Disentangling the effect of multiple ecological processes on plant trait composition is complicated by the fact that both top-down and bottom-up processes may affect similar traits. We examined the interacting role of resource variation and vole herbivory on functional trait patterns in an annual California grassland. We manipulated vole herbivory via exclosures at eight grassland sites along a steep resource gradient and measured plant composition and functional traits over three years. Plants with resource acquisitive functional traits were favored at sites with increasing resources. Vole herbivory influenced community-weighted mean (CWM) leaf nitrogen (N) and seed mass, suggesting these traits may mediate plant susceptibility to vole herbivory. After three years, CWM leaf N increased in the absence of the voles, as did CWM seed mass, although this increase in CWM seed mass only occurred at higher resource sites. Vole exclusion at high resources sites also increased the functional diversity of leaf N and seed mass by the end of the experiment. Overall, environmental filtering primarily structured the dominant plant trait strategies, but vole herbivory also influenced the functional diversity of traits that influence herbivore susceptibility, particularly at resource rich sites. Thus, habitat filtering and herbivory can operate on different dimensions of plant functional composition to influence the species and functional composition of communities.

Study Site. We conducted our experiment in the annual grasslands at the University of California Sierra Foothill Research Extension Center (SFREC), located in Browns Valley, California, USA (39º 15' N, 121º 17' W). These communities are dominated by non-native annuals with low abundances of native species. The most abundant species included Avena barbata, Elymus caput-medusae, Festuca perennis, Bromus hordeaceus, and Erodium botrys (nomenclature follows Baldwin et al. 2012). The climate is Mediterranean with cool wet growing seasons (Sept-May) and hot dry summers (June-Aug).  Growing season rainfall increased over the course of our study more than doubled from 454mm to 625mm to 978mm for the 2014-2015, 2015-2016, and 2016-2017 growing seasons, respectively. The dominant small mammal consumer is Microtus californicus, a small meadow vole that prefers habitat with a standing litter layer (Batzli & Pitelka, 1970; Ostfeld et al., 1985).  Other small mammal species at our study site predominantly reside in the more heavily cattle grazed pastures (Block & Morrison, 1990). Therefore, our design below reflects predominantly the effects of voles.

Sites. We conducted our experiment at eight experimental sites that spanned a steep resource gradient.  We described this gradient (hereafter resource gradient) using a principal component analysis (PC1 which described 47.6% of the variation in environmental variables; Supporting Information Appendix S1, Figure S1). At one end of the gradient were low productivity sites (as characterized by aboveground primary productivity) with low soil nitrogen (N), but high micronutrients (i.e., magnesium, sodium, and calcium). At the other end of the gradient, sites had high productivity and soil N.  

Experimental Design. To evaluate the effects of voles on plant trait composition, in the summer of 2014, we initiated a field experiment where we manipulated the presence of voles. At each site, we established a set of paired plots (9x9m), one fenced to exclude small mammals and one unfenced control to allow small mammal access. The exclosures were constructed of 0.64cm welded wire dug 60cm into the ground around the perimeter of the plot. The fence extended about 90cm aboveground and was topped with galvanized sheet metal with a 20cm face to prevent voles from climbing over the fence. Within each of the paired plots, we randomly established a set of six 0.5m x 0.5m subplots to assess the effects of voles on plant trait composition. We additionally set up an electric fence around the 14m x25m experimental area at each site to exclude cattle.

