Deepened winter snow increases stem growth and alters stem 13C and 15N in evergreen dwarf shrub Cassiope tetragona in high-arctic Svalbard tundra

Deeper winter snow is hypothesized to favor shrub growth and may partly explain the shrub expansion observed in many parts of the arctic during the last decades, potentially triggering biophysical feedbacks including regional warming and permafrost thawing. We experimentally tested the effects of winter snow depth on shrub growth and ecophysiology by measuring stem length and stem hydrogen ( δ 2 H), carbon ( δ 13 C), nitrogen ( δ 15 N) and oxygen ( δ 18 O) isotopic composition of the circumarctic evergreen dwarf shrub Cassiope tetragona growing in high-arctic Svalbard, Norway. Measurements were carried out on C. tetragona individuals sampled from three tundra sites, each representing a distinct moisture regime (dry heath, meadow, moist meadow). Individuals were sampled along gradients of experimentally manipulated winter snow depths in a six-year old snow fence experiment: in ambient ( c . 20 cm), medium ( c . 100 cm), and deep snow ( c . 150 cm) plots. The deep-snow treatment consistently and signi ﬁ cantly increased C. tetragona growth during the 2008 – 2011 manipulation period compared to growth in ambient-snow plots. Stem δ 15 N and stem N concentration values were signi ﬁ cantly higher in deep-snow individuals compared to individuals growing in ambient-snow plots during the course of the experiment, suggesting that soil N-availability was increased in deep-snow plots as a result of increased soil winter N mineralization. Although inter-annual growing season-precipitation δ 2 H and stem δ 2 H records closely matched, snow depth did not change stem δ 2 H or δ 18 O, suggesting that water source usage by C. tetragona was unaltered. Instead, the deep insulating snowpack may have protected C. tetragona shrubs against frost damage, potentially compensating the detrimental effects of a shortened growing season and associated phenological delay on growth. Our ﬁ ndings suggest that an increase in winter precipitation in the High Arctic, as predicted by climate models, has the potential to alter the growth and ecophysiology of evergreen shrub C. tetragona through changes in plant mineral nutrition and frost damage protection.


Introduction
The Arctic has generally warmed during the last decades, especially during the winter period (Serreze and Barry 2011).Winter precipitation in the Arctic is expected to increase during this century as a consequence of higher atmospheric temperatures, seaice loss and associated increasing air humidity, especially in the High Arctic (IPCC 2013, Bintanja andSelten 2014).Changes in winter precipitation patterns may affect vegetation composition in arctic tundra.During the last decades, shrub expansion has been observed across many arctic and alpine tundra sites (Tape et al 2006, Callaghan et al 2011, Myers-Smith et al 2011, Elmendorf et al 2012, Frost and Epstein 2014).Snow trapping by shrubs, resulting in higher soil winter temperatures and potentially enhanced plant mineral nutrition, has been suggested as a possible mechanism for shrub expansion in the Arctic (Sturm et al 2005).This snow-shrub hypothesis is supported in part by recent dendroecological studies, indicating a link between winter precipitation/ snow cover and shrub growth, although the direction and magnitude of the responses differed from positive (Hallinger et al 2010) to negative (Schmidt et al 2010, Mallik et al 2011).We studied the growth response and underlying physiological changes in Cassiope tetragona (Arctic bell heather) to experimental snow depth changes in high-arctic Svalbard tundra to improve our mechanistic understanding of the controls of snow depth on shrub growth.Cassiope tetragona is an evergreen shrub species with a circumarctic distribution and great latitudinal spread (Callaghan et al 1989, Rayback andHenry 2005), and is dominant in hemi-prostrate dwarf shrub tundra regions (Walker et al 2005).Therefore, the potential impacts of changing patterns of snow coverage on C. tetragona growth could be of particular importance for high-arctic tundra ecosystem functioning now and in the future.
An increase in winter snow precipitation may lead to higher winter soil temperatures due to the insulating capacity of snow, resulting in higher microbial activity and N mineralization rates (Schimel et al 2004).This may increase plant N availability during the growing season (Natali et al 2012) and increase leaf N concentrations (Welker et al 2005a).Experimental studies have shown that mainly shrubs may benefit from higher nutrient availability by growing denser and taller (Shaver andChapin 1980, Bret-Harte et al 2004).However, the timing of nutrient-release by winter soil mineralization may not coincide with the start of the plant growing season.Thus, deeper snow may not lead to enhanced nutrient availability for shrubs in all cases (Buckeridge et al 2010, Vankoughnett and Grogan 2014).
Deeper snow can also increase the availability of water during the growing season (Morgner et al 2010, Leffler andWelker 2013), unless snow melt water runs off the frozen surface during early spring before soil thaw commences (Löffler 2005).Ehleringer and Dawson (1992) suggested that plants growing in areas with low summer precipitation inputs, as in high-arctic Svalbard, may rely mostly on water from deeper soil layers to ensure stable provision of water and maintenance of growth during dry summers.In the High Arctic, growth of C. tetragona is generally restricted to winter snow-covered areas, which apart from being a strategy to prevent frost damage during the extreme cold winter (Billings and Mooney 1968), may prevent desiccation during the growing season (Callaghan et al 1989).A summer water manipulation study in high-arctic Svalbard did however fail to observe an effect of additional water supply on C. tetragona stem length growth (Weijers et al 2013a), in contrast to the positive growth response to increased summer precipitation observed for Betula nana in the Siberian tundra (Blok et al 2011, Keuper et al 2012).
Finally, deeper snow may also affect growth positively by providing protection against frost after episodic winter warming events (Bokhorst et al 2011), as C. tetragona growth has been found to be negatively affected by winter warming events (Weijers et al 2012).At the same time, increased snow depth may result in a delay of the growing season onset (Borner et al 2008), potentially reducing C. tetragona growth (Mallik et al 2011).The complexities of snow depth effects on C. tetragona growth thus appear to be operating across a range of processes that may be both indirect by altering the ecohydrological and nutritional conditions for growth, as well as directly via physical protection by snow from harsh winter climate conditions (Cooper 2014).
Despite the correlative evidence for a snow depth-shrub growth linkage, an experimental approach is needed to assess the direct effects of winter snow depth on shrub growth.Moreover, to unravel the underlying mechanisms, it is essential to go beyond analyses of growth metrics and utilize information on physiological responses to snow cover changes, as contained in stable isotope data (McCarroll andLoader 2004, Pattison andWelker 2014).To estimate the impacts of snow depth changes on C. tetragona growth, leaf gas exchange, nitrogen (N) acquisition and water source usage, we manipulated winter snow depth in a field experiment conducted in high-arctic Svalbard and measured C. tetragona stem length growth, stem carbon (δ 13 C), nitrogen (δ 15 N), oxygen (δ 18 O) and hydrogen (δ 2 H) isotopes.Cassiope tetragona individuals were sampled along winter snow depth gradients in a six-year old snow fence experiment: in ambient (c.20 cm), medium (c. 100 cm), and deep snow (c.150 cm) plots located in three tundra sites, each representing a distinct moisture regime (dry heath, meadow, moist meadow).The aim of our study was to improve the mechanistic understanding of the controls of winter snow depth on C. tetragona growth and ecophysiology in a high-arctic tundra ecosystem.We tested our hypotheses that C. tetragona stem length growth increases with deeper winter snow, with the magnitude of the responses depending on the contrasting effects of increased snow depth on growth: negative through shortening of the growing season and positive through frost protection.Secondly, we hypothesized deeper snow would alleviate both water and nitrogen limitations for C. tetragona growth.

