Comparison of NOX emissions and NO2 concentrations from a regional scale air quality model (CMAQ-DDM/3D) with satellite NO2 retrievals (SCIAMACHY) over the continental U.S.

1
Comparison of NOX emissions and NO2
concentrations from a regional scale air quality
model (CMAQ-DDM/3D) with satellite NO2 retrievals
(SCIAMACHY) over the continental U.S.

• Burcak Kaynak1, Yongtao Hu1, Randall V. Martin2,3
  and Armistead G. Russell1
•  1 School of Civil and Environmental
  Engineering, Georgia Institute of Technology
• 2 Department of Physics and Atmospheric Science,
  Dalhousie University
• 3 Harvard-Smithsonian Center for Astrophysics
• October 7, 2008
• 7th Annual CMAS Conference
• Chapel Hill

2
Overview

• Objective
• Improve the understanding of atmospheric
  chemistry emissions
• Regional air quality models
• Ground-based aircraft observations
• Satellite retrievals
• Why ?
• Improve emission estimates our understanding of
  atmospheric processes
• Understand strengths weaknesses of the
  satellite retrievals, get ideas to improve the
  quality of the retrievals for further use in the
  tropospheric air pollution research
• How ?
• Extensive comparison of observations model
  results
• Model advancements
• Assimilation of the observations within the model
  by an inverse modeling technique

3
SCIAMACHY

• Scanning Imaging Absorption Spectrometer for
  Atmospheric Chartography
• onboard the ENVISAT which was launched in March
  2002 into a sun-synchronous orbit
• global measurement of atmospheric NO2 columns
  through nadir observation of global backscatter
• typical spatial resolution
• 30km x 60km
• global coverage
• over 6 days
• scans through U.S.
• in the mornings
• ( 1030 local time)
• units in tot trop. columns
• (molecules/cm2)
• SCIAMACHY satellite retrievals
• from Martin et al., 2006

Martin, R. V., et al. (2006), J. Geophys.
Res.-Atmos., 111(D15308)
4
Modeling Approach
MM5 meteorology 34 vertical levels,
Four-Dimensional Data Assimilation (FDDA) SMOKE
emissions VISTAS 2002 inventory Emission
projection use growth factors from the EGAS
Version 4.0 CEM CMAQ v4.5 with DDM-3D
concentrations sensitivities SAPRC 99 Chemical
Mechanism 13 vertical layers (up to 15km)
5
Model Simulations

• 3 Simulations
• Base case
• Lightning case
• PAN photolysis case

• domain North America
• resolution 36km x 36km
• episode July-August 2004
• 3 region types selected
• urban 7 cities
• rural 11 rural areas w/o any urban area or
  large scale EGUs
• rural-point 116 large scale EGUs w/o urban
  areas

Kaynak, B., et al. (2008), ACP
6
Lightning NOx emissions

• Lightning case

Base case
Lightning increased NOx emissions around South
East especially in Florida, Mid-West and over the
Atlantic Ocean.
7
PAN Photolysis
CMAQ ICARTT comparison consistent
overestimation and high variability of PAN in
CMAQ CMAQ SCIAMACHY comparison lower NO2
columns from CMAQ, especially rural regions
even after lightning emissions PAN Photolysis
included in CMAQ Resulted minor improvement in
CMAQ ICARTT PAN comparison with similar
vertical profile (improvement up to 5 MNE and
MNB for individual flights) No significant
change obtained in CMAQ SCIAMACHY NO2 comparison
Altitude (km)
8
CMAQ vs. SCIDomain-wide

• CMAQ
• higher simulated levels in urban areas
• lower in the surrounding areas
• Possible reasons
• the pixel size of SCIAMACHY having a smoothing
  effect
• chemistry or transport problems in the model,
  e.g. NO2 oxidizing faster than actual.
• SCIAMACHY
• consistently higher around LA
• higher from NY to ocean
• Lightning reduced some discrepancy in mid-east,
  south,
• but put too much NO2 around Toronto-Illinois in
  Aug04

9
CMAQ vs. SCI Domain-wide

• West high correlation, low slope East low
  correlation, higher slope
• R20.58-0.62, Slope 0.74-0.80 R20.35-0.41,
  Slope 0.87-1.02
• CA has the highest correlation according to both
  months (R2 gt 0.70).
• WA and GA also have good correlation.

