| 引用本文: | 崔廷伟,张杰,马毅,孙凌.基于地物光谱的赤潮优势种识别研究.海洋与湖沼,2005,36(3):277-283. |
| |
|
| |
|
|
| 本文已被：浏览 1461次 下载 0次 |
码上扫一扫！ |
|
|
| 基于地物光谱的赤潮优势种识别研究 |
|
崔廷伟,张杰,马毅,孙凌
|
|
1.中国海洋大学海洋环境学院 青岛266003;2.国家海洋局第一海洋研究所 青岛266061;3.海洋环境和数值模拟国家海洋局重点实验室 青岛266061;4.中国科学院海洋研究所 青岛266071;5.中国科学院研究生院 北京100039
|
|
| 摘要: |
| 利用2001年获取的不同优势种赤潮水体的现场光谱数据,处理得到了不同优势种赤潮水体的遥感反射率光谱曲线;在此基础上,发展了基于光谱角度制图法(SpectralAngleMapping,SAM)和互相关光谱匹配法(CrossCorrelogramSpectralMapping,CCSM)的赤潮优势种识别算法。数据实验结果表明,上述两种方法对于实现基于光谱数据的赤潮优势种识别是基本有效的,且CCSM方法的识别结果略优于SAM方法。进一步提高上述方法的适用性在于大量获取不同优势种赤潮的光谱数据并建立赤潮光谱数据库。 |
| 关键词: 赤潮 优势种 光谱 识别 |
| DOI： |
| 分类号: |
| 基金项目:国家高技术研究发展计划项目——航空遥感多传感器集成与应用技术系统,2001AA633080号;国家高技术研究发展计划项目——渤海海洋生态环境海空准实时综合监测示范系统,2001AA634030号;国家重点基础研究发展规划课题——我国赤潮高发区赤潮的生物光学遥感原理与新方法,2001CB409708号 |
附件 |
|
| IN SITU SPECTRA IDENTIFICATION OF DOMINATING RED TIDE SPECIES |
|
CUI Ting-Wei1,2, ZHANG Jie3,4, MA Yi3,4, SUN Ling5,6,7
|
|
1.Ocean University of China, Qingdao 266003;2.First Institute of Oceanography, State Oceanography Administration,Qingdao, 266061;3.First Institute of Oceanography, State Oceanography Administration, Qingdao 266061;4.Key Lab of Marine Environment and Numerical Modeling, State Oceanography Administration, Qingdao,266061;5.Institute of Oceanography, Chinese Academy of Sciences, Qingdao, 266071;6.First Institute of Oceanography, State Oceanography Administration, Qingdao,266061;7.Graduate School,Chinese Academy of Sciences, Beijing,100039
|
| Abstract: |
| Red tide is an abnormal ecological phenomenon caused by excessive phytoplankton multiplication or assemblage under certain circumstances resulting in ocean color variation. It has been a serious ocean incident and received high regards from governments and scientists all over the world. Traditionally, identifying dominant red tide species is to observe the in situ water sample under microscope by counting the cells numbers and then make statistics based on the data counted. With the advancement of hyperspectral remote sensing technique, the dominating red tide species can be studied in vaster areas and more efficient way (Ma et al, 2003). However, the precondition is to get in situ spectral data of different species of red tide. In this paper, two methods of spectral angle mapping (SAM) (Ma et al, 2003) and cross correlogram spectral mapping (CCSM) (Meer et al, 1997), which are two different methods to discriminate materials by their spectra, were applied in the identification to find a better way in this field and to integrated better with the traditional biological methods.
Two-period mesocosm ecosystem experiments were conducted in enclosed bags made of special plastics in Bayuquan Port of Liaodong Bay in 2001, where the dominating species of Skeletonema costatum, Chattonellamarina and Leptocylindrus danicus were found, and occasionally Mesodinium rubrum were found in the separate cruises. Spectral data of these four different species were acquired by above-water method, using spectrometer of Field Spec Dual VNIR manufactured by ASD Company and a reference plank with 15% reflectance. For detailed acquisition and processing procedure, please refer to Cui et al (2003). After data were processed, a total of 19 remote sensing reflectance spectra were established including 4 for Mesodinium rubrum, 10 for Leptocylindrus danicus, 3 for Skeletonema costatum and 2 for Chattonella marina. Three criteria were set up for the CCSM method to identify the different species.
The spectra obtained were processed using the SAM and CCSM methods based on the total 19 spectra. Statistics results were quite satisfactory, which proved the usability of the three criteria for CCSM method. For instances, Mesodinium rubrum could be identified with 100% possibility, Leptocylindrus danicus being 80%, and Skeletonema costatum at 33% by SAM or CCSM. However, in the case of Chattonella marina, two different values, 50% or 100% were obtained by SAM or CCSM respectively, from which we can conclude that CCSM method with 3 criteria was a little better than the SAM method. The reasons for the deviation were due to the similarity between reference spectra and the spectra to be identified, and also the lack of corresponding reference spectra.
The conclusion and discussion are as follows: firstly, spectral curves of different algal species do have differences, displaying in the aspect of spectral vector directions. The same algal species have the same spectral characteristics, with the difference of vector module, which is the theoretical foundation of SAM method. SAM method is not subject to time and equipment of data acquisition, its results have no relation with the vector length, which means that the similarity between the two spectra is irrelevant to the absolute value of each one. The CCSM method had a much better performance than SAM method and the 3 criteria is appropriate. It is feasible and valid to identify dominanting red tide species based on in situ spectra using SAM or CCSM method. Our work was to test which method is more exploitable for red tide dominant species identification by hyperspectral remote sensing data, also to support and supplement the traditional biological method. Secondly, in the spectral range of 500–800nm, all the red tide spectra have two peaks and one vale. The significance in the discrepancy of the second reflectance peak wave length determines the performance of spectral identification results in certain extent: if the second reflectance peaks of two different species were close to each other, in order to identify successfully, the reference spectra of these two species must be acquired as much as possible; otherwise wrong identification could be made. Thirdly, an important factor limiting the identification precision lies on the lack of reference spectra of different dominanting species of red tide. Red tide spectral shape varies with the dominance of phytoplankton and the chlorophyll concentration, which makes the problem more complicated. Tens of species of phytoplankton have ever caused red tide near Chinese coasts and the spectral database available now cannot meet the need. Further work is suggested to focus on the spectral data acquisition of different dominanting species of red tide. The setup of a spectral library containing various patterns with different dominance and chlorophyll concentration, for deep understanding on the spectral pattern and red tide mechanism, is necessary. |
| Key words: Red tide, Dominant species, Spectra, Identification |
|
|
|
|
|