An experiment in probabilistic quantitative precipitation forecasting

To enhance its customer service, the National Weather Service Weather Forecast Office in Tulsa, Oklahoma (WFO TSA) routinely produces probabilistic quantitative precipitation forecasts (PQPF) on an experimental basis. These PQPFs provide the unconditional probability of exceedance (POE) for specified rainfall amounts. The forecast method utilizes the fact that the frequency distribution of rainfall amounts closely fits the special gamma distribution known as the exponential distribution, which decreases rapidly as rainfall amounts increase from near zero to larger amounts.

The POEs are produced in the Gridded Forecast Editor as part of the routine forecast process. The meteorologist first makes a forecast for the probability of precipitation (PoP) and a quantitative precipitation forecast (QPF). For the purpose of this experiment, the QPF is conditional upon the occurrence of rainfall greater than or equal to 0.01 inch (0.254mm). The QPF specifies the mean for the probability density function (PDF) of the exponential distribution, which is used to calculate the POEs at each grid point for the specified rainfall amounts. The POE is then multiplied by the PoP to arrive at the unconditional POE at any given grid point in the WFO TSA forecast area.

What may be unique in this study is the use of QPFs in the calculation of the POEs. When the QPF is used as the mean in the exponential distribution, it changes the shape of the distribution. The resulting PDF yields probabilities that are significantly different from the climatological values. Ultimately, when the meteorologist issues a QPF, he/she is selecting a different and distinct rainfall distribution, such that higher QPF values result in greater POEs for the larger rainfall amounts.