Temperature-induced recruitment pulses of Arctic dwarf shrub communities
Summary
- The effects of climate change on Arctic ecosystems can range between various spatiotemporal scales and may include shifts in population distribution, community composition, plant phenology, primary productivity and species biodiversity. The growth rates and age structure of tundra vegetation as well as its response to temperature variation, however, remain poorly understood because high-resolution data are limited in space and time.
- Anatomical and morphological stem characteristics were recorded to assess the growth behaviour and age structure of 871 dwarf shrubs from 10 species at 30 sites in coastal East Greenland at ˜70°N. Recruitment pulses were linked with changes in mean annual and summer temperature back to the 19th century, and a literature review was conducted to place our findings in a pan-Arctic context.
- Low cambial activity translates into estimated average/maximum plant ages of 59/204 years, suggesting relatively small turnover rates and stable community composition. Decade-long changes in the recruitment intensity were found to lag temperature variability by 2 and 6 years during warmer and colder periods, respectively (r = 0.851961–2000 and 1881–1920).
- Synthesis. Our results reveal a strong temperature dependency of Arctic dwarf shrub reproduction, a high vulnerability of circumpolar tundra ecosystems to climatic changes, and the ability of evaluating historical vegetation dynamics well beyond the northern treeline. The combined wood anatomical and plant ecological approach, considering insights from micro-sections to community assemblages, indicates that model predictions of rapid tundra expansion (i.e. shrub growth) following intense warming might underestimate plant longevity and persistence but overestimate the sensitivity and reaction time of Arctic vegetation.
Introduction
The effects of recent climate change on the productivity and functioning of Arctic ecosystems have been reported across various spatiotemporal scales within the circumpolar tundra (Serreze & Francis 2006; Bunn et al. 2007; Kaufman et al. 2009; Post et al. 2009; Walther 2010). Examples of the influence of warmer temperatures include disruptions of trophic interaction and the seasonal synchronization of ecological system components (Ims, Henden & Killengreen 2008; Post et al. 2008; Hansen et al. 2013; Høye et al. 2013), supplemented by shifts in population distribution, community composition, plant phenology, primary productivity and animal behaviour (Hope et al. 2013; Prost et al. 2013).
Modelled and observed impacts of increasing temperatures on vegetation cover and community composition (Sturm, Racine & Tape 2001; Tape, Sturm & Racine 2006; Walker et al. 2006; Verbyla 2008; Elmendorf et al. 2012a,b; Macias-Fauria et al. 2012; Hope et al. 2013; Normand et al. 2013; Pearson et al. 2013), as well as on the annual growth trends of trees and shrubs, have also been documented at many high-northern latitude sites (Harsch et al. 2009; Harsch & Bader 2011; Myers-Smith et al. 2011; and references therein). However, understanding the response of local to regional plant communities to alteration in climatological means and extremes remains challenging because the spatial and temporal extent of appropriate data sets is often limited (Büntgen & Schweingruber 2010). Imprecise assessments of past, present and projected changes in tundra expansion and subsequent biomass production translate into model uncertainty (Chapin et al. 2005; Dorrepaal 2007; McGuire et al. 2009, 2010; Schuur et al. 2009; Keenan et al. 2013; Pearson et al. 2013), which hampers the prediction of regional to global water, energy and carbon cycles and budgets.
Knowledge on the growth and age structure of Arctic dwarf shrubs was first generated by a pioneering study in Greenland (Kraus 1873), and only a few comparable approaches followed (Good 1927; Molisch 1938; Miller 1975; Parsons et al. 1994). Many current studies have considered shrubs and dwarf shrubs from lower-latitude environments in the western North American Arctic or the Scandinavian and eastern European sub-Arctic (Myers-Smith et al. 2011). Reasons for the lack of studies are technical and economical hurdles associated with the (field) collection, (anatomical) preparation and (dendrochronological) analysis of well replicated and sufficiently long dwarf shrub records.
