Envair – C. Blanchard and S. Tanenbaum DRI – E. Fujita and D. Campbell

1

Understanding Relationships Between Changes in Ambient Ozone and

Precursor Concentrations and Changes in VOC and NOx Emissions

from 1990 to 2004 in Central California

Envair – C. Blanchard and S. Tanenbaum

DRI – E. Fujita and D. Campbell

Alpine Geophysics – J. Wilkinson

Phase I Findings and Proposed Phase II ApproachMay 31, 2007

2

Acknowledgements

CARB

Dar Mims, Dwight Oda, Don Johnson, Martin Johnson, Larry Larsen, Cheryl Taylor

BAAQMD

Jim Cordova, Saffet Tanrikulu

SJVAPCD

Evan Shipp

… plus everyone else for review and assistance

3

Today’s TopicsI. Project overviewII. Trend analysis summary III. Trends in ozone precursors IV. Trends in precursors compared with county-

level emissionsV. Spatial variations of ozone precursors

=> Questions

VI. Ozone trendsVII. Meteorological classificationVIII. Do met classes shed light on ozone trends?

=> Questions

IX. Phase II analyses and schedule

4

I. Project Overview• Phase I

– Develop databases and methods– Characterize ozone and precursor trends by site,

subregion – compare precursor trends to county-level emission trends – characterize ozone trends by meteorological class

– Evaluate prospects for success in Phase II

• Phase II– Grid emissions and relate ambient primary-pollutant

trends to zone-of-influence emission trends– Relate ozone trends to ambient and emission trends of

precursors, and to meteorological conditions– Submit final report

5

II. Trends Summary

6

Sub-Region

N orthern S ierra Footh ills

Southern S ierra Footh ills

Sacram ento Valley

N orthern SF Bay A rea

Eastern SF Bay A reaSouthern SF Bay A reaN orthern San Joaquin

C entra l San Joaquin

Southern San Joaquin

AQ Sites and Subregions

7

AQ Metrics – Annual Averages• Ozone

– Annual 4th-highest daily 8-hour max, each site– Annual mean of top-60 peak 8-hour days, each site

(the top-60 days are determined for each subregion)

• CO and NOx– Annual means from top-60 days, each site– Morning (start hours 5 am – 10 am)– Time of peak 8-hour ozone maxima (“mid-day”)

• NMOC – Annual means from all days, by site– Early morning - 5 am PST (most complete sampling)

8

1990 – 2004 AQ Trends Summary Results

• No trends significantly upward*• NOx sig* down at 22 of 28 sites**• CO sig* down at 21 of 25 sites**• NMOC sig* down at 5 of 7 sites***• Ozone sig* down at 7 of 42 sites**• Annual mean top-60 ozone trends similar

to trends in annual 4th-highest 8-hour max

* p < 0.05** At least 10 years data. One or both metrics.*** 7 - 10 years data

9

III. Trends in Ozone Precursors

10Morning NOx decline: 0.66 ppbv/year (~10 ppbv over period)Mid-day NOx decline: 0.31 ppbv/year (~5 ppbv over period)

Fresno First Street - Mean Morning ( 5 am - 11 am) and Mid-day (Time of 8-hour O3 Max) NOx on Top 60 Subregional Days

0

10

20

30

40

50

60

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

NO

x (p

pb

v)

11

Parlier - Mean NOx on Top 60 Subregional High-Ozone DaysMorning ( 5 am - 11 am) and Mid-day (Time of 8-hour Ozone Max)

0

5

10

15

20

25

30

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

NO

x (p

pb

v)

Morning NOx decline: 0.39 ppbv/year (~4 ppbv over period)Mid-day NOx decline: 0.21 ppbv/year (~2 ppbv over period)

12

Morning NMOC - Fresno First Street

0

50

100

150

200

250

300

350

400

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

NM

OC

(p

pb

C)

Morning NMOC decline: 8.2 ppbC/year (~66 ppbv over period)

13

Mean Morning NMOC - Clovis

0

100

200

300

400

500

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

NM

OC

(p

pb

C)

Morning NMOC decline: 11.5 ppbC/year (~115 ppbv over period)

