RELATIONSHIP BETWEEN BIRD ABUNDANCES AND ... - Springer Link

Environmental Monitoring and Assessment (2005) 105: 209–228 DOI: 10.1007/s10661-005-3694-x

c Springer 2005

RELATIONSHIP BETWEEN BIRD ABUNDANCES AND LANDSCAPE CHARACTERISTICS: THE INFLUENCE OF SCALE SARAH P. BRENNAN and GARY D. SCHNELL∗ Department of Zoology and Sam Noble Oklahoma Museum of Natural History, University of Oklahoma, Norman, Oklahoma, USA (∗ author for correspondence, e-mail: [email protected])

(Received 26 May 2004; accepted 14 July 2004)

Abstract. Scale is important to consider when investigating effects of the environment on a species. Breeding Bird Survey (BBS) data and landscape metrics derived from aerial photographs were evaluated to determine how relationships of bird abundances with landscape variables changed over a continuous range of 16 spatial scales. We analyzed the average number of birds per stop (1985–1994) for five songbird species (family Cardinalidae) for each of 50 stops on 198 BBS transects throughout six states in the Central Plains, USA. Land along each transect was categorized into six cover types, and landscape metrics of fractal dimension (a measure of shape complexity of habitat patches), edge density, patch density, and percent area were calculated, with principal components used to construct composite environmental variables. Associations of bird abundances and landscape variables changed in accordance with small scale changes. Abundances of three species were correlated with edge density and one with component I, which subsumes initial variables of patch density for urban, closed forest, open forest, and open country. Fractal dimension and component II (summarizing amount of closed forest versus open country) were associated with the most species. Correlation patterns of fractal dimension with northern cardinal (Cardinalis cardinalis) and painted bunting (Passerina ciris) abundances were similar, with highest correlations at intermediate to small scales, suggesting indirectly that these species thrive in areas where local habitat conditions are most important. Multiscale analysis can provide insight into the spatial scale(s) at which species respond, a topic of intrinsic scientific interest with applied implications for researchers establishing protocols to assess and monitor avian populations. Keywords: blue grosbeak, Cardinalidae, dickcissel, fractal dimension, indigo bunting, landscape, multiple scales, northern cardinal, painted bunting, scale continuum

1. Introduction The scale at which the environment is studied in relation to ecological processes has an influence on our perception of patterns and of the distributions of species (Naveh and Leiberman, 1984; Wiens et al., 1987; Flather and Sauer, 1996; Bolger et al., 1997). In addition, species may respond to the surrounding environment at different scales (Böhning-Gaese, 1997; MacFaden and Capen, 2002). Understanding the effects of scale can be particularly important for those who set policies and manage lands for conservation purposes (Saab, 1999; Meyer et al., 2002). Factors at a local scale such as vegetational structure often are most important for birds (Wiens et al., 1987; Pribil and Picman, 1997), while at a broad scale

210

S. P. BRENNAN AND G. D. SCHNELL

measures like heterogeneity, amount of edge, climate, and community interactions influence bird communities (McGarigal and McComb, 1995; O’Connor et al., 1996). Studying relationships at multiple levels allows one to evaluate the effect of scale-dependent patterns on avian species (Wiens, 1989; Kotliar and Wiens, 1990). Numerous studies have looked at how the relationship between avian populations and habitat changes with scale (e.g. Böhning-Gaese, 1997; Pribil and Picman, 1997; Drolet et al., 1999; Saab, 1999; MacFaden and Capen, 2002; Tarvin and Garvin, 2002; Westphal et al., 2003) but few have incorporated more than three scale levels (Wiens et al., 1986; Hecnar and M’Closkey, 1997; Meyer et al., 2002; Brennan, 2004). In this study, we have investigated the influence of scale on the relationship between avian abundance and landscape characteristics based on abundance data for five bird species of the family Cardinalidae from Breeding Bird Surveys (BBS) (Robbins et al., 1986) and landscape variables over a continuum of 16 spatial scales from the local to regional level. Our purposes were (1) to evaluate the changes in the relationship between abundance and landscape characteristics at multiple scales; (2) to determine the scale(s) at which certain variables have the strongest relationship to abundance; (3) to observe which landscape variables have the strongest association with bird abundance; and (4) to compare findings of this investigation with those from a similar analysis of flycatchers (Tyrannidae; Brennan, 2004).

