Particle-resolvedsimulationsfor quantifying … · 2016. 11. 14. · Particle-resolvedsimulationsfor quantifying blackcarbonclimateimpactandmodel uncertainty. NicoleRiemer. Department

Particle-resolved simulations for quantifyingblack carbon climate impact and model

uncertainty

Nicole RiemerDepartment of Atmospheric Sciences, University of Illinois at Urbana-Champaign

with Matthew West

October 31, 2016

Nicole Riemer Particle-resolved simulations and black carbon impact 2016-10-31 1 / 41

Global aerosol distributions

Nicole Riemer 2016-10-31 2 / 41Particle-resolved simulations and black carbon impact

Underlying conceptual model of aerosol particles

Smoke (OC/BC)

Sea salt

Sulfate

Dust

External mixture of different aerosol typesNicole Riemer 2016-10-31 3 / 41Particle-resolved simulations and black carbon impact

Real particles in the ambient atmosphere

Li et al., Atmospheric Environment, 45, 2488--2495, 2011


Revised underlying model of aerosol particles

External mixture Complex external and internal mixtures


Transformation of black carbon in the atmosphere

Freshly emitted diesel soot “aged” diesel soot

External Internal

Freshly emitted soot is hydrophobic.Aging due to coagulation and condensation.This changes optical properties and hygroscopicity, henceclimate impacts.Nicole Riemer 2016-10-31 6 / 41Particle-resolved simulations and black carbon impact

Are these details important?

MODIS, NASA

Health Impacts

Mikhail Voskresensky/Reuters

Scatter and absorb solar radiation

Interactions with clouds

Seawifs, NASA Ackermann & Toon, NASA



Change in equilibrium annual mean surface air temperature (K)

“[…] These results confirm that the mixing state of BC with other aerosols is important in determining its climate effect.”

External

Internal

Chung and Seinfeld, JGR 2005




Microscale processes determine macroscale impacts

1

)pdgo(ln

0.01 0.10dp in µm

1.00mm

ifjficjcstotal

n(lo

g d p

)Regional scale

Global scale

Mesoscale

Microscale

Particle scale

Molecular scale


Central research question and strategy

QuestionHow important is BC mixing state for

predicting BC climate impacts?

TheoryMixing state metrics and error metrics

Simulation and dataLibrary of real-world and idealized scenarios

SynthesisImpacts and findings

ToolPartMC--MOSAIC








ToolPartMC-MOSAIC


Model aerosol representation: Modes and binsam

ount

radius radius

Evolution of mixing state is challenging to represent.Each mode or size bin is treated as internally mixed.


What are particle-resolved aerosol models?

BCRiemer et al., J. Geophys. Res.,114, D09202, 2009

SO4

• No bins or modes• Particles as vectors• Treating multidimensional size distribution

OC

Nicole Riemer 2016-10-31 14 / 41

Particle 1

Particle 2

Particle 3

BC 3 10 1SO4 12 3 4OC 5 8 2

Particle-resolved simulations and black carbon impact

Benefits of particle-resolved models

No approximation needed for mixing state.

1 Coarse graining tool: deriving parameters for more approximatemodels (e.g. BC aging).

2 Benchmark and error quantification for more approximatemodels (e.g. QMOM, MADE3, MATRIX, MOSAIC-ext).

3 Detailed studies on the particle scale and experimentalintercomparison.

Riemer et al., J. Aerosol Sci., 2010; Willis et al., Atmos. Chem. Phys., 2016;Fierce et al., Atmos. Chem. Phys., 2015, Fierce et al., Nature Comm. 2016; Ching

et al., J. Geophys. Res., 2012, 2016; Tian et al., Atmos. Chem. Phys., 2014McGraw et al., Journal of Physics: Conference Series, 2008; Kaiser et al., Geosci. Model Dev., 2014


Limitation of particle-resolved models

0

40

60

BC d

ry m

ass

frac.

wB

C,d

ry (%

)

0.01 0.02 0.05 0.1 0.2

dry diameter D (µm)

0.5

101

80

1 103

0 5 · 104 1 · 105 1.5 · 105

number conc. nB C ,d ry (D,w) (cm− 3)

102

103

104

num

ber c

onc.

n(D

)(cm

)

−3

10− 4

mass conc.m

(D)(µg

m−3)

105 24 hours (06:00 LST)

10− 2

100

102

104

number

mass

100 10 10 101 2 3 104

mass conc. mB C ,d ry (w) (µgm− 3 )

