.5, −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, and 0.5 to 1, respectively. Circle diameter scales linearly with correlation coefficient magnitude, and smaller correlations (circles) are plotted over larger ones. A histogram of the correlation coefficients with the 50th, 10th, and 90th percentile values indicated is shown below each map.

Precipitation correlation magnitudes are generally smaller than those for temperature, consistent with Briffa et al. (2002b). There is a prevailing tendency for positive correlations between summer precipitation and tree-ring width the following year that is most clearly evident at lower latitudes. There is also a latitude-dependent pattern of correlation between ring width time series and concurrent summer precipitation, but it is perhaps too noisy and weak to be of much practical use on a hemispheric scale.

Correlations between ring width and temperature or precipitation are consistent on large spatial scales, despite site-specific conditions and potentially heterogeneous species responses. Near the southern limit of our study domain, summer temperature correlations are usually negative for both the prior and concurrent years, which can be generally understood as the response of trees that are sensitive to soil moisture and therefore respond simultaneously to the influences of precipitation and temperature (via evapotranspiration) on soil moisture. Where they occur (especially in the northern reaches of our study domain), the generally opposite sign of prior and concurrent summer temperature correlations in Fig. 2 is potentially puzzling and could be related to a physiological trade-off between carbon allocation into other tissues [e.g., into cones, Woodward et al. (1994)], to the persistence of climate effects through physiological lags [e.g., budset, Littell et al. (2008)], or to the relatively strong ring width autocorrelation shown in Fig. 1.

As previously described, ring width time series exhibit substantially stronger interannual and lower-frequency autocorrelation than the meteorological data. Therefore, one thing to explore is the relationship between the excess ring width autocorrelation and the generally out-of-phase summer temperature correlations. A possible null hypothesis is that temperature affects tree physiology in only the prior summer or the concurrent summer and that strong ring width autocorrelation interacts with frequency-dependent climate sensitivity to influence out-of-phase tree–temperature correlations in the neighboring ring width. An investigation of this null hypothesis can be performed by comparing various models developed to predict local summer temperature, and the following paragraphs will summarize results from such a suite of linear regression models using lagged ring width time series as predictors. Nonlagged (concurrent) ring width time series are used as predictors of summer temperature in combination with ring width time series that begin and end a year earlier (the prior year ring width time series) or a year later (the subsequent year ring width time series) than the predicted summer temperature time series. The different regression models that result can be compared with one another and with the previous results that focused on autocorrelation and ring width–local temperature correlations. Such a comparison can help distinguish between the climate–growth response and the potentially obscuring influence of dissimilar autocorrelation in ring width and summer temperature time series.

Negative regression weights dominate in the panel of Fig. 3 devoted to the subsequent year’s ring width, consistent with the mostly negative correlations between ring width and prior summer temperature in Fig. 2. A latitude-dependent signature exists in the regression weights applied to the concurrent year’s ring width time series, consistent with the concurrent summer temperature correlations in Fig. 2. In other words, the pattern of regression coefficients applied to the concurrent and subsequent-year ring width time series within these trivariate models for local summer temperature is broadly consistent with expectations based on previous results.

Fig. 3.

Regression coefficients from trivariate linear regression models used to predict standardized local summer temperature using standardized prior-year, concurrent-year, and subsequent-year ring width time series as predictors. Circle diameter scales linearly with regression coefficient magnitude and smaller regression coefficients (circles) are plotted over larger ones. Dark blue, medium dark blue, medium light blue, light blue, yellow, orange, light red, and dark red dots correspond to regression coefficients ranging from −1 to −0.75, −0.75 to −0.5, −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, 0.5 to 0.75, and 0.75 to 1, respectively. See text for additional detail.

Fig. 3.

Regression coefficients from trivariate linear regression models used to predict standardized local summer temperature using standardized prior-year, concurrent-year, and subsequent-year ring width time series as predictors. Circle diameter scales linearly with regression coefficient magnitude and smaller regression coefficients (circles) are plotted over larger ones. Dark blue, medium dark blue, medium light blue, light blue, yellow, orange, light red, and dark red dots correspond to regression coefficients ranging from −1 to −0.75, −0.75 to −0.5, −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, 0.5 to 0.75, and 0.75 to 1, respectively. See text for additional detail.