2. Methods 2.1. B IRD

DATA

We used bird data collected on BBS, which are annual censuses that provide information on relative abundances of bird species (Robbins et al., 1986). Survey transects are 39.43 km (24.5 mile) long and visited once a year in June. Observers record all birds seen and heard for 50, 3-min point counts conducted at 0.80-km (0.5-mile) intervals along the transect (see Figure 1). Our analyses incorporated data collected from 1985 to 1994 on 198 transects located throughout Kansas, Oklahoma, Texas, Missouri, Arkansas, and Louisiana (see Figure 2) for the following species: (1) northern cardinal (Cardinalis cardinalis); (2) blue grosbeak (Passerina cerulea); (3) indigo bunting (P. cyanea); (4) painted bunting (P. ciris); and (5) dickcissel (Spiza americana). Data for these species should provide reliable indices of abundances given that these birds are readily observable if present, are not secretive, and are easily identified. The 198 transects used were surveyed a minimum of five times in the 10year period of 1985–1994. The average number of birds per stop per year for each transect provided an abundance index for each species at each of the 50 stops for each of the 198 transects. Since two species (indigo and painted buntings) occurred on less than 80% of the transects, we used a kriging method

BIRD ABUNDANCES, LANDSCAPE VARIABLES, AND SCALE

211

Figure 1. Example of land cover types adjacent to a typical Breeding Bird Survey transect. Open circles indicate the 50 stops where bird counts were made.

(van Horssen et al., 1999) to determine the approximate edge of the species’ range and, thus, which transects to include in the analysis. Kriging interpolates data values measured at irregularly spaced sample points to provide estimated values throughout an area. As a result, we used 125 of the 198 transects for the indigo bunting (Figure 2c) and 134 for the painted bunting (Figure 2d). 2.2. L ANDSCAPE

DATA

We obtained digital aerial photographs of the landscape along and adjacent to each of the 198 BBS transects from the National Aerial Photography Program, US Department of Agriculture. These photos were taken from December through March in 1989, 1990, and 1991 near the midpoint of years for which bird censuses were evaluated. For each transect, we classified cover types of land within approximately 1.20 km on either side and extending 0.40 km (0.25 miles) beyond each end of the transect. Thus, the area evaluated for each transect was 40.23 km long and 2.40 km wide. We used a habitat classification defined in Certain (2000) that categorized the landscapes along these 198 transects into six cover type: (1) urban; (2) closed forest (less than one canopy width between trees); (3) open forest (≥1 and |0.75|) in bold.

numerically consistent with average taxonomic distances among transects (Rohlf, 2003). Projections were calculated for each of the 16 spatial scales. For all calculations, we used the F-matrix from the analysis of all stops combined, but the O-matrix was changed, based on the landscape measures for the particular spatial scale being analyzed. We also calculated product–moment correlations of abundances of each bird species with each of the five landscape variables at each of the 16 scales. Transects were partitioned into 1–50 segments, a segment referring to a section that contained a determined number of point counts. For example, when investigating a scale of 20.12 km, which incorporated 25 point counts, each transect was comprised of two segments. For a spatial scale of 17.70 km (22 point counts), a single transect also contained two segments; six point counts remained at the end of the transect and were not used. In all cases for transects ending with less than the designated number of point counts for a segment for a particular scale, we omitted the leftover point counts from the analysis. For each segment of each transect we calculated the average abundance for each species, as well as each landscape metric. We employed resampling (Simon, 1997; Blank et al., 2001) to create an appropriate distribution for statistical evaluation because adjacent segments within a transect were not likely to be statistically independent. Shuffling the bird-abundance data first by entire transects and then by segments within the transect took into


215

account the spatial autocorrelation within transects (Brennan, 2004). First, we shuffled the order of transects, resulting in bird-abundance data for each transect being paired randomly with landscape data for one of the transects. Second, we shuffled abundance data according to segments within each transect, with the outcome that bird-abundance values for segments were paired randomly with segment landscape values. Each shuffle was done without replacement. Then, for the twice-shuffled data, with bird-abundance data by segment randomly associated with landscape data, we calculated the product–moment correlation based on the paired values for each segment for all transects. We repeated this overall procedure 10,000 times to create an appropriate distribution of correlations against which to evaluate the statistical significance of the correlation value obtained from the original data. 3. Results 3.1. L ANDSCAPE

VARIABLES

Based on whole transects (50 stops), we obtained high values for fractal dimension in the eastern and central regions of the study area, with lower values in the west (Figure 3a), indicating that cover types generally had less complex edges in the west. Edge density (Figure 3b) was low in the west, highest in central and northcentral areas, and moderate to low in the east. Total patch density (not figured) tended to be relatively low in the east and west, with higher values in transitional central areas. This landscape variable did not exhibit correlations above 0.15 with abundances of any of the bird species (Table II) and, thus, is not given further attention. Principal component I projections based on the full lengths of transects tended to be higher in the central region and lower values in the western and to some extent in the eastern part of the study area (Figure 4a). The variables having the highest loadings were patch densities of urban, closed forest, open forest, and open country (Table I). This component reflected the fact that transects in the central region tended to have more patches of these four cover types than do transects in the east and particularly the west. Component I projections were similar to those of total patch density. Landscape component II exhibited high values in the east and east-central portion of the study area and low values in the northwest (Figure 4b). It had high positive loadings on percent area of closed forest and a high negative loading on percent area of open country (Table I). More closed forest and less open country were found in the east-central part of the study region. 3.2. TRENDS

OVER MULTIPLE SCALES

Significant relationships between bird abundance and landscape variables over spatial scales can be subsumed under three general trends, which involved fractal

216


Figure 3. Distribution of the landscape variables fractal dimension (dimensionless) and edge density (m/ha) based on 198 Breeding Bird Survey transects. Natural-breaks method used to identify appropriate classes, with classes being represented by differentially shaded symbols.

dimension, edge density, and principal component II. In addition abundances of one species were correlated significantly with principal component I. All correlation values of bird abundances for each species with all landscape variables at each spatial scale are given in Table II.


