Relationship Between Territorial Male Golden-Winged Warblers in Managed Aspen Stands in Northern Wisconsin, USA
Roth, Amber M
BETWEEN 1966 AND 1990, mean numbers of Golden-winged Warblers (GWWA) detected on North American Breeding Bird Survey routes declined across the breeding range with significant declines in Wisconsin of nearly 4% per annum (P = 0.01; Confer 1992). Additionally, this survey indicated that GWWA declined by 8.1% annually (1966-2000) in the northeastern United States with an estimated 94% reduction in abundance (Sauer et al. 2001. Partners in Flight (Thompson et al. 1993) listed GWWA as one of the highest ranked species of management concern in the Upper Great Lakes Region. The US Fish and Wildlife Service (2002) included GWWA on its list of highest conservation priority species and is assessing its status for possible listing under the Endangered Species Act. Researchers identified maturation of eastern deciduous forests and urbanization of early serai woodlands as possible contributors to regional population declines (Confer 1992). To effectively manage GWWA, we must better understand breeding habitat requirements.
Most recent demographic studies concentrated on habitat associations in the Northeast with little emphasis on habitat types preferred by GWWA in the Midwest. In the western part of the breeding range, only three studies focused on GWWA populations and their habitat associations (Will 1986, Huffman 1997, Gumming 1998). None of these studies investigated relative abundance, density, or productivity among different habitat types or successional stages. As a result of this lack of research, resource managers in the Midwest have little information about relative habitat quality across different vegetation communities and successional stages in the western breeding range. In a pilot study in 1999, we estimated the density of territorial male GWWA in five seedling aspen stands harvested by clearcutting. When we compared GWWA densities from the 1999 surveys (0.45 ±0.13 males/ha) to GWWA densities in the literature, we discovered that the seedling aspen stands we surveyed contained high GWWA densities. The high abundance of GWWA in young, managed aspen stands led to our exploration of forest succession impacts on their habitat use (Roth 2001). We investigated the following three objectives in even-aged aspen stands in northern Wisconsin: (1) determine the aspen successional size class(es) and vegetation structure characteristics associated with high territorial male density, (2) estimate territory area and identify factors associated with territory size, and (3) identify vegetation structure and spatial characteristics associated with territory placement within aspen age classes.
We conducted our research in Lincoln and Oneida Counties in north-central Wisconsin on forests managed principally for aspen production (Figure 1). Forests on all study sites had been harvested previously and were comprised of aspen-birch forest dominated by quaking aspen (Populus tremuloides), big-tooth aspen (Populus grandidentata), and paper birch (Betula papyri/era) with scattered northern hardwood species (Acer spp. and Querem spp.). Adjacent forest types other than aspen forest included conifer swamps (Picea mariana, Abies balsamifera, and Larix laricina), pine plantations (Pinus resinosa and P. banksiana), northern hardwoods, and alder thickets.
Prominent glacial features include end and ground moraines, deposits of glacial till, eskers, kames, drumlins, drift-covered bedrock ridges, and pitted outwash plains and create an undulating to hilly landscape (Boelter 1988, Mitchell 1993). Soils comprised of sandy loam and loamy sand with scattered wet depressions dominated most sites.
Because we observed high densities of territorial males in five seedling aspen stands in 1999, we expanded our GWWA survey efforts in 2000 to examine GWWA density across a successional gradient created by 25 years of clearcutting. In 2000, we surveyed 16 additional aspen stands and grouped four stands in each of the following four size classes: Size Class 1, seedlings (
To estimate territorial male density in each stand, we used the spot mapping survey protocol of the International Bird Census Committee (1970). The purpose of this survey was not to determine the exact territorial boundaries of each male but rather to distinguish between neighboring males for an accurate estimate of male density in each stand. We marked the location of each territory occupied by a singing male on a stand map. Three stands in Size Class 1 were partially harvested on the last harvest rotation resulting in multiple clearcut units within the stand boundaries. We measured male density and vegetation structure within the clearcut boundaries of these stands excluding unharvested portions from our surveys. We estimated stand areas using a Geoexplorer II Geographic Positioning System (GPS; Trimble Navigation 1999).
Territorial male density surveys began 15 minutes prior to sunrise and lasted 4 hours between 10 May and 12 June. The observer completed a search of the entire stand area during the survey period and recorded the location of singing males on maps to minimize the chance of double counting males. In the absence of resources to capture and mark males, we differentiated individuals based on locating simultaneous singing males in adjacent territories. We assumed the male of an occupied territory regularly used a collection of singing perches not used by other males during a survey visit.
