Reconstructing High Arctic growing season intensity from shoot length growth of a dwarf shrub

Abstract

Annual shoot length of the circumarctic dwarf shrub Cassiope tetragona has proved to be a reliable proxy for past and ongoing climate change in the Arctic. This is based on its strong linear relationship with monthly climate parameters. Monthly means are, however, coarse units for prediction of growth in marginal regions with short growing seasons. An alternative to monthly averages are parameters that quantify the growing season length (GSL) and its intensity (growing degree-days; GDD₅). GDD₅ is defined as the cumulative daily mean temperature above 5°C. GSL is defined as the number of days on which the average temperature exceeds 5°C. The aims of this study were to test whether these parameters are a better predictor of growth than monthly means and to reconstruct past High Arctic growing season climate. Correlative analysis shows that GDD₅ is a better predictor of annual shoot length growth than mean monthly temperatures and GSL, both at C. tetragona’s European northern and southern distribution limit, as well as at its assumed climatic optimum. Svalbard Airport GDD₅ was reconstructed back to 1857. The reconstruction shares 61% of variance with the instrumental record. This opens the possibility to obtain an Arctic network of climate reconstructions with high temporal and spatial resolution through construction of C. tetragona shoot length chronologies.

Keywords

Arctic shrub dendrochronology growing degree-days proxy data shoot length

Introduction

Over the past century, average temperatures in the Arctic have risen at a pace almost twice as fast as the global average (Intergovernmental Panel on Climate Change (IPCC), 2007). Over land in areas north of 65°, warming has even been stronger: more than double the global average as observed in global mean temperatures from the 19th to 20th centuries (Trenberth et al., 2007). Arctic-wide proxy-based reconstructions suggest that 20th century Arctic temperatures are the highest of the past four centuries (Overpeck et al., 1997). Furthermore, decadal resolved proxy records from 60° northward imply that a millennial-scale cooling trend in the Arctic was reversed during the 20th century and that 1999–2008 was the warmest decade in the Arctic over the past 2000 years (Kaufman et al., 2009). Still, clear distinction between natural variability and climate change due to anthropogenic forcing in the Arctic is restricted by the scarcity of both instrumental and proxy climate data in this vast area. The instrumental Arctic climate record is not only spatially limited, with weather station densities varying from only 1 per 100,000 km² in Arctic Canada, northern Alaska and parts of Siberia to 2 per 1000 km² in Fennoscandinavia (McBean et al., 2005), but also temporarily, with records dating back to the 1880s from only three areas: west Greenland, Iceland and northern Europe (Overland et al., 2004). Good instrumental coverage of the Arctic climate only exists from the 1950s onward (Przybylak, 2000).

Studies on older climate variability in the Arctic rely on proxy-based reconstructions, of which only lake sediments, ice cores and tree rings provide high-resolution and continuous records (McBean et al., 2005). Although reconstructions of areal average Arctic temperatures based on multiproxy analysis are available (Kaufman et al., 2009; Overpeck et al., 1997), many of the proxy records used in such studies are from the southern periphery of the Arctic. McBean et al. (2005) concluded that, as the Arctic climate is characterized by large natural variability and regional differences, a more homogeneous coverage of past climate is necessary for quantitative detection of change and observational gaps across the Arctic need to be filled.

The climate proxy record of Svalbard, which is the main research location of this study, comes mainly from ice-core data (Divine et al., 2011; Isaksson et al., 2005a, 2005b), besides more coarse paleoclimatic estimates based on pollen assemblages (Rozema et al., 2006). The oxygen isotope records from Svalbard are, however, more strongly correlated with the instrumental record from northern Norway than with the local record. Also, despite their generally annual resolution, the high frequency climate signal in these records is weak (Divine et al., 2011). Still, lower frequency temperature variability is well captured by these records, which makes them valuable proxies. High-resolution Arctic biological temperature proxies, which generally represent climatic conditions during the growing season (Miller et al., 2010), would therefore be a welcome addition to the existing proxy record.

The long-lived (>180 years) circumarctic evergreen dwarf shrub Cassiope tetragona can provide such proxy data (Rayback et al., 2012; Weijers et al., 2010, 2012). Previous studies have shown that annually resolved growth chronologies of Cassiope are strongly linked to monthly temperature data during the growing season and/or mean summer temperatures (Aanes et al., 2002; Callaghan et al., 1989; Havström et al., 1995; Johnstone and Henry, 1997; Rayback and Henry, 2005, 2006; Rozema et al., 2009; Weijers et al., 2010, 2012).

Annual shoot length growth of this species can be measured as it forms small dark bands in its white pith, called wintermarksepta (WMS), which demark the end of each growing season (Rozema et al., 2009; Weijers et al., 2010). Annual shoot length growth is also reflected by changes in the length of the leaves attached along its stems. Smaller leaves are formed at the start and end of each growing season, resulting in wave-like leaf patterns along the stems with each wave representing one year of growth (Warming, 1908).

Shoot length growth of C. tetragona responded positively to experimental warming during the growing season at two sites in the High Arctic archipelago of Svalbard (Isdammen; 78°12′E, 15°43′E and Ny-Ålesund; 78°56′N, 11°50′E) and at a high-elevation site near Abisko, Sweden (68°19′N, 18°41′E) (Havström et al., 1993; Rozema et al., 2009; Weijers et al., 2012). A biogeographical comparison of C. tetragona growth, measured as annual number of leaves, from seven sites across the Arctic revealed a positive linear relationship between mean July temperature and growth (Havström et al., 1995). These results match with those of a dendrochronological study on C. tetragona growth at three sites ranging from the High to Sub Arctic. Here, a linear positive relationship was found between mean July temperatures and annual growth, measured as total leaf length, number of leaves, and shoot length (Weijers et al., 2012).

Temperatures preceding instrumental records and in areas without a temperature record can be reconstructed, using linear transfer functions between growth and mean summer/July temperatures (Rayback and Henry, 2006; Weijers et al., 2010). In the previously mentioned studies climate–growth relationships have been calculated with monthly climate data. However, monthly mean temperatures are a rather coarse unit, especially in marginal regions such as the Arctic, where the growing season is very short and more likely to prevail for several weeks than months (Lindholm et al., 2011).

Early June temperatures can still be too low to support substantial growth, while later that month temperatures might become sufficient. The use of climate data with a higher temporal resolution and calculation of climate variables that better match the time window when growth is realised in a certain environment can help to improve climate–growth models and thereby climate reconstructions.

An alternative to monthly averages are parameters that quantify the growing season climate, such as growing season length (GSL) and growing degree-days (GDD). GDD provides a measure for the thermal growing season intensity and is defined as the cumulative daily mean temperature above a given threshold. GSL is a measure for the growing season length and defined as the number of days on which the average daily temperature exceeds this threshold. At high latitudes, the threshold is usually set at 5°C, as plant growth is insignificant when temperatures are lower (Carter, 1998; Førland et al., 2004; Wagner-Cremer et al., 2010). Consequently, all days with average temperatures above this threshold are incorporated in these parameters and not merely temperatures of one month.

