Environmental Management

, Volume 38, Issue 3, pp 470–486

Development of a Bird Integrity Index: Measuring Avian Response to Disturbance in the Blue Mountains of Oregon, USA

Authors

ENVIRONMENTAL ASSESSMENT

DOI: 10.1007/s00267-005-0152-z

Cite this article as:
Bryce, S.A. Environmental Management (2006) 38: 470. doi:10.1007/s00267-005-0152-z

Abstract

The Bird Integrity Index (BII) presented here uses bird assemblage information to assess human impacts to 28 stream reaches in the Blue Mountains of eastern Oregon. Eighty-one candidate metrics were extracted from bird survey data for testing. The metrics represented aspects of bird taxonomic richness, tolerance or intolerance to human disturbance, dietary preferences, foraging techniques, and nesting strategies that were expected to be positively or negatively affected by human activities in the region. To evaluate the responsiveness of each metric, it was plotted against an index of reach and watershed disturbance that included attributes of land use/land cover, road density, riparian cover, mining impacts, and percent area in clearcut and partial-cut logging. Nine of the 81 candidate bird metrics remained after eliminating unresponsive and highly correlated metrics. Individual metric scores ranged from 0 to 10, and BII scores varied between 0 and 100. BII scores varied from 78.6 for a minimally disturbed, reference stream reach to 30.4 for the most highly disturbed stream reach. The BII responded clearly to varying riparian conditions and to the cumulative effects of disturbances, such as logging, grazing, and mining, which are common in the mountains of eastern Oregon. This BII for eastern Oregon was compared to an earlier BII developed for the agricultural and urban disturbance regime of the Willamette Valley in western Oregon. The BII presented here was sensitive enough to distinguish differences in condition among stream riparian zones with disturbances that were not as obvious or irreversible as those in the agricultural/urban conditions of western Oregon.

Keywords

Aquatic monitoringRiparian ecosystem conditionStream assessmentIndex of biotic integrityIBIBioassessmentRiparian birdsDisturbance

The Environmental Protection Agency and other federal and state agencies have used biological indicators to monitor the status and trends of aquatic ecosystems since the 1980s. Biological indicators are measured attributes of organisms, assemblages, or communities that record a response to conditions in the environment (Adams 1990; Cairns and others 1993; Paulsen and Linthurst 1993). A multimetric Index of Biotic Integrity (IBI) translates the assemblage composition of a regional fauna at a site into a numeric score that allows a comparison of ecological condition among sites. Typically, a site score is the sum of a number of individual metric scores calculated from taxa-richness measures or species-proportionate abundance data. Candidate metrics are drawn from species’ functional guild and life history information. They represent attributes of an assemblage that are expected to vary systematically across a gradient of human disturbance (Karr and Chu 1999; Karr and others 1986). The behavior of individual metrics comprising an index also provides diagnostic evidence to aid in the interpretation of possible causes of biological response. The strength of the IBI approach is its theoretical foundation, which is based on the structural and functional roles of native biotic assemblages (Karr and Chu 1999; Miller and others 1988).

Early IBI development focused on aquatic assemblages as indicators of stream condition, particularly fish (Karr and others 1986; Simon 1999), aquatic macroinvertebrates (Fore and others 1996; Kerans and Karr 1994), and periphyton (Fore 2002; Hill and others 2000). However, because many of the sources of stream impairment originate in the terrestrial environment, the value of assessment information gained for stream ecosystem management is enhanced by the inclusion of riparian condition assessments. Among the riparian indicators, birds are one of the most responsive and economical to sample (Brooks and others 1991, 1998; Bryce and others 2002; Hunsaker and others 1990); they are sensitive to watershed and riparian disturbances that are indirectly inferred from the response of aquatic indicators. The theoretical framework for using birds as riparian indicators is derived from a large body of research related to bird trophic and reproductive guilds. Functional guilds that classified birds according to diet, foraging behavior, and nesting preferences (Root 1967; Severinghaus 1981; Verner 1984) evolved into response guilds (Best and others 1978; Croonquist and Brooks 1991; Hansen and Urban 1992; Szaro 1986) that considered avian functional groups relative to their response to human activities.

Since the early 1990s, researchers have applied various combinations of metrics based on avian functional and response guilds to address cumulative impacts on wetland areas in Pennsylvania (Croonquist and Brooks 1991), forest fragmentation and lakeshore development in New England (Moors 1993), grazing impacts in the semiarid Great Basin of Idaho and Utah (Bradford and others 1998), the condition of military training lands in Kansas (Cully and Winter 2000), and an agriculture/urban gradient in western Oregon (Bryce and others 2002). All of these studies recorded bird response to terrestrial site, stream reach, and/or watershed impacts. Brooks and others (1998) proposed that biological indicator information could also be used at regional, population/species, and genetic levels of organization, as suggested by Noss (1990). O’Connell and others (2000) demonstrated the index approach at a regional scale by aggregating bird metric information for a large number of widely distributed sites to give a regional picture of landscape condition for the Mid-Appalachian Highland region. Once multimetric indices such as these are designed and calibrated for a particular study area, one could expect similar results when they are applied at other locations in the same ecological region (Bryce and others 1999b; Hansen and Urban 1992). However, for any of these index methods to gain wider acceptance, their applicability must be demonstrated across a variety of ecoregions and disturbance types.

