Exclusion of brown lemmings reduces vascular plant cover and biomass in Arctic coastal tundra: resampling of a 50 + year herbivore exclosure experiment near Barrow, Alaska

To determine the role lemmings play in structuring plant communities and their contribution to the ‘greening of the Arctic’, we measured plant cover and biomass in 50 + year old lemming exclosures and control plots in the coastal tundra near Barrow, Alaska. The response of plant functional types to herbivore exclusion varied among land cover types. In general, the abundance of lichens and bryophytes increased with the exclusion of lemmings, whereas graminoids decreased, although the magnitude of these responses varied among land cover types. These results suggest that sustained lemming activity promotes a higher biomass of vascular plant functional types than would be expected without their presence and highlights the importance of considering herbivory when interpreting patterns of greening in the Arctic. In light of the rapid environmental change ongoing in the Arctic and the potential regional to global implications of this change, further exploration regarding the long-term influence of arvicoline rodents on ecosystem function (e.g. carbon and energy balance) should be considered a research priority.


Introduction
Multiple lines of evidence suggest that Arctic terrestrial ecosystems are sensitive to climate change (Post et al 2009) and that the amplified warming ongoing in the Arctic (IPCC 2007) may be resulting in large-scale and rapid changes in many Arctic ecosystems (ACIA 2005, Hinzman et al 2005, Callaghan et al 2011b. While these changes have the potential to affect both the cryospheric (Hinzman et al 2005) and hydrologic (Rowland et al 2010) properties of Arctic landscapes, the degree to which warming alters ecosystem processes, such as gross ecosystem primary productivity, is less clear. Based on satellite remote sensing, Bhatt et al (2010) suggest that widespread 'greening' has occurred in many Arctic regions, particularly coastal landscapes that have warmed adjacent to areas of pronounced sea-ice loss. These results are compelling, but there is a pressing need to further validate these findings using ground-based studies, examine what shifts in ecosystem properties are associated with these changes, and the implications these changes may have on other components of the Arctic system. Additionally, the response of Arctic vegetation to warming may be confounded by other factors affecting plant growth and phenology, plant community composition and land cover change. For example, several recent studies highlight the strong impact herbivores can have on plant community composition, ecosystem function and other processes (Pajunen et al 2008, Post and Pedersen 2008, Madsen et al 2011, Villarreal et al 2011, Lara et al 2011. Such interactions between herbivory, warming and greening of the Arctic remains poorly studied, and there is uncertainty regarding the potential for herbivores to directly enhance greening (Olofsson et al 2009).
Herbivory may become intensified if warming results in higher gross primary production and increased energetic support for herbivore populations (e.g. Oksanen et al 1981, Post and Pedersen 2008, Madsen et al 2011. Recent results suggest that seemingly subtle herbivore activity can still affect ecosystem function (Wookey et al 2009). Herbivores can shape community structure directly through selective herbivory (e.g. Tape et al 2010), trampling (e.g. van der Wal et al 2001), and/or indirectly by facilitating apparent competition (e.g. Olofsson et al 2002) and altering nutrient cycling (e.g. McKendrick et al 1980). The direction and magnitude of herbivore impacts, however, can vary widely depending on the ecosystem and the plant and herbivore species present (Mulder 1999). In some cases, patterns of herbivore impacts emerge only after herbivore exclusion exceeding 10+ years (e.g. Gough et al 2008).
The boom-bust population dynamics of the brown lemming (Lemmus trimucronatus) has long intrigued researchers in northern Alaska (Thompson 1955a, Pitelka andBatzli 2007). Early observers of periodic outbreaks of lemmings noted that lemmings could denude the landscape of graminoids, resulting in the depletion of their food supply and increased rates of predation (Thompson 1955b). Thus, lemmings were thought to have an important role in regulating many tundra ecosystem properties and processes (Pitelka 1957). This view formed the foundation of the Nutrient-Recovery Hypothesis (Schultz 1964), which described ecosystem responses following peak lemming years. This hypothesis proposed that during interim years between population highs, vegetation recovers, standing dead biomass accumulates, and gross primary production returns to levels that support increased lemming densities (Schultz 1964). To assess the impact of lemmings on the tundra ecosystem near Barrow, Alaska, exclosures were built in various plant communities among different land cover types (Thompson 1955b, Schultz 1964, 1969. During periods with lemming population highs, graminoid biomass in control plots was greatly reduced relative to that of the exclosures in the short-term (Thompson 1955b). However, resampling of these plots 15 years after establishment showed that longterm exclusion of lemmings resulted in higher, rather than lower, graminoid biomass in control plots relative to exclosures (Batzli 1975), thus highlighting the difference between longterm and short-term plant community responses to lemming outbreaks.
As a contribution to the International Polar Year-Back to the Future (Proj. No. 512) project (Callaghan et al 2011a), we revisited the exclosures established by Schulz in 1959 and collected data during the 2002 and 2010 growing seasons. We used 12 of the original exclosures that remain intact and have effectively excluded lemmings for the past 50+ years (figure 1). Biomass data were collected in 2002 from these plots, and we repeated biomass sampling and estimated cover for vascular and nonvascular plant species in 2010. Unfortunately, historic biomass data collected in 1974 no longer exist in their raw format. However, we qualitatively compare mean biomass measured in 1974 by Batzli (1975) to data collected in 2002 and 2010. This study represents what we believe to be the longest herbivore exclosure experiment in the Arctic. Specifically, we examine the effects of sustained herbivory on plant community composition and the degree to which lemming herbivory may have contributed to the regional greening signal detected by Bhatt et al (2010). Thus, we asked: (i) how has long-term herbivore exclusion altered community composition and diversity?; (ii) how did this change between 1974, 2002 and 2010?; and (iii) did effects differ among exclosures in dry, moist and wet tundra?

