Abstract In the eastern Alpine region, subinversion cloudiness associated with elevated temperature inversions is a frequent phenomenon in autumn and winter, which often persists for several days. Although the prediction of fog and low stratus by numerical weather prediction (NWP) models has improved in recent years, these models still show deficiencies in the spatial and temporal evolution of such wintertime weather phenomena. In spite of sophisticated current assimilation schemes or simply due to unknown conditions, even the analysis shows large discrepancies compared to the true atmospheric state. Inversions are often “smeared out” and the moist layer below the inversion is too far from saturation. Model integration from such an initial state leads to strong biases in the total cloudiness and, due to erroneous radiative response, in 2-m temperature forecasts. In the present paper, an empirical enhancement scheme for subinversion cloudiness is introduced within the framework of Aire Limitée Adaptation Dynamique Développement International (ALADIN), the operational limited area model (LAM) at the Austrian Central Institute for Meteorology and Geodynamics (ZAMG). The scheme attempts to compensate for model deficiencies in the vertical temperature and humidity profiles in order to enhance or keep preexisting signals of inversions and associated low cloudiness. Thus, a positive feedback due to radiative reaction is activated, which finally leads to more realistic vertical profiles, low (and total) cloudiness, and improved 2-m temperature predictions. Case studies demonstrate the impacts of the scheme on predictions of the spatial distribution of low cloudiness and on the vertical profiles of temperature and humidity. Verification over stratus episodes within a 2-month period comparing a reference model run without the scheme with a modified model version with the subinversion cloudiness scheme confirms the ability of the scheme to improve stratus-related wintertime weather prediction.