FINAL REPORT
GEOGRAPHIC INFORMATION ANALYSIS OF NORTHERN GOSHAWK
(ACCIPITER GENTILIS) NEST SITES IN THE UINTA MOUNTAINS


INTRODUCTION
Since the Northern Goshawk has been designated as a Category 2 species by the U.S. Fish and Wildlife Service (U.S. Dept. Interior 1991) and as a sensitive species by the U.S. Forest Service, there has been considerable interest in goshawk nesting requirements, food habits, reproductive success, and territory size. Management recommendations have been proposed for the Northern Goshawk in the Southwestern U.S. (Reynolds et al. 1991). Before management strategies can be implemented, biologists must be able to locate goshawk nests. Survey protocals that are currently in use (Kennedy and Stahlecker 1993, and Joy et. al. 1994) are labor intensive and time consuming. The objective of this study was to produce models (Johansson et al. 1994) to accurately locate goshawk nest sites to reduce the time and cost of goshawk nest surveys by reducing the amount of area that must be surveyed. Models were constructed from the following environmental factors:Elevation
Slope
Aspect
Vegetation Type
Forest Canopy Density
Distance to Water
![]()

Figure 1. Distribution of the Northern Goshawk in the U.S.
A. g. atricapillus -- Breeding Range
A. g. laingi -- Breeding and probable wintering range
A. g. atricapillus -- Wintering Range
A. g. apache -- Breeding and probable wintering range
![]()
Figure 2. Northern Goshawk (Accipiter gentilis).
![]()
STUDY AREA: Ashley National Forest in eastern Utah.

Figure 3. Ashley National Forest scenic view.
The study area encompassed four 7.5' quads in Daggett and Uintah Counties, Utah.
LEIDY PEAK
ELK PARK
EAST PARK RESEVOIR
MOUNT LENA

METHODS
IMAGES AND COVERAGES USED FOR THE STUDY AREA:7.5' 30 Meter Digital Elevation Data (DEM)
Final GAP Vegetation Map at 30 Meters
Utah Landsat Mosaic at 30 Meter Resolution
1:100,000 Water Courses and Bodies
1:100,000 Roads
1:24,000 Quadrangle Index (ut24k-utm)
SOFTWARE:
ARC/INFO 7.0
ERDAS IMAGINE 8.2
MINITAB
CUSTOM ANSI-C PROGRAMS
COVERAGES USED TO BUILD THE MODELS:ELEVATION: Eleven Categories (Nine 100 m elevation bands, one < 2200 m band, and one > 3100 m band)
SLOPE: Eight Categories (Seven 5% slope bands and one > 35% slope band)
ASPECT: Eight Categories (compass directions)
DISTANCE TO WATER: Eight Categories (Seven 150 Meter bands of distance to water and one > 1050 m band)
GAP VEGETATION: Percent coverage of each vegetation class (aspen, lodgepole pine, ponderosa pine, etc.)
FOREST CANOPY DENSITY: Percent coverage for five tree species classes (from GAP) with three canopy densities for each (Low, Medium, and High density)
METHODS FOR BUILDING THE COVERAGES:
All Raster Coverages (IMAGINE) - were Georeferenced to UTM, Spheroid Clarke 1866, Zone 12, Datum Clarke 1866, with 30 Meter pixel sizes and units set to Meters.
NEST SITES - were digitized from locations marked on 7.5' quads. Tics were set to match the corners of the quads from the 1:24,000 Quadrangle Index.
ELEVATION - LATTICEDEM (in ARC/INFO) converted the DEM lattice into USGS format DEM. GRID (in ARC/INFO) and the RECLASS command were used to reclassify the elevation grid into the user defined categories (100 Meter elevation bands) in a REMAP_TABLE. The reclassified grid was then imported into IMAGINE as a raster image.
SLOPE - GRID (in ARC/INFO) and the SLOPE command created the Slope grid from the DEM grid. GRID (in ARC/INFO) and the RECLASS command were then used to reclassify the slope grid into the user defined categories (5% slope bands) in a REMAP_TABLE. The reclassified grid was then imported into IMAGINE as a raster image.
ASPECT - GRID (in ARC/INFO) and the ASPECT command created the Aspect grid from the DEM grid. GRID (in ARC/INFO) and the RECLASS command were then used to reclassify the aspect grid into the user defined categories (eight cardinal compass headings) in a REMAP_TABLE. The reclassified grid was then imported into IMAGINE as a raster image.
DISTANCE TO WATER - water was buffered by 35 pixels by a SEARCH command (IMAGINE - INTERPRETER - GIS ANALYSIS - SEARCH). The resulting buffered image was then RECODEd (IMAGINE - INTERPRETER - GIS ANALYSIS - RECODE) by setting zones 1 - 5 to 1, 6 - 10 to 2, ... , 31 - 35 to 7. and zone 36 (outside of buffer) to 8. This resulted in an image with eight Categories (Seven 150 Meter bands and one > 1050 M band).
FOREST CANOPY DENSITY - The 30 Meter Landsat Mosaic was reclassified (IMAGINE - CLASSIFIER - UNUPERVISED CLASSIFICATION) to 30 clusters, then reclassified again (IMAGINE - INTERPRETER - GIS ANALYSIS - NEIGHBORHOOD) using a 5 x 5 pixel window, and FUNCTION set to DIVERSITY. The resulting image was then reclassified into 3 categories (Low, Med, & High Diversity). The three category diversity image was then MATRIXed (IMAGINE - INTERPRETER - GIS ANALYSIS - MATRIX) with a reclassified GAP Vegetation image that included only forest types where goshawk nest sites had been located (PIPO, PICO, PSME, POTR, and PIPO/Shrub). This resulted in an image with five tree species classes each with three levels of diversity. Low reflectance diversity was then assumed to indicate High Density Canopy and High reflectance diversity was assumed to indicate Low Density Canopy.
METHODS FOR BUILDING THE MODELS:
Predictive Models - These types of models were constructed based upon the random selection of eleven of twenty-four nest sites. These models were then tested for their ability to locate the remaining thirteen nest sites.
Inclusive Models - These types of models were constructed based upon the use of all twenty-four nest sites to determine the total land area required to locate all nest sites.
AREAL PERCENTAGES - Percentages of each category were determined by the number of pixels in each category from the RASTER ATTRIBUTE TABLE.
DISTANCE TO WATER - Two VIEWERS were opened simultaneously in IMAGINE, with an image of the NEST SITES in one viewer and the WATER image in the other. The viewers were GEOLINKED, and the INQUIRE CURSER and the MEASUREMENT TOOL was used to measure the shortest distance from each nest site to water.
OBSERVED VALUES - A SUMMARY of nest location coverage and coverage of interest determined the values for nest sites (Appendix 1).
EXPECTED VALUES - An ANSI-C program was used to calculate expected values for each category based upon the number of selected and total nests and the percentage of the total area represented by that category.
PRIORITY RANKING FOR ORDER OF SEARCH - Categories were ranked according to the observed - expected ratios until (1) search order was completed for the predictive models, and (2) all nest sites were included.
NESTS LOCATED - The cumulative percent of all nest located by model (including selected nests) was added and the percentage of the nests that were predicted by the predictive models was computed.
AREA SEARCHED - The cumulative percent of the area searched was added for each ranked category.

