study area and study species
We conducted the study in the forest-dominated mountain range of the southern Black Forest in south-western Germany (47°51′N, 07°58′E). The study area (2812 ha) consisted of capercaillie habitat in two study plots (plot A and plot B; see Appendix S1 in the supplementary material), with altitudes ranging from 900 to 1400 m. The study plots were separated by a steep valley of 5 km width with unsuitable habitat. Plots were demarcated by altitude (> 900 m), paved roads or treeless pastures. The hilltops and valley bottoms were covered with patchy forests and interspersed with pasture. The forests were managed and used intensively for various recreation activities, and were dominated by Norway spruce Picea abies (49%), European silver fir Abies alba (19%) and common beech Fagus sylvatica (22%; Suchant, Baritz & Braunisch 2003). The winter climate in the area is temperate with high precipitation (2000 mm year−1). The area is snow-covered from November/December–April, with a snow depth of 0·5–3 m. Recreation activities in the snow-free season are mainly hiking and mountain biking, and in winter cross-country skiing, downhill skiing (only in the upper part of study plot A; see Appendix S1 in the supplementary material) and hiking. Backcountry and free-riding skiing were frequent in the down-hill skiing area between ski-runs in the upper part of study plot A (see Appendix S1 in the supplementary material). The intensively used cross-country ski-tracks, with up to 1000 visitors a day (R. Roth, personal communication), near a tourist lodge in study plot A (see Appendix S1 in the supplementary material) and a biathlon shooting stand in study plot B (see Appendix S1 in the supplementary material), were mainly confined to forest roads or were situated in unforested areas. Intensive ski tourism starts between early and late December, when the snow depth allows the preparation of ski-runs. Before the ski season starts, only a few hiking tourists visit the study area. The ski season ends with the snow melt in early spring (March/April).
In the entire Black Forest (7000 km2), during the last 30 years, the highly endangered capercaillie population has declined rapidly by about 65%, to 220 males. They persist in isolated fragments of just 510 km2 (Braunisch & Suchant 2006). The study area was located within the core population of the southern Black Forest and is currently inhabited by about 60 individuals.
radio-tracking and habitat use
In September–October 2003 and 2004, seven capercaillie males and eight females were caught in walk-in ground-nets set in bilberry Vaccinium myrtillus-rich forest patches, where capercaillie prefer to search for food (Storch 1993a). There was no bias in age or sex ratio and all were healthy individuals. Males are twice the size of females and were equipped with a 40–69-g backpack radio-transmitter; females were fitted with a 25–40-g backpack transmitter (model GPI, Titley Electronics Ltd, Ballina, Australia; model A1540, Atstrack Advanced Telemetry Systems Inc., Isanti, MN; and model PTT-100, Microwave Telemetry Inc., Columbia, MD). Transmitters were <4% of the birds’ body mass and had no noticeable effect on behaviour. We tested for a possible effect of wearing a transmitter on CM levels but none was detected (see below). During the winters 2003/04–2005/06, between 1 November and 31 March birds were located by ‘homing in’ (Kenward 2001) using a three-element hand-held antenna. Eight birds were tracked during one winter only, seven birds in either two or three winters. To minimize disturbance, most bearings were taken from forest roads or ski-tracks at <1 km distance. The forest canopy was relatively open with many leafless deciduous trees, access was easy and the spatial activity of capercaillie in winter low (Storch 1993b). All bearings were accurate and could be included in the analyses.
We defined two time periods because the area and intensity of human recreational activity changed over the course of the winter. The ‘pre-ski season’ started on 1 November (early winter with leafless deciduous trees, low snow cover) and lasted until the first heavy snow fall, when ski tourism began, usually in December. The ‘ski season’ started with the first day of ski tourism (when the number of tourists increases suddenly from almost zero to >1000) until 31 March, when the habitat use of capercaillie changes because of early lekking behaviour. In both periods, ecological and climatic conditions remain similar, with temperatures mostly <0 °C and capercaillie utilizing their winter habitats. In the ski season we could follow only 10 individuals because three were predated (male 1 and females 1 and 3) and one individual dispersed at the beginning of the study period (female 5; Table 1). All birds with more than 22 radio-locations per time period were included in the analyses. The time between two consecutive radio-locations ranged between 4 hours and 25 days (median 1 day).
