Transcript Thailand_1 - Georgia Institute of Technology

Atmospheric Modeling and its
Application to Energy and the
Environment: From Local
Impacts to Climate Change
Amit Marmur, …, K. Manomaiphiboon , …, and
Armistead (Ted) Russell
Georgia Power Professor
Georgia Institute of Technology
With Special Thanks to:
• NIEHS, US EPA, FHWA, Southern Company, SAMI
– Financial assistance
• JGSEE of Thailand
• And more…
Issues
• Energy sources contribute to local, regional and global air
pollution problems
– Local: CO, Particulate matter, air toxics
– Regional: Particulate matter, acid deposition, ozone
– Global: CO2 and climate change
• Primary and secondary pollutants impact health
– Approximately 799,000 premature deaths yearly
– Mainly due to particulate matter and ozone
• Emissions from energy sources undergo complex
atmospheric processing
– Turbulent atmospheric transport
– Non-linear chemical reactions (e.g., produce ozone)
– Deposition
• Feedbacks between meteorology and air pollution
– Climate change
Ozone Formation
Ozone Isopleth
h (sunlight)
ENOx
NOx
oxides of nitrogen
(NO + NO2)
High O3
O3
Low O3
EVOC
VOCs
Volatile organic compounds
PM Formation
h (sunlight)
NOx
PM
VOCs,
OC & EC
SO2
Sulfur dioxide
Particulate Matter
• Complex mixture of solid and
liquid particles suspended in
the ambient air
• Size classifications
–
–
–
–
“super-coarse”
“coarse” (PM10)
“fine” (PM2.5)
“ultrafine”
> 10μm
< 10μm
< 2.5μm
< 0.1μm
• Many sources
• Many chemical species:
BRIG, New Jersey (m easured)
BRIG, New Jersey (m odeled)
BRIG, New Jersey (m easured)
Sulfate
Nitrate
Ammonium
Organic Carbon
Elemental Carbon
Soils and crustals
SHRO, North Carolina (m easured)
SHRO, North Carolina (m odeled)
• Air quality modeling
Outline
– Basics
– Approaches
• Advanced approaches
• Applications
– Source impacts
– Climate impacts on air quality
Emissions-based Air Quality Model
• Representation of physical and
chemical processes
– Numerical integration
routines
• Scientifically most sound
method to link future emissions
changes to air quality
Chemistry
Emissions
Computational
Planes
5-20
50-100
50-200
ci
 ( uci )  ( Kci )  Ri  S i
t
Atmospheric Diffusion Equation
Numerics
C=AxB+E
Meteorology
Air Quality
Model
Discretize
c
 Lx,tc  fx,t
t
Operator splitting
200 species x 10000 hor. grids x 20 layers= 40 million
ct2t  Lxt Ly t Lcz2t Lyt Lxt ct
coupled, stiff non-linear differential equations
Air Quality Model
Chemistry and
Aerosol Dynamics
Heterogeneous
Processes
Photochemical
Reactions
Surface
Deposition
Homogeneous
Processes
Aerosol
Dynamics
Sink Processes
Emissions
1. Anthropogenic
2. Geogenic
3. Biogenic
Transported
Pollutants
Air Quality
Sources
(E,S, BCs)
Numerical
Solution
Techniques
Transport
(U, K, Vd)
Radiation
Topography &
Land use
Geographical
Features
Chemical
Processes
(R)
Sources
Meteorology
Temperature
Thermochemical
Reactions
Turbulence
Cloud
Cover
Wind
Computed
Concentrations
Atmospheric Modeling Process
Chemical
Mechanism
Specification
Emissions
Model
Inputs:
Emissions Inventory
Population
Roads
Land Use
Industry
Meteorology
Chemical Mechanism
Historical: Specified
Evolving: Compiler
Emissions
Inputs
Historical:
NO, NO2, HONO
Lumped VOCs
CO, SO2
Evolving:
PM, NH3,
Detailed VOCs,
Adv. Biogenics
Model
Parameter
Calculation
Air Quality
Model
Numerical
Routines
Historical:
Advection
Chem. Kinet.
Evolving
Sens. Anal.
Proc. Integ.
Unc. Anal.
