Large scale climate signals of a European oxygen isotope network from tree-rings-predominantly caused by ENSO teleconnections?

We investigate the annual variability of δ18O tree ring records from sites distributed all over Europe covering the 15 last 400 years. An Empirical Orthogonal Function (EOF) analysis reveals two distinct modes of variability on the basis of the existing δ18O tree ring records. The first mode of δ18O variability is associated with anomaly patterns of the El NiñoSouthern Oscillation (ENSO) and reflects a multi-seasonal climatic signal. The ENSO signal is visible for the last 130 years, but is found weak during the period 1600 to 1850 suggesting that the relationship between ENSO and the European climate may not stable over time. The second mode of δ18O variability, which captures an out-of-phase variability between 20 northwestern and southeastern European δ18O tree ring records, is related to a regional summer atmospheric circulation pattern revealing a pronounced centre over the North Sea. Locally, the δ18O anomalies associated with this mode show the same (opposite) sign with temperature (precipitation). We infer that the investigation of large-scale atmospheric circulation patterns and related teleconnections far beyond instrumental records can be done with oxygen isotopic signature derived from tree rings. However, the European δ18Ocel tree network needs to be consolidated and updated, as well as additional 25 research on the stationarity of reconstructed climate signals and the stationarity of teleconnections is advisable.

of δ 13 C values of tree-ring cellulose basically originates from fractionations during photosynthesis at the leaf or needle level that generally lower the δ 13 C of the atmospheric CO2 source which contains no direct climatic signal (Schleser, et al. 1995). 35 δ 18 O of tree-ring cellulose (δ 18 Ocel) is of particular interest for paleoclimate studies because it is related to source water, i.e. δ 18 O of precipitation (δ 18 OP), which is directly affected by climate processes, i.e. temperature during droplet condensation within air masses, transport distance from ocean source, type of precipitation (e.g. rain or snow) and precipitation amount (e.g. Dansgaard, 1964;Epstein et al., 1977;Rozanski et al., 1993). Within the arboreal system, δ 18 O of soil water (δ 18 OSW) constitutes the δ 18 O input and usually represents an average δ 18 OP over several precipitation events modified by partial 40 evaporation from the soil (depending on soil texture and porosity) and by a possible time lag, depending on rooting depth (Saurer et al., 2012). Representing the baseline variability, δ 18 OSW is invariably tied to the oxygen isotope signature of treering cellulose (δ 18 Ocel).
However, δ 18 Ocel is dependent on two more clusters of fractionations that reflect tree-internal processes, namely evaporative 18 O-enrichment of leaf or needle water via transpiration, and biochemical fractionations including partial isotopic exchange 45 of cellulose precursors with trunk water during cellulose biosynthesis (e.g. Saurer et al., 1997;Roden et al., 2000;Barbour, 2007;Kahmen et al., 2011;Treydte et al., 2014 and citations therein). The biochemical fractionation during cellulose biosynthesis can be largely considered as constant (27 ±4‰ (Sternberg and DeNiro, 1983)); however, varying leaf-to-air vapour pressure deficit following varying air humidity causes corresponding changes in the δ 18 O signature of leaf or needle water (e.g. Helliker and Griffiths (2007)). Although modified and dampened by physiological processes (e.g. Pèclet effect 50 (Farquhar and Lloyd, 1993) and oxygen isotope exchange with stem water during cellulose synthesis (Hill et al., 1995)) variability of 18 O enrichment of leaf-water is clearly affecting δ 18 Ocel, besides the strong signature of δ 18 OP. For example, the δ 18 Ocel are used to reconstruct precipitation (e.g. Rinne et al., 2013), air temperature (e.g. Porter et al., 2014) and drought (e.g. Nagavciuc et al., 2019). Since these quantities are largely based on transport processes within the atmosphere, the δ 18 Ocel values can be used to get detailed information about large-scale atmospheric circulation patterns (Andreu-Hayles et 55 al., 2017;Trouet et al., 2018, Nagavciuc et al., 2019. The resulting long-term perspective can be the key to identify the influence of different external forcing on, and internal variability of the behavior of large-scale atmospheric circulation. One of the most important components for the internal climate variability is the El Niño -Southern Oscillation (ENSO) which influences the atmospheric circulation globally (Allan, 1996). Since the ENSO variability is strongest in winter, 60 multiple studies have identified a significant ENSO impact on the European climate during this season. Observational studies (Fraedrich and Müller, 1992;Fraedrich, 1994;Pozo-Vazquez, 2005;Brönnimann, 2004;Brönnimann, 2007) and model studies (Merkel and Latif, 2002;Mathieu et al., 2004) suggest that an El Niño event leads to a negative phase of the North Atlantic Oscillation (NAO) with cold and dry conditions over Northern Europe and wet and warm conditions over Southeastern Europe in winter. Furthermore, it is also possible to identify a significant ENSO influence with regard to et al., 2020). To what extent these preconditions influence the climate conditions during summer is not yet known for Europe, but it can serve as a key for a better understanding of European climate. However, the relatively short period of existing instrumental data (van Oldenburgh and Burgers, 2005;Brönnimann, 2007), makes it difficult to describe the full 70 range of ENSO variability and its possible consequences for the climate of the European continent (Domeisen et al., 2019).
The aim of this study is to present a comprehensive spatiotemporal analysis of the large-scale European atmospheric circulation based on the influence of the ENSO variability over a long-term perspective. This study is based on the climatological signals of a European δ 18 Ocel network. For this purpose, Section 2 characterizes the isotope network. The used 75 climate data and methods are described in this Section as well. The results in Section 3 concentrate on the fundamental description of the dataset (Sect. 3.1) and the combined climate signal of the network (Sect 3.2). The links to ENSO are investigated using a multi seasonal perspective with different statistical approaches in Section 3.3. The multi seasonal signals of the δ 18 Ocel in combination with modelled δ 18 OP and δ 18 OSW are presented in Section 3.4. Furthermore, the second strongest climate signal is shortly introduced in Section 3.5. The final section summarizes the major results. 80 2 Data and Methods

