Predictive Modelling of Black Oystercatcher (Haemotopus Bachmani) Breeding Pair Occurrence and Prey Abundance in Haida Gwaii, British Columbia

Abstract

This thesis presents a Species Distribution Model for prediction of Black Oystercatcher breeding pair occurrence in Haida Gwaii, BC, at a landscape scale. First, a quantitative fetch index was developed and found to be a useful proxy for Black Oystercatcher prey abundance based on sampled data at 50 intertidal stations. Second, Boosted Regression Trees were used to model breeding pair occurrence compiled from two available surveys (2005 and 2010) in response to several predictors selected a priori. In order of influence, breeding pairs occurred far from the treeline, on small islands, at moderate wave exposures, at moderate intertidal area, on bedrock or gravel shoreline types, and on islands without rats. The ‘best’ model had excellent discrimination (AUC=0.89) and did not include predictors derived from in situ biological data. Models including only ShoreZone attributes performed relatively poorly. Discrimination on independent data were reasonable (AUC>0.8), although poorest when data were split temporally.