Experimental LAMP 2-h convection guidance on a 20-km grid

In late 2008, the Meteorological Development Laboratory (MDL) completed implementation of Localized Aviation MOS Program (LAMP) 2-h thunderstorm probability and categorical guidance forecasts (issued hourly) to 24 hours for 20-km grid boxes over the contiguous United States. Since then we have received feedback on the utility of the guidance for aviation operations. One criticism is that the thunderstorm probabilities tend to be low in areas of convection when convection is not accompanied with cloud-to-ground (CTG) lightning. Another criticism is the probability patterns lack desired spatial detail at the longer forecast projections (~ 6 hours and beyond).

MDL has begun development efforts to address these concerns. An aspect of LAMP that results in low probabilities is that the definition of a thunderstorm event is more restrictive than "convection" that adversely impacts aviation operations. Specifically, a LAMP thunderstorm is defined as the occurrence of one or more CTG lightning strikes [as reported by the National Lightning Detection Network (NLDN)] in a 20-km grid box for a 2-h period, whereas aviation-sensitive convection is usually defined as the occurrence of a radar reflectivity echo of ≥ 40 dbZ (VIP 3). Thus, we have defined a new convection predictand for LAMP (and the MOS input to LAMP) as the occurrence/non-occurrence of either CTG lightning or ≥ 40 dbZ radar reflectivity (or both). This predictand results in about a 60 % increase in the number of event occurrences over the CONUS, which alone increases peak forecast probabilities.

Another factor that contributes to low probabilities in LAMP thunderstorm forecasts is limited predictor input from numerical weather prediction (NWP) models; the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) is the only NWP model predictor input into LAMP. For the new convection predictand we plan to incorporate GFS-based MOS 2-h convection probabilities (to replace the current MOS probability input based purely on CTG lightning) as well as predictors from additional NCEP models such as the North American Mesoscale (NAM) and possibly others.

The LAMP convection model is being developed in two phases. In phase 1, the predictor input will be identical to that in the current thunderstorm model, but the MOS probability input uses the new convection predictand. This will allow for rapid development and testing during the 2011 warm season. In phase 2 the predictor base will be expanded to include multiple NWP models. Initial findings with the phase 1 model show a notable increase in forecast skill and peak forecast probabilities. These findings will be discussed in the conference presentation.