Tools and Techniques for Identifying
Contributions of PMfine to Regional Haze:
Source Apportionment Techniques,
CATT and FASTNET
Serpil Kayin, MARAMA
Rich Poirot, VT DEC
2
Tools and Techniques for Identifying Contributions
of PMfine to Regional Haze
Overview
• Why data analysis—Role in SIP planning
• Source Apportionment
– What is it, how does it work
– Interface with modeling & inventory development
– SA Work MANE-VU conducted
3
Data Analysis and SIP Planning
• CAAAC and new EPA guidance recommend weight-of-evidence demonstration for SIP planning and implementation
• WOE = Models + monitoring data analysis + inventory review • Specific data analysis suggested
– Conceptual model (qualitative description) – Historic trends, meteorologically adjusted – Transport assessment
– Observation-based methods
• Synthesis of Modeling/Monitoring/EI
– Identify important source sectors or specific pollutants
4
Source Apportionment of Ambient Data
• What is source apportionment?
– Mathematical technique for determining contributions of
various sources to a given sample of air
– Produces a sample-specific inventory estimate
– Can be used to develop and validate inventory information,
track effectiveness of control measures
– Can be used for various air pollutants: volatile organics,
semivolatiles, particles, toxics, or a combination of these
– Many methods: CMB (Chemical Mass Balance), PMF
5
Source Apportionment of Ambient Data
• SA is a convenient way to extract information on
pollutant sources from routinely collected ambient data
(bottom up vs. top down approach)
• EIs generally self-reported, not measured
• EIs often derived from emission factors and activity
estimates
• SA results allow independent evaluation of inventory
data and model predictions (especially source
6
Source Apportionment of Ambient Data
• Fingerprints (aka source profiles)
– Chemical patterns of source emissions
– Species should be present in both ambient air and in source emissions – Species should have limited reactivity or similar reaction rates.
• Assumptions
– Composition of source emissions is relatively constant
– Emissions do not react or selectively deposit between source and receptor (mass is conserved)
– Source profiles are linearly independent
Example Source Profiles from PMF and Unmix Modeling at Underhill, VT and Comparison of Daily Source Contributions, (right)
8
Source Apportionment of Ambient Data
• Advantages
– Models tolerate deviations in model assumptions well (often breaking sources with primary & secondary precursors into 2 or more “source components”)
– Useful for natural emission sources, & others with “strong flavor” (unique chemical composition, time series, or spatial origin)
– Can identify previously un-inventoried sources
• Disadvantages
– Usually identifies categories of sources, not individual sources – Retrospective, not predictive
– Identifies only receptor contributions, not mass emission rates
– PMF and UNMIX require large number of samples (100+); CMB can be applied to individual samples
9
Source Apportionment of Ambient Data
• What data are available
– PAMS
– PM2.5 speciation data
– Special ozone and PM studies to support photochemical models
(SEARCH, etc)
– Toxics Network data
Locations of Recent Northeastern Receptor Modeling Studies,
Conducted by MANE-VU and/or by Academic Researchers
Early work focused primarily on Rural
IMPROVE sites. More recent analyses
based on New Urban STN speciation data Comparing Urban and
Rural results shows common influences on haze and PM2.5
and
Also helps show key Local Urban Sources
Similarities & Differences between Rural Haze & Urban PM
2.5During Summer, Rural Haze and Urban PM
in the Northeast are Both Dominated by secondary Sulfates.
During the Winter, Local Urban Sources of Carbon and Nitrate
Compounds become Much More Important
12
SA Work MANE-VU conducted
• Battelle Report (May 2002)
• DRI Report (March 2005)
• NESCAUM Report (
Tools and Techniques for
Identifying Contributions to Regional Haze in the
MANE-VU Region
), Appendix B (January 2005)
• CATT and FASTNET web tools for additional SA
13
FASTNET:
Fast Aerosol Sensing Tools for Natural
Event Tracking
CATT:
Combined Aerosol Trajectory Tools
• Web-Based Data Acquisition, Visualization, Analysis Tools,
• Developed by CAPITA (R. Husar, S. Falke, K. Hoijarvi), with significant contributions from R. Poirot (VT)
• Funded by the 5 US Regional Planning Organizations,
• Managed for MANE-VU by Gary Kleiman & Serpil Kayin
• Based on Data Architecture Developed with previous support from NSF and NASA
CATT/DATAFed User Instructions
and Tutorial
• http://datafed.net
• CATT is under “web apps”
http://datafed.net/projects/CATT/CATT_Links.htm
• CATT url has resources/discussion and user manual • Tutorials available
• http://capita.wustl.edu/datafed/tutorial/Tutorial1-Basics.htm
• http://capita.wustl.edu/datafed/tutorial/Tutorial2-Cursor.htm
• http://capita.wustl.edu/datafed/tutorial/Tutorial2-Cursor.htm
They take a while to download, and need sound turned on.