Site description and experimental setup
To understand the impacts of winter snow depth on annual growth and stem isotopic characteristics (δ 2 H, δ 18 O, δ 13 C and δ 15 N) of the evergreen shrub Cassiope tetragona, individuals were harvested from a snow manipulation experiment in the Adventdalen Valley in Spitsbergen,Svalbard (78.17°N,16.27°E).Mean July temperatures (1975-2012) of 6.4 °C were recorded at the closest meteorological station at Svalbard Airport, Longyearbyen (WMO station 99849), located approximately 15 km from the research site.The coldest month is March with an average temperature of −13.7 °C.Annual precipitation sums are c.190 mm, most of which falls as snow during the November-May period (c.120 mm).
Snow fences were used to create snowdrifts on the leeward side of the fences.Nine fences of 1.5 m height and 6.2 m length were established during autumn 2006.Three fences were located within each of three sites (heath, meadow, moist meadow), characterized by differences in plant species composition, soil moisture regimes and landscape types.A detailed vegetation description of sites is given in Cooper et al (2011).All fences were located within an area of 1.5 by 2.5 km (Morgner et al 2010).The heath site is relatively dry due to its topography and is dominated by C. tetragona, whereas the meadow site is wetter due to its location in a drainage area of an adjacent hill and is dominated by Salix polaris.The moist meadow site is similar to the meadow site, but has undergone surface collapse behind one of the three snow fences during recent years and has on average higher soil moisture (Semenchuk et al 2015).Snow drifts raised winter soil temperatures in comparison to ambient snow areas, maintaining top-permafrost temperatures just below freezing point (Semenchuk et al 2013), which presumably led to deeper thaw during summer.Thaw of deeper ice-rich permafrost layers resulted in surface collapse, thus increasing topographical differences on the snowdrift side of the fences at the moist meadow site during the course of the experiment.A similar phenomenon was observed in a snow fence experiment in Northern Sweden (Johansson et al 2013).
A transect was laid out perpendicular to each snow fence to study the impacts of two incremental snow depths on soil temperature and soil moisture.Snow depth measurements were carried out during winter 2010 and 2011 using a probing stick.Snow depth decreased with distance from each snow fence (Semenchuk et al 2013).Soil moisture measurements (Theta Probe ML2x; Delta-T Devices, Cambridge, UK) were recorded at irregular intervals during carbon flux measurement campaigns between 2007 and 2011.Soil moisture was generally higher in snow drift plots (5-10 vol%) compared to ambient snow early in the growing season, with differences declining to similar values by the end of the growing season (Morgner et al 2010, Semenchuk et al 2015).Soil temperature was measured at an hourly interval at a depth of c. 2 cm in ambient (8 out of 9 fences), medium (1 fence per site) and deep snow plots (all 9 fences) using temperature probes (Tiny Tag Plus 2, Gemini Data Loggers, UK).Daily average soil temperature data from 2008-2012, measured in the snow fences, are presented in Semenchuk et al (2013).Plots for C. tetragona collection (see next section for details) were selected under three different snow depth regimes, at a distance of approximately 5 m (deep-snow, c. 150 cm) and 10 m (medium-snow, c. 100 cm) from the leeward site of the fences.In addition, plots were selected within 100 m of each snow fence from areas receiving ambient-snow levels (control, c. 20 cm).Table 1 lists winter snow depths, snow-free dates and winter soil temperatures in the three different plot types.
Five samples of complete aboveground parts of C. tetragona individuals were collected during early July 2012 at all three plots (ambient, medium and deep snow) from each snow fence at all three sites.The hemi-prostrate C. tetragona individuals at our research site were, on average, between 5 and 10 cm tall.At the heath site, collection took place along two transects behind one of the three snow fences to compensate for another snow fence at the heath site that Table 1.Average maximum winter snow depth, snow-free date, and winter soil temperature (mid-September to mid-June) measured at 2 cm depth in ambient-snow, medium-snow, and deep-snow plots.
had fallen over and laid flat during two years of the experiment and where no C. tetragona collections were made.Behind the fence that was sampled twice, transects were laid out 5 m apart to prevent sampling of individuals within an identical clonal patch.Care was taken to select only healthy individuals with green active meristems.Samples were packed in separate paper bags and air-dried before shipment to the lab for analysis.

