| 摘要: |
| 水体透明度(Secchi Disk depth,SDD)是水环境监测的重要参数,遥感技术对于监测水体透明度具有重要的应用前景。本文旨在分类和比较当前用于监测水体透明度的算法,并提出未来研究的方向,以推动水环境监测技术的进一步发展。文章对目前检索水体透明度的算法进行分类和比较。其中,经验算法、半分析算法和机器学习算法是目前研究的主要方向。通过分析算法特性和优缺点,提出未来研究的重点和方向。经验算法基于透明度与光谱数据、叶绿素a浓度等的相关性,半分析算法基于水下能见度理论,机器学习算法则基于更优的数据特征学习能力。不同算法具有各自的适用范围和限制。未来的研究应该着重于整合多源遥感数据,改进QAA (quasi-analytical-algorithm),深入分析光学参数与水体透明度的关系,将机器学习算法应用到水体透明度模型的建立中,以建立具有高精度、适用性广的反演模型。 |
| 关键词: 透明度 遥感技术 经验算法 半分析算法 QAA 机器学习 |
| DOI：10.11759/hykx20220902002 |
| 分类号:X87 |
| 基金项目:国家自然科学基金项目(42106172);山东省重点研发计划(2019GHY112017);山东省自然科学基金(ZR2021QD135);山东省科学院海洋仪器仪表研究所基金项目(HYPY202107);教育部产学合作协同育人项目(202102245036,202101044002);科教产融合创新试点工程项目(2022PY041,2022GH004) |
|
| Research progress on the remote sensing inversion algorithm for water transparency |
|
ZHAO Chun-yan1, YU Ding-feng1, ZHOU Yan1, YANG Lei1, GAO Hao1, YAO Hui-ping2
|
|
1.Institute of Oceanographic Instrumentation, Qilu University of Technology(Shandong Academy of Sciences), Qingdao 266100, China;2.College of Oceanography and Space Informatics, China University of Petroleum, Qingdao 266580, China
|
| Abstract: |
| Water transparency (Secchi Disk depth) serves as a crucial parameter in water quality monitoring. Remote sensing technology exhibits immense potential in facilitating such monitoring. This paper aims to categorize and compare the current algorithms used in monitoring the transparency of water bodies and identify prospective directions of future research to further the advancement of water quality monitoring technologies. Three algorithms constitute the main directions of current research in this field: empirical, semianalytical, and machine learning (ML). By analyzing the characteristics, advantages, and disadvantages of these algorithms, the focus and direction of future research in this domain are proposed. The empirical algorithm is based on the correlation between transparency and spectral data, along with other factors such as chlorophyll a concentration; the semianalytical algorithm is based on the underwater visibility theory; and the ML algorithm is based on superior data feature learning capabilities. Each algorithm presents a unique range of applications and limitations. Future research should focus on integrating multisource remote sensing data, improving the quasianalytical algorithm, deeply analyzing the relationship between optical parameters and water transparency, and applying ML algorithms to establish water transparency models, thereby establishing inverse models with high accuracy and wide applicability. |
| Key words: transparency remote sensing technology empirical algorithm semi-analytical algorithm QAA machine learning |
