Short-term herbivory has long-term consequences in warmed and ambient high Arctic tundra

The study site is located in the Advent Valley of Spitsbergen (Adventdalen, 78.19°N, 15.93°E), about 15 km from the town of Longyearbyen. Plots were located in a mesic tundra vegetation community in a wide valley below the Foxfønna glacier, dominated by Salix polaris, Bistorta vivipara, and Alopecurus magellanicus, and with a thick moss and lichen layer. Thirty 2 m × 2 m plots were laid out in five blocks early summer 2003, and randomly assigned to treatments. Half of the plots in each block were passively warmed by plexiglass Open-Top Chambers (OTCs) which allowed sunlight to pass into the plots unhindered (for details of OTC methods, see appendix I). OTCs accomplished a 1.2°C of mean warming of surface temperatures in July in the first summer (Cooper et al 2004). OTCs have been maintained on the site continuously every summer since their initial implementation in 2003, and removed each autumn. Within each block, plots were also randomly distributed between three levels of a grazing treatment by captive Barnacle Geese in 2003–2005: no grazing (UG), intermediate (low) grazing (LG), and high grazing, (HG). Specific treatments and their relationship to grazing pressure are described in Cooper et al (2006). Grazing manipulations were discontinued after 2005.

In the summer of 2014, community composition was recorded at each plot using a 0.75 m × 0.75 m point-frame with 100 evenly-spaced grid points. Each piece of living material at a gridpoint was tallied to the species level (as described in Shaver et al 2001), according to the Pan Arctic Flora for vascular plants (Elven et al 2011), Krypto-S (Swedish Museum of Natural History 2013) for mosses, and the Pan Arctic Checklist (Kristinsson et al 2010) for lichens. In some cases, lichens could only be identified to the genus level. Dead material, black crust, or bare substrate (i.e. rock) were also possible 'hits'. At each gridpoint, the height of the vascular plant canopy layer was measured. The measurements from all points where vascular plants were present were averaged to produce a per-plot vascular plant canopy height. Further description of the point-framing method can be found in appendix I.

Additional measurements were taken on the dominant vascular plants at each site. Plants were sampled in a systematic fashion starting from a specific gridpoint and proceeding to the right. The dominant grass species was Alopecurus magellanicus, and for each of five plants we measured the length of the longest leaf and the height of the flowering stalk. We also counted the total number of flowering stalks in the plot. For the dominant shrub Salix polaris and the dominant forb Bistorta vivipara, we measured leaf size (length and width, and calculated the resulting rectangular area) for five leaves per plot, as well as counting the total number of flowers budding, open or senescing in the plot.

All analyses were performed using R statistical software version 3.1.13 (R Core Team 2015). Point-frame data were used to calculate Shannon diversity index (diversity within each plot), Pielou's evenness index, and beta diversity (diversity between plots) using the R package 'vegan' (Oksanen et al 2012). Beta diversity was calculated using multivariate dispersion to calculate distance between samples within each treatment, and identify a centroid (spatial median) for each treatment. The response variable per plot was the distance from the plot to its centroid.

Data were analyzed using mixed-effects models in the 'lme4' package version 1.1-7 (Bates et al 2012), with experimental block as a random factor and grazing and warming as a fixed factors. Model selection was performed in a backwards stepwise manner using second-order AIC (Akaike Information Criterion) scores or 'AICc' (Mazerolle 2013), which slightly modify the AIC to adjust for small sample sizes and prevent overfitting of the model, and marginal r² (Barton 2013). When the interaction between fixed factors were included in the model, differences between treatments were compared using pairwise tests in the 'multcomp' package (Hothorn et al 2008). All response variables were assessed for normality and homogeneity of variance of the residuals using standard diagnostic procedures. Count data were analyzed using generalized linear mixed-effects models with a Poisson distribution. A few other variables were analyzed using binomial distributions of either total abundance or total grid points. Diversity, density, and size measurements were analyzed using linear models; when no interactions were included but one or both fixed factors were, the significance of individual parameters were assessed using the Satterthwaite approximation in the package 'lmerTest' (Kuznetsova et al 2015). Possible effects of sampling date on floral density were assessed in the same fashion using generalized linear mixed-effects models with a Poisson distribution and sampling date as an additional factor.

