This is to certify that the thesis entitled EDGE EFFECTS ON AVIAN AND MAMMALIAN NEST PREDATOR RELATIVE ABUNDANCE AND SPECIES RICHNESS presented by MICHELLE L. SMITH has been accepted towards fulfillment of the requirements for the MS. degree in ZOOLOGY 2 f! , E 1!! Major Professor’s Signature 3«/5—oq Date MSU is an Affirmative Action/Equal Opportunity Institution LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 6/01 chlRC/DateDuepGS—sz EDGE EFFECTS ON AVIAN AND MAMMALIAN NEST PREDATOR RELATIVE ABUNDANCE AND SPECIES RICHNESS By Michelle L. Smith A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Zoology 2004 ABSTRACT EDGE EFFECTS ON AVIAN AND MAMMLALIAN NEST PREDATOR RELATIVE ABUNDANCE AND SPECIES RICHNESS By Michelle L. Smith Few studies have systematically examined patterns of species richness and relative abundance of nest predators, although numerous studies invoke these patterns as potential causes for decreased nesting success of forest songbirds. Species richness and relative abundance of avian and mammalian nest predators were compared between forest edge and interior in two landscapes, Allegan State Game Area and Fort Custer Training Center (FCTC), in southwest Michigan. One avian predator, the American crow (Corvus brachyrhynchos), was significantly more abundant at the edge than the forest interior at FCTC. Other predators, specifically blue jays (Cyanocitta cristata) and mammals, did not demonstrate an edge response. However, abundance of these groups was highly variable across years. Abundance of brown-headed cowbirds (Molothrus ater), a brood parasite, was greater at edges in F CTC in 2003. Total species richness for both landscapes did not differ between the edge and interior. Remote camera systems captured three predation events, two involving red squirrels (T amiasciurus hudsom'cus) and one, a blue jay. The high variability in response indicates a need to assess patterns of predator distributions through long-tenn studies and at different spatial scales. ACKNOWLEDGEMENTS Funding for this project was provided by the Department of Military and Veterans Affairs, Michigan State University Graduate School, Department of Zoology, Ecology, Evolutionary Biology, and Behavior Program, and Kellogg Biological Station. Permissions to conduct research at field sites were granted from Allegan State Game Area and Environmental Office of the Department of Military and Veterans Affairs. Permits to conduct wildlife studies were acquired from the All-University Committee on Animal Use and Care (#05/03-061-00) and the Michigan Department of Natural Resources (#1508). Sherman live-traps for small mammal trapping were borrowed from The Nature Conservancy and Henry Campa III. I thank my graduate committee members, Catherine Lindell, Henry Carnpa III, and Barb Lundrigan, for their valuable input and suggestions on this project. Logistical support for statistical analyses was provided by Fernando Cardoso. Phil Myers and Scott Winterstein provided valuable insights regarding the design and implementation of the mammal surveys. Bert Skillen provided spatial data used in GIS analyses, and Tracy Aichele provided valuable assistance for those analyses. I thank my field assistants, Michelle Speicher, Amy Roosenbcrg, Sarah Stoll, and Brian Wasielewski, for their hard work and dedication in the field. I also thank the Kalamazoo Nature Center, especially Ray Adams, Mark Miller, and John Brenneman, for their continued cooperation on this project. Special thanks are extended to fellow graduate students, Amy Majchrzak, Rosa Moscarella, and Bert Skillen, for their support and friendship. iii Special thanks are extended to my family, especially Frank and Cheryl Smith, for their support. I am forever grateful for the love, patience, and support of Rob Schuster. iv TABLE OF CONTENTS LIST OF TABLES ................................................... vi LIST OF FIGURES ................................................... viii CHAPTER 1 Introduction ...................................................... Background ...................................................... Statement of Problem .............................................. Objectives ....................................................... \O\Ol\)— CHAPTER 2 Edge Effects on Avian and Mammalian Nest Predator Relative Abundance and Species Richness Abstract ......................................................... 10 Introduction ...................................................... 1 1 Methods ......................................................... 14 Study Area .................................................... 14 Survey Design ................................................. 14 Survey Methods ................................................ 17 Statistical Analyses ............................................. 21 Results .......................................................... 24 Discussion ....................................................... 37 Conclusions ................................................... 44 APPENDIX A ....................................................... 47 APPENDIX B ........................................................ SO LITERATURE CITED ................................................ 