Pollutant
Distributions
Evolving:
Sensitivities
Uncertainties
Model
Evaluation
Meteorological Inputs
Temperature,
Solar Insolation
Emissions
Inventory
Development
Emissions, Industry
and
Human Activity Data
Historical
2- or 3-D winds; Ground level T, RH;
Mixing height, Land use
Evolving:
3-D Winds, Diffusivities, Temp., RH,
D, Solar Insolation (UV & total solar)...
Meteorological Model
(Diagnostic or Prognostic)
Topographical
Data
Meteorological
Observations
Foundation
Air Quality Data Analysis and Processing
Air Quality
Observations
Grids
Adaptive
(Odman et al.)
Nested
Multiscale
(Odman et al.)
About 15
vertical
Layers up to
15 km
(many in first 1 km)
How well do they work?*
a ) O bs e rv e d
A v e r a ge P M
2.5
c o n c e n t r a t io n
2 8 .4  g/m
b) S im u la t e d
A v e r a ge P M
12%
16%
2.5
c o n c e n t r a t io n
3 1 .6  g/m
3
3
11%
21%
3%
3%
A m m o n iu m
N it r a t e
S u lf a t e
EC
28%
OC
34%
O t h er
37%
28%
4%
3%
AIRS Station 47-099-0101; Look Rock, Blount Co, TN
(high elevation)
120
100
80
60
40
20
0
Ozone (ppb)
Ozone (ppb)
AIRS Station 47-037-0011; Nashville, Davidson Co, TN (urban)
0
24
48
72
96
120
144
Time (starting July 11, 1995)
168
192
216
120
100
80
60
40
20
0
0
24
48
72
96
120
144
168
192
216
Time (starting July 11, 1995)
*Performance relies on quality of inputs. US has spent decades on
emissions inventory development. Meteorological modeling also contributes
significantly to errors
What’s next?
• Emissions-based air quality models work pretty well, how might we
use them:
– Identify, quantitatively, how specific sources impact air quality.
– Develop and test control strategies
• Decoupled direct method (implemented in CIT, URM, MAQSIP,
CMAQ, CAMX)
– Dunker: initial applications
– Yang et al.: large scale application, comp. efficient (CIT, URM)
– Hakami et al. ,Cohan et al: Higher order, with applications (MAQSIP,
CMAQ)
– Napelenok et al., : PM
• Control strategy assessment
– Least cost approach to attainment for Macon, GA (Cohan et al.)
• Assessing impacts of individual sources
• Climate impacts on air quality
Example Results : Impact of Planned
Controls: 2000 vs. 2007
Emissions reductions lead to about a 12 ppb ozone reduction:
Atlanta and Macon do not attain ozone standard (Macon by 6ppb)
Sensitivity analysis
• Given a system, find how the state
(concentrations) responds to incremental
changes in the input and model parameters:
Model
Parameters
(P)
Inputs
(P)
Model
State Variables:
 C x, t
Sensitivity
Parameters:

 Ci 
 S ij  x, t 

 Pj 

If Pj are emissions, Sij are the sensitivities/responses to emission
changes, e.g.., the sensitivity of ozone to Atlanta NOx emissions
Sensitivity Analysis with
Decoupled Direct Method (DDM):
The Power of the Derivative
(1)
• Define first order sensitivities as Sij  Ci / E j
• Take derivatives of
• Solve sensitivity equations simultaneously
Advection
Diffusion
Chemistry
Ci
  (uCi )  (KCi )  Ri
t
S ij
  (uS ij )  (KS ij )  JS j
t
Emissions
 Ei
  ij Ei
DDM-3D
NOo
NO2o
VOCio
...
T
K
u, v, w
Ei
ki
BCi
...
