Study On Recognition Models Of Urban Black And Odorous Water Bodies Based On Optical Characteristics

In the cities,urban black and odorous water bodies distribute widely,and cause serious damage to people's living environment and the overall aesthetic image of the city.Due to human screening is time-consuming,laborious and prone to omissions.In comparison,satellite remote sensing technology can achieve continuity,dynamics,and monitoring of black and odorous water bodies by points and surfaces.But at present,this technology has less practical application and recognition accuracy is not high.Therefore,taked small and medium-sized rivers in Shanghai as the research object,based on the measured hyperspectral data,fitted GF-2 multispectral data and water quality,this paper eatablished a spectral classification system,quantitatively inverted water quality parameters to determine the characteristics of black and odorous water bodies in Shanghai,and used multiple models to qualitatively distinguish general water bodies from black and odorous water bodies,provided a basis for the supervision of black and odorous water bodies by remote sensing technology in the future.The main conclusions were as follows:?1?Divided urban water bodies into three types of general water bodies?GW1,GW2,and GW3?and three types of black and odorous water bodies?BOW1,BOW2,and BOW3?.The optical characteristics of BOW1,BOW3 were different obviously from other types;for BOW1,the value of Chl-a,d₁₀,d₅₀,d₉₀ were the smallest in six kinds of water bodies,the TSM value was the largest,and the fluctuation of BOW1spectral curve was the most gentle;while for BOW3,the value of Chl-a,d₁₀,d₅₀,d₉₀were the largest in six types of water,and the spectrum curve had the largest variation.?2?The value of Chl-a and COD_Cr could be inverted directly based on satellite multispectral,then used as the characteristic indexes of Shanghai water bodies,but the errors of model prediction had obvious concentration distribution.When inverted the value of Chl-a based on fitted data,R² value was 0.660,MRE value was larger when Chl-a was less than 20ug/L,and MRE value decreased to 34.17%when Chl-a was greater than or equal to 20ug/L;When inverted the value of COD_Cr based on fitted data,R² value was 0.685,RE value was larger when COD_Cr was less than 10ug/L,and RE value decreased to below 50%when COD_Cr was greater than or equal to 10ug/L,and overall MRE value was 54.30%.The correlation between the value of DO,COD_Mnn with fitted multispectral data was low,so DO,COD_Mn only could be inverted based on measured data.R² values were0.627 and 0.628 in turn,and there were obvious concentration distributions for both indicators.When DO was greater than or equal to 4mg/L,RE values were between 0%and 50%,and overall MRE value was 45.73%;when COD_Mn was less than 4mg/L,RE values were between 50%and 120%,and when COD_Mn was greater than or equal to4mg/L,RE values were between 0%and 50%,and overall MRE value was only 24.59%.When analyzing TSM value and remote sensing reflectance,considered the change of particle size could improve the correlation,because the remote sensing reflectance of small particle size was larger than that of large particle size.Based on measured data or fitted data,constructed regression models with TSM,particle size of d₁₀,d₅₀ and d₉₀,but there was no model which R² value was more than 0.6.However,compared with TSM,the correlation of particle size of d₁₀,d₅₀ and d₉₀ with each band showed an upward trend.?3?When identified black and odorous water based on measured data,the threshold of partitioning model followed the principle of"trying to reduce the misjudgement of black and odorous water bodies to general water bodies".The validation rate of hyperspectral model for black and odorous water bodies was up to94.44%,but the validation rate for general water bodies was only 58.82%.Applied the spectral index method in other studies,100%of BOW1 could be identified,but more than 90%of GW1 would be misjudged at the same time,and validation rates of BOW2and BOW3 were only 61.54%and 40.91%.?4?When identified black and odorous water based on fitted data,the validation rate of color method and saturation method was low.The?₁ method of angle was suitable for identifying BOW1?100%?,while the ratio method was suitable for BOW2?92.31%?and BOW3?95.45%?.Therefore,an angle-ratio combination model was formed by combining the two models.The validation rate of this combination method for black and odorous water bodies was 95.92%,while increased by 6.12%than angle model?89.90%?.The misjudgment rate of this combination method for general water bodies was 39.29%,while increased by 5.36%than ratio model.The BOI method,SBWI method and NDBWI method used in other studies could only recognize BOW1type and the verification rate were less than 60%.After changed the threshold values of three methods to 0.11,0.013 and 0.15,100%of BOW1 water bodies could be identified,and the judgment rate of SBWI model for GW1 was the lowest,only 21.21%.