Air Quality Modeling

1
Air Quality Modeling

• Rosa Sohn and Sun-Kyoung Park

2
1. Introduction
Atmospheric Chemistry
Emissions Inputs
Pollutant Distributions
Numerical Routines
Meteorological Fields
Effects
Emissions Modeling
Visualization
Meteorological Modeling
InputsPopulation Roads Land Use Industry
Meteorology
Economics
InputsTopography Observed Meteorology Solar
insolation
Controls
3
2. Eulerian and Lagrangian models

• 2.1. Eulerian Model
• The behavior of species is described relative to
  a fixed coordinate system
• (1) Single box model
• Focus atmospheric chemistry
• Lack physical realism - horizontal and vertical
  transport, etc.
• (2) Multi-dimensional grid-based air quality
  model
• Potentially the most powerful
• Involving the least-restrictive assumption
• 2.2. Lagrangian Model
• The concentration changes are described relative
  to the moving fluid

4
3. Air Quality Model Formulation(1)
i1,2,3, . . . , n
ci concentration of species i.
wind velocity vector
Di molecular diffusivity of species i
Ri rate of concentration change of species i by
chemical reaction
Si source/sink of i
r air densityn number of predicted species
5
3. Air Quality Model Formulation(2)
Reynolds decomposition K theory
Assumption
Atmospheric Diffusion Equation (ADE)
6
4. Model components and process descriptions(1)

• Turbulent transport and diffusion
• K-theory
• K Function of the atmospheric stability class
  and the mixing height
• Deposition
• Dry deposition vd(ra rb rc ) 1
  ra aerodynamic resistance, controlled by the
  atmospheric turbulence rb resistance in
  the fluid sublayer very near the plant surface
  rc surface(or canopy) resistance, the
  function of pollutant, land-use,
  surface condition(dew, rain or dry..) and season
• Wet deposition
• Because of the meteorological models uncertainty
  in the formation of the clouds and the
  precipitation, wet deposition has still much
  uncertainty.

7
4. Model components and process descriptions(2)

• Chemical kinetics
• Homogeneous gas-phase chemistry
• Heterogeneous chemistry
• Acid deposition, aerosol formation
• Radiative transfer (Approach)
• To adjust the sea-level photolysis rates for
  solar zenith angle, wavelength, changes in
  altitude, haze and clouds, preferably using
  measurements.
• To use a look-up table derived from a detailed
  radiative transfer model and then modify the
  results for clouds (e.g., CMAQ)
• To use radiative fluxes calculated by
  meteorological model being used to provide other
  field

8
4. Model components and process descriptions(3)

• Particulate matter
• Impact health, visibility and gas phase species
  levels
• (e.g., In the presence of aerosol, scattering
  increase
• ? Increase in ozone formation)
• Particulate matter modeling
• Formation and growth Sectioning size
  distribution
• Size and chemical compositionCondensation,
  Coagulation, Sedimentation and Nucleation
• Individual sources are simulated to emit a set of
  aerosol packets with specific sizes and
  compositions (Cass, et al).

9
5. Mathematical and computational
implementation(1)

• Horizontal transport algorithms
• Based on Finite Difference, Finite Element and
  Finite Volume
• Spectral Method, Lagrangian approach
• Problem with solving the set of equations
• Spatial discretization artificial numerical
  dispersion, which is manifested by the formation
  of spurious waves and by pollutant peaks being
  spread out
• Currently small error/uncertainty in application
• Chemical dynamics 80 of the computer time
• QSSA(quasi-steady-state approximation)
• Hybrid method
• Gear-type method
• Good for the large integration time step (e.g.,
  1hr)

10
5. Mathematical and computational
implementation(2)

• Monoscale, nested multiscale and adaptive grids
• Large grid size is inappropriate for the
  non-linear reactions (e.g., ozone) with
  significant chemical gradient in cities
• Considering computational resources, using finer
  grids in urban area and coarser grids over rural
  area
• Plume modeling
• Concentrated sources of some pollutants in coarse
  resolution(e.g., power plant)
• Mixing is at a finite rate ? local ? volume
  average concentration.
• Assumption of immediate mixing often leads to
  overestimating the oxidation rate of NO (e.g., in
  VOC rich environment).
• Adaptive mesh technique
• Mesh is generated automatically to capture the
  fine scale features

11
5. Mathematical and computational
implementation(3)

• Mass conservation in air quality models
• Without using the continuity equation explicitly,
  models diagnose air density from pressure and
  temperature (e.g., MM5)
• Even the continuity equation is satisfied,
• Meteorological model output is stored at a
  certain interval (e.g., 1hr)
• Air Quality Models time step (e.g., 10 min) ?
  Interpolation
• The interpolation of density and momentum does
  not guarantee the mass conservation
• ?Vertical or horizontal velocity is
  recomputed.

