Introduction

The Russian and North American river systems transport large quantities of timber from the boreal forest into the Arctic Ocean. This so-called driftwood (hereinafter DW) is either composed of felled trees that were lost during river transport (e.g.. Kindle, 1921; Eggertsson, 1994a) or consists of naturally fallen trunks from river slope erosion, which enter the tributaries especially after flood and ice storm events (Alix, 2005). Moreover, discharge rates as well as riverbed and delta characteristics affect the amount of DW delivered to the ocean. After leaving the boreal rivers, the wood either sinks to the seafloor (McClain and Barry, 2014) or is rafted within the sea ice along the main ocean currents. Melted out at the ice edges in summer, it is deposited along shallow Arctic coastlines. Wood anatomical and dendrochronological techniques have proved helpful in species identification, precise dating, and provenance determination of Arctic DW (Bartholin and Hjort, 1987; Eggertsson, 1993; Hellmann et al., 2013b). However, the joint application of these techniques on a well-replicated data set of different species from widespread sampling locations, in combination with the evaluation of comprehensive networks from boreal tree-ring chronologies, is still missing (Hellmann et al., 2013a). Quantification of the amount of lost timbers is thus often challenging, and the relative contributions of the different streams and their catchment regions on the final DW composition remain unknown.

Earlier DW studies analyzed rather low numbers of samples, ranging from 13 in Fischer and Schneider (1883) to 481 in Johansen (1998, 1999). In consequence, these studies generated only limited insight on the contribution of specific source regions, as well as on historical ocean current and river discharge dynamics. DW from Iceland, Svalbard, Jan Mayen, and other islands mainly originated from the lower Angara region in central Siberia and northwest Russia (Eggertsson, 1993, 1994b; Johansen, 1998, 1999, 2001). Only a few DW samples from Greenland were so far successfully cross-dated against tree-ring widths reference chronologies from the Mackenzie and Yukon Rivers, but no material from North America was found further east, for example, on Svalbard or Jan Mayen (Giddings, 1952; Eggertsson and Laeyendecker, 1995). Until now, the vast majority of studies mainly found pine, larch, and spruce at sites outside of the present-day permanent Arctic sea ice extent (Ingvarson, 1910; Eurola, 1971; Häggblom, 1982; Eggertsson, 1994b; Eggertsson and Laeyendecker, 1995; Tremblay et al., 1997; Johansen, 1998, 1999, 2001). In contrast to overall younger dendrochronological dating achievements, most Holocene DW has been dated by ¹⁴C measurements (Stewart and England, 1983; Dyke et al., 1997; England et al., 2008; Funder et al., 2011). In fact, most Holocene DW evidence from northern Greenland and the Canadian Arctic Archipelago suggests that spruce arrived from North America and that larch originated from eastern Siberia (Funder et al., 2011). Spatio temporal patterns of the proportion of the different tree species have been related to ocean current dynamics and variations in source region forest characteristics, superimposed on changes in Arctic sea ice extent. However, the north Greenlandic sites introduced by Funder et al. (2011), in contrast to all previous studies, are located within the current-day sea ice area, making their findings less representative for most Arctic DW locations outside the range of sea ice extent.

Here, we present the so far largest collection of DW samples integrating different species from four Arctic Islands outside the current-day permanent sea ice extent. Wood anatomical characteristics were analyzed to identify the species composition of the entire data set, and ring widths were measured for each Pinus sylvestris sample, the most frequent species among all sites. Cross-dating potential of these pine DW series was assessed against a network of pine reference chronologies from the Eurasian boreal forest zone. Spatial field correlations of the DW chronologies against gridded climate indices of the 20th century were utilized to confirm our cross-dating results at the catchment-specific level.

