evaluation of a gas-phase atmospheric mechanism for low ... · pdf fileevaluation of a...

W. P. L. Carter 5/27/04 Low NOx Mechanism Evaluation 1

EVALUATION OF A GAS-PHASEATMOSPHERIC MECHANISMFOR LOW NOx CONDITIONS

• Background and Objectives

• Environmental Chambers Employed

• Chamber Effects Characterization

• Evaluation Results

• Modeling assessment of relative rates of low NOx reactions

• Discussion and Conclusions

William P. L. CarterCollege of Engineering Center for Environmental Research and Technology

University of California, Riverside, CA 92521

May 27, 2004


BACKGROUND• Ground-level ozone O3 is formed in complex reactions of

emitted VOCs with NOx in sunlight

• The mechanisms representing these reactions in models arecritical to accurate O3 control strategy predictions.

• Environmental chamber data provide the best way to test thesemechanisms independent of other uncertainties

• However, current mechanisms were evaluated using data withhigher NOx levels than most current ambient conditions.

• This is a concern because the nature of the oxidation processeschange as NOx is reduced.

• Opportunities exist for lower NOx mechanism evaluation:• Existing Low NOx data from CSIRO and TVA Chambers• Experiments from the new UCR EPA Chamber designed for

low NOx evaluations are becoming available


OBJECTIVES

• Evaluate the SAPRC-99 mechanism (the most up-to-date anddetailed mechanism used by the CARB) for low NOx conditions.

• Obtain and characterize existing TVA and CSIRO chamber datafor mechanism evaluation

• Conduct experiments in the new UCR EPA chamber mostneeded for low NOx evaluation

• Evaluate mechanism using available TVA, CSIRO, and UCREPA chamber data

• Investigate modifications to SAPRC-99 to better represent lowNOx conditions

• Recommend research needed to improve mechanismperformance


SUMMARY OF ENVIRONMENTAL CHAMBERSUSED IN THE LOW NOx EVALUATION

20031995-19961993-96Dates of Runs

O3, NO, NO2, CO,HCHO, PAN,HNO3, VOCs

O3, NO, NOy-NOO3, NO, CO,HCHO, PAN,

VOCs

MeasuredSpecies Used inEvaluation

36 Char42 Mech Eval

19 Mixtureexperiments

32 Char.48 Mech Eval

Number ofExperiments

Argon ArcSunlightFluorescentsLighting

FEP Teflon® FilmWalls

2 x 852 x 2028Volume (m3)

UCR EPACSIROTVA


DIAGRAM OF TVA INDOOR CHAMBER


DISCUSSION OF TVA INDOOR CHAMBER

• Experiments conducted by Simonaitis and Bailey of TVA in 1993through 1995 for low NOx mechanism evaluation

• Steps taken to reduce background by extensive purgingbetween runs

0.0

0.2

0.4

0.6

0.8

1.0

0.300 0.350 0.400 0.450 0.500

Wavelength (nm)

Rel

ativ

e In

tens

ity (

Nor

mal

ized

to

give

sam

e N

O2

phot

olys

is r

ate) Average TVA

BlacklightsSolar Z=0

• 3 types of fluorescentlamps used to approxi-mate solar spectrum

• NO2 actinometry resultsaveraged 0.392 min-1

• Temperature during runsvaried from ~295 - 315 oK

• TVA Chamber no longeroperational

• Data made available for modeling by Jeffries and Co-workers foran RRWG project


CSIRO OUTDOOR CHAMBERS


DISCUSSION OF CSIRO OUTDOOR CHAMBER

• Located outdoors in a suburb of Sydney, Australia

• Multiple complex surrogate – NOx experiments conducted byJohnson and co-workers to test and derive parameters forJohnson’s parameterized “extent of reaction” model

• Data used from 10 dual chamber surrogate experimentsconducted in a collaborative project with Jeffries and co-workers

• Data made available for modeling as part of an RRWG project

• Photolysis rates for modeling derived as follows:• NO2 photolysis rates as function of time estimated from TSR

data using relationship of Demerjian and Schere• Ratios of other photolysis rates to NO2 calculated using

Peterson (1977) actinic fluxes as function of time of day

• Other characterization parameters estimated based oncharacterization results for other chambers


EXAMPLES OF CSIRO TEMPERATURE AND LIGHTCHARACTERIZATION ASSIGNMENTS

Temperature(Experimental and

Model Input)

Experimental and Model Fit TSR

(Modeled TSR derived from calculated NO2photolysis rates and Demerjian and Schere

Relationship)

Experimental Model No HV adj.