55 Table 1 Table 2 Table 3 Table 4 Table 5 Table A1 Table A2 Table A3 Table Bl LIST OF TABLES Analysis of the abundance of nest predators and the brood parasite at Fort Custer Training Center (FCTC) for two years (2002 and 2003), two treatments (edge and interior), and interaction. Degrees of freedom are 1 for all factors .................................. Analysis of the abundance of nest predator groups and the brood parasite at Allegan State Game Area (ASGA) for two years (2002 and 2003), two treatments (edge and interior), and interaction. Degrees of freedom are l for all factors .................................. Vegetation characteristics for Allegan State Game Area (ASGA) by treatment (edge and interior). All sampling points, 10 interior and 18 edge, were surveyed once for vegetation. Degrees of freedom are 1 for all factors ................................................. Vegetation characteristics for Fort Custer Training Center (F CTC) by treatment (edge and interior). All but one sampling point, 12 interior and 13 edge points, were surveyed once for vegetation. Degrees of freedom are 1 for all factors .................................. Comparison of spatial statistics for both study areas, Allegan State Game Area (ASGA) and Fort Custer Training Center (FCTC) ....... Total detections of species at Allegan State Game Area (ASGA) and Fort Custer Training Center (F CTC) by treatment (edge or interior) for both years (2002 and 2003) arranged by common name ............ Total number of detections by group at Allegan State Game Area (ASGA) and Fort Custer Training Center (FCTC) by treatment (edge or interior) for both years (2002 and 2003) ....................... Analysis of total species richness for both locations (Allegan State Game Area and Fort Custer Training Center) and treatment (edge and interior). Data from 2002 and 2003 are combined for both locations. Degrees of freedom are l for all factors ......................... Total number of live-captures, recaptures, and sampling effort for white-footed mice (Peromyscus leucopus) at each point at Allegan State Game Area ........................................... vi 25 27 34 35 36 47 48 49 50 Table BZ Table B3 Table B4 Total number of live-captures, recaptures, and sampling effort for white-footed mice (Peromyscus leucopus) at each point at Fort Custer Training Center ............................................ 51 (Peromyscus leucopus) by each point at Allegan State Game Area. . . . 52 Comparison of captures and detections of white-footed mice (Peromyscus leucopus) by each point at Fort Custer Training Center. . 53 vii Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 LIST OF FIGURES Sampling points in Allegan State Game Area (ASGA) located in Allegan County, MI. Locator map (inset) of ASGA in lower peninsula of Michigan ............................................... Sampling points in Fort Custer Training Center (FCTC) located in Kalamazoo and Calhoun Counties, MI. Locator map (inset) of FCTC in lower peninsula of Michigan ................................ Schematic of the survey design for sampling points. A) Focal vegetation types of forest and grassland. B) Sampling points were located in forest stands and were categorized as either edge or interior. Edge points were within 50 m of the forest- grassland habitat boundary. Interior points were at least 200 m from any edge ......... Mean number ( _+_ 1 SE) of crow detections per point count survey by landscape for both years (2002 and 2003). Sample size differed at Allegan State Game Area (ASGA) edge (N=18) and interior (N=10) and also at Fort Custer Training Center (F CTC) edge (N=14) and interior (N=8 in 2002; N=12 in 2003). The asterisk (*) denotes significant (p<0.05) differences between the edge and interior treatments at ASGA in 2002, and at FCTC in both 2002 and 2003. . . . Mean number ( i 1 SE) of blue jay detections per point count survey by landscape for both years (2002 and 2003). Sample size differed at Allegan State Game Area (ASGA) edge (N=18) and interior (N=10) and also at Fort Custer Training Center (FCTC) edge (N=14) and interior (N=8 in 2002; N=12 in 2003). The asterisk (*) denotes significant (p<0.05) differences between years (2002 and 2003) at FCTC .................................................... Mean number ( i 1 SE) of brown-headed cowbird detections per point count survey by landscape for 2003 only. Sample size differed at Allegan State Game Area (ASGA) edge (N=18) and interior (N=10) and also at Fort Custer Training Center (FCTC) edge (N=14) and interior (N=12). The asterisk (*) denotes significant (p<0.05) differences between edge and interior treatments at FCTC .......... viii 15 l6 18 26 28 29 Figure 7 Figure 8 Mean number ( _+_ 1 SE) of small mammal detections per track plate night by landscape for both years (2002 and 2003). Sample size differed at Allegan State Game Area (ASGA) edge (N=18) and interior (N=10) and also at Fort Custer Training Center (F CTC) edge (N=14) and interior (N=8 in 2002; N=12 in 2003). The asterisk (*) denotes significant (p<0.05) differences between years (2002 and 2003) at FCTC .................................................... 31 Mean number ( i 1 SE) of medium-sized mammal detections per track plate night by landscape for both years (2002 and 2003). Sample size differed at Allegan State Game Area (ASGA) edge (N=18) and interior (N=10) and also at Fort Custer Training Center (FCTC) edge (N=14) and interior (N=8 in 2002; N=12 in 2003). The asterisk (*) denotes significant (p<0.05) differences between the edge and interior treatments at FCTC in both 2002 and 2003 ....................... 