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

We examine the influence of altered winter precipitation on a High Arctic landscape with continuous permafrost. Gas exchange, leaf tissue element and isotopic composition (N, δ13C, δ15N), and plant water sources derived from stem and soil water δ18O were examined in Salix arctica (arctic willow) following a decade of snow-fence-enhanced snow pack in NW Greenland. Study plots in ambient and +snow conditions were sampled in summer 2012. Plants experiencing enhanced snow conditions for 10 years had higher leaf [N], photosynthetic rate, and more enriched leaf δ15N. Enhanced snow did not influence stomatal conductance or depth of plant water use. We attribute the higher photosynthetic rate in S. arctica exposed to deeper snow pack to altered biogeochemical cycles which yielded higher leaf [N] rather than to enhanced water availability. These data demonstrate the complexity of High Arctic plant responses to changes in winter conditions. Furthermore, our data depict the intricate linkages between winter and summer conditions as they regulate processes such as leaf gas exchange that may control water vapor and CO2 feedbacks between arctic tundra and the surrounding atmosphere.


Introduction
Increased precipitation in the High Arctic associated with global climate change is expected to influence plant function Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. 1 Address for correspondence: Environment and Biomedical Laboratory, 3151 Alumni Drive, University of Alaska-Anchorage, Anchorage, AK 99508, USA. and ecosystem processes (Welker et al 2005, Schimel et al 2004, Buckeridge and Grogan 2008, Christiansen et al 2012. Higher snowfall has been observed in the High Arctic (Kohler et al 2006, Min et al 2008, and recent studies document increased transport of atmospheric moisture northward and greater discharge of rivers into the Arctic Ocean in Eurasia (Peterson et al 2002, Zhang et al 2013. During winter, deeper snow pack can increase soil temperature resulting in greater CO 2 flux and enhanced nutrient cycling (Brooks and Williams 1999, Schimel et al 2004, Rogers et al 2011. The effects of enhanced snow depth persist during summer, reducing soil temperature post-snowmelt, increasing soil moisture, and delaying leaf production (Wipf and Rixen 2010, Cooper et al 2011, Rogers et al 2011. Warmer soils during winters with enhanced snow clearly stimulate microbial activity and result in greater soil N availability (Schimel et al 2004, Buckeridge andGrogan 2008). The influence of greater soil moisture on carbon gain is less well understood and depends on the plant communities examined (Oberbauer et al 2007, Lafleur and Humphreys 2007, Nobrega and Grogan 2008.
Plant and ecosystem function were altered at the same study site following five consecutive winters of experimentally enhanced snow pack in northwest Greenland (Rogers et al 2011). Deep snow led to higher soil [N] and leaf [N] in Salix arctica (arctic willow) and higher over-winter CO 2 efflux, summer gross ecosystem productivity, and summer ecosystem respiration; leaf-level photosynthesis was not measured. Understanding the influence of enhanced snow pack on photosynthesis is important because plant productivity may have a greater influence on net ecosystem exchange than soil respiration in Arctic ecosystems (Griffis andRouse 2001, McFadden et al 2003) although this is not universally the case (Lafleur and Humphreys 2007). Moreover, carbon gain by Salix sp. may be a key factor in the carbon balance of High Arctic ecosystems (Steltzer et al 2008) since shrubs are becoming more abundant in the region (Wahren et al 2005, Wipf and Rixen 2010, Callaghan et al 2011a. Increased shrub abundance is likely to have a complex influence on active layer depth with potential feedbacks on climate (Blok et al 2010, Bonfils et al 2012. If greater snow pack is having an influence on photosynthesis, it could be through the alleviation of drought stress, an enhancement of nitrogen availability, or a combination of both (Osmond 1987, Evans 1989. Drought stress may be mitigated by the additional water increasing soil moisture content or by delayed melt preventing water stress from developing because leaf expansion and water use are delayed (Cooper et al 2011, Wipf and Rixen 2010). If the primary mechanism is alleviation of water stress, we should observe higher photosynthesis (A), stomatal conductance (g) and transpiration (E) under enhanced snow pack. Additionally, we may observe that plants are using water from more shallow soils in areas of deeper snow; water use from various depths can be discerned using oxygen isotope composition ( In this study, we examined photosynthetic gas exchange, leaf tissue [N] and δ 15 N, and stem and soil δ 18 O in S. arctica following ten years of experimentally enhanced snow pack in northwest Greenland. We ask if (1) gas exchange differs in this species between ambient and enhanced snow plots, and if (2) any differences in gas exchange can be attributed to a response to water availability, or a through a relationship with leaf [N].

