Brown-Headed Cowbird Parasitism of Neotropical Migratory Songbirds in Riparian Areas Along the Lower Colorado River

Brown-headed cowbird parasitism of neotropical migratory songbirds in riparian areas along the lower Colorado river

Item Type Thesis-Reproduction (electronic); text

Authors Averill, Annalaura, 1967-

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 11/10/2021 09:26:20

Link to Item http://hdl.handle.net/10150/192094 BROWN-HEADED COWBIRD PARASITISM OF NEOTROPICAL MIGRATORY SONGBIRDS IN RIPARIAN AREAS ALONG THE LOWER COLORADO RIVER

by Annalaura Averill

A Thesis Submitted to the Faculty of the SCHOOL OF RENEWABLE NATURAL RESOURCES In Partial Fulfillment of the Requirements For the Degree of MASTER OF SCIENCE WITH A MAJOR IN WILDLIFE AND FISHERIES SCIENCES In the Graduate College THE UNIVERSITY OF ARIZONA

1996 STATEMENT BY AUTHOR This thesis has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library. Brief quotations from this thesis are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgement the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: (4" u 1/

APPROVAL BY THESIS COMMITTEE This thesis has been approved on the date shown below: -

-1/Le,(4/5 /S2 Michael L. Morrison Date Visiting Associate Professor of Wildlife Biology

{, I i(jt,/I 11 Oec R. WilliaM Mannan Date Professor of Wildlife and Fisheries Science

— I 27/ 'I9L D. PhilipLGuertin Da ,e Associate Professor of Watershed Management 3 ACKNOWLEDGEMENTS Funding for this project was provided by the U. S. Fish and Wildlife Service, Southwest Region. : thank Kathy Milne and Bill Howe at the U. S. Fish and Wildlife Service Southwest Regional office in Albuquerque, NM for funding and logistical support. I also thank Wesley Martin at the Bureau of Reclamation for his support and enthusiasm for the neotropical migratory bird monitoring project along the lower Colorado River. I thank the personnel of Imperial, Cibola, Bill Williams River, and Havasu National Wildlife Refuges for their support, encouragement, and interest in this project: Andrew Loranger, Steve Hill, and Carmen Kennedy of Imperial National Wildlife Refuge; Elaine Johnson, Timothy Green, and Angel Montoya of Cibola National Wildlife Refuge; Nancy Gilbertson and Barbara Raulston of Bill Williams River National Wildlife Refuge; and James Good, Greg Wolf, and Christy Smith of Havasu National Wildlife Refuge. I also thank faculty members of The University of Arizona for providing support throughout this project. In particular, I thank Dr. Michael L. Morrison, my graduate advisor, for his support, guidance, and insightful review of all stages of my thesis. I also thank Dr. William Mannan and Dr. Philip Guertin for critical review of my study proposal and thesis. I thank Dawn Drumtra, Michael Duttenhoffer, Peter Hurley, Amy Kuenzi, and John Martin for the long hours of hard work they dedicated, without which this project would have been impossible. I thank my partner on the neotropical migratory bird monitoring project in the lower Colorado River valley, Suellen Lynn, for her help and insight. Finally, I would like to thank my friends and family for encouraging me to pursue my Master's degree and supporting me throughout this project. 4 TABLE OF CONTENTS LIST OF TABLES 6 ABSTRACT 8 INTRODUCTION 9 Study Objectives 10 Riparian Destruction: The Lower Colorado River 11 Effects of Riparian Destruction on Avifauna 14 Parasitism as an Indicator of Habitat Quality 15 Reproductive Ecology of the Brown-headed Cowbird 19 History of the Cowbird in the Lower Colorado River Valley 21 Hosts Species Breeding in the Lower Colorado River Valley 24 Impact of Cowbird Parasitism on Host Reproductive Success 27 Cowbird Management 30 STUDY SITES 32 METHODS 37 Avian Surveying: Variable Circular Plot Method 37 Nest Searching and Monitoring 39 Vegetation Sampling: Scale Considerations 41 Vegetation Sampling: Macrohabitat Features 43 Vegetation Sampling at nest sites: Microhabitat features 44 Vegetation Sampling at Random Sites: Available Plots 48 ANALYSES 49 Brown-headed cowbird and Host Species Relative Abundance 49 Parasitism Rates and Nesting Success 50 Nesting Success Calculated by the Mayfield Method 53 Macrohabitat Features Associated with Cowbirds and Hosts 56 Vegetation Analysis at Nest and Random Sites 56 RESULTS 62 Cowbird and Host Species Abundances 62 Parasitism Rates and Host Nesting Success 62 Brown-headed Cowbird Reproductive Success 75 Landscape Level Vegetation Associations 76 Nest-Site Selection Models 79 Parasitism Models 84 DISCUSSION 92 Parasitism Rates and Cowbird Abundance 92 Host Nesting Success 94 Brown-headed Cowbird Reproductive Success 102 Bird-Vegetation Associations: Macrohabitat Scale 103 Bird-Vegetation Associations: Microhabitat Scale 105 TABLE OF CONTENTS - continued Nest-site Selection by Host Species 105 Nest-site Discovery by Cowbirds 108 MANAGEMENT IMPLICATIONS 112 APPENDIX A. Codes and definitions of vegetation characteristics collected at nest and random sites 118 APPENDIX B. Nesting success of the Bell's vireo and yellow-breasted chat calculated by the Mayfield method . 122 APPENDIX C. Nesting success of the brown-headed cowbird in Bell's vireo and yellow-breasted chat nests calculated by the Mayfield method 123 APPENDIX D. Microhabitat characteristics of Bell's vireo nest sites (n = 147) 124 APPENDIX E. Microhabitat characteristics of yellow- breasted chat nest sites (n = 59) 125 APPENDIX F. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of Bell's vireo nest sites (n = 147) and random sites (n = 208) 126 APPENDIX G. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of yellow-breasted chat nest sites (n = 59) and random sites (n = 208) 128 APPENDIX H. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of successfully parasitized (n = 149) and unparasitized nests (n = 47) 130 APPENDIX I. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of Bell's vireo successfully parasitized (n = 102) and unparasitized (n = 26) 132 APPENDIX J. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of yellow-breasted chat successfully parasitized (n = 36) and unparasitized nests (n = 17) 133 LITERATURE CITED 135 6 LIST OF TABLES Table 1. Brown-headed cowbird and host species abundance per nest searching transects (n = 13) in the lower Colorado River valley, 1994 and 1995 63 Table 2. Spearman's rank correlation coefficients for cowbird abundance, host species abundance, and parasitism rates by transect, lower Colorado River valley, 1994 and 1995 64 Table 3. Parasitism rates of the Arizona Bell's vireo, yellow-breasted chat, blue grosbeak, and common yellowthroat in the lower Colorado River valley, 1994 and 1995 66 Table 4. Nesting success of Arizona Bell's vireo and yellow-breasted chats, lower Colorado River valley, 1994 and 1995 68 Table 5. Causes of nest failure for the Arizona Bell's vireo (n = 116) and yellow-breasted chat (n = 36),

lower Colorado River valley, 1994 and 1995 — 69 Table 6. Nesting and fledging success of the Arizona Bell's vireo, yellow-breasted chat, and brown- headed cowbird calculated by the Mayfield method, lower Colorado River valley, 1994 and 1995 70 Table 7. Number of species eggs, nestlings, and fledglings in parasitized and unparasitized nests of the Arizona Bell's vireo and yellow-breasted chat, lower Colorado River valley, 1994 and 1995 .. 72 Table 8. Frequency distribution of brown-headed cowbird eggs in Arizona Bell's vireo and yellow- breasted chat nests in the lower Colorado River valley, 1994 and 1995 73 Table 9. Number of host species eggs, nestlings, and fledglings in Bell's vireo and yellow-breasted chat nests that are singly and multiply parasitized, lower Colorado River valley, 1994 and 1995 74 7 LIST OF TABLES - Continued Table 10. Number of brown-headed cowbird eggs, nestlings, and fledglings in parasitized Bell's vireo and yellow-breasted chat nests, lower Colorado River valley, 1994 and 1995 76 Table 11. Expected and observed proportions of bird species detections by vegetation type, lower Colorado River valley, 1994 and 1995 77 Table 12. Bell's vireo nest site model fitted with variables that met criteria for forced entry into the model 80 Table 13. Yellow-breasted chat nest site model fitted with variables that met criteria for forced entry into the model 83 Table 14. General parasitism model fitted with variables that met criteria for forced entry into the model 86 Table 15. Bell's vireo parasitism model fitted with variables that met criteria for forced entry into the model 89 Table 16. Yellow-breasted chat parasitism model fitted with variables that met criteria for forced entry into the model 92 8 ABSTRACT Populations of several riparian-obligate, neotropical migratory songbirds have declined in recent years, partly due to brood parasitism by brown-headed cowbirds (Molothrus ater). In 1994 and 1995, 1 measured parasitism rates and nesting success of 4 riparian-obligate, neotropical migrants breeding in the lower Colorado River valley. Because vegetation characteristics of nest sites may influence cowbird parasitism and host nesting success, I also measured vegetation attributes associated with nest-site selection. Parasitism rates were 40-90%, and reproductive success was significantly lower in parasitized nests. Foliage cover may influence host nest-site selection and prevent cowbird nest discovery. The patchy distribution of native tree species may have aided cowbird nest discovery by providing elevated survey perches. Saltcedar (Tamarix ramosissima) was an important understory component providing foliage cover at nests. Understanding vegetation characteristics associated with host and cowbird nest-site selection should be useful in riparian restoration projects and cowbird control. 9 INTRODUCTION Riparian areas are characterized by deciduous trees (Populus, Salix, Acer, and Alnus spp.) and shrubby vegetation adjacent to streams, rivers, and other watercourses (Patton 1992:73). These areas provide a microclimate distinct from surrounding areas due to the tall, structurally complex vegetation, free-standing water, and high insect biomass (Rosenberg et al. 1982, Hunter et al. 1988, Patton 1992:73-74). The contrast between riparian and upland areas is especially pronounced in the arid southwest where the surrounding uplands are sparsely vegetated. Consequently, riparian areas in the southwest support among the highest diversity and density of breeding birds in North America (Johnson 1970, Johnson et al. 1977). Johnson et al. (1977) reported that 77% of breeding birds in the arid southwest were dependent on riparian vegetation at some point in their life, and 51% were completely dependent on riparian vegetation (termed riparian-obligate species). The destruction of gallery riparian forests in the western United States has been correlated with a decline in breeding populations of neotropical migratory songbirds (Klebenow and Oakleaf 1984, Rosenberg et al. 1991:25-26). In order to manage remnant riparian areas for obligate songbird species, the habitat features directly affecting reproduction and survival need to be identified (Martin 10 1992). These features include the availability and Quality of resources, intra- and interspecific competition, predation pressure, and incidence and intensity of brood parasitism by brown-headed cowbirds (Molothrus ater). My study focused on cowbird parasitism of songbird species breeding in a highly degredated riparian system in the arid southwest.

Study Objectives My general objectives were to (1) determine the impact of brood parasitism on nesting success of several riparian- obligate, neotropical migratory songbirds breeding in the lower Colorado River valley, and (2) determine vegetation characteristics associated with host nest-site selection and cowbird nest discovery. To meet these objectives, I focused on 4 species: the Arizona Bell's vireo (Vireo bellii arizonae), yellow-breasted chat (Icteria virens), blue grosbeak (Guiraca caerulea), and common yellowthroat (Geothlypis trichas). Arizona Bell's vireo and yellow- breasted chat populations are declining in the lower Colorado River valley, perhaps due to high levels of brood parasitism (Rosenberg et al. 1991:282-283, 302). My specific objectives were to: 1. Determine the incidence and intensity of brood parasitism experienced by 4 neotropical migratory songbird species in the lower Colorado River valley. 2. Determine the effect of parasitism on nesting success of the Arizona Bell's vireo and yellow-breasted chat. 3. Determine vegetation characteristics associated with nest-site selection of the Arizona Bell's vireo and yellow-breasted chat. 4. Determine vegetation characteristics associated with nest-site discovery and selection by breeding female cowbirds. 5. Provide management recommendations for cowbird control and riparian restoration.

Riparian Destruction: The Lower Colorado River Ninety percent of Arizona's historic gallery cottonwood (Populus fremontii) and willow (Salix spp.) forests have disappeared (Lofgren et al. 1990, cited in Krueper 1993). The riparian areas of the lower Colorado River valley exemplify this. Explorers of the lower Colorado River valley in the late eighteenth century described river banks and adjacent lands that were lined with thick groves of cottonwoods and willows (Bolton 1930, cited by Ohmart et al. 1977). Ohmart et al. (1977) estimated the historical extent of cottonwood-willow forest along the lower Colorado River at 2,024 hectares (5,000 acres). Destruction of gallery riparian forests along the lower Colorado River began with river navigation in the mid- nineteenth century (Ohmart et al. 1977). Fremont cottonwood, willow, and mesquite (Prosopis spp.) trees were cut down for fuel for steamboat traffic. As early as 1875, the Colorado River was examined as a possible irrigation source, but annual floods made farming along the river precarious (Ohmart et al. 1977). Massive floods in 1905 and 1922 put public pressure on Washington, D.C. to install dams for flood control. Damming of the Colorado River allowed for agricultural expansion throughout the floodplain (Ohmart et al. 1977). Concomitant was the destruction of large tracts of native Fremont cottonwood and Goodding willow (Salix gooddingii) (Ohmart et al. 1977). The introduction of saltcedar (Tamarix ramosissima) in the early twentieth century was also responsible for the decline of cottonwood-willow forests along the lower Colorado River (Ohmart et al. 1977, Rosenberg et al. 1991:21-22). Saltcedar was introduced into the United States as a soil stabilizer and an ornamental (Ohmart et al. 1977). Water management practices along the lower Colorado River created conditions that favored the spread of ' 3

saitcedar (Ohmart et al. 1977, Rosenberg et al. 1991:21-22). For example, increased soil salinity created ideal conditions for rapid spread of this salt-tolerant species. Floods became less frequent, but more severe. Saltcedar is tolerant of long-term flooding, whereas cottonwood and willow are dependent on short-term but frequent flooding. In addition, saltcedar is deciduous and the leaf litter accumulation creates highly flammable conditions. Saltcedar, unlike cottonwood and willow, regenerates rapidly after fire. Saltcedar also has a long period and high rate of seed production relative to cottonwoods and willows (Ohmart et al. 1977, Rosenberg et al. 1991:21-22). Anderson and Ohmart (1976, cited in Ohmart et al. 1977) estimated that 1,133 hectares (2,800 acres) of cottonwood- willow forest existed along the lower Colorado River in the mid-1970s. By 1986, only 307 hectares of the remaining riparian vegetation was mature cottonwood or willow (Ohmart et al. 1988). No mature cottonwood-willow patch >20 hectares currently exists along the lower Colorado River (Rosenberg et al. 1991:33). Most of the remaining cottonwood-willow forest in the lower Colorado River valley is located along the Bill Williams River, one of the principal tributaries of the lower Colorado River. The Bill Williams River also has the only large tract of naturally 14 regenerating cottonwood and willow forest in the lower Colorado River valley (pers. obs.).

Effects of Riparian Destruction on Avifauna The destruction of native riparian vegetation and the introduction of exotic plant species has important consequences for resident and migratory avifauna. Mature cottonwood-willow communities are structurally more complex (vertically and horizontally) than other riparian communities (Rosenberg et al. 1991:33). Past studies have shown a strong positive correlation between structural complexity of vegetation and avian species diversity and density (MacArthur and MacArthur 1961, Gaines 1974, Rice et al. 1983). The structural complexity of cottonwood-willow communities provides greater cover for breeding bird species during the extreme heat of summer (Rosenberg et al. 1991:33). Although mesquite communities also attract resident and migratory breeding birds, they do not support the abundance and variety of birds found in cottonwood- willow communities (Rosenberg et al. 1991:34). Saltcedar ranks third in terms of attracting and supporting breeding birds in the lower Colorado River valley, although it is more heavily used in other areas of Arizona. This regional difference in avian use of saltcedar is attributed to the 15

inability of this tree species to provide shelter from extreme heat (Hunter et al. 1988). The native riparian vegetation along the lower Colorado River is patchy in distribution and is often mixed with the invasive saltcedar species. Agricultural fields abut or are in close proximity to these riparian patches (pers. obs.). The result is a fragmented landscape with many distinct plant communities juxtaposed. The junction between plant community types may form a gradual transition zone or a well-defined boundary (Yahner 1988). These are referred to as ecotones and edges, respectively. Ecotones and edges frequently have high breeding bird diversity and density, although many of these species can be classified as habitat generalists (Gates and Gysel 1978). The brown-headed cowbird is a generalist species found in high densities in ecotones and edges (Brittingham and Temple 1983), and is an obligate brood-parasite (Friedmann and Kiff 1985). Therefore, high levels of brood parasitism are typical of edge habitats and frequently result in low reproductive success of songbirds breeding in these areas (Gates and Gysel 1978, Harris 1988, Yahner 1988).

Parasitism as an Indicator of Habitat Quality Habitat is a species or population specific concept. It is an area that provides the resources and environmental 1 6 conditions that promote occupancy and allow individuals to survive and reproduce (Morrison et al. 1992:11). Habitat duality is related to the rate of survival and reproduction of the individuals that inhabit an area (Van Horne 1983), and the ability of the young of any given year to themselves survive and reproduce (Morrison et al. 1992:12). High quality habitat provides conditions necessary for successful reproduction and survival for a relatively long time; marginal quality habitat supports individuals for relatively short periods and/or survival and reproductive rates are low (Morrison et al. 1992:11-12). Nest-site selection should be influenced by nest-site characteristics that affect number and vigor of young that successfully fledge (Martin and Roper 1988, Martin 1992). Brood parasitism typically lowers reproductive success, and thus may be a factor influencing nest-site selection in host species. A high incidence and intensity of brood parasitism may be indicative of degenerating resources (i.e., a lack of suitable nesting sites) for songbird species. Previous studies indicate that habitat quantity (patch size) affects the incidence and intensity of cowbird parasitism. Elevated levels of nest parasitism typically occur in forests with a high edge-to-interior ratio (Brittingham and Temple 1983). Riparian areas are ecotOnal, with high edge-to-a/ea ratios (Odum 1978) and high bird 17 species richness and density (Stauffer and Best 1980, Gates and Giffen 1991). The high density of breeding birds attracts predators and brood parasites, lowering reproductive success of open-nesting passerines (Johnston and Odum 1956, Gates and Giffen 1991). Many of the breeding birds in the arid southwest are riparian-obligates. Dependence on riparian forests in the xeric southwest may make these areas ecological traps for neotropical migratory songbirds (Gates and Giffen 1991). In addition, fragmentation of riparian forests creates isolated patches that may be more susceptible to cowbird parasitism. Sherry and Holmes (1992) found no parasitism in over 20 years in an unfragmented forest in New Hampshire, whereas Robinson (1992) found parasitism rates of 76% on neotropical migratory birds in a fragmented Illinois landscape. Several researchers have reported a negative correlation between distance of a nest to an edge or forest opening and the incidence of brood parasitism (Gates and Gysel 1978; Brittingham and Temple 1983), and a positive correlation between degree of habitat disturbance and parasitism levels (Airola 1986). Spatial heterogeneity within remnant patches may also affect parasitism levels (Brittingham and Temple 1983, Martin 1992). For example, studies in nest-site selection reveal that habitat complexity at the height of nest placement may be important 18 in concealing the nest from predators and cowbirds (Martin 1992). Other aspects of habitat quality, such as breeding densities of intra- and interspecifics, affect the level of brood parasitism on any one host species. Fretwell (1977) hypothesized an inverse relationship between frequency of cowbird parasitism and species-specific nest density, and field studies have supported this claim (Zimmerman 1983, Weatherhead 1989). Therefore, parasitism may be especially costly to a rare host species. Cowbirds parasitize over 220 species (Friedmann and Kiff 1985); thus, their reproductive success is not dependent on the success of a particular host. A rare host species will eventually be parasitized, and reproductive success may be lowered to a point that no longer compensates for adult mortality (Mayfield 1977). The rare species has a high chance of extirpation from an area as the result of brood parasitism, but the existence of local brown-headed cowbird populations will not be greatly affected due to the presence of alternate host species (Mayfield 1977). High parasitism rates do not necessarily threaten host population survival. It must be demonstrated that parasitism is lowering reproductive success such that a population is not replacing itself. If this is the case, it may be possible to evaluate habitat quality based on the 19 incidence (percent of nests parasitized) and intensity (percent frequency of multiple cowbird eggs in one nest with respect to total cowbird eggs laid) of brood parasitism.

Reproductive Ecology of the Brown-headed Cowbird The female brown-headed cowbird searches for nests to parasitize by quietly observing nest building behavior of a host species (Hand 1941; Mayfield 1961) and searching vegetation for available nests (Lowther 1979). Intently watching the host building a nest may stimulate the development of eggs, and thus accounts for synchronization of egg laying in the host and the parasite (Chance 1940, cited by Hand 1941). The female cowbird lays an egg in the host nest early in the morning. A host egg is usually removed and eaten on the day before, of, or after laying (Hann 1941). Occasionally nestlings are removed, perhaps in an attempt to elicit a host renest attempt (Scott and McKinney 1994). The incubation period for a cowbird egg is 10-13 days (Ehrlich 1988:616). This is shorter than the incubation period of many host species and, therefore, cowbird nestlings have a hatching advantage over host nestlings. Some bird species abandon incubation altogether once a nestling appears in the nest, triggered by the presence of the nestling to begin feeding behavior (Mayfield 1977). In 2 0 this case, the host eggs never hatch. In addition, the cowbird egg is often larger than the host egg. This is advantageous to the cowbird for two reasons: first, the cowbird nestling is more developed when it hatches; and second, the larger cowbird egg often prevents proper incubation of host eggs by blocking contact with the brood patch of the incubating female (Mayfield 1960). Cowbird nestlings are also substantially larger than most host nestlings. Cowbird nestlings beg louder and stretch higher toward parents returning with food (Walkinshaw 1972). This begging behavior extends into the fledgling period. Woodward (1983) observed hosts feeding cowbird fledglings more frequently than they fed their own young of equivalent weight. Scott and Ankney (1980) studied the fecundity of cowbirds in southern Ontario, and estimated forty eggs laid per female cowbird per breeding season. Adult female cowbirds typically breed for two seasons and, thus, could lay as many as eighty eggs in a lifetime. Scott and Ankney (1980) determined that a cowbird egg has a 0.03 probability of surviving to adulthood. Low survivorship provides an explanation for the extremely high egg production of the brown-headed cowbird. Cowbird eggs, nestlings and fledglings have a high rate of mortality. This is partly due to egg-rejection behavior 21 of hosts, but is also due to non-specific losses acting on open-nesting host and parasite alike (McGeen 1972). McGeen (1972) found that both hosts and parasites experience a 40% loss in production due to universal hazards such as weather, food limitation, and predation. In general, cowbird breeding success will be determined by the fate of eggs laid in the nests of abundant species that are heavily parasitized (Young 1963).