Run 340 P

0.00.2

0.4

0.60.8

1.01.2

1.4

300 480 660 840 1020

Run 362 P

0.00.20.4

0.60.8

1.01.21.4

300 480 660 840 1020

Run 342 L

275

280

285

290

295

300

305

300 480 660 840 1020

TSR or temperature vs. minutes after midnight


DIAGRAM OF UCR EPA CHAMBER

20 ft. 20 ft.

20 ft.

This volume kept clear to maintain light

uniformity

Temperature controlled room flushed with purified air and with reflective

material on all inner surfaces

Dual Teflon Reactors

Two air Handlers are located in the corners on each side of the light

(not shown).

Gas sample lines to laboratory below Access

Door

200 KW Arc Light

2 Banks of Blacklights

SEMS (PM) Instrument

Floor Frame

Movable top frame allows

reactors to collapse

under pressure control

Mixing System Under floor of reactors


~5 M

~2.5 M

~6 M Deep

Movable top frame holds and seals sheets and

allows reactor to collapse

2 Mil FEP Teflon® film

Rigid frame holds and seals sheets

Ports for injection, mixing, exchange, and

emptying

Firmness sensor controls top frame movement to

maintain pressure

Location of Future 2nd Reactor

Reflective Floor covered with Teflon

film

PICTURES OF UCR EPA CHAMBER

0.6 M

0

1

2

3

300 400 500 600 700

Wavelength (nm)

Nor

mal

ized

Pow

er

(arb

itrar

y un

its)

SolarChamber

Light Source and Spectrum


DISCUSSION OF UCR EPA CHAMBER

• Constructed using a $2.9 Million EPA earmark to develop a“Next Generation” chamber for mechanism evaluation• Large volume to minimize background and permit PM

studies and instrumentation with large sample requirements• Indoor chamber for maximum characterization and control• Light source simulating sunlight• Low background to permit well-characterized experiments at

low pollution levels• Advanced analytical instrumentation• Temperature control to ±1oC in ~5o - ~50oC range

• Experiments with current configuration began in early 2003:• Initial characterization and evaluation runs funded by EPA• Very low NOx surrogate experiments run for this project


DERIVATION OF MAJOR CHAMBERCHARACTERIZATION PARAMETERS

Approx. same asradical source

Assume same asradical source

Model CH3CHO -Air runs

NOx Offgasing

Believed to below in most runs

Estimated to beminor

Derived fromtracer data

Dilution

MeasuredEstimatedMeasuredO3 Decay

Model char. runssensitive to initialHONO

No data – variedin evaluationsimulations

Model CO - NOx,NOx - Air, andCH4 - NOx runs

Initial HONO

Model HCHO inchar. runs w/oHCHO source

Assumed to beunimportantcompared to R.S.

Model HCHO inCO - NOx or NOx- air runs

HCHObackground andoffgasing

Model char. runssensitive toHONO offgasing

T-Dependent rateassumed to besame as SAPRCOTC or 3 x higher

Negligiblecompared to highHCHO offgasing

ContinuousRadical source(HONO offgasing)

UCR EPACSIROTVAParameter


VALUES OF MAJOR CHAMBERCHARACTERIZATION PARAMETERS

k1 = 0.284 min-1Standard: Deriv’dfrom TSR data

High: 1.15 x Std.

k1 = 0.392 min-1Light Intensity

1.1% /hour1.3% /hour7% /hourO3 Decay

0.05 ppbStandard: ~0High RS: 2 ppb

0.5 ppbInitial HONO

10 pptAssumed to below compared toother sources

HCHO: 45 pptHCHO precursor:135 ppt

HCHO offgasing /NO2 photolysisrate ratio

Runs 55-80: 5 pptRuns 81-168:•Side A: 8.5 ppt•Side B: 12.5 ppt

Standard: Temp.-dependent fit forOTC (10-100 ppt)High RS: 3 x Std.

