Stable hydrogen isotope values of lignin methoxyl groups of four tree species across Germany and their implication for temperature reconstruction

Science of the Total Environment 579 (2017) 263–271

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Stable hydrogen isotope values of lignin methoxyl groups of four tree species across Germany and their implication for temperature reconstruction Tobias Anhäuser a,⁎, Markus Greule a, Frank Keppler a,b a b

Institute of Earth Sciences, Heidelberg University, Im Neuenheimer Feld 234-236, D-69120 Heidelberg, Germany Heidelberg Center for the Environment (HCE), Heidelberg University, D-69120 Heidelberg, Germany

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• The δ2H values of lignin methoxyl of 660 tree-ring cores were analyzed. • δ2H values differ within and between trees by ≤ 10 and ≤ 28 mUr or ‰, respectively. • Species-specific differences in the apparent isotope fractionation were found. • The potential use of δ2H of lignin methoxyl as a paleotemperature proxy was tested. • For this approach European beech trees showed most potential.

a r t i c l e

i n f o

Article history: Received 4 October 2016 Received in revised form 15 November 2016 Accepted 16 November 2016 Available online 23 November 2016 Editor: D. Barcelo Keywords: Tree-rings Sample pooling Paleoclimate proxy Apparent fractionation

a b s t r a c t Stable hydrogen isotope ratios of lignin methoxyl groups (δ2HLM values) in wood have been shown to mirror the δ2H signatures of precipitation (δ2Hprecip values). Thus, δ2HLM values were suggested to serve as a potential paleotemperature proxy since δ2Hprecip values are dominantly controlled by air temperature in the mid-latitudes. A recent study where a significant δ2HLM-temperature relationship was found for a European transect with mean annual temperatures ranging from −4 to 17 °C strengthened this assumption. However, using δ2HLM values as a paleotemperature proxy requires quantification of noise from site-, species- and biosynthetic-specific influences to determine the significance of recording smaller temperature changes. Here, we measured δ2HLM values of treering sections covering 1981–1990 and 1991–2011 of four different tree species (European beech, English oak, Scots pine, Norway spruce) at 15 sampling sites across Germany. The maximum difference in mean annual temperature between sample sites was 5 °C and all sites showed small temperature increases from 1981 to 1990 to 1991–2011 (mean Δ = 0.7 °C). For all species investigated, the maximum difference of δ2HLM within the tree was b 10 mUr or ‰ (median values) and between trees at a single site was ≤28 mUr (median values). The general pattern of the spatial δ2HLM-temperature relationship found for the European transect was confirmed here although a significant correlation was lacking. This can be explained by the lower spatial δ2Hprecip-temperature correlation (R2 = 0.39) found for sampling sites in this study and the δ2HLM differences between trees. Nevertheless, the temporal changes in δ2HLM values of European beech trees correctly reflected within ±2 °C the temperature change at every sampling site. Therefore, we suggest that δ2HLM values of European beech trees have

⁎ Corresponding author. E-mail address: [email protected] (T. Anhäuser).

http://dx.doi.org/10.1016/j.scitotenv.2016.11.109 0048-9697/© 2016 Elsevier B.V. All rights reserved.

264

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271

considerable potential for reconstructing temperature changes when applied on tree-ring chronologies and consider this approach particularly suited for Late Holocene climate studies. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Tree-ring chronologies are valuable climate archives for the reconstruction of late Holocene temperature variability as they provide continuous records at annual resolution (Büntgen et al., 2016; Esper et al., 2012; McCarroll and Loader, 2004; Wilson et al., 2016). Temperature reconstructions can be inferred from plant physiological parameters such as tree-ring width or maximum latewood density from trees growing at altitudinal or latitudinal treeline (Briffa et al., 2002; Esper, 2000; Esper et al., 2003, 2002a, 2002b; Jacoby and D'Arrigo, 1989; Wilson et al., 2016). However, for non-treeline regions additional temperature proxies are necessary to increase spatial coverage and to improve our understanding of past climate evolution/change globally. In this context, stable hydrogen isotope ratios (δ2H values) of tree-rings seem suitable since they are derived from the δ2H values of precipitation (δ2Hprecip), which are dominantly controlled by temperature in the mid-latitudes (Dansgaard, 1964; Gat, 1996). Hence, δ2H values complement and even broaden the potential of tree-rings as a climate archive (Esper et al., 2015; Hartl-Meier et al., 2014; Liu et al., 2015; McCarroll and Loader, 2004). Recently it was suggested that the δ2H values of lignin methoxyl groups (expressed here as δ2HLM values) of tree-ring wood could be used as a paleotemperature proxy (Anhäuser et al., 2016, 2015, 2014; Keppler et al., 2007; Mischel et al., 2015). Methoxyl groups in tree wood are predominantly ether bonded in lignin. Their δ2H value can be readily measured without isotope fractionation as iodomethane (CH3I) which is released upon treatment of the wood with hydroiodic acid (Greule et al., 2008). The δ2H value of CH3I is measurable by gas chromatography-high temperature conversion-isotope ratio mass spectrometry (GC-HTC-IRMS). The speed and simplicity of the procedure enables the collection of large isotope data sets. The δ2HLM values of tree wood are, at a first order of control, derived from the δ2Hprecip values and are modulated by a large uniform apparent fractionation (εapp) (Anhäuser et al., 2016; Feakins et al., 2013; Keppler et al., 2007). For trees located in the mid-latitudes it was shown that they primarily reflect the mean annual δ2Hprecip value since tree source water integrates multiple precipitation events potentially involving water from the previous year (Anhäuser et al., 2016). In the Anhäuser et al. study it was also shown that the mid-latitudinal temperature signal of δ2Hprecip is reflected by δ2HLM values of tree wood since a significant correlation between δ2HLM values and mean annual temperatures (MAT) ranging from − 4 to 17 °C was found for various tree species across a European north-south transect. However, in order to determine the significance of δ2HLM when temperature ranges are lower (b 2 °C) further assessment of the noise emerging from site-specific (prior to methoxyl formation) such as secondary δ2Hprecip alterations at climatically homogenous sites, species-specific differences in εapp also termed as “plant taxonomy effect” (Liu et al., 2016) as well as biosynthetic-specific due to isotope fractionation. Further insight regarding these issues may also help to improve tree-ring sampling strategies. Here, we collected tree-ring sections of four different species including European beech, English oak, Scots pine and Norway spruce at 15 sampling sites across Germany with MATs in the range of 6.5 to 11.5 °C. At every sampling site three core samples per tree and up to five trees per species were collected. Besides the spatial (geographic) temperature variations, at every sample site a small temperature increase from 1981 to 1990 to 1991–2011 (0.7 °C on average) was also reported. Each core was dissected for both time periods and subsequently homogenized prior to δ2HLM measurements. Hence, the sampling