History of the Cowbird in the Lower Colorado River Valley The brown-headed cowbird was formerly restricted to the grasslands of the mid-continent, but spread westward and eastward concomitant with the expansion of agriculture and the fragmentation of large forested areas (Mayfield 1965). Farming and grazing lands provide ideal forage for the cowbird. Forest fragmentation creates patchy vegetation structure, exposed perches, and high host nest densities typical of edge habitats and favorable for the reproductive strategy of the brown-headed cowbird (Mayfield 1965). Grinnell (1914:156-160) reported seeing cowbirds in the lower Colorado River valley during his 1910 exploration of the region. These were dwarf brown-headed cowbirds, known as the obscurus race. Although the cowbird was common in riparian forests in the spring, it remained unknown if they wintered locally (Grinnell 1914:156-160). By the 1970's, 22 cowbirds were known to be common year-round in the lower Colorado River valley, although exhibiting variable seasonal distribution (Rosenberg et al. 1991:335). Hunter et al. (1988) reported high densities of cowbirds in all riparian vegetation types along the lower Colorado River. The highest densities were in cottonwood-willow communities, but screwbean mesquite (Prosopis pubescens), honey mesquite (Prosopis glandulosa), and saltcedar were also used by cowbirds. Bird surveys along the Bill Williams River in 1993 determined that the total number of cowbirds in riparian areas was twice as high as the next most numerous avian species, and many times higher than most neotropical migrants (Neale and Sacks 1994). Brown-headed cowbirds are host-generalists (Friedman and Kiff 1985). The disproportionately high number of cowbirds in the lower Colorado River valley suggests that cowbirds are parasitizing a diversity of neotropical migratory species. The host species of greatest concern are those that are riparian-obligates, require large areas of contiguous forest (Robbins et al. 1989), have breeding seasons that coincide directly with that of the cowbird, and have few defense mechanisms against brood parasitism (Mayfield 1965; Rothstein 1975). 23 Mayfield (1965) attributed lack of host defenses to recent interfacing between host and parasite and, thus, the lack of evolutionary time for hosts to develop anti-parasite defenses. Local populations of some bird species breeding in the lower Colorado River valley may not have been exposed to cowbirds for the time necessary to evolve anti-parasite defenses. Such defenses include nest desertion, rebuilding the nest floor over the foreign egg, cowbird egg ejection, and aggressive nest guarding behavior targeted specifically at the cowbird (Robertson and Norman 1976). A species probably does not reject cowbird eggs if >25% of that species nests are parasitized (Rothstein 1975). Consequently, survival may be threatened by cowbird parasitism (Robinson et al. 1993). Rothstein (1975) found little direct evidence to support Mayfield's hypothesis, and outlined more probable causes for development of anti-parasite defenses in some species but not others. Two causes are related to nest concealment and bill size. Species with large, poorly concealed nests may be more likely to develop anti-parasite defenses due to the high rates of parasitism they experience. In addition, a species with a bill large enough to easily eject cowbird eggs may be more likely to develop egg-ejection behavior. Piercing the parasitic egg and leaving it in the nest, as might occur with a small-beaked bird, may not be an advantageous strategy. The broken parasitic egg could leak onto the host's eggs, causing them to stick to the nest or to each other, and thereby reduce their chance of successfully hatching (Rothstein 1975). Rothstein (1975) also suggested that anti-parasite defense behaviors may simply be extensions of established behavior patterns. For example, egg-ejection behavior could be an extension of nest sanitation behavior.

Host Species Breeding in the Lower Colorado River Valley Bird species breeding in the lower Colorado River valley that may be negatively affected by cowbird parasitism include the Abert's towhee (Pipilo aberti), Arizona Bell's vireo (Vireo bellii arizonae), black-tailed gnatcatcher (Polioptila melanura), blue grosbeak (Guiraca caerulea), common yellowthroat (Geothlypis trichas), lesser goldfinch (Carduelis psaltria), red-winged blackbird (Agelaius phoeniceus), song sparrow (Melospiza melodia), summer tanager (Piranga rubra), willow flycatcher (Empidonax traillii), yellow-breasted chat (Icteria virens), and yellow warbler (Dendroica petechia) (Friedmann and Kiff 1985, Ehrlich et al. 1988, Rosenberg et al. 1991). I concentrated on four host species that were known acceptors of cowbird eggs, were considered common cowbird hosts (Friedmann and Kiff 1985), had breeding seasons that overlapped with that 75 of the cowbird, and were neotropical migrants potentially experiencing population declines in the lower Colorado River valley (Rosenberg et al. 1991). These species were the Arizona Bell's vireo, yellow-breasted chat, blue grosbeak, and common yellowthroat. The Arizona Bell's vireo was once common in the lower Colorado River valley, but has declined in numbers due to destruction of native riparian vegetation and cowbird parasitism (Rosenberg et al. 1991:282). Rosenberg et al. (1991:282) estimated a 57 percent population decline between 1976 and 1986, and this subspecies is currently listed as endangered by the California Department of Fish and Game. The Bell's vireo is a habitat specialist in the lower Colorado River valley, breeding in tall mesquite woodlands near open water (Rosenberg et al. 1991:282) and in scattered willow forests with a shrubby understory (Serena 1986). Serena (1986) found willows to be the most important plant contributing to cover around nests in half of the territories studied, and small patches of willows were present in all territories. The destruction of cottonwood- willow riparian forest may be increasing the vireo's vulnerability to cowbird parasitism and contributing to its population decline (Rosenberg et al. 1991:283). The Bell's vireo has been heavily parasitized since at least 1900 with no serious repercussions until recently (Brown 1903, cited 26 by Rosenberg et al. 1991:283). Cowbirds are now considered a serious threat to regional survival of the Arizona Bell's vireo (Rosenberg et al. 1991:283). The yellow-breasted chat is one of the most common birds in the lower Colorado River valley, but populations have declined 30% since 1976 and it is listed as a Species of Special Concern by the California Department of Fish and Game (Rosenberg et al. 1991:302). Chats use cottonwood- willow riparian forests with a dense shrubby understory, but are somewhat flexible in their habitat requirements. Population declines have been attributed to loss of cottonwood-willow forest, but the persistence of chats in the lower Colorado River valley is due to their ability to use other vegetation, such as pure stands of saltcedar (Rosenberg et al. 1991:302). Yellow-breasted chats breed from May to August, but they are secretive birds and little is known of their breeding biology in this region (Rosenberg et al. 1991:302). They are a frequent host of the brown- headed cowbird (Ehrlich et al. 1988:548). The blue grosbeak is a common cowbird host in other regions (Ehrlich et al. 1988:556), but its exact status as a host in the lower Colorado River valley is unknown. Due to the blue grosbeak's breadth of vegetation use and apparent tolerance of habitat disturbance, it has not experienced a population decline in recent years (Rosenberg et al, 27 1991:308). Blue grosbeaks reach their highest densities in cottonwood-willow vegetation, but also breed in pure saltcedar stands. This species is associated with a dense shrubby understory with a few tall trees to use as song perches. It is often found in edges, disturbed and open areas, and isolated hedgerows (Rosenberg et al. 1991:308). The common yellowthroat occurs in the lower Colorado River valley year-round, but local breeding populations may be partly or wholly migratory (Rosenberg et al. 1991:299). Common yellowthroats are found in extensive marsh habitats, but the highest breeding densities occur in partially flooded cottonwood-willow riparian forest (Rosenberg et al. 1991:300). This species is considered one of the three most common cowbird hosts (Ehrlich et al. 1988:546).

Impact of Cowbird Parasitism on Host Reproductive Success Cowbird parasitism substantially reduces the reproductive success of host species. This has been shown in the Abert's towhee (Finch 1983), dickcissel (Spiza americana) (Zimmermann 1983), Kirtland's warbler (Dendroica kirtlandii) (Walkinshaw 1972), least Bell's vireo (Vireo bellii pusilus) (Franzreb 1990) and many other open-nesting passerines. In extreme cases, the cowbird may have been the principal cause of near extirpation of riparian-obligate or forest-dwelling species. Habitat destruction and nest 28 parasitism are attributed to such a decline in the 'east Bell's vireo in Baja California, Mexico (Olson and Gray 1989) and southern California, USA (Franzreb 1990), and the Kirtland's warbler in Michigan (Walkinshaw 1972). Mumford (1952) found that the presence of even one cowbird egg in the nest of the least Bell's vireo often resulted in no vireo fledglings. Past studies indicated that parasitism rates >25% (Robinson et al. 1993) or >30% (Layman 1987) could threaten survival of local populations of host species by destabilizing a population and increasing the probability of local extinction due to stochastic events. Finch (1983) also suggested that parasitism levels >30% could jeopardize the existence of regional riparian bird populations. High parasitism levels do not necessarily threaten host population survival. It must be demonstrated that parasitism is lowering reproductive success such that the population is not replacing itself. Most host species must produce 2.0-2.5 young/pair/season to maintain a positive population growth rate assuming 40-60% adult survival and 20-35% juvenile survival (Robinson et al. 1993). If cowbird parasitism lowers host reproductive success below this level, the host population may be in jeopardy of regional extinction. 29 Since I worked with unmarked populations, reproductive success per breeding pair could not be determined. However, the maximum rate of parasitism that a host population could withstand and still replace itself can be approximated from the equation: p c = fl - [20(1 - p o )]/(1 - 1')} (May and Robinson 1985) where A = number of young fledged by an unparasitized female in one breeding season 1/ = number of young fledged by a parasitized female in one breeding season, p = annual mortality rate of adult females, and

p o = mortality rate in the first year. This can only be an approximation because information on mortality rates for neotropical migratory songbird populations in the lower Colorado River valley is unavailable. A literature review can yield approximate values. Martin (1995) reported average annual adult mortality for shrub nesting (nests <3 m off ground), neotropical migrants as 0.48, and Robinson et al. (1993) assumed juvenile mortality rates between 65-80%. These values fall in the range of other reported studies (Nice 1937, Nolan 1978). Once population-specific information on fledging rates from parasitized and unparasitized nests is available, the number of broods per season that a pair must 3C successfully fledge for populations to sustain themselves given regional parasitism rates can be determined.

Cowbird Management Cowbird management depends on identifying those aspects of the environment that promote cowbird occupancy and allow survival and successful perpetuation of the species. Cowbird habitat should be discussed in terms of breeding and foraging habitat. Cowbirds flock and feed together in open pastures and agricultural lands, but do not actively defend feeding territories. However, female cowbirds may establish territories on nearby breeding grounds, chasing intraspecific intruders during periods of nest searching (Darley 1982). Rothstein et al. (1987) found that cowbirds travelled up to 7 km between feeding and nesting sites. Live decoy cowbird traps have been successful in managing neotropical migratory birds with small populations and narrow geographic ranges (Robinson et al. 1993). However, cowbird traps may not be effective over large fragmented landscapes that provide cowbirds with numerous feeding grounds. In this case, proper placement of cowbird traps is the key to a good cowbird management program (Robinson et al. 1993). Rothstein et al. (1987) found that trapping cowbirds at the largest and most conspicuous foraging sites in the Sierra Nevada mountains of California 3 1 had little impact on parasitism rates in the surrounding forest. Traps placed in dense riparian areas were also found to be less effective than traps placed in adjacent open areas (Robinson et al. 1993). To date, the most effective trap placement appears to be between cowbird feeding and breeding sites. Grzybowski (1990) found that such a strategy reduced parasitism on the black-capped vireo (Vireo atricapillus) from 70 percent to 20 percent, and reproductive success increased 6 to 8 fold. Other cowbird control methods include shooting cowbirds on the breeding or foraging grounds (Layman 1987), cowbird egg removal from host nests (Franzreb 1990), placement of artificial or inviable cowbird eggs in host nests to deter future parasitism (Ortega et al. 1994), elimination of livestock grazing near forests where hosts breed, and reforestation (Layman 1987). For the least Bell's vireo in southern California, cowbird trapping and removal of cowbird eggs from host nests was effective in reducing reproductive losses due to parasitism (RECON 1989, cited by Franzreb 1990). Similarly, Walkinshaw (1972) noted increased breeding success when cowbird eggs and young were removed from Kirtland's warbler nests. All of these cowbird control methods are labor intensive. The last option, habitat restoration, is the only permanent solutiun. Recovery of the Kirtland's warbler 32 exemplifies this. Cowbird removal programs were successful in raising productivity, but population size increased only slightly (Kelly and Decapita 1982). Population increases occurred following habitat restoration through sivicultural and prescribed burning programs (Probst and Weinrich 1993). It may be useful to identify specific vegetation characteristics associated with nest-site selection by female brown-headed cowbirds. If there are particular vegetation configurations that deter nest-site discovery by parasites, these characteristics could be enhanced through manipulation and revegetation efforts. Habitat restoration on the breeding grounds may lower parasitism and predation rates and, subsequently, increase nesting success and population viability of host species.

STUDY SITES My study sites were located on 4 national wildlife refuges spanning 285 kilometers of the lower Colorado River from Needles, California to Martinez Lake, Arizona. All 4 sites are extremely arid and hot, with an average annual rainfall of 6.5 cm in Yuma, Arizona to 10 cm near Bullhead City, Arizona (Sellers and Hill 1974, cited by Rosenberg et al. 1991:8). Average daily temperatures range from 3 degrees celsius low to 21 degrees celsius high in the winter, and 21 degrees celsius low to 41 degrees celsius 33 high in the spring and summer. Relative humidity ranges from 5 to 20 percent (Sellers and Hill 1974, cited by Rosenberg et al. 1991:8). The exotic saltcedar was the predominant riparian vegetation type along most of the river (Hunter et al. 1988). Native riparian vegetation consisted of remnant patches of Fremont cottonwood and Goodding willow, typically with a shrub or low tree understory (pers. obs.). Further from the river were large stands of mesquite, saltbush (Atriplex polycarpa), quail bush (A. lentiformes), inkweed (Suaeda torreyana), arrowweed (Pluchea serricea), Baccharis spp., and wolfberry (Lycium spp.). In areas where long-term flooding occurred, extensive marshes prevailed. In contrast to the riparian and marsh plant communities were the desert uplands, dominated by creosote bush (Larrea tridentata), and the desert arroyos, dominated by ironwood (Olneya tesota) and palo verde (Cercidium spp.; Rosenberg et al. 1991:39). My northern most study site was located on Havasu National Wildlife Refuge (HNWR), Mohave County, Arizona. The predominant vegetation communities on Havasu NWR were mixed saltcedar-mesquite woodlands and extensive marshes. There was one large stand of decadent willows with no regeneration occurring. Other stands of scattered willows had a dense mesquite-saltcedar understory. There was also 34 an extensive stand of athel tamarisk (Tamarix aphylla), an introduced species that reproduces vegetatively. Revegetation projects on the refuges hoped to reestablish native riparian plant communities along the lower Colorado River. At HNWR, there were two revegetation sites: a 1993 revegetation site in which sapling cottonwoods were planted in a narrow, linear plot approximately 2,000 meters long, and which was incapable of supporting breeding bird populations; and a 1980s revegetation plot, approximately 600 meters long by 50 meters wide, which was capable of supporting several breeding pairs of riparian-obligate songbird species (Averill and Lynn, University of Arizona, unpubl. data). Approximately 50 kilometers south of HNWR was my second study site, located at the Bill Williams River National Wildlife Refuge (BWRNWR), La Paz and Mohave Counties, Arizona. The Bill Williams River was the site of the largest remaining stand of cottonwood-willow riparian forest (approximately 160 hectares) in the lower Colorado River valley. However, saltcedar was a pervasive understory species on most of the refuge. Extended flooding as recent as 1993 caused extensive damage to the cottonwood-willow forest. The cottonwood-willow community has also been altered by fires and high soil salinity, situations that favor the spread of saltcedar (Rosenberg et al. 1991:21). 35 Consequently, natural regeneration of cottonwoods and willows was occurring, but few old cottonwood or willow trees remained. The BWRNWR also had a recent (1990s) cottonwood revegetation site (approximately 200 meters by 50 meters), large tracts of cattail (Typha latifolia) and giant cane (Phragmitis australis) marsh, and extensive desert uplands. Approximately 100 kilometers south of BWRNWR was my third study site, located at Cibola National Wildlife Refuge (CNWR), La Paz County, Arizona. This stretch of the river was dominated by saltcedar, occasionally inter-mixed with honey mesquite. There were small remnant stands of willows and cottonwoods with a dense understory of saltcedar and mesquite. Revegetation sites included a 30 hectare plot revegetated in 1978 with honey mesquite, cottonwood, willow, blue palo verde (C. floridum) and the exotic eucalyptus tree (Eucalyptus sp.); a 20 hectare plot of cottonwoods, willows, palo verdes, mesquites, and salt-tolerant shrub species such as quail bush, wolfberry and inkweed (Rosenberg et al. 1991:58-59); and a 1994 cottonwood revegetation site. Cibola NWR is one of the most managed stretches of the lower Colorado River. The river's course has been drastically altered, but the original channel has been maintained as relatively pristine backwater for migrating waterLirds (Rosenberg et al. 1991:106). Several large farm 36 units were also managed for wintering Canada geese (Branza canadensis) and sandhill cranes (Grus canadensis). My southern most study site was approximately 30 kilometers south of CNWR on Imperial National Wildlife Refuge (INWR), La Paz and Yuma Counties, Arizona. This portion of the Colorado River has not been channelized, and relatively pristine wetland habitats can be found (Rosenberg et al. 1991:107-108). The predominant vegetation communities on INWR were desert arroyos characterized by palo verde, ironwood, honey mesquite, smoke tree (Psorothamnus spinosus), catclaw acacia (Acacia greggii), brittlebush (Encelia farinosa), wolfberry, and saltbush; and desert bajada characterized by creosote bush, prickly-pear cactus (Opuntia basilaris), cholla (Opuntia spp.), ocotillo (Fouquieria splendens), and saguaro (Carnegiea gigantea). There were several stands of scattered willows mixed with saltcedar, each approximately 3 hectares in size. There was also a small stand of athel tamarisk, approximately 2 hectares in size. Agricultural or urban areas bordered the National Wildlife Refuges. These were potential feeding grounds for the brown-headed cowbird. Of particular concern was the high level of agriculture in the immediate vicinity of the Bill Williams River. Both Planet Ranch (on the southeastern border of BWNWR) and Parker Valley (30 kilometers southwest 37 of BWNWR) could provide ample foraging habitat for the brown-headed cowbird. In addition, there were many abandoned agricultural fields that bordered the cottonwood- willow riparian areas of the Bill Williams River. These field-forest ecotones were ideal habitat for the brown- headed cowbird. Similarly, the Blythe area (30 kilometers northeast of Cibola NWR) and the Yuma area (30 kilometers southwest of Imperial NWR) were largely devoted to agriculture, and thus provided nearby feeding grounds for the brown-headed cowbird.

METHODS Avian Surveying: Variable Circular Plot Method I used the variable circular plot method (Reynolds et al. 1980) to determine relative species abundance by vegetation type. This method was designed for rugged terrain and tall, structurally complex vegetation which is typical of the lower Colorado River valley. I established point-count stations along transects using a stratified sampling design (Bibby et al. 1992) such that each vegetation type representative of the lower Colorado River valley was sampled. I established point-count stations 200 meters apart in woody vegetation to avoid double counting birds at neighboring points (Bibby et al. 1992). On the desert upland transects, I established points 300 meters 38 apart. Point-count stations were frequently established along edges (narrow paths, roads, desert washes, field- forest boundaries) due to the high level of fragmentation in the lower Colorado River valley. I began bird surveys half an hour before sunrise and ended 3 hours after sunrise. This corresponded to the period of greatest bird activity. Four trained observers surveyed birds, each observer counting approximately 15 points per morning. Each observer alternated the transect visited throughout the season to spread bias associated with inter-observer variability evenly across all points. I surveyed each transect forwards and backwards on alternating visits to minimize the effect of time of day on bird counts. I surveyed each point-count station once every 20 days during the breeding season, beginning early April and ending late June, 1994 and 1995. Surveys were not done on mornings that were rainy or extremely windy. I allowed 1 minute to pass before beginning a survey to diminish the effect of disturbance of bird activity caused by walking to the point-count station. I recorded all birds seen and heard at each point-count station during a 5 minute interval. Five minutes allowed the majority of birds to be detected while minimizing the chance of counting an individual twice during the same period (Bibby et al. 1992). 39 This requirement is imperative if each counting station is to be statistically independent (Reynolds et al. 1980). I recorded the distance to each bird as falling into 1 of 5 concentric bands radiating out from the observer (Bibby et al. 1992): 0-30 meters, 31-60 meters, 61-100 meters, 101-150 meters, >150 meters. I recorded the location of each bird by vegetation type and the specific plant species used whenever possible. I also recorded the sex, age and activity of each bird.

Nest Searching and Monitoring I used established bird survey transects to determine areas where host species occurred, and I searched for nests along these same transects. I established additional nest- searching transects in areas not surveyed but known to contain at least 1 of the host species. I started each transect at a randomly placed point, and flagged 50 meter intervals as reference points. I searched 17 transects in 1994, and 15 in 1995. Six of the transects searched in 1994 were reused in 1995 due to incomplete coverage in 1994. I searched each transect once a week in 1994 and every 4-5 days in 1995. Visits were conducted from mid-April through early-July, 1994 and 1995. I concentrated my nest searching efforts on Arizona Bell's vireos, yellow-breasted chats, blue grosbeaks, and 40 common yellowthroats. I searched transects by observing nesting behavior of host species and by systematically

searching the vegetation. -1 used behavioral cues, such as the carrying of nesting material, food, or fecal sacs to discover nest locations (Mar-in and Guepel 1993). recorded nest location, nest contents, behavior of adult birds, and interactions with cowbirds. I searched underneath each nest for ejected cowbird eggs, and I examined the nest lining for buried cowbird eggs once each nest had fledged or failed. When a nest site was discovered, I tied orange flagging nearby and wrote directions to the nest on the flag. I did not inspect the contents of a nest when I suspected that incubation had not begun to minimize the chance of desertion due to disturbance (Ralph et al. 1993). I also did not approach a nest if cowbirds were in the vicinity. I monitored nests once a week in 1994 and every 4-5 days in 1995 to determine status (i.e., parasitism, stage in the nesting cycle, nest contents, nest success or failure; Martin and Guepel 1993). I checked high nests with the use of a mirror attached to a pole (Ralph et al. 1993). When it proved impossible to check the contents of a nest, I collected information on adult activity (or lack of activity) so that nesting stage could be determined (Martin and Guepel 1993). I assumed that a nest was successful if 41 fledglings were seen near the nest or begging was heard near the nest, if the nest had flattened rims and fecal droppings on the edge (Martin and Guepel 1993), or if the nest had large nestlings close to fledging age at the time of the last nest check and there was no evidence of predation. I assumed that the number of fledglings was equivalent to the number of nestlings present in the nest at the last visit, as long as these nestlings were within a few days of fledging. I assumed predation had occurred if the entire contents of a nest disappeared before fledging was possible, the nest was torn up or falling out of the tree, or egg fragments were in the nest or on the ground below the nest (Martin and Guepel 1993). I assumed that a nest was abandoned if nest building was incomplete or cold eggs were in the nest on two or more occasions.