Materials and methods
An increased snow depth experiment was established near Thule Air Base in northwest Greenland (76 • 32 N, 68 • 42 W, 170 m elevation) in 2002; this study was conducted in 2012. Vegetation is a dwarf-shrub/herb tundra dominated by Salix arctica and Dryas integrifolia similar to a local long-term climate change experiment (Sullivan et al 2008). Mean annual precipitation and temperature  is about 120 mm and −11 • C, respectively. June, July and August 2012 (figure 1) were warmer (5.3 versus 4 • C) and wetter (127 versus 40 mm) than average. Snow depth was enhanced using a snow fence (15 m long × 1.2 m tall), consisting of vertical slats of wood approximately 5 cm wide with similar sized spaces erected in summer 2002. Snow accumulation was enhanced on the windward and leeward side of the fence and measured near the fence on 29 May 2012. We used a 2 m grid of sample points extending 10 m from the fence in each direction, running 12 m parallel to the fence, and recorded snow depth at each point (n = 70). Snow samples were collected from a single snow pit at four depths between the surface and 1 m to determine water equivalent of depth. Snow depth was also recorded at 24 additional points around the study site not influenced by the fence. Finally, distance from the fence to where snow depth was approximately equal to that at uninfluenced points was measured (about 35 m). The locations of study plots were selected based on snow accumulation and topography. Since soils are shallow, rocky and frozen in the spring, much of the additional water accumulated near the snow fence runs off into shallow depressions on the largely flat landscape, these run-off areas were avoided when selecting deep and ambient snow accumulation zones for study. Three study plots (1.5 m × 1.5 m) were selected in deep snow areas (+snow) and three other study plots were selected in ambient snow areas. The +snow plots were between 2 and 4 m on the windward side of the fence and the ambient snow plots was about 40 m from the fence.
Beginning in late June, and continuing until early August the study plots were sampled every week (six sampling periods). Sampling began approximately four weeks following melt in the ambient plots and two weeks following melt in the +snow plots. Leaf-level gas exchange on five leaves of S. arctica per study plot was measured using a portable photosynthesis system (model 6400, Li-Cor Inc., Lincoln, NB). Randomly selected leaves were measured in the field using quantum flux similar to ambient conditions at solar noon (1250 µmol m −2 s −1 for wavelengths 400-700 nm, 10% blue light), and a reference [CO 2 ] of 400 µl l −1 inside the 2 cm 2 chamber. Leaves completely filled the chamber. Air temperature during days of measurement varied between 4.7 and 11.4 • C; soil temperature varied between 5.8 and 10.7 • C; and relative humidity varied between 57 and 89% (table 1). Several leaves from each plot were collected each measurement day for analysis of [N] (%N), carbon (δ 13 C) and nitrogen (δ 15 N) isotope composition (Dawson et al 2002). Leaf tissue was oven dried at about 70 • C and ground to a fine powder in a mechanical shaker (Mini-Beadbeater-16, Biospec Products Inc., Bartlesville, OK) with 3.2 mm ball bearings. Samples were combusted in an elemental analyzer (model 4010, Costech Inc., Valencia, CA) coupled to a mass spectrometer (model Delta V, Thermo Scientific, Waltham, MA) at the University of Alaska-Anchorage. Standard deviation of standards from this analysis for %N, δ 15 N and δ 13 C was 0.02, 0.15, and 0.03, respectively.
Gas exchange and leaf tissue data collected each week were examined with mixed model analyses using the LME4 package within the statistical computing environment R (R Development Core Team 2008). For each response variable two alternative models were compared: model 1 was a random effects model with fixed day of year and random day of year conditioned on plot, model 2 added treatment (ambient or +snow) as a fixed effect; both models were run with uncorrelated random effects and day as a factor. The most appropriate model had the lower akaike information criterion (AIC) and significance was assessed at α = 0.05 using the log-likelihood ratio of the two models. Photosynthesis (A), stomatal conductance (g) and transpiration (E) were logtransformed to meet normality assumptions. The relationship between mean photosynthesis for each measurement day and leaf temperature, air temperature, relative humidity and soil moisture were examined with Pearson correlation using the STATS package in R.