The isotope network
We investigate the dominant modes of variability of 26 δ 18 Ocel records, distributed over Europe, and their relationships with regional and large-scale climate anomalies. 22 of the 26 δ 18 Ocel records were created within the EU project ISONET (Annual Reconstructions of European Climate Variability using a High-Resolution Isotopic network) (Treydte et al., 2007a, b).

85
Furthermore, four additional sites from Bulgaria, Turkey, Southwest Germany and Slovenia were added (Hafner et al., 2014;Heinrich et al., 2013). In total, the isotope network contains eight broadleaf tree sites (Quercus) and 18 coniferous tree sites (Pinus, Juniper, Larix, Cedrus) from altitudes varying for each location from 10m up to 2200m above sea level (Fig. 1). 24 of the 26 sites are distributed over the European continent whereas two additional sites are located in the Atlas Mountains of Morocco and in the Taurus Mountains of Turkey.

90
The stable isotopes of oxygen in tree-ring cellulose, reported as δ 18 Ocel vs. SMOW (Craig, 1957) of each site were determined as described by Treydte et al. (2007a, b). At least four dominant trees were chosen per site and two increment cores were taken per tree. After the standard dendrochronological dating following Fritts (1976), the individual tree rings were dissected from the cores. However, for oak only the latewood was used for the analyses. This procedure assumed that climate signals of the current year were predominantely applied since early wood of oaks frequently contains climate 95 information of the preceding year (Hill et al., 1995). The temporal resolution of the isotope records is annually.

Climate data
We use the gridded fields of the 20th Century Reanalysis Project (20CR) version V2c (Compo et al., 2011) for the 105 climatological analysis. The 20CR version V2c covers the period from 1851-2014 and has a temporal resolution of six hours and a spatial resolution of 2°×2°. Additionally, the reanalysis data of the Global Precipitation Climatology Centre (GPCC; Schneider et al., 2014) is used to study the relationship between the δ 18 Ocel and the precipitation anomalies. Furthermore, the COBE-SST2 dataset (Hirahara et al., 2014) is included in the study. To identify the links to drought conditions in Europe we make use of the Standardized Precipitation Evapotranspiration Index (SPEI3) dataset (Vicente-Serrano et al., 2010).

110
The data from both reanalyses were seasonally averaged: DJF (December to February), MAM (March to May), JJA (June to August) and SON (September to November). Since we focus on interannual to decadal variability, the linear trends from each grid cell have been removed.

Nudged model scenarios
Beside the observational/reanalysis-based climate data, we also investigate the relation between δ 18 OP and δ 18 OSW to receive 115 further insights about the fractionation / photosynthesis processes. Here, we analyse modelled data based on nudged ECHAM5-wiso simulations with ERA-40 and ERA-Interim reanalysis fields (Uppala et al., 2005;Berrisford et al., 2011;Dee et al., 2011) for the period 1960 to 2005 for the European region (Butzin et al., 2014).

