Impact of lightning-NO on eastern United States
photochemistry during the summer of 2006 as
determined using the CMAQ model
Dale Allen
Dept. of Atmos and Oceanic Sci, UMD-College Park
Kenneth Pickering
Atmos Chem and Dyn Branch, NASA-GSFC
Robert Pinder and Thomas Pierce
Atmos Modeling and Analysis Div, U.S. EPA
Barron Henderson
Dept. of Env Sci and Eng, UNC Chapel Hill
William Koshak
Earth Science Office, NASA-MSFC
2011 CMAS Meeting
26 October 2011
Motivation for Including Lightning NOx in CMAQ
• In the summer over the US, production of NO by lightning (LNOx) is
responsible for 60-80% of upper tropospheric (UT) NOx and 20-30% of
UT ozone (Zhang et al., 2003; Allen et al., 2010).
• Mid- and upper-tropospheric ozone production rates are highly
sensitive to NOx mixing ratios.
• Inversion-based estimates of NO emissions from CMAQ simulations
w/o LNOx have large errors at rural locations (Napelenok et al., 2008).
• CMAQ-calculated N deposition is much too low without LNOx (e.g.,
Low-bias in CMAQ nitric acid wet deposition at NADP sites cut in half
when LNOx was added).
• LNOx can add several ppbv to summertime surface O3 concentrations
• CMAQ NO2 amounts are too low (high) at rural (urban) locations
(Castellanos et al., 2011; Huijnen et al., 2010) suggesting that the
lifetime of NO2 (Henderson et al., 2011) and/or the transport of NOx
(Gilliland et al., 2008) is underestimated by regional models. How will
LNOx affect these biases?
LNOx Production
•
Lightning-NO production is assumed to be proportional to convective
precip rate multiplied by a scaling factor chosen so that monthly avg
model flash rates match monthly avg NLDN-based total flash rates.
•
NLDN-based total flash rate is est by multiplying NLDN CG flash rate
by Z+1, where Z is the climatological IC/CG ratio (Boccippio et al.,
2001) determined by taking the ratio between satellite-retrieved
(Optical Transient Detector) total flash rates and NLDN CG flash rates.
•
IC and CG flashes are assumed to produce 500 moles of N per flash, a
value that is consistent with cloud resolved modeling of observed
convective events [DeCaria et al., 2005; Ott et al., 2009] & with largerscale modeling of INTEX-A [Hudman et al., 2007; Allen et al., 2010].
•
Vertical dist of emissions is assumed to be proportional to the
pressure convoluted by the segment altitude distribution of flashes in
the vicinity of the North Alabama LMA (Koshak et al., 2010)
CMAQ simulation of summer 2006
• Simulations of 2006 air quality performed at EPA under the
management of Wyat Appel and Shawn Roselle as part of the Air
Quality Model Evaluation International Initiative (AQMEII)
http://aqmeii.jrc.ec.europa.eu/doc/AQMEII_activity_description.pdf
• Version 4.7.1 of CMAQ used with CB-05 chemical mechanism
• NEI-based emissions with year specific power plant emissions from
CEMS and satellite-derived wildfire emissions
• Chemical boundary conditions from GEMS (European-led
assimilation effort)
• http://ozone.meteo.be/meteo/view/en/1550484-GEMS.html
OMI tropospheric NO2 products
1.DP-GC product [Lamsal et al., 2010]
2. v2.0 DOMINO product [Boersma et al., 2007; Boersma et al., 2011]
DP-GC and DOMINO products begin with same slant column & use
same method to remove stratospheric column.
Different methods used to convert tropospheric slant cols to overhead
cols
 Yield different tropospheric vertical column amounts
tropopause
R=0.47
R=0.79
Sensitivity of urban/rural ratio to hor/vert smoothing
1. U/R ratios ↓
when LNOx
is added
as LNOx
is larger part
of rural col
than urban
column
2. U/R ratios ↓
when mapped
onto 0.25x0.25
grid because
aggregation
mixes rural and
urban locations
3. U/R ratios ↓
when avgk is
applied because
U/R ratios are
largest near
surface where
OMI is
relatively
insensitive
4. Relative to OMI, CMAQ has high bias at urban sites.
Biases at rural sites are relatively small after accounting for LNOx
and the smoothing inherent in DOMINO averaging kernels.
How much do uncertainties in CB05 chemistry contribute to CMAQ’s
inability to capture the high amounts of UT NOx measured during the
INTEX-A period (An upper bound)?
• CMAQv4.7.1 with CB-05 chemistry and AERO5 aerosols
was used to simulate the summer of 2004.
• Three simulations: 1) standard chemistry without
lightning-NO, 2) standard chemistry with lightning-NO,
and 3) updated chemistry with lightning-NO
• Updated chemistry: Organic nitrate (ON) yield from the
oxidation of paraffins (PAR) was reduced from 15% to
3%. The decrease in ON production reduces NO
consumption, increases the NOx lifetime, and is in better
agreement with observations (Henderson et al., 2011).
LNOx
increases
UT NOx by
75-100 pptv
(X4 increase);
Adjusting chemistry increases UT NOx by
20-30 pptv reducing model biases by 5-16%
if data are unbiased and by 10-33% if data
are assumed to be 30% too high due to
MPN interference (Browne et al., 2011)
however, a
120-500 pptv
low-bias remains.