The objective of the present study was to test a Bird Integrity Index (BII), originally developed for riparian systems in the agricultural Willamette Valley of western Oregon (Bryce and others 2002), on a group of stream riparian zones in the forested Blue Mountains of eastern Oregon. It was an opportunity to test the sensitivity of the BII in a region with a narrow range of human impacts, where most disturbances were related to natural resource extraction and where most sites were in good to fair condition.

The BII is a riparian condition index that uses birds as condition indicators through the systematic response of avian guild-based metrics to a range of human disturbance factors. Although one can acquire interesting and useful information concerning the dynamics of bird assemblages from riparian bird surveys such as this, such information is secondary to the primary objective of assessing the condition of stream riparian zones.

Methods

Study Area

In 1999 the Environmental Protection Agency’s Environmental Monitoring and Assessment Program (EPA EMAP) initiated a 5-year survey of wadable and boatable streams and rivers in 11 western states as part of an effort to develop biological indicators and standardized field methods. Field crews sampled fish, aquatic macroinvertebrates, and stream physical and chemical habitat at an average of 250 randomly selected stream reaches per year across the western United States. EMAP field protocols were designed to collect standard biotic and abiotic information at randomly selected sites to give unbiased regional estimates of various aspects of stream condition (Olsen and others 1999; Peck and others 2003; Stevens 1997). This study of 28 second to fourth order, wadable stream reaches in the forested Blue Mountains of the John Day River basin of eastern Oregon was added to the broader EMAP field effort as a more intensive investigation of stream riparian condition (Figure 1). The sites were selected by intensifying the EMAP random sampling grid to produce a concentration of sites in the Blue Mountain region. In order to complete the fieldwork in a timely manner, EPA further limited the field sampling to sites on public land. Between 2000 and 2004, riparian birds and vegetation were sampled in addition to the standard EMAP aquatic indicators at each stream reach.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-005-0152-z/MediaObjects/267_2005_152_f1.jpg
Figure 1

Locations of sample stream reaches, repeat sites (circled), and Level IV ecological regions (Thorson and others 2003) in the upper John Day River basin in the Blue Mountains of eastern Oregon.

The John Day River basin encompasses a complex of mountain ranges that make up the Blue Mountains of eastern Oregon (Figure 1). The sample stream reaches are located in several forested ecoregions distinguished by differences in elevation, precipitation, geology, soil, and vegetation (Clarke and Bryce 1997; Thorson and others 2003). Dry forests of ponderosa pine (Pinus ponderosa) dominate the mid-elevation zone (Figure 1, Ecoregions 11b, 11d, and 11h). Historically, the ponderosa pine forest resembled a savanna, with widely spaced large trees growing above grasses and scattered shrubs. Frequent ground fires consumed the fine fuels, but they generally spared the large fire-resistant trees (Johnson and others 1994; Lehmkuhl and others 1994). Presently, after a century of fire suppression and logging, the open-canopied forest has been replaced by thickets of young trees (although management strategies have been changing over the last 10 years to more closely emulate the natural fire regime) (Oliver and others 1994; Robbins and Wolf 1994). A shrub cover of snowberry (Symphoricarpus alba), serviceberry (Amelanchieralnifolia), rose (Rosa woodsii), willow (Salix spp.), or black hawthorne (Crataegusdouglasii) typically grows in the riparian areas of the ponderosa pine zone, but shrub distribution is presently reduced in many areas due to grazing impacts (Irwin and others 1994). The Blue Mountains support closed-canopy forests of spruce and fir (Picea engelmannii, Abies grandis, and Abies lasiocarpa) at higher elevations, where there is more rainfall and a heavier snowpack (Figure 1, Ecoregion 11l). Grazing is widespread across the entire John Day basin although it is less prevalent in the denser high-elevation forests. Land uses include ranching at lower elevations, with logging, summer cattle grazing, and mining on public lands at higher elevations.

Field Methods

The bird survey in eastern Oregon extended across a 5-week period during the breeding season, from the first week in June through the first week of July. Each site was sampled once during this period. Six sites were revisited in 2003 and 2004 to test interannual variability (circled sites in Figure 1). The author sampled birds from sunrise to 4 hours past sunrise using a point count method to record all birds heard or seen during a 10-min time frame. Stops were placed every 100 m along a 1-km transect parallel to the stream, and data from the 10 stops were pooled for analysis. The point count circles did not have a fixed radius because the object of the survey was to characterize bird species composition for a stream reach rather than to determine avian population densities. As a result, I recorded every bird heard at a sampling point to maximize the number of observations in the database, taking care not to double count birds recorded in the previous 100-m segment. Normal sound attenuation and stream noise tended to keep detection distances within the riparian zone and within the point count circle. The plot design balanced the need to maximize the number of birds recorded while not overly exceeding the shorter section of stream reach shared with the other EMAP in-stream indicators.

Defining a Disturbance Gradient

Biological assemblages can be useful indicators of ecological condition if we are successful in untangling the effects of human disturbance from the natural stressors in their environment. It is not possible to characterize all of the numerous pathways by which impairment occurs in riparian and aquatic ecosystems, but we can identify and try to quantify human activities that contribute to the degradation of biotic integrity.

Ranking sites along a human disturbance gradient provides an independent measure of stream and riparian conditions that can be used in both multimetric and multivariate analyses. Disturbance measures are essential for evaluating the responsiveness of candidate metrics for indices of biological integrity. In this case, plotting candidate bird metrics against a measure of human disturbance reveals (1) whether indicator response covers the full spectrum of habitat quality; (2) which bird metrics are sensitive to extremes in impact; and (3) which bird taxa disappear or become more prevalent with increasing disturbance. The evaluation process aids in the selection of metrics with the strongest signal to compose a useful bird integrity index.