Here, we test whether variations in GDD and GSL can indeed better explain variations in shoot length growth of C. tetragona than mean monthly temperatures. We used previously developed shoot length chronologies from C. tetragona shrubs that were growing at three sites along a European latitudinal temperature gradient (Weijers et al., 2010, 2012). The three sites together represent the entire temperature gradient of this species. The three shoot length chronologies are compared with instrumental records of growing degree-days above 5°C (GDD₅) and growing season length (GSL) calculated from daily mean temperature data collected from weather stations nearby the three sampling locations. Furthermore, we aim to reconstruct past growing season climate at Svalbard beyond the length of the instrumental record, which started in 1912, using the 154 and 169 years long shoot length chronologies from Ny-Ålesund and Endalen, respectively. These are the longest terrestrial growth chronologies developed for the area and the longest C. tetragona growth chronologies worldwide. The Ny-Ålesund record was, up to now, not used for any reconstruction purposes.

Material and methods

Site description and sampling

Cassiope tetragona shrub samples were collected from three different climatic zones in the Arctic (see Elvebakk (1999) and Walker et al. (2005) for a definition of these zones). On 21 August 2010, 21 shrubs were sampled near C. tetragona’s cold tolerance limit, 5.5 km southeast of Ny-Ålesund, Svalbard (78°54′N, 12°10′E; 35 m above sea level (a.s.l.)), at the border between the middle arctic tundra zone (zone C) and the northern arctic tundra zone (zone B), where C. tetragona is lacking. The vegetation at this site is characterized by mixed patches of the dwarf shrubs C. tetragona, Dryas octopetala, and Salix polaris. Within these patches, either Salix or Cassiope is dominant. Mean June, July, and August temperatures between 1979 and 2008 at this site were 2.08, 5.14, and 4.19°C, respectively. The mean annual precipitation sum over the same period was 408 mm. Of the 21 harvested shrubs, a total of 111 branches were analyzed in detail.

In June and August of 2008 and at the end of August 2009, a total of 32 shrubs were sampled at C. tetragona’s assumed climatic optimum in Endalen, Svalbard (78°11′N, 15°44′E; 100 m a.s.l.), 4 km southeast of Longyearbyen. This site is located in the Arctic tundra zone C, which is characterized by the presence of C. tetragona. Here, C. tetragona covers approx. 30% of the vegetation. Salix polaris and Dryas octopetala are also present, as well as patches of Empetrum nigrum ssp. hermaphroditum and Betula nana (Weijers et al., 2010). Despite the relative short distance to Ny-Ålesund, Endalen is remarkably warmer with mean June, July, and August temperatures of 2.76, 6.43, and 5.37°C, respectively. The site is also dryer than Ny-Ålesund with an average annual precipitation sum over the period 1979–2008 of only 190 mm. A total of 213 branches of the 32 shrubs were analyzed for the construction of a shoot length growth chronology.

At the end of August 2009, 12 shrubs were sampled near the European southern limit of the species, about 2 km southeast of Abisko, Sweden (68°20′N, 18°51′; 500 m a.s.l.). This site is located in the arctic subzone E. The site is characterized by a relatively rich dwarf shrub vegetation, which consist of patches of C. tetragona, Empetrum nigrum ssp. hermaphroditum, Vaccinium uliginosum, Rhododendron lapponicum and Salix hastata (Havström et al., 1993). Abisko is warmer than both sites in Svalbard, with mean June, July, and August temperatures of 8.76, 11.72, and 10.17°C, respectively, as measured between 1979 and 2008. The average annual precipitation sum in Abisko over this period was 337 mm. A total of 74 branches were analyzed for the construction of a shoot length chronology.

The shrub samples harvested at all sites consisted of multiple long and therefore old shoots and had lengths ranging between about 0.2 and 0.9 m. To ensure accurate dating from collection year backwards we preferably collected living shoots, as indicated by the presence of green tips. To avoid repeated sampling of the same individual, several meters distance was kept between sampling spots.

Chronology development

Annual shoot length growth was measured for several branches of each shrub sample by measuring the distances between the consecutive wintermarksepta (WMS) in their piths under a stereo microscope (10 or 30× magnification, 0.1 mm precision). Annual shoot length was then plotted as time series for visual crossdating (Weijers et al., 2010). For crossdating we used information on known date of harvest, in addition to synchronous behaviour in annual growth patterns between branches from the same shrub and branches from shrubs at the same site. The presence of pointer years, such as years with extremely little growth as in 1994 at Endalen and Ny-Ålesund, facilitated this process. In Figure 1, the crossdating process is further explained. Visual crossdating was checked in COFECHA v6.06 (Grissino-Mayer, 2001; Holmes, 1983).

Figure 1.

Example of crossdating of individual branch series from Ny-Ålesund. (a) All series from shrub NA4 could be crossdated with each other, except for the oldest (NA4.41). (b) The crossdated series were then averaged into one chronology (black line). All series from NA10 could then be crossdated with each other and the NA4-average. (c) Then, the NA4 and NA10 series were averaged (black line); all NA5-series and NA4.41, could consequently be crossdated. This process was repeated for each site until all individual branch series were visually crossdated.

The individual crossdated branch chronologies were standardized with a horizontal line through their mean, in WINARSTAN (Cook, 1985). Thus, each individual annual value was divided by the mean value of the related branch series. By doing so, correction is made for both differences in shoot-growth-level and undesired age-related trends and autocorrelation, while the interannual climate signal is in principle enhanced (Weijers et al., 2010). Long-term age-related growth trends in C. tetragona shrubs are hard to observe in individual branch chronologies, because of their relatively short time span. Juvenile growth trends, however, with shorter annual shoot length increments during approximately the first 5 years of growth, are apparent in many side-branches (Rayback and Henry, 2006; Weijers et al., 2010). Shoot length series of different branches were first averaged per shrub and subsequently merged into site chronologies, to prevent overrepresentation of individual shrubs.

We calculated three different statistics to evaluate each chronology: mean sensitivity, average correlation with the master, and the Expressed Population Signal (EPS). Mean sensitivity describes the percentage change from one year to the next (Fritts, 1976) and was calculated in WINARSTAN based on the individual shrub chronologies (branch chronologies averaged per shrub) per site. The average correlation is the mean correlation strength between the individual shrub chronologies of a site and the average of all others at that site and was calculated in COFECHA. The EPS is a measure for the reliability of a mean chronology, based on the correlation between individual series and sample size. EPS values above 0.85 are regarded as reliable (Wigley et al., 1984).

Climate data

Daily and monthly mean temperatures from two Ny-Ålesund weather stations (78°55′N, 11°52′E, 42 m a.s.l.; January 1969–July 1974 and 78°55′N, 11°55′E, 8 m a.s.l.; August 1974–present) were obtained from the eKlima database of the Norwegian Meteorological Institute (DNMI, 2011).

