Regional Transport Study of Air Pollutants with Linked Global Tropospheric Chemistry and Regional Ai

1
Regional Transport Study of Air Pollutants with
Linked Global Tropospheric Chemistry and Regional
Air Quality Models
Daewon W. Byun, Nankyoung Moon, Heejin In
Institute for Multidimensional Air Quality
Studies (IMAQS) University of Houston
Daniel Jacob, Rokjin Park
Harvard University
2
Introduction
Some US regional air quality problems may be
originated from long-range transport
processes (eg. Transport of EC/OC/CO/dust from
Sahara biomass burning from Central America)
One of key problems of regional air quality
models is finding accurate initial and boundary
conditions for the simulations.
Distribution of surface air chemistry and PM
monitoring sites is limited both in the spatial
density and in the physical and chemical details.
Current method Run a regional air quality model
at a coarser resolution with seasonal profile
data and use emissions input for a long period
for the spin-up process.
The fixed profile BCs are never accurate and
cannot account for changes due to air pollution
long-range transport events.
It could be different at each side of domain
reflecting certain regional differences.
3
Three Areas of Inter-linkage Issues

• Dynamic Representations in Global and Regional
  Models
• - Chemical Representations in Global and Regional
  Models
• Mechanics of Linkage
• Linkage of scales grid structure and scales of
  data representation (generation of IC/BCs)
• Linkage of chemical species
• Linkage of dynamics

Study Objectives
Provide tools/methods to link regional and global
modeling systems
(e.g.) Set the boundary of the domain that
outside areas do not have much direct emissions
and no high concentration blobs already existing
What is the sensitivity of the simulations to the
different IC/BCs?
4
Mechanics of Linkage

• Linkage of scales Currently, grid structures of
  the global and regional models are not
  consistent
• Requires less preferable horizontal vertical
  interpolation

Implementation Example
GEOS-CHEM (Goddard Earth Observing
System-CHEMisrty)
MODEL3 CMAQ(Multi-pollutant Air Quality model)
LAMBERT CONFORMAL 108 km X 108 km 23
layers in Sigma Po
LAT-LON 2 degree X 2.5 degree 20 layers in
Sigma P
Initial Boundary Condition
IO/API Format in 108 km resolution
Future requires geocentric coordinates (from
a flat-earth to a spherical earth, if not
spheroid)
5
Horizontal distribution of O3 concentration from
GEOS-CHEM global output at Layer 1
For 2000 August episode
108km resolution
2 X 2.5 degree resolution
6
Chemical speciesCurrently, chemical mechanisms
in global and regional models are not consistent
Mechanics of Linkage
MAPPING Table
GEOS-CHEM
O3-NOX-Hydrocarbon chemistry 24 species
CMAQ
CB4 16 species
Un-used species ACET, ALD2
7
Linkage of Chemistry
GEOS-CHEM
Mapping Table
CMAQ
SAPRAC-99
8
Emission comparison
9
VOC Definition
Chemical Mechanism
VOC PAR 2OLE 2ETH 2ALD2 7TOL 8XYL
5ISOP FORM
CB4
VOC C2H6 C3H8  ALK4 PRPE ISOP CH2O
ALD2 RCHO
GEOS-CHEM
The emissions inputs used for the GEOS-CHEM and
CMAQ for the NOx and VOC species were compared.
Table presents how total VOC in mechanisms are
calculated and the values of NOx and VOC for GEIA
represent smaller than NEI99-SMOKE in maximum
values.
10
GEOS-CHEM OUTPUT, Layer-1, Ox
Summer
11
GEOS-CHEM OUTPUT, Top layer in CMAQ, Ox
Summer
12
GEOS-CHEM OUTPUT, Layer-1, Ox
Winter
13
GEOS-CHEM OUTPUT, Top layer in CMAQ, Ox
Winter
14
O3 and SO4 seasonal boundary condition time
series
(Col56,Row114)
(Col1,Row88)
(Col73,Row1)
15
O3 time series at different vertical layers
Western Boundary
Summer
Winter
16
O3 time series at different vertical layers
Southern Boundary
Summer
Winter
17
O3 time series at different vertical layers
Northern Boundary
Summer
Winter
18
SO4 seasonal boundary condition time series
Summer
Winter
19
CMAQ simulation
Emission NEI99 ( SMOKE )
Chemical mechanism CB4 / SAPRC99
MET. DATA MCIP (MM5)
Domain CONUS 36-km
Simulation IC BC with Original profile data IC
BC with GEOS-CHEM output
20
Comparison of simulated O3 concentration with
AIRS
21
Comparison of CMAQ results in different IC and BC
(2000.08.25. 09, 21UTC)
03AM CST
03PM CST
Profile Data Case
GEOS-CHEM Data Case
22
In CMAQ simulations, the results using GEOS-CHEM
output for boundary condition have smaller value
from 16 ppb to 20 ppb than the results using
profile data around western and northeast
boundary area. On the other hand, there is
opposite results at south boundary area, which is
related with positive bios of GEOS-CHEM over the
GULF of Mexico.
It is necessary to investigate the chemical
mechanism differences in CMAQ simulation with
GEOS-CHEM boundary condition .
23
Comparison of O3 production rate
profile BC
GEOS-CHEM BC
CMAQ/CB4
CMAQ/CB4
GEOS-CHEM BC
CMAQ/SAPRC
24
Comparison of wind field
MM5
NASA-GMAO
General patterns of wind fields are well
Some difference shows in circled area. -
CMAQ/MM5 shows parallel to the grid -
GEOS-CHEM/NASA-GMAO shows inflow
This difference can be cause the uncertainty to
regional air quality simulations.
Lets see how big the problem is
25
GEOSCHEM Easterly and northerly MM5 Clock
wise rotation motion
26
MM5
GMAO
27
Study importance of the dynamic consistency
Comparison of the first guess field used in MM5
between ETA and GMAO
EDAS
DAO
28
Comparison of wind fields among three different
MM5 results.
Case 1 MM5 results with EDAS first guess Case 2
MM5 results with ETA first guess and GMAO
objective analysis Case 3 MM5 results with GMAO
first guess
trying to get closer wind fields to GMAO
29
MM5 Results
August 25. 00 UTC
CASE 1
CASE 2
CASE 3
GMAO
30
Comparison of Root Mean Square Error
(RMSE) RMSE is for MM5 result of each case and
GMAO
According to the evaluation result of numerical
models, RMSE was 1.63, 1.57 and 1.41 for wind
speed and 68.37, 66.66 and 69.49 for wind
direction for RAMS, MM5 and Meso-Eta respectively
(Zhong and Fast, 2003). In that evaluation, RMSE
was for observation data and simulation results
for different meteorological model outputs.
31

