For the most recent version of this paper - e-mail
met71@hotmail.com
*** WORK IN PROGRESS ***
-------------------------------------------------------------
Forecast Applications Using Lightning Data: Nowcasting!
Charles Ryan
Electronics Engineer, Techniques Development Lab, New River,
Arizonia
Mark Mears
Software Developer Lightning 2000, President, Aninoquisi
Software Design, Huntland, Tennessee
John S. Sturtevant
Meteorological Technician, Techniques Development Lab, MRF
Co., Florence, Alabama.
-------------------------------------------------------------
Abstract
The object of this paper is to point out very obvious
citations in the literature that deal with using lightning as
a nowcast tool. True, we should use all of our tools in the
meteorological toolbox. These are radar, satellite, lightning
detection, Profilers, and other remote sensing tools. The
second purpose of this paper is to demonstrate the
applications for further development by the meteorological
community.
-------------------------------------------------------------
1. Introduction
Lightning detection is an age old tool that has just
recently gained the credibility that it deserves. From the
LDAR Projects, the National Lightning Detection Network, the
work of Dr. Krenhbiel of New Mexico Tech and the other work
done around the world, in the vein of better understanding
lightning and its relationship or correlation to severe
weather development. We have so far failed to make create a
lightning system and algorithms similar to those used in the
Storm Attributes of the WSR-88D NEXRAD weather radars.
My thrust here is to point out the more obvious
citations that lay across the literature and to create a
sense of perspective in further development that will lead to
sound lightning algorithms that can be used with our current
meteorological tool box, and used as a stand alone when
certain tools are not immediately available. Cases of this
nature may range from the radar being down, to solar flares
interfering with satellite imaging. Not to mention that
lightning could be useful for nowcasting (Holle, 1993). Which
is the thrust of this whole paper. The main objectives of
this paper is to look at Prediction, Onset, Detection,
Surveillance/Data Collection, Termination (i.e. collection of
all data related to the event) and notes of Post-Storm
Analysis. With an emphasis on Nowcasting.
2. Why is Lightning Data So Important
Real time lightning provides second by second
information about thunderstorms, that other sensors don't
provide (Holle 1993).
Lightning is usually the first indication of the
beginning of convection (Mosher 1989). Lightning data can
also detect lightning before the first echo appears on radar
in some cases (FYI#17). Lightning has a direct correlation to
the updraft which happens first in the thunderstorm sequence.
Lightning also is shows active and decaying areas of the
convection better than radar, and especially embedded
thunderstorms (Mosher, 1989). A strong updraft promotes more
frequent interactions between small and large ice
hydrometeors within the mixed phase of the thunderstorm
leading to electrification and lightning (Buechler, 1996).
When using lightning data it is good at showing
convective configuration, such as cells, lines,
redevelopment, and is usually the first indicator that
isolated convection forming into lines or an isolated cell is
forming out in front of a line. Also, the movement of this
convection is easier to determine using lightning data.
(Mosher, 1989).
Lightning generally occurs at the 0 C height and higher.
Most lightning is around the -20 C area. Charge regions are
always above the freezing level and correlate well with
precipitation regions (Schlatter, 1997). -20 C or higher
updrafts tend to increase flash rates. Updrafts greater than
2.5 m/s seem to be needed to produce lightning (Hallet, 1994)
A switch from negative to positive polarity usually
means a storm will become strong or becomes a supercell,
followed by large hail which generally occurs during positive
domination (MacGorman, 1994). Changes in polarity can also
signal a change in storm structure (Buechler, 1996). Sharp
increases in flash rate generally occurs just prior to
reports of severe weather (Knapp, 1994).
While there is a focal point for some time now on Cloud
to Ground lightning it is interesting to note that 75% of all
lightning is intracloud (SAFIR). RF Signals and visible
light from lightning travels at 186,000 miles in a second.
3. Prediction
First CG usually occur 5 to 30 minutes after the first IC
are observed, as IC decrease CG tend to increase, CG
generally follow the peak in IC by 5 to 10 minutes. Warning
times between first IC and first CG range from 5 to 30
minutes. (Richard 1990) Using LIST (Lightning Initiation
Signature) as developed at KSC, we would look for a 40 dbz
echo and a have about a 7.5 minute lead time to first CG
(Gremillion 1999). However other studies show, that we need
to look for a 30 dbz echo above 7 km and this gives us about
a 4.6 to 8 minute lead time to first CG Flash (Forbes 1999)
Another thing to look for is the 500 MB height and
lightning occurrence when the 500 MB lowers to -30 C (Sonoi
1999)
Using the hourly text RUC Analysis look at:
KI Index
SFC Dewpoint temp
Wet Bulb Zero Height
Theta-E
700-500 Lapse Rate
Vertical Velocities
-10 C Isotherm (most useful)
850 MB level
700 MB Level
Showalter Index
700 MB Wind Speed
CAPE
Lifted Index
850 MB Wind Shear
Precipable Water
Helicity
BRN
POS Shear
You can estimate cloud tops by using the following
formula: .47 x CAPE = estimated height (Wilfong 1993)
Some of the meso analysis features to look for is that
supercells are more likely in low Richardson Index areas (10
to 40 RI) with CAPE around 1500 to 3500 j/kg (Buechler 1990).