32 ix Chapter 1 INTRODUCTION Habitat edges have been implicated as a potential cause for the recent decline of songbird populations. Few studies have systematically examined the impacts of edges on nest predators and brood-parasites, although this is commonly cited as a possible explanation for decreased nesting success of many songbirds. In this background chapter, I will provide a brief review of literature on habitat edges and discuss the impacts on populations of forest songbirds. In addition, I will specifically review literature on nest predator and brood-parasite response to fragmentation and habitat edges. I will then briefly outline the research question, as well as the objectives of my research. My aim is to provide background information regarding the impact of habitat edges on nest predator and brood-parasite populations, and explicate the importance of this information for the conservation management of songbird populations. Background Land use conversion is prevalent in the Unites States, particularly in the Midwest region (Robinson et al. 1995). One result of this practice has been a dramatic increase in fragmentation. Fragmentation causes physical changes in landscapes, such as decreased effective habitat area and increased habitat edges (Marini et al. 1995, Bowers et al. 1996). These changes can alter ecological processes and impact community composition and structure (Oehler and Litvaitis 1996, Manson et al. 1999). Fragmentation also increases landscape heterogeneity, diversity and complexity, potentially leading to more complex biological communities and interactions (Fagan 1999). Despite the extent and potential implications of forest fragmentation, there is little consensus on the impacts of forest fragmentation on biological communities (Donovan et al. 1997, Fahrig 2003). Studies that examine changes in biological communities in landscapes that differ in spatial configuration, or arrangement, are necessary to assess the impacts of fragmentation (F ahrig 1999). Edges are defined as transitions between two separate vegetation types marked by an abrupt change in structural complexity (Murcia 1995). The general types of forest edges, as defined by Kremsater and Bunnell (1999) include natural (e. g. topographic changes), permanent (e. g. forests adjacent to lakes), permanent anthropogenic (forests adjacent to residential development), natural successional (e. g. edges created by fires) and anthropogenic successional (e. g. edges created by clearcuts). This transitional area of an edge, also known as an ecotone, can impact the associated communities and ecosystems both directly and indirectly. Direct effects of habitat edges include changes in air temperature, wind velocity, humidity, and radiation (Kremsater and Bunnell 1999). These abiotic changes in the microclimate can impact species distribution and abundance through differential tolerance and survival at edges (Murcia 1995, Kremsater and Bunnell 1999). Wildlife species often demonstrate different responses to edges, ranging from preference to avoidance (Murcia 1995) and therefore edge effectsare often species- specific (Manson et al. 1999). Among the direct abiotic and biotic affects, there are also indirect edge effects, such as increased species interactions (Murcia I995, Morin 1999). Species interactions (e. g. pollination, predation, parasitism) may be either beneficial or detrimental. Consequently, edges vary in the impact on species as either a negative, positive, or non-existent effect. A majority of the recent literature has focused on the negative effects of habitat edges (Murcia 1995), although it is still debated as to whether or not edge effects exist (see Paton 1994, Lahti 2001). A general consensus is that edge effects exist for ‘some species, in some landscapes, some of the time’ (Dijak and Thompson 2000). Despite the great amount of literature on the topic, there are numerous misconceptions and erroneous assumptions associated with edge effects. The four most commonly cited misconceptions include 1) similarity of edge response for related species; 2) changes in biological communities occurring at fixed distances from habitat edge; 3) comparable edge response for a species at various edge types; 4) consistent edge response for a given taxon among different geographic regions (Sisk and Haddad 2003). However, research has demonstrated that edge effects cannot be generalized by taxa (e. g. Chalfoun et al. 2002a), distance (e.g. Dale et al. 2000), edge type (c. g. Bayne and Hobson 1998), or geographic location (e. g. Marzluff and Restani 1999). Since research indicates the lack of a pattern of organismal responses to edges, it is therefore necessary to determine the causes underlying this phenomenon of ‘edge effects’ to assess the impact of edges on wildlife communities. To understand the potential impact of habitat edges, it is necessary to critically examine proposed mechanisms, or causes, underlying edge effects. Several general mechanisms of edge effects have been proposed (Pagan 1999), including: 1) Differential movement of organisms at habitat edges, for example, edges may act as a filter or banier for some species; 2) Increased mortality of organisms at habitat edges; 3) Differential reproductive subsidies at habitat edges, for example, increased food resources at edges; and 4) Increased species interactions at habitat edges, for example, increased pollination, predation or parasitism at edges. While these four hypotheses have been addressed in the literature to varying degrees, there has been a substantial amount of research on the changes in biological interactions, in particular, nest predation and parasitism of songbird populations. A recent focus in research has been the impact of fragmentation and habitat edges on songbird populations. Many songbird populations have suffered dramatic declines in recent decades (Bdhning-Gaese et al. 1993). These p0pulation declines have been attributed to effects associated with increased levels of fragmentation and subsequent increase in habitat edges (Robinson et al. 1995). Nesting success is of particular interest, as this factor is directly related to population viability (Martin 1987). Numerous studies provide evidence that nest predation and parasitism rates increase with greater extent of fragmentation (see Lahti 2001). Furthermore, there is evidence to suggest that nest predation and parasitism levels increase with proximity to habitat edges (see Paton 1994). However, these effects are not always