基于偏振反射与光谱不变量的植被氮含量遥感反演

Nitrogen is indispensable in the process of vegetation growth and development. Its content conveys information about efficiency of vegetation photosynthesis and nutritional status, and thus is critical for researches on ecosystem evaluation and global climate change, etc. Remote sensing inversion is the most effective tool for nitrogen content monitoring at a large scale. However, the developed methods are only suitable for nitrogen content inversion at a small scale. They may not work well or even be failed for large-scale inversion of nitrogen content because of the two coupled relationships, i.e., “specular reflection-interior scattering” and “leaf albedo-canopy structure”. This project thus intends to: firstly, build a physical polarization model for vegetated land surfaces using sunlit leaf area by taking polarization mechanism into consideration and accurately estimate the polarized reflectance of vegetation to decouple the “specular reflection-interior scattering” relationship; secondly, based on the spectral invariants theory, analyze the difference between multiple scattering at green and near infrared bands and propose a method for fast estimation of directional area scattering factor by decreasing data demands and computation costs with the aim to decouple the “leaf albedo-canopy structure” relationship; finally, evaluate the accuracy of remote sensing of nitrogen content at a large scale after the decoupling techniques. The project can provide new thoughts on highly accurate estimation of vegetation nitrogen content and will have scientific and practical impacts on the development of quantitative remote sensing modeling and inversion.

氮素是植被生长发育过程中不可缺少的重要元素，其含量信息体现了植被光合作用效率和营养状况，对于生态系统评估和全球气候变化等研究极其重要。遥感反演是大范围氮含量监测的最有效手段，但是由于“镜面反射-内部散射”和“叶片反照率-冠层结构”两大耦合问题的影响，现有反演方法通常仅适用于区域范围较小的情形，对于大范围氮含量监测能力不足，甚至失效。基于此，本项目拟开展以下研究：首先，考虑植被镜面反射的偏振产生机理，建立基于光照叶面的偏振反射率物理模型，精确估计植被偏振反射率，解耦“镜面反射-内部散射”；其次，利用光谱不变量理论分析植被在绿光和近红外的多次散射差异，建立方向面积散射因子快速估算方法，降低估算所需数据量和计算量，解耦“叶片反照率-冠层结构”；最后，定量评估解耦后的大范围氮含量遥感反演精度。本项目研究成果将为植被氮含量遥感反演提供新思路，对定量遥感建模与反演的发展具有重要科学意义与实用价值。

遥感是大范围地表监测的最有效手段，传统的植被遥感反演方法面临“镜面反射-内部散射”和“叶片反照率-冠层结构”两大耦合问题，导致对于大范围监测能力不足，甚至失效。基于此，项目组建立了基于光照叶面的植被偏振反射率物理模型、基于光谱不变量理论的方向面积散射因子快速估算方法以及大范围植被氮含量遥感反演方法，并探索了人工智能技术在偏振建模和遥感反演领域的应用，提出了基于人工神经网络的偏振反射率和生化参量反演模型，解耦上述两大耦合问题，提高了方向面积散射因子估算效率，提升了偏振反射和遥感反演精度。研究成果发表SCI论文7篇。研发的大范围植被氮含量遥感反演方法成功应用于我国东南地区和德国巴伐利亚国家森林公园区域。