1995  Urban Air Toxics
Monitoring Program (UATMP)
Final Report
EPA Contract No. 68-1)3-0095
Delivery Order 07

Prepared for:
Office of Air Quality and Planning Standards
U.S. Environmental Proteciton Agency
Research Triangle Park, North Carolina 27711
                          Locations participating in the 1995 program

                                                January 20, 1997

-------
1995 Urban Air Toxics Monitoring Program
                   Final Report
           EPA Contract No. 68-D3-0095
                 Delivery Order 07
                     Prepared for:

               KathyWeant and Neil Berg
         Office of Air Quality Planning and Standards
           U.S. Environmental Protection Agency
            Research Triangle Park, NC 27711
                     Prepared by:

                Eastern Research Group
                  110 Hartwell Avenue
                 Lexington, MA 02173

                   January 20, 1997

-------
                                 DISCLAIMER
Through its Office of Air Quality Planning and Standards, the U.S. Environmental Protection
Agency funded and managed the research described in this report under EPA Contract No.
68-D3-0095 to Eastern Research Group, Inc. This report has been subjected to the Agency's
peer and administrative review and has been approved for publication as an EPA document.
Mention of trade names or commercial products does not constitute endorsement or
recommendation for use of these products.

-------
                              TABLE OF CONTENTS

                                                                             Page

      List of Figures    	iv
      List of Tables    	  v
      List of Abbreviations	  vii
      Abstract         	viii

1.0   Introduction    	1-1

2.0   The 1995 UATMP	2-1

      2.1     Monitoring Locations	2-1
      2.2     Sampling and Analytical Methods	2-2
      2.3     Sampling Schedules 	2-4
      2.4     Completeness	2-4

3.0   Summary of Ambient Air Monitoring Data 	3-1

      3.1     Air Monitoring Summary Parameters 	3-1

             3.1.1    Prevalence	3-1
             3.1.2    Concentration Range	3-2
             3.1.3    Central Tendency (Annual Average Concentrations)	3-3
             3.1.4    Variability 	3-5

      3.2     VOC Summary Tables 	3-6

      3.3     Carbonyl Summary Tables 	3-8

4.0   Statistical Analysis   	4-1

      4.1     The 1995 UATMP Concentration Distributions  	4-1

             4.1.1    Statistical Parameters	4-2

                      4.1.1.1    Skewness  	4-2
                      4.1.1.2    Rurtosis  	4-2
                      4.1.1.3    Shapiro-Wilk Statistic	4-3

             4.1.2     VOC Distributions	4-4
             4.1.3     Carbonyl Distributions	4-5
                                         11

-------
      4.2    Data Correlations      	4-7

             4.2.1       Methodology	4-7
             4.2.2       Correlations Among VOCs  	4-9
             4.2.3       Correlations Among Carbonyls  	4-10

5.0   Interpreting Ambient Air Monitoring Data  	5-1

      5.1    Emissions and Dispersion	5-1

             5.1.1       Chemical Emissions	5-2

                        5.1.1.1    Industrial Emission Sources	5-2
                        5.1.1.2    Motor Vehicle Emission Sources	5-3
                        5.1.1.3    Natural Emission Sources	5-4

             5.1.2       Atmospheric Fate and Transport  	5-4

      5.2    Other Air Monitoring Studies	5-4

      5.3    Example Analysis: 1,3-Butadiene	5-5

             5.3.1       Geometric Mean Concentrations  	5-5
             5.3.2       Interpretations	5-6
             5.3.3       Conclusions	5-8

6.0   Spatial Variations            	6-1

      6.1    Spatial Variations in the Overall Magnitude of Air Pollution	6-1

             6.1.1       Ranking by Overall Level of VOC Pollution  	6-3

                        6.1.1.1    Comparison of VOC Ranks to Industrial Emissions  6-3
                        6.1.1.2    Comparison of VOC Ranks to Motor Vehicle
                                  Emissions	6-4
                        6.1.1.3    Conclusion from VOC Ranks  	6-5

             6.1.2       Ranking by Overall Level of Carbonyl Pollution   	6-5

             6.1.3       Comparison of Carbonyl and VOC Overall Program Ranks  .... 6-6
                                         m

-------
      6.2    Impact of Industrial Emissions 	6-7

             6.2.1      Industrial Emissions of VOCs	6-7
             6.2.2      Industrial Emissions of Carbonyls	6-8
             6.2.3      Emission Inventory Interpretations	6-9

      6.3    Impact of Motor Vehicle Emissions  	6-10

             6.3.1      VOCProfiles	6-11
             6.3.2      Carbonyl Profiles	6-12

      6.4    Summary             	6-13

             6.4.1      VOC Summary  	6-13
             6.4.2      Carbonyl Summary	6-13

7.0   Temporal Variations          	7-1

      7.1    Time Frames for Temporal Variations	7-1

      7.2    Seasonal Variations     	7-1

             7.2.1      Significance of Seasonal Variations	7-2
             7.2.2      Seasonal Variations in VOC Concentrations	7-3
             7.2.3      Seasonal Variations in Carbonyl Concentrations	7-4

      7.3    Annual Trends         	7-5

             7.3.1      Significance of Annual Variations	7-6
             7.3.2      Annual Variations in VOC Concentrations	7-7
             7.3.3      Annual Variations in Carbonyl Concentrations	7-7

      7.4    Summary             	7-8

8.0   Precision and Accuracy       	8-1

      8.1    Precision              	8-1

             8.1.1      Analytical Precision	8-1
             8.1.2      Sampling and Analytical Precision  	8-3

      8.2    Accuracy             	8-5
                                          IV

-------
      8.3    Comparison to UATMP Data Quality Objectives  	8-6

9.0   Conclusions and Recommendations	9-1

      9.1    Conclusions           	9-1
      9.2    Recommendations      	9-3

10.0  References                  	10-1

-------
                                 LIST OF FIGURES


2-1    Locations of the 1995 UATMP Monitoring Stations	2-6

3-1    General Features of Normal and Lognormal Data Distributions	3-10

4-1    Examples of Skewness and Kurtosis	4-11

5-1    Geometric Mean 1,3-Butadiene Concentration Calculated for the 16 Monitoring
       Stations Participating in the 1995 UATMP	5-9
5-2    Air Emissions of 1,3-Butadiene within a 10-Mile Radius of the Port Neches, Texas
       (PNTX) Monitoring Station	5-10
5-3    Total 1,3-Butadiene Air Emissions Reported by Facilities Located within a 10-Mile
       Radius of the Port Neches Monitoring Station to TRI	5-11

6-1    Facilities in the Vicinity of the Baton Rouge, Louisiana (B2LA) Monitoring Station
       Reporting VOC or Carbonyl Air Releases to TRI	6-15
6-2    Facilities in the Vicinity of the Bellingham, Washington (BEWA) Monitoring Station
       Reporting VOC or Carbonyl Air Releases to TRI	6-16
6-3    Facilities in the Vicinity of the Brownsville, Texas (BRTX) Monitoring Station Reporting
       VOC or Carbonyl Air Releases to TRI 	6-17
6-4    Facilities in the Vicinity of the Brattleboro, Vermont (BRVT) Monitoring Station
       Reporting VOC or Carbonyl Air Releases to TRI	6-18
6-5    Facilities in the Vicinity of the Burlington,  Vermont (BUVT) Monitoring Station
       Reporting VOC or Carbonyl Air Releases to TRI	6-19
6-6    Facilities in the Vicinity of the Camden, New Jersey (CANJ) Monitoring Station
       Reporting VOC or Carbonyl Air Releases to TRI	6-20
6-7    Facilities in the Vicinity of the Davidson, Tennessee (DATN) Monitoring Station
       Reporting VOC or Carbonyl Air Releases to TRI	6-21
6-8    Facilities in the Vicinity of the El Paso, Texas (EPTX) Monitoring Station Reporting VOC
       or Carbonyl Air Releases to TRI	6-22
6-9    Facilities hi the Vicinity of the Garyville, Louisiana (GALA) Monitoring Station Reporting
       VOC or Carbonyl Air Releases to TRI 	6-23
6-10   Facilities in the Vicinity of the Galveston, Texas (GATX) Monitoring Station Reporting
       VOC or Carbonyl Air Releases to TRI 	6-24
6-11   Facilities in the Vicinity of the Hahnville, Louisiana (HALA) Monitoring Station
       Reporting VOC or Carbonyl Air Releases to TRI	6-25
6-12   Facilities in the Vicinity of the Port Neches, Texas (PNTX) Monitoring Station Reporting
       VOC or Carbonyl Air Releases to TRI 	6-26
6-13   Facilities hi the Vicinity of the Rutland, Vermont (RUVT) Monitoring Station Reporting
       VOC or Carbonyl Air Releases to TRI 	6-27
                                         VI

-------
                          LIST OF FIGURES (Continued)

6-14  Facilities in the Vicinity of the Underbill, Vermont (UNVT) Monitoring Station Reporting
      VOC or Carbonyl Air Releases to TRI 	6-28
6-15  Facilities in the Vicinity of the Vancouver, Washington (VAWA) Monitoring Station
      Reporting VOC or Carbonyl Air Releases to TRI	6-29
6-16  Facilities in the Vicinity of the Winooski, Vermont (WIVT)) Monitoring Station Reporting
      VOC or Carbonyl Air Releases to TRI 	6-30
6-17  Comparison of BTEX Concentration Profile to Roadside Study	.6-31
6-18  Ratios of Acetaldehyde and Formaldehyde to Ethylbenzene	6-34
                                        Vll

-------
                                LIST OF TABLES
2-1    Site Descriptions for the 1995 UATMP Monitoring Stations	2-7
2-2    Estimated VOC Detection Limits	2-9
2-3    Estimated Caibonyl Detection Limits	2-10
2-4    Sampling Schedules and Completeness 	2-11

3-1    Summary of Prevalence of VOCs	3-11
3-2    Summary of Highest VOC Concentrations	3-12
3-3    Summary of VOC Geometric Mean Concentrations	3-13
3-4    Summary of Prevalence of Carbonyls	3-14
3-5    Summary of Highest Caibonyl Concentrations		3-15
3-6    Summary of Caibonyl Geometric Mean Concentrations 	3-16

4-1    Statistical Analysis of VOC Distributions	4-12
4-2    Statistical Analysis of Caibonyl Distributions	4-12
4-3    Significant Correlations Among VOCs 	4-13
4-4    Significant Correlations Among Carbonyls	4-13

5-1    Geometric Mean 1,3-Butadiene Concentrations Measured at Port Neches, Texas  ... 5-12

6-1    Ranking of Monitoring Stations by Overall Levels of VOCs 	6-36
6-2    Ranking of Monitoring Stations by Overall Levels of Carbonyls	6-37
6-3    Comparison of Geometric Mean Concentrations of Selected VOCs with Total Air
       Releases Reported by Facilities within a 10-Mile Radius of UATMP Monitoring
       Stations	6-38
6-4    Comparison of Geometric mean Concentrations of Selected Carbonyls with Total
       Air Releases Reported by Facilities within a 10-Mile Radius of UATMP
       Monitoring Stations	6-41

7-1    Number of Monitoring Stations at which Selected VOCs Had Highest
       Seasonal-Average Ambient Air Concentrations	7-9
7-2    Number of Monitoring Stations at which Selected Carbonyls Had Highest
       Seasonal-Average Ambient Air Concentrations	7-10
7-3    Annual Trends in Geometric Mean Concentrations Observed at the Camden, New
       Jersey (CANJ) Monitoring Station	7-11

8-1    VOC Analytical Precision	8-8
8-2    Carbonyl Analytical Precision 	8-10
8-3    VOC Sampling and Analytical Precision	8-11
8-4    Carbonyl Sampling and Analytical Precision	8-13
                                       vui

-------
                           LIST OF ABBREVIATIONS
AIRS        Aerometric Information and Retrieval System
AQS         Air Quality Subsystem (of the Aerometric Information and Retrieval System)
ATSDR      Agency for Toxic Substances and Disease Registry
BTEX       benzene, toluene, ethylbenzene, and xylenes (o-, m-, and/?-xylene)
EPA         U.S. Environmental Protection Agency
EPCRA      Emergency Planning and Community Right-to-Know Act
FID         flame ionization detection
GC/MSD     gas chromatography
HPLC       high performance liquid chromatography
MDL        minimum detection limit
MSD        mass selective detection
ND          non-detect
ppbv         parts per billion (by volume)
RPD         relative percent difference
TRI         Toxic Release Inventory
UATMP      Urban Air Toxics Monitoring Program
UV          ultraviolet
VOC         volatile organic compound
Monitoring Stations

B2LA       Baton Rouge, Louisiana
BEWA      Bellingham, Washington
BRTX      Brownsville, Texas
BRVT      Brattleboro, Vermont
BUVT      Burlington, Vermont
CANJ       Camden, New Jersey
DATN      Davidson, Tennessee
EPTX       El Paso, Texas
GALA      Garyville, Louisiana
GATX      Galveston, Texas
HALA      Hahnville, Louisiana
PNTX      Port Neches, Texas
RUVT      Rutiand, Vermont
UNVT      Underhill, Vermont
VAWA      Vancouver, Washington
WTVT       Winooski, Vermont
                                       IX

-------
                                      ABSTRACT
       The Urban Air Toxics Monitoring Program (UATMP) is designed to characterize the
magnitude and composition of air pollution in urban centers.  This report presents the results of
ambient air monitoring performed under program year 1995 of the UATMP. Over the course of
this program year, ambient air samples were collected at sixteen different monitoring locations
and were analyzed for both volatile organic compounds (VOCs) and carbonyls. Not surprisingly,
the ambient air concentrations measured in this study vary significantly from monitoring location
to monitoring location and also from season to season.  Accordingly, the analyses in this study
examine both spatial and temporal variations observed in the ambient air monitoring data.
Analysis of these variations ultimately helps determine the origin of chemicals found in urban air
pollution.  In particular, the analyses in this report present compelling evidence that ambient air
concentrations of many VOCs, particularly aromatic hydrocarbons, result primarily from motor
vehicle emissions, while ambient air concentrations of many carbonyls, especially aldehydes, result
from photochemical reactions.

       As noted throughout this report, the ambient air monitoring data collected during the 1995
UATMP serve a wide range of purposes. Not only do these data characterize the nature and
extent of urban air pollution in the vicinity of the sixteen monitoring stations participating in this
study, but these data also indicate some trends and patterns that may be common to all urban
environments.  To that end, this report presents results specific to particular monitoring stations
as well as results apparently common to urban environments. The analyses presented in this
report ultimately should provide additional insight into the complex nature of urban air pollution.

-------
Section 1

-------
 1.0    Introduction
        This report summarizes ambient air monitoring data collected in sixteen urban locations in
 1995 and 1996 as pan of the Urban Air Toxics Monitoring Program (UATMP).  Initiated by the
 U.S. Environmental Protection Agency (EPA) in 1987, the UATMP is designed to characterize
 the nature and extent of potentially toxic air pollution in urban centers.  Since 1987, several state
 and local environmental agencies have participated in the UATMP by developing and
 implementing ambient air monitoring programs. These air monitoring efforts have helped identify
 the toxic chemicals that are most prevalent in urban air and indicate emissions sources that are
 likely to be contributing to elevated chemical concentrations.  This report identifies notable trends
 and patterns in ambient air concentrations measured in 1995 and 1996 and further characterizes
 the complex nature of air pollution in urban environments.

       Urban air monitoring data are of interest to a wide range of groups, including
 environmental and public health agencies, research scientists, and concerned citizens. To
 accommodate the diverse needs and backgrounds of these groups, this report includes both
 general summaries and detailed technical analyses of the UATMP ambient air monitoring data.
 For those interested in further analyzing the monitoring results, all chemical concentrations
 measured under UATMP programs are loaded on the Air Quality Subsystem (AQS) of the
 Aerometric Information and Retrieval System (AIRS), an electronic database maintain^ by EPA.

       The remainder of this report is organized as follows. The figures and tables cited in the
 report text appear at the end of the respective report sections (figures first, followed by tables).

Section 2 ("The 1995 UATMP") describes the sites participating in the 1995 program, the
       sampling and analytical methods used to measure chemical concentrations, and the time
       period over which monitoring stations sampled ambient air.

Section 3 ("Summary of Ambient Air Monitoring Data") summarizes the chemical concentrations
       measured at each ambient air monitoring location. This section addresses the prevalence
       of chemicals in UATMP samples, the range of measured  concentrations, the central

                                          1-1

-------
       tendency of concentration distributions, and the variability in the concentration
       measurements.

 Section 4 ("Statistical Analysis") presents a detailed statistical analysis of the ambient air
       monitoring results. The analysis compares the observed chemical concentration
       distributions to both normal and lognormal standards and identifies correlations between
       concentrations measured for different chemicals.

 Section 5 ("Interpreting Ambient Air Monitoring Data") presents guidelines for interpreting
       ambient air monitoring data collected in urban environments. In particular, the section
       identifies emission sources (including motor vehicles and manufacturing facilities)
       contributing most significantly to high levels of chemicals found in urban air.

 Section 6 ("Spatial Variations") explains how chemical concentrations vary from one monitoring
       location to another.  The section relates these spatial variations to population density,
       industrial emission sources, and motor vehicle emission sources.

 Section 7 ("Temporal Variations") assesses temporal variations in the measured ambient air
       concentrations.  The section presents a quantitative review of seasonal concentration
       patterns and a qualitative review of trends observed over longer periods.

 Section 8 ("Precision and Accuracy ") reviews the replicate analysis of duplicate air monitoring
       samples and comments on the accuracy and precision of concentrations measured at each
       monitoring station.

Section 9 ("Conclusions and Recommendations") summarizes the most significant findings of this
       report and makes several recommendations for analyzing measured chemical
       concentrations in the future.

Section JO ("References") lists the references cited throughout the report.

                                          1-2

-------
Section 2

-------
 2.0    The 1995 UATMP
       Sixteen monitoring stations participated in the 1995 UATMP. These stations measured
 concentrations of 54 different chemical compounds. The measurements presented in this report
 are average air concentrations observed over 24-hour periods.  Such measurements were taken
 every 12 days for a duration close to twelve months.

       Not surprisingly, chemical concentrations observed at the 1995 UATMP monitoring
 stations strongly correlate with the proximity of these stations to chemical emission sources.
 Therefore, this section begins with a brief description of the locations and immediate surroundings
 of the 1995 UATMP monitoring stations. Next, this section provides essential background
 information on sampling and analytical methods, including the meaning and significance of their
 corresponding detection limits. Finally, this section describes the sampling schedules implemented
 at the monitoring stations and the completeness of the monitoring program.

 Note:  Historically, the UA IMP "program year" has been assigned to the year in -which
       participating stations begin sampling ambient air. The UATMP sampling schedule for
       this report ran from August 1995 through September 1996.  Even though the majority of
       data analyzed in this report-were collected in 1996, the program is titled "1995
       UATMP" to remain consistent -withprevious naming conventions.

       2.1    Monitoring Locations
       EPA encourages state and local  agencies to assess the magnitude of toxic chemicals in
urban air pollution.  To take pan in the UATMP, state or local  agencies must develop and
implement air monitoring programs in coordination with EPA and must also pay for a fraction of
the total sampling and analytical costs.

       Figure 2-1 shows the locations of and four letter codes for the 16 monitoring stations
participating in the 1995 UATMP. Of these, only Baton Rouge, Louisiana (B2LA), Camden,
New Jersey (CANJ), and Port Neches, Texas (PNTX) have participated in previous UATMP
                                         2-1

-------
 monitoring efforts. Appendix A contains detailed she descriptions of each monitoring location.
 These site descriptions were reproduced from the AIRS database and were developed from brief
 visual surveys of nearby emission sources and major roadways conducted by she engineers while
 installing monitoring stations. Table 2-1 summarizes the most relevant information for each site.
 Table 2-1 also indicates the total population residing within a 10-mile radius of each monitoring
 station, as determined from 1990 census data (USDOC, 1993). As discussed further in Sections 5
 and 6, both the proximity of industrial emission sources and population density significantly affect
 the composition and magnitude of air pollution in urban environments.

       2.2   Sampling and Analytical Methods
       The UATMP data quality objectives (ES, 1987; USEPA, 1988) identify both volatile
 organic and carbonyl compounds as "contaminants of concern" in urban air pollution.
 Accordingly, the 1995 UATMP monitoring stations measured concentrations of 38 volatile
 organic compounds (VOCs) and 16 carbonyl compounds using the EPA-approved Compendium
 Methods TO-14 and TO-11 (USEPA, 1984a; USEPA, 1984b), respectively. With one exception,
 two separate devices were installed at each monitoring site to collect ambient air samples; the
 Davidson, Tennessee, location sampled only for VOCs and therefore had just one sampling
 device. As specified in the ambient monitoring protocols, 24-hour integrated VOC and carbonyl
 samples were collected in stainless steel canisters and onto silica gel cartridges, respectively. The
 VOC samples were then analyzed by gas chromatography with mass selective detection and flame
 ionization detection (GC/MSD-FID), and carbonyl samples were analyzed using high performance
 liquid chromatography (HPLC) with ultraviolet (UV) detection. Appendices B and C describe the
 respective VOC and carbonyl sampling and analytical methods in greater detail.