Most studies to date have focused on variation in shrub cover and abundance, whereas relatively little energy has been invested in detecting changes in growth and recruitment pulses (Myers-Smith et al. 2011; Elmendorf et al. 2012a,b). Nevertheless, efforts have recently been made to detect climatological fingerprints in successive patterns of annual growth rings formed by some Arctic dwarf shrub species (see http://shrubhub.biology.ualberta.ca/ for a detailed literature overview). Warmer temperatures are mainly reflected by either increased individual growth rates or changes in the age structure of populations. Improved individual growth rates will generally lead to larger coverage, but might also affect the local community structure due to alterations in the competitive balance among species. Recruitment pulses will cause fluctuations in the age structure of populations, that is the establishment of new individuals within or even beyond a species’ current altitudinal and latitudinal distribution range. Recruitment is thus the driving mechanism for the geographical expansion of shrubs beyond their growth margin. Decoding factors controlling the growth and age structure of populations are therefore indispensable in reconstructing shrub dynamics in response to environmental change.
An active discussion of possible cross-dating issues caused by imprecise ring boundaries, irregular growth disturbances and locally absent or even completely missing rings has stimulated research at the interface of wood anatomy and dendroecology, including dendroclimatology (Büntgen & Schweingruber 2010; Hallinger & Wilmking 2011; Wilmking et al. 2012; Buchwal et al. 2013; Myers-Smith et al. 2015). This modern research avenue also underlines the need for alternative methods of estimating past vegetation dynamics and subsequent rates of dry matter production and carbon allocation in high-latitude (and -altitude) environments well above the northern (and upper) treeline. The importance of such studies is further reinforced by the fact that more than half of the Earth's landmass is covered by non-forested vegetation (Friedl et al. 2010), whereas forest trees only represent ~20% (Büntgen, Psomas & Schweingruber 2014). Unravelling the frequency and severity, as well as the causes and consequences of biotic and abiotic responses in circumpolar tundra ecosystems (Wookey et al. 2009; Bonfils et al. 2012; Cahoon et al. 2012; Heskel et al. 2013, 2014), remains a pending task for the interdisciplinary arena of global change ecology.
Here, we hypothesize that species-specific and community-wide changes in the growth and age structure of high Arctic tundra vegetation reflect trends and extremes in different seasonal temperature means. To test this assumption, we compiled wood anatomical and stem morphological information of 871 dwarf shrubs from 10 species at 30 sites in coastal East Greenland. We then analysed possible linkages between climate variability and the intensity of shrub recruitment. A literature review placed our results in the board context of past and ongoing Arctic vegetation dynamics. Finally, we discuss data and methodological caveats to provide a critical impetus for innovative efforts at the interface of wood anatomy and dendroecology. This study ultimately aims at enhancing research in functional ecology beyond the geographical limits of forest growth.
Materials and methods
Sampling Design
A total of 945 individual dwarf shrubs were sampled from 10 widespread high Arctic tundra species (Fig. 1a,b): Arctostaphylos alpinus, Cassiope hypnoides, Cassiope tetragona, Empetrum nigrum, Dryas octopetala, Rhododendron lapponicum, Vaccinium uliginosum, Betula nana, Salix arctica and Salix herbacea. The natural distribution of these species is quasi-circumpolar (Elvebakk 1999; Walker et al. 2002), and certain genera are also abundant in alpine environments across Eurasia and North America (Hulten 1968). All specimens were collected within 30 plot sites along the coastline of East Greenland near Scoresbysund, that is Ittoqqortoormiit, at 70°26′6″N and 21°58′100″W. Species- and site-specific density of the vegetation cover defined plot sizes, since about 30–40 specimens were collected. The study area, characterized by sharp biogeographic gradients (Karlsen & Elvebakk 2003), is located at the southern margin of the maximum pack-ice extent, where short vegetation periods are constrained by a cold and dry climate. Temperature means and precipitation totals between June and August are 1.9 °C and 94 mm, respectively (Schweingruber et al. 2013). Fairly rich botanical and geological descriptions of the study area have been achieved as a result of initial environmental assessments by the petroleum industry (see Karlsen & Elvebakk 2003 for a biogeographic overview). Detailed knowledge exists for the interior part of Jameson Land (Bay et al. 1984; Bay, Holt & Miljøundersøgelser 1986), as well as for the area near Ittoqqortoormiit (Hartz 1895; Kruuse 1905; Menzies 1933; Gelting 1934; Halliday 1974), including inventories of the lichen flora (Hansen 1982).
