Ulf Büntgen

Graphs

Location of 124 larch and spruce tree-ring sites in the European Alps, and the 13 instrumental stations used for comparison. Dav = Davos, Feu = Feuerkogel, Gsb = Gr. St. Bernhard, Jfj = Jungfraujoch, Lga = Lago Gabiet, Pat = Patscherkofel, Rol = Passo di Rolle, Smh = Schmittenhöhe, Seg = Sils/Segl-Maria, Snt = Säntis, Son = Sonnenblick, Via = Villacher Alpe, Zug = Zugspitze (Büntgen et al. 2008 GCB).

Regional curves of (a) 62 larch and (b) 62 spruce sites (grey), together with mean records integrating data from <1400 (blue), 1400-1900 (green) and >1900 (red) m asl. Insets show the relationship between average growth rate (AGR) and mean segment length (MSL) at the individual sites (Büntgen et al. 2008 GCB).

Maximum response of spline detrended (a) 62 larch and (b) 62 spruce site chronologies (ordered by elevation) to current year monthly and seasonal temperature targets computed over the early 1864-1933 period. Black dots indicate correlations >0.39. Histogram shows the amount of (c) larch and (d) spruce chronologies that have maximum correlation with certain temperature targets (6-8 = June-August) (Büntgen et al. 2008 GCB).

Scatter plots of early (1864-1933) versus late (1934-2003) period maximum growth/temperature response for (a) larch and (b) spruce. Results are shown for all (left), the sensitive (mid), and insensitive (right) sites, as introduced in figure 3 (Büntgen et al. 2008 GCB).

Mean (a) larch and (b) spruce RCS (black) and 300yr spline (grey) chronologies that integrate 40 larch and 24 spruce sites, respectively. Records are regressed (1864-2003) against June-July temperatures (red). Dots (black and grey) indicate 31yr-moving correlations between the chronologies (RCS and spline) and temperature. Bottom panels show residuals between the RCS records and instrumental measurements, and their 20yr low-pass filters (grey areas) (Büntgen et al. 2008 GCB).

Residuals (ordered by size) between larch and spruce chronologies (after RCS and 300yr spline detrending) and June-July temperatures (1864-2003). Orange and red bars indicate annual residuals of the late 20th century that occurred between 1960-79 and 1980-2003, respectively (Büntgen et al. 2008 GCB).

20yr low-pass filtered (a) larch and (b) spruce chronologies regressed (blue) or scaled (green) against June-July temperatures (red) over the 1864-2003 period (Büntgen et al. 2008 GCB).

Comparison between (Büntgen et al. 2008 GCB, green), the TRW-based (blue, Büntgen et al. 2005), and MXD-based (red, Büntgen et al. 2006) reconstructions of summer temperature in the greater Alpine region using 60yr low-pass filters. Data are expressed as anomalies wrt. 1901-2000.

Calibration and verification statistics of (a) larch and (b) spruce 300yr spline and RCS chronologies using regression (scaling). Correlations were computed using unfiltered (ori), and 10yr high- and low-pass filtered (hp and lp) time-series (Büntgen et al. 2008 GCB).

Characteristics of three MXD datasets used to reconstruct long-term regional-scale summer temperature variations (Büntgen et al. 2008 CD).

(A) Location of the newly developed GER and SOB chronologies in the Spanish Pyrenees (Büntgen et al. 2008 CD). Black star indicates the location of the high-elevation Pic du Midi instrumental station. (B) Location of the summer temperature reconstructions from the Pyrenees (PYR), Alps (VAL) and Scandinavia (TOR). (C) Location of currently available MXD chronologies likely suitable to reconstruct variations in summer temperatures back to at least AD 1200 (CAM, Szeicz and MacDonald 1995; TOR, Grudd 2006; POL, Briffa et al. 1995; JAM, Briffa et al. 1995; QUE, Wang et al. 2001; GOT, Esper et al. 2002; ICE, Luckman and Wilson 2005; LAU, Schweingruber et al. 1988; VAL, Büntgen et al. 2006b), and this study (PYR). The TOR, VAL and PYR chronologies all extend into the 21st century (Büntgen et al. 2008 CD).

The two timberline sites Gerber (GER) and Sobrestivo (SOB) within the Central Spanish Pyrenees, characterized by wide talus slopes and open pine forests (Büntgen et al. 2008 CD).