design allowed for four tree species a detailed assessment of the following aspects: (i) the ‘within tree’ variability as assessed circumferentially and temporally (ii) the ‘between tree’ variability of the δ2HLM values at climatically homogeneous sampling areas (iii) εapp—species-specific differences and within species variability (iv) the effect of pooling sample material on δ2HLM values. Instead of measuring δ2HLM of every core separately, it may be suitable to ‘pool’ material from multiple trees from each species at each site into one sample. This would considerably reduce the analytical workload and hence save time and resources (v) the relationship between δ2HLM values and both the spatial and temporal temperature variations 2. Materials and methods 2.1. Study sites and tree-ring sampling strategy The 15 study sites were located across Germany (Fig. 1a) with typically flat terrains in closed forests. At every study site, if available, four tree species, European beech (Fagus sylvatica), English oak (Quercus robur), Scots pine (Pinus sylvestris) and Norway spruce (Picea abies), were sampled within a few kilometers (Fig. 1b). When possible, samples from five trees of each species were taken within a radius of 200 m of each other (Fig. 1b). All trees sampled had a minimum circumference of 120 cm thus ensuring a tree age of approximately ≥60 years. From each tree, three core samples were taken (Fig. 1c) using a 5-mm diameter borer equidistant at approximately 1.2 m above ground with a minimum core length of 150 mm. In total, this led to 660 tree-ring cores of 220 trees. After collection the samples were dried in a drying cabinet (48 h at 30 °C). The annual growth rings were counted visually under a magnifier in order to separate the time sections 1981–1990 and 1991–2011. For some sections it was necessary to enhance the visibility of the ring structure using white chalk powder. Each tree-ring section was homogenized with a micro-mill (mesh size 1 mm). To generate ‘pooled samples’, sub-samples of the homogenized material were mixed (in equal amounts) so as to provide one representative sample of each investigated tree species at every site. For this we used the time section 1991–2011. Depending on the number of trees sampled at each site, a ‘pooled sample’ included material from 9 to 15 tree-ring cores. 2.2. Temperature and stable isotope data of precipitation Except for the site at Annweiler, weather stations of the Deutsche Wetterdienst (DWD) were located within 1 to 15 km of all other sampling sites which allowed comparison of the δ2HLM values with measured temperature data (Table 1). For the sampling site at Annweiler interpolated temperature data were used (Harris et al., 2014). For the δ2 Hprecip values we used measured data from stations of the Global Network of Isotopes in Precipitation (GNIP)1 that were available for the majority of the sampling sites. The δ2Hprecip and temperature data were averaged to match those of the prepared tree-ring time sections of 1981–1990 and 1991–2011. Both parameters showed an increase in 1

Data available at http://isohis.iaea.org.

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271

265

Fig. 1. a) Sampling locations in Germany are indicated by circles which are color coded depending on the tree species found at that site b) Schematic sampling procedure at each sampling site. The number of trees sampled per species varied between three and five. c) For every tree, three cores were removed equidistant at approximately 1.2 m above ground. Each core was dissected for the time periods 1981–1990 and 1991–2011 and was subsequently homogenized.

the range of 1 to 6 mUr (milliurey = 0.001 = 1‰, cf. Section 2.4) and 0.3 to 1.1 °C, respectively (Table 1). 2.3. Instrumentation for the 2H analysis of methoxyl groups Hydrogen isotope signatures of the lignin methoxyl groups from the homogenized wood samples were measured as CH3I released upon

treatment of the ground samples with hydroiodic acid (HI) by the procedure of Greule et al. (2008). HI (0.5 mL; puriss. p.a., 55–60%, not stabilized, purchased from Sigma-Aldrich, Seelze, Germany or Gillingham, UK, respectively) was added to the sample (2–10 mg) in a crimp glass vial (1.5 mL; IVA Analysentechnik, Meerbusch, Germany). The vials were sealed with crimp caps containing PTFE lined butyl rubber septa (thickness 0.9 mm) and incubated for 30 min at 130 °C. After heating,

Table 1 Tree species collected together with the temperature and δ2Hprecip values for the German sampling sites. Sample site

Görlitz Köln/Bonn Regensburg Öhringen Berlin Schleswig Bremen Leipzig Meiningen Karlsruhe Mannheimc Braunschweig Annweilerc Hohenpeißenberg Oberstdorf Mean values a

Latitude

51.161754 50.839528 49.060113 49.222428 52.421551 54.519458 52.972932 51.313194 50.564608 49.049636 49.553478 52.284100 49.190466 47.801103 47.388806

Longitude

14.842596 07.192028 12.157065 09.454049 13.117061 09.528030 08.635384 12.361556 10.361139 08.233050 08.576956 10.454700 07.980519 11.008883 10.271242

Altitude [m.a.s.l.]