217

TABLE II Correlations (∗p < 0.05;∗∗ p < 0.01) of landscape variables with average numbers of five cardinalids per stop at each of the 16 spatial scales Segment length

Fractal dimension

1 2 3 4 5 6 7 8 9 10 11 12 17 22 25 50

0.2674∗∗ 0.3491∗∗ 0.3718∗∗ 0.4031∗∗ 0.4027∗∗ 0.4174∗∗ 0.4153∗∗ 0.4143∗∗ 0.4219∗∗ 0.4057∗∗ 0.4159∗∗ 0.4209∗∗ 0.4100∗∗ 0.3995∗∗ 0.3707∗∗ 0.3002∗∗

1 2 3 4 5 6 7 8 9 10 11 12 17 22 25 50

0.1635∗∗ 0.2316∗∗ 0.2660∗∗ 0.2852∗∗ 0.3066∗∗ 0.3179∗∗ 0.3245∗∗ 0.3288∗∗ 0.3538∗∗ 0.3510∗∗ 0.3431∗∗ 0.3494∗∗ 0.3714∗∗ 0.3763∗∗ 0.3651∗∗ 0.3608∗∗

1 2

0.1158∗∗ 0.1690∗∗

Edge density

Total patch density

Northern cardinal 0.0058 −0.0041 ∗∗ 0.2469 −0.0129 0.2666∗∗ 0.0354 0.2686∗∗ 0.0430 0.2715∗∗ 0.0499 ∗∗ 0.2804 0.0501 0.2814∗∗ 0.0564 0.2848∗∗ 0.0605 0.2859∗∗ 0.0577 ∗∗ 0.2822 0.0619 0.2873∗∗ 0.0476 0.2905∗∗ 0.0579 0.3109∗∗ 0.0630 ∗∗ 0.2969 0.0637 0.2901∗∗ 0.0541 0.2972∗∗ 0.0611 Blue grosbeak −0.0077 −0.0237 0.1307∗∗ −0.0062 0.1539∗∗ −0.0161 ∗∗ 0.1595 −0.0166 0.1683∗∗ −0.0187 0.1711∗∗ −0.0177 0.1701∗∗ −0.0178 ∗∗ 0.1737 −0.0160 0.1858∗∗ −0.0170 0.1859∗∗ −0.0184 0.1778∗∗ −0.0190 ∗∗ 0.1793 −0.0211 0.2009∗∗ −0.0170 0.2016∗∗ −0.0114 0.1952∗∗ −0.0201 ∗∗ 0.1919 −0.0298 Indigo bunting −0.0118 −0.0889 0.0581 −0.0869

Component I

Component II

−0.0129 0.0152 0.0177 0.0188 0.0280 0.0259 0.0326 0.0344 0.0354 0.0346 0.0251 0.0269 0.0337 0.0314 0.0212 0.0321

0.3747∗∗ 0.4017∗∗ 0.4316∗∗ 0.4418∗∗ 0.4428∗∗ 0.4568∗∗ 0.4582∗∗ 0.4639∗∗ 0.4720∗∗ 0.4671∗∗ 0.4804∗∗ 0.4824∗∗ 0.5118∗∗ 0.5016∗∗ 0.5035∗∗ 0.5361∗∗

−0.0248 −0.0222 −0.0230 −0.0283 −0.0303 −0.0310 −0.0355 −0.0323 −0.0332 −0.0330 −0.0378 −0.0401 −0.0401 −0.0369 −0.0396 −0.0499

0.2111∗∗ 0.2600∗∗ 0.2911∗∗ 0.3159∗∗ 0.3282∗∗ 0.3382∗∗ 0.3450∗∗ 0.3565∗∗ 0.3743∗∗ 0.3638∗∗ 0.3797∗∗ 0.3744∗∗ 0.4015∗∗ 0.4154∗∗ 0.4068∗∗ 0.4765∗∗

−0.1045∗ −0.1097∗

0.3589∗∗ 0.4034∗∗

(Continued on next page.)

