Date post: | 28-Jan-2016 |
Category: |
Documents |
Upload: | raymond-boyd |
View: | 216 times |
Download: | 0 times |
1
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Air Quality Modeling
Zac Adelman and Craig MattocksCarolina Environmental Program
dx =
(Reco
sφ)d
λ e
∂C/ ∂t = - ∂(uC)/ ∂x - ∂(vC)/ ∂y - ∂(wC)/ ∂z
J = ∫ 4πIλ,T Aλ,T QYλ,T dλ ∞
0
krkr = Ar(300/T)Bexp(Cr/T) = Ar(300/T)Bexp(Cr/T)
2
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Outline Why model air pollution? Air quality modeling system components
– Meteorology– Emissions– Chemistry and transport
Technical and operational details Problems in air quality modeling Application examples Future directions in atmospheric modeling
3
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Adapted from Mackenzie and Mackenzie, “Our Changing Planet”, Prentice Hall, New Jersey, 1995.
Anthropogenic Emissions
Biogenic Emissions
Advection Diffusion
Boundary Conditions
Aerosol Processes:
Nucleation & Coagulation
Condensation & Deposition Evaporation &
Sublimation
O3 + NO NO2 + O2
O + O2 + M O3 + M
Gas Chemistry: Aqueous Chemistry:
SO2 (g) SO2 (aq)
Source Sink
Geogenic Emissions
4
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Why model air pollution?
Air pollution models are frameworks that integrate our understanding of individual processes with atmospheric measurements
Air pollution systems are non-linear– Need to establish the link between emissions sources and
ambient concentrations
5
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Air Quality Modeling Components
Meteorology modeling Emissions processing Initial/boundary conditions processing Photolysis rate processing Chemistry and transport modeling
6
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
WRF-ARW Basics Fundamentals of Numerical Weather Prediction
– Real vs. artificial atmosphere– Map projections– Horizontal grid staggering– Vertical coordinate systems
Definitions & Acronyms Flavors of WRF
– ARW core– NMM core
Other Numerical Weather Prediction Models– MM5– ARPS– Global Icosahedral
WRF Model Governing Equations– Vertical coordinate and grid discretization– Time integration– Microphysics
Current Defects of WRF
7
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Real vs. Artificial Atmosphere
FVkfdtVd
p
True analytical solutions are unknown!Numerical models are discrete approximations of a continuous fluid.
8
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Map Projections
Example of a regional high resolution grid
(projection of a spherical surface onto a 2D plane) nested within a global (lat,lon)
grid with spherical coordinates
x = r cos y = r Differences in map projections require
caution when dealing with flow of information across grid boundaries.
WRF offers polar stereographic, Lambert conformal, Mercator and rotated Lat-Lon map projections.
9
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Arakawa “A” Grid
Unstaggered grid - all variables defined everywhere.
Poor performance, first grid geometry employed in NWP models.
Noisy - large errors, short waves propagate energy in wrong direction, additional smoothing required.
Poorest at geostrophic adjustment - wave energy trapped, heights remain too high.
Can use a 2x larger time step than staggered grids.
10
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Arakawa “B” Grid
Staggered, velocity at corners. Preferred at coarse resolution. Superior for poorly resolved inertia-gravity
waves. Good for geostrophy, Rossby waves:
collocation of velocity points. Bad for gravity waves: computational
checkerboard mode. Used by MM5 model.
11
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Arakawa “C” Grid
Staggered, mass at center, normal velocity, fluxes at grid cell faces, vorticity at corners.
Preferred at fine resolution. Superior for gravity waves. Good for well resolved inertia-gravity waves. Simulates Kelvin waves (shoulder on boundary)
well. Bad for poorly resolved waves: Rossby waves
(computational checkerboard mode) and inertia-gravity waves due to averaging the Coriolis force.
Used by WRF-ARW, ARPS, CMAQ models.
12
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Arakawa “D” Grid
Staggered, mass at center, tangential velocity along grid faces.
Poorest performance, worst dispersion properties, rarely used.
Noisy - large errors, short waves propagate energy in wrong direction.
