Author Identifier
Ana do Carmo Carvalho: http://orcid.org/0000-0002-5297-2157
Date of Award
2026
Document Type
Thesis
Publisher
Edith Cowan University
Degree Name
Doctor of Philosophy
School
School of Science
First Supervisor
David Blake
Second Supervisor
Eddie Van Etten
Third Supervisor
Pierre Horwitz
Abstract
The Northern Jarrah Forest in Southwestern Australia (SWA), part of a global biodiversity hotspot, is home to fire-sensitive tree species such as marri (Corymbia calophylla) and jarrah (Eucalyptus marginata). As climate change intensifies drought and extreme fire events, understanding the drivers of fire severity is increasingly critical. This thesis analyses fire regimes within the Mundaring drinking water catchment to better understand the spatiotemporal complexity of fire severity in relation to fire history and environmental factors.
The thesis commences with a review of three Fire History Databases (FHDs) used in SWA, highlighting significant inconsistencies and data gaps across time and space. It concludes that one of them, the Landgate FHD outperforms the FHDs from two other government departments and is the most reliable source for fire regime studies. Before it could be used for analytical purposes, fire history records needed to be validated, with fire dates and perimeter boundaries identified as the most critical elements requiring validation.
The validation process uses a semi-automated method based on Landsat imagery, detecting and dating 340 historical fire events from 1990 to 2021. The approach applies multiple spectral indices to analyse post-fire canopy responses and employs random forest classification to map f ire severity at 30 meters resolution using a Python script and the Google Earth Engine API.
With this refined and validated dataset, the influence of fire history variables, namely time since f ire (TSF), penultimate fire severity, number of previous fires, and season, on the severity of subsequent fires, is investigated. Results indicate complex, non-linear relationships, with TSF and penultimate fire severity emerging as the most influential predictors. A spatio-temporal generalized mixed model (link = "log") shows that TSF over 25 years is associated with reduced fire severity.
A subsequent analysis demonstrates that even after accounting for environmental factors, TSF remains the strongest driver of recent fire severity. A regional analysis using K-means clustering and Principal Component Analysis identifies four distinct fire regimes in the Mundaring region, shaped by varying contributions from TSF, fire season, slope, and actual evapotranspiration.
These findings underscore the critical need for accurate fire history data and location-specific strategies to enhance fire management under a changing climate. They also highlight the importance of a reliable fire recording framework in SWA to support research on fire frequency, extent, and severity. Furthermore, difficulties in maintaining short fire intervals (< 6 years) have led to large forest areas with a TSF of 10–16 years, a period associated with a higher likelihood of severe fires, particularly following spring burns (August-September).
Recommended Citation
do Carmo Carvalho, A. (2026). Study of fire regimes in Southwestern Australia using geospatial techniques. Edith Cowan University. https://doi.org/10.25958/encv-9549