You can also use USAF AWS Formula's to estimate VIL's as
well.
You need to record these items along with lightning data
and storm data to create your own local climatology and
lightning forecast numbers.
Also do a routine meso-analysis of the area of concern,
when conditions warrant.
High vertical shear 20 ms or greater, a height locally
of the -10 C isotherm for lightning, and cloud tops forecast
to 9 km or greater, CAPE 400+, LI -2, DP 55 F+, Echo Tops 30
to 50,00 feet, VIL's 30 or less, 850-300 MB strong Vertical
wind shear, high Theta-e, Steep lapse rates are good
indicators that lightning will be present in the forecast
area.
Keep in mind that lightning and its onset is latitude
dependent in many ways, so please work on creating your own
local climatology. as these numbers are not a one size fits
all situation.
4. Onset - Forecasting Cloud to Ground Lightning
Lightning is usually the first indication of the
beginning of convection. Lightning has a direct correlation
to the updraft which happens first in the thunderstorm
sequence (Mosher, 1989).
It has been noticed with the LDAR system that Cloud to
Ground Flashes occur about four to five minutes after the
first intra-cloud lightning are observed.
-------------------------------------------------------------
Echo Tops, Vertically Integrated Liquids, Maximum Vertical
Velocity, and Convective Available Potential Energy
-------------------------------------------------------------
Tops in (Feet) VIL Vel CAPE
10,000 10
16,500 -- 3 100
20,000 20
30,000 30
33,000 -- 12 800
40,000 43
49,000 -- 24 1700
50,000 55
60,000 63 50 4100
-------------------------------------------------------------
One can expect correlations at the outset between total
lightning activity and the development of severe weather
(Williams 1998). Correlations between the updraft strength
and other variables such as cloud top height, cloud water
mass, and hail size have also been observed (Goodman, et
al.,1998). Lightning stroke count correlates well with
updrafts (McCaul, per. comm.) Parameterization relationships
between CAPE or Cloud top height and lightning frequency were
shown in Guyana (J. Moline)(Hallet, 1994). Vertical Velocity
is other area of correlation as it is a function of CAPE and
Cloud Top Height (Hallet, 1994).
The lightning discharge is up to 100 Km long and the
thunderstorm convection scale often exceeds the depth of the
atmosphere (Hallet, 1994). Non-Severe storms may only have 2
to 10 stokes per minute where as severe storms may have 10 to
40 strokes per minute based on the NLDN (Doswell, 1985).
Lightning generally, has to have vertical velocities of 5
to 20 m per sec, and advect vertically above the -10 C
isotherm and some times through the tropopause to -60 C.
Flash rates tend to increase as the vertical growth crosses
the -20C isotherm. E-fields of 100 -300 Kv/m must exist to
initiate lightning (Schroeder 1999)
For onset of lightning one looks for a 35 dbz echo at
the -10 C to -20C level. Look for -25 C to -30 C cloud tops.
We need moderate instability to have lightning in most cases.
Mature cells can range from 35 dbz to 65 dbz. The location of
the first pulse of lightning is generally in the updraft
region where the mixed phase hydrometeors exist and interact
(Markson 1999). A more general rule for lightning potential
or onset maybe obtained by watching the 35 to 45 dbz echo's
at the -10 C isotherm (VISIT). Either way we get the general
principle which will be latitude dependent as well.
Cloud top heights range from 6 km to 15 km, these tops are
lower over positive flash regions than over negative regions.
There is a suggestion that +CG try to avoid rainy areas, this
of course is not always the rule.
The 6 to 9 km area of the atmosphere seems to be the
most active area for lightning. With CID's occurring around
the 4 to 11 km area, or 47 to 58 dbz echo regions.
Echo Tops: (modified from Watson 1994)
<35,000 least flashes
45 to 50,000 most flashes
>55,000 least flashes
Intra-cloud lightning represents 70% to 90% of the total
lightning activity of a thunderstorm. There tends to be a lag
of about five minutes between the IC peak and the CG Peak
(Richard 1990). Rust, 1990, however found that IC flashes
were about 80% of total lightning. Maximum lightning activity
closely follows maximum vertical development of a convective
cell (Williams 1990)
Positive CG are generally associated with the growth of
an anvil.