Material and Methods

A total of 2556 DW samples were collected between 1996 and 2012 in Greenland (n = 1348), Iceland (n = 774), Svalbard (n = 390), and on the Faroe Islands (n = 44), an area in the North Atlantic roughly covering 62°–78°N and 21°W—38°E (Fig. 1). The amount of DW along the different shorelines strongly varied from island to island. Few scattered trunks were found in eastern Greenland and Svalbard, whereas high trunk densities were characteristic for most shorelines in northwest Iceland. At each of the five Arctic sites, disc samples were collected from natural and logged material including small branches and large trunks. All material was archived, sanded in several steps from 80 to 400 grains cm^-1, and macroscopically categorized. Morphological features were used to differentiate between deciduous and coniferous wood, as well as between pine, spruce, larch, and fir (Fig. 1). Microscopic techniques were applied not only to determine the genus, but also to support species identification. Radial sections were cut for all samples. Unstained microscopic sections enabled the separation between Pinus sylvestris and Pinus Sibirica, as well as Larix sp., Picea sp., Abies sp., Salix sp., Populus sp., and Betula sp. Large pinoid pits characterize the rays of Pinus sylvestris and Pinus Sibirica. Differentiation between the two pine species was possible by the occurrence of different ray-tracheid wall structures, being serrated for Pinus sylvestris but smooth for Pinus Sibirica (see Hellmann et al., 2013b, for details).

Ring width patterns of all 969 Pinus sylvestris samples were measured using the program TSAPwin and a Lintab measuring device (Rinntech, Heidelberg, Germany). At least two radii per disc were produced and averaged to sample-specific mean curves. The individual mean measurement series were compared against each other to detect groups of DW samples that share statistically significant growth patterns. Further cross-dating was realized with the program PAST4 by SCIEM (Scientific Engineering and Manufacture, Vienna), which allows the synchronization of ring-width series by setting a reference and sample curves. Statistical criteria (two t-values [Baillie and Pilcher, 1973; Hollstein, 1980], Gleichläufigkeit, and correlation coefficient) are used to evaluate the best match. The visual comparison of the series excludes mismatching due to “false correlations,” caused by single common peaks or low points. All series that showed t-values ≥6, correlation coefficients ≥0.5, and that passed visual control, were combined to develop floating mean chronologies. Finally, we compared the floating DW chronologies against absolutely dated pine reference chronologies from the Eurasian boreal forest zone. Age trend removal and chronology development followed standard techniques: Cubic smoothing splines with 50% frequency-response cutoff at 200 years (SPL; Cook and Peters, 1981) were applied after power-transformation (PT; Cook and Peters, 1997) and using the most recent version of the ARSTAN (ARSTAN_41d for Windows) software (Cook and Krusic, 2007). The corresponding index values were calculated as residuals between the transformed ring width measurements and their corresponding curve fits. The final chronologies were calculated using bi-weight robust means and considering the standard ARSTAN routine (STD).Temporal variance changes were further stabilized with respect to fluctuations in sample size and interseries correlation (Osborn et al., 1997).

FIGURE 1.

(A) Driftwood (DW) sampling sites (1: Traill [Greenland, 253 samples]; 2: Scoresbysund [Greenland, 1095]; 3: Westfjords [Iceland, 774]; 4: Ny Friesland [Svalbard, 390]; 5: Faroe Islands [44]) and Pinus sylvestris distribution area, including the major boreal river systems and reference chronologies: Dvina-Pechora (red: 11 sites and 284 series), Ob (blue: 11 sites and 615 series), Yenisei (green: 14 sites and 319 series), Lena (orange: 17 sites and 345 series). (B) DW species composition and replication (1–5 refer to the sampling sites): PISY (Pinus sylvestris; 969), PISI (Pinus sibirica; 48), LASP (Larix sp.; 747), PCSP (Picea sp.; 426), ABSP (Abies sp.; 106), and Dec. (deciduous wood; Betula sp., Salix sp., and Populus sp.; 254). (C) Characteristics of the large Eurasian river systems draining into the Arctic Ocean (Holmes et al., 2002; McClelland et al., 2004). (D) Coasts with DW on Greenland, Scoresbysund and (E) Iceland, Westfjords. (F) Macroscopic (cross section) and (G) microscopic (radial section) wood structures of Pinus sylvestris.

The resulting chronologies were truncated either at a minimum replication of five series or when the Expressed Population Signal (EPS) dropped below the commonly applied quality threshold of 0.85 (Briffa and Jones, 1990). The EPS reflects the quality of a finite chronology compared to a perfect theoretical infinite population (Wigley et al., 1984) and was herein calculated over 30-year windows with 15 years overlap.

