Tree rings, δ13C and climate in Picea glauca growing near Churchill, subarctic Manitoba, Canada

Available online at www.sciencedirect.com

Chemical Geology 252 (2008) 88 – 101 www.elsevier.com/locate/chemgeo

Tree rings, δ 13 C and climate in Picea glauca growing near Churchill, subarctic Manitoba, Canada J.C. Tardif a,⁎, F. Conciatori a , S.W. Leavitt b a

Centre for Forest Interdisciplinary Research (C-FIR), The University of Winnipeg, 515 Avenue Portage, Winnipeg, MB Canada R3B 2E9 b Laboratory of Tree-Ring Research, The University of Arizona, 105 West Stadium, Tucson, AZ 85721 USA Accepted 31 January 2008

Abstract White spruce (Picea glauca (Moench) Voss) trees were sampled to investigate the relationship between ring width and δ13C as well as between these descriptors and climate. Cross-sections from eight white spruce trees growing in the Hudson Bay Lowlands, subartic Manitoba, with an establishment date prior to 1775 were selected from our archive. For each cross-section, two radii were measured to produce ring-width chronologies. For stable-carbon isotope measurements, every fifth ring (2000, 1995, … 1750) was analyzed separately from each tree, whereas approximately equal masses from the eight trees were pooled into composite samples for each of the intermediate years mixing masses from each tree equal to that from the tree with the least amount. Alpha-cellulose was extracted and δ13C was determined using continuous-flow mass spectrometry. The ring-width residual chronology derived from eight trees was highly correlated to that previously developed for the area. A small but significant correlation was also found between the ring-width and the δ13C residual chronologies. The climate signal contained in both ringwidth and the δ13C residual chronologies suggests that warm early summers were conducive to larger rings whereas the δ13C best reflects the overall growing season temperature. Precipitation, relative humidity and/or drought index have little to no association with either chronology. Significant correlations were observed between δ13C and the historical ice records from the Hudson Bay Co. archive data suggesting that during long (warm) growing season the 13C content of alpha-cellulose increases. Our results indicate that both ring-width and δ13C provide complementary information. Future work will involve looking at long-term trends in carbon isotope ratios and the potential use of δ13C for dendroclimatic reconstruction. © 2008 Elsevier B.V. All rights reserved. Keywords: White spruce; Tree-line; Stable-carbon isotopes; Alpha-cellulose; Hudson Bay Co. archive; Forest-tundra transition

1. Introduction With the development of new mass-spectrometer analytical capabilities and novel sampling techniques (Dodd et al., 2008this volume), the use of stable isotopes in dendrochronology has been rapidly increasing (Borella et al., 1998; McCarroll and Loader, 2004, 2006). In addition to problems related to properly crossdated tree-ring series, many questions still persist as to the use of stable isotopes in dendroclimatology (Hughes, 2002). The majority of studies have related variations in stable-carbon isotope ratios to variation in moisture availability (see reviews ⁎ Corresponding author. Centre for Forest Interdisciplinary Research (C-FIR), University of Winnipeg, 515 Avenue Portage, Winnipeg, Manitoba, Canada R3B 2E9. Tel.: +1 204 786 9475; fax: +1 204 774 4134. E-mail address: [email protected] (J.C. Tardif). 0009-2541/$ - see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.chemgeo.2008.01.015

of Schleser et al., 1999; McCarroll and Loader, 2004). In support of stomatal conductance control on tree-ring δ13C, many researchers report a negative correlation between δ13C values and both relative humidity and precipitation (Leavitt and Long, 1989; Saurer and Siegenthaler, 1989; Lipp et al., 1991; Robertson et al., 1997; Treydte et al., 2001; Leavitt et al., 2002; Gagen et al., 2004). The rationale for such a control relates to the fact that during photosynthesis, plants preferentially assimilate the light carbon isotope (12CO2) and discriminate against the heavy isotope (13CO2) (Farquhar et al., 1989; McCarroll and Loader, 2006). This isotopic fractionation is mainly controlled by the CO2 diffusion through the stomata and by the carbon-fixing enzyme Rubisco, which discriminate against 13C. During moisture stress, stomatal closure interrupts the flow of CO2 and leads to greater utilization of 13C within the stomatal chamber by Rubisco and thus to higher 13C/12C ratios

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

(less negative δ13C values). While this relation to drought is fairly well documented, Saurer et al. (2004) hypothesized that the typical association between moisture and carbon isotope discrimination observed in regions with warm-dry climate may not hold in northern forests such as those in Siberia. In northern Finland, studies of Scots pine (Pinus sylvestris L.) suggested that photosynthetic rate, which is controlled primary by sunshine and temperature, dominated at sites near the distribution limit of the species whereas stomatal conductance, which is controlled by air humidity and soil moisture, was more important in southern sites (McCarroll and Pawellek, 2001; McCarroll et al., 2003). In the boreal forest of North America, few dendroisotopic studies have been conducted. In the interior boreal forest of Alaska, analysis of ring width, maximum density and δ13C in white spruce (Picea glauca (Moench) Voss) revealed that radial growth has decreased in association with increased temperatureinduced drought stress in the 20th century (Barber et al., 2000). In western Canada, Gray and Thompson (1977) found a high correlation between 18O/16O in cellulose of white spruce and mean annual temperature. In central Canada, Brooks et al. (1998) have looked at stable-carbon isotope in jack pine (Pinus banksiana Lamb). In North America, no δ13C chronologies have been developed for tree species growing at the northern tree-line. Likewise, little isotopic work (Sonninen and Jungner, 1995; McCarroll and Pawellek, 1998) has been conducted at the northern European timberline (McCarroll and Loader, 2006). In subarctic Manitoba and for the Churchill area, the latest Canadian Regional Circulation Model (modelling change in 2041–2060 relative to 1971–1990 simulated by CRCM3.6.1) predicts a significant warming with little change in precipitation (Canadian Centre for Climate Modelling and Analysis, 2005). Global warming has been linked to a lengthening of the Hudson Bay ice-free season (Gagnon and Gough, 2005) and to the fate of permafrost conditions (Gough and Leung, 2002). Potential response of the North American tree-line to climate change has been reviewed by MacDonald et al. (1998), Serreze et al. (2000) and Gajewski and Atkinson (2003). In contrast to the tree-line of the east side of the Hudson Bay where the growth response of tree species to climatic warming of the 20th century has been abundantly studied (Jacoby et al., 1988 and voluminous publications by the Subarctic and Subalpine Ecosystems research group of the Centre d'études nordiques (CEN) led by Dr. Serge Payette from the Université Laval), fewer tree-ring studies have been conducted on the west side of the Hudson Bay (e.g., Scott et al., 1988; Scott and Hansell, 2002; Girardin et al., 2005; Au and Tardif, 2007). Subarctic Manitoba has more commonly been a sampling point among a larger network of tree-ring chronologies (e.g., Guiot, 1987; Jacoby and D'Arrigo, 1991; D'Arrigo and Jacoby, 1993; Schweingruber et al., 1993; Briffa et al., 1994; Jones et al., 1995; D'Arrigo and Jacoby, 1999). In North America, white spruce achieves a transcontinental distribution from Newfoundland and Labrador west to Alaska east (Nienstaedt and Zasada, 1990). In Manitoba, white spruce reaches both its southern and northern distribution limit (more than 1000 km apart). At the southern limit of distribution of the

89

species, Chhin et al. (2004) reported that drought during the growing season was the main climate factor controlling radial growth. The authors observed a negative association with May– June–July temperature and a positive one with the same month's precipitation. Girardin and Tardif (2005) reported for the boreal plains that radial growth in white spruce was mainly associated with the previous and current year's summer precipitation (positive) and temperature (negative) and to the June drought signal. At its northern limit of distribution, ring width in white spruce was mainly associated (positively) with temperatures in June–July (Girardin and Tardif, 2005). Similar results were reported for sites near the northern tree-line on both the west and east side of the Hudson Bay (Jacoby and Ulan 1982; Jacoby et al., 1988). The objective of this study was to investigate both radial growth and δ13C in white spruce in relation to climate in the Churchill region of subarctic Manitoba. First, we developed and compared ring-width and δ13C chronologies. Second, we assessed the nature and strength of the climate signal in both chronology types. This initial study provides insight as to the utility of long-term annual-resolution δ13C chronologies in conjunction with standard ring-width chronologies in climate reconstruction studies in the forest-tundra transition. 2. Methods 2.1. Study area The study area is located in the Hudson Bay Lowlands ecoregion and south of the town of Churchill, Manitoba (58° 44'N; 94° 04'W) (Fig. 1). It lies within the Hudson Plain ecozone on the western coast of the Hudson Bay, referred to as the forest-tundra (Zoladeski et al., 1995). The forest-tundra forms a transitional boundary between the true boreal forest to the south and the tundra to the north (Ritchie, 1962). The study area is within the ecoclimatic region of the high subarctic (Scott, 1995). Ordovician and Silurian limestone and dolomite characterize the geology of the area and overlie the Precambrian Shield (Dredge and Nixon, 1992). However, glacial tills of the Pleistocene or marine deposits of the Holocene cover most of the region. As a result of the permafrost and a low slope, the landscape is a poorly drained peatland interspersed with thermokarst lakes and ponds. There are also well-drained glacial and beach formations such as moraines, beach ridges, and eskers (Scott, 1995). The entire study area is uniformly low (50 m a.s.l.) and flat, forming a vast wetland adjacent to Hudson Bay (Dredge and Nixon, 1992). In the study area, the crust is isostatically rebounding at an estimated average rate of about 1 m/century (Clarke et al., 1982). Regional climate is highly influenced by the effects of nearby Hudson Bay (Rouse, 1991), with long cold winters and short cool summers (Fig. 2). The strong arctic effect of the Hudson Bay ice pack (delaying the onset of summer conditions) persists through much of the summer (Rouse, 1984a) and could be important in maintaining high summer moisture during the snow-free season (Scott et al., 1993). During the snow-free period, air temperature can drop by as much as 20 °C as a result

