A peer-reviewed version of this preprint was published in PeerJ on 29 September 2016. View the peer-reviewed version (peerj.com/articles/2538), which is the preferred citable publication unless you specifically need to cite this preprint. Roberson EJ, Chips MJ, Carson WP, Rooney TP. 2016. Deer herbivory reduces web-building spider abundance by simplifying forest vegetation structure. PeerJ 4:e2538 https://doi.org/10.7717/peerj.2538 1 Deer herbivory reduces web-building spider abundance by simplifying forest vegetation structure 2 3 4 Elizabeth J. Roberson1, Michael J. Chips2, Walter P. Carson2, Thomas P. Rooney1 5 6 1 Department of Biological Sciences, Wright State University, 3640 Colonel Glenn Hwy., 7 Dayton OH 45435, USA 8 2 Department of Biological Sciences, University of Pittsburgh, A234 Langley Hall, Pittsburgh, 9 PA 15260, USA 10 11 Corresponding Author: 12 Rooney, Thomas P. 13 Department of Biological Sciences, Wright State University, 3640 Colonel Glenn Hwy., Dayton 14 OH 45435, USA 15 [email protected] 16 1 PeerJ Preprints | https://doi.org/10.7287/peerj.preprints.2153v2 | CC BY 4.0 Open Access | rec: 3 Sep 2016, publ: 3 Sep 2016 1 ABSTRACT 2 Indirect ecological effects are a common feature of ecological systems, arising when one 3 species affects interactions among two or more other species. We examined how browsing by 4 white-tailed deer (Odocoileus virginianus) indirectly affected the abundance and composition of 5 a web-building spider guild through their effects on the structure of the ground and shrub layers 6 of northern hardwood forests. We examined paired plots consisting of deer-free and control plots 7 in the Allegheny Plateau region Pennsylvania and Northern Highlands region of Wisconsin. We 8 recorded the abundance of seven types of webs, each corresponding to a family of web-building 9 spiders. We quantified vegetation structure and habitat suitability for the spiders by computing a 10 web scaffold availability index (WSAI) at 0.5 m and 1.0 m above the ground. At Northern 11 Highlands sites, we recorded prey availability. Spider webs were twice as abundant in deer-free 12 plots compared to control plots, while WSAI was 7-12 times greater in deer-free plots. Prey 13 availability was lower in deer-free plots. With the exception of funnel web-builders, all spider 14 web types were significantly more abundant in deer-free plots. Both deer exclusion and the 15 geographic region of plots were significant predictors of spider community structure. In closed 16 canopy forests with high browsing pressure, the low density of tree saplings and shrubs provides 17 few locations for web-building spiders to anchor webs. Recruitment of these spiders may become 18 coupled with forest disturbance events that increase tree and shrub recruitment. By modifying 19 habitat structure, deer appear to indirectly modify arthropod food web interactions. As deer 20 populations have increased in eastern North America over the past several decades, the effects of 21 deer on web-building spiders may be widespread. 22 2 1 INTRODUCTION 2 3 Indirect ecological effects due to direct interaction between two species that affect a third 4 species (Wooton 1994) often arise due to the actions of dominant species, keystone species, or 5 ecosystem engineers (Jones et al. 1994; Pringle 2008). Menge (1995) reported that indirect 6 interactions account for ~40% of the change in the abundance and percent cover of species after 7 experimental manipulations of rocky intertidal food webs. These indirect effects occurred 8 coincident with or shortly after direct effects are observed (Menge 1997). Despite their 9 importance, indirect effects can be difficult to detect, particularly in short-term studies (Hamilton 10 2000). Moreover, indirect effects can be conflated with direct effects and therefore overlooked 11 entirely (Wooton 1994). Here, we examine the indirect effects of a large mammalian generalist 12 herbivore on the structure of a web-building spider guild. 13 White-tailed deer (Odocoileus virginianus, hereafter deer) in North America have 14 increased in abundance in recent decades (Crête 1999; Ripple et al. 2010; Bressette et al. 2012). 15 In the early 20th century, deer were rare or absent from most of the United States (Leopold et al. 16 1947). Now, high deer abundance presents several management problems in much of the United 17 States (Warren 1997; Côté et al. 2004), including much of Wisconsin and Pennsylvania. Through 18 selective feeding, deer directly affect forest communities by altering species composition and 19 vegetation structure (Côté et al. 2004; Takatsuki 2009; Habeck and Schultz 2016). These direct 20 effects have the potential to indirectly alter the abundance of co-occurring animal species 21 (Rooney 2001; Rooney and Waller 2003; Sakai et al. 2012). For example, through resource 22 competition, deer can negatively affect the abundance of small granivorous mammals. McShea 