Site descriptions
All experimental plots used in this study are located within 5 km of Barrow, Alaska (71 • 18 N, 156 • 40 W) on the Arctic Coastal Plain of northern Alaska (Brown et al 1980). Mean annual temperature, precipitation, and snowfall are −12 • C, 11 cm and 69 cm respectively ; (NCDC 2005). While there has been a slight warming trend in the Barrow area over the past ten years (Shiklomanov et al 2010), neither the study years presented here, nor the year preceding, were exceptionally warm or wet. The landscape is underlain by continuous permafrost and active layer depth rarely exceeds 40 cm. The region's low topographic relief (0-10 m a.s.l.) results in a fairly 'waterlogged' environment, 65% of which is polygonized tundra (Brown et al 1980). Vegetation varies according to microtopography and soil moisture, where drier heath-like tundra with more shrubs and forbs dominates slightly elevated areas, such as old beach ridges, high-center polygons and the rims of some low center polygons; and graminoids dominate wetter areas with poor drainage, such as lower troughs, meadows and marshes. Brown lemmings are fairly common in this ecosystem, eating primarily graminoids and bryophytes in winter, and have periodic population outbreaks with very high densities every 3-5 years . While the brown lemming is the primary herbivore in this system, collared lemmings (Dicronstonyx groenlandicus) live at very low abundances . Caribou also occur in coastal tundra but they are rare at this study site due to hunting pressures from the local Barrow community.
In 2002 and 2010 we resampled twelve pairs of control and exclosure plots established by Schultz (1964) in 1959. Sampling efforts in 2002 and 2010 occurred two years following lemming population outbreaks in 2000 and 2008 respectively (Holt 2010). Exclosures were situated in dry, wet and intermediate (moist) land cover types (n = 4 in all cases, figure 1) and consisted of a 2 m × 2 m plot with posts driven to permafrost surrounded by 1.27 cm 2 wire mesh buried 10-15 cm into the soil and extending ca. 75 cm above the ground. Control plots were established within 5 m of each exclosure and consisted of a 2 m × 2 m plot marked with wooden pegs. We quantitatively determined the land cover type for each site by measuring soil moisture with at Spectrum Time Domain Reflectometer probe, and categorized sites as dry below 30% volumetric water content (VWC), moist between 30% and 60% VWC, and wet above 60% VWC.

Plant cover and biomass
In 2010, during the final week of July and close to peak growing season, we measured plant species cover by visually estimating the cover of each vascular and nonvascular species present in the center 1 m 2 of the exclosure and control plots. We also estimated the cover of bare ground, vascular plant litter and animal sign (lemming clippings, burrows, trails and feces).
Close to peak growing season in 2002 and 2010, we randomly located two 10 cm × 10 cm quadrats within all 1 m 2 plots described above and collected biomass samples by cutting through surface vegetation to below root depth and removing all plant and soil within the quadrat. An individual plant was considered 'within' the sample if the meristems were within the quadrat. All live, aboveground biomass (green material and woody stems) were sorted manually and vascular and nonvascular plants were assigned to various functional groups. We also combined standing and loose dead material as litter. All material was then oven dried at 70 • C for 48 h and weighed.