90

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

Fig. 1. Map of the study area showing the location of Churchill, Manitoba (black square). The sites where white spruce trees were sampled are indicated by black circles. The location of the Churchill A meteorological station is indicated by a black star.

of wind shift from the south to the north (Rouse, 1984b). On average, ice usually breaks up in mid-June and floating ice remains in early July. Data from the meteorological station in Churchill A (58° 44' N, 94° 3' W, elevation: 28.70 m a.s.l.) show a mean annual temperature of − 6.9 °C for the reference period 1971–2000 (Environment Canada, 2003). The mean daily temperature during the coldest month (January) is − 26.7 °C, and during the warmest month (July) it is 12.0 °C. The total annual precipitation is 431.6 mm, with 264.4 mm falling as rain (Fig. 2). In the area, snowmelt usually occurs in May/June and the first snowfall occurs in mid-September (Scott

et al., 1993). On average, 594.7 degree days above 5 °C are observed annually (Environment Canada, 2003). The vegetation of the area is dominated by white spruce, black spruce (Picea mariana [Mill] BSP) and tamarack (Larix laricina (Du Roi) K. Koch). White spruce communities are found almost exclusively on well-drained sand or gravel ridges in the forest-tundra or at the junction of sand and rock at the northern tree-line of the study area (Ritchie, 1962; Brook, 2001). The black spruce communities, referred to as the “black spruce muskeg” (Ritchie, 1962), are mainly associated with sphagnum moss over peat deposits. Tamarack grows freely throughout the study area in association with both black spruce and white spruce (Brook, 2001; Monson, 2003). Pure tamarack stands are confined to fen areas with water levels at or near the surface. The trees are usually found on hummocks that form in this type of landscape (Ritchie, 1962). 2.2. Tree selection

Fig. 2. Monthly climate means for the period 1970–2000 at the Churchill A meteorological station (58° 44′ N, 94° 3′ W, elevation: 28.70 m a.s.l.).

Samples used in this study came from those collected as part as a larger study (Monson, 2003; Girardin et al., 2005). We decided that more than the four trees suggested by Leavitt and Long (1984) for the semi-arid U.S. Southwest environment would best ensure a representative local isotope chronology in this subarctic environment. In fact, Robertson et al. (1997) and McCarroll and Pawellek (1998) had found that under some circumstances more than four trees were needed to accurately represent the tree-ring isotopic records at sites in Finland. We

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

first identified all white spruce trees from our database with both a pith date prior to 1775 and a collected cross-section. Site location, tree age, stem circumference, ring size, lack of compression wood and injuries (decay) were all criteria used to select the final set of eight trees for isotopic analysis. Individual trees were thus selected from seven sites with a mean (± standard deviation) intersite distance of 25.9 ± 5.4 km (Fig. 1). The mean (± standard deviation) establishment date of the selected trees was 1707 ± 37 yr. All white spruce trees had normal growth form and were not Krummholz. 2.3. Ring-width chronology All eight cross-sections were air dried and sanded with progressively finer sandpaper to reveal the individual cells of each ring. Once sanded, the samples were visually crossdated using a technique similar to that described by Yamaguchi (1991). A VELMEX UniSlide stage micrometer interfaced with a computer was used to measure the ring widths of each series to a precision of 0.01 mm. Two radii from each of the eight selected trees were measured for ring width. Visual crossdating of the wood samples and accuracy of the measurement was validated using COFECHA (Holmes, 1983). A standardized chronology was constructed using the program ARSTAN (Cook and Holmes, 1986). In ARSTAN, a spline function giving a 50% frequency response of 60 yr (Cook and Peters, 1981) was applied to each measurement series to remove the age/size related trend and other low-frequency trends presumed unrelated to climate. Standardization of the ring-width measurement series assured that the mean and the variance of all the series were made comparable. A residual chronology was also created in the same manner as the standard one except that the series were averaged using residuals from autoregressive modeling of the standardized measurement series (Cook and Holmes, 1986). This resulted in a chronology with a strong common signal and without persistence. The presence of autocorrelation reduces the effective number of independent observations and thus reduces the degrees of freedom used to determine the confidence in estimates of correlation coefficients (Monserud and Marshall, 2001). Along with general statistics describing the chronology, the expressed population signal (EPS) statistic was calculated. This statistic is an indicator of chronology reliability. It measures how well the chronology compares to a theoretical population chronology based on an infinite number of trees (Wigley et al., 1984). The statistic ranges from 0.0 to 1.0, i.e., from no agreement to 1.0 for perfect agreement with the theoretical population chronology. 2.4. Stable-carbon isotopes The carbon isotope ratios of alpha-cellulose were analyzed from eight white spruce trees. The previously crossdated and measured cross-sections were further reduced in size. From each cross-section, four orthogonal radial sections 5 mm width and 5 mm thick were cut using a band saw. These sections were sanded with progressively finer sandpaper up to 400 grit. Each section was vacuumed between sandings to remove wood dust

91

and sandpaper was cleaned between trees. The resulting 32 radial sections were crossdated once more using a dissecting pin to identify decades, half-century and century. For each tree, the initial growth period (innermost rings) was eliminated from analysis to minimize the “juvenile effect” (Bert et al., 1997; Robertson et al., 1997). This corresponded to the first 25– 50 years for six well preserved cross-sections but N 50 years for two sections presenting decay and suppression in the innermost portion. In each radial section, individual annual tree rings were carefully separated with a dissecting scalpel under a binocular microscope. The annual shavings from the four radii of each tree were separately pooled and put into a microcentrifuge tube previously labelled with tree number and year. Pooling assured that we had enough wood to analyze and that we get a representative yearly δ13C value (Leavitt and Long, 1984). For years with extremely narrow rings only, extra wood was obtained from a fifth ring (radial section) to assure sufficient material for carbon isotope (and eventually enough for oxygen isotope) analysis. Variation of isotopes ratios along the circumference of a single ring was shown to be lower than that among trees (Leavitt and Long, 1984) and adding a fifth ring when needed should have had a negligible impact on our results. Individual tree rings were thus separated in all eight trees for approximately 250 years. No attempts were made to separate earlywood from latewood. The white spruce showed limited radial growth, rendering partitioning of earlywoodlatewood almost impossible. In Norway spruce (Picea abies (L.) Karst), Schleser et al. (1999) observed little qualitative differences in δ13C values obtained from earlywood, latewood or complete ring. After ring separation, close to 2000 wood samples were milled finely enough to pass through a 20 mesh screen. In this study, largely because of budget limitations, we ran individual tree isotopic analysis on every fifth ring (2000, 1995, … 1750). For the intermediate years, we pooled approximately equal mass from each tree determined by the mass from the tree contributing the least amount in each year. This avoided producing a mean mass-weighted δ13C value (Borella et al., 1998) and equally weighted slow and fast growing trees. The individual isotopic analysis of 396 rings (between 45 to 51 per tree) allowed us to compare tree-to-tree variability and to place confidence limits around mean values (McCarroll and Loader, 2004). All wood samples (individual rings and pooled ones) were reduced to alpha-cellulose using the chemical extraction methods described in Leavitt and Danzer (1993). Each wood sample was transferred and sealed in previously labelled acidand solute-resistant digestion pouches (each containing eight compartments). Using two Soxhlet apparatus (128 samples), oils and resins were removed using a 24-hour extraction regime using a 2:1 toluene:ethanol mixture followed by 100% ethanol. The lignin component was then removed through oxidation in an acidified sodium chlorite solution for approximately 40 h to yield holocellulose. The hemicellulose component was further removed through reaction with sodium hydroxide (17.0 to 18.5% w/v NaOH) for 1 h with constant stirring to produce homogeneous alpha-cellulose after which samples were washed

