NPS/CIRA Group Data Analysis and Receptor Modeling Receptor
Modeling Meeting with Carol McCoy, ARD Division Chief CIRA, Fort
Collins, CO 16 June 2011 Slide 2 Data Analysis & Interpretation
Ongoing (like IMPROVE) & Special Studies Are trends in
pollutants increasing or decreasing? Why? Are there geographical
differences in the trends? What does that say about sources? Does
the monitoring data make sense? Should there be changes in
monitoring protocols or has a recent change caused unexpected
changes in the data? Do the models make sense? If not, why not?
Industry is claiming XXX. Is that a valid claim? Sources have shut
down. Was there a noticeable effect? New sources are planned. Will
it matter? Is there a better location? For the next special study
What season is best? Will locations on opposite sides of a park see
the same trends? What temporal and spatial scales are necessary for
monitoring if the data to be useful? What are the expected
concentrations? Slide 3 Some Past Topics Source attribution source
types, source locations, international sources Hygroscopicity water
uptake Smoke natural smoke vs anthropogenic, organic chemistry
issues, markers Use of Satellites fill in spatial patterns, verify
models Improving Measurement Techniques faster, cheaper, better
resolution, more accuracy, better documentation Single particle
characteristics size, shape, mixture type Data & information
dissemination web sites, databases, software, books, papers, etc.
Tracking trends emissions, concentrations, visibility Natural
Background what is it and how can we get there? Nitrogen Deposition
why is it increasing, how can we better measure it, what sources
are contributing? Human Perception what do people see, value,
remember? Slide 4 What are the Trends in Nitrogen Wet Deposition?
Slide 5 April & July Chosen for ROMANS 2006 Slide 6 Could
Climate Change Be Influencing Ammonia Deposition At Rocky Mountain?
Which Meteorological Model Does A Better Job of Predicting
Precipitation at Rocky Mountain? Slide 7 2006 Measured Wind
Directions By Height Is there more upslope (easterly) transport
during the Summer? Do the Meteorological Models Capture this? Slide
8 Model Evaluation Slide 9 Spatial Patterns in Model Statistics
More Model Evaluation Slide 10 Source-Receptor Modeling Photo by
Ralph Turcotte Where do air pollutants come from? Slide 11 Some
Special Studies Conducted for Source Attribution and/or Optical and
Physical Characteristics Grand Canyon, AZ WHITEX (1987) MOHAVE
(1991?) Mt. Rainier, WA PANORAMAS (1984) PREVENT (1990) Big Bend,
TX Scoping Study (1996) BRAVO (1999) Rocky Mountain, CO ROMANS
(2006) ROMANS II (2009) Grand Teton, WY Grand Trends (2011)
Yosemite, CA Smoke Characteristics (2002) Eastern U.S. NAPAP SEAVS
Shenandoah Assessment Typically intensive monitoring for weeks to
months, followed by months to years of data analysis and modeling.
Often involving stake holders on all sides of the visibility or
deposition issue. Slide 12 Deterministic or Source-based Models
Model(s) Emissions Chemistry Meteorology Predictions of
concentrations and source attributions at a receptor of interest
Common Problems: Lack of input data Expense Examples -SMOKE
(Emissions Model) -CAMx, CMAQ (Air Quality Models) -Boundary
conditions GOCART, GEOSCHEM -MM5 & wrf mesoscale meteorology
Slide 13 Receptor Models Simple Statistical Analysis Measured Data
at and/or near a receptor Sometimes Meteorology and/or Source
Characterization Qualitative and/or Quantitative Source Attribution
Problems: Simplifying Assumptions Sometimes long-term averages
only. Examples: -Many trajectory analyses forward & backward
-Spatial & temporal patterns (EOF) -Chemical Mass Balance (CMB)
-Positive Matrix Factorization (PMF) Slide 14 Hybrid Models Further
Analyses, more modeling, a melding of all available information.
