Research Studies

Louisiana and Southeast Texas
Fog Research and Modeling
(continued)

Conclusion

A fog model was developed to help forecast nights of fog initiation and redundancy. During a 3 month test period over the winter months, the model recorded a 96% success rate missing only two events...both were post-frontal. This is much better than the two numerical model mos products; the 12Z NGM MOS was 75% and the 00Z was 72%. The 12Z AVN MOS was 81% and the 00Z was 78%. Testing for accuracy, viability and usefulness will continue for several months. During this time, tweaking will continue.

When testing, the model is run two days prior to fog nights and the day of a fog night. The model is run again the morning after a fog night with the exact conditions input. This is used to verify the model. It was unfortunate that the MOS products could not be run as well during this 'next morning' verification phase. One big problem existed, the model uses input from questions posed to the forecaster. If the forecaster puts in erroneous data or makes a bad forecast, not only was his/her forecast off, but the model would give back the results of the bad data. The next morning when the model is run, the forecaster is shown which data input was erroneous and did not match the actual data that occurred that night. It goes without saying, the better the forecast the better the model will perform.

I have thought about using data from each numeric model to input into the fog model but this presents its own problems. The models do poorly in fog prediction because not enough information is ingested from the local environment to get an accurate reading from the algorithms used in each model. When ingesting this limited data, the fog model will only give back a prediction success ratio shown in the aforementioned numeric model percentages.

The model can only be used along the northern gulf coast. This is because of the localized input such as wind direction and speed, temperature and moisture profiles. These data for forming fog may be slightly to drastically different in other locations of the conus. Elevated terrain along the gulf coast may also play a role which has not been integrated into the model.

If an office would like to use or test the model, feel free to do so. I would love to get some responses back. The .exe file of the model can be sent via e-mail. If any office is serious about using the fog model and would like it re-sourced for a particular area, please give me a mailing at Tim.Erickson@noaa.gov and I will work with you as much as possible and make the new source code (the source code is done in C++ around 4000 lines). This would be a lengthy process as it has taken all of 2 1/2 years to conduct and bring about a working model at the LCH WFO.

What could the model be used for? The FAA could use this to find where fog is more likely to form and therefore conduct air traffic accordingly. The marine industry could benefit from this since many hours and a lot of money is wasted when loaded ships can not be piloted through channels to port during fog events. The public in general can benefit when knowing that the forecast is calling for fog that night instead of being asleep when a forecast is updated around midnight to show fog. This would allow the public to allot more time to leave home knowing that driving time will be extended the next morning which may result in less accidents due to fog.

Post face The author would like to thank David Leigh, former Student Intern at WFO Lake Charles, for his tireless assistance with data collection for this project. His help is greatly appreciated.