14

Mean Morning NMOC Parlier

0

50

100

150

200

250

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

NM

OC

(p

pb

C)

Morning NMOC decline: 5.1 ppbC/year (~51 ppbv over period)

15

Morning CO decline: 43 ppbv/year (~650 ppbv over period)Mid-day CO decline: 20 ppbv/year (~300 ppbv over period)

(1990 data suspect)

Fresno First Street - Mean CO on Top 60 Subregional High-Ozone DaysMorning ( 5 am - 11 am) and Mid-day (Time of 8-hour Ozone Max)

0

200

400

600

800

1000

1200

1400

1600

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

CO

(p

pb

v)

16

FRESNO

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

YEAR

NM

OC

/CO

Similar results for Bakersfield and Sacramento Del Paso: ~0.3 ppbC NMOC per ppbv CO (some years differed)

NMOC/CO from regression of NMOC against CO, by year.Annual r2 ranged from 0.65 to 0.94 (5 am samples)

NMOC vs. CO Fresno First Street, 5 am Samples, Summer 2000

0

100

200

300

400

500

600

700

0 500 1000 1500 2000 2500

CO (ppbv)

NM

OC

(p

pb

C)

17CO/NOx declines so does NMOC/NOx

Fresno First Street Morning CO/NOx

0

5

10

15

20

25

30

35

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

CO

/NO

x (d

imen

sion

less

)

18

FRESNO

0

3

6

9

12

15

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

YEAR

BE

NZ

EN

E/C

O

Benzene/CO from regression of benzene against CO, by year.Annual r2 ranged from 0.68 to 0.94, except 1998 (5 am samples)

19

IV. Comparison of AQ Trends to County-Level Emission Trends

20

Fresno County NOx Emissions Compared With Morning NOx Concentrations at Fresno First Street

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

No

rmal

ized

An

nu

al V

alu

e

NOx Emissions Morning NOx Linear (NOx Emissions) Linear (Morning NOx)

Emission trend (0.40) exceeds ambient trend (0.20)

21Ambient trend (0.80) exceeds emission trend (0.60)

Fresno County CO Emissions Compared With Morning CO Concentrations at Fresno First Street

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

No

rmal

ized

An

nu

al V

alu

e

CO Emissions Morning CO Linear (CO Emissions) Linear (Morning CO)

22

Ambient vs. Emissions Decrease for CO

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

Emissions Decrease (normalized difference)

Am

bie

nt

Dec

reas

e (n

orm

aliz

ed d

iffe

ren

ce)

Ambient decrease exceeds emissions decrease (13 sites)

Ambient decrease less than emissions decrease (12 sites)

23

Ambient vs. Emissions Decrease for NOx

-0.2

0

0.2

0.4

0.6

0.8

1

-0.2 0 0.2 0.4 0.6 0.8 1

Emissions Decrease (normalized difference)

Am

bie

nt

Dec

reas

e (n

orm

aliz

ed d

iffe

ren

ce)

Ambient decrease exceeds emissions decrease (15 sites)

Ambient decrease less than emissions decrease (13 sites)

24

Ambient and Emissions Decreases for CO

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Fremont-Chapel W

ay

Livermore

Jackson-Clinton Road

San Andreas-Gold Strike

Pittsburg-10th Street

Bethel Island Road

Concord-2975 Treat Blvd

Placerville-Gold Nugget

Fresno-Drumm

ond Street

Parlier

Fresno-Sierra Skypark #2

Fresno-1st Street

Clovis-N Villa Avenue

Edison

Oildale-3311 Manor Stree

Arvin-Bear Mountain Blvd

Bakersfield-Golden State

Bakersfield-5558 Califor

Merced-S Coffee Avenue

Roseville-N Sunrise Blvd

North Highlands-Blackfoo

Sacramento-Del Paso M

ano

Elk Grove-Bruceville Roa

Sacramento-T Street

Folsom

Stockton-Hazelton Street

Gilroy-9th Street

San Jose-4th Street

Vallejo-304 Tuolumne Str

Modesto-14th Street

Turlock-S Minaret Street

Visalia-N Church Street

Sonora-Barretta Street

Site

Dec

reas

e (N

orm

aliz

ed V

alu

es)

