Using handheld mobile laser scanning to quantify fine-scale surface fuels and detect changes post-disturbance in Northern California forests

The understory plays a critical role in the disturbance dynamics of forest ecosystems, as it can influence wildfire behavior. Unfortunately, the 3D structure of understory fuels is often difficult to quantify and model due to vegetation and substrate heterogeneity. LiDAR remote sensing can measure changes in 3D forest structure more rapidly, comprehensively, and accurately than manual approaches, but a remote sensing approach is more frequently applied to the overstory compared to the understory. Here we evaluated the use of handheld mobile laser scanning (HMLS) to measure and detect changes in fine-scale surface fuels following wildfire and timber harvest in Northern Californian forests, USA. First, the ability of HMLS to quantify surface fuels was validated by destructively sampling vegetation below 1 m with a known occupied volume within a 3D frame and comparing destructive-based volumes with HMLS-based occupied volume estimates. There was a positive linear relationship (R2 = 0.72) between volume estimates, and occupied volume estimated from 1-cm voxels had the best relationship with measured biomass (R2 = 0.46, RMSE = 50.76 g, p < .0001) compared to larger voxel sizes. Next, HMLS was used to scan forest plots where wildfire or timber harvest had occurred, producing bi-temporal (pre and post) structural measurements. Plot scans were voxelized and the volume occupied by surface fuels was extracted and quantified. Changes in plot-level HMLS estimates of surface fuels were compared to data collected with a standardized manual field protocol to quantify plot-level dead and uprooted vegetation (Brown’s transects). Both HMLS and Brown’s transects detected a similar decrease in surface fuels post-wildfire. However, removal of ground voxels for the HMLS analysis revealed the regrowth of live vegetation one-year post-fire that was not captured by Brown’s transects. Neither remote sensing nor field approaches detected any changes in fine-scale surface fuels post-logging. HMLS can be a valuable tool for land stewards to rapidly quantify understory vegetation, especially following disturbance. An accurate assessment of understory vegetation is crucial for management plans to reduce wildfire risk and both live and dead fuels might not be captured fully post-wildfire using non-remote sensing approaches.

Methods
Data were collected in a few different ways. 3D frame data were collected by scanning a 3D frame with a handheld mobile laser scanner (HMLS) and then destructively sampling of the vegetation inside. The scans were processed by the scanner's software (GeoSLAM, SLAM algorithm), and the vegetation samples were oven dried to get dry mass measurements. Plot-level data were collected at 11.3 m radius circle plots at 2 locations across 3 time periods, lidar scans were taken with the HMLS and Brown's data were collected using the standard Brown's transect protocol. Brown's data were processed to extract estimates of fuel mass per area for each plot. All of the lidar scans taken with the HMLS (both frame and plot scans) were further processed in Lidar360, CloudCompare, and R with the lidR package to clip scans to the frame/plot boundary, height normalize, and voxelize the scans. Frame scans were voxelized at 4 different voxel sizes (1, 5, 10, and 25 cm), while plot scans were all voxelized at 1 cm voxel size.

林下植被（understory）在森林生态系统的干扰动态中发挥着关键作用，因其能够影响野火行为。受植被与基质异质性的限制，林下可燃物的三维结构往往难以量化与建模。激光雷达（LiDAR）遥感可较人工方法更快速、全面且精准地测量森林三维结构的变化，但相较于林下植被，遥感手段更多被应用于林冠层的相关研究。本研究针对美国加利福尼亚北部森林，评估了手持式移动激光扫描（HMLS）在野火与木材采伐后，对精细尺度地表可燃物变化的测量与检测效果。首先，本研究通过在三维框架内对1米以下植被开展破坏性采样（已知其框架内占据体积），并将破坏性采样得到的体积与HMLS估算的占据体积进行对比，验证了HMLS量化地表可燃物的能力。体积估算值之间呈现显著正线性相关关系（决定系数R²=0.72）；相较于更大尺寸的体素（voxel），基于1厘米体素估算的占据体积与实测生物量的相关性最优（R²=0.46，均方根误差（RMSE）=50.76 g，p < 0.0001）。随后，本研究使用HMLS对发生过野火或木材采伐的森林样地进行扫描，获取了双时相（灾前与灾后）的结构测量数据。对样地扫描数据进行体素化处理，并提取、量化地表可燃物的占据体积。将HMLS估算的样地尺度地表可燃物变化数据，与采用标准化人工野外规程采集的样地尺度死亡与倒伏植被数据（布朗样带（Brown’s transects））进行对比。HMLS与布朗样带均检测到野火后地表可燃物出现相似程度的下降。但在HMLS分析中移除地面体素后，结果显示野火发生一年后存活植被的再生情况，而这一信息未被布朗样带捕捉到。无论是遥感手段还是野外调查方法，均未检测到采伐后精细尺度地表可燃物的变化。HMLS可成为土地管理者快速量化林下植被的有效工具，尤其在干扰事件发生后。精准评估林下植被对于降低野火风险的管理规划至关重要，且采用非遥感手段时，野火后的存活与死亡可燃物可能无法被完全捕捉。

研究方法
数据采集采用多种方式。三维框架数据通过使用手持式移动激光扫描（HMLS）扫描三维框架，随后对框架内植被进行破坏性采样获取。扫描数据通过扫描仪自带软件（GeoSLAM，同步定位与地图构建（SLAM）算法）进行处理，植被样本经烘箱烘干以获取干重测量值。在2个地点的3个时间周期内，于半径11.3米的圆形样地采集样地尺度数据：使用HMLS开展激光雷达扫描，并采用标准布朗样带规程采集布朗样带数据。对布朗样带数据进行处理，以提取每个样地的单位面积可燃物质量估算值。所有HMLS获取的激光雷达扫描数据（包括框架扫描与样地扫描）均通过Lidar360、CloudCompare软件以及R语言的lidR包进行进一步处理，包括将扫描数据裁剪至框架/样地边界、进行高度归一化以及体素化处理。框架扫描数据采用4种不同尺寸的体素（1、5、10和25厘米）进行体素化，而样地扫描数据均采用1厘米体素进行体素化。