Evaluation of the Accuracy of Atmospheric Radiation Calculations Using Different Computational Schemes in the ECRAD Radiation Model for Cloudy and Cloudless Skies

Improving the quality of modeling the net radiation components is an urgent task that contributes to the development of modern models for numerical weather and climate forecasting. This study provides estimates of the accuracy of various computational schemes in the widely used ecRad radiation model based on comparisons with measurements of the RAD-MSU (BSRN) radiation complex at the Meteorological Observatory (MO) at Moscow State University for clear and cloudy conditions for the warm, snowless period from August 2021 to October 2023. For cloudless sky conditions, the reproduction of the total net radiation using CAMS aerosol reanalysis and ERA-5 reanalysis is satisfactory (error less than 4%); however, there is some overestimation of the shortwave (SW net) and underestimation of the longwave (LW net) net radiation. At the same time, the reproduction quality of the total net radiation is mainly determined by SW net. It is shown that, in different ranges of the cloud amount, SW net can be significantly underestimated (up to ‒130 W/m2) and overestimated (up to 100 W/m2), depending on the errors in reproducing direct shortwave radiation in almost all algorithms. At the same time, the use of the McICA max-ran algorithm, in general, leads to smaller median differences with measurements. LW net is reproduced by the all methods in approximately the same way, with mostly positive differences in the range of 10 W/m2. Additionally, the three-dimensional (3D) radiation effects of clouds in the SPARTACUS algorithm were evaluated, along with the accounting of the third additional cloud state (3reg). The 3D radiation effects of clouds in the shortwave spectrum are most pronounced for cloud amount 5–7 and 7–9 (about 6 W/m2 for global radiation), while the effect of accounting for additional cloud state is significantly higher (up to 20–22 W/m2 for global radiation). In conditions of continuous cloud cover with optically dense clouds, the most effective method is the SPARTACUS algorithm with three-dimensional effects and two cloud states.