Oceanic Thunderstorm Characteristics and
Lightning Data Assimilation into NWP Models
Antti Pessi and Steven Businger
University of Hawaii
Department of Meteorology
PacNet/LLDN
DE day
LA
DE night
Relationships Between Lightning, Precipitation, and
Hydrometeor Characteristics over the North Pacific Ocean*
Oceanic thunderstorm characteristics were investigated by comparing
lightning data from PacNet and TRMM’s Lightning Imaging Sensor to
precipitation and hydrometeor data from TRMM’s Precipitation Radar
and Microwave Imager.
*Pessi and Businger 2009, JAMC, in press.
Data and Methods
• Data from over 2000 TRMM overpasses during a 3 year period
(February 2004 - February 2007).
• Lightning data from PacNet (quality controlled) and TRMM’s
Lightning Imaging Sensor.
• Precipitation and hydrometeor data from TRMM’s Precipitation
Radar (2A25 v6) and Microwave Imager (2A12 v6).
• Domain divided into 0.5° x 0.5° grid cells.
• PacNet lightning strikes counted ±15 min from satellite
overpass time.
• DE model applied to PacNet data to quantify the lightning
rates.
• Data divided to winter (October-April) and summer (JuneSeptember) storms.
Summer Thunderstorms
250 hPa
30 August 2005 1200 UTC
Associated with
upper-level lows
(TUTT cells)
Surface
Winter Thunderstorms
250 hPa
12 December 2005 0600 UTC
Associated with
cold-fronts and
kona lows
Surface
Lightning vs. Radar Reflectivity
-20°C
0°C
-20°C
0°C
Black symbols PacNet data
White symbols LIS data
Grey symbols combined PacNet/LIS
Latent heat
Precipitable ice
Precipitation water path
Convective rain
Ice water path
Stratiform rain
The Impact of Lightning Data Assimilation (LDA)
on two Winter Storm Simulations over the North Pacific
Latent Heating Assimilation Method
1.
Create an input file before the
model run starts

Apply DE model to quantify the
lightning rates (lat, lon, LT at each
flash location).



2. Convert the lightning rates to rainfall
rates over the whole domain and
each timestep using the relationship
formula.
Assimilation method
The method was programmed to MM5’s
Kain-Fritsch convective parameterization
scheme.
The method uses Newtonian relaxation
(nudging) technique to adjust the model’s
vertical latent heating profiles according
to ‘observed’ rainfall rates.
Adjustment is done in the model’s
convective temperature tendency
equations.
The method is a 4DDA-type assimilation
method, where nudging occurs during the
forecast run.
Latent Heating Assimilation Method
Go through each grid point:
Is there lightning in the grid cell?
No
Yes
No action.
No lightning ≠ no rain
Is model producing rain?
No
Search adjacent model grid points
for similar rain rates as observed.
Use modeled latent heating profile
from matching grid point. Saturate
levels where dT/dt>0.
Yes
Is observed rain>model rain?
No
No action.
Use modeled profile.
Yes
Scale latent heating rates:
R  Rm
c o
,(c  3)
Rm
Ti  (1 c)Timdl
Experiment Design

MM5 v3 model

Initial conditions from GFS-model (T254L64 ~55 km grid)

Boundary conditions every 6 hours

Horizontal resolution 27 km, 39 vertical levels, no nesting

Kain-Fritsch convective parameterization scheme
Model integration 24-48 h
Assimilation 8 h