Figure 4. Banding young Northern Goshawks in the nest.

TOTAL NEST SITES
Figure 5. Twenty-four known goshawk nest sites were located within the study area. Nest sites are shown here highlighted by 24-pixel diameter (420 acre) buffer zones.
SELECTED NEST SITES
Figure 6. Eleven of the twenty-four nest sites were randomly chosen for use in generating models to predict the location of goshawk nests. The models were tested for their ability to predict the thirteen remaining nests that were not selected.

RESULTS AND MODELS

ELEVATION
Figure 7. Elevation Coverage. Elevation (from the DEM) was divided into eleven 100 meter categories as follows: CATEGORY % AREA CATEGORY % AREA < 2200 m 1.37 2700-2800 m 10.51 2200-2300 m 3.58 2800-2900 m 13.15 2300-2400 m 6.13 2900-3000 m 13.37 2400-2500 m 10.41 3000-3100 m 8.69 2500-2600 m 14.02 > 3100 m 8.31 2600-2700 m 10.48
TABLE 1. PREDICTIVE MODEL BASED UPON ELEVATION ORDER OF CLASS PERCENT SELECT TOTAL CUM % CUM % SEARCH ELEVATION(m) AREA NESTS OBS-EXP NESTS NESTS AREA 1 2300-2400 6.13 2 1.326 5 20.83 6.13 2 2400-2500 10.41 4 2.855 6 45.83 16.54 3 2800-2900 13.15 2 0.554 4 62.50 29.69 4 2900-3000 13.37 2 0.529 3 75.00 43.06 5 2700-2800 10.51 1 -0.156 4 91.67 53.57

Figure 8. The area to be searched based upon the Predictive Elevation model (Table 1). This model predicted eleven of thirteen nests (84.6%) after searching 53.57% of the study area.
TABLE 2. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON ELEVATION MODEL ORDER OF CLASS PERCENT OBS CUM % CUM % SEARCH ELEVATION(m) AREA NESTS OBS-EXP NESTS AREA 1 2300-2400 6.13 5 3.52 20.83 6.13 2 2400-2500 10.41 6 3.50 45.83 16.54 3 2700-2800 10.51 4 1.48 62.50 27.05 4 2800-2900 13.15 4 0.84 79.17 40.20 5 2900-3000 13.37 3 -0.21 91.67 53.57 6 < 2200 1.37 0 -0.33 91.67 54.94 7 2200-2300 3.58 0 -0.86 91.67 58.52 8 2500-2600 14.02 2 -1.36 100.00 72.54

Figure 9. The area to be searched (72.54%) based upon the Inclusive Elevation model (Table 2) to locate 100% of the nests.

