Toward better use of lightning data in operational forecasting

Lightning is a well-known weather hazard; it kills roughly 50 to 100 people in the United States each year, putting it on a par with tornadoes and flash floods as a cause of weather-related fatalities. Many more people survive lightning strikes, but suffer long-term health problems afterwards. Lightning also starts many house fires and wild fires, and often causes local to widespread power outages affecting thousands of electric utility customers.

All thunderstorms by definition produce lightning; however some thunderstorms produce far more cloud-to-ground (CG) lightning than others. Of course, it takes only one CG lightning flash from one isolated thunderstorm to cause a fatality. Even so, the risk of being struck by lightning is certainly much greater in a high-CG flash rate storm, as is the likelihood of fire damage and especially the effect on electrical utilities. It would be advantageous if a probability of high-CG storms could be determined a day or even an hour in advance.

CG lightning data have been available to National Weather Service (NWS) field forecasters since the late 1990's through the Advanced Weather Information Processing System (AWIPS). High-CG storms are easily identified in real time by displaying 1-minute, 5-minute, 15-minute, etc., flash counts in combination with radar and/or satellite imagery. While references to “intense” or “frequent” lightning may be included as part of a severe thunderstorm warning or “special weather statement”, there is no NWS product that deals specifically with lightning. Also, there seems to be no general sense among forecasters as to what sort of flash rate or flash density represents a “high-CG storm” versus an “ordinary” thunderstorm. What is needed, among other things, is a lightning climatology that includes flash-rate frequency distribution.

As one attempt to address these issues, national lightning data for the summer of 2004 were archived from the AWIPS at the NWS Forecast Office in Mount Holly, New Jersey (PHI). Hourly lightning counts were made for various spatial scales including the PHI county warning area, six eastern metropolitan areas, and a multi-state area over the mid-Atlantic region. The results give a sense of the diurnal distribution of hourly lightning, and they also show that the majority of seasonal lightning tends to occur on a few “big” days.

In theory, strong instability should lead to thunderstorms with strong updrafts, leading to rapid electric charge separation, leading to high lightning rates. However, attempts to correlate CG flash counts over a fixed domain with meteorological variables such as stability parameters were not very successful, being complicated by the fact that thunderstorms may not form at all in a given area, even under very favorable conditions. In an attempt to find a conditional lightning predictability (given thunderstorms are present), about 150 hours were identified that had a large number of flashes (> 2500) over a multi-state area. Then the largest lightning cluster for each of those hours was located, and the hourly flash density was compared to the maximum analyzed CAPE in proximity to the cluster. Results showed an improved but still rather modest explanation of variance (r2 = .135). The data do suggest an upper threshold for flash density at any given hour which may be related to CAPE or some other parameter. Work is continuing to include lightning data from summer 2005, and further results will be presented in the extended abstract and at the conference.