摘要
利用遥感技术反演藻蓝蛋白(PC)浓度对监测湖泊水体中蓝藻的生长情况具有重要意义。为了能够准确反演长三角地区湖泊的PC浓度,该研究通过分析实测的高光谱数据,构建了一种光谱分类方法,将水体分为4种类型。同时,又以多个光谱形状特征(PC峰面积、PC谷面积、PC峰谷高度差、最大一阶导数、叶绿素a谷上峰面积)和反射率组合(比值、差值、波段差和比、三波段、四波段)为自变量,分别建立了面向不同类型水体的PC浓度反演模型,并对分类前后的整体精度进行了比较。研究结果显示,4类水体均能取得一定精度的PC浓度反演模型,相应的R2分别为0.616、0.925、0.967、0.742。同时,分类后的整体模型反演精度也较未分类时最佳反演模型明显提升,R2达到0.954,RMSE与MAPE分别为0.208μg/L和18.38%。该研究所构建的分类方法有效适用,各类型水体也均能匹配相对最佳的反演模型,可为今后的PC浓度反演研究提供支持。
The inversion of phycocyanin(PC)concentration through remote sensing technology holds significant importance for monitoring the growth of cyanobacteria in lake water bodies.To precisely invert the PC concentration of lakes within the Yangtze River Delta region,in this study,by means of the analysis of the measured hyperspectral data,a spectral classification approach was established to categorize the water body into four types.Concurrently,with multiple spectral shape features(PC’s peak area,PC’s valley area,PC’s peak-valley height difference,maximum first derivative,chlorophyll-a’s peak area above valley)and combinations of reflectance(ratios,differences,the ratio of band difference to band sum,three-band,four-band)serving as independent variables,the PC concentration inversion models for distinct types of water bodies were respectively developed,and the overall accuracy prior to and subsequent to classification was compared.The research findings suggest that PC concentration inversion models with a certain degree of accuracy can be acquired for all four types of water bodies,with the corresponding R2 values being 0.616,0.925,0.967,and 0.742,respectively.Simultaneously,the overall inversion accuracy of the model after classification was conspicuously enhanced compared to the optimal inversion model before classification,with R2 attaining 0.954,and RMSE and MAPE being 0.208μg/L and 18.38%,respectively.Hence,the classification method constructed in this study is effectively applicable,and each type of water body can be matched with the relatively optimal inversion model,providing support for future research on PC concentration inversion.
作者
秦淦
谢勇
陶醉
邵雯
李斌
QIN Gan;XIE Yong;TAO Zui;SHAO Wen;LI Bin(School of Geographical Science,Nanjing University of Information Science and Technology,Nanjing 210044,China;Aerospce Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China)
出处
《环境科学与技术》
北大核心
2025年第4期133-143,共11页
Environmental Science & Technology
基金
国家自然科学基金项目(42176176)
高分辨率对地观测系统重大专项(21-Y20B01-9001-19/22)
国家重点研发计划项目(2023YFB3905801)。
关键词
藻蓝蛋白
高光谱
光谱形状
分类
phycocyanin
hyperspectral
spectral shape
classification