J
Ri
k j
3-D
Air
Quality
Model
O3(t,x,y,z)
NO(t,x,y,z)
NO2(t,x,y,z)
VOCi(t,x,y,z)
...
decoupled
DDM-3D
Sensitivity
Analysis
s ij (t) 
c i (t)
p j
DDM compared to Brute Force
CA
Ci
Sij 
 j
C
C A  CB
S ij 
 A  b
Sulfate
CB
E
EB
EA
Emissions of SO2
Consistency of first-order sensitivities
Brute Force (20% change)
R2 > 0.99
Low bias & error
DDM-3D
Advantages of DDM-3D
• Computes sensitivities of all modeled species
to many different parameters in one
simulation
– Can “tell” model to give sensitivities to 10s of parameters in
the same run
• Captures small perturbations in input
parameters
– Strangely wonderful
• Avoids numerical errors sometimes present
in sensitivities calculated with Brute Force
• Lowers the requirement for computational
resources
Evidence of Numerical
Errors in BF
NH4 sensitivity
to domain-wide
SO2 reductions
NOx reductions
at a point
Efficiency of DDM-3D
Complication: Nonlinearity
• Often, only a handful of
sensitivities are modeled
(e.g., 30% NOx reduction)
• Assumptions of scaling
and additivity not
necessarily accurate
• But it may be impractical to
model all combinations
ENOx
Ozone Isopleth
High O3
Low O3
EVOC
Calculation of higher-order derivatives:
If taking the derivative once is good, twice
must be better
High-order Decoupled Direct Method [HDDM, (Hakami et al., 2003)]:
Sij(1)
t
Sij( 2 )
t
Sij( m )
t
 (uSij(1) )  (KSij(1) )  Ei  J i S (j1)
 (uSij( 2 ) )  (KSij( 2 ) )  J i S (j2 )  J *S ( 1) S (j1)
 (uS
( m)
ij
)  (KS
(m)
ij
)  J iS
 m  1  * (m k)
J S (k) S j
  
k 1  k 
m 1
(m)
j
(n.b.: > third order derivatives numerically sensitive)
Brute Force and HDDM-3D
Ozone
S (1) 
CA
CB
B
+
C

A
+
S ( 2)
∆C
a1
S
EB
-EA
EA
(1)
 2C

 2
CB  C A


EVOC
Control Strategy Development
• Macon out of attainment by
6 ppb in 2007
• Want to identify least cost
control strategy
• Process:
– Identify possible controls and
costs ($/ton of VOC or NOx)
– Simulate response to controls
([O3]/ton VOC or NOx)
– Calculate control
effectiveness([O3]/$)
– Choose most effective controls
until get 6 ppb
– Test strategy
Sources of
Macon’s ozone
A
B
S
M
8-hr ozone, Aug. 17, 2000
(2007 emissions)
Macon
Scherer
Atlanta
Branch
Sensitivity of 8-hr Ozone in Macon
0.030
Atlanta NOX
0.025
Branch
Scherer
Macon Non-mobile
0.020
Macon Mobile
0.015
0.010
0.005
0.000
-0.005
12-Aug
13-Aug
14-Aug
15-Aug
16-Aug
17-Aug
18-Aug
19-Aug
2007 Emissions
and Sensitivities
NOx emission rates
(tpd)
Year 2007 NOx Emissions (tpd)
750
600
450
Area
Point
Non-Road
On-Road
300
150
0
Plant
Rest of "Macon Atlanta
Plant
Scherer Macon Buffer" (20 cnty) Branch
Rest of
GA
Plant
Rest of
Scherer Macon
Rest of
GA
Macon ozone sensitivity
(ppt/tpd)
S(1) Macon 8-hr ozone (ppt/tpd)
300
250
200
150
100
50
0
Macon Atlanta
Plant
Buffer (20 cnty) Branch
Marginal Abatement Costs by Region
Marginal cost per ton (Year 2000$)
$60,000
NOx
$50,000
VOC
$40,000
$30,000
$20,000
Cost-optimization
$10,000
$0
0%
10%
20%
30%
Percentage emission reduction in Bibb County
Source-Receptor Response
40%
Choose options with least
marginal $/impact until:
(1) attain a.q. goal, or
(2) reach budget constraint
Annual Cost (Year 2000$, in millions)
Strategies for Macon attainment (need 6.5 ppb)
Key Measures
• Zero-cost options
(PRB coal, burning
ban, ...): 1.72 ppb, $0
$200
$180
Macon only
$160
All Georgia
$140
• Bibb industrial NOx:
0.82 ppb, $2.6 million
$120
$100
• Locomotive controls:
0.77 ppb, $7.3 million
$80
$60
$40
$20
$0
0
2
4
6
8
10
Reduction in 8-hour ozone near Macon monitor (ppb)
• SCRs at Scherer:
1.63 ppb, $20.9 million
• Vehicle I&M in Bibb:
0.25 ppb, $4.9 million
Single-Source Impact Analysis (Bergin et al.)