Regression coefficients applied to ring width predictor time series from the year prior to the modeled summer temperature are generally smaller than those applied to the concurrent and subsequent-year ring width time series. Regression coefficients applied to the prior-year ring width time series are also generally opposite in sign from those applied to the concurrent-year ring width time series; note that regression coefficients applied to the prior and concurrent year’s ring width have a latitude dependence that is mostly opposite of one another. A nontechnical explanation of these results is that because it is impossible for the prior year’s ring width to anticipate summer temperature anomalies in the subsequent year, regression coefficients applied to the prior-year ring width time series cannot result from a direct physiological response to local temperature; in fact, it must be an artifact of multicollinearity (Meko 1981). Furthermore, the fact that the regression coefficients applied to the prior-year ring width time series are consistently out-of-phase with the regression coefficients applied to the concurrent year’s ring width strongly suggests they result from ring width autocorrelation. Finally, the fact that regression coefficients applied to the prior-year ring width time series are nonzero, but small relative to the regression coefficients applied to the concurrent and subsequent year’s ring width, suggests that climate–growth relationships are generally dominant over the potentially spurious influence of biological autocorrelation in these trivariate models of summer temperature.

A comparison of regression coefficients from the trivariate models of summer temperature in Fig. 3 with regression coefficients resulting from a variety of univariate and bivariate models of summer temperature (Table 1) suggests not only that regression coefficients are generally stable across different univariate, bivariate and trivariate models (especially the coefficients applied to the concurrent and subsequent year’s ring widths), but also that the autocorrelation signal is generally concentrated in the prior year’s ring width. Both concurrent and subsequent year’s ring width therefore seem to contain useful information about the simultaneous and lagged growth responses to summer temperature, respectively. This interpretation is consistent with a more detailed site-by-site comparison of the regression coefficients resulting from the different univariate, bivariate and trivariate models summarized in Table 1 (not shown).

Table 1.

Correlation coefficients (in units of percent) between regression coefficients resulting from various linear models of local summer temperature. Tree-ring widths from the year prior to (p), concurrent with (c), and subsequent to (s) the predicted summer temperature are used as predictors in the various univariate, bivariate, and trivariate models. Correlations are calculated over the 762 models of local summer temperature using the 762 ring width time series as predictors. Correlation magnitudes below 0.2 are not shown.

Correlation coefficients (in units of percent) between regression coefficients resulting from various linear models of local summer temperature. Tree-ring widths from the year prior to (p), concurrent with (c), and subsequent to (s) the predicted summer temperature are used as predictors in the various univariate, bivariate, and trivariate models. Correlations are calculated over the 762 models of local summer temperature using the 762 ring width time series as predictors. Correlation magnitudes below 0.2 are not shown.
Correlation coefficients (in units of percent) between regression coefficients resulting from various linear models of local summer temperature. Tree-ring widths from the year prior to (p), concurrent with (c), and subsequent to (s) the predicted summer temperature are used as predictors in the various univariate, bivariate, and trivariate models. Correlations are calculated over the 762 models of local summer temperature using the 762 ring width time series as predictors. Correlation magnitudes below 0.2 are not shown.

Figure 4 shows correlations between ring width and prior and concurrent summer temperature or precipitation time series for three example species in Europe. The Norway spruce [Picea abies (PCAB)] temperature correlations follow the hemispheric-average patterns described for the all-species analysis of Fig. 2; increased ring width is associated with cooler prior summer temperatures and warmer concurrent summer temperatures. Black pine [Pinus nigra (PINI)] ring widths tend to be anticorrelated with prior summer temperature (in general agreement with the hemispheric-average pattern in Fig. 2), but concurrent summer temperature correlations also tend to be negative. One consistent interpretation is that, as summer temperature increases, water demand also increases and, because this increased water demand goes largely unmet in the semiarid Mediterranean environment, tree growth is suppressed and reflected in smaller ring widths. A similar response has been described for Douglas-fir (Pseudotsuga menziesii) in western North America (Littell et al. 2008). Much of the Mediterranean nonconformity with the hemispheric-average pattern of concurrent summer temperature correlation in Fig. 2 can therefore be attributed to the concentration of black pine ring width chronologies there. Negative correlations with both the prior and concurrent summer temperature suggest that black pine ring widths may also have increased sensitivity to low-frequency variability, compared to the other species in Fig. 4. Scots pine [Pinus sylvestris (PISY)] ring widths tend to have a positive correlation with concurrent summer temperatures in northern Europe and Scandinavia, in agreement with the hemispheric pattern from Fig. 2. Correlations also tend to be positive with prior summer temperature in northern Scandinavia, counter to the general hemispheric pattern in Fig. 2. Overall, Fig. 4 demonstrates that many of the departures from the hemispheric-average patterns in Fig. 2 can be understood as expressions of the peculiar distributions of specific species (e.g., black pine and Scots pine). The robustness of these species-dependent results will be discussed in additional detail later.

Fig. 4.

Correlation coefficients between ring width and (top three rows) temperature or (bottom row) precipitation time series as in Fig. 2, but for only a few example species in Europe.

Fig. 4.

Correlation coefficients between ring width and (top three rows) temperature or (bottom row) precipitation time series as in Fig. 2, but for only a few example species in Europe.

