Use of Airmass History Models

1
Use of Airmass History Models Techniques for
Source Attribution

• Bret A. SchichtelWashington UniversitySt.
  Louis, MO

Presentation to EPA Source Attribution
workshopJuly 16 - 18, 1997
http//capita.wustl.edu/neardat/CAPITA/CapitaRepor
ts/AirmassHist/EPASrcAtt_jul17/index.htm
2
Airmass History
Estimation of the pathway of an airmass to a
receptor (backward AMH) or from a source (forward
AMH) and meteorological variables along the
pathway.
Airmass Back Trajectory
Airmass Met. Variables
Plumes
3
Source Receptor Relationship
ReceptorConcentration
Dilution
Chemistry/Removal
Emissions



Airmass history modeling and analysis aid in the
understanding of the SRR processes and
qualitatively and quantitatively establish source
contributions to receptors.
4
Airmass History Analysis Techniques

• Individual airmass histories
• Backward and forward airmass history ensemble
  analysis
• Air quality simulation
• Transfer matrices
• Emission Retrieval

Goals of Workshop addressed

• Area of Influence
• Selecting and analyzing pollution episodes
• Selecting control strategies
• Evaluate air quality models

5
Characteristics of Airmass History Analyses to be
presented

• Regional Pollutants
• Ozone
• Fine particulates
• visibility
• Climatological analysis
• Proposed year fine particle standard
• Source attribution for typical conditions
• Source attribution for typical episodes

6
Regional Airmass History Models
- ATAD -Single 2-D back/forward trajectories
from single site -Wind fields Diagnostic from
available measured data -No Mixing -
HY-SPLIT -3-D back/forward trajectories and
plumes from single site -Wind fields NGM, ETA,
RAMS, . -Mixing for Plumes No Mixing for back
trajectories -Pollutant simulation - CAPITA
Monte Carlo Model -3-D back/forward airmass
histories and plumes from multiple sites -Wind
fields NGM, RAMS,... -Mixing for forward and
backward airmass histories -Pollutant simulation
7
Airmass Histories - Model Outputs
Multiple 3-D Back Trajectories
Airmass History Variables
2-D Back Trajectory
8
CAPITA Monte Carlo Model
Direct simulation of emissions, transport,
transformation, and removal
http//capita.wustl.edu/capita/CapitaReports/Monte
Carlo/MonteCarlo.html
9
Transport
Advection 3-D wind fields
Horizontal Dispersion Eddy diffusion Kx and Ky
vary depending on hour of day
Vertical Dispersion
Below the mixing layer particles are uniformly
distributed from ground to mixing height. No
dispersion above mixing layer.
10
Kinetics
ChemistryPseudo first order transformation
rates, function of meteorological variables,
such as solar radiation, temperature, water vapor
content
Deposition dry and wet Pseudo first order
rates equationsDry deposition function of hour
of solar radiation, Mixing Hgt Wet deposition
function of precipitation rate
11
Model Output

• Database of airmass histories
• Pollutant concentrations and deposition fields
• Transfer matrices

Computer Platform
IBM-PC
Computation Requirements
Low 3 months of back airmass histories for 500
sites 1 day3 months of sulfate simulations
over North America 2 days
User expertise
Airmass history server- Low Pollutant simulation
- High
12
Primary Meteorological Input Data
National Meteorological Centers Nested Grid Model
(NGM)
Time range 1991 - Present Horizontal
resolution 160 kmVertical resolution 10
layers up to 7 km3-D variables u, v, w, temp.,
humiditySurface variables include Precip,
Mixing Hgt,.Database size 1 year - 250
megabytes
13
Airmass History Analysis Techniques
Individual Airmass Histories Techniques -Visuall
y combine measured/modeled air quality data
with airmass history and meteorological
data Uses -Pollution episode analysis. Brings
meteorological context to air quality
data. Goals of Workshop addressed -Pollution
episode selection and analysis -Evaluate air
quality models
14
Animation of Grand Canyon Fine Particle Sulfur,
Back Trajectories Precipitation
The following day the airmass transport is still
from the south, but it encountered precipitation
near the Grand Canyon. The sulfur concentrations
dropped by a factor of 8.
On February 7, the Grand Canyon has elevated
sulfur concentrations. The back trajectory shows
airmass stagnation in S. AZ prior to impacting
the Grand Canyon.
15
Merging Air Quality Meteorological Data for
Episode Analysis
OTAG 1991 modeling episode Animation
16
Anatomy of the July 1995 Regional Ozone Episode
Regional scale ozone transport across state
boundaries occurs when airmasses stagnate over
multi-state areas of high emission regions
creating ozone blobs which are subsequently
transport to downwind states
17
Strengths

