AMAD Strategic Plan: Introduction

AIR QUALITY MODELING
Kenneth Schere, Prakash Bhave, Roger Brode
CMAS Conference
October 13, 2010
Chapel Hill, NC
Office of Research and Development
National Exposure Research Laboratory
Laboratory, Atmospheric Modeling and Analysis Division
Major Types of Air Quality Models
Statistical
Dispersion
Grid-Based
Hybrid
Attributes
• Dependent on
observations
• Steady-state plume
or puff models
• Gaussian
distributions
• Dynamic
• Volume
averages
• Complex
chemistry
• Combined
dispersion/grid
models
Space
Scales
• Variable, depending
upon obs network
• Generally, urban or
metropolitan
• Mostly source-to
urban-scale
• up to 50 km from
source
• Urban to global
• Grid resolutions
from 1 to 50 km
• Source to
global scale
Time
Scales
• Hourly or longer
averaging times
• Hourly or longer
averaging times
• Hourly or longer
averaging times
• Daily to multiannual
integrations
• Hourly or
longer
averaging
times
• Daily to
multi-annual
integrations
Examples
• Land-use or other
regression
• Neural networks
• AERMOD
• ISC
• CMAQ
• CAMx
• CMAQAERMOD
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
grid
cell
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Statistical Models
• Depend upon local air quality observations
• e.g., land-use regression models
– Rely on density of observations and land-use data
to define air quality gradients
• Specific to a given area and time
• Cannot be generalized or extrapolated
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Emissions-Based Models
• Independent of local measurements
• Can be generalized for applications in space
and time
• Can be extrapolated to future conditions
• Subject to difficult-to-quantify errors and
biases
– Emissions, meteorology,
computational/process algorithms
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Dispersion Models
• e.g., AERMOD and others
• Local-scale modeling
– Single-source or source complex; near-field (< 50 km range)
– Neighborhoods
• Explicit parameterizations of local turbulence and
dispersion
• Requires on-site meteorological data or meteorological
modeling
• Requires local source emissions
• Generally used for passive gases and aerosols
• Uses a specified receptor grid for concentration
estimates
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Model-to-Monitor Comparison – Atlanta (JST)
Grid-Based Air Quality
Modeling Systems
• e.g., CMAQ, CAMx, WRF-Chem, among others
• System of linked models
– meteorology  emissions  air quality
• Scalable
– global  continental  regional  urban
• Variability of concentration estimates increase with
increasing model resolution
• Applied to passive and reactive trace gases and
aerosols
• Used fixed 3-D grid system for concentration estimates
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Atlanta NOx Emissions
4-km grids
1-km grids
Grid-Based Air Quality
Modeling Systems Limitations
• Volume average estimates; not points
• Complex systems of models
– Subject to greater parametric uncertainty
– Applications are resource-intensive
• Model trouble-shooting/ diagnostics can be difficult
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Hybrid Modeling
Systems
• Combines strengths of different types of models
• Examples:
– Plume-in-grid techniques
• Regional grid models + subgrid plume enhancement
for major point sources (e.g., CAMx + P-in-G;
CMAQ+APT)
– Linked models
• CMAQ for regional characterization + AERMOD for
local hotspots (e.g., benzene; NO2)
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Specialized Models – e.g. near-road
600 K
300 K
Tailpipe
Tailpipe-to-Road
300 K
~ 100 m from
roadway
Curbside
Road-to-Ambient
Plume Processing
Ambient
Background
~ km
Ambient
Processing
Grid-level
Road-level
Tailpipe-level
Emission:
Emission:
Emission:
TheThe
emission
emission
profiles
profiles
on or
near
near
the
near
the
end
the
exit
ofroadway
plume
of the processing
tailpipe
curb
(particle
dynamics slows down significantly at this point)
Zhang, K. M., A. S. Wexler, et al. (2005). "Evolution of particle number distribution near roadways. Part III: Traffic analysis and on-road size resolved particulate
emission factors." Atmospheric Environment 39(22): 4155-4166.
Modeling the Particle Size Distribution Near a
Roadway (K.M. Zhang et al., Atmos. Env. 2004)
Modeled with Dilution Only
Full Particle Dynamics Model
Particle dynamics (e.g., condensation & evaporation) are important. When those
processes are neglected, size distribution is simulated poorly.
Model Evaluation
• A necessary step in building confidence in a model
application
• Requires observed data at appropriate spatial/temporal
scale as the model
– Observed data need to be characterized in terms of
uncertainties and representativeness
• Confidence in model results increase as the rigor of the
evaluation increases
– Operational  diagnostic  dynamic
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Closing Comments
• Air quality models are most adept for assessing
relative changes
– i.e., how do ambient concentrations change as a
result of changes in meteorology, climate, land-use,
emissions, etc.
• The larger the spatial and temporal scales of model
integration, the more confidence in the model results
– Absolute predictions at a particular place and time
are the most uncertain model estimates
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory
Closing Comments
• Models can predict the local concentration gradients
based on the given emissions distribution
• Probabilistic use of model results
– Model estimates as concentration distributions
– Multi-model ensembles
• Combined use of models and observations for
assessments is optimal
Office of Research and Development
Laboratory, Atmospheric Modeling and Analysis Division
National Exposure Research Laboratory