Particulate Matter Modeling: Including Nanoparticles · reaction – Sensitive to temperature K T K=[HNO3(g)][NH3(g)] Georgia Institute of Technology Diurnal Nitrate Pattern Source:

Georgia Institute of Technology

Particulate Matter Modeling:Including Nanoparticles

Ted RussellAir Resources Engineering Center

Georgia Tech

April 9, 2001


Issues

• Much of the suspected health and welfare effects from air pollution due to particulate matter– Health (the main concern)

• Morbidity and mortality concerns• Asthma• etc.

– Welfare• Visibility• Deposition

• Particulate matter modeling has significant challenges– Modeling techniques in development– input/output uncertainties impact model

evaluation


Outline• Air quality modeling

– Role– Scientific foundation– Model vs. process

• Particulate matter models– State of the science– Current research directions

• Conclusions


Role of Air Quality Modeling In Air Quality Management

Emissions

Chemistry

Air Quality/Health Impacts

Controls Pollutant Distributions

Air Quality Model

Meteorology

Air Quality Goals


Air Quality Model

• Representation of physical and chemical processes – Numerical integration

routines• Scientifically most sound

method to link future emissions changes to air quality

∂c∂t + L(x, t)c = f(x, t)

c(t+2∆t) = Lx(∆t) Ly (∆t) Lcz(2∆t) Ly(∆t) Lx(∆t) c(t)

ComputationalPlanes

5-20

50-200

Air Quality Model

100 species x 5000 hor. grids x 10 layers= 5 million coupled, stiff non-linear differential equations

∂∂ct

c c R Sii i i i= −∇ + ∇ ∇ + +( ) ( )u K

Atmospheric Diffusion EquationDiscretize

Operator splitting

50-100

Emissions

Chemistry

Meteorology

NumericsC=AxB+E


Air Quality Model

Air Quality

Temperature Radiation

CloudCover Wind

Emissions1. Anthropogenic2. Geogenic3. Biogenic

Sources

TransportedPollutants

Sources (E,S,BCs)

GeographicalFeatures

Transport(U, K, Vd)

Turbulence

NumericalSolution

Techniques

SurfaceDeposition Sink Processes

Topography &Land use

PhotochemicalReactions

ThermochemicalReactions

HomogeneousProcesses

HeterogeneousProcesses

Computed Concentrations

Meteorology

Aerosol Dynamics

ChemicalProcesses

(R)

Chemistry and Aerosol Dynamics


Evolution inAir Quality Model Development

1950 1960 1970 1980 1990 2000

Mod

el C

ompr

ehen

sive

ness

Gau

ssia

npl

ume

Phot

oche

mic

al b

ox(..

.EK

MA

in 1

977)

Mul

tidim

ensi

onal

Ph

otoc

hem

ical

Aerosols

PiG

Nes

ted

Mod

els

Mod

els

3(?)

Golden era MaturationInfancy

Adv

ance

d di

agno

stic

feat

ures


Evolution inAir Quality Model Application

1950 1960 1970 1980 1990 2000

Mod

el C

ompr

ehen

sive

ness

App

licat

ion

ofG

auss

ian

plum

efo

r so

urce

impa

ct

Golden era?Infancy

Und

erst

andi

ng

ozon

e ch

emis

try

Air

qua

lity

plan

ning

Und

erst

andi

ng u

rban

and

re

gion

al o

zone

dyn

amic

s

Und

erst

andi

ng

PM d

ynam

ics

*or denial

Regulatory focusScience focus

Qua

ntify

ing

unce

rtai

nty


Important Milestones

1970 1980 1990 2000

Second generationAQMs

Aerosol thermodynamics

Use of prognosticmet fields

Prognostic met fields with FDDA

Identifiedemissions

errors major

Organic aerosoldynamics

CAAA call for 3D AQM

Biogenicsimportant

CB Series ofchem. mechanisms

SAPRC Series ofchem. mechanisms

EKMA

RRS: Start of the 3D & UAM era

OTAG:UAM-V

NAPAP:RADM

Externalaerosol


How Good Are They?