       Tables 2-2  and 2-3 list the estimated VOC and carbonyl detection limits for the respective
 sampling methods. Estimated detection limits were determined according to EPA guidance,
'Definition and Procedure for the Determination of the Method Detection Limit" (FR, 1984). A
detection limit is an estimate of the lowest chemical concentration that a sampling and analytical
method can reliably measure. If a chemical concentration in ambient air does not exceed the
                                         2-2

-------
 method sensitivity (as gauged by the detection limit), the analytical method might not differentiate
 the chemical from other chemicals in the sample or the random "noise" inherent in chemical
 analysis. If the concentration of a chemical cannot be quantified by the laboratory analytical
 equipment, the chemical concentration is reported as "non-detect" (ND). The only conclusion
 that can be drawn from an ND result is that the actual chemical concentration in the sample lies
 somewhere between zero and the estimated detection limit. As noted throughout this report, the
 treatment of non-detects in data summaries and statistical analyses significantly influences the
 conclusions drawn from ambient air monitoring results.  Sections 3 and 4 discuss this issue
 further.

       In some cases, laboratory analysis of ambient samples measured chemical concentrations
 at levels lower than the estimated detection limits. These concentrations were considered valid
 results in the data analyses in this report.  Section 8, however, stresses that concentrations
 measured at levels near the estimated detection limit may be highly variable.

       As shown in Tables 2-2 and 2-3, the estimated detection limits for the methods used in the
 1995 UATMP range from 0.003 to 0.39 parts per billion, by volume (ppbv).  Therefore, the
 sampling and analytical methods adopted in this study can generate estimates of ambient
 concentrations to levels at least as low as 0.4 ppbv. Additional monitoring studies involving more
 sensitive sampling and analytical methods would be necessary to examine ambient air
 concentrations at lower levels.

       In addition to minimum detection limits, the silica gel cartridges have maximum capacities
 for sampling carbonyls. Appendix C discusses the reason for this upper bound. If the cartridge
 capacity is exceeded over the course of a sample, the sampling and analytical device cannot
 measure the actual carbonyl concentration. Following standard method procedures, such samples
 are declared invalid.  Only  10 carbonyl samples in the  1995 UATMP (eight at the Hahnville,
Louisiana monitoring station and two at the Baton Rouge, Louisiana monitoring station) were
invalidated for this reason. Sections 3 and 6 revisit this issue.
                                          2-3

-------
       The canisters used to measure VOCs have theoretical maximum sampling capacities. As
 noted in Appendix B, these capacities were never exceeded during the 1995 UATMP.

       2.3    Sampling Schedules
       At every monitoring location, samples were collected by continuously drawing ambient air
 into the stainless steel canisters and through the silica gel cartridges for a 24-hour period,
 beginning and ending at midnight standard time. In general, samples were collected for one
 24-hour period every 12 days throughout the sampling schedule for each site. Several duplicate
 samples were collected at each monitoring location and analyzed in replicate to gauge the
 precision of the sampling and analytical methods. Due to logistical and administrative constraints,
 the annual sampling periods vary from one monitoring location to another.  Table 2-4 lists the
 beginning and ending sampling dates for each monitoring location.

       2.4     Completeness
       The completeness of ambient air monitoring efforts refers to the fraction of attempted
 sampling events resulting hi either quantified chemical concentrations orND results. Invalid
 sampling events may result from various sampling or analytical errors. Appendices D and ฃ
 present the specific reasons for invalidating respective VOC and carbonyl samples during the 1995
 UATMP. Table 2-4 summarizes the completeness of the VOC and carbonyl data sets collected at
 each monitoring location.  As shown in the table, the completeness ranges between 52 and 100
 percent for the VOC sampling and between 68 and 100 percent for the carbonyl sampling at the
 1995 UATMP monitoring stations.  Throughout the 1995 UATMP, 389 of the 487 planned VOC
 samples and 405 of the 465 planned carbonyl samples were successfully analyzed. Therefore, the
 completeness of VOC and carbonyl sampling events for the entire program were 80 and 87
 percent, respectively.  The completeness data shown in Table 2-4 do not include the collection
and analysis of duplicate samples.

       According to the UATMP data quality objectives (USEPA, 1988), at least 15  samples
from a given monitoring station must be successfully analyzed to generate a data set complete
                                         2-4

-------
enough for estimating annual average concentrations. With the exception of the Brattleboro,
Vermont (BRVT) station, all monitoring stations participating in the 1995 UATMP meet the data
quality objectives for sampling completeness. Section 3.1.3 further discusses how to estimate
annual average concentrations from the UATMP monitoring results.
                                        2-5

-------
                                                                         Figure 2-1
                                                   Locations of the 1995 UATMP Monitoring Stations
                       Belllngham. WA
                          (BEWA)
N)
WlnoosM. VT
  (WIVT)
Burlington. VT
   (BUVT)
                                                                                                                             Underfill!. VT
                                                                                                                               (UNVT)
                                                                                                                       TN
                                                       rซ
                                                    (BRTX)    ^                            (QALA)
                                                                       Baton Roug*. LA   X
                                                                           (B2LA)
                                                                                   Hahnvfll*. LA
                                                                                     (HALA)
            Note: The four letter codes shown were used primarily to track ambient air sanqrtes when they were transferred from the monitoring station to
            the analytical laboratory.

-------
                                                             Table 2-1
                                   Site Descriptions for the 1995 UATMP Monitoring Stations
1995
UATMP
Code
B2LA
BEWA
BRTX
BRVT
BUVT
CANJ
DATN
AIRS Site
Code
22-033-0009
53-073-0016
48-061-0006
50-025-0004
50-007-0003
34-007-0003
47-037-0011
Location *
Baton Rouge, LA
Bellingham, WA
Brownsville, TX
Brattleboro.VT
Burlington, VT
Camden, NJ
Davidson, TN
Population Residing
within a 10-Mile
Radius of the
Monitoring Station b
336,577
91,274
125,547
27,862
103,912
2,021,082
426,257
Distance from Monitoring Station
to Interstate Highways Located
within a 10-Mile Radius ฐ
1.9 miles from Interstate 10;
2.2 miles from Interstate 12
0.1 miles from Interstate 5
No interstate highways within 10
miles of the monitoring station
1 .9 miles from Interstate 91
2.3 miles from Interstate 89
1.9 miles from Interstate 95;
2.9 miles from Interstate 76
0.6 miles from Interstate 40;
0.8 miles from Interstate 24;
0.8 miles from Interstate 65
Number of Facilities Located
within a 10-Mile Radius of the
Monitoring Station Reporting Air
Releases of VOCs or Carbonyls to
the Toxic Release Inventory d
9
4
2
2
3
39
13
1 The El Paso and Brownsville, Texas, monitoring stations are located within 10 miles of the Mexico border. Because only U.S. census, roadway, and
industry data were reviewed in this study, the listed site characteristics may not represent the immediate surroundings of these monitoring stations.
b Reference:  USDOC, 1993.
0 As an objective measure of nearby major roadways, this table only indicates proximity to interstate highways. The AIRS site descriptions in
Appendix A offer more detailed information for some monitoring locations.
d Refer to Section 5.1.1.1 for a detailed description of the Toxic Release Inventory.  Reference: TRI, 1994.

-------
                                                            Table 2-1 (Continued)
                                         Site Descriptions for the 1995 UATMP Monitoring Stations
1995
UATMP
Code
EPTX
GALA
GATX
HALA
PNTX
RUVT
UNVT
VAWA
WIVT
AIRS Site
Code
48-141-0027
22-095-0002
48-167-1002
22-089-0003
48-245-0017
50-021-0002
50-007-0007
53-011-0011
50-007-0010
Location *
El Paso, TX
Gaiyville, LA
Galveston, TX
Hahnville, LA
PortNeches,TX
Rutland, VT
Underbill, VT
Vancouver, WA
Winooski,VT
Population Residing
within a 10-Mile
Radius of the
Monitoring Station b
410,475
56,800
103,167
107,033
146,467
38,969
18,997
587,395
109,541
Distance from Monitoring Station
to Interstate Highways Located
within a 10-Mile Radius ฐ
0. 1 miles from Interstate 10
2.8 miles from Interstate 10
4.2 miles from Interstate 45
6.3 miles from Interstate 10;
7.9 miles from Interstate 55
10.0 miles from Interstate 10
None
9.6 miles from Interstate 89
0.9 miles from Interstate 5;
6.9 miles from Interstate 84
0.5 miles from Interstate 89
Number of Facilities Located
within a 10-Mile Radius of the
Monitoring Station Reporting Air
Releases of VOCs or Carbonyls to
the Toxic Release Inventory d
2
5
6
9
15
2
0
30
3
tsJ
io
       * The El Paso and Brownsville, Texas, monitoring stations are located within 10 miles of the Mexico border. Because only U.S. census, roadway, and
       industry data were reviewed in this study, the listed site characteristics may not represent the immediate surroundings of these monitoring stations.
       b Reference: USDOC, 1993.
       c As an objective measure of nearby major roadways, this table only indicates proximity to interstate highways. The AIRS site descriptions in
       Appendix A offer more detailed information for some monitoring locations.
       dRefertoSection5.1.1.1foradetaileddescriptionofthe Toxic Release Inventory. Reference: TRI, 1994.

-------
                                      Table 2-2
                           Estimated VOC Detection Limits
Chemical
Acetylene
Benzene
Bfomochloromethane •. .
Bromodichloromethane
Bromoforro
Bromomethane
1,3-Butadiene
Carbon tetrachloride
Chlorobenzene
Chloroethane
Chloroform \
Chloromethane
Chloroprene ^ , ,^
Dibromochloromethane
JH-Dichlorobenzene
o-Dichlorobenzene
p-Dichlorobenzene
1, 1-Dichloroethane
1.,2-Dicnlorbetnane
trans- 1 ,2-Dichloroethylene
1,2-Dichforopropane
c/5-l,3-Dichloropropylene
rrans-l,3-DIchloropropykne
Ethylbenzene
Methylene chloride
n-Octane
Propylene
Styrene
1,1,2,2-Tettachloroetnane
Tetrachloroethylene
Toluene
1,1,1 -Trichloroethane
1, 1,2-Trichloroetnane
Trichloroethylene
Vinyl chloride
w,/?-Xylene
o-Xvlene
Estimated Detection Limit (ppbv)
% v 0,12 - 	 ,
0.24
"-'^ ""^ , " *"$&?, - ' - -
6.09 ''
" ; ff ; ; 010f;> %
0.18
1 / 0;ts ^/-^ •• xs
0.07
', 0x0$ \ ^ / >•."—-
0.18
QM - \x V
0.39
//,,!. ซ.. $3ฎ V " ^ซ
0.05
' /* ' ' 
-------
                                          Table 2-3
                            Estimated Carbonyl Detection Limits
Chemical
AcetaMehyde
Acetone
Acrolein
Benzaldehyde
Butynlsobirtyraldehyde
Crotonaldehyde
2,5-Dimethylbenzaldehyde
Formaldehyde
Hexanaldehyde
Isovaleraldehyde
PropioDaldebyde
Tolualdehydes
VaJeraMehvde
Estimated Detection Limit (ppbv)
500 liter
sample volume
0,01
0.01
0.03
&ot
0.02
0.01
0,01
0.01
O.OJ
0.10
1,000 liter
sample volume
0,004 , ,
0.004
>OQ4 _, ;
0.01
;&oo<;if
0.01
x *un
0.005
0.003 '
0.01
0.004
0.05
f f f : ', \A
001
1,500 liter
sample volume
* 0.003
0.003
f ff, f, -, \
0.01
L '.*&*& ;v
0.01
0.003
, ; v> -0.002
0.005
, _ 0.003
0.03
0.005 I
Note:  The carbonyl detection limit varies with the volume of ambient air drawn through the sampling
       apparatus. Although the target sample volume for carbonyl monitoring was 1,000 liters, die
       observed sample volumes ranged from roughly 500 to 1,500 liters. Appendix C discusses this issue
       further.

Reference: FR, 1984.
                                            2-10

-------
                                                                     Table 2-4
                                                      Sampling Schedules and Completeness
Code
B2LA
BEWA
BRTX
BRVT
BUVT
CANJ
DATN
EPTX
GALA
GATX
HALA
PNTX
RUVT
UNVT
VAWA
WIVT
...
Monitoring
Location
Baton Rouge, LA
Bellingham, WA
Brownsville, TX
Brattleboro, VT *
Burlington, VT
Camden, NJ
Davidson, TN
El Paso, TX
Garyville, LA
Galveston, TX
Hahnville, LA
Port Neches, TX
Rutland, VT
Underbill, VT
Vancouver, WA
Winooski, VT
Overall Proeram
Sampling Period
Starting
Date
7/20/95
8/13/95
8/25/95
3/16/96
8/13/95
9/24/95
9/6/95
8/25/95
8/13/95
8/25/95
8/13/95
8/31/95
8/13/95
8/1/95
8/25/95
8/13/95
—
Ending
Date
8/25/96
8/31/96
8/12/96
8/25/96
8/19/96
8/25/96
8/13/96
8/25/96
8/7/96
8/25/96
8/19/96
8/25/96
8/19/96
8/19/96
5/27/96
8/19/96
—
VOCData
Samples
Attempted
25
32
33
4
35
33
30
32
37
31
33
34
33
31
31
33
487
Samples
Analyzed
18
24
19
4
29
29
25
32
24
28
21
31
27
31
16
31
389
Completeness
72%
75%
58%
100%
83%
88%
83%
100%
65%
90%
64%
91%
82%
100%
52%
94%
80%
Carbonyl Data
Samples
Attempted
24
35
35
15
34
33
NAb
32
33
31
33
33
33
33
28
33
465
Samples
Analyzed
21
30
24
13
32
28
NAb
30
30
30
23
31
31
33
19
30
405
Completeness
88%
86%
69%
87%
94%
85%
NAb
94%
91%
97%
70%
94%
94%
100%
68%
91%
87%
N)
         * At the Brattleboro, Vermont (BRVT) station, the VOC sampling device was installed several months after the carbonyl sampling device. Furthermore,
         several 3-hour average VOC samples were collected at the BRVT station, but the results of these samples were not considered in this analysis.
         b No carbonyl samples were collected at the Davidson, Tennessee (DATN) monitoring station.

-------
Section 3

-------
 3.0   Summary of Ambient Air Monitoring Data
       This section reviews the large volume of ambient air monitoring data collected during the
 1995 UATMP. The data are summarized in a series of tables, one table per chemical group (VOC
 or carbonyl) per monitoring location. Each table indicates:

 •      The prevalence of chemicals hi urban air samples collected at the site
 •      The range of measured concentrations
 •      The central tendency of the concentration distributions
 •      The variability in the concentration observations

       Section 3.1 describes the parameters used to summarize the air monitoring data with
 respect to the above characteristics.  Appendices F and G present the site-specific summary tables
 for VOC and carbonyl monitoring, respectively. Sections 3.2 and 3.3 include overall summary
 tables characterizing the prevalence, maximum, and central tendency of chemical concentrations
 observed throughout the UATMP. For reference, Appendices H and I list the results of every
 successful sampling event for VOCs and carbonyls, respectively, at every monitoring station.

       3.1    Air Monitoring Summary Parameters
       No single parameter completely characterizes the results of extensive ambient air
monitoring programs.  To provide a  full but succinct picture of the results of the 1995 UATMP,
the summary tables in Appendices F  and G present four parameters that together indicate the
nature and extent of air pollution found at each monitoring location.  These parameters are
prevalence, concentration range, central tendency, and variability. The following subsections
explain how these parameters characterize ambient air monitoring data.

       3.1.1   Prevalence
       Prevalence refers to the frequency with which an event occurs; in this case, it refers to the
frequency with which a chemical is found at detectable levels during the course of ambient air
                                         3-1

-------
monitoring. To indicate the prevalence of VOCs and carbonyls in urban ambient air, the summary
tables in Appendices F and G identify:

•      The total number of valid samples taken at the site
•      The number of ND findings
•      The frequency (%) with which detectable concentrations were found in the samples

       These prevalence data serve two useful purposes. First, they clearly indicate how often
VOC and carbonyl chemicals are present at detectable levels. This information may be useful to
agencies attempting to identify chemicals of concern in urban locations.  Second, the number of
non-detects for a given chemical can be used to estimate the accuracy of annual average
concentrations calculated from the collection of 24-hour measurements.  Sections 3.1.3 and 4.1
address this issue further.

Note:  A prevalence of zero means that the chemical was not detected in any samples.  The
       chemical could be present in the air, but at a concentration below its detection limit (see
       Tables 2-2 and 2-3).

       3.1.2  Concentration Range
       Concentration range refers to the span of the concentration data, from lowest level to
highest. To indicate the range of chemical concentrations measured at a she, the summary tables
present both the lowest and highest concentrations observed at the respective monitoring
locations.  This information, particularly the peak concentrations,  should help local agencies
assess worst-case air pollution scenarios.  For most chemicals, at least one sample at each site
resulted in a non-detect, so the lowest concentration reported is "ND."

       Because samples were collected every  12 days, the highest concentrations cited in the
summary tables only represent estimates of the actual peaks. Chemical concentrations could have
                                          3-2

-------
 risen to higher levels during nonsampling days, but it is likely that the data reported here reflect
 typical peak concentrations.

       3.1.3  Central Tendency (Annual Average Concentrations)
       As noted in Section 2, the 1995 UATMP measured 24-hour average concentrations.
 Many regulatory standards and health guidelines, however, are based on annual average
 concentrations. Furthermore, many agencies assess air pollution trends over the long term using
 annual average concentrations. Due to fluctuations in chemical emissions and local
 meteorological conditions, 24-hour and annual average concentrations can differ significantly.
 With a sufficient number of 24-hour average concentrations, the annual average concentration can
 be reasonably estimated as the central tendency of the 24-hour average concentration distribution.
 The following paragraphs carefully review the assumptions made to calculate central tendency (or
 annual average) concentrations.

       Median, arithmetic mean, and geometric mean concentrations are commonly calculated as
 a measure of the central tendency of a concentration distribution. The shape of the concentration
 distribution, however, determines the most appropriate value to use. As shown in Figure 3-1,
 normally distributed data (having bell-shaped curves) have equal arithmetic mean and median
 values. If ambient air concentrations were normally distributed, the central tendency
 concentration would roughly equal the median and arithmetic mean of an adequate number of
 24-hour average samples. Nonetheless, as shown in many air monitoring studies (Gilbert,  1987),
 including previous UATMP summary reports (Radian, 1990; 1991), ambient air concentrations
 tend to more closely fit lognormal distributions. The lognormal distribution depicted in Figure
 3-1 is asymmetrical with a higher probability of "outlier" observations (e.g., point "C4" in the
 figure) than found in normal distributions. These outliers may exert an undue influence on the
arithmetic mean.  In particular, arithmetic mean concentrations calculated for lognormally
distributed data generally overestimate the actual central tendency.  As shown in Figure 3-1, the
geometric mean more accurately represents the central tendency of lognormally distributed data.
Although the geometric mean provides the best estimate of central tendency  for UATMP
monitoring results, the arithmetic mean and median concentrations have also been included in the
                                          3-3

-------
  Treatment of Non-Detects                                                           <
                                       ""                \        >                 %     •.
  When analyzing ambient air monitoring data, something must be done with non-detect     :
  observations. These observations can be ignored, assigned a concentration of zero, or assigned
  some other proxy concentration. As mentioned previously; anon-detect result indicates Jhat
  Hie actual chemical concentration in a sample is sonievvherebetweeii2^o and the method
  detection limit As a best estimate of the actual concentration, this study assigns all non-detect
  observations a concentration equal to one-half the estimated detection limit tins approach has
  |>ieen followed by many previous monitoring studies and is 4 recommended approach ferifek J
  assessments ; involving environmental monitoring date 
-------
 VOC measurements for the same period, the median, arithmetic mean, and geometric mean values
 for VOCs at that site are only rough estimates of the actual central tendencies.

