Understanding the importance of fuel structure for fire severity through the lens of remote sensing

Abstract

Recent wildfires in southeast Australia have emphasised the growing challenge of maintaining the severity of fire within tolerable limits of ecosystems. However, there is limited information on what causes variation in remotely sensed fire severity, particularly the role of fuel. Applying remote sensing data to untangle the drivers of fire severity variation can yield insights important to managing fire-prone landscapes. This thesis evaluated remote sensing methods that estimate fire fuel and fire severity and applied this knowledge to a novel investigation of how fuel structure affects fire severity. Initially, the state of fire fuel remote sensing knowledge was analysed through a comprehensive review of the recent literature (Chapter 2). Here, the fuel attributes studied in the remote sensing literature were synthesised against the fuel attributes affecting fire behaviour. A key finding was that, although subcanopy fuel is important for fire behaviour in Australian forests, this contrasted with a relative lack of subcanopy fuel remote sensing research. However, ongoing advancements in Light Detection and Ranging remote sensing demonstrated promise for addressing challenges of subcanopy fuel estimation. Following this, and in the wake of the extensive 2019-20 Australian Black Summer fire season, the performances of several fire severity remote sensing indices were evaluated using field-based estimates of fire severity (Chapter 3). The Relative Burn Ratio was the best-performing index tested. Additional emerging remote sensing methods were evaluated, including LiDAR-derived indices, although they demonstrated empirical shortcomings in the estimation of lower severity fire. Evaluation of fire severity indices in Chapter 2 emphasised the uncertainty and biases that are inherent to remotely sensed fire severity estimates, and the limited physical meaning of these indices. However, when examining drivers of fire severity, it is useful to invoke physical descriptions of fire effects and fire behaviour related explanations for predictor effects. Robust insight on fuel structure effects on fire severity therefore required prior consideration of the likely causes of predictor effects, their relevance to fire behaviour, and the scope for confounding influences (Chapter 4). Analysis of fire severity drivers revealed substantial canopy height effects on remotely sensed fire severity estimates. Additionally, covariation existed between canopy height and environmental variables that are typically used to predict fire severity. This suggests that distinguishing fire behaviour related causes of fire severity variation requires consideration of potentially confounding covariation in canopy height. In response, several novel methodological aspects were introduced in a large-sample analysis of fire severity determinants in southeast Australia (Chapter 5). These included: (1) incorporation of various LiDAR-derived fuel structure metrics as fire severity predictors, following those applied in recent remote sensing studies (Chapter 2); (2) application of the RBR for estimating fire severity (Chapter 3); and (3) standardisation of canopy height effects on fire severity to allow greater causal insight related to fire behaviour (Chapter 4). To summarise key findings, higher fuel density in understorey and mid-canopy fuel layers promoted higher fire severity, and planned and unplanned fire reduced fire severity for 10 and 15 years, respectively. Nonetheless, fuel and fire history were generally uninfluential beyond thresholds of fire weather severity. This thesis developed useful insights in the remote sensing of fire fuel and severity. These new insights, generated from thesis Chapters 2, 3, and 4, were applied to a novel study untangling the contribution of fuel structure to fire severity. The findings of this thesis are applicable to post-fire assessment of fire effects and pre-fire planning for favourable fire severity related outcomes. Show more