Screening the stream reaches and their watersheds in the John Day basin followed iterative analyses of maps, aerial photographs (1:40,000 scale), satellite imagery land-cover data (NLCD) (Vogelmann and others 2001), and physical habitat information. All visible human alterations to native ecosystems, including their number, type, intensity, and extent of impact were recorded, ranked, and quantified where possible, following the method of Bryce and others (1999a). Disturbances smaller than the scale of the available data or otherwise invisible (toxics, excessive nutrients) were not recorded. The most convenient spatial unit for tabulating disturbances was the watershed (delineated upstream from each sampling point). Although the activities of resident birds are obviously not constrained by watershed boundaries, using the watershed as a common spatial unit allowed the direct comparison of bird survey results to those of the other in-stream indicators (i.e., fish and macroinvertebrates).

The resulting disturbance index is integrative, drawing on several hierarchical levels (watershed, riparian zone, and stream channel) as a source for disturbance metrics. Information collected at multiple scales in the region of inquiry gives a more complete representation of the local mechanisms of disturbance and their effects on the avian community. Local land uses have brought broad-ranging changes to the streams and watersheds of the region (Irwin and others 1994; Johnson and others 1994; Wissmar and others 1994). Past tractor logging created scars of parallel road networks on mountain contours. Placer and dredge mining left miles of streams without defined channels. Riparian shrubs have been depleted over 150 years of cattle grazing and cultivation of alluvial land (Elmore and Beschta 1987; Sanders and Edge 1998).

To create metrics that quantified these disturbances, (1) clearcut and partial-cut patches were delineated on aerial photography and their areal extent was estimated using a dot grid, (2) geographic information system (GIS) road coverages (Bureau of the Census 1992) were updated using recent US Forest Service fire road maps and digitized to calculate road density; and (3) road densities within a 30-m buffer along the stream network and miles of stream having a parallel road within 30 m were tallied for each watershed. Other metrics incorporated quantitative physical habitat information for riparian vegetation and stream channel condition (e.g., canopy- and mid-layer vegetation and in-stream large woody debris cover) (Kaufmann and others 1999). Ordinal metrics (scored 0, 1, 5, and 10 for zero, low, medium, and high impacts) were created for attributes that lacked quantitative data, such as grazing and mining impacts, using field knowledge, map analysis, and best professional judgment. The final disturbance index of 17 metrics contained 9 quantitative and 8 ordinal metrics, combining quantifiable aspects of watershed condition and qualitative interpretations of the degree of impact (Table 1). Raw metric scores were normalized to fall between 0 and 10 and were added together to create a total disturbance score for each stream reach and its watershed. Disturbance totals for each site were normalized again to fall between 0 and 100, with higher scores denoting increased disturbance.
Table 1

Seventeen measures of regional human disturbance comprise the disturbance index

Extent

Quantitative metrics

Watershed

percentage watershed forested area clearcut [from National Land Cover Database (NLCD)]

Watershed

percentage watershed forested area partial cut (NLCD)

Watershed

Updated road density (km/km2)

Riparian network

Road density (km/km2) in 30-m buffer along stream network (each side of stream)

Riparian network

Total length of road in 30-m buffer expressed as a percentage of total stream length

Riparian reach

Riparian trees greater than 0.3 m dbh (positive attribute, score reversed)

Riparian reach

Riparian vegetation: canopy + mid-layer + woody ground cover (positive attribute, score reversed)

Riparian reach

Riparian disturbances, sum all types

Reach–stream channel

Fish cover: large woody debris (positive attribute, score reversed)

 

Ordinal metrics

Watershed

Underground mines

Watershed

Watershed grazing pressure

Watershed

Town sites

Watershed

Riparian agriculture in the watershed

Watershed/reach

Channel modifications in reach or watershed: channelization, riprap, structures, alluvial recontouring, irrigation canals

Riparian network

Underground or placer mines and tailings in riparian zone

Riparian reach

Inadequate reach riparian woodland/shrub cover (less than 10m each side of stream)

Riparian reach

Reach grazing pressure

Note: The first nine metrics listed are quantitative; the last eight are ordinal and are scored 0, 1, 5, and 10 for zero, low, medium, and high disturbance.

Defining Reference Condition

The creation of a disturbance index that convincingly translates the mechanisms of disturbance into numerical values depends on its calibration with a model of reference condition (Bryce and others 1999b; Hughes 1995; Hughes and others 1986). Every ecological region contains a set of existing (or conceptual) characteristic natural landscapes and associated biotic communities that are minimally disturbed by human activities and that represent the range of reference conditions. It is often assumed that the associated flora and fauna existing at a minimally disturbed reference site represent biotic integrity for that region. Therefore, any site assessed for disturbance can be compared with an ecoregional reference model to assess the amount of divergence of the disturbed site from regional expectations for biotic integrity. To successfully calibrate an index of biotic integrity using a measure of human disturbance, the process requires that a sufficient number of the sites sampled should represent both minimally disturbed reference sites and highly disturbed streams and watersheds. In this study, as often happens, most of the randomly selected field sites were in fair condition. In the initial sample, only two streams met reference quality, and just three sites were highly disturbed. In 2004, six hand-picked sites (four reference and two highly disturbed) were added to the random sites and surveyed in the field in the expectation of strengthening index response. Stream reaches and catchments in old growth or mature ponderosa pine (P. ponderosa) forest were particularly hard to find, as this vegetation type has declined by approximately 85% across eastern Oregon since presettlement times (Sallabanks and others 2001). The criteria for reference site selection were that the stream reaches were representative of the ecoregions comprising the study and that they lacked all of the human disturbances common in the region. Bird response at these minimally disturbed reference sites served as an indication of the character of bird assemblages under natural conditions in the region.

