1. Forecasting Lightning Initiation and Cessation at Kennedy Space Center
Henry Fuelberg
Florida State University

2. U.S. Lightning Factoids
Each year lightning strikes U.S. ~ 25 million times
Plus 3 to 50 times as many cloud flashes
Positive vs. Negative Flashes
Voltage = 100 million V
Amperage = 5,000 - ~250,000 A
Temperatures reaching ~50,000oC
~ 80 people killed each year in U.S.
~ 300 injuries each year in U.S.
~26,500 house fires annually due to lightning
Total insurance claims ~ $ 1 Billion
Total costs ~ $ 5-6 Billion

3. Non-Inductive Charging Mechanism
Charging occurs when ice crystals and graupel collide in the presence of supercooled water (mixed phase region of storm).
Updrafts carry lighter particles with positive charge aloft.
Heavier particles with negative charge sink to bottom of storm.
Krebiehl (1986)

4. Cloud-to-Ground Lightning

5. High Speed Photography
Marcello Saba—INPE, Sao Paulo, Brazil
7000 frames per second

6. Up Close

7. Positive Flashes Deposit positive charge on surface
Usually a single stroke
Often strike away from storm core, up to 5-10 miles away
Longer continuing current
Stronger peak current
Therefore….more dangerous

8. Positive Flash Scenario
Often initiates in the upper (anvil) region of storms
After most of the heaviest precipitation has fallen from a cell, the upper positive charge is exposed to the ground.
Classical Thunderstorm
While

9. Satellite-Based Lightning Detection
Optical detectors on polar satellites
GOES-R to be launched in 2015

10. Ground-Based Lightning Detection

11. National Lightning Detection Network
• Data from 1989-Present

12. U.S. Lightning Distribution from NLDN

13. Florida Leads the Nation
Warm season (June-Aug)
Many heavily populated areas are vulnerable.
Enhanced flash densities are due to:
-- Sea breezes and lake/river breezes
-- Shape/orientation of the coastline.
Flashes/ km2/ warm season

14. Lightning Detection and Ranging (LDAR) Network
Detects VHF “sources”, or stepped-leaders of IC and CG flashes
Limited detection of sources near the ground (CG flashes)
Sources combined into flashes using spatial and temporal constraints

15. Initiation and Cessation
Typically, cloud-to-ground (CG) lightning is second-leading cause of weather-related fatalities in the U.S. behind flash floods
Most casualties occur either before or after the most intense lightning activity.
Forecasting the first and last flash of a storm is very important
Leon the Lightning Safety Lion

16. Forecasting the First Flash
Can you see the storm approaching?
Many studies have related radar parameters to first flash
Wolf [2006] studied over 1000 single cell, multicell, and supercell storms in the southern U.S.
Pretty good at forecasting lightning initiation
Is inverting lightning initiation criteria useful for forecasting total lightning cessation?
Charging is sufficient for CG flashes to occur
CG lightning initiation follows
Sufficient hydrometeors in the mixed-phase region of the cloud
Ascent of 40 dBZ echo above -10°C

17. Purpose of FSU Study Safety is critical at Kennedy Space Center
25,000 persons, $25 billion of facilities
U.S. Air Force’s 45 Weather Squadron issues lightning advisories.
Many operations must be suspended while lightning is a threat.
(From Weems et al., 2001)
(From Rudlosky)

18. Forecasting the Last Flash
Little is known about cessation
45WS maintains warnings too long
Wasted man power
Improved forecasts could produce a cost savings of $millions per year
But…safety is the primary concern

19. Warning Decision Support System – Integrated Information (WDSS-II) + LDAR
Allows you to merge and manipulate radar + LDAR data
LDAR source density grids (2 km horizontal and 1 km vertical resolution)
WSR-88D quality-controlled and gridded at ~1 km resolution.
Radar data ingested as it arrived in each elevation scan

