地理研究

地理研究 ›› 2007, Vol. 26 ›› Issue (6): 1186-1196.doi: 10.11821/yj2007060013

• 地球信息科学 • 上一篇    下一篇

海洋悬浮泥沙二元特征参数 MODIS遥感反演模型研究

王 芳1,2, 李国胜2   

  1. 1. 加拿大卡尔加里大学,卡尔加里 加拿大 T;
    2. 中国科学院地理科学与资源研究所,北京 100101
  • 收稿日期:2007-03-23 修回日期:2007-08-14 出版日期:2007-11-25 发布日期:2007-11-25
  • 作者简介:王芳(1979-),女,江苏灌云人,加拿大卡尔加里大学研究生。主要从事环境遥感研究。 *通讯作者 : 李国胜(1963-),男,江苏常州人,研究员。主要从事陆海相互作用过程遥感与GIS模拟研究。 E-mail:ligs@igsnrr.ac.cn
  • 基金资助:

    国家自然科学基金资助项目(40771030、40571020);中国科学院知识创新工程领域前沿项目。

Two parameters retrieval models of suspended sediment concentration of Bohai Sea based on MODIS data

WANG Fang1,2, LI Guo-sheng2   

  1. 1. University of Calgary, Calgary, Alberta T| Canada|
    2. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
  • Received:2007-03-23 Revised:2007-08-14 Online:2007-11-25 Published:2007-11-25
  • Supported by:

    国家自然科学基金资助项目(40771030、40571020);中国科学院知识创新工程领域前沿项目。

PDF (PC)

赞

被引次数

54

摘要：

本文提出了一种采用海面离水辐射率和泥沙粒径二元特征参数来反演研究海区(渤海)海洋表层悬浮泥沙浓度的新的遥感反演算法,以此为基础分别建立了基于MODIS遥感数据和泥沙粒径二元特征参数的主成分和神经网络两种泥沙浓度反演模型,并对比分析了两类模型的反演精度以及泥沙粒径因子对模型的影响。分析结果表明,新建立的二元特征参数反演算法在采用主成分模型和神经网络模型时的检验误差分别为0.256和0.244,而忽略泥沙粒径因子贡献的主成分模型和神经网络模型的检验误差分别为0.384和0.390,因此可以认为,在泥沙浓度反演模型中加入粒径因子时,模型的预测精度和模型稳定性均比只考虑浓度对反射率贡献的模型有显著改善。

关键词: 渤海, 悬浮泥沙浓度, MODIS, 遥感反演, 二元模型

Abstract:

In recent decades, remote sensing has been proven to be an effective method to retrieve suspended sediment concentration. However, most retrieval methods developed today only based on the direct relationship between suspended sediment concentration and the remote sensing reflectance which is not reasonable when there is a big difference among the grain size of the study area. Bohai Sea is a relative closure region in which the grain size of the suspended sediment varies widely. This paper brought out a new method on the retrieval of suspended sediment concentration of Bohai Sea, which used both the reflectance from remote sensing data and the grain size of the suspended sediment. After analyzing the spectrum characteristics of suspended sediment, MODIS data characteristics and the relationship between suspended sediment concentration and grain size, the unary-parameter and binary-parameter PCA-based and NN-based models were constructed in Bohai Sea based on the data collected. The analytical results show that when introducing grain size parameter into the PCA-based models, model's correlation coefficient was increased from 0.697 to 0.724, while its predicting error was decreased from 0.383809 to 0.256722. This phenomenon also happened in the NN-based models where the predicting error was decreased from 0.390374 to 0.244427. The stability of the models with a grain size parameter being also better than the one without the grain size. It is proved that a model's retrieval precision and stability can be improved effectively by introducing grain size into the model. Therefore, it's necessary to add the grain size into the retrieval model in order to improve the precision of the prediction. Moreover, representative remotely sensed MODIS imagery was used to validate the model built in this paper which has the same conclusion with the previous researches. Generally speaking, in Bohai Sea, suspended sediment concentration in winter low water period is higher than that in summer abundant water period; and high concentration areas are located in Bohai Bay, Laizhou Bay and Liaodong Bay and their distributions are parallel to the shoreline.

Key words: Bohai Sea, suspended sediment concentration, remote sensing, two parameters model