12
5. Mathematical and computational
implementation(4)

• Advanced Analysis routines
• Integration of specific physical and chemical
  processes terms
• Applied to Lagrangian Box Modeling studies
• Direct sensitivity analysis
• Brute Force
• Decoupled Direct Method(DDM)
• Adjoint Approach
• Limitation cannot capture nonlinear response

13
6. Model Input (1) - Meteorology

• Meteorological Input
• Horizontal and Vertical Wind fields, Temperature,
  Humidity, Mixing depth, Solar insolation fields,
  Vertical diffusivities, cloud characteristics(liqu
  id water content, droplet size, cloud size,
  etc.), rain fall
• How to prepare input fields?
• Interpolating relatively sparse observations over
  the modeling domain using the objective analysis
• Meteorological Model (e.g., MM5) output because
  of the sparseness of data

14
6. Model Input (2) Emission

• Emission
• CO, NO, NO2, SO2, VOCs, SO3, NH3, PM2.5 and PM10
• Emissions are one of the most uncertain, but the
  most important inputs into air quality models
• Temporal Processing
• The inventory is the yearly averaged data, but
  the AQM needs short interval emission input(e.g.,
  hourly).
• Spatial Processing
• The inventory is county based data, but the AQM
  needs gridded emission inventory if you are doing
  the multi-dimensional grid-based air quality model

15
UNPROJECTED LATITUDE-LONGITUDE
16
Map Projection

• Grid defined in the AQM depends on the map
  projection.
• Map Projection
• Attempt to portray the surface of the earth or a
  portion of the earth on a flat surface
• Cylindrical
• Psuedo-cylindrical
• Conical
• Azimuthal
• Other

17
(1) Cylindrical Projection
Mercator Lamberts Cylindrical Equal-area
Galls Sterographic Cylindrical Miller
Cylindrical Behrmann Cylindrical Equal-area
Peters Transverse Mercator
18
(2) Psuedo-Cylindrical Projections

• Mollweide Equal-area
• Eckert IV Equal-area
• Eckert VI Equal-area
• Sinusoidal Equal-area
• Robinson

19
(3) Conical Projections

• Albers Equal Area Conical Projection
• Lambert Conformal Conical Projection
• Equidistant Conical Projection

CONICAL TANGENT
20
(4) Azimuthal Projections
Equidistant Azimuthal Projection Lambert
Equal Area Azimuthal Projection
21
7. Air Quality Model Evaluation (1)

• Assessment of the adequacy and correctness of the
  science represented in the model through
  comparison against empirical data
• Normalized Bias, D
• Normalized Gross Error, Ed
• Unpaired Peak Prediction Accuracy

22
7. Air Quality Model Evaluation (2)

• Statistical Benchmark for the model performance
  (US EPA, 1991 Tesche et al.)
• Normalized Bias ?
  5 ? 15
• Normalized Gross Error ? 30
  ? 35
• Unpaired peak prediction accuracy ? 15 ? 20

23
7. Model Evaluation (3) Input Meteorology

• Mean Bias Error (MBE)
• Mean Normalized Bias (MNB)
• Root Mean Square Error (RMSE)
• Mean Absolute Gross Error (MAGE)
• Mean Normalized Gross Error (MNGE)

24
Statistical Benchmarks
7. Model Evaluation (4) Input Meteorology
Wind Speed RMSE ? 2 m/s Bias ? ?0.5 m/s
Wind Direction Gross Error ? 30 deg Bias ? ?10 deg
Temperature Gross Error ? 2 K Bias ? ?0.5 K
Specific Humidity Gross Error ? 2g/kg Bias ? ?1g/kg
Source Environmental Report MM5 Performance
Evaluation Project
Matthew T. Johnson, Kirk Baker (2001)
25
8. Application (1)

• Sensitivity to Process Parameterizations
• Sensitivity to Model Numerics/Structure
• Small uncertainty in numerical technique
• Grid size, number of the vertical layers.
• (e.g., Difference in ozone prediction
  Horizontal grid size 5 km, 10 km and 20km
  Number of the vertical layers 6 15
  layers)

26
8. Application (2)

