Data for article: Robust characterization of forest structure from airborne laser scanning – a systematic assessment and sample workflow for ecologists

This is a collection of scripts and research data to assess the robustness of forest structure characterization from airborne laser scanning (ALS). It accompanies the article "Robust characterization of forest structure from airborne laser scanning – a systematic assessment and sample workflow for ecologists". The entire collection of data and scripts has been anonymized to conform to double-blind peer review. In the article, we assess the derivation of canopy height models (CHMs) from point cloud data, how sensitive CHM algorithms are to point cloud degradation (pulse density thinning, large scan angles, loss of higher-order returns) and how uncertainties and biases propagate to commonly used forest structure metrics. In addition, we provide a standardized processing pipeline in R to convert point clouds into CHMs. The main data source for this study are ALS point clouds from nine Australian research sites belonging to the Terrestrial Ecosystem Research Network (TERN, 5 km x 5 km extent each). The underlying data can be found here: https://portal.tern.org.au/metadata/TERN/4ff0b4c9-cfa0-4d09-9520-b5402adc583f. For one site (Robson Creek), we also used field data to assess the sensitivity of aboveground biomass estimates to ALS point cloud characteristics. Data are available here: https://portal.tern.org.au/metadata/supersite.174. To characterize climatic/environmental differences between sites, we used climatic data from the CHELSA/BIOCLIM+ climatology 1981-2010 (Brun et al. 2022: Global climate-related predictors at kilometer resolution for the past and future. Earth System Science Data, 14(12), 5573–5603. https://doi.org/10.5194/essd-14-5573-2022; Karger et al. 2017: Climatologies at high resolution for the earth's land surface areas. Scientific Data, 4(1), 170122. https://doi.org/10.1038/sdata.2017.122). We note that the enormous size of the full set of manipulated point clouds (original + thinned + individual flightlines: ~400 GB) and the derived raster products (~200 GB) by far exceeds limits on data storage in Zenodo. However, all analyses can be recreated from scratch from the openly available data and the R code in this repository. In addition, we include derived products for the nine study sites that allow to replicate results in the main text without any point cloud processing (CHMs and other rasters across thinned point clouds + summary statistics). The different data layers are: 01_rscripts.zip: contains a sample script to test the processing pipeline (test.processing.R) as well as a collection of helper functions (ALS_processing_helperfunctions_v40.R); the script can be run directly after unzipping the folder, but an installation of LAStools (https://rapidlasso.de) is necessary (path_lastools = "PATH/TO/LASTOOLS/BIN"); we note that the script was developed on Windows PCs, its application with the recent Linux distribution of LAStools has not yet been tested contains the full set of scripts necessary to reproduce the analyses, including point cloud manipulations and derivation of CHMs from the raw data (create.CHMs.R) as well as the overall robustness analysis (analyze.CHMs.R); to replicate the processing of the raw point clouds step by step, .laz files should be downloaded from the TERN repository (cf. citation above) and placed in a "data" folder, with subfolders for each site and with the same naming conventions as in this repository (e.g., "/data/Alice Mulga") 02_reference.zip contains