238 139 475 373 143 62 43 128 481 130 104 80 250 800 684

Tree species collected

Weighted mean δ2Hprecip[mUr vs. VSMOW]b

Mean temperature [°C]a

European beech

English Norway oak spruce

Scots pine

1981–1990 1991–2011 Difference 1981–1990 1991–2011 Difference

5 5 – 5 5 5 5 3 3 4 4 4 4 3 5

5 5 5 5 5 5 5 3 – 4 4 4 4 3 5

5 5 5 5 3 5 5 – 3 4 4 5 – 4 –

8.5 10.0 8.4 9.3 9.7 8.3 9.1 9.2 7.1 10.6 10.5 9.3 9.6d 6.8 6.2 8.8

5 5 5 5 5 5 5 – 3 – – – 4 3 –

9.0 10.5 9.5 10.1 10.1 8.8 9.7 9.7 8.1 11.5 11.1 9.8 10.4d 7.5 6.5 9.5

0.5 0.5 1.1 0.8 0.4 0.5 0.6 0.5 1.0 0.9 0.6 0.5 0.8 0.7 0.3 0.7

−60[7] −70

−64

6

−60

−57[1] −47[7]

3

−64[1]

−59[10]

5

−59 −59 −56 −59 −75

−55 −55 −53 −55 −74[5]

4 4 3 4 1

−63

−58

4

Temperature data obtained from the Deutsche Wetterdienst (DWD) at www.dwd.de. b δ2Hprecip data obtained from the Global Network of Isotopes in Precipitation (GNIP). The subscript indicates the number of years that were not involved for the calculation of the mean value due to insufficient or missing δ2Hprecip data. c At this sample site the δ2Hprecip data has been used from the nearby (b50 km) GNIP station of Karlsruhe. d temperature data obtained using the CRU TS 3.10 grid-box (Harris et al., 2014).

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271

standards were calibrated against international reference substances (VSMOW2 [δ2HVSMOW = 0.0 ± 0.3 mUr] and SLAP2 [δ2 HVSMOW = − 427.5 ± 0.3 mUr]) using TC/EA-IRMS (elemental analyser-isotopic ratio mass spectrometer, IsoLab, Max Planck Institute for Biogeochemistry, Jena, Germany). The calibrated δ2H values in mUr vs. V-SMOW for the two CH3I working standards were − 173.0 ± 1.5 mUr (n = 9, 1σ) and − 66.2 mUr ± 1.2 mUr (n = 8, 1σ). Every wood sample was measured in quadruplicate followed by consecutive injections of both working standards. Standard deviations (n = 3, 1σ) were in the range of b 1 to 3 mUr. Additional uncertainty is introduced by the ‘external precision’ (also referred as the chemical replication uncertainty) and has a value of 1.6 mUr previously estimated for 10 aliquots of a single wood sample by Greule et al. (2008). Using Gaussian error propagation, the overall uncertainty in δ2H determination of the lignin methoxyl groups was in the range of ±2–5 mUr.

European beech

Temporal difference in ²HLM [mUr]

60 40 20 0

n=150

n=144

n=171

f -180 -200 -220 -240 -260

Scots pine

n=24

[mUr]

80

n=180

n=8

-30

n=50

e 100

English oak

-20

n=31

n=11

n=12

n=13

0

-10

n=34

20

0

0

n=29

40

10

10

app

60

20

δ2Hprecip

20

Apparent fractionationn,

80

c 30

n=11

d 100

30

n=48

n=50

n=48

n=57

n=60

0

40

n=57

10

50

n=13

20

Throughout this paper, the ‘delta’ (δ) notation, the relative difference of the isotope ratio of a material to that of a standard V-SMOW (Vienna Standard Mean Ocean Water) is used: values of δ2H relative to that for V-SMOW are defined by the equation: δ2H = ((2H/1H) sample / (2H/1H) standard) − 1. To comply with guidelines for the International System of Units (SI), we follow the proposal of Brand and Coplen (2012) and use the term urey, after H.C. Urey (Urey, 1948) (symbol Ur), as the isotope delta (Coplen, 2011) value unit. In such a manner, an isotopedelta value expressed traditionally as −25‰ can be written −25 mUr.

b

n=60

30

2.4. Definition of δ values and isotope fractionation

60

n=14

40

Circumferential variability '81-'90 [mUr]

50

'Between trees' variability '81-'90 [mUr]

a 60

n=14

'Between trees' variability '91-'11 [mUr]

Circumferrential variability '91-'11 [mUr]

the samples were allowed to equilibrate at room temperature (22 ± 0.5 °C, air conditioned room) for at least 30 min before an aliquot of the headspace (10–90 μL) was collected and directly injected into the analytical system using a gastight syringe (100 μL, SGE Analytical Science). The δ2H values (see below) of the CH3I formed from the methoxyl groups in the wood samples (δ2HLM) were measured using an HP 6890N gas chromatograph (Agilent, Santa Clara, USA) equipped with an auto sampler A200S (CTC Analytics, Zwingen, Switzerland), coupled to a DeltaPLUSXL isotope ratio mass spectrometer (ThermoQuest Finnigan, Bremen, Germany) via a thermo conversion reactor [ceramic tube (Al2O3), length 320 mm, 0.5 mm i.d., reactor temperature 1450 °C] and a GC Combustion III Interface (ThermoQuest Finnigan, Bremen, Germany). The gas chromatograph (GC) was fitted with a Zebron ZB-5MS capillary column (Phenomenex, Torrance, USA) (30 m × 0.25 mm i.d., df 1 μm) and following conditions were employed: split injection (4:1), initial oven temperature at 30 °C for 3.8 min, ramp at 30 °C/min to 100 °C. Helium was used as carrier gas at a flow of 0.6 mL/min constant flow. A tank of high purity hydrogen gas (H2, hydrogen 5.0, Linde, Höllriegelskreuth, Germany) with a δ2H value in the range of −195 to − 225 mUr was used as the working reference gas (V-SMOW, range provided by the supplier). The H+ 3 factor, determined daily during the measurement period, was in the range 4.1 to 4.4 ppm/nA. All δ2HLM values were normalized by a two-point linear calibration (Coplen, 1988; Paul et al., 2007) using δ2H values of two CH3I working standards relative to V-SMOW. The δ2H values of the CH3I working