92

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

thoroughly with deionized water over a 3–4 h period. The same procedure was repeated with a bath containing glacial acetic acid and samples were dried prior to analysis. For analysis, about 0.8 mg of alpha-cellulose was weighed from each sample and sealed in tin foil capsules. To estimate the reproducibility of δ13C measurements, cellulose samples were reanalyzed for a series of samples. In each analytical run, the first three samples were from an acetanilide standard of known isotopic composition followed by white spruce samples sequenced with another acetanilide standard every 10 rings. The alpha-cellulose was combusted online and the resulting CO2 was analyzed on a stable-isotope mass-spectrometer. The δ 13 C values are expressed in per mil versus the Vienna PeeDee belemnite (VPDB) standard. All isotope analysis took place at the Environmental Isotope Laboratory, Department of Geosciences, University of Arizona. The overall analytical precision of the isotope measurement was 0.11‰ (n = 105 standards). Since industrialisation, burning of 13C-depleted fossil fuel has caused a depletion of atmospheric δ13C of about 1.5‰ (McCarroll and Loader, 2006). To correct for the decline in δ13C of the atmosphere since about AD1850, we followed the methods outlined by McCarroll and Loader (2006). First, a δ13C residual chronology was calculated using the same approach as for ring width. A spline function giving a 50% frequency response of 60 yr was used to transformed δ13C values in δ13C indices. Statistical detrending has been criticized because it removes not only the direct effect of declining δ13C of the atmosphere but also any physiological response to increased atmospheric CO2 concentrations, and any change in climate that has occurred as a result of the CO2 rise (McCarroll and Loader, 2006). In order to keep low-frequency variation, we also mathematically removed the changes in atmospheric δ13C and expressed tree-ring δ13C values relative to a pre-industrial reference value of − 6.4‰ as suggested by McCarroll and Loader (2006). We used the high-precision record of atmospheric δ13C from Francey et al. (1999) to correct our data. This low-frequency chronology will be referred to as the correctedδ13C chronology. 2.5. Analysis of climate association The associations between ring-width and δ13C residual chronologies and climate were analyzed mainly using correlation analyses (Fritts, 1976). Mean monthly temperature (1929– 2000) and total monthly precipitation (1932–2000) data sets from Vincent and Gullett (1999) and Mekis and Hogg (1999), respectively, were used in these analyses (station 5060600, Churchill). The Canadian Drought Code (CDC) component of the Canadian Forest Fire Behavior System (Van Wagner, 1987; Girardin et al., 2004) was also used as a potential indicator of moisture stress. The CDC is a daily indicator of summertime moisture in deep organic layers in boreal conifer stands (Van Wagner, 1987). The CDC is a slow-drying index with a time lag estimated at 52 days. Moisture losses are the result of daily evaporation and transpiration, while daily precipitation accounts for moisture gains. Evaporation and transpiration losses are first estimated by calculating maximum potential

evapotranspiration based on daily temperature and seasonal day length. Second, the maximum potential evapotranspiration value is scaled by the available soil moisture to reflect the fact that as soil moisture decreases, evaporation is increasingly more difficult. The maximum water holding capacity of the CDC is 100 mm for a layer with a bulk density of about 18 cm deep and 25 kg/m 2 in dry weight, which amounts to approximately 400% of water per unit of mass. Daily CDC values were computed using daily maximum temperature and daily precipitation data from the above stations. The computation was conducted following the procedure of Van Wagner (1987) and mean monthly CDC averages were produced. Additional data from the Churchill A meteorological station (1957–2000) included total monthly snowfall, mean monthly minimum and maximum temperature, relative humidity, and dew point temperature. Both ring-width and δ13C residual chronologies were used in these climatic analyses and the climatic variables covered a 17month period from May of the year prior to (t− 1) ring formation to September of the year (t) of ring formation. We also calculated correlations with historical ice records from the Hudson's Bay Company's Archives and compiled by Moodie and Catchpole (1975). This data set provided freeze-up and break-up dates for the estuaries on Hudson Bay and was developed for the period 1714–1871 from analysis of historical document contents. Data relevant to Prince of Wales Fort near Churchill were extracted and used to calculate the ice-free period. The ice-free period was calculated by taking the date of complete ice freeze-up and subtracting the date of ice break-up for available years. This provided an independent data set on which to test the association between ring-width/δ13C residual chronologies and climate. In addition, we created four categories using the indices from both ring-width and δ13C residual chronologies and tested if any differed in terms of seasonalized mean temperature, total precipitation and drought code using an analysis of variance. The categories were 1: below average δ13C and ring-width indices, 2: below average δ13C indices and above average ringwidth indices, 3: above average δ13C indices and below average ring-width indices and, 4: above average ring-width and δ13C indices. We also used a Pearson Chi-Square analysis to test the distribution of both below- and above-average ring-width and δ 13 C indices in relation to categorized summer climate variables. The summer climate (June–July–August) was categorized as following 1: cool and dry, 2: cool and wet, 3: warm and wet and 4: warm and dry. These categories were determined using the averaged mean summer temperature and total precipitation. Program SYSTAT (Ver 11) was used for all of the climatic analysis (SYSTAT 2004). 3. Results 3.1. Ring-width and δ13C values No strong consistent association was observed between fifthyear ring-width and corresponding δ13C measurements (uncorrected and corrected) for the eight white spruce trees (Fig. 3).

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

93

Fig. 3. Ring-width versus δ13C measurements for each of the eight Picea glauca trees. The grey circles represented uncorrected δ13C measurements whereas the empty black circles represent δ13C values that were corrected for the depletion of δ13C in the atmosphere since 1850 (McCarroll and Loader, 2006). The two lines represent linear regressions calculated with each set of data (uncorrected δ13C: grey dash line and corrected δ13C: black line) and are indicative of the general trend in the data. The r2 related to the ring width-corrected δ13C measurements is indicated and coefficients followed by a star are significant at the 95% level.

Some individual scatter plots suggest, however, that δ13C could increase with ring width but in a non-linear way. The absence of a consistent relationship among trees suggests individualistic tree responses. The chronology statistics indicate that white spruce trees grew little in diameter with a mean annual ringwidth less than 0.5 mm (Fig. 3 and Table 1). The ring-width residual chronology developed in this study was strongly correlated (r = 0.83, p b 0.0001; n = 251) with that of Girardin et al. (2005), which had a much greater sample depth. Interestingly, the common signal statistics for the δ13C chronology (the every fifth-year measurements) were higher than for both ringwidth chronologies (fifth-year and continuous measurements; Table 1). The common signal in the annual ring-width chronology (44 continuous measurements) was also higher than in that developed from fifth-year ring width (44 fifth-year measurements). This is illustrated by a higher interseries correlation, common percent variance and EPS value in the former (Table 1). This result shows that the ring-width chronology developed from fifth-year measurements may

underestimate the common signal when compared to continuous ring-width chronology. It also suggests that the common signal reported for the δ 13 C chronology from fifth-year measurements may be underestimated when compared to that of other studies using continuous annual sequence of δ13C measurements. The δ13C chronology developed from fifth-year measurements (Fig. 4a) indicates variations in δ13C ranging from − 22.2 to − 26.4‰ (Table 1). Many trees presented a general decrease in δ13C approximately starting after 1875 (Fig. 4a). The continuous δ13C chronology shows variations ranging from − 22.60 to − 25.30‰ (Table 1) and revealed more detailed lowfrequency variation (Fig. 4b). A more or less continuous decrease in δ13C measurements was observed to start prior to the depletion of atmosphere δ13C associated with industrialisation. The highest δ13C measurements occurred in the 1750–60s, a period also characterized by higher variance in the data. An abrupt decrease in δ13C measurements was observed in the late 1810s and lasted until the 1850s. When the long-term decline in

94

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

Table 1 General statistics for both the ring-width and δ13C chronologies

Mean

a

Median a Standard deviation a Minimum a

direction of change as indicated by the Gleichläufigkeit measure (Fig. 5).

Ring-width (fifth year)

Ring-width (continuous)

δ13C δ13C (fifth year) (continuous)

0.34 mm (n = 393) 0.28 mm (n = 393) 0.22 (n = 393)

0.34 mm (n = 3863) 0.27 mm (n = 3863) 0.25 (n = 3863) 0.03 mm (n = 3863) 2.03 mm (n = 3863) 0.41 (n = 356) 42.47

− 24.0‰ (n = 396) − 23.9‰ (n = 396) 0.94‰ (n = 396) − 26.4‰ (n = 396) − 22.2‰ (n = 396) − 0.01 (n = 51) 46.25

− 24.0‰ (n = 251) − 24.0‰ (n = 251) 0.52‰ (n = 251) − 25.3‰ (n = 251) − 22.6‰ (n = 251) 0.19 (n = 251) N.A.

