Status of Breeding and Wintering Birds
in Petrified Forest National Park, Arizona
Alison Banks, Charles van Riper III, Steven S. Rosenstock(1)
     10 November 2002
     USGS Southwest Biological Science Center
     Colorado Plateau Field Station
     PO Box 5614
     Northern Arizona University, Flagstaff, AZ 86011

            1 Arizona Game and Fish Dept., 2221 W. Greenway Rd., Phoenix, AZ 85023

INTRODUCTION
Petrified Forest National Park provides several major habitat types for birds, including grassland, shrubland, pinyon-juniper and juniper woodlands, and cottonwood-dominated riparian woodland. Because of its status as a park, Petrified Forest represents a relatively stable area where management actions can be monitored and avian populations can be tracked over time. The park also represents one of the few ungrazed short-grass prairie locations in the Southwest that can provide an excellent baseline reference point from which other studies in modified habitat can be compared.
In the western US, grasslands generally host few species of birds, relative to other habitat types (Wiens 1974), but grasslands have experienced consistent and widespread avian population declines in the latter half of the past century (Peterjohn and Sauer 1999). Threats to grassland habitats include but are not limited to land conversion, degradation by livestock, lack of fire, and overgrazing (West 1984, Arnold et al. 1964, Gibson and Hurlbert 1987). The effects of such changes on grassland avifauna are still incompletely understood. For example, grazing by livestock affects some nesting bird species positively and others negatively (Bock et al. 1984, Bock and Bock 1988, Bock et al.1992). In the Southwest there are few published studies of grassland birds, and even fewer from Northern Arizona (e.g., Monson 1941, Beatty 1978, LaRue 1994, Rosenstock and van Riper 2001), and no published studies on the avifauna of Petrified Forest National Park.

Riparian woodlands in the Southwest are extremely important to breeding birds; comprising less than one percent of the land area, riparian habitat supports>75% of the breeding bird fauna (Johnson et al. 1977, Szaro 1980). Destruction of riparian habitat has been identified as the leading cause of avian declines in the West (DeSante and George 1994). Because riparian habitats are linear and limited in extent, they are susceptible to influence by outside land use practices, such as agriculture, grazing, or development (Tewksbury et al. 1998). In the Southwest, an additional conservation concern is invasion of riparian areas by non native tamarisk (i.e., salt-cedar). Tamarisk displaces native vegetation and creates expanses of monotypic vegetation, resulting in little species heterogeneity within habitats. Thus, although the riparian woodland of Petrified Forest is limited in extent, it is likely to be disproportionately valuable to birds within the park.
Petrified Forest National Park provides important avian habitats that, outside the park, are threatened by various human land-use practices. As a National Park, Petrified Forest provides an ideal location for a control site to compare the results over time of management actions taken outside the park and the impacts of those actions on avian communities. This report provides baseline data regarding breeding and wintering birds in Petrified Forest NP and recommends a monitoring plan for bird populations within the park.

METHODS
We identified eight major vegetation habitat types within the park: 1) badland, 2) cliff edge (short shrub), 3) dune, 4) grassland, 5) juniper woodland, 6) shrubland, 7) dry wash, and 8) riparian woodland. Within the eight vegetation types we established 10 transects, sampling these habitats in proportion to their overall abundance within the park. Transects varied in length from 1750-3150 m, or 7-13 points located >or= 250 m apart (Table 1). We over-sampled riparian vegetation because of its significance for birds, sampling along washes with (“riparian woodland”) and without (“dry wash”) cottonwood over story. Badlands were under sampled because our initial observations revealed an apparent lack of avifauna utilizing this expansive habitat within the park. We sampled juniper-cliffrose woodland near Pintado Point but did not sample the juniper-pinyon pine habitat represented on Chinde Mesa in the northern part of the park, because of the latter site’s remoteness.
Locations of all avian survey points were recorded in UTM coordinates with a PLGR Global Position System unit and transects recorded (Appendix I). Points were marked in the field with rebar stakes tagged with transect name and point number. To facilitate the location of survey points for future monitoring efforts, we wrote instructions for finding points along transects, using compass bearings and landmarks (Appendix II).

We established vegetation sampling plots along all transects, to describe in more detail habitat types sampled and to verify habitat delineations. Habitat measurements follow a protocol that may be replicated in the future (Appendix III; data collection sheet Appendix IV). Two to four vegetation sampling plots were measured per bird survey point (Fig. 1; Appendix IV); two plots were measured if the vegetation appeared homogenous, four if we observed substantial variation in the vegetation (“plots” column, Table 2). We quantified percentages of different ground cover types, including bare ground, rock, grass, forbs, litter/brush, short-shrubs (less than 0.5 m) and shrubs. A shrub count was used to estimate shrub density, except at Puerco River, where shrub stems were counted. We measured shrub heights, sampling all shrubs in our plots that were at least 0.5 m in height, because we felt that these presented potential nest sites for bird species. At the Puerco River, we counted all trees within 100 m of our census points, according to four diameter at breast height (DBH) size classes (see Appendix IV). We then measured percent tree canopy cover, height, and species composition. At Pintado Point all shrubs (e.g., juniper, cliffrose), greater than or equal to 2 m in height, were counted because at this location they function structurally as small trees.
Bird surveys
Historical records. We entered all park historical records into a Microsoft Excel spreadsheet database, including park museum specimen records, Natural History Observation cards, monthly bird observation notes, and correspondence regarding birds and bird specimens. In addition, we contacted 40 natural history museums throughout the country to locate bird specimens historically collected within Petrified Forest. We searched library databases for publications about birds in Petrified Forest, using the keywords: “Apache County, Arizona, aves, birds, national park, Navajo County, and Petrified Forest,” in the following databases: Biological and Agricultural Index (1983-1999), Life Sciences (1982-1999), Trends in Ecology and Evolution (1939-1999), and Zoological Record (1993-1999).