For Endalen, daily and monthly mean temperatures from the Svalbard Airport weather station were used (78°14′N, 15°28′E; 8 m a.s.l.; August 1975–present; DNMI, 2011). Daily mean temperatures preceding 1975 were obtained from the Longyearbyen record (78°13′N, 15°21′E; 1917–1919, 1922 and 1957–March 1977; DNMI, 2011) and Green Harbour record (78°02′N, 14°14′E; December 1911–September 1930; DNMI, 2011). Measured at different stations in the Isfjorden area, these series needed to be homogenized to match the Svalbard Airport series (Nordli and Kohler, 2003). Temperatures at the Longyearbyen station are generally warmer than at Svalbard Airport and Green Harbour is colder than Longyearbyen. We created a homogenized data set of daily mean temperatures for the periods 1912–1930 and 1957–2010. No daily mean temperatures were available for the period 1931–1956. Individual daily temperature series were homogenized by calculating linear transfer functions per month using overlapping periods between individual data sets. First, Longyearbyen temperatures were transferred to Svalbard Airport temperatures and then the relationship between the transferred Longyearbyen temperatures (1917–1922) were used to transfer Green Harbour to Svalbard Airport temperatures. Monthly mean temperatures prior to 1975 were obtained from the homogenized monthly Svalbard Airport record as published by NORDKLIM. Daily and monthly mean temperatures from Abisko (68°21′N, 18°49′E, 388 m a.s.l.; 1913–2009) were provided by the Abisko Scientific Research Station.

Growing degree-day sums

From the daily mean temperatures the growing season intensity was derived for each site, expressed as growing degree-days (GDD). GDD is calculated as the cumulative heat above a selected threshold temperature, usually defined as:

\begin{matrix} {GDD}_{x} = \sum_{i = 1}^{365} (T_{i} - X), & T_{i} \geq X \end{matrix}

(1)

where T_i is the daily mean temperature for day i and X the selected threshold temperature (Carter, 1998; Førland et al., 2004; Skaugen and Tveito, 2004; Wagner-Cremer et al., 2010). We chose to use 5°C as the threshold after Carter (1998), for it has become the standard definition and is used for modelled Arctic thermal growing season projections (Førland et al., 2004; Skaugen and Tveito, 2004). The growing season length (GSL) was then calculated by summing the number of days on which the average temperature exceeded the 5°C threshold.

Climate–growth analyses

We had previously calculated Pearson correlation coefficients between the best monthly climate predictor at each site (mean August temperature for Ny-Ålesund and mean July temperature for Endalen and Abisko) and the raw and standardized shoot length chronologies (see Weijers et al., 2010, 2012). Both the raw and standardized chronologies were used in climate–growth analyses, to ensure that standardization did not negatively impact the climate signal, as climatic trends are sometimes accidently removed by standardization methods designed to remove age-related growth trends. We repeated these calculations with GDD₅ and GSL as predictors. The Ny-Ålesund chronologies were also compared with the climate data from the Svalbard Airport weather station, because of its relative vicinity and the fact that this record is much longer than that of Ny-Ålesund. We calculated Pearson correlation coefficients between the separate climate variables to check the intercorrelation between the predictors. All calculations were carried out in PASW Statistics 17.0.2.

The reliability of linear transfer functions used for climate reconstructions relies on the linearity of climate–growth relationships (Loehle, 2009). We therefore plotted shoot length growth against GDD₅ over two ranges: one over all sites combined (GDD₅ between approximately 0 and 800) and one for Svalbard alone (Ny-Ålesund and Endalen combined, GDD₅ approximately 0–200) to test the linearity of the GDD₅–raw growth relationship.

GDD₅ reconstruction: calibration and verification

The overlapping period between the homogenized GDD₅ data set valid for Svalbard Airport and the Svalbard C. tetragona shoot length chronologies consist of 71 years in total (1912–1930 and 1957–2008). This period was split into two periods for independent calibration and verification on separate data sets: an early period consisting of the years 1912–1930 and 1957–1972 (34 years) and a late period consisting of the years 1973–2008 (35 years). Linear transfer functions models with GDD₅ as a function of standardized shoot length at Ny-Ålesund and Endalen were computed in PASW Statistics 17.0.2 to reconstruct GDD₅. These transfer functions were calculated over both the early and late period and verified over the late and early period, respectively. For verification we calculated the coefficients of determination (R²) between predicted and measured GDD₅ over the verification period, the Reduction of Error (RE) statistic, and the Coefficient of Efficiency (CE) statistic (Briffa et al., 1988; Fritts, 1976; Fritts et al., 1990). In addition, the first difference sign test and Product Means (PM) test were used for verification (Fritts, 1976). Values of RE and CE can range from −∞ to 1, with 1 indicating perfect agreement. The first difference sign test measures the amount of parallel run between two time series, disregarding the magnitude of change. The PM statistic is a measure for the amount of common change in two time series, both with regard to the amount of parallel run and the magnitude of change.

Results

Characteristics of the shoot length chronologies

The raw shoot length chronologies developed for the three sites differ in length and growth level (Figure 2). Chronologies calculated from the shoot length records for shrubs from Endalen and Ny-Ålesund span a period of more than 150 years of growth; the Abisko chronology from the southernmost shrubs is much shorter, partly as a result of much faster annual growth. All three raw-data chronologies are characterized by high autocorrelation, which was strongly reduced by standardization (Table 1). The average correlation values of the individual standardized shrub series with their related master chronology were high (between 0.432 and 0.477; Table 1). The effective standardized chronology length (EPS>0.85) was 1894–2010 for Ny-Ålesund, 1870–2008 for Endalen, and 1980–2009 for Abisko (Table 1). The standardized shoot length series are further characterized by intermediate to high (0.284–0.360) mean sensitivity values (Grissino-Mayer, 2001).

Figure 2.

Raw (grey lines) and standardized (black lines) shoot length chronologies for each of the three sites (a, b, and c). Sample sizes (grey shaded areas) and EPS-values (see Wigley et al., 1984) for the raw (grey lines) and standardized (black lines) chronologies and EPS threshold-value of 0.85 (solid lines) are given in e, f, and g.

Table 1.

Characteristics of the raw and standardized shoot length growth chronologies.

	Ny-Ålesund	Endalen	Abisko
Mean	0.993 (3.529 mm)	1.00 (5.05 mm)	0.988 (9.267 mm)
Standard deviation	0.443 (1.562 mm)	0.342 (1.767 mm)	0.386 (3.593 mm)
Mean sensitivity	0.360	0.284	0.353 (0.351)
Autocorrelation
First order	0.353* (0.669*)	0.321* (0.662*)	0.591 (0.730*)
Second order	0.344 (0.659*)	0.278 (0.616*)	0.203 n.s. (0.491***)
Third order	0.250* (0.622***)	0.137 n.s. (0.539***)	−0.064 n.s. (0.369***)
Average correlation with master	0.477 (0.539)	0.439 (0.432)	0.451 (0.462)
Effective chronology length, where EPS>0.85	1894–2010 (1877–2010)	1870–2008 (1873–2008)	1980–2009

Notes:

Values for the raw chronologies are placed between brackets when different from those of the standardized chronologies. Mean, standard deviation, mean sensitivity, and first-order autocorrelation values are after Weijers et al. (2010, 2012). EPS is the Expressed Population Signal (Wigley et al., 1984); ‘Average correlation with master’ is the mean correlation of each individual shrub series with the mean master chronology derived from all other series, as calculated in COFECHA (Grissino-Mayer, 2001; Holmes, 1983).

