The application of a hybrid sea surface temperature algorithm to COMS geostationary satellite data in the Northwest Pacific Ocean

• Published in: International journal of remote sensing Vol. 41; no. 15; pp. 5953 - 5973
• Main Authors: Park, Kyung-Ae, Woo, Hye-Jin, Lee, Eun-Young
• Format: Journal Article
• Language: English
• Published: London Taylor & Francis 02.08.2020; Taylor & Francis Ltd
• Subjects: Algorithms; Brightness temperature; Coefficients; Communication; Errors; hybrids; Mathematical models; Meteorological satellites; Numerical models; Oceans; Pacific Ocean; Regression; remote sensing; Satellite data; Satellite observation; Satellites; Sea surface; Sea surface temperature; Surface radiation temperature; Surface temperature; surface water temperature; Synchronous satellites; Pacific Ocean
• ISSN: 0143-1161; 1366-5901; 1366-5901
• DOI: 10.1080/01431161.2019.1688889

The hybrid sea surface temperature (SST) algorithm derives coefficients for SST estimation based on the regression procedure of incremental values between the satellite-observed brightness temperature (BT) minus the first-guess BT and the in-situ temperature minus the first-guess SST. In this study, the hybrid SST algorithm was applied to retrieve the SST coefficients of which the results were compared with extensively used regression algorithms of Multi-Channel SST (MCSST) and Non-Linear SST (NLSST) using data obtained by the Meteorological Imager of the first Korean geostationary satellite, the Communication, Ocean and Meteorological Satellite, in the Northwest Pacific Ocean in May 2014. The empirical regression SSTs and hybrid SSTs exhibited similar root-mean-squared errors in the range of 0.61-0.84°C. Although the SST errors were quite similar to each other, the comparison of the hybrid SSTs with the in-situ SSTs indicated a slightly higher accuracy of approximately 0.2°C compared with that of the empirically derived SSTs. SST retrieval using the hybrid algorithm reduced the errors near the cloud edges. The differences between the hybrid SST and regression SST tended to be greater when the distance to the nearest cloudy pixel was less than 0.2°. This study discusses the potential causes of the differences between the SST errors of the algorithms and addresses a few constraints in the execution of the hybrid algorithm related to the numerical model.