Provide a technique to evaluate the
impacts from a single large emissions
source on regional air quality, incorporating
non-linear processes and multi-day effects
in estimating pollutant responses to
relatively small emissions perturbations.
Motivation and Application
• The ability to evaluate regional secondary pollution impacts
from large single sources would provide a valuable tool for
more effective air quality management practices, such as
refining programs (e.g. emissions trading, regional
planning), and supporting more effective compliance
enforcement.
• Typical modeling approach (removing the emissions from a
single source) has numerical errors.
• Court case led to need to assess impact of a single power
plant (Sammis) in Ohio on downwind areas (a distance of
up to about 1000 km)
Average Day Elevated NOx Emissions
NOx Emissions (avg tons/day)
2500
2000
1500
1000
excess
500
allowable
0
May-95
Jul-95
Aug-00
W. H. Sammis Power Plant
Court
Estimated
from W.H.
Sammis Plant
(court
estimated
emissions)
Ohio
Elevated EGU
Jul-95
Model Inventory
Approach
Two air quality models and grids, three ozone episodes, and three sensitivity
techniques (brute-force, DDM, higher order DDM)
CMAQ, 36x36 km
Aug. 12-20, 2000
2-ord. DDM
URM, multiscale from 24x24 km2
July 11-19 and May 24-29, 1995
DDM
Maximum increase in 1-hr avg O3
Comparison of the maximum increase in hourly-averaged ozone concentrations due
to excess NOx emissions from the Sammis plant.
URM with DDM
(a) July 11-19, 1995
CMAQ with
2nd order DDM
(b) May 24-29, 1995 (c) August 12-20, 2000
1-hr O3 cell responses to excess emissions
All hours
maximum = 2.2
When O3 > 0.060 ppm
maximum = 2.2
Max.
increases
minimum = -3.6
Max.
decreases
CMAQ, 2nd ord DDM, August
minimum = -1.2
Conclusions
• Single-source simulation results agree with
past field experiments, indicating that
appropriate modeling techniques are
available for quantifying single-source
regional air quality impacts.
Climate Change Impacts on Air Quality
• Climate change is forecast to affect air temperature,
absolute humidity, precipitation frequency, etc.
• Increases in ground-level ozone concentrations are
expected in the future due to higher temperatures and
more frequent stagnation events.
• Ozone-related health effects are also anticipated to be more
significant.
• Both ozone and PM2.5 (particulate matter with
aerodynamic diameter less than 2.5 micron meters) are also
found to impact climate via direct and indirect effects on
radiative forcing.
http://www.nature.com/news/2004/040913/images/climate.jpg
Potential Climate Changes in 2050
Temperature (oC)
oC
- IPCC SRES, A1B scenario using GISS
Absolute Humidity (%)
Issues
How will climate change affect air quality
with non-projected and projected emissions?
 How well currently planned control
strategies will work if climate changes in the
future ?

Above questions can be answered by quantifying
sensitivities of air pollutants (e.g., ozone and PM2.5)
to their precursors (e.g., NOx, NH3, VOCs and SO2)
and associated uncertainties.