Lightning can occur at -4 F surface temperature (Moore
1999).
Updrafts of 2 - 3 ms are sufficient for lightning
(NWSTC) However, most the time these updrafts are much
stronger and supported for long durations of time. Keep in
mind that lightning rates tend to be proportional to the
fifth power, of the updraft velocity (Markson 1999) Higher
the flash rate higher the updraft velocity. There tends to be
a 10 minute lag time between large updrafts and high
lightning flash rates. An average updraft velocity is 7 to 8
m per second (Schroeder 1999). Lightning discharges most
often occur from 10 to 20 min after graupel region with temp
of -15 C to -20 C is observed at heights of 4 to 5 km
(Kouichirou 1999).
CG strokes occur mostly down shear of the main
reflectivity core in supercells. In supercells it is mostly
on the left side with respect of storm motion. In multi-cell
storms lightning is near the reflectivity core (Reuter 1999).
High Shear environments produce primarily +CG while, low
shear environments produce -CG (Helsdon 1990).
A. Notes on Lightning
Each Flash can contain between 1 to 8 strokes. Positive
Flashes generally have 1 - 4 strokes, with a mean
multiplicity of about 1.15. While Negative Flashes have a
mean multiplicity of about 2.25. The Overall multiplicity
seems to be around 2.08 - 2.30, depending on who's study and
data you look at. Negative Flashes seem to have 22 ka, and
Positive Flashes 10 - 40 ka. With 25 Ka, being the middle or
median current level.
Flash to Stroke criteria is that a stroke must be the
same polarity, within 2 seconds of the first stroke, and
within 10 km of the first stroke - these combined make up a
flash and the multiplicity of the flash. In short, how many
strokes were associated with a given flash (Watson 1994).
Strikes can occur up to 6 km from the outside of the
cloud base or just beyond the 20 dbz echo contour (Forbes
1995).
Thunderstorm updrafts are around 0 to 50 m/sec. The
Negative charge are is near 6 km MSL at about -15 C, and the
Positive Charge area is around 8 to 12 km. Strong electrical
field and presence of hydrometeors is enough for a lightning
discharge (Richard 1990).
B. Types of Lightning
These are the most common forms of lighting forecaster
will generally deal with:
Positive Cloud to Ground (5 Km MSL -5C charge area)
occur as the storm dissipates, or with Anvil LTG, some
times with winter storms. PSD with severe storms,
associated with Tornadoes, hail. Tend to occur mostly in
stratiform areas (MacGorman 1993).
Negative Cloud to Ground (5 Km MSL -25C charge area,
peak current tends to be 60ka (Orville 1999).
Positive Intra-Cloud (8 Km MSL)
Negative Intra-Cloud (8 Km MSL)
Leaders
CID (Compact Intra-Cloud Discharge)
CID Narrow Bipolar
CID Narrow Negative Bipolar
CID Narrow Positive Bipolar
Large Peak Current (mainly +CG)
Spider Lightning: Spider lightning can jump from cloud
to ground usually a positive stroke in the stratiform
precipitation region. However, most spider lightning is
intracloud.
C. Most Lightning detection systems work on the following
frequencies: 5-15, 60 khz, 2, 3, 4, 20-80, 30, 33.4, 34.3,
50, 66, 69, 75, 110-118, 114, 139, 250, 253, 274, 295, 327,
350, 600, 1430 Mhz.
D. Types of Lightning Detection Systems are: Magnetic
Direction Finding, Time of Arrival, Single Station Loop
Antenna, and some have a variety of technologies.
6. DETECTION
A. Severe Thunderstorms
IC and CC are the most important in initially determining
which storms will become severe. There tends to be a rapid
increase in intracloud flash rate 1 to 15 minutes in advance
of severe weather manifestation at the surface.
-CG tend to dominate near the vicinity of deep convection
(MacGorman 1993).
30 to 60 flashes per minute (62 - 125 strokes per minute
approximate) lightning jumps are generally associated with
updraft intensification and storm severity (Goodman 1999).
A majority of severe storms showed a rapid increase in total
lightning (or lightning jump) prior to the onset of severe
weather (Hodanish 1998). Lightning rates increase
significantly when there is a marked increase in updraft
velocity (MacGorman 1985). Knapp found that jumps can occur
an hour prior to severe weather occurrences, and that peaks
in flash rate tend to indicate hail begun (Knapp 1994). The
NWSTC tends to indicate that tornadoes and hail occur about
10 - 15 minutes after a pronounced peak in 5 minute CG rate,
and microburst can occur about five minutes after a
pronounced peak (NWSTC).
