The National Meteorological Center (NMC) and the Techniques Development Laboratory (TDL) of the NWS produce a number of QPFs. Generally, these forecasts cover the conterminous United States and provide lead times from 3 to 60 hours. The exact definitions of predictands, forecast issuance times, lead times, and formats vary among the products. Because the predictands and the forecasts do not match input requirements of hydrologic models, these products cannot be used directly in flood forecasting. However, they provide indispensable guidance to making QPFs for hydrologic purposes, where the predictand of interest is the basin average precipitation amount.
Krzysztofowicz and Drake [1993] surveyed nine forecasters from the NWS Pittsburgh office who had been making QPFs for river basins and, therefore, were in a position to evaluate guidance QPFs. Among 14 products, the highest rating was received by the 24-hour QPF prepared subjectively by meteorologists from the NMC Forecast Branch. Funk [1991] has described the empirical rules and techniques utilized by the NMC meteorologists. Houghton and Rubin [1990], Junker and Hoke [1990], and Junker et al. [1992] have examined the performance characteristics of the Nested Grid Model (NGM)---a numerical weather prediction model---as they affect the quality of QPFs. Finally, Junker [1993] has explained how the ``experience, knowledge of the operational models and of conceptual models for heavy rain, and various empirical rules and techniques'' enable the NMC meteorologists to improve upon forecasts coming out of NGM, despite the technological advances. Threat scores of subjective QPFs have been consistently higher than scores of model QPFs, and have shown steady, albeit slow, improvement since the early 1960s.
The TDL is in the process of developing a spectrum of new QPF guidance products. Their characteristics, input data, and forecasting techniques have been described by Glahn et al. [1992]. In general, statistical interpretive techniques will be applied to outputs from advective, mesoscale dynamic, and global dynamic models, to produce three sets of QPFs for lead times of 0--3 hours, 1--20 hours, and 6 hours to 10 days. Forecasts will be made for about 600 locations in the continental United States and will specify probabilities of exceedance of fixed precipitation amounts, such as 0.01, 0.25, 0.50, 1.00, 2.00, and 3.00 inches; there will also be an estimate of the expected amount. The primary advantages of these QPFs are (1) the probabilistic format which apprises the local forecaster of the uncertainty associated with the guidance, (2) the spatial resolution which will vary from 2--20 km for flash flood warnings, to 40--80 km for main-stem river forecasts, and (3) the frequency of updating, from 10 minutes for flash flood warnings to 2--4 per day for main-stem river forecasts.