An observer visited stands lacking GWWAs two to three times and used taped playback of GWWA vocalizations to confirm the absence of GWWA on the final visit. Stands with at least 1 male GWWA observation were visited three to five times. For stands with eight or more males, a team of three observers worked together only on the last one to two visits to ensure adequate coverage and to reduce the chance of double counting males.
For statistical analysis, we based male density on the highest count of territorial males. For stands with multiple clearcut units, mean values for male density and stand edge were calculated for each stand. To compare male use to the vegetation structure variables and the amount of stand edge (stand perimeter length (km)/stand area (ha)), we selected male density (territorial males/ha) as the dependent variable.
GWWA Territory Area
We randomly selected up to three GWWA territories in each stand to estimate territory area. Territory mapping of the selected territories began immediately following the conclusion of the density survey (approximately 4 hours after sunrise). Though the survey protocol remained the same as for the density survey, our mapping effort was more intense with the objective to accurately estimate territory area.
On the first visit to a stand, each observer randomly selected a singing male observed during the density survey for a detailed spot mapping survey. On subsequent visits, an observer surveyed the same territory to obtain several independent territory area estimates. Observers marked all singing perches used by the selected male with flagging tape using a color scheme unique to each observer. The observer used a GPS receiver to map the perimeter of the territory from one survey day to estimate an index of territory area. Observers repeated this procedure for each day the territory was successfully mapped. We used the mean among visits in our statistical analyses.
We collected vegetation data in GWWA territories and at random points from 22 june to 15 july in 2000. We used three analytical approaches to investigate variation in male densities and territory placement in the four aspen size classes. First, we compared male density and vegetation structure characteristics measured at random sample points throughout stands. Our goal in this analysis was to investigate differences in vegetation structure among size classes to understand mechanisms for male density changes due to aspen succession. second, we described the vegetation structure of territories among size classes. We used this analysis to investigate the work of Confer (1992), who believed GWWA territories had consistent vegetation pattern regardless of community type. Finally, we compared vegetation structure within territories to random sample points to determine if GWWA selected specific vegetation structure during territory placement.
To select random sample points, we created a grid of 12 sample points at least 100 m apart starting at a random point in each stand. In each stand, we measured vegetation at 10 randomly selected grid points. We centered a 1-m2 quadrant on each sample point to estimate percent ground cover, percent litter cover, litter depth, and stem density of woody vegetation groups including aspen suckers, all tree seedlings (including aspen suckers), and dead stems (shrubs and trees). To estimate shrub-layer height-density, we measured visual obstruction of the standing vegetation with a modified Robel pole (6 × 300 cm) centered on the sample point (Robel et al. 1970). From the pole, we walked 10 m in each cardinal direction and recorded the visual obstruction from a height of 1.5 m. We used a spherical densiometer to estimate canopy closure. The Robel pole was used to measure seedling tree, herb, and shrub height, and a clinometer was used to measure regenerating tree height in sapling and pole size classes. To measure woody debris, we extended a tape measure 10 m in the cardinal directions from the sample point and totaled the centimeters of woody slash intercepted by the tape measure. Woody debris was defined as dead, woody material that possessed some vertical structure but did not include standing, dead woody stems (>45° with the ground), litter, or stumps.
Within GWWA territory boundaries, we created a grid of 16 points 10 m apart starting from a random point. A visual obstruction reading was taken from one cardinal direction at each sample point. For the first point, the Robel pole was read from a random cardinal direction. For the subsequent points in the grid, the pole was read from the previous point. To measure the other vegetation structure variables, we used the same protocol for random sample points within the stand.
For comparisons of territorial male density, territory area, and vegetation structure variables across size classes, we used one-way analysis of variance (ANOVA). We also used ANOVA to compare vegetation structure variables at random sample points and from within GWWA territories. For multiple comparisons, we used Fisher’s least significant difference test. To compare vegetation structure between sample points in territories and those selected at random across stands, we used a paired r-test.
For ANOVA and paired f-tests, we determined if all test assumptions (continuous variables, normality, equal variances, and random sampling of the population) were met. The use of the maximum count of territorial males required an arccosine transformation of male density values to produce a normal distribution. We used natural logarithms or square roots to transform independent variables to meet normality assumptions. When transformation of a variable was unsuccessful in producing a normal distribution, we used the nonparametric Kruskall-Wallis test. We analyzed transformed data, but present untransformed values in the tables and text because they are more easily interpreted. We used the modified Levene equal-variance test to test for equal variances and Shapiro-Wilk W test for assessing normality (Hintz 2003). If variances were not equal, we used the Aspin-Weich unequal variances test to compare size classes.