       As noted above, the accuracy of the central tendency estimates also depends on the
 accuracy of the individual concentration measurements.  According to the method specifications,
 Compendium Methods TO-14 and TO-11 can accurately measure VOC and carbonyl
 concentrations at levels exceeding the estimated detection limits listed in Tables 2-2 and 2-3.  (See
 Section 8 for more detailed information on method accuracy and precision.) Non-detect
 observations, however, cannot be assigned an accurate concentration; the assumed concentration
 of one-half the detection limit represents only an estimate of the actual chemical concentration.
 Therefore, the accuracy of central tendency estimates decreases with an increasing number of
 non-detect observations.  For example, if a chemical has only ND readings for every sample, all
 the concentrations would be assigned a value of one-half the detection limit, and the median,
 arithmetic mean, and geometric mean would all equal this value; yet, the actual central tendency
 concentration could be anywhere between zero and the detection limit. Sections 3.2 and 3.3
 discuss this issue further in the context of the 1995 UATMP results.

       3.1.4  Variability
       Variability refers to the spread of data observations about the central tendency value.
Variability in ambient air monitoring data may be useful to researchers examining the impact of
meteorological and emission fluctuations on ambient air concentrations.  Common measures of
data variability are the standard deviation and the coefficient of variation. Even though standard
deviations may not be the most representative measure of variability for lognormally distributed
data, the parameter has been used to measure variability in many previous ambient air monitoring
efforts.  Accordingly, the summary tables include standard deviations for comparison to other air
monitoring studies.

       Because standard deviations increase with higher data values, standard deviations of
different data sets (e.g., for different chemicals) may not be comparable. The coefficient of
                                          3-5

-------
variation, on the other hand, expresses the standard deviation of a data distribution as a
percentage of the arithmetic mean. By scaling the standard deviation to the mean value, the
coefficient of variation quantifies variability on a uniform scale, allowing comparison across
distributions for different sites and chemicals. The summary tables in Appendices F and G present
both measures of variability; all values were calculated by assigning non-detect concentrations
equal to one-half the estimated detection limit.

       3.2    VOC Summary Tables
       Tables F-l through F-16 (see Appendix F) summarize the VOC monitoring results for the
16 stations participating in the 1995 UATMP. The tables present prevalence, concentration
range, central tendency, and variability data as discussed above. Note that some chemicals were
frequently found at trace levels (below their detection limits). As noted earlier, the accuracy of
central tendency estimates decreases with increasing number of non-detects.  To indicate
frequency of detection, chemicals having less than 50 percent of the observations resulting in non-
detects for a given monitoring site are presented in bold text in the summary tables. The central
tendency estimates for these chemicals should adequately represent the annual average
concentration at each monitoring location. Any comparisons to, or evaluations of, annual average
concentrations for the other chemicals (not in bold text), however, should note that central
tendency concentrations may be significantly influenced by the large number of non-detects.  The
accuracy of the central tendency concentrations for these chemicals can only be improved by
designing ambient air monitoring technologies capable of quantifying chemical concentrations at
trace levels.

       To provide a sense of overall 1995 UATMP results, Tables 3-1 through 3-3 summarize
the ranges of prevalence, maximum, and central tendency data parameters observed across all
monitoring locations. The Brattleboro, Vermont (BRVT) monitoring station collected only four
valid VOC samples. Due to the limited sample size at this station, the BRVT results are not
include in Tables 3-1 through 3-3.
                                          3-6

-------
       Table 3-1 shows the number of monitoring stations that detected VOCs in certain
 percentages of the collected samples. For example (see Table 3-1), 14 monitoring stations
 detected benzene in 75 to 100 percent of the samples collected, while one monitoring station
 detected benzene in 50 to 75 percent of the samples. Therefore, Table 3-1 can be used to identify
 those chemicals most prevalent in the ambient air surrounding the 1995 UATMP monitoring
 locations. Acetylene was observed in at least 75 percent of the samples collected at every
 monitoring station, and many other chemicals were detected in the majority of samples collected
 at most participating monitoring sites.  These chemicals include benzene, carbon tetrachloride,
 chloromethane, ethylbenzene, w-octane, propylene, toluene, 1,1,1-trichloroethane, and the xylene
 isomers.  Because some chemicals may be toxic at levels below their detection limits, prevalence
 alone should not be considered when determining chemicals of concern in urban air pollution.

       Table 3-2 presents a similar summary for the ranges of highest concentrations measured at
 the monitoring locations.  As shown in Table 3-2, the peak concentrations of many chemicals did
 not exceed 1 ppbv at any of the 1995 UATMP monitoring stations. None of the VOCs had peak
 concentrations exceeding 5 ppbv at every monitoring station. Acetylene peak concentrations
 exceeded 10 ppbv at more monitoring locations than any other VOC.

       Table 3-3 summarizes the geometric mean concentrations calculated for every 1995
 UATMP monitoring location. As shown in the table, of the 38 VOCs considered in this study, 24
 had geometric mean concentrations below 1 ppbv at all 15 monitoring stations considered. Only
 acetylene, propylene, and toluene had geometric mean concentrations exceeding 1 ppbv at more
 than five  of the participating monitoring locations.  These geometric mean data may be useful for
 establishing baseline concentrations for the urban locations participating in the  1995 UATMP.

       Although Tables 3-1 through 3-3 efficiently summarize the large volume of ambient air
monitoring data collected during the 1995 UATMP, the VOC data summaries characterize only
the ambient air quality observed at the monitoring locations participating in the UATMP program.
These summaries do not necessarily encompass the prevalence, peak concentrations, and
                                         3-7

-------
geometric mean concentrations expected in all urban environments.  Sections 4 through 7 further
analyze and interpret the VOC results summarized in Tables 3-1 through 3-3.

       3.3   Carfaonyl Summary Tables
       Tables G-l through G-15 (see Appendix G) summarize the carbonyl monitoring results for
the 15 stations measuring carbonyl compounds. Carbonyls were not monitored at Davidson,
Tennessee. Like the VOC summary tables, the carbonyl summary tables include the four
parameters described above; results for chemicals detected in over SO percent of the samples at a
monitoring station are presented in bold type.  As shown in the tables, the majority of carbonyl
compounds meet this criteria.  Therefore, the central tendency estimates for most carbonyl
compounds in this program should be reasonably representative of annual average concentrations.

       Tables 3-4 through 3-6 summarize the ranges of prevalence, maximum, and central
tendency data for the IS ambient air monitoring locations measuring carbonyl compounds. As
shown in Table 3-4, most of the carbonyls were detected in at least SO percent of the 199S
UATMP samples.  At every participating monitoring station, acetaldehyde, acetone, and
formaldehyde were detected in over 75 percent of the collected samples.  Although prevalence
ranges for carbonyl compounds generally exceed those for VOCs (compare Tables 3-1 and 3-4),
the difference partly results from the lower detection limits for carbonyl compounds (compare
Tables 2-2 and 2-3).

       Table 3-5 summarizes the highest carbonyl concentrations recorded during the 199S
UATMP.  Like the VOCs, many carbonyl compounds had peak concentrations lower than 1 ppbv
at most of the monitoring stations. Acetaldehyde and formaldehyde, however, generally had peak
concentrations exceeding 10 ppbv. As shown in the summary tables in Appendix G, these two
chemicals have maximum concentrations exceeding 100 ppbv at some of the monitoring stations.

       Table 3-6 presents the ranges of geometric mean concentrations calculated for the
carbonyl compounds.  Acrolein, benzaldehyde, crotonaldehyde, 2,5-dimethylbenzldehyde, and
isovaleraldehyde generally had geometric mean concentrations measured below 1 ppbv, while

                                         3-8

-------
acetaldehyde and formaldehyde typically had geometric mean concentrations measured above 5
ppbv. The geometric means for the remaining carbonyls varied significantly between the 1995
UATMP monitoring locations.

       As discussed in Section 3.2 for the VOC data, the carbonyl data presented in Tables 3-4
through 3-6 summarize prevalence, peak concentrations, and geometric mean concentrations for
the ambient air monitoring locations shown in Figure 2-1, but may not necessarily represent
carbonyl concentrations in all urban environments. Sections 4 through 7 offer additional insight
to the variations observed across the different monitoring locations.
                                         3-9

-------
                                          Figure 3-1
               General Features of Normal and Lognormal Data Distributions


                                      The Normal Distribution
                                                           Ci=Arith
                                                           Cis Modicn concentration
                                            Concentration
                                    The Lognormal Distribution
                                                                    UodiM of the toBBiflhnw
                                                                    of ino concentratione
                                                                C4-Eปmple'Outller'
                                                                    concentration
                                          Concentration
Notes:  This figure displays continuous probability distributions. Ambient concentrations measured at monitoring
       stations are discrete samples drawn from the concentration distribution.

       For normal distributions, the geometric mean always has a lower magnitude than the arithmetic mean.


                                             3-10

-------
                                          Table 3-1
                              Summary of Prevalence of VOCs
Chemical
Acetylene
Benzene
Bromochloroinethane
Bromodichloromethane
Bromoform ,,,..
Bromomethane
-1 3-Butadiene
Carbon tetrachloride
Chiorobenzene
Chloroethane
CidorofonB
Chloromethane
Chloroprene ..
Dibromochloromethane
o-Dichlorobenzene
p-Bichlorobenzene
1,1-Dichloroethane
1^2-Dichloroefeane
frans-l,2-Dichloroethylene
cw-l,3-Dichloropropylene
Ethylbenzene
Mefhylene chloride
n-Octane
Propyfene
Styrene
Tetrachloroethylene
Toluene
1,1,1-Trichloroethane
1,1^2-TrichIoroetiiane
Trichloroethylene
Vinyl chloride
m,/>-Xylene
o-Xylene
Number of Monitoring Stations with Prevalence in the Listed Range *
0-25 %
0
15
/ r
9
" J
0
, 14
15
4
2
14
!' : 11
10
1
15
15
15
15
15
15
0
2,
0
0
3
12
2
1
0
14
11
15
1
2
25-50%
. 9 ,/
0
0
/ 3 ฐ
?
0
1 * \
0
0
' ^.:. < "'
i
3
0
-,*/ ,/
0
0 , /'
0
0
" 4 . , :
3
3
1
3
i" : " J
i
i
i
0 ;
i
3
50-75 %
" \ 1
0
"'";r"\;
3
2
- o _\
0
'-" 4
2
s : ^ "o 	
o 	
o "_ "^
0
0
o'""
_1 / ^
0
"_"l^w/,
7
0
1
\ o ' ;
0
2
75-100%
14
, LUฎ 111
" % "** 0 '
13
0
_ 1 11 1
1 11ฐ! ,
"1 ^ 11 \
_0
'ซ}, '^.<^vl
0
	 i4 	 .
"12* v™"
8
	 ;i:\:^:
3^
14
'v / Av'V'
, 2,
13
tl i
* Because only four valid VOC samples were collected at Brattleboro, Vermont, the station was not included
in this summary table. Therefore, the table considers results from 15 monitoring stations.
                                            3-11

-------
                                         Table 3-2
                         Summary of Highest VOC Concentrations
Chemical
Acetylene
BmrniV&lrtmiTiethan
Btomodichloromethane
fiffimoiorm.
Bromomethane
1^-Butadiene '- .. •
Carbon tetrachloride
Cfclonibeazene
Chloroethane
Chloroform
Chloromethane
Chloroprene -'
Dibromochlorome thane
o-Dichlorobenzene
1 , 1 -Dichloroethane
1^-Dichloroethane
fro?w-l,2-Dichloroethylene
1^2-13ich1oropropane
CT5-l,3-Dichloropropylene
Jrpbv
0
3
4 , ' J3* ,,
15
12
' , 13 *
15
13" * *
15
' , 'U .."..
0
, IS //
14
15 " ^'
15
15
J5
15
15*" " 	
'15 ;' '
8
, 4 '
12
1
7
15
0
9
14
15
12
2
1-5 ppbv
, , 3^- -
10
0
'" "'f '\- ' "
2
0
.., J ฐ" * , ^
i-
15
0
0
o'
0
•. •, •. A -.f s'
"f 	 o"" """
0
6
'2
4 ^ ^ ^
0
9
6
6
0
11
6
5-10 ppbv
'^t/Vw
"o
%' ง ""
0
b
r ^ 50^ ^
0 ""'
^X1ซJU" "\
.".?.'.*v L.1TI
„/.,, ฐ. ,\' 1
' ! . " ฐm/
___o^
w, -<"-'0x' "V-^J>-
'D\
0
^.,rr „,
0
^ /4 % ' '
2
,0 /A 1
0
, "3
0
o
1
0 "j
>10ppbv
~ AJ' 'KZ-,^<^
ปป„ ซ. ,„ i , .„, ,
,*, v^ป^ป
""'17ฐllrll
'X"*VA"1^J
.,,.,... "1^^ Ij
n ,s,ฎ.v ป^N
',"'"7'o^%'_
,^ปฃ.
'' "" v O % f' v'
S-N vrf USV.O ">.VAW.ซ. W,S-Xซ %•>.
0

ซ,™sป,
^ ^ * „ ,
'2
",\'IT™1
0
*• , , '*,„,
0
^vA
	 o"
1\
• Because only four valid VOC samples were collected at Brattleboro, Vermont, the station was not included
in this summary table. Therefore, the table considers results from 15 monitoring stations.
                                           3-12

-------
                                         Table 3-3
                    Summary of VOC Geometric Mean Concentrations
Chemical
Acetylene
Benzene
BroinochlGFomethane
BromodichlorometHane
Broffiofixm
B th
1,3-Butadicne
Carbon tetrachloride
ChJorobenzene
Chloroethane
Chloroform
Chlorometnane
MM 4 f
v/JuM>iu{ucue ••
Dibromochloromethane
w-Dichlorobenzcne
o-Dichlorobenzene
jvDicblorobenzene
1,1-Dichloroethane
1^-Dicbforoefoane
trans- 1,2-Dichloroethylene
1,2-DichIoropropane
CT5- 1,3-Dichloropropylene
fron5-l,3-Dichloropropylene
Ethylbenzene
Methylene chloride
n-Octane
Psopylene
Styrene
1,1,2,2-Tetrachloroethane
Tetrachloroethylene
Toluene
1,1,1-Trichloroethane
1,1,2-TricUoroethane
Trichloroethylene
Vinyl chloride
m,/>-Xylene
oOtytene
Number of Monitoring Stations with Geometric
Mean Concentration in Listed Range "
0-1 ppbv
: 1' -
15
15 ;"
15
15
15
1*J
14
15
is
15
i ' 15
14
- •• 15 '
15
, "1$
15
! ' - 15 -
15
15
15
13
15
15
15
14
14
9
15
14
15
7
15
15
13
14
13
14
1-2 ppbv
,5 ', -
0
- f fl " v
""o
a
o
\/
0
D :
0
0
1
", & " ' ,' '•
0
$" -- '
0
0
0
0
0
€
0
V
0
D :
1
5
0
0
0
6
0 .
0
0
0
1
I
2-5 ppbv
ฃ ^ ,
0
"x'^ 0 -,
tf ft •. •:
0
!- ^ 0 " '
n
V
0
Q
0
o -•
0
': ("'^ 's's '* ?
0
l% '% o ,^ :
0
$ • •'
0
0
0
0,
0
0
0
0
0
••'•• •• 1-
0
0
0
2
0
0
0
0
1
0
>5 ppbv
-'-'-•& ..<.,->:.,..
0
'-;, -^. *i \t
~' P...' *...""
ซ;-•; ^1^;;^
0 ™ % "
V
	 o
C^'Jc '""
0
^*V -•• >X0;-' "? ซ
"o
" - ;' *4 ->,-.
% ""S6"
,\V %^w'<' *
6
! ^ ;^ :
"s "V
'•ฉ 1 '\
6
^_ ,"'
0
^
0
^r
0
<0~
0
1
0
$
0
ฉ
2
1
0
0 i
'Because only four valid VOC samples were collected at Brattleboro, Vermont, the station was not included
in this summary table. Therefore, the table considers results from 15 monitoring stations.
                                          3-13

-------
                                         Table 3-4
                           Summary of Prevalence of Carbonyls
Chemical
Acetaldefayde
Acetone
Acrdein
Benzaldehvde
Botyr/IsobBtyraldehyde ,
Crotonaldehyde
Formaldehyde
Hexanaldehvde
Isovaleraldehyde
IIWtBL >.ซ•<>] J^lMnlik
jreopionaiflenyoe
Tolualdehydes
Valeraldehvde
Number of Monitoring Stations with Prevalence in the Listed Range *
0-25 %
,0
0
2 ; -
0
0
,• v ' •* •. ^
0
0
0
5
0
25-50%
0
" 2 "--*•;•.
2

-------
                                        Table 3-5
                      Summary of Highest Carbonyl Concentrations
Chemical
Acetaldfihyde
Acetone
Bcnzaldehvde
Crotonaldehyde
2^-Dimedjyibenzalddiyde
Formaldehyde
Hexanaideiivde
Isovaleraldehyde
Tolualdehydes
ValeraMehvd&
Number of Monitoring Stations with
Highest Concentration in Listed Range *
0-1 ppbv
0
, , |3 * v
13
.. ? , .... .: . '•
"n 	 '"
'14 *;
0
s -'
13
9
6
9
1-5 ppbv
' S- ^ X
% " 2
	 4 	 "
3
1
7
4
5-10ppbv
4 - - -
6
^ % ^ ^
o
	 "o """""*"
1
\ <. ^ •.
1
" ^ -ft
1
' If ' *
>10ppbv
\c $ -^'
2
-. • v. •••••.-. •.'•:• '^Av. ^
0
\ •&* Q
-c^-x f &* ^s "•' AV / *•••- ^ "•% ^ J^A
0
11
0
i
* The Davidson, Tennessee (DATN) monitoring station did not sample for carbonyls. Therefore, the table
considers results from IS monitoring stations.
                                          3-15

-------
                                       Table 3-6
                 Summary of Carbonyl Geometric Mean Concentrations
Chemical
Acctaktehydc '
Acetone
AcnoJem \ - - , T
f S V
BoiTaldehvde
fiaiyc/lJMbuijTalddiyde
Crotonaldchydc
2,5-Dimdhj^bcnzaldehyde
Formaldehyde
Hexanakletovde
> WJ-Wf
Isovaleraldehyde
PrQpiQnaldehYde
Tolualdchydes
Valeraldehvde
Number of Monitoring Stations with Geometric
Mean Concentration in Listed Range *
0-1 ppbv
0
0
-, ', & * 'ฐ "
v* , **' , V
13
' *> ,-/1
11
^ ' i,$4 " " ,- -j
0
' '*,',/;
13
, ^
6
9
1-2 ppbv
*, 2 :- ..
7
^•> ^ -i ^ •*>A>/ '""'*•'•• •*
"^ % ''' r-"
2
"0* ,.1
4
V, 'X \|V" - ,-
3
^ s V '/" <, '•• *,
1
- ^^>'"^
7
4'"- 1
2-5 ppbv
' 4 ; -
6
-,.ฅ•• y^<>"-.A * , < ", *< >
, , ^8 "• „ % *
0
. f V, <. / ^,V • d- V ^ %S \ •,
-r, ^ -;>
0
' ••-•X *• •• •*'. •.
'", " s*stf*t •• *'^A*
1
-- ^.;o^..'.. i
•• /^ ^ > >^ w •$-.' •.' t
1
' % :< $
1
". '<* -"^ '
>5ppbv
'*"• •$ ^^5 ' '
••••^O -X---S.^X-"i%X-X ^s^v^-.ซX-.>-.
2
VV v^ '*? •• ^"•."''^
&&}**&?''" I"-' >'
0
f ff* ' ^ y&jc* >"*flvป' ff- ?
^ ' s*''%W*^* * $*.
6
"V /?SX; ^ X-^'X '1
11
f v x p^^ s^ ^ ^%.-Xs^^
S-.-.S f SS.Sfฃt *->.vX~. S X-V^ >
0
*;::*^: -
1
- >3--';;
' The Davidson, Tennessee (DATN) monitoring station did not sample for carbonyls. Therefore, the table
considers results from 15 monitoring stations.
                                         3-16

-------
Section 4

-------
 4.0    Statistical Analysis
        This section characterizes the distributions of data collected under the 1995 UATMP and
 examines correlations among the various chemical compounds. As noted in Section 3, the nature
 of concentration distributions determines the most appropriate methods for interpreting
 monitoring data. Accordingly, Section 4.1 presents several statistical analyses used to compare
 the 1995 UATMP concentration distributions to two common standards: normal and lognonnal
 distributions. These analyses reach the important conclusion that the UATMP monitoring data
 more closely fit lognormal distributions.  To provide additional insight into the monitoring data,
 Section 4.2 evaluates correlations between different chemical concentration distributions. These
 correlations identify data trends common to all of the 1995 UATMP monitoring locations.