The sampling sites, either located on crystalline orthogneiss with large granite boulders (22 plots) or sedimentary sandstone (eight plots) (Fig. 1a), range from 1 to 320 m asl. Orthogneiss is ample in silicates, and plant growth is generally restricted to the lee side of large boulders or between boulders. A comparatively low overall amount of plant-available soil nutrients distinguishes plots within the sandstone region from those on intrusive bedrock. Detailed information on the phenology, ecology and geology of each of the 30 individual plot sites is provided in the Supplementary Information (Table S1 in Supporting Information). Spatial patterns in permafrost distribution and thickness of the active permafrost layer are unknown. A rigorous sampling of the completely preserved above- and below-ground plant stem was targeted for each specimen. In this way, the full variety of prevailing dwarf shrub sizes and ages was captured at each individual plot site, and the thickest part of the often-complex branch and/or root systems was included. Whenever possible, the oldest remaining part of the taproot was analysed. This section ideally comprises the maximum number of rings and therefore enables the most accurate determination of plant ages and recruitment dates. Because preservation of the taproot was not always guaranteed, we additionally recorded the existence or absence of the pith for differentiation between shoots and roots, respectively. Additionally, some specimens are clonal. Their plant ages thus reflect either the germination date or the date of ramet formation (i.e. independent but genetically identical individuals of one clonal colony). The term ‘recruitment date’ is therefore used to indicate when independent individuals (i.e. roots or shoots) start growing and thus contribute to the community as a whole.
Anatomical Preparation and Morphological Assessment
After labelling and archiving all samples, sliding microtomes with disposable blades were used to prepare several cross sections at or near the taproot. Unstained cuts preserved in glycerol allowed ring counting without further preparation. Double staining with safranin and astra blue amplified anatomical structures (Gärtner & Schweingruber 2013). Additional sample preservation with Nawashin solution and extra staining with picric-aniline blue visualized cell contents and nuclei and thus enabled cell longevity to be estimated. After saturation with Nawashin solution for 10 min, sections were washed with water before simultaneously stained with safranin-astra blue and picric-aniline blue. A short heating at 80 °C prepared the slides for their subsequent dehydration and embedding process with ethanol, xylene and ‘Canada’ balsam (Gärtner & Schweingruber 2013). Nawashin solution consists of 10 parts 1% chromic acid, four parts 4% formaldehyde and one part acetic acid (Purvis, Collier & Walls 1964). Safranin dye consists of 0.8 g safranin powder in 100 mL of distilled water. Astra blue dye consists of 0.5 g of astra blue powder in 100 mL distilled water and 2 mL acids. Picric-aniline blue dye consists of one part saturated aniline blue and four parts saturated picric acid (trinitrophenol) dissolved in 95% ethanol.
Stem morphological analyses were performed for each microsection and include measuring of the longest xylem radius, counting the consecutive number of rings and recording the corresponding bark thickness. Particularly, straight taproot sections of 1 cm length were water saturated for 24 h, wet-weighed before and after removing the bark, dried for 1 day at 60 °C and again weighed with and without cortex. The volume of the remaining dry wood was also measured (see Table S2 in Supporting Information for site-specific values). Estimates of plant ages were obtained through rigorous ring counting without cross-dating (i.e. the visual synchronization of similar rings from different samples over time). All processing steps were performed at WSL and resulted in 871 samples for which recruitment dates were estimated and compared with changes in Greenlandic air temperature.