Annual stem length measurements
Leaves of C. tetragona grow in opposite pairs and form rows on four sides of the branches.A wave-like pattern in leaf lengths and leaf scar distances allows for the reconstruction of annual leaf and stem growth rates for several decades (Callaghan et al 1989, Havström et al 1995).Recently, a novel method has been developed allowing reliable reconstructions of growth on even longer time scales by the detection of dark marrow tissue bands between growth years, termed the wintermarksepta method (Rozema et al 2009, Weijers et al 2010).Annual stem length growth is measured as the distance between wintermarksepta (Rozema et al 2009).Wintermarksepta are considered to be small bands of meristem tissue in the pith, formed at the end of the growing season, and thereby demark the end of one year of stem growth.Distances between wintermarksepta have proven to be a reliable proxy for past climate (Weijers et al 2010, Weijers et al 2013b).Of each set of five individual aboveground plant individuals sampled from each plot, the two presumably oldest individuals were selected for further analysis.Thus, 54 individuals, spread evenly across sites and snow depth treatments, were measured in total.Annual growth was measured on all individuals using the wintermarksepta method with a precision of 0.01 mm from the latest year of growth (2012) up to the oldest year of growth (the oldest individual measured was 35 years old) on the dominant stem from each individual, defined as the oldest stem containing a green meristem.Growth records of individuals were visually crossdated and further checked for errors with the software program COFE-CHA (Holmes 1983).The crossdated raw stem increment growth records were detrended by applying a linear regression line with flexible slope to each individual growth curve using the program ARSTAN v6.05 (Cook 1985, Cook andHolmes 1986)  Residual growth curves were averaged across the two individuals collected per snow depth locality to avoid spatial pseudo-replication in the subsequent mixedmodel analyses (see statistical analysis section below).

Isotopic analyses
Of the two C. tetragona individuals analyzed per plot for annual stem length growth, one individual was allocated to δ 13 C/δ 15 N analysis, and the other individual to δ 2 H/δ 18 O analysis.Cassiope tetragona increments were weighed on a fine balance (precision 1 μg), cut, and packed (c. 2 mg) into tin (δ 13 C/δ 15 N) or silver capsules (δ 2 H/δ 18 O).Annual growth increments from the year before the establishment of the experiment in 2006 up until 2011 (growth increments of 2011 were bulked with underdeveloped 2012 growth increments) were analyzed for isotopic composition for one individual per plot for each snow fence.In addition, growth increments of samples from ambient snow plots were analyzed for the years 2000-2005 to study annual isotopic variations in response to fluctuations in local climate parameters.We measured isotopes on whole stem wood, as stem mass was limited, preventing extraction of sufficient amounts of cellulose for isotopic analysis.Earlier work showed that δ 13 C records derived from whole wood may show greater climate sensitivity than δ 13 C records from extracted cellulose or lignin tissue (Loader et al 2003), but this has not yet been validated for arctic shrubs.
Isotopic ratios of carbon ( 13 C: 12 C) and nitrogen ( 15 N: 14 N), nitrogen concentration (%N) and carbon concentration (%C) were determined on an Isoprime isotope ratio mass spectrometer (Isoprime Ltd, Cheadle Hulme, UK) coupled to a CN elemental analyzer (Eurovector, Milan, Italy) using continuous flow, with helium used as a carrier gas.Samples were combusted at a temperature of 1050 °C for C and 700 °C for N. Isotopic C and N ratios are presented in δ-notation (δ = [R sample /R standard -1] *1000; R depicting 13 C: 12 C and 15 N: 14 N ratios), relative to isotopic ratios of Vienna Pee Dee Belemnite and atmospheric N 2 , respectively.Samples were analyzed with reference gas calibrated against international standards IAEA C5, CH6, CH7, N1, N2 and USGS 25, 26, 32, and drift corrected using peach leaves (NIST 1547) as working standard.The standard deviation of isotope measurements of the standards was ±0.2 parts per mille (‰) for δ 15 N and ±0.1‰ for δ 13 C.