Vegetation Sampling: Scale Considerations Several researchers have highlighted the importance of scale considerations in studies of wildlife-habitat relationships (Johnson 1980, Wiens 1986, Morris 1987, Block and Morrison 1991, Block and Brennan 1993:55-57). Markedly different patterns in habitat use emerge when patterns are examined at different scales (Wiens et al. 1986, Block et al. 1991). Block and Brennan (1993:56) recommended that avian-vegetation associations be examined at the 42 macrohabizaz scale (corresponding to the array of specific vegetation types or seral stages across the landscape) and microhabizat scale (corresponding to the presence of specific features in the environment, such as perches and nest sites, that elicit a settling response; Block and Brennan 1993:38). These correspond to Johnson's (1980) second and third order of habitat selection; that is, the selection of a home range and the selection of foraging and nesting sites within the home range. Microhabitat features should also be examined at several scales (Martin and Roper 1988). For example, nest- site selection should examine vegetation characteristics of the nest site and the nest patch (a specified area that encompasses the nest). In addition, habitat quality can be examined by measuring vegetation features associated with successful and depredated nests (Martin and Roper 1988, Holway 1991, Norment 1993, Filliater et al. 1994, Kilgo et al. 1996) or unparasitized and parasitized nests (Staab 1995, Brittingham and Temple 1996). I examined habitat selection for host species and cowbirds at 3 levels: the macrohabitat level, by examining bird associations with specific vegetation types; the microhabitat level, by examining vegetation characteristics associated with nest-site selection within specific vegetation types; and at the bird (or individual) level by 43 examining differences in nest-site characteristics of parasitized and unparasitized nests.

Vegetation Sampling: Macrohabitat features Riparian restoration efforts typically occur at a landscape level. Therefore, it is important to identify the macrohabitat features associated with cowbird parasitism. Parasitism rates are a direct function of cowbird density, which in turn is a direct function of host density (McGeen 1972, Mayfield 1977). Identifying the vegetation characteristics associated with high host species and cowbird abundance will likely identify landscape features associated with high parasitism rates. In addition, the identification of cowbird foraging areas will likely indicate areas where parasitism rates will be high. I classified vegetation within 100 meters of each point-count station by general community type. I established 8 100-m transects at each point-count station, centered on the point and extended to the north, northeast, east, southeast, south, southwest, west, and northwest. I recorded vegetation community type at 5 meter intervals, categorized as cottonwood-willow, monotypic saltcedar, Athel tamarisk, monotypic mesquite, mesquite/saltcedar mixed community, desert bajada, desert arroyo, arrowweed, marsh, open water, agricultural land, or gravel bars. 44 Vegetation Sampling at Nest Sites: Microhabitat Features I measured vegetation at nest sites after each nest had fledged or failed to quantify differences between unparasitized and parasitized nests. I was unable to measure vegetation at a few nests that disappeared and exact location was unknown. I quantified vegetation within a 5 meter radius circle centered on the ground below the nest, as recommended by Petit et al. (1988). Petit et al. (1988) measured vegetation characteristics at nest sites in 1, 5 and 11.3-m radius circles, and found that the larger sampling plot did not accurately reflect nest-site selection. Five meter radius plots have been used in several other studies of nest-site selection (Martin and Roper 1988, Kilgo et al. 1996). Therefore, I used 5-m radius plots for comparative purposes as well. The area within 5 meters of the nest is referred to as the nest patch, following Martin and Roper (1988). I established 2 5-meter radii at each nest site, centered on the nest and oriented in the east-west direction. I measured percent cover by species, vertical height composition, canopy and ground cover with the point intercept method (Bonham 1989). At meter intervals, I recorded the plant species whose foliage intercepted an imaginary vertical line extending straight up from each point. The vertical line was divided into 4 height 45 categories following Anderson and Ohmart (1986): 0-0.6 m corresponding to the herbaceous layer, 0.7-4.5 in corresponding to the low canopy layer, 4.6-7.5 in corresponding to the middle canopy layer, and >7.5 m corresponding to the upper canopy layer. Estimates were aided by the use of an ocular tube with crossthreads and a paperclip weight (James and Shugart 1970). If a plant species was sighted at the point of intersection of the crossthreads, it was recorded as present in the appropriate height category. I also measured canopy and ground cover at each meter marker by sighting up and down through the ocular tube and recording presence (+) or absence (-) of vegetation cover where the threads crossed. I recorded type of ground cover as bare ground or water (no vegetation cover), leaf litter and debris, or herbaceous cover. I calculated percent cover by species as the number of points within the nest patch at which each species was recorded divided by the total number of sample points. I calculated percent cover by species for each height interval in the same manner. I calculated percent cover in each height interval by summing the number of times vegetation (regardless of species) was recorded in a particular height interval and dividing by the total number of sample points (this value could be >100%). I calculated percent canopy 46 and ground cover as the percentage of sample points with a positive (4-) recording. I used the point-centered quarter technique (Cottam and Curtis 1956) to collect information on tree and shrub dispersion at the nest site because heterogeneity of vegetation near the nest may influence nesting success (Norment 1993). Dispersion was measured as the distance from the nest to the closest live shrub (woody plant >0.6 meters tall and <2.5 cm dbh, including plants that had single or multiple central stems) and closest mature tree (>8 cm dbh) in each quadrant of a circle formed by perpendicular radii centered on the nest and extended in the cardinal directions. For each nest, I calculated the minimum, maximum, and mean distance to the closest shrub and mature tree. I calculated dispersion as the coefficient of variation (C.V.) for distance to closest shrub and mature tree. I recorded distance to the nearest opening, defined as an area with ground vegetation <1.5 meters high and no canopy cover. I classified each opening into 1 of 8 categories: road, path, desert wash, desert bajada, marsh, river channel, grassy field, or forest clearing. I recorded size of each opening as either <0.04 ha (22 m diameter circle) or >0.04 ha. I chose this criterion because the microclimatic conditions at the center of a forest opening 47 with a diameter 2 to 3 times the height of surrounding trees is similar to that of an edge (Smith 1986:206, cited in Paton 1994). The average canopy height ranged from 5 to 6 m in the mesquite and saltcedar riparian areas to 16 m in the cottonwood-willow riparian areas (pers. obs.), with an overall average of 11 m. I recorded distance from the nest to the nearest mature cottonwood or willow tree and cottonwood-willow tree density. If the nest was in a mature cottonwood or willow, this distance was set at 0. I estimated cottonwood-willow tree density by counting the number of mature trees within a 30 meter radius of the closest cottonwood or willow tree. I classified cottonwood- willow tree density into 7 categories: 0 trees, 1-10 trees, 11-20 trees, 21-30 trees, 31-70 trees, 71-100 trees, and >100 trees. At the nest site, I recorded nest substrate, height of the nest-plant, and height of the nest measured with a pole delineated by decimeter markings. Trees taller than 6.5 meters were measured with a clinometer or by ocular estimation with the pole serving as a reference. I measured percent canopy cover above the nest and ground cover below the nest by determining the amount of vegetation concealing a 25 cm diameter circle centered on the nest and measured from directly below the nest sighting up and down through an ocular tube (Martin and Roper 1988, Ralph et al. 1993). I 48 measured percent vegetation cover between the ground and the nest by lying on the ground and looking up through the ocular tube. I measured lateral nest concealment by determining percent vegetation cover of a 25 cm diameter circle centered on the nest and measured from a distance of 1 m in each of the four cardinal directions (Martin and Roper 1988, Ralph et al. 1993). Variables used in vegetation analyses were average lateral concealment, vertical concealment from above and below, percent ground cover, and total nest concealment (sum of lateral and vertical concealment).

Vegetation sampling at Random sites: Available Plots I measured vegetation characteristics within randomly placed 5 in radius circular plots to determine which attributes may be triggering nest-site selection in host species. I stratified random plots among the nest-searching transects, a single random plot per nest-site plot. Therefore, I examined nest-site selection within the general vegetation type associated with host nests. The sampling requirements were similar to that of the Breeding BIology Research and monitoring Database (BBIRD) Program protocol (Tom E. Martin, Montana Cooperative Wildlife Research Unit, University of Montana, Missoula, 1994). I determined the location of each random plot by generating 2 random decimal 49 numbers, and multiplying these numbers by the length and width of the transect. The first random number determined how many meters along the transect the plot would occur, and the second number determined how many meters into the vegetation the plot would be placed. I spun a compass to randomize the direction in which the plot was established. I measured the same vegetation characteristics at random sites except for nest concealment, distance to closest cottonwood or willow, and cottonwood-willow tree density. The latter 2 characteristics were measured at nest sites in conjunction with Lynn's (1996) study.

ANALYSES Brown-headed Cowbirds and Host Species Relative Abundance Thirteen of the bird survey transects were also used as nest-searching transects. For these transects, I calculated relative abundance of cowbirds and host species as the mean number of detections/point/count period. I only considered bird detections within 60 meters of each point count station to minimize the chances of double counting birds at adjacent points (Raphael 1987). I calculated relative abundance regardless of sex due to similarity of male and female calls in most species. I ranked species in descending order of mean detections/point/count period for each transect, and tied scores received an average group rank. I used the non- 50 parametric 2-tailed Spearman's rank test to examine correlations between host abundance and cowbird abundance (Milton 1992:446-447). I also used the 2-tailed Spearman's rank test to examine correlations between host abundance and parasitism rates. Parasitism rates were combined for yellow-breasted chats and vireos to increase the sample size on each transect (there was no statistical difference in the parasitism rates of these 2 species, Chi- square test-of-independence). Significance was set at P < 0.10 for all tests. Although an alpha level of 0.05 is frequently used, I set the alpha level at 0.10 to increase the power of my analyses (Milton 1992:214).

Parasitism Rates and Nesting Success I calculated parasitism rates for Bell's vireos, blue grosbeaks, common yellowthroats, and yellow-breasted chats by determining the percent of nests parasitized and successfully parasitized. I considered a nest parasitized if il cowbird egg or nestling was observed in the nest or cowbird eggs were discovered on the ground below the nest. Successfully parasitized nests were those in which cowbird eggs were laid at an appropriate point in the nesting cycle (egg laying or early incubation) and were not ejected from the nest. calculated nesting success for Bell's vireos and yellow-breasted chats, the only species with adequate sample sizes ( 20 nests; Hensler and Nichols 1981). I calculated nesting success as the percent of nests confirmed active that fledged at least 1 host young. I also calculated nesting success for unparasitized, parasitized, successfully parasitized, and unsuccessfully parasitized nests as the percent of nests confirmed active that fledged at least 1 host young. I tested for significant differences in nesting success between unparasitized and parasitized nests, and between unsuccessfully parasitized and successfully parasitized nests, with the Fisher exact test (Zar 1996:540- 552). I examined the mean number of host eggs, nestlings, and fledglings in nests that successfully completed the egg laying, incubation, and brooding stages, respectively, and tested for significant differences between unparasitized and parasitized nests with the Mann-Whitney U test (Sokal and Rohlf 1995:427-431). I determined the distribution of cowbird eggs in nests with completed clutches. I reported cowbird egg distribution as the frequency and percent of nests receiving 0, 1, and >1 cowbird egg, as well as the percent of nests with cowbird eggs below the nest. I calculated the mean number of host eggs, nestlings, and fledglings in successful nests that were singly (1 cowbird egg) and multiply (>1 52 cowbird egg) parasitized. I tested for significant differences between singly and multiply parasitized nests using the Mann-Whitney U test (Sokal and Rohlf 1995:427- 431). I also calculated cowbird fledging success in vireo and chat nests. I reported the mean number of cowbird eggs, nestlings, and fledglings in nests that successfully completed the egg laying, incubation, and nestling stages, respectively. I tested for significant differences in cowbird fledging success between vireo and chat nests with the Mann-Whitney U test (Sokal and Rohlf 1995:427-431). I attributed the cause of nest failure to abandonment or predation. In addition, I reported the percent of abandoned nests that were deserted due to parasitism. I assumed that a nest was abandoned due to parasitism if the nest was active up to the time parasitism occurred. I used the Chi-square goodness-of-fit test to determine if parasitized nests were more likely to be depredated than unparasitized nests (Zar 1996:457-511). I calculated parasitism rates and reproductive success for each year of the study. I used the Fisher exact test (Zar 1996:540-552) to determine if parasitism rates and nesting success for vireos and chats were independent of year. I used the Chi-square test-of-independence (Norusis/SPSS:129-131) to determine if parasitism rates were 53 independent of host species and study site (i.e., refuge). If there was no significant interaction, data was pooled across years, species, and sites.

Nesting Success Calculated by the Mayfield Method I calculated reproductive success by dividing the nesting cycle into 5 different stages: nest-building, egg- laying, incubation, hatching, and brooding stages. For the nest-building and egg-laying stages, I calculated the percent of all initiated nests that survived through each stage, respectively. I calculated reproductive success during the incubation and nestling stage using the Mayfield method (Mayfield 1961, 1975). The Mayfield method is an effective tool for calculating nesting success (Johnson 1979, Ralph et al. 1993, Martin and Guepel 1993). Hensler and Nichols (1981) recommended a minimum sample size of 20 nests for estimating reproductive success using this method. I found >20 nests for Bell's vireos and yellow-breasted chats, and subsequently calculated nesting success for these 2 species. Mayfield's method calculates a daily mortality rate (DMR) based on the amount of time each nest is under observation, referred to as nest days (Mayfield 1961, 1975). A nest seen once is not counted because the observation does not span a period of time. If a nest is depredated or 34

abandoned between nest checks, then the date of loss is -our_ at the middle of the interval between nest checks. The mortality rate is calculated by dividing the total number of nests lost by the total number of nest-day exposures (the sum of the nest days of all the nests under observation; Mayfield 1961, 1975). I assumed incubation began when the last egg was laid (Ehrlich et al. 1988:481). Final clutch size was the number of eggs in the nest after parasitism occurred, unless the nest was parasitized late in the nesting cycle (i.e., cowbird egg laid late in incubation, with nestlings in the nest, or after the nest fledged or failed). Occasionally nests were depredated between visits, and I did not know nest contents at the time of predation. In other words, the nest previously contained eggs, but they could have hatched before predation occurred. I randomly assigned half of these nests as lost during incubation and half as lost during the nestling stage. Separate daily nest survival rates (1 - DMR) can be calculated for the incubation and brooding stage of the nesting cycle if it is suspected that they differ (Mayfield 1961, 1975). Past studies indicated that nests were more frequently depredated during the nestling stage because of frequent adult visits to the nest and the loud begging of juveniles (Mayfield 1975). Therefore, I calculated separate 55 daily survival rates for incubation and brooding periods. calculated nesting success (the probability that a nest present at the beginning of incubation would survive to fledge host young) as the product of the nest survival rates during the incubation, hatching, and brooding periods (Mayfield 1961, 1975). Nest survival for each period pertained only to nests at which host young persisted. Mayfield (1961, 1975) suggested that it may be necessary to estimate DMRs of eggs and nestlings if high rates of partial clutch or brood loss was observed. I chose to calculate egg and nestling mortality rates due to frequent partial clutch/brood losses. I calculated daily egg/nestling mortality rates by dividing the number of eggs/nestlings lost without loss of the entire clutch by the number of days that all eggs/nestlings were under observation (a product of the number of observation days and the number of eggs/nestlings). I calculated fledging success (the probability that an egg present at the beginning of incubation would fledge) as the product of the nest survival rates during incubation and brooding, the egg and nestling survival rates, and the hatching rate (Mayfield 1961, 1975). I determined the hatching rate by summing the nestlings present immediately after hatching and dividing by the sum of the eggs present immediately before hatching. For this calculation, I only used nests that were checked 56 right before and after hatching dates. I calculated fledging success for Bell's vireos, yellow-breasted chats, and brown-headed cowbirds.

Macrohabitat Features Associated with Cowbirds and Hosts I calculated cowbird and host species frequency by vegetation type from recorded observations of vegetation use during point-count surveys. I determined availability of each vegetation type by summing over all points the number of times each vegetation type was encountered along the 8 sampling radii and dividing by the total number of observations. I tested for a difference between expected and observed use of vegetation types with the Chi-squared goodness-of-fit test (Norusis/SPSS 1990:225). I determined significant differences between expected and observed use of specific vegetation types with Bonferroni simultaneous 95% confidence interval estimations (Byers et al. 1984). Confidence intervals could not be calculated when observed frequencies were equal to 0 or 1.

Vegetation Analysis at Nest and Random Sites I used logistic regression to build models that would discriminate between vegetation characteristics of nest- sites and random sites, and unparasitized and successfully parasitized nests. I g/ouped nests in this manner because 57 nests that were not discovered by cowbirds until late in the nesting cycle had a success rate similar to unparasitized nests. This grouping also provided less disparity between sample sizes of the outcome groups. I built 5 models: a Bell's vireo nest-site selection model, a yellow-breasted chat nest-site selection model, a Bell's vireo parasitism model, a yellow-breasted chat parasitism model, and a general parasitism model (based on data from all 4 species). I chose logistic regression because of its appropriateness for data sets that are not normally distributed, contain binary dependent variables, and a combination of categorical and continuous independent variables (Hosmer and Lemeshow 1989:1-7). Prior to logistic regression modeling, I screened the vegetation variables with univariate analyses. Hosmer and Lemeshow (1989:83-87) recommended this strategy because models are more likely to be numerically stable and easier to interpret when fewer variables are entered. The decision of whether or not to screen variables depends on overall sample sizes and the sample size of each outcome group relative to the number of candidate variables (Hosmer and Lemeshow 1989:86). I had small sample sizes for some of my outcome groups, and therefore chose to screen variables prior to model building. For categorical variables, I tested for significant differences between outcome groups 58 with the Chi-square test-of-independence. For continuous variables, I tested for significant differences between outcome groups with the Mann-Whitney U test. I screened variables at the P < 0.25 level for entry into logistic regression (Hosmer and Lemeshow 1989:83-86). This significance level is higher than traditionally accepted values of 0.05 or 0.10 to allow for the possibility that a group of variables taken together may become an important predictor of the outcome variable (Hosmer and Lemeshow 1989:86). I also examined correlations between independent variables. I considered variables with a Pearson's correlation coefficient 10.80I to be highly correlated (Milton 1992:368). I retained one of each pair of correlated variables based on the relative ease of measurement and interpretation of each variable. I initially used forward stepwise variable selection to build the logistic regression models. I used the score statistic to evaluate variable entry into the model, and the change in likelihood (likelihood-ratio test) to evaluate variable removal from the model at each step (Norusis/SPSS 1992:15). Variables were entered into the model if the score statistic was <0.10, and variables were removed from the model if the likelihood ratio was >0.11. Because of unequal sample sizes in the categories of the dependent 59 binary variable, I ran 30 forward stepwise logistic regression analyses by subsampling from the larger category (i.e., subsampling from the random plots and successfully parasitized nests). I considered variables important in each model if inclusion of the variable raised the classification rate of either dependent category by 5%. Classification rate was determined by examining a 2 X 2 contingency table of predicted versus observed values (Norusis/SPSS 1992:8). From the contingency table, I could determine the percent of plots correctly classified as nests or random sites, and the percent of nests correctly classified as parasitized and unparasitized. Variables that were important in 20% (6) of the models were subsequently used in a forced entry logistic regression model. This model building strategy allowed me to create a final model that included only the group of variables that most consistently differentiated between outcome groups when data sets were subsampled. I examined all variables for numerical instability by inspecting the standard error of the coefficient. Large standard errors indicate numerically unstable results, which could be the result of partial correlation between independent variables or high frequencies of zero cells in a contingency table of independent variable by dependent variable (Hosmer and Lemeshow 1989:126-133). These problems 60 were avoided by running correlation analyses, and eliminating or collapsing zero count cells in a biologically meaningful manner. I verified the importance of each variable in the forward stepwise logistic regression models by examining the significance of the Wald statistic (variable coefficient/ variable standard error), and assessing goodness-of-fit of the model. I assessed fit of the models by examining the goodness-of-fit statistic, which compared observed probabilities to those predicted by the model (Norusis/SPSS 1992:10). I also examined the model Chi-square and the improvement Chi-square statistic significance levels to assess contribution of each variable to the final model (Norusis/SPSS 1992:11). I tested each variable in the forced entry model for non-linearity of the logit by quartile analysis. To do this, I ran 10 forced entry models (again subsampling from the larger outcome group) with the important variables categorized into quartiles. I inspected the estimated coefficients of the quartile groups for a linear increasing or decreasing trend (Hosmer and Lemeshow 1989:90). The most frequent trend observed was used to determine proper modeling of the variable in final analysis. The final model was based on 10 subsampled, forced entry models with important variables categorized as determined by auartile analysis. Subsampling was used for all final model building strategies except the yellow- breasted chat parasitism model where sample sizes were more equitable. I calculated the mean coefficient, standard deviation, and log odds ratio {exp(mean coefficient)} for each variable in the model. I assessed the stability of each variable in the final model by examining the C.V. of the model coefficient (Capen et al. 1986). The models were interpreted by examining the log odds ratio. These values were interpreted as the likelihood that a plot was random (nest-site selection models) or successfully parasitized (parasitism models). A greater likelihood was indicated by a log odds ratio >1, and a lower likelihood was indicated by a log odds ratio <1 (interpreted by examining the inverse of the log odds ratio). I examined the model's ability to correctly classify data (i.e., classification rate) at each step and for the final model using a 2 X 2 contingency table of predicted versus observed values. For the final model, I averaged classification rates across the 10 forced entry models. Nest-site selection models should be interpreted with caution because the dependent binary variables were not mutually exclusive (used resources are a subset of available resources). However, other researchers have used logistic 62 regression for analyses of used plots versus random plots (McShea et al. 1995).

RESULTS Cowbird and Host Species Abundances Cowbirds were 1.5 times more frequent than the most abundant host species (yellow-breasted chat), and 31 times more frequent than the least abundant host species (lesser goldfinch; Table 1). Cowbirds were 2.6 times more abundant than Bell's vireos. The abundance of all host species combined was approximately 3.4 times greater than cowbird abundance. Cowbird abundance increased as cumulative host abundance increased (r, = 0.525, P = 0.065), and parasitism rates on Bell's vireos and yellow-breasted chats decreased as cumulative host abundance increased (r, = -0.656, P = 0.015; Table 2).