10
CMAQ vs. SCI State-wide
CMAQ lower in west (CA, NV, AZ, UT, NM, CO, WY)
in a few northeastern states (ME, NH)

• State averages for July August 2004
• inconsistencies between two months (OR, ID, MT)
• OR possible overestimation of the fire emissions
  for July 2004

11
CMAQ vs. SCI Land type

• Urban
• SCIAMACHY
• Los Angeles is high
• Houston, Chicago
• Phoenix is low.
• Rural
• SCIAMACHY
• NV, WA very high
• ID, OR low
• (similar to emissions)
• Rural-Point
• have some outliers,
• But overall correlation
• is good.

12
Land type

• NO2, scia/NO2, cmaq
• (averaged for 2 months)
• Red SCIAMACHY is higher
• (Los Angeles, NV, WA)
• Green CMAQ is higher
• (ID, OR, Houston)
• Yellow comparable

13
ICARTT Intex-NA
Jul 04
Eastern, North-eastern U.S.
Aug 04
14
Ground Observations

• No negative bias in Los Angeles (21 AIRS
  stations), on the contrary CMAQ has a positive
  bias indicating overestimation.
• Atlanta is overestimated which is not observed in
  satellite.
• Houston, Chicago Phoenix are overestimated,
  similar to satellite.

15
Summary

• Lightning emissions resulted in minor
  improvements for some regions, but overall
  correlation did not improve.
• CMAQ usually is higher than the SCIAMACHY
  observations in urban centers, but lower in
  surrounding areas. Possible reasons
• the pixel size of SCIAMACHY having a smoothing
  effect,
• diagnostic biases in the SCIAMACHY retrieval
  analyses,
• biases in the emissions estimates,
• chemistry, transport problems in the model.
• Western U.S. has lower NO2 from the model, but
  high correlation.
• Eastern U.S. has comparable NO2 levels, but
  correlation is lower.
• On a state-by-state basis, most western states
  and a few eastern states have simulated NO2
  columns lower than observed.

16
Summary

• NO2 total columns from satellite correlate well
  with simulated NO2 for rural regions but less
  so with urban rural-point (even though
  power plant emissions are well known)
• Los Angeles is the major outlier between
  simulated and observed abundances in urban
  regions. This may indicate
• a retrieval/analysis error,
• a bias in emission estimates specific to that
  region (or, conversely biases in the other
  regions),
• modeling issues specific to that area.
• The potential reasons for lower correlation of
  rural-point could be the
• transport of NO2 out of the small scan area
  probably minor-
• insufficient time for conversion of NO to NO2 in
  power point plumes.
• High correlation of rural regions is helpful
  for using the satellite retrievals to obtain
  emission estimates for area sources that low in
  amounts and are sparse which is hard to capture
  otherwise.

17
Summary

• Specifically rural areas in NV, WA may have
  more and ID, OR may have less emissions than
  inventories show.
• Emission estimates from uncertain sources like
  lightning and fire can be improved using
  satellite retrievals.
• Using satellite observations is still
  problematic but comparisons are promising even
  though the uncertainties are high, using
  satellite retrievals for data assimilation can
  give more insightful information and quantitative
  results for improving emission inventories of
  some states which showed significant
  discrepancies from the satellite retrievals. More
  studies like this and with other models,
  inter-method measurement comparisons are needed.

18
Future Work
Inverse modeling using NO2 columns w/ FDDA
19
Acknowledgements

• NASA Project SV6-76007 (NNG04GE15G), and EPA
  grants (RD83096001, RD83107601 and RD83215901)
• Russell Group, Georgia Institute of Technology
• Randall Martin for SCIAMACHY NO2 retrievals
• Global Hydrology Resource Center (GHRC) for
  providing the NLDN flash data
• Kenneth Pickering for suggestions on vertical
  allocation of lightning NOx
• Bill Carter for suggestions for PAN photolysis

Thanks for your time.