Calculation Of Remote Sensing Reflectance Based On Radiative Transfer Model And Analysis Of Chlorophyll Retrieval Algorithm

This thesis modeled the remote sensing reflectance (Rrs) of case 2 waters in Yangtze estuary and its adjacent waters combined with the regional bio-optical models by radiative transfer numberical model Hydrolight. Meanwhile,some analytical Rrs model has a good fitting when compared with in-situ data. Four retrieval algorithms of Chlorophyll-a (Chla) concentration were analyzed for their adaptability and sensitivity in highly turbid waters, serving as a reference for future study.This thesis studied the absorption and scattering coefficient of main optical components of water, such as Suspended Particulate Matter (SPM), Chlorophyll-a (Chla) and yellow substances (CDOM). Three Chla absorption coefficient models were parameterized with field measured data and compared, it was found that the polynomial model with the reference wavelength at 440nm has a good performance in both high and low Chla concentration conditions, while the polynomial model with the reference wavelength at 675nm underestimated the Chla concentration, and exponential model overestimate Chla concentration at 400～550nm. So the polynomial model with the reference wavelength at 440nm was chosen to calculate the Chla absorption coefficient. Besides, three SPM backscattering coefficient models were compared, it was found that Liu model (2013) had smaller RMSE(Root Mean Square Error) when compared with Kuang(2010)’s and David(2007)’s model by applying to field measured data. So Liu model(2013) was chosen to calculate SPM back-scattering.Based on the bio-optical models metioned above as the inputs of Hydrolight, we simulated the Rrs of different water types. When compared with the in-situ data, the RMSE was less than 0.01sr-1 for all data, showing that simulated Rrs at 550～725nm bands has a high accuracy. Through compared with Rrs’s simulated models, it indicated that Kubelka-Munk(KM)(Shen et al.,2010a; Sokoletsky et al.,2013), AGM(Albert and Gege,2006) and Hydrolight numberical models have a good ability of Rrs’simulation compared with in-situ data when concentration of SPM is 7.4g/m3, concentration of Chla is 1.3mg/m3, and absorption of CDOM at the wavelength of 440nm is 0.25m-1, while LM(Lee et al.,1994) and GM(Gordon et al.,1998) models show the obvious underestimate. Rrs was impacted by SPM due to its high scattering at wavelengths higer than 550nm. Rrs was modified by Chla due to its high absorption in blue and red bands. CDOM absorption can significantly impact on Rrs in 400-700nm.With the in-situ Rrs data, we used two-band, three-band, Fluorescent Light Height (FLH), and Synthetic Chlorophyll Index (SCI) algorithms for Chla retrievals. Results indicated that the two-band, three-band and SCI algorithms show the similarity RMSE, while FLH algorithm shows much bigger RMSE.The impact of SPM on Rrs decrease with an increase of Chla concentration, which means that the sensitivity decreased. The two-band and three-band algorithms were suitable for Chla retrieval with low SPM concentration. The FLH algorithms was lose its ability under the concentration of SPM higher than 10g/m3 when retrieving Chla, while SCI exhibited a better potential of weaking the influence of SPM when retrieving Chla in highly turbid SPM rich waters. The two-band and three-band algorithms are not sensitive to the change of total absorption and total back-scattering coefficients, while the FLH and SCI algorithms show a better sensitivity.