TEMPORAL STABILITY OF CLIMATIC SIGNAL RECORDED BY STABLE CARBON ISOTOPE COMPOSITION OF TREE RINGS α-CELLULOSE—A CASE STUDY FOR SUWAŁKI REGION

Abstract

Investigations of stable carbon isotope composition in α-cellulose extracted from tree rings of pines (Pinus sylvestris L.) growing in the unpolluted Suwałki region, northeastern part of Poland, are undertaken. The presented carbon isotope record covers the period of 1931–2003. Values of δ¹³C measured in the tree ring α-cellulose are compared to meteorological data. These δ¹³C values in tree ring cellulose respond to summer temperature, insolation, relative humidity, and precipitation. The best correlation is observed between relative humidity and carbon isotope data. The August relative humidity is found more influential on δ¹³C values than relative humidity for any other month or combination of months (r = –0.65). Relations between isotopic and meteorological data demonstrate that precipitation influences the stable carbon isotopic ratios to a lower extent than humidity. The intensity and duration of summer rainfall events can determine this effect. The temporal stability of climate-proxy connections is an important issue in paleoclimatic reconstruction. Therefore, the temporal stability of climatic signals recorded by stable carbon isotopes is analyzed in this research using the moving correlation function for moving intervals with a 25-year window. Based on those investigations the highest time stability of correlation was found for the carbon isotope and the August relative humidity. More variability is observed for the correlation of δ¹³C values with precipitation.

INTRODUCTION

Tree rings represent an integration of the conditions corresponding to each growing season. Tree ring widths and maximum latewood density may be used to reconstruct some aspects of the climatic changes, e.g., summer temperature or precipitation. Such reconstructions are known mainly for high altitudes or latitudes where climatic impact is the dominating factor for tree growth (Esper et al. 2002). However, tree ring reconstruction often fails if a more complex set of growth-limiting conditions exists. Variation of stable carbon isotope composition (δ¹³C) in the subsequent rings may be a valuable parameter for reconstructing growth conditions in temperate climate regions (Loader and Switsur 1995). McCarrol and Loader (2004) included a summary of papers that have explored tree ring isotope time series for palaeoenvironmental research. Treydte et al. (2007) presented results of carbon and oxygen isotopes investigations from 23 sites from Finland to Morocco including Suwałki. Cellulose as the main and stable wood component is traditionally used for isotope analyses in paleoclimatological studies (McCarrol and Loader 2004; Boettger et al. 2007).

A tree grows under specific environmental conditions, like temperature, precipitation, humidity, light intensity, and change in soil moisture, and these conditions influence the photosynthesis process and respectively the degree of carbon isotope fractionation in corresponding tree rings (Stuiver et al. 1984; Robertson et al. 1997; Schleser et al. 1999). Tree-ring stable carbon isotope ratios (δ¹³C) in environments of low moisture stress are likely to be controlled primarily by the photosynthetic rate. Therefore, sunshine, rather than temperature, represents the more direct controlling factor. Temperature reconstructions based on tree-ring δ¹³C results thus rest on the assumption that temperature and sunshine are strongly linked (Young et al. 2010, 2019). At sites where summer moisture stress restricts stomatal conductance to CO₂, or where moisture stress is involved in the covariation of multiple climate variables (precipitation, temperature, cloudiness etc.) carbon isotope variability can be used to reconstruct summer precipitation and drought (Gagen et al. 2022). Processes controlling carbon isotopic fractionation in trees are not fully explained, yet. According to the model developed by Francey and Farquhar (1982), carbon fractionation occurs in the leaf during CO₂ diffusion through stomata and during carboxylation. Many investigations demonstrated the influence of climatic conditions on δ¹³C values in tree rings (Freyer and Belacy 1983; Stuiver and Braziunas 1987; Lipp et al. 1991; Szczepanek et al. 2006; Andreu-Hayles et al. 2017; Young et al. 2019; Chakraborty et al. 2022; Gagen et al. 2022). In trees, the dominant environmental factors that control the δ¹³C of leaf sugars are those which determine stomatal conductance and the rate of photosynthesis. Dry conditions are related to low stomatal conductance and lead to reduced carboxylation discrimination against δ¹³C (Tuzet et al. 2003).

