Estimating Hourly All-Sky Surface Longwave Upward Radiation Using the New Generation of Chinese Geostationary Weather Satellites Fengyun-4A/AGRI

Surface longwave upward radiation (SLUR) is a key parameter in studying hydrological and climate models. This study develops a framework for estimating the all-sky SLUR from the Advanced Geostationary Radiation Imager (AGRI) onboard the Chinese geostationary weather satellite FengYun-4A (FY-4A). The framework is composed of a hybrid method for estimating clear-sky SLUR and a machine learning (ML) method for estimating cloudy-sky SLUR. According to the in situ validation, the
$R^{2}$
/bias/root mean square error (RMSE) of the developed hybrid method is 0.95/0.59/18.41 W/m2, which is clearly superior to the AGRI official SLUR and ERA5 SLUR with an
$R^{2}$
/bias/RMSE of 0.95/−7.79/19.04 and 0.92/-4.94/23.2 W/m2, respectively. The developed hybrid method performs better than the classical land surface temperature-broadband emissivity (LST-BBE) method. The
$R^{2}$
/bias/RMSE of the developed cloudy-sky SLUR estimate light gradient boosting machine (LightGBM) model is 0.85/0.56/21.16 W/m2, which is also better than the accuracy of the LST-BBE method and comparable to the accuracy of the ERA5 SLUR. The
$R^{2}$
, bias, and RMES of the all-sky SLUR are 0.93, 0.57, and 19.58 W/m2, respectively. The developed framework is employed to determine hourly all-sky SLUR from AGRI data. This study provides a promising solution to obtain hourly all-sky SLUR from geostationary satellites (GOESs).