A total of 53 absolutely dated Pinus sylvestris ring-width reference chronologies were compiled for the Russian boreal forest zone. All records originate from the 40°–140°E and 60°–90°N area and were either obtained from the International Tree Ring Data Bank (ITRDB,  http://www.ncdc.noaa.gov/data-access/paleoclimatology-data/datasets/tree-ring) or kindly provided by Russian colleagues after personal request. First, we combined the series of the sites, for which older or updated data exist, reducing the number of site chronologies to 46. Second, Pearson correlation coefficients (≥0.5) between the standardized chronologies were used to select the data according to their spatially coherent growth patterns in several steps. We then employed the resulting regional master chronologies to cross-date three floating Arctic DW chronologies.

Moreover, we calculated spatial field correlations between the combined DW and reference ring width chronologies and high-resolution (0.5° × 0.5°) gridded July–August temperature means and March-May precipitation totals (CRU TS3.21). Different chronology versions solely based on DW or living reference material, as well as using slightly different detrending techniques, were thoroughly examined against various gridded climate indices. All correlation maps, calculated over the 1961— 1990 common period were generated via the KNMI Climate Explorer ( http://climexp.knmi.nl/).

Results

DW composed of Pinus sylvestris, Pinus sibirica, Larix sp., Picea sp.,Abies sp., Salix sp., Populus sp., and Betula sp. was found on all four islands (Fig. 1). Pinus sylvestris is the dominant species at all sites (37.9%), followed by Larix sp. (29.2%) and Picea sp. (16.7%), Abies sp. (4.1%), Pinus sibirica (1.9%), and deciduous wood (9.9%). Pine was further differentiated at the species level (Hellmann et al., 2013b, Fig. 1) allowing the spatial separation of Pinus sylvestris and Pinus sibirica across Eurasia. No pine-species originated from the North American continent.

Geographic groups of the boreal reference chronologies were consistent with the catchment areas of the large Russian river sys- terns: Northern Dvina together with Pechora, Ob, Yenisei, and Lena (Fig. 1). Sites within the Dvina-Pechora region range from 60°–67°N and 41°–57°E. Chronologies within the Ob catchment are located between 60°–66°N and 65°–77°E. The Yenisei sites are within 55°– 66°N and 88°–96°E. Data from the Lena represent the 60°–64°N and 125°–l33°E area. The longest tree-ring width reference chronology spans the period 1571–2008 and represents the Ob catchment, whereas the shortest chronology covers 1654–2003 and is located in the Lena region (Fig. 2). Pearson correlation coefficients between the regional reference chronologies range from 0.22 for the nearby Dvina-Pechora and Ob catchments, to 0.05 for the most western and most eastern regions: Dvina-Peehora and Lena. The reference chronologies for the Dvina-Peehora and the Ob regions include 11 sites each (284 and 615 individual tree series, respectively); the Yenisei region, 14 (319 series); and the Lena, 17 (345 series) (Table 1). The whole Siberian tree-ring network includes 1563 series.

FIGURE 2.

Reference chronologies of major Siberian catchment areas integrating several local site chronologies. Chronological tree-ring series were power transformed, detrended using a 200-year spline, variance stabilized, and normalized. Curves are shown for a minimum sample depth of 5 series. The dotted line indicates the EPS-value of 0.85. The inset is showing correlation coefficients between the different reference chronologies (A.D. 1735-1964, dashed gray lines).

Focusing on the Pinus sylvestris DW material, we developed three floating chronologies integrating 498 of the total of 969 pine DW samples (Fig. 3). One of these floating chronologies comprises 453 series, whereas the other two contain only 28 and 17 samples, respectively. Mean segment length of all Pinus sylvestris measurement series is 121.7 years, with a minimum of 20 and a maximum of 336 years. Comparison with the living reference data from Siberia enabled 51.39% of the pine samples to be dated. Most interestingly, 90.96% (453 samples) of the dated samples originate from the central Siberian Yenisei, 5.62% (28 samples) from the western Dvina-Peehora and 3.41% (17 samples) from the eastern Lena region (Fig. 4). No single DW sample was cross-dated against the Ob reference chronology. Mean interseries correlation coefficients, that is, Rbar values, for the dated DW samples range from 0.31 for the Dvina-Peehora and Yenisei region to 0.80 for the Lena region (see Table 1 for further chronology characteristics).