Parasitism Rates and Host Nesting Success Two hundred and thirty four nests were confirmed active during my study: 154 Bell's vireo, 5 blue grosbeak, 14 common yellowthroat, and 61 yellow-breasted chat nests. Nest contents (i.e., parasitized or not) were determined at 131 Bell's vireo, 5 blue grosbeak, 11 common yellowthroat, and 55 yellow-breasted chat nests. 63 Table 1. Brown-headed cowbird and host species abundance per nest searching transect (n = 13) in the lower Colorado River valley, 1994 and 1995 . a

Species SD

HOST SPECIES Abert's towhee 0.55 0.19 Bell's vireo 0.47 0.38 black-tailed gnatcatcher 0.10 0.13 blue grosbeak 0.39 0.23 common yellowthroat 0.52 0.22 lesser goldfinch 0.04 0.05 red-winged blackbird 0.51 1.11 song sparrow 0.55 0.35 summer tanager 0.15 0.11 yellow warbler 0.09 0.10 yellow breasted chat 0.84 0.29 All common host species' 4.21 1.37 BROWN-HEADED COWBIRD brown-headed cowbird 1.24 0.33

• Bird surveys were conducted on 13 of the nest searching transects which were typical of riparian areas in the lower Colorado River valley. • mean # birds/point/count period for 13 transects. Detections were within 60 meters of point count stations. ' mean abundance of all common host species listed above. 64 Table 2. Spearman's rank correlation coefficients for cowbird abundance, host species abundance, and parasitism rates by transect, lower Colorado River valley, 1994 and 1995'

Host Abundance Cowbird Abundance Parasitism Rates'

d r P _5.r PQ

Bell's vireo 0.245 0.420 -0.249 0.413

Yellow-breasted chat 0.423 0.150 0.152 0.620

Cumulative hosts' 0.525 0.065* -0.656 0.015*

a Bird surveys were conducted on 13 of the transects used for nest searching. b Parasitism rate is the percent of all vireo and chat nests found on each transect that were parasitized. • Spearman's rank correlation coefficients. d * indicates significance at the P < 0.10 level. ' Host species abundance is the cumulative abundance of common host species breeding along the lower Colorado River. For a complete list of common hosts, see Table 1.

There was no association between parasitism of yellow- breasted chat nests and year of study (P = 1.000, Fisher exact test), successful parasitism of chat nests and year of study (P = 0.5742, Fisher exact test), and chat nesting success and year of study (P = 0.3766, Fisher exact test). Similarly, there was no association between parasitism of Bell's vireo nests and year of study (P = 0.2334, Fisher exact test), successful parasitism of vireo nests and year of study (P = 1.0000, Fisher exact test), and vireo nesting success and year of study (P = 0.7862, Fisher exact test). 65 There was also no association between parasitism and study site (X' = 5.964, P = 0.1134), and no association between successful parasitism and study site (X = 2.113, P = 0.5492). Therefore, data was pooled across years and study sites to increase sample size. Parasitism was associated with host species (X' = 12.561, P = 0.0057), but successful parasitism was not associated with host species (X' = 5.145, P = 0.1615).

These Chi-square statistics are questionable because 20% of the expected values were <5. Separate parasitism models were built for yellow-breasted chats and Bell's vireos, as well as one general model in which all species were combined. Parasitism rates ranged from 40% for the blue grosbeak to 90% for the common yellowthroat and Bell's vireo (Table 3). Bell's vireo and yellow-breasted chat nests were occasionally inappropriately parasitized. That is, cowbird eggs were laid in the nest toward the end of incubation, with nestlings in the nest, or after the nest fledged or failed. These nests usually fledged host young, and were considered unsuccessfully parasitized. Seventy to 80% of vireo and chat nests were successfully parasitized, a 10% difference from the actual parasitism rate (Table 3). Further results concentrate on the Bell's vireo and yellow- breasted chat due to sample size considerations. 66 Table 3. Parasitism rates of the Arizona Bell's vireo, yellow-breasted chat, blue grosbeak, and common yellowthroat in the lower Colorado River valley, 1994 and 1995.

Host Species n= % Parasitisme Successful Parasitism'

Bell's Vireo 131 90.1 80.2 Blue Grosbeak 5 40.0 40.0 Common Yellowthroat 11 90.9 90.9 Yellow-breasted 55 80.0 69.1 Chat

a Sample size is the number of nests at which parasitism could be determined (nests abandoned during construction and nests whose contents were unknown excluded from analyses). b Percent parasitized, including nests with cowbird eggs on ground below nest. ' Successfully parasitized nests were those at which parasitism occurred during egg-laying or early incubation and the cowbird egg had a chance of successfully hatching as long as the nest remained active.

Nesting success, calculated as the percent of all nests confirmed active that fledged at least 1 host young, was 11% for Bell's vireos and 29% for the yellow-breasted chat (Table 4). Unparasitized nests had a high success rate (70%) compared to parasitized (6-22%) and successfully parasitized (0-14%) nests. There was a significant difference in the proportion of unparasitized and parasitized Bell's vireo nests that fledged host young (P < 0.0001), and the proportion of successfully and unsuccessfully parasitized vireo nests that fledged host 6 7 young (P 0.0001, Fisher exact test). There was no significant difference in the proportion of unparasitized and unsuccessfully parasitized vireo nests that fledged host young (P = 0.4283, Fisher exact test). There was a significant difference in the proportion of unparasitized and parasitized yellow-breasted chat nests that fledged host young (P = 0.0065), and the proportion of successfully and unsuccessfully parasitized chat nests that fledged host young (P = 0.0148, Fisher exact test). There was no significant difference in the proportion of unparasitized and unsuccessfully parasitized chat nests that fledged host young (P = 1.0000, Fisher exact test). Nest failure was attributed to predation and nest abandonment (Table 5). Nest desertion during construction was low across species and years: >90% of all initiated nests were completed. Sixty percent of vireo nests and 83% of chat nests completed egg laying. Nest desertion during egg laying or incubation was high, and was typically a response to cowbird parasitism. Eighty percent of vireo and 55% of chat nests that were abandoned after nest building had recently been parasitized. Predation was the primary cause of nest failure once incubation began. Parasitism did not influence whether a vireo (P = 0.841) or chat (P = 0.230) nest was depredated (Fisher exact test). 68 Table 4. Nesting success of Arizona Bell's vireo and yellow-breasted chats, lower Colorado River valley, 1994 and 1995.

Parasitism Status' nt % Successful c % Failed

ARIZONA BELL'S VIREO Unparasitized 13 69.2 30.8 Parasitized 116 6.0 94.0 Successfully 103 0.0 100.0 parasitized Unsuccessfully 13 46.2 53.8 parasitized All nests 149 10.7 89.3

YELLOW-BREASTED CHAT

Unparasitized 1 0 70.0 30.0 Parasitized 41 22.0 78.0 Successfully 35 14.3 85.7 parasitized Unsuccessfully 6 66.7 33.3 parasitized

All nests 55 29.1 70.9

'Parasitized nests received at least 1 cowbird egg. Successfully parasitized nests were parasitized early in the nesting cycle, and the cowbird egg remained in the nest. Unsuccessfully parasitized nests received a cowbird egg, but the egg was laid late in the nesting cycle or was ejected from the nest. All nests include both unparasitized and parasitized nests. • Nests used in analyses were those that were confirmed active and fledging success for host species was known. • Percent of nests that fledged at least 1 Bell's vireo or yellow-breasted chat young. • Percent of nests that fledged only cowbirds or failed due to predation or nest abandonment. 69 Table 5. Causes of nest failure for the Arizona Bell's vireo (n = 116) and yellow-breasted chat (n = 36), lower Colorado River valley, 1994 and 1995.a

Species % Depredated= % Abandoned' building egg-laying/incubations

Arizona Bell's vireo 42.2 10.3 47.4

Yellow-breasted chat 63.9 5.6 30.5

a n refers to the number of nests that failed and cause of fialure was determined. b Percent of failed nests that were depredated. ' Percent of failed nests that were abandoned. d Percent of failed nests that were abandoned before nest construction was complete. ' Percent of failed nests that were abandoned during the egg-laying or early incubation stage. The majority of these nests were abandoned after 1 cowbird egg appeared in the nest (80% for the Bell's vireo and 55% for the yellow- breasted chat).

Overall breeding success, calculated by the Mayfield method as the percent of nests present at the start of incubation that survived to fledge young, ranged from 14% for Bell's vireos in 1995 to approximately 32% for yellow- breasted chats in 1994 and 1995 (Table 6). Partial clutch loss, frequently the result of cowbird egg removal, resulted in low probabilities of eggs surviving to fledge (7-23%). In both 1994 and 1995, yellow-breasted chat eggs and nests had higher survival probabilities than Bell's vireo eggs and nests. Daily mortality rates of nests were highest during the nestling phase, with the daily mortality rate of 70 nestlings frequently exceeding that of eggs. Daily mortality rates (DMR) of eggs, nestlings, and nests are presented in Appendix B.

Table 6. Nesting and fledging success of the Arizona Bell's vireo, yellow-breasted chat, and brown-headed cowbird calculated by the Mayfield method, lower Colorado River valley, 1994 and 1995.'

Species Year % nests survive b % eggs survive'

Bell's Vireo 1994 17.9 11.6 1995 14.0 7.4 yellow-breasted 1994 32.4 23.3 chat 1995 32.6 22.0 Brown-headed 1994 16.9 cowbirdd 1995 18.4

' Only nests that successfully completed the egg-laying stage were used in calculations. • Probability that a nest present at the start of incubation will survive to fledge at least 1 Bell's vireo or yellow- breasted chat. • Probability that a Bell's vireo, yellow-breasted chat, or brown-headed cowbird egg present at the start of incubation will survive to fledge. d Overall cowbird fledging success was similar in Bell's vireo and yellow-breasted chat nests (18 and 17%, respectively).

There was a significant difference in the mean number of Bell's vireo eggs (U = 131.5, P = 0.0000), nestlings (U = 53.0, P = 0.0008), and fledglings (U = 24.0, P = 0.0005) in successful nests that were parasitized and unparasitized (Table 7). Successful nests that were unparasitized fledged 71

approximately 2 more young per nest than parasitized nests. There was also a significant difference in the mean number of yellow-breasted chat eggs (U = 77.5, P = 0.0033), nestlings (U = 24.5, P = 0.0116), and fledglings (U = 17.5, P = 0.0300) in successful nests that were parasitized and unparasitized. Successful nests that were unparasitized fledged approximately 1 more young per nest than parasitized nests. Successful, parasitized yellow-breasted chat nests fledged approximately twice the number of their own young as did successful, parasitized Bell's vireo nests. This suggests that Bell's vireos are experiencing greater reproductive losses due to parasitism. In addition, chat nestlings in parasitized nests had a high probability of surviving to fledge as indicated by the slight difference in number of nestlings and fledglings per successful nest. There was a 2-fold difference in the number of vireo nestlings and fledglings in successful, parasitized nests. The greatest proportion of nests received 1 cowbird egg, but 36% of nests received >1 cowbird egg (Table 8). Eight percent of Bell's vireo nests also had a cowbird egg below the nest. There was a significant difference in the number of Bell's vireo eggs (U = 285.0, P = 0.0001) and nestlings (U = 34.0, P = 0.003) in nests that contained one cowbird egg versus those that contained multiple cowbird eggs (Table 9). On the other hand, he number of young fledging from 72 Table 7. Number of species eggs, nestlings, and fledglings in parasitized and unparasitized nests of the Arizona Bell's vireo and yellow-breasted chat, lower Colorado River valley, 1994 and 1995.a

Parasitism Eggs # Nestlings' #Fledglings' status SD SD n SD

ARIZONA BELL'S VIREO Unparasitized 3.5 0.5 13 3.1 0.9 11 2.9 0.8 9 Parasitized 2.1* 1.2 71 1.4* 1.3 30 0.8* 1.3 22 YELLOW-BREASTED CHAT Unparasitized 3.5 1.2 11 2.7 0.9 9 2.6 1.0 7 Parasitized 2.2* 1.1 33 1.5* 1.0 14 1.4* 1.0 12

a Only nests that were not abandoned during construction or egg laying and nests at which final clutch contents were known were used in calculations. • Mean and standard deviation of number of host eggs in nests that successfully completed egg-laying. ' Mean and standard deviation of number of host nestlings in nests that successfully completed incubation. • Mean and standard deviation of number of host fledglings in nests that successfully fledged. * indicates a significant difference in the mean number of host eggs, nestlings, or fledglings between parasitized and unparasitized nests at the P < 0.10 level, Mann-Whitney U test. multiply parasitized nests did not differ significantly from those nests that received only one cowbird egg (U = 37.0, P = 0.319; calculations based on nests that were not depredated). There was no significant difference between the number of yellow-breasted chat eggs (U = 107.5, P = 0.378), nestlings (U = 20.5, P = 0.635), or fledglings (U = 11.0, P = 0.360) in multiply and singly parasitized nests. 73 Table 8. Frequency distribution of brown-headed cowbird eggs in Arizona Bell's vireo and yellow-breasted chat nests in the lower Colorado River valley, 1994 and 1995.

Parasitism Statusa Frequency (%) Cowbird eggs below nest (%)'

Arizona Bell's vireo Singly Parasitized 55 (47.8) 7.8 Multiply Parasitized 60 (52.2)

Yellow-breasted chat 0. 0 Singly Parasitized 29 (67.4) Multiply Parasitized 14 (32.6)

Singly parasitized refers to nests with 1 cowbird egg. Multiply parasitized refers to nests with >1 cowbird egg. b Percent of parasitized nests that are singly and multiply parasitized. ' Percent of host nests with a cowbird egg on the ground below nest.

However, 9% of singly parasitized vireo nests fledged host young compared to 2% of multiply parasitized vireo nests. Twenty seven percent of singly parasitized chat nests fledged host young compared to 14% of multiply parasitized chat nests. Critical parasitism levels can be estimated using May and Robinson's (1985) equation (see introduction). Annual adult and juvenile mortality was set at 48 and 70%, respectively, and the number of young fledged by 74 Table 9. Number of host species eggs, nestlings, and fledglings in Bell's vireo and yellow-breasted chat nests that are singly and multiply parasitized, lower Colorado River valley, 1994 and 1995.'

Parasitism # Eggs # Nestlings d # Fledglings' status' SD n I SD SD

ARIZONA BELL'S VIREO

1 cowbird egg 2.8 0.9 30 2.0 1.2 18 1.2 1.6 12

>1 cowbird egg 1.6* 1.2 41 0.6* 0.8 11 0.4 0.7 8 YELLOW-BREASTED CHAT

1 cowbird egg 2.4 0.8 20 1.6 0.9 8 1.6 0.9 8

>1 cowbird egg 2.0 1.5 13 1.3 1.2 6 1.0 1.2 4

a Only nests that were not abandoned during construction or egg laying, and nests at which final clutch contents were known, were used in calculations. • Includes cowbird eggs below the nest. ' Mean and standard deviation of number of host eggs in nests that successfully completed egg-laying. ' Mean and standard deviation of number of host nestlings in nests that successfully completed incubation. e Mean and standard deviation of number of host fledglings in nests that successfully fledged. * indicates a significant difference in the mean number of host eggs, nestlings, or fledglings between nests that were singly and multiply parasitized at the P < 0.10 level, Mann- Whitney U test. unparasitized and parasitized females was substituted with species-specific parameters (Table 7). Assuming that each female successfully raised 2 broods, the Bell's vireo population could withstand parasitism rates up to 61% and the yellow-breasted chat population could withstand parasitism rates up to 84%. Brown-headed Cowbird Reproductive Success Brown-headed cowbirds experienced low reproductive success in Bell's vireo and yellow-breasted chat nests (Table 10). The percent of cowbird eggs present in vireo and chat nests at the start of incubation that survived to fledge was 18 and 17%, respectively (Appendix C). There was no significant difference in the mean number of cowbird eggs (U = 1030.0, P = 0.1540), nestlings (U = 191.5, P = 0.6423), and fledglings (U = 102.5, P = 0.2234) in vireo and chat nests (calculations consider only nests that were not abandoned or depredated). However, 12% of multiply parasitized vireo nests fledged cowbird young compared to 31% of chat nests. Bell's vireos also had higher desertion rates for nests parasitized during the egg-laying stage (Table 5). Daily mortality rates of parasitized nests, cowbird eggs, and cowbird nestlings calculated by the Mayfield method are presented in Appendix C. The daily mortality rate of parasitized nests was lower during the nestling stage than during the incubation stage. Similarly, the daily mortality rate of cowbird nestlings was lower than that of cowbird eggs. 76 Table 10. Number of brown-headed cowbird eggs, nestlings, and fledglings in parasitized Bell's vireo and yellow- breasted chat nests, lower Colorado River valley, 1994 and 1995.'

Host species'' # Eggs' # Nestlings' # Fledglings' _ SD _n SD _n SD _n

Bell's vireo 1.62 0.92 71 0.75 0.62 32 0.86 0.64 22 yellow-breasted 1.53 0.86 34 0.92 0.86 13 0.53 0.52 12 chat aNests used in calculations were those at which contents (exact # of cowbirds) was known. There was no significant difference in the number of cowbird eggs, nestlings, or fledglings between vireo and chat nests at the P < 0.10 level, Mann-Whitney U test. • Mean and standard deviation of number of cowbird eggs in host nests that successfully completed egg laying. • Mean and standard deviation of number of cowbird nestlings in host nests that successfully completed incubation. ' Mean and standard deviation of number of cowbird fledglings in host nests that successfully fledged.

Landscape Level Vegetation Associations Bell's vireos, yellow-breasted chats, and brown-headed cowbirds were found in cottonwood-willow and mesquite vegetation significantly more than expected in regards to availability of these vegetation types (Chi-square test; Table 12). In addition, cowbirds were found in farm fields and saltcedar vegetation significantly more than expected while Bell's vireos were found in saltcedar significantly less than expected. All 3 species were found in desert bajada and desert wash significantly less than expected.

# 433 # 3 3 3 # 4 3 3 4 4 4, 3 3 4, 4, 3 4 ---. ...--, ---. --- CD r- ,-1 r- ,--4 4.0 r- CD ,--4 ,--t r- r- ,T OD r- .4r 01 OD nS) LO 01 40 4.(1 ,-A OD 00 ,--f r- i..0 In Cs1 , T CO ,--1 cp ,e, ,--1 OD r- d> c> ---f c‘i CD r- ,-( ar 4.0 Cl (N r-Œ CD 10 C:) CD CD r•1 r-i CD CD CD ',7' 0 CD CD 0 CD ,..-4 cr CD CD •,7 CD CD CD DD • • • , • • • • • • • • • • • • • • • • • • • • CD CD (:) CD DC 0 CD CD CD CD CD CD 0 CD C:D CD CD CD CD 0 0 (D CD DD

. . . . (:> ,-4 CO lS) LO a- r c> in tr) (n,1 ,-i Cr) (Y) (Z) (=> 01 .71' ,--1 C.1 W() ('.! rn c> (7) CD ,-.1 CO CD OD CD (,) CNJ ,--I CD CD 41) Cil ar c> r- rn ,---1 c> ,:r 1.0 C:3 , T (D (:) CD 04 CD

CO (`,1 C•1 Lt) n.0

Cl> (•1 .1" 0.4 CD CD CO CD (,) (•1 01 01 OD \4) C, J szr r-A CX) No-)m(NDcn CD CD CD C,) CD CD C:> CX) (•1 c> Ln c:> Lr) c> CNI r-4 CD VI) I-- CD c> c> c> c> c‘,1 c> c> c> c) DcCDDQr.- 0 DDCDDDDD CD - • - • • • • • • • • • • • • - • • • • • • • • • • • • • • • DccDDcc, c> c> c> c> c> c> c> c> c> c> c> c> c> c> c> c> c> c> c> c>

o CX) Cl• in CD lX) cr f-- r-4 co cf) ar tr) c> 'O r-q CX) •zr 1.0 DLO a' a) ko co is1 (LULOrH OD 4.0 4.0 OD r- ks> cn op Ln LUHOLDP) r- LULO CL CD 0 0 CD C> CD 0 Ca 0 CD r--1 CD 0 (D 0 CD Cq CD CD 0 CD CD 0 CD CD • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • r4.1 C> CD CD 0 C) CD 0 0 CD CD 0 CD CD 0 CD C> CD C> C> CD CD C> C) C> C> CD CD

m m m 3 "0 3 '0 3 '0 O 0 0 —4 M —1 m .--i mi —4 —4 —4 —4 —I —4 -H -ci -H -(5 -H -o 3 o 3 o 3 o ---- .4 ro o ------. 4 Ti o ----. 4 ro o 'O U) 4 3 P 0 0) 3 1-n 'OU) ci 3 O ro 0 $.4 1) O(U "0 i 1HW 0 CO Tl O C) 0 a) ni .4.-) 0 3 0 0 a) as 4-) 0 3 a) 0 ell 3 a) P 4-) 0 ai 3 41) P 4-) "0 ni 3 a) 4-) mi z 4-) 3 tyi -HG) 3 (U 4 -) 3 tz)-) -HC) 3 M -P 3 01 -H . -I 0 0 P 3 , U .-I 0 0 S-1 3 .. U -4 0 0 S-1 3 . a) a) o E a) ci) rr4J a) as 4--) a) o E a) cn izr -u a) oc 4--1 a) o a) cf) ' .. - r--14-) m P P P P (I) n-1 a) ,C - 1-1 4-1 Cd .-1 )-i S-1 4.-I (1) .--1 al •C 1-1 4-4 Cd 3-1 )-I 4-1 S-4 CO 4 -4 (U 0 a) S-4 (d (11 (U a) M 0., 4-) ra 0 0 P or5 ms al (ll ni sa, 4-) ai 0 41) P al al Ri a) g 01 C.) c1 ‹ GLI 01 Z Z UO 0 g CO 0 g g r-.4 cil Z Z ci) C) g CO (-) Q g DL1 01 Z Z

a) -H o a) (1

ra 0 C.) -H CD 4-4 -H 44 4 44 4 44 4 3 3 -I. # 43 a) ------— --- —--- ..-- ..------— ------..------LO HI OD HI in rn 0) r- .Q. CQ rn Q. U.) r- op cs) 01 CQ 01 CQ HI CD CQ 0.1 ...0 CD Cil Qr CQ TH OD Hi cr CD Ul ,-I CV Pl CD CD cr CD 0 CD 0..) CD CQ CD VD 0 CD CD CD ,-I CQ ...... • • • • • • (() cf) CD CD CD )CCDCD CD CD CD 0. CD 0 CD CD CD CD CD CD CD Cl (1) a) ai 0 , • • . . . - . • • • • • • • • • 0.4 co o c) op c) cr c:4 CD rn 01 o r- 4-4 4.4) cp cs CQ in LU 0 CD CNI C> DA CD C> HI QT CD 0) C> HI CD Hi l0 LU (NC) CD O.) CD CD CD (C-D r-4 CD Ln CD CD 0 DD 1--1 -1--) 4-1 a) S-1 ...... • 4-) 0 CD 0 CD CD C> CD CD CD 0 CD CD CD CD CD CD C> 0 CD al a --...• .--- — .....-.. .-... .--- .....----- ..--. --..., ..-- ...-- .----.... 000 0 -H -H - H g4 -1--) «3(0 -1-) Ln ID -H Ti r- (1) ko r- 4-1 0.4 CDŒ)