We used simple linear regression to investigate the relationship of male density across all stands with edge and vegetation characteristic independent variables. We selected these independent variables for the regression analyses by using Spearman-rank correlations to investigate the relationship between vegetation characteristics and male density. We only used vegetation variables with correlations >0.35 and examined scatter plots to assess whether the relationship was linear or nonlinear. The absence of nonlinear patterns supported using a linear model. We reviewed residual plots for all regressions to evaluate equality of error variances. All errors for each regression model were checked for independence and normality. To investigate multiple variable regression models, we used an all-possible-regressions procedure, and when multiple variables were correlated with one another, we selected the variable with the highest correlation to the dependent variable for inclusion in the procedure. We used NCSS (Hintze 2003) for allstatistical analyses and set statistical significance at [alpha] = 0.05.
Vegetation Structure of Size Classes
Five of 13 vegetation structure variables measured at random plots differed among aspen size classes (Table 1). In the seedling size class, visual obstruction was significantly higher than in the other size classes. Canopy closure was significantly lower than in the other size classes. Maximum aspen regeneration height and tree regeneration height increased with increasing size class. Aspen sucker density decreased with increasing size class.
GWWA Density Across a Successional Gradient
GWWAs colonized newly created aspen clearcuts in the first breeding season following harvest. We observed 75 territorial males across all size classes in 2000 (Table 2). Stand areas ranged from 11.8 to 132.5 ha (42.0 ± 7.1 ha; mean ± SE). Mean territorial male density (0.55 ± 0.20 males/ha) was significantly higher in the seedling size class than in the other size classes CF^sub 3,12^ = 8.87, P = 0.002). Territorial male density did not vary among the other size classes (Table 2) and was 0.04 (±0.02) males/ha.
Aspen sucker density differed among the aspen size classes (F^sub 3,12^ = 53.34, P 14 = 21.14, P
GWWA Territory Area
Territory area was determined from a mean of 5.5 (±0.2) outer perimeter points. Of the 24 GWWA territories mapped (Table 2), only one occupied territory (in Size Class 3) was abandoned after the initial territory mapping visit. Singing males occupied all other territories through mid-June. Only one territory was identified in Size Class 2, so we combined territory area data for Size Classes 2 and 3. Mean territory area was significantly larger in Size Classes 1 and 4 than in Size Classes 2 and 3 combined (F^sub 2,7^ = 6.12, P = 0.03, Table 2).
Vegetation Structure Within Territories Among Size Classes
We found that six vegetation structure characteristics within territories varied among size classes (Roth 2001). Aspen sucker density (t = 2.3-3.7, P = 0.01-0.04) and tree seedling density (F^sub 3,6^ = 7.52, P = 0.02) were highest in seedling stands. Maximum aspen regeneration height was shorter in Size Class 1 than in Size Class 4 (t = -7.1, P = 0.001). Visual obstruction was higher in Size Classes 1 and 3 than Size Classes 2 and 4 (F^sub 3,6^ = 44.33, P
Territory Placement Within Size Classes of Aspen
We found that five vegetation structure variables differed between random sample plots and territories (Table 4). In Size Class 1, we observed no difference in vegetation structure. In Size Classes 2-4, territories had significantly less canopy closure and shorter maximum tree regeneration height. Visual obstruction was higher in territories than at random sample points.
GWWA Density Across a Successional Gradient
In our study area, territorial male GWWAs were most abundant in seedling size aspen stands. The high GWWA densities observed in regenerating aspen stands concurred with findings from other studies in the Upper Great Lakes Region (Fouchi and Gullion 1984, Steifen 1985, Probst et al. 1992, Wemmer 1993, Gumming 1998). We also found that in sapling and pole-sized aspen stands, low densities of territorial males persisted until 25 years post harvest, primarily in forest gaps, areas of poor aspen regeneration, along edges bordering shrubland cover, and along utility and road rights-of-way. Gumming (1998) found territorial males in aspen stands as old as 80 years. In our study, we associated high male densities with high aspen sucker density. Aspen sucker density was highly correlated with visual obstruction, canopy closure, maximum aspen and tree regeneration height, dead woody stem density, ground cover, and maximum herb height. Aspen sucker density alone explained 60% of the variation in male density between size classes, and shrub height explained an additional 17% of the variation. Therefore, estimating aspen seedling density may be a quick and effective method for measuring habitat quality in aspen communities.
