Content area
Many eastern forest bird populations have declined over the past half-century due to reduced availability of early successional communities as forests in the region have matured following mass cutting in the 19th and early 20th century. Species recovery efforts that aim to increase the area of early successional forest on public and private lands are ongoing. However, such efforts, along with associated avian monitoring programs, are often challenged by an incomplete understanding of the amount and distribution of existing early successional communities across large landscapes comprised of multiple ownerships. This challenge has been driven by a lack of remotely sensed data that accurately identify young forest communities. A compounding problem in studying habitat associations of birds is the low detectability of some species using traditional survey methods. Working with the Ruffed Grouse (Bonasa umbellus), I employed two advancing technologies- autonomous recording units (ARUs) and Light Detection and Ranging (LiDAR)- to mitigate these issues. Using ARU recordings from 1038 unique locations surveyed in Pennsylvania from 2020-23, I used a machine-learned classifier to create drumming grouse detection histories for each survey location. I then used several LiDAR-derived forest structure variables generated from statewide LiDAR collected in 2016-2020 to model grouse occurrence. I detected grouse at 398 of 1038 (naïve occupancy = 38.3%) survey locations. Three models were developed to predict drumming grouse occurrence with (1) vegetative structure metrics, (2) landscape metrics, and (3) a combination of both (1) and (2). My best vegetative structure model included the covariates number of recording days, the height of the 90th percentile of LiDAR returns (quadratic, maximum probability of grouse occurrence at 10.9m), and the percentage of first LiDAR returns at 1 to 5m (quadratic, maximum at 29.9%). The top landscape habitat model contained the terms number of recording days (+), elevation (+), local connectedness (+), percent forest cover (+), percent conifer forest (quadratic, maximum at 9%), longitude (quadratic, maximum at -77.23057), and northern hardwood forest (quadratic, maximum at 80%). Lastly, using my models, I created a series of maps depicting Ruffed Grouse occurrence probability in Pennsylvania as well as where habitat management efforts should be prioritized. At the state-wide level, the map produced with the combined top model depicted the highest probabilities of Ruffed Grouse occurrence in north-central Pennsylvania and the lowest probabilities in the southwest and southeast- the same areas that have experienced the greatest decline in Ruffed Grouse population in the state. The maps of potential management benefit had a similar pattern, showing much of north central Pennsylvania as promising area to implement Ruffed Grouse management. By integrating the information within these maps and predictive models into management decisions, wildlife managers can better inform monitoring and habitat treatments aiding Ruffed Grouse population recovery.