Data analysis
As a first step, the characteristics of each time series of the δ 18 Ocel network and their relation to altitude and latitude are 120 investigated. For a better comparison, the linear trend of each δ 18 Ocel time series are removed and all time series are standardized (z-values).
To combine the signals of the isotope network, we use the Principle Component Analysis (PCA) and the Empirical Orthogonal Functions (EOF). These techniques were described by Pearson (1902) and Hotteling (1935) and were used for the first time by Lorenz (1956) for climatological studies (Storch and Zwiers, 1999). The EOFs are used to identify the most 125 dominant patterns of the δ 18 Ocel tree network variability. By applying the PCA, it is possible to gain a temporal perspective of these patterns since the phase and amplitude are described by the principal components (PC). In addition, the final components are checked if they can be analysed meaningfully by the rule of Kaiser (1960), which indicates that the https://doi.org/10.5194/cp-2020-39 Preprint. Discussion started: 30 March 2020 c Author(s) 2020. CC BY 4.0 License. eigenvalue (λ) of a component has to be λ> 1. The following components are further checked if they fulfil the requirements of the rule of North et al. (1982). This rule states that the pattern of the eigenvectors of one component is strongly 130 contaminated by other EOFs that correspond to the closest eigenvalues (Storch and Zwiers, 1999).
The ISONET network consists of a multi-site and multi-species tree-ring network covering more or less the period between 1600 to 2003. However, some tree-ring series cover the whole period, others cover only a shorter period. In order to be able to have a long-term perspective, one needs to find a statistically meaningful way to extend the shorter records to make use of the whole 400 years of data. Since most Multilinear Principal Component Analysis algorithms do not work with gaps in the 135 initial matrix we make use of an algorithm developed by Josse and Husson (2016) which is able to fill the temporal gaps without a change of the PC. In the first step we place the mean in the gap and execute a PCA with this dataset. Afterwards, the dataset is projected onto the new component axis. So, that the values are rotated and the value of a gap change. The new value for the gap is placed into the initial dataset. With this new dataset, a PCA is again carried out. This process is repeated until convergence is reached. The result is a gap-free dataset which can be used for PCA.

140
Composite maps of observed precipitation, air temperature, geopotential height 500mb (Z500) and SST are defined by extreme values of the network's principle component above a certain threshold. We choose events above and below one standard deviation with respect to the mean. The average climate conditions expressed by the composites allow us to analyse the general climate state occurring at times of extreme minima or maxima separately. In addition, both composites of climate conditions are combined under the assumption that the extremes of the minima and maxima show exactly the opposite 145 climate state. For this purpose, the minimum composite is subtracted from the maximum composite at each grid point.
Beside the composite maps, we extract the values of PC1 for those years for which ENSO values are higher than the average plus one standard deviation and lower than the average minus one standard deviation. The difference from the former distribution for the values of minima and maxima years is tested with the t-test. To better understand if El Niño or La Niña events coeval with extremes in the δ 18 Ocel time series, the Event Coincidence Analysis (Siegmund et al., 2017;Donges et al., 150 2016) using the PC1 and a December Nino 3.4 index is applied (HadISST1; Rayner et al., 2003). Therefore, we analyze whether the years in which the Nino 3.4 index is above the 75th percentile match the 75th percentile in PC1. In general, a significance level of α= 0.05 was used in all analyses.
Finally, we analyse the relation between seasonally averaged δ 18 OP and δ 18 OSW from winter, spring and summer, based on nudged ECHAM5-wiso simulations (Butzin et al., 2014), with PC1 based on the δ 18 Ocel values from the ISONET network. Quercus compared to Pinus. This might be determined genetically, because also δ 13 Ccel values of angiosperms co-occurring with gymnosperms are typically found more negative. Angiosperm wood tissue contains vessels, i.e. specialized waterconducting cells that are generally larger in diameter, and therefore more conductive to water, than conifer wood cells (Sperry et al., 2006).