Nonstationary relationships between climatic parameters and isotopic data are rarely discussed in the literature (Aykroyd et al. 2001; Daux et al. 2011; Naulier et al. 2015; Savard and Daux 2020). Temporal variability in the strength of proxy-climate correlations was found in several cases (Reynolds-Henne et al. 2007; Hilasvuori et al. 2009; Rinne et al. 2010; Young et al. 2010; Daux et al. 2011; Boettger et al. 2014). Identifying periods of instability in climate-isotope data correlations is very important in producing reliable reconstructions.

In addition to climatic factors, many other factors can affect measured values of δ¹³C in annual tree rings such as the choice of wood component, or the effect of respired CO₂ (McCaroll and Loader 2004). Furthermore, CO₂ assimilated by the plant comes from different sources: natural, as well as anthropogenic. The increase in the burning of fossil fuels in industrial areas since the 19th century caused the emission of CO₂ to the atmosphere and changes in carbon isotopic composition in the atmosphere and following in plants. Depression of δ¹³C values by about 1.5‰ results from the assimilation of ¹³C-depleted CO₂ due to emission from fossil fuels connected with anthropogenic activity (Donngara and Verrica 2002; McCaroll and Loader 2004; McCaroll et al. 2009). For most tree ring δ¹³C series this trend is apparent (Freyer and Balacy 1983; Treydte et al. 2001), but not for all series (Tans and Mook 1980; Robertson et al. 1997).

The main purpose of isotopic studies for the Suwałki region is to reconstruct climate changes in the past based on δ¹³C in tree rings and to check human impact on the environment. Climate reconstruction requires prior calibration to establish a statistical relationship between isotopic and climatic data. It is also important to check whether other factors, especially anthropogenic ones, affect δ¹³C values. The purpose of the research presented in here is to determine the correlation of δ¹³C with climatic factors and check whether they change over time. The next stage of the research will be to check the influence of human activities on δ¹³C values. In this case, studies of radiocarbon concentrations, which were carried out for the annual increments of a tree from neighbor Augustów (Pawełczyk and Pazdur 2004), can help in the conclusion. Such a comprehensive study will make it possible to choose the best calibration to reconstruct the climate for the Suwałki region.

DESCRIPTIVE BACKGROUND

Sampling

The research area is situated not far from Suwałki city (54°06 ’N, 22°57’E) in the northeastern part of Poland close to the Polish–Lithuanian border. There are valuable forest complexes in the vicinity of the study area relatively unchanged by human activity. Material for isotopic investigations covering the years 1931–2003 comes mainly from the Wigry National Park.

The coring of living pine trees (Pinus sylvestris L.) was taken with a Pressler increment borer (0.5 cm in diameter). For the mentioned period, 14 trees were subjected to dendrochronological studies. Cores contained 124–231 rings. The pine dendrochronological standard for the Suwałki region is fully described in Szychowska-Krąpiec and Krąpiec (2005). The quality of the tree ring chronology was evaluated by calculating the expressed population signal (EPS) statistics. The EPS was higher than 0.85 and therefore was reliable along the whole length of the series.

Each sample used for isotope analyses consists of annual tree rings pooled from the four best cross-dated trees. It is a number proven to be satisfactory for a representative isotope site record (Borella et al. 1998; Treydte et al. 2001, 2007; Leavitt 2008).

Cellulose Extraction

Isotope analyses were carried out on the α-cellulose separated from the wood samples. The technique for the extraction of α-cellulose is based on the method presented by Green (1963) and Loader et al. (1997). The extraction procedure of α-cellulose is fully described in Pawełczyk et al. (2004) and Pazdur et al. (2005)

Isotopic Measurements

The δ¹³C measurements presented in this paper were part of an isotopic study carried out for the Suwałki site. The entire project covered measurements for 400 years in the case of carbon and oxygen isotopes and for 100 years in the case of hydrogen isotopes. The use of simultaneous measurements of δ¹³C and δ¹⁸O has made it possible to significantly reduce the number of measurements.