Statistical analysis
We calculated species richness and the Shannon-Wiener index of diversity using 2010 cover estimates for individual species.
To determine how herbivore exclosures affected diversity among land cover types with different plant communities, we applied a two-factor ANOVA using treatment and plant community as main factors. We examined differences between groups using least significant differences (LSD) in a post hoc test. To see how plant functional types as a whole differed with treatment and land cover type, we combined specific data for cover into the following classes: lichens, bryophytes, forbs, deciduous shrubs, and graminoids. To calculate relative cover, which adjusts community composition for varying total cover in different samples and allows for better comparison among land cover types, we included categories for animal sign, litter, and bare ground. We only used six plant functional types; five living categories plus litter, for analysis of biomass. To determine the effect of treatment (exclosure) and land cover types on species composition, we used MANOVA with cover of plant functional types as the multiple variables and with exclosure or control (EX/CT) and land cover type (LCT) as main factors. We used a two-way repeated-measures MANOVA to test for differences in biomass among plant functional types (treatment and plant community as main factors, and year, 2002 or 2010 as the repeated factor). We examined differences among functional types using univariate ANOVAs within the MANOVAs and post hoc LSD tests. All data were checked for normality and homogeneity of variance, and to meet these assumptions cover and biomass data were arcsine and natural log transformed respectively. All statistical analyses were performed using JMP V.8.0 (Cary, NC).

Diversity
Exclusion of lemmings for over 50 years did not result in overall differences in species richness between exclosure and control plots (table 1). Species richness differed among  Table 1. Mean (±s.e.) species richness and the Shannon-Wiener Index of Diversity for herbivore exclosures (EX) and control plots (CT) in the three land cover types (LCT). ANOVA results revealed that richness differed among land cover types (F 2,18 = 26.2, p < 0.0001), but exclosures and controls did not differ significantly (F 1,18 = 8.2, p = 0.5). The ANOVA results for the Shannon-Wiener Index revealed a significant treatment*land cover type interaction (F 2,18 = 4.6, p = 0.02).

Richness
Shannon-Wiener index land cover types, with dry communities having the most species and wet communities the fewest, but no statistical difference occurred between exclosures and control plots. This pattern generally held for the Shannon-Wiener index (which mathematically incorporates evenness and richness) of diversity, except that an interaction occurred so that exclosure plots had higher diversity than control plots at dry and moist sites, but controls had slightly higher diversity at wet sites (table 1).

Cover
Unlike diversity, exclusion of lemmings resulted in strong changes in the response of functional types to land cover type (figure 2; EX/CT main effect, Wilks' λ = 0.05, F 7,12 = 9.0, p < 0.001; LCT main effect. Wilks' λ = 0.03, F 14,24 = 8.1, p < 0.001; EX/CT*LCT Interaction Wilks' λ = 0.1, F 14,24 = 3.73, p = 0.002). Univariate analysis revealed greater animal sign and bare ground in control plots than in exclosures (no animal sign was observed in the exclosures), which did not vary with land cover type (table 2). Interestingly, bryophyte, deciduous shrub and forb cover did not show treatment effects, only land cover type effects. Cover of bryophytes and forbs generally increased as soil moisture increased, whereas more deciduous shrubs occurred in drier tundra (figure 2). Graminoids increased as soil moisture increased, and decreased in exclosures, particularly in wet tundra ( figure 2, table 2). Lichens showed the opposite response, decreasing in cover with soil moisture and increasing in cover in exclosures, particularly at drier sites. More litter occurred inside lemming exclosures than in control plots, particularly in wet tundra.

