Linking wildland fuel combustibility, emission factors, biochemistry and fire behaviour using imaging spectroscopy

Remote sensing to quantify attributes of vegetation necessary for predicting regional-scale wildland fire danger lags significantly behind other applications in the wildfire management domain, largely due to insufficient connection between what is remotely sensed and what attributes are important for fire behaviour. To address this gap, we introduce and test an experimental methodology that integrates imaging spectroscopy with laboratory-based fire behaviour experiments, biochemical and calorimetry measurements, and greenhouse gas and particulate emissions sampling to link fuel bed spectra with fire behaviour, burn severity and fuel consumption. In a pilot study, three fuel types (eucalypt canopy branchlets, eucalypt litter, and annual ryegrass) were treated to produce two levels of combustibility (high, low) and burned by free-spreading fire under controlled laboratory conditions. Pre- and post-burn hyperspectral images (400–2500 nm) were collected during 26 fire experiments, alongside measurements of fuel moisture content, combustion efficiency, rate of spread, biochemistry, calorimetry, and emissions of CO2, CO, CH4, PM2.5. This dataset sumarizes this information.

在野火管理领域，用于量化预测区域尺度野外火灾危险所需植被属性的遥感技术，其发展进度显著落后于该领域内的其他应用方向，核心症结在于遥感获取的信息与火灾行为关键影响植被属性之间缺乏足够的关联机制。为填补这一研究空白，本研究提出并验证了一套实验方法：将成像光谱学（imaging spectroscopy）与实验室火灾行为实验、生物化学及量热学测量、温室气体与颗粒物排放采样相结合，以此建立燃料床光谱与火灾行为、燃烧烈度及燃料消耗之间的关联。在一项预实验中，研究人员选取3类燃料（桉树冠部细枝、桉树枯落物及一年生黑麦草），通过人工处理使其呈现两种可燃性水平（高、低），并在可控实验室条件下开展自由蔓延火势燃烧实验。在全部26次火灾实验中，研究人员均采集了燃烧前后的高光谱图像（hyperspectral images，400–2500 nm），并同步测量了燃料含水率、燃烧效率、火势蔓延速率、生物化学指标、量热学参数以及CO₂、CO、CH₄、PM2.5的排放数据。本数据集即汇总了上述全部实验相关信息。