Meteorological and precipitation isotope data
Daily air temperature and precipitation data  were obtained from the meteorological station at Svalbard Airport (WMO station 99849; www.eklima.no),approximately 10 km from the research location.Average monthly temperature and monthly precipitation sums were calculated for months with complete data records.
Monthly mean precipitation (1990-2011) δ 2 H and δ 18 O values from a meteorological station (WMO station 100 400) in Ny-Ålesund, Svalbard (78.92°N, 11.92°E), located approximately 130 km northwest of the experimental snow fence site, were obtained from the Global Network of Isotopes in Precipitation (GNIP; http://www.iaea.org/water).Monthly composite precipitation samples were collected and analyzed for δ 2 H and δ 18 O at the IAEA lab in Vienna, Austria.Precipitation source was recorded as snow, rain, or mixtures of snow and rain.Isotope contents are expressed as ‰ deviation from the standard VSMOW isotope ratio.

Statistical analyses
The effects of snow depth and site on annual C. tetragona stem length, stem %N, stem N pools, and isotopic composition (δ 2 H, δ 18 O, δ 13 C and δ 15 N) for years 2008-2011, as well as for the pre-treatment year 2006, were analyzed using the PROC MIXED procedure in SAS Enterprise Guide 6.1.We excluded growth and isotopic data from 2007 fragments from the mixed models as this may have been affected by pretreatment conditions through storage of plant resources during the end of the 2006 growing season, when snow treatments were not yet in effect.In addition, stem %N, stem N-pool and stem isotope data from 2011 growth increments were excluded from the mixed models, as these values strongly deviated due to inclusion of underdeveloped 2012 increments with the 2011 fragments.Data were natural log-transformed where needed to achieve normal distributions.Snow depth, site, and snow depth*site were selected in the models as fixed factors, with snow fence nested within site as random factor.
Fixed effects were determined using the Restricted Maximum Likelihood method, with degrees of freedom estimated by the Kenward and Rogers method.Growth year was selected as repeated factor to constrain model degrees of freedom.Toeplitz, autoregressive or heterogeneous autoregressive structures were used as covariance structure for the repeated effect to take into account potential autocorrelation of stem growth, stem N pools and isotopic composition of stems within individuals between years, depending on which structure yielded lowest Akaike Information Criterion statistics.One-way Analysis of Variance post hoc tests were performed to compare differences in stem growth, stem N pools and isotopic composition of stems between individuals growing under mediumand deep-versus ambient-snow conditions.

Cassiope tetragona annual growth and isotopic composition variability in relation to climate
We developed a growth chronology from the residuals of detrended growth curves of 18 individuals from ambient-snow plots sampled across all three sites for the period 1998-2011.Our residual chronology had an expressed population signal over 0.85 for the entire chronology period, indicating the sampled pool of individuals from ambient-snow plots expressed a growth signal that was likely representative for the regional C. tetragona population.The mean sensitivity of our residual chronology was 0.31, indicating the relative change in stem increment length from one growth year to the next.Our residual chronology correlated strongly with a growth chronology developed by Weijers et al (2010) from C. tetragona individuals collected at nearby Endalen (r = 0.85, P < 0.001, n = 11 years; figure S2 in supporting information).
Growth of ambient-snow individuals (1998-2011), measured as annual stem growth, was negatively correlated to summer (June-September) precipitation, but positively correlated to summer (June-August) air temperatures (table 2).The strongest, positive, correlation was observed between growth and mean July temperature (r = 0.82, P < 0.001, n = 12 years).Summer precipitation and summer temperature were negatively correlated (r = −0.61,P < 0.05, n = 14 years; data not shown).No significant relationships between annual stem growth and winter monthly temperature or winter monthly precipitation sums were observed (data not shown).
May-July average precipitation δ 2 H values closely matched δ 2 H values in annual C. tetragona stem increments from ambient-snow plots between 2000 and 2009 (r = 0.82, P < 0.01, n = 10 years; figure 1).From 2010 onwards this relationship became less apparent, with stem δ 2 H values sharply increasing and deviating from the precipitation δ 2 H temporal trend.The ratio between annual winter (preceding year Novembercurrent year May) and summer (current year June-August) precipitation, hereafter referred to as the 'winter:summer precipitation ratio', was closely related to C. tetragona stem δ 13 C (r = 0.88, P < 0.001, n = 12 years) and stem δ 2 H values (r = 0.84, P < 0.001, n = 12 years; figure 1) for the period 2000-2011.This relationship became non-significant when excluding the last two years with relatively high C. tetragona δ 13 C and δ 2 H values. Stem δ 13 C and δ 2 H values were significantly correlated in ambient-snow individuals for the period 2000-2011 (r = 0.94, P < 0.001, n = 12 years; figure 1).