Frequency-dependent correlations between ring widths and temperature or precipitation (Fig. 5) are somewhat different from the raw temperature and precipitation correlations shown in Fig. 2. High-frequency temperature correlations with the prior summer (JJA−) tend to be more negative than the generally negative raw correlations, and high-frequency temperature correlations with the concurrent summer (JJA) tend to be more positive than the generally positive raw correlations. The latitudinally dependent pattern of correlation between ring width chronologies and concurrent summer temperature described previously for the raw correlations in Fig. 2 is concentrated in the low frequencies. Precipitation correlations are less obviously frequency dependent, at least on a hemispheric scale.

Fig. 5.

Correlation coefficients between (left and middle) 5-year high-pass (hp) and (right) 5-year low-pass (lp) filtered temperature and precipitation time series and high-pass- and low-pass-filtered tree-ring width time series in the summer prior to (JJA−) and concurrent with (JJA) ring width formation. Circle diameter scales linearly with correlation coefficient magnitude, and smaller correlations (circles) are plotted over larger ones. Dark blue, medium blue, light blue, yellow, orange, and red circles correspond to correlation coefficient ranges from −1 to −0.5, −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, and 0.5 to 1, respectively. A histogram of the correlation coefficients is shown.

Fig. 5.

Correlation coefficients between (left and middle) 5-year high-pass (hp) and (right) 5-year low-pass (lp) filtered temperature and precipitation time series and high-pass- and low-pass-filtered tree-ring width time series in the summer prior to (JJA−) and concurrent with (JJA) ring width formation. Circle diameter scales linearly with correlation coefficient magnitude, and smaller correlations (circles) are plotted over larger ones. Dark blue, medium blue, light blue, yellow, orange, and red circles correspond to correlation coefficient ranges from −1 to −0.5, −0.5 to −0.25, −0.25 to 0, 0 to 0.25, 0.25 to 0.5, and 0.5 to 1, respectively. A histogram of the correlation coefficients is shown.

Figure 6 illustrates that the species-dependent relationships between temperature or precipitation and tree-ring widths shown previously in Fig. 4 are also frequency dependent. The out-of-phase summer temperature correlation with Norway spruce ring widths is dominated by high frequencies, whereas the summer temperature and precipitation correlations with black pine ring widths are dominated by low frequencies. The unfiltered positive correlation coefficients between Scots pine ring widths and both prior and concurrent summer temperature shown in Fig. 4 for northern Scandinavia are largely the result of strong low-frequency temperature correlation; high-pass filtered correlations between Scots pine ring width and temperature shown in Fig. 6 are out-of-phase in the prior and concurrent summers, like the all-species pattern throughout the Northern Hemisphere (Figs. 2 and 5) and the Norway spruce summer temperature correlations shown in Fig. 6.

Fig. 6.

Frequency-dependent correlation coefficients as in Fig. 5 but only for the same few example species in Europe shown in Fig. 4.

Fig. 6.

Frequency-dependent correlation coefficients as in Fig. 5 but only for the same few example species in Europe shown in Fig. 4.

4. Conclusions and discussion

This paper has described simple, but robust, regionally and hemispherically consistent correlations between standard ring width site chronology and temperature or precipitation time series in the extratropical Northern Hemisphere. The consistency of correlation patterns on these spatial scales confirms that coherent climate signals are indeed recoverable using ring widths despite diverse local site characteristics with different ecological limitations, numerous different species, heterogeneity in site chronology standardization, and the unique interplay of all these factors for each of the ring width time series. Even though tree rings exhibit temperature and precipitation correlation patterns over broad spatial scales, the results presented here suggest that reconstructions using ring widths as candidate predictors will continue to benefit from deliberate and thoughtful predictor selection criteria based on location, species, and the seasonally dependent characteristics of both the observed target (predictand) and the ring width (predictor) time series. For example, the fact that this study emphasizes associations between standardized ring width chronologies and temperature in both the concurrent (in agreement with the results of Briffa et al. 2002a) and prior summer has practical implications for proxy selection and processing in a reconstruction. Perhaps most importantly, this work illustrates the potentially dissimilar frequency fidelity in ring widths as recorders of temperature, precipitation, or other climate variables and that the degree to which autocorrelation is climatic or biological or both in nature determines how it is handled in reconstructions.