SLOPE
Figure 10. Slope Coverage. Slope from the DEM was divided into eight categories as follows:
CATEGORY % AREA CATEGORY % AREA
0- 5% 11.90 20-25% 9.37
5-10% 25.43 25-30% 7.12
10-15% 17.78 30-35% 5.48
15-20% 11.73 > 35% 11.19
TABLE 3. PREDICTIVE MODEL BASED UPON SLOPE
ORDER OF CLASS PRECENT SELECT TOTAL CUM % CUM %
SEARCH SLOPE(%) AREA NESTS OBS-EXP NESTS NESTS AREA
1 10-15 17.78 4 2.044 9 37.50 17.78
2 0- 5 11.90 2 0.691 5 58.33 29.68
3 30-35 5.48 1 0.397 1 62.50 35.16
4 25-30 7.12 1 0.217 2 70.83 42.28
5 5-10 25.43 3 0.203 5 91.67 67.71

Figure 11. The area to be searched based upon the Predictive Slope model (Table 3). This model predicted eleven of thirteen nests (84.6%) after searching 67.7% of the study area.
TABLE 4. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON SLOPE ORDER OF CLASS PRECENT OBS CUM % CUM % SEARCH SLOPE(%) AREA NESTS OBS-EXP NESTS AREA 1 10-15 17.78 9 4.733 37.50 17.78 2 0- 5 11.90 5 2.144 58.33 29.68 3 25-30 7.12 2 0.291 66.67 36.80 4 20-25 9.37 2 -0.249 75.00 46.17 5 30-35 5.48 1 -0.315 79.17 51.65 6 5-10 25.43 5 -1.103 100.00 77.08

Figure 12. The area to be searched (77.08%) based upon the Inclusive Slope model (Table 4) to locate 100% of the nests.
ASPECT

Figure 13. Aspect Coverage. Aspect from the DEM was divided into nine categories as follows:
CATEGORY % AREA CATEGORY % AREA
N 17.76 S 9.18
NE 15.57 SW 7.16
E 15.11 W 8.05
SE 11.71 NW 15.41
FLAT 0.05
TABLE 5. PREDICTIVE MODEL BASED UPON ASPECT
ORDER OF CLASS PRECENT SELECT TOTAL CUM % CUM %
SEARCH ASPECT AREA NESTS OBS-EXP NESTS NESTS AREA
1 SW 7.157 3 2.213 4 16.67 7.16
2 NW 15.406 2 0.305 4 33.33 22.56
3 NE 15.570 2 0.287 4 50.00 38.13
4 W 8.052 1 0.114 2 58.33 46.19
5 N 17.764 2 0.046 6 83.33 63.95
6 FLAT 0.503 0 -0.121 0 83.33 64.45
7 S 9.175 0 -0.202 2 91.67 73.63
8 SE 11.706 1 -0.288 2 100.00 85.33

Figure 14. The area to be searched based upon both Aspect models (Tables 5 & 6). Both models predicted all thirteen nests (100%) after searching 85.33% of the study area.
TABLE 6. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON ASPECT ORDER OF CLASS PRECENT OBS CUM % CUM % SEARCH ASPECT AREA NESTS OBS-EXP NESTS AREA 1 SW 7.157 4 2.282 16.67 7.16 2 N 17.764 6 1.737 41.67 24.92 3 NW 15.406 4 0.303 58.33 40.33 4 NE 15.570 4 0.263 75.00 55.89 5 W 8.052 2 0.068 83.33 63.95 6 FLAT 0.503 0 -0.121 83.33 64.45 7 S 9.175 2 -0.202 91.67 73.63 8 SE 11.706 2 -0.809 100.00 85.33


Figure 15. Northern Goshawks.

DISTANCE TO WATER

Figure 16. Distance To Water Coverage. Eight categories of Distance To Water were created as follows:
CATEGORY % AREA CATEGORY % AREA
0 - 150 18.37 600 - 750 10.13
150 - 300 15.87 750 - 900 7.96
300 - 450 13.76 900 -1050 6.62
450 - 600 11.85 > 1050 15.44
TABLE 7. PREDICTIVE MODEL BASED UPON DISTANCE TO WATER
ORDER OF CLASS PERCENT SELECT TOTAL CUM % CUM %
SEARCH DIST TO WATER(m) AREA NESTS OBS-EXP NESTS NESTS AREA
1 0 - 150 18.37 6 3.979 10 41.67 18.37
2 150 - 300 15.87 2 0.254 3 54.17 34.24
3 900 -1050 6.62 1 0.212 1 58.33 40.86
4 750 - 900 7.96 1 0.124 4 75.00 48.82
5 450 - 600 11.85 1 -0.304 2 83.33 60.67
6 300 - 450 13.76 1 -0.514 4 100.00 74.43

Figure 17. The area to be searched based upon both Distance To Water models (Tables 7 & 8). Both models predicted all thirteen nests (100%) after searching 74.43% of the study area.
TABLE 8. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON DISTANCE TO WATER ORDER OF CLASS PERCENT OBS CUM % CUM % SEARCH DIST TO WATER(m) AREA NESTS OBS-EXP NESTS AREA 1 0 - 150 18.37 10 5.59 41.67 18.37 2 750 - 900 7.96 4 2.09 58.33 26.33 3 300 - 450 13.76 4 0.70 75.00 40.09 4 900 -1050 6.62 1 -0.59 79.17 46.71 5 150 - 300 15.87 3 -0.81 91.67 62.58 6 450 - 600 11.85 2 -0.84 100.00 74.43

GAP VEGETATION

Figure 18. GAP Vegetation Coverage. GAP Vegetation of the study area included 26 vegetation classes of the 38 total classes.
TABLE 9. PREDICTIVE MODEL BASED UPON GAP VEGETATION ORDER OF CLASS PERCENT SELECT TOTAL CUM % CUM % SEARCH VEGETATION AREA NESTS OBS-EXP NESTS NESTS AREA 1 PSME 5.58 3 2.386 6 25.00 5.58 2 PICO 48.06 7 1.713 15 87.50 53.64 3 PIPO 3.28 1 0.639 1 91.67 56.92