Linking Recruitment Pulses with Temperature Changes
Monthly resolved, homogenized temperature measurements were selected from five meteorological stations at the Danish Meteorological Institute (DMI; www.dmi.dk). These data are distributed across Greenland and span the periods 1948–2011 (Pituffik) and 1873–2011 (Mittarfik Upernavik, Mittarfik Ilulissat, Nuuk and Ivittut). All time series were transformed into anomalies relative to the 1981–2010 period. The individual station data share a strong common signal. The grand average cross-correlation is 0.67, 0.67, 0.88 and 0.84 for the summer (June–August), warm season (April–September), winter (December–February) and annual (January–December) means, respectively (1949–2011). Mean annual and summer temperature anomalies were compared with estimated phases of shrub recruitment back to 1873 AD. Possible wavelength-dependent relationships between the time series were explored by analysing their original, high-, band- and low-pass-filtered components (Büntgen et al. 2006). That is, all records were separated into different frequency domains. Split periods and temporal lags were considered to test and account for temporal changes in the obtained growth-climate relations. Recent versions of the ARSTAN/DPL software were used for time-series analyses (Cook & Krusic 2005).
To place the East Greenlandic findings within a pan-Arctic context of shrub recruitment, we performed a comprehensive literature review, including papers from a variety of different data sources (i.e. dendroecology, repeat photography and vegetation surveys). We compiled all studies referenced in Myers-Smith et al. (2011) of shrub encroachment in tundra ecosystems and then updated this review by more recent literature on shrub recruitment (until November 2014). We therefore used the search terms ‘shrub’, ‘recruitment’, ‘tundra’ and ‘age distribution’ in a non-restrictive search using Google Scholar and Web of Science. In addition, we considered our collaborative networks (e.g. ShrubHub; http://shrubhub.biology.ualberta.ca/) to identify all recently published papers. The literature survey thus comprehensively represents studies that report time series of shrub recruitment or encroachment in Arctic ecosystems. Work based on the Normalized Difference Vegetation Index (NDVI) to attribute the timing of changes in shrub cover or abundance was not considered.
Results
Cambial Activity and Plant Longevity
Exceptionally, narrow ring widths were found in all dwarf shrub species (Fig. 2a). Most samples from the 10 species fall within the narrowest ring width class < 0.05 mm, and specimens from nine different species were represented in the next two ring width classes of 0.051–0.075 and 0.076–0.100 mm. Overall, fewer samples from fewer species were included in the wider ring width classes > 0.10 mm. More specifically, 95.4% of all samples had average ring widths between 0.00 and 0.10 mm, whereas only 4.6% of the shrubs had average ring widths > 0.15 mm (Figs 2a and 3a). Radial stem thickening ranged from 0.00 to 0.05 mm year−1 for 47.4% of all shrubs sampled. The occurrence of ring width classes and the frequency of individuals within each class varied among species. The largest rings on average (across all individuals of a given species) were formed by Betula nana (0.12 mm), whereas the lowest average ring width was found for Cassiope hypnoides (0.02 mm) (Figs 2a and 3a). Comparison among species-specific growth rates was, however, biased by different sample sizes (Fig. 3a). For example, only 12 individuals were collected for Cassiope hypnoides, whereas 156 were collected for Salix arctica.
All dwarf shrub species showed exceptionally high rates of plant longevity (Figs 2b and 3a). Samples from 10 species had ages spanning 21–40 years, whereas nine species include ages between 41 and 80 years. Eight (seven) species were aged between 81 and 100 (101 and 120). Overall fewer samples from fewer species are captured by the upper age classes >120 years. Further details on site- and species-specific variation in cambial activity (i.e. ring width) and plant longevity (i.e. ring number) are summarized in Table S2 in Supporting Information together with supplementary information on wood anatomy and dendroecology. Additional insights (mean and minimum/maximum values) on species-specific values of ring width (mm), plant longevity (years) and sample replication (plants) are represented in Fig. 3. The narrowest ring widths were found in Cassiope hypnoides, followed by Cassiope tetragona and Dryas octopetala. Maximum plant longevity was attained by Dryas octopetala, followed by Rhododendron lapponicum and Salix arctica. The most replicated species is Salix arctica followed by Cassiope tetragona and Rhododendron lapponicum.
