home > data and maps > derived data > goshawk

ForestERA Data Layer Overview - Northern Goshawk Nesting Habitat

Description

This is a layer representing predicted nesting habitat for the Northern Goshawk (Accipiter gentilis), as well as a measure of nesting habitat quality, across the assessment area. The layer was built using a combination of simple rules based on a literature review, and Mahalanobis distance based on habitat conditions at known nest sites. The layer has a resolution of 90m (0.8 ha or 2 acres), and values within the layer range from zero to 5. Higher values indicate a higher degree of habitat suitability.

Purpose

This data layer was created as part of the ForestERA project to support landscape-scale forest restoration planning efforts by a broad group of stakeholders including federal and state agencies, academic institutions, and non-governmental entities. These data are intended for regional analyses over spatial extents on the order of tens to hundreds of thousands of acres, and were not developed for use at finer spatial scales, although they may be useful for some applications at finer scales.

Development

We chose two modeling approaches to optimize the use of available information on Northern Goshawk. First, because previous studies documented characteristics of nesting habitat for these species, we used rules developed from the literature to identify the extent of potential nesting habitat on the landscape. This involved a search of the literature to identify a set of variables related to nesting habitat and a range of possible threshold values that might be used to identify areas most likely to support nesting pairs of these raptors. Second, we employed the Mahalanobis distance statistic (hereafter M-distance) to generate a dataset that could be used as a surrogate for habitat “suitability” or “preference”. Georeferenced locations of 123 nest sites for Northern Goshawk were obtained from the Arizona Game and Fish Department and studies at Northern Arizona University (e.g., Beier and Drennan 1997) for use in the modeling effort.

To identify potential nesting habitat, we used publications that described habitat characteristics around Northern Goshawk nest sites (Daw & Destephano 2001; Finn et al. 2002; Hargis et al. 1994; Hayward & Escano 1989; Joy 2002) and current management guidelines (Reynolds et al. 1992) to choose a range of threshold values (Prather et al., in review). We classified areas with both canopy cover > 40% and basal area > 18 m2/ha as potential nesting habitat for goshawks. We then eliminated patches of habitat smaller than 10ha because these patches would likely be too small to be suitable for nesting (Reynolds et al. 1992). The final extent of potential nesting habitat covered 350,000 ha (43% of the study area).

After delineation of potential nesting habitat, we used M-distance and habitat characteristics at Northern Goshawk nest sites to develop an index of habitat suitability within potential nesting habitat. M-distance has been used frequently in mathematics and physics to compare the similarity of datasets, but is rarely used in ecology (e.g., Clark et al. 1993; Corsi et al. 1999; Knick and Dyer 1997). M-distance is a multivariate statistic that provides a measure of dissimilarity between two multivariate datasets (Farber & Kadmon 2003). In this effort, one dataset is the mean vector of habitat characteristics at known Northern Goshawk nest sites while the other dataset is the range of conditions across the entire landscape (Prather et al., in review). The “distance” for any location on the landscape represents the dissimilarity between conditions at that location and the mean habitat conditions at the known nest sites. This value is scaled in multivariate space using the covariance between the measured habitat conditions at the locations used by the species (Clark et al. 1993).
We used dominant overstory vegetation, canopy cover, basal area, tree density, slope, and the sine and cosine derivatives of aspect as predictor variables for our M-distance modeling effort. We determined values for each of these variables at Northern Goshawk nest sites by extracting the values from the GIS raster layers using the “zonal statistics” function in ArcGIS Spatial Analyst (ESRI Corporation). We overlaid the nest locations with the raster layers of the predictive variables and determined the value of each raster layer in the pixel corresponding to the nest location. Analysis of characteristics of the 90m (0.81ha) pixel in which each nest fell was deemed appropriate since other studies have shown significant patterns between nest site selection by large raptors and habitat characteristics at scales of approximately 0.5 - 1ha (e.g., McGrath et al. 2002). For purposes of developing the M-distance model we used 2/3 (82) of the nest sites as training data and the remaining 1/3 (42) of the nest sites were withheld for accuracy assessment.

The final habitat layer represents a combination of our two modeling efforts. We classified this layer into 5 categories for purposes of display, analysis, and dissemination. The zero category represents areas not predicted to be suitable for nesting habitat using our simple rules. The remaining 4 categories represent areas of increasing similarity to the mean habitat characteristics at Northern Goshawk nest sites based on M-distance. Fifty percent of nest sites would be expected to fall into category 4, an additional 25% into category 3, an additional 15% into category 2, and the final 10% into category 1.

Accuracy Assessment

To test the effectiveness of using threshold values within predictor variables as a means of identifying potential habitat, we used Chi-squared goodness of fit tests (Zar 1999) to determine whether the potential habitat models did significantly better at predicting locations of Northern Goshawk nests than chance alone. For this analysis we used the entire set of nest sites (n = 123). This analysis suggested that significantly more (101 or 82%) nest sites fell within the predicted extent of potential nesting habitat than would be expected by chance alone (v = 2, X2 = 41.6, P < 0.0001).

To test whether the M-distance statistic was effective at identifying the locations of nest sites, we used Chi-squared goodness of fit tests to determine whether the M-distance statistic accurately predicted the number of nests within the test datasets that fell within the portion of the predicted habitat in each of the four categories identified above. For these analyses we used the 1/3 of the nests (n = 41) that were not used as training data for the M-distance model. This analysis showed that the distribution of nests within each habitat category fit the expected distributions (v = 3, X2 = 0.24, P = 0.97; Table 1). This suggests that the simple rules did a good job at defining potential habitat, and that M-distance was effective at predicting the area within potential habitat where a given percentage of nests would be expected to occur.

Table 1:  Distribution of Northern Goshawk nests (n = 41) in the test dataset in relation to categories generated using Mahalanobis distances and the nests in the training datasets.

Sources of errors

The potential habitat model developed above contained about 82% of the known Northern Goshawk nest sites. There are several potential reasons that some sites may fall outside of predicted habitat. First, there is some uncertainty in the basal area and canopy cover layers used in the modeling. Thus some areas that are suitable habitat may not be identified as such. Second, many of the goshawk nest sites had positioning errors of + 100m, so the location of some nests on the landscape may be inaccurate. Third, conditions may have changed at some nest sites since they were last active. All of these may lead to errors.

It should also be noted that certain assumptions must be made when using the M-distance statistic as a measure of habitat “quality” or “preference”. The most important of these assumptions are that the distribution of habitat characteristics at known nest sites is representative of the actual distribution of nest sites, and that the mean habitat characteristics at nest sites represent “optimal” or “preferred” habitat characteristics. We do not currently have the data to determine whether M-distance values can be linked to survivorship or reproductive success, and thus we do not know how good an estimator of habitat quality is provided by the M-distance statistic.

Recommendations

We recommend that this layer be used at a minimum resolution of 90m (0.8 ha or 2 acres) for purposes of analysis and display. However, ForestERA data layers were not designed for analyses at the level of individual pixels, and uncertainty in the data will generally decline over greater spatial extents. Therefore, we recommend using larger analysis units, with groupings of at least 50 cells (40 ha or 100 acres). Finally, we reiterate that ForestERA data layers were developed for the purpose of regional landscape-level planning, and we suggest that the analyses be applied over spatial extents of tens to hundreds of thousands of acres. We recognize, however, that this layer may be useful for analyses over smaller spatial extents depending on the type and purpose of those analyses.

Literature Cited

Beier, P. and J. E. Drennan. 1997. Forest structure and prey abundance in foraging areas of Northern Goshawks. Ecological Applications 7: 564-571.

Clark, J. D., J. E. Dunn, and K. G. Smith. 1983. A multivariate model of female black bear habitat use for a Geographic Information System. Journal of Wildlife Management 57: 519-526.

Corsi, F., E. Dupre, and L. Boitani. 1999. A large-scale model of wolf distribution in Italy for conservation planning. Conservation Biology 13: 150-159.

Daw, S. K., and S. DeStephano. 2001. Forest characteristics of Northern Goshawk nest stands and post-fledging areas in Oregon. Journal of Wildlife Management 65: 59-65.

Farber, O. and R. Kadmon. 2003. Assessment of alternative approaches for bioclimatic modeling with special emphasis on the M-distance. Ecological Modelling 160: 115-130.

Finn, S. P., J. M. Marzluff, and D. E. Varland. 2002. Effects of landscape and local habitat attributes on Northern Goshawk site occupancy in western Washington. Forest Science 48: 427-436.

Hargis, C. A., C. McCarthy, and R. D. Perloff. 1994. Home ranges and habitats of northern Goshawks in eastern California. Studies in Avian Biology 16: 66-74.

Hayward, G. D. and R. E. Escano. 1989. Goshawk nest-site characteristics in western Montana and northern Idaho. Condor 91: 476-479.

Joy, S. M. 2002. Northern Goshawk habitat on the Kaibab National Forest, Arizona: factors affecting nest locations and territory quality. Ph.D. Dissertation. Colorado State University, Fort Collins. 241 pp.

Knick, S. T. and D. L. Dyer. 1997. Distribution of Black-tailed Jackrabbit habitat determined by GIS in southwestern Idaho. Journal of Wildlife Management 61: 75-85.

McGrath, M. T., S. DeStephano, R. A. Riggs, L. L. Irwin, and G. L. Roloff. 2002. Spatially explicit influences on Northern Goshawk nesting habitat in the interior Pacific Northwest. Wildlife Monographs 154: 1-63.

Prather, J. W., J. L. Ganey, P. Beier, W. M. Block, M. Ingraldi, J. Jenness, Y. Xu, H. M. Hampton, and T. D. Sisk. (in review). Simple models successfully map nesting habitat for two raptors in southwestern coniferous forests. Conservation Biology.

Reynolds, R. T., R. T. Graham, M. H. Reiser, L. Bassett, P. L. Kennedy, D. A. Boyce Jr., G. Goodwin, R. Smith, and E. L. Fisher. 1992. Management Recommendations for the Northern Goshawk in the Southwestern United States. USDA Forest Service General Technical Report RM-GTR-217.

Zar, J. H. 1999. Biostatistical Analysis, 4th Edition. Prentice-Hall Inc. Englewood Cliffs, New Jersey.

Last updated February 11, 2005