Our study helps place previous regional and species-specific dendroecology and dendroclimatology work, for example Peterson and Peterson (2001, mountain hemlock), Littell et al. (2008, Douglas-fir), and Kipfmueller and Salzer (2010, whitebark pine), in the context of hemispheric and regional-scale patterns. It also puts the studies of multispecies networks within regions (e.g., Nakawatase and Peterson 2006; Holman and Peterson 2006) into a broader context. Since Briffa et al. (2002b), the nature of regional and species-dependent climate sensitivities has almost always been explored one species or one region at a time. Our results suggest that the utility of these prior studies for ecological, geophysical, and dendroclimatological tree-ring work could be better exploited if the nature of the frequency response is considered simultaneously. For example, Douglas-fir ring widths seem to have shorter autocorrelation time scales (autoregressive-1 to autoregressive-3 or 4 time scales are justified, Littell et al. 2008) than limber pine (autoregressive-6 time scales or longer are justified) in the western United States. Drought reconstructions that focus on just one or the other of these species or a poorly considered combination could be biased unless the nature of the spatial and temporal autocorrelation is considered or perhaps even exploited. Similarly, mountain hemlock (Peterson and Peterson 2001) and subalpine larch (Graumlich and Brubaker 1986; Pederson et al. 2011) both exhibit substantial decadal variance in the Pacific Northwest (PNW). However, we need to better understand whether this signature represents a coherent regional climate signal or a species-dependent response to be more confident in any reconstruction based on these time series.

Standardization practices differ according to the application, the nature of the chronology, and the researcher’s individual preference. One might expect this to result in a deterioration of regionally coherent signal. However, we find broad patterns of local temperature–ring width and local precipitation–ring width relationships in spite of the different standardizations. Furthermore, many of the region-, species-, and frequency-dependent patterns that we describe depend on chronologies contributed by several different researchers. It is unlikely that such a diverse collection of standardization approaches applied to chronologies from different species in different places with different beginning and end dates would coherently contribute to or be responsible for the regional- and hemispheric-scale patterns that we describe.

Another reasonable concern relates to the generic reliability of our claims that ring width chronology–local climate relationships are region, species, and frequency dependent, given that these assertions are based as they are on results from subsets of the entire ITRDB. Figure 7 summarizes a nested analysis of correlations between ring width site chronologies as a function of geographic distance (i.e., correlograms) as an independent measure of the region, species, and frequency dependence that we ascribe to earlier results. Coherence over distance is one indicator of the potential for tree rings to capture climate signal.

Fig. 7.

(a) Correlations between the 427 temperature (brown dots and line) or precipitation (blue-gray dots and line) time series nearest the 762 ITRDB ring width site chronology time series (black dots and line) as a function of distance are shown. (b) Four regional subsets [Scandinavia (Scan), Labrador (Lab), eastern Mediterranean (Med), and the Pacific Northwest (PNW)] are subjectively and randomly chosen from the 762 chronologies to overlap the regions and species shown in Figs. 4 and 6 to be distributed around the hemisphere and to contain a reasonable number of chronologies from at least two species. (c) Ring width correlation vs distance within the four regions from (b) are shown; species-dependent correlation vs distance for the most prominent species within the four regions is shown for (d) North America and (e) Europe, as is (f) frequency-dependent correlation vs distance for a subsample of species from (e) that overlap with the regions and species shown in Fig. 6. Correlation coefficients in (c) through (f) and in the mean lines of (a) are averaged into 100-km bins.

Fig. 7.

(a) Correlations between the 427 temperature (brown dots and line) or precipitation (blue-gray dots and line) time series nearest the 762 ITRDB ring width site chronology time series (black dots and line) as a function of distance are shown. (b) Four regional subsets [Scandinavia (Scan), Labrador (Lab), eastern Mediterranean (Med), and the Pacific Northwest (PNW)] are subjectively and randomly chosen from the 762 chronologies to overlap the regions and species shown in Figs. 4 and 6 to be distributed around the hemisphere and to contain a reasonable number of chronologies from at least two species. (c) Ring width correlation vs distance within the four regions from (b) are shown; species-dependent correlation vs distance for the most prominent species within the four regions is shown for (d) North America and (e) Europe, as is (f) frequency-dependent correlation vs distance for a subsample of species from (e) that overlap with the regions and species shown in Fig. 6. Correlation coefficients in (c) through (f) and in the mean lines of (a) are averaged into 100-km bins.

Figure 7a shows that temperature time series are more coherent over distance than precipitation time series; it also shows that both temperature and precipitation are more coherent over distance than ring width time series. This is a well-known result, but it represents a basis for comparison with the other correlograms in Fig. 7.

Figure 7c shows that regional collections of ring widths typically exhibit correlograms with systematically larger or smaller coherence over distance than the correlogram developed from the hemispheric-scale dataset as a whole. In other words, ring width coherence over distance is regionally dependent and, because ring width coherence over distance is an indicator of the potential for capturing climate signal, the regionally dependent correlograms are also consistent with regionally dependent ring width–local climate relationships. That correlation magnitudes in regional and species-specific correlograms (Figs. 7d,e) are almost always equal to or higher than for the accompanying region-specific correlograms over the range of meaningful correlation magnitudes provides strong evidence that ring width coherence over distance is also species specific. The one notable exception involves Mediterranean ring widths of Norway spruce located between 100 and 200 km of each other (Fig. 7e), but this particular data point happens to be based on the correlation between only a single pair of ring width time series. Perhaps the most striking example of species-dependent coherence is in the PNW region, where species-specific correlograms exhibit substantially larger spatial coherence than the correlogram for the region as a whole. Including multiple species in a regional correlogram apparently diminishes coherence. Finally, Fig. 7f shows that correlograms based on high-pass filtered ring widths generally exhibit higher spatial coherence than low-pass filtered ring widths for the species and regions shown in Fig. 6, except possibly for the preferentially “low frequency” black pine ring widths.

