The propagation of electromagnetic waves beyond the line of sight can be caused by atmospheric ducts, which are significant concerns in the fields of radar and communication. This paper utilizes data from seven automatic weather stations and five radio-sounding stations to statistically analyze the characteristics of the atmospheric ducts in the northwest region of the South China Sea (SCS). After verifying the practicality of numerical analysis data from NCEP CFSv2 and ECMWF in studying atmospheric ducts using measured data, we analyzed the spatial–temporal distribution characteristics of the height of the regional evaporation duct and the bottom height of the elevated duct. The study found that the NCEP CFSv2 data accurately capture the evaporation duct height and duct occurrence rate in the study area, and the elevated duct bottom height calculated from ERA5 and the measured data have good consistency. The occurrence rate and height of the evaporation duct in coastal stations in the northwest of the SCS vary significantly by month, demonstrating clear monthly distribution patterns; conversely, changes in the Xisha station are minimal, indicating good temporal uniformity. For lower atmospheric ducts, the difference in occurrence rates between 00:00 and 12:00 (UTC) is negligible. The occurrence probability of elevated ducts in the Beibu Gulf area is relatively high, mainly concentrated from January to April, and the Xisha area is dominated by surface ducts without foundation layers, mainly concentrated from June to August. Monsoons play a critical role in the generation and evolution of atmospheric ducts in the northwest of the SCS, with the height of the evaporation duct increasing and the bottom height of the elevated duct decreasing after the onset of the summer monsoon. In the end, we simulated electromagnetic propagation loss under different frequencies and radiation elevation angles in various duct environments within a typical atmospheric duct structure.

There are two complementary ways to understand the model bias in climate simulations. One is via the observation-model difference, which is widely used in model evaluation as well as in model tuning during the development phase of the models (e.g., Flato et al., 2013). The other is based on the dispersion among different model results, which is also called intermodel spread or model spread. The small intermodel spread usually implies small model bias, but not necessarily so because of the existence of structural bias in the models. While the structural bias must come from incomplete knowledge in correctly representing relevant processes in climate models, the origin of intermodel spread is more complicated. Indeed, except that the diversity in representing/ parameterizing the model processes may cause the intermodel spread, the intrinsically chaotic nature of coupled climate system may also lead to the intermodel spread. Therefore, the intermodel spread should not simply be