Candidate Metrics

We derived candidate metrics from the lists of bird species and relative abundances in the survey dataset (see the Appendix). The metrics, 81 in all, represented aspects of bird taxonomic richness, tolerance or intolerance to human disturbance, dietary and foraging preferences, and nesting strategies.

Taxonomic richness and abundance

The simplest metrics are those based on species richness and total number of individuals. Metrics incorporating information on warblers and other neotropical migratory species (long-distance migrants) were expected to decrease with increasing human disturbance (Table 2). Many neotropical migrants are considered sensitive, and declines in neotropical species due to human disturbance and habitat fragmentation have been well documented (Temple 1986; Terborgh 1989; Wilcove and Terborgh 1984).
Table 2

Metric categories, expected response, and number of metrics in each category that were responsive to disturbance in eastern Oregon and earlier western Oregon BII

Metric category

Expected response

Metrics responsive to disturbance eastern Oregon

Metrics responsive to disturbance western Oregon

Total abundance

Increases with moderate disturbance; decreases with heavy disturbance

0

0

Native species richness

Increases with moderate disturbance; decreases with heavy disturbance

0

1

Neotropical migrants

Decrease

0

2,3

Warblers

Decrease

0

1,3,4

Tolerants

Increase

3

3,4

Intolerants

Decrease

3

1,2,3

Omnivores/granivores

Increase

3

3,4

Insectivores

Decrease

4

1,3

Ground gleaners

Increase

4

3,4

Foliage gleaners

Decrease

4

2,3,4

Bark Gleaners

Decrease

0

3,4

Woodland ground nesters

Decrease

0

2,4

Cavity nesters

Decrease

0

3

Native cavity nesters

Decrease

0

1,2,3

Nest sensitive (woodland ground nesters + native cavity nesters)

Decrease

3

1,2,3,4

Nest Parasite: presence/absence

Increase

1

N/A

Nest parasite abundance

Increase

4

N/A

Nest parasite hosts

Decrease

0

N/A

Canopy granivores

Decrease

0

N/A

Hawking species (i.e., flycatchers)

Decrease

0

N/A

Interior forest species

Decrease

3,4

N/A

Habitat-specific species

Decrease

0

N/A

Shrub-dependent species

Decrease

0

N/A

Riparian obligate species

Decrease

0

N/A

Note: Four metrics were tested in each category (except total abundance, native species richness, and nest parasite): number of species, number of individuals, percent species, and percent individuals (labeled 1, 2, 3, or 4 in columns 3 and 4). Of 81 metrics tested in eastern Oregon, 11 showed a clear response to the disturbance measure (i.e., they showed a consistent increase or decline across the range of human disturbance). Of 62 metrics tested in western Oregon, 32 showed a clear response to disturbance in the first screening.

Tolerance to human disturbance

Tolerant and intolerant bird species were distinguished from ubiquitous species (widely found in both disturbed and undisturbed habitats) and generalists (adaptable to multiple habitats) by the use of literature review and field experience (species and guild assignments listed in the Appendix) (Marshall and others 2003; Poole and Gill 1999). Ubiquitous species were generally few in number, but were present in almost every habitat sampled. The majority of species detected were considered generalists. Generalist species might respond to disturbance in one of the other metric categories, but they do not have a known sensitivity (or tolerance) that would place them in the intolerant (or tolerant) group. In eastern Oregon, intolerants include species most affected by clearcut logging or the destruction of riparian shrubs and trees through intentional removal, grazing, or mining. Conversely, tolerants are those bird species that benefit from such activities and the resultant habitat changes to earlier successional stages.

Foraging and dietary guilds

Bird species were classified into foraging guilds and dietary guilds according to their feeding strategies and predominant food preferences during the breeding season, as indicated in the work of Terres (1980), Ehrlich and others (1988), and DeGraaf and others (1991) (see the Appendix). Any disagreements among the references were settled based on information specific to Oregon bird assemblages (Marshall and others 2003) or professional judgment and local field knowledge. Commonly recognized foraging guilds used in metric development were foliage gleaners, bark gleaners, ground gleaners, and aerial foragers. Metrics based on dietary guilds included omnivores (consumers of both animal and vegetable material), granivores (both ground and canopy seed eaters), insectivores, carnivores, and fruit, foliage, and nectar consumers.

Nesting strategies

Birds encountered in the survey were classified by nesting strategy into platform, low and high cup, ground, and cavity nest types (see the Appendix). Ground nesting and cavity nesting were considered the two strategies most likely to be affected by logging, predation, and trampling by grazing animals. Several cavity nesters in eastern Oregon have evolved to nest in large, old-growth snags, which have declined after years of high-grade logging. Increased levels of grazing and logging also promote expansion of the nest parasite, the brown-headed cowbird (Molothrus ater). The cowbird can have a significant impact on the nesting success of some warbler, flycatcher, and vireo species, particularly those species that are already in decline for other reasons (Muehter 2004).

Metric Development

The metric categories discussed in the previous subsection were selected in an attempt to represent important ecological characteristics of bird assemblages living in the mountains of eastern Oregon. A quartet of metrics was tested for each metric category: (1) number of species, (2) number of individuals, (3) percent species, and (4) percent individuals. Each metric underwent a screening process designed to choose the most responsive, precise, and repeatable metrics to build an index with high indicator value and a strong signal.