The fact that standard ring width time series exhibit systematically stronger temporal (Fig. 1) and weaker spatial (Fig. 7) autocorrelation than the meteorological data suggests that nonclimatological growth history information remains after standardization. Separate analyses (not shown) were conducted on prewhitened (residual) ring width time series from the ITRDB and on synthetic one-year-differenced time series constructed from the standard ITRDB chronologies. Summer temperature correlation magnitudes were not noticeably larger for either the residual or the one-year-differenced time series than for the standard time series, as one might expect if the true physiological response were concentrated in one summer (i.e., in the concurrent ring width) and autocorrelation dominated the signal in the neighboring ring widths (i.e., the prior and subsequent ring widths). As noted previously by others (e.g., Fritts 1966, 1971; D’Arrigo et al. 1992; Littell et al. 2008), this suggests that ecophysiological responses such as ring width anomalies can result from climate (e.g., summer temperature) anomalies in both the year prior to and concurrent with ring formation. Therefore, prewhitened ring width time series or any tree-ring measurement associated solely with summer temperature in one or the other of the two years might have excluded useful climate information. Methods that consider concurrent and subsequent predictors are not new in dendroclimatology (e.g., Fritts 1976; Frank and Esper 2005a,b), but this study suggests that the manner in which lagged tree-ring measurements are treated in a reconstruction can have a strong influence on the results, especially in the presence of relatively strong temporal autocorrelation.

The spatial consistency of temperature and precipitation correlations confirms that there is indeed much coherent climate information in tree-ring widths. A qualitatively similar approach could be applied to other high resolution proxies that include enough overlap with the observed meteorological data to provide sufficient degrees of freedom (e.g., ice core isotopes, ice core chemistry, annually resolved sediment cores, etc.) to determine their intrinsic spatial coherence. Understanding the influence of spatial, species, and spectral variability in climate–ring width relationships contained within a particular reconstruction could reveal whether the influences of biological autocorrelation and lagged responses in ring widths enhance or obscure authentic climate–growth signals. In some cases, proxy selections might be suggested that could make a reconstruction more economical and/or accurate. For example, certain species or measurements (e.g., latewood density and/or ring width) could be selected as candidate predictors based on a priori climate–proxy relationships, depending on the reconstruction target’s seasonality, spectral characteristics, and/or the desired degrees of freedom required for calibration and verification.

Acknowledgments

We would especially like to thank three anonymous reviewers as well as Kevin Rennert, Anne Bjune, and Camille Li for very thoughtful and helpful suggestions on different aspects of this work. We would also very much like to thank the researchers who have contributed to the International Tree-Ring Data Bank, IGBP PAGES/World Data Center for Paleoclimatology, NOAA/NCDC Paleoclimatology Program; Boulder, Colorado. Unfortunately, it is impractical to individually cite all these researchers, given the number of relevant papers associated with the data we analyzed. This research was supported by the National Science Foundation through a Graduate Research Fellowship and under Grant ATM 0812802 and by the Research Council of Norway through the COMPAS project. This publication is partially funded by the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreements NA17RJ1232 and NA10OAR4320148. JL acknowledges support from NOAA SARP, Grant Number NA07OAR4310371 to N. Mantua, J. S. Littell, and A. F. Hamlet.

