Validation and Preliminary Analysis of JATAP Air Toxics Data and Development of an Air Toxics Emissi

1
Validation and Preliminary Analysis of JATAP Air
Toxics Data and Development of an Air Toxics
Emission Inventory

• Prepared by
• Hilary R. Hafner
• Steven G. Brown
• Michael C. McCarthy
• Heather Arkinson
• Dana Sullivan
• Sonoma Technology, Inc.
• Petaluma, CA
• Presented to
• Arizona Department of Environmental Quality
• Phoenix, AZ
• June 11, 2004

2
Overview

• Background
• Data Validation Approach
• Data Validation Results
• Data Analysis Preliminary Results
• Status and Plans for JATAP Data Analysis
• Update on the Development of an Air Toxics
  Emission Inventory

3
Data Validation and Analysis
4
Background

• As part of the Joint Air Toxics Assessment
  Project (JATAP), air toxics data were collected
  between March 2003-March 2004 at three sites
• West 43rd (MCAZ)
• South Phoenix (SPAZ)
• St Johns (Gila River Indian Community) (SJAZ)
• ADEQ sites for comparison include
• Phoenix Supersite (PSAZ)
• Queen Valley (QVAZ)

5
(No Transcript)
6
Technical Approach

• Acquire and validate JATAP air toxics data for S.
  Phoenix and W. 43rd site
• Characterize the spatial and temporal variation
  in the hazardous air pollutants (HAPs).
• Compare the S. Phoenix and W. 43rd HAPs
  characteristics to other site data
• Perform preliminary source apportionment
• Prepare the data set for risk assessment
• Prepare final report and present results

7
Data Overview

• 24-hr gas-phase air toxics samples were collected
  every 6th day
• On alternating sampling days, two 12-hour samples
  (AM and PM) were collected
• Samples were analyzed for 58 air toxics
• Duplicate (collocated) and replicate (additional
  chemical analysis on canister) data also were
  collected

8
Hazardous Air Pollutants (HAPs)
EPA 33

• Toxic air pollutants are generally referred to as
  hazardous air pollutants (HAPs) or air toxics.
• Toxic air pollutants are those pollutants that
  are known or suspected to cause cancer (or other
  serious health effects) or adverse environmental
  effects.
• Although there are thousands of compounds in the
  air that could cause harm, 188 are listed in the
  1990 CAA. Generally, measurements focus on the
  33 shown.

9
HAPs List for JATAP
Mobile source air toxic (MSAT) National air
toxics trends site target HAP
10
Key Species List
11
Treating Missing Data And Data Below MDL (1 of 2)

• The selection of the treatment method for data
  below the MDL can have a significant impact on
  analysis results if a large amount of data is
  below the MDL. During computation of an annual
  average (AA)
• ignoring data below the MDL results in an AA that
  is biased high
• replacing the data below the MDL with the MDL
  value results in an AA that is biased high or
• replacing the data below the MDL with zero
  results in an AA that is biased low.
• MDL method detection limit also known as
  minimum detection limit
• 40 CFR, Chapter 1, Appendix B to Part 136
  provides methodology for computing MDLs
• Typically, the labs calculate a standard
  deviation of multiple measurements and create a
  99 confidence bound using a t-distribution
  (Bortnick, 2003).
• Which means the MDL is a threshold above which we
  can be 99 confident that a non-zero
  concentration is present.

12
How Should Missing Data And Data Below MDL be
Treated? (2 of 2)

• The process of using reported values below the
  MDL (when available) and using MDL/2 for
  nondetect data provides a defensible annual
  average for data sets with up to 50 of data
  below the MDL. If more than 50 of the data are
  below the MDL, the annual average can be biased
  by the choice of MDL substitution.
• Analyses indicated that values below the MDL
  should be reported to (flagged, but not censored
  from) the database, and used.

13
Database Preparation

• Assemble the database.
• Place data in a common data format with
  descriptive information concerning variables,
  validation level, QC codes, detection limits,
  time standard, standard units, and metadata (site
  information, etc.).
• Ensure that results of and suggestions from all
  audit reports and lab/field blanks have been
  incorporated into the database. Were there any
  problem species?

