Capture

Birds were captured at night by spotlight trapping (Wakkinen, Reese, Connelly & Fischer 1992b). During March and April, we captured 20, 21 and 24 different hens in 1998,1999, and 2000, respectively, and equipped them with 19-g battery-powered telemetry transmitters (Model 5902N, Advanced Telemetry Systems, Isanti, Minnesota, USA).

During the nesting season, we located the birds at 5–7 day intervals to estimate time of nest initiation and nest fate. We located birds with a hand-held, 2-element Yagi antenna and a Telonics TR-2 (Telonics, Mesa, Arizona, USA) receiver. We mapped nest locations using a global positioning system (Garmin GPS 12XL, GARMIN International Inc., Olathe, Kansas, USA), and avoided flushing nesting hens by circling around them at a distance of 15–20 m. To avoid nest detection by predators, we marked nests with rocks or dead shrubs placed nearby. We considered a nest successful if ≥ 1 eggshell remained that had membranes detached from the inside of the shell (Sveum 1995). If there were no eggshells remaining in the nest, we considered it an unsuccessful nest.

Habitat measurements

We used GIS software (Terrain Navigator, Maptech, Andover, Maine, USA) to estimate the distance from a nest to the nearest lek. We also classified the cover type at each grouse nest site. We randomly selected 45 sites within shrub-dominated communities to represent the available habitat. Since grouse were not selected at random from among the entire population, we sampled random sites, which were matched to nest sites (Beck & George 2000). Thus, our inferences concern the individuals sampled and not the entire population. The number of random sites was equal to the number of individual hens nesting. Thus, for each nest site we randomly selected another location within a 0.1 -1 km radius of the nest because individual sage-grouse hens show fidelity to nesting areas during the breeding season (Fischer, Wakkinen, Apa, Reese & Connelly 1993, Young 1994). At the random sites, we measured the same characteristics as at nest sites, except that we measured the shrub nearest the centre of the site for comparison to the nest shrub. This centre shrub was not necessarily the same species as the nest shrub. Random locations falling in habitats that were not used by grouse (e.g. juniper savannah) were discarded and resampled.

We measured the following 12 characteristics of nest sites: distance from nest to nearest known lek, percent cover of shrubs, forbs, perennial grasses, annual grasses, litter, bare ground and rock, nest shrub species length (north-south axis of shrub) and width (east-west axis of shrub), maximum foliage height and visual obstruction. The locations of most leks were known due to access to historic air and ground surveys as well as extensive air surveys conducted during our study period. We used the line intercept method to estimate the percent cover of shrubs (Canfield 1941) and recorded canopy cover by shrub species. Along two 30-m transects oriented in the cardinal directions, with the nest at the centre of the transects, we established line intercepts. We measured the height of the closest shrub within 2 m of each 5-m mark on the transects to the nearest centimetre excluding inflorescence (Sveum 1995). If there was no shrub within 2 m, we did not record a measurement. We also measured residual grass height of the closest perennial grass within 2 m of each 5-m mark on the transects to the nearest centimetre excluding inflorescence (Sveum 1995). This potentially yielded seven shrub and grass height measurements from each of the two transects (a total of 14 shrub and grass heights) plus the height of the nest shrub, i.e. a total of ≤ 15 shrub and grass height measurements. We averaged the shrub height measurements to represent a single shrub height for the site (Klebenow 1969), and grouped all species of shrubs for this measurement. We averaged the residual perennial grass height measurements to represent the average height for each site. We used a Daubenmire frame placed at eight random, systematic locations in each vegetation survey area to measure the percent cover of perennial grasses, annual grasses, forbs, litter, bare ground and rock (Daubenmire 1959). Only rocks of > 10 cm in diameter were recorded as rock cover. We centred the Daubenmire frame lengthwise over the 1, 5, 10, 20, 25 and 30-m marks on both transects. At the 15-m mark (which was centred on the nest), we centred the Daubenmire frame lengthwise on the transect used, and placed it adjacent to the nest (or main stem of the centre shrub for random sites) on either side. We grouped all species of shrub for measuring the total shrub height, and measured the foliage height of residual perennial grasses in the same way as we measured the shrub height. To measure visual obstructions at each vegetation plot, we used a visual obstruction pole (Robel, Briggs, Dayton & Hulbert 1970) marked in 5-cm increments, placed the pole at the 7.5 and 22.5 m marks of the two transects, and viewed it from a distance of 4.5 m and, a height of 150–160 cm. When the pole was positioned in the centre of the transect, we viewed it from four randomly determined directions, and when it was positioned on the two transects, we viewed it from each of two randomly determined directions. From each of these 12 vantage points, we recorded the lowest visible obstruction height to the nearest 5 cm. This resulted in 12 visual obstruction measurements at each site, which we averaged to estimate the visual obstruction at the site in question. We sampled all nest sites in this same manner within a week after the fate (failure or hatching) of a nest was determined.