observed. For example, edge effects are rarely observed in the western US, although they are commonly reported in the Midwest (Marzluff and Restani 1999). The negative impacts of forest edges are often attributed to increased nest predation and brood parasitism levels as compared to forest interior (Murcia 1995, Heske et al. 2001). Several hypotheses have been proposed to explain increased levels of nest predation and parasitism including increased abundance, density, activity or species richness of nest predators and brood-parasites (Chalfoun et al. 2002b). It has been suggested that nest predators and brood-parasites are more abundant or active at edges (see Lahti 2001). Evidence for differential abundance or activity at edges has been found for various taxonomic groups, including deer mice (Peromyscus maniculatus, Bayne and Hobson 1998, Kristan et al. 2003), raccoons (Procyon lotor, Dijak and Thompson 2000), brown-headed cowbirds (Molothrus ater, Coker and Capen 1995, Evans and Gates 1997) blue jays (Cyanocitta cristata, Robinson and Robison 1999, Chalfoun et al. 2002b), and crows (Andrén 1992). There are still relatively few studies that specifically examine patterns of nest predator and brood-parasite abundance, activity or density. Furthermore, most studies focus on a single taxonomic group, therefore few studies have examined species richness at edges (Chalfoun et al. 2002a). Edge and landscape effects on nest predation levels are ultimately a function of nest predator ecology (Lahti 2001). Consequently, documenting patterns of nest predator abundance, distribution, density, and species richness would improve our understanding of differential nest predation rates (Chalfoun et al. 2002a). However, there has been little research regarding the effects of edges and landscapes on nest predators (Lahti 2001 ). Edge effects, historically generalized across taxa, are now considered to be species-specific (Sisk and Haddad 2003). Many recent studies have demonstrated that edge effects vary among different taxa, in that some species respond to edges, as avoidance or affinity to edges, whereas other species are not affected. Many nest predator species are considered to be generalists and capable of using edges. For example, Dijak and Thompson (2000) determined that raccoons were more abundant in agricultural edges than in forest interior. Studies of deer mice indicated significantly higher abundance at forest-farm boundaries as compared to forest interior (Bayne and Hobson 1998). Similar observations were recorded for the white—footed mouse in a forested landscape in the Northeast U.S. (DeGraaf et al. 1999). Corvid predators, specifically crows and blue jays, are often cited as being edge species. For example, Andrén (1992) determined that crows were more abundant in forest-agricultural patches compared to forest interior. Blue jays were more abundant in both clearcut and forest edges than in forest interior (Robinson and Robinson 1999, Chalfoun et al. 2002b). Brood-parasites are also often associated with edges. The brown-headed cowbird (Molothrus ater), a brood- parasite, was more abundant in forest openings than continuous forest (Coker and Capen 1995, Evans and Gates 1997). As edge effects are species-specific, these effects also vary with different landscape types. Landscape context is described as the habitat composition of the surrounding landscape (Lindenmayer et a1. 1999). Few studies have systematically examined the influence of landscape context on nest predator and brood-parasite ecology, although this has often been cited as an important factor related to nest predation levels (Lahti 2001). Oehler and Litvaitis (1996) assessed the numerical response of medium-sized predators, such as raccoons and skunks, in relation to habitat edges and diversity in three landscapes. While raccoons did not respond to habitat edges, abundance was positively correlated with increased landscape heterogeneity (Oehler and Litvaitis 1996). Bayne and Hobson (1997) compared landscapes fragmented by silviculture and agriculture with a contiguous forest landscape. It was determined that increased predation of ground nests in the agricultural landscape was associated with increased abundance of the red squirrel and increased diversity of predators in the landscape (Bayne and Hobson 1997). Tewksbury et al. (1998) determined that parasitism was most strongly correlated with extent of human development in the landscape, as well as host abundance. Blue jays were found to be more abundant in forest fragmented by logging in comparison with uncut forest (Robinson and Robinson 1999). A meta-analysis by Chalfoun et al. (20020) indicated that nest predators most often (41.2%) responded positively to edges in landscapes dominated by agriculture. Positive responses of predators through increased actiVity, abundance, or species richness were less common (12.9%) in forest landscapes and non-existent (0%) in grasslands (Chalfoun et al. 20020). This pattern has been attributed to differential resources associated with these habitats, in that agriculture may provide additional food and hosts, whereas clear—cut or grassland habitats provide relatively few resources. A recent review provided support for this idea, indicating that edge effects were more likely to be significant in agricultural landscapes than in other landscapes (Lahti 2001). The extent of fragmentation within a landscape also has an impact on wildlife populations. Small mammals often vary in abundance or density in response to fragmentation. For example, Yahner (1992) found that white-footed mice and southern red-backed voles were more prevalent in areas with relatively more fragmentation, although this was not the case for the masked shrew (Sorex cinereus). Abundance of small mammals also varied