Ambient CO Decrease at Sites Emissions CO Decrease in County

25

Ambient and Emissions Decreases for NOx

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Fremont-Chapel W

ay

Livermore

Jackson-Clinton Road

San Andreas-Gold Strike

Pittsburg-10th Street

Bethel Island Road

Concord-2975 Treat Blvd

Placerville-Gold Nugget

Fresno-Drumm

ond Street

Parlier

Fresno-Sierra Skypark #2

Fresno-1st Street

Clovis-N Villa Avenue

Edison

Oildale-3311 Manor Stree

Arvin-Bear Mountain Blvd

Bakersfield-Golden State

Bakersfield-5558 Califor

Merced-S Coffee Avenue

Roseville-N Sunrise Blvd

North Highlands-Blackfoo

Sacramento-Del Paso M

ano

Elk Grove-Bruceville Roa

Sacramento-T Street

Folsom

Stockton-Hazelton Street

Gilroy-9th Street

San Jose-4th Street

Vallejo-304 Tuolumne Str

Modesto-14th Street

Turlock-S Minaret Street

Visalia-N Church Street

Sonora-Barretta Street

Site

Dec

reas

e (N

orm

aliz

ed V

alu

es)

Ambient NOx Decrease at Sites Emissions NOx Decrease in County

26

What Did We Learn?

• On average, ambient precursor decreases are comparable to county-level emissions decreases

• There is a possibility that emission decreases are overestimated or underestimated for some counties

• Confirmation requires comparison of site trends to spatially-resolved emission trends (Phase II)

27

V. Spatial Variations of Ozone Precursors

28

Sacramento Del Paso Mean Morning Concentrations

CO NOx

29

Roseville Mean Morning Concentrations

CO NOx

30

Fresno 1st St Mean Morning Concentrations

CO NOx

31

Fresno Sierra Skypark Mean Morning Concentrations

CO NOx

32

Phase II AQ vs. Emissions• Directional variations of primary species AQ

concentrations imply significant local influences• Comparison of site primary species AQ trends to

emission inventory trends can be improved by using gridded inventories

• The Phase II comparisons will permit more robust conclusions about the differences between site and emission trends – eliminate the mismatch between spatial scales (replace county-level emissions with emissions from local zones of influence)

• Potential limitation is accuracy of gridding

33

Questions and Comments on Primary Species Phase I Findings and

Phase II Objectives

34

VI. What Are the Ozone Trends?

35Annual 4th max decline: 0.58 ppbv/year (~9 ppbv over period)Mean Top 60 decline: 0.38 ppbv/year (~6 ppbv over period)

Fresno First Street -- Annual 4th-Highest 8-Hour Ozone and Mean of Top 60 Subregional Days

0

20

40

60

80

100

120

140

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

Ozo

ne

(pp

bv)

36

Parlier -- Annual 4th-Highest 8-Hour Ozone and Mean of Top 60 Subregional Days

0

20

40

60

80

100

120

140

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

Ozo

ne

(pp

bv)

37

Livermore Annual 4th-Highest 8-Hour Ozone and Mean of Top 60 Subregional Days

0

10

20

30

40

50

60

70

80

90

100

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2000 2001 2002 2003 2004

Year

Ozo

ne

(pp

bv)

First Street | Rincon

38* Livermore 1st St only

Ozone Trends in Subregions - Medians of Site Trends

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

EasternSFB*

NorthernSFB

SouthernSFB

NorthernSJV

Central SJV

SouthernSJV

Sacra- mentoValley

NorthernSierra

Foothills

SouthernSierra

Foothills

Subregion

Ozo

ne

Tre

nd

(p

pb

v p

er y

ear)

4th Highest 8-hour maximum Top 60 8-hour maximum

39

VII. Meteorological Classification

40

Why Examine Meteorological Data?

Meteorological information may permit more complete reconciliation of ambient ozone trends with precursor and emissions trends.

Phase I. Uses meteorological information to split days into groups with different meteorological characteristics. Initial evidence indicates that ozone trends vary by site, subregion, met type.