13
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Arakawa “E” Grid
Semi-staggered grid. Equivalent to superposition of 2 C-grids,
then rotated 45 degrees. Center set to translated (lat,lon) = (0,0) to
prevent distortion near edges, poles. Developed for Eta step-mountain
coordinate to enhance blocking, overcome PGF errors caused by sigma coordinates.
Controls the cascade of energy toward smaller scales.
Used by WRF-NMM and Eta models.
14
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
p
Continental orography
The vertical structure of a typical pressure vertical coordinatesystem (or p-system). The dashed lines represent constantpressure-levels ( p0 to p7 ). Note that p7 < p0 .
p7
p6
p5
p4p3p2p1p0
Pressure (p) coordinate system
Advantages:Density no longer appears in the pressure gradient force term:
The continuity equation is a pure diagnostic equation:
Disadvantages:Lower boundary of atmosphere is not a coordinate surface
FVkfd
Vdp
0p
Vp
15
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Sigma () coordinate system
43
2
1
0
The vertical structure of a typical sigma vertical coordinatesystem (or -system). The dashed lines represent constantsigma () levels. = 0 where p= 0 and = 1 where p=PS .
p
Continental orography
= P-Ptop / Ps-Ptop
Advantages:
Lowest sigma-level ( = 1) follows the topography.
Disadvantages:
Absence of constant pressure levels in the upper atmosphere.
Error in PGF along topography - near cancellation of large terms.
16
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Hybrid coordinate systems
4
3
2
1
0
The vertical structure of a typical hybrid vertical coordinatesystem (or -system). The dashed lines represent constant hybrid() levels.
p
Continental orography
p = AO + BPS ; PS = Surface pressure
Prescribed pre-defined conditions are required to determine thevalue of A O and B at each model level.
p-system: More suitable for upper
atmosphere simulations
-systems: More suitable for surface simulations
BPAp SO
17
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Definitions & Acronyms
WRF: Weather Research & Forecasting numerical weather prediction model
ARW: Advanced Research WRF [nee Eulerian Model (EM)] core
NMM: Nonhydrostatic Mesoscale Model core WRF-SI: Standard Initialization (4 components) -
prepares real atmospheric data for input to WRF WRF-VAR: Variational 3D/4D data assimilation
system (not used for this class) IDV: Integrated Data Viewer - Java application for
interactive visualization of WRF model output
18
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Flavors of WRF (ARW) ARW solver (research - NCAR, Boulder, Colorado)
– Fully compressible, nonhydrostatic equations with hydrostatic option– Arakawa-C horizontal grid staggering– Mass-based terrain following vertical coordinate
• Vertical grid spacing can vary with height• Top is a constant pressure surface
– Scalar-conserving flux form for prognostic model variables– 2nd to 6th-order advection options in horizontal &vertical– One-way, two-way and movable nest options– Runge-Kutta 2nd & 3rd-order time integration options– Time-splitting
• Large time step for advection• Small time step for acoustic and internal gravity waves• Small step horizontally explicit, vertically implicit• Divergence damping for suppressing sound waves
– Full physics options for land surface, PBL, radiation, microphysics and cumulus parameterization
– WRF-chem under development: http://ruc.fsl.noaa.gov/wrf/WG11/
19
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Flavors of WRF (NMM) NMM solver (operational - NCEP, Camp Springs, Maryland)
– Fully compressible, nonhydrostatic equations with reduced hydrostatic option – Arakawa-E horizontal grid staggering, rotated latitude-longitude – Hybrid sigma-pressure vertical coordinate– Conservative, positive definite, flux-corrected scheme used for horizontal and
vertical advection of TKE and water species– 2nd-order spatial that conserves a number of 1st-order and quadratic
quantities, including energy and enstrophy– One-way, two-way and movable nesting options– Time-integration schemes: forward-backward for horizontally propagating fast
waves, implicit for vertically propagating sound waves, Adams-Bashforth for horizontal advection and Coriolis force, and Crank-Nicholson for vertical advection
– Divergence damping & E subgrid coupling for suppressing sound waves– Full physics options for land surface, PBL, radiation, microphysics (only
Ferrier scheme) and cumulus parameterization– Note: Many ARW core options are not yet implemented! Nesting still under
development– NMM core will be used for HWRF (hurricane version of WRF), operational in
summer of 2007
20
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Other NWP Models (MM5) MM5 (research - PSU/NCAR, Boulder, Colorado)
– Progenitor of WRF-ARW, mature NWP model with extensive configuration options