14
Data Validation/Analysis Considerations

• Levels of other pollutants
• Time of day/year
• Observations at other sites
• Audits and inter-laboratory comparisons
• Instrument performance history
• Calibration drift
• Site characteristics
• Meteorology
• Exceptional events

15
Validation Approach (1 of 2)

• Understand the pollutant sources, lifetimes,
  temporal behavior, etc. (use developed cheat
  sheets)
• Understand site location and sampling and
  analysis techniques
• Inspect summary statistics and apply screening
  criteria (look for unrealistic maxima or minima
  and for consistency with nearby stations)
  determine data completeness

16
Validation Approach (2 of 2)

• Inspect all species (its quick and easy)
• Time series plots (are expected patterns
  evident?)
• Scatter plots (are expected relationships
  observed?)
• Inspect every sample
• Fingerprint plots (have all species been
  reported?)
• Flag data and document modifications
• Prepare seasonal and annual averages

17
Data Validation Progress/Results

• Data collected through Dec 2003 have been
  processed and validated
• Processing was not trivial because the lab format
  is difficult to use, and information was
  sometimes incorrect
• Validation on Jan-Mar 2004 data should begin
  soon, once data are received
• In general, data are of good quality
• Many species are below the MDL
• Duplicate/replicate analyses give an evaluation
  of the precision of data

18
Duplicates and Replicates
19
Replicate Analysis (1 of 2)
Compare Replicate Measurements (Good match with
bias)
20
Replicate Analysis (2 of 2)
Compare Replicate Measurements (Bad Example)
Outliers flagged as suspect (poor reproducibility)
21
Summary

• Duplicate/replicate analyses are important to
  gauge our confidence in the data
• Acetonitrile and methyl ethyl ketone (MEK)
  exhibited a high degree of error (i.e., gt 10)
• A number of samples ( 10) had very poor
  reproducibility for most species
• Further analysis will be conducted on
  duplicate/replicate samples

22
Data Validation
All units on the following plots are ppbv
23
Time Series Acetonitrile
Extremely high acetonitrile concentrations at S.
Phoenix were flagged as suspect
24
Time Series Methyl Ethyl Ketone
Extremely high methyl ethyl ketone (MEK) at W.
43rd more than twice any other MEK
concentration!
25
Scatter Plots Ethylbenzene and Xylenes
Scatter plot of ethylbenzene and mp-xylene
concentrations from S. Phoenix in 2001-2003.
These two species show a tight correlation as
expected.
S. Phoenix
26
Summary

• Overall data quality was good
• Some species were flagged as suspect for further
  investigation
• Additional analysis on duplicate/replicate
  samples will be conducted to further gauge
  confidence (i.e., precision) in the data
• Many species were consistently below MDL, meaning
  these will not be useful in quantifying trends
  and sometimes risk

27
What Is Our Confidence In The Data?

• Confidence in air toxics measurements varies by
  pollutant
• high for compounds with median concentrations
  well above detection limits
• low for compounds with median concentrations
  close to detection limits
• Note that data below the MDL tells us something
  (i.e., provides an upper limit)

28
S. Phoenix AA Decision Matrix
29
W. 43rd AA Decision Matrix
30
MDLs and Ambient Concentrations

• 24 of the 58 measured species were above MDL
  more than 75 of the time
• These species have sufficient data to quantify
  trends
• In general, these are also the most abundant
  species
• About 25-30 of the species were below detection
  most (gt 50) of the time

W. 43rd
N species
S. Phoenix
N species
31
Preliminary Data Analysis Results
32
Abundant Species (Top 10)

• S. Phoenix and W. 43rd sites are somewhat similar
  in concentrations and abundant species

33
Abundant Species (Top 10)

• Some of these abundant species (toluene,
  acetylene, xylenes, benzene) are usually from
  mobile sources

34
Interpreting SYSTAT Plots
Notched Box Whisker Plots in SYSTAT

• where
• IR interquartile range
• C.I. confidence interval

35
Diurnal Trends
S. Phoenix Toluene
W 43rd Benzene

• Many species concentrations are lower in the
  afternoon/evening than the night/morning,
  possibly indicating the influence of mobile
  sources.

36
Seasonal Variations

• Some species show small seasonal (1spring)
  differences (i.e., CFCs)
• Mobile source marker concentrations were highest
  in winter (e.g., acetylene, toluene, benzene)
• Dichlorodifluoromethane concentrations were also
  highest in winter (4)

S. Phoenix
W. 43rd
37
Comparison to Other Sites

• Mobile source markers (toluene, acetylene,
  benzene, etc.) are often lowest at St. John and
  Queen Valley sites
• Styrene concentrations are highest at the St.
  John site
• W43rd and S. Phoenix air toxics concentrations
  are
• generally similar
• often different than Phoenix Supersite
• All three urban site concentrations are higher
  than Queen Valley and St. John (for most species)

38
Annual Average Daily Traffic Volume
(vehicles per day)
39
Status and Plans for JATAP Data Analysis (1 of 3)

• Acquire and validate the HAPs samples.
• Data have been organized and compiled,
  completeness assessed, prepared for validation,
  and validation performed.
• We will provide a list of samples which we
  recommend be flagged as suspect or invalid and
  the reason for the flags.