Phase II. More detailed analyses.

41

Met Classification*:

1. Principal component analysis (PCA) of regional-scale met variables

2. K-means clustering of PCs

* Trend adjustment Forecasting Interbasin transport

PCA applied to all daysof all years from

1990 to 2004(n = 5480** days)

Clustering applied to allozone-season days

(n = 2790 days)

** 5441 with pressure gradient data; 4202 with 850 mb data

42

PCA of Regional-Scale Variables

San Francisco-to-Medford sea-level pressure gradient, daily averageSan Francisco-to-Reno sea-level pressure gradient, daily averageSan Francisco-to-Fresno sea-level pressure gradient, daily averageSan Francisco-to-Las Vegas sea-level pressure gradient, daily average

Oakland 850 mb vector component (u) wind speed and direction at 4 amOakland 850 mb vector component (v) wind speed and direction at 4 amOakland 850 mb vector component (u) wind speed and direction at 4 pmOakland 850 mb vector component (v) wind speed and direction at 4 pm

Oakland 850 mb temperature and height at 4 amOakland 850 mb temperature and height at 4 pm

43

Three PCs explain 80% of variance of eight variablesPC1 is westerly wind – PC2 is northerly wind – what is PC3?

PCA - Regional Scale Meteorological Variables

-1.5

-1

-0.5

0

0.5

1

1.5

SanFrancisco-Medfordpressuregradient

SanFrancisco-

Renopressuregradient

SanFrancisco-

Fresnopressuregradient

SanFrancisco-Las Vegaspressuregradient

Oakland850 mb u-vector windspeed anddirection, 4

am

Oakland850 mb v-

vector windspeed anddirection, 4

am

Oakland850 mb u-vector windspeed anddirection, 4

pm

Oakland850 mb v-

vector windspeed anddirection, 4

pm

Variable

Pri

nci

pal

Co

mp

on

ent

(dim

ensi

on

less

)

PC1 PC2 PC3

44

0

1

2

3

4

5

6

Res

ulta

nt W

ind

Spe

ed (

m s

-1)

-2.5 -1.5 -.5 .5 1.5 2.5Principal Component 3 (dimensionless)

d. Bakersfield California Avenue

0

1

2

3

4

5

6

Res

ulta

nt W

ind

Spe

ed (

m s

-1)

-2.5 -1.5 -.5 .5 1.5 2.5Principal Component 3 (dimensionless)

c. Clovis

0

1

2

3

4

5

6

Res

ulta

nt W

ind

Spe

ed (

m s

-1)

-2.5 -1.5 -.5 .5 1.5 2.5Principal Component 3 (dimensionless)

b. Livermore Rincon

0

1

2

3

4

5

6

Res

ulta

nt W

ind

Spe

ed (

m s

-1)

-2.5 -1.5 -.5 .5 1.5 2.5Principal Component 3 (dimensionless)

a. Sacramento Del Paso

PC3 correlates with surface wind speeds – interpret as ventilation

45

Split days into four groups using K-means clustering

46

Clusters separate days into groups with different pressure gradients

47

Clusters separate days into groups with different 850 mb wind directions

48

Clusters separate days into groups with different 850 mb T

49

12

34

S1

S2

S3

S4

0

50

100

150

200

250

300

350

400

450

Number of Days

Previous Day Cluster

Current Day Cluster

Met Clusters Compared With Previous-Day Cluster

Clusters exhibit persistence and preferred transitions(especially 4-to-1, 2-to-3, and 3-to-4)

50

SBANBA & EBA

SAC NSJV

CSJV SSJV

Daily-average surface wind speeds, all sites

51

52

What Can We Learn From Met Clusters?

