投稿中心

审稿中心

编辑中心

期刊出版

网站地图

友情链接

引用本文:高玉蓉,刘明亮,吴志旭,何剑波,虞左明.应用实测光谱估算千岛湖夏季叶绿素a浓度.湖泊科学,2012,24(4):553-561. DOI:10.18307/2012.0408
GAO Yurong,LIU Mingliang,WU Zhixu,HE Jianbo,YU Zuoming.Chlorophyll-a concentration estimation with field spectra of summer water-body in Lake Qiandao. J. Lake Sci.2012,24(4):553-561. DOI:10.18307/2012.0408
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 4977次   下载 2351 本文二维码信息
码上扫一扫!
分享到: 微信 更多
应用实测光谱估算千岛湖夏季叶绿素a浓度
高玉蓉1, 刘明亮1, 吴志旭2, 何剑波1, 虞左明1
1.杭州市环境保护科学研究院, 杭州 310014;2.淳安县环境保护监测站, 淳安 311700
摘要:
依据2010年8月的实测数据构建了千岛湖水体夏季叶绿素a浓度的实测光谱数据估算模型,并进行了验证.利用ASD FieldSpec3野外光谱仪获取高光谱数据,计算水体离水辐亮度和遥感反射率.通过寻找反演水体叶绿素a浓度的高光谱敏感波段,采用单波段相关分析、波段比值、微分光谱、三波段模型、BP人工神经网络等多种算法进行比较分析,结果表明:叶绿素a浓度与单波段光谱反射率的相关性不大;596 nm和489 nm波长处反射率比值、545 nm处光谱一阶微分与叶绿素a浓度均呈较显著相关,估测模型决定系数R2分别为0.782、0.590,RMSE分别为0.89、1.98μg/L;三波段模型的反演结果优于传统的波段比值和一阶微分法,R2为0.838,RMSE为0.71μg/L;神经网络模型大大提高了叶绿素a浓度的反演精度,R2高达0.942,RMSE为0.63μg/L.本研究为今后在千岛湖水域的夏季相邻月份进行叶绿素a浓度大范围遥感反演研究奠定了基础.
关键词:  千岛湖  叶绿素a  高光谱  遥感  BP人工神经网络
DOI:10.18307/2012.0408
分类号:
基金项目:国家自然科学基金项目(40971252);中国科学院知识创新工程重要方向项目(KZCX2-YW-QN312);杭州市环保局项目(2010006)联合资助
Chlorophyll-a concentration estimation with field spectra of summer water-body in Lake Qiandao
GAO Yurong1, LIU Mingliang1, WU Zhixu2, HE Jianbo1, YU Zuoming1
1.Hangzhou Institute of Environmental Science, Hangzhou 310014, P. R. China;2.Chun'an Environmental Monitoring Station, Chun'an 311700, P. R. China
Abstract:
Based on the in situ data collected in August 2010, hyperspectral data models estimating summer chlorophyll-a concentration in Lake Qiandao are presented. A large quantity of hyperspectral reflectance data and water quality data of the typical area of the lake were obtained. Hyperspectral data were measured using ASD FieldSpec3, and were calculated for water-leaving radiance and reflectance of water. Different methods including band ratio model, the first derivative model, three-band-model and BP neural network model were used to estimate chlorophyll-a concentration. Results showed that single band reflectance model gave the worst estimation on chlorophyll-a concentration. Band ratio model with the ratio of reflectance 596 nm/489 nm and the first derivative model of reflectance near 545 nm gave better results with high determination coefficients of 0.782 and 0.590, respectively. By comparison, the three-band-model had higher estimation precision (coefficient of 0.838) than the band ratio model and the first derivative model. BP neural network model performed best with a high determination coefficient of 0.942. The root mean square error between measured and estimated chlorophyll-a concentrations using the four models was 0.89, 1.98, 0.71 and 0.63 μg/L, respectively. Therefore, three-band-model and BP neural network model was recommended to estimate chlorophyll-a concentration with remote sensing data for large area of Lake Qiandao in the summer.
Key words:  Lake Qiandao  chlorophyll-a  hyperspectral data  remote sensing  ANN-BP
分享按钮