A Temporal distribution of 203 GER (red) and B 58 SOB (blue) core samples ordered by calendar age of their innermost ring. Note the reduction in sample size <5 series prior to AD 1260. Black dots show sample distribution if series were ordered by their outermost ring. C Mean cambial age of the GER and SOB samples for each calendar year. D Regional curves (RCs) of the age-aligned GER and SOB MXD (left) and TRW (right) series (truncated <20 series at the series outermost end) (Büntgen et al. 2008 CD).

A Comparison between the unfiltered GER (red) and SOB (blue) RCS MXD chronologies (truncated <5 series), and B their 20-year high-pass, and C 20-year low-pass bands. The unfiltered RCS MXD chronologies were scaled to the same mean and variance over the common 1517-2005 period (Büntgen et al. 2008 CD).

A EPS and Rbar statistics (calculated over 30 years lagged by 15 years) of B the PYR chronology (AD 1260-2005) without variance adjustment, and replication. The horizontal dashed line in A denotes the 0.85 EPS criterion for signal strength acceptance (Wigley et al. 1984). Grey shading in B denotes 95% bootstrap confidence levels of the smoothed RCS chronology. C The final PYR chronology after variance adjustment, and D moving 31-year standard deviations of the chronologies without variance adjustment (blue), using variance adjustment for changes in sample replication (black), and variance adjustment for changes in sample replication and Rbar (red). Smoothed curves in B and C are 20-year low-pass filters (Büntgen et al. 2008 CD).

Growth/climate response of the MXD (black), TRW (white), and their mean (grey) RCS chronologies using A maximum temperatures, B mean temperatures, C minimum temperatures, and D precipitation sums. Correlations are computed from previous year May to current year October over the common 1901-2002 period. Horizontal dashed lines denote the 99% significance levels, corrected for lag-1 autocorrelation. Temperature data are derived from the Pic du Midi station, and precipitation data from the CRUTS2.1 dataset, using the mean of 15 grid-boxes that cover 0-2°E and 42-43°N. Numbers on x-axis refer to March-May, May-September, June-September, July-September, May-August, June-August, July-August, June-July, and January-December, respectively (Büntgen et al. 2008 CD).

A Comparison between unfiltered and 15-year low-pass filtered May-September minimum, mean, and maximum temperatures from the Pic du Midi station (blue, green, and red), and mean temperatures from the corresponding CRUTS2.1 (Mitchell and Jones 2005) and HadCRUT3v (Brohan et al. 2006) grids (black and grey). B Comparison between the 15-year high-pass components. C 31-year moving inter-series correlations between the five individual series. Values are expressed as anomalies with respect to the 1961-90 period and records smoothed using cubic smoothing spline functions (Büntgen et al. 2008 CD).

A Simple scaling of the PYR chronology (black) against May-September maximum temperatures from the Pic du Midi station (red) over the 1944-2005 period, and B their residuals through time. C The 15-year low-passed, and D the 15-year high-passed components of the modeled (black) and measured (red) temperatures. E Moving 31-year correlations between the actual and modeled values without filtering (black), and 15-year high-pass filtering (grey). Temperatures are shown as anomalies with respect to the 1961-90 period (Büntgen et al. 2008 CD).

A Original reconstructions for Scandinavia, the Alps, and Pyrenees expressed as anomalies from the instrumental reference periods 1951-70, 1901-2000 and 1961-1990, respectively. Yellow and blue rectangles denote the ten warmest and coldest calendar decades of the common AD 1260-2000 period. B Moving 51-year correlations between the three reconstructions. C Reconstructions from the Pyrenees (red), Alps (green) and Scandinavia (blue) after 20-40-year band-pass filtering. Records were scaled to have the same mean and variance over their common period (Büntgen et al. 2008 CD).

Left hand side shows spatial correlations between the temperature reconstructions from A Torneträsk (northern Sweden), B Valais (Swiss Alps), and C Pyrenees (Central Spanish Pyrenees), and the gridded 5°x5° HadCRUT3v dataset of monthly surface temperatures (Brohan et al. 2006). Right hand site shows the corresponding results for the meteorological records used for calibration: A Northern Sweden, B high-elevation Alps, and C Pic du Midi. Black stars indicate the locations of the records. Insets denote reconstructed (left) and measured (right) time-series utilized for correlation over the common 1850-2003 and 1882-2003 period, respectively (Büntgen et al. 2008 CD).