Snow manipulations effects on stem length growth, isotopic composition and nitrogen content
Average C. tetragona individual age was 19 years (SE = 0.6 year), with similar mean ages across snow depths and sites.Snow depth had a significant effect on C. tetragona annual stem growth during the snow manipulation period 2008-2011, with higher growth observed in individuals growing in deep-snow plots compared to individuals growing in ambient-snow plots (table 3, figure 2).The strongest positive shrub growth response to the deep-snow treatment was observed in all three sites during the most productive growth year 2009.During time of sampling in July 2012, annual C. tetragona stem growth was uncompleted, with snow-manipulated plots showing lower growth than ambient-snow plots (figure S1 in supporting information).Growth was not significantly different between the snow depth treatments during the pre-treatment growth period 1998-2006.Growth did also not differ significantly between sites during both pre-treatment and treatment periods (P > 0.05).
Snow depth had a significant effect on stem δ 13 C and δ 15 N values for the period 2008-2010, but not on δ 2 H or δ 18 O values (table 3, figure 3).Stems from Effects of snow depth and site on stem length growth were analyzed for the period 2008-2011.Values are based on averages of nine individuals snow depth −1 , with three individuals analyzed snow depth −1 site −1 .Significant effects (P < 0.05) are marked in bold.Differences were assessed between individuals growing under three different snow depth regimes (ambient-, medium-and deep-snow) at three different sites (heath, meadow, moist meadow).The interaction term snow 'depth*site' was not significant for any parameter and thus removed from the models.

Cassiope tetragona growth responses to deepened snow
Our results show that the deep-snow treatment enhanced stem length growth of C. tetragona compared to ambient-snow individuals, despite the delayed snowmelt and reduction in growing season length.Individuals from the medium-snow plots did not show differences in growth compared to ambientsnow individuals, suggesting a snow depth threshold needs to be crossed before snow depth has a significant effect on growth (figure 2).Wipf and Rixen (2010) analyzed 39 snow manipulation studies conducted at arctic and alpine tundra sites with ambient snowmelt dates between day of year (DOY) 120-180, showing that delayed snowmelt generally decreased aboveground plant productivity, except at sites with intermediate snowmelt dates.This is consistent with our  Soil temperatures during early winter (November-December) in medium-snow plots were roughly similar to soil temperatures in ambient-snow plots, whereas deep-snow plots showed on average 5-10 °C significantly higher early winter top-soil temperatures compared to ambient-snow soils (Semenchuk et al 2013).As snowdrifts start building up close to the snow fences, deep-snow plots are covered earlier with snow than the medium-snow plots which are located further away from the fences.Individuals growing in the medium-snow plots may thus not have the same level of frost protection against extreme early-winter warming events as deep-snow individuals (Bokhorst et al 2011), as it takes a longer time for a deep snow pack to completely melt out.Other snow fence experiments performed in arctic and subarctic tundra also measured increased plant productivity in snow-accumulation plots, which was attributed to the combined effects of higher moisture and nutrient availability during the growing season (Schimel et al 2004, Natali et al 2012, Bosiö et al 2014).

Responses of photosynthesis and stomatal conductance to deepened snow
We observed a strong positive C. tetragona growth response to deep snow in 2009 at all our three sites (figure S1 in supporting information), despite this being a year with a relative thick and stable ambient snowpack of approximately 30 cm deep (Semenchuk et al 2013).We did not observe a growth response to the medium-snow treatment, suggesting that a snow depth threshold exists for the snow treatment to positively affect C. tetragona growth.Although the medium-and deep-snow treatments initially reduced stem growth early in the growing season, as observed in the 2012 growth segments (figure S1 in supporting information), this growth delay was offset in the deepsnow individuals later during the growing season.Starr and Oberbauer (2003) reported that C. tetragona starts photosynthesizing when still snow-covered during spring, although light levels received by C. tetragona under snow may be lower compared to snow-free conditions, reducing the photosynthetic capacity.Corresponding with this observation, Wipf (2010) showed that the dormancy period of C. tetragona is unaffected by snow depth.This suggests that for C. tetragona, the photosynthetic active period may not have been affected by the snow treatments, which could indicate that frost protection is an important control positively influencing growth at sites with deep-snow winter conditions (Wipf et al 2009).
We observed depleted stem-δ 13 C values in C. tetragona plants growing in medium-snow plots (figure 3).Depleted stem-δ 13 C values may be due to either higher rates of stomatal conductance or lower  photosynthetic rates (higher internal CO 2 : ambient air CO 2 ratios) when plants are growing under sufficient water supply (Farquhar et al 1989, Welker et al 2005b).We observed higher stem %N values, a surrogate for changes in leaf photosynthesis (Welker et al 1993), in individuals from deep-snow plots but not in mediumsnow individuals.This suggests that the observed lower stem-δ 13 C values in medium-snow individuals are not related to lower photosynthesis rates but are most likely associated with higher rates of stomatal conductance.
Since C. tetragona leaves have a lifespan of approximately four years (Weijers et al 2012), C assimilation and discrimination against 13 C may have continued for a few years in segments formed before establishment of the experiment, as suggested by the lower δ 13 C values in medium-snow 2006 stem segments from the meadow site.A similar observation was made in a water addition experiment nearby our research location, in which C. tetragona leaves and stem increments showed depleted δ 13 C values in water addition plots both before and after the start of the experiment (Weijers et al 2013a).Although radial growth may contribute to the isotopic values of previous years' growth, the strong linkage between the isotope time-series and the large inter-annual variations within isotope series suggest that the annual isotope signal is largely preserved in these relatively young branches.The increase in δ 13 C values (becoming less negative, figure 1) from older to recent growth segments is not coherent with trends in stem length growth over time, but may be explained by the observed declines in the relative amount of summer precipitation during the last years of growth, reducing stomatal conductance and intercellular CO 2 partial pressure.The latter will lead to a shift towards less depleted δ 13 C in newly synthesized plant tissue, closer to atmospheric δ 13 C values (Farquhar et al 1989).The observed strong correlation between stem δ 13 C and δ 18 O (figure 5) supports the suggestion that interannual variations in stomatal conductance are reflected in the observed δ 13 C trend over time (Saurer et al 1997, Scheidegger et al 2000).An alternative explanation for the δ 13 C increase with stem age could be related to lignification of cellulose compounds during secondary radial growth, since lignin is more 13 Cdepleted than cellulose (Hobbie and Werner 2004).Similar stem age-δ 13 C patterns have previously been observed for C. tetragona by Weijers et al (2013a) and Rayback et al (2012).