Biomass
The distribution of biomass among functional types differed with year (figure 3; Wilks' λ = 0.44, F 6,31 = 2.3, p = 0.06), land cover type (Wilks' λ = 2.2, F 6,31 = 11.38, p < 0.001) and treatment (Wilks' λ = 0.26, F 12,62 = 4.97, p < 0.001), however, there were no significant interactions among main effects. Similar to cover, bryophyte biomass increased with soil moisture, however, year and treatment also influenced results ( figure 3, table 3). Bryophyte biomass increased as sites increased with soil moisture in 2010, but not in 2002 (year*LCT interaction), and it was higher in exclosures than in controls in 2010, but not in 2002 (year*EX/CT interaction). Similar to results for cover, deciduous shrub biomass was higher at drier sites, no shrubs occurred in wet sites, and no significant effect of treatment was observed. Forb biomass, although never large, was generally greatest inside exclosures and in moist tundra (figure 3). Graminoid biomass increased as soil moisture increased and was lower inside exclosures than control plots in all land cover types. Lichen biomass showed opposing trends to those for graminoids (figure 3).  Litter biomass did not differ with land cover type or between exclosures and controls in either year. Although Batzli (1975) sampled 20 sets of exclosures in 1974 (4 wet, 9 mesic and 7 dry) and we only sampled a subset of those (4 in each land cover type) due to many exclosures falling into disrepair, qualitative comparisons of biomass can be made for two functional components of the vegetation with the same sampling and classification procedures. At dry sites, lichen in control plots averaged ca. 40 g m −2 in 1974, 50 g m −2 in 2002, and 30 g m −2 in 2010, whereas exclosure plots averaged ca. 80 g m −2 in 1974, 260 g m −2 in 2002, and 150 g m −2 in 2010. Thus, while control plots showed no trend over the years, lichen biomass increased two-fold in exclosure plots. In moist plots, lichen biomass increased during the three sampling years in both controls and exclosures, but was greater within exclosures (30, 40 and 75 g m −2 , respectively for controls, and 40, 80 and 150 g m −2 for exclosures, respectively). Less consistent changes in lichen biomass occurred at wet sites (5, 30 and 0 g m −2 , respectively for controls, and 10, 15 and 25 g m −2 , respectively for exclosures). Graminoids showed the opposite result with greater and more consistent increases in controls than in exclosures, except at dry sites (120, 175 and 160 g m −2 , respectively, for controls and 100, 135 and 70 g m −2 , respectively, for exclosures at wet sites; 40, 110 and 90 g m −2 , respectively, for controls and 25, 40 and 40 for exclosures at moist sites; 20, 40 and 40 g m −2 , respectively for controls and 10, 20 and 35 g m −2 , respectively, for exclosures at dry sites). No comparisons could be made for bryophytes because the two studies used different methods to determine bryophyte biomass.