by year, where variation was attributed to changes in microenvironmental conditions (Yahner 1992). Generalist predators, specifically medium-sized mammals and corvids, are often associated with increased fragmentation. This idea was supported by the research of Oehler and Litvaitis (1996) that indicated greater abundance of raccoons and wild canids with increased extent of habitat fragmentation. In a study of corvids, Andén (1992) found that predation by crows increased relative to the amount of agriculture habitat in the landscape. Furthermore, abundance of the key predator (hooded crow, Corvus corone) was greatest in a heterogeneous forest-agriculture landscape (Andrén 1992). In addition, parasitism by brown-headed cowbirds, a brood-parasite, decreased with increasing forest cover; however, extent of human development and host abundance were better predictors of parasitism rates (Tewksbury et al. 1998). While landscape fragmentation and configuration are considered to be important factors affecting abundance and species richness of predator assemblages, few studies have examined these factors at the landscape scale (Fahrig 1999). Documenting species richness and abundance of communities in landscapes that vary spatially is a necessary component of fragmentation research (Fahrig 1999). Knowledge of nest predator response to edge and landscape configuration is important to understanding the causes underlying differential predation rates on breeding birds. However, few studies have systematically examined patterns of nest predator abundance or species richness with regard to edges and landscape configuration. Statement of Problem Population declines of some songbirds have been attributed to increased forest fragmentation and amount of edge (e. g. Robinson et al. 1995). Nesting success is ultimately related to the ecology of nest predators and brood-parasites (Chalfoun et al. 2002a). Therefore, understanding patterns of nest predator and brood-parasite abundance and species richness is essential to understand effects of fragmentation and edges on breeding birds. Objectives The purpose of this research was to address knowledge gaps on nest predator and brood-parasite response to fragmentation and edges. Species richness and relative abundance of avian and mammalian species were compared between two fragmented landscapes, and relative abundance was compared between forest edge and interior as an indication of edge response. This research specifically addressed the hypotheses that 1) relative abundance of nest predator and brood-parasite species is greater at edges as compared to forest interior, and 2) nest predator species richness is greater at edges than in the forest interior. Chapter 2 Edge Effects on Avian and Mammalian Nest Predator Relative Abundance and Species Richness Abstract Few studies have systematically examined patterns of species richness and relative abundance of nest predators, although numerous studies invoke these patterns as potential causes for decreased nesting success of forest songbirds. Species richness and relative abundance of avian and mammalian nest predators were compared between forest edge and interior in two landscapes, Allegan State Game Area and Fort Custer Training Center (FCTC), in southwest Michigan. One avian predator, the American crow (Corvus brachyrhynchos), was significantly more abundant at the edge than the forest interior at FCTC. Other predators, specifically blue jays (Cyanocitta cristata) and mammals, did not demonstrate an edge response. However, abundance of these groups was highly variable across years. Abundance of brown-headed cowbirds (Molothrus ater), a brood parasite, was greater at edges in FCTC in 2003. Total species richness for both landscapes did not differ between the edge and interior. Remote camera systems captured three predation events, two involving red squirrels (T amiasciurus hudsonicus) and one, a blue jay. The high variability in response indicates a need to assess patterns of predator distributions through long-term studies and at different spatial scales. 10 Introduction Fragmentation, the isolation of viable habitat patches, decreases effective habitat area relative to edge habitat (Marini et al. 1995, F ahrig 2003). Forest fragmentation has been implicated as a potential cause of population declines of passerines in North America (e.g. Robinson et al. 1995). This is a particular concern in the midwestem US, as fragmentation is prevalent and nest predation levels are relatively high in this region (Robinson et al. 1995, Donovan et al. 1997). Nest predation is a primary cause of nest failure of songbirds (Martin 1993), and therefore a key factor in population viability (Marini et al. 1995). Numerous studies have demonstrated an association between fragmentation and increased nest predation in the Midwest (e. g. Gates and Gysel 1978, Andrén and Angelstam 1988, Robinson et a1. 1995). Specifically, increased levels of landscape fragmentation negatively impact nesting success for many forest songbirds, with higher nest predation in more fragmented landscapes (Wilcove 1985, Donovan et al. 1997, Harley and Hunter 1998, but see Paton 1994, Lahti 2001). Previous research indicated that nest predators are influenced by landscape fragmentation, however relatively few studies examine the impact of fragmentation levels on nest predators (Chalfoun et al. 2002a). Edge habitat, the interface of adjacent land cover types, is a direct result of fragmentation. Ecological consequences of increased edges include abiotic and biotic changes, such as changes in wind disturbance or dispersal ability (Murcia 1995). Consequently, the creation of edges may also change community dynamics and biological interactions (Fagan et al. 1999), such as predator-prey dynamics (Mahan and Yahner 1999). Edges often have different effects on associated communities, although 11 recent research has focused on negative effects (Murcia 1995). In particular, forest songbirds are negatively affected by amount and proximity of edges (Hoover et a1. 