Flashes Strokes LTG Notes
-------------------------------------------------------------
>300 >600 High Straight line winds
100 - 300 200 - 600 Mod Tornado Potential
< 100 < 200 Weak Hail Potential
(adapted from NWSTC)
-------------------------------------------------------------
When outflow dominates there seem to be a decrease in CG
LTG and when this occurs downburst, tornadoes, or hail at the
surface are possible (NWSTC). Positive cloud to ground
flashes sometimes dominate vigorous storms that produce
tornadoes and large hail (.75 inch plus) at the surface
(Macgorman 1994). Large hail and tornadoes are often reported
for storms dominated by positive CG lightning (Gilmore 2000).
Most severe weather is west - northwest of the lightning
maxima. 60 to 68 strokes per minute have a close correlation
to severe weather occurrences. Strokes are generally
correlated or concentrated near the mesocyclone core when a
meso-cyclone is present (NWSTC).
CG LTG is most frequent prior to the occurrence of
severe weather when the inflow to the cell is greater than
the outflow (NWSTC).
B. Positive Stroke Dominated Storms
Storms that dominate with +CG for tens of minutes almost
always produce severe weather. When relatively high flash
rates and densities of positive CG lightning occur they are
associated with severe weather (MacGorman 1993).
PSD storms tend to have 40 to 55,000 ft echo tops, with
VIl's ranging around 15 to 30. 31 to 33,000 mean echo tops
are the median for most PSD storms (Watson 1994). With most
CG to strokes having a maximum density at about 6 km above
MSL (Solomon 1999). (Stolznburg, 1994) found that these echo
tops generally have a 50 plus dbz intensity.
Strong shear may be necessary for Tornadic +CG PSD
storms.
+CG tend to come from (NWSTC):
1) High on the back of the main storm tower.
2) Through the wall cloud.
3) Downshear of the anvil.
4) Stratiform precipitation region.
When updrafts are very strong predominate polarity tend
to be positive, further PSD tend to have a high flash
density. (NWSTC)
Most all PSD storms have some rotation, and most storms
produce tornadoes (NWSTC).
+CG are generally in the area of maximum reflectivity
and +CG storms are generally strong mesocyclones, they tend
to be west - northwest of the -CG centers. Sometimes flash
rates decrease to near zero for tens of minutes prior to
polarity changes. This is repetitive but during this time
long lived violent tornadoes can begin! (NWSTC).
PSD cells generally display a well defined peak-lull-peak
with most significant amplitudes occurring 20 minutes prior
to the report of a tornado. Polarity shifts at time of
report. NSD tend to shift to PSD at time of tornado
occurrence (Knapp 1994).
PSD storms are generally accompanied by large hail, as PSD
decrease to NSD the hail decreases, +CG in dense clusters
frequently are associated with large hail and many produce
tornadoes (MacGorman 1993, Shafer 2000).
PSD storms tend to have vertically integrated liquids that
tend to be higher during +CG domination as well as hail seems
to be of greater size during +CG phase. At the change in
polarity hail size tends to decrease (MacGorman 1994).
C. Forecasting Mesocyclones and Tornadoes
It is important to note that when trying to use
lightning to forecast onset of tornadoes is to use intra-
cloud and cloud to cloud lightning. McCaul cites that in some
tornadoes only a few Cloud to Ground strokes have occurred
(McCaul, per. comm.)
Positively dominated storms produce rotation most of the
time. 1.5 or more CG strikes per minute with polarity change
from positive to negative polarity produce the most damaging
tornadoes (Check it Out 96-06). 10 KHZ Sferic rates peak
about 1.5 hours before tornadoes and decrease during
tornadoes to 40% of the peak value. 150 KHZ Sferic rate were
found to increase until tornadoes were produced and are
exceptionally high during periods when tornadoes occurred and
begin decreasing when severe weather is ending. 3 MHz sferic
rates are much higher in tornadoes than non-severe storms.
Horizontal polarization increases as vertical polarization
decreases as the storm severity increases. Flash rates range
from 60 to 90 per minute before tornadoes and increase with
tornadoes. All storms with frequent positive ground flashes
produced large hail and may produced tornadoes (MacGorman et
al., 1993). During lulls large hail and long lived tornadoes
occur (MacGorman, 1994).