REFERENCES

REFERENCES
Biondi
,
F.
,
A.
Gershunov
, and
D. R.
Cayan
,
2001
:
North Pacific decadal climate variability since 1661
.
J. Climate
,
14
,
5
10
.
Briffa
,
K. R.
,
T. S.
Bartholin
,
D.
Eckstein
,
P. D.
Jones
,
W.
Karln
,
F. H.
Schweingruber
, and
P.
Zetterberg
,
1990
:
A 1400-year tree-ring record of summer temperatures in Fennoscandia
.
Nature
,
346
,
434
439
.
Briffa
,
K. R.
,
P. D.
Jones
,
T. S.
Bartholin
,
D.
Eckstein
,
F. H.
Schweingruber
,
W.
Karlen
,
P.
Zetterberg
, and
M.
Eronen
,
1992
:
Fennoscandian summers from AD 500: Temperature changes on short and long timescales
.
Climate Dyn.
,
7
,
111
119
.
Briffa
,
K. R.
,
F. H.
Schweingruber
,
P. D.
Jones
,
T. J.
Osborn
,
S. G.
Shiyatov
, and
E. A.
Vaganov
,
1998
:
Reduced sensitivity of recent tree-growth to temperature at high northern latitudes
.
Nature
,
391
,
678
682
.
Briffa
,
K. R.
,
T. J.
Osborn
,
F. H.
Schweingruber
,
I. C.
Harris
,
P. D.
Jones
,
S. G.
Shiyatov
, and
E. A.
Vaganov
,
2001
:
Low-frequency temperature variations from a northern tree ring density network
.
J. Geophys. Res.
,
106
,
2929
2941
.
Briffa
,
K. R.
,
T. J.
Osborn
,
F. H.
Schweingruber
,
P. D.
Jones
,
S. G.
Shiyatov
, and
E. A.
Vaganov
,
2002a
:
Tree-ring width and density data around the Northern Hemisphere. Part 1: Local and regional climate signals
.
Holocene
,
12
,
737
757
.
Briffa
,
K. R.
,
T. J.
Osborn
,
F. H.
Schweingruber
,
P. D.
Jones
,
S. G.
Shiyatov
, and
E. A.
Vaganov
,
2002b
:
Tree-ring width and density data around the Northern Hemisphere. Part 2: Spatio-temporal variability and associated climate patterns
.
Holocene
,
12
,
759
789
.
Cook
,
E. R.
,
1985
:
A time series approach to tree-ring standardization
.
Ph.D. thesis, Univeristy of Arizona, 171 pp
.
Cook
,
E. R.
,
K. R.
Briffa
,
S.
Shiyatov
,
V.
Mazepa
, and
P. D.
Jones
,
1990
:
Data analysis
.
Methods of Dendroclimatology: Applications in the Environmental Sciences, E. Cook and L. Kairiukstis, Eds., Kluwer Academic Press, 97–162
.
Cook
,
E. R.
,
K. R.
Briffa
,
D. M.
Meko
,
D. A.
Graybill
, and
G.
Funkhouser
,
1995
:
The segment length curse in long tree-ring chronology development for paleoclimatic studies
.
Holocene
,
5
,
229
237
.
Cook
,
E. R.
,
R. D.
D’Arrigo
, and
K. R.
Briffa
,
1998
:
A reconstruction of the North Atlantic Oscillation using tree-ring chronologies from North America and Europe
.
Holocene
,
8
,
9
17
.
Cook
,
E. R.
,
D. M.
Meko
,
D. W.
Stahle
, and
M. K.
Cleaveland
,
1999
:
Drought reconstructions for the continental United States
.
J. Climate
,
12
,
1145
1162
.
Cook
,
E. R.
,
R. D.
D’Arrigo
, and
M. E.
Mann
,
2002
:
A well-verified, multiproxy reconstruction of the winter North Atlantic Oscillation index since A.D. 1400
.
J. Climate
,
15
,
1754
1764
.
Cook
,
E. R.
,
C. A.
Woodhouse
,
C. M.
Eakin
,
D. M.
Meko
, and
D. W.
Stahle
,
2004
:
Long-term aridity changes in the western United States
.
Science
,
306
,
1015
1018
.
Crowley
,
T. J.
, and
T. S.
Lowery
,
2000
:
How warm was the Medieval Warm Period?
Ambio
,
29
,
51
54
.
Cullen
,
H. M.
,
R. D.
D’Arrigo
,
E. R.
Cook
, and
M. E.
Mann
,
2001
:
Multiproxy reconstructions of the North Atlantic Oscillation
.
Paleoceanography
,
16
,
27
39
.
D’Arrigo
,
R. D.
,
G. C.
Jacoby
, and
R. M.
Free
,
1992
:
Tree-ring width and maximum latewood density at the North American tree line: Parameters of climatic change
.
Can. J. For. Res.
,
22
,
1290
1296
.
D’Arrigo
,
R. D.
,
R.
Villalba
, and
G.
Wiles
,
2001
:
Tree-ring estimates of Pacific decadal climate variability
.
Climate Dyn.
,
18
,
219
224
.
D’Arrigo
,
R. D.
,
E. R.
Cook
,
M. E.
Mann
, and
G. C.
Jacoby
,
2003
:
Tree-ring reconstructions of temperature and sea level pressure variability associated with the warm-season Arctic Oscillation since A.D. 1650
.
Geophys. Res. Lett.
,
30
,
1549
,
doi:10.1029/2003GL017250
.
Erlandsson
