The Meteorological Development Laboratory (MDL) is responsible for producing the Model Output Statistics (MOS) guidance. As part of this guidance, equations to predict the probability of thunderstorms were derived for the contiguous U.S. (CONUS) by using cloud-to-ground lightning data and output from the numerical weather prediction models. Output from both the Global Forecast System (GFS) and the Eta model was used to generate forecast equations. We currently produce thunderstorm probability forecast guidance on a 40-km grid for 6-, 12-, and 24-h periods. As the horizontal resolution of the numerical models is increasing, we are updating our thunderstorm equations to reflect the increase in resolution in both time and space. Our latest research concentrates on issues related to shorter temporal and smaller spatial resolutions. We are developing a set of Eta-based probability forecast equations for both a 40-km grid spacing and a 20-km grid spacing over the CONUS. In addition, we are increasing the temporal resolution of the forecasts by generating forecast equations for 3-h forecast periods, beginning with the 3-6 h forecast projection.

Many interesting issues have arisen from the increase in the temporal and spatial resolution. Different user communities have different guidance needs. Guidance developed to support the Storm Prediction Center’s convective outlook product may not need the same resolution as guidance intended for the aviation community. Should we try to meet the needs of all users with one complete system, or is it more practical to develop packages with different resolutions customized for different user communities? In previous work, we have seen that as the resolution of the guidance increases in time or space, the range of the probability values becomes smaller. In general, as the grid box and time period become smaller, the probability of the event occurring in the grid box during the time period is also smaller. The question becomes then, at what point is the probability value too small to be useful. Another issue is the areal coverage of the forecasts. Will the forecast patterns between a 40-km and a 20-km product look the same, even if the actual value of the probabilities are different? Are seasonal and diurnal variations in lightning climatology important in determining how far we can increase both the temporal and horizontal resolution? Finally, comparing the skill of the guidance at different resolutions is difficult. Can a meaningful objective measure be employed when the guidance is on different grids? As we increase the resolution of the guidance, we may produce more detail in the forecasts, but the verification scores may not reflect an increase in skill if the higher resolution forecasts have position or timing errors.

These questions will be examined in this paper. We will present some objective verification scores, and examples of the guidance at different resolutions valid during a thunderstorm outbreak will be shown.