Burn severity assessment is critical for understanding the pattern of post-fire vegetation recovery and ecosystem resilience. Previous studies proposed various field criteria (e.g., Composite Burn Index (CBI)) to quantify burn severity from strata level to total site level, yet suffering from surveyors' subjective interpretation across site conditions. High-resolution passive remote sensing allows for more objective assessment based on the strong relation between fire damages and spectral features. Importantly, burn severity generally characterizes differences between forest overstory and understory layers due to their discrepancies in vegetation structures and environmental conditions. Spatially explicit mapping of strata-level burn severity is vital for post-fire forest management and ecological function evaluation. Until now, almost no available spaceborne remote sensing method can concurrently assess overstory and understory burn severity over heterogeneous forests. In this study, we proposed a Hybrid Composite Burn Index (HCBI) that comprehensively indicates the fire-induced spectral and structural changes along the vertical profile by integrating spectral and waveform attributes from both active and passive remote sensing data. Firstly, we introduced two remote sensing-based rating factors namely relative spectral change (RSC) and relative waveform change (RWC), and established HCBI through weighting and scoring the rating factors. Subsequently, we evaluated the effectiveness and generality of overstory, understory, and total site HCBI based on various pairs of simulated remote sensing data of pre-fire and post-fire forest scenes. Thirdly, we derived the spatially discontinuous maps of overstory, understory, and total site HCBI of the Xiushan fire site in Great Xing'an Mountain using WorldView-2 (WV-2) multispectral imagery and Global Ecosystem Dynamics Investigation (GEDI) full-waveform LiDAR data. Finally, we produced wall-to-wall HCBI maps using multiple predictive variables from pre-fire and post-fire Sentinel-2 MultiSpectral Instrument (MSI) images with a Random Forest (RF) algorithm. The predicted overstory, understory, and total site HCBI were validated by field-surveyed CBI. The assessment results based on simulation data showed HCBI was sensitive to the fire damage regardless of burn severity levels and vegetation cover levels (R of larger than 0.97 and RMSE of <0.15). The prediction results based on RF models achieved reliable HCBI of the total site, overstory, and understory levels (R of around 0.85 and RMSE of around 0.30). We found the wall-to-wall maps of HCBI captured the subtle horizontal and vertical variation even in the case of understory burn alone. We concluded that the newly proposed HCBI can advance the remote sensing-based assessment of stratified burn severity, offering opportunities for making fine-scale forest management decisions.

中文翻译：

---

大兴安岭星载光谱波形属性分层烧伤严重程度评估

烧伤严重程度评估对于了解火后植被恢复模式和生态系统恢复能力至关重要。先前的研究提出了各种现场标准（例如综合烧伤指数（CBI））来量化从地层到整个场地水平的烧伤严重程度，但受到测量员对场地条件的主观解释的影响。高分辨率被动遥感可以根据火灾损失和光谱特征之间的密切关系进行更客观的评估。重要的是，烧伤严重程度通常表征森林上层和林下层之间的差异，因为它们在植被结构和环境条件方面存在差异。地层烧伤严重程度的空间明确绘图对于火后森林管理和生态功能评估至关重要。到目前为止，几乎没有可用的星载遥感方法可以同时评估异质森林的林上和林下烧毁严重程度。在本研究中，我们提出了混合复合燃烧指数（HCBI），通过集成主动和被动遥感数据的光谱和波形属性，综合指示火灾引起的沿垂直剖面的光谱和结构变化。首先引入相对光谱变化（RSC）和相对波形变化（RWC）两个基于遥感的评级因子，并通过对评级因子进行加权和评分建立HCBI。随后，我们根据火灾前和火灾后森林场景的多对模拟遥感数据评估了林下、林下和全站点 HCBI 的有效性和通用性。第三，利用WorldView-2（WV-2）多光谱影像和全球生态系统动力学调查（GEDI）全波形，得出了大兴安岭秀山火灾现场的上层、林下和全点HCBI的空间不连续图。激光雷达数据。最后，我们使用来自火灾前和火灾后 Sentinel-2 多光谱仪器 (MSI) 图像的多个预测变量以及随机森林 (RF) 算法生成了全面的 HCBI 地图。预测的上层、下层和总场地 HCBI 通过现场调查的 CBI 进行了验证。基于模拟数据的评估结果表明，无论烧伤严重程度和植被覆盖水平如何，HCBI 对火灾损失都很敏感（R 大于 0.97，RMSE <0.15）。基于 RF 模型的预测结果实现了总场地、上层和下层水平的可靠 HCBI（R 约为 0.85，RMSE 约为 0.30）。我们发现，即使在仅林下燃烧的情况下，HCBI 的全面地图也捕捉到了微妙的水平和垂直变化。我们的结论是，新提出的 HCBI 可以推进基于遥感的分层烧伤严重程度评估，为制定精细的森林管理决策提供机会。