218


TABLE II (Continued ) Segment length

Fractal dimension

3 4 5 6 7 8 9 10 11 12 17 22 25 50

0.1826∗∗ 0.2012∗∗ 0.2232∗∗ 0.2303∗∗ 0.2309∗∗ 0.2448∗∗ 0.2613∗∗ 0.2684∗∗ 0.2508∗∗ 0.2616∗∗ 0.2726∗∗ 0.2638∗∗ 0.2686∗∗ 0.2907∗∗

1 2 3 4 5 6 7 8 9 10 11 12 17 22 25 50

0.1223∗∗ 0.2045∗∗ 0.1888∗∗ 0.2077∗∗ 0.1988∗∗ 0.1996∗∗ 0.1919∗∗ 0.1962∗∗ 0.1991∗∗ 0.1909∗∗ 0.2049∗∗ 0.1903∗ 0.1769∗ 0.1434 0.1473∗ 0.1179

1 2 3 4 5

−0.0521 −0.0797 −0.0901∗ −0.0884 −0.0672

Edge density

Total patch density

−0.0896 −0.0936 −0.0865 −0.0935 −0.0914 −0.0921 −0.0850 −0.0923 −0.0950 −0.1006 −0.1099 −0.1065 −0.1125 −0.1361 Painted bunting 0.0186 0.0960 0.2380∗∗ −0.0114 0.2117∗∗ 0.1354 0.2246∗∗ 0.1223 0.2260∗∗ 0.1198 0.2232∗∗ 0.1195 0.2331∗∗ 0.1279 ∗∗ 0.2323 0.1315 0.2269∗∗ 0.1307 0.2323∗∗ 0.1268 0.2389∗∗ 0.1170 ∗∗ 0.2336 0.1277 0.2424∗∗ 0.1125 0.2442∗∗ 0.1204 0.2429∗∗ 0.1241 ∗∗ 0.3190 0.1130 Dickcissel 0.0065 0.1105 −0.0139 −0.0095 −0.0155 0.1124 −0.0123 0.1101 0.0006 0.1090 0.0527 0.0520 0.0547 0.0522 0.0542 0.0509 0.0671 0.0527 0.0492 0.0517 0.0418 0.0679 0.0466 0.0474

Component I

Component II

−0.1114 −0.1128 −0.1118 −0.1205 −0.1216 −0.1196 −0.1192 −0.1236 −0.1250 −0.1348 −0.1344 −0.1341 −0.1516 −0.1571

0.4191∗∗ 0.4310∗∗ 0.4337∗∗ 0.4442∗∗ 0.4476∗∗ 0.4574∗∗ 0.4576∗∗ 0.4552∗∗ 0.4552∗∗ 0.4678∗∗ 0.4984∗∗ 0.4697∗∗ 0.4728∗∗ 0.4616∗∗

0.0566 −0.0183 0.2117 0.0784 0.0701 0.0738 0.0725 0.0816 0.0837 0.0754 0.0727 0.0742 0.0666 0.0586 0.0663 0.1099

0.1651∗∗ 0.1713∗∗ 0.1945∗∗ 0.1919∗∗ 0.1975∗∗ 0.2042∗∗ 0.2076∗∗ 0.2047∗∗ 0.2140∗∗ 0.2014∗∗ 0.2086∗∗ 0.2118∗∗ 0.2121∗∗ 0.2040∗∗ 0.2034∗∗ 0.1691∗∗

0.0971∗ 0.0675 0.1086∗ 0.1120∗ 0.1106

−0.3244∗∗ −0.3574∗∗ −0.3890∗∗ −0.3981∗∗ −0.3980∗∗

(Continued on next page)


219

TABLE II (Continued) Segment length

Fractal dimension

6 7 8 9 10 11 12 17 22 25 50

−0.0768 −0.0631 −0.0530 −0.0435 −0.0353 −0.0399 −0.0323 0.0121 0.0071 0.0650 0.1293

Edge density 0.0039 0.0060 0.0127 0.0176 0.0189 0.0238 0.0215 0.0539 0.0440 0.0614 0.0785

Total patch density

Component I

0.1218 0.1176 0.1185 0.0896 0.1123 0.1323 0.1306 0.1479 0.1340 0.1380 0.1372

0.1244∗ 0.1189∗ 0.1258∗ 0.1280∗ 0.1238∗ 0.1437∗ 0.1502∗ 0.1659∗ 0.1308∗ 0.1612∗ 0.1668∗

Component II −0.4110∗∗ −0.4066∗∗ −0.4223∗∗ −0.4372∗∗ −0.4150∗∗ −0.4448∗∗ −0.4355∗∗ −0.4645∗∗ −0.4552∗∗ −0.4291∗∗ −0.4668∗∗