Plant age was found to be negatively associated with the average ring width across data from all samples. Young plants generally produce wide rings, whereas old plants tend to form narrow increments. Average ring width was positively correlated with the length of the longest xylem radius (Fig. 3b), the total bark thickness and the length of the longest xylem radius as well as the total bark thickness and the average ring width. These relationships, although not surprising, underscored the dependence of bark thickness on growth (Table S2 in Supporting Information). Slower-growing and older specimens were protected by thicker bark. No differences were observed in the range of ring width classes and the frequency of individuals within each class between sediment or intrusive bedrock (Tables S1 and S2 in Supporting Information). All species except for, Cassiope hypnoides and Vaccinium uliginosum, which only occurred in one of the geological settings, showed similar rates of cambial activity and plant age (Figs 2 and 3; Fig. S1 in Supporting Information). In addition to an assessment of the effect of different nutrient levels, the geological grouping also suggests that there was no grazing bias among sites.
Temperature Variability and Recruitment Intensity
Summer and annual temperatures have warmed since the 1870s in Greenland (Fig. 4a). Temperature anomalies for these variables, however, indicate two periods of relatively cold conditions before ~1930 and from ~1970 to 1990. Warmer conditions occurred between ~1930 and 1950 as well as after ~2000. Highest summer temperatures peaked in 2008, 1931 and 2010 (i.e. anomalies of 1.6, 1.4 and 1.3 °C, respectively, with respect to 1981–2010), whereas the coldest June–August means were recorded in 1884, 1914 and 1873 (i.e. anomalies of −2.6, −2.5 and −2.5 °C, respectively, with respect to 1981–2010). Increased year-to-year variability was observed across the annual mean record, in which the three warmest and coldest years are 2010, 1947 and 1941, and 1884, 1894 and 1898 (i.e. anomalies of 3.3, 2.5 and 2.2 °C and −3.8, −3.2 and −3.1 °C, respectively, with respect to 1981–2010). Both the records of annual and summer temperature shared a significantly high fraction of intra-annual to decadal variability (r = 0.65–0.851873–2011) (Fig. S2 in Supporting Information). Disagreement among all high-, low- and band-passed temperature measurements is, however, most evident between ~1930 and the 1970s (Fig. S2b in Supporting Information).
The temperature increase at the onset of the 20th century coincided with intensified shrub recruitment on both geological bedrocks (Fig. 4a,b). Changes in recruitment intensity across all species and sites showed a consistent signal over time. Generally, low levels of recruitment were found before 1920, which were followed by enhanced recruitment that reached its maximum intensity between 1935 and 1985 (Fig. 4b). The 10 to 15-year band-passed temperature time series was highly correlated with recruitment intensity (r = 0.71–0.85), though revealed distinct lags through time (Fig. 4c). A 6-year lag is identified before 1920, a 3-year lag from 1921 to 1960 and a 2-year lag between 1961 and 2000. Correlations and lags were similar when analysed separately for sediment and intrusive sites. Surprisingly, the period of prolonged lag response before 1920 matched with overall lower temperatures, whereas shorter delays in temperature-driven recruitment arose during periods of overall higher temperatures. Nevertheless, no increase in recruitment intensity in response to increasing temperatures was observed during recent decades when recruitment intensity gradually declined.
A literature review for changes in the intensity of Arctic shrub recruitment provided a circumpolar perspective. Most studies indicated maximum recruitment from the mid-20th century onward (Fig. 5a,b). Temperatures during this period were relatively warm compared with periods of low recruitment intensity. The most intensive recruitment pulse in Greenland began before many other regions, and there was no clear spatiotemporal pattern in the occurrence of pronounced recruitment phases at the larger scale. Although most of the studies indicate enhanced shrub recruitment during the second half of the 20th century, the northward and upward expansion of many shrub and dwarf shrub communities was often assumed to have occurred earlier in the 20th century.
