Evaluation of a spatial habitat suitability model for the northern goshawk (Accipiter gentilis) in West-central Alberta

Abstract

The current era of forest management has expanded the array of values, spatial dimensions and temporal horizons beyond those formerly considered in the planning process. The desire to evaluate the impact of various management scenarios on biodiversity has prompted the development of tools such as habitat suitability models. The implicit but frequently untested assumption in the application o f these models is that their outputs accurately reflect real-world habitat use. As part of the Biodiversity Assessment Project, Millar Western Forest Products Ltd. developed spatially-explicit habitat supply models for 17 wildlife species. The objective o f this study was to evaluate the performance of the previously untested habitat suitability model for the Northern Goshawk (Accipiter gentilis). The model was first reviewed for agreement with current literature in terms of computational methods and outputs across forest cover-types. A sensitivity analysis of the model directed field studies and illustrated that forest cover-type was the most influential variable within the nesting component of the model. Field investigations revealed the model’s inability to identify goshawk nesting habitat correctly. Identified goshawk nest sites (n = 15) had an average nesting suitability value of only 0.25 (on a scale o f 0 to 1). A negative correlation between modelled nesting suitability values and actual habitat use was also observed. Improvements to the nesting component of the model were made in regard to forest cover-type, canopy closure, and proximity to roads. A modification to the method used to calculate foraging habitat was also suggested. These modifications improved model performance with respect to known nest sites, but only within the limits of input data. Discrepancies between forest inventory and field data were frequent as field-collected data commonly reported trembling aspen cover where the forest inventory did not. These discrepancies resulted in model modifications improving nesting scores for known sites from 0.25 to 0.45 using inventory data and from 0.35 to 0.94 using data collected directly from nest sites. Model outputs were examined across a 200-year planning horizon using three forest-management scenarios to test the relative impact of model alterations. Results from this procedure illustrated similar trends within and between scenarios using both the original and modified models. Consequently the modified model would be unlikely to alter management decisions under a system where scenarios are evaluated in relation to each other.