Modeling Procedure
Leung and Gustafson (2005)
NASA GISS
IPCC A1B
With 2050 climate
MM5
With 2001 & 2050
climate
MCIP
SMOKE
(w/ 2001 EI)
SMOKE
(w/ 2050 EI)
CMAQ-DDM
*Leung and Gustafson (2005), Geophys. Res. Lett., 32, L16711
Global and Regional Climate Models*
GISS GCM: grid spacing = 4º x 5º
9 levels
output every 6 hours
MM5 Domain 1: dx = 108 km
67x109 points
output hourly
MM5 Domain 2: dx = 36 km
115x169 points
output hourly
*Leung and Gustafson (2005), Geophys. Res. Lett., 32, L16711
Air Quality Simulation Domain
 147 x 111 grid cells
 36-km by 36-km grid
size
 9 vertical layers
 U.S. regions:
• West (ws)
• Plains (pl)
• Midwest (mw)
• Northeast (ne)
• Southeast (se)
•Also investigating
Mexico and Canada
Emission Inventory Projection
•Accurate projection of emissions key to comparing relative
impacts on future air quality
Step 1. Use latest projection data available for the near future
- Use EPA CAIR Modeling EI (Point/Area/Nonroad, from 2001
to 2020)
- Use RPO SIP Modeling EI (Mobile, from 2002 to 2018)
Step 2. Get growth data for the distant future
- Use IMAGE model (IPCC SRES, A1B)
- From 2020(2018 for mobile activity) to 2050
- Use SMOKE/Mobile6 for Mobile source control
Woo et. al, 2006
Emission Inventory Projection
Calculate combined factor of
growth/control from EPA base
year(2001) vs. future year
(2020) emissions inventory
Calculate growth
factor for Y2020Y2050 (A1B) from
IMAGE
Calculate growth factor
for Y2001-Y2050 for
Canada/Mexico from
IMAGE
RPO 2018
Activity data
(On-road
mobile)
Use EPA 2020 CAIR-case inventory
Compare EPA CAIR vs.
IMAGE for Y2001-Y2020
Develop SCC to IMAGE
fuel/sector x-reference
Update cross-references
Check/apply growth factors to
2020 EPA CAIR EI to get 2050 EI
SMOKE/M6ready activity
data for 2050
Regional Emissions
Year 2001
Year 2020
25
Millions TPY
20
15
10
5
0
2001
Year 2050
Pnt
2020
Area
Nonroad
2050
Onroad
Present and future years NOx emissions by state and by source types
Emission Changes
6.0E+04
SO2
NOx
8.0E+04
5.0E+04
tons/day
tons/day
2001
4.0E+04
3.0E+04
2.0E+04
2050_np
2050
4.0E+04
2.0E+04
1.0E+04
0.0E+00
1.5E+04
6.0E+04
ws
pl
mw
ne
se
0.0E+00
1.5E+05
us
ws
pl
ne
se
us
ne
se
us
VOCs
NH3
1.0E+04
tons/day
tons/day
mw
5.0E+03
0.0E+00
1.0E+05
5.0E+04
0.0E+00
ws
pl
mw
ne
se
us
ws
pl
mw
Summary of Air Quality Simulations
Scenario
Emission
Inventory (E.I.)
Climatic Conditions
Future Air Quality
Impacting Factors
2001
Historic (2001)
Historic
(2001 whole year)
N.A.
2000-2002 summers
Historic (2000-2002)
Historic
(2000-2002 summers)
N.A.
2050_np (non-projected
emissions, but
meteorologically
influenced for
consistency)
Historic (2001)
Future
(2050 whole year)
Potential future climate
changes
2049-2051_np summers
Historic (2001)
Future
(2049-2051 summers)
Potential future climate
changes
2050
Future (2050)
Future
(2050 whole year)
Potential future climate
changes & projected E.I.
2049-2051 summers
Future (2049-2051)
Future
(2049-2051 summers)
Potential future climate
changes & projected E.I.
