This paper describes the development and testing of enhanced local guidance for Probability of Precipitation (PoP) forecasting. A statistical technique known as logistical regression has been applied to forecast output from NCEP's operational Eta model to produce Pop forecast equations for Philadelphia, PA. Separate equations were developed for each of three twelve-hour forecast periods, beginning at 12 hours after model initialization time. Development of the technique began in the warm season (April through September) of 1998, and has continued to the present. A separate set of equations was also developed for the cool season (October through March).

The predictors screened for the regression include Eta-forecast total precipitation, low to mid-level vertical velocities, low to mid-level relative humidity, surface pressure change, wind direction, lifted index, and low-level moisture convergence. The predictor values were extracted manually from graphical displays of Eta forecast grids on the National Weather Services's Advanced Weather Information Processing System. When sufficient data were collected, the regression analysis was performed using a standard statistical software package.

Verification of the forecasts against the operational NGM-based MOS equations began with the warm season of 1999 and has continued to the present for both cool and warm seasons. The verification has generally shown a modest improvement over the operational MOS guidance, in spite of the frequent changes to the Eta grid resolution and model physics. This result may be due to better forecasts of the predictor variables by the Eta, and it also suggests that logistical regression may be a better method for developing PoP forecast equations.