At the UFZ-Umweltforschungszentrum in Leipzig-Halle, δ¹³C measurements in α-cellulose were carried out simultaneously with δ¹⁸O measurements using an online technique with an IRMS-DELTA plus XL spectrometer (Thermo Finnigan) connected to an HTP reactor (HEKAtech, Germany), in which the pyrolysis process took place (Knöller et al. 2005).

The δ¹³C ratios in cellulose are classically obtained by combusting the samples to CO₂ in an Elemental Analyzer (EA) coupled to an isotope ratio mass spectrometer. In previous studies, the high-temperature pyrolysis (HTP) technique has been reported as useful for determining δ¹⁸O and δ¹³C isotopic ratios at the same time from a single cellulose sample operating at 1400°C or above. Despite the presence of reactive carbon, HTP produces reasonably good δ¹³C values as confirmed by higher agreement with measurements obtained by the elemental analyzer system and the HTP (Knoller et al. 2005; Andreu-Hayles et al. 2019).

The results are given as in the usual δ notation in ‰ versus VPDB scale calibrated using IAEA-CH-3 (δ¹³C = –24.72‰) and IAEA-CH-6 (δ¹³C = –10.45‰) standards. The error margins of these methods are ±0.2‰ for carbon.

Site Description and Climatic Data

The area of Wigry National Park is covered by glacial sediments from the Würm Glaciation (tills and fluvioglacial sands with cobbles and boulders) reaching 150 m deep. The spectrum of soils developed on this allochthonous material includes rusty (aerosols) and podzolic soils. Tree stands encompass mainly Norway spruce and Scots pine (Migaszewski et al. 2005).

Climatically, the Suwałki Region substantially differs from other regions of Poland. It is the coldest part of Poland, apart from the mountains, with annual mean temperatures around 6.3ºC (for years 1931–2003). The climate of the Suwałki Region is quite harsh, with some continental features. The spring reaches this region usually three weeks later than in southwestern Poland. The annual sum of precipitation is relatively low (591 mm/year—mean value for years 1931–2003), and it achieves the highest value in the summer season (80 mm/month). Predominates, the summer rainfall, which is characteristic of areas with continental climate features. On average, the summer rainfall consists of 63% of the total annual rainfall, and for different years it ranges from 46 to 81%. On average, there are 163 days per year with more than 0.1 mm of precipitation. The highest number of these is in winter and there are fewer in the warm season, meaning that rainfall is more abundant then and is often stormy. Heavy rain of more than 10.0 mm happens 14 days a year, mostly in the summer, from June to September. The winters are long and frosty, with long-lasting snow and ice cover (between 90 and 110 days in a year), and the number of days with average temperatures below the freezing point (0ºC) amounts to 108–119 a year. The vegetation period lasts for about 180–190 days. The highest relative humidity is here, as in Poland, in the cool season, especially in winter, when the average monthly relative humidity exceeds 90%. The driest period is in May and June—when the monthly average relative humidity drops below 70% (Kondracki 1998; Krzysztofiak and Olszewki 1999; Woś 1999).

The instrumental meteorological data used for the investigation started in 1931, in the case of temperature and precipitation, and from 1954, in the case of humidity and insolation. The data were obtained from the meteorological observatory in Suwałki (54°07’N, 22°58’E), which is located within 40 km of the sampling sites. An ombrothermic diagram of the region of Suwałki is presented in Figure 1.

Figure 1 An ombrothermic diagram of the region of Suwałki.

Ensuring the quality of climate data is of key importance when using long data series to carry out analyses related to the impact of variability of meteorological conditions over a selected area on other measured values. The Institute of Meteorology and Water Management – National Research Institute undertook activities aimed at developing complete information related to the daily variability of thermal and precipitation conditions in Poland. The official way of calculating characteristics, especially mean daily air temperature, changed over different periods and depended on the measuring program of the station. For example, to calculate the official value of mean daily air temperature since 1966, the following formula was used:

(1) $$\rm {Tsr = (T00 + T03 +T06 + T09+ T12 +T15 +T18 +T21)/8,} $$

where:

Txx = air temperature at measurement date xx (UTC),

For earlier data, a slightly different formula was used:

(2) $$\rm{Tsr = (T06 + T12+ 2 * T18)/4}. $$

Other climate data were also prepared with similar care.