40
Status and Plans for JATAP Data Analysis (2 of 3)

• Characterize the spatial and temporal variation
  in the HAPs.
• investigate the relationship among species
  collected at the two ADEQ sites using scatter
  plots and correlation matrices,
• determine the most abundant species,
• investigate the variation in concentrations by
  month and day of week,
• compare HAPs concentrations and spatial/temporal
  patterns at the study sites to data collected at
  the Phoenix Supersite and Queen Valley sites
• use supplemental data collected at the study
  sites including continuous measures of
  meteorology, gaseous air pollutants, PM mass, and
  PM components.

41
Status and Plans for JATAP Data Analysis (3 of 3)

• Compare with Arizona Ambient Air Quality
  Guidelines.
• We will compare 24-hr concentrations of HAPs with
  Arizonas guidelines.
• Perform preliminary source apportionment.
• Further investigate the relationships among
  species by performing factor analysis (or
  principal component analysis) with the data.
  This technique provides qualitative information
  regarding potential emission source types of the
  toxics.

42
Development of an Air Toxics Emission Inventory
Update
43
JATAP Toxics Emissions InventoryPhase I
Existing Emissions

• Acquire existing PM10 and ozone precursor
  emissions
• By source category
• Spatially allocated
• Select appropriate toxics speciation profiles
• PM10 emissions
• Volatile organic compound (VOC) emissions
• Apply speciation profiles to PM10 and VOC
  emissions
• Particulate matter toxics
• Gaseous toxics
• Characterize speciated toxics emissions
• Identify potentially missing sources
• Prioritize potentially missing sources

44
JATAP Toxics Emissions InventoryDomain
45
JATAP Toxics Emissions InventoryPhase I
Completed Tasks

• Acquired PM10 Salt River State Implementation
  Plan (SIP) inventory files
• Point sources spatially allocated PM10
  emissions (MCESD)
• On-road sources spatially allocated activity
  data (ADEQ)
• Area sources spatially allocated activity data
  (ADEQ)
• Non-road sources methods and emission factors
  (ADEQ)
• Surveyed the domain for potentially important
  sources
• Point source PM10 and VOC emissions
• On-road diesel exhaust (hauling)
• Non-road diesel exhaust (agricultural and
  construction)

46
JATAP Toxics Emissions InventoryPictures
Petroleum Liquid Storage
Construction Equipment
Metals Processing
Agricultural Equipment
47
JATAP Toxics Emissions InventoryPhase I
Current Tasks

• Acquiring 1999 Ozone Precursor inventory files
• Point sources spatially allocated VOC emissions
  (MCESD)
• On-road sources methods and emission factors
  (MAG)
• Area sources methods and emission factors (MAG)
• Non-road sources methods and emission factors
  (MAG)
• Selecting PM10 speciation profiles for source
  types
• Industrial processes
• Materials handling
• Diesel engine exhaust

48
JATAP Toxics Emissions InventoryPhase I Future
Tasks

• Apply speciation profiles to PM10 emissions
• Select speciation profiles for VOC emissions
• Apply speciation profiles to VOC emissions
• Characterize speciated toxics emissions
• Identify and prioritize missing sources
• Summary of project status and prioritization
  (June 25th)

49
JATAP Toxics Emissions InventoryPhase II
Missing Sources

• Acquire PM10 and VOC emissions for missing
  sources
• Source category
• Spatially allocated
• Select appropriate toxics speciation profiles
• PM10 emissions
• VOC emissions
• Apply speciation profiles to PM10 and VOC
  emissions
• Particulate matter toxics
• Gaseous toxics
• Summarize the complete toxics inventory
• Characterize toxics emissions over entire domain
• Format estimated emissions for Industrial Source
  Complex (ISC) dispersion modeling

End
50
Appendix

• Data validation and analysis tools
• Data validation summary sheet overview
• Discussions of national and regional spatial and
  temporal distributions of air toxics
• Products available from http//www.ladco.org/toxic
  s.html

51
Data Validation Summary Sheets for 19 HAPs

• Sheets were prepared to assist in the validation
  and understanding of HAP data.
• Sheets were prepared for 19 HAPs and elemental
  carbon (EC).

52
37
53
38