1. MANE-VU Contribution Assessment Efforts: Trajectory Cluster Analysis
Goal of presentation:
Present some information on modeling visibility conditions and hopefully stimulate discussion about how this information should be applied in developing a workplan for the MANE-VU planning process.
Gary Kleiman, Iyad Kheirbek, Nicolas Hamel, John Graham, and Ingrid Ulbrich
NESCAUM
MANE-VU/MARAMA Science Meeting, January 27, 2004

2. Contribution Assessment
1. Introduction
Overview of RH regulations, RPO process and timeline, purpose of contribution assessment
2. Conceptual model of regional haze in MANE-VU region
a. NARSTO PM and O3 Assessments
b. Previous NESCAUM & Other RPO Reports
c. Watson’s critical review (AWMA)
d. IMPROVE, NAPAP Reports
3. Overview of monitoring results
a. Spatial patterns, temporal patterns
b. Routine monitoring, special programs
c. Draw on NESCAUM “2002: A Year in Review” Report

3. Contribution Assessment
4. Modeling/data analysis techniques
a. Trajectory Analysis
i NESCAUM HYSPLIT analysis
ii VIEWS/ATAD analysis
b. Source Apportionment
i Phase I and II studies via CATT tool
ii CARAPACE
iii SOAP
c. Chemical Transport Models
i REMSAD 7.10 w/source tagging
ii CMAQ for episodes (August 2002)
iii CALGRID (Eastern U.S.)
d. Dispersion Modeling
i CALPUFF (Eastern U.S.)

4. Contribution Assessment
5. Synthesis of Trajectory, Receptor, Chemical Transport and Source Tagging
Models
a. Synthesize results
b. Reconcile with observations
6. Conclusions
Summarize findings with respect to contributions

5. PATH Analysis Patterns in Atmospheric Transport History
Based on prior work by Jenny Moody (UVA) at Harvard Forest in Central MA
Groups similar trajectories in 3-dimensions

6. PATH Algorithm Similarity quantified as distance between trajectories:
(Z represent normalized coordinates with 0 mean
and standard deviation of 1)

7. Cluster Definition
Clusters are defined by establishing a radius of proximity R:
The maximal cluster (largest number of elements) is
removed from the analysis and the process is repeated

8. Acadia 1997
ALL TRAJECTORIES
CLUSTER 3 TRAJECTORIES

9. Acadia: 6 Clusters Acadia, 1997-2002, 48hr, 500m, R=10 1 2 3 4 5 6 7
(hr/104km2)

10. Free Parameters Radius of proximity Height of starting point
Length of cluster calculation (e.g. 24 hr, 48 hr, 120hr)
Temporal range (e.g. 97 vs )
Receptor site
Resolution/completeness of met data

11. Radius of Proximity Acadia, 1997-2002, 48hr, 500m, R=10 1 2 3 4 5 6 7
N1=4294
N2=1493
N3=1009
N4=695
N5=632
N6=405
N7=337
N8=284
(hr/104km2)

12. Radius of Proximity Acadia, 1997-2002, 48hr, 500m, R=6 1 2 3 4 5 6 7
N1=1165
N2=710
N3=588
N4=511
N5=500
N6=467
N7=369
N8=331
(hr/104km2)

13. Starting height Acadia, 1997-2002, 48hr, 200/500/1000m, R=10 Cluster 4
Receptor Level= 200m
Mean height= 435 m
Cluster 2
Receptor Level= 500m
Mean height= 564 m
Cluster 3
Receptor Level= 1000m
Mean height= 823 m
(hr/104km2)

14. Starting height Acadia, 1997-2002, 48hr, 200/500/1000m, R=10 Cluster 5
Receptor Level= 200m
Mean Height= 892 m
Cluster 4
Receptor Level= 500m
Mean Height= 867 m
Cluster 5
Receptor Level= 1000m
Mean Height= 1247 m
(hr/104km2)

15. Length of cluster Acadia, 1997-2002, 24/48/120hr, 500m, R=6,10,16
Total Trajectories in 24 hr Analysis = (99-02)
Mean Height= 1286 m
Total Trajectories in 48 hr Analysis = 10592
Mean Height= 1916 m
Total Trajectories in 120 hr Analysis = 4429
Mean Height= 1692 m
(hr/104km2)

16. Temporal range Acadia, 97/98/99/00/01/02/97-02, 48hr, 500m, R=10 1997
1998
1999
Cluster 3
Cluster 2
Cluster 3
1223 m
1728 m
1500 m
2000
2001
2002
Cluster 3
Cluster 2
Cluster 3
1916 m
1479 m
1890 m
Cluster 3
1916 m

17. Site Differences Acadia, 1997-2002, 48hr, 500m, R=10 1 2 3 4 5 6 7
(hr/104km2)

18. Site Differences Shenandoah, 1999-2002, 48hr, 500m, R=10 1 2 3 4 5 6 7
(hr/104km2)

19. Site Differences Baltimore, 1997-2002, 48hr, 500m, R=10 1 2 3 4 5 6 7
(hr/104km2)

20. Atmospheric Modes vs. Clusters
Clusters are defined as groupings of similar trajectories during a specific time period
Atmospheric modes describes the underlying meteorological phenomena that explain the observed clusters