0)0) • _Q -c 0 CQ Ul t-i TC> CO 1-1 Ln 01 70 Cf) ,i CQ .17D CD C> CD C) 0) CD u) al CD CD cp r- H1 'Gr CD 0 QI in CO CD CD CD Q C...) CD CD rn op co c) -H -H (ND CD CD in CD CD CD CD 0.) CD ,-I CD G> CD CD 0 0 CD CD CD CD CD CD CD C> CD CD C> 0) a a a)

CD SH -H a) 4-) ca r- 0:10 ,T Ln CD rH CO Cf) Q Lf) Hl CO II) lD Ln rH 00 Lf) k.0 rH co Ln rn 14> Ln r-1 CX> -H-H O rH (ND C> 0 C> C> CD Cs/ C> CD Hi Hi CD 10 0 CD 0 CD CQ CD U) • • • • . • • . • • • • • • • • • • • • • • • • a) 0 DC) CD CD 0 CD CD CD 1:D CD C> CD CD 0 C> C> C> (D CD C> 0 ,H 0 0 -H ca -H -H (4-i -4-) 4-) 0J c.)o a) a) cn u) 3 1J 3 TI o ç o U) a) ow (1) 4-) ,H ro .--t cc) 0 -I 4---i 4-4 -4 ,Q -0 a) - H 7, - H 'Cl 00 Lo 0 0 3 u) (D cil - '--__,ç: -CI 0 ---- .0 -cl 0 44-4:iI-4 Ti ci) 0 3 )-I 'Ci CO ç: $-1 • C) .-1 414 0 M Pt i a) 0 cd 7:1 l l-W C) C) C) (a. 4-) 0 3 0.) 0 a) 171 -1.> 0 4) 0 cl) ai -p 4a) , cri Ti (71 3 1.) $.4 4--) -0 (Q 3 a) S-I 4--) -0 ni v W 3 (Q ç:4-)4-4 3 0\ -HQ) 3 ro 4-4 3 Cs -HQ) 3 04 / to U -4 'd0 3 .0 c.) H -0 0 14 3 .0 0 U _oH _ OH La 4-3 0 a) ro 4..) a) 0 E a) cf) V 4- -) (1.1 Ri -1-) a) 0 a) cc tr -1-) 4i4-)r1-4 a) 4 1-4 4-4 CO S.-1 3-I S.4 S.-I U) .--1 CD .. 1--1 -ta CO l-i S-i S.-i S-4 CO ,-I 414 Hi $4 ni a 4-4 41 0 a) 4-1 M M rd 41) rd a 4-4 ai 0 Cll S441 cd ni al a) ai a4 (1) a) uo 40 g4 44 0 q ‹4 4, cci X X ci) C) <4 Ca C.) (21 g4 44 01 X X u) C) 00 4-) 0,--) 0 0 (/) 4_) -H '0 4 -u a) a a-H (15 7J ro -p Cl TI 0 u) -ci cj 0 (1) a) -,-1 $.4 (a -0 ,Q ..r a) 4--) a) a) ° a) (Q :a 4-) 4-1 T5 0)0 3 0 54 -H .0 C) 0 10 (1) a) '4- 1 $"4 4-4 I 0 H 3 0 OH 0 (i) 3 H X _Q o 4-1 0 0 cx1 o , () 6 ., ce -H ,, 0 a) .0 0 >4 ru .0 0 124 79 Nest-Site Selection Models Vegetation was sampled at 147 Bell's vireo nests, 59 yellow-breasted chat nests, and 208 random sites. Univariate analyses indicated significant differences between Bell's vireo nest sites and random sites for several variables. Vireo nests were: (1) farther from openings, (2) closer to mature trees, and (3) had lower horizontal dispersion of vegetation. Vireo nest patches had: (1) higher percent canopy and ground cover (primarily leaf litter cover), (2) higher percent foliage cover in the low (0.6-4.5 m) and middle (4.6-7.5 m) canopy layer, (3) higher species richness in the low canopy layer, (4) higher percent saltcedar and mesquite foliage cover, and (5) higher percent saltcedar and mesquite foliage cover in the middle canopy layer (Appendix F). The 3 vegetation characteristics that best differentiated between vireo nest sites and random sites in forward stepwise logistic regression modeling were: (1) mean distance to closest tree, (2) foliage cover in the low canopy layer, and (3) saltcedar foliage cover in the middle canopy layer (Table 12). Inspection of the log odds ratio for the variable mean distance to closest tree indicates that the distance was greater at random sites than at Bell's vireo nest sites (log odds ratio >1; Table 12). Plots with mean distance to tree between 7.8 and 12.4 m were 1.5 times more likely to be 80 Table 12. Bell's vireo nest-site model fitted with variables that met criteria for forced entry into the model.

Variable Stf CV'

Mean distance to closest tree (0-7.7 m)e 7.8-12.4 m 0.387 0.129 33.30 1.47 >12.4 m 1.091 0.129 11.85 2.98 Foliage cover 0.6-4.5 m high (0-80%)e 90-100% -1.321 0.158 11.95 0.27 >100% -1.921 0.130 6.76 0.15 Saltcedar cover 4.5-7.5 m high (0%)e 10-100% -0.566 0.095 16.69 0.57 Constant 1.149 0.124 10.83

Correctly classified: f Nests: 78% Random: 58% (n = 147) (n = 208)

• p is the mean coefficient of the 10 models that were generated by subsampling from the random plots data set. b Standard deviation of model coefficient. ' Coefficient of variation = 100 X (SD/mean). d T is the log odds ratio = Exp (mean coefficient). This value is interpreted as the likelihood that a plot is a random site rather than a nest site. e Reference category. Other categories are interpreted relative to this one. f Percent of nest sites and random sites correctly classified with model.

random sites than plots with distances <7.8 m; plots with mean distances >12.4 m were 3 times more likely to be random sites than plots with mean distances <7.8 m. The log odds ratio for foliage cover in the low canopy layer is <1, demonstrating that as foliage cover increased, plots were less likely to be random sites. Plots with 90-100% foliage cover were 3.7 times less likely to be random sites than 81 plots with <90% foliage cover; plots with >100% foliage cover were 6.7 times less likely to be random sites than plots with <90% foliage cover. The log odds ratio for percent saltcedar cover in the middle canopy layer was also <1. Plots with >0% saltcedar cover were 1.8 times less likely to be random sites than plots with 0% saltcedar cover. The CV was low for all variables (6.76 - 33.30), indicating that the model coefficients and log odds ratios were similar among the 10 fitted logistic regression models. Classification rates were high for nest sites (78%), but were lower for random sites (58%). The low classification rate of random sites is not unexpected because random sites are potential nest sites. Univariate analyses also indicated significant differences between yellow-breasted chat nests and random sites for several variables. Chat nests were: (1) farther from openings, (2) closer to shrubs and mature trees, (3) had lower dispersion of mature trees. Chat nest patches had: (1) higher percent ground cover (principally leaf litter cover), (2) higher percent foliage cover in the low and middle canopy layer, (3) higher species richness in the middle canopy layer, (4) higher percent shrub cover (especially of Baccharis spp.), (5) higher saltcedar foliage cover, (6) lower mesquite foliage cover, and (7) higher 82 percent saitcedar cover in the middle canopy layer (Appendix G). The 5 vegetation characteristics that best differentiated between chat nest sites and random sites in forward stepwise logistic regression modeling were: (1) mean distance to closest shrub, (2) foliage cover in the low canopy layer, (3) total saltcedar foliage cover, (4) saltcedar foliage cover in the middle canopy layer, and (5) leaf litter ground cover (Table 13). Inspection of the log odds ratio for the variable mean distance to closest shrub indicates that the distance was greater at random sites than at yellow-breasted chat nest sites (log odds ratio >1; Table 13). Plots with mean distance to shrub between 0.6 and 1.5 m were 2 times more likely to be random sites than plots with distances <0.6 m; plots with mean distances between 1.6 and 2.8 m were 11 times more likely to be random sites than plots with distances <0.6 m; and plots with mean distances >2.8 m were 27 times more likely to be random sites than plots with distances <0.6m. The log odds ratio for foliage cover in the low canopy layer is <1, demonstrating that as foliage cover increased, plots were less likely to be random sites. Plots with 90-100% foliage cover were 2 times less likely to be random sites, and plots with >100% foliage cover were 4.3 times less likely to be random sites than plots with 0-80% foliage cover. The log odds ratio for percent saltcedar 83 Table 13. Yellow-breasted chat nest-site model fitted with variables that met criteria for forced entry into the model.

Variable spb CV

Mean distance to closest shrub (0-0.5 m)e 0.6-1.5 m 0.825 0.513 62.19 2.28 1.6-2.8 in 2.399 0.464 19.34 11.01 >2.8 in 3.312 0.454 13.70 27.44 Foliage cover 0.6-4.5 in high (0-80%)e 90-100% -0.706 0.318 44.96 0.49 >100% -1.484 0.303 20.39 0.23 Saltcedar foliage cover (0%)e 10-40% 0.368 0.420 114.34 1.44 >40% -1.125 0.476 42.32 0.32 Saltcedar cover 4.6-7.5 in high (0%)e 10-100% -0.938 0.527 56.20 0.39 Leaf litter ground cover (<100%)e 100% -1.002 0.335 33.41 0.37 Constant 1.188 0.456 38.40

Correctly classified: f Nests: 81% Random: 76% (n = 59) (n = 208)

• p is the mean coefficient of the 10 models that were generated by subsampling from random plots data set. b Standard deviation of model coefficient. ' Coefficient of variation = 100 X (SD/mean). • T is the log odds ratio = Exp ( mean coefficient). This value is interpreted as the likelihood that a plot is a random site rather than a nest site. e Reference category. Other categories are interpreted relative to this one. f Percent of nest sites and random sites correctly classified with model. foliage cover within the nest patch was variable. Plots with 10-40% saltcedar foliage cover were 1.5 times more likely to be random sites than plots with <10% saltcedar 84 cover; plots with >40% saltcedar cover were 3 times less likely to be random sites than plots with <10% saltcedar cover. Plots with 10-100% saltcedar cover in the middle canopy layer were 2.5 times less likely to be random sites than plots with 0% saltcedar cover. Plots with 100% leaf litter ground cover were 2.7 times less likely to be random sites than plots with <100% leaf litter cover. The CV was low (13.7) for some variables, but was highly variable. Classification rates were high (76-81%), but the classification rate of random sites was again lower than that of nest sites.

Parasitism Models Nest contents were determined at 196 nest sites of the Bell's vireo, blue grosbeak, common yellowthroat, and yellow-breasted chat. One hundred and forty nine nests were successfully parasitized and 47 were unparasitized or inappropriately parasitized. Univariate analyses indicated significant differences between successfully parasitized and unparasitized nests for several variables. Successfully parasitized nests had: (1) less ground cover below the nest, (2) less canopy cover above the nest, (3) were farther from shrubs, (4) closer to mature trees, and (5) closer to a mature cottonwood or willow tree. Successfully parasitized nest patches had: (1) lower percent ground 85 cover, (2) higher percent canopy cover, (3) lower percent foliage cover in the herbaceous layer, (4) higher percent foliage cover in the middle and upper (>7.5 m) canopy layer, (5) higher percent cottonwood-willow foliage cover, (6) lower percent foliage shrub cover 0-0.6 m high, and (7) lower overall shrub foliage cover (Appendix H). The 5 vegetation characteristics that best differentiated between successfully parasitized and unparasitized nests were: (1) percent ground cover below nest, (2) percent canopy cover above nest, (3) mean distance to closest shrub, (4) shrub foliage cover 0-0.6 m high, and (5) foliage cover in the high canopy layer (Table 14). Inspection of the log odds ratio for the variables percent ground cover below nest and percent canopy cover above nest indicates that nests with greater cover are less likely to be parasitized (Table 14). Nests with 100% ground cover are 3 times less likely to be parasitized than nests with <100% ground cover. Similarly, nests with >60% canopy cover are 3 times less likely to be parasitized than nests with <60% canopy cover. The log odds ratio for mean distance to closest shrub was >1, indicating that as this distance increased, likelihood of parasitism also increased. Nests at which mean distance to closest shrub was >2.2 m were approximately 3 times more likely to be parasitized than nests with mean distance to closest shrub <0.5 m. In

86 Table 14. General parasitism model fitted with variables that met criteria for forced entry into the model.'

Variable SD: CV'

% ground cover below nest (<100%)' -1.192100% 0.299 25.09 0.30 % canopy cover above nest (0-60%)" >60% -1.130 0.302 26.76 0.32 Mean distance to Closest shrub (0-0.5 m)f 0.6-2.2 in 0.189 0.305 160.94 1.21 >2.2 in 1.295 0.534 41.25 3.65 Shrub cover 0-0.6 in high (0%) f >0% -0.778 0.272 35.01 0.46 Foliage cover >7.5 in high (0%) f >0% 0.731 0.322 44.06 2.08 Constant 1.323 0.297 22.36

Correctly classified:g Unparasitized: 69% Parasitized: 72% (n = 47) (n = 149)

'Model based on differences between all unparasitized and parasitized nests of Bell's vireos, blue grosbeaks, common yellowthroats, and yellow-breasted chats. Dp is the mean coefficient of the 10 models that were generated by subsampling from the successfully parasitized nest data set. ' Standard deviation of model coefficient. d Coefficient of variation = 100 X (SD/mean). e is the log odds ratio = Exp (mean coefficient). This value is interpreted as the likelihood that a plot is a successfully parasitized nest rather than an unparasitized nest. f Reference category. Other categories are interpreted relative to this one. g Percent of unparasitized and successfully parasitized nests correctly classified with model. 87 Inspection of the log odds ratio for the variables percent ground cover below nest and percent canopy cover above nest indicates that nests with greater cover are less likely to be parasitized (Table 14). Nests with 100% ground cover are 3 times less likely to be parasitized than nests with <100% ground cover. Similarly, nests with >60% canopy cover are 3 times less likely to be parasitized than nests with <60% canopy cover. The log odds ratio for mean distance to closest shrub was >1, indicating that as this distance increased, likelihood of parasitism also increased. Nests at which mean distance to closest shrub was >2.2 m were approximately 3 times more likely to be parasitized than nests with mean distance to closest shrub <0.5 m. In addition, nests with >0% shrub cover between 0 and 0.6 m high were 2 times less likely to be parasitized, and nests with >0% foliage cover in the high canopy layer were 2 times more likely to be parasitized. The CV was fairly low (22.36-44.06) for all variables except mean distance to closest shrub (0.6-2.2 m category), indicating that the model coefficients and log odds ratio were similar for the 10 fitted logistic regression models for most variables. Classification rates ranged from 69% for unparasitized nests to 72% for successfully parasitized nests. 88 Bell's vireo and yellow-breasted chat parasitism models fit with these same 5 variables resulted in similar classification rates, but the magnitude and direction of some coefficients varied. Species-specific models were built to determine vegetation characteristics of successfully parasitized and unparasitized nests of these 2 species. Univariate analyses indicated significant differences between successfully parasitized and unparasitized Bell's vireo nests. Successfully parasitized vireo nests had: (1) higher lateral vegetation concealment, (2) lower dispersion of vegetation, (3) were farther from shrubs, and (4) closer to a mature cottonwood or willow tree. Successfully parasitized vireo nest patches had: (1) higher percent foliage cover in the middle canopy layer, (2) higher percent cottonwood-willow foliage cover, (3) lower percent shrub foliage cover 0-0.6 m high, and (4) lower percent shrub foliage cover (Appendix I). The 3 vegetation characteristics that best differentiated between successfully parasitized and unparasitized vireo nests were: (1) mean distance to closest shrub, (2) distance to a mature cottonwood or willow tree, and (3) shrub foliage cover (Table 15). 89 Table 15. Bell's vireo parasitism model fitted with variables that met criteria for forced entry into the model.

Variable SD CV=

Mean distance to closest shrub (0-1.6 m)e 1.7-3.0 in 0.320 0.552 172.18 1.38 >3.0 in 1.637 0.394 24.10 5.14 Distance to cottonwood or willow (0-5 m)e 5.1-12 m 1.244 0.327 26.32 3.47 12.1-30 in -0.141 0.435 309.45 0.87 >30 in -1.599 0.679 42.46 0.20 Shrub foliage cover (0%)e 10-50% -0.135 0.351 259.81 0.87 60-100% -0.767 0.691 90.12 0.46 Constant -0.003 0.510 16990.00 Correctly classified: Unparasitized: 67% Parasitized: 74% (n = 26) (n = 102)

• p is the mean coefficient of the 10 models that were generated by subsampling from the successfully parasitized nest data set. • Standard deviation of model coefficient. • Coefficient of variation = 100 X (SD/mean). d T is the log odds ratio = Exp (mean coefficient). This value is interpreted as the likelihood that a plot is a successfully parasitized nest rather than an unparasitized nest. e Reference category. Other categories are interpreted relative to this one. f Percent of unparasitized and successfully parasitized nests correctly classified with this model.

Inspection of the log odds ratio for the variable mean distance to closest shrub indicates that the distance was greater at successfully parasitized nests. Vireo nests with mean distance to shrub >3.0 in were 5 times more likely to be successfully parasitized (Table 15). Nests with the closest 90 mature cottonwood or willow tree between 5 and 12 m were 3 times more likely to be parasitized than nests with a mature cottonwood or willow within 5 meters. However, nests with the closest mature cottonwood or willow 12-30 m away had a slightly lower likelihood of being parasitized than nests with the closest cottonwood or willow within 5 m. Nests with the closest cottonwood or willow >30 meters away were 5 times less likely to be parasitized than nests with the closest cottonwood or willow within 5 m. The log odds ratio for shrub foliage cover indicated that nests with >50% shrub foliage cover were 2 times less likely to be parasitized. The CV ranged from 24.10-309.45 for these variables, indicating that the model coefficients and log odds ratios were not consistent for some of the categories of these variables. Important information can still be gained from those variable categories that had low CVs. Sixty seven percent of unparasitized nests and 74% of successfully parasitized nests were correctly classified with this model. Univariate analyses indicated significant differences between successfully parasitized and unparasitized yellow- breasted chat nests. Successfully parasitized chat nests had: (1) lower percent ground cover below the nest, (2) lower percent cover between the nest and the ground, (3) lower dispersion of vegetation, (4) were closer to mature trees, and (5) closer to a mature cottonwood or willow tree. 91 Successfully parasitized chat nest patches had: (1) lower percent ground cover, (2) lower percent foliage cover in the herbaceous layer, (3) higher percent foliage cover in the high canopy layer, (4) higher percent arrowweed foliage cover, and (5) lower percent live herbaceous cover (Appendix J). The 3 vegetation characteristics that best differentiated between successfully parasitized and unparasitized chat nests were: (1) percent vegetation cover between the nest and the ground, (2) percent foliage cover in the high canopy layer, and (3) percent live herbaceous cover (Table 16). Inspection of the log odds ratio indicates that successful parasitism is 11 times less likely at chat nests with >40% concealment below the nest (Table 16). The log odds ratio for foliage cover in the high canopy layer is >1, indicating that as foliage cover increased, nests were more likely to be successfully parasitized. Successful parasitism is 10 times more likely at nests with >0% foliage cover in the high canopy layer. However, parasitism is 11 times less likely at nests with >0% live herbaceous cover below nest. The standard errors for the model coefficients are low (0.617-1.164), indicating that the model is numerically stable. Classification rates for this model were high, ranging from 77% for unparasitized nests and 81% for successfully parasitized nests. 92 Table 16. Yellow-breasted chat parasitism model fitted with variables that met criteria for forced entry into the model.'

Variable SE' Waldo P'

% vegetation cover under nest (0-40%)g >40% -2.432 0.870 7.821 0.005 0.09 Foliage cover >7.5 m high (0%)° >0% 2.344 1.164 4.055 0.044 10.42 Live herbaceous cover (0%)g >0% -2.486 0.873 8.105 0.004 0.08 Constant 1.951 0.617 9,997 0.002

Correctly classified: h Unparasitized: 77% Parasitized: 81% (n = 17) (n = 36)

' No subsampling was performed for forced entry logistic regression model. b p is the model coefficient. ' Standard error of model coefficient. d Wald statistic. ' Significance value. f T is the log odds ratio = Exp (mean coefficient). This value is interpreted as the likelihood that a plot is a successfully parasitized nest rather than an unparasitized nest. g Reference category. Other categories are interpreted relative to this one. h Percent of nests correctly classified with model.

DISCUSSION Parasitism Rates and Cowbird Abundance Parasitism rates were higher in the lower Colorado River valley (40-90%) than in protected natural areas along the Colorado River in Grand Canyon National Park (Brown 1994). Brown (1994) reported parasitism rates of 7% for the 93 Bell's vireo (n = 57), 60% for the blue grosbeak (n = 5), 56% for the common yellowthroat (n = 9), and 11% for the yellow-breasted chat (n = 37). However, high levels of parasitism have been documented for other species, including the Kirtiand's warbler (Mayfield 1977), black-capped vireo (Grzybowski 1990), and least Bell's vireo (Franzreb 1987), and the cowbird has been implicated in the near extirpation of these species. Martin (1992) proposed that cowbirds respond to cumulative host densities rather than the density of particular host species, resulting in higher parasitism rates in host-rich areas. Moreover, Lowther and Johnston (1978) found the highest densities of cowbirds in areas with the highest cumulative host densities. In the lower Colorado River valley, cowbird abundance increased as cumulative host abundance increased, suggesting that cowbirds are attracted to areas where host species are concentrated. Parasitism rates on the Arizona Bell's vireo and yellow-breasted chat decreased as cumulative host abundance increased. An inverse relationship between host- specific density and parasitism rates has been documented in several studies (Zimmerman 1983, Weatherhead 1989). However, the effect of cumulative host density on species- specific parasitism rates is not well understood. For example, Fretwell (1977) reported higher parasitism rates on 94 dickcissels in areas with high red-winged blackbird densities. Clark and Robertson (1979) reported lower parasitism rates on yellow warblers in areas with high densities of blackbird nests.

