Tree-ring anatomy of Pinus cembra trees open new avenues for climate reconstructions in the European Alps

Tree rings form the backbone of high-resolution palaeoclimatology and represent one of the most frequently used proxy to reconstruct climate variability of the Common Era. In the European Alps, reconstructions were often based on tree-ring width (TRW) and maximum latewood density (MXD) series, with a focus on European larch. By contrast, only a very limited number of dendroclimatic studies exists for long-lived, multi-centennial Pinus cembra , despite the widespread occurrence of the species at treeline sites across the European Alps. This lack of reconstructions can be ascribed to the difficulties encountered in past studies in extracting a robust climate signal from TRW and MXD chronologies. In this study, we tested various wood anatomical parameters from P. cembra as proxies for the reconstruction of past air temperatures. To this end, we measured anatomical cell parameters and TRW of old-growth trees from the God da Tamangur forest stand, known for being the highest pure, and continuous P. cembra forest in Europe. We demonstrate that several wood anatomical parameters allow robust reconstruction of past temperature variability at annual to multidecadal timescales. Best results are obtained with maximum latewood radial cell wall thickness (CWTrad) measured at 40 µm radial band width. Over the 1920–2017 period, the CWTrad chronology explains 62% and more than 80% of interannual and decadal variability of air temperatures during a time window corresponding roughly with the growing season. These values exceed those found in past work on P. cembra and even exceed the values reported for MXD chronologies built with L. decidua and hitherto considered the gold standard for dendroclimatic reconstructions in the European Alps. The wood anatomical analysis of P. cembra records therefore unveils a dormant potential and opens new avenues for a species that has been considered unsuitable for climate reconstructions so far.


Introduction
Annually resolved and absolutely dated tree-ring chronologies form the backbone of climate reconstructions and represent the most important proxy archive to reconstruct climate variability over past centuries to millennia (Frank et al., 2010;Masson-Delmotte et al., 2013;Büntgen et al., 2009).
The tree-ring parameters that were most frequently and successfully used to study temperature variations at high latitudes and altitudes are tree-ring width (TRW) and maximum latewood density (MXD) (e.g., Esper et al., 2018).In the case of the European Alps, temperature reconstructions are relying mainly on a suite of multi-centennial TRW and MXD chronologies from conifers growing close to timberline (1600-2500 m asl) (Frank and Esper, 2005 a,b).Among the available tree species, longlived and temperature-sensitive European larch (Larix decidua Mill.) and Swiss stone pine (Pinus cembra L.) have the greatest potential for millennium-long reconstructions (see e.g., Corona et al., 2010a,b;Büntgen et al., 2005Büntgen et al., , 2006)).In addition, they often co-occur in the same environment (EUFORGEN, 2014).Yet, whereas L. decidua has been used extensively for climate reconstructions (Büntgen et al., 2005(Büntgen et al., , 2006)), only very few dendroclimatic studies were based on P. cembra, despite the availability of numerous (n=120) multi-centennial (45% and 15% exceed 330 years and 500 years, respectively, Fig. 1A) and high-elevation (80% of the existing tree-ring records have been sampled above 2000 m asl; Fig. 1B) chronologies.This imbalance is related to difficulties in extracting a robust signal from P. cembra.ring width series.
Based on a survey of published dendroclimatic studies, TRW chronologies of P. cembra tend to contain a summer temperature (often June -August, JJA) signal, but with correlations that only rarely exceed r=0.4.They also tend to record a previous-year (n-1)mostly autumnprecipitation or temperature signal (Table 1).The imbalance holds true also for the density signal.The thin latewoodcharacteristic of the specieshas been reported repeatedly to limit its suitability for climatic J o u r n a l P r e -p r o o f Journal Pre-proof reconstructions based on MXD (Schweingruber and Johnson 1993;Schweingruber 1985), a finding that was confirmed more recently by Frank et al. (2005), showing that correlations between MXD and climate were consistently lower than those found in other widespread Alpine conifer species.
Quantitative wood anatomy (QWA)defined as the analysis of the variability of xylem anatomical features along dated tree-ring seriescould alleviate this problem (Von Arx et al., 2016;Björklund et al., 2019) as it operates at the cellular level, analyzes a wide number of anatomical parameters (i.e.lumen area, lumen diameter, cell wall-thickness) and showcases a strong summer temperature signals of latewood cell wall thickness (Carrer et al., 2017;Björklund et al., 2020;Seftigen et al., 2022).Following studies by Carrer et al. (2018) in the Italian Alps and Stirbu et al. (2022) in the Carpathians, showing significant correlation between cell wall thickness and summer temperatures, we hypothesize that QWA could offer an alternative to MXD series to trace past temperature variability from P. cembra trees.
In this study, we therefore employ QWA with the objective to test the suitability of P. cembra series for climate reconstruction.Using increment cores from trees growing in the highest continuous and one of the oldest P. cembra forests of the European Alps, we (i) construct chronologies for four different wood anatomical parameters (i.e.tangential and radial cell wall thickness, cell lumen area and anatomical MXD); (ii) quantify climate-tree growth relations by contrasting daily meteorological records from the E-OBS dataset and the QWA chronologies; so as to (iii) test correlation robustness and suitability of P. cembra QWA series with a calibration-verification procedure.

