Date of Award


Degree Type


Degree Name

Master of Science (Biological Sciences)


School of Natural Sciences


Faculty of Health, Engineering and Science

First Advisor

Dr Eddie Van Etten

Second Advisor

Professor Will Stock


Aim and background

The aim of this study was to reconstruct fire regimes/history of an extensive arid to semi-arid region of Western Australia for the last 40 years by studying temporal and spatial fire scar patterns using remote sensing in a Geographic Information System environment.


A fire history database was constructed, for the years 1973-2012 of an area of Western Australia between the Murchison and Minilya Rivers, of all bushfires greater than 1 ha that were visible from Landsat satellite imagery ranging from 30 – 60m ground resolution. The data was analysed to characterise a general bushfire regime for the region which included: mean and median size of fire scars; range and mean count of bushfire numbers; and the effect that land-systems, vegetation types/structures, land tenure and climate had on the bushfire regime.

The 82 land-systems were grouped into 4 main land types and the 141 vegetation types were grouped into 7 vegetation groups and 7 vegetation structural groups. The percentages of areas burnt were calculated and a correlation coefficient was calculated between total area and total area burnt for each variable within the study area. Land tenure was divided into 3 types and two discrete areas in the northern study area were compared to determine if land tenure had an impact on fire regimes. To assess the effect climate had on the bushfire extent I calculated their correlation to both cumulative rainfall and mean maximum temperature. A generalised linear model was used to determine which climatic variables impacted on the bushfire regime.


A total of 23.8% of the total study area burnt during the study period. The individual fire scars ranged from4000 km2. The frequency of fires ranged from 10-40+ years across the different vegetation types.

The land type that burnt the most extensively in both the southern and northern study areas was Sandplains/coastal (southern — 16.6%; northern — 57%). The land type that burnt the least extensively in the southern study area was Alluvial/plains with eucalypts/claypans group (2.3%), and in the northern study area it was the Hills/mesas group (~1%). The Heathland vegetation (54%) burnt the most extensively in the southern study area and Hummock grassland (68%) in the northern. The northern study area had a greater variety of vegetation groups that burnt than the southern area (northern — 37; southern — 22). The Acacia Shrubland vegetation group contained the greatest number of fire fronts that stopped shortly after burning into this group. Only Shrub and Mallee structural groups burned in the southern study area (grass, tree, samphire shrub, chenopod shrub and bare did not burn); all structural groups except Mallee burnt in the northern study area.

Two regions in the northern study area with structurally similar vegetation showed some differences in fire frequency when grazing activity by feral goats was restricted vs. unrestricted, though there was insufficient data for statistically relevant findings. The variables retained in the generalised linear models of fire extent were; rain in the previous two years to year of fire (positive relationship) and rain in the previous three years to fire (negative relationship) for the northern study area, and number of annual days over 35°C (positive relationship), rain in the year of fire (negative relationship) and rain in the three years prior to fire (positive relationship) in the southern study area.


Bushfire size distribution and frequency are comparable to other regimes in similar climates; however on average they tended to be larger. Infrequent, extensive (>4000 km2) bushfires can skew the statistics since the area burnt contributes to mean values to a greater degree as the individual fires get larger. Such large fires are strongly linked to 2 year antecedent above average rainfall episodes, especially in the northern study area and particularly when pluvial periods are followed by droughts and above-average temperatures. The extent of a bushfire is partially dependent on vegetation type and structure, and land-system. More research needs to be undertaken to see how much of an influence (if any) grazing practices leads to difference in bushfire regimes. Predicted future changes in climate may lead to more frequent and higher intensity bushfires in the study area.

This research furthers the understanding of bushfire regimes in arid Western Australia. It also contributes to expanding current knowledge of bushfires in an arid Acacia shrubland ecosystem and it should improve predictions and management of bushfires in these extensive regions of Australia.


Paper Location