0.82 0.37

0.83 0.38

N.A. N.A.

0.05 mm (n = 393) Maximum a 1.17 mm (n = 393) Autocorrelation 0.30 1st order b (n = 51) Variance in PC-1 30.13 (%) c EPS signal c 0.64 0.18 Interseries correlation c

The continuous ring-width chronology covers the period 1645–2000 whereas the continuous δ13C chronology covers the period 1750–2000. Both fifth-year chronologies were developed from individual ring measurements (2000, 1995, … 1750). a Calculated from untransformed data (raw data). b Calculated from the standardized chronology. c Calculated from the residual chronology (n = 44 for all chronologies) and the common period for the continuous ring-width chronology was 1957–2000.

δ13C associated with industrialisation is removed (Fig. 4c), the δ13C chronology shows more mid-frequency fluctuations and a rise in δ13C measurements is observed toward the end of the 20th century. The residual chronologies (δ13C and ring-width) developed after standardizing the data with a 60-yr spline were weakly correlated despite greater year-to-year agreement in

3.2. Ring-width and δ13C association with climate 3.2.1. Correlation analyses The Pearson correlation analyses revealed that both ringwidth and δ 13 C residual chronologies were significantly associated with climate parameters (Fig. 6). However, no strong correlations were observed between either ring-width or δ13C residual chronologies and monthly climate variables indicative of moisture availability such as total precipitation, drought code and relative humidity. The absence of significant negative associations with variables related to evapotranspiration suggests that stomatal closure due to moisture stress may not be the main control regulating 13C/12C ratio. Temperatures (and dew point temperature) were the most strongly correlated variable to both residual chronologies. In particular, July temperatures (minimum, mean, maximum and dew point) were positively correlated with both chronologies suggesting that warm July promotes xylem production and photosynthesis. The dew point (June–July–August), which represents the temperature at which the atmosphere is saturated in water, was also strongly correlated to δ13C and to a lesser extent to ring-width indices. Interestingly, the ring-width residual chronology was more strongly correlated to early summer conditions (June–July) whereas the δ13C chronology better reflected the condition in July–August (Fig. 6). In contrast May minimum temperatures were negatively correlated with both ring-width and δ13C residual chronologies. Chronology-specific correlations were also observed (Fig. 6). The δ13C residual chronology had negative correlations with total precipitation in May and snowfall in August. The negative

Fig. 4. Carbon isotope chronologies. a) Individual fifth-year δ13C measurement for each of the eight Picea glauca trees. b) Continuous δ13C measurement for the period 1750–2000. The 95% confidence interval was calculated from the fifth-year δ13C measurements presented in (a). The upper dark grey line represents the decline in δ13C of the atmosphere as per McCarroll and Loader (2004). Grey circles in (a) and (b) represent duplicate measurements. c) Continuous δ13C measurements for the period 1750–2000 after correction for the depletion of δ13C in the atmosphere since 1850 (McCarroll and Loader, 2006).

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

95

Fig. 5. Ring-width (black line) and δ13C (grey line) residual chronologies for the period 1750–2000. The Pearson correlation coefficient between the two residual chronologies for the period 1750–2000 is 0.198 (p = 0.002) and the Gleichläufigkeit (Glk) measure is 58.8%.

correlation with August snowfall supports the idea that the δ13C chronology reflects the length of the growing season. Under the stomatal conductance hypothesis, a negative association between relative humidity and δ13C would be expected except that our results show a positive correlation between both

variables in months from the year prior to ring formation. The ring-width residual chronology showed a negative correlation with October snowfall (inverted with temperatures) suggesting that early snow accumulation may negatively affect growth in the following year by influencing the early growing season

Fig. 6. Pearson correlation coefficients between both the δ13C (left panel) and ring-width (right panel) residual chronologies and the monthly climatic variables. From top to bottom, the climate variables are Min Temp: mean of minimum temperature, Mean Temp: mean of average temperature, Max Temp: mean of maximum temperature, CDC: Canadian drought code, Snow: total snow precipitation, Prec: total precipitation, DP: mean dew point and RH: mean relative humidity. The months are from May of the year prior to ring formation (large caps) to September of the year of ring formation (small caps). Correlation coefficient with the previous summer (June–July–August, P) and the current summer (c) conditions are also presented. Threshold for significant correlations (p b 0.05) is indicated by the dotted lines. The period of analysis was 1929–2000 for the temperature (minimum, average, maximum), 1933–2000 for precipitation (snow and total), 1933–1999 for CDC, and 1953– 2000 for both DP and RH.

96

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

Fig. 7. Linear regression between both the ring-width (data represented by grey circles and regression line) and δ13C (data represented by black circles and regression by broken line) residual chronologies and the ice records from the Hudson's Bay Company's Archives (source of data: Moodie and Catchpole, 1975).

conditions. The ring-width residual chronology was also correlated to more variables from the year preceding ring formation than the δ13C residual chronology (Fig. 6). The relationships between both ring-width and δ13C residual chronologies and seasonalized climate data indicated that only mean summer temperatures of the year of ring formation were significantly and positively associated with both ring-width and δ13C (Fig. 6). The dew point temperature was positively related to δ13C chronology. Highly significant linear associations were also observed between the δ13C residual chronology and 1) first break-up date (r = 0.47, p = 0.004, n = 37), 2) complete freeze-up date (r = 0.40, p = 0.030, n = 29), and 3) length of the ice-free season (r = 0. 61, p b 0.0001, n = 29) at Prince of Wales Fort near Churchill using available data from 1750 to 1800 (Fig. 7). Early break-up, late freeze-up and thus long ice-free season were all associated with increased δ13C in alpha-cellulose. No significant associations were, however, observed between these ice records and the ring-width residual chronology (all p N 0.320).

with climate, we tested for differences in climate for years when: 1) both chronologies had below average indices, 2) δ13C indices were below average (inverse for ring-width indices), 3) δ13C indices were above average (inverse for ring-width indices) and 4) both chronologies showed above average indices. The analysis of variance failed to detect any statistical differences among the four categories in regards to seasonalized total precipitation and drought code (Table 2). A unique significant difference with summer temperature was observed. Years with low ring-width and δ13C indices were associated with cool summers (9.5 °C) compared to years with high ringwith and δ13C indices being associated with warmer summers (10.6 °C). These results again stress the dominant control of temperature over ring-width formation and carbon assimilation. A less significant difference was also observed with winter precipitation (Table 2). Years with high ring-width and low δ13C indices registered less winter precipitation than years with low ring-width indices (low and high δ13C indices).

3.2.2. Analysis of variance In order to better understand the interrelationship between ring-width and δ13C residual chronologies and their association

3.2.3. Pearson Chi-Square analysis Taking another perspective, the Pearson Chi-Square statistic was calculated to see how years with low and high δ13C (ring-

Table 2 Mean and standard deviation of climate variables for the tree-ring residual chronology categories: 1) below average δ13C and ring-width indices, 2: below average δ13C indices and above average ring-width indices, 3: above average δ13C indices and below average ring-width indices and, 4: above average ring-width and δ13C indices Category 1

Category 2

Category 3

Category 4

p value

Mean temperature (°C) Fall (Sep–Nov) Winter (Dec–Feb) Spring (Mar–May) Summer (Jun–Aug)

− 2.5 ± 1.6 − 24.6 ± 1.9 − 9.9 ± 1.6 9.5 ± 1.1a

− 2.8 ± 2.0 − 26.4 ± 2.6 − 11.1 ± 2.4 9.7 ± 1.1ab

−3.0 ± 2.0 −24.6 ± 2.0 −10.8 ± 1.7 10.2 ± 1.2ab

− 3.1 ± 1.5 − 25.2 ± 1.8 − 11.0 ± 1.4 10.6 ± 1.0b

0.724 0.091 0.127 0.004

Total precipitation (mm) Fall (Sep–Nov) Winter (Dec–Feb) Spring (Mar–May) Summer (Jun–Aug)

180.1 ± 65.1 90.1 ± 33.1b 105.5 ± 52.7 175.9 ± 52.7

168.1 ± 52.2 59.1 ± 29.8a 76.4 ± 45.7 179.7 ± 67.8

196.7 ± 57.5 91.0 ± 29.5b 103.4 ± 54.2 185.5 ± 51.9

169.7 ± 50.5 83.6 ± 28.2ab 101.7 ± 50.0 169.8 ± 55.7

0.504 0.030 0.425 0.857

Canadian Drought Code Fall (Sep–Nov) Winter (Dec–Feb) Spring (Mar–May) Summer (Jun–Aug)

132.7 ± 56.6 N.A. 4.8 ± 5.6 154.5 ± 49.0

153.0 ± 49.1 N.A. 3.61 ± 4.42 154.5 ± 55.6

139.0 ± 78.7 N.A. 4.6 ± 4.9 156.0 ± 47.2

136.3 ± 56.5 N.A. 2.1 ± 3.2 158.9 ± 53.2

0.826 N.A. 0.253 0.992

Significant differences (p b 0.05) identified by the ANOVA appear in bold. For these variables, the different letters indicate significant differences (p b 0.05) among categories after a Tuckey post-hoc test.