Significance levels: *** p < 0.0001; ** p ≤ 0.001; * p = 0.005.

Climate–growth analysis

At all sites, the correlations between the shoot length chronologies and GDD₅ were stronger (r between 0.61 and 0.74) than those with the best mean monthly climate predictor (r between 0.53 and 0.64; Table 2). In contrast, the correlations between GSL and shoot length (r between 0.38 and 0.61) were the weakest correlations at all sites. While the relationship between GSL and shoot length is strongest near the northern European range limit of C. tetragona (r=0.61, p<0.0001), it is only just significant near its European southern distribution limit (r=0.38, p=0.045).

Table 2.

Climate–growth relationships between GDD₅, GSL, and the best mean monthly climatic predictor for each site, and the raw/standardized shoot length chronologies from Ny-Ålesund, Endalen, and Abisko.

	GDD₅	GSL	Best mean monthly predictor
Ny-Ålesund (with Ny-Å)			August temperature
r	0.69 (0.68)	0.60	0.60 (0.58)
R ²	0.47 (0.46)	0.36	0.36 (0.34)
n	42	42	42
p	<0.0001	<0.0001	<0.0001
Endalen			July temperature
r	0.72	0.55 (0.56)	0.64
R ²	0.51	0.30 (0.31)	0.41
n	71	71	97
p	<0.0001	<0.0001	<0.0001
Abisko			July temperature
r	0.61 (0.67)	0.38 (0.43)	0.53 (0.59)
R ²	0.37 (0.45)	0.14 (0.18)	0.28 (0.35)
n	30	30	30
p	<0.001 (<0.0001)	0.045 (0.021)	0.002 (0.001)
Ny-Ålesund (with SA)			July temperature
r	0.74 (0.73)	0.61 (0.60)	0.62 (0.59)
R ²	0.55 (0.53)	0.37 (0.36)	0.39 (0.35)
n	73	73	99
p	<0.0001	<0.0001	<0.0001

Notes:

Values partly after Weijers et al. (2010, 2012) and the raw/standardized shoot length chronologies from Ny-Ålesund, Endalen and Abisko. Values for the raw chronologies are placed between brackets when different from those of the standardized chronologies.

r: Pearson correlation coefficient; R²: coefficient of determination; n: sample size (number of years in analysis); ‘with Ny-Å’: comparison with the Ny-Ålesund record; ‘with SA’: comparison with the Svalbard Airport record.

Standardization hardly affected the strength of the climate signal in the Ny-Ålesund and Endalen shoot length chronologies, but led to a decrease in the strength of the climate–growth correlations at Abisko. The relationship between GDD₅ and shoot length growth was strongest at Endalen (r=0.72, p<0.0001), when growth is correlated with the local instrumental record of each site. However, the GDD₅–growth relationship is strongest at the northernmost site, when Ny-Ålesund shoot length growth is compared with the longer Svalbard Airport record (r=0.74, p<0.0001). The weakest GDD₅–growth relationship was found at Abisko (r=0.61, p<0.001). The intercorrelations between the studied climatic variables at each site are given in Table 3. All climatic variables are highly correlated at all sites (r between 0.60 and 0.90), except mean July temperature and GSL at Abisko (r=0.17, p>0.05).

Table 3.

Intercorrelations (Pearson correlation coefficients) between the climatic parameters for each site.

	GDD₅	GSL
Ny-Ålesund
GDD₅	–	0.88**
GSL	0.88**	–
Mean August temperature n = 42	0.71**	0.70**
Svalbard Airport (Endalen)
GDD₅	–	0.89**
GSL	0.89**	–
Mean July temperature n = 71	0.90**	0.84**
Abisko
GDD₅	–	0.60*
GSL	0.60*	–
Mean July temperature n = 31	0.69*	0.17^n.s.

Notes:

n: sample size (number of years in analysis); ** p <0.0001; * p < 0.001; ^n.s. not significant.

The relationship between GDD₅ and raw C. tetragona shoot length growth over the entire range of GDD₅ (0–800) is not linear, but it is linear over the range of GDD₅ occurring on Svalbard, (0–200, Figure 3).

Figure 3.

Relationship between GDD₅ (x-axis) and raw shoot length (mm; y-axis) across all sites (a) and on Svalbard alone (b). Diamonds, squares and circles represent data points from Ny-Ålesund, Endalen, and Abisko respectively. * p<0.0001.

GDD₅ reconstruction

The standardized Ny-Ålesund and Endalen shoot length chronologies correlated strongly with Svalbard Airport GDD₅ (R²=0.55 and R²=0.51, respectively; p<0.0001). Together the chronologies share 61% variance with Svalbard Airport GDD₅. Consequently, two transfer function models were developed, one for the early (1912–1930 and 1957–1972) calibration period and one for the late (1973–2008) calibration period (Table 4a). Both models were capable of producing reliable GDD₅ estimates as shown by the verification statistics (Table 4b). The amount of common variance between estimates resulting from both models and measured GDD₅ was lower over the early period. The early model performed better than the late calibration model as shown by higher RE and CE values and a better sign test statistic (Table 4b). This model was therefore used for reconstruction of GDD₅ (Figure 4).

Table 4.

(a) Multiple linear regression transfer functions models calculated over the early and late calibration periods with standardized shoot length growth at Ny-Ålesund (NyA) and Endalen (End) as predictors of GDD₅. * p < 0.0001. (b) Calibration and verification statistics for the reconstruction of Svalbard Airport GDD₅ based on linear transfer functions between the standardized shoot growth chronologies of Ny-Ålesund and Endalen. RE: Reduction of Error; CE: Coefficient of Efficiency; PM: Product Means statistic (see Briffa et al., 1988; Fritts, 1976; Fritts et al., 1990 for details on these statistics); *p < 0.05, **p < 0.01, ***p ≤ 0.0001. (c) Pearson correlation coefficients (r) and coefficients of determination (R²; values between brackets) between instrumental July temperatures and instrumental/reconstructed GDD₅. *p < 0.0001.