n=12

266

Norway spruce

Fig. 2. Box plots of the ‘within tree’ and ‘between trees’ δ2HLM variability as wells as the εapp variability. The ‘within tree’ variability is indicated by the maximum difference between the three circumferentially determined δ2HLM values per tree shown in a) for 1991–2011 and in b) for 1981–1990 (n indicates the number of trees). c) Temporal δ2HLM difference between 1991 and 2011 and 1981–1990 for every tree-ring core. Also shown is the temporal difference of δ2Hprecip for the same time sections (cf. Table 1). The ‘between trees’ variability is indicated by the maximum difference between the mean δ2HLM values of every tree per site (n indicates the number of sites) shown in d) for 1991–2011 and in e) for 1981–1990. f) Variability of the apparent isotopic fractionation (εapp) between δ2Hprecip (GNIP) and δ2HLM (using the mean value of every tree with n showing the number of trees). Only sites were used that were close to a GNIP station (10 out 15 sampling sites). Box plots describe the spread of data: median value (bold black horizontal line), upper and lower quartile (boxes) and maximum and minimum value (vertical lines). Outliers exceed the upper/lower quartile by 1.5 times of the interquartile range.

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271

The apparent fractionation between δ2Hprecip and δ2HLM (εapp) was calculated according to the following equation: εapp = (δ2Hprecip + 1)/(δ2HLM + 1) − 1. As done for the δ2H values we report the εapp values equivalently in Ur. 3. Results and discussion 3.1. ‘Within tree’ δ2HLM variability As indicator for the circumferential δ2HLM variability we used the maximum difference between the three determined δ2HLM values (summarized in box plots for each tree species for 1991–2011 and 1981–1990 time periods in Fig. 2a and Fig. 2b, respectively). For the 1991–2011 period all tree species showed a median circumferential variability of ≤10 mUr. A similar pattern was noted for the 1981–1990 period. The lowest circumferential variability was found for English Oak (Quercus robus) ranging from 1 to 18 mUr with a median of 7 mUr. Similar circumferential variabilities have been found for the δ2H values of cellulose with values of 10–30 and 5–20 mUr for Abies pindrow and Quercus petraea, respectively, reported when investigations were conducted with individual tree-rings (Leavitt, 2010 and references therein) instead of homogenized time sections used in this study. The causes of circumferential δ2H variability which have been previously discussed for cellulose and cellulose nitrate (Leavitt, 2010 and references therein) are also now considered appropriate for lignin methoxyl groups. It has been argued that the circumferential variability reflects heterogeneous source water δ2H values that may be present in different root directions (Edwards, 1993). Further influences on circumferential δ2HLM differences may be related to normal/compression wood (Luckman and Gray, 1990) which, however, has not yet been investigated for the lignin methoxyl groups. Since in our study we used decadalscale wood sections instead of annual tree-rings, some mismatches in δ2HLM may arise by a circumferential asymmetric tree-ring width pattern. Unfortunately, such influences are not only challenging to quantify but their relative contribution may also vary for every tree. However, the ‘overall uncertainty’ in the δ2HLM determination of ± 2–5 mUr (refer to Section 2.3) can explain the circumferential δ2HLM variability for the majority of trees as they show values of ≤10 mUr. The remaining noise likely arises from non mass-weighted tree-ring sections and may be reduced by measuring δ2HLM annually. We further calculated the difference in δ2HLM between time periods 1991–2011 and 1981–1990 of every tree-ring core to evaluate the temporal δ2HLM variability (summarized for each tree species in box plots in Fig. 2c). The European beech samples indicate a temporal increase in δ2HLM with a median value of 4 mUr. English Oak, Scots pine and Norway spruce samples show values of −1, −1 and 0 mUr, respectively, indicating a minor decrease or no temporal change in δ2HLM. Thus, the δ2HLM increase shown for the European beech trees reflects the δ2Hprecip increase seen for Germany with a mean value of 4 mUr (Table 1) whilst the other three tree species do not resolve this change in δ2Hprecip. When using tree-rings as climate archives, ‘juvenile’ trends in stable isotope signals are considered to potentially obliterate climate signals (Leavitt, 2010; Lipp et al., 1993; McCarroll and Loader, 2004). However, since the cores investigated here covered a minimum tree age of 60 years and where only the most recent 30 years were used for our investigations a ‘juvenile’ effect for the δ2HLM values should be at least reduced but, may differ in persistence for the four tree species examined. Nonetheless, it appears that European beech trees reflect temporal differences in δ2Hprecip as low as 4 mUr and is further discussed in Section 3.5. 2

3.2. ‘Between trees’ δ HLM variability As an indicator for ‘between trees’ variability of a tree species, we used the maximum difference between the mean δ2HLM values of