Composite analysis between the new regional-scale summer temperature history of the Pyrenees and the European-scale reconstruction of 500 hPa geopotential height by Luterbacher et al. (2002), using the 34 (A) coldest and (B) warmest years (1659-1999). The black star indicates the location of this study, and the black lines are bootstrap significance levels (Büntgen et al. 2008 CD).

Comparison between 20-year low-pass filtered A regional and B Northern Hemisphere temperature reconstructions scaled to have the same mean and variance over the AD 1260-2003, and 1000-1979 period, respectively. C Correlations (1260-1979) between the unfiltered and smoothed regional- and large-scale records (Büntgen et al. 2008 CD).

Location of TRW sites (triangles) and instrumental stations (circles) used in this study (Büntgen et al. 2008 TRACE). While each triangle represents one beech stand, the larger VOS triangle integrates samples from various stand locations along the Central Vosges Mountains.

Characteristics of site chronologies (Büntgen et al. 2008 TRACE) after spline detrending. Loc=Location (Lat/Lon), Ele=Elevation (m asl), Pep=Replication (Series), Per=Period, Per >5=Period >5 series, MSL=Mean Segment Length (Years), AGR=Average Growth Rate (mm/year), Lag-1=autocorrelation at year one. Bold chronologies were considered in the mean Swiss-French record.

Characteristics of instrumental station data used in  Büntgen et al. (2008 TRACE). Ele=Elevation (m asl), T-record=Period covered by temperature measurements (monthly), P-record=Period covered by precipitation measurements (monthly).

Summer (June-August) instrumental temperature data expressed as anomalies with respect to 1971-2000. Grand average correlation between the 10 stations for the annual, winter and summer data is 0.88, 0.93 and 0.94, respectively. Inter-station correlations range from 0.53 (Cht/Str) to 0.97 (Bas/Kar) for annual, from 0.81 (Chat/Str) to 0.99 (Kar/Str) for winter and from 0.88 (Nan/Nch) to 0.97 (Bas/Cht) for summer. Lag-1 autocorrelation of the mean summer (June-August) temperature record is 0.20. Smoothed mean curve is a 20yr low-pass filter (Büntgen et al. 2008 TRACE).

Characteristics of the (a) mean Swiss-French beech chronology and their (b) RBAR, (c) EPS and (d) replication. RBAR and EPS values were calculated over 30yr windows lagged by 20 years along the chronology. Lag-1 autocorrelation of the record is 0.24 (Büntgen et al. 2008 TRACE).

Comparison between the mean Swiss-French beech chronology and (a) monthly temperature means and (b) monthly precipitation sums from previous year April to current year September. Correlations were computed over the full 1806-2003 and four split periods of equal length. Significant levels are not provided as lag-1 autocorrelation varies between each monthly climate target and the different periods considered (Büntgen et al. 2008 TRACE).

Calibration/verification statistics of the 150yr spline TRW chronology and June-August scPDSI data. RE, and CE statistics refer to the regression model. Methodological explanation of the DW, RE, and CE values is given in the text (Büntgen et al. 2009 CD).

Central European hydro-climatic series used in Büntgen et al. (2009 CD). Lag-1 indicates 1st order autocorrelation (1744-1978). Correlations were computed against the JJA mean of the nearest 0.5° CRU grid box (1901-1978). Stars denote reconstructions that include instrumental data. Average correlation of the seven JJA precipitation grid boxes is 0.56 ranging from 0.31 (PFI/WMS) to 0.85 (BRA/WMS). Average correlation of the seven JJA scPDSI grid boxes is 0.55 ranging from 0.14 (TST/GLA) to 0.77 (BRA/WMS).

The 20 most negative/positive TRW-based JJA scPDSI values (1744-2006) for Slovakia in comparison with other related sources (seasonal precipitation totals during the instrumental period are expressed in % of the corresponding 1901-2000 means) (Büntgen et al. 2009 CD).

Temporal distribution of the Slovakian TRW compilation with grey bars representing the 86 individual measurement series and black bars indicating the corresponding pith-offset estimations (Büntgen et al. 2009 CD).

Regional-scale variability of June-August (A) temperature, (B) precipitation, and (C) scPDSI anomalies with respect to the 1961-1990 mean. Grey shadings denote decadal-scale variability after 10yr low-pass filtering. Inter-parameter correlations were computed over 1901-2002 and brackets refer to the low-passed time-series. Right-hand side shows autocorrelation structure of the target temperature, precipitation and scPDSI data (black) compared to the proxy TRW chronology (red) computed over the full (bold), and two early/late split periods (dashed/thin) (Büntgen et al. 2009 CD).

