复杂地形区多尺度叶面积指数遥感估算方法研究

The landscape structure shows very heterogeneous characteristics and the difference of growth status among either different vegetation types or the same type in different regions is very obvious due to effects of topographic relief in the rugged surface. It is necessary to monitor spatial difference and temporal changes of leaf area index (LAI) using remote sensing data. Furthermore, the topography induces the radiometric error of spectral signals and affects the performance of models (e.g. ecological model and canopy reflectance model), which increases the uncertainty of LAI estimation using remotely sensed observations in a mountainous area. Taking topographical effects on remotely sensed data, ecological model, canopy reflectance model into account, the data assimilation scheme was developed to retrieve multiscale LAI in time series in a mountainous area combining multi-resolution remotely sensed observations, ecological model, canopy reflectance model and prior knowledge. The proposed scheme was used for time series LAI estimation at multi-resolution scale for forest, crop, and grass in rugged terrain. The uncertainties of multiscale estimation results were assessed based on LAI reference values, and comparative analysis of different LAI results were done in order to explore their spatial and temporal consistencies. In addition, the topographic effects on estimating time series LAI using multi-resolution remotely sensed data were also investigated. This research mainly innovated in the retrieval of multiscale (e.g. 30m, 250, and 500m) LAI from multi-source data and accuracy evaluation of LAI estimations. The successful implementation of this project is helpful to deepen understanding of topographic effects on multiscale LAI inversion in complicated topography. The study could also serve as a good scientific support for improvements of both retrieval algorithm of LAI products and accuracy of LAI estimation from remotely sensed observations in a mountainous region.

受地形起伏影响，山地地表覆被和景观结构呈现较强的空间异质性，且不同植被类型、不同区域同种植被类型物候特征差异明显，山地LAI遥感估算需兼顾空间差异和时间动态，加之地形易造成遥感光谱信号失真以及模型模拟能力下降，增加了遥感反演的不确定性。选取山区森林、作物和草地植被为研究对象，在不同山地环境梯度下联合不同分辨率遥感数据、生态模型、冠层反射模型以及山地先验知识等，考虑地形对遥感观测和模型的影响，借助数据同化方法，实现山地多尺度时间序列LAI遥感反演，评估不同分辨率LAI估算结果的不确定性及其对山地植被生态系统时空变化特征的表征能力，揭示地形对LAI反演的影响。本研究在复杂地形区多尺度（30m、250m、500m）LAI遥感估算以及结果验证等方面均具有创新性。本项目的成功实施有助于认知山地多尺度LAI遥感估算的地形效应问题，可为LAI遥感产品算法在山区的改进及反演精度的提高提供科学理论依据。

叶面积指数（leaf area index, 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。受地形起伏影响，山地地表覆被和景观结构呈现较强的空间异质性，且不同植被类型、不同区域同种植被类型物候特征差异明显，山地LAI遥感估算需兼顾空间差异和时间动态，加之地形易造成遥感光谱信号失真以及模型模拟能力下降，增加了遥感反演的不确定性。选择典型山地区域及植被类型为研究对象，本项目主要开展了已有LAI遥感产品在复杂山地的对比分析与质量评估、复杂地形区多空间尺度LAI遥感反演，以及多尺度时间序列LAI遥感估算等工作。. 研究发现，中国西南山区地表起伏差异较大，导致各LAI遥感产品表现不一，同源产品生长季LAI差值低于非同源产品，地形因子影响着山地各产品间的差异；目前全球尺度LAI产品算法忽略了地形影响且空间分辨率较粗，无法准确刻画复杂山地高度空间异质性，由此导致的产品不确定性将传播至GPP模型，并对估算结果产生较大影响；不同传感器之间的空间分辨率差异能够为变量反演提供互补的有效信息，LAI反演中提供其他遥感信息作为约束条件，将有助于提高反演的稳定性和合理性；消除地形对遥感信号的辐射畸变能够显著改善不同地形条件情况下LAI反演精度，极大提升多尺度LAI估算结果的空间一致性及表达能力；将LAI的时间维变化规律引入瞬时遥感模型的反演算法中，增加LAI反演的时间维约束信息，将会提高LAI遥感数据产品的精度和时空连续性；发展的基于数据同化的LAI时空动态逐级降尺度方法，能够提高LAI遥感反演同化模型的适用性，有效改善估算结果对陆地生态系统时空格局的表征能力。. 本研究在复杂地形区产品验证、多空间尺度LAI遥感估算以及时间序列LAI估算等方面均具有创新性。本项目的成功执行有助于认知已有遥感产品质量在山地的表现以及多尺度LAI遥感估算的地形效应问题，可为LAI遥感产品算法在山区的改进及反演精度的提高提供科学理论依据。