AMERICAN WOODCOCK ECOLOGY AND BIRD CONSERVATION IN RELATION TO FOREST MANAGEMENT

The extent of shrubland and young forest in the Northeast, USA, has declined rapidly since the mid-1900’s. Accordingly, the abundance of wildlife that depends on young forest has also declined. For example, American woodcock (Scolopax minor), an upland shorebird species, require an appropriate spectrum and spatial configuration of young forest to thrive and their populations have declined significantly since at least 1968. Active forest management is required to conserve populations of American woodcock and other young forest wildlife, but the importance of young forest management to some aspects of the ecology of key wildlife are not fully understood. I investigated three aspects of American woodcock ecology in relation to young forest management in Rhode Island, USA. First, I monitored the daytime locations of radiomarked American woodcock to assess habitat selection at multiple scales in relation to young forest management. Second, I also monitored American woodcock movements between daytime and nighttime locations and quantified food availability and predator activity at these sites to test the foraging-benefit and predation-risk hypotheses that were proposed to explain American woodcock commuting behavior. Third, I compared landbird communities at managed forest openings used by breeding American woodcock and nearby random forest sites to determine whether American woodcock habitat management benefits non-target landbirds and so verifies adopting American woodcock as an umbrella species useful for conservation. Daytime habitat selected by American woodcock comprised areas of younger forest where the biomass of preferred food (i.e., earthworms [Haplotaxida]) was 1.7 – 3.1 times greater, and the density of shrub and sapling stems was two times greater, compared to random sites. American woodcock home ranges were typically <50 ha and encompassed wetland forests and deciduous or mixed upland forests on flatter slopes ≤1.5 km from streams, agricultural openings, upland young forests, and moist soils. Across Rhode Island, most forested land was in the low – moderate classes of relative probability of use, but young forest management in key areas effectively increased relative use. I illustrated how land managers can use resource selection functions to predict the response of American woodcock to young forest management and so maximize conservation benefits. All of the American woodcock I monitored commuted between dense forest stands and forest openings during the day and night, respectively. I found no support for the foraging-benefit hypothesis because individuals moved from daytime locations where earthworms were 3 – 4 times more abundant to nighttime locations where preferred food was scarce. Soil moisture content was greater at daytime than nighttime locations which may explain why earthworms were more prevalent at those sites. In contrast, I found support for the predation-risk hypothesis because individuals moved from daytime locations where mammalian predators were more active to nighttime locations where mammalian predators were less active. Thus, American woodcock commuted between daytime and nighttime locations to avoid predators and not to feed. Maintaining forest openings is an important part of American woodcock habitat management so that individuals can eat by day and stay safe by night. I identified 38 – 51 bird species during 10-minute point counts at American woodcock singing grounds and random forest sites, and 62 – 73% of the more frequently occurring species were more common at American woodcock singing grounds. On average, 55% of the more common species at American woodcock singing grounds were of high regional or local conservation priority. Young forest species such as prairie warbler (Setophaga discolor) and gray catbird (Dumetella carolinensis) were more abundant at American woodcock singing grounds and scarce or absent at random forest sites while the opposite was true for more mature forest species such as ovenbird (Seiurus aurocapilla) and red-eyed vireo (Vireo olivaceus). Moreover, the total number of birds (all species combined) and diversity of birds were ≥1.5 times greater at American woodcock singing grounds than random forest sites. Critical breeding sites for American woodcock can be managed by clearcutting ≥2-ha patches of older secondary forest and many other young forest bird species of conservation priority inhabit these managed areas. Thus, the American woodcock can serve as an effective umbrella species for young forest birds in the Northeast, USA, but complementary umbrella species such as the ovenbird should be considered to aid in the conservation of more mature forest birds.

studies comparing vegetation structure at nest or roost sites and random sites using traditional null hypothesis testing help describe habitat features associated with reproduction or occupancy for forest birds (e.g., McAuley et al., 1996;Miller and Jordan, 2011;Zahner et al., 2012) and mammals (e.g., Hackett and Pagels, 2003;O'Keefe et al., 2009). More recently, studies of habitat selection have transitioned towards using resource selection functions (RSFs) to understand how probability of use by target species is influenced by environmental covariates (Johnson et al., 2006;Manly et al., 2002;McDonald, 2013). Importantly, these analysis methods allow multiple competing hypotheses to be easily tested using an information-theoretic approach (Anderson et al., 2000;Johnson et al., 2006), facilitate studies of habitat selection across multiple spatial scales (e.g., Johnson et al., 2004), and can be used to predict shifts in probability of use by target species in response to environmental change (e.g., Brown et al., 2007). We investigated habitat selection by American woodcock (Scolopax minor) using both traditional and contemporary analysis methods in order to inform young forest management in the Northeast.
The American woodcock (hereafter woodcock) is a key indicator species of young forest because populations thrive only in landscapes with an appropriate mixture of young forest ranging from forest openings to approximately 30-year-old forest stands (Kelley et al., 2008). Woodcock breed across the eastern USA and adjacent southern and southeastern Canada and winter mainly across the southern half of the eastern USA (Sheldon, 1967), and their populations have declined significantly since 1968 (Cooper and Rau, 2012). Although woodcock are a popular game bird, woodcock survival is similar between hunted and non-hunted sites and so recreational hunting is widespread wetland forest type and Atlantic white cedar (Chamaecyparis thyoides) swamps occurred locally (Enser and Lundgren, 2006).
During 1995, the Rhode Island Department of Environmental Management initiated a forest cutting program to benefit declining populations of woodcock and other wildlife associated with young forest. A series of 2 -5-ha clearcuts in older second-growth forest (e.g., 60 -100 years) were initially made at Great Swamp followed by additional forest management at that site during 2007 and 2012. Similar forest management began at Arcadia and Big River during 1996 and 2006, respectively. Future management at each site is expected to involve additional forest cutting at regular (e.g., 10-year) intervals and, where appropriate, the creation of larger (e.g., 10-ha) young forest patches. At the time of this study, Great Swamp contained the highest proportion of combined upland and wetland young forest (15%) whereas young forest was uncommon at Arcadia (2%) and Big River (1%). Forest openings in the form of abandoned meadows and agricultural fields were also maintained to benefit woodcock and other wildlife, but the relative proportions of these at each site were low (i.e., 1 -2%).

Woodcock capture, marking, and tracking
We caught woodcock from 2 April -16 May 2011 and 2012 (IACUC protocol AN10-02-017) by placing mist-nets at known singing grounds where males engaged in crepuscular courtship displays to attract females for breeding (McAuley et al., 1993;Sheldon, 1967). We included only male woodcock in our study because females are difficult to catch with mist-nets during spring (McAuley et al., 1993). We caught 50 males during 2011 and 42 males during 2012, and determined age by examining plumage characteristics of wings (Sheldon, 1967). After capture, we fitted each woodcock with an Advanced Telemetry Systems two-stage transmitter (Model A5400) using cattle tag cement and a wire belly-band for attachment (package weight ≤4.0 g; McAuley et al., 1993) and released birds on site.
From 23 May -25 August 2011 and 2012, we monitored the daytime locations of each bird 3 -4 times per week. We tracked radio-marked birds on foot using a threeelement antenna and approached each bird until the receiver gave an audible signal without the use of the antenna or headphones. On average, this method allowed us to approach to ≤18 m (Masse et al., 2013) and we marked exact locations in the field using a handheld GPS unit. We located each bird once during each monitoring day (0600 -1900 EST) and stratified our telemetry schedule by time of day during subsequent weeks to ensure that the majority of the daytime period was represented in the sample of telemetry locations for each bird. Male woodcock generally concentrate daytime activity within small areas (Hudgins et al., 1985), called diurnal coverts, and so we approached marked birds from different directions on subsequent visits in order to circumscribe selected coverts. Since we were interested in summer habitat selection we included in our study only those individuals for which we obtained >25 locations throughout each monitoring period. Consequently, we excluded 40 woodcock because they died (2011: n = 4; 2012: n = 4), slipped their transmitters (2011: n = 3; 2012: n = 4), or moved away from study sites and could not be relocated (2011: n = 16; 2012: n = 9) prior to obtaining sufficient numbers of telemetry locations (Table 1).