Figure 19. The area to be searched based upon the Predictive GAP model (Table 9). This model predicted eleven of thirteen nests (84.6%) after searching 56.92% of the study area.
TABLE 10. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON GAP VEGETATION ORDER OF CLASS PERCENT OBS CUM % CUM % SEARCH VEGETATION AREA NESTS OBS-EXP NESTS AREA 1 PSME 5.58 6 4.66 12.50 5.58 2 PICO 48.06 15 3.47 87.50 53.64 3 POTR 0.84 1 0.80 91.67 54.48 4 PIPO/SHRB 1.16 1 0.72 95.83 55.64 5 PIPO 3.28 1 0.21 100.00 58.92

Figure 20. The area to be searched (58.92%) based upon the Inclusive GAP model (Table 10) to locate 100% of the nests.

FOREST CANOPY DENSITY

Figure 21. Forest Canopy Density Coverage. This coverage consisted of five GAP vegetation tree species classes where goshawk nests occurred. Each class was divided into three canopy density classes (Low, Medium, & High) for a total of 15 classes.
CATEGORY % AREA CATEGORY % AREA
LOWDEN PIPO 0.47 LOWDEN POTR 0.00+
MEDDEN PIPO 2.15 MEDDEN POTR 0.28
HIGHDEN PIPO 0.67 HIGHDEN POTR 0.55
LOWDEN PICO 4.27 LOWDEN PIPO/SHRB 0
MEDDEN PICO 24.60 MEDDEN PIPO/SHRB 0.94
HIGHDEN PICO 19.36 HIGHDEN PIPO/SHRB 0.22
LOWDEN PSME 0.60
MEDDEN PSME 2.65
HIGHDEN PSME 2.35
TABLE 11. PREDICTIVE MODEL BASED UPON FOREST CANOPY DENSITY
ORDER OF CLASS PERCENT SELECT TOTAL CUM% CUM %
SEARCH CANOPY AREA NESTS OBS-EXP NESTS NESTS AREA
1 MEDDEN PICO 24.596 5 2.294 7 29.17 24.60
2 MEDDEN PSME 2.654 2 1.721 4 45.83 27.25
3 LOWDEN PICO 4.268 2 1.531 5 66.67 31.52
4 MEDDEN PIPO 2.152 1 0.763 1 70.83 33.67
5 HIGHDEN PSME 2.346 1 0.742 2 79.17 36.02

Figure 22. The area to be searched based upon the Predictive Forest Canopy Density model (Table 11). This model predicted eight of thirteen nests (61.5%) after searching 36.02% of the study area.
TABLE 12. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON FOREST CANOPY DENSITY ORDER OF CLASS PERCENT OBS CUM% CUM % SEARCH CANOPY AREA NESTS OBS-EXP NESTS AREA 1 LOWDEN PICO 4.268 5 3.98 20.83 4.27 2 MEDDEN PSME 2.654 4 3.36 37.50 6.92 3 HIGHDEN PSME 2.346 2 1.44 45.83 9.27 4 MEDDEN PICO 24.596 7 1.10 75.00 33.86 5 MEDDEN POTR 0.277 1 0.93 79.17 34.14 6 MEDDEN PIPO/SHRB 0.942 1 0.77 83.33 35.08 7 MEDDEN PIPO 2.152 1 0.48 87.50 37.25 . . . 22 HIGHDEN PICO 19.364 3 -1.64 100.00 67.02
Figure 23. The area to be searched (67.02%) based upon the Inclusive Forest Canopy Density model (Table 12) to locate 100% of the nests.


Figure 24. Young Northern Goshawks (between four and seven days old).

ASPECT AND DISTANCE TO WATER
Figure 25. Aspect and Distance to Water Coverage. A total of 72 classes were created (ASPECT-8 and DIST TO WATER-9).
TABLE 13. PREDICTIVE MODEL BASED UPON ASPECT AND DISTANCE TO WATER ORDER OF CLASS PERCENT OBS TOTAL CUM % CUM % SEARCH ASPECT DIST TO WATER(m) AREA NESTS OBS-EXP NESTS NESTS AREA 1 SW 0-150 1.225 2 1.865 2 8.33 1.23 2 NE 0-150 3.335 2 1.633 4 25.00 4.56 3 W 450-600 0.834 1 0.908 1 29.17 5.39 4 SW 150-300 1.132 1 0.875 1 33.33 6.53 5 N 750-900 1.305 1 0.856 3 45.83 7.83 6 SE 300-450 1.575 1 0.827 1 50.00 9.41 7 NW 150-300 2.280 1 0.749 1 54.17 11.69 8 NW 0-150 2.435 1 0.732 1 58.33 14.12 9 N 0-150 3.559 1 0.609 2 66.67 17.68