Results of Deterministic Models Results of Receptor Models Better
Predictions of concentrations & source attributions, Insight
into problems Examples -Model reconciliation -Tests against tracer
data -Tests against synthetic data -Spatial & temporal patterns
of error using BRAVO, IMPROVE, & CASTNet Data -Synthesized
REMSAD & CMAQ -Scaled receptor techniques More Monitoring Data
Slide 15 Saharan Dust and Mineral Ratios Slide 16 Factor Analysis
Type 1 (Species by Time, 1 Site) What species vary similarly? Do
they suggest source types? Factor 2 - 37% of variance Soil
elements, Zn, Pb, SO 2 (Power plant, smelter) Factor 1 - 38% of
variance S, Se, Na (TX, marine, industry) Factor 3 - 13% of
variance V, Ni & OC, K (Oil & fires) Factor 4 - 12% of
variance As, Cu (smelter) Example from Big Bend 1996 Scoping Study
Slide 17 BRAVO 1999 Slide 18 Factor Analysis Type 2 (Site by Time,
1 Species) What times, sites vary similarly? Do patterns suggest
source areas? Example from BRAVO 1999 Slide 19 Trajectory Analysis
Methods Trajectory Mass Balance Model Residence Time Type Analyses
Quadrant Assignment Cluster Analysis Hit - No Hit Emissions
Estimation Residence Time Conditional Probability Differential
Probability Source Contribution Function Average and Maximum Fields
Concentration Weighted Residence Time Slide 20 Qualitative back
trajectory analyses Where does air come from on high and low
pollution days ? Several statistical techniques used. This is the
simplest. High SulfurLow Sulfur QUALITATIVE RECEPTOR MODELING Slide
21 ROMANS II 2009 Ammonia Slide 22 Winter Spring Summer Fall 2009
Mean Ammonia Slide 23 Five Transport Patterns 1996 Scoping Study,
Big Bend National Park Cluster Analysis Slide 24 Concentrations by
Transport Patterns Based on Clustered Trajectories 1996 Scoping
Study - Big Bend National Park Slide 25 Transport from Northeastern
Colorado Slide 26 Regression Techniques of Source Apportionment
Assumption: The concentration measured at the receptor is some
linear combination of the contributions of several sources
Concentration = a 1 Source 1 + a 2 Source 2 + Variations: CMB
(Chemical Mass Balance), PMF (Positive Matrix Factorization), UNMIX
Use source profiles and concentrations of several species to
predict attributions for 1 measurement period. Source profiles are
inferred from the measured concentration data in the latter two.
TrMB ( Trajectory Mass Balance) Use many concentrations of 1
species and counts of trajectory endpoints in source regions to
predict average attributions over a long period. Others... Slide 27
TrMB Source Regions How do we know TrMB works? Inert Tracers from
BRAVO. Slide 28 TrMB modeling of BRAVO tracers at K-Bar, 9/17-10/28
(42 days). Negative concentrations were set to zero before summing.
Attributions accurate to within the uncertainty of the measurement
and standard error of the regression coefficients shown in bold and
a larger font. Slide 29 ROMANS 2006 Automating thousands of TrMB
Runs Slide 30 Sensitivity Analyses To Source Area Selection To
Height of Trajectories Slide 31 WRF vs MM5 ROMANS I April 23, 2006
Episode Slide 32 Should be S Westerly Should be Easterly Slide 33
Should be S Westerly Should be Easterly Slide 34 Ammonia 2009 By
Season WinterSpring Summer Fall Slide 35 Ammonia 2009 By Season
Winter Spring Summer Fall Slide 36 Receptor Models Have Been Used
For Peer-reviewed literature Regulatory decisions (RHR) Education
& Outreach Interagency Task Forces Interpretive Exhibits
Booklets (Intro To Visibility) Understanding Trends Reality checks
on models Counterbalance to industry Etc