Study site
Our study site is located in God da Tamangur (46.68 °N; 10.36 °E)meaning the « forest back there» in Vallader Romanschknown as the highest, pure, and continuous P. cembra forest in Europe (Fig. 1A, C, D).The forest is located at an elevation of up to 2,300 m asl, at the furthest end of the Val S-charl, south of Scuol, Grisons, Switzerland, on the NW-facing slope of Piz Starlex (3,075 m asl).
Due to its remote location and high elevation, the forest has been spared from massive logging which was quite widespread in the valley because of intense mining activities.To the Romansch culture and language, the Tamangur forest has become a symbol of tenacity, strength and also symbolizes the will to survive.The forest nature reserve God da Tamangur stretches across 86 ha (Fig. 1C).Annual mean

J o u r n a l P r e -p r o o f
Journal Pre-proof temperature and precipitation sums measured at the Buffalora meteorological station (1,968 m asl), located 7.6 km SW from our study site, amount to 5. 75 °C and 677 mm, respectively (Meteo Swiss, 2022).

Sample collection and wood processing
Samples were collected during two field campaigns in summer 2017 and 2018.During the first campaign and with the purpose to build a robust TRW chronology, we sampled 45 long-lived P. cembra trees, growing between 2100 and 2300 m asl, without any visible faults such as nodes, reaction wood, or rotten parts.From each tree we extracted two increment cores at breast height (at ca.130 cm above ground).Ring widths were measured to the nearest 0.01 mm using TSAPWin (Rinntech, Germany), crossdated using standard dendrochronological procedures (Stokes and Smiley 1968) and checked for dating and measurement errors with the COFECHA software (Holmes 1983).
Based on the preliminary analyses of the tree-ring width series, we selected 9 individual trees showing the best inter-series correlation and covering the period 1920-2017.To perform wood cell anatomical measurements, these trees were sampled in summer 2018 using a 12 mm increment borer.The 12 mm cores were split into 4-5 cm long pieces to obtain 15-micron-thick cross-section with a Rotary Microtome (Leica RM 2255/2245).The sections were stained with Safranin and Astra blue to increase contrast and fixed with Canada balsam following standard protocols (Gärtner & Schweingruber, 2013;von Arx et al., 2016).Digital images of the microsections -at a resolution of 2.27 pixels/μm -were produced at the Swiss Federal Research Institute WSL (Birmensdorf, Switzerland), using a Zeiss Axio Scan Z1 (Carl Zeiss AG, Germany).

Wood anatomical measurements and chronology statistics
For the 9 selected trees, we used the ROXAS (v3.1) image analysis software (von Arx and Carrer, 2014) to automatically detect anatomical structures for all tracheid cells and annual ring boundaries for the period 1920-2017.We excluded measurements of samples with cell walls damaged during sampling or preparation.The output was summarized for four cell anatomical parameters (Fig. 2), including cell lumen area (LA), cell wall thickness oriented in radial (CWT rad ) and tangential (CWT tan ) directions (Prendin et al., 2017;Von Arx and Carrer, 2014) as well as anatomical maximum latewood density (aMXD) representing the maximum ratio between the cell wall area and the total cell area (the sum of cell wall and cell lumen area) throughout the ring (Björklund et al., 2020;Edwards et al., 2022).