• Splitting days into groups having similar meteorological conditions is useful for reducing meteorological “noise”

• Potentially may reveal differences in response of ozone to precursor changes under different source-receptor conditions or different degrees of photochemical activity and “aging”

53

VIII. Do Met Classes Shed Light on Ozone Trends?

54

Mean peak 8-hour ozone concentrations varied among sites, subregions, and met types – these days are all top-60 peak 8-hour days

55

The change in mean peak 8-hour ozone concentrations from 1995-1999 to 2000-2004 varied among sites, subregions, and met types

56

Change in Mean Peak Daily 8-Hour Ozone, 1995-99 to 2000-04

(by Meteorological Class and Subregion)

-15

-10

-5

0

5

10

15

Subregion

Ozo

ne

(p

pb

v)

SBA NBA EBA NSJ SAC NSF CSJ SSJ

Change in Mean Peak Daily 8-Hour Ozone, 1995-99 to 2000-04(by Meteorological Class and Subregion)

-15

-10

-5

0

5

10

15

Subregion

Ozo

ne

(p

pb

v)

SBA NBA EBA NSJ SAC NSF CSJ SSJSBA NBA EBA NSJ SAC NSF CSJ SSJ

57

Change in Mean Peak Daily 8-Hour Ozone, 1995-99 to 2000-04(by Meteorological Class and Subregion)

-15

-10

-5

0

5

10

15

Ozo

ne

Ch

ang

e (p

pb

v)

Type 1 Type 2 Type 3 Type 4

Change in Mean Peak Daily 8-Hour Ozone, 1995-99 to 2000-04(by Meteorological Class and Subregion)

-15

-10

-5

0

5

10

15

Ozo

ne

Ch

ang

e (p

pb

v)

Type 1 Type 2 Type 3 Type 4Type 1 Type 2 Type 3 Type 4

(Order is SBA, NBA, EBA, NSJ, SAC, NSF, CSJ, SSJ)

58

For all met classes, top-60 mean peak 8-hour ozone is down at Clovis but up at Parlier … the differences in ozone trends at Clovis and Parlier do not appear to be due to changes in meteorology or in the frequencies of occurrence of different meteorological types

59

Mean Top-60 Peak 8-hour Ozone - Cluster 1

0

20

40

60

80

100

120

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

Ozo

ne (p

pb

v)

Clovis Parlier Linear (Clovis) Linear (Parlier)

60

Mean Top-60 Peak 8-hour Ozone - Cluster 2

0

20

40

60

80

100

120

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

Ozo

ne (p

pb

v)

Clovis Parlier Linear (Clovis) Linear (Parlier)

61

Mean Top-60 8-hour Ozone - Cluster 3

0

20

40

60

80

100

120

1994 1995 1996 1997 1998 1999 2000 2001 2002 2004

Year

Ozo

ne (p

pb

v)

Clovis Parlier Linear (Clovis) Linear (Parlier)

62

Mean Top-60 Peak 8-hour Ozone - Cluster 4

0

20

40

60

80

100

120

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Year

Ozo

ne (p

pb

v)

Clovis Parlier Linear (Clovis) Linear (Parlier)

63

Next Steps – Phase II Ozone Analyses

• Phase I analyses demonstrate variations of ozone and of ozone trends but do not explain them

• Site-to-site variations and directional variations of mean concentrations imply significant local ozone formation

• Need to analyze ozone formation rates

• Potential limitation is signal-to-noise

64

Questions and Comments on

Ozone Phase I Findings and

Phase II Objectives

65

Phase II

• Phase II schedule – complete by December 2007 with final report and draft manuscript

• Why would Phase II be useful?– Identify zones of emission influence - more

accurate comparison of sites’ AQ trends with “zone-of-influence” emission trends

– Better understanding of ozone trends at sites within each subregion and the relation of ozone to precursor trends, differentiated by met class and subclass

66

Task 5

• Generate gridded inventories from county-level inventories and historical surrogate files

• Develop monitor-specific “zone-of-influence” emission trends using 3x3 to 7x7 arrays of grid cells around each long-term monitor

67

Task 6

• Compare “zone-of-influence” emission trends to ambient primary-pollutant trends. Identify consistencies and discrepancies. Evaluate evidence for inaccuracies in emission estimates.

• Subdivide met classes. Determine peak ozone changes by site, met class and subclass, and (if warranted) month and day of week. Relate ozone changes to precursor trends and meteorology.

68

Reporting

• Task 7: Prepare final report and draft manuscript

• Task 8: Provide data, documentation, and software

• Task 9: Present findings at meetings