        4.1    The 1995 UATMP Concentration Distributions
        Characterizing statistical distributions of urban air pollutants provides a concise and
 familiar means to describe air monitoring data. Two distributions commonly used to characterize
 urban air pollution are normal and lognonnal distributions (see Figure 3-1). As discussed earlier,
 normal distributions exhibit smooth "bell-shaped" curves that are symmetrically centered, and
 lognormal distributions are skewed toward larger values. In terms of air monitoring data, if
 measured concentrations are randomly scattered about a central tendency (or mean) value, the
 monitoring data are said to be normally distributed.  On the other hand, if the logarithms of the
 measured concentrations are randomly scattered about a central tendency, the monitoring data
 are said to be lognormally distributed.  This subtle distinction plays a very important role in
 properly interpreting ambient air monitoring data.

       Section 4.1.1 describes three statistical parameters-skewness, kurtosis, and the Shapiro-
Wilk statistic-that can be used to determine whether data distributions more closely fit normal or
lognormal distributions. Sections 4.1.2 and 4.1.3 summarize the statistical parameters calculated
for the VOC and carbonyl data distributions, respectively.
                                          4-1

-------
       4.1.1   Statistical Parameters
       A number of parameters and statistical tests are commonly used to characterize data
 distributions.  This report considers the skewness, kurtosis, and Shapiro-Wilk statistic in
 comparing air monitoring data to normal and lognormal standards.  These parameters are
 generally not  of interest for regulatory purposes. The interpretations based on these parameters,
 however, have significant implications for regulations addressing air monitoring results. For
 example, the most appropriate measure of annual average concentrations depends on the nature of
 the corresponding concentration distributions.  The following subsections describe how the
 skewness, kurtosis, and Shapiro-Wilk statistic relate to the UATMP concentration distributions.

       4.1.1.1       Skewness
       Skewness, the third moment of a distribution, measures the symmetry of data about their
 median value  (Harnett, 1982). As shown in Figure 4-1, positively skewed distributions have long
 tails extending toward higher values, while negatively skewed distributions have long tails
 extending toward lower values.  The skewness of normally distributed data equals zero. Lithe
 case of lognormally distributed data, however, the skewness of the logarithms of the data points
 equals zero.  The magnitude of the skewness indicates the extent to which distribution symmetries
 deviate from normality.

       Skewnesses were calculated twice for each distribution of ambient air monitoring data:
 once for the distribution of the measured concentrations and once for the distribution of the
logarithms of the measured concentrations.  By definition, the distribution with skewness of
lower magnitude  has symmetry more representative of normally distributed data. Sections 4.1.2
 and 4.1.3 summarize the skewnesses calculated for the UATMP monitoring results.

       4.1.1.2       Kurtosis
       Kurtosis,  the fourth moment of a distribution, measures the peakedness or flatness of data
at their median value.  By definition, the kurtosis of normally distributed data equals three. Many
statistical software packages, however, subtract three from the calculated kurtosis and report an
                                          4-2

-------
 "adjusted" kurtosis. This convention makes the adjusted kurtosis have interpretations
 qualitatively similar to those for the skewness:

 •      Positive adjusted kurtoses indicate peaked distributions with thin tails (see Figure 4-1)
 •      Negative adjusted kurtoses indicate flat distributions with broad tails (see Figure 4-1)
 •      Normally distributed data have adjusted kurtoses equal to zero
 •      Logarithms oflognormalfy distributed data also have adjusted kurtoses equal to zero

       As with skewnesses, adjusted kurtoses were calculated both for the distribution of
 measured concentrations and for the distribution of the logarithms of the measured
 concentrations. The distribution with an adjusted kurtosis of lower magnitude has peakedness
 more representative of normally distributed data. Sections 4.1.2 and 4.1.3 summarize the kurtosis
 calculations for VOCs and carbonyls, respectively.

       4.1.13        Shapiro-Wilk Statistic
       Unlike the skewness and kurtosis, the Shapiro-Wilk statistic does not measure a single
 "shape" characteristic, such as symmetry or peakedness, of a distribution. Rather, the Shapiro-
 Wilk statistic measures the extent to which data distributions deviate from normality (Shapiro and
 Wilk, 1965). Many statistics textbooks, software packages, and literature references describe
 how to calculate the Shapiro-Wilk statistic.  By definition, the Shapiro-Wilk statistic for a data
 distribution is some value between zero and one.  The magnitude of the Shapiro-Wilk statistic
varies with the number of observations for a given data distribution. Accordingly, comparison of
 Shapiro-Wilk statistics across data distributions for different chemicals may be invalid.  Normally
 distributed data and the logarithms oflognormalfy distributed data have Shapiro-Wilk statistics
 equal to one.

       As with the other statistical parameters, Shapiro-Wilk statistics were  calculated for both
the distributions of measured concentrations and the distributions of the logarithms of the
measured concentrations. The distribution having a Shapiro-Wilk statistic closer to unity can be
                                          4-3

-------
 said to be more representative of normally distributed data. The following subsections summarize
 Shapiro-Wilk statistics calculated for both VOC and carbonyl data distributions.

       4.1.2  VOC Distributions
       A total of 570 VOC concentration distributions (38 chemicals measured at IS monitoring
 locations) were initially considered in the statistical analysis. As noted in Section 2.4, however,
 the Brattleboro, Vermont (BRVT) station did not collect enough VOC samples to meet the
 UATMP data quality objectives for completeness.  The VOC concentrations for this station were
 therefore not included in the statistical analysis.

       Skewnesses, kurtoses, and Shapiro-Wilk statistics were calculated both for the
distributions of measured concentrations and for the distributions of the logarithms of the
measured concentrations. In all calculations, non-detect observations were assigned values one-
half the estimated detection Umit. For
distributions having a large number of
non-detects, the statistical parameters
may not accurately characterize the
actual ambient air concentration
distributions (see sidebar). Therefore,
the results in this section only
consider the 181 VOC concentration
Detection limits in enYkonmentd,BiQnh^ring^die$ '.'.
eflFe<2ttve|^ttuncate thelower ซnd ofnaeasured data';f'"
distributions (Giftxert, 1987). Tttstnmcatioii    ;:.  ,;
introduces biases into n^sntiM^pR^isI^^  ~
calculated ^ eavinmmen^ iiu}mioring der^. In this
stiidv: non^deleid: observatian&'hsve beett fls^cnied fite
distributions that contained no more
than 25 percent non-detect
observations.

       For these 181 distributions,
Table 4-1 compares the skewnesses,
kurtoses, and Shapiro-Wilk statistics
calculated both for the measured
opposed to their true value, which is unknown. For
have an artificial clustering of concefttnttioos at ooe~
not rq>resent the distribution of actual chemical
concentrations ai levels not detectable by the
sampling and analytical methods. Therefore, the
frequency of non-detects ultimately fimtts the
accuracy of statistical analyses of environmental
monitoring data.  Interpretations of monitoring data
involving many noa-detect observations drauid be
made wtth caution.
                                          4-4

-------
 concentrations and for the logarithms of the measured concentrations. As shown in the table, the
 distributions of the logarithms of the concentrations meet the following criteria:
 •      The magnitudes of the skewness for 89 percent of the distributions of the logarithms are
       less than those for the corresponding distributions of measured concentrations
 •      The magnitudes of the adjusted kurtoses for 77 percent of the distributions of the
       logarithms are less than those for the corresponding distributions of the measured
       concentrations

 •      Shapiro-Wilk statistics for 86 percent of the distributions of the logarithms are closer to
       unity than those for the corresponding distributions of the measured concentrations

       As discussed in the previous subsections, these three observations all indicate that the
 logarithms of the measured concentrations fit normal data distributions better than the measured
 concentrations themselves. Otherwise stated, VOC concentrations measured during the 1995
 UATMP more closely fit lognormal (rather than normal) data distributions. This results suggests
 that the geometric means presented in Section 3 are more representative of annual average
 concentrations than the corresponding arithmetic mean or median values.

       4.1.3   Carbonyl Distributions
       The statistical analysis of the carbonyl monitoring data closely parallels the statistical
 analysis of the VOCs.  A total of 182 carbonyl concentration distributions were initially
 considered (13 chemicals measured at 14 monitoring stations). The Davidson, Tennessee
 (DATN) monitoring station did not collect carbonyl samples and was therefore not considered in
this analysis. As noted in Section 2.2, eight samples collected at the Hahnville, Louisiana
(HALA) monitoring station were invalidated because the sampling capacity of the silica gel
cartridge was exceeded. Because elevated carbonyl concentrations caused these samples to be
invalidated, the measured carbonyl  concentration distribution at HALA likely does not include
several of the highest concentrations. This suggests that the upper end of the HALA carbonyl
distribution may be artificially truncated,  and the statistical parameters for HALA may not
                                          4-5

-------
 necessarily represent the actual distribution characteristics. Accordingly, the HALA
 concentration distributions were not considered in the carbonyl statistical analysis.

       As with the VOC analysis, only the carbonyl concentration distributions having no more
 than 25 percent non-detect observations were compared to normal and lognormal standards. This
 eliminated 70 of the 182 carbonyl concentration distributions from the statistical analysis. Table
 4-2 summarizes the skewness, kurtosis, and Shapiro-Wilk statistic results for the remaining 112
 carbonyl distributions. The table shows that distributions of the logarithms of carbonyl
 concentrations meet the following criteria:

 •      The magnitudes of the skewness for 73 percent of the distributions of the logarithms are
       less than those for the corresponding distributions of measured concentrations

 •      The magnitudes of the adjusted kurtoses for 63 percent of the distributions of the
       logarithms are less than those for the corresponding distributions of measured
       concentrations

 •      Shapiro-Wilk statistics for 73 percent of the distributions of the logarithms are closer to
       unity than those for the corresponding distributions of measured concentrations

       As with the VOC results, these observations again indicate that the logarithms of the
measured concentrations fit normal data distributions better than the measured concentrations
themselves.  Therefore, both the VOC and carbonyl concentration distributions fit lognormal
distributions more closely than normal distributions, and the concentration measurements should
be interpreted accordingly.

Note:  Neither the VOC nor carbonyl monitoring data exactly fit lognormal or normal
       distributions.  Nonetheless,  the previous statistical analyses clearly indicate that both the
       VOC and carbonyl monitoring data more closely fit the shape and characteristics of
       lognormally distributed data.  This result suggests that geometric mean concentrations
                                          4-6

-------
       are better indicators of annual average concentrations than either the arithmetic mean or
       median concentrations (see Section 3).

       4.2    Data Correlations
       Data correlations measure the direction and strength of linear relationships between two
 variables (Harnett, 1982).  In the context of the UATMP statistical analysis, data correlations
 among ambient air concentrations identify notable trends and patterns that may not be readily
 apparent. These trends and patterns, or lack thereof, ultimately characterize the extent to which
 emission sources common to all urban environments contribute to the composition and magnitude
 of urban air pollution.  This study uses Pearson correlation coefficients to evaluate the
 relationships between concentrations of different chemicals. Subsection 4.2.1 summarizes
 relevant properties of the Pearson correlation coefficients and outlines how these parameters are
 applied to the UATMP monitoring data.  Subsections 4.2.2 and 4.2.3 then discuss the nature and
 extent of correlations observed among VOC and carbonyl monitoring data, respectively.

       4.2.1  Methodology
       Pearson correlation coefficients measure the degree of correlation between two variables.
 The methodologies used to calculate these correlations are presented in most common statistics
 references (Snedecor and Cochran,  1980). By definition, the Pearson correlation coefficients
 always tie between -1 and 1; where a correlation coefficient of-1 indicates a perfectly linear
 "negative" relationship  between two variables and a correlation coefficient of 1 indicates a
 perfectly linear "positive" relationship. Negative relationships occur when increases in the
 magnitude of one variable are associated with decreases in the magnitude of the other; and
 positive relationships occur when the magnitudes  of two variables both increase or both decrease
 accordingly. Data that  are completely uncorrelated have Pearson correlation coefficients of zero.
Pearson correlation coefficients closest to -1 or 1  indicate highly correlated data.
                                          4-7

-------
       In this analysis, correlation coefficients were calculated between concentrations measured
for different chemicals. Unlike the distribution analysis in Section 4.1, the correlation analysis
considers the results from all monitoring stations combined when evaluating Pearson correlation
coefficients. Following this approach, the calculated correlations indicate trends and patterns in
the chemical concentrations measured across all monitoring stations. Any significant correlations
observed for all monitoring stations should indicate whether emission sources common to urban
environments, such as motor vehicles, contribute most significantly to the measured
concentrations.

       Although correlations could also be calculated between the chemicals for particular
monitoring stations, such correlations would have a greater chance of being influenced by the
limited number of observations at each monitoring site. This limitation ultimately leads to a
higher probability of spurious correlations among the site-specific data and, therefore, does not
necessarily provide for a meaningful statistical analysis. For qualitative comparisons of the
similarities and differences between concentrations measured at each monitoring location, Section
6 summarizes the spatial variations in chemical concentrations and includes some evaluations of
site-specific concentration profiles.

       In all correlation calculations, non-detect observations were replaced with concentrations
of one-half the estimated detection limit. As with the statistical parameters considered in Section
4.1, the Pearson correlation coefficients are biased for chemicals having large numbers of non-
detects. As an extreme example, both vinyl chloride and bromodichloromethane were detected in
less than 25 percent of the 1995 UATMP samples (see Table 3-1). Because non-detect results are
all substituted with values of one-half the estimated detection limit, the calculated Pearson
correlation coefficient for these chemicals would indicate a very high degree of correlation, even
though there is insufficient information available to reach such a conclusion. To avoid problems
such as this, only those chemicals detected in over 75 percent of the UATMP samples (see Tables
3-1  and 3-4) were considered in the correlation analysis.
                                          4-8

-------
       4.2.2  Correlations Among VOCs
       Of the 38 VOCs measured in the 1995 UATMP, 11 were evaluated for data correlations
 (i.e., 11 were detected in over 75 percent of the samples collected throughout the program). For
 these 11 chemicals, Pearson correlation coefficients were calculated between the 55 possible
 chemical pairings.  The correlations were evaluated by aggregating the concentration data for
 every chemical across the 15 monitoring stations measuring VOCs.  Table 4-3 summarizes the
 results of the Pearson correlation coefficients found to exceed 0.5, a level chosen as indicative of
 a strong correlation between chemical pairs. This level also represents a natural break among the
 calculated correlation coefficients. More specifically, the correlation coefficients for chemical
 pairs not shown in Table 4-3 generally fell between 0 and 0.3. This lower range of correlation
 coefficients is assumed indicative of uncorrelated data.

       Table 4-3 recognizes two chemical pairings (benzene and ethylbenzene, and m,p-xy\ene
 and o-xylene) that have Pearson correlation coefficients greater than 0.9.  This high degree of
 correlation across all monitoring sites suggests that concentrations of these chemical pairs likely
 result from emission sources consistent across urban environments.  Section 6 presents compelling
 evidence that motor vehicle emissions significantly contribute to ambient air concentrations of
 these aromatic hydrocarbons. The remaining correlations listed in Table 4-3 are significantly
 weaker than those for benzene/ethylbenzene and m,/?-xylene/0-xylene. The weaker correlations
 suggest only subtle connections between concentrations of the xylene isomers and w-octane,
 acetylene, and toluene. These weaker correlations indicate greater variability among the
 concentrations measured at the different monitoring stations and, therefore, are not necessarily the
result of emission sources common to all urban environments.

       Of particular importance, Table 4-3 shows that just seven of the 55 possible chemical
pairings exhibit correlations at levels indicative of underlying trends or patterns in the monitoring
data.  The lack of significant correlations among the UATMP monitoring data suggests that the
composition of air pollution significantly varies from one urban location to another.
                                          4-9

-------
       4.2.3  Correlations Among Carbonyls
       Of the 16 carbonyl compounds considered in the 1995 UATMP, nine were detected in at
least 75 percent of the collected samples. Pearson correlation coefficients were calculated
between the 32 possible chemical pairings in a fashion similar to the VOC calculations. Table 4-4
presents the 6 carbonyl pairings having Pearson correlation coefficients greater than 0.5.  Unlike
the VOC results, none of the carbonyl pairings have Pearson correlation coefficients exceeding
0.9, suggesting that sources of carbonyls in ambient air, whether from emissions or photochemical
reactions, are not consistent among the 1995 UATMP monitoring locations. As with the VOC
analysis, the carbonyl correlation analysis strongly suggests that the composition of air pollution is
not uniform across all urban environments.
                                         4-10

-------
                                  Figure 4-1

                       Examples of Skewness and Kurtosis
          (A) Positive Skewness
(B) Negative Skewness
            Concentration
   Concentration
         (Q Positive Kurtosis
(D) Negative Kurtosis
I
ซ
2
             Concentration
                                                         Concentration
                                     4-11

-------
                                    Table 4-1
                      Statistical Analysis of VOC Distributions
                           (181 Distributions Considered)
          Criteria
   Distribution of the
Measured Concentrations
(% of Total Distributions)
    Distribution of the
Logarithms of the Measured
      Concentrations
 (% of Total Distributions)
Distributions that have a
lower magnitude skewness
       20(11%)
        161 (89%)
Distributions that have a
lower magnitude adjusted
kurtosis
       42 (23%)
        139 (77%)
Distributions that have a
Shapiro-Wilk statistic closer
to one
       25 (14%)
        156 (86%)
                                    Table 4-2
                    Statistical Analysis of Carbonyl Distributions
                          (112 Distributions Considered)
          Criteria
   Distribution of the
Measured Concentrations
(% of Total Distributions)
    Distribution of the
Logarithms of the Measured
      Concentrations
 (% of Total Distributions)
Distributions that have a
lower magnitude skewness
       30 (27%)
        82 (73%)
Distributions that have a
lower magnitude adjusted
kurtosis
       41 (37%)
        71 (63%)
Distributions that have a
Shapiro-Wilk statistic closer
to one
       30 (27%)
        82 (73%)
                                      4-12

-------
               Table 4-3
  Significant Correlations Among VOCs
     (55 Possible Chemical Pairings)
Chemicals
Benzene
w,/>-Xylene
m,/>-Xylene
o-Xylene
o-Xylene
m,/>-Xylene
o-Xylene
Ethylbenzene
o-Xylene
w-Octane
w-Octane
Acetylene
Toluene
Toluene
Pearson Correlation
Coefficient
0.98
0.96
0.78
0.73
0.56
0.54
0.53
               Table 4-4
Significant Correlations Among Carbonyls
     (32 Possible Chemical Pairings)
Chemicals
Formaldehyde
Butyr/Isobutyraldehyde
Butyr/Isobutyraldehyde
Benzaldehyde
Butyr/Isobutyraldehyde
Benzaldehyde
Benzaldehyde
Propionaldehyde
Benzaldehyde
Propionaldehyde
Formaldehyde
Hexanaldehyde
Pearson Correlation
Coefficient
0.81
0.80
0.73
0.60
0.56
0.53
                 4-13

-------
Section 5

-------
 5.0    Interpreting Ambient Air Monitoring Data
        This section presents methods for interpreting trends and patterns in ambient air
 monitoring data.  As shown in the summary tables in Section 3 and the statistical analyses in
 Section 4, the composition of urban air pollution varies greatly between the 1995 UATMP
 monitoring locations.  The presence of a wide range of industrial and motor vehicle emission
 sources in urban environments complicates efforts to make sense of air pollution measurements.
 Nonetheless, Section 5.1 discusses how air monitoring data can be put into perspective by
 analyzing emission inventories and atmospheric fete and transport mechanisms.  Section 5.2
 describes the context that historical monitoring data can give to current air monitoring results. To
 illustrate appropriate factors to consider when interpreting measured air pollutant concentrations,
 Section 5.3 provides an example analysis of spatial variations in 1,3-butadiene concentrations
 observed during the 1995 UATMP. Following this example, interested readers can analyze and
 interpret concentration trends for other chemicals not directly addressed in this report. Sections 6
 and 7 examine additional spatial and temporal variations in ambient air concentrations using the
 methods presented in this section.

       5.1    Emissions and Dispersion
       Chemicals found in urban air pollution come from a wide range of emission sources. The
 nature and magnitude of these chemical emissions largely determine the chemical composition of
 urban air pollution. Local meteorology and atmospheric chemistry, on the other hand, determine
 how quickly chemical emissions disperse in ambient air. Sections 5.1.1 and 5.1.2 review
 resources available for determining how chemical emissions and atmospheric fate and transport
 affect urban air pollution.  Although sophisticated models for attributing ambient air pollutant
 concentrations to specific emission sources have been reported in the literature (Scheff, 1993), the
 resources presented in the following paragraphs should suffice for explaining general trends in air
monitoring data.
                                          5-1

-------
       5.1.1  Chemical Emissions
       Industrial, motor vehicle, and natural emission sources account for most chemicals found
in urban air pollution (Graedel, 1978).  The following paragraphs characterize each of these
sources and present references for determining the magnitude and composition of their chemical
emissions.  In some cases, accurate emissions data may help identify the origin of chemicals found
in urban air pollution.  The complex mixture of emission sources hi urban environments, however,
makes source attribution efforts extremely difficult.  Section 5.3 and Section 6 present specific
examples of using emissions data to interpret air monitoring results.