21. Generalized Westerly Transport (GW)
Acadia
Typically associated with low wind speeds, low mean heights
Wide pattern indicative of varied, meandering path
730 m
Baltimore
746 m

22. Southwest Coastal (SWC)
Acadia
639 m
Lyebrook
578 m

23. Southwest Inland (SWI)
Acadia
948 m
Lyebrook
1013 m

24. Southwest Bi-modal (SWB)
Baltimore
Represents a mix of southwest coastal and southwest inland
737 m
Shenandoah
932 m

25. Northwest Low (NWL)
Acadia
Predominant in Fall
1066 m
Baltimore
698 m

26. Northwest High (NWH) High altitude version of Northwest cluster Acadia
1630 m
Baltimore
2401 m

27. Northeast Maritime (NEM)
Acadia
544 m
New York City
870 m

28. Southeast Maritime (SEM)
Acadia
459 m
New York City
432 m

29. Predominant Meteorological Pathways and Influence on Air Quality
Back-trajectory Clusters
Pollution-Date Database
Correlate dates to establish relationships between atmospheric modes and pollution tendencies

30. Meteorological Influence on Air Quality: Acadia
Receptor Level= 500m, R=6, 48hr Backtrajectory
N1=1165
N2=710
N3=588
N4=511 31
4.7 57
7.8 39
5.8 66
8.8
730 m
1067 m
473 m
544 m
N5=500
N6=467
N7=389
N8=331 65
8.7 62
8.3 39
5.4 40
5.4
8 panel chart with table including statistics and an extra column for PM2.5, Reconstructed Extinction and Ozone
1262 m
710 m
1630 m
459 m
Key: Number of Trajectories
Reconstructed Extinction (Mm-1)
PM2.5 (ug/m3)
Mean pressure height (meters)

31. Meteorological Influence on Air Quality: Lye Brook
Receptor Level= 500m, R=10, 48hr Backtrajectory
N1=5686
N2=1551
N3=1141
N4=967 60
9.6 52
7.9 29
4.8 46
7.1
722 m
1215 m
418 m
1013 m
N5=666
N6=444
N7=432
N8=327
8 panel chart with table including statistics and an extra column for PM2.5, Reconstructed Extinction and Ozone 29
5.2 39
5.9 18
3.2 55
10.3
1513 m
1907 m
700 m
578 m
Key: Number of Trajectories
Reconstructed Extinction (Mm-1)
PM2.5 (ug/m3)
Mean pressure height (meters)

32. Meteorological Influence on Air Quality: NYC
Receptor Level= 500m, R=10, 48hr Backtrajectory
N1=5843
N2=2242
N3=1081
N4=979
17.9
13.4
17.0
9.3
725 m
1785 m
560 m
897 m
N5=574
N6=357
N7=354
N8=341
10.3
13.4
8 panel chart with table including statistics and an extra column for PM2.5, Reconstructed Extinction and Ozone
18.4
8.7
1212 m
1313 m
432 m
1800 m
Key: Number of Trajectories
Reconstructed Extinction (Mm-1)
PM2.5 (ug/m3)
Mean pressure height (meters)

33. Meteorological Influence on Air Quality: Baltimore
Receptor Level= 500m, R=10, 48hr Backtrajectory
N1=6402
N2=2513
N3=1093
N4=667
17.9
14.3
17.0
8.4
1531 m
737 m
746 m
966 m
N5=579
N6=398
N7=307
N8=302
8 panel chart with table including statistics and an extra column for PM2.5, Reconstructed Extinction and Ozone
14.7
10.9
12.3
9.2
1719 m
458 m
2401 m
698 m
Key: Number of Trajectories
Reconstructed Extinction (Mm-1)
PM2.5 (ug/m3)
Mean pressure height (meters)

34. Meteorological Influence on Air Quality: Shenandoah
Receptor Level= 500m, R=10, 48hr Backtrajectory,
Cluster 1
Cluster 2
Cluster 3
Cluster 4 62
8.2 91
12.7 68
9.1 68
9.8
930 m
836 m
932 m
924 m
Cluster 5
Cluster 6
Cluster 7
Cluster 8
8 panel chart with table including statistics and an extra column for PM2.5, Reconstructed Extinction and Ozone 65
8.8 48
5.8 69
10.0 46
7.0
875 m
916 m
951 m
808 m
Key: Number of Trajectories
Reconstructed Extinction (Mm-1)
PM2.5 (ug/m3)
Mean pressure height (meters)

35. Next Steps Complete trajectory database to fill in holes
Compare membership in similar clusters
Incremental probabilities (take out central tendency)
Average incremental probabilities across multiple sites (based on similarity of atmospheric modes)