reference digital surface models (DSMs), canopy height models (CHMs) and digital terrain models (DTMs) for all nine TERN sites, based on the original ALS point clouds note that these reference layers are produced with the "CHMhighest" algorithm, which provides an easily interpretable canopy description as long as pulse densities are high (>= 20 shots per squaremetre) 03_climate.zip contains site coordinates contains the climate layers from the CHELSA climatology (cf. citation above, only used to evaluate climatic ranges of sites) 04_robson_additional.zip contains biomass estimates for Robson Creek contains shapefiles for large trees at Robson Creek (only used for visualization purposes) 05_downsampling_pulse_[Site name].zip [Site name] is a stand-in for the nine TERN sites (e.g., "Alice Mulga.zip", "Credo.zip", etc.) contains the data necessary to reproduce results in the main text of the study, i.e. DSMs, CHMs, and DTMs for all nine TERN sites, and at different pulse density levels (from 16 down to 0.5 laser shots per squaremetre) also contains calculated summary statistics for each site All zip files should be extracted into the same folder, except for 05_downsampling_pulse_[Site name].zip which should all be moved to a subfolder called "downsampling_pulse".本数据集为一组脚本与研究数据集合，用于评估机载激光扫描（Airborne Laser Scanning, ALS）森林结构表征的鲁棒性。本数据集配套发表于论文《机载激光扫描森林结构的鲁棒表征——面向生态学家的系统评估与样本工作流》（"Robust characterization of forest structure from airborne laser scanning – a systematic assessment and sample workflow for ecologists"）。为符合双盲同行评审要求，整套数据与脚本已完成匿名化处理。 在该论文中，我们评估了从点云数据衍生冠层高度模型（Canopy Height Model, CHM）的方法，探究冠层高度模型算法对点云退化（脉冲密度降采样、大扫描角、高阶回波损失）的敏感性，以及不确定性与偏差如何传播至常用森林结构指标。此外，我们提供了一套基于R语言的标准化处理流程，可将点云转换为冠层高度模型。 本研究的核心数据来源为隶属于陆地生态系统研究网络（Terrestrial Ecosystem Research Network, TERN）的9个澳大利亚研究样地的机载激光扫描点云，每个样地范围为5 km × 5 km。公开可获取的基础数据链接为：https://portal.tern.org.au/metadata/TERN/4ff0b4c9-cfa0-4d09-9520-b5402adc583f。针对Robson Creek样地，我们还使用了野外实测数据，以评估地上生物量估算结果对机载激光扫描点云特征的敏感性，相关数据可通过以下链接获取：https://portal.tern.org.au/metadata/supersite.174。 为表征各研究样地间的气候与环境差异，我们使用了CHELSA/BIOCLIM+气候数据集1981-2010年版本（Brun et al. 2022：《过去与未来千米分辨率全球气候相关预测因子》，Earth System Science Data, 14(12), 5573–5603. https://doi.org/10.5194/essd-14-5573-2022; Karger et al. 2017: 《地球陆地表面区域高分辨率气候数据集》，Scientific Data, 4(1), 170122. https://doi.org/10.1038/sdata.2017.122）。 需要说明的是，完整的人工修改点云数据集（原始点云+降采样点云+各飞行航线数据：约400 GB）及衍生栅格产品（约200 GB）体积过大，远超Zenodo的数据存储限制。不过，所有分析均可通过公开可得的基础数据与本仓库中的R代码从头复现。此外，我们还提供了9个研究样地的衍生产品，无需进行点云处理即可复现论文正文结果，包括不同降采样点云对应的冠层高度模型及其他栅格数据，以及汇总统计量。 本数据集包含以下不同数据图层： 1. 01_rscripts.zip：包含用于测试处理流程的示例脚本（test.processing.R）与一组辅助函数（ALS_processing_helperfunctions_v40.R）。解压文件夹后可直接运行脚本，但需提前安装LAStools（https://rapidlasso.de），并配置`path_lastools = "PATH/TO/LASTOOLS/BIN"`。需要注意的是，本脚本开发于Windows系统，暂未测试其在LAStools的最新Linux版本上的兼容性。本压缩包同时包含复现全部分析所需的完整脚本集，包括点云处理与从原始数据衍生冠层高度模型的代码（create.CHMs.R），以及整体鲁棒性分析代码（analyze.CHMs.R）。若需逐步骤复现原始点云的处理流程，需从TERN仓库下载.laz格式文件（参见上文引用），并将其放置于名为"data"的文件夹中，每个样地对应一个子文件夹，且需遵循本仓库的命名规范（例如："/data/Alice Mulga"）。 2. 02_reference.zip：包含9个TERN样地的参考数字表面模型（Digital Surface Model, DSM）、冠层高度模型与数字地形模型（Digital Terrain Model, DTM），基于原始机载激光扫描点云生成。需注意，这些参考图层采用"CHMhighest"算法生成，当脉冲密度较高（≥20次/平方米）时，该算法可提供易于解读的冠层表征结果。 3. 03_climate.zip：包含样地坐标，以及来自CHELSA气候数据集的气候图层（参见上文引用，仅用于评估样地的气候范围）。 4. 04_robson_additional.zip：包含Robson Creek样地的生物量估算结果，以及该样地大型树木的形状文件（仅用于可视化）。 5. 05_downsampling_pulse_[Site name].zip：其中[Site name]为9个TERN样地的占位符（例如"Alice Mulga.zip""Credo.zip"等），包含复现本研究正文结果所需的数据，即9个TERN样地在不同脉冲密度水平下（从16次/平方米降至0.5次/平方米）的数字表面模型、冠层高度模型与数字地形模型，同时包含各研究样地的计算汇总统计量。 所有压缩包均需解压至同一文件夹，仅05_downsampling_pulse_[Site name].zip需全部移动至名为"downsampling_pulse"的子文件夹中。