For each of the 81 metrics tested, the scores for each site were plotted against the disturbance index scores for each stream reach and its upstream watershed. To qualify for inclusion in the final index, metrics had to show a consistent increase or decline across the range of human disturbance. Responsive metrics were also tested for covariation with elevation and watershed area, natural factors that might have masked or accentuated the relationship of metrics with the human disturbance measure.

Metrics that remained after the first screening were tested for redundancy and year-to-year variability. A correlation matrix was examined to test for highly correlated (and thus redundant) metrics. Then, for six stream reaches that had been resurveyed in 2002, scatterplots of visit 1 versus visit 2 were created for each metric and the full BII to test for year-to-year variability. Finally, analysis of variance (ANOVA) was used to evaluate the precision of the BII scores.

Results and Discussion

Bird Assemblages

Species richness varied from 14 to 27 among the 28 sampled stream reaches. Low species numbers occurred at very dry sites with a south or southwestern aspect or in denser closed canopy forest sites at higher elevations. The highest species richness occurred at sites with a wet meadow/woodland interface. A number of bird species were abundant across xeric forest, mesic forest, and meadow edges at all elevations: American robin (Turdus migratorius), dark-eyed junco (Junco hyemalis), mountain chickadee (Poecile gambeli), western tanager (Piranga ludoviciana), and yellow-rumped warbler (Dendroica coronata) (abundance figures listed in the Appendix). Conversely, species seldom encountered in the survey included large tree or snag nesters, such as Lewis’ woodpecker (Melanerpes lewis), Williamson’s sapsucker (Sphyrapicus thyroideus), and the white-headed woodpecker (Picoides albolarvatus); inhabitants of riparian shrubs removed by grazing (particularly willow), such as yellow warbler (Dendroica petechia) and willow flycatcher (Empidonax traillii); and birds adapted to the historical fire regime, such as olive-sided flycatcher (Contopus bocooperialis) and Townsend’s solitaire (Myadestes townsendi).

Metric Selection and Performance

This subsection describes the general response and utility of the major metric categories following the first screening for responsiveness. It is useful to consider unresponsive metrics as well as responsive ones; occasionally, they might be modified to work more effectively. Closer examination might enlighten us about metric application under various conditions (e.g., why a metric might be responsive in one region and not another).

Taxonomic richness and abundance

Species richness behaved as expected, with lower species numbers in dense, closed-canopy forests with minimal disturbance, increasing numbers at sites with moderate disturbance, and decreasing species numbers at highly disturbed sites. Although this response has been described by Odum and others (1979) and Odum (1985), a metric showing a peaked response is not amenable for use in an index. Also, species richness is not a particularly informative metric for condition assessments; we are more concerned with the identity and role of individual species or functional groups at each site. Metrics based on the relative abundances of warbler species and neotropical migratory bird species were not responsive to the disturbance gradient probably because these assemblages were relatively evenly distributed across a forested landscape, where there was often little distinction in vegetation type between riparian and upland areas.

Tolerance to human disturbance

Tolerant and intolerant species represent the tails of a hypothetical distribution of bird species inhabiting a region; the majority of species encountered are typically habitat generalists. However, a limited number of sensitive (or insensitive) species responding at each end of the disturbance gradient clarifies the overall assemblage response to regional human disturbance and contributes to a useful index of biotic integrity. Of the species detected during the bird survey, 27.3% were considered intolerants (see the Appendix), although a number of these species occurred only as one or two individuals in the database. Intolerant individuals comprised 8.9% of the total number of individuals observed. Intolerant species included mature forest cavity nesters and reclusive or area-dependent species sensitive to logging disturbances, such as the hermit thrush (Catharus guttatus) and Cassin’s vireo (Vireo cassinii). Riparian shrub inhabitants, such as the yellow warbler (Dendroica petechia) and orange-crowned warbler (Vermivora celata), comprised another category of intolerant species. Although the metric percent intolerant species was responsive to the disturbance gradient, the metrics referring specifically to shrub-dependent species were not responsive.

Tolerant species made up 14.7% of the total number of species detected, and tolerant individuals comprised 7.3% of the total number of individuals observed. Tolerant species, such as the Dusky flycatcher (Empidonax oberholseri) and Nashville warbler (Vermivora ruficapilla), inhabit early successional conifer forest and cut-over land (DeGraaf and others 1991; Marshall and others 2003). Percent tolerant species was the most responsive metric in the tolerant category.

Foraging and dietary guilds

Ground and foliage gleaning metrics responded to disturbance, but bark gleaning metrics did not. Although ground gleaning species occurred in both undisturbed and disturbed habitats, the proportion of ground gleaning individuals increased noticeably in fragmented forest habitats, making it a useful metric. The metric percent foliage gleaning individuals, although responsive, registered a negative response only at high disturbance levels, where tree canopy along the reach was sparse or lacking. The failure of the bark gleaning metric might have been due to the fact that auditory detections typically dominate bird sampling in point counts, and it was likely that some bark gleaners were overlooked. The most vocal bark gleaner, the red-breasted nuthatch, was common in any wooded setting in the study area.

