Effects of Residential Development and Landscape Composition on the Breeding Birds of Placer County’s Foothill Oak Woodlands

This study examines the effect of rural residential development and landscape composition on breeding birds in Placer County’s foothill oak woodlands. Point count survey data were used to construct generalized linear models for individual species’ abundance or probability of occurrence, based on two sets of variables: GIS-derived landscape characteristics, including development density, oak woodland proportion, and habitat diversity; and field-collected local habitat parameters. We found that many species examined were sensitive to either development density or landscape composition at some distance between 250 and 4,000 m. Of the 48 breeding species common enough to analyze statistically, the occurrence of 24 species was significantly associated with landscape characteristics. Species shown to be associated with development density and/or urban edge proximity included the lark sparrow (-), Rufouscrowned sparrow (-), western meadowlark (-), black Phoebe (+), house finch (+) and western scrub-jay (+). Several other species were not development-sensitive but were positively associated with the proportion of oak woodland found in the surrounding landscape. For a subset of locations, some species also exhibited responses to local habitat variables, suggesting that further investigation of the importance of landscape vs. local factors is warranted. The diversity of responses observed across a range of species requires the recommendation of a multifaceted conservation strategy for oak woodland birds and their habitat. Introduction Placer County’s human population is the fastest growing in California, with a growth rate of 3.5 percent in 2000 (California Department of Finance 2001). Much of this population growth is occurring in the county's foothill oak woodlands, 93 percent of which are privately owned and over 50 percent of which (30,000+ acres) have rural residential or urban land-use designations (Placer Legacy 2000). Concern about this rapid growth and the loss of open space and rural character led to the development of the Placer Legacy Open Space and Agricultural Conservation Program, which seeks to balance growth with the conservation of open space and wildlife resources. Because foothill oak woodlands are rapidly urbanizing and poorly protected, though treasured for their scenic and wildlife values, much of the program's early emphasis has focused on acquiring one or more large parcels to preserve oak woodlands. In addition, the County is interested in understanding how 1 An abbreviated version of this paper was presented at the Fifth Symposium on Oak Woodlands: Oaks in California’s Changing Landscape, October 22-25, 2001, San Diego, California. 2 GIS Specialist, Point Reyes Bird Observatory, 4990 Shoreline Highway, Stinson Beach, CA 94970 (email: [email protected]) 3 Consulting Wildlife Ecologist, 8200 Turner Dr., Granite Bay, CA 95746 (e-mail: [email protected]) USDA Forest Service Gen. Tech. Rep. PSW-GTR-184. 2002. 341 Placer County Breeding Birds—Stralberg and Williams the rural residential landscape can be better managed to preserve wildlife, sensitive resources and water quality. This project was initiated as a part of the Placer Legacy Program as an effort to assess the effects of rural residential development and habitat fragmentation on breeding birds as indicators for oak woodland habitat. Habitat suitability for wildlife is an important consideration in reserve design, and local habitat relationships are relatively well studied in California's foothill oak woodlands (Avery and Van Riper 1990, Block 1989, Block and Morrison 1990, Block and others 1994, Tietje and others 1997, Verner and others 1997, Wilson and others 1991). Recently, much attention has also been focused on the potential effects of rural residential development, vineyard expansion and other human modifications to oak woodland landscapes. In Sonoma County, Merenlender and others (1998) found that the level of development of a parcel influences bird community composition and that neotropical migrants in particular demonstrate reduced abundances in suburban areas and, to a lesser extent, rural residential areas. Several recent studies of birds in other California habitats have suggested that characteristics of the surrounding landscape may influence habitat quality for many species and, in some cases, may even be better predictors of species occurrence than local habitat structure (e.g., Bolger and others 1997, Stralberg 1999). Currently, a need remains for a better understanding of landscape-scale processes that affect habitat suitability of oak woodlands beyond local habitat structure (Bell 1997, Garrison and Davis 1997, Thomas 1997). Such knowledge may be particularly valuable when candidates for reserves are structurally similar, as they are in foothill oak woodlands of Placer County. We initiated this study to test the hypothesis that some birds will be affected by landscape-scale patterns of development irrespective of local habitat. One primary