Development, operational use, and evaluation of the perfect prog national lightning prediction system at the Storm Prediction Center
Phillip D. Bothwell, NOAA/NCEP/SPC, Norman, OK
Since the first experimental Perfect Prog (PP) cloud-to-ground lightning predictions started at the SPC in 2003, the program has grown to play a larger and larger role as a decision support aid for those with thunderstorm and fire weather forecast responsibilities. The basics of the forecasts remain the same; development of lightning prediction equations for 3 hour periods on a 40 x 40 km grid using input data from operational NWP models. In the PP approach, several hundred candidate predictors are reduced to about a dozen primary factors using principal component analysis. Logistic regression is then used to produce the forecasts for each grid point. While originally designed to aid in guidance for predicting dry thunderstorms in the western U.S. as part of the SPC Fire Weather Outlooks, these forecasts cover all of the lower 48 states and adjacent coastal waters. Since 2006, the forecasts for the western U.S. have been available through a web page to NWS WFOs and fire fighting agencies each summer. The lightning forecasts were expanded on an experimental basis in 2008 to cover Alaska.
The perfect prog system has evolved to use four different and unique input gridded data sets depending on the length of forecast needed. A zero to three hour forecast is available about 7 minutes past the top of each hour at the SPC, using as input the hourly SPC 3-D Mesoscale Analysis data set. Also, RUC-based PP forecasts are available each hour out to 12 hours. The NAM forecasts initially produced out to 48 hours (originally the Eta) have expanded to use the full 84 hour NAM data as input to the PP scheme along with a Dry Thunderstorm Potential index (DTPI). In 2008, a new experimental 8 year climatology and development data set was used to produce guidance forecasts for thunderstorms in Alaska to support their fire weather forecast program. For Alaska, GFS input data is interpolated to a 45 km grid and experimental forecasts are available out to seven and a half days.
Forecasts for two different significant lightning categories (defined as 10 or more flashes and 100 or more flashes per 3 hours in a 40 km grid box) are also produced and verified for the lower 48 states. Forecasts for 10 or more flashes are being tested at the Salt Lake City, Utah, Weather Forecast Office (WFO) as a means of identifying areas of more significant dry thunderstorm potential during the summer. The Raleigh, North Carolina WFO continues for a second year to use the forecasts of 100 or more flashes as guidance for highlighting significant lightning events in their early morning hazardous weather outlook.
The verification of these experimental forecasts along with examples using the various model inputs for different time periods, geographical areas, and flash rates (1, 10 and 100)per three hours will be presented at the conference. Implications for what these forecasts offer now and in the future as improvements are made will be discussed.
Extended Abstract (736K)
Session 6, Lightning Warning and Prediction Using Observations and Models
Tuesday, 13 January 2009, 3:30 PM-5:30 PM, Room 131A
Previous paper Next paper
Browse or search entire meeting
AMS Home Page