104 105 106

number conc. nB C ,d ry (w) (cm− 3)

107

number

mass

Resolution: Only finitenumber of computationalparticles available per gridcell (104 − 107).On-going research to develop more efficientalgorithms, e.g.“weighted particles”(DeVille et al., J. Comp.Phys., 2011) and parallelmethods.


condensation coagulat

VOCs diesel sootNOx gasoline sootSO2 cooking POANH3 road dust

Typical model setup

windaged urban aerosol

polluted residual layer

wind

background aerosol

stable nocturnal layer

polluted mixed layer

photooxidation +

ion

Zaveri, Easter, Riemer, West, JGR 2010





TheoryMixing state metricsand error metrics

Simulation and dataLibrary of real worldand idealized scenarios


ToolPartMC-MOSAIC


Hypothesis: Error versus mixing state metric

external internalaerosol mixing state

error in simulated aerosol property


Hypothesis: Error versus mixing state metric




Mixing state terminology

Population mixing state and Morphological mixing stateHere we will only consider the population mixing state.

1 How “complex” are the particles, i.e. how many species arepresent in one particle?

2 How different are the particles from each other?Nicole Riemer 2016-10-31 21 / 41Particle-resolved simulations and black carbon impact

Problem solved in ecology: Species diversity

Externally mixed waterholes

Internally mixed waterholes

Good, Biometrika, 1953MacArthur, Ecology, 1955Whi> aker, Ecol. Monogr., 1960; Science 1965; Taxon, 1972


Mixing state metric χ

x = 100%

Internally mixed

Externally mixed

x = 0%

Based on diversity metrics framework used in ecology.Need to know per-particle mass fractions.

Riemer and West, Atmos. Chem. Phys., 2013; Healy et al., Atmos. Chem. Phys., 2014; O’Brien et al., J. Geophys. Res., 2015


Application to field data: MEGAPOLI

NO3

SO4

OA Di = 4.0

BCNH4

SO4 NO3

OA

BC

Rel

ativ

e pe

ak a

rea

(arb

itrar

y un

its)

200180140 160120100m/z

806040200

Rel

ativ

e pe

ak a

rea

(arb

itrar

y un

its)

100 120 140 160 180200m/z

0 20 40 6080

+

-

K /C3H3+ +

H(NO3)3-

NO3-

NH4+ C2H3O

+

NO2-

HSO4-

C2H3+

Rel

ativ

e pe

ak a

rea

(arb

itrar

y un

its)

200180160140120100m/z

806040200

+

Di = 2.0

dva = 163 nm

-

C+

C3+

C4 C5+ +

C2-

C3-

C5-

+C2

Ca+

C4-

HSO4-

dva = 935 nm

ed Di values of 2.0 (top

Collaboration with RobertHealy, University of Toronto.

Paris, France, winter 2010.

Mass fractions estimated basedon ATOFMS data,supplemented by AMS, SMPSand MAAP.

Five species: OA, BC, SO ,4

NO3, NH4.

Healy et al., Atmos. Chem. Phys., 4, 6289–6299, 2014

Nicole Riemer 2016-10-31 24 / 41

+C3

NH4 BC +

+C +

C2H3O+

5NO3 OA

SO4

Di = 3.0dva = 436 nm

+NH4

3 -

- -

NO2 HSO4 H(NO3)3


Mixing state parameters during the campaign

70

60

50

401.00.80.60.40.20.0B

ulk

popu

latio

n m

ass

fract

ion

Date

4.0

3.53.0

2.52.0

4.54.03.53.0

DD

C M C

D

D

NH4NO3

SO4

OABC

Healy et al., Atmos. Chem. Phys., 4, 6289–6299, 2014








ToolPartMC-MOSAIC






1. CCN concentration2. Optical properties


ToolPartMC-MOSAIC


Back to the hypothesis




Importance of mixing state for CCN properties

Use χ fromreference population

calculate difference in CCNconc. after compositionaveraging

10−8

10−7

10−6

10−2

10−1

100

101

dry diameter D / m

crit.

supe

rsat

urat

ion

S

/ %

c

num

ber

conc

. /

cm−3

1e+01

1e+02

1e+03

1e+04

1e+05

10−8

10−7

10−6

10−2

10−1

100

101

dry diameter D / m

crit.

supe

rsat

urat

ion

S

/ %

c

CCN inactive

CCN active

CCN inactive

CCN active

composition-averaging

Error in CCNconcentration?