Host Nesting Success Estimates of nesting success for the lower Colorado River valley vireo and chat populations were lower than those reported in other studies. Martin (1992) reviewed published reports of nesting success for 17 species, and the average nesting success calculated by the Mayfield method was 42%. Daily mortality rates of Bell's vireo and yellow- breasted chat nests were highest during the nestling stage. This is consistent with the findings of other studies and suggests increased nest vulnerability during the nestling phase (Mayfield 1975). Adults visit the nest more frequently during the nestling stage, and either the adults or begging juveniles may alert predators to the presence of the nest (Gates and Geysel 1978). Vireos and chats also experienced high rates of partial clutch loss, frequently the result of egg removal by cowbirds. Most egg removal by cowbirds occurred during the egg-laying stage (pers. obs.). Cowbird parasitism frequently lowers the reproductive success of open-cup nesters. Rothstein (1975) reported a 40-98% decrease in nest success in parasitized nests of 4 95 common host species. In the lower Colorado River valley, the Arizona Bell's vireo and yellow-breasted chat had lower nesting success and fledged fewer young from parasitized nests. However, yellow-breasted chats were better able to withstand the pressures of parasitism. Parasitized yellow- breasted chats were 3.5 times more likely to fledge host young than parasitized Bell's vireos. This disparity was even greater if only successfully parasitized nests were examined. Less than 1% of successfully parasitized vireo nests fledged young, compared to 14% in the yellow-breasted chat. The yellow-breasted chat is a large warbler, comparable in size to the brown-headed cowbird (6.25 versus 6.50 inches in length; Robbins et al. 1983:289, 301). The incubation period for chats is 11 days, similar to that of the cowbird. On the other hand, chat nestlings grow rapidly and leave the nest as early as 8 days after hatching (Ehrlich et al. 1988:548). Therefore, chat nestlings are better able to compete with aggressive cowbird nestlings. Bell's vireos are small birds (4.25 inches in length; Robbins et al. 1983:265), with a longer incubation period than cowbirds (14 versus 10 - 13 days, respectively) and a brooding period comparable in length (11 days; Ehrlich et al. 1988:492, 616). Cowbird young have a clear advantage over vireo young. 96 The mean number of host eggs, nestlings, and fledglings was significantly lower in parasitized nests of both species. The Bell's vireo experienced high losses in parasitized nests at all stages of the nesting cycle, whereas the yellow-breasted chat experienced most losses during the incubation stage. Chat nestlings had a high probability of fledging from parasitized nests that were not depredated or abandoned. Marvil and Cruz (1989) also found significantly lower clutch sizes and fledging success in parasitized solitary vireo nests. There was a high frequency of multiple parasitism in my study area: 37.6% of vireo and chat nests (combined) at which nest contents were known received >1 cowbird egg. Mayfield (1965) suggested that female cowbirds may be more likely to deposit eggs in previously parasitized nests that were not abandoned due to proven host acceptance of the eggs. Alternatively, multiple parasitism may be indicative of high cowbird densities. Cowbirds are known to lay up to 40 eggs per season (Scott and Ankney 1980). In the lower Colorado River valley, cumulative host abundance was only 3.5 times greater than cowbird abundance. Therefore, it is not surprising that many nests were multiply parasitized. Singly parasitized nests had a higher probability than multiply parasitized nests of fledging host young. However, the number of host young fledging from singly and multiply 9 7 parasitized nests was not significantly different. It appears that 1 cowbird egg in the nests of these host species can substantially lower reproductive success. Other studies confirm that 1 cowbird egg is enough to cause no host young to fledge from parasitized nests of small host species, such as the vireo (Mumford 1952). Even a small effect of parasitism on nesting success could be detrimental to populations experiencing high predation (Brittingham and Temple 1983). Predation rates were high for both vireos and chats in the lower Colorado River valley. Predation was the primary mortality agent for the yellow-breasted chat, as is true for most songbird species (Martin 1992) and other chat populations (Thomson and Nolan 1973). However, nest abandonment was the primary cause of nest failure for the Bell's vireo, and parasitism was responsible for most cases of desertion. Barlow (1962) also reported parasitism as the primary mortality agent affecting nesting success of Bell's vireos. Cowbird eggs were discovered on the ground below Bell's vireo nests. There are 2 possible explanations for this: 1) vireos may have recognized cowbird eggs as foreign and ejected them from the nest or 2) cowbirds may have ejected eggs of conspecifics from multiply parasitized nests. The first explanation is probably unlikely. Bell's vireos are considered acceptors of cowbird eggs (Ehrlich et al. 98 1988:492) and may be unable to eject cowbird eggs due to the small size of their bills (Rothstein 1990). The second explanation is the most probable, especially considering the high concentration of cowbirds in the lower Colorado River valley. Due to the low probability of fledging more than 1 cowbird from a multiply parasitized nest, it would be advantageous for female cowbirds to remove the eggs of conspecifics, in addition to or instead of removing host eggs. On several occasions, vireos and chats buried cowbird eggs within the nest lining (pers. obs.). Bell's vireos also displayed nest-guarding or preincubation behavior (sitting tightly on an incomplete clutch) in response to an intruder's presence. Such behavior may prevent parasites from entering a nest, and has been observed in other studies of small host species (Burgham and Picman 1989, Hobson and Sealy 1989, Uyehara and Narins 1995). Bell's vireos and yellow-breasted chats also abandoned nests in response to parasitism. Most nests were abandoned during the early egg- laying stage, before high parental investment in the nesting attempt. Nest desertion is the most common rejection behavior in small host species (Friedmann 1963, Graham 1988), and may be a function of nest-site availability (Petit 1991). Nest abandonment delays breeding, which for double-brooded 99 species may lower total reproductive success by limiting a female to rearing 1 brood. Therefore, nest-abandonment may be a better strategy for smaller host species that have higher fitness by raising 1 unparasitized brood rather than 2 parasitized broods (Petit 1991). This appears to be the case for the Bell's vireo, and may explain the high desertion rate seen in this species. The yellow-breasted chat populations in the lower Colorado River valley may be able to replace themselves despite high observed parasitism rates. Mayfield (1977) noted that some species may be able to withstand high parasitism rates by raising multiple broods. Assuming annual adult and juvenile mortality rates of 48 and 70%, respectively, these yellow-breasted chat populations could withstand current parasitism rates if each female successfully raised 2 broods (May and Robinson 1985, eq.4). Ehrlich et al. (1988:548) reported that yellow-breasted chats can successfully raise 2 broods per season. The high parasitism rates and low reproductive success of Bell's vireos in the lower Colorado River valley suggest that these populations may be unable to replace themselves. These populations could withstand current parasitism rates if each female successfully raised 3 broods. This is unlikely considering the high rate of nest desertion and failure in parasitized Bell's vireo nests. In addition, 100 neither Rosenberg et ai. (1991:283) nor ever observed Bell's vireos raising 3 broods in the lower Colorado River valley. Therefore, these Bell's vireo populations may not be able to maintain themselves without recruitment from other populations. The Bell's vireo is common in isolated pockets in the lower Colorado River valley despite high parasitism rates. One possible explanation is the length of time it takes to drive robust populations to extinction even under high parasitism pressure (Trail and Baptista 1993). Another explanation is that migrating and dispersing birds are attracted to riparian areas due to the high vegetative diversity, high insect biomass, and cool microclimate. Predators and parasites are also attracted to these areas, forming an ecological trap for breeding birds (Gates and Gysel 1978, Gates and Giffen 1991). A third explanation is that my study design resulted in inflated parasitism levels. Rothstein (1994) stressed the importance of edge effect in explaining the persistence of Bell's vireos and willow flycatchers along the lower Colorado River despite the high parasitism rates reported by early researchers in the region (Brown 1903). Rothstein (1994) suggested that these parasitism rates applied to the nests most easily located, i.e., on the edge of dense riparian forests. However, the landscape of the lower 101 Colorado River valley has changed dramatically since 1900: the riparian corridor is narrow and fragmented, with little interior forest remaining. My nest searching activities were extensive, covering a large proportion of the remaining area in which Bell's vireos were likely to be found. Therefore, observed parasitism rates may not differ much from actual parasitism rates. Finally, the possibility remains that each vireo pair, through multiple renesting and brooding efforts, fledged enough young to maintain the population. On the other hand, the lower Colorado River Bell's vireos may form sink populations that are maintained by continual immigration of individuals from more productive, source areas (Pulliam 1988). Other studies have confirmed the presence of sink populations of neotropical migrants in fragmented woodland landscapes (Gibbs and Faaborg 1990, Robinson 1992). Howe et al. (1991) suggest that a large proportion of a species population might reside and reproduce in sink areas. These sink areas protect the persistence of metapopulations by buffering against stochastic extinction (Goodman 1987) and contributing to a more diverse gene pool (Lande and Barrowclough 1987). Therefore, the identification and protection of both source and sink populations may be important to the long-term survival of Arizona Bell's vireo populations. 102 Critical parasitism levels obtained from May and Robinson's (1985) equation are sensitive to variations in the mortality and fecundity values (Trail and Baptista 1993). The mortality values I used were estimates, but were similar to those found in the literature for other populations (Nolan 1978, Trail and Baptista 1993, Martin 1995). In addition, I assumed that parasitized and unparasitized females raised the same number of broods per year. Therefore, the critical parasitism values I obtained are useful only as rough estimates.

Brown-headed Cowbird Reproductive Success Fledging success, defined as the probability that a cowbird egg present at the start of incubation would fledge, was low (18%). This is probably a high estimate because many cowbird eggs were abandoned or ejected from the nest before incubation began. In addition, newly deposited cowbird eggs were found in previously abandoned nests and in nests that already fledged. Other researchers have observed cowbirds dumping their eggs in inappropriate places when disturbed during egg laying (Sealy et al. 1995). If all cowbird eggs were considered in calculations, fledging success would probably approach the level (3%) reported by Scott and Ankney (1980). 103 Daily mortality rates were frequently higher during the incubation stage than during the brooding stage. High cowbird losses during the incubation stage were probably the result of egg rejection by hosts (nest abandonment, egg burial, egg ejection), or increased nest predation due to aggressive anti-parasite defense that alerted predators to nest location. However, parasitized nests during my study were not more likely to be depredated than unparasitized nests. The low fledging success of cowbird eggs was probably compensated for by high egg production in breeding female cowbirds. Cowbird fledging success was equivalent in Bell's vireo and yellow-breasted chat nests. However, vireos were more likely to abandon nests parasitized during the egg-laying stage. Therefore, yellow-breasted chats were better cowbird hosts than Bell's vireos.

Bird-Vegetation Associations: Macrohabitat Scale Bell's vireos, blue grosbeaks, common yellowthroats, and yellow-breasted chats were significantly associated with cottonwood-willow and mesquite vegetation in the lower Colorado River valley. On the other hand, monotypic saltcedar was not an important vegetation type for most neotropical migratory songbirds breeding in the lower Colorado River valley (Lynn et al. 1996). In the southwest, 104 breeding birds frequently reach their highest densities riparian areas due to the high vegetative diversity, high invertebrate biomass, and high canopy cover that provides shelter from the extreme heat of the surrounding desert environment (Carothers et al. 1974, Rosenberg et al. 1982, Hunter et al. 1988). Riparian areas comprised of monotypic saltcedar stands may not provide the cool microclimate (Anderson et al. 1977) and high insect abundance (Cohan et al. 1978) provided by cottonwood or willow-dominated riparian areas in the lower Colorado River valley. Hubbard (1977) documented the importance of cottonwood- willow riparian forest for breeding neotropical migrants. Other studies have also documented a correlation between the decline in native riparian vegetation and the decline in neotropical migratory songbird populations (Klebenow and Oakleaf 1984). The lower Colorado River valley appears to be another example of this trend (Rosenberg et al. 1991). The brown-headed cowbird was positively associated with cottonwood-willow riparian forests, mesquite woodlands, and saltcedar in the lower Colorado River valley. Cowbirds were most abundant in cottonwood-willow forests, probably due to the high concentration of hosts in these areas. Cowbird use of monotypic saltcedar stands may be attributed partly to the presence of some host species within this vegetation type (Lynn et al. 1996), and to the use of saltcedar as 105 roosting and foraging sites. The low observed frequency of cowbirds in farm fields was probably a time-specific phenomenon. Cowbirds roost in the riparian areas at night and search for nests there in the early morning; in the afternoon, cowbirds congregate and forage in farm fields or urban areas. All of my surveys were conducted in the early morning hours when cowbird concentrations in farm fields was low.

Bird-Vegetation Associations: Microhabitat Scale Nest-site Selection by Host Species.-- Within the vegetation types selected on a macrohabitat scale, specific vegetative features were associated with nest-site selection. Bell's vireos and yellow-breasted chats selected nest sites that were farther than random sites from forest edges or clearings. Predators and parasites concentrate their nest-searching activities along forest edges or corridors (Bider 1968, Payne 1973, Norman and Robertson 1975). In addition, small forest openings (such as tree fall gaps) may serve as focal points from which cowbirds can search for host nests (Brittingham and Temple 1983, 1996). Therefore, it may be advantageous for birds to place nests farther from edges or forest canopy openings. 106 Bell's vireos and yellow-breasted chats selected nest patches with higher percent foliage cover in the low canopy layer ( 0.6-4.5 in high) than randomly located plots. :n the lower Colorado River valley, Bell's vireos nested between 0.5 and 3.8 in in height (average 1.4 m), and yellow-breasted chats nested between 0.8 and 4.2 m in height (average 2.0 m; Appendix D and E). Foliage cover in the low canopy layer may conceal nests from predators and brood parasites. Other studies have found higher nest concealment or vegetation volume/density within the nest patch than at random sites (Holway 1991, Kilgo et al. 1996). High vegetation volume at nest height may be indicative of the number of potential nest sites within the nest patch, and perhaps within the territory. This could facilitate renesting following nest failure. In addition, a high density of potential nest sites may lower predator or parasite search efficiency by increasing the number of sites that must be searched (Martin 1988, Martin and Roper 1988). There was also less dispersion of vegetation at vireo and chat nest sites than at random sites. This suggests a preference for uniform distribution of vegetation at the nest site. Horizontal homogeneity of vegetation could increase the number of potential nest sites within the vicinity of the nest, again lowering predator or parasite search efficiency (Norment 1993). 107 Bell's vireo and yellow-breasted char nest batches had higher percent foliage cover in the middle canopy layer than

random plots. Rosenzweig (1981) suggested that neg .= placement may be primarily influenced by proximity to abundant supplies of preferred resources. Selection for dense foliage cover in nest patches may be associated with high food abundance (Conner et al. 1986) and, therefore, increased foraging opportunities near the nest. Foraging near the nest not only maximizes food delivery rates to young (Norment 1993), it also prevents extended forays away from the nest that leave it unprotected (Arcese and Smith 1988). Percent cottonwood-willow cover was not significantly different between nest patches and random plots for both species. However, both vireo and chat nest patches had higher percent saltcedar and mesquite foliage cover than random patches, especially within the middle canopy layer (4.5-7.5 m high). In addition, vireos and chats placed most of their nests in either saltcedar, mesquite, or arrowweed (Appendix D and E). Therefore, saltcedar and mesquite appear to be important understory components of the lower Colorado River riparian forests for nesting avifauna. Bell's vireos selected nest sites that were closer to mature trees than random sites, and nest patches that had higher percent canopy cover by mature trees. Yellow- 108 breasted chats selected nest sites that were closer to shrubs and mature trees than random sites, and nest patches that had higher percent Baccharis foliage cover (a shrub species frequently used as a nesting substrate). The proximity of mature trees to chat nests may be related to the use of tall trees as display perches by the males (Thompson and Nolan 1973). Otherwise, nest-site selection for these 2 species appears to be associated with vegetative features affecting nest concealment directly at the nest and within the nest patch. Nest-site Discovery by Cowbirds.-- Martin (1992) stressed the importance of determining habitat features that lower the fitness of breeding populations of neotropical migratory songbirds. Factors that play a large role in nesting mortality, such as predation and parasitism, should influence individuals to select nest sites with features that minimize the risk of mortality (Martin and Roper 1988). In the lower Colorado River valley, cowbird parasitism appears to be a primary mortality agent for open-cup nesters. Therefore, natural selection should favor individual birds that select habitat features that minimize the chance of nest discovery by cowbirds. On the other hand, high quality nest sites may be limited, forcing individuals to select sites that are of marginal quality (Martin and Roper 1988). 109 The general parasitism model suggests that nest concealment may be important in preventing early discovery of nests by cowbirds. The presence of a shrubby unders ory may decrease the likelihood of parasitism, perhaps by providing nest concealment and decreasing parasite search efficiency. Successfully parasitized nests had less shrub cover near the nest (indicated by a larger mean distance to the closest shrub), less shrub foliage cover between 0 and 0.6 m high within the nest patch, less ground cover below the nest, and less canopy cover above the nest. Other studies have reported less concealment at parasitized nests (Staab 1995) and at depredated nests (Martin and Roper 1988). On the other hand, several studies have shown concealment to differ between nests and random sites, but not between successful and unsuccessful nests (Holway 1991, Kilgo et al. 1996). The general parasitism model also indicated the importance of foliage cover in the upper canopy layer. Successfully parasitized nests had higher percent foliage cover in the upper canopy layer, which is most likely due to the fact that parasitized nests were closer to mature trees, such as cottonwoods or willows. Cowbirds use tall trees as perches from which to survey the surrounding area for nests to parasitize (Norman and Robertson 1975). Anderson and Storer (1976) found higher parasitism rates when Kirtland's 110 warbler nests were located near exposed perches (dead snags) in the relatively open jack-pine (Pinus banksiana) forests of Michigan. Much of the riparian area in the lower Colorado River valley is fragmented and relatively open, and mature trees (e.g., cottonwoods and willows) near nests may have been used by cowbirds for displaying and nest searching. The species-specific parasitism models show similar patterns. Successfully parasitized Bell's vireo nests had lower shrub foliage cover within the nest patch and lower shrub cover at the nest (as indicated by a greater distance to the mean closest shrub). On the other hand, parasitized nests were generally closer to mature cottonwood or willow trees than unparasitized nests. The modeling of this variable was not straight forward. Likelihood of parasitism decreased when the closest mature cottonwood or willow was at a distance of 12 m or more. However, parasitized nests with a mature cottonwood or willow tree 5-12 m away were more likely to be parasitized than nests with cottonwoods or willows within 5 m. Therefore, mature cottonwoods and willows may have provided nest concealment when located within 5 m of the nest, but at slightly greater distances they provided elevated perches for cowbird surveying. Nest concealment was also an important factor affecting parasitism for yellow-breasted chats. Successfully parasitized yellow-breasted chat nests were less concealed below the nest. Staab (1995) found a similar pattern in parasitized nests of common cowbird hosts along Walnut and Apache Creeks, Yavapai County, Arizona: nests were less likely to be parasitized if obstructing vegetation was immediately below the nest. Parasitized yellow-breasted chat nests also had greater foliage cover in the upper canopy layer, and univariate analyses showed that parasitized nests were significantly closer to mature cottonwood or willow trees. This suggests, once again, the use of tall trees as cowbird surveying perches. Brittingham and Temple (1996) found that parasitized nests in a Wisconsin deciduous forest had a lower percent cover in the sub-canopy (3-10 m high) and canopy (>10 m high) layers than unparasitized nests. They suggest that an open sub-canopy and canopy increases parasite search efficiency. Differences in general vegetation structure of the Wisconsin and lower Colorado River sites can explain the disparate results of these 2 studies. Nest patches in the Wisconsin site had a higher mean percent canopy cover above 3 m (46-88%; Brittingham and Temple 1996) than nest patches in the lower Colorado River valley (3-37%). Most nests along the lower Colorado River were already located in areas with low percent foliage cover in the upper canopy layers. Therefore, grater canopy cover near parasitized nests may 2 represent the presence of trees that served as cowbird survey perches. Note: percent canopy cover was measured in a similar manner in both studies (see Ambuel and Temple 1983), with minor modifications that probably increased my estimates of percent cover in each canopy layer.

MANAGEMENT IMPLICATIONS Cowbird control (e.g., live-decoy trapping and shooting) is costly and not a permanent solution to lowering parasitism rates (Robinson et al. 1993). In addition, these techniques are not effective over large fragmented landscapes that provide cowbirds with numerous feeding grounds and easy access to nesting avifauna. The only permanent method of cowbird control is habitat restoration on the breeding grounds (Robinson et al. 1993). The nest- site selection and parasitism models we developed indicated that vegetation management can be used to enhance breeding habitat for neotropical migratory songbirds while reducing brood parasitism by the brown-headed cowbird. The patchiness of cottonwood-willow vegetation in the lower Colorado River valley is a primary concern. Bell's vireos and yellow-breasted chats were attracted to areas predominated by cottonwoods and willows, but experienced high parasitism rates due to the patchy distribution of these trees along much of the river. Therefore, management 113 should focus on restoring large tracts of contiguous cottonwood-willow forest. However, this may only be feasible along the Bill Williams River where lack of channelization and periodic flooding creates favorable soil conditions for cottonwood-willow persistence and regeneration. Cottonwood-willow revegetation efforts should aim at widening the riparian corridor along the Bill Williams River, thereby decreasing the amount of edge habitat available to brown-headed cowbirds. The increase in saltcedar-dominated riparian areas along the lower Colorado River is inevitable due to current water management practices that lower the water table, increase soil salinity, and reduce the frequency and intensity of flooding (Anderson et al. 1977). Monotypic saltcedar stands, which predominate much of the lower Colorado River drainage, have relatively low value to breeding neotropical migrants in this region (Lynn et al. 1996). However, the nest-site selection models indicate that saltcedar is an important understory component of the remaining cottonwood-willow forest. Saltcedar may provide suitable nesting sites when an overstory of cottonwood or willow provides shelter from the extreme summer heat characteristic of the lower Colorado River valley. Other studies report that saltcedar can provide (even enhance) 7 14 breeding habitat for riparian-obligate bird species in the southwest (Hunter et al. 1988, Brown and Trosset 1989). Saltcedar can be managed along the lower Colorado River to benefit breeding bird populations in 1 of 2 ways. First, riparian areas can be managed for mature stages of saitcedar (those stages providing the greatest percent foliage density above 4.5 meters in height; Hunter et al. 1988). Mature, structurally complex stands of saltcedar are capable of attracting song sparrows, Abert's towhees, summer tanagers, and yellow-breasted chats (Anderson et al. 1977). Second, the addition of native tree species to saltcedar-dominated areas may provide conditions (i.e., shade) necessary to attract breeding bird species (Anderson et al. 1977, Ellis 1995). For example, Bell's vireos and yellow-breasted chats breeding in the lower Colorado River valley selected nest sites that were close to mature, structurally complex trees. Therefore, it may be possible to provide additional nesting habitat for breeding birds by clearing areas within monotypic saltcedar stands and planting willows and cottonwoods. These clearings should be arranged such that a closed canopy is formed as the cottonwoods and willows mature. Because saltcedar can reestablish itself following removal, the cottonwoods and willows should be monitored until they are established and not outcompeted by saltcedar for light and water (Sudbrock 1993, Busch and Smith 1995). 115 If this strategy is incorporated over a large area, the result will be a cottonwood-willow forest with a shrubby saitcedar understory. The current management practice of revegetating xeric, high-salinity soil sites with cottonwood saplings will not provide for suitable, long-term nesting habitat for breeding birds. These soil conditions are not favorable to persistence and regeneration of cottonwoods (Busch and Smith 1995). Revegetated sites can be improved by simulating natural flood regimes to promote regeneration of cottonwoods. In addition, the nest-site selection and parasitism models indicate that a shrubby understory is an important component of breeding habitat for several neotropical migrants. Therefore, revegetated sites should include shrubby understory species as well as cottonwoods and willows. Many shrub species, such as arrowweed, quail bush, Baccharis spp., and saltcedar, are able to grow in dry, saline soils (Rosenberg et al. 1991, Busch and Smith 1995). Mesquite and palo verde trees may also serve as understory species. These tree species are non- phreatophytes and can be successfully established in areas with lowered water tables (Briggs et al. 1994). Finally, revegetated sites should consist of willows and cottonwoods instead of monotypic cottonwood stands. Willows have a lower osmotic potential than cottonwoods, a characteristic 116 that appears to favor persistence under conditions of moisture stress and high salinity (Busch and Smith 1995). Cowbird control (live-decoy trapping, shooting), although only a temporary solution, may be necessary to protect local breeding populations of some neotropical migratory passerines (Robinson et al. 1995). Live decoy cowbird trapping has been an effective means of cowbird control in other locations (RECON 1989 cited by Franzreb 1990, Grzybowski 1990). Regardless, intensive trapping efforts are necessary if a decline in parasitism rates is expected. For example, 3,500 cowbirds were removed from Kirtland's warbler breeding grounds before parasitism rates declined from 60% to 3% (Kelly and Decapita 1982). Placement and timing of trapping efforts will be crucial for attracting female cowbirds. Trapping efforts may be most successful right after cowbirds arrive on the breeding grounds but before intensive breeding begins (Kelly and DeCapita 1982). Beezley and Rieger (1987) found that more females were caught when traps were placed in foraging areas rather than in the riparian (breeding) areas. Grzybowski (1990) achieved the greatest reduction in parasitism rates when traps were set on the perimeter of black-capped vireo nesting areas. If traps are set in the nesting area, trapping sites should coincide with areas where cumulative host abundance (therefore, cowbird 117 abundance) is high. Site-specific information on species abundances in the lower Colorado River valley is given in Lynn et al. (1996). 118 APPENDIX A. Codes and definitions of vegetation characteristics collected at nest and random sites.

