Table W-values for 3 planned comparisons when limiting experimental error rates to 5, 10, 15, and 20 percent

Transcription

1 between the two sample years. We did not evaluate observer bias directly because it is evaluated and discussed elsewhere in this volume (see Manuwal, Gilbert and Allwine b, and Carey and others). Potential effects of sampling biases on the results are evaluated in the discussion. Analytical Methods This study was designed to be exploratory, thus no a priori hypotheses were identified. In light of this objective, the statistical tests were used to elucidate trends and consistent patterns of variation in the data rather than for testing specific hypotheses. We report exact P-values to evaluate the strength of the patterns, except where we needed to define a level of significance to proceed with further analyses. The exceptions were using a P-value of <O.05 to indicate a significant main effect or an interaction between main effects of two-way analysis of variance before we proceeded with appropriate planned comparisons, and using a P-value of <0.151 with the Kruskal-Wallis procedure to identify which bird species showed strong differences in abundance among forest age-classes in at least two of the three provinces. Bird abundance was calculated as the number of birds detected per survey per stand for each of the two sample years, where one survey day equals 12 counting stations x 8 minutes. We calculated species richness as the number of unique species detected in a stand over all surveys during one year. Only detections estimated to be < 50 m from the counting station were used. Nocturnal and raptor species were poorly sampled by the counting technique and were not included in the analysis. To remove vagrants and species detected in few stands, we used two criteria: a species was included in the count for a particular stand if it was detected during at least two surveys in that stand, and, to be included, a species had to be detected in >8 percent of the stands within a subprovince. Thus, rare species (that is, species detected in low numbers) were included in the analysis if detection criteria were met at the stand and subprovince level. This procedure removed species that commonly occurred in habitats substantially different from ones dominated by Douglas-fir (for example, red-winged blackbird, northern oriole, belted kingfisher, mourning dove). We did not distinguish between hermit and Townsend s warblers (hereafter, hermit warbler) or between dusky and Hammond s flycatchers (hereafter, Empidonax flycatchers) during our analyses because these species were difficult to distinguish in the field. Because our primary objective was to evaluate regional patterns of species composition and abundance, and not yearly variation, we used the mean of the two sample years in our analysis. Thus, mean bird species richness and mean abundance values represent the average of two sample years. Mean abundance values were calculated for individual bird Table W-values for 3 planned comparisons when limiting experimental error rates to 5, 10, 15, and 20 percent Experimental error P-value for 3 rate (percent) planned comparisons a a' = l- (1-a)l/K, where a = significance level of each individual comparison for an experimental error rate of a, and K = the number of comparisons (Sokal and Rohlf 1981). species, all species combined, all resident species, all migrant species, and for each of four spatial-foraging guilds: bark, aerial, canopy, and understory foragers (see appendix table 8 for list of species associated with each group). Province and age comparisons-to evaluate whether bird abundance or number of species differed among forest ageclasses, species richness values and mean abundance values for all species combined, residents, migrants, and the four foraging guilds were analyzed by a two-way analysis of variance (ANOVA) for unbalanced designs (Wilkinson 1988). For all ANOVA s, stand age (that is, young, mature, oldgrowth) and province (that is, southern Washington Cascades, Oregon Cascades, Oregon Coast Ranges) were used as the independent factors (the main effects). Normality and homogeneity of variance of richness and abundance values were improved by performing a square-root transformation before the ANOVA. Multiple comparisons were planned; thus, we used F-tests to determine differences among age-classes and among provinces (Sokal and Rohlf 1981). Because the planned comparisons were nonorthogonal (that is, lacked independence), we used the Bonferroni procedure (Klockars and Sax 1986, Wilkinson 1988) to reduce the probability of interpreting a difference as related to ages and provinces when in fact the patterns were a result of random variation. Given that the interaction between the main effects was not significant, we tested three planned comparisons per main effect (Sokal and Rohlf 1981). Instead of setting a specific experimental error at 5 percent, for example, we reported the exact P-value for each comparison, and provide a list of what the acceptable P-values are for experimental errors of 5, 10, and 15 percent (table 3). Abundance patterns of individual bird species-species were categorized into four groups based on general knowledge of territory size, percentage of all detections that were < 50 m versus >50 m from the counting station, and consistency of detection within a stand (see appendix table 8). 181