Discussion and conclusion
The 50+ year exclusion of lemmings from the coastal tundra near Barrow resulted in dramatic changes to plant community structure, but the specific effects varied with vegetation type. Results are largely consistent with prior studies that suggest herbivores can have a positive effect on graminoids (e.g. van der Wal and Brooker 2004); although for an exception see (Virtanen 2000) and may control transitions between moss and lichen dominated tundra and graminoid tundra states (e.g. Zimov et al 1995). As expected, animal signs only occurred in the control plots, verifying the sustained effectiveness of the exclosures. Sampling methods also appeared to be coherent, indicated by the generally close agreement between data for cover and biomass. Although cover data showed some interactions between land cover type and treatment that biomass data did not, the exclusion of lemmings resulted in substantial changes to the plant communities investigated. In dry tundra exclosures, a lichen dominated community has developed, similar to that seen with the exclusion of other mammalian herbivores (e.g. van der Wal 2006, Gough et al 2008. Vegetation changes in wet tundra exclosures resulted in a moss-dominated community with higher amounts of standing litter and a lower abundance of graminoids, an effect similar to that reported by van der Wal and Brooker (2004).
Exclusion in moist tundra appeared to produce an intermediate response with increases in lichens, forb and bryophyte biomass, and a decrease in the abundance of graminoids. Despite these shifts in functional types, exclosures had little effect on species diversity, a result that largely masks shifts in the species pool away from a diverse vascular flora toward a lichen and bryophyte dominated flora within the exclosures. However, there was a significant land cover type by treatment interaction for the Shannon-Wiener Index of Diversity. Exclusion of lemmings resulted in higher Shannon-Wiener values in dry, generally low productive, plots and lower values in wet, generally higher productive, plots is consistent with the grazer-reversal hypothesis (sensu Proulx and Mazumder (1998)). The increasing difference between controls and exclosures from 1974 to 2002/2010 indicated a continuing impact from the exclosure treatments almost 50 years after their construction. Differences in the distribution of biomass among functional plant groups in 2002 and 2010 suggests that changes in vegetation in response to removal of grazers are still occurring, but whether this reflects continuing development of treatment effects after 50 years or interannual variability remains unknown.
In work previously published on these exclosures and control plots, Batzli (1975) reported more graminoids in the moist and dry tundra control plots relative to exclosures, more litter in moist site exclosures, and more lichen in dry site exclosures. We found more graminoids in control plots than exclosures for all land cover types, suggesting that brown lemmings facilitate the production of their preferred forage, which includes several common species of graminoids in such as Carex spp., Eriophorum spp. and Dupontia fisherii (Batzli and Pitelka 1983). Herbivores usually influence vegetation through direct removal of biomass of palatable species (Diaz et al 2007), but many graminoid species with evolutionary histories closely tied to those of herbivores have compensatory growth strategies (Ferraro 2002). Although this seems to be rare in Arctic graminoid species (e.g. Elliot and Henry (2011)), tolerance (Bråthen and Odasz-Albrigtsen 2000) may be more common among graminoids at Barrow (Chapin et al 1980). While intensive grazing by lemmings may reduce production over short time scales of a few years (Thompson 1955b), sustained episodic herbivory appears to increase graminoid abundance over much longer time scales (Batzli 1975). Schultz (1969) noted that production and decomposition inside exclosures steadily declined with the age of the herbivory exclosures studied, whereas organic matter decomposed faster in the presence of lemmings outside the exclosures. As a result of these and other studies, Batzli (1975) concluded that extensive clipping of monocot shoots by episodic lemming outbreaks increased rates of nutrient cycling, and that this accounted for a greater production of graminoid biomass over long time periods. Batzli also noted that lemmings disrupt moss and lichen vegetation, particularly when they grub for rhizomes when population densities are high, which could account for the low abundance of moss and lichen cover documented for control plots in this study. Lemmings also eat large amounts of moss, particularly in winter (Batzli and Pitelka 1983), which may partially explain the decline of mosses in control plots.
The distinction between short-term and long-term effects highlights the importance of addressing the timing of vegetation sampling in relation to episodic changes in lemming population densities. Thompson (1955aThompson ( , 1955b and Schultz (1964) showed that lemmings reduce the production of graminoids during growing seasons with high lemming population densities, but that graminoids largely recover during the following growing season. As a result, vegetation in control plots reflects the phase of the lemming cycle. Thus, sampling for long-term effects, as in this study, should occur at least one and preferably two seasons after the year of a lemming high. Our sampling occurred in 2002 and 2010, both during the second growing season following lemming population highs in 2000 (Holt 2010) and 2008(Villarreal et al 2011. Batzli's (1975) sampling occurred in 1974, the third year after a lemming high in 1971. Thus, this and Batzli's 1974(Batzli 1975) study do not reflect a temporary depression of biomass associated with high lemming populations, but rather consistent long-term patterns.
A review of the effects of experimental warming on tundra plants concluded that higher temperatures leads to increased cover of deciduous shrubs and graminoids and to decreased cover of mosses and lichens (Walker et al 2006). Our results indicate that the activity of lemmings near Barrow facilitates three of those four effects, the exception being shrubs, which brown lemmings largely ignore. A considerable challenge remains in understanding the spatial and temporal extent of lemming effects on vegetation. Although brown lemmings appear to reach very high densities only within a 100 km or so of Barrow, other herbivorous arvicoline rodents become more abundant at other sites where up to five sympatric species can be found (Pitelka and Batzli 1993). Several species of arvicoline consume deciduous shrubs, and in winter they may even eat bark and girdle shrubs Jung 1980, Batzli andHenttonen 1990). Of course, consumption and girdling of deciduous shrubs would reduce, rather than increase the prevalence of shrubs, an effect already reported for large herbivores (Post andPedersen 2008, Olofsson et al 2009). In any event, a significant impact of arvicoline rodents probably occurs in a variety of habitats and geographic areas. Because functional attributes of ecosystems appear to be strongly affected by the timing and duration of the lemming cycles near Barrow (Lara et al 2011), we urge greater attention to sampling of ecosystem properties and processes with respect to arvicoline rodent populations in general.
In conclusion, our resampling of the historic herbivore exclosures in the coastal tundra near Barrow, Alaska revealed that lemming exclusion decreases the cover and biomass of graminoids markedly and increases the biomass of lichens and bryophytes, particularly in wet tundra.
Because these plants respond similarly to warming, lemmings may have partially contributed to the recent greening of Arctic landscapes. Exploring these relationships in more depth over a range of spatio-temporal scales, including more mechanistic experiments detailing the responses of plants to arvicoline rodents in other tundra communities, examining the response of tundra to warming with and without herbivores, and assessing response of herbivores to climate change is much needed.