objective is to provide specific recommendations to the County of Placer regarding priorities for management, conservation and acquisition of foothill oak woodlands, as well as future zoning decisions and general plan revisions. We also hope to gain a better general understanding of the features of habitat and landscape patterns that determine species occupancy, in order to inform land-use planning and conservation, as well as wildlife management on private and public land. This may be accomplished in part through future revisions to the California Partners in Flight (CPIF) Oak Woodland Bird Conservation Plan (Zack and others 2000), an interagency effort to promote the conservation of migratory birds and their habitats throughout the Americas. Methods Study Area Our study area in western Placer County ranged in elevation from 70 to 480 meters, and encompassed an area of approximately 550 km (fig. 1). Dominant tree species included blue oak (Quercus douglasii), which occurs primarily on drier sites (especially ridges that were historically difficult to irrigate), and interior live oak (Quercus wislizenii), which tends to occur in more mesic areas such as drainage basins and north-facing slopes. A complex human history has altered the distribution and structure of many of these oak woodlands, including their understory structure and composition. Interspersed with oak woodlands is a combination of orchards, cropland, dry pasture rangeland, irrigated pasture, rural residential development (“ranchettes”), and urban and suburban development. USDA Forest Service Gen. Tech. Rep. PSW-GTR-184. 2002. 342 Placer County Breeding Birds—Stralberg and Williams Figure 1—Study area and point count locations. In Placer County, large intact blocks of oak woodland are rare, and habitat patches are not easily defined or necessarily isolated from other habitat patches. We therefore chose a point-based approach for sampling habitat, rather than surveying entire habitat patches. Our intent was to sample bird species at random throughout a representative cross-section of the County’s development spectrum: from urban park to rural residential to largely undeveloped rangeland. Study Design and Point Selection A total of 75 points was surveyed. We began by selecting a stratified random sample of 80 potential survey points by generating a 500-m sampling grid of points superimposed over the county's oak woodlands within the study area. Suitable points (>2,600) were defined as meeting one of the following CWHR habitat classifications (Mayer and Laudenslayer 1988) according to GIS vegetation data (Forest Service 2000): blue oak woodland or blue oak foothill pine, but also montane hardwood, urban, annual grassland, valley-foothill riparian or agricultural cropland if our familiarity with the area suggested that the amount of oak woodland at or adjacent to the point was underestimated. Selected points were stratified by general plan land-use categories (Placer County General Plan 1994): (1) Low Density Residential (LDR) and Rural Residential (RR) 1-2.5 acres; (2) RR 2.5-5 and RR 5-10; (3) RR 10, Agricultural (Ag) 10, and Ag 20; and (4) Ag 40, Ag 80, and Open Space (OS). Actual land use varied considerably from the General Plan designation because many USDA Forest Service Gen. Tech. Rep. PSW-GTR-184. 2002. 343 Placer County Breeding Birds—Stralberg and Williams parcels were not yet “built out,” but this method allowed representation of a range of land uses and parcel sizes, as well as geographic area, among sample points. To improve our sampling effectiveness, we developed a random clustering technique that began by randomly choosing one of nine 7.5 ft topographic map quadrangles followed by the random selection of a legal section (1 mi or 2.59 km) as a starting point. We then randomly chose points from the 500-m sampling grid within immediately adjoining sections, expanding the radius by 1 section as each layer of sections was exhausted. Constraints were that no more than 4 points could fall within one section and no more than 2 points of a given land-use category could occur in any one section. This process was repeated to produce two random sets of 40 semi-clustered points, each at least 500 m from the nearest sampling point. Because the first 80 points included no parcels zoned for 80 acres and very few parcels smaller than 5 acres, we randomly selected an additional 22 points within these parcel size ranges (10 and 12 points, respectively). Each point was groundtruthed to meet the following minimum criteria: (1) oak woodland as the dominant habitat type, with at least two oaks within the 50-m radius and a canopy cover ≥ 10 percent; (2) no house or other large building within a 50-m radius; (3) not within 500 m of a major highway; (4) not excessively time-consuming to access; and (5) <5 percent paved two-lane public road within a 50-m radius (private single-lane dirt roads were fairly common and practically impossible to exclude). We did not exclude any points based on other habitat characteristics such as presence of water, understory composition, slope or aspect. Of these original 