Discussion
Our study of average annual growth and longevity of dwarf shrubs in coastal East Greenland ~70°N indicates the following: (i) Low cambial activity and high species longevity, as shown by an average annual ring width and average minimum plant age of 0.07 mm and 60 years, respectively. (ii) High agreement between temperatures and recruitment intensity, given some delay in the recruitment of new individuals after temperature changes. (iii) Consistency in shorter delays of 2–3 years in temperature-driven recruitment pulses during overall warmer periods and a prolonged lag response of 6 years throughout cooler intervals. These key findings (1–3) emerge from the joint application of wood anatomical and dendroecological techniques. They imply low turnover rates and an overall stable community composition of the East Greenlandic shrub community. At the same time, they also suggest the ability of tundra vegetation to respond rapidly to climatic variation. This has been shown by the short delay time in the recruitment of new individuals after the onset of warmer temperatures (Fig. 4).
The observed heterogeneity in growth rates and longevity of all species may have some ecological implications. Since recruitment rates differ among species and sites, the community composition could shift in response to future climate change (i.e. warmer and wetter), which would additionally influence the amount of accumulated biomass. If faster growing species such as Betula nana, possibly also preferring specific site conditions, will be favoured over the slower-growing species, such as Cassiope hypnoides, Cassiope tetragona and Dryas octopetala, remains debatable. Not only must local site conditions be considered, but also higher temperatures. A precipitation-induced larger snow pack will also lead to warmer soil temperatures in winter, thereby increasing nutrient availability and plant growth (Hagedorn et al. 2014). These changes may in turn promote the accumulation of more snow. Mounting evidence recently suggested increasing precipitation totals over the high-northern latitudes following local evaporation and sea-ice retreat (Bintanja & Selten 2014).
The temporal offset, that is observed lag effects between the climatological and dendroecological time series could be due to ecological interaction between climatic conditions and plant recruitment dynamics that play out over multiple years. Although annual temperature extremes can lead to seedling mortality through drought or frost damage (Holtmeier & Broll 2007), we speculate that more than 1 year with favourable climatic conditions would be needed to produce high-quality seeds in sufficient quantities to trigger high numbers of successful germinants with the potential to survive and mature during the subsequent years. These combined factors could create reproductive lags of multiple years between warm summers and successful regeneration pulses during generally colder periods. Otherwise, recruitment-climate lags could relate to the backwards accumulation of missing annual shrub rings (Büntgen & Schweingruber 2010; Hallinger, Manthey & Wilmking 2010; Hallinger & Wilmking 2011).
Soil moisture and air humidity together with an overall lack of summer frost continuing for at least several years are critical for the establishment of new plants in harsh Arctic tundra ecosystems (Holtmeier & Broll 2007). The successful germination, survival and growth of young shrub seedlings require especially favourable circumstances until the root system is sufficiently developed. Nonetheless, other factors, such as light and nutrient availability, often become limiting for older plants (Wicklein et al. 2012; von Arx et al. 2013). At the initial stages of life, many seedlings die within hours if conditions are unfavourable, and temperature can affect rates of metabolism and growth and thus also control the plant-specific water demand. Soil moisture, soil temperature, permafrost and snow cover are also known to influence the survival of tundra seedlings by altering root water uptake and transpiration rates (Munier et al. 2010; Aune, Hofgaard & Söderström 2011; Wilmking et al. 2012; Frost et al. 2013). In addition, differences in the timing of snowmelt at the site level affect growing season length and influence growth at early life stages (Høye et al. 2013).
Since mortality of tundra saplings is high, only the fittest individuals survive (see Schweingruber & Poschlod 2005 for a dendroecological review of herbs and shrubs). With regard to tree species from the high-northern latitudes or certain high-elevation sites, the establishment and survival of seedlings has been described as a decades-long process of suppressed annual increments before an abrupt growth release finally occurs (Esper & Schweingruber 2004; Harsch et al. 2009; Harsch & Bader 2011). For these reasons, while annual temperature extremes can have distinct effects on tundra recruitment, we speculate that more than 1 year with favourable climatic conditions would be needed for high-quality seeds to ripen and accumulate in sufficient quantities to germinate successfully and produce healthy seedlings with the potential to survive and mature during the subsequent years. This hypothesis coincides with the tentative result that more time (6 years) was necessary for a successful regeneration during colder conditions before ~1920, whereas more rapid responses have been observed in warmer years.