Figure 26. The area to be searched based upon the Predictive Aspect and Distance to Water model (Table 13). This model located an additional five nest sites (38.5%) after searching 17.68% of the study area.
TABLE 14. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON ASPECT & DISTANCE TO WATER ORDER OF CLASS PERCENT OBS CUM % CUM % SEARCH ASPECT DIST TO WATER(m) AREA NESTS OBS-EXP NESTS AREA 1 NE 0- 150 3.335 4 3.200 16.67 3.34 2 N 750- 900 1.305 3 2.687 29.17 4.64 3 SW 0- 150 1.225 2 1.706 37.50 5.87 4 N 0- 150 3.559 2 1.146 45.83 9.42 5 S 750- 900 0.756 1 0.819 50.00 10.18 6 SW 450- 600 0.806 1 0.807 54.17 10.99 7 W 450- 600 0.834 1 0.800 58.33 11.82 8 SW 150- 300 1.132 1 0.728 62.50 12.95 9 S 300- 450 1.217 1 0.708 66.67 14.17 10 NW 900-1050 1.252 1 0.700 70.83 15.42 11 W 0- 150 1.380 1 0.669 75.00 16.80 12 SE 300- 450 1.575 1 0.622 79.17 18.37 13 SE 150- 300 1.874 1 0.550 83.33 20.25 14 NW 300- 450 2.081 1 0.501 87.50 22.33 15 NW 150- 300 2.280 1 0.453 91.67 24.61 16 NW 0- 150 2.435 1 0.416 95.83 27.04 17 N 300- 450 2.600 1 0.376 100.00 29.64
Figure 27. The area to be searched (29.64%) based upon the Inclusive Aspect and Distance to Water model (Table 14) to locate 100% of the nests.

ELEVATION AND SLOPEA total of 88 classes were created (ELEVATION-11 and SLOPE-8).
TABLE 15. PREDICTIVE MODEL BASED UPON ELEVATION AND SLOPE ORDER OF CLASS PERCENT SELECT TOTAL CUM % CUM % SEARCH ELEVATION(m) SLOPE(%) AREA NESTS OBS-EXP NESTS NESTS AREA 1 2400 10-15 1.914 2 1.789 2 8.33 1.91 2 2300 25-30 0.319 1 0.965 2 16.67 2.23 3 2400 30-35 0.391 1 0.957 1 20.83 2.62 4 2700 5-10 1.701 1 0.813 1 25.00 4.33 5 2800 0- 5 1.717 1 0.811 1 29.17 6.04 6 2300 5-10 1.755 1 0.807 1 33.33 7.80 7 2400 5-10 2.031 1 0.777 2 41.67 9.83 8 2800 10-15 2.233 1 0.754 2 50.00 12.06 9 2900 10-15 2.482 1 0.727 2 58.33 14.54 10 2900 5-10 4.679 1 0.485 1 62.50 19.22

Figure 28. The area to be searched based upon the Predictive Elevation and Slope model (Table 15). The model located an additional four nest sites (30.8%) after searching 19.22% of the study area.
TABLE 16. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON ELEVATION & SLOPE ORDER OF CLASS PERCENT OBS CUM % CUM % SEARCH ELEVATION(m) SLOPE(%) AREA NESTS OBS-EXP NESTS AREA 1 2300 25-30 0.319 2 1.92 8.33 0.32 2 2300 20-25 0.381 2 1.91 16.67 0.70 3 2700 0- 5 0.840 2 1.80 25.00 1.54 4 2400 10-15 1.914 2 1.54 33.33 3.46 5 2400 0- 5 2.031 2 1.51 41.67 5.49 6 2800 10-15 2.233 2 1.46 50.00 7.72 7 2900 10-15 2.482 2 1.40 58.33 10.20 8 2500 10-15 2.798 2 1.33 66.67 13.00 9 2400 30-35 0.391 1 0.91 70.83 13.39 10 2700 10-15 1.639 1 0.61 75.00 15.03 11 2700 5-10 1.701 1 0.59 79.17 16.73 12 2800 0- 5 1.717 1 0.59 83.33 18.45 13 2300 5-10 1.755 1 0.58 87.50 20.20 14 2400 5-10 2.795 1 0.33 91.67 23.00 15 2800 5-10 3.714 1 0.11 95.83 26.71 . . . 36 2900 6-10 4.679 1 -0.12 100.00 36.40

Figure 29. The area to be searched (36.4%) based upon the Inclusive Elevation and Slope model (Table 16) to locate 100% of the nests.

GAP VEGETATION, ELEVATION, AND SLOPE
A total of 2288 classes were created (GAP-26, ELEVATION-11, and SLOPE-8).
TABLE 17. PREDICTIVE MODEL BASED UPON GAP VEGETATION, ELEVATION, AND SLOPE ORDER OF CLASS PERCENT SELECT TOTAL CUM % CUM % SEARCH VEGETATION ELEVATION(m) SLOPE(%) AREA NEST OBS-EXP NESTS NESTS AREA 1 PSME 2400 10-15 0.303 2 1.967 2 8.33 0.30 2 PICO 2400 30-35 0.167 1 0.982 1 12.50 0.47 3 PICO 2400 0- 5 0.184 1 0.980 2 20.83 0.65 4 PSME 2300 25-30 0.188 1 0.979 2 29.17 0.84 5 PIPO 2300 5-10 0.222 1 0.976 1 33.33 1.06 6 PICO 2800 0- 5 0.979 1 0.892 1 37.50 2.04 7 PICO 2700 5-10 1.050 1 0.885 1 41.67 3.09 8 PICO 2900 10-15 1.669 1 0.816 2 50.00 4.76 9 PICO 2800 10-15 1.796 1 0.802 2 58.33 6.56 10 PICO 2900 5-10 2.585 1 0.716 1 62.50 9.14