– Support terminated, no future enhancements by NCAR’s MMM division
– Nonhydrostatic and hydrostatic frameworks– Arakawa-B horizontal grid staggering– Terrain following sigma vertical coordinate– Unsophisticated advective transport schemes cause dispersion,
dissipation, poor mass conservation, lack of shape preservation – Outdated Leapfrog time integration scheme– One-way and two-way (including movable) nesting options– 4-dimensional data assimilation via nudging (Newtonian
relaxation), 3D-VAR, and adjoint model– Full physics options for land surface, PBL, radiation,
microphysics and cumulus parameterization
21
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Other NWP Models (ARPS) ARPS (research - CAPS/OU, Norman, Oklahoma)
– Advanced Regional Prediction System– Sophisticated NWP model with capabilities similar to WRF– Primarily used for tornado simulations at ultra-high (25 meter)
resolutions and assimilation of experimental radar data at mesoscale
– Elegant, source code, easy to read/understand/modify, ideal for research projects, very helpful scientists at CAPS
– Arakawa-C horizontal grid staggering– Currently lacks full mass conservation and Runge-Kutta time
integration scheme– ARPS Data Assimilation System (ADAS) under active
development/enhancement (MPI version soon), faster & more flexible than WRF-SI, employed in LEAD NSF cyber-infrastructure project
– wrf2arps and arps2wrf data set conversion programs available– http://www.caps.ou.edu/ARPS/arpsdownload.html
22
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Global Icosahedral Model
23
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
WRF Model Governing Equations(Eulerian Flux Form)
∂U/∂t + ( · Vu) − ∂(pφ∇ η)/∂x + ∂(pφx)/∂η = FU
∂V/∂t + ( · Vv) − ∂(pφ∇ η)/∂y + ∂(pφy)/∂η = FV
∂W/∂t + ( · Vw) − g(∂p/∂η − μ) = F∇ W
∂Θ/∂t + ( · Vθ) = F∇ Θ
∂μ/∂t + ( · V) = 0∇
∂φ/∂t + μ−1[(V · φ) − gW] = 0∇
Momentum:
Potential Temperature:
Continuity:
Geopotential Height:
∂φ/∂η = -μ
where:μ = column massV = μv = (U,V,W)Ω = μ d(η)/dtΘ = μθ
Diagnostic Hydrostatic (inverse density ):
24
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
WRF Vertical Coordinate
25
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Vertical Grid Discretization
26
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Φ∗ = Φt + t/3 R(Φt )
Φ∗∗ = Φt + t/2 R(Φ∗)
Φt+t = Φt + t R(Φ∗∗)
“2.5” Order SchemeLinear: 3rd orderNon-linear: 2nd order
Runge-Kutta Time Integration
Square Wave Advection Tests:
27
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Runge-Kutta Time Step Constraint RK3 is limited by the advective Courant number (ut/x) and the
user’s choice of advection schemes (2nd through 6th order) The maximum stable Courant numbers for advection in the RK3
scheme are almost double those in the leapfrog time-integration scheme
Maximum Courant number for 1D advection in RK3
Time Scheme
Spatial order
3rd 4th 5th 6th
Leapfrog Unstable 0.72 Unstable 0.62
RK2 0.88 Unstable 0.30 Unstable
RK3 1.61 1.26 1.42 1.08
28
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Microphysics Includes explicitly resolved water vapor, cloud and precipitation processes Model accommodates any number of mixing-ratio variables Four-dimensional arrays with 3 spatial indices and one species index Memory (size of 4th dimension) is allocated depending on the scheme Carried out at the end of the time-step as an adjustment process, does not
provide tendencies Rationale: condensation adjustment should be at the end of the time step
to guarantee that the final saturation balance is accurate for the updated temperature and moisture
Latent heating forcing for potential temperature during dynamical sub-steps (saving the microphysical heating as an approximation for the next time step)
Sedimentation process is accounted for, a smaller time step is allowed to calculate vertical flux of precipitation to prevent instability
Saturation adjustment is also included
29
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
WRF Microphysics Options Mixed-phase processes are those that result from the interaction of ice
and water particles (e.g. riming that produces graupel or hail) For grid sizes ≤ 10 km, where updrafts may be resolved, mixed-phase
schemes should be used, particularly in convective or icing situations For coarser grids the added expense of these schemes is not worth it
because riming is not likely to be resolved
Scheme Number of Moisture Variables
Ice-Phase Processes
Mixed-Phase Processes
Kessler 3 N N
Purdue Lin 6 Y Y
WSM3 3 Y N
WSM5 5 Y N
WSM6 6 Y Y
Eta GCP 2 Y Y
Thompson 7 Y Y
30
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Current Defects of WRF Serious deficiencies in PBL parameterizations and land surface models
produce biases/errors in the predicted surface and 2-meter temperatures, and PBL height. WRF cannot maintain shallow stable layers.