       5.1.1.1        Industrial Emission Sources
       Airborne chemical emissions from industrial facilities significantly contribute to the
pollution observed in urban centers. Industrial air emissions of VOCs and carbonyls are generally
classified into two categories: stack emissions and fugitive emissions. Stack emissions include all
chemical releases through confined air streams, such as process vents and smoke stacks. Fugitive
emissions account for all other chemical releases to the air, typically, VOC and carbonyl fugitive
emissions result from evaporative losses from equipment leaks and chemical processing areas.
The total amount of chemicals released into the air through stack and fugitive emissions varies
from urban area to urban area.

       Federal, state,  and local agencies compile air emission inventories to estimate total
quantities of air releases within their jurisdictions.  At the federal level, the Toxic Release
Inventory (TRI) (see sidebar on next page) contains extensive emissions data for a wide range of
industries and provides an excellent reference for evaluating how industrial emissions might
impact urban air concentrations in different parts of the country.  The accuracy of TRI data is not
known, but it is certainly limited by the accuracy of emission estimates provided by industrial
facilities. Section 6 presents an analysis of how TRI data can be used to evaluate UATMP
results.  Readers interested in evaluating results observed at specific monitoring locations should
also consider emission inventories prepared by state and local agencies.
                                           5-2

-------
  Toxic Release Inventory                                                 -    -
                                                                                  \
  Section 315 of the Emergency Planning and Community Rj^Ho-Know Act (BK^) requires
  industrial facilities to disclose information characterizing environmental releases of "hazardous'*
  chemicals. The TRI hazardous chemical fist currently includes over 600 chemicals.  The XRI
  reporting requirement applies to fecilities that:
 •ซ       Have at least 10 foll-tnne employees                                 •.,,'•..• ^
  *      M into Standard Industrial Classification (SIC) Codes 20 through 39 (most industrial
    '     inanufactunn2 facuitiesj    '       s      %                  s           '
  ป  -     ManufecUn-eT process, or otherwise use hazardous cheim
 Facilities meeting these criteria must submit "Fonn-K" reports to EPA specifying &e quantifies
 of hazardous chemicals released to the environment or transferred to offsite locations.  Every
 year, nearly 80,000 form-k emisstem reports are safeinitted to EPA and loaded into ^etUl
 database.  At the writing of this report, TRI data from reporting years 19E7 through 1994 were
 publicly avaOable (TRI, 1994).
       5.1.1.2        Motor Vehicle Emission Sources
       In addition to industrial emission sources, motor vehicles significantly contribute to air
pollution observed in urban environments. Accordingly, analysis of urban air monitoring data
must consider contributions from motor vehicle sources. Chemicals found in motor vehicle
exhaust generally result from incomplete combustion of vehicle fuels.  Although modern vehicles
and, more recently, vehicle fuels have been engineered to minimize air emissions, all motor
vehicles with internal combustion engines emit a wide range of chemical pollutants. The
magnitude of these emissions in urban areas primarily depends on the volume of traffic; whereas
the chemical profile in these emissions depends more on vehicle design and fuel content.  Several
studies have generated chemical profiles of motor vehicle exhaust (Conner, Lonneman, Seila,
1995), and such profiles provide additional context for analyzing spatial variations in air pollutant
concentrations (see Section 6.3).
                                          5-3

-------
       5.1.1.3        Natural Emission Sources
       Natural decay processes in oceans and forests also emit certain chemicals into the air.
Their contribution to air pollution in urban areas is generally minimal compared to industrial and
motor vehicles sources, but can be significant in waste-laden areas, such as landfills. According to
the AIRS site descriptions in Appendix A, none of the sites participating in the 1995 UATMP are
located near large landfills. Therefore, natural emission sources are not considered in the
following analyses of UATMP results.

       5.1.2  Atmospheric Fate and Transport
       The fete of chemicals emitted into the air is primarily determined by atmospheric
dispersion and photochemical reactions. Atmospheric dispersion refers to the natural phenomena
by which chemical emissions are gradually transported to regions of lower concentration by
convection and diffusion.  Dispersion behavior depends on both emission source characteristics
(such as stack height and exh velocity) and local meteorological conditions (such as cloud cover
and wind velocity).  Photochemical reactions both consume and produce pollutants in urban
environments, and the rate constants for the various reactions vary with temperature and chemical
composition.

       Atmospheric dispersion and photochemical reactions must be considered when analyzing
ambient air monitoring data, since both factors help explain the fete and transport of chemical
emissions in urban environments. Although dispersion models have been developed to evaluate
the fete and transport of chemical emissions from a wide range of sources (USEPA, 1986), a
detailed dispersion analysis is beyond the scope of the current work. The analyses in Section 5.3
and Section 6 briefly consider the fate and transport of chemical emissions with respect to the
UATMP monitoring results, and the analyses in Sections 6 and 7 relate ambient air concentrations
of carbonyls to photochemical reactivity.

       5.2    Other Air Monitoring Studies
       Ambient air monitoring  data can also be put into perspective by comparing measured
concentrations to results of other studies.  The comparisons must be made carefully, however, to
                                          5-4

-------
 evaluate other studies in proper context. In particular, sampling and analytical methods, detection
 limits, and proximity of monitoring locations to emission sources may differ from study to study,
 making direct comparison of measured concentrations difficult. Some excellent sources of air
 monitoring data include previous UATMP reports (Radian, 1990; 1991), lexicological profiles
 published by the Agency for Toxic Substances and Disease Registry (ATSDR), and monitoring
 reports published by state and local environmental agencies. The scientific literature contains
 many additional references to previous monitoring efforts. The analyses in Section S.3, Section 6,
 and Section 7 all use other air monitoring studies to interpret trends and patterns in the 1995
 UATMP monitoring data.

       5.3    Example Analysis: 1,3-Butadiene
       As an example of interpreting air monitoring data, the following discussion evaluates the
 1,3-butadiene concentrations measured at the 16 monitoring stations participating in the 1995
 UATMP. The example has been provided to illustrate appropriate issues to consider when
 analyzing air monitoring results and should prove useful to readers interested in interpreting
 results for other chemicals. It attempts to answer one general question: What do the 1995
 UATMP monitoring results say about 1,3-butadiene concentrations in urban environments?

       5.3.1  Geometric Mean Concentrations
       Figure 5-1 displays 1,3-butadiene geometric mean concentrations calculated for the 1995
UATMP monitoring stations. For reference, the figure also shows the magnitude of the estimated
 1,3-butadiene detection limit (0.15 ppbv, see Table 2-2) with respect to the geometric mean
concentrations.  Clearly, the geometric mean concentrations observed at all monitoring stations,
except for Port Neches, Texas (PNTX), fall below the estimated detection limit.  The
 1,3-butadiene geometric mean concentration for PNTX, on the other hand, exceeds the detection
limit and is at least an order of magnitude greater than geometric mean concentrations observed at
the other UATMP monitoring locations.
                                          5-5

-------
       5.3.2  Interpretations
       As noted in Section 3, ambient air concentration measurements at levels below the
estimated detection limit may be unreliable.  As a result, the 1,3-butadiene geometric mean
concentrations for all monitoring stations (except for PNTX) should only be considered estimates
of actual annual average concentrations. For these monitoring locations, additional air monitoring
efforts involving more sensitive sampling and analytical methods would be necessary to measure
the ambient 1,3-butadiene concentrations more accurately.

       In the case of the PNTX station, however, the high geometric mean concentration allows
for a more detailed analysis using the factors described in Sections 5.1 and 5.2.  The significant
difference between 1,3-butadiene concentrations measured at PNTX and at the other 1995
UATMP monitoring stations suggests that 1,3-butadiene emission sources unique to the Port
Neches area likely contribute to the observed elevated concentrations. Indeed, previous studies
indicate that 1,3-butadiene is manufactured primarily at petrochemical facilities in Texas and
Louisiana and is used (generally in smaller quantities) almost exclusively in rubber and plastic
manufacturing facilities located throughout the country (ATSDR,  1992). According to TRI data
for reporting year 1994, nine industrial facilities within a 10-mile radius of the PNTX monitoring
station released 1,3-butadiene in stack or fugitive air emissions. Figure 5-2 indicates the locations
of, and total 1,3-butadiene air emissions from,  these nine facilities  (TRI, 1994).  Atmospheric fate
and transport data suggest that 1,3-butadiene readily participates in photochemical reactions.
More specifically, previous studies report a 1,3-butadiene half-life in urban ambient air of
approximately two hours (ATSDR, 1992), suggesting that distant emission sources of
1,3-butadiene contribute little to the observed concentrations at PNTX. Therefore, the
1,3-butadiene measured at PNTX likely originates in large part from the facilities shown in Figure
5-2.

       Of particular interest, a different UATMP monitoring station collected ambient air data at
another location in Port Neches, Texas for the 1990 and 1991 UATMP program years (Radian,
1991). Table 5-1 lists the geometric mean 1,3-butadiene concentrations measured in Port Neches,
Texas during the 1990, 1991, and 1995 UATMP programs.  As shown in the table, the geometric

                                          5-6

-------
 mean 1,3-butadiene concentration measured at PNTX in 1995 clearly exceeds those measured
 during 1990 and 1991. This trend likely results from a combination of factors, including changes
 in 1,3-butadiene emissions over the long term and the slightly different locations of the PNTX
 monitoring stations.

       Figure 5-3 summarizes the history of 1,3-butadiene air releases reported to TRI for
 facilities within a 10-mile radius of the PNTX monitoring station. According to the TRI data
 shown in Figure 5-3, the air emissions of 1,3-butadiene from facilities in the immediate vicinity of
 Port Neches have decreased dramatically since 1987. On the other hand, the UATMP monitoring
 in PNTX (see Table 5-1) suggests that ambient air concentrations of 1,3-butadiene have increased
 significantly between 1990 and 1995.  Although inaccurate TRI reporting may account for the
 increased ambient concentrations despite the apparent decrease in emissions, the following
 paragraphs suggest that differing locations of the PNTX monitoring stations may offer a better
 explanation of the ambient 1,3-butadiene concentration trend. Section 6.2.3 offers additional
 explanations for discrepancies between emission inventories and ambient air monitoring
 measurements.

       Figure 5-2 shows the location of the 1990 and 1991 PNTX monitoring station with
 respect to both the  1995 PNTX monitoring station and the facilities reporting 1,3-butadiene
 releases to TRI. The industrial facilities shown in the figure all have reported 1,3-butadiene
releases to TRI for every reporting year since 1987.  Figure 5-2 indicates that the 1995 PNTX
monitoring station is located within one mile of two industrial 1,3-butadiene emission sources, but
the 1990 and  1991 monitoring station is located slightly more distant from the most significant
 1,3-butadiene emission source in the Port Neches urban area.  This added distance from the
emission source may account for the lower concentrations observed during the 1990 and 1991
UATMP program years, despite the higher 1,3-butadiene emissions at the time. Atmospheric
dispersion models can  assess the impact of distance from emission sources on ambient air quality
(USEPA, 1986), but a detailed dispersion analysis is beyond the scope of this report. In any case,
                                          5-7

-------
extended ambient air monitoring efforts at fixed locations around Port Neches would be necessary
to more accurately evaluate 1,3-butadiene concentrations over the long term.

       53.3   Conclusions
       1,3-Butadiene concentrations at almost every monitoring station participating in the 1995
UATMP were no greater than 0.15 ppbv and were too low to be reliably quantified by the
sampling and analytical equipment used in this study.  Therefore, emission sources common to
urban environments, such as motor vehicles, likely do not cause 1,3-butadiene concentrations in
ambient air to exceed the estimated detection limit.

       At the PNTX station, elevated 1,3-butadiene concentrations apparently result from air
emissions from petrochemical and rubber manufacturing facilities located within a 10-mile radius
of the monitoring station. Although previous UATMP monitoring in the Port Neches area found
significantly lower 1,3-butadiene concentrations than did the 1995 UATMP study, the difference
appears to result from the placement of the 1995 station closer to facilities known to emit greater
amounts of 1,3-butadiene (see Figure 5-2).
Note:  The methodology used in this analysis can be readily extended to other chemicals
       considered in the 1995 UATMP.  To highlight the most general trends and patterns in the
       UATMP data, Sections 6 and 7provide additional information on the spatial variations
       and temporal variations of chemicals measured in ambient air.
                                          5-8

-------
              2.00
(J\
vb
          o
                                                        Figure 5-1
                               Geometric Mean 1,3-Butadiene Concentrations Calculated for
                               the 16 Monitoring Stations Participating in the 1995 UATMP
                                      (Refer to Table 2-1 for monitoring station codes)
                                 Minimum detection limit - 0.15 ppbv
              0.00
                                                             i     i
                                                        Monitoring Station

-------
                                   Figure 5-2
             Air Emissions of 1,3-Butadiene within a 10-Mile Radius of the
                   Port Neches, Texas (PNTX) Monitoring Station
                   (Emissions Data from TRI Reporting Year 1994)
 Q-jBeaumont
                   \7,8171bs

             / 6,600 Ibs
   Orange County
                                                                        Area of
                                                                        Detail
                                                Port Neches, TX   %
                                                Monitoring Station \
                                      ,300 Ibs     (1995 UATMP)
             Port Neches, TX
            Monitoring Station
         (1990 and 1991 UATMP)

                Jefferson County
  Key:
                    x
  * Monitoring station *-x^
  — County boundary
  QCity
 - • US Interstate
                                 8,600 Ibs e
                                     3,870 Ibs
6,300 Ibs
Source: TRI, 1994.
                                       5-10

-------
                                                   Figure 5-3
                       Total 1,3-Butadiene Air Emissions Reported by Facilities Located
                     within a 10-Mile Radius of the Port Neches Monitoring Station to TRI
2,000,000
=• 1.500,000


|

ฃ 1.000,000
0
I
  500.000  ;
           1.750,730 Ibs.
 ; * "• •.' f \


i-<--:> v  ,,
• <'•. 4 " •  \,s, -
  ~.  ^<' ^ 'ซ,>N .
                        1,309,954 Ibs.
                                     1.072.756 Ibs.
                                                    799.252 Ibs.
                                                                    366.700 Ibs.    372.371 Ibs.    361,656 Ibs.
                                                                                                          133.556 Ibs.
              1987
                               1988
                                1989
1990          1991

TRI Reporting Year
1992
1993
1994

-------
                                    Table 5-1
                   Geometric Mean 1,3-Butadiene Concentrations
                          Measured at Port Neches, Texas
UATMP
Program Year
1990
1991
1995
Geometric Mean 1,3-Butadiene
Concentration (ppbv)
0.39
1.11
1.82
Number of Valid
Samples Collected
29
30
31
Note: As shown in Figure 5-2, the 1995 UATMP monitoring station in PNTX was at a different
      location than the 1990 and 1991 UATMP monitoring station.
                                      5-12

-------
Section 6

-------
 6.0    Spatial Variations
       This section describes the extent to which measured urban air concentrations vary from
 one UATMP monitoring station to another and offers several explanations for the observed
 spatial variations. As shown in this section, analysis of these spatial variations helps identify the
 sources of chemicals found in urban ambient air. Using a simple ranking scheme, Section 6.1
 identifies the monitoring stations that consistently measured elevated concentrations of VOCs and
 carbonyls. The section qualitatively considers how industrial emissions and population density
 correlate with the magnitude of urban air pollution measured at the different monitoring stations.
 Using the data interpretation guidelines outlined in Section S, Sections 6.2 and 6.3 present more
 detailed analyses of the spatial variations observed for selected chemicals. These analyses
 examine the impact of industrial and motor vehicle emission sources on the measured ambient air
 concentrations. Section 6.4 then summarizes the important conclusions drawn from evaluating
 spatial variations among concentrations measured during the 1995 UATMP.

       6.1    Spatial Variations in the Overall Magnitude of Air Pollution
       A simple ranking scheme was developed to show how chemical concentrations vary at the
 different UATMP monitoring locations.  Each location was ranked according to its overall level
 of urban air pollution relative to the other monitoring locations. Because the ranking scheme
 characterizes the overall magnitude of air pollution rather than focusing on the concentrations of
individual chemicals, the site ranks are a convenient indicator of how VOC and carbonyl
concentrations as a whole vary among the monitoring stations participating in the 1995 UATMP.
(Refer to Sections 6.2 and 6.3 for trends and patterns in the spatial variations observed for
specific chemicals.)

       The 1995 UATMP monitoring stations were ranked as follows:
 1.      For a given chemical,  each monitoring station was assigned a unique chemical rank.
       The chemical rank indicates the relative magnitude of the station's measured geometric
       mean concentration with respect to those measured at the other monitoring stations. For
       example, the monitoring station recording the highest geometric mean concentration for a
       given chemical was assigned a chemical rank of one, the monitoring station recording the
                                         6-1

-------
       second highest geometric mean concentration was assigned a chemical rank of two, and so
       on. This step was repeated for every chemical considered in the 1995 UATMP.

 2.     The chemical ranks for a given monitoring station -were averaged to determine a site-
       average rank.  The magnitude of the site-average rank indicates how often chemicals were
       measured at elevated geometric mean concentrations. More specifically, monitoring
       stations measuring elevated concentrations of many different chemicals have she-average
       ranks closer to one, and monitoring stations consistently measuring low concentrations (or
       non-detects) have site-average ranks closer to sixteen.

 3.     The site-average ranks for the different monitoring stations -were ordered in a list, from
       lowest to highest. An overall rank was assigned to each monitoring station based on its
       position in the list. More specifically, an overall rank of one was assigned to the
       monitoring station with the lowest site-average rank (Le., the station with the greatest
       magnitude of air pollution). An overall program rank of two was assigned to the
       monitoring station with the second-lowest site-average rank, and so on. The overall rank,
       therefore, provides a convenient indicator for comparing the magnitude of air pollution
       observed at different monitoring stations.


       Following this ranking scheme, overall ranks were determined separately for VOCs and

 carbonyls.  The ranking results are presented in subsections 6.1.1 and 6.1.2, respectively. To

 provide additional context, the subsections also present data summarizing industrial emissions and

 population density  in the immediate vicinity of the monitoring locations. Finally, subsection 6.1.3
 compares the VOC and carbonyl ranks to determine whether the spatial patterns for these two
 chemical groups differ.


Note:  The overall ranks only characterize the extent of air pollution -with respect to the

       monitoring locations participating in the 1995 UATMP. Accordingly, stations-with ranks

       closer to one have higher levels of air pollution than the other monitoring locations

       shown in Figure 2-1, but do not necessarily have higher levels than other urban
       environments.
                                           6-2

-------
       6.1.1  Ranking by Overall Level of VOC Pollution
       Table 6-1 ranks the 1995 UATMP monitoring stations by overall level of VOCs. The
 ranks suggest that the monitoring station at Bellingham, Washington (BEWA) generally measured
 higher VOC concentrations than the other monitoring stations.  On the other hand, the Underbill,
 Vermont (UNVT) monitoring station consistently measured lower VOC concentrations than the
 other stations. For the most part, the monitoring stations located in Vermont measured lower
 VOC concentrations than the UATMP monitoring stations located elsewhere. To give additional
 context to these spatial variations, Table 6-1 also presents industrial emissions and population
 density data and rankings. The subsections below identify the sources of these data and explain
 how they relate to the overall ranks.

       6.1.1.1       Comparison of VOC Ranks to Industrial Emissions
       The VOC industrial emissions data shown in Table 6-1 summarize total air releases
 reported by industrial facilities located within 10 miles of each monitoring station. Emissions data
 were downloaded from the TRI database (TRI, 1994). The data listed in Table 6-1 are the sum of
 the emissions reported only for those VOCs considered in the 1995 UATMP. For reference,
 Figures 6-1 through 6-16 display the locations of the industrial emission sources with respect to
 the 16 monitoring stations participating in the 1995 program. To facilitate comparison with the
 overall VOC pollution ranks, the emissions in Table 6-1 were also ranked. The area with highest
 overall VOC emissions was assigned an industrial emission rank of one, and the other areas were
 assigned other industrial emission ranks accordingly.