2 Group 1 species were detected more often 150 m of a counting station and, typically, were the ones detected most consistently and often on the surveys. To determine their abundance, we used all detections 150 m from a counting station, the same criteria used for calculating abundance throughout this study. Species included in groups 2 and 3 were detected more often >50 m from a counting station and detected less often than group 1 species. Group 3 species were more difficult to detect than group 2 species. Group 3 species have large territories and were detected less consistently within a stand. If single occurrences accounted for >30 percent of the stands in which a species was detected, then that species was put in group 3. Detections < 150 m from a counting station were used for calculating abundance of both group 2 and 3 species. Group 2 species had to be detected > 2 times within a stand but group 3 species only once. If a species was detected in <3 percent of the stand counts, it was placed into group 4 and no abundance was calculated. We gauged our interpretation of the results according to the reliability of the count data. We assumed group 1 was the most reliable abundance data and group 3 the least. Within each province, mean abundance of individual species was compared between stand age-classes by the Kruskal- Wallis procedure. Experimental error rates were evaluated by comparing the number of significant comparisons to the total number of comparisons made. Of all the species evaluated, 16 had P-values that were <0.151 in at least two of the three provinces. Because of their strong trends among forest ageclasses, these species were discussed in further detail. Ordination of bird data-we used detrended correspondence analysis (DECORANA or DCA), an ordination procedure, to describe differences in bird communities of the sampled stands. DECORANA uses species composition and abundance data to determine a sample s position in relation to all other samples (Gauch 1982, Hill 1979a). Thus, samples (that is, stands) that are similar are close together and dissimilar samples are far apart To simplify interpretation, we removed the effects of abundance differences between stands and provinces by relativizing the data, so that species abundances totaled 100 in each stand (Mohler 1987). We plotted the position of each stand using DECORANA scores for the first and second axes, where the scores represented the position of a stand along some identifiable environmental gradients (that is, each axis represents an environmental gradient). Median scores for age-class and province categories were calculated, and the position of each age-class and province category was plotted such that the boundaries included at least 90 percent of the stands. Spearman rank (r s ) correlations and scatterplots were used to evaluate relations between all DECORANA scores for the first and second axes, and environmental (elevation, latitude, stand age) as well as biological (species richness, bird abundances) variables. Relation to habitat-of the vegetation and site characteristic variables measured by Spies, we selected 143 as potentially important to our analysis. Variables selected included 58 live-tree, 13 stand-condition, 32 snag, 25 log, and 15 understory-plant variables (see appendix table 9). Each variable was tested for normality by the Kolmogorov-Smimov test (Zar 1984), and, if necessary and useful, transformations were applied. No one transformation worked well on all variables; thus, we used three: log (X + l), square root (X ), and arcsine (square root X) on some of the percentage variables (see appendix table 9). We used all possible subsets multiple regression (Dixon 1985) to evaluate which vegetation variables explained the most variation in species richness, total bird abundance, abundance of resident and migrant species, and abundance of each of the four foraging guilds. The analysis was approached from two levels. Multiple regressions were performed using data from all three provinces together, and then on each province individually. Pearson correlations of vegetation and bird abundance variables were done as a first step before regression analysis. This analysis determined which variables were most associated with bird abundance; variable selection was based on strength of the correlations with dependent variables; and the correlations were generally low so that any variable with an absolute value >0.200 was considered; low intercorrelations with other vegetation variables; and ease of measurement and applicability to management decisions. Vegetation variables that were intercorrelated but were highly correlated with the dependent variable were not removed at this stage because of their potential importance to the total equation. No intercorrelated variables were selected in the final regression equations. The best multiple regression equation was selected from all possible subsets by evaluating which group of variables explained the most variance (adjusted R ), low inter-correlation among vegetation variables, and ease of interpretation. The number of independent variables in an equation (see Johnson 1981) was limited to no more than four for the withinprovince regressions, and no more than six for the all-stands regressions. Results Over 115,000 birds were detected during more than 16,000 station counts in the three physiographic provinces. Ninetythree bird species were detected. We detected 41 species that regularly used Douglas-fir forests >40 years old, after the 1 This term refers to the analysis criteria established for this study: bird species within a province detected < 50 m from a counting station on > 2 sample days per stand in >3 stands (>8 percent). 182

