Fuelberg Lab

Go Back a Page
Current Research :: Past Research :: Dr. Henry E. Fuelberg :: Fuelberg Lab Home

Preparing a Statistical Tool for Forecasting Warm Season Afternoon and Early Evening Lightning in Florida

A Research Project for Florida Power and Light Company

Cloud-to-ground (CG) lightning is a major cause of power disruptions in areas served by Florida Power & Light Company (FP&L). These disruptions are especially problematic when they occur after normal working hours. Each day around 1:30 PM FP&L personnel decide whether lightning is likely during the later afternoon and early evening hours. If a lightning threat is perceived, then line crews are retained after the end of their normal shifts. If the threat is mis-judged, the Company either will not be able to respond to un-forecast power outages in a timely manner, or, conversely, resources will be wasted for a threat that does not occur. Clearly, accurate lightning forecasts in the 3-12 hour time range have great economic benefit to the Company. A program of research at Florida State University (FSU) seeks to provide more accurate lightning forecasts for this two county area served by FP&L. Phase I of the project began in August 2002; Phase II began in June 2003; Phase III began in June 2004; Phase IV began in June 2005; and the positive lightning study began in June 2006.

Phase I

Phase I of the FSU research focused on describing the spatial and temporal distribution of lightning in Miami-Dade and Broward Counties and its relation to atmospheric conditions. CG lightning data and 1200 UTC radiosonde soundings for Miami and West Palm Beach were acquired for the 14 year period between 1989-2002. The patterns and the timing of lightning were examined and related to radiosonde-derived values of atmospheric winds, humidity, and stability. A composite view of warm season lightning over the entire Florida peninsula is shown in Fig. 1.

fig1; lightning climo
Fig. 1. Composite view of warm season lightning over the entire Florida peninsula.

Warm season lighting over the Florida Peninsula mostly is due to the inland penetrating sea breeze front. Sea breezes are an almost daily occurrence during the summer. However, their strength and degree of inland penetration vary daily depending on the direction and speed of the prevailing wind (as determined from radiosondes). The location and amount of lightning associated with the sea breeze also varies due to these atmospheric conditions.

Phase II

Based on our preliminary findings, we developed statistical guidance (equations) that enable FP&L officials to estimate at their 1:30 PM meeting the lightning threat that will occur later in the day over eastern Miami-Dade and Broward Counties during the warm season. These forecast equations utilize input from the unaltered 1200 UTC Miami radiosonde sounding that is available to FP&L personnel.

We developed and evaluated two types of equations for the eastern portion of each county:

We used logistic regression techniques to derive both types of forecast guidance equations. Specifically, a spread sheet was populated with the amount of lightning that occurs each day over each area and period of interest, along with a host of pertinent radiosonde-derived parameters describing atmospheric winds, humidity, stability, etc. The regression procedure was then applied to the data set, selecting the several parameters (Parms) that explain the most variance, and assigning an appropriate coefficient (C) to each parameter.

It is important to evaluate the utility of the guidance equations that are developed. Therefore, we calculated various statistics for this purpose, including the probability of detection, probability of false detection (false alarms), true skill statistics, rms errors of forecast vs. observed lightning, etc. Thus, the success of the equations was quantified. We quantified the degree to which results from the forecast equations exceed the results expected from climatology and persistence.

This research is being collaborated with the National Weather Service Offices in Miami and Tallahassee. The NWS brings tremendous operational expertise to the research effort. The Miami office prepares NWS forecasts for part of the FP&L service area, including Miami-Dade and Broward Counties. Pablo Santos, the Science and Operations Officer at NWS-MIA, is the focal point at that office. Andrew I. Watson, the Science and Operations Officer at NWS-TLH, performed considerable research on the sea breeze and associated weather in South Florida during his previous employment at the National Severe Storms Laboratory. Since NWS-TLH is located in the same building as the Florida State University Department of Meteorology, the collaborations are easy to facilitate.

Phase III

We developed guidance equations for eight additional service areas that were specified by FP&L. Some of these areas are individual counties, while others are combinations of counties. The areas are:

Our goal for these new areas was similar to that for the two Phase II regions. That is, we first performed sufficient climatological studies to understand the factors producing lightning in each area. Then, we developed an equation for each area that predicts whether one or more lightning flashes will occur during the afternoon and evening hours. Morning radiosonde data are used to develop these equations. We also developed for each area a guidance equation that predicts the number of flashes that is expected.

Input for the guidance equations is the morning radiosonde releases from either Miami, Tampa/Ruskin, Cape Canaveral, or Jacksonville. Although only the Miami sounding was needed in the Phase II counties, some of the new Phase III areas are closer to Jacksonville, Tampa, or Cape Canaveral.

Phase IV (research in progress)

The lightning guidance that we developed during Phases II and III utilized 1200 UTC radiosonde soundings as input. This was based on the assumption that atmospheric conditions do not vary significantly from the sounding time through the end of the forecast period. In addition, we assumed that atmospheric conditions in the forecast area are similar to those at the radiosonde site which may be hundreds of miles away. Ease of operation also is a major factor in the use of radiosonde data. However, whenever these assumptions are violated, which may happen frequently, there will be errors in the lightning forecast. Therefore, afternoon input data, local to the area of interest, should yield improved lightning forecasts. Those improved forecasts are the goal of Phase IV.

The main objective of the Phase IV research is to use NWS numerical guidance to create statistically-derived forecast spatial fields of:

This research constitutes the Ph.D. dissertation topic of graduate student Phil Shafer.

The prediction maps will be generated for the entire FP&L service territory, i.e., all of the peninsula plus parts of North Florida. We will develop forecast maps for hourly periods. FSU and FP&L personnel agree that spatial fields of lightning forecasts will be more useful than those for specifically defined regions as proposed during Phase III. The lightning guidance products will be valid for the warm season months when the sea breeze is the dominant forcing mechanism for thunderstorms. During Phase III, FP&L defined the warm season to be June, July, and August; however, May and September can be added by mutual consent of both parties.

Several caveats about the proposed research are mentioned below:

Positive Lightning Research

Phases I - IV of the research have considered all cloud to ground flashes, i.e., the combination of positive and negative discharges. However, FP&L employees have observed that positive flashes are the most damaging. This research focuses on the positive flashes alone. We will develop a climatology of positive flashes, and then seek methods for forecasting the amount of positive flashes that will occur. The topic constitutes the M.S. thesis research for graduate student Scott Rudlosky.

Graduate Students on this Project

Phil Shafer
Scott Rudlosky

Click here to view Phil Shafer's abstract & article that was published in the October 2006 issue of Weather and Forecasting entitled A Statistical Procedure to Forecast Warm Season Lightning over Portions of the Florida Peninsula.

Click here to view Phil Shafer's Master's thesis entitled Developing Statistical Guidance for Forecasting the Amount of Warm Season Afternoon and Evening Lightning in South Florida.

Click here to view Justin Winarchick's Master's thesis entitled Developing Statistical Guidance for Afternoon Lightning Activity in Portions of Two South Florida Counties.

Click here to reach the web page of the Florida Power & Light Corporation.

last updated October 31, 2006