20. Storm Cell Tracking Using WDSS-II clustering and tracking algorithm:
Identify storms based on a selected parameter
Set clustering thresholds low enough to track storm cores through cessation
Data mine each cluster
120 statistics are computed within each cluster to produce time-series
Clustering small, isolated, weakening cells = major challenge!
Maximum values are used, as in prior studies
VIL Density field with 4 associated storm clusters on 1 Sep 2005

21. Assess Temporal Relations
Describe lightning variability using radar characteristics 
Describe the variability in radar parameters using lightning data

22. Do Frequency/ Altitude Hold Answer ? IC Flashes CG Flashes
Time-Normalized Initiation Heights
Total IC Flashes = 14581
Height AGL (km)
IC Flashes
Do Frequency/
Altitude Hold
Answer ?
Normalized Time [0.0 = Time of first flash, 1.0 = Time of last flash]
Time-Normalized Initiation Heights
Total CG Flashes = 2238
CG Flashes
Height AGL (km)
Normalized Time [0.0 = Time of first flash, 1.0 = Time of last flash]

23. Do Peak Currents Hold Answer?
Time-Normalized CG Flash Peak Currents for 116 storms
Total CG Strikes = 2524
Peak Current (kA)
Normalized Time [0.0 = Time of first flash, 1.0 = Time of last flash]

24. Do Multiplicities Hold Answer?
Time-Normalized CG Flash Multiplicities for 116 storms
Total CG Strikes = 2524
Multiplicity
Normalized Time [0.0 = Time of first flash, 1.0 = Time of last flash]

25. Gradual or Abrupt Decay
Figure 18 is the time-height plot of several parameters for a typical single cell on 17 May The WDSS-II cluster-derived radar and LDAR source density data encompass the duration of lightning activity, from before the first flash at 1940:47 UTC until after the last flash at 2044:57 UTC. Greatest source densities occur at ~1956 UTC, ~33% through the storm’s duration. Thus, greatest source densities in the storm’s core occur near the middle of its life cycle when reflectivities greater than 40 dBZ extend higher than 7 km (the height of the -10°C isotherm). Most IC flashes initiate above the freezing level (0°C), while CG flashes originate at lower altitudes, generally below the height of the -10°C isotherm. The lapse rate of reflectivity at t = 0.5 (halfway through storm duration) is 4.05 dBZ km-1, increasing to 5.85 dBZ km-1 at cessation. The reflectivity contours remain at a relatively constant height during most of the storm’s lifetime, but descend during the last several minutes. The 40 dBZ echo descends below the height of -10°C by the time of the last lightning flash, an IC flash at 2044:57 UTC.

26. Outliers 20 min “outlier”
Figure 19 illustrates a single cell storm that occurred on 13 August WDSS-II cluster statistics begin ~2 min after lightning initiation. One should note the absence of lightning activity between 2049 and 2107 UTC. While no lightning occurs during this 18 min period, a final IC flash does occur at 2107:29 UTC. This is one of the 6 storms in the original 116 storm dataset that experiences an “outlier”-type of decay [Stano et al., 2009], with an apparent cessation at ~2047 UTC and one last “surprise” IC flash at 2107:29 UTC. IC flash initiations again are located above the 0°C isotherm (Figure 18); however, a comparison of reflectivity profiles (Figures 18 and 19) shows that the altitudes of flash initiations and source density contours are more dispersed in the current storm (Figure 19). The greatest source density again occurs at ~2027 UTC, about 33% through the storm’s duration. One should note that the height of -20°C is 9 km (1 km higher than in the previous case). The lapse rate of reflectivity is 4.80 dBZ km-1 at t = 0.5, increasing to 6.39 dBZ km-1 at the last flash. Contours of reflectivity exceeding 35 dBZ are at a nearly constant altitude below -10°C throughout the last 40% of the storm duration, from ~2046 to 2107 UTC, with downward sloping occurring only after the last flash. There is no clear indication that this storm will experience an “outlier” cessation behavior; however, reflectivity contours near the end of its lifecycle are more constant in altitude than the other two cases in this section. We hypothesize that the nearly constant reflectivity contours below -10°C indicate a small rate of electrification. Charge may build up gradually in the storm until the breakdown threshold is finally met, resulting in the final IC flash.
20 min
“outlier”