• Sensitivity to Model Input
• Emissions, meteorological conditions, boundary
  conditions, initial conditions, HONO formation
  rate and deposition
• Emission control ? Ozone and PM control (e.g.,
  SO2 control ? sulfate decrease, but nitrate
  increases)
• VOCs with different reactivity ? Ozone, PM,
  (e.g., Methanol based fuels would be beneficial
  for ozone control because of its atmospheric low
  reactivity)

27
9. Current Status of AQM

• 1st generation simple chemistry at local scales
• 2nd generation local, urban, regional addressing
  each scale with a separate model and often
  focusing on a single pollutant.
• 3rd generation multiple pollutants
  simultaneously up to continental scales and
  incorporate feedbacks between chemical and
  meteorological components.
• Models3 (SMOKE, MM5 and CMAQ(Community Multiscale
  Air Quality) Modeling system urban to regional
  scale air quality simulation of tropospheric
  ozone, acid deposition, visibility and fine
  particulate).
• 4th generation (Future) extend linkages and
  process feedback to include air, water, land, and
  biota to simulate the transport and fate of
  chemical and nutrients throughout an ecosystem.

28
Step 1. Getting Started with AQM

• Download the existing model with free of charge
• Models3
• MM5 (http//www.mmm.ucar.edu/mm5/mm5-home.html)
• PSU/NCAR mesoscale model version 5
• Meteorological Field
• SMOKE (http//edge.emc.mcnc.org/uihelp/docs/smoke.
  html)
• Sparse Matrix Operator Kernel Emissions modeling
  system
• Converting emissions inventory data into the
  formatted emission files required by an AQM
• CMAQ (http//www.epa.gov/asmdnerl/models3/)
• Community Multiscale Air Quality Modeling System
• Atmospheric chemistry combined with the numerical
  routine

29
Step 2. Meteorological Input MM5 (1)

• MM5 (PSU/NCAR mesoscale model version 5) download
• http//www.mmm.ucar.edu/mm5/mm5-home.html
• MM5 Input Data http//dss.ucar.edu/catalogs/
• Topography and Landuse data
• Gridded atmospheric data with sea-level pressure,
  wind, temperature, relative humidity and
  geopotential height
• Observation data that contains soundings and
  surface reports

30
Step 2. Meteorological Input MM5 (2)

• MM5 Output Manipulation for the evaluation, etc.
• MM5toGrADS http//www.mmm.ucar.edu/mm5/mm5v3/tutor
  ial/mm5tograds/mm5tograds.html
• GrADS(Grid Analysis and DisplaySystem)
  http//grads.iges.org/grads/
• Converting the MM5 output to the format required
  in SMOKE and CMAQ (NetCDF format)
• MCIP2(Meteorology-Chemistry Interface Processor
  Version 2) Inside of the CMAQ model system

31
Step 3. Emission Input SMOKE

• National Emission Inventory (1996 yr, 1999 yr)
  (http//www.epa.gov/ttn/chief/net/index.html)
• Projection to the model year
• EGAS 4.0 (http//www.epa.gov/ttn/chief/emch/projec
  tion/egas40/)
• Spatial Processing
• Converting the county based inventory data to the
  gridded emission inventory which fits to the
  multi-dimensional grid-based air quality model
• ESRI ArcGIS ArcView, ArcInfo, ArcMap,
  ArcToolbox and ArcCatalog
• (Architecture Computer Lab (Rm 359))
• SMOKE (http//edge.emc.mcnc.org/uihelp/docs/smoke.
  html)
• Converting emissions inventory into the formatted
  emission files required by an AQM (NetCDF format)

32
Step 4. Air Quality Modeling - CMAQ

• CMAQ Modeling system
• (Community Multiscale Air Quality)
• http//www.epa.gov/asmdnerl/models3/
• CMAQ Document
• http//www.epa.gov/asmdnerl/models3/doc/science/sc
  ience.html
• CMAQ Tutorial
• http//www.epa.gov/scram001/cmaq.htm

33
Step 5. Tools

• PAVE
• The Package for Analysis and Visualization of
  Environmental Data
• Visualization and the analysis of NetCDF data
• http//www.epa.gov/asmdnerl/models3/vistutor/pave.
  html
• NetCDF IO/API
• The Models-3 Input/Output Applications
  Programming Interface
• The standard data access library for EPA's
  Models-3 available from both C and Fortran.
  (ASCII or Binary ?? NetCDF)
• http//www.emc.mcnc.org/products/ioapi/AA.html
• AQM testing with arbitrary input (e.g., zero
  emission)
• Evaluation

34
Questions Comments Reference NARSTO
critical review of photochemical models and
modeling, Armistead Russell and Robin Dennis,
2000 Atmospheric Environment 34. 2283 - 2324