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

width) indices were distributed according to years with 1) warm-dry, 2) warm-wet, 3) cool-dry and 4) cool-wet summers. If stomatal conductance is a main factor controlling δ13C, higher δ13C indices would be expected in years with warm-dry summers than in years with warm-wet summer. The results revealed that in warm-dry and warm-wet summers, higher-thanexpected δ13C indices were observed, and the opposite was true for cool-dry and cool-wet summers (Pearson Chi-Square significant at p = 0.001, n = 68). Further analysis revealed no statistical differences between wet and warm conditions. During cool (dry or wet) years, a depletion of 13C was observed and during warm (dry or wet) years, 13C enrichment was observed. These results suggest that during warm years, alpha-cellulose is enriched in 13C but apparently not because of moisture stress. The Pearson Chi-Square test was not significant when used with ring-width indices (p = 0.34, n = 68) suggesting less specificity to the overall growing season condition. 4. Discussion 4.1. Chronologies and signal strength Our results showed that ring-width and δ13C measurements were not highly associated. In the literature, no common relationship exists between ring-width and δ13C measurements. Brooks et al. (1998) found no significant correlation (p = 0.82) between δ13C and ring width in P. banksiana. McCarroll et al. (2003) found a positive (p b 0.01) correlation between δ13C and ring width as well as with maximum density in P. sylvestris. The absence of a consistent relationship between ring-width and δ13C measurements suggests independence between the two processes and individual tree response. Similarly the absence of a strong correlation between both ring-width and δ13C residual chronologies suggests that the temporal windows in which radial growth and carbon fixation take place are not totally overlapping. The signal strength in the white spruce δ13C residual chronology was higher than in the ring-width residual chronology. The stronger signal in stable-carbon isotope series and the need for fewer trees to attain a given signal threshold compared to ring-width series appears to be a characteristic of tree-ring isotope studies (Robertson et al., 1997; McCarroll and Pawellek, 1998; Gagen et al., 2004; McCarroll and Loader, 2004). It has been proposed that a sample size leading to an EPS value of 0.85 may be minimally acceptable in tree-ring isotope studies (McCarroll and Loader, 2004). For oak, EPS values above 0.85 were reported for a δ13C chronology made up of four to five trees (Robertson et al., 1997). Near the northern distribution limit of P. sylvestris, McCarroll and Pawellek (1998) observed EPS values of 0.89 to 0.95 with chronologies developed from five trees, and two to four trees were usually sufficient to get an EPS value of 0.85. Stronger results were obtained after correcting for atmospheric decline in δ13C and removal of juvenile effect (McCarroll and Pawellek, 2001). Our current assessment of the common signal in white spruce δ13C residual chronology suggests that sampling eight to ten trees may be necessary to reach an EPS value of 0.85.

97

Based on the comparison made of the δ13C and ring-width chronologies, we speculate, however, that the common signal evaluated from fifth-year measurements may not allow adequate quantification/comparison of the EPS. The ring-width residual chronology developed in this study was significantly correlated with that used in Girardin et al. (2005) and also revealed a similar association to climate. The statistical properties of the ring-width residual chronology developed using fifth-year measurements (1750, 1755 … 2000) differed, however, from that developed using continuous annual measurements for the period 1957– 2000. Both inter-tree correlation and EPS values substantially rose when continuous ring-width measurements were used. If we assume the same properties held for the δ13 C fifth-year measurements, the statistic presented may underestimate the common signal. It is thus possible that a representative δ13C chronology for white spruce could be developed from fewer than eight trees and that higher EPS values could be obtained from continuous δ13C measurements. The white spruce δ 13 C series derived from fifth-year measurements was characterized by large tree-to-tree variation. It is possible that at the forest-tundra transition near Churchill no single environmental factor dominates isotopic fractionation thus lowering the common signal among trees. Variations in δ13C measurements could be related to differences in microsite or in tree and stand history. The common signal may have been diluted by the large area in which trees were sampled (white spruce trees came from seven sites). Changes in light conditions associated with stand dynamics (canopy closure) or damage caused by snow could also influence the photosynthetic capacity of trees and contribute to lowering of the common signal. Few studies have, however, examined the influence of disturbances and/or stand dynamics on δ13C variations in trees. Leavitt and Long (1987) examined the impact of forest fire, selective cutting and insect defoliation with δ13C of 5-yr ring groups and concluded that the impact of these disturbances was relatively minor. McDowell et al. (2003), however, analyzed δ13C in individual rings and found evidence that after thinning the δ13C decreases, related to more available soil moisture and increased stomatal conductance. The range of δ13C measurements observed in white spruce corresponded to that observed in other isotope studies using various species. The reported range in δ13C measurements among individual trees at a site has been typically on the order of 2–3‰ (Leavitt and Long, 1984, 1988; Leavitt, 1993). In P. abies, Treydte et al. (2001) observed a maximum isotope difference of 3.1‰ (−24.9 and −21.8‰) between individuals. In white spruce, Sun et al. (1996) reported that there was a significant variation in δ13C measurement in needle (1.6–2.0‰) owing to genetic variations among trees. Despite more variable low-frequency fluctuation, the overall high-frequency variation was similar among white spruce trees. The continuous decline in δ13C observed since about 1750 (Fig. 4 b) was also unusual and superposed on the well documented decline of the atmospheric δ13C induced by fossil fuel burning since about 1850 AD (Treydte et al., 2001; McCarroll and Loader, 2004). Further evidence that the δ13C signal in white spruce is representative of a larger area comes from the strong correlation

98

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

observed between our δ13C uncorrected chronology and that developed for jack pine about 400 km south of Churchill, Manitoba (Brooks et al., 1998; Ehleringer et al., 1998). The firstdifferences derived from each δ13C chronology were significantly correlated for the period 1962–1982 (Spearman correlation 0.618, p = 0.003, n = 21). The Gleichläufigkeit value of 86.8% also indicates a strong year-to-year agreement in direction of change. The use of first-difference has been advocated by Schleser et al. (1999) to minimize problems related to changing atmospheric CO2 concentration and δ13C values. It should be noted that the jack pine δ13C chronology was offset soon after 1982, perhaps linked to dating issues or some real environmental lag, so the portion 1983–1992 was not included. The low-frequency variations observed in the white spruce δ13C corrected-chronology suggest, assuming it is related to temperature, that warm conditions may have prevailed from mid-1700s to early 1800s. This was followed by a cooling trend and subsequent warming from the 1860s to the 1920s. A cooling trend was again observed from the 1930s to the 1960s followed thereafter by continuous warming conditions. Using tree-ring data from the Churchill area, Scott et al. (1988) argued that largescale changes in atmospheric circulation occurred in the late 1700s. The authors speculated that during the period 1760–1820 the summer position of the Arctic Front moved north (that is toward or beyond Churchill) bringing more favourable growth conditions. They referred to this period as a “weak interruption of the Little Ice Age from the 1760s to 1820s”. In a reconstruction of summer temperatures using white spruce near the northern tree-line on the eastern-side of the Hudson Bay, Jacoby et al. (1988) reported an abrupt cooling of the summer starting in 1810s and its persistence for three decades. “This abrupt cooling in the early 1800's is a ubiquitous feature in highlatitude North American trees” (Jacoby et al., 1988). The Briffa et al. (1994) reconstruction of summer temperature in the Mackenzie Valley revealed that warm decades occurred in the 1780s and 1790s and cold summers in the 1810s and most of the 1830s and 1880s. 4.2. δ13C and climate The δ13C at the site and species levels reflects the relative contributions of both stomatal conductance and photosynthesis rate (McCarroll and Loader, 2004, 2006). Warren et al. (2001) compiled data from conifers worldwide and concluded that δ13C was most probably an indicator of drought stress in seasonally dry climates. They reported that variation in irradiance and N concentration could have as large an affect on discrimination as water availability. Saurer et al. (2004) hypothesized that the typical association between moisture and carbon isotope discrimination observed in regions with warm-dry climate may not hold for northern forests. In the forest-tundra transition of the Hudson Bay Lowland, we found the dominant controls over δ13C in tree rings from white spruce to be climatic factors typically associated with photosynthetic rate. We speculate that the enrichment in 13C observed in white spruce trees during warm summers was caused by increased photosynthesis and overall 13C uptake. A greater control of photosynthesis relative to stomatal