       Comparing the overall VOC pollution ranks to the industrial emission ranks (both listed in
 Table 6-1) roughly indicates the extent to which industrial emissions contribute to the observed
 spatial variations. With three notable exceptions (the Bellingham, Washington; ฃ1 Paso, Texas;
and Hahnville, Louisiana monitoring stations), the VOC ranks are similar to the industrial
emission ranks.  These three exceptions may be explained as follows:

•      As indicated in Table 2-1 and shown  in Figure 6-2, the Bellingham, Washington (BEWA)
       monitoring station is located less than one tenth of a mile from Interstate 5, the main highway
       connecting the Seattle and Vancouver (Canada) urban centers. The proximity to the heavily

                                          6-3

-------
       traveled highway may account for the high VOC rank, despite the low industrial emission
       rank.
       As indicated in Table 2-1 and shown in Figure 6-8, the El Paso, Texas (EPTX) monitoring
       station is also located less than one tenth of a mile from a major interstate highway. This
       proximity to a major source of motor vehicle emissions likely accounts for the discrepancy
       between the industrial emission and VOC pollution ranks.
       As shown in Table 6-1, the Hahnville, Louisiana (HALA) station has the second highest
       industrial emission rank but the fourth lowest VOC pollution rank. No single explanation
       accounts for the difference between these ranks.  Section 6.2.3, however, presents several
       reasons for apparent discrepancies between ambient  air monitoring results and TRI
       emissions data.
       Despite these exceptions, the general agreement between the two sets of ranks suggests
that industrial emissions may contribute significantly to the VOC concentrations observed in urban
ambient ah*. The exceptions do, however, indicate that motor vehicle emission sources also must
be considered when examining the spatial variations of VOC concentrations. Section 6.2
considers the impact of industrial emissions on ambient air concentrations in greater detail.

       6.1.1.2       Comparison of VOC Ranks to Motor Vehicle Emissions
       Data on the volume of motor vehicle traffic would be most useful for comparing overall
VOC pollution ranks to motor vehicle emission sources.  The AIRS she descriptions in Appendix
A, however, generally do not include detailed information on roadways near the participating
monitoring stations. In the absence of such data, this study uses population density as an
indicator of motor vehicle traffic, since more densely populated regions generally have more
motor vehicles on their roadways. Table 6-1 lists the population, as determined from 1990 U.S.
Census data, residing within a 10-mile radius of each monitoring station. The population density
data in Table 6-1 are ranked as the industrial emission ranks were; the monitoring stations with
the highest population residing within a 10-mile radius was assigned a population density rank of
one, with the other stations being assigned other population density ranks accordingly.
                                          6-4

-------
       As shown in Table 6-1, the population density ranks, like the industrial emission ranks,
 generally agree well with the overall VOC pollution ranks. The most significant exception again
 is the Bellingham, Washington (BEWA) monitoring station. As noted previously, the close
 proximity to a major interstate highway likely explains the elevated VOC concentrations measured
 at this location. The exception also illustrates the trouble with using population density as an
 indicator of motor vehicle emissions: major highways are found in urban centers as well as
 suburban and rural areas. Accordingly, population data alone does not necessarily indicate
 regions with higher motor vehicle emissions. Nonetheless, the correlation between the population
 density and overall VOC pollution ranks in Table 6-1 suggests that motor vehicle emissions
 contribute somehow to the ambient concentrations measured during the 1995 UATMP. Section
 6.3 further considers the impact of motor vehicle emissions on ambient concentrations of VOCs.

       6.1.13        Conclusions from VOC Ranks
       As discussed above, data on industrial emissions and surrogate data for motor vehicle
 emissions both correlate well with the overall VOC pollution ranks assigned to the various
 UATMP monitoring stations.  This preliminary analysis suggests that both industrial and motor
 vehicle emission sources may be contributing to elevated concentrations of VOCs in urban
 ambient air.  Sections 6.2 and 6.3 provide more detailed analyses of the impact of various
 emission sources on urban air quality.

       6.1.2   Ranking by Overall Level of Carbonyl Pollution
       Table 6-2 ranks the 1995 UATMP monitoring stations by overall level of carbonyls.  Like
the VOC summary table, Table 6-2 includes industrial emission and population density ranks. The
industrial emission ranks are based on total air releases of carbonyls (as opposed to VOCs) and
therefore differ from the industrial emission ranks listed in Table 6-1.

       During the 1995 UATMP program, the Garyville, Louisiana (GALA) monitoring station
recorded the highest carbonyl concentrations, while the Underbill, Vermont (UNVT) monitoring
station recorded the lowest. As with VOCs, Vermont monitoring stations generally measured
                                          6-5

-------
lower concentrations of carbonyls than stations in other geographic regions, although the

difference is not as pronounced. Unlike the VOC results, the overall carbonyl pollution ranks

shown in Table 6-2 do not correlate with the industrial emission ranks or the population density

ranks. This suggests that emissions of carbonyls from industrial and motor vehicle sources are not
the largest contributors to the carbonyl concentrations measured at the UATMP monitoring

stations.  A considerable proportion of the carbonyls found in urban ambient air must therefore

originate from other sources.  Sections 6.2 and 6.3 provide compelling evidence that carbonyls

produced by photochemical reactions likely make up a large portion of the carbonyls found in

urban air pollution.


      6.1.3  Comparison of Carbonyl and VOC Overall Program Ranks

      To provide additional insight into the observed spatial variations in urban air pollution, the
overall VOC pollution ranks were compared to the carbonyl ranks to determine whether elevated

VOC concentrations tend to be accompanied by elevated carbonyl concentrations, and vice versa.
With three exceptions, the overall VOC and carbonyl pollution ranks are similar.  The three

exceptions may be explained as follows:
       The Camden, New Jersey (CAN!) monitoring station has an overall VOC rank of two and
       an overall carbonyl rank of fourteen.  The high rank for VOCs is likely explained by the
       large number of industrial and motor vehicle emissions found in this densely populated
       urban area. The low rank of carbonyls is less easily explained. Nevertheless, close
       examination of invalid sample data indicates that a series of carbonyl samples collected at
       CANJ were invalidated during September and October, 1995 (see Appendix E). As
       shown hi Section 7.2, the carbonyl chemicals generally have highest concentrations during
       wanner months, and the gap in sampling data at CANJ partly explains the relatively low
       concentrations of carbonyls measured at the monitoring station.

       The Garyville, Louisiana (GALA) monitoring station has an overall VOC rank of eleven
       and an overall carbonyl rank of one. Unlike the case for CANJ, this apparent discrepancy
       cannot be explained  by incomplete sampling data.  Nonetheless, Sections 6.2 and 6.3
       provide compelling evidence indicating that carbonyls, particularly aldehydes, are formed
       in the air typically as products of photochemical reactions.  As discussed in Section 6.4,
       oxidation of airborne hydrocarbons is the predominant reaction mechanism forming
       aldehydes in urban ambient air. Accordingly, elevated carbonyl  concentrations at GALA
       may result from elevated ambient levels of hydrocarbons known to participate in
       photochemical reactions. Many of these hydrocarbons, particularly methane and ethane,
                                         6-6

-------
       were not monitored in the vicinity of GALA. Therefore, not enough information is
       available to determine conclusively whether the elevated levels of carbonyls measured at
       the GALA station indeed result from photochemical reactions.
 •      The Hahnville, Louisiana (HALA) monitoring station has an overall VOC rank of twelve
       and an overall carbonyl rank of three. As shown in Figure 6-9, the HALA monitoring
       station is located roughly fifteen miles from the GALA monitoring station. Due to the
       close proximity of the stations, the explanations provided for GALA (above) also apply to
       the HALA monitoring station.

       Despite these exceptions, the relative agreement between overall program ranks for VOCs
 and carbonyls indicates that concentrations of the two appear to be positively correlated. The
 detailed analyses in Section 6.3 provide further explanations for the correlations between these
 concentrations.
       6.2    Impact of Industrial Emissions
       Although overall pollution ranks provide a simple and effective means of examining spatial
variations in pollution patterns, they do not reveal trends and patterns among the individual
chemicals. To complement the general analysis in Section 6.1, this section examines the impact of
industrial emissions on ambient air levels of specific VOCs and carbonyls. In particular,
subsections 6.2.1 and 6.2.2 compare TRI emissions data for selected VOCs and carbonyls to the
ambient concentrations measured at the 1995 UATMP monitoring stations. As these subsections
demonstrate, TRI emissions data generally cannot explain the spatial variations seen in the
concentrations of the individual chemicals most prevalent in urban ambient air.

       6.2.1   Industrial Emissions of VOCs
       This subsection compares the observed geometric mean concentrations  of selected VOCs
to the emissions reported by industrial facilities. As discussed in Section 3, the geometric mean
concentrations of chemicals detected in the majority of the UATMP samples are more accurate
than those of rarely detected chemicals. Accordingly, this analysis considers only those VOCs
detected in over 75 percent of the  samples collected from at least 10 of the 1995 UATMP
monitoring stations.  The following VOCs meet this criterion:
                                          6-7

-------
Acetylene                  Chloromethane       Propylene                 w,/>-Xylene
Benzene                   Ethylbenzene         Toluene                   o-Xylene
Carbon tetrachloride        w-Octane            1,1,1-Trichloroethane
       With two exceptions, emissions data for these chemicals are available from TRI (TRI,
 1994). According to TRI reporting guidelines (USEPA, 1995), industries are not required to
 disclose releases of acetylene or n-octane to TRI.  These two chemicals are therefore not
 considered here.

       Table 6-3 compares the emissions data for each of the chemicals listed above (excluding
 acetylene and n-octane) to the respective geometric mean concentrations measured at every
 monitoring station. The emissions data in Table 6-3 indicate total air releases of specific
 chemicals reported by facilities located within 10 miles of the 1995 UATMP monitoring stations.
 For ease of interpretation, the data in Table 6-3 are listed in order of decreasing geometric mean
 concentration.

       For all the chemicals considered, the emissions data and ambient air concentrations shown
 in Table 6-3 have no notable correlations.  For example, some chemicals have relatively low
 concentrations in monitoring locations with many nearby industrial emission sources, while other
 chemicals have high concentrations in locations with no industrial emissions at all.  Therefore,
 although industrial emissions certainly contribute in part to the concentrations measured in this
program, the TRI emissions data alone do not explain the spatial variations seen in the levels of
VOCs prevalent in urban ambient air. Subsection 6.2.3 presents several reasons why TRI
emissions data may not correlate with measured ambient air concentrations.

       6.2.2 Industrial Emissions of Carbonyls
       Using the approach presented in subsection 6.2.1, TRI emissions data also were compared
to geometric mean concentrations of carbonyls measured at the various UATMP monitoring
stations.  Of all the carbonyls, only acetaldehyde, butvr/isobutyraldehyde, formaldehyde, and

                                           6-8

-------
 propionaldehyde met the chemical selection criterion noted in subsection 6.2.1. More specifically,
 these chemicals both were detected in over 75 percent of the samples collected from at least 10 of
 the monitoring stations and are among the chemicals required to be reported to TRI (USEPA,
 1995). For these chemicals, Table 6-4 compares the TRI industrial emissions data to the
 geometric mean concentrations measured at the UATMP monitoring stations.  Like the data in
 Table 6-3, the data in Table 6-4 are organized in order of decreasing geometric mean
 concentration.

       For the chemicals shown in Table 6-4, TRI emissions data do not explain the varying
 concentrations of carbonyls measured during the 1995 UATMP. For example, the geometric
 mean concentrations for propionaldehyde and butyr/isobutyraldehyde clearly differ from
 monitoring location to monitoring location, but none of the industrial facilities located within 10
 miles of the UATMP monitoring stations reported air releases of these chemicals to TRI. The
 emissions data for acetaldehyde and formaldehyde also do not correlate with the measured
 geometric mean concentrations.  These observations indicate that emissions of carbonyls from
 industrial facilities are not a significant source of the carbonyls detected at the  1995 UATMP
 monitoring locations. Section 6.3 offers additional insight into the origin of carbonyls in urban air
 pollution.

       6.2.3   Emission Inventory Interpretations
       The previous discussion concludes that spatial variations in the levels of individual VOCs
 and carbonyls cannot be explained solely by spatial variations in industrial emissions. Recall that
overall VOC levels were correlated to industrial emissions (see subsection 6.1.1.1), while overall
 carbonyl  levels were not (see subsection 6.1.2). The apparent discrepancies between the TRI
emissions data and the geometric mean concentrations of individual VOCs and carbonyls can be
explained by any combination of the following factors:
       TRI emissions data may not accurately represent industrial emissions of individual
       chemicals.
                                          6-9

-------
 •      Spatial variations in local meteorological conditions may cause emissions to disperse
       differently from urban area to urban area.
 •      Quantities of chemicals emitted from motor vehicles or formed in photochemical reactions
       may outweigh quantities of chemicals in TRI releases.

       Because the accuracy of TRI emissions data is not known and a detailed dispersion
 modeling analysis is beyond the scope of this project, the impact of the first two factors cannot be
 readily assessed. The following section considers whether motor vehicle emissions might account
 for the spatial variations in the ambient air concentrations of individual VOCs and carbonyls better
 than industrial emissions.
       6.3    Impact of Motor Vehicle Emissions
       As noted in Section 5.1.1.2, the magnitude of emissions from motor vehicles generally
depends on the volume of traffic in urban areas, but the composition of these emissions depends
more on vehicle design. Because the volume of traffic clearly varies from urban area to urban
area, the magnitude of ambient air concentrations resulting from motor vehicle emissions is also
expected to vary. On the other hand, the distribution of motor vehicle design (i.e., the relative
number of cars of different styles) may be quite similar across urban areas.  Accordingly, the
composition of ambient air pollution resulting from motor vehicle emissions should exhibit less
variability.

       To determine the extent to  which motor vehicle emissions contribute to air pollution hi
urban centers, this section examines how the composition of air pollution varies from one
monitoring location to the next. For additional context, the observed compositions are compared
to those reported in other studies.  Subsections 6.3.1 and 6.3.2 analyze the compositions of VOCs
and carbonyls commonly found in motor vehicle exhaust.
                                          6-10

-------
       6.3.1   VOC Profiles
       Hydrocarbon emissions from motor vehicles generally result from the incomplete
 combustion of vehicle fuels (Wayne, 1991). Because these fuels contain a wide range of
 hydrocarbon species, the associated emissions also have many chemical constituents, including
 many VOCs.  Nonetheless, a previous air monitoring study has indicated relatively constant
 compositions of ambient air samples collected on the shoulder of a heavily traveled urban
 roadway (Conner, Lonneman, Seila, 1995). In particular, the VOCs benzene, toluene,
 ethylbenzene, and the xylene isomers (commonly referred to as BTEX compounds) were detected
 at levels well exceeding the method detection limits for the study, and they occurred in about the
 same proportions throughout the monitoring program (Conner, Lonneman, Seila, 1995). The
 composition of the BTEX compounds in this roadside study also exhibit excellent agreement with
 compositions reported in other studies, including previous UATMP summary reports (Radian,
 1996).

       To compare 1995 UATMP monitoring results to the BTEX concentration profiles
 reported in the roadside study, the data from the two studies were converted to a common unit of
 measurement. The roadside concentration profiles were reported as BTEX compositions (i.e.,
 mole fraction or volume fractions), while the UATMP measured concentrations, which cannot be
 converted to compositions without more information. Composition ratios and concentrations
 ratios are equivalent, however.  For this analysis, therefore, the UATMP data were converted to
 ratios of the geometric mean concentration of a given BTEX compound to that of ethylbenzene.
Because ethylbenzene generally has the lowest geometric mean concentration of the BTEX
 compounds, expressing ratios in terms of ethylbenzene helps ensure that all concentration ratios
are greater than one.

       Figure 6-17 displays the ratios described above. For comparison, the results of the
roadside study (Conner, Lonneman, Seila, 1995) are shown next to the UATMP monitoring
 station results. All ratios are displayed on the same scale.  Figure 6-17 clearly indicates that, with
the possible exception of the Brownsville, Texas (BRTX) BTEX profile, the profiles for all of the
                                        6-11

-------
 1995 UATMP monitoring stations are at least similar in shape, with the profiles for many stations
 nearly identical. The similarities between these profiles and the reference roadside profile strongly
 suggest that emissions from motor vehicles are the primary source of BTEX compounds in urban
 ambient air. With regard to the BRTX monitoring station, it should be noted that ethylbenzene
 was detected in only 42 percent of the samples collected.  Therefore, the accuracy of the
 geometric mean concentration for ethylbenzene at BRTX may be limited by the frequency of
 detection (see discussion in Section 3.1.3).

 Note:  The nearly uniform BTEX concentration profiles for many of the UATMP monitoring
       stations confirms one of the major findings of the statistical analyses presented in Section
       4.2.2: the most significant Pearson correlation coefficients between VOC compounds
       generally are observed for pairs of aromatic hydrocarbons.

       6.3.2  Carbonyl Profiles
       The previous analysis links air concentrations of certain VOCs (BTEX compounds) to
 emissions of these compounds from motor vehicles.  This section presents a similar analysis for
 carbonyls.  The roadside study, however, did not consider carbonyls. Nevertheless, if carbonyls,
 like BTEX compounds, are directly correlated to motor vehicle emissions, the ratios of
 concentrations of carbonyls to concentrations of ethylbenzene should be similarly consistent
 across monitoring locations.

       Figure 6-18 shows the ratios of the geometric mean concentrations of the two most
 prevalent carbonyls, acetaldehyde and formaldehyde, to that of ethylbenzene.  Unlike the VOC
 profiles, the carbonyl profiles for the 1995 UATMP monitoring locations are not similar.
 (Comparisons were performed for many of the other carbonyls, and none appeared to have trends
 consistent across all monitoring stations.) This suggests that the presence of carbonyls in urban
 ambient air cannot be explained as simply as the VOCs. Although emissions from motor vehicles
likely contain carbonyl constituents, the data in Figure 6-18 clearly show that the ambient levels of
carbonyls cannot be linked strictly to motor vehicle emissions.
                                         6-12

-------
       Although Figure 6-IS does not indicate similarities across all monitoring locations (as
 does Figure 6-17), the carbonyl profiles in Figure 6-18 do exhibit similarities between monitoring
 stations located within close proximity. For example, the Baton Rouge, Garyville, and Hahnville
 (Louisiana) monitoring stations all exhibit similar aldehyde profiles. Similarities are also observed
 between the Vermont monitoring stations. As discussed in the next section, the ambient air
 concentrations of carbonyls appear to photochemical reactions involving airborne VOCs, possibly
 explaining the regional trends mentioned here.

       6.4    Summary
       As shown throughout this section, spatial variations in ambient air concentrations offer
 insight into the potential sources of chemicals detected in urban air pollution. The following
 conclusions can be made:

       6.4.1  VOC Summary
       Although the general analysis in Section 6.1 indicates the possibility of a correlation
between VOC industrial emissions and VOC ambient air concentrations, the detailed analysis in
 Section 6.2 reveals significant discrepancies between TRI emissions data and measured ambient
air concentrations for individual chemicals. These discrepancies may result from inaccuracies in
the TRI emissions data, influences of local meteorological conditions, or contributions from
emission sources not reportable to TRI. In the case of certain VOCs (aromatic hydrocarbons),
however, Section 6.3 provides compelling evidence that ambient concentration  profiles clearly
resemble those found in motor vehicle exhaust.  Thus, motor vehicle emissions generally outweigh
industrial emissions in terms of their contributions to ambient air concentrations of aromatic
hydrocarbons. Because other VOCs were not examined, it cannot be determined if this trend
applies to other VOCs.

