Implementing an Outgoing Longwave Radiation Climate Dataset from Fengyun 3E Satellite Data with a Machine-Learning Algorithm
China’s FengYun 3E (FY3E) meteorological satellite, launched in 2021, is equipped with advanced instruments for comprehensive Earth observations. In this study, we compared outgoing longwave radiation (OLR) measurements from the FY3E satellite (FY3E OLR) and from a series of satellites operated by the National Oceanic and Atmospheric Agency (NOAA, United States of America; hereafter NOAA OLR) and analyzed the spatiotemporal differences between the datasets. We designed a new correction model, “DeepFM”, implementing both a factorization machine algorithm and a deep artificial neural network to minimize daily mean differences between the datasets. Then, we evaluated the spatiotemporal consistency between the corrected FY3E OLR and NOAA OLR data. The DeepFM model effectively reduced daily mean differences: after correction, the daily mean absolute bias and root-mean-square error decreased from 7.4 W/m2 to 4.2 W/m2 and from 10.3 W/m2 to 6.3 W/m2, respectively, indicating a notably improved spatiotemporal consistency between the corrected FY3E OLR and NOAA OLR data. Subsequently, we merged these datasets to generate a long-term OLR dataset suitable for climate analyses. This study provides a robust technological basis and innovative methodology for the dedicated application of China meteorological satellites to climate science.