Table 1. Number of bearings collected during the pre-ski and ski seasons for seven radio-tracked capercaillie males and eight females, and the size of their home ranges (minimum convex polygon, MCP, in ha), during the pre-ski and ski season for birds with more than 22 bearings per period. Individuals with 22 bearings or less (1–22) were not included in statistical analyses. Bearings were taken over 72 different days during the pre-ski season and over 147 different days in the ski season
|Individuals||Number of bearings||Home range (MCP, ha)|
|Male 1|| 27||(5)|| 59||–|
|Male 2|| 23|| 25||141|| 22|
|Male 3|| 30|| 31||185||137|
|Male 4|| 41|| 63||171||125|
|Male 5|| 49|| 95||208||245|
|Male 6|| 51||102||548||219|
|Male 7||(17)|| 0||–||–|
|Female 1|| 29|| 0||179||–|
|Female 2|| 39|| 26|| 79|| 66|
|Female 3|| 45||(21)||130||–|
|Female 4|| 45|| 67||134|| 58|
|Female 5||(9)|| 69||–||128|
|Female 6|| 43|| 65||118|| 71|
|Female 7|| 52|| 96||367||104|
Home ranges were determined by the minimum convex polygon method (MCP; GIS software ArcView 3·2) separately for the two time periods, pre-ski and ski season (Table 1). The extents of the study area and home ranges were calculated by excluding non-forested areas and very steep slopes >40°, which were unsuitable habitat for capercaillie; no radio-locations were recorded in such habitats.
Capercaillie prefer flat snow-rich mountain ridges and hilltops, and avoid steep slopes (Klaus et al. 1989; Graf et al. 2005). As a measure of habitat quality, we used the steepness of the slope with two categories (SLO1 = 0–10°, SLO2 = 10–40°, based on 50 × 50-m grid cells; DEM, Land Survey Office of Baden-Wuerttemberg, AZ:2851·9/3). We treated the intensity of recreation, and thus the potential of human disturbance, as another component of habitat quality. We defined three classes of recreation intensity. All forested areas without human presence during the winter (e.g. inaccessible areas and away from any ski-tracks) were digitized as REC1 (low recreation intensity). All regularly used tourism infrastructure in forests, such as ski-tracks, ski-lifts, ski-runs, hiking trails, roads, the tourist lodge and the biathlon shooting stand, with a buffer of 50 m, were digitized as REC3 (high recreation intensity; see Appendix S1 in the supplementary material). A previous study had revealed a 90-percentile flushing distance of 50 m (752 flushing events; Thiel et al. 2007a) between capercaillie and an off-trail hiker. All habitats that did not fall in one of the two categories were classified as areas with moderate recreation intensity (REC2). All GIS analyses for the habitat use study were conducted with GIS software ArcGis 9·1.
As the recreation intensity classes did not reflect the same intensities in both time periods, we defined them separately except for REC1. REC1 was associated with no or very low recreation intensity in both time periods. During the pre-ski season, REC2A accounted for off-trail recreation activities with a low intensity level, REC2B in the ski season for off-trail recreation activities with a moderate intensity level. REC3A during the pre-ski season included the area of all on-trail recreation activities within a buffer of 50 m, mostly forest roads with only a few hikers. In the ski season, REC3B included on-trail recreation activities within a 50-m buffer along all intensively used ski-tracks, ski-runs and hiking trails with several hundred recreationists per day.
sampling of droppings
All droppings were sampled from 1 November to 31 March in 2003/04 and 2004/05 in the same study area. We located transmitter-equipped capercaillie every 3–5 weeks to sample their fresh droppings. In addition, fresh droppings from capercaillie without transmitters were collected by walking along contour lines crossing forests and searching for droppings by eye on the surface of the snow (Thiel et al. 2007b). When searching for droppings of capercaillie without transmitters, the same location was visited only three times per winter. We determined the sex of the capercaillie from the size of the intestinal droppings, i.e. the dropping diameter of males is >10 mm, and that of females <8 mm (K. Bollmann, unpublished data). As home ranges of capercaillie in winter are small (Storch 1993b; Table 1), the droppings were spatially clumped. We only sampled fresh droppings and considered droppings from the same sex within a circle of 300 m (28·3 ha) as originating from the same individual, because at low densities capercaillie only occasionally aggregate in flocks (Klaus et al. 1989). Therefore we assigned each dropping of an individual without a transmitter to a ‘potential individual’, as an approach to prevent pseudoreplication in statistical analyses. Moreover, all repeated sampling of droppings from the same location were treated as originating from the same potential individual. As we only sampled fresh droppings and we knew at the time of sampling where the transmitter-equipped birds were located, we could avoid unintentionally sampling droppings from transmitter-equipped birds. We sampled 5–15 droppings at each location (= 1 dropping sample).
