The NLDN.YYYY.MM.ioapi files contain monthly average NLDN-based CG flash rates
(flashes km-2 day-1) for January 2002 through December 2014 on the 12-km CONUS CMAQ
grid.
These files were created using raw files of the form
NflashYYYY.JDAY_daily_v2_lit.raw (1 Jan 2002 to 29 Feb 2008) and
NflashYYYY.JDAY_daily_v3_lit.raw (1 Mar 2008 to 31 Dec 2014) as input.
Where YYYY is 4-digit year and JDAY is 3-digit Julian Day. These raw files are available
from Vaisala.
The V2 files do not contain explicit typing information.
Flashes within the V3 files are classified as cloud-to-ground (CG) or intra/inter cloud (IC).
In order to create a consistent flash rate data set, we used the V3 data for March 1, 2008 to
December 31, 2013 to estimate the likelihood that a given flash was CG given its peak
amplitude (kA).
The table below shows the fraction of V3 flashes during that time period that are typed as CG
for the peak amplitude bins we considered.
kA range
[-∞,-60]
[-59, -50]
[-49, -40]
[-39, -30]
[-29, -25]
[-24, -20]
[-19, -15]
[-14, -10]
[ -9, -5]
[ -4, 0]
[ 1, 15]
[ 16, 20]
[ 21, 25]
[ 26, 35]
[ 36, ∞]
CG-fraction
0.985
0.969
0.953
0.927
0.899
0.867
0.815
0.723
0.582
0.483
0.000
0.264
0.311
0.404
0.631
For each month between January 2002 and December 2013, we summed the number of CG
flashes in each grid box (number of flashes multiplied by the likelihood that each flash was CG)
and estimated the mean CG flash density (flashes km-2 day-1) and obtained adjusted V2 flash
densities.
Comparisons of the adjusted V2 flash densities between the v2 (pre-March 1 2008) and v3
(March 1 2008 on) time periods showed a discontinuity across the v2/v3 time period. The
discontinuity was removed by dividing the flash rates during the v3 time period by 1.16, where
the factor 1.16 ensured that the total number of adjusted V2 flash rates during the V3 period
matches the total number of V3 flashes. Adjusted V2 flash densities for 2002-2013 are
included in the CMAS data set.
We used the same approach to estimate the CG flash densities in 2014. However, this approach
led to a likely non-physical increase in flash density. Therefore, beginning in 2014, monthly
average V3 flash densities as opposed to adjusted V2 flash densities are included in the CMAS
data set.
Note: The gridded data sets do not include a correction for CG detection efficiency. During this
time period, detection efficiencies were believed be ~93% over the continental U.S. For these
locations, detection efficiency can be taken into account by dividing the gridded flash rates by
0.93.
The nldnmonthlyflashrate_usa_YYYY.pptx files contain 4-panel plots of
UL) NLDN-based CG flash rate as determined using above algorithm
UR) NLDN-based CG + IC flash rate
CG flashes are calculated as described above and then adjusted for detection efficiency
using the Vaisala detection efficiency model of Ron Holle.
Monthly OTDLIS climatology is used for grid boxes where the NLDN DE is < 0.2
Monthly OTDLIS climatology is used for Canadian grid boxes
IC flash rates are estimated by multiplying CG flashes by (1 + R) where R is the
climatological IC/CG ratio, which are from Boccippio et al. (2000)
LL) V2.2 OTDLIS climatology
LR) NLDN-based CG flash rate calculated using V3 typing information (only available for Mar
2008 to present; note: This plot is shown for Jan and Feb 2008 but contains all zeros).
Additional information on the lightning-NO algorithm in CMAQ is available in Allen et al.
(2012).
This work was done as part of a NASA Applied Sciences Decision Support System project
under the direction of PI Kenneth Pickering
This file was updated on November 15, 2015.