(A) 31yr moving STDEV of (B) the 150yr spline chronology (black) scaled over 1901-2002 to June-August scPDSI data (grey), their 10yr low-pass filters, and (C) 31yr moving correlations. Droughts are expressed as anomalies with respect to the 1961-1990 mean. Circles denote those years that range within the ten most extremes common to proxy and target data (1901-2002). Reconstructed driest and wettest years are 1947 and 1970, whereas the measured extremes are 1987 and 1916, respectively. (D) The new JJA scPDSI estimate (black plus 20yr low-pass filter) back to 1744. Rbar and EPS statistics additionally indicate the robust signal strength of the TRW-based chronology (Büntgen et al. 2009 CD).

(A) MTM spectra using 2yr resolution/ 3 tapers of the 150yr spline chronology (black) scaled against JJA scPDSI data (grey) calculated over 1901-2002, and (B) the new drought reconstruction (1744-2006). Thin lines are 90% confidence limits. (C) Wavelet (Morlet 6.0/6) power spectrum of the reconstruction (1744-2006), with contour levels chosen to be at 75%, 50%, 25%, and 5% of the wavelet power above each level. Black contour is the 10% significance level using a red-noise (autoregressive lag-1) background spectrum (Büntgen et al. 2009 CD).

(A) Spatial field correlation of the new Slovakian record (black star) against JJA drought data (0.5 x 0.5° scPDSI grid-boxes) computed over the 1901-1978 period. Circles indicate locations of the ten hydro-climatic records used for comparison. Dashed lines indicate studies of less defined regional-scale. (B) Spatial field correlation of the Tatra summer temperature reconstruction (white star; Büntgen et al. 2007) using 0.5 x 0.5° temperature grid-boxes and the 1901-1978 period (Büntgen et al. 2009 CD).

Composite 500 hPa geopotential height (gpm) for (A) the 26 wettest and (B) the 26 driest summers (1744-1999) of the Slovakian JJA scPDSI reconstruction (black star) (Büntgen et al. 2009 CD).

Fluctuation of Central European hydro-climatic records scaled to have a mean of zero and variance of one over 1700 to present, with the bold lines being 20yr low-pass filters, and the grey shading indicating the 1880s that are characterized by reduced coherency between this study and the other records. Bottom part indicates 31yr moving correlations (1744-1978) between this study (TST) and the nine hydro-climatic records used for comparison (green line represents their mean). Inter-series correlation (Rbar) of the nine hydro-climatic records computed over 31yr moving windows excluding the new Slovakian record (blue line) (Büntgen et al. 2009 CD).

(A) Standard deviations of the individual (standardized) Central European hydro-climatic records (as introduced in Table 3 and Figure 7) calculated over 61yr moving windows stress increasing variability back in time. Red and black curves are simple averages of the three orange and seven grey lines, respectively. (B) Power spectra (Morlet 6.0/6 global wavelets) of the Central-European hydro-climatic records computed over the 1744-1978 period stress decreasing variability at lower frequencies. Dashed lines are 10% significance levels using red-noise (autoregressive lag-1) background spectra (Büntgen et al. 2009 CD).

Temporal distribution of the 2610 living and relict Alpine larch sample used for temperature reconstruction (Büntgen et al. 2009 TRACE).

(A) Six chronologies (grey) after various detrendings and their mean (black). The black solid line shows chronology replication ranging from 2-6. (B) Moving 31-year inter-chronologies correlation (Rbar) and standard deviation (STDEV) of the six unfiltered chronologies. (C) Chronologies after 60-year low-pass filtering (Büntgen et al. 2009 TRACE).

(A) Overlap between the instrumental (grey) and reconstructed (black) temperatures, with the solid line indicating 31-year moving correlations between both records. Temperatures are expressed as anomalies with respect to the 20th century. The grey shading indicates offset between warmer instrumental measurements and cooler proxy estimates. (B) The June-July temperature reconstruction after regressing the mean chronology over the 1864-2003 period. Inset denotes calibration and verification statistics. Series are 20-year low-pass filtered, and grey boxes show the 10 warmest and coldest calendar decades (Büntgen et al. 2009 TRACE).