An array of decisive factors determining the annual ring width patterns operates on local scales. Temperatures increase from the cold coast towards inland areas, from mountains towards lowlands and from flat and wind-exposed areas towards wind-protected areas on slopes with favourable exposure. Hydrology, or rather topography and substrate texture governing hydrology, varies widely within short distances. The interplay of microclimatic and microsite factors may additionally change during the growing season, as does the impact of these factors on plant processes (Ogle et al. 2012; Wu, Jansson & Kolari 2012). Similarly, the requirements and limitations of plant growth follow seasonal patterns. Early in the season, for example, germinating seeds and young seedlings may depend on the moderating capacity of existing dwarf shrubs with respect to temperature extremes, whereas desiccation may pose a problem in summer, particularly in the light of seedling survival under predicted future warming. A thorough understanding of how different microenvironmental factors, including fungi control variation in the growth and recruitment of Arctic shrub communities is not yet available.
We observed no difference in cambial activity (i.e. ring width) or plant longevity (i.e. ring number) that could be attributed to site-specific factors. Changes in the recruitment, growth and longevity of tundra plant communities depend on microsite conditions, including incoming solar radiation, precipitation, soil moisture and temperature, nutrient availability, snow pack and melt timing, active layer depth, and growing season length (Myers-Smith et al. 2011), and also on herbivory and fungi (Post & Pedersen 2008; Graae et al. 2011). Our data collection from two different geological settings did not indicate an effect of the different site conditions and also suggests that there is no grazing bias between intrusive and sediment sites (Fig. S1 in Supporting Information). Although East Greenland reindeer (Rangifer tarandus eogroenlandicus), one of three relatively isolated, small-bodied, high-arctic subspecies (Gravlund et al. 1998), became extinct approximately 100 years ago (Degerbøl 1957), muskox (Ovibos moschatus) populations still frequently feed on the willow-dominated, smooth sediment plateaus of Jameson Land (Thing et al. 1987). However, we did not observe any dissimilarity between the two sites that could be attributed to differential herbivory.
Although our study was based on almost 900 individual shrubs, several data-related and methodological-induced uncertainties remain. In case of such massively replicated data sets, careful cross-dating cannot be conducted for all individuals (Schweingruber et al. 2013), especially given the high degree of local-scale variation in secondary growth (Buchwal et al. 2013). The herein reported plant ages therefore only constitute minimum estimates of the preserved tissues (Schweingruber & Büntgen 2013), and the application of cross-dating would certainly have resulted in a more accurate determination of plant ages.
In consequence, time-series analysis was applied to estimate the climate sensitivity of recruitment pulses. This technique allows differing lags between measured ring widths and climatic changes to be considered over time. It is important to note that sample size is lower for the earlier portion of the record prior to ~1920, during which differences in species composition, due to differential longevity of species, might affect the recruitment-temperature relationship. In addition, the observed offset might be related, in part, to inconsistency in the instrumental temperature measurements themselves. A possible bias in seasonal station readings is indicated by a significant decline in the coherency between the annual and summer means (Fig. S2 in Supporting Information). This type of ‘meteorological target error’ (Frank et al. 2007), however, is expected to be more randomly distributed in time and thus should not result in a systematic and temporally stable pattern as herein observed. Further ambiguity might result from overlooking the very youngest and smallest shrub individuals during fieldwork. If this situation indeed occurred, an artificial underrepresentation of the youngest specimens after ~1980 would explain the lack of modern recruitment in response to recent warming (Fig. 4). Despite the fact that the abovementioned limitations may influence the magnitude of the observed temperature-recruitment lag/delay pattern with different intensity throughout time, the successful establishment and survival of shrub recruits was mainly driven by longer term temperature changes that occurred on interannual time scales (Fig. 4).