Habitat sampling
We sampled woodcock habitat at two spatial scales in order to investigate thirdand second-order selection. Third-order selection pertains to specific sites selected by individuals within their home ranges whereas second-order selection pertains to the positioning of home ranges within a broader landscape or geographical range (Johnson, 1980).

Third-order selection
For analysis of third-order selection, we considered diurnal coverts represented by clusters of telemetry locations for each bird similar to Hudgins et al. (1985). For each woodcock, clusters of five or more locations in which each location was ≤100 m of another location were defined as a diurnal covert and we delineated the boundaries of diurnal coverts using minimum convex polygons (MCP; Mohr, 1947). We delineated 1 -3 diurnal coverts for each woodcock, but each bird generally showed preference for a single diurnal covert and so we designated for each bird a primary diurnal covert that contained the most telemetry locations. If an individual woodcock selected multiple diurnal coverts with equal frequency then we randomly selected one to represent the primary diurnal covert. Primary diurnal coverts for 16 of 52 woodcock overlapped to some degree and so in situations where overlap was >20% we randomly selected one woodcock's primary diurnal covert for inclusion. In addition, one woodcock was tracked during both years so we randomly selected one year to include for this individual. As a result, we promoted independence among the primary diurnal coverts included in this portion of our study and ensured that each woodcock (2011: n = 11 at Arcadia, 8 at Big River, and 3 at Great Swamp; 2012: n = 4 at Arcadia, 7 at Big River, and 9 at Great Swamp) was represented equally.
We assessed third-order selection by measuring habitat variables in 5-m radius (0.008-ha) circular plots that were randomly located inside (n = 5) and outside (n = 5) each bird's primary diurnal covert from 24 August -30 September 2011 and 2012.
We used Geospatial Modeling Environment (Beyer, 2013) to randomly select plot locations up to 500 m outside each primary diurnal covert. We enforced a minimum distance of 15 m between plot locations to ensure that plots did not overlap. At the center of each plot, we collected a 10-cm deep soil core and determined soil pH, soil moisture content (% by weight), and soil organic matter content (% by weight) following Masse et al. (2013). We also dug a 900-cm 2 soil pit to 10-cm deep at the center of each plot and collected all earthworms by hand sorting soil pit contents (Dangerfield, 1997). We estimated earthworm density (#/m 2 ) and measured fresh and freeze-dried earthworm weight (g/m 2 ) to the nearest 0.001 g. We calculated canopy closure (%) at the center of each plot using a spherical densiometer (Lemmon, 1957) and visually estimated overstory height class (i.e., 0 -3 m, 3 -9 m, or >9 m) for each plot. We measured diameter at breast height ( , Hudgins et al., 1985;Sepik and Derleth, 1993) and so we also calculated these home range estimates to facilitate comparisons with other studies.
We used a design I study with sampling protocol A (Manly et al., 2002) to assess second-order selection. For each site, we delineated a composite area of available woodcock habitat by pooling individual kernel home ranges across years and circumscribing them with a MCP. We delineated available habitat in this way because individuals frequently moved across the landscape during crepuscular periods (Masse et al., 2013) and so encountered, and selected against, areas outside of their diurnal home ranges. We delineated a composite area of used woodcock habitat for each site by pooling individual kernel core-use areas across years. Since woodcock remain in forested coverts during the day (Dessecker and McAuley, 2001;Hudgins et al., 1985) we clipped composite areas of available and used habitat by forest boundaries using  (Kramer, 1956). We also tested for interactions between plot location and other main effects, but dropped interaction terms that were not significant (i.e., P > 0.05). We verified model assumptions of residual normality using the Shapiro-Wilk test (Shapiro and Wilk, 1965) and by inspecting normal probability plots, and homogeneity of variance by inspecting residual plots. We used multinomial logistic regression (Agresti, 2007) to test the main effects of plot location, age, and year on generalized overstory cover type. We set mixed upland forest as the reference category, specified a mixed model by treating bird identity as a random effect, and used the Gauss-Hermite quadrature approximation method to obtain maximum likelihood estimation (SAS, 2011; PROC GLIMMIX). We did not test the main effect of site on generalized overstory cover type because not all cover types were represented at each site and cover type differences were already evident among management areas (see 2.1.) so we were not interested in further quantifying these differences.
Kernel home range and core-use area size were strongly correlated (r = 0.99) so we tested for differences in home range size only. Kernel home ranges and core-use areas were often divided into multiple parts as a result of woodcock movement patterns and so we counted the number of home range and core-use area divisions for each bird to help characterize this aspect of second-order selection. The number of home range and core-use area divisions were moderately correlated (r = 0.51) so we retained both variables. We log-transformed kernel home range size in order to improve normality and used analysis of variance to test the main effects of age, site, and year on home range size. We adjusted for multiple comparisons, tested for interactions between main effects, and assessed model assumptions as before. We used log-linear regression assuming a Poisson distribution (Agresti, 2007) to test the main effects of age, site, and year on the number of home range and core-use area divisions.
We also tested for interactions between main effects, but dropped interactions that were not significant. We adjusted for slight underdispersion in the number of home range divisions, and slight overdispersion in the number of core-use area divisions, using the deviance scale parameter (SAS, 2011; PROC GENMOD).
We used logistic regression to derive the coefficient values for the exponential form of the RSF [w(x) = exp(β 1 x 1 + … + β p x p )] based on available and used habitat (Manly et al., 2002). Johnson et al. (2006) found this approach to be both theoretically appropriate and quantitatively robust to sample contamination (i.e., available sample containing used and unused resource units) and overlap (i.e., resource units occurring in the available sample and used sample). Contamination of our sample of available habitat was low (6%) and overlap among our samples of available and used habitat was minimal (<1% distance to the nearest upland young forest reduced the relative probability of use by woodcock and the maximum distance within composite areas of used woodcock habitat was 1,000 m, then areas ≤1,000 m from upland young forest were considered more suitable for woodcock while those >1,000 m were considered less suitable. We used the maximum value for each of these variables to select older second-growth upland forest that might be most beneficial for woodcock habitat management. For the second illustration, we considered a 4-km 2 case study area because management of a woodcock habitat mosaic is recommended at this scale (Williamson, 2010). We chose a site in Arcadia where woodcock were known to occur, forest management practices to improve woodcock habitat have previously been implemented, and the estimated relative use by woodcock varied from low to high. We simulated the creation of about 30 ha of upland young forest patches (n = 7; range = 2 -10 ha) and 10 ha of herbaceous forest openings (n = 3; range = 2 -6 ha) within areas of deciduous, mixed, or coniferous second-growth forest deemed most beneficial for habitat management, and then re-calculated the RSF to illustrate how relative use changed in response to forest cutting. For simplicity, we placed hypothetical management units adjacent to roads (i.e., access points) and ≥100 m from the nearest stream. We ignored other criteria which might influence where forest cutting can occur, but vary from region to region (e.g., state or local ordinances).

Third-order selection
We identified 46 diurnal coverts during 2011 and 36 diurnal coverts during 2012.  (Table 2). Overstory height class was within the 3 -9 m interval inside primary diurnal coverts while height class was >9 m outside primary diurnal coverts.
Overstory size class was within the 11.5 -26.7 cm (i.e., pole) interval inside and outside primary diurnal coverts, but overstory trees outside primary diurnal coverts tended towards the 26.8 -41.9 cm (i.e., small sawtimber) interval ( Regardless of site, soil moisture content and soil organic matter content were similar inside and outside primary diurnal coverts whereas shrub and sapling density was 46% greater inside primary diurnal coverts (Table 2).
Generalized overstory cover type differed by plot location (F 4, 248 = 4.58, P = 0.001), but we found no evidence for significant effects of age or year (P ≥ 0.124).
This model suggested that the relative probability of use by woodcock 1) increased with increasing elevation, 2) decreased with increasing slope, 3) was higher in deciduous upland forest, mixed upland forest, deciduous wetland forest, mixed wetland forest, and wetland young forest, but lower in coniferous upland forest, upland young forest, and coniferous wetland forest, and 4) decreased with increasing distance to the nearest stream, agricultural opening, upland young forest, and moist soil ( Table 3). The vast majority of forested land in Rhode Island occurred in the low (445 km 2 ), low-moderate (234 km 2 ), moderate (533 km 2 ), and moderate-high (444 km 2 ) classes of relative use whereas areas of high relative use (46 km 2 ) were widely scattered ( Fig. 2a). Our validation of the RSF revealed adequate fit between observed and expected proportions of pixels in each ordinal class (Χ 2 4 = 0.083, P = 0.999). In addition, the linear regression model relating observed and expected proportions of pixels in each ordinal class (y = 0.921x + 0.016) had an intercept similar to 0 (P = 0.718), a slope >0 (P = 0.011), but near 1, and a high R 2 (0.912) indicating that the RSF was proportional to true probability of use.