7.2 ppt

(as NOx offgasingonly)

HONO or NOxoffgasing / NO2photolysis rateratio

UCR EPACSIRO (varied)TVAParameter


UCR-EPA HONO INPUT PARAMETER VS. RUN NO.

0

5

10

15

20

25

50 75 100 125 150 175

EPA Run Number

HO

NO

Inpu

t / N

O2

Pho

toly

sis

Rat

e R

atio

(ppt

)Rad. Source (Side A) Rad. Source (Side B)NOx Source (Side A) NOx Source (Side B)Assigned (Side A) Assigned (Side B)


COMPARISONS OF RADICAL SOURCEPARAMETERS FOR UCR EPA AND

PREVIOUS CHAMBERS

1

10

100

1000

290 300 310 320

Average Temperature (K)

HO

NO

offg

asin

g / N

O2

phot

olys

is r

ate

ratio

(pp

t)ETC

DTC

DTC-3

DTC-10

XTC

CTC1

CTC2

OTC

Fit for OTC

UCR EPA 1

UCR EPA 2A

UCR EPA 2B


MEASURE OF MODEL PERFORMANCE FOR OZONE

Model performance for simulating both ozone formation and NOoxidation is measured by ability to predict ∆([O3]-[NO]):

∆([O3]-[NO]) = ([O3]-[NO])FINAL- ([O3]-[NO])INITIAL

This gives auseful measurefor both high NOand high O3conditions

Time

Con

cent

ratio

n

O3

NO

∆([O3]-[NO])

Measures O3formation onceNO is consumed

Measuresinitial NOoxidationrate


UCR EPA RUNS AND FITS TO ∆([O3]- [NO])

10%-10%15-202Formaldehyde - CO - NOx

ErrorBias

15%-11%2 - 11024Ambient Surrogate - NOx

18%-17%5 - 306Aromatic - NOx + CO

10%10%5 - 254Toluene or m-Xylene - NOx

16%16%5 - 252Propene - NOx

15%-15%10 - 252Ethene - NOx

23%-23%8 - 252Formaldehyde - NOX

31%-2%0 - 20032Characterization

Average ModelFits

NOxRange(ppb)

RunsRun Type


SUMMARY OF TVA RUNS AND FITS TO ∆([O3]- [NO]

ErrorBias

9%-8%25 - 16912Ambient Surrogate - NOx

7%6%50 - 10023Simple Mixture - NOx

8%1%50 - 2665Toluene or m-Xylene - NOx

10%10%22 - 547Ethene, Propene, ortrans-2-Butene - NOx

28%-28%181Isopentane - NOx

10%-4%39 - 424Formaldehyde - NOx

15%1%0 - 5432Characterization

Average ModelFits

NOxRange(ppb)

RunsRun Type


CSIRO RUNS AND FITS TO ∆([O3]- [NO]

ErrorBias

42%27%33%

-42%-26%-33%

17 - 10020

Ambient Surrogate – NOx

Standard Char. ModelHigh Radical SourceHigh Light Intensity

Average ModelFits

NOxRange(ppb)

RunsRun Type


EFFECTS OF ALTERNATIVE CHARACTERIZATIONASSUMPTIONS ON CSIRO SIMULATIONS

-20%

0%

20%

40%

60%

80%

0 2 4 6 8 10 12

ROG / NOx Ratio

∆([

O3]

-[N

O])

Mod

el U

nder

pred

ictio

n E

rror

(Exp

erim

ent -

Mod

el)

/ Exp

erim

ent

Standard Model

High Radical Source

High Light Intensity

∆([O3]-[NO]) Underprediction error vs. Surrogate / NOx ratio


MODEL BIASES FOR PREDICTION OF ∆([O3]-[NO])