• All evidence suggests that they describe the processes most affecting the evolution of ozone and (if equipped) particulate matter (o.k., many components of PM) after pollutant emission

• Current limitations– Input errors

• Emissions• Meteorology

– Monitoring data• Sparse, ground level

– Don’t effectively use upper-air data– Model components/formulation/design


Particulate Matter Dynamics• Particulates are exceptionally complex

– Complete description must include size and chemical composition

• Continuous size distribution• Chemical composition varies continuously with

size• Phase conversion important

– PM function concentration function is more complex:

• C(x,t,dp): space, time and particle diameter– Composition may not be uniform for a

given size: C(x,t,dp, si): si is source i– Makes ozone modeling look really easy


Sources of Particulate Matter

• PM has both primary and secondary components– Primary

• Organic & elemental carbon (OC/EC), crustals, metals, water• Mobile sources, industry, utilities, dust

– Secondary• Sulfate, nitrate, ammonium and organic carbon• Utilities, mobile sources, industry, biogenic, fertilizer,

emissions control equipment


Particulate Sulfate Formation

SO2H2O2

O3

OH

Cloud and gas-phase processing:

Oxidants all impacted byNOx emissions

Emissions:Utilities, mobile sources

H2SO4

H2SO4

NH4HSO4

(NH4)2SO4

Condensation &ammonia

scavenging

NH3(g)


Particulate Nitrate & OC Formation

NH3(g)

O3+…+HNO3(g)+Org (pm)NOx

oxides of nitrogen(NO + NO2)

ROGreactive organic gases


PM Nitrate Formation

• Gaseous nitric acid formed from NOx emissions

• Ammonia derived directly from emissions

• Combine via equilibrium reaction– Sensitive to temperature K

T

K=[HNO3(g)][NH3(g)]


Diurnal Nitrate Pattern

Source: R. Weber


PM Nitrate Formation

• Nitric acid reacts with (free) ammonia– As ammonia emissions

increase, nitrate will increase until gas phase nitric acid depleted

– Sulfate reduces free ammonia/increases acidity, reducing nitrate formation

• Reducing NOx will decrease HNO3 formation, but may not decrease PM nitrate much

Aer

osol

Nitr

ate

SO2 NH3 NOx


Particulate Matter Dynamics

EC,Org

H+, NO3-,SO4=

Org, EC:SO2+H2O2==>

H2SO4

HNO3(g)SO2(g)

In dropletchemistry

ThermodynamicsMasstransfer

Coagulation

Internal Mixture

External Mixture

Semivolatileorganics


Particle size distribution

Particle diameter (dp)

Mas

s di

stri

butio

n (d

M/d

(dp)

)

Accumulationmode

Coarsemode

Nucleationmode

Aitkinmode

Sulfate fraction changes with size


Particle number distribution


Num

ber d

istr

ibut

ion

(dN

/d(d

p))

Accumulationmode Coarse

mode

Nucleationmode Aitkin

mode


Particulate Matter Modeling Approaches

• Size distribution– None– Sectional– Modal

• Gas-to-particle conversion– Inorganic– Organic

• Aerosol particle description– Internal mixture– External mixture


Size Distribution

• No description– All PM is together: no information on size or

composition as a function of size• Sectional

– Size distribution made up of a user defined number of bins

• Can have many bins (>20) or few (4)• Describes compositional changes as a function of size

• Modal– Size distribution made up of 2-4 “modes”

corresponding to modes in size distribution– Mode shaped like log-normal profiles


Sectional Approach

Aerosol Histogram

0

2

4

6

8

10

12

14

16

18

0.0 0.1 1.0 10.0 100.0

Diameter (micrometers)

dCi/d

logD

pi

(mic

rog

ram

s/m

3/lo

g(m

icro

met

ers)

)


Modal Approach


Mas

s di

stri

butio

n (d

M/d

(dp)

)

Sulfate fraction


Gas-to-Particle Conversion

• Gas phase species can condense upon, and volatilize from, particulate matter– Inorganics

• Thermodynamic equilibrium usually assumed– ISORROPIA– AIM

– Organics• Semivolatiles and low-vapor pressure organics• One and two-step approach

– One step: gas phase chemistry leads to condensable species which goes to particulate phase

– Two step: gas phase chemistry leads to semivolatile product that partitions between gas and condense phases


Organic Partitioning Coefficient

Temperature

Frac

tion

cond

ense

d

Yield=M0[αKi/(1+KiM0)]

Ki=1/Csat,i =RT/(γpimi)