3D/4D Variational data assimilation and Ensemble Kalman Filtering (EnKF) still under development, EnKF available to community from NCAR as part of the Data Assimilation Research Testbed (DART).
Not clear yet what to do in “convective no-man’s land” – convective parameterizations valid only at horizontal scales > 10 km, but needed to trigger convection at 5-10 km scales.
Multi-species microphysics schemes with more accurate particle size distributions and multiple moments should be developed to rectify errors in the prediction of convective cells.
Heat and momentum exchange coefficients need to be improved for high-wind conditions in order to forecast hurricane intensity. Wind wave and sea spray coupling should also be implemented. Movable, vortex-following 2-way interactive nested grid capability has recently been incorporated into the WRF framework.
Upper atmospheric processes (gravity wave drag and stratospheric physics) need to be improved for coupling with global models.
31
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Processing
Area Mobile Point Biogenic
Emissions Processing Steps
AQM-ready Emissions:
32
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Terminology
Inventory: estimate of pollutant emissions at a given spatial unit
Model Grid: 3-d representation of the earths surface based on discrete and uniform spatial units, i.e. grid cells
Speciation: conversion of inventory pollutant species to model pollutant species
Gridding: conversion of inventory spatial units to model grid cells
Temporalization: conversion of inventory temporal units to those requires by an air quality model
33
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Terminology
Plume rise: calculation of the vertical distribution of emissions from point sources and the subsequent allocation of the emissions to the model layers
Spatial surrogate: GIS-based estimate of the fraction of a grid cell covered by a particular land-use category (e.g. population or rural housing)
Profiles: emissions distributions in space, time, or to chemical species
Cross-referencing: relating profiles to specific emissions sources
34
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Area sources
Most basic inventory unit Country/province/municipality wide estimate Requires spatial surrogates to map to a model
grid Examples
– Construction and agricultural emissions– Road dust– Fires
35
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Mobile sources
On-road: Estimate by road way and vehicle type– Requires emissions factors for local vehicles and activities/speeds for
local roads– Gridding by road way distribution or links– Can use local meteorology to adjust emissions factors for temperature
and humidity– Examples: Heavy-duty diesel trucks on primary highways, light-duty
gasoline cars on rural roads
Non-road: Area-like estimates– Examples: Construction and mining vehicles, recreational vehicles
(boats, ATV’s),
36
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Point sources
Emissions at specific latitude-longitude coordinates Often elevated sources that require stack parameters (e.g.
