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
DRI
Alpine Geophysics
May 25, 2006
2
Today’s Meeting – Progress Report
Data acquisition and evaluation Met classification results (version 2) Ambient AQ trends Next steps Feedback?
3
Data Acquisition and Evaluation
4
Emissions Inventories
Obtained ARB monthly historical emissions Have checked completeness & readability Status: ready for re-aggregation into county or
greater total VOC, NOx, CO
5
Ambient AQ Data
Hourly gas-phase (O3, NO, NOx, CO, HC)– Completeness– Rounding conventions
VOC: PAMS, ARB, SJVAQS (1990) – No combination of previous VOC validation provides
complete spatial & temporal coverage– Start with EPA archives and validate data
6
Meteorological Data
Upper air: Oakland soundings 4 am & 4 pm Pressure gradients: San Francisco to Medford,
Reno, and Fresno Surface wind speed and direction (applied QA
steps recommended by BAAQMD)
7
Surface Met Data QA Issues Identified by BAAQMD
Site locations Time zone Zero values vs. missing data Sudden increases or decreases in WS, T Long periods of constant WS, WD, or T Incorrect flag values Redundant data
8
Network
N CD C
C IM IS
BAAQ M D
Sites w ith 10 + years of M eteorological Data (daily w indspeed and w ind direction)
Surface MeteorologicalStations
9
Met Classification Methods
10
Met Classification Approach
Goal is to stratify days, not to adjust ozone Stratifications related to transport directions Primary clustering method is based on met
variables related to transport Secondary method is spatial correlation of
peak ozone (e.g., DRI CCOS clustering algorithm, BAAQMD analysis)
11
Met Classification Details
K-means clustering Use all days May through October each year Standardized variables 850 mb T, height, u and v wind vector components Pressure gradients SFO to Medford, Reno, Fresno Surface u and v wind vector components, 53
stations Limited to 12 cluster types
12
Classification Results
13
Summary of Findings
Met classifications yield clusters having differences in transport direction or distance
Different clusters exhibit different mean concentrations of air pollutants
Pollutant trends at a site may but do not always differ among clusters
Clusters account for about 30 – 40% of ozone variance
14
Four Met Types Dominate Jul - Oct
15
Which Clusters Have Highest Ozone?
Clusters 1, 3, 5, 6 (summer and fall) Sometimes cluster 9 (Sep and Oct) In southern SJV, also clusters 2, 7, 8, 11
(occur most frequently May and June)
16
Daily 8-Hour Peak Ozone, 1990-2004Livermore
Meteorological Type (cluster number)
17
Daily 8-Hour Peak Ozone, 1990-2004Folsom
Meteorological Type (cluster number)
18
Daily 8-Hour Peak Ozone, 1990-2004Parlier
Meteorological Type (cluster number)
19
Daily 8-Hour Peak Ozone, 1990-2004Arvin
Meteorological Type (cluster number)
20
Do clusters exhibit differences in transport direction and distance?… most easily seen using plots ofU and V wind vector components(these show direction of origin)
21
Directional Differences
22
Directional Differences
23
Directional Differences
24
Directional Differences
25
Mean Directional Differences
26
850 mb Directional Differences
27
Directional Differences
28
Main Finding of Cluster Analysis
Clusters exhibit differences in transportdirections or distances. The differencesare more pronounced at some sites than others.
29
Can Clusters Be Improved?
Maybe -
Direct comparisons of met types (clusters) to groupings determined from wind speed and direction at eachsite suggest room for improvement.
30
Calm<1m/s
NE1-2m/s
SE1-2m/s
SW1-2m/s
NW1-2m/s
NE>2m/s
SE>2m/s
SW>2m/s
NW>2m/s
Chabot
83 8 1 12 34 3 2 5
6 41 12
12 64 109 44 55
4 7 4 16
1 3
16 5
1 5
1 2 21 4
36 24 20 1 3 5
1 42 10
1 4 2 10 11 1 4
1 4 2 8
1
2
3
4
5
6
7
8
9
10
11
12
Direction/Windspeed
Clu
ster
2
100
62 5
31 126
2
166 23
31
Sacramento
212 1 13 27 41 6 1
5 157
148 26 207 19 14 131
16 1 10 7 25
22 16 286
6 5 59 30 319
1 3 18 233
20 5 85 23 7 82 12
24 9 55 1 99
7 4 9 1 5 69 2
1 62
2 1 38
1
3
4
5
6
7
8
9
10
11
12
Direction/Windspeed
Clu
ster
Calm<1m/s
NE1-2m/s
SE1-2m/s
SW1-2m/s
NW1-2m/s
NE>2m/s
SE>2m/s
NW>2m/s
32 5
12
SW>2m/s
21
32
Fresno
290 3 1
1 2 4 71 59
346 1 2 2 186 5
38 1 15 5 13 3
79 2 7 196
25 1 1 226 1 159
3 18 2 231
2 30 198
77 1 1 90 15
3 1 12 79
39 1 13 6
1 40
1
3
4
5
6
7
8
9
10
11
12
Direction/Windspeed
Clu
ster
Calm<1m/s
NE1-2m/s
SE1-2m/s
SW1-2m/s
NW1-2m/s
NE>2m/s
SE>2m/s
SW>2m/s
NW>2m/s
302
6
3
50
33
Preliminary Trends Assessment
34
Trends Methods
Split subregional top 60 days by cluster For each site and each cluster, compute linear
regression of AQ metric against day Determine regression slope and SE
35
Do trends differ among clusters?
