I. Probability Forecast Advantages

The primary advantage of the probability forecast is that it can be used to illustrate the likelihood of an event occuring.

Categorical, or yes-no type forecasts force the forecaster to pick one event or the other and typically convey less information to the public or end user.

The public understands probability more than forecasters think. Most individuals have developed probability thresholds concerning decisions such as whether or not to carry an umbrella, to have a picnic, or go on a hike.

II. Issues to consider:

Probability more precisely defines the forecaster certainty than words or phrases such as "chance of" or "likely."

Probability is the same as chance. The probability at a point is the same as the chance of the event at a point.

The average point probability over the forecast area is used in most zone forecasts.

Using average point probability is difficult if the probability varies drastically over the forecast area. This is particularly problematic in areas of commplex terrain. Splitting the forecast area up into parts is recommended when the point probability varies by more than 20%.

Average point probability is not the same as areal probability. Areal probability is the chance of an event occuring in an area and is generally larger than the average point probability. As the areal size increases, the areal probability also tends to increase. For example, during the summer convective season the chance of rain at SLC airport is less than the chance of rain occuring somewhere in Salt Lake County which is less than the chance of rain occuring in the state of Utah.

Probability also tends to increase with the length of the time period.

Forecast skill is limited by lead time and as the lead time increases the forecast probability must converge on climatology.

New forecasters tend to overestimate thier skill and use excessively high or low probabiliities.

Consensus forecasts are generally better than most individual forecasts over long time periods.

III. Improving or "beating" MOS

For better or worse, forecaster skill is frequently compared with that of MOS.

When Brier skill scores are used, changing MOS can be worth more at some probabilities than at others and in one direction more than the other.

For example, if a 10% MOS POP is changed to 0% you gain only .01 no precip is observed and and loose .19 if precip does occur. If you change the 10% to 20%, you loose .03 if it doesn't precip and gain .17 if it does.

Understand these aspects of playing the forecast skill game!

V. References and on-line links

Hughts, L. A., 1980: Probability forecasting. NOAA Technical Memorandum NWS FCST 24, NWS Central Region, Central Region, Kansas City, MO.

Murphy, A. H., and R. W. Katz, 1985: Probability, Statistics, and Decision Making in the Atmospheric Sciences. Westview Press, Boulder, Co. 545 pp.

Updated May 8, 1997