J o u r n a l P r e -p r o o f
Journal Pre-proof In addition, and with the goal to compare our findings with conventional tree-ring widths analyses, we also established a ring width (TRW) chronology.
For each cell for which measurements were performed, we recorded the absolute and relative positions within each dated tree ring (Prendin et al., 2017;von Arx and Carrer, 2014).Following Bjorklund et al. (2020), we assigned each cell to tangential bands varying between 10 and 100 µm in radial width (with distances measured parallel to ring boundaries).In addition, we determined the transition from earlywood to latewood cells for each ring using Mork's index = 1 at a 10 μ radial resolution (Denne, 1988).For each ring, maximum values of CWT tan , CWT rad and aMXD were extracted from the bands identified as belonging to the latewood.For LA, maximum values were extracted from each ring irrespective of their position within earlywood or latewood.
The conventional ring-width measurements were detrended with a negative exponential function so as to eliminate non-climatic (e.g., age-related growth trends and other biological disturbances) effects from the series (Fritts 1976;Cook and Kairiukstis,1990).The detrended series were then aggregated into a TRW chronology using a biweight robust mean which reduces the influence of outliers (Cook and Peters, 1981).Given the absence of any evident long-term ontogenetic trend in anatomical series, detrending is not normally considered necessary (Carrer et al., 2018) in QWA studies.Yet, to increase interannual climatic signal and to remove multi-decadal trends in wood anatomical and meteorological time series, we standardized time series using a 30-year cubic smoothing spline, in addition to raw series.In a next step, empirical measures of dendroclimatic signals (Hughes et al., 2011) -average inter-series correlation (RBAR EFF ) and expressed population signal (EPS) (Wigley et al., 1984) -were computed to test the strength of the environmental information embedded in the chronologies using the maximum overlap of pairwise correlations (Bunn et al., 2014).All analyses were performed in R Studio (R Studio Team, 2020) using the R package dplR (Bunn, 2008;Bunn et al., 2014).

Assessing climate influence on tree-ring width and wood anatomical parameters
QWA data were correlated with the gridded (0.1° × 0.1° lat/long) E-OBS daily temperature dataset (Cornes et al., 2018).We selected the four grid-points centered over the sampling site (46.6-46.7°N; 10.3-10.4°E)and retrieved data for the period 1920-2017.Monthly temperatures from the E-OBS dataset are significantly correlated (r>0.87,p<0.01) with the meteorological series of Buffalora over the 1920-2017 period.For the growing season (April-September; or A-S) and summer (June-August; JJA) period, both meteorological series are correlated at 0.95 (p<0.01).

J o u r n a l P r e -p r o o f Journal Pre-proof
To test for the robustness of the newly developed QWA series for dendroclimatic reconstructions, records (TRW and QWA) were scaled against the E-OBS dataset.We preferred scaling over regression to avoid artificial variance reduction (Esper et al., 2005).In addition and to go beyond the conventional monthly climate data aggregation (Fonti et al., 2010) and striving for a higher detail in the resolution of the most important associations, we calibrated regression models on temperatures averaged over one-to-365-days windows starting on January 1 of the year preceding ring formation (n-1) and ending on December 31 of the year in which the ring was formed (n) using the R package dendroTools (Jevsenak, 2022).For each P. cembra parameter and optimal temperature time window identified from climate-growth relationships (see above), we computed 10,000 temperature reconstructions using a split calibration-verification procedure coupled with a bootstrap approach in which 50% of the years covered by both the meteorological and tree-ring datasets were randomly extracted for calibration while the remaining years were used for validation over the 1920-2017 period.
For each sampling, the root mean-square error (RMSE), coefficient of determination (r 2 for the calibration and R 2 for the verification periods), reduction of error (RE) and coefficient of efficiency (CE) statistics (Cook et al., 1994) were applied to test the predictive capacity of the transfer function.
Calibration and validation statistics are illustrated for each parameter with their 5 th , 50 th and 95 th percentiles.