Statistical analysis

We evaluated cover type selection by comparing used to available sites (Manly, McDonald & Thomas 1993). As neither grouse nor study area was selected at random (although our study area encompassed most of the range of this grouse species in northeastern California), we matched random sites to nest sites. We only used one nest location from each hen for assessment points rather than general areas, as we did not have a complete inventory of used sites. We used logistic regression to estimate a resource selection function for nesting cover type use by omitting both the intercept and the following ratio: probability of sampling a used resource/ probability of sampling an available resource, while leaving only parameter estimates of shrub type in the model (Manly et al. 1993:129). The logistic regression model was structured so that site status (nesting = 1, or random = 0) was the response variable, while low sagebrush, mixed shrub and big sagebrush were categorical predictor variables. The general structure of our model was:

where logit (π′) = log_e[π′/( 1 -π′)]. In this example, π′ represents the estimated probability of a site being used for nesting; β₁, β₂, and β₃ represent the parameter estimates for the three shrub-dominated cover types; LOW = 1 if the site was a low sagebrush type and 0 if it was a mixed shrub or big sagebrush type; MIXED = 1 if the site was a mixed sagebrush type and 0 if it was a low or big sagebrush type; BIG = 1 if the site was a big sagebrush type and 0 if it was a low or mixed sagebrush type. Thus, we modelled the probability of a site being used for nesting given shrub cover type. This analysis was relevant to inferences about general habitat use patterns, but not to structural components of habitats, which are probably important to sage-grouse habitat selection (see below).

To maintain independence of samples, we randomly selected one nest site for each of the radio-marked 45 hens that nested during the three years for analyses of both cover type selection and microhabitat selection. From the microhabitat analysis we excluded five nests that had been abandoned by hens because desertion may have been caused by the observers. Thus, our sample size was 40 independent nests. We first randomly selected a single successful nest without replacement for individual birds; we then randomly selected one failed nest per bird from the remaining individuals. To minimize multicollinearity and redundancy of variables, we excluded some habitat variables that were correlated. We used the variable that was most easy to interpret biologically. Thus, we used distance from nest to lek, perennial grass cover, litter cover, bare ground cover, rock cover, sagebrush cover, other shrub cover, total shrub height, residual perennial grass height and visual obstruction height for analyses. However, we omitted distance from nest to lek when comparing random and nest sites because random sites were spatially associated with the nest sites. We tested the normality of the variables by examining normal probability plot and homoscedasticity using the Modified-Levene Equal-Variance Test (Hintze 1997). We used the following transformations for variables that were not normally distributed: log transformation for distance to lek; and arcsine square root transformation for all percentage cover variables (Zar 1996). Total shrub height had unequal variance, so we transformed this variable using the reciprocal of the square root (Zar 1996). With the transformed values, we used two-factor multivariate analysis of variance (MANOVA) to test for year effects and to compare nest sites with random sites (Zar 1996). We used MANOVA rather than matched pairs logistic regression to make our results more comparable to the Oregon studies as suggested by the Western States Sage and Columbian Sharp-tailed Grouse Workshop (Schroeder et al. 1999:17). We performed another two-factor MANOVA to compare successful and unsuccessful nests and effects of year. However, we examined year effects only for 1999 and 2000 because we changed some of the vegetation sampling procedures. We resampled the 1998 nest sites in 1999 on the same date (±1 week) that we determined its fate in 1998. If the resulting Wilks' Lambda test statistic for each of the MANOVAs resulted in a P-value of 0.05, we tested each of the individual variables with an ANOVA. We also computed the descriptive statistics for characteristics of successful, unsuccessful and random sites using the statistical program NCSS 97 (NCSS Statistical Software, Kaysville, Utah, USA).

Nesting activity

We radio-marked 65 different female greater sagegrouse, and monitored 20, 37 and 46 hens during March-August in 1998,1999 and 2000, respectively. In the three years, 27, 13 and five hens (i.e. N = 45 independent grouse nests) attempted to nest, while 20 hens were not known to nest. Over the three years, we found 88 nests as some hens renested, and of these five, 14 and 16, respectively, were successful nests, resulting in an overall nest success of 39.8% (26.3,35.9 and 45.7% for 1998, 1999 and 2000, respectively).