The applied sampling scheme including small branches and big trunks resulted in a wide age distribution range: 109 pine samples (11.25%) contain less than 60 rings and are therefore hardly datable. A mean segment length of 144 years of all dated material, compared to only 98 years of the undated samples supports the finding that the dating of young trees often remains challenging (Table 1). Moreover, the average growth rate of the undated material is higher (1.136 mm), compared to the dated samples (0.860 mm), which results from wider rings and thus younger trees. Mean tree ages of 128, 130, and 138 years for Dvina-Peehora, Lena, and Yenisei collections are rather similar among the different catchments.

TABLE 1

Characteristics of the boreal reference chronologies, as well as the DW measurements herein separated by sampling site and source area (No: number; MSL: mean segment length (years); AGR: average growth rate (mm); AC: lag-1 autocorrelation; Rbar: mean interseries correlation). MSL and AGR were calculated for the entire length of the chronologies, whereas AC and Rbar were calculated for the periods replicated by more than five series.

A total of 324 dated samples most likely originate from logged trees (65.06%), 67 had a natural mortality (13.45%), and 107 were not classified (21.49%) (Fig. 4). The mean ages (average growth rates) do not differ much between these classes, with 124 years (0.873 mm) for the logged material, 134 years (0.808 mm) for natural, and 155 years (0.850 mm) for nonclassified samples. For all three source areas, more than half of the samples were logged with 67.86% of the Dvina-Pechora material. 65.34% of the Yenisei samples, and 53.00% of the Lena wood.

The outermost rings of all dated pine samples fall in the 1614 to 1999 period (Fig. 4). The oldest pine grew from 1614–1810, originated from the Yenisei region, and was deposited on Svalbard. The youngest end year is 1999, originating from the Yenisei and found for samples on Greenland (Scoresbysund) and Iceland, where wood was collected in 2012. The minimum transport time from the falling of the tree until reaching a coast is thus constrained to 13 years. Caution is advised because bark and even sapwood were rarely preserved.

The spatial correlation patterns obtained from calibrating the combined DW plus reference chronologies against gridded climate data reinforce our cross-dating results (Fig. 3). Spatial field correlations are concordant with the respective river catchment area. The final chronologies of combined DW and living-tree series reveal significant positive spatial field correlations (≥0.4) with either variation in temperature means (Dvina-Pechora and Yenisei data) or precipitation totals (Lena data) for the time period from 1961 to 1990. The Dvina-Pechora and Yenisei data correlate with July-August summer temperature. The Lena River chronology shows positive correlations with March-May precipitation totals. Calculations using either only DW or only reference series do not show significant correlation patterns.

FIGURE 3.

(A) Siberian reference chronologies (black) and cross-dated DW chronologies (color) truncated at a minimum of 5 series. Correlation coefficients between the reference and DW chronologies were calculated over their common period A.D. 1735-1964 (dashed gray lines). (B) Spatial correlation fields (r ≥ 0.4) over the period A.D. 1961–1990 (pink lines) between the combined DW-reference chronologies and July–August temperatures (for the Dvina-Pechora and Yenisei region) or March–April–May precipitation (for the Lena region), respectively.

Discussion

BOREAL RIVER DISCHARGE

Boreal river discharge is expected to increase in the 21st century due to global warming (McClelland et al., 2004; Shiklomanov and Lammers, 2009). Implications are manifold and include changes in ocean current circulation (Peterson et al., 2002). Within the Eurasian Arctic, the central Siberian Yenisei River, draining into the Kara Sea, has the highest discharge of 620 km³ yr^-1. River-specific discharge rates coincide with the number of transported DW logs (Fig. 1). The Yenisei has not only the highest runoff of all large Eurasian rivers, but also represents the source of 91% of all cross-dated pine DW samples. Since pine is the most frequent species among our DW compilation, we define the Yenisei catchment as the main origin area of the recent DW found on the beaches of Greenland, Iceland, and Svalbard. Cross-dating results from earlier studies in Norway, Svalbard, Iceland, and Jan Mayen also found DW mainly to originate from this region (Eggertsson, 1993, 1994b; Johansen, 1998,1999).