χ = 0.28105

104

103

102

101

(a) (b)

0.4 0.6 0.8Mixing state index χ

Erro

rin

CCN

co

ncen

trat

ion

0 0.2 1

80%60%40%

20%

(c) senv=0.3%

Ching et al., ACP, in prep., 2016


Scenario library to sample relevant parameter rangeenvironments.

Fierce et al., ACP, 2015; Fierce et al., Nature Comm., 2016; Ching et al., JGR, 2016


Relationship of χ and error in CCN concentration

Mixing state index χ

Erro

rin

CC

Nco

ncen

tratio

n/%

Senv = 0.1 % Senv = 0.3 %

Senv = 0.6 % Senv = 1.0 %

Ching et al., ACP, in prep., 2016


Back to the multiscale model hierarchy

)pdgo(ln

0.01 0.10dp in µm

1.00mm

ifjficjcstotal

n(lo

g d p

)Regional scale

Global scale

Mesoscale

Microscale

Particle scale

Molecular scale


Back to the multiscale model hierarchy

1


)pdgo(ln

0.01 0.10dp in µm

1.00mm

ifjficjcstotal

n(lo

g d p

)Regional scale

Global scale

Mesoscale

Microscale

Particle scale

Molecular scale

Diversity in composition affects BC absorption

absorption enhancement by BCin example particles

average composition

Per-particlecomposition

Per-particleabsorption

particle-resolved composition abs. enhancement

compositionof 30 exampleBC-containing

particles

Fierce et al., Nature Comm., 7, 12361, 2016.






particles

abs. enhancementFierce et al., Nature Comm., 7, 12361, 2016.




average composition


particles

average composition abs. enhancement

average composition

population-level absorption enhancement




Why this bias?

mass of BC contained in each particle (pg)

Uniform composition assumption: All particles contain the samevolume fraction of each aerosol component.Causes an artificial redistribution of coating material ontoparticles containing large amounts of BC.

dist

ribu

tion

in a

eros

ol m

ass

with

re

spec

t to

per-

part

icle

BC

mas

s ["

g]

0

0.25

0.50

0.1

uniform0.75

no change in per-particle BC mass

10-210-4 10-3

NO3NH4SO4SOA POA BC

particle-resolved



Coarse-graining of particle-resolved data

target hygroscopicity of dry coating

btarget volume fraction dry coating

environmental relative humidity

target relative humidity

predicted absorption enhancement

distribution in composition of BC-containing particles

BC populations 1 2 … N

kernel density regression

nonparametric relationship

a

volume fraction dry coatingBC populations 1 2 … N

hygroscopicity of dry coatingBC populations 1 2 … N

absorption enhancementBC populations 1 2 … N

Fierce et al., Nature Comm., 7, 12361, 2016, Supplemental Material.


Taking this to the global scale

a b c

Accounting for particle-level composition diversity

Accounting for particle-level composition diversity

Assuming uniform composition

Neglecting water uptake Accounting for water uptake

our best estimate

Accounting for water uptake



Outcomes

1 Framework to quantify the impact of aerosol mixing state ondifferent target quantities (CCN properties and opticalproperties).

Mixing state metric and error metricsCoarse-graining method: Use particle-resolved simulations todevelop parameterization with inputs that large-scale modelsalready track.

2 CCN properties: Small errors (< 10 %) are caused byinternal-mixture assumption for populations with χ > 0.6. Largeerrors (up to 150 %) for χ < 0.2.

3 Optical properties: If particle-diversity is neglected,population-level absorption enhancement is overpredicted (up to 70%).


Impacts on community

BCM

assF

ract

ion

Optical and CCN Act ivat ion Properties

With PNNL:MOSAIC-mix

multisectional

2 µm

ECINOC

Scenario libraries Mixing state metrics

With O’Brien, Moffet, Gilles, Laskin: Spectro-microscopy and mixing state

With Tami Bond andLaura Fierce: Predictive models of aging t ime scales

Fierce et al., ACP, 2015Fierce et al., BAMS, 2015

Ching et al., JGR, 2015 O’Brien et al., JGR, 2015Nicole Riemer 2016-10-31 41 / 41

x = 100%

Internallymixed

Externallymixed

x = 0%


Documents

Particle-resolvedsimulationsfor quantifying … · 2016. 11. 14. · Particle-resolvedsimulationsfor quantifying blackcarbonclimateimpactandmodel uncertainty. NicoleRiemer. Department