September 2026: Post by Dr Felicity Charles
For most Australian ecosystems, understanding the best way to manage them requires knowledge of fire history. Yet for so many areas, there is no detailed and accurate fire mapping available. We may look to satellite-derived fire mapping products like MODIS, but these are often too coarse to capture small fires relevant to management, or they miss low intensity fires which burn only understorey vegetation.
My team and I worked across public land with access to on-ground mapped fire history data, and private properties with often only verbal accounts of fire history. We realised if we could fit a model that showed the relationship between fire history data from the public estate and satellite imagery, we could project fire history estimates to areas outside the public estate and create better estimates of landscape-scale fire history.
For most Australian ecosystems, understanding the best way to manage them requires knowledge of fire history. Yet for so many areas, there is no detailed and accurate fire mapping available. We may look to satellite-derived fire mapping products like MODIS, but these are often too coarse to capture small fires relevant to management, or they miss low intensity fires which burn only understorey vegetation.
My team and I worked across public land with access to on-ground mapped fire history data, and private properties with often only verbal accounts of fire history. We realised if we could fit a model that showed the relationship between fire history data from the public estate and satellite imagery, we could project fire history estimates to areas outside the public estate and create better estimates of landscape-scale fire history.
A management burn in south-east Queensland (photo: Felicity Charles)
We soon found that simple linear regression models didn’t improve accuracy of fire frequency estimates. So this turned into a 2-year project which harnessed the relationship between fire, climate and environmental attributes. We used a novel application of species distribution modelling to produce a generalisable workflow to improve the accuracy of fire history estimates derived from satellite imagery.
A generalisable workflow to improve the accuracy of fire history estimates from satellite imagery.
Using this innovative species distribution modelling approach and leveraging the relationship between satellite-derived fire data and other variables – fire history data from public estates, climate, terrain, and vegetation productivity – our workflow could be applied to improve fire history estimates wherever fire data are available. This offers several benefits for fire history applications:
- Improving landscape-scale fire history estimates for areas which are unmapped or incorrectly mapped as unburnt due to undetectable burnt areas,
- Improving estimates of satellite-derived fire history where discrepancies exist with data produced from on ground mapping,
- Understanding whether the land has burnt recently or not,
- Providing more in-depth information on the fire regime, such as the number of times a given part of the land has burnt over a period of time (fire frequency).
Our customisable modelling workflow includes accessing and reformatting data to calculate a fire history metric of interest, modelling the relationship between fire history and other variables, and producing spatial predictions for the fire history metric of interest. Our modelling workflow performed well in both fire-prone and fire-sensitive vegetation in a case study of eastern Australia – showcasing its applicability across diverse ecological contexts. The figure above compares fire frequency estimates for 1987 to 2023 from two observed data sources – satellite-derived imagery and on-ground mapped public estate data – against spatial predictions from a generalised linear model (GLM) and generalised additive model (GAM).
For more information, check out the full paper:
Charles FE, Reside AE, Moss PT, Smith AL (2026). Integrating public land fire data and satellite imagery improves fire frequency estimates across the landscape. International Journal of Wildland Fire 35, WF25076.
The customisable code are available on GitHub along with the spatial prediction rasters of fire frequency from 1987 to 2023 for southeast Queensland, Australia.