267

every tree per site (summarized in box plots for 1991–2011 in Fig. 2d and for 1981–1990 in Fig. 2e). Depending on tree availability the number of trees per species per site varied between three and five. For both time periods the median ‘between trees’ variability of European beech, Scots pine and Norway spruce is ≤28 mUr whilst English oak showed elevated variabilities. Thus, the median ‘between trees’ variability is at least twice that of the circumferential variability. Studies investigating δ2H values of cellulose showed similar variabilities in the range of 5 to 30 mUr (Leavitt, 2010 and references therein). The individual tree species were located within a radius of about 200 m. Hence, secondary δ2Hprecip alterations (prior to methoxyl formation) may to some degree contribute to the ‘between trees’ variability. Such alterations could arise by different soil and hydrological properties known to cause differences in water accumulation and/or influences of isotopically different groundwater (Anhäuser et al., 2016). This suggestion is supported by investigations of Feakins et al. (2013) who measured δ2H values of the xylem water which is a direct proxy for the δ2H value of the source water (Dawson and Ehleringer, 1991; White et al., 1985) that is used for the lignin methoxyl formation. They sampled five freshwater plants of Coccoloba diversifolia at two sites (50 m apart) showing a maximum difference of 17 mUr in the δ2H values of the xylem water. Hence, it is likely that more than half of the ‘between trees’ variability of ≤28 mUr shown in this study is due to site-specific δ2Hprecip alterations. The remaining variability could arise due to an effect of the biosynthetic lignin methoxyl formation (discussed in Section 3.3). 3.3. εapp—species-specific differences and within species variability The εapp values have been determined for every tree species (Fig. 2f; Table 2). For δ2Hprecip only measured GNIP data from German stations that were close to our sampling sites were used (10 out of 15; cf. Table 1). For the δ2HLM value the mean of every tree was used. For English oak and Scots pine almost identical εapp values have been calculated with a mean value of −212 ± 14 and −213 ± 16 mUr, respectively. In comparison to these tree species a somewhat higher εapp value was observed for European beech (− 205 ± 15 mUr) and a much lower value was found for Norway spruce (−237 ± 19 mUr). The causes of these species-specific differences in εapp are not yet fully understood but may be considered due to either a different seasonal timing of water uptake, different depth of water usage (root system) and/or genetic related differences in lignin methoxyl biosynthesis. Nevertheless, differences in εapp exist and care must be taken when, for instance, averaging δ2HLM values of different tree species, particularly so if Norway spruce is included. Similarly Liu et al. (2016) also advised caution for nalkanes when using taxonomically blind average εapp values. The variabilities of εapp within a species as expressed by the standard deviations (1σ) are in the range of ±14 to 19 mUr (Table 2) and reflects the ‘between trees’ variability. Secondary δ2Hprecip alterations at a single site (as evaluated in Section 3.2) seems a plausible noise considering the magnitude of the εapp variability. However, noise induced by the biosynthetic fractionation (εbio) in the course of lignin methoxyl formation might be involved in the variability of εapp. Recently, the assumption that εbio is a species-specific constant was challenged for other Table 2 Mean apparent fractionation (εapp) and standard deviation (1σ) between δ2Hprecip and δ2HLM for the four tree species. Tree species

εapp with SD [mUr]

European beech English oak Scots pine Norway spruce Various trees (Anhäuser et al., 2016)

−205 −211 −213 −237 −213

± ± ± ± ±

15 14 17 19 17

Number of trees

εapp (site mean) with SD [mUr]

Number of sites

37 41 35 27 111

−205 −211 −213 −236

9 10 8 6

± ± ± ±

11 7 12 17

268

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271

Fig. 3. Comparison between the δ2HLM values of mean site single measurements (black circles) and samples when ‘pooled’ (white circles) prior to isotope analysis for each investigated tree species. Standard deviations (1σ) are shown for the single measurements and are within the symbol for the ‘pooled’ samples.

compound-specific δ2H proxies. Differences in εbio have been shown on a sub annual level for leaf wax n-alkanes (Newberry et al., 2015) and leaf cellulose (Kimak et al., 2015). An equivalent noise affecting εbio for the lignin methoxyl groups in tree-rings (not leaf derived) and particularly when using annually resolved samples has yet to be quantified. When the mean δ2HLM value per site is used for calculation of εapp the standard deviation is reduced for each tree species and indicates that this mean δ2HLM value captures a more site representative isotope signature (Table 2). Interestingly, in Anhäuser et al. (2016) an εapp with a similar variation in magnitude (17 mUr) was reported when 8 different tree species were investigated. However, we consider that the similarity resulted from the much larger sample set Anhäuser et al. (2016) used in their calculations. 3.4. Comparison between the δ2HLM value of the ‘pooled’ samples versus the mean of the single measurements Comparing ‘pooled’ δ2HLM values (cf. Section 2.1) with the site mean δ HLM values of the single measurements shows almost exclusively an agreement between both values when the standard deviation (1σ) of the single measurements is taken into account (Fig. 3). Mean differences between both values for all sampling sites are: European beech (1 mUr), English oak (1 mUr), Scots pine (6 mUr) and Norway spruce (4 mUr). Therefore, making use of a pooling strategy for sample preparation to considerably reduce the analytical workload would appear useful, most especially for European beech and English oak. Furthermore, the standard deviation of the site mean δ2HLM value of the single measurements reflects the ‘within tree’ and especially the ‘between trees’ 2

variability. This uncertainty estimation should also be considered valid for the ‘pooled’ δ2HLM value as its standard deviation solely represents the analytical uncertainty (McCarroll and Loader, 2004).

3.5. The relationship between δ2HLM values and both the spatial and temporal temperature variations 3.5.1. The spatial relationship Recently, Anhäuser et al. (2016) showed that spatial δ2HLM variations of various tree species reflected the spatial δ2Hprecip-MAT relationship for a temperature range from − 4 to 17 °C across Europe (Fig. 4, upper linear relationship). The studies conducted on the tree ring samples in this study enable these relationships to be discussed on a smaller MAT range (6.5 to 11.5 °C) for four tree species. In addition to the GNIP stations that were close to our sampling sites we used additional sites2 to obtain a more representative relationship of the spatial δ2HprecipMAT relationship for the 1991–2011 time period. For the δ2HLM values we used the mean δ2HLM values per site of each of the four investigated tree species (colored circles in Fig. 4). From Fig. 4 it can be noted that both precipitation and lignin methoxyl isotope values for the German sampling sites in this study are in broad agreement with the isotope data pattern shown by Anhäuser et al. (2016). Only the Norway spruce which shows a systematic offset in the δ2HLM values in the range of 20– 40 mUr stands out from the data for the other tree species. However, the coefficients of determination are low for the δ2Hprecip-MAT relationship 2

Bad Salzuflen, Stuttgart, Konstanz, Hof-Hohensaas, Trier, Würzburg, Cuxhaven.

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271

269

Fig. 4. Upper linear relationships: δ2Hprecip values versus MAT for the European north-south transect (white diamonds; Anhäuser et al., 2016) and Germany (blue diamonds). Lower linear relationships: δ2HLM values versus MAT for the European north-south transect (white circles; Anhäuser et al., 2016) and the German tree sampling sites (colored circles) which show no significant spatial δ2HLM-MAT relationship for each tree species (p values ≥ 0.19). Black box shows the mean δ2HLM of all core samples (excluding Norway spruce). For clarity purposes standard deviations are not shown.