stack height, exit gas velocities, exit gas temperatures, etc.) Can use annual, daily, or hourly emissions estimates Examples:
– Electricity generating units (EGU’s)– Smelters– Wildfires
37
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Biogenic sources
Estimates of emissions from vegetation and soils Uses gridded land-use data and emissions factors by
vegetation type Uses local meteorology to calculate emissions based on
photosynthetically active radiation (PAR) and to adjust for temperatures
Examples:– VOC emissions from specific tree species– Soil NO
38
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Gridded sources
Pre-gridded emissions from global databases Normalize to the model grid to combine with
other sources Can encompass any of the emissions categories Top-down vs. bottom-up emissions estimate
39
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Processing
Purpose: convert emissions data to formats required by air quality model
Primary functions– Import data into system– Spatial allocation (gridding)– Chemical allocation (speciation)– Temporal allocation– Merge– Quality assurance
40
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Processing Steps Data Import
– Inventory categories
• Area
• Point
• Mobile
• Biogenic
• Gridded– ASCII or gridded binary– Country/state/county estimates– Annual estimates– Pollutants include bulk VOC and
PM2.5
Spatial Allocation– Inventory spatial units model
grid– Requires spatial surrogates
EI Grid Cell Mapping
41
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Processing Steps Chemical Allocation
– Tons Moles– Converts inventory
pollutants to air quality model species
– Model-dependent speciation profiles
– NOx NO + NO2
– VOC PAR, OLE, etc.– PM2.5 NO3, SO4, etc.
Temporal allocation– Inventory units
hourly emissions– Requires temporal
profiles
0
0.02
0.04
0.06
0.08
0.1
0.12
J F M A M J J A S O N D
% E
mis
sion
s0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
M T W Th F Sa Su
% E
mis
sion
s
0
0.01
0.02
0.03
0.04
0.05
0.06
% E
mis
sion
s
Monthly Weekly
Diurnal
42
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Processing Steps
Merging– Combine all
intermediate steps to create AQM-ready emissions
– Combine individual source categories
– Formatting– Units– Output file naming
Quality Assurance– Report base inventory
values and changes at each processing step
– Customize reporting
• e.g. by state and SCC, by SCC and temporal profiles, by grid cell and surrogate I.D.
– Means to determine why a result occurred
43
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Other Emissions Processing Steps
Plume Rise– Allocate elevated
emissions sources to vertical model layers
– Compute layer fractions
– Require meteorology to calculate plume buoyancy
– Stationary point, fires, in-flight aircraft
Projections– Grow and/or control
inventories for future year modeling
– Source-based projection information
44
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Emissions Processing Paradigms
Linear– Sequential steps that follow a particular order– Requires completing one step before completing the next
Parallel– Flexible sequence with steps in any order
Import Grid Speciate Temporal Merge
Import
Grid
Speciate
Temporal
Merge AQM
AQM
45
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Initial and Boundary Conditions
Initial conditions define the chemical conditions at the start of a simulation
– Defined using vertical profiles of clean background concentrations
Boundary conditions define the chemical conditions on the horizontal faces of the modeling domain
– Static and dynamic boundaries are possible Initial conditions decay exponentially with
simulation time; boundary conditions on the upwind boundary continue to affect predictions through an entire simulation.
46
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Initial/Boundary Condition Processing
Processing requires generating IC/BC estimates on a model-grid
Nested simulations extract BC’s from a parent grid
Multi-day simulations extract IC’s from the last hour of the previous day
47
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Photolysis Rate Processing
Photolysis: Chemical dissociation caused by the absorption of solar radiation
Photolysis rate: rate of reaction for pollutants that undergo photolysis
Processing calculates clear sky photolysis rates at different latitudes and altitudes
Air quality models adjust rates with cloud cover estimates from meteorology
J = ∫ 4πIλ,T Aλ,T QYλ,T dλ ∞
0
48
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
What are air quality models?