165
The overall variance of the datasets is not dependent on the type of tree species. The highest scattering is found for the Beside the latitudinal effect, altitude also influences the oxygen isotope ratios as shown in Fig. 2B which can likewise be 175 described by a linear regression. It should be noted, that the southern sites are found at higher altitudes than the northern sites, which could influence the relation by a latitudinal effect. Therefore, we show that the δ 18 Ocel network is influenced by a latitudinal and an altitudinal effect, thereby supplying a proof of concept for the often-stated effects of latitude, altitude and continentality on the isotopic source value of water (McCarroll and Loader, 2004).

180
Performing PC analyses, the first component of the isotope network explains 16.2%, the second 9.1%, the third 6.4%, the fourth 5.5%, and the fifth 5.2% of the variance. Therefore, the first five components explain a cumulative variability of around 43%. Concerning the rule of Kaiser (1960), the first nine components can thus be meaningfully analysed. Since the first two components also fulfilled the requirements of the rule of North et al. (1982), they are investigated by their temporal and spatial characteristics.

185
The dominant pattern (EOF1), which describes 16.2% of the total variance, shows a spatially homogeneous pattern (Fig.   3A). The majority of time series in Europe are characterized by negative eigenvectors. The pole of this EOF pattern is centred over France and Germany. In contrast, tree sites close to the Mediterranean Sea and the northernmost site in Finland are characterized by eigenvectors close to zero. Therefore, these locations contribute little to the first component's time series (PC1) shown in Fig. 3C. Furthermore, the highest interannual variability is found for the beginning of the 18th and 19th century. The highest values of the second component (PC2) are reached at the beginning of the 18th century, whereas the smallest values are identified for the beginning of the 19th century (Fig. 3 D).  (Fraedrich, 1994;Fraedrich and Müller, 1992). 210 Furthermore, a negative phase of the winter NAO is also visible in the composite maps with cold and dry conditions over Northern Europe and wet and warm conditions over Southeastern Europe. In addition to elevated near surface temperature anomalies over Europe, the El Niño signal is visible in northern hemispheric temperature anomalies (Ineson and Scaife, 2008). That is also visible in density plots of PC1 (Supp. 1), which show that during El Niño years, the distribution of the PC1 is shifted towards higher values, whereas the distribution of the PC1 is shifted towards lower values for La Niña years.

215
According to the t-test, both shifts are significantly different (p < 0.05).
In composite maps for spring (Fig. 4B, 4E, 4H, 4K), the atmospheric circulation remains in a similar configuration over Europe and the North Atlantic compared to the winter season. In contrast, the equatorial Atlantic and the coast of South America show an enhanced warming (cooling) signal. A lagged warming of the equatorial Atlantic is identified up to 6 months after an El Niño event (Latif and Grötzner, 1999).
(lower) precipitation in Central Europe as well as lower (higher) surface air temperatures in summer corresponding to low (high) δ 18 Ocel values. Furthermore, the Northeast Atlantic shows cold (warm) temperatures whereas the North Atlantic south of Greenland shows a warming (cooling) tendency. This particular SST anomaly pattern is often associated with droughts 225 and heatwaves in Europe (Feudale and Shukla, 2011;Ionita et al., 2017).
To test if there is a significant relationship between the variability of our δ 18 Ocel records and drought variability, we correlated the JJA PC1 time series with the Central Europe JJA SPEI3 drought index (Vicente-Serrano et al., 2010; Longitude -5° to 10°/Latitude 46° to 52°). The SPEI3 index is suitable for this analysis because we to take into account the climate conditions of the pre-season to make it comparable to PC1. For the period from 1901 to 2005, the correlation is 230 significant (R=0.49; p < 0.01). To better understand whether extremes in δ 18 Ocel time series co-occurring with El Niño or La Niña events, we apply the Event Coincidence Analysis (Siegmund et al., 2017;Donges et al., 2016)

Comparison of δ 18 Ocel with modelled δ 18 O in precipitation and soil water
By employing nudged climate simulations with ECHAM5-wiso (Butzin et al., 2014), we aim at gaining better insight into how well the δ 18 Ocel tree signature is able to capture a multi-seasonal signal. A significant correlation between PC1 and δ 18 OP 240 is shown in the correlation maps for winter, spring and summer, where Central Europe is characterized by a moderate correlation (Fig. 6). A similar pattern can be identified for the correlation between δ 18 OSW and PC1. Compared to our previous analysis, the correlation between these quantities is increasing from winter to summer where it reached the maximum correlation. Since the δ 18 Ocel ratio is largely dependent on the isotopic composition of soil water, the correlation is even stronger with the δ 18 OP. We suggest that the δ 18 Ocel signal is a multi-seasonal signal which captures the hydroclimatic 245 conditions from winter towards summer. The reason for this situation is that depending on the root system, winter snowfall and groundwater storage, trees use water from previous seasons for photosynthesis (Treydte et al., 2006). Additionally, it is thereby implied that the duration of the water uptake should also be considered (Tang and Feng, 2001).