• Case 2 (ETA first guess and objective analysis
  with GMAO) shows the most closest results to
  GMAO filed in three cases from RMSE analysis.
• Even if MM5 use GMAO data for the first guess in
  case3, MM5 can not simulate closer values to
  initial filed (GMAO) with the lower resolution of
  GMAO in time(6 hourly) and space(2X2.5).
• The best case is use of ETA data of high
  resolution in time and space for the first guess
  and use of objective analysis with GMAO data in
  INTERPF.

32
CMAQ Simulation
Emission EPA-NEI99 Chemical Mechanism
CB4 Meteorology MM5 results of three cases
(Each case has corresponding
case of MM5)

• Comparison of Ozone difference
• CASE2 CASE1
• (ETA_first guess GMAO_objective
  analysis) (ETA_first guess)
• CASE3 CASE1
• (GMAO_first guess) (ETA_first guess)

33
CMAQ Simulation Results Ozone Concentration
Differences MM5 with DAO - MM5 with EDAS for
August 25, 2000
34
Comparison of O3 difference
Case2 Case1
35
Biomass burning due to ENSO-related drought in
Mexico and Central America during April June
1998
TOMS Aerosol Index
36
GEOS2CMAQ Interface
Aerosol species Mapping
GEOS-CHEM
CMAQ
OC
AORGIAORGJAORGPAIAORGPAJ AORGBIAORGBJ
OC_hydrophilic OC_hydrophobic EC_hydrophilic
EC_hydrophobic
EC
AECIAECJ

Coordinate transformation
36 km 36 km Rambert Conformal

2.5 2
Simple Interpolation
23 Sigma vertical layer (Ptop 50 mb)

30 vertical layer (Ptop 10 mb)

37
GEOS-CHEM Global simulation ( 2.5 2 )
GEOS2CMAQ Interface
OC
EC
ICON
BCON
38
EMISSION
DATA SOURCE US EPA NEI 99 Processed with SMOKE
OC
EC
39
Spatial Evolutional Feature of OC
CMAQ ver 4.3 Grids 133 91 23
Resolution 36 km 36 km Science Process
CB4-AERO3- EBI solver Meteorogical data from MM5
ver3.6
40
Simulated Monthly CON Evaluation
by IMPROVE Monitoring
W/ fixed profile BC
EC
OC
W/ GEOS-CHEM BC
Difference
A
B
B
C
41
Daily concentrations of Simulated vs. Observed
OC
Region A
OR
WA
NV
CA
Region B
CO
UT
FL
VT
42
IMPROVE Network
Improving of OC concentration
Region C
TX
AZ
TN
AR
43
Daily Mean CON
A
OC
EC
B
Simulation
R 0.59
G
G
R 0.60
C
R 0.33
P
P
R 0.35
Observation
44
Monthly concentrations of Simulated vs. Observed
EC
OC
G
R 0.75
G
R 0.79
Simulation
P
P
R 0.33
R 0.68
Observation
45
Conclusion
Linkage issues between global tropospheric
chemistry model and regional air quality model
has been studied.
To investigate the effects of using GEOS-CHEM
output as initial and boundary conditions instead
of the profile data on regional simulations, we
have conducted 4 sensitivity CMAQ simulations
with the CB4 and SAPRC99 as the chemical
mechanisms.
We observe significant differences between
profile vs. GEOSCHEM IC/BC.
Global-regional scale linking is the best when
direct emission source is little outside the
regional domain boundary e.g., US-continental
domain.
It is necessary to quantify and minimize the
effects of different dynamics between the global
and regional meteorological data used and to
study the issues of consistency in chemical
mechanisms.