Though the distinctly higher territorial male densities in the seedling size class of aspen may reflect habitat quality, nest productivity along with density were recommended as better indicators (Van Home 1983, Vickery et al. 1992, Donovan et al. 1995, Purcell and Verner 1998). However, we did not possess the resources to examine nest productivity and only two other studies investigated GWWA nest success in the western part of the breeding range. In northern Minnesota, one of three GWWA nests found in young aspen stands successfully fledged at least one young (Hanski et al. 1996). In Michigan, Will (1986) identified 11 of 13 (85%) nests as successful in an aspen forest-abandoned field ecotone. This limited nest productivity information for GWWA in young aspen forests was inadequate to assess the relationship between population densities and habitat quality (Will 1986, Hanski et al. 1996). However, results from nest productivity studies of shrubland bird communities indicated that nest predation was considerably lower in recent aspen clearcuts than in older stands (Yahner and Wright 1985, Yahner and Cypher 1987, Yahner 1991, Hanski et al. 1996). In addition, ground-nesting species like the GWWA were more likely to be successful than aboveground-nesting species in recent deciduous forest clearcuts (Yahner and Cypher 1987, Yahner 1991, Rudnicky and Hunter 1993, Seitz and Zegers 1993, Vander Haegen and DeGraaf 1996). The consistently higher nest success rate for ground-nesting species in recent clearcuts across studies suggested the high male densities of GWWAs observed in our study could be associated with high productivity.
Despite our efforts to avoid double counting territorial male GWWA during density surveys, the possibility existed given the lack of color marking to differentiate males. Will (1986) observed at least one male defending two territories, one on either side of another male’s territory. This “leapfrogging” behavior potentially led to an overestimate of male density in our study, particularly in Size Class 1 stands. However, we believe this behavior was rare and did not influence our survey results.
GWWA Territory Area
Generally, territory area on our study sites was smaller than reported in other studies. Confer (1992) reported a territory area range of 0.4-6.0 ha. In Michigan, Will (1986) measured mean territory area as 1.0 ha (range 0.41-3.25 ha; n = 32) in an aspen forest-field ecotone, and Murray and Gill (1976) observed a range of 1.4-5.2 ha (n = 9). Only Ficken and Picken ( 1968a) in New York documented territory areas (range 0.4-0.8 ha) comparable to those in our study. We believe that the relatively small territories observed in seedling aspen stands may indicate higher quality habitat than the cover types investigated in these other studies. Studies of ovenbirds (Seirus auropallis) reported an inverse relationship between territory area and prey abundance (Stenger 1958, Smith and Shugart 1987). These authors and others (see Newton 1998) suggested that relatively smaller territories were associated with higher quality habitat.
On our study sites, two factors appeared to impact territory area. In the seedling size class, territory areas may have been constrained by neighboring GWWA males. In older aspen stands, territories occurred in openings within the stand and were approximately the same size as the openings occupied. We believe that the distribution of shrubby openings within older stands was an important determinant of territory size for GWWAs in our study. Likely the relatively high standard error associated with territory area in Size Class 4 reflects variation in opening size. The larger territory sizes in the seedling stands relative to the other size classes may not reflect a difference in habitat quality but more likely resulted from these other limiting factors.
Vegetation Structure Within Territories Among Size Classes
Confer (1992) described GWWA territories rangewide as consistently having “patches of herbs, shrubs, and scattered trees, plus a forested edge.” In our study, territorial males preferred the youngest aspen stands characterized by dense aspen sucker regeneration. Our results contradict speculation that GWWA avoided clearcuts with dense regeneration (Confer and Knapp 1981). Furthermore, territories in our young stands consisted of high stem densities (seedlings, shrubs) and rarely contained patches of herbs without woody vegetation as observed by Confer (1992) in successional habitat on abandoned farmland in the Northeast. Studies in the western part of the breeding range also associated the presence of GWWA with relatively dense shrub cover in aspen stands (Huffman 1997, Cumming 1998). Regional variation in the vegetation structure preferences of GWWA may relate to differential availability of various shrub communities across the range rather than to differences in vegetation structure preferences.
Territory Placement Within Size Classes of Aspen
In aspen seedling stands, GWWAs preferred stands with short, dense woody stem densities. In the older stands of our study, we expected territory placement in areas with relatively short regeneration height and vegetation structure like that in Size Class 1 (Confer and Knapp 1981, Collins et al. 1982, Confer 1992, Wemmer 1993). Males selected areas with lower canopy closure, higher visual obstruction, and shorter tree regeneration height. Other researchers generally concurred on a negative association between GWWA site use and increasing canopy closure (Collins et al. 1982, Confer 1992, Huffman 1997, Gumming 1998). The low canopy closure in forest openings allowed the development of a dense shrub layer relative to surrounding vegetation resulting in relatively high visual obstruction readings. The dense vegetation structure near ground level may be important to conceal GWWA nests.