Deep snow increases Cassiope tetragona stem nitrogen concentrations
Stem δ 15 N values were less depleted in medium-snow and deep-snow individuals compared to ambientsnow individuals (figure 3), suggesting a higher uptake of inorganic-N over organic-N compounds under deep-snow than under ambient-snow conditions (Michelsen et al 1996b, Yano et al 2010).This potential shift in inorganic-N over organic-N may be due to greater winter decomposition of organic matter in relatively warm deep-snow covered soils (Schimel et al 2004, Natali et al 2012).Higher soil-extractable ammonium and nitrate concentrations were observed in deep-compared to ambient-snow plots at our site (Semenchuk et al 2015), which were reflected in higher C. tetragona stem %N values in deep-snow individuals in our study.Our findings suggests that soils in deepsnow plots had a higher soil N-availability resulting from greater winter soil microbial activity under deep snow (Schimel et al 2004).In line with our findings of higher C. tetragona stem %N under deep snow, Salix polaris leaves collected at our site in deep-snow plots also showed increased leaf %N (Semenchuk et al 2015).Similar findings were reported for Salix arctica leaf tissue in a snow fence experiment in the high-arctic of NW-Greenland (Rogers et al 2011, Leffler and Welker 2013).
Evergreen species like C. tetragona have the advantage of long (leaf) life spans and low nutrient loss rates, at the cost of low growth rates and limited responsiveness to environmental changes (Aerts 1995).A longterm fertilization experiment in a subarctic heath of Northern Sweden showed that higher N-availability initially stimulated C. tetragona growth (Michelsen et al 1996a).However, after 22 years of fertilization, C. tetragona cover declined, with graminoid species increasing in cover (Campioli et al 2012).These findings highlight the importance of long-term studies to understand plant community responses to environmental changes in slow-growing arctic tundra environments.Cassiope tetragona's poor competitive ability for nutrient acquisition in comparison with fast-growing deciduous shrub or graminoid species may eventually lead to an incremental decline in C. tetragona growth rates under higher winter-released soil N (Havström et al 1993, Kielland 1994, Molau 1997, Aerts and Chapin 2000).However, large differences may exist in growth limitations for C. tetragona between high-, low-, and sub-arctic regions, with presumably greater temperature limitation for growth in the high-arctic and greater N-limitation in the subarctic (Havström et al 1993).

Source water use of Cassiope tetragona not affected by snow depth
We observed a strong correlation between May-July precipitation δ 2 H and C. tetragona stem δ 2 H (figure 1), suggesting that most of the water supply to C. tetragona originated from precipitation falling during the growing season.Neither δ 2 H nor δ 18 O C. tetragona stem values were affected by snow depth, suggesting that source water usage by C. tetragona was unaffected, similar to observations by Leffler and Welker (2013) for Salix arctica in high-arctic NW-Greenland.However, we cannot exclude the possibility that local differences in source water usage were canceled out by differences in C. tetragona transpiration rates under different snow regimes, providing an alternative explanation for similar δ 2 H and δ 18 O values.Stem δ 18 O values were enriched in the heath site compared to the meadow and moist meadow sites, suggesting that C. tetragona individuals growing in the relatively dry and elevated heath site are more dependent on water derived from rain than snow precipitation, in contrast to the meadow and moist meadow sites which likely have a greater inflow of more δ 18 O-depleted snowmelt water from a nearby ridge.
The marked enrichment in δ 2 H during 2010 and 2011 in C. tetragona stems growing under ambientsnow conditions coincided with a shift towards greater precipitation inputs during winter versus summer.The summers of 2009-2011 were remarkably dry with average June-August precipitation sums of 26 mm.Greater winter snow precipitation inputs normally lead to more δ 2 H-depleted source water, since snow has depleted δ 2 H compared to rain (Welker et al 1995, Welker 2000, Welker 2012).This contradicts our observation of enriched stem δ 2 H values during the last two years.One explanation for this offset between stem-and precipitation-δ 2 H may be the incremental depletion in 2 H during wood formation in C. tetragona stems, with more recently formed wood containing less 2 H-depleted lignified compounds than old wood (White 1989).An alternative explanation for 2 H-enriched stem tissue during these years with low summer precipitation could be the result of stomatal closure in response to drier summer conditions, leading to reduced transpiration rates and associated leaf heating, causing leaf water evaporation and enrichment of newly synthesized leaf and stem tissue.Both factors are considered to lead to leaf water 2 H-enrichment (Farquhar et al 2007).This explanation is supported by observed enrichment in stem δ 13 C during the last two years of growth (figure 1), which may indicate higher water-use-efficiency with lower internal CO 2 vapor pressure after stomatal closure (Farquhar and Sharkey 1982) and reduced discrimination against 13 CO 2 by enzymes involved in the photosynthesis pathway (Farquhar et al 1989).In agreement with our conclusion that growth was indeed water-limited during the last two growth years, we observed relatively low annual stem growth during 2010 and 2011, especially in the medium-and ambient-snow plots (figure 2).However, C. tetragona growth in a heath site close to our research location in Svalbard did not respond to an experimental doubling of ambient summer precipitation over four years (Weijers et al 2013a), suggesting that C. tetragona growth is generally not waterlimited in our study area (Callaghan et al 1989).Instead, we observed a negative relationship between summer precipitation and growth under ambient snow conditions, which has been noted in an earlier study comparing inter-annual climate and C. tetragona growth patterns (Callaghan et al 1989).We conclude, however, that the correlation between precipitation and growth in our study may be corroborated by the inverse relationship observed between summer precipitation and air temperature between 1998 and 2011.

