基于激光雷达技术的树木三维构型空间格局及驱动力分析：以蒙古栎为例

Effective and accurate extraction of three-dimensional (3D) tree architecture in different forest environments is the prerequisite for understanding its large-scale spatial pattern and driving factors, which is of great significance for fulfilling the sustainable forest management goal of our nation. Currently, field-based manual measurement is one of the most frequently used methods for tree architecture parameter extraction. However, taking field measurements is very labor-intensive and time-consuming with limited accuracy. Near-surface light detection and ranging (LiDAR) technology has a strong capability to penetrate forest canopy and obtain 3D tree structures, which provides a new way to obtain tree architecture parameters. Taking Quercus mongolica as an example, this study aims to address the above-mentioned problem by using LiDAR data derived from multiple LiDAR platforms, including unmanned aerial vehicle (UAV) and backpack. Specifically, this study will 1) explore the strategy for acquiring LiDAR data from multiple near-surface remote sensing platforms in forest environment; 2) develop the algorithm for automatically registering and merging multi-platform LiDAR data; 3) establish a complete algorithm suite for effective, accurate and automatic extraction of 3D tree architecture parameters; and 4) analyze the spatial patter and driving factors of Quercus mongolica architecture. The expected fruits of the proposed project can help to understand the mechanism of how Quercus mongolica responds to climate change and provide a new technical paradigm for studying the spatial pattern and driving factor of tree architecture.

快速、准确地获取不同立地条件下树木三维构型参数，掌握其空间格局及空间变异规律，对于维护我国森林健康、实现森林可持续发展目标至关重要。目前，树木三维构型的提取依赖于手工实测数据，耗时费力且精度有限。近地面激光雷达技术在获取植被三维结构信息上快速、准确的优势为解决上述问题提供了可能。本项目拟以蒙古栎（Quercus mongolica）为例，通过引入多平台近地面激光雷达技术，突破树木构型参数获取的技术瓶颈，为研究树木构型空间格局及其驱动力提供新手段。拟开展的研究内容包括探讨适用于树木构型参数提取的激光雷达多平台协同扫描控制方案，开发林区环境下多源点云数据的自动匹配融合算法；建立基于激光雷达的树木三维构型参数快速、准确、批量提取的技术体系；分析蒙古栎构型的空间分布规律及空间变异的驱动因素。本项目的预期成果将揭示蒙古栎对气候变化响应机制，为未来树木构型空间格局及驱动力研究提供新的技术范式。

快速、准确地获取不同立地条件下树木三维构型参数，掌握其空间格局及空间变异规律，对于维护我国森林健康、实现森林可持续发展目标至关重要。目前，树木三维构型的提取依赖于手工实测数据，耗时费力且精度有限。近地面激光雷达技术在获取植被三维结构信息上快速、准确的优势为解决上述问题提供了可能。针对上述问题，本项目研发了双传感器背包激光雷达系统，解决了单传感器背包激光雷达系统在树木冠层顶部的信息丢失问题，论证了低成本无人机激光雷达系统在森林冠层结构参数获取方面的有效性，大幅提高了林区激光雷达数据的获取效率，降低了数据获取成本；研发了以树木位置为匹配特征的林区多源激光雷达数据自动配准融合算法，解决了林区环境下多源激光雷达数据自动配准融合的难题；研发了基于深度学习的植物茎叶分离与分割算法，实现了树木枝干构型参数（如枝长、枝粗、枝干夹角、分枝数、枝干体积等）的自动提取，解决了器官水平树木三维构型参数无损、高精度测量的难题；提出了一种深度学习指导的空间内插模型，结合近地面激光雷达数据、星载激光雷达数据和其他多源遥感数据研发了大尺度森林冠层高度绘制模型，实现了我国国家尺度30米分辨率森林冠层高度的高精度绘制；揭示了分形维数在表征森林冠层结构复杂性方面的尺度不可拓展性，利用信息熵理论构建了多源激光雷达数据适用的森林冠层结构复杂性定量化指标，揭示了森林冠层结构复杂性协同物种多样性对森林生产力的影响；揭示了蒙古栎冠层构型的空间分布格局，发现了蒙古栎冠层构型会随着水热条件的变化而变化，且这一变化会耦合叶片性状的变化来共同平衡树木对光水资源的需求以适应环境条件的变化。项目执行期间共发表论文18篇，申请核心专利1项；项目负责人获得李小文遥感科学青年奖等奖项，项目团队成员获得中国科学院朱李月华奖学金和国家奖学金等奖项。