The main technique that is currently used in fog forecasting for Melbourne Airport is one that predicts the onset time of fog at Melbourne Airport. The technique is graphical and, in application, the previous afternoon’s wet bulb depression and low level wind at a nearby station are combined with indicators of the synoptic-scale circulation to provide an estimate of whether or not fog is likely to form and, if so, at what time.
The purpose of the present paper is two-fold.
Firstly, a synoptic classification of fog situations is presented. It is shown that fog occurs quite frequently at Melbourne Airport with some synoptic classes. By contrast, there has never been a fog in association with some other synoptic classes.
Secondly, the paper describes a new technique to predict the probability of occurrence of fog that is based on aspects of the onset time technique. The new technique uses logistic regression and was developed using 22 years of data between 1971 and 1993. The skill displayed by the logistic regression technique is verified using the Brier Skill Score on over 4 years of independent data between 1994 and 1998.
In a logit regression model the predicted values for the dependent variable will never be less than or equal to 0, nor greater than or equal to 1, regardless of the values of the independent variables. This is accomplished by applying the following regression equation
y=(exp(a+S bixi))/(1+(exp(a+S bixi)))
where ‘y’ is the dependent variable, the xi are the independent variables, and a and the bi are constants. In operation, where ‘y’ is a yes/no variable, the equation yields the probability of occurrence of a particular phenomenon.
In developing the new technique, the previous afternoon’s dewpoint (x1) and temperature (x2) are combined with a "length of night" parameter, (x3), defined as |month – 6|, in a set of regression equations.
The equations are developed on data associated with synoptic classes for which there are sufficient data and sufficient "fog" cases, and for which the derived statistical relationship is significant. For these synoptic classes, which encompass 53% of all cases and 62% of "fog" cases, the relevant equation is then solved to provide a forecast of fog probability. For the other synoptic classes, the fog probability is set at the climatological normal for that class.
Several illustrative examples of applying the technique are given
The 8th Conference on Aviation, Range, and Aerospace Meteorology