Emission Changes
6.0E+04
SO2
NOx
8.0E+04
5.0E+04
tons/day
tons/day
2001
4.0E+04
3.0E+04
2.0E+04
2050_np
2050
4.0E+04
2.0E+04
1.0E+04
0.0E+00
1.5E+04
6.0E+04
ws
pl
mw
ne
se
0.0E+00
1.5E+05
us
ws
pl
ne
se
us
ne
se
us
VOCs
NH3
1.0E+04
tons/day
tons/day
mw
5.0E+03
0.0E+00
1.0E+05
5.0E+04
0.0E+00
ws
pl
mw
ne
se
us
ws
pl
mw
Impact of Future Climate Change on
Ground-level Ozone and PM2.5
Concentrations
Daily maximum 8 hour ozone concentration CDF plots in 2001, 2050 and 2050_np
2001
2050
2050_np
CDF
Southeast
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
NOx limitation sharpening “S”,
reducing peak
Small increase in O3 due to climate
Substantial decrease in O3 due to climate
Reduced NOx scavenging
0
20
40
60
80
100
120
M8hO3 (ppb)
2001
2050
2050_np
CDF
US
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Peaks (ppb)
2001: 141 (actual= 146)
2050_NP: 152
2050: 120
0
20
40
60
M8hO3 (ppb)
80
100
120
Summer Average Max 8hr O3
O3_2000-2002summers
O3_FutureSummers - O3_HistoricSummers
O3_2049-2051summers
O3_FutureSummers - O3_FutureSummers_np
np: Emission Inventory 2001, Climate 2050
PM2.5_2001
PM2.5_2050 - PM2.5_2001
Annual PM2.5
PM2.5_2050
PM2.5_2050 - PM2.5_2050np
np: Emission Inventory 2001, Climate 2050
Impact of Potential Climate Change on
Average Max8hrO3
0.100
All grid averages (not just monitor locations)
0.090
Max8hrO3 (ppbV)
(
0.080
2001
2050
2050_np
Summers 2000-2002
Summers 2049-2051
Summers 2049-2051_np
0.070
0.060
0.050
0.040
0.030
0.020
0.010
0.000
West
Plains
Midwest
Northeast
Southeast
- 3-8 ppbV lower in 2050
- Only +/- 1ppbV difference without considering future emission
controls (2050_np)
- More significant reductions in summers.
US
Impact of Potential Climate Change on PM2.5
2001
2050
2050_np
Summers 2000-2002
Summers 2049-2051
Summers 2000-2051_np
20
18
16
PM2.5 (μg/m3)
(
14
12
10
8
6
4
2
0
West
Plains
Midwest
Northeast
Southeast
US
- about 0.3-3.8 µg/m3 lower in 2050
- maximum 0.6 µg/m3 difference without considering future emission
controls (2050_np)
-Usually np is slower in summer, though can be higher on average
Annual Averaged Changes in Averaged
Max8hrO3 & PM2.5
Max8hrO3 (%)
PM2.5 (%)
2050
2050np
2050
2050np
West
-6.5
0.2
-9.2
2.9
Plains
-7.9
1.4
-22.0
-0.8
Midwest
-10.5
-0.2
-22.7
4.2
Northeas
t
-10.0
-0.5
-28.5
6.5
Southeas
t
-14.8
2.3
-31.4
-2.4
US
-9.2
0.9
-23.4
1.1
Regional Predicted Max8hrO3
Characteristics Unit of 99.5% and peak: ppbV
2000-2002 summers
# of days
over 80
ppb
# of days
over 85
ppb
(sim/act)
Peak
2049-2051 summers
# of days
over 80
ppb
# of
days
over 85
ppb
2049-2051_np summers
Peak
# of days
over 80
ppb
# of
days
over 85
ppb
Peak
West / Los Angeles
149
95/85
119
31
6
97
221
186
146
Plains / Houston
127
107/87
127
29
10
94
165
146
143
Midwest / Chicago
78
66/32
138
19
12
106
59
44
152
Northeast / New York
51
38/46
112
1
0
81
82
60
121
Southeast / Atlanta
199
182/54*
124/
139
0
0
78
195
177
131
Significant improvement
Stagnation events Increase in some areas
* 1998-2000: 137
Conclusions
• Climate change, alone, with no emissions growth or controls has a
mixed effects on the ozone and PM2.5 levels as well as their sensitivities
to precursor emissions.
– Ozone generally up some, PM mixed
• The impact of changes in precursor emissions due to planned controls
and anticipated changes in activity levels have a much greater effect
than the impact of climate change for ozone and PM2.5 levels.
– Carefully forecasting emissions is critical to result relevancy
• Spatial distribution and annual variations in the contribution of
precursors to ozone and PM2.5 formation remain quite similar.
– Sensitivities of ozone to NOx increase on a per ton basis mostly due to
reduced NOx levels, a bit due to climate
– Sensitivities of PM2.5 to precursors similar on per ton basis
• Lower NOx and higher NH3 emissions increase sensitivity of NO3 to NOx in
2050 projected emissions case
Thanks… Questions?