Significant amplitudes occur during the 20 minutes prior
to the tornado report. Knapp goes on to say this is a well-
defined Peak-Lull-Peak rate pulsing along with the polarity
change we have already discussed. Once the Tornado is ending
or has ended the polarity switches back. Regardless of the
standardized flash rate there is a distinct peak at tornado
report time. Then the flash rate rises for the next ten
minutes with a slight lull between peak and continuing rise
or peaking to around 228 strikes per minute on the NLDN
system. 44,900 to 50,000 foot echo tops are average for
Tornadoes (Knapp, 1994). In short a 230 hour flash rate (62
to 125 strokes per minute with 80% PSD tendency) on average
may indicate potential tornadogensis (Goodman 1999).
Another way to look at this is that tornadoes begin
after the PSD change to NSD, or PSD decrease as -CG increase.
Some storms decrease from PSD to NSD tens of minutes before
NSD begin, during this period sometimes long lived violent
tornadoes begin (MacGorman 1993). CG peak 10 to 20 minutes
prior to tornado, there is a polarity change from POS to NEG,
and IC increase, also NSD tend to show PSD as tornado begins
(Perez 1997). Decrease in total flash rate just prior to or
during the tornado relates possibly to the collapse of the
bounded weak echo region which may be indicated by a average
208 stroke per minute decrease (Goodman 1999).
Tornadoes occur at the peak and each peak is larger than
the preceding peak. However it is also observed that the
tornadoes occur 20 minutes after the maximum peak was
observed. This is heavily dependant on the type of detection
system that is being utilized. Wall Clouds are often observed
just prior to the maximum observed peaks (MacGorman,
1994). Another example may be that tornado production began
during or just after relative maximum in the -CG strokes, and
declines in lightning activity appear to accompany the
initiation of each tornado. Comparison of VIL to CG rate may
be a useful way to identify tornado episodes (Buechler,
1996). There are cases where VIL remains high, even if CG
rate decreases at onset of tornadogenisis.
Sudden lightning jumps nearly always precede the descent
of a mesocyclonic circulation. These jumps are typically 30
to 60 flashes per minute. These jumps may be associated with
rapid vertical growth of storm updrafts and circulation tends
to appear and descend at the peak of the jump. Intracloud
generally dominate at this point, just prior to the descent
of the mesocyclone circulation (Goodman 1998). Intracloud
flash rates tend to be unusually large during violent
tornadoes (MacGorman 1994).
Sudden lightning jumps are a signature of rapid
intensification of the vortex stretching and associated
concentration of angular momentum which becomes the tornado
(Goodman, et al., 1998). Dominant polarity of ground flashes
changes from positive to negative, with peak negative ground
flash rate comparable or larger than the earlier positive
peak. In such cases, the first change occurred as the storm
begins intensifying; VIL and storm Height both increased
shortly before +CG flashes became dominant (MacGorman, 1993).
One must note that the lull is when the most violent tornado
began. A forecaster may expect a tornado occurrence from
positively dominated cells that display a rapid increase
after an earlier lull, the polarity switch is generally 10
minutes prior to the tornado (Knapp, 1994) however in the
example above the tornado can occur at the peak or as the
peak is being achieved.
Another trait to watch is for storm movement in which a
potentially tornadic cell is turning right. Right turners
are often produce very violent tornadoes. When lightning or
storm mass is concentrated higher in the storm the better
the chance of it becoming tornadic (Buechler, 1996)
Strength of the updraft and strength of the mesocyclone
appear to affect how charge is transported (Bluestein 1990).
Using a five minute Storm Cell graph may be a very
important tool in assessing the present storm, however, a one
needs to look at the last 20 to 65 minutes for traits that
may indicate much of what just discussed. The most
significant amplitudes occur within 20 minutes of the report
of a Tornado.
Summarizing, One needs to watch the five minute stroke
rate. Look for Peak - Lull - Peak signatures. Most
significant amplitudes generally occur within 20 minute of
the report of a tornado. Rapid increases in stroke rate after
an earlier peak and tightly grouped may indicate very rapid
intensification. In the southeast, on might look for a
transition of positive stroke dominated storms to a negative
stroke dominated storm, which may indicate that a tornado has
or is about to be produced. It is important to use weather
radar along side lightning data.
A switch from positive to negative is considered
potential tornadic signature. Switch is polarity may indicate
rapid growth or substantial change in storm structure
(NWSTC).
HP supercells when they change from positive to negative
polarity they become tornadic.
When tornadoes dissipate the -cG flash rate may increase
to very large values (MacGorman 1994).
Intra-cloud flash rates tend to increase during
tornadoes (MacGorman 1985).
Most significant amplitudes in flash rate occur 20
minutes prior to first touchdown, and -CG shift polarity at
time of touch down (NWSTC).
Peak CG rate precedes tornado formation most all of the
time and stroke rate decrease with touchdown of the tornado
(NWSTC).
