A predictive habitat occupancy model of northern bobwhite in the Delmarva Peninsula, USA

Date
2010
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Literature on predicting species presence contains numerous methodological recommendations that reduce bias associated with imperfect detection of individuals, measurement scale of model variables, model selection uncertainty, and spatial autocorrelation. My objective was to incorporate and test recent modeling advances to predict potential habitat occupancy of northern bobwhite. From 15 May-15 August, 2008 and 2009, I conducted repeat-visit surveys at 360 sites within Delaware to sample the presence of bobwhite. I randomly selected half the data to model the scale-dependent relationships of bobwhite presence with metrics of site scale (500 m radius) and landscape scale habitat composition and configuration. At the site scale, bobwhite presence was negatively related to interspersion and juxtaposition of early successional and agriculture habitat, early successional to forest edge density, and agriculture to forest edge density. At the landscape scale, bobwhite presence was negatively related to cohesion of human development within 2.5 km, positively related to cohesion of early successional habitat within 2.0 km, and positively related to percentage of shrub habitat within 1.0 km. The habitat occupancy model fit the validation dataset moderately well with an area under the receiver operating characteristic curve value of 0.631. I applied my habitat occupancy model to map the predicted presence of breeding bobwhite within the Delmarva Peninsula, USA. The modeling results and distribution map will be used to guide future habitat management efforts. I also hope my methodology can serve as a basis for future habitat modeling of bobwhite and other grassland species across their range. I then tested the efficacy of a broadcast caller for estimating density and occupancy of northern bobwhite. Density estimates while using a broadcast caller were higher and increasing passive listening survey duration from 3 to 7 min did not change density estimates. However, increasing survey duration or using a broadcast caller led to a higher detection probability. Use of a broadcast caller is inappropriate for determining density estimates through distance sampling, but may be appropriate for determining occupancy under limited conditions. Finally, I tested if a habitat occupancy model could predict the change in occupancy of Breeding Bird Survey (BBS) routes over time in the Delmarva Peninsula, USA. I used 50 stop BBS data to calculate the percentage of stops occupied by bobwhite during 1992, 1996, 2001 and 2005. I then calculated the probability of breeding-season bobwhite occupancy at each BBS stop for each time period using the occupancy model applied to National Oceanic and Atmospheric Organization Coastal Change and Analysis Program (NOAA CCAP) land cover data. The average change in observed occupancy per year was -8.7% (± SE 1.3%) while the average predicted change in occupancy was 1.0% (± SE 0.8%). Predicted route occupancy was not related to observed route occupancy across sampling periods. Change in predicted occupancy and observed occupancy were also not related. I consider two broad reasons why observed results did not correlate to my predictive model. The first is methodological (low predictive success of model, error from applying model from one year to other years, and imperfect detection of bobwhite during BBS surveys) while the second is biological (slack in habitat configurations, differential habitat use between breeding and nonbreeding season, impact of predation and hunting, and impact of agricultural chemicals). Biologists should use caution when applying a static multi-scale occupancy model to large scale temporal habitat-population processes.
Description
Keywords
Citation