• Applicable to particulates, ozone and visibility
• Informed decision - Brings multiple variables
  and views of data for selection and analysis
  of episodes
• High user efficiency - Visualize large
  quantities of data quickly
• Low computer resources

Weaknesses

• Single trajectories prone to large errors.
• Potential for information overload.

18
Airmass History Analysis Techniques
Ensemble Analysis Techniques - Cluster
analysis forward and backward AMH - Residence
time analysis Backward AMH - Source Regions
of Influence Forward AMH Uses - Qualitative
source attribution - Transport
climatology Goals of Workshop addressed - Area
of Influence - Pollution episode
representativeness - Selecting control
strategies
19
Residence Time AnalysisWhere is the airmass most
likely to have previously resided
Whiteface Mt. NY, June - August 1989 - 95
Back Trajectories
Residence Time Probabilities
Wishinski and Poirot, 1995 http//capita.wustl.ed
u/otag/Reports/Restime/Restime.html Airmass
histories from HY-SPLIT model
20
Airmass History Stratification
Whiteface Mt. NY- Residence Time Probabilities
Ozone lt 51 ppb
June - August 1989 - 95
Low ozone concentrations are associated with
airflow from the northeast
High ozone concentrations are associated with
airflow from the east to southeast

• Technique identifies airmass pathways not the
  source areas along the pathway
• Central bias - all airmass histories must pass
  through receptor grid cell

21
Removing the Central Bias Incremental
Probability Analysis
Incremental Probability
Stratified Probability
Everyday Probability

-

• High ozone is associated with airflow from the
  central east
• Regions implicated increase from south to north

22
Identifying Unique Source RegionsIncremental
Probabilities from 23 Combined Receptor Sites
Upper 50 Ozone
Lower 50 Ozone
June - August 1989 - 95
June - August 1989 - 95

• High ozone is associated with airflow from the
  Midwest
• Implies that Midwest is source of high ozone
  to many receptors. This region would be good
  source area to focus control strategies on.

23
Strengths

• Applicable to particulates, ozone, visibility
• Ensemble analysis reduces trajectory error
• Does not include a prior knowledge of emissions
  and kinetics
• Receptor viewpoint Which sources contribute to
  favorite receptor region
• Regional scale analysis and climatology

Weaknesses

• Qualitative
• Not suitable to evaluate local scale influences
• Does not implicate specific sources or source
  types

24
Source Region of InfluenceThe most likely region
that a source will impact
St. Louis Source
Transfer Matrix
Forward Airmass Histories

• St. Louis emissions can impact anywhere in the
  Eastern US. The impact tends to decrease with
  increasing transport distances.
• The source region of influence is defined as the
  smallest area encompassing the source that
  contains 63 of ambient mass. Note, this is a
  relative measure.

25
Source Region of Influence - St. Louis, MO
Quarter 3, 1992
Quarter 3, 1995
The shape and size of the region of influence is
dependent upon the pollutant lifetime, wind speed
and wind direction. The longer the lifetime,
higher the wind speed the larger the region of
influence. The elongation is primarily due to
the persistence of the wind direction.
26
Transport Climatology - Summer

• Resultant transport from Texas around Southeast
  and eastward.
• Region of influence is 40 smaller in Southeast
  compared to rest of Eastern US.

Schichtel and Husar, 1996 http//capita.wustl.edu
/otag/reports/sri/sri_hlo3.htm
27
Transport Climatology - Local Ozone Episodes
High ozone in the central OTAG domain occurs
during slow transport winds. In the north and
west, high ozone is associated with strong winds.
Low ozone occurs on days with transport from
outside the region. The regions of influence
(yellow shaded areas) are also higher on low
ozone days.
28
OTAG Modeling Episodes Representativeness
Transport winds during the 91,93,95 episodes
are representative of regional episodes.OTAG
episode transport winds differ from winds at high
local O3 levels.
Comparison of transport winds during the 91,
93, 95 episodes with winds during regional
episodes in general.
Comparison of transport winds during the 91,
93, 95 episodes with winds during locally high
O3.
29
Strengths

