Occupancy Models of Nesting-Season Habitat Associations of Red-Shouldered Hawks in Central Minnesota

ABSTRACT Red-shouldered hawks (Buteo lineatus) are a species of special conservation concern in much of the Great Lakes region, and apparent population declines are thought to be primarily due to habitat loss and alteration. To evaluate red-shouldered hawk-habitat associations during the nesting season and at the landscape scale, we conducted repeated call-broadcast surveys in central Minnesota, USA, across 3 landscapes that represented a range of landscape conditions as a result of differing management practices. In 2004, we conducted repeated call-broadcast surveys at 131 locations in 2 study areas, and in 2005, we surveyed 238 locations in 3 study areas. We developed models relating habitat characteristics at 2 spatial scales to red-shouldered hawk occupancy and assessed support for these models in an information-theoretic framework. Overall, a small proportion of nonforest (grass, clear-cut area, forest <5 yr old), and a large proportion of mature deciduous forest (>40 yr old), had the strongest association with red-shouldered hawk occupancy (proportion of sites occupied) at both spatial scales. The landscape conditions we examined appeared to contain a habitat transition important to red-shouldered hawks. We found, in predominately forest landscapes, the amount of open habitat was most strongly associated with red-shouldered hawk occupancy, but in landscapes that included slightly less mature forest and more extensive open habitats, the extent of mature deciduous forest was most strongly associated with red-shouldered hawk occupancy. Our results suggested that relatively small (<5 ha) patches of open habitat (clear-cuts) in otherwise forested landscapes did not appear to influence red-shouldered hawk occupancy. Whereas, in an otherwise similar landscape, with smaller amounts of mature deciduous forest and larger (>15 ha) patches of open habitat, red-shouldered hawk occupancy decreased, suggesting a threshold in landscape composition, based on both the amount of mature forest and open area, is important in managing forest landscapes for red-shouldered hawks. Our results show that during the nesting season, red-shouldered hawks in central Minnesota occupy at similar rates landscapes with different habitat compositions resulting from different management strategies and that management strategies that create small openings may not negatively affect red-shouldered hawk occupancy.     

Red-Shouldered Hawk Occupancy Surveys in Central Minnesota,USA

ABSTRACT Forest-dwelling raptors are often difficult to detect because many species occur at low density or are secretive. Broadcasting conspecific vocalizations can increase the probability of detecting forest-dwelling raptors and has been shown to be an effective method for locating raptors and assessing their relative abundance. Recent advances in statistical techniques based on presence—absence data use probabilistic arguments to derive probability of detection when it is < 1 and to provide a model and likelihood-based method for estimating proportion of sites occupied. We used these maximum-likelihood models with data from red-shouldered hawk (Buteo lineatus) call-broadcast surveys conducted in central Minnesota, USA, in 1994–1995 and 2004–2005. Our objectives were to obtain estimates of occupancy and detection probability 1) over multiple sampling seasons (yr), 2) incorporating within-season time-specific detection probabilities, 3) with call type and breeding stage included as covariates in models of probability of detection, and 4) with different sampling strategies. We visited individual survey locations 2–9 times per year, and estimates of both probability of detection (range = 0.28-0.54) and site occupancy (range = 0.81-0.97) varied among years. Detection probability was affected by inclusion of a within-season time-specific covariate, call type, and breeding stage. In 2004 and 2005 we used survey results to assess the effect that number of sample locations, double sampling, and discontinued sampling had on parameter estimates. We found that estimates of probability of detection and proportion of sites occupied were similar across different sampling strategies, and we suggest ways to reduce sampling effort in a monitoring program.