Statistical models: describe concentrations in the future as a statistical function of current chemical and/or meteorological conditions
Chemistry-transport models (CTM): based on fundamental descriptions of physical and chemical processes in the atmosphere
49
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
How do CTMs work? Air Quality Model processes
– Dynamical/thermodynamical: meteorology, land surface conditions (soil, water)
– Transport: emissions, advection, diffusion, dry deposition, sedimentation
– Gas phase chemistry: photochemistry, phase changes
– Radiative: optical depth, visibility, energy transfer– Aerosol/clouds: nucleation, coagulation,
heterogeneous chemistry, aqueous chemistry
50
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Overlay 3-D boxes on a grid
Lagrangian/Trajectory Models
Moves relative to the coordinate
Different locations at different times
Only emissions enter the cell
No material leaves the cell
51
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Overlay 3-D boxes on a grid
Eulerian Models
Fixed relative to the coordinate
All locations at all times
Materials move through all cell faces*
52
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Conceptual approach to CTMs
Extend the 2-D box model to three dimensions
∆x
∆y
u1C1 u2C2
∆y
∆x ∆zu2C2u1C1Ci Ci
2-D 3-D
u = wind vector
Ci = concentration of species i
Basic Continuity Equation (flux in 1 direction):
∆C ∆x ∆y ∆z = u1C1∆y ∆z ∆t – u2C2 ∆y ∆z ∆t
Divide by ∆t and volume: ∂C/ ∂t = - ∂(uC)/ ∂x
53
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Expanded Continuity Equation
∆y
∆x ∆z
E
u2C2u1C1
Ri
3-D
S
u,v,w = wind vectors
E = emissions
S = loss processes
Ri = Chemical formation of species I
D = Molecular diffusion coefficient
Expanded Continuity Equation Derivation:
Expand to flux three dimensions:
∂C/ ∂t = - ∂(uC)/ ∂x - ∂(vC)/ ∂y - ∂(wC)/ ∂z
= - ∙ (vC) (flux divergence form)
Add additional production and loss terms:
∂C/ ∂t + ∙ (vC) = D 2 C + R + E - S
w1C1
w2C2
X v1C1
54
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Advection: - ∙ (vC)
Boundary Conditions
Aerosol Processes:Gas Chemistry: Aqueous
Chemistry:
Remis
Rdep
Rchem
Rwash
Rnuc + Rc/e + Rdp/s + Rds/e + Rhr
Rhr
Diffusion: D 2 C
R = Rate
chem=chemical production/loss hr=heterogeneous reactions nuc=nucleation c/ev=condensation/evaporation dp/s=depositional growth/sublimation ds/e=dissolution/evaporation wash=washout dep=deposition emis=emissions
55
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Full Continuity Equations Gas Continuity Equation
∂C/ ∂t + ∙ (vC) = D 2 C + Rchemg + Remisg + Rdepg+ Rwashg + Rnucg + Rc/eg + Rdp/sg + Rds/eg + Rhrg
Particle Continuity Equation (number)
∂n/ ∂t + ∙ (vn) = D 2 n + Remisn + Rdepn+ Rsedn + Rnucn
+ Rwashn + Rcoagn
Particle Continuity Equation (volume concentration)
∂V/ ∂t + ∙ (vV) = D 2 V + Remisv + Rdepv+ Rsedv + Rnucv
+ Rwashv + Rcoagv+ Rc/ev + Rdp/sv + Rds/ev
+ Regv + Rqgv + Rhrv
56
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
CTM Coordinate Systems
Convert all motion equations from Cartesian to spherical coordinates
Horizontal grids typically on the order of 1 to 36 km
Recent applications extending to 500m and 108 km
Lambert conformal, polar stereographic, and Mercator are the most common modeling projections
dx = (Recosφ)dλe dy = Red φ
57
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
CTM Coordinate Systems
Vertical grids extend from the surface to 10 km Altitude coordinate: layers are defined as
surfaces of constant height with variable pressure
Pressure coordinate: layers are defined as surfaces of constant pressure with variable height
Sigma-pressure coordinate: layers defined as surfaces of constant σ, where
ppaa – p – pa,topa,top
pa, surf - pa,topσσ = = (0,1)
58
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Boundary Layer Processes Boundary layer more difficult
to model because of greater turbulence and larger emissions forcing terms than in free troposphere; land surface interactions and planetary boundary layer (PBL) dynamics dominate
Surface temperature and soil moisture affect energy and moisture flux; affect mixing heights, winds, and pollutant concentrations
59
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Modeled cloud processes
Radiative transfer: reflecting, scattering, and trapping heat
Atmospheric component of the hydrologic cycle
Wet deposition of gases and particles