Calculation Methods

The autocorrelation analyses were performed using MATLAB. An unbiased statistic provides a more accurate estimate of the population parameter. A biased statistic will either underestimate or overestimate the population parameter. When the bias is removed, the autocorrelation appears noisier at higher lags.

In our study, the correlations between isotopic and climatic data were calculated using the DendroClim2002 program. Each bootstrap estimate is obtained by generating 1000 samples (selected as random with replacement), and then running numerical computations for each sample (Biondi and Waikul 2004).

A simple, but effective way to explore the stability of calibration models is to compute them for multiple periods or intervals, which, ideally, should be systematically selected to minimize bias. Using a moving interval technique (constant length progressively slid by one year) the temporal stability of the correlation between isotope chronology and climate was investigated by Biondi and Waikul (2004), and Cullen and Grierson (2006). For the Suwałki region investigations of the temporal stability of climatic signal were performed for 25-year moving windows with the first period being 1931–1955, the second 1932–1956, and so on. To perform moving correlation functions DendroClim2002 requires an interval greater or equal to twice the number of climatic predictors (for this case—temperature, precipitation, humidity, or insolation of selected months or seasons). To meet this requirement the number of monthly predictors was reduced to 12 by choosing periods (months, seasons), based on results shown in Table 1.

Table 1 Correlation coefficients (R) estimated using DENDROCLIM2002 (Biondi and Waikul 2004) between isotopic data for pine (Pinus sylvestris L.) and meteorological data (T—temperature, P—precipitation, H—humidity, S—insolation). Bolded values denote the significant values (at the 95% confidence level). Capital letters in the month descriptions denote the previous year.

* Years 1932–2003.

** Years 1955–2003.

Values of δ¹³C in tree rings from Suwałki are presented in Figure 2. For the carbon, there are values of raw δ¹³C and corrected according to McCarroll and Loader (2004), due to changes in stable carbon isotope ratios of atmospheric CO₂ since industrialization.

Figure 2 Isotopic (δ¹³C uncorrected and corrected) and meteorological records (T—temperature, P—precipitation, H—humidity, S—insolation, data covers the period May–August) for Suwałki region.

RESULTS AND DISCUSSION

Autocorrelation Analyses

Biased and unbiased autocorrelation function estimates for isotope chronology of raw and CO_2atm corrected δ¹³C, were determined. Although the unbiased estimator is more appreciated from the statistics point of view, it increases in an unjustified manner for large lags. The autocorrelation analysis is therefore performed for lags, where the two estimators are similar, i.e., for about 25% of total lags, as presented in Figure 3. It can be concluded that isotope chronologies are not fully random, and they exhibit significant deterministic components because the envelopes of the estimates do not rapidly decay. For all cases, significant first-order autocorrelation could be seen. This points to a permanent influence on isotope values for the current year by those of the previous year. This is probably due to the tree’s use of reserves accumulated during the previous growing season. Reserves stored late in the growing season are used for early wood formation the following spring.

Figure 3 Autocorrelation functions for (a) δ¹³C, (b) corrected δ¹³C.