Surprisingly and contrary to our hypothesis, total live vascular plant, bryophyte, and lichen abundance in warmed plots did not differ significantly from ambient plots. This may be attributed to harsh environmental limitations in the High Arctic: temperature, nutrients, and moisture. The Svalbard permafrost keeps soil temperatures cooler than in the Low Arctic and contributes to determining plant community composition (Brooker and Van Der Wal 2003). Warming soil temperature, for example through increased insulative snow depth, has been shown to have a positive effect on aboveground productivity in the High Arctic, while merely warming summer air temperature may not (Natali et al 2012). Thus Open Top Chambers placed only during the growing season, as in this experiment, do not necessarily release plants from the cold soil temperatures to the extent that winter warming might. Other environmental limitations include the comparatively nutrient-poor substrate in the High Arctic. Nitrogen and phosphorus are co-limiting and many responses depend on increasing phosphorus supply (Gordon et al 2002, Robinson et al 1998). Lack of additional fertilization may have limited our plant community's response to warming. Finally, soil moisture is also frequently a limiting factor in the Arctic (Elmendorf et al 2012a), and OTC warming dries soils (Bokhorst et al 2013).

Despite few changes in live plant abundance, we found that the amount of dead vascular plant material increased significantly, sometimes by a factor of three or more, in the OTC plots. This pattern is common in many previous experiments (Wahren et al 2005, Klein et al 2004, Oberbauer et al 2007, Elmendorf et al 2012a), but not all (Post and Pedersen 2008). Previous grazing history had no effect on live abundance in the plots, even though initial data from this study shows that the goose grazing had reduced graminoid leaf length (Cooper et al 2006). A long-term herbivore exclosure experiment elsewhere on Svalbard also shows that biomass is greatly reduced with grazing, however the complete exclosure of herbivores in the latter study allowed tundra vegetation to fully recover from the effects of grazing (Sjögersten et al 2011). Thus the primary driver of this accumulation of dead plant material seems to be temperature, rather than grazing history.

Some litter appeared new, as if it was from the previous growing season, but it was not possible to age the litter and determine when the increased buildup began (i.e. early or late in the warming experiment). In low arctic sites vascular plant material decays about 30–60% over two-three years (Hobbie and Gough 2004, Cornelissen et al 2007), while in a polar desert on Svalbard shrub litter decomposed 15%–40% in the first 15 months (Robinson et al 1995). This suggests that litter detected in the plots is likely from recent growing seasons. In fact, in 2013, summer precipitation was three times higher than the average of the previous five years (Abbandonato 2014). This likely released the plants from moisture stress, allowing a strong short-term growth and biomass response to the OTC treatment within a single season and separate from the long-term effects; the resulting biomass would produce more litter for the following season (2014, when we measured the plots). This does not preclude additional long-term mechanisms which could increase litter deposition, but it is important to note that high-precipitation years may become more frequent on Svalbard due to the predicted increase in precipitation due to sea ice melting (Kopec et al 2015) as well as the increasing variability of climate and extreme climate events (IPCC 2014).

The buildup of a thicker litter layer may eventually become a limitation for growth and decomposition by shading prostrate leaves and cooling the soil. More likely, however, is that as has been shown in Greenland (Blok et al 2016), decomposition rates are slowed by the warming and soil drying which is typical in OTCs (Robinson et al 1995). Additionally, a substantial portion of litter in our study came from the increase in dead shrub material, which is typically the most recalcitrant type of vascular plant litter (Myers-Smith et al 2011). Because warming of a greater magnitude can increase the speed of decomposition (Hobbie 1996), this may be a medium-term transient response as climate change accelerates. While the increase in dead plant material could have been aided by the OTC structures themselves, which reduce movement of litter between plots, the removal of the OTCs during the winter season should have minimized this effect.

The increase in litter has major implications for ecosystem functioning, as decomposition of dead material is the main source for carbon and nitrogen cycling. However, since much of the additional litter was comparatively recalcitrant shrub material this effect may be muted (Blok et al 2016). Previous research also suggests that arctic leaf litter has lower nitrogen and phosphorus content due to high nutrient resorption efficiencies during leaf senescence (Aerts et al 2012). Future studies should explicitly include measures of bioavailable nutrient stores in the soil at long-term OTC sites, in parallel with monitoring litter accumulation.

As hypothesized, we found some shifts in functional group abundance, although these changes were more subtle than expected. Evergreen shrub abundance (D. octopetala) was similar in ambient and OTC plots after 12 growing seasons of warming, while deciduous shrub abundance (S. polaris) was lower in warmed plots. Because evergreen shrubs in particular are as limited by soil moisture as they are by summer air temperature (Elmendorf et al 2012a), soil drying by the OTCs could explain this pattern. More broadly, greening has been widely detected experimentally as well as by remote sensing across the Low Arctic (Myers-Smith et al 2011) and occasionally in the High Arctic (for example, Hudson and Henry 2009), but meta-analyses of OTC experiments across biomes indicate that the High Arctic is more resistant to shrub expansion (Elmendorf et al 2012a). This may be because high arctic shrubs are dwarf forms rather than the taller forms which respond more positively to experimental warming (Elmendorf et al 2012a). Our results suggest that the ecosystem services and regulation currently provided by dwarf shrubs (Myers-Smith et al 2011) may remain relatively unchanged in the High Arctic.