Horizontal radius of influence 0.125-0.25°

Time window of influence ±15 min

Model time step 81 sec

Vertical latent heating profiles adjusted every timestep

Three forecast lengths: 24, 36, and 48 hours
12-Hour Forecast valid 1200 UTC 19 Dec. 2002
Sea-Level Pressure and 3-h Accumulated Rainfall
CTRL
LDA
982 hPa
977 hPa
Analyzed pressure
was 972 hPa
Difference
LDA-CTRL
Central Pressure - 24-h Run
CTRL
Analysis
LDA
Assimilation period
Central Pressure - 36-h run
CTRL
Analysis
LDA
Assimilation period
Central Pressure - 48-h Run
LDA
CTRL
Analysis
Assimilation period
How Did the Assimilation Affect the Storm
Dynamics to Achieve an Improved Forecast?
Lightning was observed hundreds of
kilometers away from the storm center over
the cold front and cold pool.
Advection of Warm Air over the Storm Center
Vertically integrated
virtual temperature
(shaded, LDA-CTRL)
SLP (contours, LDA)
03h
06h
09h
12h
Upper figure:
(a) CTRL, (b) LDA
Wind speed at 400 hPa (m/s, shaded)
Temperature at 400 hPa (K, contours)
Latent heating, as indicated by the high
lightning rates, increased temperature and
T across the front.
This resulted in increased along-front winds,
consistent with thermal wind balance.
Lower figure:
Difference between LDA and CTRL in:
Virtual temperature (K, shaded)
Geopotential height (m, contours)
Enhanced advection of warm air over the
storm center dropped the surface pressure
hydrostatically.
Advection of Warm Air over the Storm Center
pslv  pz exp(
g0 Z
__
)
Rd T v

pslv = sea-level pressure
pz = pressure at level z
g0 = gravitational constant
Z = height of the pz pressure surface
Rd = gas constant for dry air
Tv = average virtual temperature
between sea-level and pz
Plugging in the values results in
SLP of 982 and 977 hPa for
CTRL and LDA, respectively.
Sensitivity Studies
+stdev LDA
How do the errors in lightning
rates and/or DE model, and
lightning-rainfall relationship
affect the model results?
Standard LDA -stdev LDA
standard LDA
-stdev LDA
Standard LDA +stdev LDA
Sensitivity Studies
Model insensitive to assimilated rainfall rates and
very insensitive to errors in lightning rates.
Squall-line over Hawaii,
28 February 2004
30ºN
20ºN
Hawaii
06 UTC 28 February 2004
Satellite IR
12 UTC 28 February 2004
IR
Another Approach:
Four Dimensional Data Assimilation (FDDA)





Lightning observations are converted to vertical moisture profiles
using lightning - rainfall - moisture profile relationship
Vertical moisture profiles are assimilated using MM5 FDDA
Newtonian nudging (or relaxation) nudges the model state toward the
observations by adding artificial tendency terms to prognostic
equations based on the difference between model- and observed
states.
Disadvantage: Moisture is strongly nonlinear function of temperature need knowledge of temperature field prior to run
Using too high nudging coefficient or too high moisture values may
result in model instability.
Four Dimensional Data Assimilation (FDDA):
Results from the same two cases
Results qualitatively similar than using latent heat nudging,
but needed either high nudging coefficient (instability issues) or
unrealistically high moisture values (x2) to be efficient
Lightning Data Assimilation Using LAPS
• LDA in LAPS is being investigated using
lightning rate - radar reflectivity relationship
(Pessi and Businger 2009, JAMC)
• LAPS can ingest a radar data file (Z in x, y, z
coordinates) that is used to create a 3-D cloud
analysis
• Physically balanced method
Nudging Method - Discussion
• Introduces artificial tendency terms.
• Changes mass and may result in model instability
• How “safe” it is in operations?
Improved nudging method?
• Latent heat nudging method can be “tuned” in several ways
• Scale rain rates downwards as well as upwards
• Remove the latent heating coefficient limit (<3) because in case
of light model rainfall the method does not have much effect
• Still does not solve the instability issue
Summary
• Three years of lightning data from PacNet and LIS were compared to
precipitation and hydrometeor data from TRMM’s Precipitation Radar and
Microwave Imager.
• Convective rainfall rates were well correlated with lightning rates in
climatological time scale.
• Radar reflectivity and echo tops showed logarithmic increase with lightning
rate.
• Some other hydrometeor showed logarithmic increase with lightning rate
as well.
• Lightning data assimilation method nudges model’s vertical latent heating
profiles according to convective rainfall rates derived from lightning data.
• Two cases were analyzed: an extratropical storm over the northeast
Pacific and a squall-line over Hawaii.
• Lightning data assimilation resulted in enhanced advection of warm air into
the storm center and reduced central pressure forecast error.
• Timing of the squall-line passage was improved.