,
S.
,
1936
:
Dendroclimatological studies
.
Stockholm’s College Geochronological Institute Rep. 23, 16 pp
.
Esper
,
J.
,
E. R.
Cook
, and
F. H.
Schweingruber
,
2002
:
Low-frequency signals in long tree-ring chronologies for reconstructing past temperature variability
.
Science
,
295
,
2250
2253
.
Esper
,
J.
,
E. R.
Cook
,
P. J.
Krusic
,
K.
Peters
, and
F. H.
Schweingruber
,
2003
:
Tests of the RCS method for preserving low-frequency variability in long tree-ring chronologies
.
Tree-Ring Res.
,
59
,
81
98
.
Esper
,
J.
,
D. C.
Frank
, and
R. J. S.
Wilson
,
2004
:
Climate reconstructions: Low-frequency ambition and high-frequency ratification
.
Eos, Trans. Amer. Geophys. Union
,
85
,
12
,
doi:10.1029/2004EO120002
.
Esper
,
J.
,
P. J.
Krusic
,
K.
Peters
, and
D.
Frank
,
2008
:
Exploration of long-term growth changes using the tree-ring detrending program “Spotty”.
Dendrochronologia
,
27
,
75
82
.
Frank
,
D. C.
, and
J.
Esper
,
2005a
:
Characterization and climate response patterns of a high-elevation, multi-species tree-ring network in the European Alps
.
Dendrochronologia
,
22
,
107
121
.
Frank
,
D. C.
, and
J.
Esper
,
2005b
:
Temperature reconstructions and comparisons with instrumental data from a tree-ring network for the European Alps
.
Int. J. Climatol.
,
25
,
1437
1454
.
Fritts
,
H. C.
,
1966
:
Growth rings of trees: Their correlation with climate
.
Science
,
154
,
973
979
.
Fritts
,
H. C.
,
1971
:
Dendroclimatology and dendroecology
.
Quat. Res.
,
1
,
419
449
.
Fritts
,
H. C.
,
1976
:
Tree Rings and Climate
.
Academic Press, 567 pp
.
Gedalof
,
Z.
, and
D. J.
Smith
,
2001
:
Interdecadal climate variability and regime-scale shifts in Pacific North America
.
Geophys. Res. Lett.
,
28
,
1515
1518
.
Glueck
,
M. F.
, and
C. W.
Stockton
,
2001
:
Reconstruction of the North Atlantic Oscillation, 1429–1983
.
Int. J. Climatol.
,
21
,
1453
1465
.
Graumlich
,
L. J.
, and
L. B.
Brubaker
,
1986
:
Reconstruction of annual temperature (1590–1979) for Longmire Washington derived from tree rings
.
Quat. Res.
,
25
,
223
234
.
Grissino-Mayer
,
H. D.
, and
H. C.
Fritts
,
1997
:
The International Tree-Ring Data Bank: An enhanced global database serving the global scientific community
.
Holocene
,
7
,
235
238
.
Guiot
,
J.
,
1985
:
The extrapolation of recent climatological series with spectral canonical regression
.
Int. J. Climatol.
,
5
,
325
335
.
Guiot
,
J.
,
1986
:
ARMA techniques for modelling tree-ring response to climate and for reconstructing variations of paleoclimates
.
Ecol. Modell.
,
33
,
149
171
.
Holman
,
M. L.
, and
D. L.
Peterson
,
2006
:
Sensitivity of forest growth to climatic variability in the Olympic Mountains, Washington
.
Can. J. For. Res.
,
36
,
92
104
.
Huang
,
S.
,
2004
:
Merging information from different resources for new insights into climate change in the past and future
.
Geophys. Res. Lett.
,
31
,
L13205
,
doi:10.1029/2004GL019781
.
Jones
,
J. M.
, and
M.
Widmann
,
2003
:
Instrument- and tree-ring-based estimates of the Antarctic Oscillation
.
J. Climate
,
16
,
3511
3524
.
Jones
,
P. D.
, and
M. E.
Mann
,
2004
:
Climate over past millennia
.
Rev. Geophys.
,
42
,
RG2002
,
doi:10.1029/2003RG000143
.
Jones
,
P. D.
,
K. R.
Briffa
,
T. P.
Barnett
, and
S. F. B.
Tett
,
1998
:
High-resolution palaeoclimatic records for the last millennium: Interpretation, integration, and comparison with general circulation model control-run temperatures
.
Holocene
,
8
,
455
471
.
Jones
,
P. D.
, and
Coauthors
,
2009
:
High-resolution palaeoclimatology of the last millenium: A review of current status and future prospects
.
Holocene
,
19
,
3
49
.
Kipfmueller
,
K. F.
, and
M. W.
Salzer
,
2010
:
Linear trend and climate response of five-needle pines in the western United States related to treeline proximity
.
Can. J. For. Res.
,
40
,
131
142
.
Littell
,
