modellingorganicaerosol · 2008-07-29 · decomposition products droplet aerosol jang et al. 2002...

Modelling Organic Aerosol

David Simpson

EMEP, MET.NO & Chalmers

– p.1/78

Overview

Introduction

Approaches to modelling

SOA models

Examples of model usage + evaluation

Conclusions

– p.2/78

What is EMEP?

Cooperative Programme for Monitoring and Evaluation ofthe Long-Range Transmission of Air Pollutants in Europe

(European Monitoring and Evaluation Programme)

– p.3/78

What is EMEP?

Cooperative Programme for Monitoring and Evaluation ofthe Long-Range Transmission of Air Pollutants in Europe

(European Monitoring and Evaluation Programme)

Aims: To provide sound scientific support for theConvention, in particular in the areas of:

Atmospheric monitoring and modelling

Emission inventories and emissions projections

Integrated assessment

– p.3/78

EMEP . . .

CLRTAP: Convention on Long Range Transboundary AirPollution

Adopted 1979

51 Parties

Eight Protocols

EMEP, 1984

Last one: Göteborg,1999

Contribution to EU NEC Directives + CAFE

– p.4/78

Air Pollution Modelling

– p.5/78

Purpose of Modelling

1. Policy - emissioncontrol

2. Scientificresearch

3. Both!

Example: d(Dep(N))/d(ENOx),Austria

– p.6/78

Complete approach

Detailed understanding might require:

size distributions

complex chemical pro-cesses

– p.7/78

Policy/Global Models

Typically require:

simpler - only masses (PM2.5, PM10)

Well evaluated (trustworthy) models

Concentrate on main processes:

– p.8/78

Box model

The simplest model:

em

Emissions

< >

><

dC

dt=

E

h− u.(C − C0)

– p.9/78

Box model

Adding Chemistry:

em

Emissions

CHEMISTRY

< >

><

dC

dt=

E

h− u.(C − C0) + P − L.C

– p.10/78

Box model, cont.

... and more terms

-Vg. C : dry deposition

-L . C : wet deposition

entrainment

. . .

Allow the box to move?

⇒ Lagrangian

– p.11/78

The Eulerian 3D model

Represents all mainphysical and chemicalprocesses

Numerical integration

Scientifically most sound method of calculating air pollution

– p.12/78

Eulerian model, cont.

3D models are CPU-expensive:

170 × 130 × 20 = 440 000 gridcells×100 species⇒ 44 million concentrations

Typically requires supercomputer for long simulations.

– p.13/78

Eulerian model, cont.

3D models are CPU-expensive:

170 × 130 × 20 = 440 000 gridcells×100 species⇒ 44 million concentrations

Typically requires supercomputer for long simulations.

(But, the times they are a changing.....)

– p.13/78

Aerosol Extras

⇒ Many other ‘effective’ species

– p.14/78

Aerosol Extras


Nucleation

– p.14/78

Aerosol Extras


Nucleation

Coagulation

– p.14/78

Aerosol Extras


Nucleation

Coagulation

Condensation

– p.14/78

Aerosol Extras


Nucleation

Coagulation

Condensation

Cloud-processes

– p.14/78

Aerosol Extras


Nucleation

Coagulation

Condensation

Cloud-processes

Size-resolved emissions

– p.14/78

Aside: Is complexity good?

Acc

ura

cy

Complexity– p.15/78

Aerosol modelling

Two main approaches:

1. Modal models

2. Sectional models

– p.16/78

Modal models

Make use of log-normal distribution

n(ln Dp) =N

√2π ln σg

exp

[

−1

2

(ln D − ln Dpg)2

ln2 σg

]

– p.17/78

Modal models, cont.

Break-down atmosphericdistribution in 2-3 modes

– p.18/78

Modal models, cont.

The kth moment is defined as:

Mk =

∫

+∞

−∞

Dkpn(ln Dp)d(ln Dp)

with solution

Mk = N.Dkpg

exp

[

k2

2ln2 σg

]

M0 = total particle number concentration

M2 ∝ surface area

M3 ∝ volume, mass

– p.19/78

Modal models, cont.

Advantages

Requires very few parameters (σ, Dpg)

Computationally inexpensive

Dis-Advantages

Has no explicit size-distribution, therefore conditionsassumed uniform within a mode

– p.20/78

Sectional models

Divide aerosol distribution into ‘bins’ or ‘sections’.Typically 4-100, e.g.

– p.21/78

Sectional models, cont.

Advantages

State-of-the-art description (with many sections)

Allows different chemical mixtures at different sizes

Flexible

Dis-Advantages

Computationally expensive

Physics/chemical basis not always known

– p.22/78

Belief in models?