Plant traits. To examine the distribution of functional traits in communities we sampled 5-10 individuals of the dominant and subdominant resident species  across the  sites for plant functional traits (following Pérez-Harguindeguy et al. 2013). The samples were collected from within the non-cattle grazed experimental area but not from within any experimental plots. Samples were collected at peak biomass from April-May in 2015 and 2016. We measured maximum plant vegetative height, specific leaf area (SLA; leaf area/dry leaf mass), leaf water content (LWC; 1- leaf dry weight/leaf fresh weight) and seed mass.  The sampled leaves were additionally processed for tissue carbon and nitrogen content to estimate leaf N content and carbon to nitrogen ratios. These traits are strong indicators of resource use and plant growth. SLA is positively correlated with a species relative growth rate and tissue N (P B Reich, Walters, & Ellsworth, 1997; Westoby, Falster, Moles, Vesk, & Wright, 2002); leaf water content is negatively correlated to water stress (Farooq, Wahid, Kobayashi, Fujita, & Basra, 2009); plant height is often indicative of competitive interactions for light (Westoby, 1998); and greater leaf N and lower carbon to nitrogen ratios can be linked to higher food quality (Westoby, 1999). To account for potential trait differences in species that occurred across the environmental gradient, we sampled individuals for as many species as possible at both the low and high end of the gradient. To estimate a species’ seed mass, we first took 10 samples with the same number of seeds (i.e. the number of seeds was either 50 or 75 seeds for a species, depending on seed availability). For each sample, we calculated the average seed weight by dividing the total weight of the sample by the number of seeds. The species level average was then the average of those 10 estimates. (All species trait data are available in Table S1.) In total, we sampled traits on 54 different grassland species (24 of these in both habitat types), which made up on average 98% of the species composition in a given plot (mean, range of species cover: 2015: 99.25 (89-100), 2016: 98.9 (93.5-100), 2017: 98.75 (84-100).

Community sampling. From 2015-2017, at peak biomass (Apr-May) we sampled the plant species composition within each subplot. To estimate vole activity, in 2016 and 2017, we recorded the frequency of vole activity (i.e. runways, burrows, and droppings), along 8 10m long transects within each larger control plot at each site. For each subplot, we then calculated the community weighted mean (CWM) for each individual plant trait and the functional dispersion of each trait. CWM is measured as the mean of species trait values present in the community, weighted by the relative abundance of each species (Lavorel et al., 2008). Functional richness (FRic) estimates the dispersion of species in trait spaces without accounting for species abundance and is estimated as the convex hull volume (Villeger, Mason, & Mouillot, 2008). Functional dispersion (FDis) is the average distance to the centroid in multivariate trait space that is weighted by species relative abundances (Laliberte & Legendre, 2010). For those species that occurred across the environmental gradient we used the species level trait data for a plot that best matched its position along the environmental gradient for these calculations.

There are four files for this data set: 1) Site environmental data, 2) Plot community species cover data 3) Plant species trait data and 4) Vole runway activity data.

Site Environmental Data: Three soils samples up to 10cm in depth were collected from each large 9mx9m plot (n=16) and processed for soil resources. These data are reported here. Additionally, the mean live aboveground biomass and mean litter from six 10cmx50cm subplots are reported. These data were included in a principal components analysis that generated two PCA axis. The first, PCA1 was used as the resource gradient described in the manuscript.

Plot Community Species Cover Data: With each block, 12 0.5m 0.5m plots were sample in each of three years (2015, 2016, 2017). For this sampling the visual cover of each observed species was recorded, with cover adding over 100 to allow for multiple canopy layers. These data are the cover estimates for each plant species in each plot over time. The Site.Prod column indicates the how these data were stratified to match up with species trait that were collected at the low or high end of the resource gradient.

Plant species trait data: These data are species-level trait data for species observed at the field site. Specific Leaf Area (SLA), leaf water content (LWC), leaf area (Area.cm2), leaf dry matter content (LDMC), maximum vegetative height (Height.max.cm) were sampled on 5-10 different individuals in the field. These plant leaves were used for leaf nitrogen (Leaf.N), leaf carbon (Leaf.C), and carbon to nitrogen ratios (C.N.Ratio). Seed mass was extracted from the Kew database, Baker seed collection at the Jepson Herbarium, site collected samples, or purchased seed.

Vole Runway data: These data are the average runways observed in 8 transects sampled in the presence of voles (i.e. open plots).