Wildlife Ecology Unit

Goshawk Bioregional Monitoring Design

A Brief Description of the Proposed Bioregional Monitoring Design For Northern Goshawks

Christina Hargis and Brian Woodbridge
May, 2004

Information on northern goshawk (Accipiter gentilis) status is generally obtained by observing nesting activity at local scales. Although nest observations provide breeding information for specific territories, this approach is unable to provide information on the abundance of goshawks over broad spatial extents. To address the need for information on goshawk population trends, the Forest Service assembled a working group in 2002 that developed a design for monitoring goshawks at a "bioregional" scale (e.g., northern Rockies, Cascade-Sierra, Intermountain Great Basin). The working group consisted of statisticians and goshawk researchers within and outside of the Forest Service. The group was chartered to create a monitoring design to be implemented on National Forest Service land, but the Forest Service is interested in collaboration with other landowners and state natural resource agencies to provide a more complete picture of goshawk status.
The working group created a monitoring design with the following objectives: 1) to estimate the relative abundance of territorial adult goshawks within a bioregion; 2) to assess changes in goshawk relative abundance over time; and 3) to determine whether changes in relative abundance, if any, are associated with changes in habitat.

The sample population for each bioregion is a grid of square, 688 ha primary sampling units (PSUs) across all potential goshawk habitat owned or managed by the collaborating parties. The sampling frame is stratified to obtain a reasonable estimate of goshawk relative abundance with an efficient use of funds. The monitoring indicator is the proportion of sampled PSUs with goshawk presence, based on the broadcast acoustical survey method. After the sampled PSUs are surveyed two times (nestling and fledgling periods), the frequency of goshawk presence within the bioregion is estimated using a maximum likelihood estimator. Changes in frequency of goshawk presence will be assessed after a minimum of five years, using a logistic model with habitat parameters entered as covariates. Information from bioregional monitoring will help determine the status of goshawk populations and their habitats over a spatial extent that is meaningful for goshawk conservation.

For each bioregion, the first step in implementing this design is to identify a bioregional coordinator to oversee the entire program. The responsibilities of the bioregional coordinator are to establish the sampling frame, determine sampling intensity, oversee data collection, conduct data analysis, prepare annual reports, and administer the budget. The coordinator can be affiliated with any agency, research facility, or university.

To establish the sampling frame, the bioregional coordinator acquires a base map with layers for vegetation, roads, and landownership. A grid of PSUs can be automated in GIS using a random UTM as a starting point. All PSUs that fall on lands owned or managed by the Forest Service and any other collaborators are identified, and from these, all non-habitat (non-forested) lands are removed. The remaining PSUS are assigned unique identifiers and are the sampling frame for the bioregion.

The PSUs are then stratified into 4 strata:
1) primary habitat, easy to access
2) primary habitat, difficult to access
3) marginal habitat, easy to access
4) marginal habitat, difficult to access

Criteria for primary and marginal, easy and difficult, are developed by the bioregional coordinator with assistance from the literature, expert opinion, or modeling. It is expected that there will be errors in assigning PSUs to these 4 strata, especially if goshawk habitat is poorly understood in a bioregion and/or if accessibility is unknown. Nevertheless, even crude stratification can provide a more efficient design than simple random sampling. Systematic or simple random sampling would result in a large commitment of monitoring funds in areas that are likely not used by goshawks, with the inherent risk that little would be learned about goshawk population status. The purpose of stratification is to provide a reasonable estimate of goshawk abundance with an efficient use of funds.

After the sampling frame is stratified, a specific algorithm (Jim Baldwin, Pacific Southwest Research Station) is used to 1) calculate sample size and 2) allocate the sample size among the 4 strata. The equation does not result in proportional sampling among the 4 strata. Rather, it allocates the majority of the sample to primary habitats and to habitats that are easy to access, while ensuring that marginal habitats and habitats that are difficult to access are also included. Areas that are currently surveyed for goshawks tend to be restricted to primary habitat in places that are easy to access, so this stratification will ensure that the sample includes areas with lesser habitat potential and less accessibility.