J. S.
,
D. L.
Peterson
, and
M.
Tjoelker
,
2008
:
Douglas-fir growth in mountain ecosystems: Water limits tree growth from stand to region
.
Ecol. Monogr.
,
18
,
349
368
.
MacDonald
,
G. M.
, and
R. A.
Case
,
2005
:
Variations in the Pacific decadal oscillation over the past millennium
.
Geophys. Res. Lett.
,
32
,
L08703
,
doi:10.1029/2005GL022478
.
Mann
,
M. E.
, and
P. D.
Jones
,
2003
:
Global surface temperatures over the past two millennia
.
Geophys. Res. Lett.
,
30
,
1820
,
doi:10.1029/2003GL017814
.
Mann
,
M. E.
,
R. S.
Bradley
, and
M. K.
Hughes
,
1999
:
Northern Hemisphere temperatures during the past millennium: Inferences, uncertainties, and limitations
.
Geophys. Res. Lett.
,
26
,
759
762
.
Mann
,
M. E.
,
E.
Gille
,
R. S.
Bradley
,
M. K.
Hughes
,
J.
Overpeck
,
F. T.
Keimig
, and
W.
Gross
,
2000
:
Global temperature patterns in past centuries: An interactive presentation
.
Earth Interactions
,
4
.
[Available online at http://EarthInteractions.org.]
Meko
,
D. M.
,
1981
:
Applications of Box-Jenkins methods of time-series analysis to reconstruction of drought from tree rings
.
Ph.D. thesis, University of Arizona, 149 pp
.
Mitchell
,
T. D.
,
T. R.
Carter
,
P. D.
Jones
,
M.
Hulme
, and
M.
New
,
2004
:
A comprehensive set of high-resolution grids of monthly climate for Europe and the globe: The observed record (1901–2000) and 16 scenarios (2001–2100)
.
Tyndall Centre Tech. Rep. Working Paper 55, 30 pp
.
Moberg
,
A.
,
D. M.
Sonechkin
,
K.
Holmgren
,
N. M.
Datsenko
, and
W.
Karlen
,
2005
:
Highly variable Northern Hemisphere temperatures reconstructed from low- and high-resolution proxy data
.
Nature
,
433
,
613
617
.
Nakawatase
,
J. M.
, and
D. L.
Peterson
,
2006
:
Spatial and temporal variability in forest growth-climate relationships in the Olympic Mountains, Washington
.
Can. J. For. Res.
,
36
,
77
91
.
Osborn
,
T. J.
, and
K. R.
Briffa
,
2000
:
Revisiting timescale-dependent reconstruction of climate from tree-ring chronologies
.
Dendrochronologia
,
18
,
9
25
.
Pederson
,
G. T.
, and
Coauthors
,
2011
:
The unusual nature of recent snowpack declines in the North American Cordillera
.
Science
,
333
,
332
335
,
doi:10.1126/science.1201570
.
Peterson
,
D. W.
, and
D. L.
Peterson
,
2001
:
Mountain hemlock growth responds to climatic variability at annual and decadal scales
.
Ecology
,
82
,
3330
3345
.
Schmutz
,
C.
,
J.
Luterbacher
,
D.
Gyalistras
,
E.
Xoplaki
, and
H.
Wanner
,
2000
:
Can we trust proxy-based NAO index reconstructions?
Geophys. Res. Lett.
,
27
,
1135
1138
.
Schweingruber
,
F. H.
,
O. U.
Bräker
, and
E.
Schär
,
1987
:
Temperature information from a European dendroclimatological sampling network
.
Dendrochronologia
,
5
,
9
33
.
Stahle
,
D. W.
, and
Coauthors
,
1998
:
Experimental dendroclimatic reconstruction of the Southern Oscillation
.
Bull. Amer. Meteor. Soc.
,
79
,
2137
2152
.
Woodhouse
,
C. A.
,
1997
:
Tree-ring reconstructions of circulation indices
.
Climate Res.
,
8
,
117
127
.
Woodhouse
,
C. A.
,
2003
:
A 431-yr reconstruction of western Colorado snowpack from tree rings
.
J. Climate
,
16
,
1551
1561
.
Woodhouse
,
C. A.
, and
J. T.
Overpeck
,
1998
:
2000 years of drought variability in the central United States
.
Bull. Amer. Meteor. Soc.
,
79
,
2693
2714
.
Woodward
,
A.
,
D. G.
Silsbee
,
E. G.
Schreiner
, and
J. E.
Means
,
1994
:
Influence of climate on radial growth and cone production in subalpine fir (Abies lasiocarpa) and mountain hemlock (Tsuga mertensiana)
.
Can. J. For. Res.
,
24
,
1133
1143
.

Footnotes

*

Bjerknes Centre for Climate Research Publication Number A 332 and the Joint Institute for the Study of the Atmosphere and Ocean Contribution Number 1860.

1

The ITRDB is maintained by the Paleoclimatology Branch of the National Climatic Data Center of the National Oceanic and Atmospheric Administration.