The interplay of temperature variability and recruitment intensity of the East Greenlandic shrub community is well in line with independent evidence of intense mid-20th century shrub recruitment as reported from many high-northern ecotones (Fig. 5). Annual mean temperatures during the 1930s were most likely warm enough to trigger the establishment of new shrub seedlings in northern Norway (Bär, Bräuning & Löffler 2006; Bär & Löffler 2007; Bär et al. 2008; Hallinger, Manthey & Wilmking 2010). For many tree species, new seedlings associated with the early to mid-20th century warming have been reported from central Sweden (Kullman 2001), northern and central Finnish Lapland (Holtmeier et al. 2003), the Kola Peninsula (Gervais & MacDonald 2000), the Polar Urals (Shiyatov 1992; Shiyatov, Terent'ev & Fomin 2005; Devi et al. 2008; Hagedorn et al. 2014), northern central (Kirdyanov et al. 2012) and eastern (MacDonald, Case & Szeicz 1998) Siberia, as well as northern Quebec (Payette & Filion 1985). Several dendroecological studies, including comparisons of repeated photographic images and studies of tree growth forms and sapling germination, further revealed treeline positional and structural changes during the mid-20th century in Scandinavia (Hustich 1958; Aas 1969; Kallio & Lehtonen 1975; Kullman 1986, 2001, 2002; Hofgaard, Kullman & Alexandersson 1991), the Polar Urals (Gorchakovsky & Shiyatov 1978; Shiyatov 1992; Devi et al. 2008), Quebec (Morin & Payette 1984; Payette et al. 1985; Lavoie & Payette 1994), the Canadian Rocky Mountains (Luckman & Kavanagh 1998) and across most of western Siberia (Esper & Schweingruber 2004; Kharuk et al. 2006), for example.
In the light of the above, we advocate a different type of dendroecology from that conducted in the past to foster our understanding of the response of tundra shrub communities to climate change. Previous dendroecological investigations have often been based on relatively few selected individuals from few species at few locations with a focus on dendroclimatology rather than on understanding age distributions and recruitment pulses over different spatial scales. In contrast, sampling representatively across the landscape among different age classes and species can gain valuable insight on variation in growth rates and age structures. This community ecological perspective on dendroecology allows for the identification of recruitment phases, as well as the analysis of their biotic and biotic drivers across space and time. By considering many individuals and species in communities across environmental gradients, community-based dendroecology thus fills a gap in our knowledge on the variation of growth and reproduction relative to a wide range of diverse influences over large temporal and spatial scales. If successfully applied, this approach will help to improve model predictions of warming-induced rapid tundra expansion because simulations often underestimate the longevity and persistence of individual plants and species but might overestimate the sensitivity level and reaction time of entire communities.
Finally, we prioritize five research avenues at the interface of wood anatomy and dendroecology that will ideally enhance future endeavours in functional ecology (Calow 1987), functional biogeography (Violle et al. 2014) and range dynamic modelling (Normand et al. 2014), with all of them going beyond the geographical limits of forest growth and ranging from the scale of cells to communities: (i) Determination of individual plant ages and age structures of communities for various tundra species and ecological systems. (ii) Consideration of site- and species-specific differences in plant morphological characteristics and stem anatomical features together with diverse dendroecological techniques to facilitate the exploration and separation of biotic and abiotic controls on the growth and recruitment phases of tundra vegetation that may range from individual specimens to community patterns. (iii) Expansion of stem anatomical and dendrochronological investigations from tundra shrubs to herbs with distinct annual growth increments. (iv) Estimation of historical changes in biomass production and carbon allocation within and among plant species. (v) Application of community-based dendroecological studies to investigate changes in age structure, recruitment intensity, growth level and wood anatomy across gradients with varying microsite conditions.
Acknowledgements
L.H. was supported by the Eva Mayr-Stihl Foundation, and U.B. was supported by the Operational Programme of Education for Competitiveness of Ministry of Education, Youth and Sports of the Czech Republic (Project: Building up a multidisciplinary scientific team focused on drought, No. CZ.1.07/2.3.00/20.0248). A. Ivanova, L. Schneider and S. Braun contributed to the laboratory work. We thank N.E. Zimmermann for making the digitalized (species-specific shrub) range maps available. The Danish Council for Independent Research – Natural Sciences (10-085056) and the Villum Foundation's Young Investigator Programme (VKR023456) supported SN.
Data accessibility
All dwarf shrub dendroecological data herein presented are freely available at http://www.wsl.ch/fe/landschaftsdynamik/dendroecology/index_EN