The basic rule:

– p.23/78

Belief in models?

The basic rule:

Garbage in ⇒ ⇒ Garbage out:

– p.23/78

Belief in models?

The basic rule:


SOA twist:

– p.23/78

Belief in models?

The basic rule:


SOA twist:

Garbage in the middle!

– p.23/78

Sources of OC

– p.24/78

Organic Aerosol

OA: Subject=Horrendous!! 1000s of compounds, mainlyunknown. Formation mechanisms complex and unkown!

– p.25/78

Partitioning

Gas-Particlepartitioning:

Ai

Gi

=COA

C∗

i

where

C∗

i is saturationconcentration,= f (Vapourpressure)

– p.26/78

G/P cont.

A multitude of eqns found, e.g.

Ai

Gi

= Ki.COA =RT

MWζip0L,i

.COA

Ai

Gi

=COA

C∗

i

=RT

MWiζ′

ip0L,i

.COA

Smog-chambers:

Y = COA

∑ αiKi

1 + αiKiCOA

– p.27/78

Chemistry < − > SOA?

Stolen from Neil...

– p.28/78

SOA: α-K approaches

Smog-chamber data could be explained with:

VOC + Ox ⇒ α1 P1 + α2 P2

– p.29/78

G/P approaches

COA(µg/m3) ⇒

Pi = Ai + Gi

Ai

Gi

= Ki.COA

Ai

Gi

=COA

C∗

i

– p.30/78

α-K approaches, cont.

Pros:

Easy-to-use

Available for many compounds

– p.31/78


Pros:

Easy-to-use


Cons:

Derived from smog-chambers, often 40◦C, 100s ppb,v.low RH

Not flexible/mechanistic

– p.31/78


Pros:

Easy-to-use


Cons:

Derived from smog-chambers, often 40◦C, 100s ppb,v.low RH

Not flexible/mechanistic

Coefficients used so far, wrong?

– p.31/78

New Evaluations

m-xylene

(Griffin, 1999)

a-pinene

(Griffin, 1999)

Isoprene

(Pre-2005)m-xylene

(Ng, 2007)(high-NOx)

a-pinene

(Chan2007/,Ng2006)

Isoprene

(Henze,2006)m-xylene

(Ng, 2007)

(low-NOx)

Isoprene

(Chan, 2007)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

Yie

ld (

Fra

cti

on

)

Yield (Y) at M0 = 5 ug/m3

’OLD’ ’NEW’

Changes in Yield Estimates

e.g. Ng et al. (2006, 2007), Chan et al. (2007)– p.32/78

Volatilty Approach

Donahue, Robinson....

The Volatility Basis Set

!! !" !# $ # "$

$%#

$%"

$%!

$%&

$%'

$%(

$%)

$%*

$%+

#

,-.# $/01/2!./3

!!4

!/2567-8-,/97:;<=->4

!"#$%$&'($!)$*

!+

C∗i =

{

0.01,0.1,1,10,100,1000,104,105,106}

µgm

α-Pinene + Ozone Mass Balance

!"!#

!""

!"#

!"$

!"%

"

" '#

" '$

" '%

" '&

!

!'#

!'$

!'%

()*+,!-+.

!/0

!"+,123.456789+*832:25+;4::+<34=>62?0

C10=1

C10O4 = 1.4

O3+

O

O

OOH

100 ppt

10-7 torr

• Mass yields α′i = {.004,0, .05,.09, .12, .18,...}

• Only around 0.055 SOA formation from α-pinene in the LVOC range at low NO

• Mass balance for ‘nominal product’ demands ξmax =∑

i αi ≃ 1.4

– p.33/78

Volatility methods?

Pros:

Flexible framework

Maps more of parameter space

Easier to link new data/experiments

Efficient for global models

Cons:

Still not mechanistic

– p.34/78

EMEP approach

EMEP Kam-2 Method: ‘Explicit’, extended from Kamens et al.:

21 reactions, 15 products, dimer, .... Andersson-Sköld andSimpson, JGR, 2001

– p.35/78

EMEP BSOA, Kam-2

Evaluated against smog-chamber:

0 50 100 150 200 250

Time (mins)

0

10

20

30

40

50

Ae

roso

l m

ass (

ug

/m3

)

Hoffman H6

T=43-48 C, light, +NOx

0 50 100 150 200

Time (mins)

0

500

1000

1500

2000

2500

3000

Ae

roso

l m

ass (

ug

/m3

)

Kamens K3

T= 15-6 C

– p.36/78

BSOA; Kam-2 Method

Comparison with Smog-Chambers good(Andersson-Sköld and Simpson, 2001):9-820 ppb α-pinene, 0-240 ppb NOx

– p.37/78

EMEP Kam-2(X)

Pros:

Flexible framework

‘Real’ species (surrogates anyway)

Linked with gas-phase chemistry

Evaluated against several smog-chamber exps.