Forecasters can expect tornado occurrence from PSD cells
that display a rapid increase from an earlier Peak-Lull
(Knapp 1994).
D. Forecasting Hail
Frequent Positive flashes are a good indicator of large
hail, especially during the time in which the cell is
positive strike dominated (Check it Out, 96-06, Hallet,
1994). Peak flash rates occur just before the hail begins
(Knapp, 1994).
Significant VIL drops from peaks tend to indicate large
hail production. 3.5 VIL Density is a good indicator of
severe hail (Check it Out, 96-06) VIL Density is simply the
VIL divided by the Echo Height.
One might watch for "Sharp" increases in flash rate,
with peaks just prior to hail at the surface.
Generally speaking, CG lightning centers tend to develop
9 minutes before hail at surface. These centers tend to be 3
miles upstream of the hail fall. With hail fall occurring 5
kilometers down shear from the storm's track. Lightning
diminishes generally after a hail fall. Most hail fall have a
average duration of about 26 minutes. (NWSTC).
There may be a 146 strokes per minute or greater
associated with a cell producing cell. There is generally a
rapid increase in flashes just prior to the hail. There is
generally a peak just before the hail begins and a uniform
flash rate in the cell during the hail fall (NWSTC). Look for
frequent +CG and +CG density for possible hail, the presence
of multiple CID's are a good indicator of hail as well.
Hail Size
--------------------------------
Lightning Jump Hail Size
Strokes Per minute inches
65 - 70 1"
70 - 85 1.5" to 1.8"
85 > 2.0" >
Sparity of CG strokes in a storm having a mesocyclone is
a good indicator of hail (Shafer 2000).
E. Compact Intrcloud Discharges
CID's may be very good potential hail indicators.
Compact Intracloud Discharges (CID's) occur between 8
and 11 km AGL at the 40 to 58 dbz level and are generally at
the 5 to 25 Mhz detection range. Positive CID's are most
common. Negative CID's occur at higher altitudes. CID's track
well with the movement of the storm and are generally
centered on the high reflectivity core (Smith 1999).
F. Forecasting Microbursts
Total lightning flash rates tend to track the updraft
with rates increasing as the updraft intensifies, and
decreasing rapidly with cessation of vertical growth which
signals the onset of the microburst or downburst (Goodman, et
al, 1998)
Studies in the southeast US have found that -CG flash
rates tend to increase as reflectivity cores develop
downwards and tend to peak a few minutes before the
microburst and sometimes as the peak occurs. Damaging winds
tend to occur as minimums are reached (MacGorman, 1996).
We all know that wind speeds at the surface at 600 or
even a 1000 strokes per minute rarely exceed the speeds found
in this table but it useful in terms of downburst estimates,
but used very cautiously at best if used at all.
Wind Speed
---------------------------------
Lightning Jump MPH
60 58
160 182
600 260
Peak IC tend to occur at the time of maximum
reflectivity aloft. The IC LTG data tend to show potential
for short term predictor of microbursts (Rust 1990). CG tend
to increase with development of strong downdraft. Peak CG
follow about 10 minutes later after IC peak (Richard 1990).
Microbursts seem to be driven by the collapse of the hail
core (Hoffert 1996). Peak lightning rate dominates 4 to 10
minutes ahead of the microburst (Williams 1990).
The peak in the intracloud activity near the time of the
peak vertical development precedes the maximum outflow at the
ground (28 m/sec) by about five minutes (Williams 1990).
Decreasing CG indicate decay or collapse 3 to 5 minutes
ahead of a microburst at the surface, providing up to five
minute lead time (NWSTC). In HP storms high volumes of LTG
may be present (Knupp 1994).
Some Thermodynamics to look for are a 18 gm/kg
approximate mixing ratio, 1500 to 3000 CAPE, and 20 m/sec
outflow (Williams 1990).
G. Rainfall
There have been may studies done on the rainfall to
lightning stroke/flash relationships. The values from study
to study vary greatly. Some values follow:
.002" per stroke
.17" per stroke
.28" per stroke
.0025" per stroke
.00001" per stroke
.01987" per stroke
The median seems to be .07906 inches per stroke. .00001
inch per stroke is the number referenced in the literature
(Sheridan 1997).
My gut observation is that all these values are latitude
and geographically dependent and sometimes "system type"
dependent. Therefore knowing the actual rain rate via other
sensors is important, along with radar derived rain rates,
and then lightning rain rates may be useful where radar is
limited or not able to interrogate a certain area of
interest.
However, convective rainfall is not limited to lightning, and
great caution must be used in lightning rain rates, which can
be at times fictional at best.