       6.4.2  Carbonyl Summary
       As shown both in the general analysis in Section 6.1 and the detailed analyses in Sections
6.2 and 6.3, spatial variations in carbonyl concentrations measured during the 1995 UATMP do
                                         6-13

-------
not correlate with carbonyl emissions from either industrial or motor vehicle sources indicating
that these emissions are not the primary sources of carbonyls found in urban ambient air. Rather,
carbonyls found in urban air probably originate in large part from some other source, most likely
photochemical reactions.  The seasonal variations presented in Section 7 support this hypothesis.
Many atmospheric chemistry textbooks also indicate that carbonyls, particularly aldehydes, are
formed as products in photochemical reactions (Wayne,  1991; Brimblecombe, 1986). In
particular, the aldehydes typically are formed by the oxidation of hydrocarbons in the atmosphere
(Wayne,  1991). Therefore, ambient concentrations of carbonyls may be better correlated to
ambient concentrations of hydrocarbons than to carbonyl emissions. Although a detailed analysis
of photochemical reaction mechanisms is beyond the scope of this project, the reader can refer to
atmospheric chemistry references for additional information on how carbonyls are formed in urban
ambient air (Wayne, 1991; Seinfeld, 1986).
                                          6-14

-------
                                     Figure 6-1
     Facilities in the Vicinity of the Baton Rouge, Louisiana (B2LA) Monitoring Station
                    Reporting VOC or Carbonyl Air Releases to TRI
                                     10-mile radius
                                                 East
                                             Baton Rouge
                                                Parish
                               Baton Rouge   Baton Rouge, LA
                                        ~*~Monitormg Station
                West
         \  Baton Rouge
                Parish
  O
                                                                      Livingston
                                                                        Parish
•frXMonitoring station

D
     County boundary
     City
     US Interstate
                          Iberville Parish
                                                                    \
   Release Categories:
                        O Aromatics          & Olefins             O Carbonyls
                        • Chlorinated Aromatics  A Chlorinated Olefins    T Halogenated Paraffins
Source: TRI, 1994.
                                       6-15

-------
                                    Figure 6-2
   Facilities in the Vicinity of the Bellingham, Washington (BEWA) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
                  \
                     \
\
             10-mile radius
                                      Whatcom
                                       County
                            Area of Detail
                                             Bellingham, WA
                                          i   Monitoring Station
                                                 Skagit
                                                 County
                                                     T-V
                                        *  Monitoring station
                                    •   —  County boundary

                                     I
                                                                   City
                                                                   US Interstate
   Release Categories:
 O Aromatics          ^ Olefins             O Carbonyls
 • Chlorinated Aromatics A Chlorinated Olefins    y Halogenated Paraffins
Source: TRI, 1994.
                                       6-16

-------
                                      Figure 6-3
       Facilities in the Vicinity of the Brownsville, Texas (BRTX) Monitoring Station
                    Reporting VOC or Carbonyl Air Releases to TRI
                                      10-mile radius
                                             Brownsville, TX
                                            . Monitoring Station
   Key:
  * Monitoring station v^%
  — County boundary
  D City
 - -  US Interstate
   Release Categories:
O Aromatics          ^ Olefins            O Carbonyk
• Chlorinated Aromatics A Chlorinated Olefins   V Halogenated Paraffins
Source: TRI, 1994.
                                        6-17

-------
                                     Figure 6-4
     Facilities in the Vicinity of the Brattleboro, Vermont (BRVT) Monitoring Station
                    Reporting VOC or Carbonyl Air Releases to TRI
                                     10-mile radius
                  Windham County, VT
                                                              Cheshire Bounty, NH
                                           Brattlebbro, VT
                                           Monitoring Station
                                                                     NEW HAMPSHIRE
                                                                       SETTS
  Key:
     lonitoring station %%
 — County boundary
 D City
- •  US Interstate
                                                  \
                               Franklin County, MA*,
   Release Categories:
                         O Aromatics          ^ Olefins            O Carbonyls
                         • Chlorinated Aromatics A Chlorinated Olefins   T Halogenated Paraffins
Source: TRI, 1994.
                                       6-18

-------
                                     Figure 6-5
      Facilities in the Vicinity of the Burlington, Vermont (BUVT) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
                          Grand Isle
                         County, VT rfO-mile radius
 Clinton County, NY
                     Monitoring Station      ^T\Vinooski, VT
                                      ""-ND     *   Monitoring Station
                                         "*Burlinton           <
          'ป Essex County, NY
                                   Chittenden County, VT
                                                                              \
   Key:
  * Monitoring station
  — County boundary
  D City
  - - US Interstate
   Release Categories:
O Aromatics          ^ Olefins            O Carbonyls
• Chlorinated Aromatics A Chlorinated Olefins   T Halogenated Paraffins
Source: TRI, 1994.
                                       6-19

-------
                                    Figure 6-6
     Facilities in the Vicinity of the Camden, New Jersey (CANJ) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
                       Montgomery County, PA
                                    10-mile radius
                        J     O  Philadelphia  /
                                                            Burlington County, NJ
                                                                        \
                                                              O
   Delaware
  County, PA
                                        * Camden, NJ
                                          Monitoring Stati
                        Gloucester County, NJ
  if Monitoring station
  — County boundary
  D City
 - •  US Interstate
   Release Categories
O Aromatics          ^ Olefins            O Carbonyls
• Chlorinated Aromatics ฑ Chlorinated Olefins   T Halogenated Paraffins
Source: TRI, 1994.
                                       6-20

-------
                                    Figure 6-7
      Facilities in the Vicinity of the Davidson, Tennessee (DATN) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
        AiaofDcaul
                               >r
                                     10-mile radius
                                                            Sumner County
                                   \
                                     •<*
                                                R
                                             Hendersonville
           / Davidson County
                           OAT
                J    ^
               V
                 i
                                              tV
                                  /
                           "
        Nashville-Davidson
             ^> 1/Davidson, TN
              '^Ivlpnitoring Station
              /
                  \^    \*
                                                         OAT
            V
            . \
                   it
\
  * Momtormg stafion—-
                                                          \   /
                                                           .-^
                                                               ^
                                                                   \
   Release Categories:
O Aromatics          ^ Olefins
• Chlorinated Aromatics A Chlorinated Olefins
         O Carbonyls
         T Halogenated Paraffins
Source: TRI, 1994.
                                      6-21

-------
                                     Figure 6-8
        Facilities in the Vicinity of the El Paso, Texas (EPTX) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
 Dona Ana
County, NM /
NEWMEXICO
  MEXICO
               El Paso, TX
               Monitoring Station
   Key:
                      ^
  * Monitoring station s%ป^
  — County boundary
  D City
 — •  US Interstate
   Release Categories:
OAromatics          AOlefins             QCarbonyls
• Chlorinated Aromatics A Chlorinated Olefins    T Halogenated Paraffins
Source: TRI, 1994.
                                       6-22

-------
                                     Figure 6-9
      Facilities in the Vicinity of the Garyville, Louisiana (GALA) Monitoring Station
                    Reporting VOC or Carbonyl Air Releases to TRI
                    Livingston Parish
                                                       no
                •  Ascension Parish
                                                   I

                                             i
                                             \   I ป**
                                               *     ^i
                                               *    \ ^*
                                               %    \

                                               \   \
                                                *
                                                    \
                                                            OTA
                                           Garyville, LA
                                           Monitoring Station
                                                              —-
                                                              St Charles Parish
                                                          o
                                  St John the Baptist
                                        Parish
   Key:
  * Monitoring station  **ป
  — County boundary
  D City
 — •  US Interstate
                                 •'   Hahnville,
                                     Monitoring Station
   Release Categories:
OAromatics          AOkfins             QCarbonyls
• Chlorinated Aromatics 4 Chlorinated Olefins    T Halogenated Paraffins
Source: TRI, 1994.
                                       6-23

-------
                                 Figure 6-10
   Facilities in the Vicinity of the Galveston, Texas (GATX) Monitoring Station
                Reporting VOC or Carbonyl Air Releases to TRI
                               OQAT"!
                                         OTA
                                 \
                                     \
               Galveston County
                                               alveston, TX
                                             Monitoring Station
                                                      Gulf of Mexico
   Key?
     Monitoring station
  — County boundary
  DCity
 - •  US Interstate
   Release Categories:
O Aromatics          ^ Olefins             O Carbonyls
• Chlorinated Aromatics A Chlorinated Olefins    V Halogenated Paraffins
Source: TRI, 1994.
                                       6-24

-------
                                     Figure 6-11
     Facilities in the Vicinity of the Hahnville, Louisiana (HALA) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
                    OTA
T Garyville/LA
  Monitoring Station
     StfJohn
    the Baptist
      Parish

                                                               N.
              &hnville, LA
              Monitoring Station
                                 o
                                                                         Kenner
                                               Jefferson
                                                Parish
   Key:
  * Monitoring station
  — County boundary
  D City
 - •  US Interstate
                       St Charles Parish
                                                     OT
   Release Categories:
O Aromatics          ^ Olefins            O Carbonyk
• Chlorinated Aromatics A Chlorinated Olefins   T Halogenated Paraffins
Source: TRI, 1994.
                                      6-25

-------
                                     Figure 6-12
      Facilities in the Vicinity of the Port Neches, Texas (PNTX) Monitoring Station
                    Reporting VOC or Carbonyl Air Releases to TRI
                                     10-mile radius
                                         Orange County
                       OTA
                  OY
                 Jefferson County
   Key:
  * Monitoring station
  — County boundary
  DCity
 - •  US Interstate
                                                                             Area of
                                                                             Detail
                                                                              O+A
                 -Monitorm&~Station
                 OA     OA
                                                                          Gulfof
                                                                          Mexico
   Release Categories:
O Aromatics          ^ Olefins
• Chlorinated Aromatics A Chlorinated Olefins
OCarbonyls
T Halogenated Paraffins
Source: TRI, 1994.
                                       6-26

-------
                                     Figure 6-13
      Facilities in the Vicinity of the Rutland, Vermont (RUVT) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
                                    10-mile radius
                                  Rutland County
                                                                       Windsor
                                                                        County
  Key:
 * Monitoring station %x
 — County boundary
 D City
- • US Interstate
                                          Rutland, VT
                                        ^"Monitoring Station
                                             o
  Release Categories:
O Aromatics          ^ Olefins             O Carbonyls
• Chlorinated Aromatics A Chlorinated Olefins    V Halogenated Paraffins
                                       6-27

-------
                                    Figure 6-14
      Facilities in the Vicinity of the Underbill, Vermont (UNVT) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
                      Franklin County
                                                 LamoiUe County
Chittenden County
                                          Underbill, VT
                                        *~Monitonng Station
   Key:
  *  Monitoring station  s~
 —  County boundary
  D  City
 -   US Interstate
                                         Washington County
   Release Categones
       O Aromatics          ฃ Olefins            O Carbonyls
       • Chlorinated Aromatics A Chlorinated Olefins   V Halogenated Paraffins
Source: TRI, 1994.
                                      6-28

-------
                                  Figure 6-15
   Facilities in the Vicinity of the Vancouver, Washington (VAWA) Monitoring Station
                  Reporting VOC or Carbonyl Air Releases to TRI
  AreaofDetail
Washington \
  County
                                  \10-mileradius

                                 "\	
                                    Key:
                                    * Monitoring station
                                   — County boundary
                                    D City
                                  - •  US Interstate
                                              Clark County
                                       Vancouver, WA
                                       Monitoring Site
                                        o
                                                                         TA
                                             184/US 30
                              <>O  \ [fortland, OR     ,'
                                             -
            ijjeaverton
                             •~-yv-^
                                            V
                                 x
                               x
                              *
                                      Greshamrr]

                             Multnomah County
   Release Categories:
O Aromatics          & Olefins           O Carbonyls
• Chlorinated Aromatics A Chlorinated Olefins   T Halogenated Paraffins
Source: TRI, 1994.
                                    6-29

-------
                                    Figure 6-16
      Facilities in the Vicinity of the Winooski, Vermont (WIVT) Monitoring Station
                   Reporting VOC or Carbonyl Air Releases to TRI
                  Grand Isle
                  County, VT
  Clinton
County, NY,
                                                              ?ranklin County, VT
          10-mile radius
                                          \
                                          X
                                        ;\IWinooski,VT
             Burlington, VT    _=      *~jMonitoring Station
             ป/r   M.  1   cu *:  ^Burlington           <>b
             Monitoring Station        /
     Essex'County, NY
   Key:
  * Monitoring station
  — County boundary
  DCity        \
 - •  US Interstate   \
                                                     IS9
                                                                \
                            Chittenden County, VT
                                                                      \
   Release Categories:
O Aromatics          ^ Olefins             O Carbonyls
• Chlorinated Aromatics  A Chlorinated Olefins    Y Halogenated Paraffins
Source: TRI, 1994.
                                      6-30

-------
                                      Figure 6-17
             Comparison of BTEX Concentration Profile to Roadside Study
    Roadside
B2LA
BB/VA          BRTX
 Monitoring Location
BUVT
CANJ
B Benzene/Bhylbenzene    • Toluene/Bhylbenzene   om,p-Xylene/Bthylbenzene    @ o-Xylene/Bhylbenzene

-------
                                                     Figure 6-17 (Continued)
                                   Comparison of BTEX Concentration Profile to Roadside Study
Os
                       Roadside
DATN
BTX           GALA
 Monitoring Location
GATX
HA LA
                         I Benzene/Bhylbenzene  •Toluene/Bhylbenzene  Qm.p-Xylene/Bthylbenzene  B o-Xylene/Bhylbenzene

-------
                                                     Figure 6-17 (Continued)
                                  Comparison of BTEX Concentration Profile to Roadside Study
Oi
            8
            8
                2
                       Roadside
PNTX
RUVT            UNVT
 Monitoring Location
VAWA
WIVT
                         iBenzene/Bhylbenzene  • Toluene/BhyIbenzene Bm,p-Xylene/Bthy(benzene E o-Xylene/Bhylbenzene

-------
                                                          Figure 6-18
                                    Ratios of Acetaldehyde and Formaldehyde to Ethylbenzene
a\
4
               120
            O
                40
                20
                        B2LA
BEWA
BRTX         BUVT         CANJ

       Monitoring Location
BปTX
GALA
                               I Formaldehyde/Bhylbenzene
                                         I Acetaldehyde/Bhy [benzene

-------
                                    Figure 6-18 (Continued)
                    Ratios of Acetaldehyde and Formaldehyde to Ethylbenzene
40
20
        GATX
HALA
PNTX         RUVT         UNVT

       Monitoring Location
VAWA
WIVT
                I Formaldehyde/Bhylbenzene
                                         I Acetaldehyde/Bhy Ibenzene

-------
                                              Table 6-1
                  Ranking of Monitoring Stations by Overall Levels of VOCs
Rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Monitoring
Location
Bcllineham, WA
V '
Camden,NJ
El Paso, TX
Galveston,TX
Davidson, TN
Baton Rouge, LA
PortNeches,TX
Vancouver, WA
Brownsville, TX
Burlington, VT
Garyville.LA
Hahnvillc, LA
Rutland, VT
Winooski,VT
Underhill,VT
VOC Industrial Emissions
within a 10-Mile Radius of the
Monitoring Location
Pounds in
1994
280,000
1,200,000
110,000
1,100,000
1,900,000
1,500,000
5,600,000
1,200,000
29,000
16,000
550,000
2,200,000
81,000
16,000
0
Emissions
Rank
9
5
10
7
3
4
1
6
12
13 (tie)
8
2
11
13 (tie)
15
Population Residing within a
10-Mile Radius of the
Monitoring Location

Persons
91,274
2,021,082
410,475
103,167
426,257
336,577
146,467
587,395
125,547
103,912
56,800
107,033
38,969
109,541
18,997
Population
Rank
12
1
4
11
3
5
6
2
7
10
13
9
14
8
15
Notes:   Because the Brattleboro, Vermont (BRVT) station only collected four valid samples, the station was not
        included in the VOC ranking.

        Industrial emissions were calculated from emissions data for TRI reporting year 1994 (TRI, 1994). The
        emissions reported in the table are the total air releases of the VOCs that were considered in this monitoring
        program. The following five VOCs were considered in the monitoring program, but were not reportable to
        TRI in 1994:  acetylene, bromochloromethane, bromodichloromethane, dibromochloromethane,
        1,1-dichloroetbane, andn-octane. All emissions totals are rounded to two significant digits.

        Population was determined from 1990 U.S. Census data (USDOC, 1993). Because the BRVT monitoring
        station was not included in this table, while the DATN station was not included in Table 6-2, the assigned
        population ranks in the two tables differ slightly.
                                                6-36

-------
                                               Table 6-2
                 Ranking of Monitoring Stations by Overall Level of Carbonyls
Rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Monitoring
Location
Garyville,LA
El Paso, TX
Hahnvffle,LA
Bellingham, WA
o ^
Brownsville, TX
Baton Rouge, LA
Rutland, VT
Burlington, VT
PortNeches,TX
Galveston,TX
Vancouver, WA
Brattleboro, VT
Winooski,VT
Camden,NJ
Underbill, VT
Carbonyl Industrial Emissions
within a 10-Mile Radius of the
Monitoring Location
Pounds in
1994
0
0
52,000
44,000
0
29,300
0
0
15,000
23,000
130,000
0
0
2,400
0
Emissions
Rank
8 (tie)
8 (tie)
2
3
8 (tie)
4
8 (tie)
8 (tie)
6
5
1
8 (tie)
8 (tie)
7
8 (tie)
Population Residing within
a 10-Mile Radius of the
Monitoring Location
Persons
56,800
410,475
107,033
91,274
125,547
336,577
38,969
103,912
146,467
103,167
587,395
27,862
109,541
2,021,082
18,997
Population
Rank
12
3
8
11
6
4
13
9
5
10
2
14
7
1
15
Notes:  The Davidson, Tennessee (DATN) station did not sample for caibonyls and was not included in the carbonyl
        ranking.

        Industrial emissions were calculated from emissions data for TRI reporting year 1994 (TRI, 1994). The
        emissions reported in the table are the total air releases of only those carbonyls that were considered in this
        monitoring program. The following carbonyls were considered in the monitoring program, but were not
        reportable to TRI in 1994:  acetone, benzaldehyde, crotonaldehyde, 2,5-dimethylbenzaldehyde, hexanaldehyde,
        isovaleraldehyde, tolualdehydes, and valeraldehyde. All emissions totals are rounded to two significant digits.

        Population was determined from 1990 U.S. Census data (USDOC, 1993). Because the DATN monitoring
        station was not included in this table, while the BRVT station was not included in Table 6-1, the assigned
        population ranks in the two tables differ slightly.
                                                 6-37

-------
                                                      Table 6-3
                Comparison of Geometric Mean Concentrations of Selected VOCs With Total Air Releases
                      Reported by Facilities Within a 10-Mile Radius of UATMP Monitoring Stations
Benzene
Station
Code
EPTX
BEWA
GATX
BUVT
VAWA
B2LA
CANJ
RUVT
BRTX
DATN
HALA
PNTX
GALA
WIVT
UNVT
Geometric Mean
Concentration
(ppbv)
0.99
0.85
0.63
0.61
0.53
0.52
0.46
0.44
0.43
0.41
0.40
0.37
0.36
0.24
0.12
Total Air
Releases
(Ibs)
0
0
44,307
0
4,148
172,055
118,702
0
0
0
124,241
657,379
5,383
0
0
Carbon Tetrachloride
Station
Code
GALA
PNTX
GATX
HALA
VAWA
CANJ
BEWA
DATN
WIVT
B2LA
BRTX
EPTX
BUVT
RUVT
UNVT
Geometric Mean
Concentration
(ppbv)
0.10
0.09
0.09
0.09
0.09
0.08
0.08
0.08
0.08
0.07
0.07
0.07
0.07
6.07
0.07
Total Air
Releases
(Ibs)
0
146,233
0
0
0
10
0
0
0
2,133
0
0
0
0
0
Chloromethane
Station
Code
GALA
VAWA
BRTX
GATX
HALA
PNTX
CANJ
EPTX
DATN
B2LA
BUVT
BEWA
UNVT
RUVT
WIVT
Geometric Mean
Concentration
(ppbv)
1.03
0.86
0.84
0.79
0.77
0.68
0.68
0.65
0.63
0.62
0.60
0.60
0.49
0.46
0.45
Total Air
Releases
(Ibs)
6
0
0
0
138,775
6,250
99,870
0
0
41,095
0
0
0
0
0
Source of emissions data: TRI, 1994
Note:   The data in the table are displayed in order of decreasing geometric mean concentration.

-------
                                                         Table 6-3 (Continued)
                        Comparison of Geometric Mean Concentrations of Selected VOCs With Total Air Releases
                             Reported by Facilities Within a 10-Mile Radius of UATMP Monitoring Stations
Ethylbenzene
Station
Code
BEWA
BRTX
EPTX
BUVT
GATX
DATN
RUVT
B2LA
PNTX
HALA
CANJ
GALA
VAWA
WIVT
UNVT
Geometric Mean
Concentration
(ppbv)
0.62
0.37
0.35
0.24
0.22
0.19
0.18
0.15
0.14
0.14
0.14
0.13
0.12
0.11
0.08
Total Air
Releases
(H>s)
0
0
0
5
48,441
0
0
52,424
152,641
16,200
70,116
4,032
28,594
5
0
Propylene
Station
Code
GATX
EPTX
CANJ
B2LA
PNTX
BUVT
BEWA
VAWA
HALA
GALA
DATN
RUVT
BRTX
WIVT
UNVT
Geometric Mean
Concentration
(ppbv)
2.09
1.57
1.24
1.20
1.19
1.19
1.00
0.76
0.67
0.66
0.59
0.56
0.27
0.18
0.06
Total Air
Releases
(Ibs)
544,340
0
136,261
382,055
2,298,278
0
0
0
1,059,842
2,770
0
0
0
0
0
Toluene
Station
Code
BEWA
EPTX
BUVT
VAWA
DATN
GATX
RUVT
CANJ
HALA
B2LA
GALA
PNTX
WIVT
UNVT
BRTX
Geometric Mean
Concentration
(ppbv)
3.18
2.11
1.40
1.27
1.23
1.13
1.08
1.04
0.92
0.89
0.81
0.52
0.47
0.20
0.14
Total Air
Releases
(Ibs)
0
25,063
14,965
478,585
4,758
109,144
80,820
412,743
145,926
342,156
46,500
1,216,349
14,965
0
9,100
a\
w
so
        Source of emissions data: TRI, 1994
        Note:  The data in the table are displayed in order of decreasing geometric mean concentration.