Figure 30. The area to be searched based upon the Predictive Gap Vegetation, Elevation, and Slope model (Table 17). This model located an additional four nest sites (30.8%) after searching 9.14% of the study area.
TABLE 18. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON GAP VEGTATION, ELEVATION, AND SLOPE ORDER OF CLASS PERCENT SELECT CUM % CUM % SEARCH VEGETATION ELEVATION(m) SLOPE(%) AREA NEST OBS-EXP NESTS AREA 1 PICO 2400 0- 5 0.184 2 1.956 8.33 0.18 2 PSME 2300 25-30 0.188 2 1.955 16.67 0.37 3 PSME 2400 10-15 0.303 2 1.927 25.00 0.68 4 PICO 2900 10-15 1.669 2 1.599 33.33 2.34 5 PICO 2800 10-15 1.796 2 1.569 41.67 4.14 6 PPSB 2700 0- 5 0.004 1 0.999 45.83 4.14 7 PICO 2300 20-25 0.012 1 0.997 50.00 4.16 8 POTR 2500 10-15 0.039 1 0.991 54.17 4.12 9 PICO 2400 30-35 0.167 1 0.960 58.33 4.36 10 PSME 2300 20-25 0.184 1 0.956 62.50 4.54 11 PSME 2400 5-10 0.210 1 0.950 66.67 4.76 12 PIPO 2300 5-10 0.222 1 0.947 70.83 4.98 13 PICO 2700 0- 5 0.391 1 0.906 75.00 5.37 14 PICO 2800 0- 5 0.979 1 0.765 79.17 6.34 15 PICO 2700 5-10 1.050 1 0.748 83.33 7.40 16 PICO 2700 10-15 1.200 1 0.712 87.50 8.60 17 PICO 2500 10-15 1.537 1 0.631 91.67 10.14 18 PICO 2900 5-10 2.585 1 0.380 95.83 12.72 19 PICO 2800 5-10 2.641 1 0.366 100.00 15.36

Figure 31. The area to be searched (15.36%) based upon the Inclusive Gap Vegetation, Elevation, and Slope model (Table 18) to locate 100% of the nests.

FOREST CANOPY DENSITY, ELEVATION, AND SLOPEA total of 1320 classes were created (FOREST CANOPY-15, ELEVATION-11, and SLOPE-8).
TABLE 19. PREDICTIVE MODEL BASED UPON FOREST CANOPY DENSITY, ELEVATION, AND SLOPE ORDER OF CLASS PERCENT SELECT TOTAL CUM % CUM % SEARCH CANOPY ELEVATION(m) SLOPE(%) AREA NESTS OBS-EXP NESTS NESTS AREA 1 HIGHDEN PSME 2400 10-15 0.103 2 1.989 2 8.33 0.10 2 LOWDEN PICO 2400 0- 5 0.064 1 0.993 2 16.67 0.17 3 MEDDEN PICO 2400 30-35 0.074 1 0.992 1 20.83 0.24 4 MEDDEN PSME 2300 20-25 0.083 1 0.991 1 25.00 0.32 5 MEDDEN PIPO 2300 5-10 0.163 1 0.982 1 29.17 0.48 6 LOWDEN PICO 2800 0- 5 0.219 1 0.976 1 33.33 0.70 7 LOWDEN PICO 2900 5-10 0.315 1 0.965 1 37.50 1.02 8 MEDDEN PICO 2700 5-10 0.698 1 0.923 1 41.67 1.72 9 MEDDEN PICO 2900 10-15 0.880 1 0.903 2 50.50 2.60 10 MEDDEN PICO 2800 10-15 0.985 1 0.892 1 54.17 3.58

Figure 32. The area to be searched based upon the Predictive Forest Canopy Density, Elevation, and Slope model (Table 19). This model located an additional two nest sites (15.4%) after searching 3.58% of the study area.
TABLE 20. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON FOREST CANOPY DENSITY, ELEVATION, AND SLOPE ORDER OF CLASS PERCENT OBS CUM % CUM % SEARCH CANOPY ELEVATION(m) SLOPE(%) AREA NESTS OBS-EXP NESTS AREA 1 LOWDEN PICO 2400 0- 5 0.064 2 1.985 8.33 0.064 2 MEDDEN PSME 2300 20-25 0.093 2 1.978 16.67 0.157 3 HIGHDEN PSME 2400 10-15 0.103 2 1.975 25.00 0.260 4 MEDDEN PICO 2900 10-15 0.880 2 1.789 33.33 1.140 5 MEDDEN PPSB 2700 0- 5 0.003 1 0.999 37.50 1.143 6 MEDDEN POTR 2500 10-15 0.011 1 0.997 41.67 1.154 7 MEDDEN PICO 2300 15-20 0.012 1 0.997 45.83 1.166 8 LOWDEN PICO 2700 0- 5 0.063 1 0.985 50.00 1.229 9 MEDDEN PICO 2400 30-35 0.074 1 0.982 54.17 1.303 10 MEDDEN PSME 2300 25-30 0.083 1 0.980 58.33 1.386 11 MEDDEN PSME 2400 5-10 0.122 1 0.971 62.50 1.508 12 MEDDEN PIPO 2300 5-10 0.163 1 0.961 66.67 1.671 13 LOWDEN PICO 2800 0- 5 0.219 1 0.947 70.83 1.890 14 LOWDEN PICO 2900 5-10 0.315 1 0.924 75.00 2.205 15 HIGHDEN PICO 2700 10-15 0.355 1 0.915 79.17 2.560 16 HIGHDEN PICO 2500 10-15 0.443 1 0.894 83.33 3.003 17 HIGHDEN PICO 2800 10-15 0.627 1 0.850 87.50 3.630 18 MEDDEN PICO 2700 5-10 0.698 1 0.832 91.67 4.328 19 MEDDEN PICO 2800 10-15 0.985 1 0.764 95.83 5.313 20 MEDDEN PICO 2800 5-10 1.553 1 0.627 100.00 6.866

