Generation of global 1-km daily top-of-atmosphere outgoing longwave radiation product from 2000 to 2021 using machine learning

Abstract

Top-of-atmosphere (TOA) outgoing longwave radiation (OLR), a key component of the Earth’s energy budget, serves as a diagnostic of the Earth’s climate system response to incoming solar radiation. However, existing products are typically estimated using broadband sensors with coarse spatial resolutions. This paper presents a machine learning method to estimate TOA OLR by directly linking Moderate Resolution Imaging Spectroradiometer (MODIS) TOA radiances with TOA OLR determined by Clouds and the Earth’s Radiant Energy System (CERES) and other information, such as the viewing geometry, land surface temperature and cloud top temperature determined by Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2). Models are built separately under clear- and cloudy-sky conditions using a gradient boosting regression tree. Independent test results show that the root mean square errors (RMSEs) of the clear-sky and cloudy-sky models for estimating instantaneous values are 4.1 and 7.8 W/m2, respectively. Real-time conversion ratios derived from CERES daily and hourly OLR data are used to convert the instantaneous MODIS OLR to daily results. Inter-comparisons of the daily results show that the RMSE of the estimated MODIS OLR is 8.9 W/m2 in East Asia. The developed high resolution dataset will be beneficial in analyzing the regional energy budget.

Keywords

• toa outgoing longwave radiation
• modis
• ceres
• machine learning
• earth’s energy budget