Applications of the resource selection function
Older second-growth upland forest (e.g., 60 -100 years) in Rhode Island where woodcock habitat management was deemed most beneficial was within the maximum values of used woodcock habitat (i.e., composite core-use areas; see 2.3.2.) for each quantitative variable that reduced relative probability of use in the top-ranked RSF.
Generally, management of older second-growth upland forest was deemed most beneficial on slopes ≤53% and within 1,211 m of the nearest stream, 1,314 m of the nearest agricultural opening, 1,498 m of the nearest upland young forest, and 639 m of the nearest moist soil. Most (1,281 km 2 ) older second-growth upland forest was located in areas where woodcock habitat management was classified as most beneficial while only 109 km 2 was located in areas where management was classified as least beneficial (Fig. 2b). Across the 4-km 2 case study area, clearcutting 40 ha (10%) to produce young forest and forest openings reduced the 210 ha of forested land in the low class of relative use to 118 ha, increased the 77 ha in both the low-moderate and moderate classes of relative use to 103 ha and 115 ha, respectively, and increased the 22 ha in the moderate-high class of relative use to 38 ha (Fig. 3).

Third-order selection
We found that daytime activity by male woodcock in Rhode Island was highly localized within areas of their home range. In Pennsylvania, USA, the diurnal coverts of male woodcock during April -May were about 0.1 -1.0 ha (Hudgins et al., 1985).
Most (74%) of the diurnal coverts that we identified were within this range, but some were as much as four times larger. Adult females caring for young concentrated daytime activity within areas that were approximately 0.8 -2.6 ha in Minnesota, USA (Wenstrom, 1974), and 1.0 -2.8 ha in Pennsylvania (Caldwell and Lindzey, 1974) so localized habitat selection is not specific to males. Localized habitat selection has also been found during winter months in Alabama, USA, where woodcock activity centers were from 0.4 -5.7 ha (Horton and Causey, 1979). Historically, young forest likely occurred as relatively small, isolated patches resulting from localized natural disturbances (Askins, 2001) so woodcock and other young forest birds likely adapted to exploit small areas of preferred habitat (Askins et al., 2007).
The structure of preferred young forest provides woodcock protection from predators (Dessecker and McAuley, 2001;Keppie and Whiting, 1994;McAuley et al., 1996;Straw et al., 1986), but older forest might also be selected for nesting, broodrearing, or feeding if the density of shrub or sapling stems is sufficient to provide similar protective cover (Dessecker and McAuley, 2001;Williamson, 2010). On average, tree density inside the primary diurnal coverts that we investigated was about 466 stems/ha which was less than the tree density associated with aspen (Populus spp.; mean = 760 stems/ha) and mixed deciduous forest types (mean = 890 stems/ha) selected by woodcock in Michigan, USA (Rabe, 1977), but similar to the tree density associated with nesting and brood-rearing habitat selected by female woodcock (range = 400 -783 stems/ha; Dessecker and McAuley, 2001). In Pennsylvania, optimum basal area of trees and sapling density for daytime habitat was estimated to be 14.3 m 2 /ha and 4,900 stems/ha, respectively, and woodcock generally avoided areas where basal area was ≥20.0 m 2 /ha and sapling density was <1,500 stems/ha (Straw et al., 1986). We found that mean basal area inside primary diurnal coverts was 22.1 m 2 /ha and shrub and sapling density was 21,452 stems/ha. While overstory trees inside primary diurnal coverts tended to be shorter and smaller in diameter (i.e., younger) than those outside (Table 2), woodcock in Rhode Island may currently be selecting the best available forest rather than optimum forest.
The high shrub and sapling density typical of diurnal coverts in Rhode Island may protect woodcock from predators even though the structure of selected coverts differs from young forests that woodcock typically select in other areas of the Northeast.
Indeed, in some areas, understory structure rather than species composition may be most useful for identifying sites selected by woodcock (Rabe, 1977). The shrub and sapling density that we observed inside primary diurnal coverts was over four times greater than the sapling density at sites selected by woodcock in Pennsylvania (Straw et al., 1986) and similar to shrub and sapling densities in areas selected by female woodcock in Minnesota (Morgenweck, 1977) and Maine, USA (McAuley et al., 1996). Moreover, shrub and sapling density was nearly two times greater inside than outside primary diurnal coverts. We only documented eight mortalities among the 60 woodcock that we monitored during summers 2011 and 2012 so woodcock survival is relatively high in Rhode Island. Woodcock survival is also relatively high in Maine (Derleth and Sepik, 1990) where optimum habitat is more widespread. Thus, the shrub and sapling density typical of diurnal coverts in Rhode Island apparently provides similar protective cover as forests selected by woodcock in other parts of the Northeast.
Woodcock typically feed in forested coverts during the day (Masse et al., 2013) and so our findings that woodcock consistently selected forest stands where earthworm availability was at least 46% greater than random sites (  (1993) found no relationship between earthworm dry weight and woodcock habitat selection in Maine, but noted that mean earthworm dry weight was 8.9 g/m 2 at sites selected by woodcock. Mean earthworm density (23.7 earthworms/m 2 ) and dry weight (1.8 g/m 2 ) inside primary diurnal coverts in Rhode Island were generally lower than those found elsewhere in the Northeast. However, earthworm availability was even more limited outside primary diurnal coverts (Table 2) suggesting that woodcock selected areas that could maximize feeding opportunities.

Second-order selection
We found that the size of kernel home ranges and core-use areas for male woodcock in Rhode Island were highly variable, but did not differ by age, site, or year. showed some capacity to re-visit sites used in previous years while also exploiting apparently new areas on the surrounding landscape since one of the four divisions of his 2012 home range overlapped with one of the divisions of his 2011 home range.
All else being equal, relative use by woodcock of forested land tended to be greatest in wetland forest and lowest in coniferous upland forest (Table 3). Wetland forest may be particularly attractive to woodcock as daytime habitat during summer because the moist soils typically associated with this cover type tend to promote higher densities of earthworms and shrub or sapling stems (Williamson, 2010). The fact that relative use was most negatively influenced by coniferous upland forest coincides with our finding that this cover type was less likely to occur inside primary diurnal coverts. Consequently, relative use by woodcock can effectively be increased if older, second-growth, coniferous upland forests are harvested and replaced with upland young forest, or deciduous or mixed upland forest. However, coniferous upland forest may be selected by woodcock during periods of summer drought (Sepik et al., 1983) so some of this forest type should be maintained on landscapes within or around woodcock habitat mosaics.
We also found that relative use by woodcock of forested land decreased at higher slopes and farther distances from the nearest stream, moist soil, upland young forest, and agricultural opening. Woodcock habitat suitability also declined on steeper slopes in West Virginia, USA (Steketee, 2000). In general, woodcock habitat management is considered most beneficial on flatter slopes (Dessecker and McAuley, 2001) perhaps because these areas can better support earthworm populations (Steketee, 2000). Our findings that proximity to streams and moist soils influences relative use supports the views that creating woodcock habitat closer to streams (Williamson, 2010) or across moisture gradients (Dessecker and McAuley, 2001) is most beneficial. The affinity of woodcock to young forest has been well-documented across the Northeast (Hudgins et al., 1985;McAuley et al., 1996;Sheldon, 1967) and so we expected relative use to decrease as distance to the nearest upland young forest increased. However, we were somewhat surprised to find that relative use also decreased as distance to the nearest agricultural opening increased because greater proportions of agriculture on the surrounding landscape reduced woodcock habitat suitability in West Virginia (Steketee, 2000) and were not associated with areas used by woodcock during spring in Pennsylvania (Klute et al., 2000). Variation in the response of woodcock populations to agricultural openings likely relates to the predominant type of agriculture in a region or considered in a given study, but this subject has yet to be investigated. Some agricultural openings provide critical breeding sites for woodcock during spring (Sheldon, 1967) and roosting sites during summer (Dunford and Owen, 1973;Masse et al., 2013) so the proximity of these landscape features to forests used by woodcock has some ecological relevance. Declines of woodcock populations in Pennsylvania from the 1960's -1970's mirrored declines in the extent of pastureland and other cover types used by woodcock (Gutzwiller et al., 1980) so it seems reasonable that forests farther from certain agricultural openings are generally less ideal than those closer to these forest openings.