3 q = Oregon cbast Range 0 = Oregon Cascade flange A = Southern Washington Cascade Range 1 = Oregon coast Range 0 = Oregon Cascade Range * = Southern Washington Cascade Range 8 z Latitude exclusion of raptorial, nocturnal, and poorly detected species. Of these 41 species, 56 percent regularly used all three pmvinces, and 73 percent regularly used two provinces. Six speties regularly used only the Oregon Coast Range: the blackcapped chickadee, wrentit, purple fiich, rufous-sided towhee, song sparrow, and orange-crowned warbler. The Townsend s solitaire regularly used only the Oregon Cascades. The red crossbill and Vaux s swift regularly used only the southern Washington Cascades. Elevation, Latitude, and Longitude Stand elevation varied with latitade and longitude within and between provinces (fig. 1). Study stands low and high in elevation tended to be in the Oregon Coast Ranges and Oregon Cascades, respectively. Total bird abundance was correlated negatively with elevation and latitude and correlated positively with longitude (fig. 2).

4 Table 4-The percentage of young, mature, and old-growth stands in which 16 species were detected that showed abundance differences in at least 2 of the 3 provinces omg tco 80 lco ca _1 ll&gl-ow w Table S-Number and percentage (in parentheses) of bird species with significantly different abundances among age-claws at P-values of 0.05, 0.10, 0.15, and (21%) 5 (15%) 18 (51%) (27%) 7 (21%) 20 (57%) (42%) 9 (26%) 26 (74%) (48%) 11 (32%) 26 (74%) Total number of species Abundance Patterns of Individual Bird Species Chestnut-backed chickadees and hermit warblers, detected in all 132 study stands, and winter wrens, western flycatchers, and golden-crowned kinglets, detected in >95 percent of the study stands, were the most abundant species over all three provinces (table 4). Bird abundance differences among stand age-classes were detected for 29 species at P Twice as many species had differences in abundance at P < among stand age-classes in the Oregon Coast Ranges than in the southern Washington Cascades or Oregon Cascades (table 5). Sixteen species showed abundance differences in at least two of the three provinces at P < (figs. 3-5, table 4). yet despite the large sample size, pauems were not similar for any species in all three provinces. Abundance of all sixteen species, except Vaux s swift, showed age-class differences in the Oregon Coast Ranges. Six hole nesters-the chestnutbacked chickadee, red-breasted nuthatch, hairy woodpecker, pileated woodpecker, red-breasted sapsucker, and Vaux s swift-were most abundant in old-growth stands in at least two provinces (figs. 3-5). Brown creeper abundance was highest in old-growth and mature stands (fig. 3). but western flycatcher (fig. 3) and hermit thrush (fig. 4) were most abundant in old-growth and young stands. Black-throated gray and hermit warblers tended to be most abundant in young stands (fig. 3). Evening grosbeak and American robin abundance was highest in young and mature. stands (fig. 4). No clear pattern was observed for gray jay, Steller s jay, and northern flicker (figs. 3-5). Abundance of only the red-breasted sapsucker differed in all three provinces at P < (fig. 5). 184

5 Black-throated gray warbler I.:...-..: IL- I 1 Chestnut-backed chickadee 185

6 .;-; Evening grosbeak r Steller s jay Hairy woodpecker

7 Northern flicker 7 Province and Age Comparisons Interaction and main effectsno province x age interaction effects were detected in the eight two-way ANOVA tests at P < except for the comparison of migrant species (P = 0.060) (figs. 6-13). The main effect of province was different in all ANOVA tests at P < Only half the tests on stand age-namely, total abundance and abundance of resident species, bark foragers. and aerial foragers-were different at P < (figs. 6-13). All other stand age cornpafisons were different at P > Pairwise comparisons of provinc+all multiple painvise comparisons of mean abundance and species richness were higher in the Oregon Coast Ranges than in the Oregon and southern Washington Cascades at P < (figs. 6.13).- Bird abundance was generally higher in the Oregon than southern Washington Cascades, except for resident species and understory foragers, which were more abundant in the southern Washington Cascades, and bark foragers. whose abundance did not differ between the two provinces (figs. 6-13). Pairwise comparisons of stand agestand-age pairwise comparisons showed that total abundance and abundance of resident species and bark foragers were highest in old-growth stands (figs. 7,8, 10). Abundance of aerial foragers was highest in young and old-growth stands (fig. 11). The only substantial and consistent increase with stand age-class was the abundance of bark foragers (fig. 10). In six of the eight ANOVA comparisons, mean abundance and richness were lowest in mature stands (figs. 6-13). No comparisons were lowest in old growth.

8 B 188

9 r t 189

10 B i,,i tz CR oc SW c