Nesting Success of Yellow-breasted Chats: Effects of Nest Site and Territory Vegetation Structure

—The effects of habitat and vegetation characteristics on the reproductive success of Yellowbreasted Chats (Icteria virens) were examined in central Kentucky. During the 1998 breeding season, 49 nests were located and monitored and the characteristics of nest sites and territories determined. Habitats where nests were located were categorized as old field, linear, or clump, and nests were classified as early or late. Chat nests were located in areas with more foliage and lateral cover than unused sites. However, most nests (55%) were not successful, and variables that differed between nest sites and random locations did not appear to influence nesting success. A diverse and, in an evolutionary sense, novel community of predators may eliminate predictably safe nest sites for chats on our study area. Chats in territories with more foliage cover and less canopy cover were more likely to fledge young. Dense foliage may lower the chances of nest predation by increasing the number of potential nest sites in a territory and may also provide better foraging habitat. Received 29 Feb. 2000, accepted 18 July 2000. When choosing a nest site, songbirds may select habitat patches (Martin and Roper 1988) that improve their chances of successfully fledging young. For example, large shrub patches may contain more potential nest sites for a visually searching predator to investigate (Martin and Roper 1988) and more effectively screen nests and the actions of parents than smaller patches (Holway 1991). Habitat features within patches may also influence nesting success. For example, successful Hermit Thrush (Catharus guttatus) nests were characterized by a greater density of white fir (Abies concolor) saplings and greater concealment than unsuccessful nests (Martin and Roper 1988), and successful Hooded Warbler (Wilsonia citrina) nests had more fern (primarily Woodwardia areolata and Polystichum acrostichoides) cover than unsuccessful nests (Kilgo et al. 1996). In contrast to these results, other investigators have failed to detect any relationship between the characteristics of either nest patches or nest sites and nesting success (Filliater et al. 1994, Howlett and Stutchbury 1996, Braden 1999). Clearly, additional data are needed concerning the possible relationship between nest-site selection and nesting success. Habitat features within breeding territories independent of nest sites may also affect reproductive success. For example, nesting suc1 Dept. of Biological Sciences, Eastern Kentucky Univ., Richmond, KY 40475. 2 Corresponding author; E-mail gritchis@acs. eku.edu cess among Northern Cardinals (Cardinalis cardinalis) in Texas was positively correlated with the presence of patchy understory foliage and arthropod biomass (Conner et al. 1986). Similarly, California Gnatcatchers (Polioptila californica) nested earlier and produced more fledglings when territories included more grass and forb cover, perennial structure, and horizontal perennial homogeneity, with less vertical perennial homogeneity and perennial diversity (Braden et al. 1997). Yellow-breasted Chats (Icteria virens) breed in early successional habitats with an abundance of weedy cover and scattered trees (Palmer-Ball 1996). Dense thickets of blackberry (Rubus sp.), multiflora rose (Rosa multiflora), and Japanese honeysuckle (Lonicera japonica) are commonly used for nesting (Dennis 1958, Thompson and Nolan 1973). Although chats typically nest in areas with dense thickets, little is known about the specific vegetational features at nest sites and within territories that might influence their reproductive success. Burhans and Thompson (1999) reported that chats experienced less predation and higher rates of parasitism in large (.5.5 m diameter) nest patches; however, few specific nest site characteristics were measured. The objective of our study was to examine nest site selection by Yellow-breasted Chats in central Kentucky at the territory, patch, and nest site levels and, specifically, to quantify the vegetation structure at chat nest sites and within territories and to determine which, if any, habitat features were correlated with reproductive success. 511 Ricketts and Ritchison • CHAT NESTING SUCCESS METHODS AND MATERIALS We studied chats from April–July 1998 at the Central Kentucky Wildlife Management Area, 17 km southeast of Richmond, Madison County, Kentucky. The area (621 ha) consists of a mosaic of deciduous woodlots, old fields, and fencerows. Old fields were dominated by various herbaceous species plus thickets of smooth sumac (Rhus glabra) and blackberry (Rubus allegheniensis); fencerows included eastern redcedars (Juniperus virginiana) and black locusts (Robinia pseudoacacia) plus thickets of multiflora rose (Rosa multiflora), Japanese honeysuckle (Lonicera japonica), smooth sumac, and blackberry. Beginning in late April, we captured chats in mist nets and banded them with a numbered USFWS leg band and a unique combination of colored plastic bands. We delineated territorial boundaries by monitoring the movements of males