(a)
	Period		Model		d.f. (regression/residual)		F	R² (R²-adj)
Early calibration	1912–1930 and 1957–1972		GDD₅ = −87.06 + (52.81 × NyA) + (108.93 × End)		34 (2/32)		13.600	0.46 (0.43)*
Late calibration	1973–2008		GDD₅ = −79.70 + (87.20 × NyA) + (68.00 × End)		35 (2/33)		25.545	0.61 (0.58)*
(b)
Chronologies used; calibration period	Verification period	R² total period (1912–1930 and 1957–2008)	R² calibration	R² verification	RE	CE	Sign test (hit/miss)	PM
Ny-Ålesund and Endalen; 1912–1930 and 1957–1972	1973–2008	0.61	0.46	0.58	0.75	0.58	30/5**	4.37***
Ny-Ålesund and Endalen; 1973–2008	1912–1930 and 1957–1972	0.61	0.61	0.43	0.71	0.43	24/9*	5.41***
(c)
	Mean July temperature 1912–1930 and 1957–2008		Mean July temperature 1931–1956			Mean July temperature 1912–2008
Instrumental GDD₅	0.90 (0.81)*		–			–
Reconstructed GDD₅	0.70 (0.49)*		0.70 (0.49)*			0.69 (0.47)*

Figure 4.

(a) Reconstruction of Svalbard Airport GDD₅ (black dotted line) with error (± standard error of the prediction; grey dotted lines), 5 yr moving average (black line), and instrumental GDD₅ (grey line). (b) Mean July temperatures (black dotted line) with 5 yr moving average (black line) and instrumental GDD₅ (grey line).

Most fluctuations present in measured GDD₅ are also present in the reconstructed series (Table 4b and Figure 4). The largest offsets between reconstructed and measured GDD₅ occur during the early period of the instrumental record (1912–1930), with both under- and overestimation of GDD₅, and in the period 2005–2007, when GDD₅ was underestimated. Linear regression through reconstructed GDD₅ between 1885 and 2008 shows a positive trend, suggesting an increase in GDD₅ values (annual cumulative heat sum >5°C) with 2.65 per decade on average. Linear regression through measured data complemented with the GDD₅ reconstruction suggests an average increase of GDD₅ values with 3.12 per decade over the same period. The reconstruction suggests that GDD₅ was slightly decreasing between 1885 and 1912. Afterwards, GDD₅ increased until 1941, followed by a decrease, which lasted until the early 1960s. Since then GDD₅ values increased with 14.86 per decade on average (1960–2010).

Discussion

We found that growing season intensity measured as growing degree-days above 5°C (GDD₅) is a better predictor of annual shoot length of Cassiope tetragona than mean monthly temperature data at its northern distribution boundary, at its assumed climatic optimum, and near its European southern distribution limit. While not fully linear over the entire gradient studied here, the relationship between GDD₅ and C. tetragona shoot length is linear on the High Arctic archipelago of Svalbard. Through a linear transfer function derived from calibration between the two longest growth chronologies developed for the area thus far and local instrumental GDD₅, reliable estimates of past Svalbard Airport GDD₅ were calculated, which share over 60% of variance with the instrumental record. This shows that information on past climate change in the High Arctic is recorded by C. tetragona growth. A more uniform coverage of past Arctic climate can therefore be obtained through the measurement of past growth of this dwarf shrub, which is a common component of Arctic tundra ecosystems.

Climate–growth analysis

GDD₅ is the stronger predictor of growth at all sites compared with GSL and monthly mean temperatures. This is probably caused by the fact that warm days outside July or August are incorporated in GDD₅. The GDD₅–shoot length growth relationship was the weakest near Abisko, the southern distribution limit of C. tetragona in Europe. Potentially, the daily mean temperatures we used for the calculation of Abisko GDD₅ are too high, as our research location there was located at a higher elevation (500 m a.s.l.) than the weather station (388 m a.s.l.). Temperature correction for elevation would result in a shift of the Abisko data points to the left in Figure 3 (left), but would still not result in a linear relationship. For this study, we followed the reasoning of Havström et al. (1993), who argued that the cooling effect of Lake Torneträsk nearby the meteorological station at Abisko is likely similar to the altitudinal effect at 450 m a.s.l. and assumed that no correction for the altitude effect was required. The weaker GDD₅–growth correlation at Abisko suggest that GDD₅ becomes less important for C. tetragona growth at its southern distribution limit, which is in line with Havström et al. (1993), who suggested competition for light and nutrients as more limiting factors for growth near Cassiope’s southern limit. Another explanation for the observation that the GDD₅–shoot length growth relationship at Abisko was not located on the same linear line as the one at Svalbard, could be that the definition of GDD₅, which we used, overestimates the actual biological GDD₅. Therefore, we also calculated GDD₅ for Abisko using a different definition of the growing season (GS), in which the start and end of the GS is defined as the last day of the first/last 5-day period with average temperature above 5°C after/before the last/first occurrence of late/early frosts (Jones et al., 2002; Wagner-Cremer et al., 2010). GDD₅ is then the cumulative temperature above 5°C during the defined GS. Although the use of this alternative definition of GDD₅ slightly improved the GDD₅–standardized growth relationship at Abisko (R²=0.40 instead of R²=0.37), it did not improve the GDD₅–raw growth relationship. Also, the GDD₅–growth relationship over the complete studied gradient (0–800) remained non-linear.

Non-linear climate–growth relationships are potentially problematic for climate reconstructions based on growth parameters, as they can lead to an underestimation of the predicted climate variables when linear transfer functions are used (Esper and Frank, 2009; Loehle, 2009). The growth response of C. tetragona to mean July temperatures over the same High to Sub Arctic gradient as studied here (Ny-Alesund–Endalen–Abisko) has been found to be linear (Weijers et al., 2012). Thus, while mean July temperatures appear to remain equally important to C. tetragona shoot length growth over a broad gradient, temperature increases at the southern limit of the species’ range during other parts of the growing season, and a lengthening of the growing season, seem to be less important to growth. These factors may, however, be more important for reproduction, as flower production in C. tetragona has previously been demonstrated to be more sensitive to summer temperatures than shoot length growth (Johnstone, 1995; Johnstone and Henry, 1997). To prevent underestimation of past warm events, it is therefore better to reconstruct mean July temperatures instead of GDD₅, in case strong fluctuations in past raw shoot length growth are found, comparable with shoot length currently occurring at Abisko. Still, the Ny-Ålesund and Endalen shoot length chronologies are suitable for the reconstruction of GDD₅, as past growth was smaller than current growth in Svalbard and the GDD₅–growth relationship there is linear.

Standardization

After standardization some autocorrelation was still present in the chronologies (Table 1). However, the Svalbard Airport and Abisko GDD₅ series are characterized by autocorrelation as well (first-, second- and third-order autocorrelation at Svalbard Airport between 1957 and 2010 is 0.281, 0.319, and 0.352, respectively (p<0.05); first-order autocorrelation at Abisko between 1978 and 2009 is 0.425 (p<0.05), second- and third-order autocorrelations are not significant there). Part of the autocorrelation present in the raw chronologies can therefore have been caused by these long-term trends in climate. When all autocorrelation is removed from the shoot length chronologies through application of different detrending methods (e.g. cubic splines), part of the climate signal is lost (see also Weijers et al., 2012). One reason to believe the correlations found between the autocorrelated standardized shoot length and GDD₅ series on Svalbard are not (partly) spurious, is that the raw shoot length series seem to be capable of capturing trends over a range in GDD₅ between 0 and 200 (Figure 3). A second reason is that C. tetragona shoot length growth has been proven to respond positively to artificially increased summer temperatures (and consequently GDD₅), both at a research site near Endalen (Rozema et al., 2009) and near Ny-Ålesund (Havström et al., 1993; Weijers et al., 2012). Finally, the autocorrelation patterns in the standardized Ny-Ålesund and Endalen chronologies are similar to those found in the Svalbard Airport GDD₅ record.

