FUTURE DIRECTIONS FOR IMPROVING AIR QUALITY PREDICTIONS

1
FUTURE DIRECTIONS FOR IMPROVING AIR QUALITY
PREDICTIONS Gregory R. Carmichael University
of Iowa
Center for Global Regional Env. Research
68 faculty/16 departments/ 6
colleges

• Outline
• Overview of AQ modeling status.
• How good are the modeling tools?
• Constraining models with observations.
• Future directions.

College of Engineering
2
Air Quality Modeling Improving Predictions of
Air Quality (analysis and forecasting
perspectives)
Met model
Predicted Quantity e.g., ozone AQ violation
Chemical, Aerosol, Removal modules
CTM
How confident are we in the models predictions?
Emissions
Observations
3
Chemical Weather Prediction Improvements
through Closer Integration of Models and
Measurements
Chemical
4
How do we build upon what is done and move beyond
to improve air quality prediction?

• Informed by model inter-comparison studies.
• Informed by comparisons of predictions with
  observations.
• Informed by process studies.

5
Experiments such as TRACE-P and ACE-Asia employ
mobile Super-Sites and study pollution outflow
from source regions
China
Spring 2001
6
Predictability as Measured by Correlation
Coefficient Met Parameters are Best
lt 1km
O3 predicted better than CO
Performance decreases with altitude
Carmichael et al., JGR, 2003
7
Lessons from the ICARTT Experiment
Experiments such as these employ mobile
Super-Sites and study pollution outflow from
source regions
8
Extensive Real-Time Evaluation of Regional
Forecasts Stu McKeen
http//www.etl.noaa.gov/programs/2004/neaqs/verifi
cation/
9
Forecasting Air Quality an Important Activity in
Air Quality Management Persistence
Single Forward Model w/o
assimilation
Ensemble forecast (8 models) w/o assimilation
(further improvements with bias corrections based
on obs)
McKeen et al., JGR, 2005
10
Ensemble Methods Also Work for PM2.5 Forecasting
McKeen et al., JGR, 2007
11
MICS-Asia lt Model InterComparison Study in Asia gt
Evaluation of model performance to make an
international common understanding and improve
for air pollution modeling in East Asia
Main goal

• Nine different regional models
• Observations
• EANET (47 sites) (gas, aerosol, deposition)
• Ozonesondes
• Trace-P Obs.
• Special obs. (aerosols)
• Met obs (sondes and surface)
• (daily monthly analysis)
• Special Section of Atmospheric Environment (8
  papers)

12
The ensemble mean near surface monthly mean total
sulfur deposition amounts (as sulfate) for the
different seasons.
March 2001
July 2001
SO42- wet deposition (mg m-2)
(mg m-2)
Nitrate quantities typically underestimated.
13
Special Section Atmospheric Environment

• The Model Intercomparison Study for Asia Phase
  II, Methodology and Overview of Findings
• Model Intercomparison and Evaluation of Ozone and
  Relevant Species
• Model Intercomparison and Evaluation of
  Particulate Sulfate, Nitrate and Ammonium
• Impact of Global Emissions on Regional Air
  Quality in Asia
• An intercomparison study of emission inventories
  for the Japan region.
• Sensitivity analysis of predicted aerosol
  composition to the aerosol module formulation.
• Model Intercomparison and Evaluation of Acid
  Deposition
• Evaluating Gaseous Pollutants in East Asia Using
  An Advanced Modeling System Models-3/CMAQ System

14
Regional-Scale Chemical Analysis for Air Quality
Modeling A Closer Integration Of Observations
And Models
Transport Meteorology

• Improved
• forecasts
• science
• field experiment design
• models
• emission estimates
• S/R relationships

15
Air Quality Prediction A Challenge of Scales and
Integration
Modified after Pierce NASA/Langley
16
Mega-City Footprints Can Be Large
Percentage Contribution to Total Sulfur
Deposition due to SO2 Emissions from Megacities
in Asia, (1975-2000)
Megacities account for 11 of emissions and
occupy lt1 of land area
Guttakundi et al., Atmos. Env., 2003 JGR, 2006
17
Integrated Science Studies Impacts of Global
Composition on Regional Air Quality
MODIS Aerosol Optical Depth (AOD) July 17-20, 2004
July 17, 2004
July 18, 2004

• The largest Alaskan wild fire event on record
  occurred in late June-July 2004.
• Satellite, remote and in-situ airborne, and
  ground measurements collected during INTEX-A
  quantify the impacts of the Alaskan fires on US
  air quality.