Low-bias is still
60-300 pptv
after accounting for
MPN interference
Mean summer 2006 enhancement of
8-hr maxO3 in CMAQ due to LNOx
O3 enhancement (ppbv)
(nitrate)
Adding LNOx
eliminates
low-bias
Adjusting
for precip
bias leads
to better fit
Eastern US: Longitudes east of 100W
Nitrate
Biases
NE: +5%
SE: +19%
West –(5-10)%
Ammonium
biases
NE: +20%
SE: +40%
West: -(20-25)%
Summary
•
•
•
•
•
•
•
For a 500 mole per flash lightning-NO source, mean tropospheric NO2
columns agree with satellite-retrieved columns to within -5 to +13%.
Contribution of LNOx to mean model column is ~25%, ranging from
~10% in the northern states to >45% along the Gulf of Mexico and in
the southwestern states.
CMAQ columns have a high-bias wrt DOMINO columns over urban
areas. Biases at other locations were minor after accounting for the
impacts of lightning-NO emissions and the averaging kernel on model
columns.
Chemistry explains less than 1/3 of upper tropospheric NO2
underpredictions by CMAQ during the INTEX-A period
UT O3 is biased high wrt eastern U.S. sonde data. While LNOx
contributes to bias, most of it is likely due to the specification of BC
and noise introduced by vertical velocity calculation within CMAQ.
LNOx increases wet dep of nitrate by 43%, total dep of N by 10%, &
changes 30% low-bias wrt NADP measurements to 2% high-bias.
On poor AQ days (O3>60 ppbv), LNOx contributes >6.5 ppbv to 8hrO3
at 10% of western sites and 3% of eastern sites
Acknowledgements
Wyat Appel & Shawn Roselle of EPA: AQMEII simulations
Ana Prados of UMBC: Gridding OMI std product
Anne Thompson: IONS ozonesonde data,
L. Lamsal: DP-GC NO2 data.
OTD/LIS data are from NASA/MSFC.
NLDN data are collected by Vaisala Inc
NASA Applied Science Air Quality Program
Adding LNOx
eliminates
low-bias
Adjusting for
precip bias
lessens scatter
but increases
bias.
Processing of DOMINO & CMAQ fields
•
•
•
•
•
•
Gridded DOMINO fields created by mapping version 2.0 level 2 DOMINO
fields onto 0.25°x0.25° grid.
DOMINO retrievals over snow/ice or with cloud radiance fractions > 50%
filtered out (Boersma et al., 2009)
Mean value in each grid box obtained using algorithm that gives more
weight to near-nadir pixels and to pixels with low geometric cloud fractions
(Celarier and Retscher, 2009).
CMAQ profiles extracted at location of high-quality DOMINO pixels &
weighted in same manner. CMAQ output interpolated onto TM4 vertical grid
(TM4 model used to obtain a priori profiles for DOMINO product) .
When appropriate, averaging kernel is applied to tropospheric model subcolumns before weighting is performed (Allen et al., 2010; Boersma et al.,
2009).
CMAQ tropospheric NO2 column determined by summing sub-columns
within the troposphere, where the number of tropospheric layers is included
in DOMINO data product.
CMAQ Lightning-NO Parameterization
k:
PROD:
LF:
Precon:
threshold:
G:
αi,j:
LNOx = k* PROD*LF, where
Conversion factor (Molecular weight of N / Avogadros #)
Moles of NO produced per flash
Total flash rate (IC + CG), where
LF = G * αi,j * (preconi,j – threshold), where
Convective precipitation rate from WRF
Value of precon below which the flash rate is set to zero.
Scaling factor chosen so that domain-avg WRF flash rate
matches domain averaged observed flash rate.
Local scaling factor chosen so that monthly avg modelcalc flash rate for each grid box equals local observed
flash rate
For these retrospective simulations, the observed flash rate is the NLDNbased total flash rate for June, July, and August 2006.
Operational forecasts could use satellite-retrieved or NLDN-based
climatological flash rates for a season as observations.
Vertical partitioning of lightning-NO emissions
Vertical distribution of
flash channel length
in the vicinity of the
North Alabama LMA
is used along with a
direct relationship with
pressure to determine
the fraction of
emissions to put into
each layer from the
surface to the CMAQpredicted cloud top
Segment altitude distribution for
all flashes from Koshak et al. [2010]
Comparison of CMAQ & OMI tropospheric column O3
Trpps =
150 hPa
OMI Level-2 daily ozone profile data courtesy of X. Liu
Bias: -1.6 DU
(nitrate)
Note:
Adding LNOx causes
Wetdep(OxN) 50%↑
Totdep(OxN) 22%↑
Totdep(N) 11%↑
Adding LNOx
NE US: -14%  +11%
SE US: -19%  +28%
MW/GP US: -32%  +2%
RM/W US: -31%  -2%
SE bias is reduced to
-3% if adjustment is made
for a high-bias in SE US
CMAQ precip
Lightning-NO
increases UT
ozone by 5-7 ppbv
Adjusting chemistry
increases UT O3
by 2-2.5 ppbv.
Changes small in
lower troposphere
Low-bias
is reduced
with LNOx
However
many
factors
contribute
to UT
biases in
ozone
In general, the
contribution of
LNOx to 8hrO3
decreases on
bad AQ days
over the
eastern U.S.
Outline
• Motivation & Background
• Describe method used to parameterize lightning-NO
within CMAQ
• Show impact of lightning-NO on tropospheric
composition, air quality, and nitrogen deposition
over the U.S. during the summer of 2006
• Use OMI NO2 fields to investigate the cause of biases
between modeled and “observed” NO2 mixing ratios
at urban and rural locations
• Investigate the impact of uncertainties in chemistry
on upper tropospheric NOx distributions in the
context of the INTEX-A mission