The Meteorological Development Laboratory plans to implement a new guidance product consisting of 2-h thunderstorm probabilities for 20-km gridboxes over the continental United States (CONUS). In addition to the probabilities, the forecasts will also be produced in categorical form. The thunderstorm event is defined as one or more cloud-to-ground lightning strikes in a gridbox during the valid period. The forecasts are for 0-2, 2-4, …, 22-24 h projections, and when the program is fully implemented they will be issued at hourly intervals. The 20-km grid on which the forecasts reside is compatible with the grid used for the National Digital Forecast Database (NDFD) of the National Weather Service (NWS).

The thunderstorm probabilities are based on the Localized Aviation MOS Program (LAMP) concept, whereby model output statistics (MOS) thunderstorm probabilities issued four times daily are updated hourly. The MOS probabilities are based on the NWS National Centers for Environmental Prediction Global Forecast System, and the predictand is identical to that for LAMP except the valid period is 3 hours instead of 2 hours. The LAMP updates are produced by regression equations, in which the predictor input includes gridded observed and advected fields from lightning strike and quality controlled radar reflectivity measurements, observations of surface weather variables, topography, and lightning climatology, in addition to the MOS forecasts. In experimental regression equations, the lightning and radar predictors dominate in the first two forecast projections, whereas MOS probability and lightning climatology predictors dominate thereafter.

Verification of experimental LAMP thunderstorm probabilities for one, hourly cycle and one season shows a strong skill improvement on MOS (as calibrated to a 2-h valid period) in the 0-2 h period. Thereafter, the LAMP improvements fall off rapidly with time such that they are small after about 8 hours. The substantial improvements in the first few periods result mainly from lightning and radar predictors.

Real-time test production of the LAMP thunderstorm forecasts for two, hourly cycles is planned for the 2004-05 cool season. Both the probability and categorical forecasts will be produced for the CONUS area. The forecasts cannot be produced for NWS locations outside the CONUS because a sufficient lightning dataset does not presently exist there.

The conference presentation will include expanded information concerning the LAMP predictors and their rankings relative to MOS as a function of forecast projection, and additional verification results.