GDD₅ reconstruction

We reconstructed Svalbard Airport GDD₅ for the period 1857–2008 (Figure 4). We consider the reconstruction reliable from 1885 onward, despite the fact that EPS values of the standardized Ny-Ålesund chronology are below 0.85 between 1885 and 1894 (0.82–0.85). The standardized Ny-Ålesund chronology corresponds largely with the standardized Endalen chronology from 1885 onward, suggesting a strong common external signal in both chronologies. Reconstructed GDD₅ before 1885 is not further discussed below.

We believe that the relatively lower amount of common variance between measured and reconstructed GDD₅ over the early period (1912–1930 and 1957–1972) can largely be ascribed to the fact that daily mean temperatures over this period needed to be transferred, which inevitably led to loss of information.

The Svalbard Airport temperature record starts in August 1972, but the homogenised Svalbard Airport mean monthly temperature record has been extended back to December 1911 through the inclusion of other series from neighbouring sites. The period 1931–1956 was mainly completed with data from the Russian settlement in Barentsburg and the Longyearbyen weather stations (Kohler et al., 2002; Nordli and Kohler, 2003). However, daily mean temperature data from these stations are not publicly available for this period, resulting in a gap in instrumental GDD₅. Still, Svalbard Airport mean July temperatures share a large amount of variance with measured GDD₅ over the periods 1912–1930 and 1957–2008 (R²=0.81, n=70, p<0.0001; Table 4c). This enables verification of reconstructed GDD₅ over the period with missing data through linear regression analysis with mean July temperature. The amount of common variance over the periods 1912–1930 and 1957–2008 between reconstructed GDD₅ and mean July temperature is lower (R²=0.49, n=71, p<0.0001; Table 4c) than that between measured GDD₅ and mean July temperature. Over the ‘gap’ (1931–1956) the amount of common variance is also 49% (n=26, p<0.0001). Therefore, we regard this part of the reconstruction to be just as reliable as the parts between 1912 and 1930, and 1957 and 2008.

The period of warming present in the GDD₅ reconstruction between 1915 and 1941, the subsequent cooling into the 1960s, and the warming thereafter, are also present in observational annual mean surface air temperatures (SAT) between 70 and 90°N (Johannessen et al., 2004), 60 and 90°N (McBean et al., 2005) and from 62° northward (Polyakov et al., 2003). The early 20th century warming with peak temperatures in the 1930s and 1940s has also been observed in a homogenized Arctic temperature record, based on instrumental data from Greenland’s west coast (Vinther et al., 2006), in some varved lake sediments from Baffin Island in the Canadian Arctic (Thomas and Briner, 2009), but not in other varved sediments from the same area (Moore et al., 2001). The increase in reconstructed GDD₅ from the 1960s onward is in line with Kaufman et al. (2009), who concluded that over the past five decades a long-term cooling trend in the Arctic was reversed. Recently, it has been shown that this trend has likely been stronger than previously reported, but was still reversed over the past five decades (Esper et al., 2012). Another Arctic-wide, multiproxy temperature reconstruction, however, shows cooling from the 1950s into the 1980s (Overpeck et al., 1997).

We also compared the GDD₅ reconstruction with monthly North Atlantic Oscillation (NAO) and Arctic Oscillation (AO) indices, but found no impact of these circulation patterns on GDD₅. This is perhaps not surprising as monthly surface air temperatures (SAT) between 1982 and 1999 were found to be positively correlated to AO indices in Northern Europe and north Russia, and negatively in Greenland and northern Canada. In Svalbard, more or less located between those areas, there was no significant relationship between AO and monthly SAT during this period (Wang and Key, 2005). Also, SAT variance between Iceland and Svalbard was previously found to be practically uncorrelated with NAO (Wood and Overland, 2010).

Potentially, the principal weakness of the use of shoot length chronologies of C. tetragona as an Arctic climate proxy is their relatively short time span (> 180 yr) compared with ice-core records, for example, which often span multiple centuries (e.g. Isaksson et al., 2005a). Still, C. tetragona shoot length chronologies are longer than most other shrub-based growth chronologies (e.g. based on growth rings from Salix alaxensis (Zalatan and Gajewski, 2006), Salix arctica (Schmidt et al., 2006, 2010; Woodcock and Bradley, 1994), Salix lanata (Forbes et al., 2010), Empetrum hermaphroditum (Bär et al., 2006, 2007, 2008), and growth and reproduction chronologies of Cassiope mertensiana (Rayback et al., 2010)). Dendrochronological studies on shrubs are, however, invaluable for studies on the impact of climate change on these species, such as shrub expansion in tundra ecosystems (Hallinger et al., 2010; Myers-Smith et al., 2011). Furthermore, as most Arctic instrumental records only span some decades, the extension of these records with several decades with estimates based on C. tetragona growth and/or other shrubs could provide important insights on past non-anthropogenic climate variability across the Arctic.

Shoot length chronologies of C. tetragona have the advantage over several Arctic climate proxies, such as pollen records (Birks, 1991; Rozema et al., 2006) and sediment cores (Svendsen and Mangerud, 1997), being precisely dated, annually resolved, and well replicated. Ice-core records and varved lake sediments are invaluable sources of past (Arctic) climate, able to capture low and high frequency climate changes, with (near) annual resolution (e.g. Divine et al., 2011; Thomas and Briner, 2009). However, such records are not easily replicated. Also, they do not necessarily best represent the local climate. The Svalbard δ¹⁸O records, for example, correlate better with temperatures from Vardø, mainland Norway, than with the Svalbard Airport record (Divine et al., 2011; Isaksson et al., 2003, 2005a, 2005b). Still, high frequency changes in winter temperatures on Svalbard are recorded in the oxygen isotope composition of ice-cores (Divine et al., 2011). Past inter- and intra-annual (summer and winter) climate change can thus potentially be studied by combining ice-core with C. tetragona proxy data.

Conclusions

We have shown that growing season intensity measured as GDD₅ is a better predictor of C. tetragona shoot length growth than mean monthly temperatures and GSL at its European northern and southern distribution limit, as well as at its assumed climatic optimum. This greater predictive capacity of GDD₅ allows for an extension of local High Arctic instrumental records with accurate estimates of past GDD₅ based on C. tetragona shoot length chronologies. Our reconstruction of Svalbard Airport GDD₅ (1885–2008) shows that the growing season in Svalbard has become more intense from the early 1960s onward. Before that time, there were both positive and negative trends in growing season intensity.

