Particulate Backscattering Characteristics and Remote Sensing Retrieval in the Zhanjiang Bay
Based on the in-situ investigation of Zhanjiang Bay in January 2018,the in-situ remote sensing reflectance(Rrs),particulate backscattering(bbp),chlorophyll a concentration(Chl a)and inorganic suspended matter concentration(ISM)were obtained.The backscattering characteristics of particulates in Zhanjiang Bay were analyzed,and the backscattering coefficients of particulates were retrieved by remote sensing.The research results showed that the coefficients of variation(CV)of bbp in the six bands(420,442,470,510,590 and 700 nm)were between 50％～60％in surface water,and the variation range was 0.026 1～0.211 2 m-1,which also mean the complexity of optical properties in water.In order to better quantify the spectral characteristics of bbp,the power function spectral model of bbp was constructed with 510 nm as the reference band,and the slope index of the spectral model was 1.55.In the meantime,the bbp(510)had a power relationship with ISM and an exponential relationship with particulate composition(Chl a/ISM),while the determination coefficient(R2)was 0.74 and 0.62,respectively.It indicated that the first-order driving factor of particulate backscattering in the bay was mainly the concentration of inorganic suspended matter,and the second-order driving factor of particulate composition also contributed to the variation of bbp(510).In addition,in order to accurately estimate the particulate backscattering coefficient in Zhanjiang Bay,a random forest model was constructed based on in-situ remote sensing reflectance,and compared with three semi-analytical algorithms such as QAA-v6,QAA-RGB and QAA-705.The R2 of rando m forest model was 0.86,the mean absolute percentage error(MAPE)was 12％,the root mean square error(RMSE)was 0.02 m-1,the R2 of QAA-v6,QAA-RGB and QAA-705 was 0.63,0.71 and 0.53,the MAPE was 186％,117％and 243％,and the RMSE was 0.16,0.09 and 0.18 m 1,respectively.Although the three semi-analytical algorithms also had high R2,there were significant differences between the estimated and measured values,and the MAPE and RMSE were also large.The retrieval accuracy of three semi-analytical algorithms was significantly lower than that of the random forest method,which indicated that the random forest method had great potential application when using remote sensing to retrieve the bbp in Zhanjiang Bay.
万方数据
维普
