Retrieval of outgoing longwave radiation at top-of-atmosphere using Himawari-8 AHI data

Abstract

This study proposes an algorithm to retrieve the outgoing longwave radiation at the top of the atmosphere (TOA OLR) using the advanced Himawari imager (AHI) unit that is a narrowband sensor installed in Himawari-8, a geostationary satellite providing window and water vapor imagery. The outgoing longwave radiation at the top of the atmosphere (TOA OLR) was retrieved by an empirical method where a radiative transfer model (RTM) was used to simulate various atmospheric conditions, such as the surface temperature, water vapor, and cloud characteristics (e.g., cloud optical thickness and cloud height). The results of an algorithm using a single channel (OLR12.4) and two channels (OLR6.2+12.4) were converted into Terra Cloud and Earth's Radiant Energy System (CERES) TOA OLR for the daytime (0105–0135 UTC) and nighttime (1325–1355 UTC) cases of a scene analysis (typhoon Goni on August 18, 2015). Long-term analysis (August 2015–July 2016) and statistical analysis (i.e., mean, bias, root mean square error (RMSE), and correlation coefficient) were conducted. The analysis results showed that the largest error was produced in the cloudy areas (overcast areas), and the minimum RMSE and a high correlation coefficient were observed over an ocean area. In the scene analysis, however, the daytime and nighttime average RMSE of the CERES TOA OLR and OLR6.2+12.4 was 12.21Wm−2 and showed the correlation coefficient of 0.971. It was an improved result over the RMSE (12.33Wm−2) and correlation coefficient (0.967) shown in the analysis of OLR12.4. In the long-term analysis, the average RMSE was 11.83Wm−2 and showed a correlation coefficient of 0.972 (OLR12.4 showed 12.17Wm−2 and 0.969). Furthermore, to minimize the difference in the cloud area that had the largest error, the study analyzed TOA OLR at the significance level of 95% (±2 RMSE). As a result, the scene analysis showed an RMSE of 10.54Wm−2 and correlation coefficient of 0.977, and the long-term data showed an RMSE of 10.23Wm−2 and correlation coefficient of 0.979.