Rainfall tends occur in the following manner:
0 minute - Max Updraft
10 minute - Max Lightning
16 minutes - Max Rainfall
Maximum rainfall tends to coincide well with those areas
that receive the highest concentration of CG strokes. My
personal observation has been that areas of highest IC/CG
concentrations are the areas of highest rainfall.
The areas of highest precipitation generally are the
areas of highest density of CG lightning (Orville 1997).
Clearly more research has to be done to understand the
relationship between precipitation and lightning (Orville
1997).
H. Forecasting Flash Floods
The relationship of lightning to rainfall rates has been
made time after time in the literature (Mosher, 1989).
Precipitation generally starts at the surface about 10 to 20
minutes after the first intra-cloud lightning is observed
(Schlatter, 1997). Maximum flash rate may infer flash
flooding (Holle, 1994). Greatest rainfall rate seems to occur
5 to 40 minutes after peak stroke production period (Henz,
1996). For individual cells, peak lightning activity is
expected to precede the peak rainfall observed at the surface
by about 2 to 15 minutes (Buechler, 1994).
Forecasters need to look for storms that move 14 miles
per hour or slower, with high stroke density, which is often
associated with heavy rainfall.
Predominate stroke density maxima, coincide well peak
rainfall rate maxima. This can be easily correlated in most
cases to NEXRAD Storm Total Precipitation product.
In flash flood situations the storm generally runs
through three phase:
Rain Rate 15 Minute
mm per min Peak Stroke Rate
-------------------------------------------------------------
1) Storm Growing 25 - 100 52
2) Three Peak Maxima 100 - 250 300 +
3) Weaker 50 - 100 260
-------------------------------------------------------------
Look for lightning repeated over the same general area,
with same similar patterns or training affect. This can be a
good clue of potential onset of flash flooding from
convective rainfall.
A potential rain rate per hour algorithm would look like
this:
Correlated Rainfall Rates
----------------------------
Strokes Per Rainfall Rate
Minute inches per Hour
200 1/2"
400 1.0"
600 1.5"
800 2.0"
1000 2.5"
I. Winter
Updrafts of 2 to 3 m/sec are sufficient for lightning.
-10 C is:
<4600 ft 1.4 km 1400 m no LTG
>5800 ft 1.8 km 1800 m Strong LTG
(adapted from NWSTC)
Snow storms with Lightning are possibly producing snow
bursts of 2 to 4 inches per hour. They are also likely to be
+CG polarity. (R-SCAN)
7. Dissipation
When the flashes change from mostly negative
polarization to Positive Polarization then cells will
dissipate in about 30 minutes. This because the updrafts
cease, this can be seen on radar shortly after the lightning
stops. The intensities of the cells will decrease after about
ten to twenty minutes. (Mosher 1989).
8. Termination
When an event ends it is best to archive the lightning
data, the radar data, satellite data, and rainfall, and
NEXRAD storm total precipitation products and other data that
is useful in post analysis and creating site specific
lightning climatology for use in future nowcasting scenarios.
9. Post Storm Analysis
There will in time be tools for more post analysis of
lightning data of all types which will certainly allow for us
to be able to better determine what lightning signatures to
look for and for software based systems to alert the
forecaster to potential trouble ahead of time, providing
significant lead times in some instances for better warnings
and fewer false alarms.
10. Surveillance
Most LTG data is captured via the NLDN and utilized by
many entities. However, single station sensors exists now
that use software to display lightning such as NEXSTORM,
Lightning 2000 and Boltek and others. There a other systems
such as GP-1, LDAR, and space borne optical sensors.
Regardless much can be gained by using lightning data of all
types and this will certainly lead to better algorithms and
much better detection of severe weather in the future.
11. Summary and Conclusions
-------------------------------------------------------------
APPENDIX
Adopted from Devore using Flashes:
Flash Prob Radar Radar Rainfall Est
Density Hail VIP dbz Rate VIL
--------------------------------------------------------
.0002 0% 0 0 0 0
.0035 0% 1 23.5 .055 0 First IC
.0093 0% 2 35 .295 5 IC Present
.0178 10% 3 43 .745 10
.0241 50% 4 47.5 1.495 28
.0334 75% 5 53 3.495 48
.0498 100% 6 60+ 5.000 > 70 Large Hail
--------------------------------------------------------
-------------------------------------------------------------
Corresponding Author: John S. Sturtevant, 326 Crestwood
Drive, Florence, Alabama 35633-1468, met71@hotmail.com
-------------------------------------------------------------
Acknowledgements: Would like to thanks, our working team
consisting of Charles Ryan, Mark Mears, John Sturtevant for
their outstanding work. University of North Alabama Library
in Florence, Alabama for their assistance in locating
research materials. For all those we may have failed to name
that have contributed to this paper, including the countless
authors in the reference section for their time, effort and
energy as well.