Applications of the resource selection function
Given the link between declines of woodcock populations and young forest, the American Woodcock Conservation Plan (AWCP) established habitat goals for restoring woodcock densities to those of the 1970's (Kelley et al., 2008). Across the Northeast, >22,000 km 2 of young forest is needed to restore woodcock densities (Kelley et al., 2008) so widespread, active forest management will be required if the goals of the AWCP are to be met. Forest clearcutting is generally regarded as the most efficient method for creating quality woodcock habitat (Dessecker and McAuley, 2001;McAuley et al., 1996;Williamson, 2010). On the one hand, non-game birds which require similar young forest would likely benefit from woodcock habitat management. On the other hand, removing all trees from select areas may be aesthetically displeasing (Gobster, 2001) or viewed as harmful to populations of wildlife that require more mature forest (Wallendorf et al., 2007). Consequently, forest management efforts to create quality woodcock habitat should be strategically-coordinated and scientifically-informed so that conservation benefits are maximized while negative impacts are minimized.
Managing young forest to increase relative probability of use by woodcock of surrounding landscapes can help improve connectivity between habitat patches thereby reducing the negative impacts of habitat patch isolation. Moderate-high and high classes of relative use were widely scattered across Rhode Island (Fig. 2a) and metapopulation theory dictates that immigration to habitat patches decreases as isolation of habitat patches increases (Hanski, 1998). We used our RSF to identify 1,281 km 2 of older second-growth upland forest where habitat management might be most beneficial for increasing relative use by woodcock (Fig. 2b). About 377 km 2 of young forest must be managed in Rhode Island to restore woodcock population densities (Kelley et al., 2008), but this represents roughly 14% of the total land area and is about four times larger than the current extent of young forest in the state (Buffum et al., 2011). A more feasible goal might be to first stabilize the extent of non-coastal upland young forest by clearcutting about 136 ha of older second-growth forest per year over the next 20 years (Buffum et al., 2011). We recommend that forest clearcutting to create habitat for woodcock and other young forest wildlife (e.g., New England cottontail [Sylvilagus transitionalis]) should take place in areas identified as most beneficial for management in order to help meet the goals of the AWCP. In addition, other land management practices such as allowing ≥30-m buffers around agricultural openings to regenerate into young forest benefit woodcock (Williamson, 2010) and increase the extent of young forest without requiring older forest to be cut down.
In the Northeast, woodcock best management practices focus on creating habitat mosaics that provide all necessary components of quality habitat within a 4-km 2 landscape (Williamson, 2010). About 25% of each habitat mosaic should be maintained as young forest by clearcutting blocks >2 ha on a 40-year rotation (McAuley et al., 1996), and occasional herbaceous forest openings (e.g., wildlife openings or old fields) should be maintained to provide breeding sites (e.g., >0.2 ha each; about eight per 40 ha) and roosting sites (e.g., >2 ha each; about one per 40 ha; Williamson, 2010). Clearcutting forest blocks >1 ha has also been recommended to conserve other species of young forest birds (Schlossberg and King, 2007). Moreover, wildlife openings such as old fields provide adequate habitat for some of these species (King et al., 2009). We used our RSF to show that creating 30 ha of upland young forest and 10 ha of herbaceous forest openings at key sites in a 4-km 2 case study area increased relative use by woodcock of surrounding forested land (Fig. 3).
Clearcutting older second-growth upland forest to enhance woodcock habitat is not suitable in all areas so tools that can distinguish where management efforts are likely to be most effective will be useful in forest management decision-making. own RSFs provided they have basic data on used and available or unused sites.
Employing quantitative tools such as RSFs during the decision-making process will help to maximize conservation benefits and facilitate more efficient and effective forest management planning.         (Sheldon 1967;Dwyer et al. 1988). Male woodcock perform courtship flights over these forest openings, called singing grounds, whereas females typically nest and rear young in nearby forests (Sheldon 1967). During summer, woodcock typically spend the day feeding in moist, young, deciduous or mixed hardwood-conifer forests (Sheldon 1967;Straw et al. 1986;Keppie and Whiting 1994;McAuley et al. 1996;Dessecker and McAuley 2001), called diurnal coverts, and then some fly to natural or maintained forest openings at dusk (Mendall and Aldous 1943;Sheldon 1961;Krohn 1971 (Sheldon 1967;Keppie and Whiting 1994). Because woodcock are opportunistic feeders (Sheldon 1967) it is likely that many of these prey items could be obtained at diurnal coverts. Moreover, no one to date has directly tested the foraging-benefit hypothesis by quantifying the availability or diversity of woodcock foods at both the diurnal coverts and nocturnal roost fields used by individuals.
Subsequent research found that little, or no, feeding occurred by woodcock at nocturnal roost fields during summer (Krohn 1970). Furthermore, nocturnal roost fields used by woodcock during summer typically are not conducive to feeding because the soils at these sites tend to be dry, hard, and lacking in potential prey items (Sheldon 1961;Krohn 1970;Wishart and Bider 1976). After flying to nocturnal roost fields, woodcock are usually inactive throughout the night during summer (Dunford and Owen 1973;Owen and Morgan 1975;Wishart and Bider 1977) so these areas likely provide some benefit other than feeding opportunities. Dunford and Owen (1973) suggested that woodcock flew from diurnal coverts to nocturnal roost fields during summer because these areas provided safer refuge from predators. While it is generally accepted (e.g., Williamson 2010), no one to date has directly tested the predation-risk hypothesis.
Our objective was to simultaneously test the foraging-benefit and predation-risk hypotheses for woodcock that fly between diurnal coverts and nocturnal roost fields during summer to determine the benefit afforded to individuals engaging in this behavior. Specifically, we compared the availability and diversity of woodcock foods, soil properties, and mammalian predator activity at both the diurnal coverts and nocturnal roost fields used by individually-marked woodcock. The foraging-benefit hypothesis predicts greater availability and diversity of soil macrofauna at woodcock nocturnal roost fields than diurnal coverts. The predation-risk hypothesis predicts greater mammalian predator activity during the night at woodcock diurnal coverts than nocturnal roost fields.