23 (2000) observed that in years of low food (acorn) abundance, deer reduced the abundance of two 3 1 common species of rodent by 50%. Similarly elk (Cervus elaphus) reduced vegetation cover, 2 thereby causing a decline in the abundance of woodrats, voles, and two species of mice (Parsons 3 et al. 2013). Additionally, deer herbivory can alter resource quality for other herbivores by 4 altering plant species composition, or increasing secondary metabolites of particular species 5 (Vourc’h et al. 2001; Nuttle et al. 2011). A reduction in vegetation cover and vertical complexity 6 alters habitat for birds and other flying species (Rooney 2001). The removal of deer can lead to 7 increased vertical structure and ground cover (Rooney 2009). In studies where deer are removed, 8 ground and shrub-nesting birds increase in abundance (McShea and Rappole 2000; Holt et al. 9 2011). 10 The indirect effect of deer on arthropods may be strong for species that depend on 11 vegetation for habitat (Stewart 2001), because deer browsing reduces the three dimensional 12 structure of the ground and shrub layers of forest habitats (Habeck and Schultz 2015). Vegetation 13 structure is important for web-building spiders, which use woody and herbaceous surfaces as 14 anchoring points for their webs. These anchoring points can be a limiting resource for web 15 builders (Rypstra 1983; Uetz 1991; Gómez et al. 2016). Miyashita et al. (2004) examined this 16 relationship in forested regions of Japan. They report that the abundance and richness of web- 17 building spiders increased in areas without deer browsing. They attributed this to an increase in 18 vegetation cover, or more specifically, physical structures for anchoring webs. In a follow-up 19 study, Takada et al. (2008) found that web-building spiders were more vulnerable than non-web 20 builders to deer browsing. 21 In this study, we determined how deer affected assemblages of web-building spiders and 22 their habitat structure. We built on previous work (Miyashita et al. 2004; Takada et al. 2008) by 23 identifying responses of a broader range of web-building spiders to browsing effects. We 4 1 examined web-building spider assemblages with and without deer, using a paired exclosure- 2 control design. We surveyed webs, documented vegetation structure, and inventoried spider prey 3 to determine the degree to which deer alter the abundance and composition of a web-building 4 spider guild, their habitat structure, and their prey abundance. 5 6 MATERIALS AND METHODS 7 Field Methods 8 9 We surveyed ten paired exclosure-control study plots located in the north-central and 10 northeastern United States. Four paired plots were located in the Northern Highlands region of 11 northern Wisconsin in Vilas County (46°9’ N, 89°51’ W) on a 2500 ha property owned by 12 Dairymen’s Inc (Rooney 2009). This site supported high densities of deer throughout most of the 13 20th century, greatly altering plant community composition (Rooney 2009; Begley-Miller et al. 14 2014). In 1990, four deer exclosures were constructed in a 5 ha, old-growth hemlock-hardwood 15 stand (predominantly Tsuga canadensis, Acer saccharum, and Betula alleghaniensis). Exclosures 16 are 1.8m tall, constructed of wire mesh, and range in size from 169 m2 to 720 m2. Each exclosure 17 has an adjacent control plot of the same area, but with ambient browsing pressure. The 18 exclosures are separated from one another by a mean distance of 195 ± 15 m (Rooney 2009). The 19 remaining six paired plots were located in the Allegheny Plateau region in north-central 20 Pennsylvania in Elk County (41°25’ N, 78°50’ W). In the early 2000s, the Pennsylvania Game 21 Commission constructed and maintained an array of six deer exclosures in State Gamelands 44 22 and 28 across a 200-km2 area. This forest is part of the Hemlock-Northern Hardwood 23 Association (Whitney 1990), and is composed of second and third growth forests (predominantly 5 1 Acer rubrum, Prunus serotina, and Acer saccharum). For a more detailed description of the 2 region, see Horsley et al. (2003) and Chips et al. (2015). All exclosures were approximately 2.25 3 m tall, ranged in size from 500 m2 to 900 m2, and had an adjacent control plot in a randomly 4 selected location within 20 m of the edge of each fence. 5 We surveyed our plots for spider webs, and classified spider webs according to their 6 structure (Fig. 1). Spider families can often be identified by the types of webs they build. We did 7 not always identify the spider that created the web because they were not always present. 8 However, we identified the putative family of spider that created each type of web we tallied 9 (Bradley 2013). We classified webs as: funnel web (Agelenidae), (b) sheet web (Linyphiidae), 10 (c) mesh web (Dictynidae), (d) reduced orb web (Uloboridae) (e) vertical orb web (Aranaeidae), 11 (f) tangle web (Theridiidae), (g) horizontal orb web (Tetragnathidae).