1995). For example, King et al. (1998a) found that nest predation increased with proximity to clearcut edges. Nest predation is also greater in relatively small forest tracts in comparison with larger patches, as smaller forest tracts have a greater proportion of edge habitat (Hoover et al. 1995). However, recent reviews of edge effects on nest predation indicated mixed results, suggesting that edge effects are not consistent (Paton 1994, Hartley and Hunter 1998, Lahti 2001). Nest predation levels are ultimately a function of nest predator ecology; therefore, understanding edge effects on nest predator communities may prove to be important in establishing the relationship of edges and associated effects on breeding birds (Chalfoun et al. 20020). The pattern of elevated nest predation at forest edges has been attributed to increased abundance of nest predators at edges (Lahti 2001). While many studies have supported this hypothesis, other studies have found no differences in nest predator abundance (Paton 1994). For example, Dijak and Thompson (2000) determined that raccoons (Procyon lotor) were more abundant at forest edges within close proximity (less than 25 m) to agricultural fields compared to the forest interior (greater than 100 m). However, Chalfoun et al. (2002b) did not detect any significant differences in the relative abundance of raccoons in a similar landscape and geographic region. King et al. (1998b) found no difference in abundance of blue jays (Cyanocitta cristata) at a forest clear-cut edge as compared to the forest interior. In contrast, other studies have detected greater abundances of a crow (Andrén 1992) and other avian predators (Chalfoun et al. 2002b) at forest edges. Few studies have systematically examined nest predator ecology, and most 12 focus on a single response variable in a single taxon (Chalfoun et al. 20020). Consequently, most studies have not examined species richness, although greater species richness is one potential explanation for increased predation at edges (Chalfoun et al. 2002b). Studies that simultaneously examine avian and mammalian predators are necessary to determine causes underlying differential depredation of songbird nests, as birds and mammals are important predators of nests (Bollinger and Peak 1995) and often demonstrate differing patterns in relative abundance (Nour et a1. 1993). The main objective of this study was to determine the relative abundance and species richness of avian and mammalian nest predators at edges in comparison to the forest interior in two landscapes in southwest Michigan during 2002 and 2003. This study tested the predictions that 1) nest predators would be relatively more abundant at forest edges, although it was expected that this would vary by taxonomic group; and 2) species richness of nest predator assemblages would be greater at forest edges. Total abundance of all nest predators was also compared between two forested landscapes, which provided an indication of spatial variability in the region. In addition, relative abundance of the brown-headed cowbird (Molothrus ater), a brood parasite, was assessed in 2003. Based on previous research (Coker and Capen 1995, Evans and Gates 1997), it was predicted that this species would also be more abundant at habitat edges. In 2003, I sought to identify local nest predators in one of the landscapes. 13 Methods Study Area Study areas were located in Allegan State Game Area (ASGA; Figure 1; 42° 32' N, 85° 59' W) and Fort Custer Training Center (FCTC; Figure 2; 42° 18' N, 85° 19' W) in southwest Michigan. Both areas consist of a heterogeneous matrix of forest and other land cover types, such as grasslands and wetlands. ASGA and FCTC differ in total area, approximately 20,000 ha and 3,000 ha respectively. Both areas are subject to management activities such as timber harvesting and vegetation management. However, potentially invasive activities, such as timber-harvesting in the vicinity of sampling sites, were suspended for the duration of this study. Common tree species in the study areas included oak (Quercus spp.), maple (Acer spp.), aspen (Populus spp.), and pine (Pinus spp.). Understory largely consisted of sassafras (Sassafras 01bidum), dogwood (Camus spp.), elderberry (Sambucus canadensis), spice bush (Lindera benzoin), and multiflora rose (Rosa multiflora). Survey Design Sampling was conducted in deciduous/mixed forest stands. The adjacent edge habitat, defined by Paton (1994) as an opening of three times the canopy height, consisted of forest adjacent to grasslands at least 3,600 m2 in area. The physical appearance of the edge was considered to be a canopy dripline edge, or edges with an abrupt change in plant biomass and height with dense understory (Murcia 1995). Random sampling points were selected in forested areas using a stratified scheme so that points were within two distance-from-edge classes: 1) distances less than 50 m 14 .5353 he «Sufi—Eon 532 05 E 5 89¢ E com ~32 “a 0.63 .3an 55:: 53562 H853 cam—mg -388 05 Co E om 53:3 803 $58 owvm .uotoEm 8 owvo 556 we. wontowofio 203 we.» 35?. “mono.“ E @882 225 $58 wEEEwm Am .28?me ES 58o.