Spatially explicit influences on northern goshawk nesting habitat in the interior Pacific Northwest

We compared northern goshawk (Accipiter gentilis atricapillus) nesting habitat within 1 ha of nest sites and at landscape scales of 10, 30, 60, 83, 120, 150, and 170 ha in 4 study areas east of the Cascade Mountains in Oregon and Washington. Our objective was to describe goshawk nesting habitat at biologically relevant scales and to develop models capable of assessing the effects of forest management alternatives on habitat suitability. We evaluated habitat at 82 active goshawk nests and 95 random sites. Productivity (voting fledged per nest) was evaluated at 81 nests. We collected data on forest structure within 1 ha of nests and random points. At scales ranging from 10 to 170 ha, we recorded the abundance and spatial distribution of several forest stages of stand development (i.e., stand initiation, stein exclusion, understory reinitiation, old growth) on aerial photographs. We used logistic regression and classification and regression trees (CART) to (1) evaluate habitat selection, (2) construct models to calculate the probability of nesting, and (3) explore relationships between reproductive output and habitat conditions. We assessed model accuracy via bootstrap and jackknife cross-validation techniques. By examining goshawk habitat relationships it multiple spatial scales across several study areas, we detected unifying spatial patterns and structural conditions surrounding goshawk nesting habitat. Our ability to discriminate goshawk nest sites from available habitat decreased as landscape scale increased, and different factors influenced goshawks at different scales. The presence and arrangement of forest structural types interacted to influence site suitability for nesting. At the 1-ha scale, the stage of stand development (i.e., stand initiation, stem exclusion, understory reinitiation, old growth), low topographic position, and tree basal area reliably discriminated between nests and random sites. Low topographic position and basal area were more influential than stand structure. At the landscape scale, modeling indicated that conditions at different scales interact to influence selection of habitat for nesting. A core area exists surrounding goshawk nests in which stem exclusion and understory reinitiation stands with canopy closure greater than or equal to50% serve as apparent protection against potentially detrimental effects associated with more open forest (e.g., predators and micro-climate). Among several models tested, the model that best discriminated between nests and random sites encompassed 83 ha surrounding the nest and incorporated habitat characteristics from multiple scales nested within that range. This model had a cross-validated classification accuracy of 75%. Positive correlations were found between fledging rate and tree basal area within 1 ha of the nest (F-3,F- 77 = 2.89, P = 0.0407), and between fledging rate and the percentage of landscape occupied by "stem exclusion" stands of low canopy closure (i.e., <50%) at landscape scales >60 ha (F-3,F- 77, 0.041 less than or equal to P less than or equal to 0.089). Spatial modeling also showed that timber harvest can be managed to maintain or enhance goshawk nest site suitability over time in the Interior Northwest, and that a non-harvest strategy can be just as detrimental to nesting habitat as can be aggressive, maximum-yield forestry. We conclude that (1) northern goshawk nesting habitat becomes less distinguishable from the landscape with increasing area, and (2) habitat management based on exclusionary buffers should be re-evaluated in light of the way different habitat factors interact across spatial scales. We present case studies illustrating how landscape scale modeling can be implemented to (1) predict the influences of alternative silvicultural prescriptions on the suitability of potential nesting habitat over time, and (2) characterize large landscapes with respect to abundance and distribution of suitable nesting habitat.