Trends for Stable Carbon Isotope Ratios

For the entire period from 1931 to 2003, the raw δ¹³C values for the pine tree ring cellulose from the Suwałki Region show no significant trend (slope = –0.0003, R² = 5 × 10⁻⁵). However, for the corrected δ¹³C values, an increasing trend is observed (slope = 0.0187, R² = 0.2528). A decreasing trend in the raw δ¹³C values, reflecting the global trend of δ¹³C in atmospheric CO₂ due to fossil fuel burning, was expected. When examining two shorter periods within the study, 1963–1983 and 1984–2003, changes in trends for both δ¹³C and δ¹³C_corrected are evident. For the period 1963–1983, a decreasing trend is observed (δ¹³C: slope = –0.508, R²= 0.4799; δ¹³C_corrected: slope = –0.227, R² = 0.1516), aligning with expectations. The Suwałki Region is relatively clean and free from significant industrial impact, so the local effect of δ¹³C decrease should be minimal, although the global effect should be visible. In contrast, for the period 1984-2003, an increasing trend is observed (δ¹³C: slope = 0.0389, R² = 0.0901; δ¹³C_corrected: slope = 0.067, R² = 0.2264). This change is likely related to the increase in SO₂ emissions, also evident in the Suwałki region. There was an EMEP (European Monitoring and Evaluation Programme) station in Suwałki from 1979 to 1992, and later a station in Diabla Góra. These stations, located in relatively clean areas, were meant to record background conditions for northeastern Poland. Nevertheless, an increase in SO₂ concentrations in the air was recorded in this area in the 1980s and early 1990s. Gaseous SO₂ can cause the closure of stomata, resulting in less negative carbon isotope composition (Rinne et al. 2010).

Additionally, in the last 20 years of the 20th century and the early 21st century, the Suwałki region experienced significant fluctuations in temperature and precipitation. During this period, an increase in temperature (slope = –0.0817, R² = 0.3691) and sunshine (slope = 2.5631, R² = 0.2883) was observed, along with a decrease in humidity (slope = –0.2975, R² = 0.3639) and annual precipitation (slope = 2.487, R² = 0.1256). Less negative δ¹³C values correlate with warm and low humidity conditions in summer, while more negative values are associated with cooler and humid summers (Cullen and Grierson 2006).

Relationship between Climate and Stable Carbon Isotope Ratios

Relationships between isotope values for tree rings and monthly climate data were modeled using a bootstrapped correlation function in DendroClim2002 (Biondi and Waikul 2004). Each bootstrap estimate is obtained by generating 1000 samples (selected as random with replacement), and then running numerical computations for each sample (Biondi and Waikul 2004). The correlation coefficients between annual isotope data (δ¹³C, δ¹³C_corrected) in tree ring cellulose and temperature, insolation, sum of precipitation as well as humidity are presented in Table 1.

It was found that in the same case, a stronger correlation between climatic and isotopic data is observed in the case of uncorrected carbon isotopic data. Therefore, correlation is presented for corrected as well as for uncorrected data. In the case of carbon isotopes, the highest correlation coefficient due to temperature (r=0.38, n=72 for correlation with August temperature) for the period 1932–2003 exists for uncorrected δ¹³C data, but when considering the shorter time interval 1955–2003 a higher correlation coefficient occurs for the corrected δ¹³C (r=0.51, n=49, for correlation with August temperature. If periods combined from several months are considered, higher correlation coefficients (r=0.54, n=72) are obtained when analyzing the relationship between uncorrected δ¹³C data and the mean temperature of the May–August period. The positive correlation between summer temperature and δ¹³C is consistent with results from the literature (McCaroll and Loader 2004). The correlation between carbon isotope and insolation is a little stronger than in the case of correlation with temperature (r=0.47, n=49 for insolation in June in case of uncorrected δ¹³C, and r=0.60, n=49 for the insolation in May–August in the case of the corrected value of δ¹³C). Insolation, rather than temperature, represents the more direct controlling factor. The correlation between the sum of precipitation and carbon isotopic data is stronger for the corrected data in the case of the longer period 1932–2003 (r=–0.35, n=72, p<0.005 for precipitation in August), but for the shorter period 1955–2003 is stronger for corrected data (r=–0.45, n=49, p<0.005 for precipitation in August). If periods combined from several months are considered, higher correlation coefficients are obtained also for May–August precipitation (n=–0.46, n=72) for the corrected δ¹³C data. The relative humidity values yield the best correlation between climate and carbon isotope data. The August relative humidity value is more influential than the relative humidity of any other month or combination of months, in determining the δ¹³C values of cellulose (r=–0.65, n=49, p<0.001 in the case of uncorrected data). Many investigators reported that the δ¹³C of organic matter is strongly related to water status and humidity (Roden and Ehleringer 2007). Water stress or high vapor pressure deficits cause stomatal closure and reduce intercellular CO₂ concentrations and discrimination against ¹³C by Rubisco. Thus, δ¹³C in organic matter has been used to estimate plant water-use efficiency (Farquhar et al. 1989). For carbon, it was found that the photosynthetic rate is controlled primarily by photon flux (sunshine) and temperature, whereas stomatal conductance is by air humidity and soil moisture (McCarroll and Pawellek 2001; Gagen et al. 2007). Values of δ¹³C in environments of low moisture stress are likely to be controlled primarily by photosynthetic rate. The results do not make it clear which factor—stomatal conductance or photosynthetic rate—is the dominant factor controlling δ¹³C in the study area. Based on the presented results it can be proven that precipitation influences the stable isotopic carbon ratios less than the relative humidity especially in the case of uncorrected δ¹³C.