Northern cardinals and painted buntings had a similar pattern of correlations with fractal dimension, as did blue grosbeaks and indigo buntings (Figure 5). Correlations for both northern cardinals and painted buntings increased to 3.22 km (segment length 4), after which values remained fairly constant until approximately 9.66 km (segment length 12), where correlations decreased slightly. Blue grosbeaks and indigo buntings had similar patterns in that correlations increased to 13.68 km (segment length 17) and then were fairly level. These four species tended to have higher abundances in areas with irregularly shaped patches, which generally are found in the eastern and central regions of the study area when considering entire transects (Figure 3a). Northern cardinal correlations were higher than those for the other species. The correlation patterns of edge density with numbers of northern cardinals, blue grosbeaks, and painted buntings were very similar (Figure 6). Correlations for all three species increased until 4.83 km (segment length 6), after which values tended to level off. For the painted bunting the correlation increased between 20.12 and 40.23 km (segment lengths 25 and 50). These three species tended to be more abundant in areas with greater amounts of edge, which at the 50-segment-length level (i.e. entire transect) generally were found in the central and north-central region, and to a lesser extent in the eastern region of the study area (Figure 3b). The correlations of principal component I and abundances of dickcissels were relatively weak but in general increased from 0.80 to 13.68 km (segment lengths 1– 17), after which values remained fairly constant (Figure 7). Dickcissels were more abundant in the central region of the study area where there was more patches of urban, closed forest, open forest, and open country. Associations were statistically significant at 14 of 16 spatial scales analyzed.

220


Figure 4. Projections of 198 Breeding Bird Survey transects onto first two principal components based on 10 landscape variables measured for total length of transect (i.e. segment length 50). Natural-breaks method used to identify appropriate classes, with classes being represented by differentially shaded symbols.


221

Figure 5. Correlations over range of segment lengths of fractal dimension with average number of birds per stop for northern cardinals, blue grosbeaks, indigo buntings, and painted buntings. Open symbols indicate nonsignificant values ( p > 0.05), gray symbols significant values ( p < 0.05), and black symbols highly significant values ( p < 0.01).

Figure 6. Correlations over range of segment lengths of edge density with average number of birds per stop for northern cardinals, blue grosbeaks, and painted buntings. Open symbols indicate nonsignificant values ( p > 0.05) and black symbols highly significant values ( p < 0.01).

222


Figure 7. Correlations over range of segment lengths of landscape principal component I and average number of dickcissels per stop. Open symbols indicate nonsignificant values ( p > 0.05) and gray symbols significant values ( p < 0.05).

Correlation patterns of component II projections with abundances of northern cardinals and indigo buntings were very similar through 13.68 km (segment length 17; Figure 8). For the northern cardinal, the correlations leveled off and then rose slightly at the largest scale (40.23 km, segment length 50), while for the blue grosbeak they dropped slightly and then remained relatively constant. The correlations increased until 13.68 km (segment length 17), after which the curve remained fairly level. The correlations of blue grosbeak abundance and principal component II showed a general increase from 0.80 to 40.23 km (segment lengths 1–50). Painted buntings also had a positive association with component II but correlations were low and relatively similar at 4.83 km (segment length 6) and beyond. For these species, abundances tended to be higher in areas where there was more closed forest relative to open country. The dickcissel had negative instead of positive correlations with component II (Figure 8). The negative correlations gradually increased until 9.66 km (segment length 12) and were most pronounced at 40.23 km (segment length 50). Dickcissels tended to be more abundant in northwestern region of the study area, which had more open country and less closed forest. 3.3. SCALE

AND CLOSEST ASSOCIATIONS

For the 13 sets of statistically significant correlations of bird abundances with landscape measures, seven reached a maximum correlation at a low to intermediate scale (ca. 3.22–17.70 km, segment lengths 4–22) and then remained relatively constant. Four patterns involved increases in correlations from the smallest scale to the largest. In contrast, two graphs – those for northern cardinals and painted buntings with fractal dimension (Figure 5) – were notable in that the highest correlations were obtained for intermediate scales, with a marked drop in correlations with both smaller and larger scales.


223

Figure 8. Correlations over range of segment lengths of values of landscape principal component II and average number of birds per stop for northern cardinals, blue grosbeaks, indigo buntings, painted buntings, and dickcissels. Open symbols indicate nonsignificant values ( p > 0.05) and black symbols highly significant values ( p < 0.01).

Edge density was significantly correlated with northern cardinals, blue grosbeaks, and painted buntings, with the highest correlation occurring at intermediate and larger scales. Correlations for edge density and abundance of northern cardinal were highest at 13.68 km (segment length 17), blue grosbeak at 17.70 km (segment length 22), and painted bunting at 40.23 km (segment length 50; Figure 6). Component I was significantly correlated to dickcissels at eight spatial scales with the highest correlation at 40.23 km (segment length 50; Figure 7). The greatest correlations with component II also occurred at 40.23 km (segment length 50)

224


for northern cardinals, blue grosbeaks, and dickcissels, while the highest correlation occurred at an intermediate scale for indigo and painted buntings (13.68 km, segment length 17; Figure 8). 3.4. I NFLUENTIAL

LANDSCAPE VARIABLES

Of the five landscape variables, we investigated (fractal dimension, edge density, total patch density, component I, and component II) only two – fractal dimension and principal component II – were significantly associated with abundances of at least four species at a majority of the spatial scales. Abundances of all except the painted bunting significantly correlated with fractal dimension at all spatial scales (Figure 5) and abundances of four of the five cardinalids were strongly associated with component II (Figure 8).

