• 关键词：太湖  藻蓝蛋白  反射率光谱  光谱分类  遥感  反演模型

• 基金项目：国家自然科学基金（No.41101378）；国家高分辨率对地观测重大专项（No.05-Y30B02-9001-13/15）；“十二五”国家水体污染控制与治理科技重大专项（No.2014ZX07508-1）

• 郭一洋
• 1. 辽宁工程技术大学测绘与地理科学学院, 阜新 123000;2. 环境保护部卫星环境应用中心, 北京 100094;3. 中国科学院遥感与数字地球研究所 数字地球重点实验室, 北京 100094

• 朱利
• 1. 环境保护部卫星环境应用中心, 北京 100094;2. 国家环境保护卫星遥感重点实验室, 北京 100094

• 吴传庆
• 1. 环境保护部卫星环境应用中心, 北京 100094;2. 国家环境保护卫星遥感重点实验室, 北京 100094

• 李俊生
• 中国科学院遥感与数字地球研究所 数字地球重点实验室, 北京 100094

• 张方方
• 中国科学院遥感与数字地球研究所 数字地球重点实验室, 北京 100094

• 摘要：藻蓝蛋白是蓝藻的标志性色素，利用遥感反演藻蓝蛋白浓度的时空分布对于蓝藻水华监测和预警具有重要意义.太湖水体光学特性时空差异较大，传统的藻蓝蛋白遥感反演方法在太湖各湖区各季节的适用性有限.因此，本文采用先分类再反演的策略，基于水面反射率光谱分类进行太湖藻蓝蛋白浓度反演建模.首先采用逐步迭代的K均值聚类方法实现光谱分类；然后分别利用每一类的训练样本光谱数据建立最适用于该类的藻蓝蛋白反演模型；最后利用每一类的检验样本光谱数据进行反演模型精度评价.为了对比，同时采用不分类的传统方法进行反演建模和精度评价.检验结果表明：基于不分类的传统建模方法得到的均方根误差RMSE=14.14μg·L^-1，平均相对误差σ=59.1%，反演结果和实测数据可决系数为0.46；基于光谱分类的建模方法得到RMSE=8.47μg·L^-1，σ=31.3%，反演结果和实测数据可决系数为0.87.因此，基于光谱分类的藻类蛋白反演方法有效地提高了反演精度，可以为其它水体的藻蓝蛋白浓度反演提供借鉴.

• Abstract：The phycocyanin is a symbolic pigment of cyanobacteria. Using remote sensing to retrieve the spatial and temporal distribution of algal blue protein concentration has important significance for the monitoring and early warning of cyanobacteria bloom. The temporal and spatial difference of the optical properties of the water body in Taihu is relatively large, and the application of the traditional remote sensing inversion method in Taihu lake is limited. In this paper, we use the method of classification and retrieval, based on the water surface reflectance spectroscopy, to model the concentration of Taihu's blue algae. The K mean clustering method is used to realize the spectral classification. Then the model is established by using each kind of training sample data. Finally, the accuracy of the model is evaluated by using the test data of each class. the traditional method is used to retrieve modeling and precision evaluation.Test results show that:Based on traditional modeling method of no classification get RMSE=14.14 μg·L^-1, σ=59.1%, the correlation coefficient of the retrieval results and the measured data is 0.46, the method based on spectral classification obtain the RMSE=8.47 μg·L^-1, σ=31.3%, and the correlation coefficient retrieval results and the measured data is 0.87. Therefore, the method of the spectral classification based on the method of algal protein inversion can effectively improve the inversion precision, and can provide a reference for the inversion of the concentration of the algae blue protein in other waters.

• Key words：Taihu Lake  phycocyanin  reflectance spectra  spectra classification  remote sensing  retrieve model

• 
  摘要点击次数： 1139 全文下载次数： 1794