Predicting Winter Fog over Complex Terrain Using Deep Learning

Fog forms in high-elevation complex terrain as frequently as it does over bodies of water but is less understood and harder to predict. Forecasting winter cold fog over complex terrain is particularly difficult due to the complex interactions between land, water, snow cover, and atmospheric conditions in the process of fog formation. Traditional physical and numerical models have a limited ability to represent various conditions associated with fog formation; thus, fog prediction remains a challenge in weather prediction. This study aims to evaluate the effectiveness of machine learning methods in predicting winter fog over complex terrain (e.g., Salt Lake City and Heber City in Utah). We will utilize multiple years of surface meteorological observations and available reanalysis data products. Emphasis will be placed on examining various meteorological variables for their effectiveness in machine learning to help produce meaningful forecasts. Details and results will be presented.