MATERIALS AND METHODS
We conducted this field study within and around three state wildlife management

Woodcock movements and data collection
We used mist nets to catch adult woodcock on singing grounds, where males perform courtship flights and copulate with females (Sheldon 1967 and nine second-year [SY] males; 2012: ten ASY and 13 SY males and one ASY female) flew from diurnal coverts to nocturnal roost fields on some nights.
For the present study, we monitored the daytime and nighttime locations of radiomarked woodcock 1-3 times per week from 1 July-20 August each year. We tracked each individual on foot using a three-element antenna and used a GPS to determine exact locations once each day (0600-1900 hrs EST) and once each night (2030-0240 hrs). The location of each bird was determined by slowly moving in the direction of the radio signal while reducing the gain of the receiver until the receiver began giving an audible signal without the use of the antenna or headphones. We quantified the accuracy of this technique by placing five transmitters on the ground, approaching each transmitter from each cardinal direction, then measuring the distance to the transmitter once the receiver first started giving an audible signal without the antenna or headphones. On average, we were (mean ± SD) 17.7 ± 8.3 m from transmitters using this technique. Because we were interested in determining which variables cause woodcock to fly between diurnal coverts and nocturnal roost fields, we identified the location of the diurnal covert and nocturnal roost field used during a 6-day period for each bird included in this study. This paired design allowed us to directly compare the foraging-benefit and predation-risk associated with each bird's diurnal covert and nocturnal roost field.
We collected soil macrofauna at the nocturnal roost field and diurnal covert of each woodcock by digging five 900-cm 2 soil pits to 10-cm deep. We flushed each woodcock once at its nocturnal roost field from 8-20 August 2011 and from 9 July-7 August 2012 and centered the first soil pit on the flush point. Four additional soil pits were located 5 m in each cardinal direction from the flush point to provide an overall average density of soil macrofauna at each site. We stored soil pit contents in plastic bags that were tied shut, returned early the following morning, then collected all soil macrofauna by hand sorting similar to Dangerfield (1997) except we sorted pit contents over white plastic bags. On subsequent days we flushed each woodcock once at its diurnal covert and collected potential prey in the same manner, but immediately hand sorted soil pit contents after digging. We counted all soil macrofauna and identified individuals to Order except centipedes (Chilopoda) and millipedes (Diplopoda). Since earthworms are the dominant prey of woodcock (Sheldon 1967) we also weighed fresh and freeze-dried earthworm samples from each site. We tested the foraging-benefit hypothesis using 38 of the available radio-marked woodcock (2011: nine ASY and eight SY males; 2012: eight ASY and 12 SY males and one ASY female) because we had complete information on food abundance at both their diurnal coverts and nocturnal roost fields.
We collected a 10-cm deep soil core from the flush point and two randomly chosen soil pits at each diurnal covert and nocturnal roost field to determine soil moisture content and soil pH during 2011 and these variables along with soil organic matter content during 2012. We measured soil moisture content gravimetrically by drying to a constant weight at 105º C. We measured soil pH using a 1:5 soil/water (mass/vol) ratio (Hendershot et al. 1993) with a glass pH electrode and a pH meter (model UB-10; Denver Instruments). We measured soil organic matter content using the loss-on-ignition method via combustion of oven-dry (105 o C) soil in a furnace at 550º C for 4 hr. We expressed soil moisture content and soil organic matter content as percent by weight.
We quantified mammalian predator activity at the diurnal covert and nocturnal in order to minimize the influence of these activities on predator behavior and ensure that all sites were monitored for predator activity during a similar time period within each year. Because several woodcock might roost in the same forest opening at night (Sheldon 1967) we randomly selected one woodcock from forest openings where >1 radio-tagged bird was present to include in our sample. This reduced our sample size, but was necessary to ensure independence between pairs of observations. We tested the predation-risk hypothesis using 23 of the 38 woodcock (2011: six ASY and five SY males; 2012: four ASY and eight SY males) included in the test of the foragingbenefit hypothesis because these individuals satisfied our independence criteria and we had complete information on mammalian predator activity at both their diurnal coverts and nocturnal roost fields.
We acknowledge that raptors are also important predators of woodcock. Great

Statistical analysis
We calculated the population density of each potential prey item at each diurnal covert and nocturnal roost field used by each woodcock. Since woodcock might not consume all macrofauna found in the soil we also calculated the cumulative density of known woodcock foods (see Keppie and Whiting 1994) at each site. We estimated the richness of soil macrofauna at each site by counting the number of unique taxonomic groups and estimated diversity by calculating the Shannon Index, H', (Magurran 2004) and then converting to diversity (Jost 2006). We used either paired t-tests or Wilcoxon signed-rank tests (Ott and Longnecker 2010) to compare population densities of potential prey, cumulative densities of known prey, richness, and diversity depending on the normality of paired differences. We assessed normality using a combination of histograms, boxplots, or normal probability plots. We also used paired t-tests or Wilcoxon signed-rank tests to compare earthworm weight, soil moisture content, soil pH, and soil organic matter content. We tested for a difference in the number of nights that baited track stations were visited by any mammalian predator during 2011 using  During 2011, all potential prey items found at woodcock nocturnal roost fields were also found at diurnal coverts (Table 1). Average population densities of millipedes and earthworms were about 49 times greater (V = 78.00, P < 0.01) and 3 times greater (t 16 = 2.14, P = 0.02), respectively, at diurnal coverts whereas the average population density of ants was about 10 times greater (V = 5.00, P = 0.04) at nocturnal roost fields. Average population densities of all other soil macrofauna were similar between sites (P ≥ 0.09; Table 1). During 2012, cockroaches (Blattodea), centipedes, and butterfly/moth larvae were unique to nocturnal roost fields, but average population densities of these were extremely low ( Table 2). The average population density of beetles was nearly 3 times greater (V = 26.00, P = 0.01) at nocturnal roost fields while average population densities of earthworms and pillbugs (Isopoda) were approximately 4 times greater (t 20 = 2.52, P = 0.01) and 8 times greater (V = 60.00, P = 0.02), respectively, at diurnal coverts. Average population densities of all other soil macrofauna were similar between sites (P ≥ 0.09; Table 2). During both years, earthworm fresh weight (P < 0.03) and dry weight (P < 0.03) were greatest at diurnal coverts (Figure 1a), and the cumulative density of known woodcock foods was similar between sites (P ≥ 0.39; Figure 1b). During 2011 we found greater richness (t 16 = 2.85, P = 0.01) and diversity (t 16 = 2.30, P = 0.04) of soil macrofauna at diurnal coverts, but these measures were similar between sites during 2012 (P ≥ 0.46; Figure   1c).

Woodcock
Generally, radio-marked woodcock spent the day in forested wetlands, floodplain forests, or moist upland forests and flew to small forest clearcuts, maintained or abandoned herbaceous meadows, or other idle agricultural fields to spend the night. At diurnal coverts, soil moisture content during 2011 (41.6 ± 25.1%) and 2012 (43.3 ± 28.5%) was 1.7 times greater (t 16 = 2.97, P < 0.01) and 1.5 times greater (t 20 = 2.67, P = 0.01), respectively, than at nocturnal roost fields. Soil pH was similar between sites during both years (P ≥ 0.22) and we found no evidence that soil organic matter content differed between sites (P = 0.09).
During 2011, nocturnal mammalian predators visited baited track stations at diurnal coverts more frequently than nocturnal roost fields for about 73% (8 of 11) of the woodcock that we monitored (F 1, 10 = 8.11, P = 0.02; Figure 2 During 2012, the number of days until initial predator visit was approximately 1.8 times greater at nocturnal roost fields than diurnal coverts (t 11 = 2.02, P = 0.03; Figure   3). We photographed raccoon, fisher (Martes pennanti), coyote, red fox, Virginia opossum (Didelphis virginiana), striped skunk, domestic cat, and long-tailed weasel (Mustela frenata) at sites used by woodcock. We also photographed one broad-winged hawk (Buteo platypterus) at a diurnal covert and one red-tailed hawk (Buteo jamaicensis) at a nocturnal roost field.

DISCUSSION
Our results show that the benefit afforded to woodcock that fly between diurnal coverts and nocturnal roost fields during summer is one of reduced predation risk and not novel feeding opportunities. Several lines of evidence support this conclusion.
First, nearly all soil macrofauna that we found at nocturnal roost fields were also found at diurnal coverts and the population densities of potential prey were not consistently greater at nocturnal roost fields. Second, preferred woodcock foods (i.e., earthworms) were always more abundant at diurnal coverts, the cumulative density of known woodcock foods was similar between sites, and the richness and diversity of soil macrofauna was similar or greater at diurnal coverts depending on the year. Third, two separate indices of predator activity suggest that nocturnal mammalian predators are more active at diurnal coverts. Taken together, this evidence provides the first empirical support for the predation-risk hypothesis and against the foraging-benefit hypothesis to explain the function of woodcock commuting between forests and fields during summer.