Footnotes

Acknowledgements

We thank James Weedon for collecting plants near Abisko and Lennard van Rij for his support with the collection of plants near Ny-Ålesund. We thank the anonymous reviewers whose comments led to great improvement of the manuscript.

Funding

The research presented forms part of the project 851.40.051 ‘Long-lived evergreen shrubs from polar ecosystems as monitors of present and past climate change’, funded by IPY-NWO.

References

Aanes

Saether

Smith

. (2002) The Arctic Oscillation predicts effects of climate change in two trophic levels in a high-arctic ecosystem. Ecology Letters 5: 445–453.

Bär

Bräuning

Löffler

(2006) Dendroecology of dwarf shrubs in the high mountains of Norway – A methodological approach. Dendrochronologia 24: 17–27.

Bär

Bräuning

Löffler

(2007) Ring-width chronologies of the alpine dwarf shrub Empetrum hermaphroditum from the Norwegian mountains. Iawa Journal 28: 325–338.

Bär

Pape

Bräuning

. (2008) Growth-ring variations of dwarf shrubs reflect regional climate signals in alpine environments rather than topoclimatic differences. Journal of Biogeography 35: 625–636.

Birks

(1991) Holocene vegetational history and climatic change in west Spitsbergen – Plant macrofossils from Skardtjorna, an Arctic lake. The Holocene 1: 209–218.

Briffa

Jones

Pilcher

. (1988) Reconstructing summer temperatures in northern Fennoscandinavia back to AD 1700 using tree-ring data from Scots Pine. Arctic and Alpine Research 20: 385–394.

Callaghan

Carlsson

Tyler

NJC

(1989) Historical records of climate-related growth in Cassiope tetragona from the Arctic. Journal of Ecology 77: 823–837.

Carter

(1998) Changes in the thermal growing season in Nordic countries during the past century and prospects for the future. Agricultural and Food Science in Finland 7: 161–179.

Cook

(1985) A time series analysis approach to tree-ring standardization. PhD Thesis, University of Arizona.

10.

Divine

Isaksson

Martma

. (2011) Thousand years of winter surface air temperature variations in Svalbard and northern Norway reconstructed from ice core data. Polar Research 30, 7379, DOI: 10.3402/polar.v30i0.7379. Available at: http://www.polarresearch.net/index.php/polar/article/view/7379

11.

DNMI (2011) eKlima, Free Access to Weather- and Climate Data from Norwegian Meteorological Institute from Historical Data to Real Time Observations. DNMI.

12.

Elvebakk

(1999) Bioclimatic delimitation and subdivision of the Arctic. In: Nordal

Razzhivin

(eds) The Species Concept in the High North – A Panarctic Flora Initiative. Oslo: The Norwegian Academy of Science and Letters, pp. 81–112.

13.

Esper

Frank

(2009) Divergence pitfalls in tree-ring research. Climatic Change 94: 261–266.

14.

Esper

Frank

Timonen

. (2012) Orbital forcing of tree-ring data. Nature Climate Change DOI: 10.1038/nclimate1589.

15.

Forbes

Fauria

Zetterberg

(2010) Russian Arctic warming and ‘greening’ is closely tracked by tundra shrub willows. Global Change Biology 16: 1542–1554.

16.

Førland

Skaugen

Benestad

. (2004) Variations in thermal growing, heating, and freezing indices in the Nordic Arctic, 1900–2050. Arctic, Antarctic, and Alpine Research 36: 347–356.

17.

Fritts

(1976) Tree Rings and Climate. Caldwell NJ: The Blackburn Press.

18.

Fritts

Guiot

Gordon

(1990) Verification. In: Cook

Kairiukstis

(eds) Methods of Dendrochronology: Applications in the Environmental Sciences. Dordrecht: Kluwer Academic, pp. 178–185.

19.

Grissino-Mayer

(2001) Evaluating crossdating accuracy: A manual and tutorial for the computer program COFECHA. Tree-Ring Research 57: 205–221.

20.

Hallinger

Manthey

Wilmking

(2010) Establishing a missing link: Warm summers and winter snow cover promote shrub expansion into alpine tundra in Scandinavia. New Phytologist 186: 890–899.

21.

Havström

Callaghan

Jonasson

(1993) Differential growth responses of Cassiope tetragona, an arctic dwarf-shrub, to environmental perturbations among three contrasting high- and subarctic sites. Oikos 66: 389–402.

22.

Havström

Callaghan

Jonasson

. (1995) Little Ice-Age temperature estimated by growth and flowering differences between subfossil and extant shoots of Cassiope tetragona, an arctic heather. Functional Ecology 9: 650–654.

23.

Holmes

(1983) Computer-assisted quality control in tree-ring dating and measurement. Tree-Ring Bulletin 43: 69–78.

24.

Intergovernmental Panel on Climate Change (2007) Climate Change 2007: Synthesis Report. An Assessment of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press.

25.

Isaksson

Divine

Kohler

. (2005a) Climate oscillations as recorded in Svalbard ice core delta O-18 records between AD 1200 and 1997. Geografiska Annaler Series A-Physical Geography 87A: 203–214.

26.

Isaksson

Hermanson

Sheila

. (2003) Ice cores from Svalbard – Useful archives of past climate and pollution history. Physics and Chemistry of the Earth 28: 1217–1228.

27.

Isaksson

Kohler

Pohjola

. (2005b) Two ice-core delta O-18 records from Svalbard illustrating climate and sea-ice variability over the last 400 years. The Holocene 15: 501–509.

28.

Johannessen

Bengtsson

Miles

. (2004) Arctic climate change: Observed and modelled temperature and sea-ice variability. Tellus A 56: 328–341.

29.

Johnstone

(1995) Responses of Cassiope tetragona, a high arctic evergreen dwarf shrub, to variations in growing season temperature and growing season length at Alexandra Fiord, Ellesmere Island. M.Sc. thesis, University of British Columbia.

30.

Johnstone

Henry

GHR

(1997) Retrospective analysis of growth and reproduction in Cassiope tetragona and relations to climate in the Canadian High Arctic. Arctic and Alpine Research 29: 459–469.

31.

Jones

Briffa

Osborn

. (2002) Relationships between circulation strength and the variability of growing-season and cold-season climate in northern and central Europe. The Holocene 12: 643–656.

32.

Kaufman

Schneider

McKay

. (2009) Recent warming reverses long-term Arctic cooling. Science 325: 1236–1239.

33.

Kohler

Nordli

Brandt

. (2002) Svalbard Temperature and Precipitation, Late 19th Century to the Present. Final Report on ACIA-funded Project. Oslo: Norwegian Polar Institute.

34.

Lindholm

Jalkanen

Salminen

. (2011) The height-increment record of summer temperature extended over the last millennium in Fennoscandia. The Holocene 21: 319–326.

35.

Loehle

(2009) A mathematical analysis of the divergence problem in dendroclimatology. Climatic Change 94: 233–245.

