地球信息科学学报 ›› 2011, Vol. 13 ›› Issue (5): 687-694.doi: 10.3724/SP.J.1047.2011.00687

• 模型与算法应用 • 上一篇    下一篇

混合光谱分解算法的太湖水体组分浓度反演分析

陆超平, 吕恒, 李云梅   

  1. 南京师范大学虚拟地理环境教育部重点实验室, 南京 210046
  • 收稿日期:2011-05-12 修回日期:2011-08-10 出版日期:2011-10-25 发布日期:2011-10-25
  • 通讯作者: 吕恒(1977-),博士,副教授,主要从事水色遥感研究。E-mail: Henglu@njnu.edu.cn E-mail:Henglu@njnu.edu.cn
  • 作者简介:陆超平(1987-),男,硕士研究生,主要研究方向为水环境遥感、高光谱遥感、地理信息系统。E-mail:coollcp@126.com
  • 基金资助:

    国家自然科学基金项目(40701136,40971215);国家科技支撑计划项目(2008BAC34B05);国家水环境遥感技术体系研究与示范(2009ZX07527-006)。

Algorithms Based on Spectral Decomposition Algorithm for Retrieval of Constituents in Taihu Lake

LU Chaoping, LV Heng, LI Yunmei   

  1. Key Laboratory of Virtual Geographic Environment, Ministry of Education, Nanjing Normal University, Nanjing 210046, China
  • Received:2011-05-12 Revised:2011-08-10 Online:2011-10-25 Published:2011-10-25

摘要: 二类水体组份的遥感定量反演一直是水色遥感的难点和热点问题,原因在于其水体组分(纯水、叶绿素、悬浮物及CDOM)之间复杂的相互作用。本文引入光谱分解算法,通过Hydrolight软件模拟叶绿素、悬浮物和CDOM的标准反射率光谱,解决光谱分解算法中"纯端元"难以获取的问题。在此基础上建立了二类水体组分光谱分解反演模型。模型表明,光谱分解系数之间有很强的独立性,能够作为独立变量估算对应组分浓度。对光谱分解建模数据建立波段比值模型,用不同时期的同步实测数据,对光谱分解算法和比值模型进行比较验证,结果表明,光谱分解模型和波段比值模型的悬浮物反演平均相对误差分别为:22.4%和37.7%。验证结果:光谱分解算法相对于传统的经验模型,具有一定的季节通用性。叶绿素浓度估算模型的平均相对误差为31.7%,反演精度不高。其因是由于叶绿素浓度相对于非藻类悬浮物浓度过低,平均浓度分别为17.3μg/L和78.6mg/L,叶绿素的光谱信息在一定程度上被高悬浮物掩盖。

关键词: 光谱分解, 太湖水体, Hydrolight

Abstract: Quantitative remote sensing inversion of constituents in Case II waters has been a difficult and hot issue. It is due to the complex interaction among the constituents (e.g. pure water, phytoplankton, non-phytoplanktonic suspended solids and CDOM). In this paper we analyze the spectral decomposition algorithm of the optically active substances (OAS) in Taihu Lake. It contains two steps. First step is to obtain standard reflectance spectra of end-members (Chl a, NPSS, CDOM and pure Water) by simulation experience using Hydrolight; second step is to establish a number of equations (according to the numbers of end-members) obtained from different wavelengths and derive the decomposition coefficients to be used as the independent variables in the SDA-based estimation model. The results show that the decomposition coefficients obtained from spectral decomposition algorithm are highly independent. Moreover, we build the band ratio model based on the same data as the spectral decomposition algorithm. Then we apply the two models to the experimental data in November, 2008, and get the average relative error of 22.4% and 37.7%, which shows that to some extent the spectral decomposition algorithm has more seasonal versatility than that of band ratio model. The average relative error of chlorophyll concentration estimation model is 31.7%, which shows low inversion accuracy. The reason may be due to the concentration of chlorophyll is too low compared to non-algal suspended solids concentration. The average concentrations were 17.3μg/L and 78.6mg/L. So, to some extent, the spectral information of chlorophyll is concealed by the high suspended solids.

Key words: spectral decomposition, Taihu waters, Hydrolight