山区森林时间序列叶面积指数遥感反演及地形效应分析

Monitoring time-series changes of leaf area index (LAI) is very important to study terrestrial ecosystem carbon cycling, effect of global change and so on. The topographic relief induces the radiometric error of remotely sensed signals and affects the quality of remote sensing observations. Thus, it is difficult to retrieve LAI from remotely sensed data in a mountainous area at landscape scale. At present few researches have been done on time-series inversion of LAI from remote sensing data in complicated topography. Taking the typical forest vegetation in a mountainous area as the research object, the study constructs time-series data sets by using domestic HJ-1A/B satellite images. An ecological process model is coupled with a canopy reflectance model. Taking effects of topography on remotely sensed observations and the coupled model, the data assimilation method is developed to retrieve LAI based on remotely sensed observations, field measurements, and the coupled model. Then the effects of topography are eliminated through several data assimilation strategies, and the LAI is retrieved from remotely sensed data in a mountainous area at landscape scale. At last, combining digital elevation model, the study investigates the effects of topography on estimating time-series LAI of typical forest vegetation using remotely sensed data in complex terrain, and assesses uncertainty in time-series LAI retrieval. The successful implementation of this project not only is helpful to deepen understanding of topographic effects on time-series LAI inversion in complicated topography, but also gradually improve the method system of LAI inversion from remotely sensed observations. The study could also serve as a good reference and scientific support for eco-environmental monitoring using remote sensing data and terrestrial ecosystem simulation in a mountainous region.

监测植被冠层叶面积指数（LAI）的时间序列变化对于研究陆地生态系统碳循环及全球变化的影响等具有非常重要的意义。在山区景观尺度地形起伏引起遥感观测的辐射误差，影响了遥感数据质量，增加了遥感反演山区LAI的难度。目前尚缺乏山区时间序列LAI遥感反演的相关研究。本项目以山区典型森林植被为研究对象，利用国产环境减灾卫星数据构建时序遥感数据集；耦合生态机理模型与冠层反射模型，考虑地形对遥感观测和耦合模型的影响，发展基于遥感观测、地面实测和耦合模型的LAI同化反演方法；采用多种同化策略校正地形效应，实现山区景观尺度时间序列LAI的遥感反演；结合数字高程信息，分析地形对山区时序LAI遥感反演的影响，评估LAI反演结果的不确定性。本项目的成功实施将有助于认知山区时间序列LAI遥感反演的地形效应问题，也有利于逐步完善LAI遥感反演方法体系，可为山地生态环境遥感监测与生态系统模拟研究提供方法借鉴与科学依据。

叶面积指数（LAI）遥感估算是植被定量遥感研究的热点之一，监测植被LAI时空变化对于研究陆地生态系统碳循环及全球变化等具有非常重要的意义。项目申请时主要针对森林生态系统，不过在介绍主要研究进展时又不局限于森林类型。在系统梳理山地生态参量遥感反演以及森林生态机理模型等方面研究进展的基础上，在我国西南山区设置10个观测样区作为研究区，其中包括5个森林生态系统样区、3个农田生态系统样区和2个草地生态系统样区，结合地面观测与同期获取的遥感数据，考虑地形因素影响，开展各样区LAI遥感反演研究，探讨地形起伏对LAI反演结果的影响；系统分析LAI遥感产品随地形因子的时空变化规律，阐明地形对LAI产品可靠性的影响程度；发展模型参数敏感性分析方法，借助数据同化算法，耦合生态模型与多源遥感观测，在不同地形梯度（如平原、高山峡谷区）开展时间序列LAI遥感同化估算研究，评估LAI同化结果的时空表征能力。.研究发现，考虑了海拔、坡度和坡向等地形因子的森林LAI遥感反演模型与未考虑地形变量的模型相比，其验证精度有所提高；LAI遥感产品质量易受复杂地形；相比地形，产品间时空一致性更易受植被类型支配；山区不同LAI产品季相表征差异显著，尤其在常绿阔叶林区域，且产品无法准确提取低山丘陵区作物物候特征；与地面实测数据相比，产品在山区的验证精度较差；基于遥感与模型同化的方法能够生成时空连续的LAI数据集，相比MODIS LAI产品、遥感反演LAI以及生态模型模拟的LAI，同化后的LAI结果无论在估算精度还是在时间序列尺度均有明显改善。在复杂山区考虑地形影响，遥感观测与生态模型两者结合既能提供LAI宏观监测信息，又可动态反映LAI时序变化，有利于实现优势互补，提升应用潜力。项目发展的山地LAI遥感估算及产品评价方法有助于认知山地植被LAI反演的地形效应问题，可为山地LAI遥感产品改进及植被长势监测提供科学依据。