基于无人机多光谱数据的三峡库区支流叶绿素a浓度估算——以小江为例  

Estimation of Chlorophyll a Concentration in Three Gorges Reservoir Branch based on UAV Multispectral Data——A Case Study of Xiaojiang River

作　　者：王会 王永前[1,4,5] 李剑锋 杨世琦[2] 吴雷祥 WANG Hui;WANG Yongqian;LI Jianfeng;YANG Shiqi;WU LeiXiang(College of Resources and Environment,Chengdu University of Information Technology,Chengdu 610225,China;Chongqing Meteorological Service,Chongqing 401147,China;China Institute of Water Resources and Hydropower,Beijing 100038,China;China Meteorological Administration key Laboratory of Atmospheric Sounding,Beijing,100081,China;China Meteorological Administration Meteorological Detection Engineering Technology Research Center,Beijing,100081,China)

机构地区：[1]成都信息工程大学资源环境学院,四川成都610225 [2]重庆市气象科学研究所,重庆401147 [3]中国水利水电科学研究院,北京100038 [4]中国气象局大气探测重点开放实验室,北京100081 [5]中国气象局气象探测工程技术研究中心,北京100081

出　　处：《成都信息工程大学学报》2024年第2期233-239,共7页Journal of Chengdu University of Information Technology

基　　金：中国气象局大气探测重点开放实验室联合基金开放课题资助项目(L2021Z07)。

摘　　要：以三峡库区重要支流——小江为研究区,基于大疆P4M无人机多光谱数据与原位测量叶绿素浓度数据,建立叶绿素a浓度估算模型。结果表明,采样点叶绿素浓度值在5~260μg/L,水体光谱曲线表现为各自的光谱特征,高叶绿素浓度水体的蓝(450 nm)、红(650 nm)波段附近出现叶绿素吸收峰,且随叶绿素浓度增加而加深;高悬浮物浓度水体光谱反射峰值高于其他水体,反射峰位置存在向长波方向移动的“红移”现象;使用MERIS陆地叶绿素指数(MTCI)建立的叶绿素a浓度估算模型精度最高,模型决定系数R2为0.89,平均绝对误差MAE为7.72μg/L,均方根误差RMSE为9.34μg/L,平均绝对误差百分比MAPE为35.72%;无人机水体成像应在环境光照稳定、太阳高度角适宜的时间段进行。This paper takes Xiaojiang River,an important tributary of the Three Gorges Reservoir,as the study area,establishes a chlorophyll a concentration estimation model based on the multi-spectral data of P4M UAV in Xinjiang and the chlorophyll concentration data measured in situ.The results showed that the chlorophyll concentration at the sampling point was between 5-260μg/L,and the spectral curves of water showed their own spectral characteristics.Chlorophyll absorption peaks appeared near the blue(450 nm)and red(650 nm)bands of water with high chlorophyll concentration,and grow with the increase of chlorophyll concentration.The spectral reflection peak of water body with high suspended matter concentration is higher than that of other water bodies,and the position of reflection peak has a“red shift”phenomenon moving to the long wave direction.The MERIS land chlorophyll index(MTCI)was used to establish the chlorophyll a concentration estimation model with the highest accuracy.The model determination coefficient R2 was 0.89,the average absolute error MAE was 7.72μg/L,the root mean square error RMSE was 9.34μg/L,and the mean absolute percentage error MAPE was35.72%.UAV water body imaging should be carried out in the period when light is stable and the solar altitude angle is appropriate.

关 键 词：无人机 多光谱 三峡重要支流 叶绿素A 回归模型 

分 类 号：X821[环境科学与工程—环境工程]