Scene-based turbid water atmospheric correction for GOCI

Abstract

This study follows up on the study of Lee et al. (2012) for the scene-based turbid water atmospheric correction for the GOCI (Geostationary Ocean Color Imagery) theoretically based on Ruddick et al. (2000), so-called MUMM. The MUMM atmospheric correction algorithm which was developed for the SeaWiFS data processing accurately derives the water-leaving reflectance (ρw) over moderately turbid waters with the applicable range of ρw (745) < 0.01. However in extremely turbid water with 0.01 < ρw (745), it underestimates the water-leaving reflectance and overestimates the aerosol reflectance because the ratio between two NIR water-leaving reflectance (α) is assumed to be constant whereas it can be changed with a concentration of suspended sediment in the extremely turbid region (Ruddick et al., 2006 Doron et al., 2011). To improve this phenomenon, some studies have been conducted using an polynomial NIR-model (Wang et al., 2012 Ahn et al., 2015). In the case of the updated MUMM algorithm (uMUMM), calculates appropriate α value by using an NIR model. The objective of this study is that analyze the overall performance of the uMUMM algorithm. We compare the polynomial model of uMUMM with others (i.e. the previous study, in-situ data, and radiative transfer code). Also, we conduct match-up analysis using In-situ Rrs (VIS) dataset. Lastly, we discuss the limitations of the algorithm.heric correction algorithm which was developed for the SeaWiFS data processing accurately derives the water-leaving reflectance (ρw) over moderately turbid waters with the applicable range of ρw (745) < 0.01. However in extremely turbid water with 0.01 < ρw (745), it underestimates the water-leaving reflectance and overestimates the aerosol reflectance because the ratio between two NIR water-leaving reflectance (α) is assumed to be constant whereas it can be changed with a concentration of suspended sediment in the extremely turbid region (Ruddick et al., 2006 Doron et al., 2011). To improve this phenomenon, some studies have been conducted using an polynomial NIR-model (Wang et al., 2012 Ahn et al., 2015). In the case of the updated MUMM algorithm (uMUMM), calculates appropriate α value by using an NIR model. The objective of this study is that analyze the overall performance of the uMUMM algorithm. We compare the polynomial model of uMUMM with others (i.e. the previous study, in-situ data, and radiative transfer code). Also, we conduct match-up analysis using In-situ Rrs (VIS) dataset. Lastly, we discuss the limitations of the algorithm.