Study area

We conducted research from 1996 to 1999 on the Clinch Mountain Wildlife Management Area (CMWMA) located in the counties of Smyth, Washington, Russell and Tazewell, southwestern Virginia, USA. It encompassed 10,343 ha within the Ridge and Valley province (J. Baker, Virginia Department of Game and Inland Fisheries, unpubl. report). The topography is rugged and diverse with elevations ranging within 600–1,400 m a.s.l.. The dominant cover types in the management area were xeric and mesic deciduous stands that included mixed oak Quercus spp., mixed hardwood, yellow poplar Liriodendron tulipifera and northern hardwoods (J. Baker, Virginia Department of Game and Inland Fisheries, unpubl. report).

Data collection and analysis

From 1996 through 1998 we trapped grouse during fall (September–November) and spring (February–April) using lily-pad style traps with drift fences (Gullion 1965). Sex and age were determined for each bird using standard feather criteria (Servello & Kirkpatrick 1986, Gullion 1989a,b). Each bird was fitted with a 10–11 g necklace-style radio transmitter with a frequency range of 150–151 MHz, a battery life of 12 months and an 8-hour delay mortality sensor (ATS Inc., Isanti, MN). All birds were released at their capture sites.

We estimated the location of each bird by triangulation 2–3 times per week throughout the year. Triangulations consisted of ≥3 azimuths taken within intervals of 30 minutes from telemetry stations with known universal transverse mercator (UTM) coordinates and were taken at all times of the day. We used program LOCATE II (Pacer 1990) to determine the UTM coordinates for each triangulation. When > 3 azimuths were taken, those that best estimated the bird's location were chosen using a Lenth estimator and a goodness-of-fit test (White & Garrott 1990). We calculated the home range size using the adaptive kernel method (Worton 1989) in the animal movement program extension (Hooge & Eichenlaub 1997) designed for ArcView (Environmental Systems Resource Institute 1996, Fearer 1999).

We developed a GIS coverage based on LANDSAT imagery with 22 cover types for CMWMA for our analyses (Fearer 1999). To provide information on habitat structure and species compositions for the GIS cover types, we established 201 stratified random habitat plots with a 20 m radius size in the study area. We chose the plot size to characterize data at a large scale and to approximate the pixel size (30 m) used in the GIS database. We established habitat plots before the GIS database was complete and thus used the following general cover types as strata: deciduous forest, deciduous forest with evergreen understory, mixed forest, riparian zones, open field, shrub/scrub and regeneration cuts (clear cuts 5–20 years old). Habitat data gathered in the plots were based on the methods described by Noon (1981). From the plot centre, we established four transects in the cardinal directions. Using the point intercept method at 4-m intervals along each transect, we recorded canopy cover of deciduous trees, coniferous trees and shrubs as well as percent ground cover of dead woody debris. The height of shrubs to the nearest 0.5 m also was recorded at each 4-m interval. We counted stems (< 8 cm dbh) that intersected a 2-m pole held parallel to and 1.5 m above the ground and perpendicular to the transect while walking the transect. We recorded shrub species within the plot in order of dominance based on percent cover. We also recorded basal area with an angle gauge and the stand height within the circular plot. We overlaid plot locations onto the GIS database upon its completion and summarized the habitat data by cover types. We determined the dominant tree and shrub species within each cover type based on the percent occurrence of each species across all plots within a cover type as well as the relative abundance within each plot.

Table 1.

Landscape metrics (see McGarigal & Marks (1995) for detailed descriptions and computational formulas) used to evaluate landscape characteristic selection by ruffed grouse at Clinch Mountain Wildlife Management Area, Virginia, during 1996–1998.

We tested for selection of nine landscape characteristics (Table 1) and preferential use of 15 cover types at the home range scale and study area scale (Table 2). For the home range scale analysis, we established a 300 m buffer zone outlining the shape of the 95% contour boundary of each home range to examine what was available to the grouse in their immediate area. We chose 300 m based on the average home range size (33 ha) for the birds analysed in our study. Assuming a circular shape, a 33-ha home range would have a radius of approximately 300 m. Thus, we assumed that an area within 300 m of a home range boundary was readily available to, but not used, by the birds. Using ArcView, we determined cover type coverage within each home range based on the home range's 95% contour boundary, and within the total area encompassed by the home range plus its respective buffer (HR+B) from the GIS database. We used FRAGSTATS/ARC to calculate the landscape metrics for each of these coverages (McGarigal & Marks 1995). We treated home ranges and home ranges plus their buffers as paired samples and tested the hypothesis of no difference in landscape characteristics between the samples using the Wilcoxon signed-rank test (Hollander & Wolfe 1973, Minitab Inc. 1996).