-40% -20% 0% 20% 40%

All Characterization

HCHO - NOx

HCHO - CO - NOx

Isopentane - NOx (TVA)

Ethene - NOx

Propene - NOx

trans-2-Butene - NOx (TVA)

Toluene - NOx

m-Xylene - NOx

Toluene - CO - NOx (UCR)

m-Xylene - CO - NOx (UCR)

Simple Mixes (TVA)

Surrogate - NOx

UCR EPA

TVA

CSIRO

-42%


LOWEST NOx UCR EPA SURROGATE EXPERIMENT(ROG SURROGATE = 300 PPBC, NOx = 2 PPB)

Concentration (ppm) vs Time (minutes)

O3

0.00

0.02

0.04

0.06

0 120 240 360

NO

0.000

0.001

0.002

0 120 240 360

NO2

0.0000

0.0004

0.0008

0.0012

0.0016

0 120 240 360

M-XYLENE

0.000

0.002

0.004

0.006

0.008

0 120 240 360

PAN

0.0000

0.0002

0.0004

0.0006

0.0008

0 120 240 360

NOy - HNO3

0.000

0.001

0.002

0.003

0.004

0 120 240 360

Experimental Standard ModelNo HONO Offgasing Maximum HONO Offgasing


MODEL UNDERPREDICTION BIASES FORSURROGATE – NOx RUNS VS. ROG/NOx RATIO

-20%

0%

20%

40%

60%

80%

0.1 1 10

ROG / NOx Ratio Relative to ROG / NOx Giving Maximum O3

∆([

O3]-

[NO

]) M

odel

und

erpr

edic

tion

erro

r

(Exp

erim

enta

l - M

odel

) / E

xper

imen

tal

UCR EPA

CSIRO

TVA

MOIR

Approximate MIR


MODEL UNDERPREDICTION BIASES FORSURROGATE – NOx RUNS VS. ROG/NOx RATIO

-20%

0%

20%

40%

60%

80%

0.1 1 10

ROG / NOx Ratio Relative to ROG / NOx Giving Maximum O3

∆([

O3]-

[NO

]) M

odel

und

erpr

edic

tion

erro

r

(Exp

erim

enta

l - M

odel

) / E

xper

imen

tal

UCR EPA

CSIRO

TVA

UCR EPA (New Runs)

MOIR

Approximate MIR


MODEL UNDERPREDICTION SENSITIVITIESFOR UCR EPA RUNS

This includesmore recentruns carriedout for EPAOBM project

-10%

0%

10%

20%

30%

40%

50%

0.1 1 10

(O3 -

NO

) Mod

el u

nder

pred

ictio

n B

ias

SAPRC-99 (Range gives low and high Walls->HONO)

Reduced OH + NO2

ROG/NOx Ratio relative to ratio giving maximum O3


COMPARISON OF CB4 AND SAPRC-99MODEL ERRORS FOR UCR EPA SURROGATE RUNS

This includesmore recentruns carriedout for EPAOBM project

ROG/NOx Ratio relative to ratio giving maximum O3

-20%

0%

20%

40%

60%

80%

0.1 1 10

(O3 -

NO

) Mod

el u

nder

pred

ictio

n B

ias

SAPRC-99 CB4


COMPARISON OF CURRENT SURROGATE FITSWITH DATA USED IN PREVIOUS EVALUATION

ROG (OH reactivity as propene equivalents) / NOx (ppm)

Mod

el U

nder

pred

ictio

n E

rror

(E

xper

imen

tal -

Cal

cula

tion)

/ E

xper

imen

tal

-50%

-25%

0%

25%

50%

0 2 4 6

Mini-Surrogate

-50%

-25%

0%

25%

50%

0 2 4 6

4 Comp. Surrogate(1982)

-50%

-25%

0%

25%

50%

0 2 4 6

EC 7 HC Mix

-50%

-25%

0%

25%

50%

0 2 4 6

8-ComponentSurrogate (1993-99)

