Abstract Aim To test the hypothesis that ‘islands’ of fire‐sensitive rain forest are restricted to topographic fire refugia and investigate the role of topography–fire interactions in fire‐mediated
alternative stable state models. Location A vegetation mosaic of moorland, sclerophyll scrub, wet sclerophyll eucalypt forest and rain forest in the rugged, fire‐prone landscapes of south‐west Tasmania, Australia. Methods We used geospatial statistics
to: (1) identify the topographic determinants of rain forest distribution on nutrient‐poor substrates, and (2) identify the vegetation and topographic variables that are important in controlling the spatial pattern of a series of very large fires (> 40,000 ha) that were
mapped using Landsat Thematic Mapper (TM) satellite imagery. Results Rain forest was more likely to be found in valleys and on steep south‐facing slopes. Fires typically burned within highly flammable treeless moorland and stopped on boundaries with less flammable surrounding
vegetation types such as wet sclerophyll forest and rain forest. Controlling for the effect of vegetation, fires were most likely to burn on flats, ridges and steep north‐facing slopes and least likely to burn in valleys and on steep south‐facing slopes. These results suggest
an antagonism between fire and rain forest, in which rain forest preferentially occupies parts of the landscape where fire is least likely to burn. Main conclusions The distribution of rain forest on nutrient‐poor substrates was clearly related to parts of the landscape
that are protected from fire (i.e. topographic fire refugia). The relative flammability of vegetation types at the landscape scale offers support to the proposed hierarchy of fire frequencies (moorland > scrub > wet sclerophyll > rain forest) that underpins
the ecological models proposed for the region. The interaction between fire occurrence and a range of topographic variables suggests that topography plays an important role in mediating the fire–vegetation feedbacks thought to maintain vegetation mosaics in south‐west Tasmania.
We suggest that these fire–topography interactions should be included in models of fire‐mediated alternative stable vegetation states in other fire‐prone landscapes.