Figure 33. The area to be searched (6.87%) based upon the Inclusive Forest Canopy Density, Elevation, and Slope model (Table 20) to locate 100% of the nests.

ASPECT, ELEVATION, AND SLOPEA total of 704 classes were created (ASPECT-8, ELEVATION-11, and SLOPE-8).
TABLE 21. PREDICTIVE MODEL BASED UPON ASPECT, ELEVATION, AND SLOPE ORDER OF CLASS PERCENT SELECT TOTAL CUM % CUM % SEARCH ASPECT ELEVATION(m) SLOPE(%) AREA NESTS OBS-EXP NESTS NESTS AREA 1 SW 2400 30-35 0.020 1 0.998 1 4.17 0.02 2 N 2300 25-30 0.138 1 0.985 2 12.50 0.16 3 SW 2800 0- 5 0.242 1 0.973 1 16.67 0.40 4 NE 2400 10-15 0.296 1 0.967 1 20.83 0.70 5 SW 2800 10-15 0.310 1 0.966 1 25.00 1.01 6 NE 2700 5-10 0.317 1 0.965 1 29.17 1.32 7 NW 2300 5-10 0.327 1 0.964 1 33.33 1.65 8 SE 2900 10-15 0.347 1 0.962 1 37.50 2.00 9 NW 2400 10-15 0.362 1 0.960 1 41.67 2.36 10 W 2900 5-10 0.363 1 0.960 1 45.83 2.72 11 N 2400 0- 5 0.364 1 0.960 1 50.00 3.09

Figure 34. The area to be searched based upon the Predictive Aspect, Elevation, and Slope model (Table 21). This model located one additional nest site (7.7%) after searching 3.09% of the study area.
TABLE 22. SEARCH ORDER OF INCLUSIVE MODEL BASED UPON ASPECT, ELEVATION, AND SLOPE ORDER OF CLASS PERCENT OBS CUM % CUM % SEARCH ASPECT ELEVATION(m) SLOPE(%) AREA NESTS OBS-EXP NESTS AREA 1 N 2300 25-30 0.138 2 1.967 8.33 0.14 2 SW 2400 30-35 0.020 1 0.995 12.50 0.16 3 N 2700 0- 5 0.058 1 0.986 16.67 0.22 4 NW 2300 20-25 0.085 1 0.980 20.83 0.30 5 SE 2700 0- 5 0.112 1 0.973 25.00 0.31 6 N 2300 20-25 0.150 1 0.964 29.17 0.46 7 SW 2500 10-15 0.190 1 0.954 33.33 0.65 8 S 2800 10-15 0.227 1 0.946 37.50 0.88 9 SW 2800 0- 5 0.242 1 0.942 41.67 1.12 10 NW 2400 0- 5 0.246 1 0.941 45.83 1.37 11 NE 2400 10-15 0.296 1 0.929 50.00 1.66 12 S 2900 10-15 0.305 1 0.927 54.17 1.97 13 NE 2500 10-15 0.308 1 0.926 58.33 2.28 14 NE 2700 10-15 0.309 1 0.926 62.50 2.59 15 SW 2800 10-15 0.310 1 0.926 66.67 2.90 16 NE 2700 5-10 0.317 1 0.924 70.83 3.21 17 NW 2300 5-10 0.327 1 0.922 75.00 3.54 18 SE 2900 10-15 0.347 1 0.917 79.17 3.89 19 NW 2400 10-15 0.362 1 0.913 83.33 4.25 20 W 2900 5-10 0.363 1 0.913 87.50 4.61 21 N 2400 0- 5 0.364 1 0.913 91.67 4.98 22 W 2800 5-10 0.376 1 0.910 95.83 5.35 23 N 2400 5-10 0.758 1 0.818 100.00 6.11

Figure 35. The area to be searched (6.11%) based upon the Inclusive Aspect, Elevation, and Slope model (Table 22) to locate 100% of the nests.


Figure 36. Adult female Northern Goshawk defending her nest.

