LI Hao, a postdoctoral student in our laboratory, published a paper on the atmospheric correction model of geostationary water color satellites under large solar zenith angles in the top international remote sensing journal "Remote Sensing of Environment". The corresponding author is researcher HE Xianqiang of our laboratory.
The hourly observation data of the geostationary water color satellite GOCI have been widely used to monitor high-frequency dynamic changes in the offshore environment. However, standard atmospheric correction algorithms have large uncertainties and are even ineffective when applied to GOCI data under large solar zenith angles at dawn and dusk. In this study, the author proposes a new neural network atmospheric correction model to process GOCI morning and evening observation data. Different from previous training data sets obtained from radiative transfer simulations, this study proposes to directly use a large amount of GOCI observation data to construct a training data set, which overcomes the difficulty of the current radiative transfer model in effectively reproducing actual satellite observations under large solar zenith angles.
Verified by simulated data sets, field observation data and GOCI observation data (Figures 1-3), it is shown that the neural network atmospheric correction model is reliable and can handle GOCI data with solar zenith angles up to 85°. The remote sensing reflectance obtained by inversion based on the neural network algorithm can be further inverted to obtain hourly water color component information from morning to night (such as chlorophyll concentration, suspended matter concentration, etc.), which provides a basis for studying intraday changes in the marine ecological environment. In short, the new neural network algorithm can effectively process the twilight and twilight observation data of stationary water color satellites, and can recover twilight and twilight water color products that cannot be processed by standard atmospheric correction algorithms. In addition, this method can also be used to solve the problem of polar-orbiting water color satellite data processing in high-latitude seas in winter.
Figure 1 Comparison of remote sensing reflectance retrieved from three atmospheric correction algorithms and AERONET-OC field measurement data. (a)-(e) Comparison of individual bands. (f) Comparison of measured data for all solar zenith angles greater than 70°.
Figure 2 The hourly 443nm band remote sensing reflectance product obtained by GOCI through the neural network atmospheric correction algorithm on March 16, 2018.
Figure 3 GOCI’s remote sensing reflectance product in the 555nm band of the East China Sea on February 23, 2018. (a)-(h) are the inversion results of the neural network atmospheric correction model; (i)-(j) are the inversion results of the KOSC atmospheric correction model; (k)-(l) are the inversion results of the NIR atmospheric correction model result.
Citation: Li, H., X. He*, Y. Bai, P. Shanmugam, Y.-J. Park, J. Liu, Q. Zhu, F. Gong, D. Wang, and H. Huang (2020) , Atmospheric correction of geostationary satellite ocean color data under high solar zenith angles in open oceans, Remote Sensing of Environment, 249, 112022.
