Conclusions
Both models (ECMWF and AROME) performed adequately well in forecasting the fog, though both were far from perfect and had their own unique problems. ECMWF did quite well on the location and extend of the fog cover but didn't model the advection of the fog inland until late evening. AROME on the other hand modelled the advection well but had more problems with the extend of the fog area, especially on the South-West coast of Finland. The major problem with ECMWF is the too wide grid spacing and with AROME the tuning of the cloud microphysics to describe convection better. Although no generalization can be made about the performance of these two used numerical models with the results from this single case study, the problem of fog forecasting with NWP models is known. Bearing in mind that neither of the used models is specially designed or tuned for fog forecasting, the results aren't that bad. Considering that advection fog should be the easiest fog type for numerical models to predict, the forecasters' experience in predicting fog situations more precisely is extremely important.