基于DSCOVR/EPIC多角度“热点”数据的植被参数角度归一化及估算研究

Vegetation, as one of key land surface components, is usually described by remote sensing vegetation parameters such as vegetation indices, Fractional Vegetation Cover (FVC), leaf area index in the regional/global scale. However, due to sun-sensor-geometry, vegetation parameters often influenced by the angular effect which make them cannot reflect the real vegetation situation. In order to remove this effect, some researches treat it as random noise and remove it by filtering algorithm, which doesn't consider its regularity; others remove it by normalizing the angular effect in spectral reflectance, which can lead to the accumulation of errors. The newly published multi-angle “hotspot” dataset from DSCOVR/EPIC opens up new possibilities for solving this problem. In this study, we would like to provide an algorithm which directly normalizes the angular effect in vegetation parameters. The algorithm is supposed to be simple and available for all kinds of vegetation types and parameters. Firstly, the angular effect in vegetation parameters is analyzed based on model simulation. Secondly, the angle normalization algorithm is provided with the help of DSCOVR/EPIC dataset. Thirdly, take FVC estimation as an example, we will provide an estimation algorithm which can take advantage of “hotspot” reflectance. Finally, the normalized vegetation parameters will be validated according to other remote sensing products and field measurements. This project is supposed to obtain vegetation parameters which are not affected by the angular effect and can be better used to the research on dynamic changes of ecosystem and vegetation phenology.

植被作为地表的重要组分，其变化特征在区域/全球尺度上常由星载遥感获取的植被指数、植被覆盖度、叶面积指数等植被参数来表征。然而受太阳/观测几何的影响，植被参数包含角度信息，无法准确反映地表植被状况。现有研究或视角度效应为随机误差借助滤波去除，未考虑其规律性；或对反射率进行纠正，存在累积误差。美国最新发布的DSCOVR/EPIC数据可提供全球多角度“热点”观测，为大范围地表植被监测提供了新思路。本项目旨在提出一种简单的植被参数角度归一化算法，用于植被参数的精确估算。首先，借助模型研究不同植被参数及其角度效应；其次，借助DSCOVR/EPIC数据实现植被参数的角度归一化；再次，以植被覆盖度为目标，充分发挥“热点”的优势，提出适用于DSCOVR/EPIC的估算算法；最后，借助同类产品及地面实测对上述结果进行真实性检验。本项目预期可得到不受角度影响的植被参数，用于更好的服务生态系统及植被物候等研究。

多角度遥感由于丰富了单一角度光谱信息，被广泛应用于植被分类、植被结构参数提取等研究。然而，考虑到植被本身的各向异性及其空间分布的异质性，从不同角度观测到的植被状况有所不同，使得不同传感器或同一传感器在不同区域、不同时间获取的植被参数不具有可比性，无法真实表征地表植被变化状况。现有研究或视角度效应为随机误差借助滤波去除，未考虑其规律性；或对反射率进行纠正，存在累积误差。美国最新发布的DSCOVR/EPIC数据可提供全球多角度“热点”观测，将传统的需要4个角度表征的太阳观测几何简化为2个角度，为大范围地表植被监测提供了新思路。本项目旨在借助DSCOVR/EPIC“热点”数据优化植被参数的估算。首先，项目借助模型研究不同植被参数及其角度效应，发现与传统植被指数相比，天顶“热点”植被指数，特别是天顶“热点”EVI2与植被结构参数一次线性相关性更强；接着，以植被覆盖度（FVC）为目标，充分发挥“热点”的优势，研发全球近逐日FVC估算算法，算法经检验理论误差小于0.05，在承德的直接检验表明其误差约为0.043，与同类卫星遥感产品（误差0.049~0.087不等）相比精度较高；最后，对算法的副产品（Vv和Vs）进行时空扩展，借助MODIS BRDF数据以及Landsat-8、GF-1以及ZY-3等数据，证明在考虑光谱差异并借助精准的地类数据对副产品进行尺度转换的前提下，其可以适用于多源卫星FVC估算。本项目预期可应用于全球植被变化研究、全球气候变化研究、全球“碳中和”评估等领域，更好的监测大尺度植被结构快速变化。

内容获取失败，请点击重试