-50%

-25%

0%

25%

50%

0 2 4 6

8-ComponentSurrogate (1983-84)

-50%

-25%

0%

25%

50%

0 2 4 6

UCR EPA Surrogate(2003)

ROG/NOx Offscale (9-14)


MODEL PERFORMANCE IN SIMULATING OTHERMEASUREMENTS IN SURROGATE RUNS

Generally consistent withavailable data ([NOx]>50 ppb)

No dataHNO3

Tendency to underpredict finalconsumption rates

Generally consistentwith data

ReactantVOCs

Tendency to underpredictdepends on NOx levels

Tendency tounderpredict (dependson offgasing model)

Formaldehyde

Reasonably well simulated inruns where O3 well simulated

Very significantlyunderpredicted

PAN

Tends to underpredictconsumption following maximum

No data (NOY-NOonly)

NO2

UCR EPATVACompounds

Note: Data for above compounds not available for CSIRO Runs


MODEL PERFORMANCE SIMULATINGFORMALDEHYDE IN UCR EPA SURROGATE RUNS

-25%

0%

25%

50%

75%

- 50 100 150 200

ROG/NOx

For

mal

dehy

de In

crea

se

Und

erpr

edic

tion

Bia

s

-25%

0%

25%

50%

75%

- 50 100 150

NOx (ppb)

For

mal

dehy

de In

crea

se

Und

erpr

edic

tion

Bia

s

Formaldehyde Underprediction bias vs. ROG/NOx or NOx(experimental - calculated) / experimental ∆[HCHO]


MODEL PERFORMANCE SIMULATING PANIN UCR EPA SURROGATE RUNS

PAN Underprediction bias vs. ROG/NOx or NOx(experimental - calculated) / experimental final PAN

-50%

0%

50%

100%

- 50 100 150 200

ROG/NOx

PA

N U

nder

pred

ictio

n B

ias

-50%

0%

50%

100%

-25% 0% 25% 50%

∆([O3]-[NO]) Underprediction BiasP

AN

Und

erpr

edic

tion

Bia

s


EFFECT OF CO ON AROMATIC - NOx RUNSToluene m-Xylene

OzoneOzone Change(effect of CO)

OzoneOzone Change(effect of CO)

EPA077 (~25 ppb NOx) EPA067 (~20 ppb NOx)

Co

nce

ntr

atio

n (

pp

m)

Time (minutes) Time (minutes)

EPA066 (~5 ppb NOx)

Co

nce

ntr

atio

n (

pp

m)

Time (minutes)

0.00

0.05

0.10

0.15

0.20

0.25

0 120 240 360

Aromatic - NOx experiment

Aromatic - CO - NOx experiment

Aromatic - NOx model

Aromatic - CO - NOx model

0.00

0.04

0.08

0.12

0 120 240 3600.00

0.03

0.06

0.09

0.12

0 120 240 3600.00

0.02

0.04

0.06

0 120 240 360

0.00

0.02

0.04

0.06

0.08

0.10

0 180 360 540 -0.01

0.00

0.01

0.02

0.03

0.04

0 180 360 540


EFFECTS OF ADJUSTMENTS TO PARAMETERIZEDTOLUENE MECHANISM

Effect of CO (EPA074) Toluene - NOx Runs

Time (minutes)

∆([O

3]-[

NO

]) C

hang

e (p

pm)

∆([O

3]-[N

O])

(ppm

)

CTC079

0.0

0.1

0.2

0.3

0.4

0 120 240 360

EPA074A

0.00

0.05

0.10

0.15

0 120 240 360

0.00

0.05

0.10

0.15

0 120 240 360

Time (minutes)

Data

Standard Mechanism

Direct Reactivity Adjusted

Higher Radical Yield

∆([

O3]

-[N

O])

Cha

nge

(ppm

)

∆([

O3]

-[N

O])

(pp

m)


ASSESSMENT OF NEED TO MODIFY MECHANISMFOR LOW NOx CONDITIONS

• Lower NOx levels result in increased importance of peroxy +peroxy reactions, which form different organic products that theperoxy + NO reactions that dominate when NOx is higher.

