The Advantages of Probabilistic Versus Deterministic Radiation Fog Forecasts Using the Non-Hydrostatic Eta Model: A Case Study
Widespread fog occurred under broad anticyclonic conditions over northern France in December 2016. Detailed measurements from the SIRTA field program were available for analysis. The main objective of this study was to investigate whether, and to what extent, the ensemble modeling approach can improve forecasts of onset, evolution, and dissipation of radiation fog based on statistical scores, and to evaluate predicted visibility. Ensemble simulations using the regional Eta model driven by the ECMWF 31-member ensemble were analyzed. To account for spatial uncertainty, a five-point spatial extension surrounding the station was used, providing 155 members. Thresholds for liquid water content (LWC) as a fog predictor were defined against observed visibility. Deterministic and probabilistic verification—encompassing the Brier and continuous ranked probability scores, reliability diagram, and relative operating characteristic curve—was performed. This case study demonstrated that the probabilistic framework optimizes the ensemble spread and offers added value over single deterministic forecasts by improving fog detection thresholds and providing more precise fog timing. Timing evaluations revealed a systematic median onset delay of two hours and a four-hour dissipation delay, although 50% of the members outperformed the control run. An ensemble Kalman Filter applied as a postprocessing bias correction method reduced the LWC overestimation.
Authors
- Darko R. Koracin (ORCID: https://orcid.org/0000-0002-2043-5429)
- Dragana Mujković
- Katarina Veljović Koračin
Institutions
- University of Notre Dame (US)
- University of Belgrade (RS)
- Desert Research Institute (US)
- University of Split (HR)
Publication Details
- Journal
- Atmosphere
- Published
- 2026-11-19
- DOI
- https://doi.org/10.3390/atmos17111148
- Primary Topic
- Meteorological Phenomena and Simulations
- Type
- article
- Field-Weighted Citation Impact
- 0.00