A spatial model of bird abundance as adjusted for detection probability

Authors

  • P. Marcos Gorresen,

  • Garnett P. McMillan,

  • Richard J. Camp,

  • Thane K. Pratt


P. M. Gorresen (mgorresen@usgs.gov) and R. J. Camp, Hawaii Cooperative Studies Unit (PACRC, UH Hilo), USGS Pacific Island Ecosystems Research Center, Kilauea Field Station, Hawaii National Park, HI 96718, USA. – G. P. McMillan, Behavioral Health Research Center of the Southwest, Pacific Inst. for Research and Evaluation, 612 Encino Pl. NE, Albuquerque, NM 87102, USA. – T. K. Pratt, USGS Pacific Island Ecosystems Research Center, Kilauea Field Station, Hawaii National Park, HI 96718, USA.

Abstract

Modeling the spatial distribution of animals can be complicated by spatial and temporal effects (i.e. spatial autocorrelation and trends in abundance over time) and other factors such as imperfect detection probabilities and observation-related nuisance variables. Recent advances in modeling have demonstrated various approaches that handle most of these factors but which require a degree of sampling effort (e.g. replication) not available to many field studies. We present a two-step approach that addresses these challenges to spatially model species abundance. Habitat, spatial and temporal variables were handled with a Bayesian approach which facilitated modeling hierarchically structured data. Predicted abundance was subsequently adjusted to account for imperfect detection and the area effectively sampled for each species. We provide examples of our modeling approach for two endemic Hawaiian nectarivorous honeycreepers: ‘i‘iwi Vestiaria coccinea and ‘apapane Himatione sanguinea.

Ancillary