基于红(R)绿(G)蓝(B)三波段的典型港湾透明度反演研究

ESTIMATING TRANSPARENCY IN TYPICAL BAY USING RGB SPECTRAL BANDS

基于卫星遥感的透明度(Secchi depth,SD)反演方法已经在大洋、近海、湖泊等不同水体开展了广泛应用。富营养化河口港湾等水体具有水域面积小且受陆地影响大等特点,并且在此类水体开展环境监测具有时效性要求。受时间空间分辨率以及云层覆盖、大气校正等影响,基于卫星遥感方法在河口港湾等小水域进行透明度反演会受到一定的应用限制,因此在此类特征水体建立一种高效便捷的透明度反演方法作为卫星遥感方法的有益补充就十分迫切。本研究尝试通过无人机和智能手机应用程序HydroColor APP搭载的普通光学相机构建基于红(red,R)绿(green,G)蓝(blue,B)三波段的象山港透明度反演方法。结果表明,无人机和HydroColor的红蓝波段比值(R/B)和红绿波段比值(R/G)与透明度具有显著负相关,相关系数R为–0.88~–0.93(n=16,P<0.001)。根据相关性分析结果构建透明度反演模型并基于独立数据库对模型进行精度评估。结果显示,(1)指数反演模型要优于线性经验模型,(2)基于R/G反演模型要优于R/B模型,(3)HydroColor反演模型要优于无人机反演模型。通过以上结果分别构建基于无人机DJI-R/G和HydroColor-R/G最优透明度指数反演模型。DJI-R/G模型平均相对误差和均方根误差为29%和0.3 m,HydroColor-R/G模型为21.9%和0.27 m。以上结果表明,通过无人机和手机获取RGB信息均可用于对象山港透明度进行反演。该方法的建立为快速便捷开展河流港湾等水体的水质监测和赤潮防控提供了新的技术支持。

Satellite remote sensing-based transparency(as in the Secchi depth(SD))inversion methods have been widely applied for oceans,offshore areas,and lakes.However,eutrophic estuaries,harbors,and other water bodies are relatively small in water area with significant land impact,and environmental monitoring in these waters requires timeliness.Due to the restrictions from temporal and spatial resolution,cloud shadowing,atmospheric correction,and other factors,the application in such small water bodies are limited.Therefore,it is urgent to establish an efficient and convenient transparency inversion method for these areas to complement satellite remote sensing methods,which calls for a robust SD inversion specific to these areas.This study attempts to construct a transparency inversion method for Xiangshan Bay,Zhejiang,East China,based on the red(R),green(G),blue(B)three bands using a conventional optical camera co-worked with smartphone HydroColor app and a drone,focusing on the red(R),green(G),and blue(B)bands.We observed that the drone and the app’s red-to-blue(R/B)and red-to-green(R/G)ratios had a strong negative correlation with SD,and the correlation coefficients R was between-0.88 and-0.93(n=16,P<0.001).A transparency inversion model was established based on the results of correlation analysis and the accuracy of the model was evaluated based on an independent database.After the SD inversion model was calibrated,its accuracy was assessed using an independent dataset.Results show that the exponential inversion model is superior to the linear empirical model;the R/G inversion model is superior to the R/B model;and the HydroColor-based inversion model is superior to the drone-based inversion model.Based on these results,optimal SD inversion models were constructed for drone R/G and HydroColor-R/G determination.The average relative error and root mean square error of the drone model were 29%and 0.3 m,while those of the HydroColor model were 21.9%and 0.27 m,respectively.The above results indicate that both drones and smartphone apps can be used to obtain RGB data for SD inversion in Xiangshan Bay.The establishment of this method provides a new technical support for rapid and convenient water quality monitoring and red tide prevention and control in rivers,bays,and similar water bodies.