Evaluation of radiation schemes in the CMA-MESO model using high time-resolution radiation measurements in China: I. Long-wave radiation

Abstract. Downward long-wave irradiance (DnLWI) is a variable that directly influences the surface net radiation, which in turn affects weather and climate. Due to the lack of abundant DnLWI observations, numerical weather prediction (NWP) models provide a very effective way to yield the DnLWI. Nevertheless, the reliability of the DnLWI predicted by the NWP models needs the evaluation based on the observations or accurate radiative transfer models. In this study, the DnLWI product of the China Meteorological Administration mesoscale model (CMA-MESO) was extensively validated using long-term high time-resolution (1 min) DnLWI measurements carried out at 42 sites in China. The results showed that the DnLWIs predicted by the CMA-MESO model generally agreed well with the observations, i.e., with a relative mean bias error (rMBE) of –2.0 %, but overestimated them under overcast (3.1 %) and underestimated them under dry (–5.3 %) and cloudless (–5.2 %) conditions. It is also found that the discrepancies in the DnLWI predicted by the CMA-MESO model exhibited spatial and diurnal variations, e.g., the discrepancies were significantly smaller at night than those during the day due to the stable nocturnal boundary layer. According to the results of the partial least squares analysis, the high cloud cover, medium cloud cover, planetary boundary layer height, and integrated cloud ice were the most important factors affecting the accuracy of the DnLWIs predicted by the CMA-MESO model under all sky conditions. By comparing the outputs of the CMA-MESO model and the MODerate resolution atmospheric TRANsmission (MODTRAN) model, it was found that the uncertainties in the DnLWI predicted by the CMA-MESO model mainly arose from an inappropriate consideration of the high and medium clouds under all sky conditions as well as shortcomings in the radiation scheme under cold dry cloudless conditions.