July 19, 2004
July 20, 2004
18
Integrated Science Studies Impacts of Global
Composition on Regional Air Quality
DC-8 LIDAR backscatter, MODIS AOD and EPA AIRNow
PM2.5 on July 20, 2004
DC8 LIDAR shows that the high aerosols seen by
MODIS were at several layers in the troposphere.
19
Regional Transport Is a Major Fraction of PM2.5
and Ozone
Top Urban Bottom Rural
12-month average PM2.5 mass from speciation
samplers
Reference 2002 EPA Trends Report
http//www.epa.gov/air/airtrends/chem_spec_of_pm2.
5_b.pdf
20
NAMMA/TexAQS Integrated Science Studies Saharan
Dust Transport Global Source/Receptor Studies
CALIPSO Observations Link Texas Dust event to
Saharan Source region
NASA NAMMA Flight tracks shown in black (08/19)
and white (08/20)
Boundary Layer back trajectories from August 28
CALIPSO track shown in red
21
Air Quality Prediction A Challenge of Scales and
Integration
Modified after Pierce NASA/Langley
22
Integrated Science Studies Impacts of Global
Composition on Regional Air Quality
Global-Regional-Urban nesting of CTMs
Effects of Boundary Conditions are significant
and improve predictions (Tang et al., JGR 2007).
Alaskan BB Impacts Northern Boundary
Assessment of continental inflow/outflow requires
unified modeling/measurement strategy to
accurately characterize coupling between the
continental boundary layer, free troposphere, and
long-range transport.
23
Megacities in Aggregate Impact the Global
Environment
Large spatial differences
Transport from the tropics significant
Fig. 5.1 Annual mean plots of the sum of all of
the (10 d) MPC tracers for the model surface
layer density (10-12 kg/m3) and the column above
5km (10-9 kg/m2). From Lawrence et al., 2007.
24
Estimates of S-R relationships Annual mean
surface O3 decrease from 20 reductions in
anthropogenic NOx emissions
Source region
Full range of 15 individual models
Receptor
Largest source-receptor pair NA?EU
? more extensive information is found
onhttp//aqm.jrc.it/HTAP/.)
25
HTAP -- Experiments
more extensive information is found
onhttp//aqm.jrc.it/HTAP/.) Study Year
2001 20 anthropogenic emission perturbations
Table 3 Multi-model derived import sensitivities
for surface concentration, deposition and column
load of anthropogenic black carbon (BC),
particulate organic matter (POM), sulphate (SO4)
and all sulphur species (SO2SO4) to 20
reductions of anthropogenic emissions in the
source regions. Domestic contributions are shown
in bold. The standard deviation is shown in
parentheses.
26
Advanced Data Assimilation Techniques Provide
Data Fusion and Optimal Analysis Frameworks
Example 4dVar
Cost function
Observations information consistent with reality
Current knowledge of the state
Model information consistent with
physics/chemistry
The system is very under-determined need to
combine heterogeneous data sources with limited
spatial/temporal information
27
Intensive Field Experiments (e.g., ICARTT)
Provide Our Best Efforts to Comprehensively
Observe a Region
O3
28
Assimilation Produces An Optimal State Space
w/o assimilation
with assimilation
Ozone predictions
All Data Used
Example July 20, 2004
Chai et al., JGR 2007
Region-mean profile
29
Information content of various observations
evaluated by different combinations of data sets
assimilated
the importance of measurements
above the surface.
Surface-only
Lidar-DC8
30
Verification Ron Brown Observations Independent
Data
Predicted uncertainties estimated from background
(B) error estimates
31
Source/Receptor Calculations Perturbation
approaches
Source-oriented approach - Direct sensitivity
analysis.
Receptor/target-oriented approach - Adjoint
sensitivity analysis.
32
The Adjoints Are Themselves Very Valuable
CO as a tracer of fossil fuel
CO2Caveat Fire, chemistry, LPS (Campbel et al,
In Press)
Flight 10 Time Series
Adjoint Sensitivity
Provide insights into the footprint of an
observation
33
Sensitivity of ozone violations wrt emissions
Adjoint Analysis of the Contribution of Different
Emissions to Ozone Violations July August 2004
Hakami et al., EST 2006
34
In AQ Predictions Emissions Are A Major Source Of
Uncertainty Data Assimilation Can Produce
Optimal Estimates (Inverse Applications)
Li et al., Atmos. Env., 2007
35
What parameters should be target for adjustment?
emissions, initial conditions, boundary
conditions? and What Species?
36
Emission Inversion with Satellite Data
4D-Var setup
Time window 1200 UT- 2000 UT July 20,
2004 Control Initial ozone, and NOx
emissions Observations Ozone from different
platforms, and SCIAMACHY tropospheric NO2 columns
Scaling factors
E only
E IC
37
Q-Q plots
Results of Consideration of Emissions only and
Emissions and Initial Conditions
r20.76
r20.57
Points out model deficiencies
38
Air quality monitoring and ensemble forecasting
framework.
Feedbacks (setting standards and assess
achievements)
Region
policy
Prediction by CTM
Forecasts
Information on observations and model calculations
Super sites monit. st. (many)
Modified from Y. Zhang , Guangzhou Meeting 2007
39
Linking Emissions to Aerosol Trace Gas
Distributions and Subsequent Effects -- Summary