conductance was also observed in P. sylvestris growing at sites near the European timberline (McCarroll and Pawellek, 2001). As synthesized by McCarroll and Pawellek (2001), δ13C in tree rings should reflect air humidity and precipitation at dry sites because of the dominant control exerted by stomatal conductance, whereas it should reflect temperature and amount of sunlight at sites not limited by moisture stress because of the dominant control exerted by photosynthetic rate. In Churchill, no variables indicative of water deficit (drought code, precipitation, relative humidity) were found to correlate to the δ13C residual chronology suggesting that stomatal conductance (moisture stress) is not a major factor controlling δ13C in tree rings. Relative humidity has been reported to be a good indicator of the potential for evapotranspiration and to play a major role in controlling stomatal conductance (McCarroll and Loader, 2004). Low atmospheric humidity is known to increase the vapour pressure deficit around the leaf and lower external water vapour is associated with high rate of water loss. In the Churchill area, the cooling influence of the Hudson Bay may be important in maintaining higher summer moisture levels enabling trees to persist on the forest-tundra (Scott and Hansell, 2002). The cold maritime air originating over Hudson Bay has a smaller vapour pressure deficit compared to inland air thus inhibiting evapotranspiration (Rouse et al., 1992). During summers, cooler temperatures and greater cloud cover often occurs in Churchill as a result of frequent sea breezes (Vickers et al., 2001). The distance to which cold, moist marine air extends inland is unknown (Timoney, 1995) but our results suggest that the seven sites sampled are probably under this influence. In Churchill, the main correlation observed between δ13C in white spruce and climate was with temperature of the summer months and especially July–August. Noticeable was the strong correlation observed between δ13 C index and dew point temperature from June to August. In northern Europe, July conditions were also predominantly correlated with δ13C in P. sylvestris (Sonninen and Jungner, 1995; McCarroll and Pawellek, 1998). If the collinear nature of temperature and hours of sunshine reported by McCarroll et al. (2003) applies to the Churchill region, we can speculate that the association with temperature may also reflect the influence of photon flux on photosynthetic rate and therefore on the concentration of CO2 in the stomatal chambers. In addition, our results indicated that not only warm summers but also longer growing seasons were positively correlated to δ 13 C in white spruce. This was highlighted by the strong correlation observed between the ice-records data derived from the Hudson's Bay Company's Archives (Moodie and Catchpole, 1975) and the δ13C residual chronology. These results suggest that during long ice-free season (early break-up, late freeze-up) the uptake of 13C in alpha-cellulose increases in relation to 12C. This interpretation was also supported by the negative correlation observed between the δ13C residual chronology and August snowfall. 4.3. Ring width and climate In contrast to δ13C, ring-width indices were less integrative of the overall growing season climatic conditions. Our results

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

indicate that near the tree-line, the climatic window for radial growth is much narrower than that for carbon fixation. This was illustrated by, among others, the absence of a significant correlation between the ring-width residual chronology and the ice records when compared to δ13C. In contrast to Jacoby and Ulan (1982) who reported a positive correlation between first complete freeze-up date and white spruce radial growth the following year at a site about 8 km south of the estuary, we did not observe such an association (not presented) either using the data subset (1741–1764) used by Jacoby and Ulan (1982) or the entire data set available in Moodie and Catchpole (1975). The ring-width residual chronology was mainly correlated with early growing season June–July temperatures, and this agreed with both Jacoby and Ulan (1982) and Girardin et al. (2005). This distinction between δ 13 C and ring-width residual chronologies indicates that white spruce produces photosynthates over a longer period than June–July, i.e., at the time where most xylem cells are produced. The month of July in Churchill includes most of the growing season for white spruce (Scott et al., 1987). In a study of white spruce islands in the Churchill area, Scott and Hansell (2002) observed for the year 1998 that bud swell started on June 16 (June 24 in 1989) and that stem elongation continued until August 3 (July 26 in 1989). These data indicate that most of the xylem cell production in a year occurs in a brief period of time while secondary walls deposition may continue for a longer period. A short period of xylem production was also observed in white spruce trees growing in the boreal forest (Ko Heinrichs et al., 2007). Also noticeable in this study were the correlations observed between both residual chronologies (ring width and δ13C) and conditions either in October of the year prior to ring formation or May of the year of ring formation, two potentially determinant months when looking at snow accumulation (Fig. 2) and ice freeze-up or break-up. The observed positive association between ring width and October temperatures was also noted by both Jacoby and Ulan (1982) and Girardin et al. (2005). The same response was also found in Dryas integrifolia, a dwarf shrub growing in subarctic Manitoba (Au and Tardif, 2007). The negative association between ring width and October snowfall observed in this study as well as in D. integrifolia suggests that the onset of snow accumulation and its potential impact on soil conditions or timing of the growing season in the subsequent year may be an important component influencing radial growth. The late-spring conditions (warm and/or wet May) were also found to be correlated to either δ13C or ring-width residual chronologies. Similar responses were observed by both Girardin et al. (2005) and Au and Tardif (2007). This relationship is, however, less easily comprehensible. Girardin et al. (2005) speculated that these negative relationships may be associated with an early snowmelt at a time when the ground has not yet thawed. In spring, the melting of snow on the exposed tundra usually occurs in late May and usually three weeks later in the open forest (Rouse, 1982). In woodland, snowmelt is usually completed by mid-July and snowfall typically starts in midSeptember (Scott et al., 1993).

99

5. Conclusion This study presents the first δ13C tree-ring record for subarctic North America. Our results support the hypothesis that near the tree-line, photosynthesis is a major control determining stablecarbon isotope ratio in tree rings. No significant correlations were found between δ13C and ring width with variables controlling moisture availability, which suggests that stomatal conductance (drought) is not a major limiting factor for radial growth and photosynthesis. The complementary signal in each chronology indicates that xylem cells are produced early in the growing season whereas carbon fixation continues later into the growing season. The positive correlation of both chronologies with summer temperatures was consistent with results from the literature. Warm growing season temperature maximizes ring width (xylem production) and photosynthesis (uptake of 13C). The significant correlation observed between ice records independently developed from the Hudson's Bay Company's Archives and the δ 13 C residual chronology support the interpretation that low-frequency variations in δ13C are related to temperature and length of the growing season. Results indicated that both ring-width and δ13C measurements provide complementary information that could be used for long-term climatic reconstructions. Further analyses of the low-frequency fluctuations in the δ13C chronology will take place. Acknowledgements We thank J. Burns, C. Eastoe, M. Munro, I. Panyushkina (facilitated by NSF ESH Grant #ATM-0213696), R. Touchan and M. Salzer from the University of Arizona for their support during this project. Many thanks to field and/or laboratory assistants: E. Berglund, G. Rauh, J. Dubois, D. Wright, M. Thiessen, G. Shutler, J. Joa, D. Ko Heinrichs, B. Jones and helicopter pilots and to K. Monson. We thank Dr. D. McCarroll for making available the data from Table 2 from McCarroll and Loader (2004). We also thank Dr. M.-P. Girardin for his comments on an earlier version of the manuscript and for calculating the Canadian Drought Code. We also thank the two reviewers and the associate editor for their comments which improved the manuscript. This research was undertaken, in part, thanks to funding from the Canada Research Chairs Program. The Natural Sciences and Engineering Research Council of Canada, the Churchill Northern Research Centre, Parks Canada, Manitoba Climate Change Action Fund and the University of Winnipeg also supported this research. References Au, R., Tardif, J.C., 2007. Allometric relationships and dendroecology of the dwarf shrub Dryas integrifolia near Churchill, subarctic Manitoba. Canadian Journal of Botany 85 (6), 585–597. Barber, V., Juday, G., Finney, R., 2000. Reduced growth of Alaska white spruce in the twentieth century from temperature-induced drought stress. Nature 405, 668–672. Bert, D., Leavitt, S.W., Dupouey, J.-L., 1997. Variation of wood δ13C and wateruse efficiency of Abies alba during the last century. Ecology 78 (5), 1588–1596.