Medium for aqueous phase chemistry
Vertical transport/convective mixing
Energy balance: temperature effects and photolysis rates
Aerosol Processes Condensation
Evaporation
Gas Phase
pA pA(a) A(a) A(r)
B(a) B(r)
pC pC(a) C(a) C(r)
Water Droplet
Aerosol particle
New aerosol particle
60
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Energy/Radiative Effects Visibility
Optical depth scattering and absorption between top of atmosphere and altitude x
Photolysis rates
Radiative transfer
Sun
Single scattering
Multiple scattering
Direct
Beam
DiffuseΘs
Θ(μ,Θ) Fs (-μ,Θ)
dI/dx = σ bIB - σ ext I
σ = extinction coefficient
I = visible radiance
J = ∫ 4πIλ,T Aλ,T QYλ,T dλ ∞
0
4πI = actinic flux A = absorption cross section QY = quantum yield
61
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Gas Phase Chemistry
Thousands of different organic and inorganic gases react to form smog and PM
Gas phase chemistry is a “stiff” system
Parameterized chemistry mechanisms represent the system with a few surrogate organic pollutants
Surrogates based on molecular or atomic structures of pollutants
Carbon bond lumpingPropane = 3 PAR:
H3C-CH2-CH3
1-Butene = 2 PAR + 1 OLE:
H2C=CH-CH3-CH3
– averaged reaction rates
Molecular lumping– Surrogates represent
similarly reactive species
– Explicit or averaged reaction rates
62
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Gas Phase ChemistryImportant Inorganic Reactions
NO + O3 NO2 + O2
NO2 + hv NO + O (λ<420 nm)
O + O2 + M O3 + M
O3 + hv O2 + O(1D) (λ<310 nm)
O3 + hv O2 + O3P (λ>310 nm)
O(1D) + H2O 2OH
OH + O3 HO2 + O2
OH + NO HONO
HONO + hv OH + NO (λ<400 nm)
NO2 + OH HNO3
HNO3 + hv NO2 + OH (λ<335 nm)
Important Organic Reactions
(methane example)
OH + CH4 H2O + CH3∙
CH3∙ + O2 CH3O2∙
CH3O2∙ + NO NO2 + CH3O∙
CH3O∙ + O2 HO2 + HCHO
CH3O2∙ + HO2 O2 + CH3O2H
CH3O2H + hv CH3O∙ (λ<360 nm)
CH3O2H + OH H2O + CH3O2∙
63
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Gas Phase Chemistry
Key reaction sequence for smog
ROG∙ + NO NO2 + ROG∙∙
NO + O3 NO2 + O2
NO2 + hv NO + O
O + O2 + M O3 + M
64
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Gas Phase Chemistry Kinetics
aA + bB eE + fF Rate = kr[A]a[B]b
Rate constant calculations
d[A]t/dt = -kF[A]t = -kS[A]t[B]0 = -kT[A]t[B]0[C]0
1st order: A D + E kF = -(1/t) ln [A]t/[A]0
2nd order: A + B D + E kS = -(1/[B]0t) ln [A]t/[A]0
3rd order: A + B + C D + E kT = -(1/[B]0[C]0t) ln [A]t/[A]0
Arrhenius equation for temperature dependence
kr = Ar(300/T)Bexp(Cr/T)
Troe equation for temperature and pressure dependence
kr = {(k∞,T k0,T [M])/(k∞,T+k0,T[M])}Fc [1+(logk0[M]/k∞)^2]^-1
Photolysis rate equation
J = ∫ 4πIλ,T Aλ,T QYλ,T dλ ∞
0
65
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Aqueous Chemistry
Gases equilibrate with the aqueous phase by Henry’s law:
[A(aq)] = HApA
Dissolved gases react in solution to form new compounds
Sequence: droplet formation gases dissolve in droplet chemical reactions evaporation
66
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Aerosol Dynamics
Size distribution: ratio of # aerosols in a diameter range to the size of the range; discrete function of the number of particles
Ni = ni∆Dp
Number distribution: continuous function of the diameter of the particles
N = ∫ nN(Dp)dDp
Aerosol moments: properties of the distributions (e.g. mean, variance)
∞
0
67
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Aerosol Dynamics
Mass transfer solutions to transport mass between gas/aqueous/solid phases
Solve gravitational settling, diffusion, and advection for moving particles around
4 classes of nucleation for particle formation:– Homogenous Homomolecular– Homogenous Heteromolecular– Heterogenous Homomolecular– Heterogenous Heteromolecular
68
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
CTM Setup and Configuration All CTM’s are free and open source Compile on UNIX/Linux with Fortran Script interfaces for compiling and running Begin with download, installation, and compilation Set up computing environment
– I/O directories– Check for processor availability and disk space
69
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
CTM Setup and Configuration
CTM’s are at the end of a long sequence of preprocessing steps
– Prepare meteorology inputs with research and forecast met models
– Prepare emissions inputs with specialized emissions processors
– Prepare initial and boundary condition inputs with preprocessors packaged with CTM
