基于自动监测和Sentinel-2影像的钦州湾溶解氧反演模型研究

钦州湾是广西近岸海域水质较差的区域，为及时了解钦州湾海域水质状况，本文借助遥感水质监测技术范围广、快速、连续、可视化程度高的优势，以高分辨率遥感卫星Sentinel-2影像数据为数据源，结合广西近岸海域自动监测数据，通过一元与多元，线性与非线性的回归分析方法建立钦州湾溶解氧浓度反演模型。研究表明，在构建的370个波段组合中，最佳波段组合分别是1/B3、lnB3/（lnB1+lnB2）和B3/（B1+B2），其Pearson相关系数（R）分别为0.905 2，-0.897 0和-0.889 2。钦州湾溶解氧浓度反演模型中，逐步回归模型拥有最低的平均相对误差MRE （6.47%），最低的均方根误差RMSE （0.584 5），同时模型验证精度R²为0.654 3，稳定性较佳。本文突破了广西近岸海域传统监测的局限性，同时为钦州湾溶解氧遥感监测提供参考。

Research on Dissolved Oxygen Inversion Model of Qinzhou Bay Based on Automatic Monitoring Data and Sentinel-2 Data

The Qinzhou Bay is an area with poor water quality in the coastal waters of Guangxi.To understand the status of the seawater quality in Qinzhou Bay in time,a dissolved oxygen concentration inversion model of Qinzhou Bay is necessary.The inversion model was set up with the aid of remote sensing water quality monitoring technology which had advantages of wide rage,fast speed,high continuity and high visual degree,using the high resolution remote sensing satellite Sentinel-2 image data as the data source,and combining with the automatic monitoring data of the coastal waters of Guangxi.The study showed that among the 370 band combinations constructed,the best band combinations were 1/B3,lnB3/(lnB1+lnB2) and B3/(B1+B2),and the Pearson correlation coefficient (R) were 0.905 2,-0.897 0 and -0.889 2,respectively.The inversion model of dissolved oxygen concentration in Qinzhou Bay was constructed based on the best sensitive band and linear and nonlinear regression analysis methods.The stepwise regression model had the lowest mean relative error (MRE,6.47%) and the lowest root mean square error (RMSE,0.584 5),and the model verification accuracy R² was 0.654 3,and it showed better stability.This article breaks through the limitation of traditional monitoring in Guangxi coastal waters and provides a reference for remote sensing monitoring of dissolved oxygen in Qinzhou Bay.