In contrast, the Ob (404 km³ yr^-1), Dvina/Pechora (241 km³ yr^-1), and Lena (523 km³ yr^-1) river systems all have lower discharge rates than the Yenisei (Rachold et al., 2004). Most likely not only the lower discharge rates, but also the direction of Arctic Ocean currents is a reason for fewer DW samples delivered by these river systems (Fig. 5). Despite having the highest discharge, the Yenisei contains a relatively low sediment flux (4.7 Mt yr^-1) compared to its runoff. Only the Northern Dvina has a lower sediment discharge of 4.1 Mt yr^-1. All other large rivers transport more sediment to the Arctic Ocean (ranging from 9.4 Mt yr^-1 at the Pechora to 20.7 Mt yr^-1 at the Lena River) (Holmes et al. 2002). Sediment flux rates result from riverbed and watershed properties and influence the capability of rivers to transport other material, that is, naturally eroded trees as well as logged timbers. Moreover, a high sediment flux leads to the formation of a delta system (as, for example, at the Lena and Kolyma Rivers). Even though it is not clear if a delta indeed prevents transported sediment from reaching the Arctic Ocean (Holmes et al., 2002), observations of subfossil trunks in the Lena delta (72°N, 126°E) support the assumption that a huge number of logs get trapped in the delta system (Büntgen et al., 2014).

FIGURE 4.

Dated DW samples per source region (a: Dvina-Pechora, b: Yenisei, c: Lena), separated into logged (black), natural (gray), and nonclassified (light gray) material. Series are sorted by end-year with A.D. 1804 as the oldest (dashed line). n is the number of series included in each group. The gray shading refers to the period of high logging activities in central Siberia ca. 1920–1975. The vertical black lines refer to the DW sampling years: Ny Friesland, 1996; Traill, 2010; Scoresbysund, 2011 and 2012; and Iceland, 2012. On the Faroe Islands wood was collected over several years.

An improved understanding of long-term Arctic Ocean dynamics is important as changes in current strength and orientation can affect the expansion of the northern polar sea ice sheet (Zhang et al., 1999; Perovich and Richter-Menge, 2009). The spatial extent of the sea ice sheet is again sensitive to atmospheric climate changes through the varying ice/ocean albedo, greenhouse-gas exchange, and freshwater pathways (Morison et al., 2012; Parmentier et al., 2013; Screen et al., 2013; Pistone et al., 2014). The Arctic Ocean is frozen during winter with its maximum sea ice extent in March, but partly ice-free episodes occur during summer with the shrinking minimum sea ice extent in September (Stroeve et al., 2007). Similarly, the discharge rates of boreal rivers vary substantially throughout the year, with minimum values between November and April but with about 60% of the total discharge occurring from April to July (Gordeev. 2000; Lammers et al., 2001). High fluctuations of the water level, particularly during spring flood and storm events, may cause severe erosion of riverbanks and thus encourage naturally fallen trunks to enter the streams. Permafrost thawing can be an additional trigger for natural DW deposits. These DW trunks typically contain a root collar and/or branch bases, and can easily be distinguished from logged timbers free of rootstocks and snags (see Hellmann et al., 2013b, for details). Variation in seasonal runoff depends on the hydrological retention characteristics of the river. Headwater lakes as at the Mackenzie River, for instance, reduce seasonal variability, showing a fivefold variation from summer to winter compared to a tenfold variation for rivers without major headwater lakes such as the Yenisei and Lena (Carmack. 2000).

FIGURE 5.

(A) Pinus sylvestris distribution area (green) with pine-dominated forests (red) shown only for Russia (Bartalev et al., 2004). Large river catchments corresponding to DW source regions are specified by different colors of the rivers. Numbers of dated samples per river are indicated for each sampling site (black dots) by the pie charts, with the size corresponding to the total number of collected samples (1+2, Greenland; 3, Iceland; 4, Svalbard; 5, Faroe Islands). Black boxes locate the sampling sites of Funder et al. (2011). Arrows illustrate the main Arctic Ocean currents (ACIA, 2005), and the white area refers to the 2010 summer sea ice extent (MDA Information System, USGS, NASA). (B) Quantity of DW samples per year (last ring) and source area. The gray shading refers to the period of high logging activities in central Siberia, which are associated with poor rafting and log driving techniques and high loss of timber (ca. 1920–1975).