Bird checklist. The current park bird species checklist (1984) was revised to conform with standards set by other national parks and by The National Park Service Inventory and Monitoring NP Species Data Dictionary. The checklist was improved by incorporating several standard list features, such as distinguishing winter residents from summer residents and noting birds that breed in the park. We used museum specimen records, historical records of bird observations, the 1963 bird checklist (which indicated seasonal occurrence and nesting for some species), and observations from the 1998-1999 surveys to assign seasonal occurrence and relative abundance of birds. Common names of species follow the American Ornithologists' Union (AOU) Checklist of North American Birds (AOU, 1998).
Field technique Comparisons. We compared two survey methods: strip-transects and variable circular plot point-counts. Strip-transects were conducted by walking at a constant pace along the transect, recording all birds seen or heard from the transect (Emlen 1971). Point-counts were conducted by walking to each point on a transect, and then after a one minute stationary period, counting all birds seen or heard from that point for five minutes (Reynolds et al. 1980). A sample of the data sheet that we used is provided in Appendix V.
A single observer conducted 3 sets of strip-transect surveys and 1 set of point-counts along each transect during the 1998 breeding season (2 June - 1 July; n= 40 surveys). In 1999, 3 sets of each survey type were completed by two observers (31 May - 26 June; n=60); each observer counted every transect. The order in which transects were counted and by whom was randomly assigned. Three sets of strip-transect surveys were conducted in winter 1998-1999 (2 December - 5 February; n=30).
For all surveys, distances of birds from the transect were measured with a laser range-finder. Breeding behaviors (e.g. carrying nesting material or food, feeding young) or nests found were noted during surveys. Because birds flying over a transect (flyovers) might not be associated with that specific habitat, these birds were distinguished from others that perched within sight of the observer. Flyovers and birds seen at greater than 100 m from survey transects or points were all entered into the bird check list information, but were excluded from relative abundance analyses.

We did not survey if it rained, or if wind speeds exceeded an average of 8 mph (5 km) or gusted at 10 mph (6 km). Wind speed and temperature (shaded) were taken with a digital anemometer. Wind speed was recorded with the anemometer at arm’s length in the direction of maximal wind; the average and maximal wind speeds were recorded after one to two minutes. We found that a digital anemometer was superior to the old-fashioned ball-in-the-tube model, which performed erratically at speeds around 6 km/hr.
To augment our strip-transect and point-count surveys in dense riparian vegetation along the Puerco River, we captured birds using mist-nets on five days in June 1998, using four to six nets per day, for a total of 137 mist-net hours. We banded birds with USFWS metal bands, took standard morphological measurements, and examined birds for evidence of breeding activity (cloacal protuberance or brood patch).
Analyses.
To determine if the eight habitats that we designated within the park were distinct, we used a multivariate ordination procedure on avian relative abundance data. Ordination reduces multiple, correlated variables (e.g., percent grass cover and percent bare ground) to a few axes, allowing one to view differences and similarities among sampling points. We used percent ground cover data and number of trees to ordinate survey points by vegetative characteristics using non-metric multi-dimensional scaling, AMDS@; program DECODA (Minchin 1993). Following the ordination, we tested whether groups of points within different habitat types were distinct with analysis of similarities (Clarke 1993; program ANOSIM, Minchin 1993). ANOSIM R values vary from 0 (no difference) to 1 (maximal difference), and are reported as significant at an α<.05 after sequential Bonferroni correction for multiple tests (Rice 1989). We excluded the cliff-side transect (NINE) from vegetation analyses because this transect characterized a topographic feature and not a vegetation type.

We tested for differences among years and between observers, with paired tests for each transect. We were able to combine data from both years, because neither numbers of birds nor numbers of species that we counted differed between observer or year (t=-1.584, 0.243, p=0.148, 0.813, respectively). The two observers did not differ in their average walking rates during strip transect surveys (1.67 km/hr vs. 1.79 km/hr; range= 1.07 - 2.27; ANOVA F=1.61, p =0.21). Species lists differed little between observers (30 vs. 29 species), with 14 of the 15 most abundant species shared. The numbers of birds counted did not differ between observers (t=-0.968, p=0.340), nor did estimated densities of Horned Larks from point-counts (D=.411, s.e.=.065 vs. D=.403, s.e.= .093). The only difference was that one observer counted more Horned Larks per strip-transect survey, while the other observer counted more species per point during point-count surveys (t=-2.515, p=0.033). To account for these variances, we included observer as a covariate for statistical tests of species richness or bird abundance among transects.
Paired t-tests were used to compare effectiveness of the two survey methods: each point count survey was paired with a strip transect survey conducted during the same week (1998) or on the same day (1999). We compared the number of birds and species counted per hour of survey effort.

We used Analysis of Variance (ANOVA) to determine if numbers of birds and species varied among habitat types; post-hoc comparisons among habitats were t-tests with Bonferroni corrections for multiple testing. Pearson correlation was utilized to determine if species abundances were correlated with specific habitat characteristics. To determine if bird community composition differed among habitat types, we ordinated all bird detections from both years of strip-transect surveys, using MDS. Then we used ANOSIM to determine if avian community composition differed among habitat types.