Table 2.

Mean use (± SE) of cover types by ruffed grouse (N = 23) and mean availability (± SE) of cover types at home range (HR plus 300-m buffer zone) and study area scales in the Clinch Mountain Wildlife Management Area, Virginia, during 1996–1998.

For analysis at the study area scale, we established 50 random 33-ha circular plots across the study area to estimate average landscape characteristics. We determined cover types within random plots and calculated landscape metrics using FRAGSTATS/ARC. We treated the home range and random plot coverages as two independent samples and tested the hypothesis of no difference in landscape characteristics between the samples using the Wilcoxon rank-sum test (Hollander & Wolfe 1973, Minitab Inc. 1996).

Table 3.

Mean (±SE) values of random plot habitat data by cover type (in %) in the Clinch Mountain Wildlife Management Area, Virginia, during 1996–1998. Data were collected in 20-m radius circular plots located throughout the study area.

We used compositional analysis (Aebischer, Robertson & Kenward 1993) to assess use availability. We determined use and availability of cover types at the home range and study area scales. At the home range scale, used habitat was that located within each bird's home range, and available cover was that located in the home range plus its surrounding 300 m buffer. At the study site scale, used habitat was that located within each bird's home range, and that available was averaged across the entire study area. We determined the percent coverage of each cover type from the Arc View database, and used a S AS algorithm for the compositional analysis calculations (SAS Institute 1989, Ott & Harvey 1997). For compositional analyses, we set α = 0.10 to simplify reporting of the data, but report the P-values for the specific relationships we discuss (Johnson 1999). For the remaining analyses, we chose to report P-values rather than setting an α-value (Johnson 1999).

Table 4.

Dominant tree and shrub species by cover type in the Clinch Mountain Wildlife Management Area, Virginia, during 1996–1998.

Landscape characteristic selection

Study area scale

Ruffed grouse used cover types disproportionately to their availability at the study area scale (F_{14, 9} = 16.2, P < 0.01). Regeneration cuts had the highest rank relative to their availability at the study area scale followed by MDMU and MDEU (see Table 7). Flowever, we did not detect differences in selection among these cover types (P > 0.10). Meso-xeric deciduous and MD ranked fourth and fifth in preference at the study area scale (see Table 7), and both of these cover types were more preferred than XD (P < 0.05).

Table 7.

Ranked preference of cover types by ruffed grouse based on a compositional analysis for home range and study site scales. Data were collected in the Clinch Mountain Wildlife Management Area, Virginia, during 1996–1998.

Landscape characteristic selection

At home range and study area scales, grouse selected areas containing smaller patches, thereby minimizing the core area (> 50 m from the patch edge) within these patches. Grouse home ranges also had high patch densities at both scales, which further illustrates the selection for smaller patches. Grouse were selecting areas where patches were of relatively uniform size and regular (e.g. square) shape and contained more high contrast edge. In fact, the majority of grouse in our study centered their home ranges on these high contrast edges, such as those between regeneration cuts and mature, deciduous stands rather than in the centre of early successional cover.

The majority of the study site was mature, contiguous deciduous forest, and grouse were choosing areas within this landscape that had a greater diversity of cover types. This was reflected by the higher SHDI and number of cover types per hectare in home ranges relative to random plots. Within the areas, however, small-scale differences in cover diversity did not appear to be important.

The study site contained several large, contiguous patches of mature forest that dominated the landscape, and this was reflected in the large LPI value for the study site. Grouse home ranges tended to contain several small, less extensive patches, resulting in a smaller LPI value for home ranges relative to the study area. Often, more than half of a grouse's home range contained one or two patches that provided one or several key habitat requirements, such as cover and food. For example, almost 50% of a grouse's home range may have been in a regeneration cut or some other habitat patch that provided cover and some food sources for most of the year. The rest of its home range contained smaller parts of other patches that provided the remainder of its habitat requirements. When the bird's home range and surrounding area are considered (i.e. the home range plus its 300 m buffer), the size of the regeneration cut would not be different from the average size of the neighbouring patches. However, because about half of the bird's home range was in the regeneration cut patch and only relatively smaller parts of the other patches, the LPI value within the bird's home range would be larger than LPI for the home range and buffer area.

Areas containing several small patches of cover types preferred by grouse should provide the habitat diversity that grouse require throughout the year (Bump et al. 1947, Gullion 1972, 1989c). Patches of relatively uniform size and shape can increase the interspersion of different cover types, providing more cover types within a smaller area. The presence of several small patches vs a few large ones also can maximize the amount of edge in the landscape, another important habitat component for ruffed grouse (Bump et al. 1947, Kubisiak et al. 1980, De-Stefano & Rusch 1984, Schulz, Bakke & Gulke 1989).

