 |
 |
|
Abstract for the 2002 Hawaii Gap Analysis Project report Hawaiian Forest Bird Species Modeling Richard J. Camp, Marcos Gorresen, Bethany L. Woodworth, and Thane K. Pratt. USGS Pacific Island Ecosystem Research Center, P.O. Box 44, Hawaii National Park, HI 96718. Monitoring bird populations remains a critical component in conservation efforts and much effort has been expended on the development of appropriate survey methods. Although many survey techniques have been developed and employed (Thomas 1996, Rosenstock et al. 2002), bird surveys in Hawaii have the distinct benefit of having been established with a consistent and standardized methodology. Distance sampling has been applied to almost all forest bird surveys conducted on the six main islands since the inception of the HFBS in 1976. The procedures presented here utilize developments in distance sampling analysis previously unavailable. Earlier methods for producing density estimates from count sampling involved model selection procedures similar to those described for this study. However, previous methods (e.g., Sauer et al. 1994, Pendleton 1995, Kendall et al. 1996, Fancy 1997) did not directly adjust for sampling covariables (e.g., weather conditions or observer effects) that can significantly bias density estimates. Fancy (1997) detailed a procedure where covariate parameters were quantified in separate analyses and used to adjust density estimates a posteriori. This approach represented a marked improvement over previous methods that did not adjust for sampling conditions. However, this method adjusts the actual distance estimates, and results in interpretative difficulties similar to that of transforming data prior to analysis; i.e., the association of response and transformed predictor variables are not straightforward. Modeling covariate parameters can now be conducted in Distance 4.0 through a log link function that directly adjusts the detection function, thereby avoiding the need of manipulating the actual detection distances (Buckland et al. in preparation). The species-habitat models developed here are an extension of the widely used regression techniques applied elsewhere (see Stauffer 2002, and reference cited therein). The models operationally depend on the assumption that bird populations respond in a predictive manner to combinations of biotic (e.g., vegetation type; canopy closure) and abiotic (e.g., elevation, precipitation) factors. Kriging methods are used to interpolate densities predictions developed with regression modeling. A geographic information systems (GIS) approach permitted interpolation to be applied in a systematic manner across the entire study area. Estimates of population size and variance are automated, reproducible and applicable to other study areas and species. Our aims are to develop and describe methods for species-habitat modeling using bird survey, statistical analyses, GIS techniques, and remote sensing imagery. More specifically, we apply regression and interpolation methods to a case study for the Hawaii Amakihi (Hemignathus virens) utilizing variable circular plot bird survey data and classified Landsat 7 imagery to generate maps of density distribution and a regional population estimate. In addition, we outline topics of further research for producing annual population estimates. |
|
Department of the Interior |
| Pacific Island Ecosystems Research Center 3190 Maile Way, St John Hall, Room 408 Honolulu, (Oahu) HI 96822 Phone: (808) 956-5668 |
This page maintained by: Hawaii Forest Bird Interagency Database Project Kilauea Field Station Building 344, PO Box 44 Hawaii National Park, HI 96818 Phone: (808) 967-7396 ext 252 |
|