SEARCH

SEARCH BY CITATION

Keywords:

  • Australian Wet Tropics;
  • historical refugia;
  • palaeovegetation modelling;
  • Pleistocene refugia;
  • Quaternary climate fluctuations;
  • vegetation modelling

Abstract

Aim  We created spatially explicit models of palaeovegetation stability for the rain forests of the Australia Wet Tropics. We accounted for the climatic fluctuations of the late Quaternary, improving upon previous palaeovegetation modelling for the region in terms of data, approach and coverage of predictions.

Location  Australian Wet Tropics.

Methods  We generated climate-based distribution models for broad rain forest vegetation types using contemporary and reconstructed ‘pre-clearing’ vegetation data. Models were projected onto previously published palaeoclimate scenarios dating to c. 18 kyr bp. Vegetation stability was estimated as the average likelihood that a location was suitable for rain forest through all climate scenarios. Uncertainty associated with model projections onto novel environmental conditions was also tracked.

Results  Upland rain forest was found to be the most stable of the wet forest vegetation types examined. We provide evidence that the lowland rain forests were largely extirpated from the region during the last glacial maximum, with only small, marginally suitable fragments persisting in two areas. Models generated using contemporary vegetation data underestimated the area of environmental space suitable for rain forest in historical time periods. Model uncertainty resulting from projection onto novel environmental conditions was low, but generally increased with the number of years before present being modelled.

Main conclusions  Climate fluctuations of the late Quaternary probably resulted in dramatic change in the extent of rain forest in the region. Pockets of high-stability upland rain forest were identified, but extreme bottlenecks of area were predicted for lowland rain forest. These factors are expected to have had a dramatic impact on the historical dynamics of population connectivity and patterns of extinction and recolonization of dependent fauna. Finally, we found that models trained on contemporary vegetation data can be problematic for reconstructing vegetation patterns under novel environmental conditions. Climatic tolerances and the historical extent of vegetation may be underestimated when artificial vegetation boundaries imposed by land clearing are not taken into account.