Cons:

No aqueous/heterogenous chemistry

One (α-pinene!) species

Old

– p.38/78

The Big Stuff

MCM, 1000s reactions, 200 SVOC species (Jenkin et al.,JGR, 2004)

(or CACM, Griffin et al.) – p.39/78

MCM-type

Pros:

Explicit framework

‘Real’ species

1000s of reactions - as realistic as possible

Cons:

No aqueous/heterogenous chemistry

Two (α, β-pinene) species

Needs very large (100-500) correction factors forvolatility

– p.40/78

MCM-type, cont.

Cons:

Heavy!

– p.41/78

MCM-type, cont.

Cons:

Heavy!Pro:

Attempt to incorporate best-understanding

– p.41/78

New issues....

HULIS

H2SO4

isoprene

carbonyls

OH, H2O2

Lignin pyrolysis products

co

nd

en

sati

on

Decomposition

products

droplet

aerosol

Jan

g e

t a

l. 2

00

2

Lim

be

ck e

t a

l. 2

003

Za

ppo

li e

t a

l. 1

99

9,

Ma

yo

l-B

race

ro e

t a

l. 2

00

2

Gel

encs

ér e

t al.,

200

3

Gelencsér et al., 2002

heterogeneous

direct emission multiphase

Working hypotheses for HULIS formation

Still changing - e.g. Warneck, Ervens, Jang, Griffin,suggest aqueous/heterogenous pathways as source ofSOA. Isoprene, glyoxal, oxalic acid, ....

Do we know which pathway to follow?

– p.42/78

Sensitivity

All models sensitive to:

Vapour pressure

∆H assumptions

Activity coefficients

Deposition assumptions

Emissions

– p.43/78

Sensitivty: ∆H

Tsigaridis+Kanakidou, ACP, 2003

– p.44/78

Emissions – IIASA

Fine-particle emissions - Kupiainen, and Klimont, 2007

– p.45/78

Dry Deposition

Problems of Theory vs. Measurements:

From PhD Thesis, Rick Thomas

– p.46/78

Summary of Models

Many models

Little basis for choosing!

Little basis for evaluation!

– p.47/78

Summary of Models

Many models

Little basis for choosing!

Little basis for evaluation!

Unconstrained!Need Observations!

– p.47/78

Results

Results: Annual Average OC, year 2002 (ugC/m3)

Kam2X

– p.48/78

BSOA contribution

BSOA/OC (%)

ASOA/OC (%)

Kam2X– p.49/78

OC, take 2

OC with alternative vapour pressures

Kam-2X

Kam-2

– p.50/78

Sensitivity of OC: Birkenes

Jul02 Oct02 Jan03 Apr03 Jul03 Oct030

1

2

3

4

5

6

7

8

9

BSOAASOAWOODFFUELBGND

Jul02 Oct02 Jan03 Apr03 Jul03 Oct030

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

BSOAASOAWOODFFUELBGND

– p.51/78

Birkenes EC, TC

Jul02 Oct02 Jan03 Apr03 Jul030

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Site: Birkenes EC

EC−modEC−obs

Jul02 Oct02 Jan03 Apr03 Jul030

1

2

3

4

5

6

7

8

9

10

Site: Birkenes TC

TC−modTC−obs

Model performance - quite good at all Northern Europeansites

– p.52/78

Aveiro SO−2−4

, TC

Jul02 Oct02 Jan03 Apr03 Jul03 Oct03 Jan04 Apr04 Jul040

2

4

6

8

10

12

14

Site: Aveiro SO4

SO4−obsSO4−mod


2

4

6

8

10

12

14

16

18

20

Site: Aveiro TC

TC−obsTC−mod

Model performance - quite bad for TC at all southernEuropean sites

– p.53/78

CARBOSOL Project

1-week filters (PM2), analysed for:

cellulose ⇒ biological parti-cles

levo-glucosan

⇒ biomass-burning

OC/EC ⇒ primary emis-sions

14C ⇒ modern/fossil

16 papers: Present and Retrospective State of Or-ganic Aerosol Over Europe, J. Geophysical Research,VOL. 112, D23, 2007

– p.54/78

Aveiro revisited...