100

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101

Borella, S., Leuenberger, M., Saurer, M., Siegwolf, R., 1998. Reducing uncertainties in δ13C analysis of tree rings: pooling, milling, and cellulose extraction. Journal of Geophysical Research 103 (D16), 19519–19526. Briffa, K.R., Jones, P.D., Schweingruber, F.H., 1994. Summer temperatures across northern North America: regional reconstructions from 1760 using tree-ring densities. Journal of Geophysical Research 99 (D12), 25835–25844. Brook, R.K., 2001: Structure and dynamics of the vegetation in Wapusk National Park and the Cape Churchill Wildlife Management Area of Manitoba: Community and landscape scales. MSc Thesis, University of Manitoba. Winnipeg, Manitoba. Brooks, J.R., Flanagan, L.B., Ehleringer, J.R., 1998. Responses of boreal conifers to climate fluctuations: indications from tree-ring widths and carbon isotope analyses. Canadian Journal of Forest Research 28, 524–533. Canadian Centre for Climate Modelling and Analysis, 2005. Diagnostic plots from CRCM3.6. CCCma Online. http://www.cccma.bc.ec.gc.ca/diagnostics/crcm36/crcm36.shtml2005(Last modified: 2005-09-22, date accessed: 2006-11-02). Chhin, S., Wang, G.G., Tardif, J., 2004. Dendroclimatic analysis of white spruce at its southern limit of distribution in the Spruce Woods Provincial Park, Manitoba, Canada. Tree-Ring Research 60, 31–34. Clarke, K.E., Martini, I.P., Glooschenko, W.A., 1982. Sedimentary characteristics of the coastal environment of North Point. Ontario. Le Naturaliste Canadien 109, 385–397. Cook, E.R., Peters, K., 1981. The smoothing spline: a new approach to standardizing forest interior tree-ring width series for dendroclimatic studies. Tree-Ring Bulletin 41, 45–53. Cook, E.R., Holmes, R., 1986. Guide for computer program ARSTAN. Laboratory of Tree-Ring Research. Tucson, Arizona, University of Arizona. D'Arrigo, R.D., Jacoby, G.C., 1993. Secular trends in high northern latitude temperature reconstructions based on tree rings. Climatic Change 25 (2), 163–177. D'Arrigo, R.D., Jacoby, G.C., 1999. Northern North American tree-ring evidence for regional temperature changes after major volcanic events. Climatic Change 41 (1), 1–15. Dodd, J.P., Patterson, W.P., Holmden, C., Brasseur, J.M., 2008. Robotic micromilling of tree-rings: a new tool for obtaining subseasonal environmental isotope records. Chemical Geology 252, 21–30 (this volume). doi:10.1016/j.chemgeo.2008.01.021. Dredge, L.A., Nixon, F.M., 1992. Glacial and Environmental Geology of Northeastern Manitoba. . Memoir, vol. 432. Geological Survey of Canada. 80 pp. Environment Canada, 2003. Canadian Climate Normals 1971–2000. Canadian Climate Program. Environment Canada. Atmospheric Environment Service, Downsview, Ontario. Ehleringer, J., Brooks, J.R., Flanagan, L., 1998. BOREAS TE-05 Tree Ring and Carbon Isotope Ratio Data. Data Set. . Available on-line [http://www.daac. ornl.gov] from Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, Tennessee, U.S.A. (revision date: June 16, 2004; date accessed: 2006-05-05). Farquhar, G.D., Ehleringer, J.R., Hubick, K.T., 1989. Carbon isotope discrimination and photosynthesis. Annual Review of Plant Physiology and Plant Molecular Biology 40, 503–537. Francey, R.J., Allison, C.E., Etheridge, D.M., Trudinger, C.M., Enting, I.G., Leuenberger, M., Langenfelds, R.L., Michel, E., Steele, L.P., 1999. A 1000year high precision record of δ13C in atmospheric CO2. Tellus 51B, 170–193. Fritts, H.C., 1976. Tree Rings and Climate. Academic Press, New York. Gagen, M., McCarroll, D., Edouard, J.-L., 2004. Latewood width, maximum density and stable carbon isotope ratios of pine as palaeoclimate indicators in a dry sub-alpine environment in the southern French Alps. Arctic, Antarctic and Alpine Research 36 (2), 166–171. Gagnon, A.S., Gough, W.A., 2005. Trends in the dates of ice freeze-up and breakup over Hudson Bay, Canada. Arctic 58 (4), 370–382. Gajewski, K., Atkinson, D.A., 2003. Climatic change in northern Canada. Environmental Reviews 11 (2), 69–102. Girardin, M.P., Tardif, J., 2005. Sensitivity of tree growth to the atmospheric vertical profile in the boreal plains of Manitoba, Canada. Canadian Journal of Forest Research 35, 48–64.

Girardin, M.P., Tardif, J., Flannigan, M.D., Wotton, M.B., Bergeron, Y., 2004. Trends and periodicities in instrumental Canadian Drought Code monthly mean indexes, southern Canadian boreal forest. Canadian Journal of Forest Research 34, 103–109. Girardin, M.P., Berglund, E., Tardif, J.C., Monson, K., 2005. Radial growth of tamarack (Larix laricina) in the Churchill Area, Manitoba, Canada, in relation to climate and larch sawfly (Pristiphora erichsonii) herbivory. Arctic, Antarctic, and Alpine Research 37 (2), 206–217. Gough, W.A., Leung, A., 2002. Nature and fate of Hudson Bay permafrost. Regional Environmental Change 2 (4), 177–184. Gray, J., Thompson, P., 1977. Climatic information from 18O/16O analysis of cellulose, lignin and whole wood from tree rings. Nature 270, 708–709. Guiot, J., 1987. Reconstruction of seasonal temperatures in Central Canada since A.D. 1700 and detection of the 18.6- and 22-year signals. Climatic Change 10 (3), 249–268. Holmes, R.L., 1983. Computer assisted quality control in tree-ring dating and measurement. Tree-Ring Bulletin 43, 69–78. Hughes, M.K., 2002. Dendrochronology in climatology — the state of the art. Dendrochronologia 20 (1–2), 95–116. Jacoby, G.C., Ulan, L.D., 1982. Reconstruction of past ice conditions in a Hudson Bay estuary using tree-rings. Nature 298, 637–639. Jacoby, G.C., D'Arrigo, R.D., 1991. Global change and thermal history as recorded by Northern North American tree-ring data. In: Weller, G., Wilson, C.L., Severin, B.A.B. (Eds.), Proceedings of the International Conference on the Role of Polar Regions in Global Change. University of Alaska Fairbanks, pp. 559–605. Jacoby, G.C., Ivanciu, I.S., Ulan, L.D., 1988. A 263-year record of summer temperature for northern Quebec reconstructed from tree-ring data and evidence of a major climatic shift in the early 1800's. Palaeogeography, Palaeoclimatology, Palaeoecology 64, 69–78. Jones, P.D., Briffa, K.R., Schweingruber, F.H., 1995. Tree-ring evidence of the widespread effects of explosive volcanic eruptions. Geophysical Research Letters 22 (11), 1333–1336. Ko Heinrichs, D., Tardif, J., Bergeron, Y., 2007. Xylem production in six boreal tree species growing on an island in the boreal forest region of western Quebec, Canada. Canadian Journal of Botany 85 (5), 518–525. Leavitt, S.W., 1993. Seasonal 13C/12C changes in tree rings: species and site coherence, and a possible drought influence. Canadian Journal of Forest Research 23, 210–218. Leavitt, S.W., Long, A., 1984. Sampling strategy for stable carbon isotope analysis of tree rings in pine. Nature 311, 145–147. Leavitt, S.W., Long, A., 1987. Influence of site disturbance on δ13C isotopic time series from tree rings. Proceedings of the International Symposium on Ecological Aspects of Tree-Ring Analysis, U.S. Dept. of Energy/U.S. Environ. Protection Agency/U.S. Dept. of Agriculture, Publ. CONF8608144, pp. 119–129. Leavitt, S.W., Long, A., 1988. Intertree variability of d13C in tree rings. In: Rundel, P.W., Ehleringer, J.R., Nagy, K.A. (Eds.), Stable Isotopes in Ecological Research. Springer-Verlag, New York. Chapter 7. Leavitt, S.W., Long, A., 1989. Drought indicated in carbon- 13/carbon-12 ratios of southwestern tree rings. Water Resources Bulletin 25, 341–347. Leavitt, S.W., Danzer, S.R., 1993. Method for batch processing small wood samples to holocellulose for stable-carbon isotope analysis. Analytical Chemistry 65, 87–89. Leavitt, S.W., Wright, W.E., Long, A., 2002. Spatial expression of ENSO, drought and summer monsoon in seasonal δ13C of ponderosa pine tree rings in southern Arizona and New Mexico. Journal of Geophysical Research 107 (D18), 4349. doi:10.1029/2001JD001312. Lipp, J., Trimborn, P., Fritz, P., Moser, H., Becker, B., Frenzel, B., 1991. Stable isotopes in tree ring cellulose and climate change. Tellus 43B, 322–330. MacDonald, G.M., Szeicz, J.M., Claricoates, J., Dale, K.A., 1998. Response of the central Canadian treeline to recent climatic changes. Annals of the Association of American Geographers 88 (2), 183–208. McCarroll, D., Pawellek, F., 1998. Stable carbon isotope ratios of latewood cellulose in Pinus sylvestris from northern Finland: variability and signalstrength. The Holocene 8 (6), 675–684. McCarroll, D., Pawellek, F., 2001. Stable carbon isotope ratios of Pinus sylvestris from northern Finland and the potential for extracting a climate