“ mo womb 538%? Room A< .3an mam—9:3 8m swamp >955 05 mo oumfionom .m oSwE Eo Sow Econ T O 9 V Am— 2 cue—mmauw awoken 18 Point counts were used to provide the measure of relative abundance of the avian nest predators and brood parasites (Thompson et al. 1998). Fixed-radius (50 m) 10- minute point counts were used according to the protocol in Ralph et a1. (1993). Point counts were conducted in weather with little or no wind, from sunrise to three hours after sunrise (Ralph et al. 1993). Each point was sampled at least three times throughout the field season in 2002, and six times in 2003. Mammals were surveyed with track stations (Thompson et al. 1998), which were set up in close proximity to the established sampling points. Survey design and operation for aluminum track stations follows protocol of Justice (1961). Track plates (0.30 m x 0.45 m) were prepared with a modified chalk substrate as described by Drennan et al. (1998). Plates were left unbaited at sampling points for approximately 24 hours of exposure (Justice 1961). Animal tracks left in the substrate were preserved and cataloged. Tracks from track stations were identified using guides to mammal tracks (e. g. van Apeldoom et a1. 1993). Each point was sampled at least five times (except one point which was sampled only four times) in 2002, and eight times in 2003. In August of 2003, small mammals were also surveyed in two trapping sessions using Sherman live-traps baited with black-oil sunflower seeds and com-oat-molasses mix. A total of four traps were set randomly placed in close proximity (within 5 m) to each sampling point overnight, checked and rebaited daily for each 3-day trapping session. Data recorded included condition of traps (closed, open), identification of individuals, and number of individual captures per sampling point. Traps that were closed and empty were not counted for sampling effort. Captured animals were marked by 19 clipping a small patch (3 mmz) of the guard hairs on the dorso-caudal end. Live trapping and track plate surveys were not conducted simultaneously. Nest predators were identified by monitoring active nests of wood thrush (Hylocichla mustelina) with a remote camera system (TrailMaster® Passive Garne Monitor 550). Nests were located by a collaborating research group (Kalamazoo Nature Center). Cameras were set up in close proximity to nests (within 2 m) shortly after nests were located. Of the 8 nests that were monitored, 6 were still in the egg stage and 2 were in the nestling stage. Cameras remained on nests until the nest finished, either because of predation or a successful fledging event. Nest predators were identified from photographs after predation events. Vegetation composition and structure was characterized once at each sampling point. Approximately half of the points were surveyed in July of 2002, and all but one of the remainder were surveyed in July of 2003. Protocol was a modified version of the James and Shugart (1970) method of habitat characterization. Variables measured within 11.3 m radius plots included the number of trees by species by diameter breast height (DBH) classes (8-23, 23-38, and > 38 cm DBH) and canopy height. Variables measured within 5 m radius circular plots were: number of logs (2 7.5 cm in diameter, and _>_ 1.0 m long), number of stumps (2 7.5 cm in diameter, and 3 0.25 m tall), number of woody stems (> 0.5 m tall, < 8.0 cm DBH in size classes O-2.5 cm and > 2.5 cm), number of fallen branches (> 0.5 m long), and litter depth. Fifty meter line intercepts (Canfield 1941) were used to quantify percent bare ground, grasses, leaf litter, saplings, shrub, logs, snags, and canopy cover. 20 Land cover data for landscape metrics were acquired from the southern Lower Peninsula land cover 2000 data set (DNR). Land cover categories were originally in an Anderson Level 11 classification, but were reclassified for this analysis to an Anderson Level I classification. This classification consisted of seven land cover categories including forest, agriculture, shrub, urban, water, wetland and barren. A 15 km radius area centered on each study site was used to describe each landscape. Patch Analyst 3.0 (Rempel and Carr 2003) was used to calculate landscape and class level metrics. Landscape metrics included number of patches, mean patch size (ha), median patch size (ha), total edge (m), edge density (m/ha), and mean patch edge (m/patch); class metrics included total (ha) and percentage of area by land cover category. Landscape metrics were used to provide an indication of extent of fragmentation, and class metrics were used to describe the landscape context as relative proportions of cover types. Statistical Analyses The number of detections was used as the measure of abundance, although this does not necessarily indicate number of individuals. Abundance data did not meet normality assumptions, and transformations did not improve normality. Therefore differences in abundance between locations (i.e. ASGA or FCTC) or distance-classes (i.e. edge or interior) were tested using a repeated measures generalized log-linear model (PROC GENMOD; SAS Institute, 1999). The Pearson scale was used in all models to minimize overdispersion. All models met convergence criteria, which validates assumptions of this test. Total abundance was compared between sites using the model of location (ASGA or FCTC), year and interaction. This test was based on the number of 21 total detections from both avian point counts and mammal track plates. The number of detections by taxonomic group was used to compare edge response. This model included treatment (edge or interior), year, and interaction. Avian abundance was analyzed by species for both predators (American crow, Corvus brachyrhnchos; blue jay) and the brood parasite (brown-headed cowbird). The brown-headed cowbird data were analyzed only for 2003 due to low numbers of detections in 2002. For track plate surveys, small- mammals and medium-sized mammals were grouped by family. Small mammal families included Muridae, Soricidae, and Dipodidae. Medium-sized mammal family groups included Mustelidae, Sciuridae, Didelphidae, Procyonidae, Erethizontidae, Canidae and Felidae. Some mammal tracks