Also there's the new "user file submittal option" described at: •
Datasets Used in FASTNET
• Data are accessed from autonomous, distributed providers • DataFed ‘wrappers’ provide uniform geo-time referencing • Tools allow space/time overlay, comparisons and fusion
Near Real Time Data Integration
Delayed Data Integration
Surface Air Quality
AIRNOW O3, PM25
ASOS_STI Visibility, 300 sites METAR Visibility, 1200 sites
VIEWS_OL 40+ Aerosol Parameters
Satellite
MODIS_AOT AOT, Idea Project GASP Reflectance, AOT TOMS Absorption Indx, Refl.
SEAW_US Reflectance, AOT
Model Output
NAAPS Dust, Smoke, Sulfate, AOT WRF Sulfate
Fire Data
HMS_Fire Fire Pixels MODIS_Fire Fire Pixels Surface Meteorology
RADAR NEXTRAD
SURF_MET Temp, Dewp, Humidity… SURF_WIND Wind vectors
A Sample of
Datasets
Accessible through ESIP –
DataFed
Mediation
Near Real Time (~ day)
It has been demonstrated (project FASTNET) that these and other datasets can be accessed, repackaged and delivered by AIRNow through ‘Consoles’
MODIS
Reflectance MODIS AOT TOMS Index
GOES
AOT 1km ReflecGOES
NEXTRAD Radar MODIS Fire Pix NRL MODEL NWS Surf Wind, Bext
Next Process Next Process
Why?
How?
When? Where?CATT: A Community Tool!
Part of an Analysis Value Chain
Aerosol Data Collection IMP. EPA Aerosol Sensors Integration VIEWS Integrated AerData
AEROSOL
Weather Data Assimilate NWS Gridded Meteor. Trajectory ARL Traject. DataTRANSPORT
TrajData Cube Aggreg. Traject. AerData CubeCATT
Aggreg. Aerosol CATT-In CAPITA CATT-In CAPITA There! Not There!Further Analysis
GIS Grid Processing Emission ComparisonAerosol Event
Catalog: Web pages
• Catalog of generic ‘web
objects’ – pages, images,
animations that relate to
aerosol events
• Each ‘web object’ is
cataloged by location,
time and aerosol type.
Evaluation and Interpretation of Receptor Model Results by
Local Surface Winds or Ensemble Trajectory Techniques
Sea Salt Source from Both Unmix and PMF Modeling at Brigantine
Since Sea Salt comes from the Sea (Well, Duh!), It
tends to contribute most on days when man-made
pollutants (from inland) are lowest.
But Battelle PMF results suggest sea salt is
Receptor Model Results show Sea Salt Source (with High Na) at NE Sites CATT applied to Entire IMPROVE Network Shows High Na for the Sea(s)
Like Sea Salt, Windblown Dust at Brigantine also seems to come from the
Sea. Weird!
Highest Dust at all Eastern IMPROVE sites comes from the Sahara Desert
Receptor Models indicate a Source of Fine Windblown Dust from an Unexpected Distant Origin. It’s a Minor Contributor to Haze and PM2.5,
25
SA Work MANE-VU conducted
Surface Met and Trajectory Evaluation of Oil Source Identified by Unmix & PMF modeling at Brigantine, NJ
Trajectory Incremental Probability for Oil Source Identified by Unmix or PMF modeling at 4 MANE-VU sites (sources are within region)
Receptor Model Results show “Local” Oil Source (with High Ni) at NE Sites CATT applied to Entire IMPROVE Network Shows High Ni for East Coast
Receptor Model Results show Wood Smoke Sources, which tend to be from Canadian summer Forest Fires & winter Residential Burning in New England,
and more often from Southeastern Fires in Southern MANE-VU
Quebec Fires of July 2002 were a Big Example of a Smaller, Common summer influence in
Receptor Model Results show Wood Smoke Sources, which tend to be from Canadian summer Forest Fires & winter Residential Burning in New England,
Back Trajectories for All IMPROVE Sites on 7/7/02 Unweighted (top left), & color-weighted for OC (top right), SO4 lower right & Cl (lower left)
Color weighted ATADs for High SO4 (top) & Se (bottom) for 8/12-15/02 Haze event;
Receptor Models indicate a Large source of Secondary Sulfate (i.e. Coal) at Northeast Receptor Sites. Same area is upwind of Highest SO4 throughout
the IMPROVE Network. Selenium is a good “Primary” Coal Tracer
PMF Sulfate (coal) sources, 7 NE sites
SO4 > 15 ug/m3, all IMPROVE sites
Receptor Models show Secondary Nitrate Source which appears to be
influenced by Midwestern areas of High Agricultural NO3 Emissions (at rural IMPROVE sites in MANE-VU and throughout the East)
Several Large-Scale Northeastern NO3 Events observed in recent years, illustrated here by ASOS visibility data from FASTNET for 2/18-21/04. Are these Winter NO3 events becoming more frequent in the Northeast?
NO3 at WASH
SO4 at WASH
SO4 sources similar for Rural sites like Shenandoah & Urban sites like DC
But Urban areas have larger and different local NO3 sources
NO3 at SHEN
SO4 at SHEN
Changes in Average Upwind Sulfate (Left) and Nitrate (Right) from 1992-95 (Top) and 2000-03 (Bottom) averaged for 42 IMPROVE sites