Climate and vegetation structure shape ant communities along elevational gradients on the Colorado Plateau

Abstract Terrestrial animal communities are largely shaped by vegetation and climate. With climate also shaping vegetation, can we attribute animal patterns solely to climate? Our study observes ant community changes along climatic gradients (i.e., elevational gradients) within different habitat types (i.e., open and forest) on the Colorado Plateau in the southwestern United States. We sampled ants and vegetation along two elevational gradients spanning 1,132 m with average annual temperature and precipitation differences of 5.7°C and 645mm, respectively. We used regression analyses and structural equation modeling to compare the explanatory powers and effect sizes of climate and vegetation variables on ants. Climate variables had the strongest correlations and the largest effect sizes on ant communities, while vegetation composition, richness, and primary productivity had relatively small effects. Precipitation was the strongest predictor for most ant community metrics. Ant richness and abundance had a negative relationship with precipitation in forested habitats, and positive in open habitats. Our results show strong direct climate effects on ants with little or no effects of vegetation composition or primary productivity, but contrasting patterns between vegetation type (i.e., forested vs. open) with precipitation. This indicates vegetation structure can modulate climate responses of ant communities. Our study demonstrates climate‐animal relationships may vary among vegetation types which can impact both findings from elevational studies and how communities will react to changes in climate.

in the southwestern United States. We sampled ants and vegetation along two elevational gradients spanning 1,132 m with average annual temperature and precipitation differences of 5.7°C and 645mm, respectively. We used regression analyses and can impact both findings from elevational studies and how communities will react to changes in climate.

K E Y W O R D S
abundance, arthropods, communities, precipitation, primary productivity, richness, structural equation model, temperature climate effects occur directly through physiological effects or indirectly through changes to plant communities and food resources.
However, the relative effects of climate and plants in shaping higher trophic communities remain vague, especially when considering biogeographical distributions across climatic gradients (Hortal et al., 2012;Wisz et al., 2013).
Climate affects organisms largely through precipitation and temperature (Pörtner & Farrell, 2008). Temperature predictably declines an average of 6°C with every 1,000 m gained in elevation (Barry, 1992), and elevational gradients are commonly used to study climate effects on terrestrial communities (Rahbek, 1995).
As temperature decreases, so do metabolic processes which may decrease niche space and speciation rates (i.e., metabolic theory, Brown, Gillooly, Allen, Savage, & West, 2004). This produces monotonic declines in richness with increased elevation for most taxa (Rahbek, 1995). Unlike temperature, precipitation may increase or decrease with elevation depending on local climate. Precipitation, especially in arid ecosystems where it is a limited resource, acts as a trophic currency influencing species patterns and interactions (Allen, McCluney, Elser, & Sabo, 2014;McCluney et al., 2012).
Studies in arid regions where precipitation increases with elevation observe variable richness patterns (reviewed by Szewczyk & McCain, 2016), even increases in richness with elevation (Sanders, Moss, & Wagner, 2003). Incorporating precipitation with temperature refines elevational predictions (i.e., the elevational climate model hypothesis, McCain, 2007), and suggests that optimal climate ranges exist on elevational gradients where high precipitation and temperatures coincide. The location along a given elevational gradient where this optimal climate occurs is likely to have the highest productivity and diversity. Most elevational studies on animal communities and climate attempt to frame results within one of the aforementioned hypotheses (Szewczyk & McCain, 2016). However, the above hypotheses fail to separate vegetation and climate effects on higher trophic levels.
Our study's biophysical setting is the Colorado Plateau, a large arid region of the southwestern United States interspersed with mountains that form significant elevational gradients. This region contains life zones (i.e., biological communities found in similar elevations/latitudes) representative of most of western North America (Merriam, 1898) within a localized species pool. Several of these life zones include both forest and open habitats, which differ enormously in vegetation structure, composition, and productivity. This provides an opportunity to compare vegetation effects on animal communities within the same climate zone. We focus on ant (Hymenoptera: Formicidae) communities, which are dominant in most terrestrial food webs (Hölldobler & Wilson, 1990). Ants are commonly used as indicators of general ecological responses to disturbance (Andersen & Majer, 2004) and may be indicative of environmental stress (Tiede et al., 2017).
Elevational gradients patterns of ant richness are well-documented: richness usually declines monotonically with elevation, but can peak at mid-elevations (reviewed by Szewczyk & McCain, 2016).
Ants are either directly or indirectly dependent on vegetation for food and often for nesting space (Hölldobler & Wilson, 1990).
Vegetation structure controls levels of insolation creating microclimates, modulating environmental stress (Bolger, Kenny, & Arroyo, 2013). This is best demonstrated between open and forested habitats, where ant communities vary markedly (Lassau & Hochuli, 2004). Yet many elevational studies designed to assess climate effects are carried out in different vegetation types (e.g., Nakamura et al., 2016). With increased elevation, ants nest in more insolated locations (Plowman et al., 2020) and along elevational gradients ant community composition can closely follow dominant vegetation (e.g., Andersen, 1997). Lasmar et al. (2020)  Here, we investigate the relative effects of climate and vegetation on ant communities across two elevational gradients. We asked:

| Study sites and design
To examine variation in richness, abundance, and composition of ant communities across habitat types, we selected 12 sites across two elevational gradients located on the Colorado Plateau with considerable climate differences ( Figure 1, Table S1). Both gradients encompass the following life zones spanning 1,556-2,688 m above sea level: cool desert, pinyon/juniper, ponderosa pine, and mixed conifer. From the lowest to highest elevation sites, average annual temperature decreases from 13.6 to 6.7°C and average precipita-  Figure S1).

| Ant community
A pitfall trap (Thomas & Sleeper, 1977) was dug in each subplot, a reliable method for sampling different ground-dwelling arthropod communities with equal sampling intensity (Andersen, Hoffmann, Müller, & Griffiths, 2002). Pit traps consisted of a long borosilicate glass tube measuring 32 mm diameter and 200 mm length filled with 100 mm of propylene glycol fitted within PVC sleeves with a rain cover (Higgins, Cobb, Sommer, Delph, & Brantley, 2014). Ant communities were sampled for 7-day periods during the dry season (9-16 June 2015) on 89 subplots and during the rainy season (6-13 August 2015) on the same subplots, minus 14 which were flooded (Table S1). Pit trap samples were sorted, and voucher specimens of each species and site occurrence were identified with specialist help and deposited at the Colorado Museum of Arthropod Biodiversity at Northern Arizona University (Table S2), yielding richness and abundance measures for each trap. Species accumulation curves show most sites approach an asymptote in for sampling ( Figure S2). Vegetation was identified to species for richness and composition.

| Vegetation, weather, and climate measurements
To estimate productivity, we used normalized difference vegetation index (NDVI, Pettorelli et al., 2005) calculated from satellite imagery at the 250 m 2 resolution for each site and sampling date (Dimiceli et al., 2015). Weather was measured either by on-site weather stations or temperature data loggers at each plot. Climatic factors were established for each site using 30-year averages of annual precipitation and F I G U R E 1 Location and climate/biome descriptions of elevational study sites in northern Arizona, on the Colorado Plateau region in the southwestern United States. Sites are coded by shape and life zone by shape fills G r a n d C a n y o n N a ti o n a l P a r k  May 2015). PRISM data were downloaded at a spatial resolution of 800 meters and extracted using R version 3.2.3 (2015-12-10) © 2015 The R Foundation for Statistical Computing Platform: x86_64-apple-darwin13.4.0 (64-bit) and the package "raster" version 2.5-2 based on observed latitude and longitude of each site.

| Structural Equation Modeling (SEM)
To test the relative direct and indirect effects of climate and vegetation on the ant community, structural equation modeling (SEM, Grace, 2006) was used. An a priori model of our system was first created based on the simple assumptions that vegetation influences ants, and climate influences both vegetation and ants. A measurement model incorporating climate and vegetation data was developed and suggested by literature (Figure 2). Vegetation composition was represented as two separate variables to include both axes of the NDMS ordination. A separate model was tested for each ant community measure: richness, abundance, composition one (NMDS axis one), and composition two (NMDS axis two). To achieve necessary sample sizes for SEM, analysis was conducted with plot-level data instead of site averages. This approach results in pseudo-replicates for average annual temperature and precipitation, and NDVI (Schank & Koehnle, 2009) with Joreskog's goodness-of-fit (GIF) and X 2 tests. Contrary to most tests, a high p value indicates a good probability that a model fits the data and is thus desired. GIF > 0.95 are considered a good fit (Grace, 2006). Effect sizes are estimated with standardized path coefficients, analogous to weighted regressions, which show effect direction (positive or negative) and effect size (the further the value is from 0, the stronger the effect).