• Models measurements have improved
  substantially.
• Further improvements will require reductions in
  key uncertainties (e.g., emissions, better basic
  understanding of some processes).
• Closer integration of observations.
• Need to develop better strategies for providing
  uniqueness to targeted applications (e.g.,
  sources sectors).

40
Improving Regional Predictions Of Air Quality
Through The Closer Integration Of Observations
And Models Summary

• Models and measurements have improved
  substantially.
• Further improvements will require reductions in
  key uncertainties (e.g., emissions) and closer
  integration of observations.
• Advanced data assimilation techniques provide
  data fusion and optimal analysis frameworks.
• The system is very under-determined need to
  combine heterogeneous data sources with limited
  spatial/temporal information.
• Recovery of initial conditions, emissions and
  boundary conditions are necessary.
• Global analysis is necessary.
• These techniques can help in the analysis of
  hemispheric transport.

41
(No Transcript)
42
Improving Air Quality Predictions

• Models and measurements have improved
  substantially.
• Improvements will require reductions in key
  uncertainties (e.g., emissions, boundary
  conditions) and closer integration of
  observations.
• Ensemble methods improve forecasts of AQ.
• Formal data assimilation feasible necessary.
• Important implications for measurement systems
  and models mesonets for air quality prediction.

43
Aerosol, NO2, and SO2 from Space
AOD in 2002 (Mao and Li)
O3 (Zhu et al.)
NO2 in Oct. 2004
NO2 (Nature)
44
Atmospheric Brown Cloud - Asia
total
Anthro-fraction
Ramanathan et al., JGR, 2007
45
The Role of ABCs at Megacities in Asia
46
Optimal Interpolation
47
Preliminary Results Optimal Interpolation
48
IGACO a Strategy for Integration of Models and
Measurements

• Integration of Models and Measurements
• Provide 4-Dimensional context of the observations
• Facilitate the integration of the different
  measurement platforms
• Evaluate processes (e.g., role of biomass
  burning, heterogeneous chemistry.)
• Evaluate emission estimates (bottom-up as well as
  top-down)
• Emission control strategies testing
• Air quality forecasting