(R2 = 0.39, p b 0.05) and not significant for the δ2HLM-MAT relationship (Fig. 4). The results indicate that the spatial temperature influence on δ2 Hprecip is strongly reduced when using sampling sites on a regional scale (Germany) with a considerably smaller MAT range, as investigated in this study (6.5 to 11.5 °C). The δ2HLM differences among the four tree species reflect their differences in εapp (Fig. 2f) and explain the offset of the δ2HLM values of Norway spruce. No significant correlation was found between δ2HLM and MAT for any of the species investigated and suggests that the (spatial) temperature signal cannot be resolved for the smaller MAT range in Germany. This might be explained by the following considerations. The linear δ2Hprecip-MAT correlation is low for the German sampling sites (R2 = 0.39) when compared to the European transect (R2 = 0.95). In addition, when the standard deviation (1σ) of εapp of the investigated tree species in the range of ±7 to ±17 mUr is considered (Table 2), it appears plausible that δ2HLM cannot resolve spatial δ2Hprecip differences of Δ = 27 mUr. Nevertheless, the mean of all δ2HLM values (excluding Norway spruce) match very well the slope of the linear regression of the European δ2HLM-MAT relationship. This indicates that increasing the sample size leads to a highly representative δ2HLM signature that reflects accurately the absolute MAT within Germany (black box in Fig. 4).

3.5.2. The temporal relationship All sampling sites showed an increase in the mean temperature and δ2Hprecip values for the time periods 1981–1990 to 1991–2011, with a range between 0.5 and 1.1 °C and 1 and 6 mUr, respectively (Table 1). This suggests that temperature is not only a controlling factor of δ2 Hprecip values spatially but also temporally in Germany. We assume that δ2Hprecip values increased in general in Germany, i.e., also for the remaining sampling sites lacking a nearby GNIP station.

The temporal increase of δ2HLM of European beech trees has been shown to match well with the temporal increase of δ2Hprecip (cf. Section 3.1 and Fig. 2b) and, thus, is considered most suitable to reflect the observed temperature change in the range of 0.3 and 1.1 °C (Table 1). When the δ2HLM-temperature sensitivity as presented in Fig. 4 is assumed (~ 4 mUr/°C) the temperature change can be determined from European beech for every sampling site. The difference between the predicted (using δ2HLM) and the instrumental observed temperature change is presented in Fig. 5. With respect to the observed temperature change at all 14 sampling sites the predicted temperature change differs in the range of − 0.6 and 1.8 °C and for 12 sampling sites in the range of − 0.4 and 0.8 °C. The mean difference between

Fig. 5. Graph showing the difference between the temperatures predicted from the δ2HLM change observed in European beech trees and the instrumentally observed mean temperature change between time periods 1981–1990 and 1991–2011 at every sampling site. A δ2HLM-temperature sensitivity of 4 mUr/°C was assumed (cf. Fig. 4) for the prediction. The grey horizontal line represents the mean difference.

270

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271

predicted and measured was 0.4 °C (grey horizontal line in Fig. 5). This indicates that small temporal (relative) temperature changes (in the range of 1–2 °C) are reflected by δ2HLM much more precisely as observed for spatial (absolute) temperature differences (Fig. 4). Hence, δ2HLM values of European beech show the most potential to reflect Late Holocene temperature changes when using annually resolved tree-ring chronologies.

Acknowledgements

4. Conclusion

References

The δ2HLM values have been estimated for two homogenized time sections (1981–1990 and 1991–2011) for four tree species at 15 sampling sites within Germany. Based on the results, we were able to address the five aspects outlined in the introduction of this paper. For all tree species the circumferential variability of δ2HLM was lower than 10 mUr (median value) and is suggested mainly to be the result of both the uncertainty in the determination of δ2HLM and potential mismatches due to circumferential ring-width asymmetries. The temporal difference between the 1981–1990 and 1991–2011 time periods in δ2HLM of European beech trees (median value = 4 mUr) closely reflected the observed δ2Hprecip change for the same period (4 mUr). The ‘between trees’ δ2HLM variability at a single site was ≤28 mUr for the four tree species. When investigations from other studies (Feakins et al., 2013) are taken into consideration, we suggest that this variability is mainly induced by heterogeneous source water δ2H values at a single site although some input of noise arising from biosynthetic isotope fractionation should not be ignored. The ‘between trees’ variability might be reduced by increasing the number of trees for the determination of δ2HLM. Differences in εapp were found for all the four tree species investigated, but particularly for Norway spruce. Hence, care must be taken when, for instance, averaging δ2HLM values of different tree species. The variability in εapp of ±14 to 19 mUr (1σ standard deviation) reflects the ‘between trees’ δ2HLM variability. The comparison between the ‘pooled’ δ2HLM values and the site mean δ2HLM values of the single measurements showed almost exclusively an agreement between both values. Almost identical values were found for European beech and English oak (mean deviation 1 mUr) whilst Scots pine and Norway spruce showed somewhat higher deviations. Hence, conducting a pooling strategy in order to considerably reduce the analytical workload and hence save time and resources would seem both viable and attractive. The δ2HLM values have been shown to significantly reflect spatial MAT variations when the δ2Hprecip values are primarily controlled by MATs such as across Europe in the range of − 4 and + 17 °C (Fig. 4). The general pattern for the relationship between MAT and both these δ2H values has been reproduced for the German-wide collected tree samples. However, when the δ2Hprecip values are assessed within an area of Germany showing a lower range of the MATs of 6.8 to 11.5 °C the correlation between both parameters decreases indicating obliteration of the temperature signal in δ2Hprecip and consequently in δ2HLM. The temperature significance for predicting absolute temperatures may be improved by increasing the sampling size (N 5trees). The observed mean temperature difference between time periods 1981– 1990 and 1991–2011 was much lower (0.3 to 1.1 °C) in comparison to the differences in the MAT of the sampling sites (6.8–11.5 °C). The δ2HLM values of European beech trees were applied to calculate the temporal temperature change for all 14 sampling sites. The calculated mean temperature change only differed slightly from the observed temperature change with 0.4 °C (Fig. 5). The results obtained in this study detail the potential of the use of δ2HLM values for use as a paleotemperature proxy for four tree species. European beech has been shown to record temporal changes in δ2Hprecip as well as temperature most accurately in mid-latitudes and, thus, appears to be a suitable species to likely resolve paleotemperature changes as low as 1–2 °C.