122 locations, we were able to obtain access to 57 survey sites. The other 45 were either unsuitable (n = 16), or we were unable to contact the landowners (n = 15), or we were denied access (n = 8), or negotiating access was too time-consuming (n = 6). We repeated the selection process to find 6 stratified-random substitute points, identified 6 points semi-randomly (randomly chosen within a nonrandom area that was convenient to access), and added 6 points at widely separated urban parks for which we knew we could get immediate access. Within each park, the specific sample point was randomly determined in the field. The exact location of each site was later recorded with a Global Positioning System (to within approximately 5-15 m). Bird Counts Six-minute, unlimited-distance point counts (Ralph and others 1993) were done twice at each of the 75 sites, 7-28 days apart (mean = 16 days, SD = 4.4). Detections were recorded as within or beyond a 50-m radius. We assumed that 2 counts would be necessary to avoid problems with seasonal variation in vocalization frequency and hence detection probability. Point counts were done in the morning beginning at official sunrise to 4 hours after sunrise, during appropriate weather conditions (Verner 1985) between 17 May and 14 June 2001 by a single expert observer (B. Williams). Habitat Parameters We measured or estimated parameters describing the local habitat and physical conditions at 32 of the 75 point count locations (appendix 1). The intent was to statistically control for habitat-specific variation to focus on landscape-scale effects. USDA Forest Service Gen. Tech. Rep. PSW-GTR-184. 2002. 344 Placer County Breeding Birds—Stralberg and Williams Landscape Parameters Obtained from GIS Data We calculated several urbanization and landscape composition metrics (appendix 2) for each point-count location using ArcView 3.2a and the Spatial Analyst Extension (ESRI 2000). The County's parcel base map and associated Assessor's database were used to determine the parcel size and overall property size of each site surveyed. In addition, the number of structures within various buffer distances (250 m, 500 m, 1000 m, 2000 m and 4000 m) of each point-count location was estimated based on the development status of each parcel centroid. Because structure locations were estimated based on parcel centroids, there is some uncertainty in the number of structures counted within point count radii. Furthermore, while the bird and vegetation data were collected in 2000, the parcel base map represents July 1998 parcels and the Assessor’s database contains 1999 ownership information (matching years were not available at the time of analysis). Thus, parcels that were subdivided and developed after 1998 are not correctly represented in this database. The parcel data are nevertheless a major improvement over any other available urbanization measure (i.e., Forest Service vegetation data), particularly in the rural residential zone, where the built footprint is not easily discernable even from aerial photos. As an index of housing density, we feel this measure is the best available, short of ground-based inventories. Geographic information system (GIS) vegetation data (Forest Service 2000, 2.5 acre minimum mapping unit) were then used to coarsely determine the landcover composition of each point count location within circles of increasing radius: 250 m, 500 m, 1,000 m, 1,500 m, 2,000 m and 4,000 m. A proportion was obtained for each cover category within each radius. For analysis, the following CWHR categories were combined to calculate oak woodland coverage: blue oak woodland, blue oakfoothill pine, valley oak woodland, montane hardwood, and montane hardwoodconifer. Finer distinctions between oak woodland categories were not made due to the presumed low accuracy of the vegetation layer beyond cover class. The other category used for analysis was annual grassland, some of which was actually open oak savanna. The urban classification was not used in analysis due to the coarseness of this vegetation layer with respect to rural residential development patterns, particularly in comparison with the more accurate parcel base map and Assessor’s database (fig. 2). To evaluate the influence of landscape-level habitat diversity, we also calculated a Shannon-Wiener diversity index (Krebs 1989) for each of the above-listed radii (H = Σi pi ln(pi), where pi = area of i habitat type). For this metric, each CWHR category was treated separately to reflect diversity among oak woodland as well as other habitat types. Finally, for each point we estimated the distance to the nearest structure or urban edge, using a combination of field notes, aerial photos and the GIS parcel basemap with associated Assessor’s ownership database (in that order, depending on availability). We also used 1:100,000 scale road and hydrography GIS datasets (Teale Data Center 1997, 1999) to calculate the distance from a point count to the nearest stream and nearest paved road. USDA Forest Service Gen. Tech. Rep. PSW-GTR-184. 2002. 345 Placer County Breeding Birds—Stralberg and Williams