Abstract for presentation at Biodiversity Extinction Crisis Conference - A Pacific Response

Testing Biodiversity Toolkits - How well do they predict vertebrate species richness?

  • Anthony Weinberg, Science and Research Division, NSW Department of Primary Industries, Australia
  • Dr Rodney Kavanagh, Science and Research Division, NSW Department of Primary Industries, Australia
  • Dr Bradley Law, Science and Research Division, NSW Department of Primary Industries, Australia
  • Dr Trent Penman, Science and Research Division, NSW Department of Primary Industries, Australia
  • In the last five years, State Government agencies across Australia have rapidly developed Biodiversity Toolkits as devices for estimating the locations of species-rich areas. Biodiversity Toolkits work by combining a number of key vegetation and landscape attributes into a single index which, when properly constructed, represents the habitat requirements of a broad range of species. However, to date there has been insufficient testing of the ecological basis that underlies Biodiversity Toolkits. We compared the predictions from four Toolkits, based on measurements collected at 120 sites throughout the South West Slopes (NSW), against an existing data-set of vertebrate species collected at the same sites. Using Spearman’s rank correlations, we assessed whether sites with ‘high’ toolkit scores corresponded to sites with the greatest vertebrate species richness. Using GLMs and sensitivity analyses, we looked at ways to improve Toolkits by adding attributes and adjusting weightings. Overall, we found that Biodiversity Toolkits gave an inadequate representation of vertebrate species richness. While their performance was better in remnant vegetation, Toolkits were very poor in representing vertebrate species in planted sites because the types of attributes included were unsuitable and the use of reference sites (mature and unmodified vegetation communities) were inappropriate for assessing vertebrate species in plantings. Biodiversity Toolkits gave better predictions for fauna groups which depend on structurally complex vegetation, such as woodland-dependant birds, arboreal mammals and reptiles. Conversely, bats and non-woodland dependant birds were poorly represented. We identified five key vegetation and landscape attributes, some of which were not included in some of the Toolkits (e.g. presence of water), and show how specific adjustments to the weightings of certain attributes can improve their overall performance. This research is intended to inform Toolkit developers and users on their limitations and reliability as well as practical ways of improving their design.

    Conference Organiser - ICMS Pty Ltd