Anhäuser, T., Sirocko, F., Greule, M., Esper, J., Keppler, F., 2014. D/H ratios of methoxyl groups of the sedimentary organic matter of Lake Holzmaar (Eifel, Germany): a potential palaeoclimate/-hydrology proxy. Geochim. Cosmochim. Acta 142:39–52. http://dx.doi.org/10.1016/j.gca.2014.08.001. Anhäuser, T., Greule, M., Zech, M., Kalbitz, K., Mcroberts, C., 2015. Stable hydrogen and carbon isotope ratios of methoxyl groups during plant litter degradation. Isot. Environ. Health Stud. 51:143–154. http://dx.doi.org/10.1080/10256016.2015.1013540. Anhäuser, T., Greule, M., Polag, D., Bowen, G.J., Keppler, F., 2016. Mean annual temperatures of mid-latitude regions derived from δ2H values of wood lignin methoxyl groups and its implications for paleoclimate studies. Sci. Total Environ. 574: 1276–1282. http://dx.doi.org/10.1016/j.scitotenv.2016.07.189. Brand, W.A., Coplen, T.B., 2012. Stable isotope deltas: tiny, yet robust signatures in nature. Isot. Environ. Health Stud. 48:393–409. http://dx.doi.org/10.1080/10256016.2012. 666977. Briffa, K.R., Osborn, T.J., Schweingruber, F.H., Jones, P.D., Shiyatov, S.G., Vaganov, E.a., 2002. Tree-ring width and density data around the northern hemisphere: part 2, spatiotemporal variability and associated climate patterns. The Holocene 12:759–789. http://dx.doi.org/10.1191/0959683602hl588rp. Büntgen, U., Myglan, V.S., Ljungqvist, F.C., McCormick, M., Di Cosmo, N., Sigl, M., Jungclaus, J., Wagner, S., Krusic, P.J., Esper, J., Kaplan, J.O., de Vaan, M.A.C., Luterbacher, J., Wacker, L., Tegel, W., Kirdyanov, A.V., 2016. Cooling and societal change during the Late Antique Little Ice Age from 536 to around 660 AD. Nat. Geosci. 9:231–236. http://dx.doi.org/10.1038/ngeo2652. Coplen, T.B., 1988. Normalization of oxygen and hydrogen isotope data. Chem. Geol. 72: 293–297. http://dx.doi.org/10.1016/0168-9622(88)90042-5. Coplen, T.B., 2011. Guidelines and recommended terms for expression of stable-isotoperatio and gas-ratio measurement results. Rapid Commun. Mass Spectrom. 25: 2538–2560. http://dx.doi.org/10.1002/rcm.5129. Dansgaard, W., 1964. Stable isotopes in precipitation. Tellus 16:436–468. http://dx.doi. org/10.3402/tellusa.v16i4.8993. Dawson, T.E., Ehleringer, J.R., 1991. Streamside trees that do not use stream water. Nature 350:335–337. http://dx.doi.org/10.1038/350335a0. Edwards, T.W.D., 1993. Interpreting past climate from stable isotopes in continental organic matter. Geophys. Monogr. 78:333–341. http://dx.doi.org/10.1029/GM078p0333. Esper, J., 2000. Long-term tree-ring variations in Juniperus at the upper timber-line in the Karakorum (Pakistan). The Holocene 10:253–260. http://dx.doi.org/10.1191/ 095968300670152685. Esper, J., Cook, E.R., Schweingruber, F.H., 2002a. Low-frequency signals in long tree-ring chronologies for reconstructing past temperature variability. Science 295: 2250–2253. http://dx.doi.org/10.1126/science.1066208. Esper, J., Schweingruber, F.H., Winiger, M., 2002b. 1300 years of climatic history for Western Central Asia inferred from tree-rings. The Holocene 12:267–277. http://dx.doi. org/10.1191/0959683602hl543rp. Esper, J., Shiyatov, S.G., Mazepa, V.S., Wilson, R.J.S., Graybill, D.A., Funkhouser, G., 2003. Temperature-sensitive Tien Shan tree ring chronologies show multi-centennial growth trends. Clim. Dyn. 21:699–706. http://dx.doi.org/10.1007/s00382-003-0356-y. Esper, J., Frank, D.C., Timonen, M., Zorita, E., Wilson, R.J.S., Luterbacher, J., Holzkämper, S., Fischer, N., Wagner, S., Nievergelt, D., Verstege, A., Büntgen, U., 2012. Orbital forcing of tree-ring data. Nat. Clim. Chang. 2:862–866. http://dx.doi.org/10.1038/ nclimate1589. Esper, J., Konter, O., Krusic, P.J., Saurer, M., Holzkämper, S., Büntgen, U., 2015. Long-term summer temperature variations in the Pyrenees from detrended stable carbon isotopes. Geochronometria 42:53–59. http://dx.doi.org/10.1515/geochr-2015-0006. Feakins, S.J., Ellsworth, P.V., Sternberg, L.S.L., 2013. Lignin methoxyl hydrogen isotope ratios in a coastal ecosystem. Geochim. Cosmochim. Acta 121:54–66. http://dx.doi.org/ 10.1016/j.gca.2013.07.012. Gat, J., 1996. Oxygen and hydrogen isotopes in the hydrologic cycle. Annu. Rev. Earth Planet. Sci. 24, 225–262. Greule, M., Mosandl, A., Hamilton, J.T.G., Keppler, F., 2008. A rapid and precise method for determination of D/H ratios of plant methoxyl groups. Rapid Commun. Mass Spectrom. 22:3983–3988. http://dx.doi.org/10.1002/rcm.3817. Harris, I., Jones, P.D., Osborn, T.J., Lister, D.H., 2014. Updated high-resolution grids of monthly climatic observations—the CRU TS3.10 dataset. Int. J. Climatol. 34:623–642. http://dx.doi.org/10.1002/joc.3711. Hartl-Meier, C., Zang, C., Büntgen, U., Esper, J., Rothe, A., Göttlein, A., Dirnböck, T., Treydte, K., 2014. Uniform climate sensitivity in tree-ring stable isotopes across species and sites in a mid-latitude temperate forest. Tree Physiol. 35:4–15. http://dx.doi.org/10. 1093/treephys/tpu096. Jacoby, G.C., D'Arrigo, R., 1989. Reconstructed northern hemisphere annual temperature since 1671 based on high-latitude tree-ring data from North America. Clim. Chang. 14:39–59. http://dx.doi.org/10.1007/BF00140174. Keppler, F., Harper, D.B., Kalin, R.M., Meier-Augenstein, W., Farmer, N., Davis, S., Schmidt, H.L., Brown, D.M., Hamilton, J.T.G., 2007. Stable hydrogen isotope ratios of lignin