MTM power spectra of the (A) reconstructed (black) and measured (grey) June-July temperatures (1864-2003) using 3 tapers and a resolution of two years with robust background noise estimation. Smoothed lines are 95% confidence limits. The upper right inset denotes the corresponding global wavelet power spectra, with the dashed lines being 90% significance levels. (B) Global wavelet power spectra of the TRW-based reconstruction computed over two early/late split periods. Dashed lines are 90% significance levels using a white noise background (Büntgen et al. 2009 TRACE).

1 Location of host larch and non-host spruce TRW chronologies in the European Alps and Tatra Mts. classified according to elevation (<1400, 1400-1900, >1900 m asl) and tree-ring measurements (TRW and MXD). Grey inset frame denotes the greater Alpine region (43.0° - 48.5° N and 4.0° - 16.5° E) covered by the 0.5°x0.5° climate grid (Casty et al., 2005). Shaded insets indicate the five geographical sub-regions used for comparison with the corresponding tree-ring data (S Alps, 12 grids 43.5° - 45.5° N and 6.0° - 7.5° E, 4 host-sites; W Alps, 9 grids 45.5° - 48.0° N and 7.0° - 8.5° E, 13 host-sites; C Alps, 35 grids 45.5° - 48.0° N and 9.0° - 12.5° E, 23 host-sites; E Alps, 40 grids 45.5° - 48.0° N and 12.5° - 16.5° E, 27 host-sites; Tatra, 16 grids 48.0° - 50.0° N and 19.0° - 21.0° E, 3 host-sites) (Büntgen et al. 2009 NP).

(a) Distribution of 70 larch sites relative to elevation (<1400, 1400-1900, >1900 m asl) and parameter (TRW and MXD). Green color refers to site chronologies from which a distinct ~8 year cycle was detected along the reconstructed LBM outbreak time-series (see also supplementary table S1). (b) Temporal extent of the original (grey) and truncated (<5 series, black) chronologies. Orange (original) and red (truncated <5 series) curves describe density changes as a function of site-to-site distance (Büntgen et al. 2009 NP).

Maximum correlation with temperature of the host chronologies ordered by elevation and parameter, with colors referring to the five sub-regions. Correlations are computed over the 1901-2000 or maximum period of overlap. See supplementary Table S1 for details on the individual chronologies (Büntgen et al. 2009 NP).

Summary of the six (i-vi) detection methods (DM) applied. Black lines indicate the cumulative percentage of the detected outbreaks per year. The sum of all six methods (i-vi) shows outbreak evidence ranging from 1-98% in 295 years, whereas 0% outbreak evidence is found in 6 years, and the maximum of 98% outbreak evidence is reported for AD 1881. Maximum outbreak evidence as defined by method six (vi) reaches 24% in AD 1794, whereas 0% of evidence for this highest outbreak class (vi) is found for 191 years between 1700-2000 (Büntgen et al. 2009 NP).

Time-series of reconstructed LBM outbreaks (the cumulative percentage of the detected outbreaks per year following six intensity-levels) over the 1700-2000 period and split into five geographical sub-regions: (a-e) south, west, central, east >1500 and <1500 m asl. (f) Site replication per sub-region, and (g) outbreak patterns averaged over the entire Alpine arc. The six different colors refer to the six different outbreak intensities (ranging from low to high) that are based on the six detection methods (i-vi) as detailed in figure 4 (Büntgen et al. 2009 NP).

Comparison between three cycles of reconstructed LBM outbreaks (This Study), counted larvae population density, and discolored forest area (both from Baltensweiler & Rubli, 1999). Data are averaged over the entire Alpine arc and shown over their common period 1960-1990 (Büntgen et al. 2009 NP).

Wavelet power spectra of Alpine-wide reconstructed outbreak time-series based upon (a) 110 events during which detection method vi (all routines were positive) indicated an outbreak for 1.5-24.0% of affected site chronologies, and (b) 295 events during which the sum of the six detection methods indicated an outbreak for 1.7-98.0% of affected site chronologies. Contour levels are chosen so that 75%, 50%, 25%, and 5% of the wavelet power is above each level, respectively. Black contour is the 10% significance level, using a white-noise background spectrum. Right side shows the corresponding global wavelet power spectra (black line). Dashed lines are significance assuming the same significance level and background spectrum as indicated above (Büntgen et al. 2009 NP).

Characteristics of 70 larch sites sorted by elevation, and statistics referring to 300-yr spline chronologies. Green shadings refer to sites from which a distinct ~8 year cycle was detected along the reconstructed LBM outbreak time-series (Büntgen et al. 2009 NP).