However, shrub abundance did differ with previous grazing manipulation. The evergreen shrub D. octopetala was less abundant in highly grazed plots than in ungrazed plots in both ambient and OTC treatments, even nine years after the cessation of goose grazing. Meanwhile, live S. polaris declined in the low-intensity grazing treatment, but only in warmed plots. While geese seldom graze on these species, other associated effects such as trampling could have caused long-term declines which would have consequences for ecosystem services and community structure. Conversely, even though graminoids are a common food item for some geese on Svalbard (Fox et al 2006), grasses showed no difference in abundance by treatment, most likely due to compensatory growth (Cooper et al 2006).

Rushes had higher abundance in warmed plots, regardless of grazing. In fact, the only functional groups which showed higher abundance in the warmed plots were forbs and rushes, each of which only made up a modest amount of the total live abundance in the community. In many systems, an increase in vascular plant abundance excludes mosses and lichens by shading and litter deposition (i.e. Chapin et al 1995), although other mosses such as Sphagnum species are photoinhibited and may benefit from shading (Zona et al 2011). Interestingly, the increase in forb and rush abundance was not coupled with any change in lichen or moss abundance, much less the decline found in the Low Arctic by Cornelissen et al (2001). Instead the canopy remained relatively open, with ∼15%–20% of the area of the plots still lacking a canopy layer after 12 seasons of warming. Canopy cover and height were also unaffected by the previous grazing treatment.

Contrary to our predictions, we found no difference in diversity between warmed and ambient plots or between different grazing treatments, as measured by α-diversity (Shannon's index), β-diversity, and community evenness (Pielou's index). All the species in the mesic meadow are extremely long-lived (i.e. Parsons et al 1994, Eriksson and Froberg 1996); barring a mortality event, substantial recruitment of new species may take even longer than a decade or two. Environmental factors are also different in the Adventdalen mesic meadow compared to low arctic and subarctic sites, where long-term declines in diversity have previously been documented primarily at wet sites (Elmendorf et al 2012a).

We hypothesized that OTC warming would have a generally positive effect on dominant species in the tundra communities, but results showed much more variable and nuanced responses. Twelve seasons of warming caused a small but significant decrease in the live abundance of the dominant shrub S. polaris, a weak but significant increase the live abundance of the dominant forb B. vivipara, and no effect on the live abundance of the dominant grass A. magellanicus. Grazing history had a nuanced effect on S. polaris abundance (discussed earlier), no effect on abundance of A. magellanicus, and led to an increase in B. vivipara abundance only in the warmed, low-intensity grazing treatment. Neither warming nor grazing had any effect on leaf size in any of the three focal species.

Grazing history generally promoted flowering, and to a greater extent in warmed than ambient-temperature plots. All three species produced more abundant flowers in the OTC plots, and in A. magellanicus flowering stalks were taller in OTC than in ambient plots even though there was no difference in the height of vegetative leaves. Increases in reproduction in response to warming in the High Arctic have been found previously in several species (Wookey et al 1995, Wada and Kanda 2000, Buizer et al 2012, Klady et al 2011), theorized to be due to the abiotic limitations which may continue to constrain vegetative growth even as conditions are ameliorated by warmer temperatures. Plants in the High Arctic have evolved to prioritize reproduction under favorable conditions. Many vascular plants, including B. vivipara and S. polaris, pre-form embryonic flower buds which do not open until the next, or several more, growing seasons. In fact, sexual reproduction is quite unsuccessful in B. vivipara, which reproduces primarily by bulbils (Diggle et al 1998). Alopecurus magellanicus reproduces both sexually and vegetatively, as does S. polaris (Brochmann and Steen 1999).

One major question is whether investing in reproductive effort as the climate warms will have long-term costs: will increasing flowering and fruit production for several consecutive favorable seasons take a toll? Classic theory as well as recent experimental work suggests that reproduction sustains costs to future fecundity, vegetative growth and survival (Alvarez-Cansino et al 2010, Bell 1980), although compensatory mechanisms have also been demonstrated. In a climate change context, recent evidence shows that as growing seasons become warmer and longer, the survival costs of increasing reproduction shrink over three to five years of manipulating the level of reproduction (Sletvold and Ågren 2015). Thus even if sexual reproduction is not the main mode of propagation in these plants, the increase in flowering may become a less costly way to supplement total reproduction as the climate warms. Increased sexual reproduction will also provide plant populations with more possibilities for evolutionary adaptation to changing environmental conditions (Hedhly et al 2009). Increasing sexual reproduction in species which have multiple reproductive modes at their disposal is a particularly interesting potential eco-evolutionary feedback which should be further explored.