• SAPRC-99 uses an approximate “chemical operator” method forRO2 reactions that neglects this change in products with NOx

• Representing RO2+RO2 reactions more explicitly requiresadding many reactions to the mechanism.

• Process analysis calculations were carried out to assess therelative importance of the different types of competing RO2reactions at low NOx

• The results can then be used to assess priorities for mechanismmodifications for more accurate low NOx predictions


MODEL SIMULATIONS OF PEROXY REACTIONINTEGRATED REACTION RATE RATIOS

Ratios of integrated reaction rates vs.NOx inputs relative to NOx giving maximum O3

Averaged Conditions EKMA5-Day Stagnation with Continuous Emissions

5-Day Stagnation withEmissions on Day 1 only

0.001 0.01 0.1 10%

20%

40%

60%

80%

100%

0.001 0.01 0.1 1 0.001 0.01 0.1 1

RO2 + NO RO2 + HO2 RO2 + RCO3 RO2+RO2


• Results of test calculations of integrated peroxy radical reactionrates with three types of low NOx scenarios indicate that:• Major low-NOx sink for peroxy radicals is reaction with HO2

• Maximum importance of Alkyl Peroxy + Acyl Peroxy reactionis ~10%

• Alkyl peroxy + Alkyl peroxy reactions negligible

• Adding reactions or species to improve representations oforganic peroxy + peroxy reactions probably not worthwhile

• Higher priority is improving representation of:

RO2 + HO2 → ROOH + O2ROOH + OH or hν → Products

• This requires adding more hydroperoxide species to themechanism, whose reactions are uncertain.

IMPLICATIONS CONCERNING HOW TO MODIFYMECHANISMS FOR LOW NOx CONDITIONS


DISCUSSION AND CONCLUSIONS:Good News

• Range of conductions where mechanisms have been evaluatedhas been significantly expanded• TVA data of comparable or better quality than previous runs

in other chambers, despite formaldehyde contamination• CSIRO data proved to be useful in this evaluation, but more

characterization information would increase its utility• Lower background in UCR EPA chamber permitted useful

mechanism evaluation with NOx as low as ~2 ppb andimproves precision of evaluation

• Simulations of lower NOx characterization and most simpleVOC – NOx experiments generally satisfactory

• No apparent mechanism problem simulating very low NOxconditions where maximum O3 formation potentials achieved.

• Inaccuracies caused by approximate treatment of RO2+RO2reactions in current mechanism are probably not important


DISCUSSION AND CONCLUSIONS:Bad News

• SAPRC-99 mechanism has consistent bias in underpredictingNO oxidation and O3 formation rates at low ROG/NOx ratios

• Underprediction bias for CB4 even worse

• Significant problems with current aromatics mechanisms:• Ozone increase caused by adding CO “radical amplifier” to

aromatic – NOx systems underpredicted by factor of ~2• Adjustments to the mechanisms as currently parameterized

cannot correct this problem• Aromatics mechanism problems may be the cause of the low

ROG/NOx underprediction bias, but this is not certain.

• Improving accuracy of low NOx organic product predictions forrequires explicit treatment of uncertain organic hydroxides


RESEARCH NEEDS

• The current parameterized aromatics mechanisms need to bereformulated and made consistent with available data

• The underprediction problem at low ROG/NOx needs to beinvestigated and resolved

• Data are needed on effects of temperature on ozone and othersecondary products. (UCR EPA chamber suitable for this)

• Well-characterized, low NOx chamber data are needed todevelop and test models for secondary PM formation. (UCREPA chamber also suitable for this)

• SAPRC-99 needs to be updated to be consistent with latestrecommendations and made more compatible with SOA models

• A new condensed mechanism, traceable to an updated andevaluated detailed version, needs to be developed to finallyreplace the out-of-date CB4 mechanism

evaluation of a gas-phase atmospheric mechanism for low ... · pdf fileevaluation of a...

Documents