Why woodcock commute during summer
Previous studies have indicated that woodcock do not move to forest openings at night to feed during summer (Krohn 1970;Dunford and Owen 1973;Owen and Morgan 1975;Wishart and Bider 1977). At a field in Maine, USA, only one earthworm and few other potential woodcock foods including ants, beetle larvae, and spiders (Araneae) were found in soil collected at night at ten woodcock flush points and 20 random points (Krohn 1970). Further, the stomach contents of most birds collected from ten fields at various times during the night contained few if any earthworms or other soil macrofauna (Krohn 1970). In contrast, earthworms were prevalent in the stomachs of birds collected immediately before or after landing in forest openings at night in Maine (Krohn 1970) and Massachusetts, USA (Sheldon 1961). This suggests that feeding occurs predominantly at diurnal coverts prior to flying to nocturnal roost fields.
However, an important difference between our study and previous ones is that our paired design allowed us to directly compare food availability at both diurnal and nocturnal sites for individuals that commuted. Since earthworms are the dominant prey of woodcock (Sheldon 1967) our findings of greater earthworm availability at diurnal coverts further support the conclusion that most feeding likely occurs at these sites. Although some woodcock have been observed feeding soon after moving to forest openings at night during summer (Sheldon 1961) this may simply represent infrequent opportunistic foraging (Sheldon 1967). Generally, woodcock were sedentary after moving to fields at night during summer in Quebec, Canada (Wishart and Bider 1977) and Maine (Dunford and Owen 1973;Owen and Morgan 1975). This contrasts with behaviors observed in forest openings at night during fall and winter in New Jersey (Krohn et al. 1977), North Carolina (Stribling andDoerr 1985), and Louisiana, USA (Glasgow 1958), where woodcock actively fed at night. The reasons for this seasonal difference in behavior are not well understood, but higher food availability at nocturnal roost fields during fall (Krohn et al. 1977)  Fully understanding the function of woodcock commuting behavior during summer is complicated by differential habitat use patterns between age-sex classes.
On average, juvenile males fly to forest openings at night more often than all other age-sex classes from June-October and males tend to fly to forest openings at night more often than females (Sepik and Derleth 1993). In contrast, female woodcock may remain at diurnal coverts or fly to different forested sites at night (Sepik and Derleth 1993). However, females frequently move to forest openings at night during July (Sepik and Derleth 1993). We were not able to determine the regularity with which female woodcock fly to forest openings at night in Rhode Island because of the difficulty associated with catching females (McAuley et al. 1993). Nonetheless, moving to forest openings at night during summer must provide some benefit to both males and females, especially during periods when this behavior is prevalent.

Testing hypotheses about the trade-offs between foraging and predation risk in ecological systems: insights provided by commuting behavior.
Organisms that move between sites within each day (i.e., those that commute) provide  1977;Sepik and Derleth 1993;this study). However, our data show that predation risk during the night is elevated at diurnal coverts compared to nocturnal roost fields.
During periods of high risk, prey species are expected to allocate more time to antipredator behaviors and less time to feeding whereas feeding effort should be increased during periods of lower risk (Lima and Bednekoff 1999). Rather than remaining at diurnal coverts throughout the day and night, woodcock appear to balance the trade-off between feeding and avoiding predators by feeding at diurnal coverts during the day, a time when nocturnal mammalian predators are usually less active, and then moving to nocturnal roost fields at night.
Consistent movement by woodcock to forest openings at night during summer is influenced by the end of the breeding season, the independence of broods, and the postnuptial molt period (Sheldon 1967;Krohn 1971;Owen and Krohn 1973). Thus, during some nights, female woodcock may favor remaining at diurnal coverts, where preferred food availability is greatest, in order to rebuild energy reserves after reproducing (Sepik and Derleth 1993). Male woodcock do not produce eggs or rear young so their decision to move to forest openings at night during summer may be less  Whiskers represent 95% confidence intervals.

Figure 2
Difference in the number of nights that any mammalian predator visited a baited track station at the diurnal covert and nocturnal roost field for each of 11 radio-marked American woodcock males during September 2011 in Rhode Island, USA. Positive bars indicate more nights with a predator visit at diurnal coverts. Negative bars indicate more nights with a predator visit at nocturnal roost fields.

Figure 3
Difference in the number of days until initial predator visit at the diurnal covert and nocturnal roost field for each of 12 radio-marked American woodcock males during August 2012 in Rhode Island, USA. Positive bars indicate more days until initial predator visit at diurnal coverts. Negative bars indicate more days until initial predator visit at nocturnal roost fields. Table 1 Average density (no./m 2 ) of potential prey found in the soils at the diurnal coverts and nocturnal roost fields of 17 radio-  Table 2 Average density (no./m 2 ) of potential prey found in the soils at the diurnal coverts and nocturnal roost fields of 21 radiomarked American woodcock (20 males and one female) during July-August 2012 in Rhode Island, USA

Introduction
Declines of early-successional forests and shrublands (hereafter young forest) and populations of associated wildlife are major conservation concerns in the Northeast, USA (Buffum et al., 2011;Dettmers, 2003;Litvaitis, 2001). Young forest is an important vegetation type that historically was maintained by natural and biological disturbances from wind and ice storms, fires, beavers (Castor canadensis), insect or pathogen outbreaks, and Native Americans (Askins, 2001;Day, 1953;Foster and Aber, 2004;Lorimer, 2001). From the late-1800's to mid-1900's, young forest expanded rapidly across the Northeast in response to widespread abandonment of farmlands that were initially converted from forest to agriculture by European settlers (Foster and Aber, 2004;Foster et al., 1998). However, the extent of young forest in the Northeast declined from highs of about 30 -35% of the land area during the 1960's to ≤3% in some regions by the early-to mid-2000mid- 's (Buffum et al., 2011Trani et al., 2001). Populations of young forest birds and mammals simultaneously declined since at least the 1960's (Dettmers, 2003;Litvaitis, 2001;Sauer et al., 2012). Given that former natural and biological disturbances are unable to maintain sufficient amounts of young forest on contemporary landscapes, conservation planning to support populations of young forest wildlife requires active forest management (DeGraaf and Yamasaki, 2003;Schlossberg and King, 2007).
Conservation planning that requires active habitat management is often complicated because not all species can be managed for simultaneously. Land managers must often set priorities with limited resources, and therefore, may target conservation shortcuts such as indicator, flagship, keystone, focal, or umbrella species to maximize conservation benefits (Lambeck, 1997;Niemi et al., 1997;Noss, 1990;Simberloff, 1998). Umbrella species are usually depicted as large-bodied, wideranging species with vast area requirements (Caro and O'Doherty, 1999;Noss, 1990) although they seldom fit a single description (Fleishman et al., 2000). More generally, umbrella species are simply those whose conservation works to conserve populations of sympatric non-target wildlife (Fleishman et al., 2000). However, the effectiveness of managing for umbrella species has been debated (Andelman and Fagan, 2000;Sattler et al., 2013;Simberloff, 1998;Suter et al., 2002) Rowland et al., 2006;Rubinoff, 2001) may not be as effective. Habitat specialists may be best suited to serve as umbrella species, but ubiquity (or rarity), sensitivity to disturbance, and ease of monitoring should also be considered when identifying potential umbrella species (Andelman and Fagan, 2000;Caro and O'Doherty, 1999;Fleishman et al., 2000). Importantly, the co-occurrence of diverse non-target species is a necessary criterion for an effective umbrella species (Fleishman et al., 2000). Consequently, we investigated one of these criteria, the cooccurrence of non-target landbirds, with managed forest openings used by breeding American woodcock (Scolopax minor) to help verify the status of this bird as an umbrella species.
The American woodcock (hereafter woodcock), a 116 -279-g migratory upland shorebird, might represent an especially effective umbrella species because woodcock populations require a mixture of young forest ranging from forest openings to 30-yearold forest stands (Kelley et al., 2008). Forest openings such as recently managed clearcuts, maintained or abandoned agricultural fields, and wildlife openings composed of scattered shrubs and trees provide necessary singing grounds during spring where males engage in crepuscular courtship displays to attract females for breeding (Sheldon, 1967). Similar forest openings also provide safe nocturnal roost sites during summer (Dunford and Owen, 1973;Masse et al., 2013) and nocturnal feeding or roosting sites during fall and winter (Blackman et al., 2012;Connors and Doerr, 1982;Krohn et al., 1977). In contrast, regenerating forest stands from 2 -30 years old provide nesting and brood rearing habitat for females and daytime cover for both sexes (Kelley et al., 2008). High densities of sapling or small tree stems typical of young forest protect woodcock from diurnal predators (Dessecker and McAuley, 2001;Keppie and Whiting, 1994;McAuley et al., 1996;Straw et al., 1986), but once regenerating forests are >30 years old they become less suitable (Kelley et al., 2008).
In this study, we compared landbird communities at managed forest openings used by breeding woodcock (i.e., woodcock singing grounds) to those at nearby random forest sites. Woodcock singing grounds typically comprise forest openings > 0.2 ha with areas of low herbaceous ground cover and scattered shrubs or small trees (Williamson, 2010). Our objectives were to 1) determine the composition of the breeding bird communities, and 2) compare the abundance and diversity of breeding birds at woodcock singing grounds and nearby random forest sites. If woodcock represent an effective umbrella species then we predict that non-target young forest birds occur more frequently and are more abundant at woodcock singing grounds, and the overall number and diversity of birds is greater at these managed forest openings.