Chronology characteristics
We measured wood anatomical features for the period 1920-2017 CE on all samples for a total of 75-100 radial files per ring, with anatomical information catalogued by its position in each dated tree ring.
After the exclusion of cells with walls damaged during sampling or preparation, a total of 1,393,924 tracheid cells were measured over the period 1920-2017.Journal Pre-proof reduction at the end of the ring.We also observe that the peak of CWT rad and aMXD is much more prominent than CWT tan and best captured in radial ba d w d h a g g f 10 40 μ .
The statistical characteristics of the Tamangur chronologies are summarized in Table 2.The EPS and Rbar values show that TRW has a stronger common signal (Rbar=0.39,EPS=0.85)than the wood anatomical chronologies (LA, CWT tan , CWT Trad and aMXD), with the Rbar for the latter ranging between 0.10 (f LA a 10 μ radial band width) and 0.34 (for a CWT Trad a 30 μ radial band width; Table 2).The QWA parameter correlating best with TRW over the period 1920-2017 is LA with correlations ranging from 0.37 (for a radial band w d h f 20 μ ) 0.44 (f a radial band width of 10 μ ).Correlations between TRW and CWT tan , CWT rad and aMXD are only slightly lower with a range going from 0.31 (for aMXD with a radial band w d h f 10 μ ) 0.41 (for CWT rad with a radial band w d h f 30 μ ).

Climate-growth relationships
Correlation coefficients between different tree-ring width and wood anatomical parameters with daily climate data retrieved from E-OBS over the period 1920-2017 point to temperature as the main driver of growth, yielding positive and significant values for both TRW and QWA parameters.By contrast, much weaker, mostly non-significant correlations can be found with precipitation (not shown).
Table 2 provides key insights into the time windows for which temperatures can best explain P. for the start and September 7 to 9 for the end dates) irrespective of the radial band width selected for analysis.
In the case of detrended time series, correlations are systematically lower compared to those obtained for raw data.In the case of TRW, for which only a detrended chronology could be analyzed, the highest correlation with temperature is obtained for a fairly long time window of 118 days comprised between March 8 and July 4 (r=0.57;p<0.05).In the case of CWT rad , detrending results in lower, yet very significant correlations between temperatures and different CWT rad chronologies with values ranging between r=0.57(p<0.05; 10 μ radial band w d h) a d 0.66 (p<0.05;40 μ radial band width).1945-1954 and 1963-1972 by +0.8 and +0.85°C, respectively.At the same time, it fails to reproduce the anthropogenic warming as observed in the instrumental records since at least the mid-1990s.

Statistical quality of the chronologies
In this study, we use P. cembra tree-ring and wood anatomical parameters to test the suitability of the species for robust temperature reconstructions.By contrast to previous work, where the potential of this very widespread, timberline-forming species has been considered to be too weak (Schweingruber, 1985), we demonstrate that wood anatomical parameters, and especially radial cell wall thickness (CWT rad ), can indeed serve as very promising and accurate proxies to reconstruct past summer air temperatures.These findings do not only bring P. cembra back to the spotlight of dendroclimatic research, but also opens new perspective in paleoclimatology as the species is very long-lived, abundant at high elevations (Carrer et al., 2018, Fig. 1A) and was sampled extensively in the past with more than 120 chronologies existing to date.
Similarly, we call for more wood anatomical chronologies of P. cembra to be built as they represent suitable proxies for climate reconstruction in the Alps provided that (1) correlations among series is high (Fonti and García-Gonzalez (2008), (2) the climatic fingerprint in series of anatomical parameters

J o u r n a l P r e -p r o o f
Journal Pre-proof is stronger and less noisy than the signal found in TRW (García-Gonzalez et al., 2016) and ( 3) that the wood anatomical reconstructions pass the calibration/verification statistics.In this work, and as frequently reported in other studies (Balanzategui et al., 2021;Ziaco et al., 2016), inter-series correlation for the wood anatomical parameters are lower than those found for TRW.They exceed the values reported by Carrer et al. (2016) for spruce (0.13) and larch (0.18) trees in the Alps and are comparable to those (0.23 for CWT tan , 0.24 for CWT rad ) reported for P. cembra in the Carpathians (Stirbu et al., 2022).The weaker empirical signal strength that we observe in chronologies built from CWT tan , CWT rad and aMXD could be explained by the heterogeneity induced by intra-annual biological processes controlling carbon assimilation and allocation in tree rings (Eilmann et al., 2006;Fonti and García-Gonzalez, 2004).