Low vapor pressure


Internal vs.External Mixtures

• Particle composition can be very inhomogenous even in the same size distribution– Traditional approaches assume homogenous

mixture in each size range/bin: C(x,t,dp): • “Internal mixture”• See prior slides

– Theory and evidence suggests that particle composition varies within a size range

• Source-based differences: C(x,t,dp, si)


Internal Mixture

Sulfate

Nitrate

OC

EC

Metals

SulfateNitrateOCECMetals


External Aerosol Modeling

From Kleeman, Cass and Eldering, 1997


Nanoparticle Modeling• Nanoparticles represent an important fraction, in

terms of total number but not total mass, of PM, and are unique– Very short lifetimes– Directly emitted or, possibly, due to nucleation in short term

events

• Models have not dealt so much with nanoparticles because of their short lifetimes and small fraction of the mass

• Three approaches for modeling:– Sectional

• Can add multiple sections in the nano-modes– Modal

• Adding a new nucleation mode in addition to Aitkin mode– External, source oriented approach


Modal Approach to Nanoparticles

Particle diameter (dp, um)

Num

ber d

istr

ibut

ion

(dn/

d(d p

))

0.001 0.01 0. 1 1.0 10

Nucleation mode

Aitkin mode


Sectional Approach for Nanoparticles


Num

ber d

istr

ibut

ion

(dn/

d(d p

))

0.001 0.01 0. 1 1.0 10


Strengths and Weaknesses• Modal

– Strengths• Computationally efficient

– Weaknesses• Lack of detailed information (all nano-particles similar)

• Sectional– Strength: Tremendous capacity for detail, very flexible– Weakness: Computationally time consuming

• External mixture– Strengths: Tremendous detail, particles tied directly to

sources– Weakness: Computationally expensive

• All– Nucleation theory is highly uncertain

• Not a major factor if particles are primary in origin


Particulate Matter Sensitivity Analysis and Source Attribution

• AQM’s major function is to link source emissions to air quality: Source attribution– Individual vs. regional/category analysis

• Assessing impact of individual sources difficult– Small perturbation to noisy process– Small difference between two large numbers

» e.g.: 10 Ton/day source in a 1000 ton/day area(10 ton/day/1000 ton/day)*0.1(% change in O3/%change in NOx)*120 ppb=0.12 ppb

Can a model “see” this accurately?• Assessing categories/regions/complete strategies more

appropriate for typical approach if reductions are reasonable– Unrealistic changes to minimize noise raises additional

issues• New approach: direct sensitivity analysis


Source Attribution using Direct Sensitivity Analysis

3-DAir

QualityModel

NOo

NO2o

VOCio

...TKu, v, wEikiBCi...

O3(t,x,y,z)NO(t,x,y,z)NO2(t,x,y,z)VOCi(t,x,y,z)...

DDM-3DSensitivityAnalysis

s (t)c (t)piji

j=

∂∂

J

decoupled

∂∂

Rk

i

j

Concentrations

Response toemissions changes:accurate forvery smallchanges

Simultaneously


Response of Fine Nitrate to SO2 reductions


Source Attribution:Sulfate by Source Region

Atlanta Daily Average Sulfate Sensitivity(10% Reduction in SO2 Emissions)

0.01.02.03.04.05.06.07.08.0

7/11/95 7/12/95 7/13/95 7/14/95 7/15/95 7/16/95

Sul

fate

Red

uctio

n (%

)

WI NI SI EI AO

DirectSensitivityAnalysis


Particulate Matter Modeling and Chemical Mechanisms

• Current generation of gas-phase mechanisms (e.g., SAPRC99+, RACM) in pretty good shape for ozone– Flexible– Evolutionary– Appear to adequately describe gas phase kinetics for ozone,

etc– Limited information for determining organic composition

of PM• Important information for identifying sources and impacts lost

• Aqueous phase mechanisms – Likely adequate for inorganics and ozone– Questions about organic oxidation


PM Modeling State of the Science: Where are We?