Preferential use of cover types

Cover type preference ranking at the home range scale was similar to that at the study area scale. At both scales, small discrepancies between the ranking of the cover types and differences detected between ranks were the result of variation in selection for cover types by individual birds. Although regeneration cuts were preferred most, relative to availability at the study site scale, the amount of regeneration cuts in the home ranges was variable. This produced a large confidence interval for regeneration cut selection, decreasing the probability of detecting significant differences in selection between this and the other cover types.

Early successional forest structure provided by regeneration cuts is an important habitat component for ruffed grouse (Bump et al. 1947, Thompson & Fritzell 1989, Wiggers, Laubham & Hamilton 1992, Thompson & Dessecker 1997). In the absence of such successional habitats, thick shrub understories in mature stands, such as those provided by rhododendron Rhododendron spp. or mountain laurel Kalmia latifolia, can provide adequate habitat for ruffed grouse (Bump et al. 1947, Hale, Johnson & Landers 1982, Epperson 1988, McDonald, Storm & Palmer 1998). The mesic deciduous with evergreen and mixed understory cover types were preferred over the meso-xeric deciduous with mixed and evergreen understories and xeric deciduous with mixed and evergreen understories at both scales. Because all these cover types provided the rhododendron/laurel understory cover component, we expected the selection for these cover types to be similar. Differences in habitat structure may provide some additional explanation for differing ranks. The preferred types had higher shrub cover (mesic deciduous with evergreen understory) or higher stem densities (mesic deciduous with mixed understory) than the less preferred types. Given these differences, mixed deciduous cover types probably provided more adequate habitat than xeric deciduous types.

Differences in the habitat structure and composition between mesic deciduous and xeric deciduous cover types likely also contributed to their different ranks, as the mesic types had greater shrub cover and stem densities than the xeric types. Also, the dominant tree species in the mesic types (black birch, red oak and black cherry) provide a nutritional and diverse food source for grouse (Servello & Kirkpatrick 1987,1989). Thus, the structure and plant species composition of mesic deciduous cover types likely provided better cover and forage than the xeric deciduous cover type.

The preference for regeneration cuts is consistent with the known habitat preferences of ruffed grouse elsewhere within their range. Grouse require a diversity of cover types throughout the year to meet seasonal variations in habitat requirements (Bump et al. 1947, Kubisiak 1978, Gullion 1989c, Thompson & Dessecker 1997). Early successional cover, such as regeneration cuts, or other covers that provide similar understory structures, such as mature forest stands with a thick shrub understory, are the most preferred cover types of ruffed grouse and provide several of the grouse's habitat requirements (Bump et al. 1947, Hale etal. 1982, Epperson 1988, Thompson & Fritzell 1989, Wiggers et al. 1992, Thompson & Dessecker 1997, McDonald et al. 1998). In the eastern and southeastern parts of the ruffed grouse range in North America, rhododendron or laurel can also be important habitat components in winter, as the evergreen leaves of these shrubs provide thermal and escape cover (Bump et al. 1947, Barber 1989). Providing adjacent areas of mature forest that contain several food sources, such as birch catkins, grapes, acorns and cherries, also is important for maintaining a healthy grouse population, especially if rhododendron and/or laurel thickets are the primary source of cover (Norman & Kirkpatrick 1984, Servello & Kirkpatrick 1987,1989, Hewitt & Kirkpatrick 1997).

Based on our results, we recommend creating and maintaining a landscape with high densities of small patches that are of relatively uniform size and shape. Several of these patches should be early successional cover such as regeneration cuts. Regeneration cuts are one of the most preferred cover types, and they also provide an abundance of high contrast edge when interspersed with mature deciduous forest cover. Rhododendron or laurel thickets in moist deciduous forests, also a preferred cover type, may be beneficial as supplemental cover, as they provide thermal cover in winter and a structure similar to early successional forest cover. Mesic stands of mature hardwoods should be well interspersed with these cover types, and the stands should contain several food sources, such as oaks, grapes and birches, to supplement the grouse's diet. There is a multitude of possible landscape configurations that could contain the characteristics and mixtures of cover types we describe (Guthery 1999). Based on the results of our study, we believe creating and maintaining a landscape with these characteristics would provide the highest quality ruffed grouse habitat in the southern Appalachian region.

We believe the approach we have taken here has utility for analysing habitat use by other grouse species in other geographic regions. Comparing habitat within the home range to a buffer surrounding the home range allows determination of relatively small-scale habitat selection. Because the adaptive kernel home range was used rather than individual locations to determine habitat use, the potential effects of telemetry error in determining individual locations, and the associated habitat, was minimized.