Suitability of wood anatomical parameter series for climate reconstruction in the Alps
Our study confirms the limited suitability of TRW chronologies of P. cembra for climate reconstructions (Schweingruber, 1985) and corroborates that one should refrain from reconstructing climate using P.
cembra TRW chronologies (Fig. 4).By contrast, we find highly significant, positive correlations between CWT rad (with a radial band w d h f 40 μ ) and mean air temperatures.Interestingly, the period for which the highest correlations are obtained matches closely with the assumed length of the vegetation period at the study site (end of April to early-September; Fig. 5; Gruber et al., 2018).Our pre-1950 version) and is not updated regularly.The authors of the database state themselves that quality control has been minimal for the pre-1950 period and that station density is low, such that erroneous values cannot be excluded for data-poor areas like the Alps (Cornes, et al., 2018).This may also explain to a certain extent the discrepancies found between the observed and reconstructed air temperatures, especially the systematic overestimation of reconstructed temperatures between the 1920s and the late 1930s, as well as the extremes observed in 1932, 1933, and 1939.

Wood anatomical reconstructions of P. cembraa new gold standard in Alpine dendroclimatology?
A major advantage of wood anatomical reconstructions based on long-lived P. cembra trees over classic TRW or MXD reconstructions is the fact that they are built from raw data, that is without undergoing any standardization procedures.This is in contrast to most reconstructions where standard individual-series detrending is applied to remove age trends (Cook et al., 1990;Lindholm and Eronen, 2000;Grudd et al., 2002).Yet, such a removal can be problematic as it limits the conservation of climatic information over periods equivalent to the mean segment length, a phenomenon described a h ' g g h cu ' ( k a .,1995).Wh a app ach c ud g h w d y p y d Regional Curve Standardization (Briffa et al., 1998;Esper et al., 2003;Briffa and Melvin, 2011;Melvin and Briffa, 2014) may offer significant improvements over traditional detrending methods, this may be limited to datasets with high sample replication and heterogeneous age structure (Gagen et al., 2007).
The fact that statistical detrending can be avoided is a great advantage as it ultimately allows reconstruction of low-frequency, long-term changes in past climate from trees (Carrer et al., 2017(Carrer et al., , 2018)).
Another advantage of wood anatomical reconstructions is the lack of any divergence in the record (Cerrato et al., 2019), by contrast to the TRW chronology constructed at God da Tamangur or various Journal Pre-proof anatomical parameters will likely fail to record the trajectory of summer temperatures in the future (IPCC, 2019).In addition of potentially losing its ability to accurately record extremely warm summer temperatures, P. cembra could also find itself in the climate trap as the species could face difficulties to keep the pace with climate when it comes to adaptation to climate change (Dauphin et al., 2021).
The findings reported in this study clearly showcase the large potential of P. cembra for dendroclimatic purposes.Last but not least, it seems worthwhile to mention that P. cembra is a non-host species for the larch budmoth (LBM; Baltensweiler, 1993;Nola et al., 2006) which affects L. decidua trees cyclically (each 8-9 years; Baltenschweiler et al., 1977;Esper et al., 2007) to defoliate their needles.
Insect defoliation thereby leads to markedly reduced ring increments and mass over two to five years (Kulman 1971;Battipaglia et al., 2014;Arbellay et al., 2018, Castagneri et al. 2020)  decidua in such a way that TRW chronologies have been considered unsuitable for high-frequency reconstructions (Weidner et al., 2010).Likewise, the very low ring increments observed during LBM outbreaks (Schweingruber, 1979) have often been removed in MXD chronologies and densities were estimated with a gap filling procedure (Büntgen et al., 2006), leaving, however, considerable uncertainties in reconstructions.P. cembra, by contrast, has been used as a reference species to disentangle the effect of insect defoliation on TRW or MXD in L. decidua from that of adverse climatic conditions (Dormont et al., 2006;Büntgen et al., 2009).Our study shows that in addition to its potential to allow disentangling non-climatic effects in L. decidua chronologies, wood anatomical parameters of P. cembra clearly have the potential to yield highly valuable and solid chronologies of their own at annual to decadal timescales and for most parts of the European Alps due to their ubiquitous occurrence at treeline sites.

Conclusions
The wood anatomical chronologies based on P. cembra traits unveil a dormant potential and open new avenues for a species that has hitherto been considered unsuitable for climate reconstructions.
By comparing wood anatomical parameters with daily temperature records, the wood anatomical approach also allows improving the quality of inferences compared to classical dendroclimatic approaches (TRW or MXD; Carrer et al., 2018), as the latter usually compare monthly temperature The construction of wood anatomical reconstructions has a cost and requires efforts that remain more substantial than what is currently needed for TRW or MXD reconstructions (Björklund et al., 2020).As the wood anatomical series of P. cembra represent a substantial of proxy dataand could not be obtained from ring widths or densitiesthe effort is nonetheless justified, especially as wood anatomical procedures are constantly improving and the need for fewer samples will likely render the approach more efficient and effective in the near future.