36.

McBean

Alekseev

Chen

. (2005) Arctic climate: Past and present. In: Symon

Arris

Heal

(eds) Arctic Climate Impact Assessment. Cambridge: Cambridge University Press, pp. 21–60.

37.

Miller

Brigham-Grette

Alley

. (2010) Temperature and precipitation history of the Arctic. Quaternary Science Reviews 29: 1679–1715.

38.

Moore

Hughen

Miller

. (2001) Little Ice Age recorded in summer temperature reconstruction from vared sediments of Donard Lake, Baffin Island, Canada. Journal of Paleolimnology 25: 503–517.

39.

Myers-Smith

Forbes

Wilmking

. (2011) Shrub expansion in tundra ecosystems: Dynamics, impacts and research priorities. Environmental Research Letters 6: 045509.

40.

Nordli

Kohler

(2003) The Early 20th Century Warming. Daily Observations at Green Harbour, Grønfjorden, Spitsbergen. Norwegian Meteorological Institute, Klima Report 12/03, pp. 1–20.

41.

Overland

Spillane

Percival

. (2004) Seasonal and regional variation of pan-Arctic surface air temperature over the instrumental record. Journal of Climate 17: 3263–3282.

42.

Overpeck

Hughen

Hardy

. (1997) Arctic environmental change of the last four centuries. Science 278: 1251–1256.

43.

Polyakov

Bekryaev

Alekseev

. (2003) Variability and trends of air temperature and pressure in the maritime Arctic, 1875–2000. Journal of Climate 16: 2067–2077.

44.

Przybylak

(2000) Temporal and spatial variation of surface air temperature over the period of instrumental observations in the Arctic. International Journal of Climatology 20: 587–614.

45.

Rayback

Henry

GHR

(2005) Dendrochronological potential of the Arctic dwarf-shrub Cassiope tetragona. Tree-Ring Research 61: 43–53.

46.

Rayback

Henry

GHR

(2006) Reconstruction of summer temperature for a Canadian High Arctic site from retrospective analysis of the dwarf shrub, Cassiope tetragona. Arctic, Antarctic, and Alpine Research 38: 228–238.

47.

Rayback

Henry

GHR

Lini

(2012) Multiproxy reconstructions of climate for three sites in the Canadian High Arctic using Cassiope tetragona. Climatic Change 114: 593–619.

48.

Rayback

Lini

Berg

(2010) Multiple climate signals characterize Casiope mertensiana chronologies for a site on Mount Rainier, Washington, USA. Physical Geography 31: 79–106.

49.

Rozema

Boelen

Doorenbosch

. (2006) A vegetation, climate and environment reconstruction based on palynological analyses of high arctic tundra peat cores (5000–6000 years BP) from Svalbard. Plant Ecology 182: 155–173.

50.

Rozema

Weijers

Broekman

. (2009) Annual growth of Cassiope tetragona as a proxy for Arctic climate: Developing correlative and experimental transfer functions to reconstruct past summer temperature on a millennial time scale. Global Change Biology 15: 1703–1715.

51.

Schmidt

Baittinger

Forchhammer

(2006) Reconstructing century-long snow regimes using estimates of high arctic Salix arctica radial growth. Arctic Antarctic and Alpine Research 38: 257–262.

52.

Schmidt

Baittinger

Kollman

. (2010) Consistent dendrochronological response of the dioecious Salix arctica to variation in local snow precipitation across gender and vegetation types. Arctic Antarctic and Alpine Research 42: 471–475.

53.

Skaugen

Tveito

(2004) Growing-season and degree-day scenario in Norway for 2021–2050. Climate Research 26: 221–232.

54.

Svendsen

Mangerud

(1997) Holocene glacial and climatic variations on Spitsbergen, Svalbard. The Holocene 7: 45–57.

55.

Thomas

Briner

(2009) Climate of the past millennium inferred from varved proglacial lake sediments on northeast Baffin Island, Arctic Canada. Journal of Paleolimnology 41: 209–224.

56.

Trenberth

Jones

Ambenje

. (2007) Observations: Surface and atmospheric climate change. In: Solomon

Qin

Manning

. (eds) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge and New York: Cambridge University Press.

57.

Vinther

Andersen

Jones

. (2006) Extending Greenland temperature records into the late eighteenth century. Journal of Geophysical Research 111: D11105.

58.

Wagner-Cremer

Finsinger

Moberg

(2010) Tracing growing degree-day changes in the cuticle morphology of Betula nana leaves: A new micro-phenological palaeo-proxy. Journal of Quaternary Science 25: 1008–1017.

59.

Walker

Raynolds

Daniëls

FJA

. (2005) The circumpolar Arctic vegetation map. Journal of Vegetation Science 16: 267–282.

60.

Wang

Key

(2005) Arctic surface, cloud, and radiation properties based on the AVHRR Polar Pathfinder dataset. Part II: Recent trends. Journal of Climate 18: 2575–2593.

61.

Warming

(1908) The Structure and Biology of Arctic Flowering Plants. I. Ericineae (Ericaceae, Pyrolaceae). I. Morphology and Biology. Copenhagen: Commission for Scientific Research in Greenland.

62.

Weijers

Broekman

Rozema

(2010) Dendrochronology in the High Arctic: July air temperatures reconstructed from annual shoot length growth of the circumarctic dwarf shrub Cassiope tetragona. Quaternary Science Reviews 29: 3831–3842.

63.

Weijers

Greve Alsos

Bronken Eidesen

. (2012) No divergence in Cassiope tetragona: Persistance of growth response along a latitudinal temperature gradient and under multi-year experimental warming. Annals of Botany 110: 653–665.

64.

Wigley

TML

Briffa

Jones

(1984) On the average value of correlated time-series, with applications in dendroclimatology and hydrometeorology. Journal of Climate and Applied Meteorology 23: 201–213.

65.

Wood

Overland

(2010) Early 20th century Arctic warming in retrospect. International Journal of Climatology 30: 1269–1279.

66.

Woodcock

Bradley

(1994) Salix arctica (Pall.): Its potential for dendroclimatological studies in the High Arctic. Dendrochronologia 12: 11–22.

67.

Zalatan

Gajewski

(2006) Dendrochronological potential of Salix alaxensis from the Kuujjua River area, western Canadian Arctic. Tree-Ring Research 62: 75–82.

Reconstructing High Arctic growing season intensity from shoot length growth of a dwarf shrub

Abstract

Keywords

Introduction

Material and methods

Site description and sampling

Chronology development

Climate data

Growing degree-day sums

Climate–growth analyses

GDD5 reconstruction: calibration and verification

Results

Characteristics of the shoot length chronologies

Climate–growth analysis

GDD5 reconstruction

Discussion

Climate–growth analysis

Standardization

GDD5 reconstruction

Conclusions

Footnotes

Acknowledgements

Funding

References

GDD₅ reconstruction: calibration and verification

GDD₅ reconstruction

GDD₅ reconstruction