DISCUSSION AND SUMMARY
In order to rate the Predictive Models, we divided the percentage of nest sites predicted by the area searched
to derive performance indices which were ranked (Table 23). The performance indices for the Inclusive Models were
derived by dividing 100% by the area searched. These performance indices were ranked for their ability to locate all
nest sites.
The best overall Predictive Model was the Forest Canopy, Elevation, and Slope Model, which predicted 15.4% of
the nest sites in only 3.6% of the area (Performance Index = 4.28). The best two-factor Predictive Model was the
Aspect and Distance to Water Model, which predicted 38.5% of the nest sites in 17.7% of the area (Performance Index =
2.18). The best single-factor Predictive Model was the Forest Canopy Model which predicted 61.5% of the nest sites
in 36.0% of the area (Performance Index = 1.71).
The best overall Inclusive Model was the Aspect, Elevation, and Slope Model, which required 6.11% of the area
to locate all nest sites (Performance Index = 16.39). The best two-factor Inclusive Model was also the Aspect and
Distance to Water Model, which required 29.6% of the area to locate all nest sites (Performance Index = 3.38).
The best single-factor model was the GAP Model, which required 58.9% of the area to locate all nest sites (Performance
Index = 1.69).
The Inclusive Models will be tested this summer in Ashley National Forest for their ability to locate
additional (previously unknown) nest sites.
TABLE 23. COMPARISON OF PREDICTIVE AND INCLUSIVE MODELS.
% OF NESTS AREA PERFORMANCE PERFORMANCE AREA SEARCHED PERFORMANCE PERFORMANCE
MODEL PREDICTED SEARCHED INDEX RANK TO FIND ALL NESTS INDEX RANK
(P) (S) (P/S) (AS) (T*/AS)
ELEVATION 84.6 53.6 1.58 7 72.5 1.38 8
SLOPE 84.6 67.7 1.25 10 77.1 1.30 10
ASPECT 100.0 85.3 1.17 11 85.3 1.17 11
DIST TO WATER 100.0 74.4 1.34 9 74.4 1.34 9
GAP 84.6 56.9 1.48 8 58.9 1.69 6
FOREST CANOPY 61.5 36.0 1.71 5 67.0 1.49 7
ASPECT &
DIST TO WATER 38.5 17.7 2.18 4 29.6 3.38 4
ELEVATION &
SLOPE 30.8 19.2 1.60 6 36.4 2.75 5
GAP, ELEVATION,
& SLOPE 30.8 9.1 3.38 2 15.4 6.49 3
FOREST CANOPY,
ELEVATION,
& SLOPE 15.4 3.6 4.28 1 6.9 14.49 2
ASPECT,
ELEVATION,
& SLOPE 7.7 3.1 2.48 3 6.1 16.39 1
* - T represents all nests (100%).

LITERATURE CITEDJohansson, C., P.J. Hardin, and C.M. White. 1994. Large-area goshawk habitat modeling in Dixie National Forest using vegetation and elevation data. The Northern Goshawk: Ecology and Management. W.M. Block, M.L. Morrison, and M.H. Resier, Eds. Studies in Avian Biology No. 16:24. Joy, S.M., R.T. Reynolds, and D.G. Leslie. 1994. Northern Goshawk Broadcast Surveys: Hawk Response Variables and Survey Costs. The Northern Goshawk: Ecology and Management. W.M. Block, M.L. Morrison, and M.H. Resier, Eds. Studies in Avian Biology No. 16:24. Kennedy, P., and D. Stahlecker. 1993. Responsiveness of nesting Northern Goshawks to taped broadcasts of three conspecific calls. J. Wildl. Manage. 57:249-257. Reynolds, R.T. et al. (eds. committee). 1991. Management recommendations for the Northern Goshawk in the southwestern United States. Southwest Region, Albuquerque, N.M. U.S. Dept. Interior. 1991. Endangered and threatened wildlife and plants; Animal Candidate Review... Federal Register 56(225): 58804© 58836.

APPENDIX 1 HABITAT DATA FOR NEST SITES
DISTANCE TO
NEST ELEVATION-CLASS SLOPE-CLASS ASPECT GAP-VEG WATER ROAD
1 2571 - 5 14.8 - 3 SW HD PICO 492 440
2 2397 - 3 20.3 - 5 N MD PICO 766 855
3* 2346 - 3 25.5 - 6 N MD PSME 787 350
4 2932 - 9 10.6 - 3 S MD PICO 810 17
5* 2857 - 8 12.7 - 3 SW MD PICO 285 43
6 2875 - 8 12.1 - 3 S MD PICO 351 164
7 2447 - 4 8.8 - 2 N MD PSME 61 609
8* 2387 - 3 7.7 - 2 NW MD PIPO 256 515
9* 2414 - 4 13.8 - 3 NW MD PSME 0 365
10 2339 - 3 21.9 - 5 NW HD PSME 322 275
11 2730 - 7 3.00 - 1 SE MD PIPO/SH 176 14
12 2731 - 7 5.00 - 1 N LD PICO 344 14
13 2756 - 7 12.1 - 3 NE MD PICO 16 523
14* 2725 - 7 8.4 - 2 NE MD PICO 31 107
15 2484 - 4 2.4 - 1 NW LD PICO 925 1252
16 2539 - 5 10.1 - 3 NE MD POTR 49 1593
17* 2476 - 4 3.4 - 1 N MD PICO 92 518
18* 2451 - 4 12.7 - 3 NE HD PSME 15 195
19* 2907 - 9 10.7 - 3 SE MD PICO 346 752
20* 2890 - 8 3.7 - 1 SW LD PICO 0 25
21 2899 - 8 10.0 - 2 W MD PICO 13 261
22* 2999 - 9 8.5 - 2 W LD PICO 459 14
23 2384 - 3 26.0 - 6 N MD PSME 806 573
24* 2467 - 4 30.1 - 7 SW MD PICO 131 478
* - Indicates nests selected to build predictive
models.