Annually resolved maps of the last Alpine-wide synchronized LBM outbreak event during 1981-1984. Thin black triangles show the existing site chronologies per year, and colors refer to the reconstructed outbreak intensity ranging from heavy (purple) to low (grey). The six different colors are based on the six detection methods (i-vi) as detailed in figure 4. Corresponding numbers in the lower right summarize each year’s data availability and outbreak intensity, with the bottom graphs describing these numbers (outbreak sum per intensity-level and cumulative percentage of the intensity-levels) over time. Annual maps for 1700-2000 are provided in supplementary figure S1 (Büntgen et al. 2009 NP).

Location of the 22 high, and 6 low elevation instrumental stations, and the tree-ring datasets used. Homogenized temperature records (>1,500 m asl): Dav Davos (1901), Feu Feuerkogel (1930), Gsb Gr. St. Bernhard (1818), Jfj Jungfraujoch (1933), Lga Lago Gabiet (1928), Pak Patscherkofel (1931), Smh Schmittenhöhe (1880), Snt Säntis (1864), Son Sonnenblick (1887), Via Villacher Alpe (1851), Zug Zugspitze (1901), and low elevation temperature records (<600 m asl, underlined): Aos Aosta (1841), Ber Bern (1864), Gnv Geneva (1760), Mil Milano (1763), Sio Sion (1864), Tor Torino (1760). Homogenized precipitation records (>1,300 m asl): Aro Arosa (1890), Bal Balme (1914), Bar Bardonecchia (1914), Cer Ceresole Reale (1927), For Formazza Ponte (1901), Kal Kals (1896), Mai Marienberg (1858), Nau Nauders (1896), Sam Samedan (1861), Stm Sta. Maria (1901), Zer Zermatt (1892). Parentheses indicate start years of instrumental measurements, ending in 2002. Tree-ring datasets (>1,500 m asl): LOE Lötschental, GOM Goms, ENG Engadine (Büntgen et al. 2006 Trees).

A Temporal distribution of the 208 recent and historic samples. B Individual 300yr splines fit to each series. C Age-aligned mean curve (RC, calculated after power-transformation), and their replication. D Residuals between the 300yr spline and RCS chronology (Büntgen et al. 2006 Trees).

Comparison between annual mean (black), minimum, and maximum (grey) A precipitation, B temperature, and C the RCS ring width records, with replication at the right hand axes. Means were 20yr and D 40yr low-pass filtered for trend illustration (Büntgen et al. 2006 Trees).

Monthly growth response to A temperature (1933-2002) and B precipitation (1927-2002) of the previous and current year (maximum common periods). Grey bars show correlations with individual stations, and black bars denote correlations with the mean of these stations. Horizontal lines indicate 90% significance levels, corrected for lag-1 autocorrelation (Büntgen et al. 2006 Trees).

Moving 51yr correlations between RCS detrended ring width data from A this study, B this study plus ITRDB, C Frank05 (30 spruce series) and D Frank05 (23 non-spruce series), and previous year (left) and current year (right) June (green), July (blue), and August (red) temperatures. Horizontal lines indicate 90% significance levels, corrected for lag-1 autocorrelation (Büntgen et al. 2006 Trees).

Moving 51yr correlations between RCS detrended ring width data from A this study, B this study plus ITRDB, C Frank05 (spruce) and D Frank05 (non-spruce), and previous year June (green), July (blue), August (red), and September (dark red) precipitation. Horizontal lines indicate 90% significance levels, corrected for lag-1 autocorrelation (Büntgen et al. 2006 Trees).

Comparison of the spruce RCS chronology (Büntgen et al. 2006 Trees) with regional-, and large-scale temperature reconstructions. A This study, Frank05, and Mai-June and June-August high elevation mean temperatures (Böhm et al. 2001), after 40yr low-pass filtering. B This study and temperature reconstructions for the Alps (Büntgen et al. 2005a, tree-rings), Northern Hemisphere (Mann et al. 1999, multi-proxy), Northern Hemisphere extra-tropics (Esper et al. 2002, tree-rings) and the latitudinal band north of 60°N (Briffa 2000, tree-rings). Mann99 is likely weighted towards annual, and Büntgen05, Esper02 and Briffa00 towards warm season temperatures. Records were normalized over the 1108-1980 period of overlap and C 40yr low-pass filtered. Inset table shows correlations computed before and after 1900, with values in parentheses deriving after 40yr low-pass filtering (Büntgen et al. 2006 Trees).