4. Discussion 4.1. TRENDS

OVER MULTIPLE SCALES

Throughout the ranges of scales studied, associations of bird abundances and landscape variables changed only gradually with small changes in scale. Our findings for the five species and five landscape variables showed no abrupt changes in the strength of associations with small changes in scale, and this held over the entire range of scales. While studies investigating two or three scales (e.g. Saab, 1999; MacFaden and Capen, 2002) have shown general trends in relationships between species and landscape variables, the gradual nature of changes in the strength of associations, such as those found in our study might have not have been appreciated had only two or three spatial scales been analyzed. 4.2. I NFLUENTIAL

LANDSCAPE VARIABLES

The four species having positive correlations with fractal dimension – northern cardinal, blue grosbeak, indigo bunting, and painted bunting – appeared to prefer areas with some woody vegetation that had irregular edges and thus comparatively high fractal dimensions. The northern cardinal typically is found in areas with shrubs and or small trees, including forest edges and forest interior (Dow, 1969; Halkin and Linville, 1999); the blue grosbeak is characteristic of forest edges, hedgerows, stream edges, and multistage pine forests (Ingold, 1993); the indigo bunting often inhabits brushy and weedy areas along edges, including riparian habitats and clearings in open deciduous forests (Payne, 1992); and the painted bunting usually is found in partially open areas containing scattered brush and trees, riparian thickets and weedy and shrubby areas (Lowther et al., 1999). In contrast, areas mostly comprised of agriculture fields typically have low shape


225

complexity, and these four cardinalids usually do not occur there or they are in very low densities. As noted in Section 3, two species, the northern cardinal and painted bunting, showed a pattern of correlations with fractal dimension where the highest values occurred at intermediate scales. For the northern cardinal, correlations increased readily to a scale of 3.22 km (segment length 4) and then remained at a relatively high level until 17.70 km (segment length 22), after which they declined. Likewise with the painted bunting the correlations peaked at 3.22 km (segment length 4); they remained relatively constant until 9.66 km (segment length 12), after which they declined. Thus, these two species typically were at their highest densities in areas with irregularly shaped patches and at their lowest in regularly shaped plots based on measurements made at relatively small and intermediate scales. It is clear that these species thrive in areas where relatively local habitat conditions are most influential. As quantified by Kopachena and Crist (2000), the largest wooded areas in northeastern Texas supporting painted buntings often were irregular clumps of trees or long, narrow strips of land like those found along intermediate streams; they noted that the patches of woodland associated with this species were of uneven age and had a high ratio of edge to area. According to Parmalee (1959), the painted bunting in Oklahoma was most common in open areas dissected by small stands or strips of land. The situations described by these authors would result in relatively high fractal dimension when measured at intermediate to small scales in our study. Three species, the northern cardinal, blue grosbeak, and painted bunting, were positively correlated with edge density, a result not surprising given that they are known to commonly inhabit areas with edge (Ingold, 1993; Halkin and Linville, 1999; Lowther et al., 1999). Although indigo and painted buntings share considerable overlap in their ranges throughout our study region, along with similar habitat descriptors (Payne, 1992; Lowther et al., 1999), indigo buntings did not show a significant relationship to edge density. One reason may be because the song perches occupied by indigo buntings are more often found near linear edges of woodlands, while painted buntings usually are associated with areas of irregularly clumped trees or trees along intermittent streams (Kopachena and Crist, 2000). Both intermittent streams and irregularly shaped areas tend to have a relatively high amount of edge. Component II, representing the percent of closed forest versus open country, was significantly correlated with abundances of all five species (Figure 8). These associations highlight important habitat requirements of these birds. The northern cardinal had a relatively strong positive association with principal component II and tends to be found at the edge of forests, as well as the interior (Halkin and Linville, 1999). Birds such as the blue grosbeak, indigo bunting, and painted bunting, which frequent partly open and shrubby habitats (Payne, 1992; Ingold, 1993; Lowther et al., 1999), also had a positive relationship, albeit somewhat weaker in the case of the painted bunting abundance. Low correlations with painted buntings may be a result of the species occurring in wooded areas in an otherwise mostly open region

226


(Kopachena and Crist, 2000). Areas with these characteristics to some extent are subsumed within the limited number of broad cover types we used to define the landscape, resulting in low correlations for this species. The fifth species – the dickcissel – showed a negative association to landscape principal component II. This bird is an obligate grassland species (Vickery et al., 1999) and commonly found in a variety of open grassland habitats including hayfields, lightly grazed pastures, restored grasslands, and fallow areas of agricultural landscapes (Temple, 2002). In fact, a study by Zimmerman (1993) found that males occupied fallow fields and unmowed hayfields earlier in the spring then when they claimed territories in prairie grassland areas, reinforcing the notion that agriculture fields provide highly suitable habitats for breeding dickcissels. Overall, principal component II subsumed some of the most important habitat factors influencing the range limits, distributions, and abundances of the majority of the birds examined. 4.3. C OMPARISON