Conclusions
We show that our experimental deep-snow treatment increased C. tetragona stem growth and stem %N.Enriched stem δ 15 N values in deep-snow individuals suggest that an increased N-availability from higher soil winter mineralization rates may have contributed to the observed increase in growth under deep snow.Deeper snow may also have increased frost protection and thereby compensated for the potential negative effects of a shortened growing season, providing an additional explanation for the observed positive growth response to our deep-snow treatment.The importance of both optimal winter and summer conditions for C. tetragona growth may partly explain the frequently observed lack of direct correlations between winter climate conditions and shrub growth.Combined, our results show the importance of multiyear experimental snow manipulation studies to improve our understanding of the complex interactions between the factors regulated by snow depth changes affecting shrub growth: growing season length, winter frost protection, and water-and Navailability.A better understanding of the impacts and interactions of these complex factors is needed to predict the effects of changing snow conditions on shrub growth in the High Arctic.
to remove age-related growth trends, a common procedure in shrub dendroecology (Myers-Smith et al 2015).The final chronology was developed by averaging the residual growth records from ambient snow plots of all three sites by biweight robust mean (n = 19 individuals).The residual chronology was correlated with a standard C. tetragona chronology for the overlapping period 1998-2008, developed from nearby Endalen (figure S2 in supporting information) located within 5 km of our research site (Weijers et al 2010).

Figure 2 .
Figure 2. Annual average detrended Cassiope tetragona stem length (residuals) from 1998-2011 per snow treatment averages across sites (n = 9 replicates snow treatment −1 + SE).The vertical dotted line represents the time of establishment of the snow fences during fall 2006.The dashed line shows mean July temperatures recorded at Svalbard Airport meteorological station.Circles indicate annual growth increments of C. tetragona individuals grown under ambient-snow (c.20 cm), triangles medium-snow (c. 100 cm), and diamonds deep-snow (c.150 cm) conditions.Results of the mixed-model analysis on the effects of snow depth and site on growth are presented in table 3.

Figure 3 .
Figure 3. Isotope carbon (δ 13 C), nitrogen (δ 15 N) and oxygen (δ 18 O) natural abundances of annual (2006-2011) Cassiope tetragona stem increments plotted snow depth −1 site −1 .The year 2007 was the first growth year after snow manipulations started fall 2006.The circles represent isotope discrimination values of C. tetragona individuals grown under ambient-snow (c.20 cm), triangles mediumsnow (c. 100 cm) and diamonds deep-snow (c.150 cm) conditions.Values are means; n = (1-)3 individuals snow depth −1 site −1 ± SE.Results of the mixed-model analysis on the effects of snow depth and site on isotopic values are presented in table 3.
Contribution of working group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change ed T F Stocker et al (New York: Cambridge University Press) Johansson M, Callaghan T, V, Bosiö J, Åkerman H J, Jackowicz-Korczynski M and Christensen T, R 2013 Rapid responses of permafrost and vegetation to experimentally increased snow cover in sub-arctic Sweden Environ.Res.Lett.8 035025 Keuper F, Parmentier F-J, Blok D, van Bodegom P, Dorrepaal E, van Hal J, van Logtestijn R and Aerts R 2012 Tundra in the rain: differential vegetation responses to three years of experimentally doubled summer precipitation in Siberian shrub and Swedish bog tundra AMBIO: J. Hum.Environ.41 269-80 Kielland K 1994 Amino acid absorption by Arctic plants: implications for plant nutrition and nitrogen cycling Ecology 75 2373-83 Leffler J A and Welker J M 2013 Long-term increases in snow pack elevate leaf N and photosynthesis in Salix arctica: responses to a snow fence experiment in the High Arctic of NW Greenland Environ.Res.Lett.8 025023 Loader N J, Robertson I and McCarroll D 2003 Comparison of stable carbon isotope ratios in the whole wood, cellulose and lignin of oak tree-rings Palaeogeogr.Palaeoclimatol.Palaeoecol.196 395-407 Löffler J 2005 Snow cover dynamics, soil moisture variability and vegetation ecology in high mountain catchments of central Norway Hydrol.Process.19 2385-405 Mallik A U, Wdowiak J V and Cooper E J 2011 Growth and reproductive responses of Cassiope tetragona, a circumpolar evergreen shrub, to experimentally delayed snowmelt Arctic Antarct.Alpine Res.43 404-9 McCarroll D and Loader N J 2004 Stable isotopes in tree rings Quat.Sci.Rev. 23 771-801 Michelsen A, Jonasson S, Sleep D, Havström M and Callaghan T V 1996a Shoot biomass, δ 13 C, nitrogen and chlorophyll responses of two arctic dwarf shrubs to in situ shading, nutrient application and warming simulating climatic change Oecologia 105 1-12 Michelsen A, Schmidt I K, Jonasson S, Quarmby C and Sleep D 1996b Leaf 15 N abundance of subarctic plants provides field evidence that ericoid, ectomycorrhizal and non-and

