Cascading effects of grazing on predatory arthropod and parasitoid densities

An important goal in arthropod conservation is to understand how arthropods are affected by anthropogenic activities. Livestock graze 29% of the land area in the U.S., which can result in both top-down and bottom-up effects on ecosystems that are grazed. Grazing often reduces species richness and abundance of arthropods; however, these trends depend upon context and taxa, making generalization difficult. Grazing can either increase or decrease species richness and abundance of arthropods, depending on context and taxa, making generalization difficult. The impacts of grazing on different taxa may also be indirect, depending upon trophic interactions with other members of the community. We propose that considering trophic relationships will help clarify the effects of grazing on arthropods. Here, we study the effects of grazing by ungulates on pompilid wasps (Hemipepsis and Pepsis hawk wasps) and tarantulas (Aphonopelma hentzi). We do this by comparing the cover of forbs (flowering plants that provide nectar to hawk wasps), and the densities of hawk wasps and of tarantula burrows in areas grazed by cattle to areas with light grazing by wild ungulates in the shortgrass prairie. Grazed areas had lower cover of flowering plants, fewer parasitoids, and more tarantulas, while lightly grazed areas had higher cover of flowering plants, more parasitoids, and fewer tarantulas. We propose that hawk wasp abundance may track floral resources, enabling parasitoids to exert strong top-down pressure on tarantulas in lightly grazed areas. Thus, grazing may benefit tarantulas by reducing the abundance of their parasitoids.

Methods
We sampled forb cover, wasp densities, and tarantula densities at five shortgrass prairie sites in southeastern Colorado, USA (Cook & Redente, 1993, Figure 1a), which has a semi-arid, continental climate with hot summers and cold winters. Hardy, drought-resistant grasses dominate most of the region, which is commonly used for cattle grazing. For example, in 2019 was recorded as supporting 452,400 cattle on 27 million acres (Meyer, 2019).  

Our sites were located on Southern Plains Land Trust property, which manages the grazing levels there. Domestic cattle grazed three of the sites at moderate to heavy levels (one head per 30 acres or 0.082 head/hectare, year-round). The remaining two lightly grazed sites were not stocked with cattle and only grazed by deer (Odocoileus spp.), pronghorn (Antilocapra americana), occasional elk (Cervus canadensis), and, rarely, escaped cattle. Wild cervids occurred in grazed sites as well. Because most of the region is grazed, only two lightly grazed areas were available. Grazed and lightly grazed sites were similar with respect to topography (i.e., slopes and elevations) and climate (i.e., temperature and precipitation) (Supplemental). 

We visually estimated the percentage of forb cover using a wooden quadrat with inner dimensions of 30 cm x 30 cm. We collected two types of samples: at tarantula burrows and away from burrows. For burrow samples, we placed the quadrat over the burrow with the entrance in the center of the frame. Estimates of forb cover away from burrows were taken every 200 m along the transect while conducting Distance sampling.  

We analyzed forb cover using the lmer and emmean packages in R-studio to perform a linear mixed effect model and to calculate the estimated marginal means, respectively. We evaluated percent forb cover using grazing level as a fixed effect and site as a random effect as follows: 

Percentage forb coverGrazing Level∗Observation Type+(1|Site)Percentage forb coverGrazing Level∗Observation Type+1Site

We considered significance P < 0.05 and ensured that our variables met all the assumptions of a linear mixed effect model using a Residuals vs Fitted plot and QQ-plot. 

We performed a statistical analysis of hawk wasp and tarantula densities using the interactive program DISTANCE (Royale et al., 1982). We considered differences to be significant if their 95% confidence levels do not overlap (Thompson et al., 1994). 

We conducted Distance sampling surveys during the active season for Hemipepsis and Pepsis wasps, between June and August in 2022. We sampled all five sites each month by walking 2000 m transects at each site and recording the perpendicular distances to all observations of hawk wasps. Due to small sample sizes, we combined observations from June, July, and August, and used those data to estimate the density of wasps at each site. 

To estimate the density of tarantulas, we surveyed each study site once between June and August of 2021, estimating the densities of burrows used by tarantulas using Distance Sampling (Royale et al., 1982). Distance sampling followed Royale et al. (1982), Thompson et al. (1994), and Miller (2016) and operates on the concept that the further away from a transect an animal is, the more difficult it is to detect. At each site, we walked transects until we identified 60 tarantula burrows. The openings of tarantula burrows are easily distinguished from other small holes by their circular shape, being flush with the ground, and covered by silk. One tarantula occupies each burrow at a time (excluding mothers with new offspring) (Punzo & Henderson, 1999), making burrows an excellent proxy for surveying the animals (Hembree, 2017). We confirmed occupancy by fishing the tarantula to the surface or near the entrance using a blade of grass. If fishing was unsuccessful, burrows were deemed occupied if they showed clear signs of habitation, such as silk coverings, molts, or egg sacs. While false negatives were possible, the distinctive appearance of tarantula burrows minimized this risk. For each burrow, we recorded its perpendicular distance from the transect, diameter, and GPS coordinates.