The dietary preferences most useful in metric development were omnivores (consumers of both animal and vegetable material), granivores (seed eaters), and insectivores. Both the omnivore/granivore and insectivore metrics were responsive to disturbance, but the metric representing the “hawking” foraging technique was not responsive. Hawking species were fairly evenly divided between open- and closed-canopy and/or early and late successional forest types, indicating that the metric was too broadly defined. Mannan and others (1984) discussed the problem of scale in guild definition, noting that habitat change might be captured more accurately by making the criteria for species inclusion in guilds more restrictive (i.e., constructing guilds at a “finer scale”). In this case, the hawking metric might be improved if it were tailored to an expected disturbance response of a subgroup of fly-catching species.

Nesting strategies

The response of ground and cavity nesting metrics (r value of −0.30 and −0.42 respectively) improved when they were added together to form a nest-sensitive metric (r value of −0.47). The numbers of ground and cavity nesters recorded in this study might have been somewhat too low to justify separate metrics.

Two metrics related to cowbird presence and abundance were more responsive than metrics related to cowbird host species. The cowbird (Molothrus ater), normally common in rangeland, has extended its range to higher elevations as logging expanded road networks and human-created openings. In this survey, the cowbird was recorded as high as 1695 m (5560 ft) in a subalpine meadow–forest–clearcut habitat. The nest parasite metric plot shows an on–off or step pattern indicating a possible threshold in disturbance intensity or extent (Figure 2). Brown-headed cowbirds appeared at sites with disturbance scores greater than 25, where roads and logging fragmentation began to occur on the landscape. Cowbirds might have been present but undetected at sites with higher disturbance scores (30s and low 40s) and no cowbirds recorded.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-005-0152-z/MediaObjects/267_2005_152_f2.jpg
Figure 2

Final nine metrics comprising the BII plotted against the reach/watershed disturbance index. Disturbance scores were based on percent forested area partial and clearcut, road density metrics, riparian vegetative cover, and ordinal estimates of mining and grazing impacts.

Final Index and Distribution of Site Scores

In the initial screening, those metrics that showed a consistent increase or decline across the range of human disturbance were considered responsive. Those metrics that showed a definite response also had the greatest range between high and low values. Of 81 individual metrics tested in 22 metric categories, 11 responded well to the disturbance index (Table 2, column 3). The visual assessment of the plots was corroborated by examining the correlation between each responsive metric and the disturbance index. Two of the 11 responsive metrics were from the same metric category (percent interior forest dwelling species and individuals); the one with the highest correlation to disturbance was chosen (percent interior forest dwelling species).

This group of 10 metrics was then tested for statistical redundancy and year-to-year variability. Examination of an 11 × 11 correlation matrix (10 metrics plus disturbance index values) revealed that the metrics cowbird presence/absence and percent cowbird individuals were highly correlated with each other (Pearson r > 0.80). The metric percent cowbird individuals was retained because it contained both presence/absence and abundance information. Thus, the final BII contained nine metrics after unresponsive and redundant metrics were removed (Figure 2).

To test year-to-year variability, visit 1 × visit 2 plots were constructed for each metric for the six sites that had been revisited in 2002 (Figure 3). Although several metrics, particularly percent cowbird individuals (P_PARAIN), percent intolerant species (P_INTOLSP), and percent ground gleaning individuals (P_GGIN), were more variable than the others, a portion of this noise disappeared when the metrics were combined into the full BII index. Other authors have done quantitative analyses of this reduction in error variance between individual metrics and the final index (Hughes and others 1998, Kimberling and others 2001). The revisit plot for the BII (BIISCORE) shows good repeatability (bottom row of Figure 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-005-0152-z/MediaObjects/267_2005_152_f3.jpg
Figure 3

Visit 1 × visit 2 plots for each metric composing the BII as well as a repeat plot for the BII itself (last plot, bottom row). Most of the metrics and the index show good repeatability.

Individual metric scores were calculated by dividing the raw metric value by the highest recorded value for that metric and multiplying by 10. Five metrics, representing the abundances of tolerants, nest parasites, omnivore/granivores, and ground gleaners increased with higher disturbance, so their metric values were replaced with their complements. The BII scores for each site were calculated as the sum of the metric scores for the nine metrics comprising the index. The BII scores for the 28 stream reaches ranged from 30.4 to 78.6, with an average of 52.1 (Figure 4A). When plotted against the disturbance index, the BII has a stronger relationship (r = −0.81) than any of the component metrics alone (r = 0.33 to −0.72). The outlier in Figure 4A has a low BII score not consistent with its disturbance score. This is partly due to the fact that the stream reach is located in a landscape drier than the rest of the sample at the border of ponderosa pine (P. ponderosa) woodland and juniper/sage (Juniperus/Artemisia spp.) shrubland (Ecoregions 11b and 11a in Figure 1). Grazing, one of the more difficult activities to measure quantitatively, was more prominent than logging in this watershed.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-005-0152-z/MediaObjects/267_2005_152_f4.jpg
Figure 4

Two plots of the BII: (A) original plot of BII scores for 28 stream reaches plotted against disturbance indexand (B) residuals of BII scores versus the disturbance index scores after correcting for natural variability in elevation. The pattern of the adjusted scatterplot remains similar to the original, suggesting that the BII can be applied across the study area’s elevational gradient.

The site scores fell into three general groups (Figure 4A): (1) minimally disturbed reference streams and watersheds in or near wilderness areas with BII scores above 60; (2) disturbed reaches and watersheds with varying road-building, logging, grazing, and mining intensities and BII scores between 40 and 60; and (3) highly disturbed streams with BII scores below 40; these experienced loss of riparian vegetation, extensive reach and watershed mining, and/or channel modifications. These groups suggest thresholds for setting good, fair, and poor condition classes; however, it would be premature to do so before the index has been tested on an independent set of sites.

