基于Sentinel-2卫星遥感影像的巢湖及南淝河叶绿素a浓度反演
Estimation of Chlorophyll-a Concentration in Chaohu Lake and Nanfei River Based on Sentinel-2 Satellite Remote Sensing Imagery
投稿时间:2021-09-17  修订日期:2021-12-03
DOI:10.15928/j.1674-3075.202109170328
中文关键词:叶绿素a  卫星遥感  浓度反演  巢湖  南淝河
英文关键词:chlorophyll-a  satellite remote sensing  concentration inversion  Chaohu Lake  Nanfei River
基金项目:国家自然科学基金(43971311);安徽省科技重大专项(201903a07020014)
作者单位
孙世举 安徽大学资源与环境工程学院安徽 合肥 230601 
徐 浩 北京空间飞行器总体设计部北京 100094 
吴艳兰 安徽大学资源与环境工程学院安徽 合肥 230601 安徽省地理信息智能技术工程研究中心安徽 合肥 230601信息材料与智能感知安徽省实验室安徽 合肥?230601 
吴鹏海 安徽大学资源与环境工程学院安徽 合肥 230601
安徽省地理信息智能技术工程研究中心
安徽 合肥 230601 
杨 辉 安徽大学物质科学与信息技术研究院安徽 合肥 230601 
摘要点击次数: 1325
全文下载次数: 1089
中文摘要:
      叶绿素a是反映水生态环境污染状况的重要指标。定量反演叶绿素a浓度有助于及时监测水体营养状态变化,对富营养化水体治理具有重要意义。以巢湖及南淝河支流下游为研究区域,利用Sentinel-2卫星遥感数据源,构建其叶绿素a浓度反演模型,探究叶绿素a浓度的时空变化规律。结果显示,构建的深度神经网络(DNN)模型反演精度较高(R2=0.96,MRE=31.62%,RMSE=24.4 μg/L)。通过减少训练样本量对DNN模型精度的影响分析,发现训练样本较少时,模型仍具有较高的精度;根据其精度的敏感模型训练样本个数,将训练集按组等分,模型呈现较好的稳定性并具有一定的适用性。分析表明,研究区叶绿素a浓度在时间上呈现夏秋季上升、春冬季下降的规律,在空间上呈现湖区西高东低、局部近岸区分布较高的特点。
英文摘要:
      Chlorophyll-a concentration is an important water quality parameter, reflecting the pollution status of aquatic ecosystems. Quantitative inversion of chlorophyll-a concentration is useful for following water eutrophication status over time and is crucial for developing plans to improve eutrophic water bodies. In this study, Chaohu Lake and the lower reaches of Nanfei River and tributaries were selected for investigation. The temporal and spatial variation of chlorophyll-a concentration in the study area was analyzed using a chlorophyll-a concentration inversion model. It aimed to provide scientific evidence for intelligent and dynamic monitoring of the nutrient status of water in the Chaohu Lake basin. First, a depth neural network (DNN) model was constructed based on Sentinel-2 satellite remote sensing data from August 2 in 2018, December 27 in 2019, June 25 in 2020, November 2 in 2020 and November 13 in 2020, as well as field measurement data of chlorophyll-a concentration. The stability and applicability of the model was tested by reducing the training sets. The depth neural network (DNN) model constructed for this study had high inversion accuracy (R2=0.96, MRE=31.62%, RMSE=24.4 μg/L). Analysis of the impact of reducing the training sample sets on DNN model accuracy shows that the model maintained accuracy with fewer training samples. Based on the number of sensitive model training samples, the training sets were divided into equal parts. To summarize, the model developed in this study displayed good stability and conditional applicability. After developing and testing the model, it was used to estimate chlorophyll-a concentrations in the study area and the average chlorophyll-a concentration ranges for spring, summer, autumn and winter were, respectively, 141-168, 175-195, 172-218 and 143-164 μg/L. The concentration of chlorophyll-a increased in summer and autumn and decreased in spring and winter. Spatially, chlorophyll-a concentrations were higher in the western lake and some nearshore areas, and lower in the eastern lake.
孙世举,徐 浩,吴艳兰,吴鹏海,杨 辉.2023.基于Sentinel-2卫星遥感影像的巢湖及南淝河叶绿素a浓度反演[J].水生态学杂志,44(5):58-66.
查看全文  查看/发表评论  下载PDF阅读器   HTML
Copyright © 2008 《水生态学杂志》编辑部 地址:湖北省武汉市雄楚大街578号 
邮编:430079 电话:027-82926630 E-mail: sstx@mail.ihe.ac.cn  京ICP备09084417号