Variable code Variable definition

NESTPLSP Plant species that nest was in NESTHGT Height of nest NESTPLHGT Height of nest plant GRDCVR % ground cover within 25 cm diameter circle below nest UNDERCOV % cover within 25 cm diameter circle between ground and nest OVERCOV % canopy cover within 25 cm diameter circle above nest AVELATCOV Average % cover within 25 cm diameter circle centered at nest and measured from a distance of 1 m in each of the 4 cardinal directions TOTCOV Total vegetation cover at nest (TOTCOV = UNDERCOV + OVERCOV + AVELATCOV) DISTOPEN Distance from nest to nearest opening SIZEOPEN Size of nearest opening TYPEOPEN Type of opening DISTSGPF Distance from nest to nearest mature cottonwood or willow tree NUM30 Number of mature cottonwoods or willows within 30 meters of above tree MINSHRUB Minimum distance from nest to closest shrub in 4 quadrants MAXSHRUB Maximum distance from nest to closest shrub in 4 quadrants 119 Appendix A - continued

Variable code Variable definition

RNGSHRUB Range of distance from minimum to maximum closest shrub MEANSHRUB Average distance to closest shrub in 4 quadrants MINTREE Minimum distance from nest to closest mature tree in 4 quadrants MAXTREE Maximum distance from nest to closest mature tree in 4 quadrants RNGTREE Range of distance from minimum to maximum closest mature tree MEANTREE Average distance to closest mature tree in 4 quadrants MINCOMB Minimum distance from nest to closest vegetation in 4 quadrants MAXCOMB Maximum distance from nest to closest vegetation in 4 quadrants RNGSEP Range of distance from minimum to maximum closest vegetation MEANCOMB Average distance to closest vegetation in 4 quadrants DS Shrub dispersion = 100 X (SD/MEANSHRUB) DT Tree dispersion = 100 X (SD/MEANTREE) DC Vegetation dispersion = 100 X (SD/MEANCOMB) M5GRDCVR % ground cover in nest patch' M5CANCVR % canopy cover from mature trees in nest patch M5HERB % herbaceous cover in nest patch 120 Appendix A - continued

Variable code Variable definition

M5LEAFLITT % leaf litter cover in nest patch M5BAREGRD % bare ground in nest patch HGT1 % live foliage cover 0.0 - 0.06 m high in nest patch' HGT2 % live foliage cover 0.7 - 4.5 in high in nest patch b HGT3 % live foliage cover 4.6 - 7.5 in high in nest patch' HGT4 % live foliage cover > 7.5 in high in nest patch' SPDIV1 Plant species richness in height category 1 SPDIV2 Plant species richness in height category 2 SPDIV3 Plant species richness in height category 3 SPDIV4 Plant species richness in height category 4 SXPF1 - SXPF4 % cottonwood-willow live foliage cover in height categories 1 - 4 in nest patch SXPFTOT % total cottonwood-willow live foliage cover in nest patch TR1 - TR4 % saltcedar live foliage cover in height categories 1 - 4 in nest patch TRTOT % total saltcedar live foliage cover in nest patch PS1 - PS4 % mesquite live foliage cover in height categories 1 - 4 in nest patch 121 Appendix A - continued

Variable code Variable definition

PSTOT % total mesquite live foliage cover in nest patch

DWTREE1 - DWTREE4 % desert wash tree spp. live foliage cover in height categories 1 - 4 in nest patch DWTREETOT % total desert wash tree spp. foliage cover in nest patch BS1 - 3S2 % Baccharis live foliage cover in height category 1 and 2 in nest patch BSTOT % total Baccharis live foliage cover in nest patch TS1 - TS2 % arrowweed live foliage cover in height category 1 and 2 in nest patch TSTOT % total arrowweed live foliage cover in nest patch ASTOT % total quail bush live foliage cover in nest patch SHRUB1 - SHRUB2 % live shrub foliage cover in height category 1 & 2 in nest patch SHRUBTOT % total live foliage cover by shrubs in nest patch HETOT % total live herbaceous cover in nest patch

Nest patch refers to the 5 m radius circular plot centered on the nest. b This value can be >100% due to live foliage cover from multiple plant species.

• - ▪•

-H a) 0 oa c\1 0 ks-.) co CD 0

o 0 CD wo, 0 0 • 14-1

-H 0 775 0 Lc) CD 0 >1 ( -) if) 4-) 4--) 0 0,1 CV 0 CD CD CD fil X -H o

4-) -H

a) a) 0 0

O C (a - H - H >-.1 -H -(--2 -Q -P (13 (._) U) • -H U) (1) P 4-) -H S-4 • (1) (1) a) -H CO -H a) o., a) „Q ,C -H 0 P 4-3(Li-P (1) ▪ 5-4 fa, 7) CtT Q.) C - H C Q) (Iii -H - H • $-1 tY1 3-1

(1) • (3) 4-) • (I) (3) U) (0 4-) (1) (1) 0.1 (0 C -P al 05 (ti Ci) 11 -) Cr) c0 • (1) 44 44 44 4-4 CNI 0 0 C 0000 0 0 0 CD 0 (n (1) (1) 0 CD 0 0 >1 -P 4-3 "Li 0, (0 (0 (a

0 0 0 0 0 LU 0 0

lp 1-1 CO CO lO r-I CV CO r-1 C`..1 • P0 • • •

0 (1) -H Q Cr) (-) 4-3 -4) tyn

O Ci - H -P (jr 14-1 Q p cf) Q) • H D -H

O 0 4-) 0 Q -H 0 O 4-) 4-) i-4 • • a) - • Er) 4--) G) -HG) (r) C t)' • a) 0 0 a) 0 -4--) 01-H 0 0 (0 4-) 0 0 Q. 4-3 a) - H W tr4--) C 0 0 • tr-H -H cri tr tr (1) -H (1) 7:5 (1) - H (r) 4-) a) -H ri) 4 ) • tr (-15 rCi 0 ).4 tr • W Ti Q. - H 1:13 (0 $4 Ti 73 -H al i4 tri )4 C rC) -H -H • -H 3 .(-.2 O -H 3 3 Ti 4-) U 0 -P 0 tr cr) Ti (I) c Cr a) (1) -H 4--) 4-) Ci .4) a (a (1) -H -H ▪ tlY) 4-) ,C) 4-) 4-) U) 3 (r) 3 (0 4-) (1)• 0 W a) .-8 o - 4-3 .r-4 03 • (0 (0 t31 Q Q0)tf-44-l-44-4 3 43e:1'0000 • O 0 CU tn a) a) a) a)

ci:Co ai - H «sas as co o H o 4--) -4) 4_) .4) -H -H -H -H 4-) - 4p 4rHr4 (0 $-1 (0(0(-0(13 • 4-) W .0 ° 4-3 4-) -P 5-4 5-15-1 5-4 (j) -Hj.0000 _ •H tZT) °

(15 n-1 4-2,4 >1>1>1>1

4-1 0 4-) -H -H O mitas Ri • 4-) '—i (-) (:) ai -H ,a 3 o - H .o 124 APPENDIX D. Microhabitat characteristics of Arizona Bell's vireo nest sites (n = 147).

Variable SD

Height of nest 1.4 0.5 Distance from nest to closest opening 9.4 9.9 Distance from nest to closest shrub 1.1 1.3 Distance from nest to closest mature tree 3.1 3.2 Distance to closest cottonwood-willow 23.8 34.4 Canopy cover (%) in nest patch 69.0 30.0 Cottonwood-willow cover (%) in nest patch 24.0 35.7 Saltcedar cover (%) in nest patch 50.3 40.4 Mesquite cover (%) in nest patch 30.0 37.0 Shrub cover (%) in nest patch 23.0 30.0 Nesting substrate' Cottonwood or willow (%) 7.5 Saltcedar (%) 49.7 Mesquite (%) 24.5 Shrub species (%) 18.4

' Percent of nests that were placed in each vegetation type. 125 APPENDIX E. Microhabitat characteristics of yellow-breasted chat nest sites (n = 59).

Variable SD

Height of nest 2.0 0.8 Distance from nest to closest opening 13.7 12.9 Distance from nest to closest shrub 0.4 0.6 Distance from nest to closest mature tree 3.6 3.0 Distance to closest cottonwood-willow 12.0 10.1 Canopy cover (%) in nest patch 46.0 37.0 Cottonwood-willow cover (%) in nest patch 25.1 30.8 Saltcedar cover (%) in nest patch 64.2 33.5 Mesquite cover in (%) nest patch 7.1 17.6 Shrub cover (%) in nest patch 35.1 36.4 Nesting substrate' Cottonwood or willow (%) 1.7 Saltcedar (%) 62.7 Mesquite (%) 0.0 Shrub species (%) 35.6

' Percent of nests that were placed in each vegetation type. 126 APPENDIX F. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of Bell's vireo nest sites (n = 147) and random sites (n = 208).'

Variable b Plot type P. SD Test statistic'

DISTOPEN Nest 9.36 9.93 10680.0 0.000* Random 6.50 7.12 RNGSHRUB Nest 3.16 4.32 13942.5 0.190 Random 4.94 21.12 MINTREE Nest 3.09 3.24 11536.5 0.000* Random 4.35 3.98 MEANTREE Nest 7.70 5.58 10231.5 0.000* Random 12.58 12.48 MEANCOMB Nest 1.66 1.58 13676.5 0.112 Random 1.84 1.51 DS Nest 47.74 24.26 13908.5 0.178 Random 52.07 26.83 DC Nest 37.05 17.13 12976.0 0.020* Random 43.88 23.77 M5CANCOVER Nest 69.00 30.00 10526.5 0.000* Random 48.00 38.00 M5GRDCOVER Nest 95.00 10.00 12608.0 0.001* Random 91.00 15.00 M5GRDHERB Nest 7.80 16.20 13760.5 0.053* Random 12.80 22.50 M5GRDLITTR Nest 92.10 14.70 11972.5 0.000* Random 82.80 23.10 HGT2 Nest 118.57 30.02 9097.0 0.000* Random 95.05 34.61 HGT3 Nest 35.17 31.78 11170.5 0.000* Random 21.92 29.19 TR3 Nest 12.24 23.17 12692.5 0.000* Random 4.04 11.59 127 APPENDIX F - continued

Variable' Plot type SD Test statistic'

TRTOT Nest 50.34 40.42 12706.5 0.006* Random 37.88 35.62 PS3 Nest 7.28 15.94 12499.5 0.000* Random 1.92 8.96 PSTOT Nest 30.00 37.01 12089.5 0.000* Random 15.48 28.04 SPDIV2 CATEGORICAL VARIABLE' 7.5 0.113 a Other variables showed significant differences between nest sites and random sites, but were not used in model building because of high intercorrelation. b Variable definitions are listed in Appendix A. Statistical differences based on Mann-Whitney U statistic for continuous variables and Pearson's Chi-square statistic for categorical variables. Q * indicates significance at the P < 0.10 level. e Means and standard deviations cannot be calculated for categorical variables. 128 APPENDIX G. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of yellow-breasted chat nest sites 59) and random plots (n = 208).'

Variable' Plot type SD Test ID° statistic'

DISTOPEN Nest 13.70 12.90 3470.5 0.000* Random 6.50 7.12 SIZEOPEN CATEGORICAL VARIABLES 1.9 0.167 TYPEOPEN CATEGORICAL VARIABLE' 17.1 0.017* MINSHRUB Nest 0.42 0.59 2972.0 0.000* Random 1.19 1.42 MEANSHRUB Nest 0.80 0.81 2562.0 0.000* Random 3.03 6.10 MEANTREE Nest 7.42 6.66 3562.0 0.000* Random 12.58 12.48 DT Nest 47.06 20.63 4180.5 0.004* Random 59.92 30.23 M5GRDCOV Nest 97.00 7.00 4644.0 0.001* Random 91.00 15.00 M5HRDHERB Nest 4.90 10.90 5149.0 0.024* Random 12.80 22.50 M5GRDLITT Nest 95.90 7.70 4142.5 0.000* Random 82.80 23.10 HGT2 Nest 132.88 44.72 3158.5 0.000* Random 95.05 34.61 HGT3 Nest 31.69 34.00 5104.0 0.035* Random 21.92 29.19 TR3 Nest 16.44 28.69 4695.0 0.000* Random 4.04 11.59 TRTOT Nest 64.24 33.54 3630.5 0.000* Random 37.88 35.62 PSTOT Nest 7.12 17.62 5191.0 0.023* Random 15.48 28.04 129

APPENDIX G - continued

Variable Plot type SD Test P' statistic'

BSTOT Nest 12.54 22.86 4862.5 0.000* Random 3.70 11.60 SHRUB1 Nest 2.54 4.77 5643.0 0.148 Random 2.07 5.65 SHRUBTOT Nest 35.08 36.41 4794.0 0.006* Random 19.90 28.49 SPDIV3 CATEGORICAL VARIABLE' 5.4 0.068*

' Other variables showed significant differences between nest sites and random sites, but were not used in model building because of high intercorrelation. • Variable definitions are in Appendix A. ' Statistical differences based on Mann-Whitney U statistic for continuous variables and Pearson's Chi-square statistic for categorical variables. d * indicates significance at the P < 0.10 level. ' Means and standard deviations cannot be calculated for categorical variables. 130 APPENDIX H. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of successfully parasitized (n = 149) and unparasitized nests (n = 47).'

Variable' Parasitism x SD Test status' statistic'

GRDCOV No 95.53 11.71 2799.0 0.012* Yes 85.03 26.62 OVERCOV No 77.39 15.80 2940.0 0.161 Yes 71.82 20.28 MEANSHRUB No 1.24 1.32 2739.5 0.046* Yes 2.02 2.43 MEANTREE No 9.09 6.56 2311.0 0.020* Yes 7.14 5.29 DISTSGPF No 30.97 44.57 2458.0 0.012* Yes 16.92 22.98 M5GRDCOV No 97.00 8.00 3127.5 0.160 Yes 93.00 14.00 M5CANCOV No 54.00 35.00 2878.5 0.064* Yes 64.00 33.00 HGT1 No 16.17 17.88 2942.5 0.082* Yes 11.81 17.52 HGT3 No 27.02 30.50 2938.5 0.091* Yes 37.05 33.52 HGT4 No 3.40 10.48 3044.0 0.062* Yes 9.53 19.77

SPDIV1 CATEGORICAL VARIABLE f 4.6 0.200

SPDIV4 CATEGORICAL VARIABLE f 3.0 0.081* SXPFTOT No 6.17 28.40 3047.5 0.128 Yes 25.97 35.45 SHRUB1 No 3.83 7.09 2944.0 0.010* Yes 1.48 4.25 131 APPENDIX H - continued

Variable Parasitism SD Test status' statistic'

SHRUBTOT No 31.28 33.14 3070.5 0.178 Yes 23.96 31.73

a Nests of Bell's vireos, blue grosbeaks, common yellowthroats, and yellow-breasted chats used for generalized parasitism model. Other variables showed significant differences between unparasitized and successfully parasitized nests, but were not used in model building because of high intercorrelation. b Variable definitions are in Appendix A. • No refers to unparasitized nests and nests that were inappropriately parasitized, Yes refers to successfully parasitized nests. d Statistical differences based on Mann-Whitney U statistic for continuous variables and Pearson's Chi-square statistic for categorical variables. e * indicates significance at the P < 0.10 level. f Means and standard deviations cannot be calculated for categorical variables. 127

APPENDIX I. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of Bell's vireo successfully parasitized (n = 102) and unparasitized nests (n = 26).'

Variable' Parasitism SD Test status' statistic'

AVELATCON No 12.12 11.96 930.0 0.084' Yes 17.92 15.63 MEANSHRUB No 1.67 1.61 957.5 0.054* Yes 2.59 2.69 RANGESEP No 17.93 18.13 949.0 0.240 Yes 13.36 10.23 DISTSGPF No 42.99 56.75 858.0 0.016* Yes 19.93 26.40 HGT3 No 28.08 29.67 1123.5 0.224 Yes 37.16 32.65 SXPFTOT No 13.08 28.53 1062.5 0.067* Yes 25.29 36.58 SHRUB1 No 3.85 8.04 1142.5 0.058* Yes 1.27 4.14 SHRUBTOT No 29.62 29.86 1047.5 0.078* Yes 20.10 28.89

' Other variables showed significant differences between unparasitized and successfully parasitized nests, but were not used in model building because of high intercorrelation. b Variable definitions are in Appendix A. • No refers to unparasitized nests and nests that were inappropriately parasitized, Yes refers to successfully parasitized nests. • Statistical differences based on Mann-Whitney U statistic for continuous variables and Pearson's Chi-square statistic for categorical variables. • * indicates significance at the P < 0.10 level. 133 APPENDIX J. Vegetation variables selected for use in forward stepwise logistic regression models based on univariate analyses of yellow-breasted chat successfully parasitized (n = 36) and unparasitized nests (n = 17).a

Variable' Parasitism SD Test status' statisticc

GRDCOV No 95.29 15.05 211.0 0.032* Yes 86.25 23.92 UNDERCON No 50.59 32.45 207.5 0.059* Yes 31.94 25.50

NESTPLSP CATEGORICAL VARIABLE f 5.9 0.118 MEANTREE No 8.97 6.97 156.0 0.021* Yes 6.02 5.65 DT No 52.01 21.12 194.0 0.134 Yes 43.88 20.16 DISTSGPF No 16.34 14.12 191.0 0.118 Yes 9.54 6.70 M5GRDCOV No 99.00 2.00 236.0 0.060* Yes 95.00 8.00 HGT1 No 20.59 19.52 172.5 0.007* Yes 8.33 14.24 HGT4 No 0.59 2.43 244.5 0.086* Yes 6.94 14.89 TSTOT No 15.29 24.01 250.0 0.243 Yes 28.33 36.99 HETOT No 16.47 25.72 202.0 0.011* Yes 2.50 6.49

Other variables showed significant differences between unparasitized and successfully parasitized nests, but were not used in model building because of high intercorrelation. b Variable definitions are in Appendix A. No refers to unparasitized nests and nests that were inappropriately parasitized, Yes refers to successfully parasitized nests. 134 APPENDIX J - continued d Statistical differences based on Mann-Whitney U statistic for continuous variables and Pearson's Chi-square statistic for categorical variables. e * indicates significance at the P < 0.10 level. f Means and standard deviations cannot be calculated for categorical variables. 135 LITERATURE CITED Airola, D. A. 1986. Brown-headed cowbird parasitism and habitat disturbance in the Sierra Nevada. J. Wildl. Manage. 50:571-575. Ambuel,B, and S. A. Temple. 1983. Area-dependent changes in the bird communities and vegetation of southern Wisconsin forests. Ecology 64:1057-1068. Anderson, B. W., A. E. Higgins, and R. D. Ohmart. 1977. Avian use of saltcedar communities in the lower Colorado River valley. Pages 128-136 in R. R. Johnson and D. A. Jones, tech. coords. Importance, preservation and management of riparian habitat: a symposium. U.S. For. Serv. Gen. Tech. Rep. RM-43. , and R. D. Ohmart. 1986. Vegetation. Pages 639-659 in A. Y. Cooperrider, R. J. Boyd, and H. R. Stuart, eds. Inventory and monitoring of wildlife habitat. U.S. Bur. Land Manage. Service Center. Denver, CO XVIII, 858pp. Anderson, W. L., and R. W. Storer. Factors influencing Kirtland's warbler nesting success. Jack-Pine Warbler 54:105-115. Arcese, P., and J. N. M. Smith. 1988. Effects of population density and supplemental food on reproduction in song sparrows. J. of Animal Ecology 57:119-136. Barlow, J. C. 1962. Natural history of the Bell's vireo, Vireo Bellii Audubon. Univ. Kans. Publ. Mus. Nat. Hist. 12:241-296. Beezley, J. A., and J. P. Rieger. 1987. Least Bell's vireo management by cowbird trapping. Western Birds 18:55- 61. Bibby, C. J., N. D. Burgess, and D. A. Hill. 1992. Bird census techniques. Academic Press, Harcourt Brace and Co., London. 257pp. Bider, J. R. 1968. Animal activity in uncontrolled terrestrial communities as determined by a sand transect technique. Ecol. Monogr. 38:269-308. 136 Block, W. M., and L. A. Brennan. 1993. The habitat concept in ornithology: theory and applications. Pages 35-91 in D. M. Power, ed. Current ornithology, Vol. "- , and R. J. Gutierrez. 1991. Ecomorphological relationships of a guild of ground-foraging birds in northern California, USA. Oecologia 87:449-458. , and M. L. Morrison. 1991. Influence of scale on the management of wildlife in California oak woodlands. Pages 96-104 in R. B. Standiford, tech. coord. Proceedings of the symposium on oak woodland and hardwood rangeland management. U.S. For. Serv. Gen. Tech. Rep. PSW-126. Bonham, C. D. 1989. Measurements for terrestrial vegetation. John Wiley and Sons, New York. 338pp. Briggs, M. K., B. A. Roundy, and W. W. Shaw. 1994. Trial and error: assessing the effectiveness of riparian revegetation in Arizona. Restor. Manage. Notes 12:160- 167 Brittingham, M. C., and S. A. Temple. 1983. Have cowbirds caused forest songbirds to decline? BioScience 33:31- 35. , and . 1996. Vegetation around parasitized and non-parasitized nests within deciduous forest. J. Field Ornithol. 67:406-413. Brown, B. T. 1994. Rates of brood parasitism by brown- headed cowbirds on riparian passerines in Arizona. J. Field Ornithol. 65:160-168. , and M. W. Trosset. 1989. Nesting-habitat relationships of riparian birds along the Colorado River in Grand Canyon, Arizona. Southwest. Nat. 34:260-270. Brown, H. 1903. Arizona bird notes. Auk 20:43-50. Burgham, M. C., and J. Picman. 1989. Effect of brown- headed cowbirds on the evolution of yellow warbler anti-parasite strategies. Anim. Behav. 38:298-308. Busch, D. E., and S. D. Smith. 1995. Mechanisms associated with decline of woody species in riparian ecosystems of the southwestern U.S. Ecol. Monogr. 65:347-370. I