| RE SULTS
A total of 5,300 ants were collected from the two sampling periods in 2015, with 36 identifiable taxa [34 species including one morpho-species, two species complexes, and one genus with multiple unidentified species (Myrmica)]. One species (Monomorium cyaneum) and one species group (Solenopsis fugax group) were ubiquitous (Table S4). Richness and abundance were both highest at mid-elevation sites ( Figure S3). In general, climate variables and vegetation composition were strong predictors for ant community metrics, while vegetation richness, NDVI, and weather (real-time measures of temperature and precipitation) variables were not (Tables 1 and   2). All ant community metrics showed relationships with annual average precipitation. Ant richness and abundance were associated with vegetation composition, while ant composition was closely tied to annual average temperature (Tables 1 and 2). One part of ant composition (NMDS axis one) responded to precipitation, while another (NMDS axis two) responded to temperature (Figure 4), each axis was associated with a different suite of ant species (Table   S4) Table S5). Indicator species were found in each life zone and habitat combination (Table S5) (Table 1).  temperature was the best predictor for ant NMDS axis two (0.64).

| Climate has strong direct effects on ants relative to vegetation
Several moderate effects on ants were also evident: temperature on ant richness (0.21) and abundance (0.22) and primary productivity on ant NMDS axis one (0.22).

| D ISCUSS I ON
Climate had strong direct effects on ants with little or no effects of vegetation composition or primary productivity, but vegetation type (i.e., forested vs. open) had contrasting relationships with precipitation. This indicates the influence of vegetation on ant communities along our elevational gradients is primarily through vegetative structure rather than plant composition or productivity. Our approach is correlative, and caution must be used when assigning causality.
However, within our life zones and habitats, the model fit well, with strong explanatory variables for most ant community metrics. Our results are therefore indicative of how large-scale climate and vegetation mechanisms shape animal communities along elevational gradients.
Hypothesized mechanisms shaping elevational distributions of animals are largely climate and geography based, mainly considering vegetation a by-product (Szewczyk & McCain, 2016). Comparisons of vegetation and climate effects across elevations or latitudes usually find that climate is the strongest driver (Donoso, Johnston, & Kaspari, 2010;Sanders, Lessard, Fitzpatrick, & Dunn, 2007). Yet vegetation structure can modulate climate-animal relationships (Carneiro, Mielke, Casagrande, & Fiedler, 2014;Lasmar et al., 2020), and similar patterns are well known for climate-plant relationships (Michalet, Schöb, Lortie, Brooker, & Callaway, 2014).  (Table S5). Significant correlations of average annual air temperature (R 2 = −.0.607), average annual precipitation (R 2 = .643), and elevation are shown as solid lines. Each axis is associated with a different set of species (Table  S4) Temperature  (Michalet et al., 2014). This same climate stresses likely shape animal-communities and seem to be behind our observed patterns.
Links between climate and ants are clear (Szewczyk & McCain, 2016), but effects of vegetative influences on ant communities are less so. Assumingly then, the interactive effects of climate and vegetation are even less clear. Ants rely on vegetation for resources such as food and shelter and in some instances ant species can be dependent on particular species of vegetation (Hölldobler & Wilson, 1990). Vegetation composition, richness, and productivity had small effects on ants, while vegetation structure (i.e., open vs. forested structure), which modulates climate, was the most influential aspect of vegetation. By offering ants more nesting opportunities and retaining more soil moisture than adjacent open habitats (Michalet et al., 2014), forests may buffer the effects of low precipitation on ants. Ants are adapted to certain nesting conditions and nests in open habitats must largely be constructed in soil (Lassau & Hochuli, 2004). Soil nesting success is highly contingent on soil moisture (Johnson, 1998) ours and most other temperate/arid gradients forests that exist at higher elevations and become too cold for most dominate ant species (Andersen, 1997).

CO N FLI C T O F I NTE R E S T
None to declare.