Considering Natural Covariates

The sample site distribution spanned two distinct mountainous ecoregions: a lower elevation [610–1524 m (2000–5000 ft)] xeric forest dominated by ponderosa pine (P. ponderosa) (Figure 1, Ecoregion 11b, 11d, and 11h) and a higher elevation [1372–1982 m (4500–6500 ft)] close-canopied spruce–fir forest (Picea engelmannii and Abies lasiocarpa) (Figure 1, Ecoregion 11l). The sample size was not large enough to develop a bird integrity index for each ecoregion separately; the objective of the BII project was to build an index that was robust enough to characterize the riparian condition across this elevation range.

Bird assemblage richness and diversity have been shown to vary with elevation in other regions of North America (Able and Noon 1976; Knopf 1985; Waterhouse and others 2002). Therefore, the BII metrics might have responded to an elevation (and moisture) gradient as well as varying human disturbance. It was necessary to remove the influence of elevation to confirm that a single BII could be applied to both the pine and spruce/fir ecoregions. First, each BII metric was plotted against elevation. Several of the plots had a correlation coefficient > 0.50, suggesting a possible covariation with elevation. Then each BII metric was regressed against elevation and the resulting residuals were added together to create an adjusted BII score for each site. The adjusted BII scores were plotted against the disturbance index scores. When the influence of elevation was removed, the resulting plot for the adjusted BII (Figure 4B) showed a similar pattern to the original index plot (Figure 4A) in BII response to disturbance, although with a somewhat lower correlation coefficient (r = −0.81 vs. r = −0.70). The plots suggest that the BII can be applied to both pine- and spruce/fir-dominated ecoregions in the Blue Mountains.

Although the stream sites were considered forested at a regional level, at a finer scale the forests were often open-canopied or broken by dry or wet meadows. Bird sampling was systematic along the kilometer-length transect, and point counts were conducted regardless of changes in habitat from forest to meadow. Two of the 17 metrics in the disturbance index were measures of existing riparian vegetation. These two riparian metrics did not distinguish whether a lack of riparian vegetation was due to the presence of a natural meadow or the consequence of human activity. As a result, disturbance index scores might have been slightly depressed (1–3 points) for stream reaches with meadow inclusions (estimated by comparing scores of reference sites with meadow inclusions to those reference sites that were completely forested).

Interspersed meadow habitats also affected BII scores. Obviously, birds of meadow habitats differ in guild membership from forest birds. One can expect increased numbers of granivores, ground gleaners, and tolerant birds in meadow habitats, and these are all metric categories that receive lower scores in this index as a result of human disturbance in forested habitats. As a result, stream reaches with significant meadow portions had slightly depressed BII scores. However, even though three of the reference sites had large meadow patches along the transect and a slightly depressed disturbance index and BII scores as a result, the scores of the group of minimally disturbed reference sites clearly exceeded those of the moderately and highly disturbed sites.

Precision of the Bird Integrity Index

The root mean square error (RMSE) of within-site variance (BII scores from revisits to six sites in subsequent years) was 3.2, which is 6.6% of the observed BII range (48.2) among sites (n = 28). If the BII scores differ by 6.4, or two times the RMSE (which approximates the standard deviation of repeat measurements at the same site), then one can be 95% certain that the scores are truly different values rather than random measurement error. The signal-to-noise ratio of the BII (variance among streams/variance of repeat visits) was 14.8, meaning that the interannual variability in BII scores between visits to the same stream were small compared to the differences in BII scores among all streams in the sample. Thus, measurement error is small compared to the variability in avian response that results from either natural or human disturbance factors. An index with a signal-to-noise ratio as high as 14.8 allows one to explain about 90% of the variance between the index and a predictor variable of similar precision (Kaufmann and others 1999).

Comparison of Two Oregon Bird Integrity Indices

An earlier BII was developed for stream reaches in the urbanizing, agricultural Willamette Valley in western Oregon (Bryce and others 2002). There the BII clearly reflected the effects of intensive agriculture and urbanization on sensitive bird species in the limited riparian habitat found along streams of the region. In the Blue Mountains of eastern Oregon, on the other hand, human disturbance is mainly confined to natural resource utilization and extraction (logging, grazing, and mining). The impacts are extensive and can be locally severe; however, although the forest is fragmented, the overall land-cover matrix remains forest and the end points of change are not as irreversible as with urbanization.

Fewer bird species and individuals were encountered in the present survey than the number recorded in western Oregon (Table 3). In terms of field effort, there were detection differences between the two regions: (1) Birds in the Blue Mountains spent more time quietly foraging; (2) the peak morning time frame for bird song was shorter in the Blue Mountains than in the Willamette Valley, as auditory detections often waned by 8:30 AM (rather than 9:30 AM), particularly in hot weather; and (3) as a result, in the Blue Mountains, a larger proportion of detections were visual, in contrast to the Willamette Valley, where detections were almost entirely auditory.
Table 3

Regional differences in number of species, average number of species, number of individuals, and average number of individuals for bird surveys in the mountains of eastern Oregon compared to an agricultural valley in western Oregon

Geographic region

Range in number of species

Average number of species

Range in number of individuals

Average number of individuals

Eastern Oregon

14–27

19

39–99

65

Western Oregon

16–35

25

105–258

151

Applying the BII in a different ecological region required rethinking the ecological relationships important to bird assemblages in riparian habitats there and hypothesizing assemblage response to common human disturbances. Metrics for the standard dietary, feeding, and nesting strategies were expected to apply in both regions. Additional metrics pertaining specifically to conditions in the Blue Mountains tested species’ riparian obligacy, shrub dependency, area sensitivity (interior and edge species), cowbird parasitism, and the abundance of fly-catcher species and canopy granivores. However, although 81 metrics were tested in eastern Oregon (compared to 62 in western Oregon), just 11 responded strongly enough to the disturbance index to be included in a final index. In western Oregon, 32 metrics responded in the first screening, requiring a further winnowing process to select the best and final metrics. Although individual metrics differed between the two regions, the metric categories were similar overall, except for species richness and abundance metrics (Table 2).