Figure 3
Figure 3 showcases the evolution of wood anatomical parameters as a function of relative distance to ring border.Based on Mork' d x, w b ha a w d p ugh y 10% f h g width on average.P. cembra trees feature the classical ring structure of conifers growing in cold, temperate environments, with a monotonic reduction in cell lumen area (LA) on the one hand and an increase in cell wall thickness (CWT) and anatomical maximum latewood density (aMXD) from earlywood to latewood on the other hand.These trends differ just at the start of a ring for LA, with an evident increase at the very first part, and for tangential and radial CWT in latewood, showing a cembra growth.All statistics provided in the following are for correlations between raw data of anatomical traits and temperature data.For LA, the best correlation (r=0.56;p<0.05) is found with temperatures from May 20 to September 6 (109 days) occurred in the year preceding ring formation (n-1) for a radial band w d h f 20 μ .d ably higher correlations are obtained for the various QWA chronologies, namely between CWT tan (10 μ radial band width) and aMXD (20 μ radial band width) with temperatures for the time windows comprised between May 14 and September 4 (113 days; r=0.71, p<0.05) and July 15 and August 30 (46 days; r=0.74, p<0.05), respectively.The highest correlations are obtained between the CWT rad chronology and temperatures over a 130-day time window centered between April 30 and September 7 (r=0.79;p<0.05).If the period considered is restricted to the period 1950-2017 for which the E-OBS network is considered very robust by the data providers, we even obtain a correlation of r=0.87 (p<0.05).Noteworthy, correlation between this last wood anatomical parameter and the time window of air temperatures considered varies only very little (i.e.comprised between r=0.76 to 0.79 and time windows fluctuating only between April 26 to May 2 J o u r n a l P r e -p r o o f Journal Pre-proof Temperature reconstructionBased on the above statistics, Fig.4presents the temperature reconstruction based on a conventional TRW chronology for which P. cembra has hitherto been shown to provide only poorly significant results.The graph shows reconstructed air temperatures from March 8 to July 4 computed as the median of 10000 realizations given with the 95% bootstrap confidence interval and compared to the E-OBS dataset over the period 1920-2017.At interannual timescales, the reconstruction underestimates temperatures at several instances, with the most notable differences observed in 2004 (underestimation of -2.85°C), 1975 (-1.78°C) and 1987 (-0.68°C).By contrast, the reconstruction also overestimates air temperatures in a considerable manner, namely in 1949 (overestimation of +2.4°C) and 1971 (+2.36°C).Over the period 1920-2017, the RMSE computed between the observed and TRW reconstructed air temperatures is 0.84.At decadal timescales, the adjusted coefficient of J o u r n a l P r e -p r o o f Journal Pre-proof determination between the 11-year spline smoothed observed and reconstructed temperatures is r(adj) 2 =0.55, yet the TRW-based reconstruction overestimates air temperatures over the periods

