Numerical weather prediction (NWP) of radiation fog, particularly over complex terrain, remains a formidable challenge. Many operational NWP models often struggle with slow or no fog formation after sunset and too rapid dissipation in the morning. This study investigates the role of physical processes in the atmospheric boundary layer (ABL) in shaping the limitations of fog and low stratus (FLS) representation within the operational ICON model. Specifically, it evaluates the effects of turbulence parameterizations and vertical resolution on fog simulations. ICON simulations were conducted for selected winter periods characterized by persistent radiation fog, nocturnal fog, low stratus, and high pollutant concentrations over the Swiss Plateau. The simulations involved different configurations of the operational turbulence scheme (ICON-TKE) and the newly developed two-energies turbulence scheme (ICON-2TE). The performance of these model configurations was assessed using ABL profiler and surface observations from the Payerne weather station in Switzerland. The results indicate that ICON-2TE, with its refined turbulence representation, allows fog to persist longer and aligns more closely with observations compared to ICON-TKE. This improvement is attributed to a more sophisticated treatment of stability dependence and turbulence length scale in the ICON-2TE scheme. Notably, an increase in vertical resolution improves fog representation in the ICON-2TE scheme, while it shows almost no effect in the ICON-TKE scheme. The lack of improvement in ICON-TKE is likely due to an overestimation of turbulence mixing, which overrides the effect of changes in vertical resolution.