Byers, C. R., R. K. Steinhorst, and P. R. Krausman. 1984. Clarification of a technique for analysis of utilization-availability data. J. Wildl. Manage. 48:1050-1053. Capen, D. E., J. W. Fenwick, D. B. Inkley, and A. C. Boynton. 1986. Multivariate models of songbird habitat in New England forests. Pages 171-175 in J. Verner, M. L. Morrison, and C. J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. Univ. Wisconsin Press, Madison. Clark, K. L., and R. J. Robertson. 1979. Spatial and temporal multi-species aggregations as anti-parasite and anti-predator defense. Behav. Ecol. Sociobiol. 5:359-371. Carothers, S. W., R. R. Johnson, and S. W. Aitchison. 1974. Population structure and social organization of Southwestern riparian birds. Am. Zool. 14:97-108. Cohan, D. R., B. W. Anderson, and R. D. Ohmart. 1978. Avian population responses to saltcedar along the lower Colorado River. Pages 371-383 in R. R. Johnson and F. J. McCormick, tech. coords. Strategies for protection and management of floodplain wetlands and other riparian ecosystems. U.S. For. Serv. Gen. Tech. Rep. WO-12. Conner, R. N., M. E. Anderson, and J. G. Dickson. 1986. Relationships among territory size, habitat, song, and nesting success of northern cardinals. Auk 103:23-31. Cottam, G., and J. T. Curtis. 1956. The use of distance measures in phytosociological sampling. Ecology 37:451-460. Darley, J. A. 1982. Movements and activities of radio- tracked brown-headed cowbirds. Auk 99:316-327. Ellis, L. M. 1995. Bird use of saltcedar and cottonwood vegetation in the Middle Rio Grande Valley of New Mexico, USA. J. Arid Environ. 30:339-349. Ehrlich, P. R., D. S. Dobkin, and D. Wheye. 1988. The birder's handbook. Simon and Schuster, New York. 785pp. 138 7i7liater, 7. S., R. Breitwisch, and P. M. Nealen. 1994. Predation on northern cardinal nests: does choice of nest site matter? Condor 96:761-768. Finch, D. M. 1983. Brood parasitism of the Abert's towhee: timing, frequency, and effects. Condor 85:355-359. Franzreb, K. E. 1987. Endangered status and strategies for conservation of the least Bell's vireo (Vireo bellii pusillus) in California. Western Birds 18:43-49. . 1990. An analysis of options for reintroducing a migratory, native passerine, the endangered least Bell's vireo Vireo bellii pusillus in the Central Valley, California. Biol. Conserv. 56:105-123. Fretwell, S. D. 1977. Is the dickcissel a threatened species? Am. Birds 31:923-932. Friedmann, H. 1963. Host relations of the parasitic cowbirds. U.S. Natl. Mus. Bull. 37:21-29. , and L. F. Kiff. 1985. The parasitic cowbirds and their hosts. Proc. Western Found. Vert. Zool. 2:226- 302. Gibbs, J. P., and J. Faaborg. 1990. Estimating the viability of ovenbird and Kentucky warbler populations in forest fragments. Conserv. Biol. 4:193-196. Gaines, D. 1974. A new look at the nesting riparian avifauna of the Sacramento Valley, California. Western Birds 5:61-78. Gates, J. E., and N. R. Giffen. 1991. Neotropical migrant birds and edge effects at a forest-stream ecotone. Wilson Bull. 103:204-217. , and L. W. Gysel. 1978. Avian nest dispersion and fledging success in field-forest ecotones. Ecology 59:871-883. Goodman, D. 1987. The demography of chance extinction. Pages 11-34 in M. E. Soulé, ed. Viable populations for conservation. Cambridge Univ. Press, New York. Graham, D. S. 1988. Responses of five host species to cowbird parasitism. Condor 99:588-591. 139 Grinnell, J. 1914. An account of the mammals and birds of the lower Colorado River Valley with especial reference to the distributional problems presented. Univ. California Publ. Zool. 12:51-294. Grzybowski, J. A. 1990. Ecology and management of the Black-capped vireo (Vireo atricapillus) in the Wichita Mountains, Oklahoma - 1990. Wichita Mountains, Wildlife Refuge. U.S. Fish and Wildlife Service, Indiahoma, OK. 28pp. Hann, H. W. 1941. The cowbird at the nest. Wilson Bull. 53:211-221. Harris, L. D. 1988. Edge effects and conservation of biotic diversity. Conserv. Biol. 2:330-332. Hensler, G. L., and J. D. Nichols. 1981. The Mayfield method of estimating nesting success: a model, estimators and simulation results. Wilson Bull. 93:42- 53 Hobson, K. A., and S. G. Sealy. 1989. Responses of yellow warblers to the threat of cowbird parasitism. Anim. Behav. 38:510-519. Holway, D. A. 1991. Nest-site selection and the importance of nest concealment in the black-throated blue warbler. Condor 93:575-581. Hosmei, D. W., and S. Lemeshow. 1989. Applied logistic regression. John Wiley and Sons, New York. 307pp. Howe, R. W., G. J. Davis, and V. Mosca. 1991. The demographic significance of 'sink' populations. Biol. Conserv. 57:239-255. Hubbard, J. P. 1977. Importance of riparian ecosystems: biotic considerations. Pages 14-18 in R. R. Johnson and D. A. Jones, tech. coords. Importance, preservation and management of riparian habitat: a symposium. U.S. For. Serv. Gen. Tech. Rep. RN-43. Hunter, W. C., R. D. Ohmart, and B. W. Anderson. 1988. Use of exotic saltcedar (Tamarix chinensis) by birds in arid riparian systems. Condor 90:113-123. 1 A 0 James, F. C., and H. H. Shugart, Jr. 1970. A duantHtatve method of habitat description. Audubon Field Notes 24:727-736. Johnson, D. H. 1979. Estimating nest success: the Mayfield method and an alternative. Auk 96:651-661. . 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65-71. Johnson, R. R. 1970. Tree removal along southwestern rivers and effects on associated organisms. Report of Comm. on Res. Yearbook of the Amer. Phil. Soc. , L. T. Haight, and J. M. Simpson. 1977. Endangered species vs. endangered habitats: a concept. Pages 68- 74 in R .R. Johnson and D. A. Jones, tech. coords. Importance, preservation and management of riparian habitat: a symposium. U.S. For. Serv. Gen. Tech. Rep. RM-43. Johnston, D. W., and E. P. Odum. 1956. Breeding bird populations in relation to plant succession on the piedmont of Georgia. Ecology 37:50-62. Kelly, S. T., and M. E. DeCapita. 1982. Cowbird control and its effect on Kirtland's warbler reproductive success. Wilson Bull. 94:363-365. Kilgo, J. C., R. A. Sargent, B. R. Chapman, and K. V. Miller. 1996. Nest-site selection by hooded warblers in bottomland hardwoods of south Carolina. Wilson Bull. 108:53-60. Klebenow, D. A., and R. J. Oakleaf. 1984. Historical avifaunal changes in the riparian zones of the Truckee River, Nevada. Pages 203-210 in R. E. Warner and K. N. Hendrix, eds. California riparian systems. Univ. California Press, Berkeley. Krueper, D. J. 1993. Effects of land use practices on western riparian ecosystems. Pages 321-330 in D. M. Finch and P. W. Stangel, eds. Status and management of neotropical migratory birds. U.S. For. Serv. Gen. Tech. Rep. RN-229. 141 Lande, R. L., and G. F. Barrowciough. 1987. Effective population size, genetic variation, and their use in management. Pages 87-123 in M. E. Soulé, ed., Viable populations for conservation. Cambridge Univ. Press, New York. Layman, S. A. 1987. Brown-headed cowbirds in California: historical perspectives and management opportunities in riparian habitats. Western Birds 18:63-70. Lowther, P. E. 1979. Nest selection by brown-headed cowbirds. Wilson Bull. 91:118-121. , and R. F. Johnston. 1978. Influences of habitat on cowbird host selection. Kans. Ornith. Soc. Bull. 28:36-40. Lynn, S. 1996. Bird use of cottonwood-willow patches in the Lower Colorado River Valley. M.S. Thesis, Univ. Arizona, Tucson. 114pp. , A. Averill, and M. L. Morrison. 1996. Neotropical migratory bird monitoring project in the Lower Colorado River Valley. Final Report. U.S. Fish and Wildlife Service, Southwest Research Station, Albuquerque, NM. 229pp. MacArthur, R. H., and J. W. MacArthur. 1961. On bird species diversity. Ecology 42:594-598. Martin, T. E. 1988. On the advantage of being different: nest predation and the coexistence of bird species. Proc. Natl. Acad. Sci. 85:2196-2199. . 1992. Breeding productivity considerations: what are the appropriate habitat features for management. Pages 455-473 in J. M. Hagan and D. W. Johnston, eds. Ecology and conservation of neotropical migrant landbirds. Smithsonian Institution Press, Washington, D.C. . 1995. Avian life history evolution in relation to nest sites, nest predation, and food. Ecol. Monogr. 65:101-127. , and G. R. Geupel. 1993. Nest-monitoring plots: methods for locating nests and monitoring success. J. Field Ornithol. 64:507-519. 142 , and J. J. Roper. 1988. Nest predation and nest-site selection of a western population of the hermit thrush. Condor 90:51-57. Marvil, R. E., and A. Cruz. 1989. Impact of brown-headed cowbird parasitism on the reproductive success of the solitary vireo. Auk 106:476-480; May, R. M., and S. K. Robinson. 1985. Population dynamics of avian brood parasitism. Am. Nat. 126:475-494. Mayfield, H. 1960. The Kirtland's warbler. Cranbrook Institute of Sciences, Bloomfield Hills, MI. 242 pp. . 1961. Nesting success calculated from exposure. Wilson Bull. 73:255-261. . 1965. The brown-headed cowbird with old and new hosts. Living Bird 4:13-28. . 1975. Suggestions for calculating nesting success. Wilson Bull. 87:456-466. . 1977. Brown-headed cowbird: Agent of extermination? Am. Birds 31:107-113. McGeen, D. S. 1972. Cowbird-host relationships. Auk 82:360-380. McShea, W. J., M. V. McDonald, E. S. Morton, R. Meier, and J. H. Rappole. Long-term trends in habitat selection by Kentucky warblers. Auk 112:375-381. Milton, S. J. 1992. Statistical methods in the biological and health sciences. McGraw-Hill, New York. 526pp. Morris, D. W. 1987. Ecological scale and habitat use. Ecology 68:362-369. Morrison, M. L., B. G. Marcot, and R. W. Mannan. 1992. Wildlife-habitat relationships: concepts and applications. Univ. Wisconsin Press, Madison. 343pp. Mumford, R. 1952. Bell's vireo in Indiana. Wilson Bull. 64:224-233. 143 Neale, J. C. C., and B. N. Sacks. 1994. Distribution and abundance of breeding and migratory birds on the Bill Williams River National Wildlife Refuge, Parker, Arizona, spring, 1993. Nice, M. M. 1937. Studies in the life history of the song sparrow. Int. Trans. Linn. Soc. N.Y. Vol. 4. 247pp. Nolan, V., Jr. 1978. Ecology and behavior of the prairie warbler, Dendroica discolor. Ornithol. Monogr. 26:1- 595 Norman, R. F., and R. J. Robertson. 1975. Nest-searching behavior in the brown-headed cowbird. Auk 92:610-611. Norment, C. J. 1993. Nest-site characteristics and nest predation in Harris' sparrows and white-crowned sparrows in the Northwest Territories, Canada. Auk 110:769-777. Norusis/SPSS, Inc. 1990. SPSS base system user's guide. Chicago, IL. 520pp. . 1992. SPSS/PC+ advanced statistics, version 5.0. Chicago, IL. 481pp. Odum, E. P. 1978. Opening address: ecological importance of the riparian zone. Pages 2-4 in R. R. Johnson and J. F. McCormick, tech. coords. Strategies for protection and management of floodplain wetlands and other riparian ecosystems. U.S. For. Serv. Gen. Tech. Rep. WO-12. Ohmart, R. D., B. W. Anderson, and W. C. Hunter. 1988. The ecology of the lower Colorado River from Davis Dam to the Mexico-United States International Boundary: a community profile. U.S. Fish and Wildlife Service Biological Report 85 (7.19). 296pp. , W. O. Deason, and C. Burke. 1977. A riparian case history: the Colorado River. Pages 35-47 in R. R. Johnson and D. A. Jones, tech. coords. Importance, preservation and management of riparian habitat: a symposium. U.S. For. Serv. Gen. Tech. Rep. RM-43. 144 Olson, T. E., and M. V. Gray. 1989. Characteristics of least Bell's vireo nest sites along the Santa Ynez River. Pages 278-284 in D. L. Abell, tech. coord. California riparian systems conference. U.S. For. Serv. Gen. Tech. Rep. PSW-110. Ortega, C. P., J. C. Ortega, and A. Cruz. 1994. Use of artificial brown-headed cowbird eggs as a potential management tool in deterring parasitism. J. Wildl. Manage. 58:488-492. Paton, P. W. C. 1994. The effect of edge on avian nest success: how strong is the evidence? Conserv. Biol. 8:17-26. Patton, D. R. 1992. Wildlife habitat relationships in forested ecosystems. Timber Press, Portland, OR. 392pp. Payne, R. B. 1973. The breeding season of a parasitic bird, the brown-headed cowbird, in central California. Condor 75:80-99. Petit, K. E., D. R. Petit, and L. J. Petit. 1988. On measuring vegetation characteristics in bird territories: nest sites vs. perch sites and the effect of plot size. Am. Midl. Nat. 119:209-215. Petit, L. J. 1991. Adaptive tolerance of cowbird parasitism by prothonotary warblers: a consequence of nest-site limitation? Anim. Behav. 41:425-432. Probst, J. R. and J. Weinrich. 1993. Relating Kirtland's warbler population to changing landscape composition and structure. Landscape Ecol. 8:257-271. Pulliam, H. R. 1988. Sources, sinks, and population regulation. Am. Nat. 132:652-661. Ralph, C. J., G. R. Geupel, P. Pyle, T. E. Martin, and D. F. Desante. 1993. Handbook of field methods for monitoring landbirds. U.S. For. Serv. Gen. Tech. Rep. PSW-GTR-144. 41pp. Raphael, M. G. 1987. Estimating relative abundance of forest birds: simple versus adjusted counts. Wilson Bull. 99:125-131. 145 Reynolds, R. T., J. M. Scott, and R. A. Nussbaum. 1980. A variable circular-plot method for estimating bird numbers. Condor 82:309-313. Rice, J., B. W. Anderson, and R. D. Ohmart. 1983. Habitat selection attributes of an avian community: a discriminant investigation. Ecol. Monogr. 53:263-290. Robbins, C. S., B. Bruun, and H. S. Zim. 1983. Birds of North America: a guide to field identification. Golden Press, New York. 360pp. , D. K. Dawson, and B. A. Dowell. 1989. Habitat area requirements of breeding forest birds of the middle Atlantic states. Wildl. Monogr. 103:1-34. Robertson, R. J., and R. F. Norman. 1976. Behavioral defenses to brood parasitism by potential hosts of the brown-headed cowbird. Condor 78:166-173. Robinson, S. K. 1992. Population dynamics of breeding birds in a fragmented Illinois landscape. Pages 408- 418 in J. M. Hagan III and D. W. Johnston, eds. Ecology and conservation of neotropical migrant landbirds. Smithsonian Institution Press, Washington, D.C. , J. A. Gryzbowski, S. I. Rothstein, M. C. Brittingham, L. J. Petit, and F. R. Thompson. 1993. Management implications of cowbird parasitism on neotropical migrant songbirds. Pages 93-102 in D. M. Finch and P. W. Stangel, eds. Status and management of neotropical migratory birds. U.S. For. Serv. Gen. Tech. Rep. RM- 229. , S. I. Rothstein, M. C. Brittingham, L. J. Petit, and J. A. Grzybowski. 1995. Ecology and behavior of cowbirds and their impact on host populations. Pages 428-460 in T. E. Martin and D. M. Finch, eds. Ecology and management of neotropical migratory birds: a synthesis and review of critical issues. Oxford Univ. Press, New York. Rosenberg, K. V., R. D. Ohmart, and B. W. Anderson. 1982. Community organization of riparian breeding birds: response to an annual resource peak. Auk 99:260-274. 146

W. C. Hunter, and B. W. Anderson. 1991. Birds of the Lower Colorado River Valley. Univ. Arizona Press, Tucson. 416pp. Rosenzweig, M. L. 1981. A theory of habitat selection. Ecology 62:327-335. Rothstein, S. I. 1975. An experimental and teleonomic investigation of avian brood parasitism. Condor 77:250-271. . 1990. A model system for coevolution: avian brood parasitism. Annu. Rev. Ecol. Syst. 21:481-508. • 1994. The cowbird's invasion of the far west: history, causes and consequences experienced by host species. Pages 301-315 in J. R. Jehl, Jr., and N. K. Johnson, eds. A century of avifaunal change in western North America. Studies in avian biology No. 15. Allen Press, Lawrence, KS. 348pp. , J. Verner, E. Stevens, and L. V. Ritter. 1987. Behavioral differences among sex and age classes of the brown-headed cowbird and their relation to the efficacy of a control program. Wilson Bull. 99:322-337. Scott, D. M., and C. D. Ankney. 1980. Fecundity of the brown-headed cowbird in Southern Ontario. Auk:677-683. Scott, P. E. and B. R. McKinney. 1994. Brown-headed cowbird removes blue-gray gnatcatcher nestlings. J. Field Ornithol. 65:363-364. Sealy, S. G., D. L. Neudorf, and D. P. Hill. 1995. Rapid laying by brown-headed cowbirds Molothrus ater and other parasitic birds. Ibis 137:76-84. Serena, M. 1986. Distribution, habitat preferences, and reproductive success of Arizona Bell's vireo (Vireo bellii arizonae) along the lower Colorado River in 1981. Final report to California Department of Fish and Game, Endangered, Threatened, and Rare Wildlife, Project E-W-5, Job IV-38.1. 147 Sherry, T. W., and R. T. Holmes. 1992. Population fluctuations in a long-distance Neotropical migrant: demographic evidence for the importance of breeding season events in the American redstart. Pages 431-442 in J. M. Hagan and D. W. Johnston, eds. Ecology and conservation of neotropical migrant landbirds. Smithsonian institution Press, Washington, D.C. Sokal, R. R., and F. J. Rohlf. 1995. Biometry: the Principles and practice of statistics in biological research. W. H. Freeman and Company, New York. 887pp. Staab, C. 1995. Host and nest selection by brown-headed cowbirds within a riparian area in Arizona. M.S. Thesis, Univ. Arizona, Tucson. 75pp. Stauffer, D. F., and L. B. Best. 1980. Habitat selection by birds of riparian communities: evaluating effects of habitat alterations. J. Wildl. Manage. 44:1-15. Sudbrock, A. 1993. Tamarisk control. I. Fighting back: an overview of the invasion, and a low-impact way of fighting it. Restor. Manage. Notes 11:31-34. Thompson, C. F., and V. Nolan, Jr. 1973. Population biology of the yellow-breasted chat (Icteria virens) in southern Indiana. Ecol. Monogr. 43:145-171. Trail, P. W., and L. F. Baptista. 1993. The impact of brown-headed cowbird parasitism on populations of the Nuttall's white-crowned sparrow. Conserv. Biol. 7:309- 315 Uyehara, J. C., and P. M. Narins. 1995. Nest defense by willow flycatchers to brood-parasitic intruders. Condor 97:361-368. Van Horne, B. 1983. Density as a misleading indicator of habitat quality. J. Wildl. Manage. 47:893-901. Walkinshaw, L. H. 1972. Kirtland's Warbler-endangered. Am. Birds 26:3-9. Weatherhead, P. J. 1989. Sex ratios, host-specific reproductive success and impact of brown-headed cowbirds. Auk 106:358-366. 148 Wiens, J. A. 1986. Spatial scale and temporal variation in studies of shrub-steppe birds. Pages 154-172 in J. Diamond and T. J. Case, eds. Community ecology. Harper and Row, New York. , J. F. Addicott, T. J. Case, and J. Diamond. 1986. Overview: the importance of spatial and temporal scale in ecological investigations. Pages 145-153 in J. Diamond and T. J. Case, eds. Community ecology. Harper and Row, New York. Woodward, P. W. 1983. Behavioral ecology of fledgling brown-headed cowbirds and their hosts. Condor 85:151- 163 Yahner, R. H. 1988. Changes in wildlife communities near edges. Conserv. Biol. 2:333-339. Young, H. 1963. Breeding success of the cowbird. Wilson Bull. 75:115-122. Zar, J. H. 1996. Biostatistical analysis. Third ed. Prentice Hall, Upper Saddle River, NJ. 662pp plus appendices. Zimmerman, J. L. 1983. Cowbird parasitism of dickcissels in different habitats and at different nest densities. Wilson Bull. 95:7-22.