PARENT SESSION
Poster Session # 22: Late-Breaking Newsworthy Posters
Friday, August 8 Presentation from 8:30 AM to 10:00 AM. SITCC Exhibit Hall B.
Modelling the habitat distribution of threatened species using California Natural Heritage occurrence data.
Hernandez, Pilar*,1, Fortin, Marie-Josee1, Graham, Catherine2, Master, Lawrence3, Albert, Deborah3, 1 University of Toronto, Toronto, Ontario, Canada2 University of California, Berkeley, California3 NatureServe, Arlington, Virginia
ABSTRACT- A vast amount of digital point occurrence data is becoming available as museum of natural history collections are being geo-referenced and groups such as the Natural Heritage Network actively develop biodiversity databases. The capability for this type of information to improve our knowledge of the distribution of biodiversity is increasingly being realized through the development of modelling procedures that predict species habitat distributions from incomplete locality data. Even though the integrity of these data sets is high since they are regularly assessed by specialists and maintain strict validation procedures, many localities were collected decades and centuries ago and as a result the precision of many of these points is low. The California Natural Diversity Database (CNDDB) developed by the California Natural Heritage Program will be used to evaluate the effects of modeling with occurrence data with large spatial uncertainties. This database contains both precise data from recent field surveys and a compilation of recorded observations where the exact location within a bounding area in unknown. Here habitat distributions for the federally-listed endangered California tiger salamander (Ambystoma californiense) and the federally-listed threatened California gnatchatcher (Polioptila californica) were modelled with multiple logistic regression using subsets of the available point occurrence data selected with increasingly less rigorous inclusion criteria with regards to its spatial uncertainty. Though no apparent effects of using localities with large spatial zones of uncertainty were detected with the available evaluation methods, the exercise highlighted the large amount of variability in the resulting models that can occur with subsets of the same occurrence data set. The use of a local indicator of spatial association (LISA) to assist in the determination of the validity of a predicted occurrence was investigated.
Key words: conservation, species distribution models