Various avenues of research are underway to understand lightning better and to make forecasts of where, when, and how much lightning will occur. The Florida State University (FSU) lightning group is approaching forecasting from a statistical perspective. Most of our efforts have focused on Florida where the sea breeze is the dominant forcing mechanism for warm season thunderstorms. This paper will describe our recent results, ongoing research, and plans for the future.

Before beginning statistical developments, a detailed climatology of lightning occurrence and amount is needed to explore the associated atmospheric and geographical conditions. Our climatologies are functions of time of day and the prevailing winds. Our initial statistical approach for Florida was to relate lightning over a region to historical 1200 UTC radiosonde-derived variables. Although that approach provided reasonable forecasts, we needed to account for temporal changes in the atmosphere as well as spatial variations between radiosonde sites. Therefore, our second approach related 3-hourly gridded lightning amounts to corresponding RUC-derived parameters. The Perfect Prognosis approach was used with logistic regression to select the optimum set of RUC-derived variables and climatological parameters. The ensuing guidance product was evaluated on independent data sets from the NCEP 13-km Rapid Update Cycle, the NCEP 12-km NAM, and locally prepared high-resolution runs of the WRF model with LAPS input. The models were initialized as early as 1200 UTC. Compared to forecasts based on climatology and persistence alone, lightning forecasts from all three mesoscale model inputs generally showed positive skill through 0000 UTC of the same day. Guidance from this statistical procedure currently is being used operationally by National Weather Service offices in Florida and by Florida Power & Light Corp. Our paper will describe these statistical procedures in more detail and provide verification statistics. We currently are re-deriving the guidance equations for the Pendleton, OR, Pueblo, CO, and Sterling, VA regions to determine whether similarly good results can be obtained. Thunderstorms in these areas are triggered by processes that are more complex than Florida's sea breeze. Our paper will describe these efforts.

The FSU group also is investigating statistical approaches to nowcast when the last flash of a thunderstorm has occurred at NASA's Kennedy Space Center (KSC). Improved cessation forecasts will lead to large cost savings and a reduction in lost man hours since outdoor activities at KSC stop when lightning is in the area. Our first efforts investigated whether there was a “marker” for lightning cessation in total lightning data from KSC's LDAR network. Several approaches were pursued, and some showed considerable promise. However, each technique generally failed to accurately nowcast situations in which the last flash occurred considerably after the next to last flash, often in the anvil region of the decaying storm. This scenario could lead to disastrous consequences. We currently are investigating the combination of real time radar and LDAR data to determine if a robust cessation signal can be detected. The WDSS-II software is being used extensively to isolate this signal. The conference paper will describe these efforts toward statistically forecasting lightning cessation.