BETWEEN TAXONOMIC GROUPS

Brennan (2004) conducted similar analyses for eight species of tyrannid flycatchers for the same study area; there were notable similarities in results. In both investigations, fractal dimension and principal component II were the landscape variables most closely correlated with abundance data for the species evaluated, indicating that these factors encapsulated aspects of the landscape important to species in both groups. The landscape variable edge density was significantly correlated to three species of cardinalids as well as three species of tyrannids. This suggests that amount of edge was an important habitat characteristic for certain species in both groups of breeding birds. Both studies indicate that a multiscale approach allows investigators to determine how changes in spatial scale affect the relationship between bird abundances and landscape variables. Investigations such as that by Fuhlendorf et al. (2002) suggest that multiscale studies can be important when attempting to best manage and monitor populations and habitats. Our current study indicates that spatial aspects of habitat configurations have a notable influence on bird abundances. While abundances of some species are similarly associated with particular landscape characteristics there are notable interspecific differences, even among relatively closely related species. Documenting such similarities and differences can provide new insight concerning a complex of factors that affect avian abundances across species’ ranges.

Acknowledgements Support for the first author was provided through a George Miksch Sutton Scholarship in Ornithology, and a M. Blanche Adams and M. Frances Adams


227

Memorial Scholarship, as well as by the University of Oklahoma Graduate Student Senate. We thank Keith L. Pardieck for providing the BBS raw data, David L. Certain for delimited orthophotographs, Todd D. Fagin and May Yuan for ArcView and ArcInfo guidance and advice, Daniel J. Hough for computer assistance, Kanwaljit Aulakh and Simone D. Norman for entering bird data, Joseph P. Roberts for helping to proof bird data, and Victoria L. Adams for assistance with data organization. May Yuan, William J. Matthews, and Peter D. Vickery provided helpful comments on the manuscript.

References Blank, S., Seiter, C. and Bruce, P.: 2001, Resampling Stats in Excel, Resampling Stats Inc., Arlington, VA. Böhning-Gaese, K.: 1997, ‘Determinants of avian species richness at different scales’, J. Biogeogr. 24, 49–60. Bolger, D. T., Scott, T. A. and Rotenberry, J. T.: 1997, ‘Breeding bird abundance in an urbanizing landscape in coastal southern California’, Conserv. Biol. 11, 406–421. Brennan, S. P.: 2004, ‘Relationship of bird abundances and landscape characteristics at multiple scales’, M.S. Thesis, University of Oklahoma, Norman OK. Certain, D. L.: 2000, ‘A multiscale analysis of nest-site choice, nest success, and population abundance in the scissor-tailed flycatcher’, Ph.D. Dissertation, University of Oklahoma, Norman, OK. Dow, D. D.: 1969, ‘Habitat utilization by cardinals in central and peripheral breeding populations’, Can. J. Zool. 47, 409–417. Drolet, B., Desrochers, A. and Frotin, M.: 1999, ‘Effects of landscape structure on nesting songbird distribution in a harvested boreal forest’, Condor 101, 699–704. Flather, C. H. and Sauer, J. R.: 1996, ‘Using landscape ecology to test hypothesis about large-scale abundance patterns in migratory birds’, Ecology 71, 28–35. Fuhlendorf, S. D., Woodward, A. J. W., Leslie, D. M. Jr. and Shackford, J. S.: 2002, ‘Multi-scale effects of habitat loss and fragmentation on lesser prairie-chicken populations of the US Southern Great Plains’, Landscape Ecol. 17, 617–628. Halkin, S. L. and Linville, S. U.: 1999, ‘Northern Cardinal (Cardinalis cardinalis)’, in: A. Poole and F. Gill (eds), The Birds of North America, no. 440, Academy of Natural Sciences, Philadelphia, PA, and American Ornithologists’ Union, Washington, DC. Hecnar, S. J. and M’Closkey, R. T.: 1997, ‘Spatial scale and determination of species status of the green frog’, Conserv. Biol. 11, 670–682. Ingold, J. L.: 1993, ‘Blue Grosbeak (Passerina caerulea)’, in: A. Poole and F. Gill (eds), The Birds of North America, no. 79, Academy of Natural Sciences, Philadelphia, PA. Kopachena, J. G. and Crist, C. J.: 2000, ‘Macro-habitat features associated with painted and indigo buntings in northeast Texas’, Wilson Bull. 112, 108–114. Kotliar, N. B. and Wiens, J. A.: 1990, ‘Multiple scales of patchiness and patch structure: A hierarchial framework for the study of heterogeneity’, Oikos 59, 253–260. Lowther, P. E., Lanyon, S. M. and Thomspon, C. W.: 1999, ‘Painted Bunting (Passerina ciris)’, in: A. Poole and F. Gill (eds), The Birds of North America, no. 398, Academy of Natural Sciences, Philadelphia, PA, and American Ornithologists’ Union, Washington, DC. MacFaden, S. W. and Capen, D. E.: 2002, ‘Avian habitat relationships at multiple scales in a New England forest’, For. Sci. 48, 243–253.