Table 2 .
Pearson's correlation coefficients (r), regression slopes and corresponding P-values between climate parameters and the Cassiope tetragona stem length residual growth chronology (1998-2011) constructed from individuals growing in ambient-snow plots (n = 14 years).Significant correlations are marked with asterisks.Stem length residual data are averages of 18 C. tetragona individuals.Daily average climate data were derived from the Svalbard Airport meteorological station.Regression slopes were calculated by plotting the residual growth chronology against each climate variable.Significant correlations are indicated with asterisks ( * P < 0.05, ** P < 0.01, *** P < 0.001).
).The vertical dotted line represents the time of establishment of the snow fences during fall 2006.The dashed line shows mean July temperatures recorded at Svalbard Airport meteorological station.Circles indicate annual growth increments of C. tetragona individuals grown under ambient-snow (c.20 cm), triangles medium-snow (c. 100 cm), and diamonds deep-snow (c.150 cm) conditions.Results of the mixed-model analysis on the effects of snow depth and site on growth are presented in table 3.
addition, NSF Major Research Instrumentation award (MRI 0923571) supporting the UAA Stable Isotope Lab, awarded to JMW, contributed to the project.and Holmes R L 1986 Users manual for program ARSTAN Tree-Ring Chronologies of Western North America: California, Eastern Oregon and Northern Great Basin with Procedures Used in the Chronology Development Work Including Users Manuals for Computer Programs COFECHA and ARSTAN ed R L Holmes et al (Tucson, AZ: Laboratory of Tree-Ring Research, The University of Arizona) Cooper E J 2014 Warmer shorter winters disrupt arctic terrestrial ecosystems Annu.Rev. Ecology Evol.Systematics 45 271-95 Cooper E J, Dullinger S and Semenchuk P 2011 Late snowmelt delays plant development and results in lower reproductive success in the High Arctic Plant Sci. 180 157-67 Ehleringer J R and Dawson T E 1992 Water uptake by plants: perspectives from stable isotope composition Plant Cell Environ.15 1073-82 Elmendorf S C et al 2012 Plot-scale evidence of tundra vegetation change and links to recent summer warming Nat.Clim.Change 2 453-7 Farquhar G D, Cernusak L A and Barnes B 2007 Heavy water fractionation during transpiration Plant Physiol.143 11-8 Farquhar G D, Ehleringer J R and Hubick K T 1989 Carbon isotope discrimination and photosynthesis Annu.Rev. Plant Physiol.Plant Mol.Biol.40 503-37 Farquhar G D and Sharkey T D 1982 Stomatal conductance and photosynthesis Annu.Rev. Plant Physiol.33 317-45 Frost G V and Epstein H E 2014 Tall shrub and tree expansion in Siberian tundra ecotones since the 1960s Glob.Change Biol.20 1264-77 Hallinger M, Manthey M and Wilmking M 2010 Establishing a missing link: warm summers and winter snow cover promote shrub expansion into alpine tundra in Scandinavia New Phytologist 186 890-9 Havström M, Callaghan T V and Jonasson S 1993 Differential growth responses of Cassiope tetragona, an Arctic dwarfshrub, to environmental perturbations among three contrasting high-and subarctic sites Oikos 66 389-402 Havström M, Callaghan T V, Jonasson S and Svoboda J 1995 Little ice age temperature estimated by growth and flowering differences between subfossil and extant shoots of Cassiope tetragona, an Arctic Heather Funct.Ecology 9 650-4 Hobbie E A and Werner R A 2004 Intramolecular, compoundspecific, and bulk carbon isotope patterns in C3 and C4 plants: a review and synthesis New Phytologist 161 371-85 Holmes R L 1983 Computer-assisted quality control in tree-ring dating and measurement Tree-Ring Bull.43 69-78 IPCC 2013 Climate change 2013 The Physical Science basis.