– Generate clear sky photolysis rates with a preprocessor packaged with CTM
70
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
CTM Process Schematic
MetModeling
Emissions Modeling
IC Preparation
BC Preparation
PhotolysisRates
2-D/3-D Met Files
Model Ready Emissions
IC Input File
BC Input File
Clear SkyJ Rates
CTM Modeling
HourlyConcentrations
CumulativeWet Dep
CumulativeDry Dep
VisibilityMetrics
DiagnosticOutputs
71
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Modeling Conventions
Establish base case model performance Simulate sensitivities off the base case Select episodes or time periods that illustrate the
problem being addressed– O3 episodes– PM episodes– High flow regimes/transport scenarios– Annual episodes for one-atmosphere modeling
72
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
One Atmosphere Approach
Changing paradigm from multiple models that address individual process to a single unified model
Conceptually more realistic: all atmospheric processes are coupled
In practice very difficult because of confounding errors
“Right answer for the wrong reasons”
73
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
What is the right answer for CTMs?
Evaluation techniques – Comparisons to ambient measurements– Sanity checks– Looking for known trends (diurnal/seasonal patterns,
chemical signatures (ratios) Comparison to measurements
– Paired in space and time– Compare predicted vs observed maximums– Paired in space but not in time– Statistical metrics include paired/unpaired peak
prediction accuracy, mean normalized bias, mean error
74
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Problems in air quality modeling
Inconsistencies in spatial scales and speciation when comparing models to measurements
Incomplete measurement database (PBL, radiation budget, short lived pollutants, observations aloft)
Huge uncertainties in all input data
Garbage in = Garbage out
75
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Problems in air quality modeling
Met and chemistry models are tuned for certain conditions– Meteorology models generally don’t work well
under stagnant, low flow conditions– Chemistry models break down at night and
during background ambient conditions Incomplete science
– For various reasons, some important atmospheric processes are either not represented at all or are using crude approximations
76
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Significant CTM Studies
Regional Ozone Model: early 1980’s, first regional scale model, early studies on regional transport
Regional Acid Deposition Model: mid 1980’s, 1st multi-pollutant model, predecessor to current modeling systems
National Acid Precipitation Assessment (NAPAP): 1980’s, established links between S emissions in the Midwest to acid rain in the Northeast; first modeling studies of emissions trading programs
77
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Significant CTM Studies Ozone Transport
Assessment Group (OTAG): mid 1990’s, established multi-scale problem of ozone, relationships between 1-hr vs. 8-hr ozone standard and transport, prevailing regional conditions for poor air quality, weekday-weekend trends in air quality
78
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Model Application ExamplesRegional Ozone Sensitivities - NCDAQ
Bas
e Ju
ne,
1996
Acr
oss
the
boar
d A
rea,
M
obile
Red
uctio
ns
Com
bine
d A
rea,
Mob
ile,
Poi
nt R
educ
tions
Com
bine
d A
rea,
Mob
ile,
Poi
nt R
educ
tions
79
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Model Application Examples
Intercontinental transport of pollutants
80
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Model Application Examples
Continental one atmosphere modeling
1 day, 24-hour average ozone
Observations overlaid on plot
81
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Model Application Examples
Summer, 2002 Sulfate Winter, 2002 Nitrate
Continental one atmosphere modeling
82
Air Quality Modeling
University of North Carolina Community Modeling and Analysis System
Future Directions
Quantify model uncertainties Expand the ability of the models to represent and
integrate all atmospheric processes 2-way coupling between global, regional, and
neighborhood scale models Source apportionment technologies Couple with other media/disciplines
(water,soil,risk, economics) Community/open source development