Regional ModelingUpdate and Issues

May 6, 2003

Air Resources Board

California Environmental Protection Agency

Luis F. Woodhouse, Ph.D.

RegionalModeling

IntegratedResults

Risk Assessment

Mapping andVisualization

MicroscaleModeling

Emissions andMeteorology

Outline

• Review of last meeting

• Regional modeling update

• Model evaluation

• Comparison with previous studies

• Integrating microscale and regional modeling

• Future analysis

• Future statewide modeling considerations

3

Review of Last Meeting (September 12, 2002)

• Previous studies– UAM and CAMx with Carbon Bond IV

– Select toxics

– Small domain

• Present study– CALGRID and CMAQ with SAPRC99

– Over 30 toxics

– Large domain• Note: CAMx not used since it’s implementation mechanism

software is not publicly available

4

Toxics– 1,3-butadiene– Formaldehyde– Acetaldehyde– Acrolein – Benzene– Carbon tetrachloride– Chloroform– Dichloromethane– 1,2-Dichloroethane– o-Dichlorobenzene– p-Dichlorobenzene– Ethylene oxide– Styrene

– Toluene– Vinyl Chloride– Xylenes– Hexavalent Chromium

– Diesel PM10

– PM10 Arsenic

– PM10 Beryllium

– PM10 Cadmium

– PM10 Lead

– PM10 Manganese

– PM10 Mercury

– PM10 Nickel

– PM10 Zinc

5

Regional Modeling Domain

San Diego

Riverside

Los AngelesSan BernardinoVentura

Orange

Mexico

6

Regional Modeling Domain

93,264 km2

87 x 67 grids(4 km x 4 km)

Model Inputs • Emissions

– SCOS97 adjusted to 1998– seasonal inventories (weekday/weekend)– latest profiles, surrogates, and EMFAC2000

(with DTIM4)

• Meteorology– CALMET: diagnostic model using data from over

200 sites– MM5: prognostic model

• Boundary conditions– same for each month, based on SCOS97

7

Regional Modeling Update

• CALGRID– January 1 to December 31, 1998

• CMAQ– January, April, August and November 1998

8

Model Performance

• Verify model’s ability to reproduce measured concentrations– Ozone: Performance standards are well

established

– Toxics: No established performance standards

9

Model PerformanceConclusions

• Iterative process is needed to improve ozone performance

• In general, model predicted annual average toxics concentrations are comparable with observations for most species

• Results comparable with previous studies

10

Ozone Model Evaluation• Compared daily ratios of model-predicted to

measured maximum ozone concentrations– CALGRID closer to observations– CMAQ over predicts

11

0

0.5

1

1.5

2

2.5

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

Day of Month (August 1998)

Ratio

CMAQ CALGRID

Ozone Model Evaluation (cont.)• Calculated daily average gross errors:

– Measure model’s overall ability to reproduce observed hourly ozone at each site above a specified threshold concentration

– Iterative process

12

0

0.2

0.4

0.6

0.8

1

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

Day of Month (August 1998)CMAQ CALGRID

Toxic VOCs Model Evaluation

• Annual average concentrations– In general, model predictions are

comparable with the measured annual concentrations for most toxics VOC species

– Some species are significantly under predicted by both models: carbon tetrachloride, chloroform, ethylene chloride, styrene

