Secure Consistency of COMS/MI and GK-2A/AMI Shortwave Radiation product based on machine learning

Abstract

 Satellite-based solar radiation data is widely used to monitor global climate and environmental changes and is also actively used to analyze weather data and predict particulate matter. Korea can continuously retrieval solar radiation in the observation area due to the generational shift of COMS (Communication, Ocean and Meteorological Satellite)/MI (Meteorological Imager sensor) and GK-2A (GEO-KOMPSAT-2A)/AMI(Advanced Meteorological Imager sensor). However, The quality of each solar radiation output is different due to difference between the algorithms, input data and resolution. Therefore, it is possible to produce a climate resource map for the Korean Peninsula for continuous climate change monitoring by analyzing the error characteristics between the solar radiation of COMS/MI and GK-2A/AMI and expanding the retrieval period through correction between the two products. In this study. We analyzed the error characteristics of the two satellites compare to the meteorological observation data of Korea and the satellite CERES solar radiation data in overlapping periods. As a result of error analysis, the RMSE of COMS/MI was 85.6 (W/m2), lower than the RMSE of GK-2A/AMI, 95.6 (W/m2). Considering the solar radiation data error characteristics of these satellites, a correction model based on machine learning techniques was created to secure the consistency of solar radiation data. When this was verified with in situ data for a period of 10 years, RMSE was 89.21 (W/m2) and Bias was 17.39 (W/m2), which was stable in the temporal consistency test, and the annual increase in solar radiation on the Korean Peninsula was confirmed. ※ This work was supported by the "Graduate school of Particulate matter specialization." of Korea Environmental Industry & Technology Institute grant funded by the Ministry of Environment, Republic of Korea.