复杂地形及植被覆盖区LAI产品地面验证研究

Ground validation is the essential step to evaluate the performance of available LAI products. The ground validation scheme that was used to validate the LAI maps is significantly impacted by the terrain slope and the heterogeneity of vegetation canopy, however, the reliability and accuracy of the ground validation scheme that was applied to the terrain and contrasting plant functional types area was not comprehensively evaluated yet. In this project, a method that can be used to generate the large-area and highly realistic vegetation scenes planning to be developed based on the data sets, which were collected using terrestrial LiDAR, UAV Lidar and portable scanner etc. The generated vegetation scenes are conform to the main structural characteristic parameters and the 3D distribution of the PAI and LAI of vegatation canopy in the study area. A typical virtual vegetation scene library with different plant functional types, forest canopy structural characteristics, understory species, terrain slope and phenological periods would be build next based on the abovementioned vegetation scene construction method, and the library used as a unified platform for validating the ground validation scheme. The in-situ indirect LAI measurements and upscaling schemes would be simulated under this platform based on the inverse ray tracing algorithm, Monte Carlo ray tracing algorithm, GPU, parallel computing etc. We tried to remove the four main sources of errors (clumping effects, non-photosynthetic component, terrain slope and observation schemes) for the indirect optical methods if they are applied to the ESU vegatation canopy, and validate and improve the reliability and accuracy of upscaling schemes based on the LAI and PAI of the simulated vegetation scene. In the end, we tried to propose a new ground validation scheme for LAI products with high precision and reliability, which can be applied to terrain and contrasting plant functional types area.

地面验证为LAI产品精度及可靠性评价的重要手段。复杂地形及植被覆盖区LAI产品地面验证受地形起伏及植被冠层异质性影响显著，其验证方案可靠性及精度目前尚未系统地定量分析及评估。本项目以地面激光雷达、无人机激光雷达、手持扫描仪等手段为基础，开发可顾及植被场景冠层几何结构特征及LAI、PAI三维分布的大范围、高逼真度虚拟植被场景构建方法，同时以此为基础形成覆盖典型植被类型、冠层几何特征、地形条件、林下植被及物候期下的典型虚拟植被场景库，将其作为统一的LAI产品地面验证平台。在虚拟植被场景下联合逆向光线跟踪算法、蒙特卡洛光线跟踪算法、GPU、并行计算等手段实现地面LAI间接测量方法及尺度变换方案模拟。以虚拟植被场景LAI、PAI真实参考值开展验证分析，研究消除ESU尺度植被冠层地面LAI间接测量四大误差来源的方法，定量评估、提高尺度变换方案精度及可靠性，最后提出高精度的LAI产品地面验证方案。

LAI遥感产品可广泛应用于农业、林业、自然资源监测和全球变化等领域，因此对LAI产品进行精度评估是验证LAI产品是否符合用户需求的重要手段。地面验证是LAI产品验证的基础，地面验证涉及到的尺度变换及ESU尺度地面LAI间接测量精度问题仍有待进一步研究及细化。本项目以地面观测和模拟验证两条技术路线对ESU尺度地面LAI间接测量及粗像元尺度变换精度进行精度验证，项目组首先开展长时间序列LAI地面观测及虚拟森林场景建模参数收集，之后以TLS点云数据为基础开展观测样地三维几何重建并进一步形成虚拟森林场景库，在虚拟森林场景下模拟地面LAI间接测量方法及其尺度变换方案，结合地面观测数据分析验证ESU尺度地面LAI间接测量及粗像元尺度变换精度。..研究结果表明：采用落叶期DHP影像计算样地WAI，并将WAI计算结果用于木质总面积比参数测量会导致其计算结果显著偏大（与直接测量法木质总面积比参数对比），此结果表明传统的木质总面积比参数测量方案存在较为显著的高估，因此在森林冠层LAI地面测量中不宜采用。通过与直接测量方法木质总面积比参数结果对比发现，利用MCI仅在茂盛期开展观测可得到高精度的木质总面积比参数测量结果（误差小于20%）。采样方案对间隙率、聚集指数、PAI和LAI均存在显著的影响。提高采样点数目可一定程度上提高LAI地面测量精度，但其精度并不随采样点数目增加而一直增加。只有在考虑采样方案的条件下，采用DHP方法开展森林冠层LAI测量才能获取误差小于20%的LAI测量结果。针对ESU尺度森林LAI测量给出了相应的推荐方案。光学测量方法ESU LAI测量精度受反演模型、聚集指数算法和木质总面积比参数测量方法影响显著。除DCP方法外，DHP、MCI、TRAC均可获得测量误差小于20%的ESU尺度LAI。四种典型光学间接测量方法（DHP、DCP、TRAC和MCI）均无法获得测量误差小于5%的ESU尺度LAI。