• Ozone models are “mature”

• PM Models still evolving

• “One atmosphere”/“3rd generation” urban-to-regional models are at the forefront– Combined

gas/aerosol/deposition & nested/multiscale

– Some built in diagnostic features

• Sensitivity analysis

Regional Multiscale Model


Attributes of Advanced Models:Internal Mixture Models

• Usual attributes of advanced internal mixture models– Advanced chemical mechanism– Sectional or modal approach– Thermodynamic inorganic– One-step organic formation

• Two step on the way, but large uncertainties– Advanced diagnostic features– Examples: URM, CMAQ, CIT-AERO, UAM-AERO

• URM extensively evaluated over eastern US as part of SAMI• CMAQ is to become the community model


Attributes of Advanced Models:External Mixture Models

• Features– Limited applications to date– Very time and resource consuming– AIM thermodynamics/growth– Trajectory and grid-based versions of CIT model

• See Cass, Kleeman and co-workers

– Expect wider application in next 10 years


Example Model Application: SAMI

• Southern Appalachians Mountains Initiative (SAMI)– Stakeholder process to develop regional strategy

to deal with:• Ozone (Sum06), PM, haze, acid deposition• Single model applied to suite of 5, 10 day episodes

– Episodes chosen to represent typical year


SAMI Modeling

• Air Quality: URM-1ATM (Urban-to-Regional Multiscale One Atmosphere) Model– Horizontal cells of varying dimensions (12 - 192 km)– 7 vertical layers extending from surface to 12.8 km

• Meteorology: RAMS (Regional Atmospheric Modeling System)

– temperature, air density, wind speed and direction, total solar radiation, ultraviolet radiation, mixing height, turbulent momentum diffusivity, precipitation, cloud parameters

• Emissions: EMS-95 (Emission Modeling System)– Gas: NOx, VOCs, CO, NH3, SO2

– Aerosols: OC, EC, Ca, Mg, K, NO3, SO4, “other” PM


Urban-to-Regional Multiscale One Atmosphere (URM-1ATM) Model

• Three-dimensional Eulerian photochemical model– Finite element, multiscale transport scheme (Odman & Russell, 1991)

– Gas-phase chemistry• SAPRC-93 mechanism (Carter, 1994)

– Aqueous-phase heterogeneous sulfate chemistry– Aerosol dynamics

• Sectional approach (Gelbard and Seinfeld, 1980)• ISORROPIA thermodynamic equilibrium (Nenes, et al., 1998)• Organic aerosol yields (Pandis, et al., 1992)

– Acid deposition• Wet: Reactive Scavenging Module (Berkowitz, et al., 1989)• Dry: three-resistance approach

• “One atmosphere” modeling approach


Aerosol Module• Inorganic aerosols - ISORROPIA

– sulfate, nitrate, ammonium, chloride, sodium, hydrogen ion– condensation/evaporation (thermodynamic equilibrium)

• Organic aerosols– experimental and estimated aerosol yields from VOC oxidation

• Inert Species– EC, Mg, Ca, K, other PM

• Sectional Size Distribution

0

2

4

6

8

1 0

1 2

1 4

1 6

1 8

0 . 0 0 . 1 1 . 0 1 0 . 0 1 0 0 . 0

D i a m e t e r ( m i c r o m e t e r s )

dC

i/d

log

Dp

i

(mic

rogr

ams/

m3/

log(

mic

rom

eter

s))

< - - - - - - - - - P M< - - - - - - - - - P M 1 01 0 - - - - - - - - >- - - - - - - - >

< - - - - - - P M< - - - - - - P M 2 . 52 . 5 - - - - >- - - - >


SAMI Modeling Domain and Grid

*

*


AIRS Station 47-037-0011; Nashville, Davidson Co, TN (urban)

020406080

100120

0 24 48 72 96 120 144 168 192 216

Time (starting July 11, 1995)

Ozo

ne (p

pb)

AIRS Station 47-099-0101; Look Rock, Blount Co, TN(high elevation)

020406080

100120

0 24 48 72 96 120 144 168 192 216

Time (starting July 11, 1995)

Ozo

ne (

ppb)

a) Observed

12%

3%

28%

4%

37%

16%

Average PM 2.5 concentration

28.4 µ g/m3

b) Simulated

11%

3 %

34%

3%

28%

21%

Ammonium

Nitrate

Sulfate

EC

OC

Other

Average PM 2.5 concentration

31.6 µg/m3

Performance


Performance on July 15, 1995

Simulated (L) and Observed (R) PM2.5 for July 15, 1995

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

Station

PM

2.5

(µg

/m3 )