• Source viewpoint Which receptors are impacted
  by favorite source region
• Applicable to particulates, ozone, and
  visibility
• Applicable to climatology and episode analysis
• Direct measure of a sources region of influence
  if pollutant lifetime is known

Weaknesses

• Pollutant lifetime varies with time space -
  often ill-defined
• Simplified kinetics - can only define a
  boundary, not a source contribution field
• Does not account for vertical distribution of
  pollutants

Future Development

• Include vertical distribution of pollutants
• Enhance kinetics - add removal and
  transformation processes
• define contribution field within the region of
  influence

30
Complementary Analyses

• Forward and backward airmass history analysis
  techniques
• Analyses incorporating measured meteorology and
  receptor data

Ozone roses for selected 100 mile size
sub-regions.
Calculated from measured surface winds and ozone
data. At many sites, the avg. O3 is higher when
the wind blows from the center of the domain.
Same conclusion drawn from forward and backward
airmass history analyses.
31
Airmass History Uncertainty

• Sources of uncertainty
• Meteorological data
• Physical assumptions of airmass history model
• Horizontal and vertical transport dispersion
• Airmass starting elevations
• Inclusion of surface affects

• Uncertainty Quantification
• 20 - 30 /day trajectory error.HY-SPLIT model
  and NGM winds evaluated during the ANATEX tracer
  experiments (Draxler (1991) J. Appl. Meterol.
  301446-1467).
• 30 - 50 /day trajectory error
• Several models and wind fields evaluated during
  the ANATEX tracer experiments (Haagenson et
  al., (1990) J. Appl. Meterol. 291268-1283)
• Uncertainties can be reduced by considering
  ensembles of airmass histories, assuming errors
  are stochastic and not biased

32
Airmass History Model ComparisonHY-SPLIT Vs.
CAPITA Monte Carlo Model
HY-SPLIT NGM wind fields, no mixing Monte
Carlo Model NGM wind fields, mixing
At times individual Airmass histories compared
very well
At times individual Airmass histories compared
very poorly
33
The three month aggregate of airmass histories
produced similar transport patterns.
34
Airmass History Analysis Techniques
Pollutant Simulation and Transfer
Matrices Technique -Airmass Histories
Emissions KineticsUses - Quantitative source
attribution (transfer matrix) - Long-term and
episode pollutant simulation Goals of Workshop
addressed - Area of Influence - Selecting
control strategies
35
http//capita.wustl.edu/capita/CapitaReports/Monte
Carlo/MonteCarlo.html
36
Kinetic Processes Applied to Single Airmass
History
Variation of rate coefficients along trajectory,
and corresponding sulfur budget.
St. Louis airmass history
37
Comparison of simulated Sulfate to Measured
38
Comparison of simulated Wet Deposited Sulfate to
Measured
39
Transfer Matrices - Massachusetts Receptor, Q3
1992
Transit Probability
SO2 Kinetic Probability
SO4 Kinetic Probability
Likelihood an airmass from a source is
transported to the receptor
Likelihood SO2 emissions into the airmass impact
the receptor as SO2
Likelihood SO2 emissions into the airmass impact
the receptor as SO4
40
Quantitatively Define Source Receptor Relationship
SO2 and SO4 Source Attribution to Massachusetts
Receptor, Q3 1992
1985 NAPAP SO2 Emissions
41
Strengths

• Applicable to particulates and visibility
• Applicable to climatology and episode analysis
• Regional scale analysis
• Quantitative
• Applicable to what if analyses

Weaknesses

• Cannot simulate coupled non-linear chemistry
• Kinetics most appropriate for time periods used
  for tuning
• Low spatial resolution - not suitable for
  evaluation of near field influences

42
Summary

• Airmass history models and analysis can and have
  been be used to qualitatively and quantitatively
  perform source attribution.
• Airmass history models and analysis are suitable
  for addressing regional air quality issues, such
  as ozone, fine particulates and visibility
  degradation.
• Airmass history models and analysis are
  applicable to long term analysis, so can be used
  for source attribution for the proposed year fine
  particle standard.
• Many of these analyses are qualitative in nature
  and are appropriate as support for other analysis
  procedures.