Location Central Arnhem Land in the Australian monsoon tropics, a stronghold of traditional Aboriginal culture.
Methods Geographic information system (GIS) and global position systems (GPS) technologies were used to measure spatial and temporal changes in fire patterns over a one decade period in a 100 x 80 km area that included a cluster of Aboriginal settlements and a large uninhabited area. The major vegetation types were mapped by visual interpretation of black and white aerial photography and Landsat-5 TM satellite imagery, aerial reconnaissance and field survey. Fire activity over one decade was based on visual interpretation of georeferenced lattice points overlain on sequences cloud-free Landsat satellite images acquired for the early, mid and late dry seasons. This mapping technique was 90% accurate when checked against helicopter survey immediately after a satellite pass in the late dry season. Actual fire activity in the middle and end of one dry season near an Aboriginal settlement was mapped along a 90 km field traverse and the canopy scorch height was determined by sampling 62 burnt areas beside vehicle tracks.
Results Satellite fire mapping was reliable if the satellite pass followed shortly after a fire event, but the reliability decayed dramatically with increasing time between the fire event and the satellite pass. Thus the satellite mapping provided a conservative index of fire activity. In the early and mid dry season (July) there was only a small amount of landscape burning that was restricted to areas near Aboriginal settlements and this fire activity increased in the late dry season (September-October), when uninhabited areas were also burnt. There was a mean fire index of 3 fire records per decade in the plateau savanna, 2 per decade in the sandstone and 5 per decade in the wet savanna, the dominant vegetation types in the study area. The spatial and temporal variability of Aboriginal burning apparent in the satellite analyses was verified by field traverses. Fires set by Aborigines had low scorch height of tree crowns reflecting low intensity, despite generally occurring late in the dry season.
Conclusions Our findings support the idea that Aboriginal burning created a fine-scale mosaic of burnt and unburnt areas but do not support the widely held view that Aboriginal burning was focused primarily in the early dry season. Aboriginal fires were less frequent than the annual or biennial fires characteristic of European fire regimes.
The European fire regime appears to have triggered a positive feedback cycle while the Aboriginal regime maintained a negative feedback cycle between fire frequency and flammable grass fuels. Allied studies found very low average grass-fuel loads in both managed and unmanaged areas compared to areas under European fire management. The widely advocated management objective of early dry season burning provides one of the few options to control fires once heavy grass fuel loads have become established, however we suggest it is erroneous to characterise such a regime as reflecting traditional Aboriginal burning practices. The preservation of Aboriginal fire management regimes should be a high management priority given the difficulty of breaking the grass-fire cycle once it has been initiated.