To explore the source of water used by arctic willow we measured oxygen isotope composition (δ 18 O) of water from soil and S. arctica stems collected weekly. A small soil pit was dug to about 20 cm. Soil samples were collected from the wall of the pit using a 1.25 cm dia. metal tube at depths of 5, 10, 15 and 20 cm below the soil surface. A woody stem sample, lacking green tissue, was also collected. Soil and xylem water δ 18 O were examined using a cavity ring-down spectrometer coupled with a heat induction module (IM-CRDS, Picarro Inc., Santa Clara, CA) which simultaneously extracts and analyzes extremely small samples of water (<3 µl). Samples were loaded into metal sample holders (heavy foil clips for stem cross sections, small tubes for soil), placed into a glass vial with a septum and inserted into the IM. The IM heated the sample vaporizing the water which was passed to the CRDS for analysis. Data are expressed in the familiar δ notation for 18 O using the VSMOW standard (Dawson et al 2002). Data were processed to remove any memory effects among samples and to adjust for drift from the standard (supplement 1 available at stacks.iop.org/ERL/8/025023/mmedia). Soil samples were also weighed and dried to determine water content.
We used xylem and soil water δ 18 O to estimate the fraction of water in S. arctica taken from 5, 10, 15 or 20 cm soil depth. Data were examined with a Bayesian stable isotope mixing model (Parnell et al 2010) implemented in the SIAR package of R using 200 000 iterations with 50 000 discarded as burn-in. Soil gravimetric water content was incorporated into the model as prior information making the assumption that plants preferentially extract water from the wettest soil layers. The SIAR package uses a Dirichlet prior distribution which requires that the proportions of all isotopic sources (i.e., soil depths) sum to unity.
The response of A to quantum flux was measured in late July and early August. Measurements were made on two leaves per plot between 29 and 30 July, and again on two leaves per plot between 4 and 5 August. Reference CO 2 was where A max is the maximum photosynthesis rate at high light, A qe is the apparent quantum yield or the slope of the curve at low light and LCP is the light compensation point (x-intercept) or the minimum light required to achieve net positive A. Incident light during the experiment is PPF (photosynthetic photon flux), and A is measured net photosynthesis. Each parameter of the equation was modeled using the NLME (non-linear mixed effects) package in R with a random plot effect. Separate models were run for each of the two days of data collection and for each treatment. Significant treatment effects were determined by comparing the overlap of 95% confidence intervals for each parameter during each measurement date.

Results
Snow depth on 29 May 2012 exceeded 1 m near the fence. Two meters leeward from the fence mean snow depth was 114 ± 12 cm (mean ± SD); 10 m from the fence, snow depth was 96 ± 13 cm. On the windward side of the fence, two meters from the fence the snow depth was 81 ± 7.3 cm and 10 m from the fence, the snow depth was 54 ± 5.6 cm (see also supplement 2 available at stacks.iop.org/ERL/8/ 025023/mmedia). Conversely, snow depth varied between 15 and 40 cm (mean 23 ± 6.1 SD) at snow sample points not influenced by the fence. Water content of the snow pack near the snow fence was about 37%. Snow depth in the high snow accumulation plots was 74 ± 9.0 cm. Data collected in previous years suggest soil temperature can be up to 15 • C warmer under deep compared to ambient snow (figure 1). Leaf-level instantaneous gas exchange differed between the ambient and +snow conditions. Significant differences among treatments were evident in A (figure 2, table 2). During late June and early July (DOY 182 and 187), A was nearly identical in +snow and ambient plots but differences became apparent during the height of the growing season in mid-July (DOY 195 and 202) and remained higher through early August (figure 2). For the entire growing season, mean  Table 2. Mixed model analysis of gas exchange and leaf tissue response variables. Model 1 for each response variable included day as a fixed factor and day|plot as a random effect; model 2 for each response variable included treatment * day as a fixed factor and day|plot as a random effect. Response variables include net photosynthesis (A), stomatal conductance (g), sub-stomatal [CO 2 ] (ci), transpiration (E), total leaf [N] (N), nitrogen isotope composition (δ 15 N), and carbon isotope composition (δ 13 C). Subscripts indicate the models detailed above. Significant P-values (P < 0.05) indicate that model 2 is preferred over model 1, hence the treatment effect is significant. A in +snow areas was 26% higher than in ambient areas.

