Inverse Estimation Of Chlorophyll-a Concentration In Taihu Lake From Remote Sensing Based On Simulation And Correction Of Backscatter

In turbid water,different optical active substances influences the remote sensing reflectance measured above the surface water.Radiative transfer of light in water body mainly includes absorption and scattering.Backward scattering is an important inherent optical properties,and closely related to the water surface spectrum.As an important water quality parameters,chlorophyll-a concentration is the main indicator of water eutrophication.Inversing estimation of chlorophyll-a concentration base on spectrum is an important work content in water color remote sensing.To improve application accuracy of the chlorophyll-a concentration remote sensing inversion model,this study proposed a backscattering correction method based on the relationship between backscattering spectrum and water surface spectrum.Using the 14 field observation data and laboratory experimental water quality parameters from 2010?2016 in Taihu Lake,the paper discussed the relations among the backscattering spectrum,suspended sediment concentration and chlorophyll-a concentration,analyzed influence of the backscattering spectrum to water surface spectrum,and proposed a spectrum correction method,then built the chlorophyll-a concentration inversion model and validate the model.The main contents and conclusions are as follows:(1)Backward scattering coefficient simulation and its applicability in Taihu LakeFour methods of backscattering coefficient simulation were used,including the optical closure model by Gordon in 1988,QAA model by Lee in 2002,the improved QAA model by Le in 2009 and the IIMIW model by Li in 2013,to calculate the backscattering coefficient of the water body in Taihu Lake,the similarity and the difference between the simulated and the measured were compared.The results showed that the simulated have the same tendency with the measured,i.e,the backscattering coefficient is higher in the blue band and lower in the red band.The values range from the different simulation methods are different,even for the same water area,some are 0?0.4m-1,and some 0?4m-1.The simulated from the improved QAA model and the IIMIW model has a better consistency with the measured.(2)The backward scattering correction on the water surface spectrum and chlorophyll-a concentration inversionAccording to the change of backscattering spectrum,water bodies were divided into two kinds:the suspended sediment dominantly and phytoplankton dominantly.The backscattering coefficient of suspended sediment is higher,the backscattering coefficient of planktonic algae is lower.When the suspended sediment concentration is higher than 75mg/L or the chlorophyll-a concentration is higher than 90?g/L,the reflection of 690nm will move to red band and the offset has a good positive correlation with the backscattering coefficient.According to the relationship between backscattering spectrum and water surface spectrum,the backscattering correction method and the chlorophyll-a concentration inversion model of the spectral are developed respectively.Rrs(705)/Rrs(670)was used as the criterion for different correction.When Rrs(705)/Rrs(670)<1.6,the method 1 is used,and Rrs(705)/Rrs(670)1.6,the method 2 is used.For the suspended sediment dominantly water,the backscattering spectrum is normalized as the baseline at 675nm of water surface spectrum.After the water surface spectrum is subtracted from the baseline(method 1),the optimal band combination r'(702)/r'(668)was used to build the model,and the mean relative error of the model validation is 23.87%.For the planktonic algae dominantly water,the backscattering spectrum was normalized at 443nm of water surface spectrum(method 2).After the water surface spectrum was subtracted from the baseline,the optimal band combination r'(678)/r'(668)was used to build the model,and the mean relative error of the model validation is 11.10%.The model diagnosis shows that the residuals are of the normal distribution,and the confidence interval of the residuals is also improved.(3)Inversion model based on the backscattering correctionData set were divided into two parts,103 samples to build the model,52 samples to validate the model.Both the model accuracy(determination coefficient=0.95,root mean square error=10.66,mean absolute error=7.46?g/L,mean relative error=27.38%)and validation accuracy(determination coefficient=0.85,root mean square error=18.77,mean absolute error=8.85?g/L,mean relative error=25.39%)are better than the three-band model and band ratio model used commonly.The proposed method of this study decreased the overestimation of low chlorophyll-a concentration(under high suspended sediment concentration)and high chlorophyll-a concentration(under low suspended sediment concentration)in previous research and is useful to build the estimation model of chlorophyll-a concentration in other turbid water bodies.