Study On Remote Sensing Identification And Classification Of Urban Black And Odor Water Based On Water Absorption Coefficient

The urban black and odorous water presents black or gray with loss of ecological function with a pungent,disgusting smell.As urbanization process accelerates,the ecosystem of urban rivers also faces great challenges.Due to the discharge of industrial sewage and domestic wastewater,many water bodies are black and smelly,which seriously affects people's living environment.The current researches on monitoring urban black and odorous water bodies via remote sensing reflectance mostly recognize it from the water color.The remote sensing reflectance of the water surface is affected by the inherent optical properties such as absorption and scattering.Therefore,the change in the inherent optical properties of the water can directly express the difference in water color.In this study,the water body absorption coefficient is taken as the main index to explore the best characterization parameters of water color difference,and the identification and classification model of urban black and odorous water is constructed.This study takes urban water in Nanjing,Wuxi,Yangzhou and Changsha as research objects,analyzes the optical characteristics of black and odorous water by using field measured data,and constructs the identification and classification algorithm of urban black and odorous water based on absorption coefficient.An estimation model of total particulate matter and CDOM absorption coefficient based on hyperspectral data is established,and the model is applied to the hyperspectral image of the drone to classify the level of the urban river section.The main contents and conclusions of this paper are as follows:(1)Analysis of differences in absorption characteristics of water bodies with different levelsCompared with other water bodies,the urban black and odorous water bodies have the lowest remote sensing reflectance,which is lower than 0.03sr-I.The reflectance of normal rivers is the highest,followed by mild black-odorous.Between 400 and 550nm,as the level of black odor increases,the slope of the reflectance raises more slowly.The spectrum of serious black-odorous appears low peak at the range of 550nm-580nm and the peak width of serious black-odorous is larger than other types of water.There are differences in the absorption coefficients of urban black and odorous water bodies with different levels.The absorption coefficient of non-pigment particles(ad),pigment particles(aph)and CDOM(ag)of serious black-odorous are higher than those of mild black-odorous and urban normal water bodies.The average absorption coefficient of the pigment at 440nm in serious black-odorous is approximately 1.1 times that of mild black odorous and 1.3 times that of normal water bodies.The average absorption coefficient of non-pigment at 440nn in serious black-odorous is roughly 1.4 times that of mild black-odorous and 1.8 times that of normal water bodies.The average absorption coefficient of CDOM at 440nm in serious black-odorous is roughly 1.0 times that of mild black-odorous and 1.4 times that of normal water bodies.(2)Identification and classification index of black and odorous waterThe high CDOM absorption characteristics of black and odorous water has potential application in the identification and supervision of black-odor water bodies.The slope value S obtained by fitting the CDOM absorption coefficient parameterized model in different wavelength ranges could distinguish black and odorous water bodies from normal water bodies.As the wavelength increases,the discrimination of S value reduces:The overall accuracies of S1 value fitting from 275 to 295nm,S2 value fitting from 350 to 400nm and ag(440)are 79.39%?72.12%and 76.36%,respectively.The absorption characteristics of particulate matter in serious black-odorous are significantly different from those in mild black-odorous.The high particulate matter absorption characteristics of serious black-odorous play an important role in the classification of black-odorous.Three parameters are proposed to represent the absorption characteristic of particulate matter including falling rate of total particulate matter in the range of 400?500 nm(K400-500),curve integral area in the range of 400?500nm(SA(400-500))and the absorption coefficient of total particulate matter at 440nm(ap(440)).The accuracy of each is greater than 80%in distinguishing serious black-odorous and mild black-odorous.Among them,the curve integral area(SA(400-500))and the absorption coefficient(ap(440))have the best effect,reaching 87.65%.Considering the simplicity and operability of the actual process,the CDOM absorption coefficient at 440nm(threshold is 1.25m-1)is selected as the identification index of black odorous water and non-black odorous water,the total particulate absorption coefficient at 440 nm(threshold is 7m-1)as a classification index for mild black-odorous and serious black-odorous.(3)Inversion of absorption coefficient of black and odorous waterThe ratio of remote sensing reflectance between 720nm and 560nm is selected,and a linear exponential model is established to estimate the total particle absorption coefficient.The root mean square error is 2.14m-1,and the mean absolute percentage error is 27.26%.The multivariate regression method established by the ratio of remote sensing reflectance at 665nm/720nm and 560nm/665nm to estimate CDOM absorption coefficient at 440nm.The root mean square error of the estimated CDOM absorption coefficient is 0.30m-1,and the mean absolute error percentage is 45.27%.The improved QAA-640 algorithm is used to invert the absorption coefficient,and the root mean square error of the total particle absorption coefficient is 2.67m-1,the mean absolute percentage error is 54.34%.The root mean square error of CDOM absorption coefficient is 0.73m-1.The mean absolute percentage error is 51.81%.Therefore,for urban black and odorous water bodies,the empirical algorithm is better than the QAA semi-analytic algorithm.(4)Identification and classification of black and odorous water based on hyperspectral data of the droneBased on the remote sensing image of the UAV on December 22,2017 and September 27,2018,the absorption coefficient is inverted by the empirical algorithm,and the water body of the rivers is identified and classified according to the threshold of index.Among the four rivers,Jinchuan River,Longyuan West Road River and Xuanwu Lake are normal water bodies.The Tunliang River is divided into three sections,and the water quality of the three sections varies greatly.The southern part is a type of mild black-odorous and serious black-odorous;the middle part has good water quality and is normal water;the northern part of the river has the worst water quality,and is basically serious black-odorous.