OtherSoilsECNO3NH4ORG

SO4

BRIG DOSO GRSM JEFF LYBR MACA OKEF ROMA SHEN SHRO SIPS UPBU WASH


Sensitivity Analysis to Emissions• DDM - Decoupled Direct Method: Extended to

Particulate Matter– Use direct derivatives of governing equations– Perform numerous sensitivity calculations in one model

run.– Inaccurate sensitivities may result due to non-linear

response– Assessed response of PM to emissions

• Regionally• By source region


Sulfate Sensitivity to SO2 Emissions


Geographic Sensitivity Regions

19

6

13

10

12

7 85

32

4

11

9


SO4 & its Change on July 15, 1995 for a 10% Reduction of 2010-OTW SO2 Emissions from Different Regions

SAMI

2010-OTW Midwest Northeast

Central FL + MS

µg/m3+0.3 +0.1 - 0.1 - 0.3 - 0.5 - 0.7 - 0.9 - 1.1


SO4 & its Change on July 15, 1995 for a 10% Reduction of 2010-OTW SO2 Emissions from SAMI States

KY WV VA

TN2010-OTW NC

AL GA SC

+0.15

+0.05

- 0.05

- 0.15

- 0.25

- 0.35

- 0.45

- 0.55µg/m3



Sulfate Sensitivity at SHEN to 10% SO2 Emission Reductions

-12.0

-10.0

-8.0

-6.0

-4.0

-2.0

0.0 3/31/93

2/09/94

3/24/93

4/26/95

4/29/95

5/03/95

5/15/93

8/04/93

5/27/95

7/19/95

7/24/91

7/31/91

6/24/92

6/27/92

5/12/93

8/07/93

8/11/93

5/24/95

7/12/95

7/15/95

Su

lfat

e S

ensi

tivi

ty (

%)

0.0

2.0

4.0

6.0

8.0

10.0

12.0 Su

lfate Co

ncen

tration

(g/m

3)

AOSENEMWCNWVVATNSCNCKYGAALSO4

Class 1 Class 5Class 4Class 3Class 2

Sulfate Sensitivity at GRSM to 10% SO2 Emission Reductions

-10.0

-9.0

-8.0

-7.0

-6.0

-5.0

-4.0

-3.0

-2.0

-1.0

0.0 3/27/93

2/09/94

3/24/93

4/26/95

5/15/93

8/04/93

8/07/93

4/29/95

6/24/92

8/11/93

5/27/95

7/12/95

7/27/91

7/31/91

7/15/95

Su

lfat

e S

ensi

tivi

ty (

%)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0 Su

lfate Co

ncen

tration

(g/m

3)

AOS ENEMWC NWVV ATNS CNCK YGAA LSO4

Class 1 Class 5Class 4Class 3Class 2


Summary• Ozone models are “mature”• PM Modeling is developing

– Much more involved– More uncertainties– Performance is acceptable

• Main Features– Internal mixtures

• External computationally huge– Sectional distribution

• More flexible than modal– Inorganic thermodynamics

• Useful features– Direct sensitivity analysis

• Future– External mixture approach more common– more detailed organic chemistry


Great Smoky Mt. National Park

0.0

10.0

20.0

30.0

40.0

50.0

7/24/91

7/27/91

7/31/91

6/24/92

6/27/92

3/24/93

3/27/93

3/31/93

5/12/93

5/15/93

8/4/93

8/7/93

8/11/93

2/9/94

2/12/94

4/26/95

4/29/95

5/3/95

5/24/95

5/27/95

7/12/95

7/15/95

7/19/95P

M 2

.5 C

on

c. (

µg

/m3 )

Observations Model

Shenandoah National Park

0.0

10.0

20.0

30.0

40.0

50.0

7/24/91

7/27/91

7/31/91

6/24/92

6/27/92

3/24/93

3/27/93

3/31/93

5/12/93

5/15/93

8/4/93

8/7/93

8/11/93

2/9/94

2/12/94

4/26/95

4/29/95

5/3/95

5/24/95

5/27/95

7/12/95

7/15/95

7/19/95P

M 2

.5 C

on

c. (

µg

/m3 )

Observations Model

Documents

Particulate Matter Modeling: Including Nanoparticles · reaction – Sensitive to temperature K T K=[HNO3(g)][NH3(g)] Georgia Institute of Technology Diurnal Nitrate Pattern Source: