An Ensemble-Based Algorithm to Predict Warm Season Fog Occurrence across the Hudson Valley in East Central New York
An ensemble-based algorithm is introduced utilizing an artificial neural network database of 33 fog events that occurred during the 2012 warm season. The algorithm compares NAM, GFS, RAP, and WRF numerical model output, while recomputing the boundary layer profile from the surface up to 850 hPa at increments based on the vertical resolution of each model. Favorability of fog development is determined through a three-step process. The first step assesses the synoptic pattern by quantitatively comparing individual model output at 925, 850, 700, and 500 hPa against NCEP/NCAR reanalysis data. The second step determines the likelihood that a boundary layer is supportive for fog development by quantitatively comparing the static stability, moisture, and momentum profiles from each model against observed sounding profiles. The third step computes the time required for fog to form using an equation that relates the saturation process to the nighttime cooling rate and length of night. Final output from each model is compared to produce a probabilistic forecast of fog occurrence and onset timing.
Three cases are depicted outlining the methodology of the algorithm and its applications for aviation forecasting. Algorithm performance and limitations are also assessed. Additional fog events from the 2013 and 2014 warm seasons will likely be added to the artificial neural network database in the future along with an expansion of algorithm physics to predict fog duration and dissipation.