Further climate signals in δ 18 Ocel
Besides the multi-seasonal signal, the second component of the δ 18 Ocel values significantly relates to the summer climate 250 (Fig. 5). The second EOF of the δ 18 Ocel values, which explains 9.5 % of the total variance, shows a dipole-like pattern between northern and south-eastern Europe (Fig. 3B). This phenomenon is strongly visible in all composite maps (Fig. 5) and known as the dominant summer atmospheric circulation of the European climate . A positive https://doi.org/10.5194/cp-2020-39 Preprint. Discussion started: 30 March 2020 c Author(s) 2020. CC BY 4.0 License.
(negative) geopotential height anomaly in northern Europe co-occurs with a negative (positive) Z500 anomaly in southeastern Europe. The positive geopotential height anomaly is often described as atmospheric blocking-like pattern which 255 is related to climate extremes like floods and droughts for European mid-latitudes (Sillmann and Croci-Maspoli, 2009).
Moreover, this circulation anomaly pattern has been identified as the primary driver for extreme dry periods over the Eastern Mediterranean (Oikonomou et al., 2010) as well as the dominant driver for summer air temperature variability in Greece (Xoplaki et al., 2003a, b).
The temporal distribution of extremes in the PC2 time series, indicates that the 19th century has experienced increased 260 drought in northern Europe and enhanced precipitation in the Adriatic region (Fig. 3D). In addition, climate reconstructions also indicate that the Central part of Europe and parts of southern Sweden were relatively dry during the 19th century (Seftigen et al., 2013;Hanel et al., 2018). Since temperature extremes and precipitation patterns are largely determined by atmospheric blocking activity in these latitudes (Pfahl and Wernli, 2012), we suggest that the drought like conditions in northern Europe during summer are based on an enhanced atmospheric blocking activity.

4 Summary and conclusions
We present a δ 18 Ocel isotope network from tree rings for the last 400 years which was used to investigate the large-scale climate signals in the European mid-latitudes. According to our analysis, the climate signals of the network indicate a relation exists for winter, spring and summer with ENSO. The nudged model suggests that the summer signal is still dominating δ 18 Ocel but is partly influenced by lagged winter and spring signals. We argue that this is based on hydroclimatic 270 feedback processes as well as characteristics of the water reservoirs of the different sample sites. The ENSO signal is clearly visible for the last 130 years. However, no significant links can be deduced during the period 1600 to 1850 which is indicating that the relationship between ENSO and the European climate is not stable over time. The teleconnection changes between the tropical Pacific and Europe during the pre-instrumental period were also identified by other climate proxies (Rimbu et al., 2003).
275 Furthermore, our study shows the second mode with a dipole between North and Southeast Europe. Since this mode is highly relevant for the summer climate conditions on the entire European continent, the temporal perspective gives new insights about how the frequency of this mode changed through time. Our findings suggest that there is a tendency towards a situation whereby Southeast Europe is predominantly characterized by a high-pressure system and North Europe by a lowpressure system starting at the beginning of the 20th century.

280
Future research is required to provide additional insights into the stationarity of reconstructed climate signals and also the stationarity of teleconnections. Moreover, the European δ 18 Ocel tree network needs to be updated to capture the climate changes of the last 15 years. In addition, it is imperative to extent the isotope network by collecting more δ 18 Ocel records from Eastern Europe to improve the validity of our results for this region. In the context of the ongoing discussion about the   The first row shows the characteristics of the climate in DJF, the second in MAM and the third row in JJA, whereas the first line shows the results for surface temperature, the second for precipitation, the third for Z500 and the fourth for SST. The Z500 maps show similar characteristics in winter 565 and spring, whereas a pressure regime is directly located over Central Europe in summer. The SSTs in winter, spring and summer are characterized by ENSO activity. The Z500 and surface temperature dataset from 20CRv2c (Compo et al., 2011) and the COBE-SST2 dataset (Hirahara et al., 2014) Figure E, F, G, H are the correlation maps for EOF1 and theδ 18 O in soil water for winter, spring, summer and autumn. In all maps, the significant grid cells are coloured.