Mammalian herbivores restrict the altitudinal range limits of three alpine grass species (transplant and herbivore exclusion experiment and demographic data from natural populations), West Elk Mountains, Colorado, USA 2015-2018

Study species and sites We focused on three native, caespitose (non-clonal and only reproduce by seed), alpine-restricted grass species: Poa alpina, Festuca brachyphylla, and Elymus scribneri (Shaw 2008). More detail on the focal species life history strategies, importance, and information on herbivores can be found in the Supporting Information Methods. We established experiments in the West Elk Mountains, Colorado, USA, in which temperature, atmospheric pressure, and plant available N and P decline with elevation (Kittel et al. 2002; Dunne et al. 2003). The regional lapse rate is about 1degreeC decrease in temperature with about 140 m increase in elevation (Pepin and Losleben 2002). We used previous vegetation surveys (Lynn et al. 2019) to locate experimental habitats in the range core (core), at the low range limit (limit), or below the range limit (novel; about 230 m a.s.l. below limit sites). Novel habitats reflected about 2degreeC greater mean annual temperature than the range edge of each plant species, consistent with predicted climate warming for the region by 2050 (Pepin and Losleben 2002; Rangwala and Miller 2012). The three habitat types were replicated on three mountains for a total of nine sites (site coordinates and map in Table S1 and Fig. S1).

Experimental design Experimental implementation. Plants were reared in a greenhouse prior to transplantation in the field (details in Supporting Information Methods). We used a 2x2 factorial design to allow/exclude aboveground ungulates and allow/exclude belowground gophers using fencing, and replicated this design at core, limit, and novel habitats. Gophers were excluded by inserting wire mesh about 20 cm into the ground around a plot (Fig. S2). Ungulates were excluded using 40 cm x 40 cm fences of 20-gauge chicken wire that was 30 cm tall (mean plant height across species was 8.8 cm; Fig. S2). At each site, we marked 20 plots (30 cm x 30 cm) and randomly planted one individual of each species per plot into an equilateral triangle at 15 cm spacing between plants. All vegetation was removed and plots trenched prior to planting and weeded monthly during the growing season until harvest to control for effects of competitors in transplant sites and isolate the influence of herbivory. Herbivory treatments (no exclosure controls, above ground exclosure, below ground exclosures, and both) were randomly assigned to plots within a site and replicated in five plots per site. Experimental set-up occurred during 22 June - 20 July 2016. Experimental data collection. Starting in August 2016, we took monthly measurements of plant size and foliar damage by ungulate and insect herbivores (see background insect herbivory in natural populations Fig. S3). Plant size was determined by counting the number of vegetative tillers. We counted the number of inflorescences to estimate reproduction but collected them prior to pollen dispersal. We took visual percentage estimates of the amount of damage present on each plant. Mammal damage was assessed as even clipping by grazing animals with whole leaf/tiller damage traced back to the crown. Insect herbivory (e.g., grasshopper) was assessed as stippling and partial leaf chewing damage. We harvested plants from 14 - 27 August 2018. Prior to harvest, all plants were assessed for size and damage as described above. Plant aboveground biomass was cut at the meristem just below the soil surface and immediately separated into litter and live biomass in the lab. All biomass was dried for 48 h at 60degreeC prior to weighing to the nearest 0.001 g on a mass balance (Mettler-Toledo MS104S and PL303, Columbus, OH, USA).

Demographic data from natural populations In 2015, we set up five 1mx5m permanent plots that contained natural populations of the three focal species at one core site given the time demands of collecting demographic data (Avery core; Table S1). We marked every individual with metal tags and measured height (nearest mm), tiller number, and inflorescence number in August 2015 and repeated annual censuses through August 2018. Survival was determined for marked individuals in each year. New recruits within the plots were identified and marked each year. In 2015, we sampled an additional 30 individuals of each focal species outside the permanent plots to estimate average seed production per inflorescence for each species without affecting recruitment potential.

From supplemental methods:

Focal plant species and herbivores We focused on three native, caespitose (non-rhizome or stolon forming and reproduce by seed), alpine-restricted grass species: Poa alpina, Festuca brachyphylla, and Elymus scribneri (Shaw 2008). Festuca brachyphylla (Frederiksen 1981) and P. alpina (Pierce 2003) can be viviparous or pseudo-viviparous (produce plantlets or bulbils instead of seeds), but this behavior was not observed in over six years of observations in our populations. Grasses have been underrepresented in past studies of how biotic interactions influence geographic ranges (e.g., Alexander et al. 2015; Louthan et al. 2018) despite their diversity (Barker et al. 2001) and ecological role across the globe (Shantz 1954). Additionally, the focal species were dominant and occurred across replicated mountain gradients, and many forbs in the system would not meet this condition. Our system contained diverse native ungulate herbivores: mountain goat (Oreamnos americanus) and sheep (Ovis canadensis canadensis) tend to graze in alpine plant communities, while moose (Alces alces shirasi), elk (Cervus elaphus nelsoni), and deer (Odocoileus hemionus hemionus) forage in montane forests and meadows (Armstrong et al. 2011). Additionally, cattle move up-valley seasonally in late August to graze. The northern pocket gopher (Thomomys talpoides) disturbance peaks at about 3200 m a.s.l. in the region (Lynn et al. 2018), which is around the minimum elevation of occurrence for the focal species (about 3400 m a.s.l. mean across focal species; Lynn et al. 2019).

Greenhouse rearing We used greenhouse reared individuals which allowed us to control for age, history (including previous herbivory), and abiotic conditions (e.g., soil nutrients and water availability) prior to implementing experimental treatments. However, there is evidence that germination conditions can have long-term consequences for the traits of an individual (Donohue et al. 2010) that, in our experiment, may alter experimental individual traits relative to natural individuals. Focal species were grown from seed in the greenhouse at about 20C for about 12 weeks at the University of New Mexico. Seeds were collected in 2015 from each core site population. The three species' range core and limits matched well (mean occurrence range across species about 3700 m a.s.l.; Lynn et al. 2019), justifying site selection. In February 2016, we germinated seeds in flats using Metro-Mix 360 potting soil (sun gro Horticulture, Agawam, MA, USA). Germination rates of the species and populations were high (about 80% and above) consistent with past greenhouse work using these species germinated on sand, which is generally less favorable for germination (Lynn et al. 2019a). Once individuals were 2-6 leaves in size, we transferred them to about 150 ml root trainer pots (Stuewe and Sons, Inc., Tangen, Oregon, USA), in Metro-Mix 360, where they grew until planted into the field experiment. Each pot was top fertilized with about 15 ml of Osmocote Plus 15:9:12 N:P:K pellets (Scotts Miracle-Gro Company, Marysville, OH, USA) and watered three times daily with overhead sprayers (about 50 mm per pot). Plants were split into three equally sized clones to be transplanted one per habitat per population. Plants were transported to the Rocky Mountain Biological Laboratory in May 2016 and acclimated to the colder climate in their pots for about 20 d within a fenced area.