Appendix C. Results for input models, and methods and results of corvid and exotic presence risk model evaluations.
Input models results
Synanthropic predators - The corvid presence risk model (Fig. C1A) suggested that corvid species were ubiquitously distributed across the western United States. Areas of high corvid presence risk were associated with urban and agricultural areas. About 76% of the western United States overlapped with the medium (42.9%, 1,316,149 km2) and low presence risk class (32.6%, 1,001,328 km2). The high presence risk class (17.1%, 524,153 km2) was more than twice the size of the negligible presence risk class (7.5%, 229,182 km2).
For the cat model (Fig. C1B), the zero presence risk class covered 98.0% (3,008,749 km2) of the western United States and 97.1% (2,983,790 km2) for the dog model (Fig. C1C). Overall, 49,525 km2 and 62,868 km2 of the western United States were in the highest probability class in the cat and dog presence risk model, respectively.
|
| FIG. C1. Spatial extent of synanthropic predator input models: (A) corvid presence risk model based on anthropogenic features (populated areas, campgrounds, rest stops, agricultural land, landfills, and density of liner features) utilized by American Crows (Corvus brachyrhynchos), Common Ravens (C. corax), and Black-billed Magpies (Pica hudsonia) (areas with negligible corvid presence risk represent corvid population levels in the absence of anthropogenic features); and domestic mammalian predator presence risk models for (B) house cats, based on populated areas (>1 person/ha); and (C) domestic dogs, based on populated areas (>1 person/ha) and campgrounds. Input models were classified into risk classes for illustrative purposes. |
Habitat change/loss - The majority (71%) of the exotic plant invasion risk model (Fig. C2A) was dominated by the negligible (35.8%; 1,099,488 km2) and low invasion risk classes (35.1%; 1,078,750 km2). For the high-intensity risk classes, the high-risk class covered twice (18.2%; 558,686 km2) as much land as the medium-risk class (10.9%; 333,889 km2).The spatial distribution of anthropogenic fragmentation of wildlands, as measured by percent area containing at least one of eight anthropogenic features within a 2975 km2 area, varied highly across the western United States (Fig. C2B). Overall, <1% of this region contained no anthropogenic features where as 42.8% (1,313,958 km2) had between >0 to 10%, 28.2% (865,560 km2) had between >10 to 20%, and 11.4% (349,127 km2) had between >20 to 30% of area occupied by at least one anthropogenic feature. Nearly 10,000 km2, or 0.3% of the western United States, had >90% of area containing at least one of the eight anthropogenic features. Energy extraction was confined to areas in Montana, Wyoming, Colorado, Utah, and New Mexico (Fig. C2C). The majority of the western United States, 94% (2,875,009 km2), was not exposed to oil and gas well development. However, about 6%, or 189,997 km2, of the western United States had <5 wells/km2 and 0.2%, or 5,807 km2, had >5 wells/km2. The highest density, 13.4 wells/km2, was located in the southeastern corner of New Mexico.Human-caused fire densities seemed highest near human populated areas and along road networks (Fig. C2D). The vast majority of the western United States (94.4%, 2,899,727 km2) was in the zero ignition point density class. However, about 5.5%, or 170,660 km2, of the western United States had <10 point ignitions/km2, and 0.01%, or 427 km2, had >10 point ignitions/km2. Of the latter area, almost half of it was in the state of Arizona (48.2%), followed by California (25.6%), Montana (11.4%), New Mexico (8.5%); all other states overlapped <3% of this area, with the least observed in Utah (0.1%). The highest density of 247.4 point ignitions/km2 was located in east-central Arizona.
|
| FIG. C2. Spatial extent of habitat change/loss input models: (A) exotic plant invasion risk model based on road-influenced exotic plant dispersal; (B) anthropogenic fragmentation model based on percentage cells containing an anthropogenic feature (agricultural land, populated areas, power lines, railroads, and roads) within a 2975-km2 moving window (home range of female grizzly bear [Ursus arctos] and migratory Greater Sage-grouse [Centrocercus urophasianus]); (C) energy extraction model based oil and gas well densities within 1 km2; and (D) human-caused fire model based on ignition densities (1 km2) over the period of 1986 to 2001. Input models were classified into risk classes for illustrative purposes. |
Evaluation of input models
Corvid presence risk model - We tested the predictions of this input model using Breeding Bird Survey (BBS) data (Sauer et al. 2002) for three corvid species: American Crow (Corvus brachyrhynchos), Common Raven (C. corax), and Black-billed Magpie (Pica hudsonia). The corvid presence risk model was classified into four equal classes: high, medium, low, and negligible. Only routes that were completely within one of the four risk classes were included in the analyses. To avoid biases associated with BBS population trend analyses(Geissler and Sauer 1990; Sauer et al. 1994; Kendall et al. 1996), we developed a detection index based on the number of years at least one of the three species was detected (percent years) between 1991 and 2001 and included only those routes that were sampled at least seven times within this period. We performed ANOVA with Tukey posthoc analyses to test