Development of Sea Surface Temperature Retrieval Algorithms for Geostationary Satellite Data (Himawari-8/AHI)

Abstract

This study provides an overview of the development of sea surface temperature retrieval algorithms by using Himawari-8/AHI data as a proxy data of GK-2A with quite similar spectral bands except for 2.26-μm and 1.38-μm bands. For contingency preparation, several potential algorithms, such as Multi-channel SST (MCSST), Non-linear SST (NLSST), Hybrid SST, and Multi-band SST, were developed over the full disk region. The accuracy of each algorithm was assessed by determining the root mean square error (RMSE) and bias errors from the regression procedure of the matchup database between satellite data and quality controlled drifter temperature in-situ data for a year, from August 2016 to July 2017. Comparison of the four algorithms revealed that the Multi-band algorithm performed markedly well, with the smallest RMSE of ~0.4 °C. Time-varying validation of the estimated SST accuracy highlighted consistently low RMSE as well as the stability of the Multi-band algorithm. In addition, it is suggested that SSTs with a satellite zenith angle exceeding 60° tended to have relatively large errors which degraded the quality of the estimated SSTs. It is concluded that the SST coefficients should be updated each day, based on the previous one-month matchup database, contributing to the expected SST accuracy in the future with the degradation of the sensor or other aging effects. Further, this work discusses the importance of cloudy or cloud-contaminated pixels for the better performance of SST retrieval procedures and their real-time operational use.