Study area
We conducted this study at three forest-dominated wildlife management areas and Big River (41°37′0″N, 71°36′60″W) was 33 km 2 dominated by coniferous (45%) and mixed upland forests (31%) while deciduous upland forests (8%) and wetland forests (6%) were infrequent (RIGIS, 2012). Coniferous upland forests in the region were dominated by eastern white pine (Pinus strobus) or a mix of eastern white pine and pitch pine (Pinus rigida), mixed upland forests typically contained these species along with various oaks (Quercus spp.), and deciduous upland forests were dominated by red maple (Acer rubrum), hickories (Carya spp.), and oaks (Enser and Lundgren, 2006). Red maple swamps were the typical wetland forest type in the region (Enser and Lundgren, 2006 1995, 1996, and 2006, respectively. Managed clearcuts initially provide woodcock necessary singing grounds during spring (Dessecker and McAuley, 2001;Sheldon, 1967) and roosting sites during summer (Dunford and Owen, 1973;Masse et al., 2013), and also provide important nesting and diurnal cover as saplings, shrubs, and trees regenerate (Kelley et al., 2008;McAuley et al., 1996) and future cutting is expected to include larger (e.g., 10-ha) patches at more regular intervals (e.g., 5 -10 years). Through 2020, about 40 ha of the demonstration area at Great Swamp will be managed for young forest by clearcutting blocks of older secondary forest, and patches of young forest will be maintained at each site on a 30 -40-year rotation. Given past habitat management, the proportion of young forest was highest at Great Swamp (15%) followed by Arcadia (2%) and Big River (1%).
Maintained or abandoned agricultural fields and other herb-dominated forest openings comprised 1 -2% of each management area.

Data collection
We used standard 10-minute, 50-m radius point count surveys (Ralph et al., 1993) to determine the abundance and diversity of bird species at woodcock singing grounds and at randomly selected forest sites. We identified woodcock singing grounds from 2 April Each spring, we identified 15 -20 woodcock singing grounds at Arcadia, 10 -13 at Great Swamp, and 14 at Big River. Woodcock singing grounds were generally located in either 4 -7-year-old clearcuts or wildlife openings with scattered shrubs and trees, but some were located in grasslands or near the margins of agricultural fields. Male woodcock conduct courtship displays to attract females for breeding so the quality of surrounding habitat for nesting and brood rearing (by females) likely influences singing ground use (Dessecker and McAuley, 2001 From 27 May -2 July, we conducted one point count per year at each woodcock singing ground and random forest site. In order to eliminate potential bias from differences in observer ability (Alldredge et al., 2007), the same experienced observer conducted all surveys from 0510 -1045 (EST) during mornings with calm wind and no rain. We navigated to point count locations on foot using a handheld GPS unit and conducted 4 -6 surveys during a given morning. We alternated the timing of point counts at woodcock singing grounds and random forest sites to ensure that surveys at both treatment types were conducted at various times throughout the morning period.
We identified bird species and counted the number of individuals seen or heard within 50 m of each point count location and excluded 'fly-by' species that were observed above the height of the surrounding canopy.

Statistical analysis
We calculated the frequency of occurrence and relative abundance for each species across point counts at each of the three sites and for each year. We also summarized data from each point count location in two separate ways. First, we calculated the total number of birds (all species combined). Second, we estimated the diversity of birds by calculating the Shannon-Weiner Index (H′; Magurran, 2004) and converting to Diversity (D;Jost, 2006). We used a mixed model to test the main effects of location (i.e., woodcock singing ground vs. random forest), site, year, and all interactions on the number and diversity of birds. Interactions that were not significant (P > 0.05) were subsequently dropped from the model. We specified a random intercept corresponding to individual point count locations to account for the repeated structure of our data and we used the Gauss-Hermite quadrature approximation method to obtain maximum likelihood estimation (SAS, 2011; PROC GLIMMIX). We assumed a normal distribution because Shapiro-Wilk tests (Shapiro and Wilk, 1965) and normal probability plots suggested that the number and diversity of birds were normally distributed, and we adjusted for multiple comparisons using the Tukey-Kramer method (Kramer, 1956). For each dependent variable, we ran a separate model for because the former provided us the strongest test for annual differences while the latter provided us the strongest test for site differences.
We used sample-based rarefaction (Colwell et al., 2004) to generate species accumulation curves for woodcock singing grounds and random forest sites at each management area. Woodcock singing grounds that were surveyed during only one year were automatically included in this analysis. For woodcock singing grounds that were surveyed during 2 -3 years, we randomly selected one year to include so that each point count location in this analysis was represented by equal sampling effort.
Likewise, we randomly selected one year to include for each random forest site. We used the program EstimateS 9.1.0 (Colwell, 2013) to extrapolate rarefaction curves to 20 point count surveys and assessed differences in the expected number of species by examining the overlap of 95% confidence intervals (Colwell et al., 2012).

Results
We documented a total of 46 bird species at Arcadia, 38 at Big River, and 51 at Great Swamp (Table A.1). Nineteen species were unique to woodcock singing grounds and ten species were unique to random forest sites at both Arcadia and Big River, whereas 27 species were unique to woodcock singing grounds and 11 species were unique to random forest sites at Great Swamp. Most species were infrequent, but the mean frequency of occurrence was ≥0.20 for 11, 15, and 13 species at Arcadia, Big River, and Great Swamp, respectively; 62 -73% of these species were more common at woodcock singing grounds. At Arcadia, 5 of 7 more common species at woodcock singing grounds and 3 of 4 more common species at random forest sites were either young forest species or species of high conservation priority (Table 1). At Big River, 7 of 11 more common species at woodcock singing grounds met these criteria compared to 1 of 4 more common species at random forest sites (Table 2). In contrast, at Great Swamp, nearly all more common species at woodcock singing grounds (7 of 8) and random forest sites (4 of 5) were young forest species or species of high conservation priority (Table 3). At each site, the relative abundances of the more common species at woodcock singing grounds and random forest sites were dissimilar (Table A.2).
Generally, young forest species such as prairie warbler (see Table A.1 for scientific names), gray catbird, and common yellowthroat were more abundant at woodcock singing grounds and scarce or absent at random forest sites while the opposite was true for more mature forest species such as ovenbird, veery, and red-eyed vireo (Fig. 1).
During 2012 -2013, the number of birds (mean ± SE) was similar at Arcadia (5.15 ± 0.46) and Big River (4.52 ± 0.46; P = 0.590), but at least 1.3 times greater at Great Swamp (6.84 ± 0.46; P ≤ 0.033). The number of birds was 1.7 -2.7 times greater at woodcock singing grounds than random forest sites (P ≤ 0.007; Fig. 2) and we found no evidence for significant effects of year or interactions (P ≥ 0.149). Bird diversity was also similar at Arcadia (3.63 ± 0.28) and Big River (3.21 ± 0.28; P = 0.546), but at least 1.3 times greater at Great Swamp (4.64 ± 0.28; P ≤ 0.035). Bird diversity was 1.5 -2.4 times greater at woodcock singing grounds than random forest sites (P ≤ 0.036; Fig. 2) and we found no evidence for significant effects of year or interactions (P ≥ 0.489).
The cumulative numbers of species expected at woodcock singing grounds was always higher than random forest sites, but the 95% confidence intervals for these estimates overlapped at Arcadia and Big River. At Arcadia, the species accumulation curve for woodcock singing grounds approached an approximate asymptote at about 32 species while the curve for random forest sites began leveling off around 25 species (Fig. 3a). Similarly, the curves for woodcock singing grounds and random forest sites at Big River approached asymptotes at 29 and 23 species, respectively (Fig. 3b). In contrast, rarefaction curves at Great Swamp showed more divergence with approximate asymptotes of 38 species at woodcock singing grounds and 16 species at random forest sites (Fig. 3c).