J.C. Tardif et al. / Chemical Geology 252 (2008) 88–101 signal from long Fennoscandian chronologies. The Holocene 11, 517–526. McCarroll, D., Loader, N.J., 2004. Stable isotopes in tree rings. Quaternary Science Reviews 23, 771–801. McCarroll, D., Loader, N.J., 2006. Isotopes in tree rings. In: Leng, M.J. (Ed.), Isotopes in Palaeoenvironmental Research Series: Developments in Paleoenvironmental Research, 10, pp. 67–116. McCarroll, D., Jalkanen, R., Hicks, S., Tuovinen, M., Pawellek, F., Gagen, M., Eckstein, D., Schmitt, U., Autio, J., Heikkinen, O., 2003. Multi-proxy dendroclimatology: a pilot study in northern Finland. The Holocene 13 (6), 829–838. McDowell, N., Brooks, J.R., Fitzgerals, S.A., Bond, B.J., 2003. Carbon isotope discrimination and growth response of old Pinus ponderosa trees to stand density reductions. Plant, Cell and Environment 26, 631–644. Mekis, É., Hogg, W.D., 1999. Rehabilitation and analysis of Canadian daily precipitation time series. Atmosphere-Ocean 37, 53–85. Monserud, P.A., Marshall, J.D., 2001. Time-series analysis of δ13C from tree rings. I. Time trends and autocorrelation. Tree Physiology 21, 1087–1102. Monson, M.M.K., 2003. Fire history and secondary vegetation succession in the forest tundra near Churchill, Manitoba. MSc. thesis, University of Manitoba. 118 pp. Moodie, D.W., Catchpole, A.J.W., 1975. Environmental data from historical documents by content analysis. freeze-up and break-up of estuaries on Hudson Bay 1714–1871. Manitoba geographical studies, 5. University of Manitoba, Manitoba. Nienstaedt, H., Zasada, J.C., 1990. Picea glauca (Moench) Voss. In: Burns, R.M., Honkala, B.H. (Eds.), Silvics of North America Volume 1. Conifers. United States Department of Agriculture, Forest Service Agriculture Handbook 654, Washington, DC, pp. 204–226. Ritchie, J.C., 1962. A geobotanical survey of northern Manitoba. Arctic Institute of North American Technical Paper, vol. 9. Montreal, Quebec. Robertson, I., Rolfe, J., Switsur, V.R., Carter, A.H.C., Hall, M.A., Barker, A.C., Waterhouse, J.S., 1997. Signal strength and climate relationships in the 13 12 C/ C ratios of tree ring cellulose from oak in southwest Finland. Geophysical Research Letters 24 (12), 1487–1490. Rouse, W.R., 1982. Microclimate in low arctic tundra and forest at Churchill, Manitoba. Proceedings Fourth Canadian Permafrost Conference, pp. 68–80. Rouse, W.R., 1984a. Microclimate at Arctic tree line 2. Soil microclimate of tundra and forest. Water Resources Research 20 (1), 67–73. Rouse, W.R., 1984b. Microclimate at Arctic tree line 3. The effects of regional advection on the surface energy balance of upland tundra. Water Resources Research 20 (1), 74–78. Rouse, W.R., 1991. Impacts of Hudson Bay on the terrestrial climate of the Hudson Bay Lowlands. Arctic and Alpine Research 20, 56–66. Rouse, W.R., Carlson, D.W., Weick, E.J., 1992. Impacts of summer warming on the energy and water balance of wetland tundra. Climatic Change 22 (4), 305–326. Saurer, M., Siegenthaler, U., 1989. 13C/12C isotope ratios in trees are sensitive to relative humidity. Dendrochronologia 7, 9–13. Saurer, M., Siegwolf, R.T.W., Schweingruber, F.H., 2004. Carbon isotope discrimination indicates improving water-use efficiency of trees in northern Eurasia over the last 100 years. Global Change Biology 10, 2109–2120. Schleser, G.H., Helle, G., Lücke, A., Vos, H., 1999. Isotope signals as climate proxies: the role of transfer functions in the study of terrestrial archives. Quaternary Science Reviews 18, 927–943. Schweingruber, F.H., Briffa, K.R., Nogler, P., 1993. A tree-ring densitometric transect from Alaska to Labrador. Comparison of ring-width and maximum-

101

latewood-density chronologies in the conifer belt of northern North America. International Journal of Biometeorology 37 (3), 151–169. Scott, G.A.J., 1995. Canada's Vegetation: A World Perspective. McGill-Queen's University Press. Scott, P.A., Hansell, R.I.C., 2002. Development of white spruce tree islands in the shrub zone of the forest-tundra. Arctic 55, 238–246. Scott, P.A., Bentley, C.V., Fayle, D.C.F., Hansell, R.I.C., 1987. Crown forms and shoot elongation of white spruce at the treeline, Churchill, Manitoba, Canada. Arctic and Alpine Research 19, 175–186. Scott, P.A., Fayle, D.C.F., Bentley, C.V., Hansell, R.I.C., 1988. Large-scale changes in atmospheric circulation interpreted from patterns of tree growth at Churchill, Manitoba, Canada. Arctic and Alpine Research 20, 199–211. Scott, P.A., Hansell, R.I.C., Erickson, W.R., 1993. Influences of wind and snow on northern tree-line environments at Churchill, Manitoba, Canada. Arctic 46 (4), 316–326. Serreze, M.C., Walsh, J.E., Chapin, F.S., Osterkamp, T., Dyurgerov, M., Romanovsky, V., Oechel, W.C., Morisson, J., Zhang, T., Barry, R.G., 2000. Observational evidence of recent change in the northern high-latitude environment. Climatic Change 46 (1–2), 159–207. Sonninen, E., Jungner, H., 1995. Stable carbon isotopes in tree rings of a Scot pine (Pinus sylvestris L.) from northern Finland. In: Frenzel, B., Staufer, M.M., Weiss, M.M. (Eds.), Problems of stable isotopes in tree rings, lake sediments and peat-bogs as climatic evidence for the Holocene, Paläoklimaforschung (Palaeoclimate Research), 15, pp. 121–128. Sun, Z.J., Livingston, N.J., Guy, R.D., Ethier, G.J., 1996. Stable carbon isotopes as indicators of increased water use efficiency and productivity in white spruce (Picea glauca (Moench) Voss) seedlings. Plant, Cell and Environment 19, 887–894. SYSTAT, 2004. SYSTAT 11: Statistics I. Systat Software Incorporated, Richmond, California. Timoney, K., 1995. Tree and tundra cover anomalies in the subarctic foresttundra of Northwestern Canada. Arctic 48 (1), 13–21. Treydte, K., Schleser, G.H., Schweingruber, F.H., Winiger, M., 2001. The climatic signal of δ13C in subalpine spruces (Lotschental, Swiss Alps). Tellus 53B, 593–611. Van Wagner, C.E., 1987. Development and structure of the Canadian Forest Fire Weather Index System. Canadian Forest Service, Forestry Technical Report, 35. 37 pp. Vickers, G., Buzza, S., Schmidt, D., Mullock, J., 2001. The Weather of the Canadian Prairies Graphic Area Forecast 32. NAV CANADA and Meteorological Service of Canada, Ottawa. Vincent, L.A., Gullett, D.W., 1999. Canadian historical and homogeneous temperature datasets for climate change analyses. International Journal of Climatology 19, 1375–1388. Warren, C.R., McGrath, J.F., Adams, M.A., 2001. Water availability and carbon isotope discrimination in conifers. Oecologia 127, 476–486. Wigley, T.M.L., Briffa, K.R., Jones, P.D., 1984. On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. Journal of Climate and Applied Meteorology 23, 201–213. Yamaguchi, D.K., 1991. A simple method for cross-dating increment cores from living trees. Canadian Journal of Forest Research 21, 414–416. Zoladeski, C.A., Wickware, G.M., Delorme, R.J., Sims, R.A., Corns, I.G.W., 1995. Forest Ecosystem Classification for Manitoba: Field Guide. UBC Press, Vancouver, British Columbia.