and noting the location of singing males and aggressive encounters. Chat nests were located by following females carrying nest material and searching areas where we observed chats. Once located, nests were checked every three to four days until young fledged or the nest was lost to predation. Nests (and territories) from which at least one young fledged were classified as successful, while those from which no young fledged were classified as unsuccessful. To evaluate the potential influence of time on nesting success, nests were also classified, based on the date of initiation, as either early (before 15 June) or late (after 15 June). We evaluated chat reproductive success using both simple nesting success (number of successful nests/total nests) and the Mayfield method (Mayfield 1975). Survival probabilities between early and late nests and among patch categories (below) were compared using CONTRAST (Sauer and Williams 1989, Hines and Sauer 1989). Nest sites were assigned to one of three habitat or patch categories: old field, linear, or clump. Old fields were larger than 1 ha and included various grasses and forbs plus scattered small trees and shrubs. Linear habitats were narrow strips (,10 m wide and .0.5 ha) along fencerows, roadsides, and the edges of woodlots. Clumps were defined as areas smaller than 0.5 ha consisting of small trees and shrubs (,10 m tall) and surrounded by old field habitat. Vegetation structure at chat nests and within territories was quantified using 0.04 ha circular plots (James and Shugart 1970). Nest sites were sampled 2– 30 days after either fledging or nest failure to minimize disturbance. Nest site plots were centered at chat nests. To sample vegetation within territories, the approximate center of each territory was located and three circular plots were then located at random distances from the center at compass bearings of 908, 2108, and 3308 (Conner et al. 1986). Data from these plots were averaged. Variables measured in each plot included percent foliage cover at vertical intervals of below 1 m, 1–2 m, and 2–3 m, number of trees less than and greater than 8 cm diameter at breast height (dbh), percent canopy cover, percent ground cover, foliage height, percent lateral cover at vertical intervals of less than 1 m, 1–2 m, and 2–3 m, and percent cover of dominant understory plants such as grasses, forbs, shrubs, and bare ground (see Larison et al. 1998 for methods of estimation). In addition, nest height, substrate, and concealment were measured in nest-site plots. Percent concealment was calculated by estimating how much of a nest was obscured by foliage when viewed at nest height level from 1 m. Each nest was viewed from the four cardinal directions and an average percent concealment determined. We used stepwise logistic regression to determine which habitat variables best distinguished successful from unsuccessful nests and territories. Successful and unsuccessful nests and territories were first compared using univariate Wilcoxon 2-sample tests (Nadeau et al. 1995, Rabe et al. 1998). Variables with P , 0.15 were included in the initial regression model (Nadeau et al. 1995). Subsequently, variables were included (score x2 statistic) or removed (Wald’s x2 statistic) from logistic regression models using a criterion of P , 0.25. Kendall correlation coefficients were calculated to ensure that variables were not highly correlated (tb . 0.40; Nadeau et al. 1995). Wald’s x2 statistics were used to assess the contribution of individual variables to the model. Overall model significance was based on log-likelihood x2 statistics, classification accuracy (based on a logistic cutpoint of 0.5 to classify nests/ territories as successful or unsuccessful), and the Hosmer-Lemeshow lack-of-fit test (Rabe et al. 1998). Positive parameter coefficients in the logistic regression equations indicated that an increase in the value of a variable increased the probability of a nest-site/territory being successful. Conversely, a negative coefficient indicated that as the value of the variable increased, the probability of the nest site/territory being successful decreased. We also compared nest site and territory vegetation for 19 territories. When a pair made more than one nesting attempt, data were pooled. Values were compared using Wilcoxon 2-sample tests. A stepwise logistic regression model comparing nest site and territory vegetation data was then built using procedures described previously. Using univariate Wilcoxon 2-sample tests, we also compared the characteristics of early and late nests. Nesting attempts that spanned early and late nesting periods (n 5 10) and early nests not sampled until the late period (n 5 7) were deleted to permit better comparison of vegetation around early and late nests. Characteristics of successful and unsuccessful nests for both early and late nests (i.e., successful vs unsuccessful early nests and successful vs unsuccessful late nests) were compared using Wilcoxon 2-sample tests. All analyses were performed using SAS software (ver. 6.09 for VAX Alphaserver; SAS Institute 1989). All values are reported as mean 6 one standard error.