for difference in percent detections, square-root arcsine transformed data (Zar 1984), among corvid presence risk classes.The probability of corvid presence risk differed significantly among the risk classes (ANOVA F3,220 = 5.36, p = 0.001; Fig. C3). The high presence risk class was significantly higher compared to the low and negligible risk classes, whereas the medium risk class was significantly higher compared the negligible risk class.Exotic plant invasion risk model - We tested the predictions of the exotic plant invasion risk model using a database of plot level vegetation data covering most of the arid western United States. These datasets were provided by various mapping projects, including components of the Southwest Regional GAP project (Utah, Arizona, Colorado, and Nevada; USGS National Gap Analysis Program. 2004), U.S. Geological Survey Plots, Regional USFS Ecodata Plots, Oregon Natural Heritage Program Data, and Washington Natural Heritage Program Data. A pivot table analysis was used to classify each vegetation-sampling plot based on the presence (1) or absence (0) of 18 introduced graminoid species: Agropyron cristatum, Agrostis gigantea, Avena fatua, Bromus hordeaceus spp., Bromus inermis, Bromus japonicus, Bromus mollis, Bromus rigidus, Bromus tectorum, Dactylis glomerata, Eremopyrum triticeum, Hordeum murinum, Phleum pratense, Poa bulbosa, Poa pratensis, Taeniatherum caput-medusae, Thinopyrum intermedium, and Triticum aestivum.. This table was imported into ARC/GIS and subsequently converted to an exotic species presence/absence grid layer. We calculated the number of locations with each unique combination of the exotic plant invasion risk and the exotic species presence/absence grid and calculated the ratio of presence/absence of exotics for each risk class within each exotic plant invasion risk class. We used a contingency table analysis to test whether the distribution of presence/absence of exotic grasses is independent of exotic plant invasion risk class.
|
| FIG. C3. Test of the corvid presence risk model using Breeding Bird Survey (BBS) data. Shown are mean (± SE) percent years for which at least one corvid species (American Crow [Corvus brachyrhynchos], Common Raven [C. corax], and or Black-billed Magpie [Pica hudsonia]) was detected on a BBS route (1991–2001). As predicted, the high corvid presence risk class differed significantly from the low and negligible risk classes (bar graphs not sharing letter differ significantly [P < 0.05] from each other). Sample sizes represent number of routes completely contained within each of the four-corvid presence risk classes. |
The proportions of plots containing exotic grasses differed significantly among the four exotic plant invasion risk classes (G-test = 1487.3, df = 3, P < 0.0005) (Fig. C4). Exotic plant invasion risk was lowest in the negligible risk class (negligible-low: G-test = 572.3, df = 1, P < 0.0005), higher in the low (low-medium: G-test = 97.7, df = 1, P < 0.0005), but highest in the medium and high-risk classes (medium-high: G-test = 0.18, df = 1, P = 0.67). We found no significant difference between the medium and high-risk classes. This may be due to the stratified sampling approach of mapping project data collection, focusing on plant communities which are more difficult to classify from satellite images compared to highly disturbed areas, such urban or agricultural areas, which are readily classified.
|
| FIG. C4. Test of the exotic plant invasion risk model using 49,278 field data points (shown are sample sizes and percent of plots containing at least one exotic grass species) distributed across the shrubland ecosystem in the Intermountain West (Arizona, Colorado, Idaho, Montana, Nevada, Oregon, Utah, and Washington). As predicted, presence of exotic grasses differed between negligible and low, low and medium, but not medium and high exotic plant invasion risk (bar graphs not sharing letter are significantly [P < 0.05] different from each other). Depending on the geographical range, one or more of the following species were included in the analysis: Agropyron cristatum, Agrostis gigantea, Avena fatua, Bromus hordeaceus spp. hordeaceus, Bromus inermis, Bromus japonicus, Bromus mollis, Bromus rigidus, Bromus tectorum, Dactylis glomerata, Eremopyrum triticeum, Hordeum murinum, Phleum pratense, Poa bulbosa. |
LITERATURE CITED
Geissler, P. H., and J. R. Sauer. 1990. Topics in route-regression analysis. Pages 54–57 in J. R. Sauer and S. Droege, editors. Survey designs and spatial methods for the estimation of avian population trends. U.S. Fish and Wildlife Service, Biological Report 90 (1).
Kendall, W. L., B. C. Peterjohn, and J. R. Sauer. 1996. First-time observer effects in the North American Breeding Bird Survey. Auk 113:823–829.
Sauer, J. R., B. C. Peterjohn, and W. A. Link. 1994. Observer differences in the North American Breeding Bird Survey. Auk 111:50–62.
Sauer, J. R., J. E. Hines, and J. Fallon. 2002. The North American Breeding Bird Survey, results and analysis 1966–2001. Version 2002.1. U.S. Geological Survey Patuxent Wildlife Research Center, Laurel, Maryland, USA.
USGS National Gap Analysis Program. 2004. Southwest Regional Gap Analysis Project Field Sample Database. Version 1.1. RS/GIS Laboratory, College of Natural Resources, Utah State University, Logan, Utah, USA.
Zar, J. H. 1984. Biostatistical Analysis, 2nd edition. Prentice-Hall, Inc., Englewood Cliffs, New Jersey, USA.