Model df AIC
In contrast, g did not differ significantly between treatments.
Curiously, a small difference in E was significant, but a similar difference in ci was only marginally significant (table 2). During the first half of the growing season gas exchange typically increased, then became more variable later in the summer (figure 2), but A was consistently higher and ci was consistently lower in +snow plots. We observed a significant correlation between A and relative humidity (r = 0.833, P = 0.040, n = 6) during summer 2012. Correlations between A and air temperature (r = −0.807, P = 0.052), or soil temperature (r = 0.424, P = 0.402), or soil moisture (r = 0.007, P = 0.984) were not significant. Analysis of dried leaf tissue suggests increased snow depth altered leaf [N] and leaf δ 15 N (figure 3). For the entire growing season, mean leaf [N] in +snow plots was 20% greater than in the ambient areas and this difference was significant (figure 3, table 2). Differences between treatments were most pronounced during the height of the growing season in mid-July (DOY 195 and 202) when leaf [N] was 27% higher. Leaf δ 15 N was 2.5 enriched in the +snow plots compared to ambient plots, also a significant difference (figure 3, table 2). No treatment effect was observed for leaf tissue δ 13 C. Leaf [N] declined considerably from June  through August while little seasonal trend was observed in leaf δ 15 N or δ 13 C. Light curves further suggest greater photosynthetic capacity by plants in +snow plots (figure 4). While the overall curves do not differ significantly between +snow and ambient conditions, A max was 22% higher in +snow plots in late July and 39% higher in early August, with the latter being significant (table 3). Light compensation point and A qe were similar among treatments.
We observed few clear patterns in xylem or soil δ 18 O, or soil moisture during the growing season (supplements 3 and 4 available at stacks.iop.org/ERL/8/025023/mmedia). Xylem water δ 18 O in S. arctica was relatively constant in +snow areas with a slight positive trend during the season; δ 18 O was more variable in S. arctica from the ambient snow area. Overall, δ 18 O of xylem water was −15.9 ± 3.4 in ambient areas and −17.2 ± 2.0 in +snow areas. Soil gravimetric moisture content, averaged across the summer, was 21 ± 13% and 16 ± 14% in the +snow and ambient plots, respectively. Water extraction by soil depth did not differ between +snow and ambient treatments and declined with depth in each (figure 5). The uppermost 5 cm of soil contributed 38% and 39% in +snow and ambient treatments, respectively. Soils at 10, 15 and 20 cm contributed 24, 20 and 18% of the stem water in the +snow plots, and 24, 21 and 16% in the ambient plots.

Discussion
Nearly ten years of increased snow depth has enhanced photosynthesis (A) in S. arctica, which we attribute to higher leaf [N]. Plants in plots with greater snow accumulation had higher A and maximum photosynthesis (A max ), and marginally lower sub-stomatal [CO 2 ] (ci), but did not differ from ambient plots in stomatal conductance (g) or depth of water extraction. Hence, we cannot conclude that additional water provided by the enhanced snow pack near the snow fence alone had a substantial influence on A. Rather, greater snow accumulation increased winter soil temperature and likely altered N-cycling (Schimel et al 2004, Buckeridge andGrogan 2008) leading to greater N uptake by S. arctica, higher leaf [N], and ultimately higher A.
High leaf [N] is consistently linked to high A because leaf [N] is indicative of RUBISCO content, an N-dense protein (Evans 1989). In the High Arctic of Svalbard, leaf [N] was closely associated with differences in A max between Salix polaris, Dryas octopetala and Saxifraga oppositifolia (Muraoka et al 2008). Moreover, large scale arctic-wide studies indicate a tight linkage between biomass, leaf N and ecosystem photosynthetic capacity, regardless of species composition (Williams andRastetter 1999, van Wijk et al 2005). Leaf [N] and A responded positively to experimental additions of nitrogen (Baddeley et al 1994) as did gross ecosystem productivity in a polar semi-desert ecosystem within 10 km of our study site in NW Greenland (Arens et al 2008). Our observation of a close linkage between photosynthetic performance and leaf [N] further emphasizes the important role N limitation plays in arctic ecosystem processes (Schimel et al 2004).