13

1998 Annual Average Concentration in Los Angeles

0

1

2

3

4

5

6

7

AC

ET

BU

TD

C6H

6

C7H

8

CC

HO

CC

L4

CH

LO

DIC

M

HC

HO

MEK

PDC

B

PER

C

STY

R

TED

C

VC

HL

ppb

CALGRID CMAQ Observed

Annual Averages of Toxic VOCs

14

1998 Annual Average Concentrations in Anaheim

0

1

2

34

5

6

7

AC

ET

BU

TD

C6H

6

C7H

8

CC

HO

CC

L4

CH

LO

DIC

M

HC

HO

MEK

PD

CB

PER

C

STY

R

TED

C

VC

HL

ppb

CALGRID CMAQ Observed

1998 Annual Average Concentration in Chula Vista

00.5

11.5

22.5

33.5

AC

ET

BU

TD

C6H

6

C7H

8

CC

HO

CC

L4

CH

LO

DIC

M

HC

HO

MEK

PD

CB

PER

C

STY

R

TED

C

VC

HL

ppb

CALGRID CMAQ Observed

15

Annual Averages of Toxic VOCs

Annual Averages ofInert Toxics

• Diesel PM10

– Model predictions are comparable to observed elemental carbon results

• Hexavalent Chromium– Model predictions are below detection limit

• PM10 components

– Performance depends on species

16

Annual Average of Inert Toxics

1998 Annual Concentrations at Los Angeles

0

50

100

150

200

ARSE CADM CRVI DIES LEAD MERC NICK ZINC

ng/m

3

CALGRID CMAQ Observed

*

* DIES in ug/m3 compared to elemental carbon 17

1998 Annual PM10 Concentrations at Anaheim

0

50

100

150

ARSE CADM CRVI DIES LEAD MERC NICK ZINC

ng/m

3

CALGRID CMAQ Observed

1998 Annual Concentration at Chula Vista

0

20

40

60

ARSE CADM CRVI DIES LEAD MERC NICK ZINC

ng/m

3

CALGRID CMAQ Observed

18

*

*

*

* DIES in ug/m3 compared to elemental carbon

Annual Average of Inert Toxics

19μg/m3

CALGRID (1998)

ppb

Diesel PM10 Benzene

Comparison withPrevious Studies

• MATES II– April 1998 to March 1999 field study– Models

• UAM and recently CAMx• Carbon Bond IV reaction mechanism

• Our results are comparable

20

0

0.5

1

1.5

2

2.5

ppb

CALGRID CMAQ MATESII (UAM) Observed

Comparison with MATES IIBenzene

21

Comparison with MATES II Diesel PM10 vs. Elemental Carbon

02468

101214

ug/m

3

CALGRID CMAQ MATESII (UAM) Observed

22

Comparison with MATES II Formaldehyde

01

234

56

ANAH

BLO

G

BUR

K

CEL

A

CH

VA

CO

MP

FON

T

HPR

K

LGBH

PIC

O

RIV

R

SIM

I

UPL

A

WIL

M

ppb

CALGRID CMAQ MATESII (UAM) Observed

23

Integrating Microscale and Regional Modeling Results

• Microscale modeling estimates near source impacts (meters)

• Regional modeling estimates impacts from sources in a large area (km)

• Issue– double-counting

24

25

Barrio Logan Modeling Results

ISCST3 CALINE CALGRID BARRIO CHULA EL -----------BARRIO LOGAN-------------- LOGAN VISTA CAJON

DIESEL PM10

26

ISCST3 CALINE CALGRID BARRIO CHULA EL -----------BARRIO LOGAN-------------- LOGAN VISTA CAJON

BENZENE

Barrio Logan Modeling Results (cont.)

27

ISCST3 CALINE CALGRID BARRIO CHULA EL ---------BARRIO LOGAN-------------- LOGAN VISTA CAJON

HEXAVALENT CHROMIUM

Barrio Logan Modeling Results (cont.)

NA

Sensitivity Simulations Double Counting*

• In Barrio Logan, local emissions contribute less than 1% of the annual average concentration of most toxic species.

• In Wilmington, local emissions contribute 15%-90% of the annual average concentrations– Benzene (47%)– Diesel exhaust (40%)– 1,3-butadiene (16%)

28* simulations for all 1998 were done in each case with CALGRID

Sensitivity SimulationsBarrio Logan

• Changing boundary conditions has very small impact on annual average toxic concentrations

• Choosing different averaging periods – 12-month average toxic concentrations can be

significantly different from 4-month average concentrations

– 4-month average cumulative risk is about 10% higher than the 12-month average cumulative risk

29

Future Analysis• Improve estimates of background toxic

concentrations – Omit all toxics emissions in a cell– Omit toxic emissions from selected categories in a cell– Evaluate procedures for estimating contributions of

secondary species

• Evaluate deposition effect

• Run CALGRID using MM5 winds

• Conduct spatial analysis

30

Future Statewide Modeling Considerations

• Air quality model selection– CALGRID, CMAQ, CAMx, other – Atmospheric reaction mechanism– Run time (e.g., CALGRID with SAPRC99

and 4 km x 4 km grids, at least 6 months)

• Period simulated– Every day in a year or selected episodes

31

Future Statewide Modeling Considerations (cont.)

• Input preparation– Emissions – Meteorology (CALMET, MM5)

• Other considerations– Baseline year– Multiple year simulation– Storage requirements

32