Precipitation forecasting using a neural network

A neural network, using input from the Era Model and upper air soundings, h as been developed for the probability of precipitation (PoP) and quantitati ve precipitation forecast (QPF) for the Dallas-Fort Worth, Texas, area. For ecasts from two years were verified against a network of 36 rain gauges. Th e resulting forecasts were remarkably sharp, with over 70% of the PoP forec asts being less than 5% or greater than 95%. Of the 436 days with forecasts of less than 5% PoP, no rain occurred on 435 days. On the 111 days with fo recasts of greater than 95% PoP, rain always occurred. The linear correlati on between the forecast and observed precipitation amount was 0.95. Equitab le threat scores for threshold precipitation amounts from 0.05 in. (similar to 1 mm) to 1 in. (similar to 25 mm) are 0.63 or higher, with maximum valu es over 0.86. Combining the PoP and QPF products indicates that for very hi gh PoPs, the correlation between the QPF and observations is higher than fo r lower pops. In addition, 61 of the 70 observed rains of at least 0.5 in. (12.7 mm) are associated with PoPs greater than 85%. As a result, the syste m indicates a potential for more accurate precipitation forecasting.