基于高光谱数据藻类水华重要参数——叶绿素a的定量反演模型研究  

Quantitative inversion model of chlorophyll a,an important parameter of algal blooms,based on hyperspectral data.

作　　者：洪琴 胡清华 陈文惠[1] 李翠萍 陈增文 雍凯婷 李师炜 林晓杰 HONG Qin;HU Qing-hua;CHEN Wen-hui;LI Cui-ping;CHEN Zeng-wen;YONG Kai-ting;LI Shi-wei;LIN Xiao-jie(College of Geographical Sciences,Fujian Normal University,Fuzhou 350007,China;Fujian Environmental Monitoring Center,Fuzhou 350003,China)

机构地区：[1]福建师范大学地理科学学院,福建福州350007 [2]福建省环境监测中心站,福建福州350003

出　　处：《环境生态学》2023年第2期23-31,共9页Environmental Ecology

基　　金：福建省科技计划项目(2020Y0070);福建省环保科技计划项目(2022R001)资助。

摘　　要：水库富营养化导致的水华问题严重影响了饮用水的利用和保护,而水体叶绿素a在营养化水体中比例稳定,可以在很大程度上反映水体的营养状况,是水华监测的重要参数。目前,大多数研究集中在较为典型的富营养化水体,对中/贫营养化水体的研究较少。以莆田东圳水库为研究区,随机选取112组叶绿素a浓度数据和高光谱数据中的72组数据进行模型构建,剩余40组数据用于模型验证。在对叶绿素a浓度和光谱反射率曲线进行相关分析的基础上,通过对水体光学特性的分析和统计分析,筛选出单波段、波段比值、三波段、四波段、荧光峰高度、荧光峰位置、峰谷距离等特征参数建立反演模型。通过模型精度验证的对比分析,确定了最优叶绿素a浓度的反模型。结果表明,经过900 nm处归零化预处理的单波段模型和四波段模型对东圳水库水体叶绿素a浓度的反演效果最佳,其决定系数R 2均为0.91,平均绝对误差MAE分别为1.82μg/L和2.81μg/L,MAPE分别为22.53%和21.53%。对于叶绿素a浓度较低、跨度较小的库区水域,波段算法类模型建模精度优于荧光算法类模型。该研究结果可以为后续东圳水库的常规水质监测和水华预警提供一定的理论依据和技术参考。Water bloom caused by reservoir eutrophication seriously affects the utilization and protection of drinking water resources.In nutritional water,chlorophyll a,the proportion of stable,can largely reflect the nutritional status of water,and it s an important parameter for water bloom monitoring.At present,most of the studies focus on typical eutrophic water bodies,while the studies on mesotrophic/oligotrophic water bodies are less common.This study takes Dongzhen Reservoir in Putian.as the research area.Among the 112 sets of chlorophyll a concentration data and hyperspectral data,the author randomly selected 72 sets of data for model construction,and the remaining 40 sets of data were used for model verification.Based on the correlation analysis of chlorophyll a concentration and spectral reflectance curves,through the analysis and statistical analysis of the optical properties of water bodies,this study selected single-band,band ratio,three-band,four-band,FLH,fluorescence peak position,peak-valley distance characteristic parameters to establish an inversion model.Through the comparative analysis of model accuracy verification,the inverse model of optimal chlorophyll a concentration was determined.The results showed that the single-band model and the four-band model after the normalization at 900 nm had the best inversion effect on the chlorophyll a concentration in the water body of Dongzhen Reservoir.The R 2 was 0.91,the MSE was 6.24μg/L and 6.81μg/L,and the MAPE was 24.47%and 21.53%.For the reservoir area,its chlorophyll a concentration is in a low and the span is small,the modeling accuracy of the band algorithm model is better than that of the fluorescence algorithm model.The results of this study can provide a theoretical basis and technical reference for the subsequent conventional water quality monitoring and algal bloom warning of Dongzhen Reservoir.

关 键 词：叶绿素a浓度反演 高光谱 中/贫营养水体 东圳水库 

分 类 号：X832[环境科学与工程—环境工程]