Spatially continuous canopy height maps of forested ecosystems of Canada

This dataset contains two canopy height maps from forested ecosystems of Canada at 250m spatial resolution — one using information from the spaceborne LiDAR GEDI, and the other from ICESat-2. GEDI and ICESat-2 are particular in acquiring canopy height information in Canada — the former providing more accurate information of vegetation, yet not reaching full coverage in Canada, whilst the latter is not specifically designed to provide vegetation information but has a global coverage. We created wall-to-wall maps using ATL08 LiDAR product from the ICESat-2 satellite, and GEDI L2A from GEDI. The data were download for the mid growing season (June and August 2020). Points were filtered regarding solar background noise and atmospheric scattering, totaling 208,554 points from ICESat-2, and 1,249,354 points for GEDI after filtering and point thinning. These points were associated with 14 ancillary variables primarily corresponding to structure information, such as seasonal Sentinel-1 VV and VH polarization, seasonal Sentinel-2 red and NIR bands, and annual PALSAR-2 HH and HV polarization. Afterwards, the random forest algorithm was used to extrapolate LiDAR observations and develop regression models with the abovementioned spatially continuous variables. GEDI had a better performance than ICESat-2 with a mean difference (MD) of 0.9 m and 2.9 m in relation to ALS data used for validation, and a root mean square error (RMSE) of 4.2 m and 5.2 m, respectively. However, as both GEDI and ALS have no coverage in most of the hemi-boreal forests, ICESat-2 captures the tall canopy heights expected for these forests better than GEDI.

本数据集包含两份加拿大森林生态系统的冠层高度图，空间分辨率为250米。其中一份数据采用星载激光雷达GEDI的观测信息，另一份则源自ICESat-2卫星。GEDI与ICESat-2在加拿大冠层高度信息获取方面各有特点：前者植被信息反演精度更高，但在加拿大境内未实现全覆盖；后者虽非专为植被信息监测设计，却具备全球覆盖能力。本研究利用ICESat-2卫星的ATL08激光雷达产品以及GEDI的L2A数据，生成了全域无缝冠层高度图。数据采集时段为2020年生长季中期（6月与8月）。我们针对太阳背景噪声与大气散射效应对原始点云数据进行了滤波处理，经滤波与点云抽稀后，ICESat-2剩余有效点共208554个，GEDI剩余有效点共1249354个。将上述有效点与14项主要反映地表结构信息的辅助变量进行关联，这些变量包括季节性哨兵1号（Sentinel-1）的VV、VH极化数据，季节性哨兵2号（Sentinel-2）红波段与近红外（NIR）波段数据，以及年度PALSAR-2的HH、HV极化数据。随后，本研究采用随机森林算法，以上述空间连续型变量为自变量，对激光雷达观测数据进行空间外推并构建回归模型。经用于验证的机载激光扫描（ALS，Airborne Laser Scanning）数据检验，GEDI的模型表现优于ICESat-2：两者的平均差（MD）分别为0.9米与2.9米，均方根误差（RMSE）分别为4.2米与5.2米。但由于GEDI与ALS在多数半寒带森林区域均无观测覆盖，ICESat-2能够更准确地反演该类森林预期的高大冠层高度，表现优于GEDI。