We thank Daniela Polag, Christoph Hann, Natalie Schröter, Fabien Koch and Felicia Haase for supporting field campaigns and wood preparation as well as John Hamilton for comments on an earlier version of the manuscript. This study was supported by the Deutsche Forschungsgemeinschaft (DFG; KE 884/6-2, KE 884/8-1 and KE 884/9-1).

T. Anhäuser et al. / Science of the Total Environment 579 (2017) 263–271 methoxyl groups as a paleoclimate proxy and constraint of the geographical origin of wood. New Phytol. 176:600–609. http://dx.doi.org/10.1111/j.1469-8137.2007.02213.x. Kimak, A., Kern, Z., Leuenberger, M., 2015. Qualitative distinction of autotrophic and heterotrophic processes at the leaf level by means of triple stable isotope (C-O-H) patterns. Front. Plant Sci. 6:1008. http://dx.doi.org/10.3389/fpls.2015.01008. Leavitt, S.W., 2010. Tree-ring C-H-O isotope variability and sampling. Sci. Total Environ. 408:5244–5253. http://dx.doi.org/10.1016/j.scitotenv.2010.07.057. Lipp, J., Trimborn, P., Graff, W., Becker, B., 1993. Climatic significance of D/H ratios in the cellulose of late wood in tree rings from spruce (Picea abies L.). Proceedings International Symposium on Applications of Isotopic Techniques in Studying Past and Current Environmental Changes in the Hydrosphere, pp. 395–405 Vienna. Liu, X., An, W., Treydte, K., Wang, W., Xu, G., Zeng, X., Wu, G., Wang, B., Zhang, X., 2015. Pooled versus separate tree-ring δD measurements, and implications for reconstruction of the Arctic oscillation in northwestern China. Sci. Total Environ. 511:584–594. http://dx.doi.org/10.1016/j.scitotenv.2015.01.002. Liu, J., Liu, W., An, Z., Yang, H., 2016. Different hydrogen isotope fractionations during lipid formation in higher plants: implications for paleohydrology reconstruction at a global scale. Sci. Rep. 6:19711. http://dx.doi.org/10.1038/srep19711. Luckman, B., Gray, J., 1990. Oxygen isotope ratios from tree rings containing compression wood. Quat. Res. 33:117–121. http://dx.doi.org/10.1016/0033-5894(90)90090-8. McCarroll, D., Loader, N.J., 2004. Stable isotopes in tree rings. Quat. Sci. Rev. 23:771–801. http://dx.doi.org/10.1016/j.quascirev.2003.06.017.

271

Mischel, M., Esper, J., Keppler, F., Greule, M., Werner, W., 2015. δ2H, δ13C and δ18O from whole wood, α-cellulose and lignin methoxyl groups in Pinus sylvestris: a multi-parameter approach. Isot. Environ. Health Stud. 51, 553–568. Newberry, S.L., Kahmen, A., Dennis, P., Grant, A., 2015. N-alkane biosynthetic hydrogen isotope fractionation is not constant throughout the growing season in the riparian tree Salix viminalis. Geochim. Cosmochim. Acta 165:75–85. http://dx.doi.org/10. 1016/j.gca.2015.05.001. Paul, D., Grzegorz, S., Fórizs, I., 2007. Normalization of measured stable isotopic compositions to isotope reference scales—a review. Rapid Commun. Mass Spectrom. 21, 3006–3014. Urey, H.C., 1948. Oxygen isotopes in nature and in the laboratory. Science 108:489–496. http://dx.doi.org/10.1126/science.108.2810.489. White, J.W.C., Cook, E.R., Lawrence, J.R., Wallace, S.B., 1985. The ratios of sap in trees: implications for water sources and tree ring ratios. Geochim. Cosmochim. Acta 49: 237–246. http://dx.doi.org/10.1016/0016-7037(85)90207-8. Wilson, R., Anchukaitis, K., Briffa, K.R., Büntgen, U., Cook, E., D'Arrigo, R., Davi, N., Esper, J., Frank, D., Gunnarson, B., Hegerl, G., Helama, S., Klesse, S., Krusic, P.J., Linderholm, H.W., Myglan, V., Osborn, T.J., Rydval, M., Schneider, L., Schurer, A., Wiles, G., Zhang, P., Zorita, E., 2016. Last millennium northern hemisphere summer temperatures from tree rings: part I: the long term context. Quat. Sci. Rev. 134:1–18. http://dx. doi.org/10.1016/j.quascirev.2015.12.005.