Bird communities differ between woodcock singing grounds and random forest sites
We found that bird communities at managed forest openings used by breeding woodcock were largely distinct from those at random forest sites. Most or all of the more frequently occurring species that were unique to woodcock singing grounds were young forest species whereas those unique to random forest sites were typically mature forest species (Tables 1 -3). However, at each management area, 2 -3 species (including 1 -2 young forest species) occurred at ≥1 woodcock singing ground and random forest site each year. These species included cedar waxwing and red-eyed vireo at Arcadia (Table 1); black-capped chickadee, chipping sparrow, and eastern towhee at Big River (Table 2); and black-and-white warbler, American robin, and gray catbird at Great Swamp (Table 3). Forest generalists such as black-capped chickadee and red-eyed vireo were detected equally in clearcut stands (ca. 4 -29 years old) and forest reserves (ca. 85 -140 years old) in New Hampshire, USA, whereas cedar waxwing, chipping sparrow, American robin, and black-and-white warbler were more common in managed stands (Welsh and Healy, 1993). Young forests occupy only 3% of Rhode Island's land area (Buffum et al., 2011) and about 86% of the secondary forest in the state is ca. 40 -100 years old (Butler et al., 2012). Given insufficient habitat, some young forests species might select older, less optimal forests or specific forest types (e.g., wetland forests) which may provide similar young forest structure (e.g., American woodcock; Masse, 2014). While some young forest species occurred at ≥1 woodcock singing ground and random forest site each year the relative abundances of these were nearly always higher at woodcock singing grounds ( Fig. 1; Table A.2).
We also found some indication that landscape composition of forest cover types likely influences the composition of bird communities. Indeed, some species that were common at one management area were uncommon at others (Fig. 1). The composition of bird communities often differs between forest types (DeGraaf and Chadwick, 1987; coniferous, mixed, and deciduous upland forest whereas Big River was dominated by coniferous and mixed upland forest, and Great Swamp was dominated by wetland forest (Fig. 2a). Accordingly, species with an affinity to coniferous or mixed upland forests (e.g., chipping sparrow, prairie warbler, and pine warbler [Middleton, 1998;Nolan et al., 1999;Rodewald et al., 2013]) were more common at Arcadia or Big River. In contrast, species favoring dense undergrowth provided by wetland forests (e.g., veery, northern waterthrush, and black-and-white warbler [Bevier et al., 2005;Golet et al., 2001;Kricher, 1995;Whitaker and Eaton, 2014]) were more abundant at Great Swamp. As a result, land managers should consider how the composition of forest cover types influences the composition of bird communities when setting conservation priorities during forest management planning.

The number and diversity of birds differ between woodcock singing grounds and random forest sites
We found that the total number (all species combined) and diversity of birds differed among the three study sites, but were always at least 1.5 times greater at managed forest openings used by breeding woodcock than at random forest sites ( Fig.   2b -c). In New York, USA, bird abundance and diversity were >2 times greater in 6year-old forest clearcuts than more mature even-aged stands (Keller et al., 2003) and bird diversity was greater in forests subjected to clearcutting than forest reserves in New Hampshire (Welsh and Healy, 1993). Forest clearcuts 3 -12 years old also contained greater bird diversity than pole-sized or mature forests in Virginia, USA (Conner and Adkisson, 1975). Managing young forest for woodcock and other species necessarily results in the creation of habitat edges which often enhance wildlife diversity due to increased vegetative complexity or close proximity of disparate vegetation types (Johnston, 1947;Leopold, 1933;Yahner, 2000). Some edge effects (e.g., increased predation or brood parasitism) may be detrimental to forest birds in more agricultural landscapes (Donovan et al., 1997;Hoover et al., 2006), but in forestdominated regions of the Northeast these edge effects may have less of an impact (Rudnicky and Hunter, 1993;Yahner, 2000). Forest clearcuts and wildlife openings provide necessary habitat for young forest species (Chandler et al., 2009;King et al., 2009) and forest generalist or edge-species (this study) which further increases bird diversity at woodcock singing grounds.
Cumulative numbers of bird species expected at woodcock singing grounds and random forest sites were similar at Arcadia and Big River, but greater at woodcock singing grounds than random forest sites at Great Swamp (Fig. 3). Coniferous, mixed, and deciduous upland forests were more prevalent at Arcadia and Big River than Great Swamp (Fig. 2a) (Dessecker and McAuley, 2001;McAuley et al., 2005). Consequently, widespread, active forest management is required to conserve woodcock populations (Kelley et al., 2008). Clearcutting patches of older secondary forest is suggested as the most efficient method for increasing the extent of young forest (Dessecker and McAuley, 2001;McAuley et al., 1996;Williamson, 2010) and >22,000 km 2 of young forest needs to be managed in the Northeast to meet woodcock population goals (Kelley et al., 2008). On the one hand, populations of many other young forest birds have also declined as a result of habitat loss and degradation and are therefore likely to benefit from such extensive young forest management (Brawn et al., 2001;DeGraaf and Yamasaki, 2003). Indeed, of the 22 young forest bird species that we observed, 55 -77% occurred at woodcock singing grounds whereas only 14 -32% occurred at random forest sites (Table A.1). On the other hand, populations of more mature forest species may decline in response to disturbances such as timber harvest (Gram et al., 2003;Wallendorf et al., 2007). Our results suggest that woodcock can serve as an umbrella species for the young forest bird assemblage in the Northeast. Moreover, detailed best management practices provide a specific prescription that public and private land managers can follow to improve woodcock habitat (Williamson, 2010) which further enhances the efficacy of woodcock to serve as a conservation shortcut.
In the Northeast, woodcock habitat can effectively be improved by creating a mosaic of ≥2-ha clearcuts on about 25% of a 200 -400-ha landscape (Williamson, 2010). Maintaining clearcuts on a 40-year rotation provides the necessary spectrum of young forest for woodcock populations to thrive (McAuley et al., 1996;Williamson, 2010). The productivity and survival of young forest songbirds is typically not reduced in smaller patch sizes (Lehnen and Rodewald, 2009;Rodewald and Vitz, 2005), but clearcuts >1 -4 ha are likely to be used by a greater proportion of young forest bird species (Schlossberg and King, 2007). Young forest birds are typically less common or absent in older secondary forest whereas mature forest birds generally avoid young forest during the breeding season (Keller et al., 2003;Wallendorf et al., 2007;Welsh and Healy, 1993). However, recent research suggests that patches of young forest provide important habitat for some mature forest species during the post-fledging period (Chandler et al. 2012;Marshall et al., 2003). Thus, maintaining some young forest on a given landscape should be viewed as a means to maximize biological conservation. We suggest that woodcock can serve as an effective umbrella species in the Northeast, especially for birds breeding in young forests, but complementary umbrella species such as the ovenbird should be considered to aid in the conservation of birds breeding in more mature forests.   Whiskers represent 95% confidence intervals. Dashed lines represent 95% confidence intervals. black-and-white warbler 1, 2, 3 0.04 ± 0.04 0.23 ± 0.07 scarlet tanager 2, 3 0.00 ± 0.00 0.20 ± 0.06 Table 2 Bird species with a mean ± SE frequency of occurrence ≥ 0.20 at woodcock singing grounds or random forest sites based on 10minute, 50-m radius point counts conducted from 27 May -2 July 2012 -2013 at Big River Wildlife Management Area in Kent and Washington Counties, Rhode Island, USA. Scientific names are provided in Table A.1.