Because GWWAs are generally associated with forest edges (Ficken and Picken 1968b, Will 1986, Frech and Confer 1987), we expected to find high male densities in aspen stands with high perimeter length to area ratios (or high edge ratios). We expected aspen seedling sites to exhibit the most obvious edge effect on GWWA use because they provided a sharper edge transition between the stand areas and adjacent residual forest than sites in the other size classes. In our study, we found no relationship between male density and the amount of edge, however the sample size was small in the analysis of seedling sites. Contrary to Confer’s (1992) observation that all GWWA territories in his study included a forested edge, we observed GWWA territories in the middle of large aspen stands independent of stand edges. For example, in seedling stands, 10% of territories did not include forested edges and these were only in stands >55 ha. We suggest that researchers associated GWWAs with edges in other studies because most shrublands were small and located along the periphery of openings unlike regenerating aspen clearcuts that have contiguous shrubby cover. For example, in abandoned fields commonly used by GWWA in the eastern part of the range, shrubby cover often invaded from the wooded edges and resulted in large areas of grasses and herbs. We speculate that use of edges by GWWA could depend on the distribution pattern of shrubs at a site and may vary across geographic areas.
GWWA Management in Forested Landscapes
Recovery of GWWA populations will require maintenance and creation of early serai plant communities across the breeding range. We suggest that what constitutes early serai plant communities may vary across the geographic range of GWWA and therefore so will GWWA habitat associations. In our work, we limited our surveys to aspen serai stages. We have continued our work with GWWA habitat associations by surveying in other plant community types and found consistently high abundance of GWWA in seedling aspen stands (K. Martin and S. Lutz, unpublished data). In Wisconsin, breeding GWWAs were observed most frequently in brushy clearcuts, lowland shrublands, and edges of hardwood stands, especially aspen (Wisconsin Breeding Bird Atlas, unpublished data). Lowland shrublands could be maintained through protection from development and allowing the natural disturbance regime (e.g. flooding) to occur. Mechanical manipulation of lowland shrub communities will be limited due to federal and state laws protecting wetlands and shorelines, difficulty in using heavy machinery on wet soils, and the cost of habitat manipulation in these areas. Creation of upland early serai plant communities will be logistically easier, cost-effective because of timber revenue, and more easily implemented on a large scale. Extensive early serai aspen stands attractive to GWWAs also can be created in less time and with less effort than is possible through management of lowland shrub habitats.
On our study area, young aspen stands remain attractive to GWWA for a short time. For GWWAs, entire landscapes must be considered if maintaining local populations is a goal. Managers should think of GWWA habitat as ephemeral and plan to maintain a certain percentage of a landscape in early serai forests to provide habitat. On land managed for commercial timber involving heavy machinery, managers should rotate GWWA habitat creation spatially rather than maintaining it at a site indefinitely to prevent site degradation and reduced timber productivity from intensive, short rotation logging systems (Giese et al. 1976, Powers 1989). For example, a staggered 40-year rotation of several adjacent aspen stands would maintain some suitable habitat for GWWAs in an area. This rotation will ensure that GWWA habitat is well distributed across the landscape and maintain timber productivity.
In Wisconsin, aspen forest area declined by 8% between the 1983 and 1996 Forest Inventories (Schmidt 1997). Little aspen forest remains in southern Wisconsin, and the most significant loss occurred in central Wisconsin with a 36% decline. If this pattern of loss continues to shift northward, substantial habitat loss in the core of the GWWA breeding will occur. In Wisconsin, private individuals and public agencies own 46% and 40% of aspen forest, respectively (Schmidt 1997). GWWA population recovery depends on the involvement of these two groups. With the continued negative perception of clearcutting and increasing public pressure on public agencies to reduce clearcutting practices, young aspen forests will continue to decline. Likely, GWWA populations will decline as aspen forests decline. Effective GWWA habitat management will require public education to dispel clearcutting myths.
A majority of the GWWA population breeds in the Upper Great Lakes Region. Therefore Wisconsin, Minnesota, Michigan, and Ontario possess a majority of the responsibility for maintaining breeding habitat for this species. Increasing and improving shrubland cover may be integral to stabilizing the population. Additionally, there are other declining bird species associated with early serai forests (Hunter et al. 2001, Dessecker and McAuley 2001) that will benefit from such management practices. Clearcutting aspen forests will be one of the more effective strategies for creating shrubland cover across the landscape.
Copyright Society of American Foresters Apr 2004
Provided by ProQuest Information and Learning Company. All rights Reserved.