36
CO Trends
37
Daily Morning CO – All Sites
38
Fremont
-140
-120
-100
-80
-60
-40
-20
0
20
40
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
39
Sacramento - T St.
-140
-120
-100
-80
-60
-40
-20
0
20
40
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
40
Fresno - Drummond
-140
-120
-100
-80
-60
-40
-20
0
20
40
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
41
Bakersfield
-140
-120
-100
-80
-60
-40
-20
0
20
40
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
42
Sacramento and Northern Sierra Foothills - Cluster 1
-120
-100
-80
-60
-40
-20
0
20
40
60
80
Sacramento-TStreet
Elk Grove-Bruceville
Road
NorthHighlands-
Blackfoot Way
Sacramento-Del Paso
Manor
Roseville-NSunrise Blvd
Folsom-Natoma Street
Auburn-Dewitt-CAvenue
Grass Valley-Litton Building
Placerville-Gold Nugget
Way
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
43
Sacramento and Northern Sierra Foothills - Cluster 3
-120
-100
-80
-60
-40
-20
0
20
40
60
80
Sacramento-TStreet
Elk Grove-Bruceville
Road
NorthHighlands-
Blackfoot Way
Sacramento-Del Paso
Manor
Roseville-NSunrise Blvd
Folsom-Natoma Street
Auburn-Dewitt-CAvenue
Grass Valley-Litton Building
Placerville-Gold Nugget
Way
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
44
Sacramento and Northern Sierra Foothills - Cluster 5
-120
-100
-80
-60
-40
-20
0
20
40
60
80
Sacramento-TStreet
Elk Grove-Bruceville
Road
NorthHighlands-
Blackfoot Way
Sacramento-Del Paso
Manor
Roseville-NSunrise Blvd
Folsom-Natoma Street
Auburn-Dewitt-CAvenue
Grass Valley-Litton Building
Placerville-Gold Nugget
Way
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
45
Sacramento and Northern Sierra Foothills - Cluster 6
-120
-100
-80
-60
-40
-20
0
20
40
60
80
Sacramento-TStreet
Elk Grove-Bruceville
Road
NorthHighlands-
Blackfoot Way
Sacramento-Del Paso
Manor
Roseville-NSunrise Blvd
Folsom-Natoma Street
Auburn-Dewitt-CAvenue
Grass Valley-Litton Building
Placerville-Gold Nugget
Way
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
46
Sacramento and Northern Sierra Foothills - Cluster 9
-120
-100
-80
-60
-40
-20
0
20
40
60
80
Sacramento-TStreet
Elk Grove-Bruceville Road
NorthHighlands-
Blackfoot Way
Sacramento-DelPaso Manor
Roseville-NSunrise Blvd
Folsom-NatomaStreet
Auburn-Dewitt-CAvenue
Placerville-GoldNugget Way
Mo
rnin
g C
O (
pp
bv/
year
)
Error bars = 2 SE of slope
47
NOx Trends
48
Daily Morning NOx – All Sites
49
Fremont
-10
-8
-6
-4
-2
0
2
4
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g N
Ox
(pp
bv/
year
)
Error bars = 2 SE of slope
50
Sacramento - T St.
-10
-8
-6
-4
-2
0
2
4
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g N
Ox
(pp
bv/
year
)
Error bars = 2 SE of slope
51
Fresno - Drummond
-10
-8
-6
-4
-2
0
2
4
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g N
Ox
(pp
bv/
year
)
Error bars = 2 SE of slope
52
Bakersfield
-10
-8
-6
-4
-2
0
2
4
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Mo
rnin
g N
Ox
(pp
bv/
year
)
Error bars = 2 SE of slope
53
Ozone Trends
54
Daily Peak 8-Hour Ozone – All Sites
55
Livermore
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Pea
k 8-
ho
ur
Ozo
ne
(pp
bv/
year
)
Error bars = 2 SE of slope
56
Folsom
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Pea
k 8-
ho
ur
Ozo
ne
(pp
bv/
year
)
Error bars = 2 SE of slope
57
Parlier
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Pea
k 8-
ho
ur
Ozo
ne
(pp
bv/
year
)
Error bars = 2 SE of slope
58
Arvin
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Cluster 1 Cluster 2 Cluster 3 Cluster 5 Cluster 6 Cluster 7 Cluster 8 Cluster 9 Cluster 11
Pea
k 8-
ho
ur
Ozo
ne
(pp
bv/
year
)
Error bars = 2 SE of slope
59
What Have We Learned So Far?
Met clusters ~30 – 40% of ozone variability Clusters exhibit differences in mean transport
directions and distances. Might be able to improve distinctions among clusters.
Trends vary among clusters & sites Hypothesis: variations of ambient trends are a
function of spatial variations of emission trends
60
Next Steps –
Compare ambient to emission trends – county level or groups of counties
Draft Phase I report & Phase II plan – July 1 Task 4 meeting – ? Final Phase I report & Phase II plan If Phase II approved, compare ambient trends to
emission trends occurring in the upwind areas of influence (need to grid emission estimates)
Top Related