were difficult to differentiate between families, particularly when only partial prints were obtained. In these instances, tracks were identified to group of small mammal (fore-print < 2cm in length) or medium-sized mammals (fore-print > 2cm in length) where possible. Tracks that were identifiable only to group comprised a large portion of the detections. The exclusion of this data would have dramatically reduced the total number of detections and therefore these data were included in analyses. Tracks that could not reliably be assigned to a group, such as tracks that consisted only of scratch marks, were considered unidentified and not included in analyses. Live-trapping data were analyzed to test for differences in the abundance of Peromyscus leucopus using a contingency table analysis of total number of individual live captures by treatment (PROC LOGISTIC; SAS Institute, 1999). The number of live captures was also compared to the number of track detections from track plate surveys by sampling point for detections of P. leucopus using the Pearson Correlation (PROC 22 CORR; SAS Institute, 1999). This was to test for a correlation of estimated abundances between the two survey methods. Species richness was calculated by tallying the total number of nest predator species detected by treatment at each study site for both years. Data were analyzed using the generalized log-linear model (PROC GENMOD; SAS Institute, 1999) of location, treatment, and interaction (location x treatment). A total of 17 vegetation variables were measured for this study. The distributions of most vegetation data were not normally distributed or transforrnable. Therefore the Wilcoxon-Mann-Whitney test was used to test for differences in vegetation variables between treatments within each study site. The sequential Bonferroni test was applied to table-wise p-values to reduce type I error inflation of multiple testing (Rice 1989). An alpha level of 0.05 was considered statistically significant, and means are presented with standard errors (SE) unless indicated otherwise. 23 Results Overall number of detections of avian and mammalian predators was significantly greater in FCTC (2002 total: 144; 2003 total: 187) in comparison with ASGA (2002 total: 100; 2003 total: 147) for both years (2002: 78:11.53, DF=1, p<0.001; 2003: x2=6.37, DF =1, p=0.011). Overall abundance was also significantly greater in 2002 than in 2003 (78:24.81, DF=1, p<0.001). The interaction between year and location was not significant (x2405, DF=1, p=0.152). A total of 252 and 222 point counts were conducted in ASGA and FCTC respectively. The American crow was significantly more abundant at the forest edge than in the forest interior at F CTC in both years (Table 1, Figure 4). In ASGA, crows were more abundant at the edge in 2002, but not in 2003 (Table 2). Blue jay abundance did not differ significantly between the habitat edge and forest interior for either ASGA or FCTC (Figure 5). The brown-headed cowbird was significantly more abundant at edges in FCTC (Table 1, Figure 6). Other potential avian nest predator species present at study sites included red-tailed hawks (Buteojamaicensis) and barred owls (Strix varia). These species were rarely observed, and therefore were not included in analyses. In both years, there were a total of 364 and 317 track plate nights for ASGA and F CTC respectively. A total of 130 and 272 mammal tracks were observed with track plates at ASGA and FCTC, respectively. Of these tracks, approximately 70% were identifiable; the remaining tracks were considered unidentified and not included in analyses. Small mammal species identified included the white-footed mouse (P. leucopus.), house mouse (Mus musculus), voles (Genus: Microtus; Microtus pennsylvanicus, M. ochrogaster, and M. pinetorum), northern short-tailed shrew (Blarina 24 Table 1. Analysis“ of the abundance of nest predators and the brood parasite at Fort Custer Training Center (FCTC) for two yearsb (2002 and 2003), two treatmentsb (edge and interior), and interaction‘. Degrees of freedom are 1 for all factors. Factor Chi-square P-value American Crow Treatment 6.09 0.01 3 Year 0.03 0.870 Treatment*Year 0.36 0.545 Blue Jay Treatment 2.06 0. 1 50 Year 13.95 < 0.001 Treatment*Year 0.09 0.764 Brown-headed Cowbird Treatment 4.53 0.033 Small Mammals Treatment 0.89 0.346 Year 17.71 < 0.001 Treatment*Year 0.12 0.727 Medium-sized Mammals Treatment 0.00 0.965 Year 2.20 0.138 Treatment*Year 6.34 0.01 1 a Repeated measures generalized log-linear model. b Difference of Least Square Means Statistics. c Type 3 GEE Analysis Statistics. 25 .38 EB Noam £3 E UHUm “a we.» .93 E 255 Esoo ESQ “on muouoowov 305 .«o Amm _ H v 825:. :82 .v oSwE UHUH— 255 :58 “Eon you 83823 bx. 0:3 .«0 Gm g H v 038:: :32 .m 83me 089m N O N O O O O 5 5.55385 5.238: V55585.: 5025.5: 5 O O O O O O 5 5.538388%: 05%: 50055.58: O 5 O 5 5 N O 5 5.8.888 0258325 50:58: 58: 2 ON 0 0 N O 5 : 8:5 2909i :0000:M O O 5 O O O O O 850:8: 20:385.: 05::08m O O O O O O O 5 235:8 55385: 88850 O O N O O O O 5 8.585ng 8.205: 32550. 550559-855: 5:25:02 O O 5 O O O O O 5.520835 .335 8:02 55:55. 30:32 O 5 N N O O O O 88:05:53.5 5883 50055085 N 5 O v 5 N O m 5.5.838 32 00:02 00:02 O O 5 O O O O O 5.55055503000850 2900.5 8: >50 5 5 5 5 O O O 5 0:5: 028.5. 59:55.50: 58: O 5 O O O O N O 5505508: 38.50.: 505:0m 5.5 5830 O5 5 5 a O O m 5 O 3:25.... 8585 055:85.50 55 8 5. N 55 ON 5 O as 3.5832 5:338 85:25-58: O 555 N5 ON O5 ON N ON 325.50 25588.6 3 85: O O O 5 O O O O 385 08.35 005:: O N 5 O O O O 5 85.5; 0505355 05:52 508:2. 55 45 5 N5 55 5.5 5. ON 8885588 380 320 555.508: 8:85 0w:m 8:85 0w:m 8:85 0w:m 8:85 0w:m 05: Z 0:55:85 0:: Z :0500 NOON NOON mOON NOON 05.05 <0m< 05:: :05500 53 :0w::5.5: AmOON ::: NOONO 80% Son 8.5 €5.85 8 0w:0v 5:05:05 3 6.505 8500 5555.5. 88:0 58: ::: A<0m