Temporal Stability of Climatic Signal Recorded by Stable Carbon Isotope Composition of Tree Rings α-Cellulose

The temporal stability of climate-proxy connections is an important issue in paleoclimatic reconstruction. The temporal stability of estimated correlation coefficients is presented in Figure 4 for the months and seasons of the highest correlation between climatic and isotopic data. In Figure 4 changes in the correlation coefficient for the calculated 25-year interval from 1966–1990 to 1979–2003 can be observed.

Figure 4 Temporal stability of correlations between δ¹³C and climate factor for (a) uncorrected values of δ¹³C (b) corrected values of δ¹³C in 25-year time windows. Asterisks denote statistically significant correlations (p<0.05). The description on the vertical axis indicates which meteorological data and from which months were used to determine the correlation with the isotopic data (numbers denote months, and letters: T—temperature, P—precipitation, H—humidity, S—insolation).

Various potential reasons are provided for the instability of the strength of the isotope data-climate correlation. The article by Savard and Daux (2020) enumerates the following:

1. 1. sampling and data treatment artifacts

2. 2. stand dynamics

3. 3. effects of rising CO₂

4. 4. climate change

5. 5. pollution

The method of sample preparation and handling was designed to mitigate the first two factors. To address the third factor, a correction for CO₂ emissions was applied. Consequently, the last two factors could have influenced the lack of stability in the signals.

Non-stationary relationships between climatic parameters and isotopic data encompass climatic data from the final two decades of the twentieth century and the early years of the twenty-first century. Many elements demonstrate significant year-to-year variability and temporal fluctuations. Changes of particular climate elements are interrelated. The primary causative factors are both anthropogenic changes (greenhouse gas emissions resulting in increased greenhouse effect and global warming, local sources of air pollution) and natural changes: (1) circulation factors: changes in the intensity and location of atmospheric activity centers, changes in the frequency of advection from a specific sector, and the frequency of cyclonic and anticyclonic systems over Poland and (2) radiation factors (changes in values of global solar radiation, sunshine duration and cloudiness). These changes, especially visible after the 1980s, affect the trends of most climatic elements, meteorological phenomena and indices (Falarz et al. 2021). It is also noteworthy that δ¹³C values may contain an anthropogenic signal associated with elevated SO₂ emissions, which could have overlapped with the climate signal in the 1980s and early 1990s. Similar observations on the impact of SO₂ are presented in the article Boettger et al. (2014).

CONCLUSION

• The relationships between δ¹³C in tree-ring cellulose and meteorological data discussed above showed that, in the case of the Suwałki region, temperature and precipitation affect stable carbon isotopes less than insolation and humidity.

• The results do not make it clear which factor—stomatal conductance or photosynthetic rate—is the dominant factor controlling δ¹³C in the Suwałki region because a similar coefficient factor can be observed for humidity and insolation.

• In most cases, for the combined periods (several months) larger correlation coefficients than for one month were observed.

• The reported results confirm that stable isotope compositions of carbon can be regarded as indicators of climatic changes.

• Temporal variability in the strength of δ¹³C-climate correlations was found.