27. Lightning Bubbles

28. Numerous Statistical Approaches Tried
1. Forward Stepwise Regression
-- Sounding only parameters
-- Storm (lightning +radar) and sounding
parameters
2. Event Time Trends
-- Storm duration vs. Maximum interval
3. Lag Time of Radar Values to Cessation
-- Compare time from maximum height of maximum DBZ to time of cessation
4. Percentile Method ***
-- Break maximum interval values into
percentiles
The first four were trying to forecast the maximum interval, while the fifth forecast the time delay between the radar parameter and cessation. I am not sure if you would like to add a bullet point here stating that every method typically under-forecast cessation, except for the percentile method.

29. Most Successful – Percentile Method
Maximum Interval Percentiles
116 storms
1. Most Successful
-- Superior accuracy
-- Performed well with boot-strap method
-- Only method successful with “outliers”
2. Smaller Time Savings
-- ~5 minute savings
-- Accuracy from extended
wait times
-- Limited by “outliers”
-- However, provides certainty to advisories
Maximum Interval (min)
The chart simply presents the maximum interval for all 116 thunderstorms in the study with various percentiles indicated. This clearly indicates that the few “outlier” events tend to skew the wait time to being longer than necessary for most storms. However, the Percentile Method is the only method capable of maintaining a lightning advisory for the “outlier” events. The result is that it is the most accurate method (especially using the boot-strap technique), but it comes at the cost of a limited time savings. While the time savings is not as much as has been hoped for, it provides the 45WS forecasters certainty with the advisory and it is very easy to use. Additionally, you may tailor your risk, by selecting lower percentiles for shorter advisories if the danger of lighting is not as critical to operations (i.e. people are not involved outside). To use this, all the forecaster has to do is choose a percentile, and see if any more lightning occurs during the time interval when a flash is observed. If not, the advisory is cancelled. If another flash is observed, the clock is reset.

30. Provisional KSC Cessation Rule
1. Provisional Rule
-- Decaying storm with no expected redevelopment
-- Consider cancellation at 15 min
-- If redevelopment possible OR
-- Attached anvil cloud
-- Extend advisory up to 25 min and continue to observe before cancellation
2. Potential Time Savings
-- Current advisories maintained ~20-25 minutes
-- As Provisional Rule gains confidence may
decrease to min
-- ~22% time savings
-- Significant monetary savings over year
The Provisional Rule is based on the Percentile Method, but incorporates real-time observations and forecaster input. Essentially, storms with little chance of redevelopment (no imminent boundaries, decaying radar signature, etc.) will use the shorter 95th Percentile (15 min). Currently the WS is a little wary of 15 minutes as they want to see the Rule in action a little longer and are still erring on the side of caution. As a result, these shorter advisories are still running ~20 minutes. However, if a storm has an attached anvil, the full 99.5 Percentile (25 min) will be enforced or even extended as these appear to be the storms with large delays.
Point 2 outlines how this benefits the WS. As stated in the last slide, the Percentile Method (and thus Provisional Rule) support the existing minute wait times. However, as confidence improves, this may drop to minutes. Bill Roeder has said that even a 5 minute savings is still an approximate 22% savings in time. Combined over the entire KSC/CCAFS for a year, this will yield significant monetary savings.

31. Current Research Looking more closely at anvil region
Looking more closely at outliers (common features ?)
Include dual polarimetric radar in research
Explore hydrometeor profiles prior to cessation.

32. We Can’t Stop Lightning, But…

33. Acknowledgements Holly Anderson Melvin M.S. Geoffrey Stano Ph.D.
Funding Sources

34. Thank You for Inviting Me !!
Go Buckeyes—As long as your are not playing FSU

35. Lightning vs. Wind Direction
Wind from East Warm Season Wind from West