Cover type selection

The greater sage-grouse selected low sagebrush, big sagebrush and mixed shrub cover types for nesting. Of the 45 nest sites, 30 (67%) were in big sagebrush cover type, 13 (29%) in mixed shrub and two nests (4%) were in low sagebrush. In contrast, of the 45 random locations, 58,13 and 29% were in big sagebrush, mixed shrub and low sagebrush cover types, respectively. The resource selection function for grouse nesting habitat was:

The 95% profile likelihood confidence intervals (Cl) indicated that grouse avoided low sagebrush (95%CI = -4.48, -0.25). In contrast, parameter estimates were positive for mixed shrub and big sagebrush, but the confidence intervals included zero (95% Cl for MIX = -0.43,2.18; BIG = -0.54,0.84), indicating that these habitats were used as expected based on their availability within their potential breeding home range.

Table 1.

Habitat characteristics of successful (N = 20) and unsuccessful (N = 20) sage-grouse nests and random sites (N = 40) in the county of Lassen, California, during March-July 1998–2000. Numbers in parentheses are standard errors of the mean.

Microhabitat selection

There was no year effect between 1999 and 2000 (F_[df=9,68] = 0.89; P = 0.48), and the habitat structure between nests and random sites was similar (F_[df=9,68] = 0.96; P = 0.54).

There was no difference between successful and unsuccessful nests due to year for 1999 and 2000 (F_[df=10,27] = 1.55; P = 0.19). However, the habitat characteristics differed between successful and unsuccessful nests (F_[df=10,27] = 2.39; P = 0.12). Successful grouse nests were farther (3,588 m, SE = 811 m) from the nearest lek (F_[df=10,27] = 3.85; P = 0.06) than unsuccessful nests (1,964 m, SE = 384 m). In addition, successful nests had greater rock cover (F_[df=10,27] = 4.70; P = 0.04), greater total shrub height (F_[df=10,27] = 10.63; P < 0.01) and greater visual obstruction height (F_[df=10,27] = 5.69; P = 0.02) than unsuccessful nests (Table 1).

Table 2.

Plant species under which successful (N = 31) and unsuccessful (N = 57) sage-grouse nests were placed in the county of Lassen, California, during March-July 1998–2000. N gives the number of nests in each category, the outer right %-column in each category gives the percentage of nests in the respective category relative to all (88) nests located, and the other %-column gives the percentage of nests relative to the number of total successful (N = 31) or unsuccessful (N = 57) nests.

Management implications

Guidelines for managing greater sage-grouse habitats emphasize the importance of maintaining sagebrush for nest sites (Braun, Britt & Wallestad 1977, Connelly et al. 1991, Gregg 1991, Hanf et al. 1994, DeLong et al. 1995, Sveum et al. 1998, Connelly, Schroeder, Sands & Braun 2000). Further, Schroeder et al. (1999) suggest that greater sage-grouse habitat management should be directed toward specific nesting habitats. The birds in our study exhibited different habitat selection preferences at different spatial scales because they chose nest sites in cover types with taller shrubs rather than low sagebrush, and habitat structure used by successful nesting hens differed from that of unsuccessful hens. The hens in our study nested under big sagebrush less frequently than in other studies, which could be due to a lack of sagebrush habitat or a difference in selection patterns by these birds (i.e. both sagebrush and mixed shrub cover types were selected in proportion to availability).

The native shrub steppe in our study area has been degraded by excessive grazing, juniper encroachment, agriculture and anthropogenic development. Therefore, our results are equivocal regarding the following two alternative explanations for our results. First, habitat quality is lower in northeastern California because of natural conditions or human impact, and, consequently, grouse nest in areas that are not typical of other areas within the species' range. Second, grouse habitat is naturally more heterogeneous (i.e. more variable) at the edge of its range in northeastern California, and, consequently, grouse habitat selection patterns reflect this variation. Nevertheless, our results demonstrate the use of variable nesting sites. We suggest some habitat heterogeneity (i.e. structural and shrub species variability within habitats) and landscape heterogeneity (i.e. different shrub cover types) may be desirable in northeastern California. Our results show that not only sagebrush, but also a diversity of shrub species and structural variability can be important for nesting. However, we caution that continued juniper and exotic grass (e.g. Bromus tectorum) encroachment in this area may further degrade these shrub communities. Finally, while we attempted to make our methods comparable to other studies within our region, there is no standard methodology for examining greater sage-grouse habitat. Thus, we suggest that a meta-analysis of habitat relationships throughout the range of greater sage-grouse could be fruitful for clarifying habitat relationships of the bird and variation among studies and methods (e.g. see Anderson, Burnham, Franklin, Gutiérrez, Forsman, Anthony & Shenk 1999).