Few studies have examined the influence of altered snow conditions on leaf-level A in the Arctic. At Toolik Lake, Alaska, reduced snow pack had little influence on A of several species including Salix pulchra (Starr et al 2008). However, A max was lower following five years compared to three years of snow removal. More recently, long-term snow reduction (about 14 years) led to lower rates of A max , but added snow did not significantly increase A max in Betula nana or S. pulchra (Pattison and Welker, unpublished). Differences between our findings and these may arise from differences in growing season length or temperature between the High Arctic of Greenland and the Low Arctic of Alaska (Billings 1973).
We have experimental evidence that enhanced snow accumulation influences leaf [N]. Leaf [N] in S. arctica individuals from deep snow areas compared to ambient snow zones was 50% greater in early July 2007 at the same study site in Greenland (Rogers et al 2011). Here, following several more years of enhanced snow accumulation, we observed higher leaf [N] throughout the growing season. In a separate study in NW Greenland, S. arctica in areas of naturally deep snow pack had high leaf [N], but not necessarily higher A (Sullivan and Welker 2007). Near Toolik Lake, Alaska, higher leaf [N] was observed in several species including woody plants such as Betula nana and Dryas octopetala following up to six years of enhanced snow (Welker et al 2005) indicating similar processes operate at both study sites.
Enriched leaf nitrogen isotope composition (δ 15 N) in +snow compared to ambient plots is consistent with more rapid nitrification with deeper snow pack ( . If S. arctica is using primarily NH + 4 , or if NO − 3 is rapidly leached, leaf tissue in +snow plots should be enriched in 15 N as observed here. In another study, warmer soils associated with greater insulation during winter increased microbial biomass and activity (Schimel et al 2004, Buckeridge andGrogan 2008). This higher activity appears to have slightly increased available NH + 4 , but dramatically increased available NO − 3 at Toolik Lake, Alaska (Schimel et al 2004). A pulse of NH + 4 and NO − 3 was observed during snow-melt in high snow plots at Daring Lake, Canada (Buckeridge and Grogan 2010); this pulse, however, was largely removed from the system prior to leaf out.
Despite the enhanced snow areas yielding about double the melt water as ambient areas, water alone had little influence on differences in gas exchange physiology or depth of water extraction between treatments. Other studies report minor influence of enhanced snow on subsequent soil moisture (Buckeridge and Grogan 2008, Rogers et al 2011), while larger differences have been observed elsewhere (Cooper et al 2011). The direct role of water in this study may have been limited for several reasons including: much of the excess water in the enhanced snow plots simply running-off during snow-melt (Dunne and Black 1971) since soils are frozen; summer 2012 was characterized by repeated rains which kept soil moisture high, preventing a strong contrast between gas exchange physiology under wet and dry soil conditions; and we have only measured leaf-level g and not accounted for differences in leaf area index (Steltzer and Welker 2006) between treatments which may indicate an influence of water availability at the ecosystem scale.
We observed limited influence of atmospheric moisture deficit on the seasonal pattern of A in summer 2012. Since relative humidity and A were positively correlated among measurement days we conclude that atmospheric water deficit was having a minor influence on carbon gain. High vapor pressure deficit can reduce A in arctic and alpine plants, even without inducing stomatal closure (Johnson and Caldwell 1975); similar to our observation that g changed little throughout the summer. Although leaf [N] differs between +snow and ambient plots, it appears to have little influence on the seasonal course of photosynthesis which is likely controlled by phenology.
Our data support the contention that a changing arctic climate is having widespread influence on ecosystem and plant processes (Post et al 2009, Callaghan et al 2011b. Enhanced A in S. arctica following a decade of increased snow accumulation likely resulted from increased microbial activity when snow was present, promoted by warmer soil temperatures from insulation (Schimel et al 2004, Buckeridge and Grogan 2008, Christiansen et al 2012. Consequently, winter conditions influence summer plant processes and carbon gain may be enhanced by additional snow in the Arctic. Conversely, delayed melt may shorten the time available for carbon acquisition and cooler soils in the early summer can reduce growth and N-uptake potentially offsetting higher A (Wipf 2010, Leffler et al 2011