For each dropping sample, we determined the following predictor variables for the analyses: RADIO (without or with transmitter), INDIVIDUAL (potential individual for birds without a transmitter or individual transmitter-equipped bird), SEX (male or female), SEASON (pre-ski or ski season) and minimum daily temperature, TEMPMIN, from the nearest meteorological station (DWD Deutscher Wetterdienst) corrected for altitude by 0·6 °C per 100 m. Ambient temperature is known to affect energy metabolism, food intake, dropping production and therefore steroid measurements (Goymann et al. 2006). Furthermore, we determined the type of droppings, DROPTYPE, in three categories (night-roost droppings, droppings excreted during foraging, and droppings excreted during walking or day roosting on the ground; see definitions in Thiel et al. 2007b). Caecal droppings were not sampled because their different composition of micro-organisms affects enzymatic steroid metabolism (Klasing 2005). Each dropping sample was assigned to a slope category, SLOPE, (SLO1, SLO2) and a recreation intensity class, RECREATION (REC1–3). In total, we obtained 106 dropping samples from 14 transmitter-equipped individuals (n = 2–19 per individual, from all individuals except female 1; Table 1). The 290 droppings from birds without a transmitter were assigned to 53 potential individuals (n = 1–14 per potential individual) according to the method described above. All samples were stored at –23 °C until analysis. We did not have to be concerned about diet composition as a determinant of glucocorticoid metabolite concentration (Goymann 2005) because in winter capercaillie feed exclusively on conifer needles (Klaus et al. 1989), particularly on Norway spruce needles in the southern Black Forest (Lieser 1996). Other factors that might affect faecal CM, such as sex (Touma et al. 2003) and temperature (Goymann et al. 2006), were included as factors in our model. Sampling was restricted to winter before the reproduction period started, therefore life-cycle stage was held constant (Huber, Palme & Arnold 2003). The age and storage of droppings (Millspaugh & Washburn 2004) was not a concern because collection took place when temperatures were <0 °C and CM was stable (Thiel, Jenni-Eiermann & Palme 2005).
The faecal concentration of CM reflects the level in the plasma and can therefore be used to biomonitor the endocrine status (Touma & Palme 2005). Glucocorticoids are metabolized in various organs and therefore droppings contain a mixture of several different metabolites with a wide range of polarities. Because glucocorticoid metabolism is often species- and sometimes even sex-specific (Palme et al. 2005), a careful physiological validation of any method to measure faecal CM must be undertaken (Touma et al. 2003). We measured CM in droppings after extraction (60% methanol) with a previously described cortisone enzyme immunoassay (EIA), detecting steroids with a 3,11-dioxo structure (Rettenbacher et al. 2004). This EIA has been proven to measure corticosterone metabolites reliably in capercaillie droppings (Thiel, Jenni-Eiermann & Palme 2005). CM levels in capercaillie droppings remain stable for at least 21 days as long as the ambient temperature is <9 °C, which was the case throughout our study period; time of day does not influence faecal CM level (Thiel, Jenni-Eiermann & Palme 2005).
Because we were interested in measuring the potential long-term effects of recreation intensity on the baseline corticosterone, we wanted to eliminate potential short-term effects of other stressors, such as predator appearance. Moreover, because of differences in corticosterone metabolite concentrations between droppings (Baltic et al. 2005), probably caused by the pulsed excretion of corticosterone by the bile (K. Klasing, personal communication), we collected and homogenized the 5–15 droppings per individual and sampling location to obtain a mean concentration of these metabolites over a longer time span.
We used mixed models (residual maximum likelihood analysis, REML; Patterson & Thompson 1971) to test whether individual home range size varied with the number of bearings per individual (BEARINGS) or with season (SEASON). INDIVIDUAL was included as a random effect, BEARINGS, SEASON and its interaction term as fixed effects.
We combined the categories of the two habitat classifications (REC and SLO), resulting in six habitat types: REC1SLO1, REC1SLO2, REC2SLO1, REC2SLO2, REC3SLO1 and REC3SLO2. To test whether habitat use by capercaillie was influenced by recreation intensity, we applied compositional analysis (Aebischer, Robertson & Kenward 1993) using an Excel macro (Smith 2005) for two types of analyses.
First, we compared the composition of the six habitat types in the home ranges with the availability of the habitat types in the entire study area. This was done for the pre-ski and ski season periods separately, to test whether the home ranges of capercaillie were preferentially located in areas of particular habitat types.
Secondly, we compared habitat use within each home range, as revealed by the composition of the habitat types from the telemetry locations, with availability (habitat composition of the corresponding home range). For both periods separately, we tested whether capercaillie preferred or avoided certain habitat types within their home ranges.
Following Aebischer, Robertson & Kenward (1993), we substituted missing values by a small proportion (0·001%) for available but unused habitat types; 1000 iterations for randomizations were used. For each compositional analysis of capercaillie habitat use, Wilk's lambda (λ) and randomized P-values were reported.
We used REML to identify factors affecting CM levels. Data from all years were pooled, and data from birds with and without transmitters were analysed in the same model, because we tested for any effects of year and of wearing a transmitter by the variable RADIO. The model contained all six predictor variables as fixed effects and the variable INDIVIDUAL as a random effect. Furthermore, we included the three interaction terms SEX × SEASON, SEX × RECREATION and SEASON × RECREATION because we expected them to be biologically relevant. Non-significant interaction terms were omitted from the final model. We used GenStat for Windows version 7·3 (Payne 2003) for the analysis.