A final difference in index development between the two regions was in the application of species expectations for a minimally disturbed reference condition. The development of individual metric scoring resulting in a rigorous index of biotic integrity depends on comparison and calibration with a reference condition to determine species expectations under minimally disturbed conditions (Davis and Simon 1995; Hughes 1995; Hughes and others 1986). For the western Oregon BII, species numbers (used as the denominator in species-proportional metrics) were augmented by including riparian woodland species either extirpated or in serious decline over the last 50 years in the Willamette Valley. This adjustment was applied so that the scores would more accurately reflect conditions in a highly developed region where minimally disturbed reference conditions were lacking (Bryce and others 2002; Hughes and others 1998). If not adjusted in this way, indices of biotic integrity developed from least disturbed (rather than minimally disturbed) reference sites result in lowered expectations for stream condition, because disturbed sites might not differ appreciably from the available reference sites. The present BII for the Blue Mountains was not adjusted because acceptable reference conditions were still available in wilderness areas there. Six minimally disturbed reference sites were sampled in the Blue Mountains (21% of the sample) as opposed to two least disturbed reference sites in the more developed Willamette Valley (15% of the sample).

Conclusions

It is both a positive and negative attribute of ecological indicators that they integrate the effects of natural and anthropogenic stressors in their environment. It is our challenge in developing an index of biotic integrity to disentangle the complex interrelationships of natural variability in the environment and human disturbance on the landscape while selecting attributes of the indicator assemblage that we expect to vary systematically across a regional gradient of disturbance. We can clarify biological response at several points in the process by the following:
  1. 1.

    Limiting the geographic extent of the study to an ecological region with a limited number of disturbances. As stated earlier, this study was part of a large probabilistic survey of the western United States (EMAP-West); but it had an intensified sampling grid that concentrated sites into a single mountainous ecological region. In the larger 11-state survey, the clarity of biological response could be reduced by high variability in scores across multiple, often unmeasured, disturbance types in aggregated reporting regions (mountains, aridlands, etc.). Examining a smaller number of streams at a higher resolution and across clearer stressor gradients offers an opportunity to better understand the interrelationships of land use, landscape and in-stream stressors, and biological response.

     
  2. 2.

    Sampling a sufficient number of reference and highly disturbed sites, even if it means adding hand-picked sites to a random sample. A “sufficient number” of sites means enough sites to capture the natural variability in a group of minimally disturbed sites and to represent the most intensive disturbances in the region. The added information from reference and highly disturbed sites clarifies the response by setting clear end points for the range of conditions.

     
  3. 3.

    Proposing and evaluating candidate metrics that represent the response of specialized or sensitive species and assemblages to the range of disturbances present in the study area. One can best achieve this objective by using response guilds (e.g., tolerant, intolerant, nest parasite) in addition to functional guilds (e.g., foliage gleaners, omnivores) and tailoring metrics to regional ecological conditions.

     

This study demonstrated that the BII was adaptable to a region with very different physical characteristics, land-use activities, and disturbance factors than the one in which it had been originally developed. In the forested Blue Mountains of eastern Oregon, the disturbances were pervasive, but not as severe as those in the agricultural and urbanizing Willamette Valley. Further verification of the index is warranted, either with another group of sites in the same ecoregion or in a different western mountainous ecoregion.

The BII has been developed as an indicator of riparian conditions to relate it to ongoing efforts at in-stream bioassessments. However, from a different perspective, the index also indicates the condition of the bird assemblage itself. Birds were not the assessment end point in this study, but the individual metrics comprising the index illuminate the dynamics of bird assemblage sensitivity to a range of regional stressors. The BII can help to guide management of the riparian zones of the Blue Mountains by indicating the condition of stream riparian zones relative to one another, suggesting priorities for restoration or mitigation of the effects of particular disturbances, and raising questions for future research.

Acknowledgments

I am particularly grateful to Paul Ringold for supporting the project, to Ted Ernst and Tom Kincaid for patient statistical counseling, to Randy Hjort for timely information management output, to Colleen Burch Johnson and Donovan Reves for producing the GIS coverages and metrics, and to Suzanne Pierson for print-worthy illustrations. Also, thanks go to project research partners Mike Bollman and Teresa Magee for efficient field planning and vegetation data coordination. Three reviewers, Joshua Lawler, Bob Hughes, and Brenda McComb, and two anonymous reviewers each provided helpful comments to improve the manuscript.

The research in this article was funded by the US Environmental Protection Agency. This document was prepared at the EPA National Health and Environmental Effects Research Laboratory (NHEERL), Western Ecology Division (WED), in Corvallis, Oregon, through contract 68-D01-005 to Dynamac Corporation. It has been subjected to the agency’s peer and administrative review and approved for publication.

Copyright information

© Springer Science+Business Media, Inc. 2006