Close Window

Texas A&M University: "Fog forecasting using a one-dimensional model"

Final Report


The objective of this project was to develop and implement a radiation fog forecasting system for the National Weather Service Forecast Office in Wilmington, Ohio. The forecasting system was to be based on a state-of-the-art one-dimensional numerical model, known as COBEL, tailored to the forecasting of radiation fog and low stratus. The idea behind using a one-dimensional model was to implement an ensemble forecasting system spanning the range of variables likely to affect fog formation on a given day and thereby span the range of likely outcomes.

The specific tasks were as follows: (1) obtain data for driving the COBEL model on past cases; (2) become familiar with the COBEL model; (3) compile and run the COBEL model; (4) convert archival rawinsonde data to COBEL input format; (5) generate output display interface; (6) run COBEL on archived cases; (7) establish settings and ensemble members for the COBEL model; (8) couple the COBEL model to the Eta Model; (9) run COBEL as a real-time forecasting tool; and (10) evaluate the results of the real-time forecasting test. Tasks 1 through 3 were completed prior to the beginning of funding.

The primary responsibility for WFO Wilmington (ILN) during the first part of the project was to provide Texas A&M (TAM) with archived sounding data for input into the COBEL model. These data were to be coincident with cases that involved both the occurrence and non-occurrence of radiation fog at Lunken Municipal Airport (LUK), whose location was in close proximity to the Ohio River in southwest Ohio. To select these cases, surface data from LUK and sounding data from Dayton and ILN for the period June 1995 through June 1997 were acquired from the National Climatic Data Center. The data was screened using Excel to identify days which satisfied the following criteria: visibility below 1 mile for at least one hour; nights with light winds and clear skies; and dates later than Sept. 25, 1995 (when the Dayton sounding site was moved to ILN). Of the fifteen days that passed the screening criteria, four were selected with varying fog intensities and durations. Two additional days were selected with light winds and clear skies but no occurrence of fog.

In order to convert the selected soundings into the specified input file format needed for use in the COBEL model, ILN's versatile upper air radiosonde/forecast sounding analyzer program required augmentation. Specifically, an algorithm was written which enabled the program to accurately compute numerous meteorological parameters (temperature, dew point temperature, absolute humidity, u and v wind components, height, and more) at a resolution of one tenth of a millibar. Once the specified input data files for each case were created, a further step to massage the data was implemented in order for the COBEL model to run successfully. This step involved the manual removal of any extraneous characters such as carriage returns.

At TAM, the COBEL model was run on four cases, three of which had fog. The model output was converted to Excel format for display and analysis; consequently no additional user interface was required at this point. A series of sensitivity tests were performed to understand the effect of winds, temperatures, moisture, soil temperatures, soil moisture, radiation, pressure, and other quantities on fog formation. In the course of conducting these sensitivity tests, problems with certain aspects of the model were identified and fixed so that the model would function properly for an environment such as LUK.

The sensitivity tests yielded the following results. First, it was verified that the COBEL model was capable of accurately simulating radiation fog at LUK, with the proper initial conditions and settings. Second, the factors that caused significant changes in the fog forecasts were identified. In some cases, this required the gathering of additional data, such as soil moisture and soil type, in order to determine the likely values of sensitive parameters. Third, we developed techniques for adjusting the initial and nighttime large-scale conditions to apply to the specific river-bottom environment found around LUK. Finally, a prototype ensemble forecasting system was developed that included a set of sensitive values of various parameters and the likely ranges of these parameters.

Of these accomplishments, the second is most significant. In addition to providing important information for ensemble design, the sensitivity tests gave useful insight into the processes of fog formation. Even without running the model, forecasters now know that under synoptically favorable conditions, soil moisture is the single most important environmental variable controlling fog formation. Other parameters of importance include wind speed, low-level moisture, soil moisture, and cloud cover. These insights should provide beneficial improvement to fog forecasting at ILN.


Throughout the project, frequent communication was maintained between TAM researchers and ILN partners. This communication included exchange of input data files and discussion of formatting issues. ILN provided the raw input data files to TAM and performed comparison model runs to check the specific implementations of the COBEL model at the two locations.

Near the end of the project, the TAM researchers (Lionel Peyraud and John Nielsen-Gammon) traveled to ILM, where they presented a two-hour seminar/workshop on the results of the COBEL sensitivity tests and the insights they provided regarding fog formation. Following the seminar, discussion focused on the next steps for the research project and possible strategies for implementing the COBEL model in an operational setting, in a manner that would allow for objective measurement of the impact of COBEL in fog forecasting.


Peyraud, L., 2001: Radiation Fog Forecasting Using a 1-Dimensional Model. M.S. Thesis, Texas A&M University, 82 pp.


4.1) University partner

This particular forecast project was initiated by a contact from ILN. Benefits to TAM included exposure to professor and graduate student to a particular operational forecasting problem, experience in running boundary layer model, and insight into the dynamics of fog formation. It is unlikely that TAM would have been involved in this kind of research without ILN's participation, and the research has served to broaden the expertise at TAM and will impact the education of undergraduates and graduate students as a result. The specific benefit to the graduate student, Lionel Peyraud, was that it helped him obtain a job in operational forecasting, and it should serve to improve his forecasting on the job as well.

The primary difficulty encountered by TAM was that the COBEL model was not as straightforward and transparent as we would have liked. Much more time than expected was spent debugging the model and developing flowcharts to understand its algorithms. We are still not satisfied that the land surface model and atmospheric model are coupled in such a way as to conserve water, so the model is not quite ready to be deployed in an operational setting. Another potential difficulty might have been the fact that the COBEL model originated in France and has been further developed in Quebec, so that most of the model variable names and comments are in French. Fortunately, Lionel Peyraud was fluent in French, and it is doubtful that this project would have been successful without the participation of this individual. Mr. Peyraud has agreed to remain in contact to help with the project in the future, and the primary COBEL developer has moved to NCAR where we can also remain in contact.

4.2) Forecaster partner

The initial benefit for ILN has been an improved understanding of the processes associated with fog forecasting. This was brought about specifically through utilization of the COBEL model and in conjunction with TAM's presentation on their research findings. Long term benefits will hopefully be realized in improved aviation and public forecasts. This, however, will be dependent on the ability to transform not only the current COBEL model itself, but also the processes involved in running this model, into a much more user friendly operational forecasting tool. This will require improved streamlining of the current procedures. Yet prior to this, ILN will need to run the COBEL model in non-real time on a number of cases to determine ways to beneficially adjust the model parameters in order to ultimately enhance its operational utility. Incorporation of supplemental data (such as soil moisture) from available BUFR files should also help in achieving better results. The processes involving advection/fluxes of such parameters as moisture will also need to be investigated.

The main problem ILN ran into with the COBEL model was the fact that it runs on a unix platform, and was not very user friendly. Not only was it a non-trivial task to initially compile COBEL on ILN's Science and Application Computer (SAC), but there were too many steps involved in the processes of converting the data into the proper input format and transferring the data to the SAC. In addition, difficulty was also noted in ILN's ability to view COBEL model output. These will all be major obstacles in ILN's quest to incorporate the use of this model into its operational forecast process. The ultimate desire would be to have the ability to run the COBEL model on a PC. Note that a windows-compiled COBEL executable could be integrated seamlessly into ILN's upper air radiosonde/forecast sounding analyzer program, and could then be used NWS-wide.