Williams-Climate-change-refugia-for-terrestrial-biodiversity_0

3.3.4 Species data
The study generated species distribution models (SDMs) for Australian vertebrates. We
selected species for which adequate samples of presence records (i.e. locations at
which the species has definitely been observed) were available. Species data were
accessed from the Australian Atlas of Living Australia (ALA: http://www.ala.org.au/), the
Centre for Tropical Biodiversity and Climate Change (CTBCC:
https://plone.jcu.edu.au/researchatjcu/research/ctbcc) species data base (Williams et
al. 2010) and from the Queensland Museum (http://www.qm.qld.gov.au/).
Only occurrence records which had been identified to species level were used,
excluding undefined species and reassigning subspecies to species level. Aquatic
species, amphibians, birds and mammals with fewer than five records, and reptiles with
fewer than four records were excluded from our analyses. We chose to retain species
of amphibians, mammals and reptiles with extremely restricted distributions and very
small sample sizes (fewer than 10 records) on the basis that these taxa are of high
conservation concern and that few records may still be adequate to characterise
available environmental conditions within restricted ranges (for details see APPENDIX
3. Species distribution modelling data and model results). For species with adequate
data for modelling, there were 239 mammal, 599 bird, 218 amphibian and 625 reptile
species (Table 3). Vertebrate data were vetted by excluding points that fell into states
or bioregions (Interim Biogeographic Regionalisation for Australia, Version 7;
Environment Australia 2000) in which the species was known not to occur. Species’
known occurrence range was taken from relevant field guides (Menkhorst and Knight
2001, Churchill 2008, Tyler and Knight 2009, Wilson and Swan 2010), online
databases (http://www.arod.com.au/arod/) and from expert opinion.
Table 1: Vertebrate species presence data compiled for the project. ‘Total
records’ is the total number of records for the whole class, ‘mean records’ is the
mean number of records across all species within a class.
Class
3.3.5 Species distribution modelling
Total
records
14 Climate change refugia for terrestrial biodiversity
Mean
records
amphibians 151 943 394
birds 5 873 874 9806
mammals 355 580 1451
reptiles 249 551 664
Species distribution models incorporating baseline climate data at 0.01 degree (~1x1
km) resolution and species occurrences were created using the Maxent package
(Phillips et al. 2006). Maxent uses presence-only data to statistically relate distribution
records to environmental variables and uses the principle of maximum entropy to
develop the best model fit. Studies have compared techniques across species and
biomes to rank their performance (Elith et al. 2006, Guisan et al. 2007), and although
different techniques vary in their performance, there is generally more variation in
performance across species within technique than across techniques (Guisan et al.
2007). However, detailed comparisons of techniques found that the newer methods,
and in particular Maxent, consistently outperformed other techniques (Elith et al. 2006).
SDMs are increasingly being used to predict species responses to anthropogenic
climate change. There is evidence to show that this approach is likely to reflect realistic
changes, such as the documented shifts in species distributions in recent times
(Thomas and Lennon 1999, Parmesan and Yohe 2003, La Sorte and Thompson 2007,
Maclean et al. 2008). In addition, the few new phenotypes found in the Pleistocene
fossil record corresponding with rapid temperature shifts suggests that species are