Fig. 5
Fig.5provides the best reconstruction obtained from wood anatomical parameters, that is a reconstruction based on a raw, non-detrended CWT rad chronology and a radial band w d h f 40 μ .We show reconstructed air temperatures from April 30 to September 7 computed as the median of 10000 realizations given with the 95% bootstrap confidence interval and compared to the E-OBS dataset over the period 1920-2017.At interannual timescales, the RMSE is much better than that obtained for TRW with 0.42, with a maximum discrepancy between the observed and reconstructed air temperatures during the heatwave of 2003 (with the recorded temperatures exceeding the reconstructed values by +1.8°C) and during the 1930s, for which the reconstruction overestimates air temperatures by roughly 1.4°C in 19324°C in  , 19334°C in   and 1939.At decadal timescales, the r(adj)  . A decadal timescales, the r(adj)   2 between observed and reconstructed air temperatures is 0.85.Whereas the raw, non-detrended CWT rad chronology also captures the recent warming adequately, it overestimates observed temperatures by 0.7°C over the 1925-1934 period.
findings agree withCarrer et al. (2018) stating that correlations >0.6 could be obtained between P. cembra CWT rad measured in earlywood cells and June temperatures as well as between CWT rad measured in latewood cells and August temperatures.They are also in line withStirbu et al (2022)    who reported significant correlation (r=0.65 over the 1961-2013 period) between CWT chronologies of P. cembra and summer (JJA) temperatures.The correlations obtained for God da Tamangur exceed those (r=0.6)reported for the Upper Val di Pejo (Italian Alps) MXD chronology, located 40 km SE of God da Tamangur, and summer (JJA) air temperatures over the period 1901-2015(Cerrato et al., 2019)..When comparing our results to reconstructions based on MXD chronologies of L. decidua, our CWT rad reconstruction reaches values (r=0.79 and CE=0.55 for the period 1920-2017) that exceed those reported in one of the most widely used European Alps temperature proxy records(Büntgen et al., 2006; r=0.73 with a CE=0.47 for the period 1911 r=0.73 with a CE=0.47 for the period  -2003)).In the case of the smoothed reconstructions, a value of r=0.69 (1818-2003)  is computed between the 20-year smoothed MXD L. decidua chronology and the JJAS temperature from the Histalp dataset(Auer et al., 2007), whereas we obtain a correlation of r=0.9 (0.92) between the 11-(20)-year smoothed CWT rad chronology and J o u r n a l P r e -p r o o f Journal Pre-proof the E-OBS dataset.Noteworthy, if the reconstruction of mean air temperatures is limited to 1950, correlation values even reach r=0.86 and CE=0.69.Such a limitation in time can be justified as the E-OBS dataset with daily temperature values still has a research status for the period 1920-1949 (i.e. the TRW and MXD chronologies existing at other high-altitude / high-a ud (Jac by a d D'anatomical parameters analyzed is able to capture the amplitude of the heatwave of summer 2003 (Schär and Jendritzky, 2004).Should heatwaves like the ones recorded in summers 2003, 2009, 2011 or 2015 become the new normal (Perkins-Kirkpatrick & Lewis, 2020), it seems likely that even J o u r n a l P r e -p r o o f J o u r n a l P r e -p r o o fJournal Pre-proof values with yearly ring widths or maximum latewood density.Findings of this paperas well as those ofCarrer et al. (2018) are encouraging and call for further reconstructions based on P. cembra wood anatomical parameters.In addition, P. cembra could become a valuable addition or alternative to L. decidua reconstructions as they conserve decadal to multi-decadal climate variability without any detrending.

Fig. 4 .
Fig.4.(A) Correlations of the spline detrended ring widths (TRW) chronology and E-OBS air temperatures over time windows ranging from 1 to 365 days.(B) Regression between detrended TRW data and E-OBS air temperature for a time window extending from March 8 to July 4 for which correlations (r=0.57) were found to be highest.(C) Comparison of interannual and decadal

Fig. 5 .
Fig.5.(A) Correlations of the raw radial cell-wall thickness (CWT rad ) chronology and E-OBS air temperatures over time windows ranging from 1 to 365 days.(B) Regression between detrended CWT rad data and E-OBS air temperature for a time window extending from April 30 to September 7 for which correlations (r=0.64) were found to be highest.(C) Regression between raw CWT rad data and E-OBS air temperature for a time window extending from April 30 to September 7 for which correlations (r=0.79) were found to be highest.(D) Comparison of interannual and decadal (i.e.smoothed with a 11-yr spline) E-OBS and CWT rad reconstructed air temperatures for a time window extending from April 30 to September 7 over the period 1920-2017.
widths ranging from 10 to 100 µm as well as optimal time windows used in reconstructions.EPS = expressed population signal, Rbar= average inter-series correlation, RMSE= root mean-square errors, r= coefficient of correlation, RE = reduction of error, CE = coefficient of efficiency.For details see text.
Fig. 1 Table 2 evidences that the TRW and LA reconstructions are poorly robust with a RE of 0.13 [-0.33 to Myers and Kitajima 2007).As a result, LBM outbreaks will largely suppress climate signals in L.

Table captions Table 1 .
Overview of dendroclimatic studies from the European Alps based on Pinus cembra L. and correlations obtained with precipitation (P) and temperature (T) records.Positive correlations are shown in red, negative correlations in blue.

Table 2 .
Statistics of reconstructions based on LA, CWT tan ,, CWT rad , aMXD and TRW for radial band

Table 1 Reference Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Significance Type
Esper et al., 2008