228


McGarigal, K. and Marks, B. J.: 1995, ‘FRAGSTATS: Spatial Pattern Analysis Program for Quantifying Landscape Structure’, General Technical Report PNW-GTR-35, Pacific Northwest Research Station, Forest Service, US Department of Agriculture, Portland, OR. McGarigal, K. and McComb, W. C.: 1995, ‘Relationships between landscape structure and breeding birds in the Oregon coast range’, Ecol. Monogr. 65, 235–260. Meyer, C. B., Miller, S. L. and Ralph, C. J.: 2002, ‘Multi-scale landscape and seascape patterns associated with marbled murrelet nesting areas on the U.S. west coast’, Landscape Ecol. 17, 95–115. Miller, J. N., Books, R. P. and Croonquist, M. J.: 1997, ‘Effects of landscape patterns on biotic communities’, Landscape Ecol. 12, 137–153. Naveh, Z. and Lieberman, A. S.: 1984, Landscape Ecology: Theory and Application, Springer Verlag, New York, NY. O’Connor, R. J., Jones, M. T., White, D., Hunsaker, C., Loveland, T., Jones, B. and Preston, E.: 1996, ‘Spatial partitioning of environmental correlates of avian biodiversity in the conterminous United States’, Biodivers. Lett. 3, 97–110. Parmelee, D. F.: 1959, ‘The breeding behavior of the painted bunting in southern Oklahoma’, BirdBanding 30, 1–19. Payne, R. B.: 1992, ‘Indigo bunting (Passerina cyanea)’, in: A. Poole and F. Gill (eds), The Birds of North America, No. 4, Academy of Natural Sciences, Philadelphia, PA. Pribil, S. and Picman, J.: 1997, ‘The importance of using proper methodology and spatial scale in the study of habitat selection by birds’, Can. J. Zool. 75, 1835–1844. Robbins, C. S., Bystrak, D. and Geissler, P. H.: 1986, ‘The breeding bird survey: Its first fifteen years 1965–1979’, USFWS Resource Publication no. 157, Washington, DC. Rohlf, F. J.: 2003, NTSYSpc: Numerical Taxonomy and Multivariate Analysis System, Exeter Software, Setauket, NY. Saab, V.: 1999, ‘Importance of spatial scale to habitat use by breeding birds in riparian forest: A hierarchical analysis’, Ecol. Appl. 9, 135–151. Simon, J. L.: 1997, Resampling: The New Statistics, Resampling Stats Inc., Arlington, VA. Tarvin, K. A. and Garvin, M. C.: 2002, ‘Habitat and nesting success of blue jays (Cyanocitta cristata): Importance of scale’, Auk 119, 971–983. Temple, S. A.: 2002, ‘Dickcissel (Spiza americana)’, in: A. Poole and F. Gill (eds), The Birds of North America, no. 703, Academy of Natural Sciences, Philadelphia, PA, and American Ornithologists’ Union, Washington, DC. Van Horssen, P. W., Schot, P. P. and Barendregt, A.: 1999, ‘A GIS-based plant prediction model for wetland ecosystems’, Landscape Ecol. 14, 253–265. Vickery, P. D., Tubaro, P. L., Cardosa, J. M., Peterjohn, B. G., Herkert, J. R. and Cavalcanti, R. B.: 1999, ‘Conservation of grassland birds in the Western Hemisphere’, Stud. Avian Biol. 19, 2–26. Westphal, M. I., Field, S. A., Tyre, A. J., Paton, D. and Possingham, H. P.: 2003, ‘Effects of landscape pattern on bird species distribution in the Mt. Lofty ranges, South Australia’, Landscape Ecol. 18, 413–426. Wiens, J. A.: 1989, ‘Spatial scaling in ecology’, Funct. Ecol. 3, 385–397. Wiens, J. A., Addicott, J. F., Case, T. J. and Diamond, J.: 1986, ‘Overview: The Importance of Spatial and Temporal Scale in Ecological Investigations’, in: J. Diamond and T. J. Case (eds), Community Ecology, New York, NY, pp. 145–153. Wiens, J. A., Rotenberry, J. T., and Van Horne, B.: 1987, ‘Habitat occupancy patterns of North American shrubsteppe birds: The effects of spatial scale’, Oikos 48, 132–147. Zimmerman, J. L.: 1993, The Birds of Konza, University of Kansas Press, Lawrence, KS.