Use levoglucosan to ‘correct’ WOOD


0.5

1

1.5

Site: Aveiro LEVO


5

10

15

20

25

30

35

Site: Aveiro TCxwood

TCxwood−obsTCxwood−mod

Promising :-)

– p.55/78

cf CARBOSOL

K-Puszta (Hungary), Summer

Obs.-Derived EMEP Model

(5–95th %ile) (Kam2 - Kam2X)

TC 5.2 1.6 – 2.7

WOOD 0.3 – 0.5 0.05

EC 0.4 – 0.7 0.4

FFUEL 0.2 – 0.5 0.4

BSOA 2.9 – 3.6 0.2 - 1.4

ASOA 0.05 – 0.7 0.03 - 0.04

Units: µg C m−3

Simpson et al., JGR, 2007

– p.56/78

Use of tracers, cont.

SORGA: Norwegian project

Tove Svendby, Karl-Espen Yttri, ...

David Simpson, MET.NO

Hans Puxbaum + co. (TUV)

Kristina Stenström, Lund Univ.

– p.57/78

SORGA

– p.58/78

Source-Apportionment, cont

Other tracers:14C ⇒ modern/fossil

cellulose ⇒ plant matter, ...

sugars/alcohols ⇒ fungi, ...

OC/EC ⇒ primary emissions

levoglucosan ⇒ biomass-burning

- all factors approximate.

- some ‘traps’, e.g. some modern 14C could be fromcooking oils, tyres, etc.

(e.g. Gelencsér et al., JGR, 2007)

– p.59/78

SORGA

Sources of PM1, Summer:

(K.E. Yttri et al., 2008, Prelim)– p.60/78

SORGA

aKT-EMEP model, Hurdal:

– p.61/78

SORGA

Kam-2X-EMEP Model, Hurdal:

– p.62/78

The Target

Do we know how much OC we want?

Key words:Artifacts (EC/OC, -ve, +ve, ...)– can be of order 50% ?

Representativity - what does e.g.[OC] mean?

– p.63/78

Artifacts

EUSAAR result:

: Fig. from Jean-Philippe Putaud

– p.64/78

Other studies

Volkamer et al., GRL, 2006

– p.65/78

Other studies

Volkamer et al., GRL, 2006

– p.66/78

Conclusions

State of OC science ‘in infancy’ (Donahue et al., 2005)

– p.67/78

Conclusions

State of OC science ‘in infancy’ (Donahue et al., 2005)

. . . because as we know, there are known knowns;there are things we know we know. We also knowthere are known unknowns; that is to say we knowthere are some things we do not know. But there arealso unknown unknowns - the ones we don’t know wedon’t know.

- (Donald Rumsfeld, 2005)

– p.67/78

Conclusions

Modellers have no way to ‘solve’ SOA modelling untilchemists have understood the basics.

But, model’s can serve to test theories and emissions

Emissions? Primary OC/BC + precursor (terpenes!)emissions need verification (near-sourcemeasurements?)

– p.68/78

Conclusions 2

Measurements are required to develop and constrainmodels and validate emissions

Needs chemical speciation, tracers, many locations

Long-term field data + campaigns+supersites ideal

AMS, C14, ......

– p.69/78

Wishes...

Would be good to specify:

Emissions (AVOC, BVOC, volatility)

Volatile E(PM)? Or Condensible E(VOC)?

Source of Atmos. Aerosol:

How much is modern/fossil

How much is biomass/BSOA

How much is through aqueous pathway

Acidity/S ?

Mixing polar/nonpolar/liquid/other??

Which smog-chamber data are relevent?

Link smog/flow-chambers – atmosphere

– p.70/78

Garbage Avoidance

Strategies:

Check basics - does the model work for anything?

Check other pollutants - SO2, SO4, NOx, NOy,

Check emissions!

Check PCM tracers - EC, levoglucosan, C14

Check measurements - what do they mean?!

Be humble.....

– p.71/78

The End...

– p.72/78

Other studies

The simplest result – all PM from forests:

See: Tunved et al., Science, 2006

– p.73/78

Other studies

The simplest result – all PM from forests:

See: Tunved et al., Science, 2006

NB: Applies to clean air, selected air masses

– p.73/78

Evaluation

Mainly by comparison with:

More Complex models (e.g. for chemical schemes)

Measurements - the main test!!

– p.74/78

Ozone

– p.75/78

Model vs. Model

EMEP vs IVL (HCHO):

Andersson-Skold & Simpson, Atmos.Env., 1999

– p.76/78

HCHO Cont. Field Comp:

Donon, France:

– p.77/78

Isoprene

(Nice result for precursor too:)

Lessons?

Combination - lab (via. MCM) + field data verypowerful – tests kinetics, emissions and chemistry

(New comparisons in progress, Tack SCARP, Tellus,FZJ!)

– p.78/78

modellingorganicaerosol · 2008-07-29 · decomposition products droplet aerosol jang et al. 2002...

Documents