Each PSU is sampled twice using the broadcast acoustical survey method developed by Kennedy and Stahlecker (1993) and Joy et al. (1994). Transects are 250 m apart and call points are 200 m apart. The surveyors begin with the best habitat in the PSU and survey until a detection is made or all suitable habitat is surveyed, whichever comes first. Surveyors do not need to survey unsuitable habitat such as cliffs, talus slopes, non-forested areas, and water bodies. We anticipate 1-7 person days to survey a PSU.
Each survey results in one of two possible outcomes: presence (1) or absence (0). After both surveys, each PSU has four possible combinations of presence or absence outcomes: 00, 01, 10, or 11. The frequencies of these combinations over all sampled PSUs are used in an equation called the "likelihood function", along with the detection rate coefficients and the probability of presence. The values of the detection rate coefficients and the probability of presence that maximize the likelihood function are used as estimates for those parameters, hence the name "maximum likelihood estimates." The estimated probability of presence is the estimate of how many of the total number of PSUs in the sampling frame are likely to have goshawks present.

The bioregional plan also describes the collection of habitat data. A number of variables are collected at the landscape scale for every sampled PSU, whether or not a goshawk was detected. From these, an assessment can be made of landscape variables in PSUs with and without detections. The variables are: 1) number of vegetation patches; 2) number of vegetation cover types; 3) size of largest vegetation patch (including patch area that extends beyond the PSU boundary); 4) percent of PSU in primary, marginal, and non habitat as defined by the initial PSU stratification process; 5) estimated proportion of PSU that has been thinned and/or burned under prescription in the last 20 years; 5) estimated proportion of PSU that has been harvested in the last 20 years (commercial thinning, overstory removal or clearcut) and 6) straight-line distances from the PSU center to the nearest permanent water (including springs), road (regardless of use status), trail, and meadow edge.

Forest Inventory Analysis (FIA) data are also useful for evaluating habitat across the entire bioregion. It would not be appropriate to correlate goshawk detections with the nearest FIA point, because the FIA point might be too distant from the goshawk detection to be meaningful. Each FIA point covers approximately 6,000 ac. However, changes in bioregional FIA data over 5 years can be compared to changes in goshawk prevalence over 5 years, to see if there is any correlation between habitat data and goshawk abundance over time.
An important caveat is that the bioregional monitoring plan provides an estimate of goshawks for the entire area, but not for any individual forest. The sample size on any individual forest would be too small to estimate goshawk prevalence. Because of this, there is concern that a bioregional monitoring plan might not meet the legal requirement for monitoring MIS at the forest scale. However, goshawks within a specific national forest are not isolated from goshawks on adjacent forests and other neighboring lands, so "population" trends for a given forest are likely not meaningful.

We feel that the bioregional monitoring is the best scale for evaluating goshawk population trends. If each national forest contributes to the bioregional monitoring effort, it could be argued that they are meeting their monitoring requirements. Each national forest in the bioregion should continue nest observations, however, at least for a sample of known territories.

This monitoring plan was tested in 2003 on the San Juan and Rio Grande National Forests in southwestern Colorado. Twenty PSUs were surveyed, resulting in two detections. The average cost to survey a PSU was approximately $1,000, which included training, pre- and post-field mapping, filling out forms, and report writing, as well as the actual field work.

There is substantial momentum for implementing the bioregional design, both within the Forest Service and in several state agencies. Planning efforts are underway in the Forest Service Northern Region, in Nevada (a collaborative effort between Nevada Department of Wildlife and the Humboldt-Toiyabe National Forest), in portions of the Forest Service Rocky Mountain Region, and in the Great Lakes area (multi-states, multi-forests). These efforts include establishment of the sampling frame and field testing to provide estimates of detection probabilities.

The goshawk bioregional monitoring design will be distributed as a Forest Service report, The Northern Goshawk Inventory and Monitoring Technical Guide. A condensed version will also appear as a chapter in a forthcoming book, The Northern Goshawk: A Technical Assessment of its Status, Ecology, and Management, edited by Michael Morrison and published in the series, Studies in Avian Biology. Both publications will be available in 2004.

Literature Cited

Kennedy, P. L., and D. W. Stahlecker. 1993. Responsiveness of nesting Northern Goshawks to taped broadcasts of 3 conspecific calls. Journal of Wildlife Management 57:249-257.

Joy, S. M., R. T. Reynolds, and D. G. Leslie. 1994. Northern Goshawk broadcast surveys: hawk response variables and survey cost. Pp. 24-30 in W. M. Block, M. L. Morrison, and M. H. Reiser, (editors). The Northern Goshawk: Ecology and management. Studies in Avian Biology No. 16, Cooper Ornithological Society, Allen Press, Inc., Lawrence, Kansas, USA.