49
NAMMA/TexAQS Integrated Science Studies Saharan
Dust Transport Global Source/Receptor Studies
RAQMS1 PM2.5 forecast predicts significant dust
contribution to observed2 PM2.5
RAQMS 06hr Fx vs AIRNow PM2.5
Real-time MODIS AOD Assimilation
HSRL3 flight region
RAQMS uses lateral boundary conditions from GMAO
aerosol forecast (A. da Silva, GSFC)
1Real-time Air Quality Modeling System
(RAQMS) 2US EPA AIRNow PM2.5 Network 3High
Spectral Resolution Lidar (HSRL)
50
NAMMA/TexAQS Integrated Science Studies Saharan
Dust Transport Global Source/Receptor Studies
HSRL Aerosol Extinction and Depolarization Ratio
observations confirm dust forecast
RAQMS 06hr Fx vs AIRNow PM2.5
RAQMS Dust Forecast with GMAO Boundary Conditions
HSRL flight track
51
NAMMA/TexAQS Integrated Science Studies Saharan
Dust Transport Global Source/Receptor Studies
CALIPSO Observations Link Texas Dust event to
Saharan Source region
Airborne data provides in-situ characterization
of aerosol optical properties
Assessment of global source receptor
relationships requires unified modeling/measuremen
t strategy to accurately characterize global
source regions, long range transport, and
boundary layer entrainment processes.
Dust high scattering coefficients that are
constant with wavelength (low angstrom exponents)
52
Increase is important from pollution and climate
perspectives
53
EU/USA
54
To assess the magnitude of import from other
major emission areas in the Northern Hemisphere,
we compare three simulations 1) MOZART with all
emissions included 2) without emissions from
North America and 3) without emissions from
Europe. Simulations without emissions from the
MICS domain were also examined, but are not
discussed here.
EANET sites
How to evaluate ?
55
Improving Regional Predictions Of Air Quality
Through The Closer Integration Of Observations
And Models Summary

• Models and measurements have improved
  substantially.
• Further improvements will require reductions in
  key uncertainties (e.g., emissions) and closer
  integration of observations.
• Advanced data assimilation techniques provide
  data fusion and optimal analysis frameworks.
• The system is very under-determined need to
  combine heterogeneous data sources with limited
  spatial/temporal information.
• Recovery of initial conditions, emissions and
  boundary conditions are necessary.
• Global analysis is necessary.
• These techniques can help in the analysis of
  hemispheric transport.

IGACO GURME
56
Summary Air Quality Topic, Future Airborne
Missions

• Science Questions
• Transition from local to regional/hemispheric
  influence
• LRTAP contributions
• Hg/BB long-range transport
• Climate change heat waves, increased BB
• Validation/update emission estimates (coupled
  with satellite)
• Processes impacting CONUS AQ
• Export out of BL (Aerocom)
• Import into BL (subsiding airmass/boundary layer
  inversions)
• Maturation of AQ FX capability/incorporation of
  observations
• Suborbital contribitions
• Integrated Science studies
• Integrated airborne/sat emission estimates
• BL processes (low and slow)
• LBC constraints (high and long)
• Suborbital needs to link surface NETWORKS to
  satellite obs (US GEO)

57
MICS-Asia lt Model InterComparison Study in Asia gt
Evaluation of model performance to make an
international common understanding and improve
for air pollution modeling in East Asia
Main goal

• Nine different regional models
• Observations
• EANET (47 sites) (gas, aerosol, deposition)
• Ozonesondes
• Trace-P Obs.
• Special obs. (aerosols)
• Met obs (sondes and surface)
• (daily monthly analysis)
• Special Section of Atmospheric Environment (8
  papers)

58
SO2 Monthly Means
59
Improving Regional Predictions/Assessment Of Air
Quality Through The Closer Integration Of
Observations And Models
Future Airborne Air Quality missions should be
used with global and regional chemical data
assimilation/prediction systems to interpret/
integrate long-term satellite and surface
monitoring networks

60
The ensemble mean near surface monthly mean total
sulfur deposition amounts (as sulfate) for the
different seasons.
March 2001
July 2001
SO42- wet deposition (mg m-2)
(mg m-2)
Nitrate quantities typically underestimated.
61
Ensemble Ozone Monthly Mean
March 2001
July 2001
Japanese sites
CoV
Fig. 11
62
Contribution of Megacities(March/April 2001)
(12 of Emissions in 2 of the Area)
sulfate
ozone
How to evaluate ?
63
Air Quality Prediction A Challenge of Scales and
Integration
Modified after Pierce NASA/Langley