-------------------------------------------------------------
REFERENCES
(References in-complete)
Buechler, D.E., H. Christian, S. Goodman, 1994, Rainfall
Estimation Using Lightning Data, 7th Conference on Satellite
Meteorology, AMS, Boston, Mass.
Buechler, D.E., S.J. Goodman, E.W. McCaul, K. Knupp, 1996,
Cloud to Ground Lightning Activity, within Tornadic
Supercells in the Tennessee Valley, 18th Conference on Severe
Local Storms, AMS, Boston, Mass., 499 - 503.
Buechler, D.E., R.J. Blakeslee, H.J. Christian, R. Creasey,
K. Driscoll, S.J. Goodman, D.M. Mach, 1996, Lightning
Activity in Tornadic Storm Observed by Optical Transient
Detector (OTD), 18th Conference on Severe Local Storms, AMS,
Boston, Mass., 494 - 498.
Check It Out - Severe Weather, 96-06, HQ AWS, XON, Scott AFB.
Doswell, Charles A., III, 1985, The Operational Meteorology
of Convective Weather, Volume II, Stormscale Analysis, NOAA
Technical Memorandum ERL ESG15, 193 - 203.
FYI #17, Lightning Detection Systems, FYI Series, US Air
Force Weather Agency (DOD).
Goodman, Steven, J. R. Raghaven, R. Ramachandran, D,
Buechler, S. Hodanish, D. Sharp, E. Williams, B. Bold, A
Matlin, M. Weber, 1998, Total Lightning and Radar Storm
Characteristics Associated with Severe Storms In Central
Florida, AMS 19th Conference on Severe Local Storms, 639 -
642.
Hallet, John, 1994, Conference on Atmospheric Electricity:
Thunderstorm Electrification, St. Louis Missouri, 4-8 October
1993, Bulletin of the American Meteorological Society, Volume
75, No. 11, November 1994. American Meteorological Society,
Boston, Mass., 2159 - 2163.
Henz, John F., 1996, Innovative Re-Construction of a Western
Flash Flood using Cloud to Ground Lightning, Radar
Reflectivity, and Surface Relationships, 18th Conference on
Severe Local Storms, AMS, Boston, Mass., 474 - 478.
Holle, Ronald L., R. Lopez, 1993, Overview of Real Time
Lightning Detection Systems and their Meteorological Uses,
NOAA Technical Memorandum, ERL NSSL 102, NSSL / NOAA.
Knapp, David I., 1994, Using Cloud to Ground Lightning Data
to Identify Tornadic Thunderstorms Signatures, and Nowcast
Severe Weather, National Weather Digest, Volume 19, No. 2,
National Weather Association, Montgomery, Alabama, 35 - 42.
MacGorman, Donald R., C. Church, D. Burgess, C. Doswell, R.
Davies-Jones, 1993, Lightning in Tornadic Storms: A Review,
The Tornado, It's Structure, Dynamics, and Hazards,
Geophysical Monograph 79, American Geophysical Union, 173 -
182.
MacGorman, Donald R., D. Burgess, C. Morgenstern, 1993,
Positive Cloud to Ground Lightning in Tornadic Storms. AMS
Conference on Atmospheric Electricity, J-34 - J-39.
MacGorman, Donald R., D. Burgess, 1994, Positive Cloud to
Ground Lightning in Tornadic Storms and Hail Storms, Monthly
Weather Review, Volume 122, No. 8, AMS, Boston Mass, 1671 -
1697.
MacGorman, Donald R., M. Filiaggi, K. Hondl, 1996, Cloud to
Ground Lightning in the Lahoma Oklahoma Storm of August 17,
1994, 18th Conference on Severe Local Storms, American
Meteorological Society, Boston, Mass., 57-59.
McCaul, Bill, USRA, Global and Hydrology Center, Huntsville,
Alabama. Personal Communications.
Mosher, F. R., 1989, Uses of Real Time Lightning Data,
National Severe Storms Lab, Norman, OK.
Safir Commercial Literature.
Schlatter, Thomas, 1997, Weather Queries: When Lightning
Strikes, Weatherwise, June/July 1997, P. 40.
Williams, Earle, B. Bodi, A. Matlin, M. Weber, S. Hodenish,
D. Sharp, S. Goodman, R. Raghavan, D. Buechler, 1998, Total
Lightning as a Severe Weather Diagnostic in Baroclinic
Systems in Central Florida, 19th Conference on Severe Local
Storms, AMS, Boston, Mass, 643 - 647.
**** Note - References are incomplete at this time ****