-------
                                                  Table 6-3 (Continued)
                 Comparison of Geometric Mean Concentrations of Selected VOCs With Total Air Releases
                      Reported by Facilities Within a 10-Mile Radius of UATMP Monitoring Stations
1,1,1 -Trichloroethane
Station
Code
WIVT
BEWA
DATN
GALA
EPTX
CANJ
B2LA
BUVT
BRTX
PNTX
HALA
UNVT
RUVT
VAWA
GATX
Geometric Mean
Concentration
(ppbv)
0.31
0.30
0.28
0.22
0.21
0.21
0.21
0.18
0.18
0.18
0.17
0.17
0.16
0.15
0.15
Total Air
Releases
(Ibs)
1,292
0
198,705
213
0
8,961
16,100
1,292
0
19,495
0
0
0
57,000
0
Xylenes (Total)
Station
Code
BEWA
EPTX
BUVT
DATN
RUVT
GATX
CANJ
B2LA
VAWA
HALA
GALA
PNTX
WIVT
BRTX
UNVT
Geometric Mean
Concentration
(ppbv)
4.28
1.97
1.39
1.13
0.99
0.95
0.92
0.92
0.84
0.76
0.65
0.41
0.38
0.36
0.17
Total Air
Releases
(Ibs)
0
0
5
94,321
0
243,830
350,335
152,136
332,274
64,012
7,500
428,537
5
13,500
0
Source of emissions data:  TRI, 1994
Note:   The data in the table are displayed in order of decreasing geometric mean concentration.
Note for total xylenes: Emissions and concentrations are both the totals of the o-, m-t andp-xylene isomers.

-------
                                                                  Table 6-4
                        Comparison of Geometric Mean Concentrations of Selected Carbonyls With Total Air Releases
                               Reported by Facilities Within a 10-Mile Radius of UATMP Monitoring Stations
Acetaldehyde
Station
Code
BRTX
BEWA
EPTX
HALA
GALA
RUVT
B2LA
BUVT
PNTX
WIVT
BRVT
GATX
VAWA
UNVT
CANJ
Geometric Mean
Concentration
(ppbv)
9.69
5.91
5.05
4.21
4.00
2.99
2.95
2.69
2.58
2.08
1.92
1.58
0.77
0.54
0.25
Total Air
Releases
(Ibs)
0
39,000
0
23,338
0
0
0
0
773
0
0
23,334
0
0
0
Butyr/Isobutyraldehyde
Station
Code
GALA
HALA
BEWA
EPTX
RUVT
BUVT
GATX
BRTX
B2LA
PNTX
BRVT
WIVT
VAWA
UNVT
CANJ
Geometric Mean
Concentration
(ppbv)
0.95
0.67
0.56
0.52
0.30
0.26
0.22
0.19
0.17
0.15
0.10
0.08
0.05
0.02
0.02
Total Air
Releases
(Ibs)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Formaldehyde
Station
Code
HALA
B2LA
EPTX
GALA
BEWA
BUVT
RUVT
GATX
WIVT
BRTX
PNTX
BRVT
VAWA
CANJ
UNVT
Geometric Mean
Concentration
(ppbv)
14.38
12.91
10.50
10.21
9.05
5.98
5.93
4.29
4.11
2.91
2.22
2.05
0.92
0.92
0.79
Total Air
Releases
Obs)
19,522
29,300
0
0
5,260
0
416
0
0
0
15,000
0
129,300
2,346
0
ON
         Source of emissions data: TRI, 1994
         Note:   The data in the table are displayed in order of decreasing geometric mean concentration.
         Note:   No releases of butyraldehyde or isobutyraldehyde were reported by any facility located within 10 miles of the 1995 UATMP monitoring stations.

-------
                                                        Table 6-4 (Continued)
                     Comparison of Geometric Mean Concentrations of Selected Carbonyls With Total Air Releases
                            Reported by Facilities Within a 10-Mile Radius of UATMP Monitoring Stations
h
Propionaldehyde
Station
Code
GALA
HALA
EPTX
BEWA
B2LA
GATX
RUVT
BUVT
BRVT
PNTX
BRTX
WIVT
VAWA
UNVT
CANJ
Geometric Mean
Concentration
(ppbv)
0.89
0.76
0.59
0.43
0.37
0.37
0.32
0.25
0.16
0.14
0.12
0.10
0.06
0.02
0.02
Total Air
Releases
(Ibs)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
      Source of emissions data: TRI, 1994
      Note:   The data in the table are displayed in order of decreasing geometric mean concentration.
      Note:   No releases of propionaldehyde were reported by any facility located within 10 miles of the 1995 UATMP monitoring stations.

-------
Section 7

-------
 7.0    Temporal Variations
       This section describes the extent to which ambient air concentrations measured during
 the 1995 UATMP vary over time.  Temporal variations in ambient air concentrations result  .
 primarily from changes in industrial and motor vehicle emissions and fluctuations in local
 meteorological conditions.  With these factors considered in proper context, temporal
 variations can provide additional insight into the origin of chemicals found in urban air
 pollution. For reference, Section 7.1 discusses the different time frames over which temporal
 variations typically are analyzed in ambient air monitoring studies. Section 7.2 then identifies
 noteworthy seasonal trends in the VOC and carbonyl concentrations measured during the 1995
 UATMP, and Section 7.3 evaluates annual trends in concentrations by comparing UATMP
 results to those of previous program years.  Despite the extensive monitoring data collected in
 the current program  year, the data are not sufficient for interpreting long term air pollution
 trends which apply to all urban environments.

       7.1    Time  Frames for Temporal Variations
       Previous studies measuring urban air pollution have identified diurnal,  seasonal, and
 annual variations in ambient air concentrations. To assess diurnal variations (changes over the
 course of a day), monitoring programs must collect air samples at different times of the day.
 Because only  24-hour integrated samples were collected during the 1995 UATMP, the air
 monitoring data in this study are not sufficient for evaluating diurnal variations.  The
 monitoring data are,  however, sufficient for evaluating seasonal and annual variations.
 Sections 7.2 and 7.3 present the seasonal and annual trends observed in the ambient air
 concentrations measured during the 1995 UATMP.  These sections relate the observed
 seasonal and annual trends to changes in industrial emissions, motor vehicle emissions, and
photochemical reactions.

       7.2    Seasonal Variations
       This section describes the seasonal variations in ambient air concentrations measured at
the 16 monitoring locations participating in the 1995 UATMP.  For reference, subsection
                                        7-1

-------
 7.2. 1 first presents the significance of seasonal variations observed in urban air pollution.

 Subsections 7.2.2 and 7.2.3 then summarize how ambient air concentrations of VOCs and

 carbonyls, respectively, changed from season to season over the course of this monitoring
 program.
                    Significance of Seasonal Variations
       As noted in Section 5, industrial emissions, motor vehicle emissions, and

photochemical reactions largely determine the magnitude and composition of air pollution in

urban centers. The extent to which these factors change, or fail to change, from season to

season also should determine seasonal variations in ambient air concentrations.  The following
discussion summarizes the seasonal changes in these factors:
•      Industrial emissions.  The magnitude of industrial emissions in urban locations depends
       on the nature and production schedules of local industrial facilities.  Because many
       industrial facilities, particularly large chemical manufacturers, have continuous process
       operations, total industrial emissions in some urban environments may not change
       significantly from season to season. Nonetheless, many of these facilities emit
       chemicals primarily through evaporative losses, and these emissions increase with
       increasing ambient temperature.  Accordingly, such facilities typically have increased
       emissions during the wanner months of the year.  For a given urban environment,
       therefore, seasonality in industrial emissions primarily depends on the profile of the
       local industrial facilities.

•      Motor vehicle emissions. In urban environments, motor vehicle emissions generally are
       greatest during the rush hour traffic of morning and evening commutes.  Because the
       volume of traffic at these times does not change significantly over the course of a year,
       total emissions from motor vehicles may not vary significantly from month to month.

•      Photochemical reactions.  By definition, photochemical reactions are initiated by light.
       The rate constants of these reactions generally increase with ambient temperature.
       Accordingly, photochemical reactivity is at a peak during the summer months, when
       ambient temperatures and the number of daylight hours  also are at a peak.


       Of the factors listed above, photochemical reactions exhibit the greatest variations from

season to season, and chemicals generated by these reactions can be expected to have highest
                                          7-2

-------
 concentrations during warmer months. Accordingly, significant seasonal variations in ambient
 air concentrations can confirm a hypothesis presented in Section 6:  ambient air concentrations
 of carbonyls result primarily from photochemical reactions.

       It should be noted that some factors not considered above also affect seasonal variations
 in ambient air concentrations. Most importantly, changes in local meteorological conditions,
 particularly prevailing wind patterns, contribute to seasonal trends in urban air pollution.
 Although atmospheric dispersion models can account for these seasonal changes in local
 meteorology, site-specific dispersion modeling assessments are beyond the scope of this
 project.  Subsections 7.1.2 and 7.1.3 present the seasonal variations for VOCs and carbonyls,
 respectively. These subsections refer to the discussion above to interpret the observed seasonal
 trends, or lack thereof, in ambient air concentrations.

       7.2.2 Seasonal Variations in VOC Concentrations
       Seasonal variations in VOC concentrations were evaluated by comparing "season-
 average" geometric mean concentrations between spring, summer, fall, and winter.  The
 season-average geometric mean concentrations reflect the magnitude and composition of air
 pollution averaged over an entire season.  If a monitoring station did not collect enough valid
 samples to estimate any of the four season-average concentrations accurately, the monitoring
 station was excluded from this analysis.  As indicated both in the invalid sample results in
 Appendix B and in the completeness schedule shown in Table 2-4, the Brownsville, Texas
 (BRTX), Brattleboro, Vermont (BRVT), and Vancouver, Washington (VAWA) monitoring
 stations had at least one season with too few valid results to  estimate season-average
 concentrations reliably. Results from these monitoring stations were not considered further.

       For the remaining  13 monitoring stations, season-average geometric mean
 concentrations were calculated for every chemical. As with the geometric mean calculations
discussed in Section 3.1.3, non-detect results in the season-average calculations were assigned
a concentration of one-half the detection limit.  For chemicals not detected in a majority of the
                                        7-3

-------
 samples, however, the season-average concentrations calculated in this analysis generally equal
 one-half the method detection limit, even though the actual season-average concentration likely
 has a different value. To remove the impact of non-detects on seasonal variations, the season-
 average geometric mean concentrations were only compared for chemicals detected in over 75
 percent of the ambient air samples collected from at least ten of the 1995 UATMP monitoring
 stations.  Section 6.2.1 lists the 11 VOCs meeting this criterion.

       Table 7-1 summarizes the results of the season-average concentrations for these VOCs.
 For each VOC, the table indicates the number of monitoring stations (out of the 13 considered
 in this analysis) that observed highest season-average concentrations in spring, summer, rail,
 and winter.  As shown in the table, the VOCs most prevalent in urban ambient air generally
 have highest season-average concentrations for spring and summer, with lower season-averages
 observed for fall and winter. These results indicate that ambient concentrations of VOCs may
 exhibit seasonal trends, but the seasonal differences are not particularly pronounced. Because
 the carbonyl concentrations have a notably different seasonal profile, the significance of the
 observed seasonal variations is discussed at the end of the next subsection.

       7.2.3  Seasonal Variations in Carbonyl Concentrations
       Seasonal trends in ambient air concentrations of carbonyls were evaluated using the
 same methodology presented in the previous subsection. Of the 16 carbonyl compounds
 considered in the 1995 UATMP, nine were detected in at least 75 percent of the samples
collected from at least 10 of the monitoring stations.  Table 7-2 lists these chemicals and the
number of monitoring stations that observed highest season-average concentrations in spring,
summer,  fall, and winter.

      Comparing Table 7-2 to 7-1, season-average concentrations of carbonyls show a much
stronger tendency to peak in the summer than do those of VOCs. The significantly different
seasonal trends between carbonyls and VOCs indicates that ambient air concentrations of these
chemical categories likely originate from different sources. This observation was also suggested
                                         7-4

-------
 in Section 6.1, which indicated that ambient air concentrations of VOCs, as a whole, correlated
 both with industrial emissions and population density, while such correlations did not exist for
 caibonyls, as a whole.  The absence of correlations between ambient concentrations of carbonyls
 and common emission sources, combined with the presence of peak concentrations of carbonyls
 during summer months, strongly suggest that carbonyls in urban ambient air are formed primarily
 as products of photochemical reactions.

       7.3    Annual Trends
       Of the 16 monitoring locations participating in the 1995 UATMP, only the Baton Rouge,
 Louisiana (B2LA), Camden, New Jersey (CANJ), and Port Neches, Texas (PNTX) locations
 participated in previous UATMP monitoring efforts.  As shown in Figure 5-2, ambient air
 monitoring performed at PNTX in previous UATMP program years occurred at a location
 roughly two miles away from the location of the current PNTX monitoring station. Similarly,
 ambient air monitoring performed at B2LA in previous UATMP program years occurred at a
 location nearly four miles from the location of the current B2LA monitoring station.  The analyses
 in Section 5.3 indicate that ambient  air monitoring at different locations within the same urban
 center may exhibit significant variations between concentrations of specific chemicals. Therefore,
 this section does not consider annual trends between ambient air concentrations measured at the
 different locations in Port Neches or Baton Rouge.

       The location of the monitoring station in Camden, New Jersey, however, has not changed
 since it was installed in 1989. Valid monitoring results for this station have been reported for
UATMP program years 1989,1990,1991, and  1994. With the 1995 monitoring data, therefore,
five years (not consecutive) of ambient air monitoring results are available to evaluate annual air
pollution trends for the CANJ urban area. Before analyzing the results for this monitoring station,
subsection 7.3.1 first discusses the significance of annual trends in ambient air monitoring data.
Subsection 7.3.2 then summarizes the annual trends observed for the most prevalent VOCs in the
ambient air in Camden, New Jersey, and subsection 7.3.3 discusses annual trends for carbonyls.
                                         7-5

-------
 Note:  In UATMP program years 1992 and 1993, ambient air monitoring -wasperformed by
       another contractor. The monitoring results for CANJ in these program years are
       available from the Air Quality Subsystem (AQS) of AIRS.


       7.3.1  Significance of Annual Variations

       Annual variations in ambient air concentrations reflect changes over the long term in
 industrial emissions, motor vehicle emissions, and photochemical reactions.  The following
 discussion summarizes how these factors may change from year to yean


 •      Industrial emissions.  In urban environments, changes in industrial emissions over the
       long term may reflect the changing profile of industrial facilities or the effectiveness of
       pollution prevention initiatives. The extent of these changes clearly varies from one
       urban location to another.

 •      Motor vehicle emissions.  Improved design of motor vehicles and, more recently,
       vehicle fuels have helped minimize emissions from mobile sources in previous years.
       These design efforts ultimately should lead to a decrease in the total motor vehicle
       emissions in large metropolitan areas. In a particular ambient  air monitoring location,
       however, motor vehicle emissions may increase due to increased population or
       changing traffic patterns, despite the improved vehicle design.

 •      Photochemical reactions. As noted earlier, ambient temperature and the number of
       daylight hours determine the extent of photochemical reactions in urban environments.
       Clearly, these factors do not change significantly over the long term.  Nonetheless, the
       composition and magnitude of urban air pollution may change  from year to year, and
       these changes can affect the photochemical reactivity in urban atmospheres, despite the
       consistency of patterns in ambient temperature and daylight hours.

       As indicated in the previous discussion,  factors specific to a particular urban
environment should determine the magnitude of annual variations in urban air pollution.
Therefore, analyses of annual variations in ambient air concentrations should be interpreted
with respect to the local changes in industrial and motor  vehicle emissions, and the extent to
which these changes affect photochemical reactions.  Subsection 7.3.2 presents the annual
variations in VOC concentrations observed at the Camden, New Jersey (CANJ) monitoring
                                          7-6

-------
 station.  Subsection 7.3.3 discusses why annual variations cannot be evaluated adequately for
 ambient concentrations of carbonyls.

       7.3.2  Annual Variations in VOC Concentrations
       As shown throughout this analysis, the treatment of non-detects significantly affects the
 analysis of ambient air monitoring data. Different methods were used to treat non-detects in
 the previous UATMP summary reports. For example, some studies completely ignored non-
 detects, while others assigned non-detects proxy concentrations.  To compare historical data on
 a uniform basis, this subsection presents annual trends only for those chemicals detected in all
 of the samples collected at CAN! since the beginning of the UATMP monitoring.  These
 chemicals are toluene, benzene, and ethylbenzene.

       As shown in Table 7-3, the ambient concentrations of these aromatic hydrocarbons
 have steadily decreased in the CANT ambient air over the last six years.  These annual trends
 may result from a variety of influences. Section 6.3, however, presents strong evidence that
 ambient concentrations of aromatic hydrocarbons largely result from motor vehicle emissions.
 If this is the case, the downward trend in ambient concentrations of aromatic hydrocarbons at
 CANT may result from changes in traffic patterns, improvements in motor vehicle design, or
 improvements in motor vehicle fuels. Detailed review of emission sources particular to the
 Camden, New Jersey area can provide additional insight to this downward trend.

Note:  The annual trends discussed in this section are specific to the Camden, New Jersey
       monitoring station.  Trends in other regions of the country, possibly in other areas near
       Camden,  may not necessarily follow the trends listed here.

       7.3.3   Annual Variations in Carbonyl Concentrations
       As noted in Section 6.1.3, the CANJ monitoring station ranked very high for VOC
concentrations, but very low for carbonyls. The discussion in that subsection attributed the
discrepancy in the ranks to a gap in sampling of carbonyls during the summer months. As
                                         7-7

-------
shown in Section 7.2.3, ambient air concentrations of carbonyls peaked in the summer for
most of the monitoring stations considered in this program. The absence of carbonyl data for
two months at CANJ likely biases the geometric mean concentrations for carbonyls at that site
to lower values.  Therefore, comparison to data collected in previous years are not necessarily
valid (i.e., differences in concentrations may result from the air being "cleaner" or "dirtier," or
differences in concentrations may result from the absence of sampling data for two months in
the 1995 program).  Accordingly, annual variations for the carbonyl data are not presented.

      7.4    Summary
      Like the spatial variations analyzed in Section 6, the temporal variations evaluated in
this section offer insight into the origin of chemicals detected in urban air pollution.  As shown
in Section 7.2, ambient air concentrations of VOCs tend to be higher in spring,  summer, and
fall, and significantly lower in winter.  The seasonal trends of carbonyls are similar to those
for VOCs, but ambient air concentrations of carbonyls measured in the 1995 UATMP have a
much stronger tendency to peak during the summer months. This difference suggests that, to a
certain extent, VOCs and carbonyls in urban air pollution may originate from different
sources. In addition to the evidence presented in Section 6, the strong tendency for carbonyls
to peak during summer months indicates that the primary source of these chemicals in the
atmosphere may be photochemical reactions.  .

      The 1995  UATMP monitoring data are only sufficient for evaluating annual trends of
VOC concentrations at the Camden, New Jersey (CANJ) monitoring station.  As shown in
Section 7.3, the most prevalent VOCs in the ambient air measured at CANJ have decreased
steadily  over the last seven years.  A variety of factors might explain this downward trend, but
the trend likely results from changes in motor vehicle emissions over the long term.
                                        7-8

-------
                                     Table 7-1
                Number of Monitoring Stations at which Selected VOCs
              Had Highest Seasonal-Average Ambient Air Concentrations
Chemical
Acetylene
Benzene
Carbon tetrachloride
Chloromethane

Ethylbenzene
n-Octane
Propylene
Toluene
1,1,1-Trichloroethane
ซ,/>-Xylene
o-Xylene
TOTALS
Number of Monitoring Stations with Seasonal-Average
Ambient Air Concentration in Listed Season *
Spring
2
5
6
7
7
6
3
6
3
7
4
56 (39%)
Summer
2
3
5
4
4
3
3
5
8
3
3
43 (30%)
Fall
4
1
1
1
2
2
5
2
1
3
5
27 (19%)
Winter
5
4
1
1
0
2
2
0
1
0
1
17(12%)
* Results from 13 monitoring stations were considered in this analysis. The Brattleboro, Vermont
(BRVT), Vancouver, Washington (VAWA), and Baton Rouge, Louisiana (B2LA) monitoring
stations had incomplete data for at least one season and were therefore not included in this table.
                                        7-9

-------
                                      Table 7-2
              Number of Monitoring Stations at which Selected Carbonyls
              Had Highest Seasonal-Average Ambient Air Concentrations
Chemical
Acetaldehvde

Acetone
Butyr/Isobutyraldehyde
Crotonaldehyde
Formaldehyde
Hexanaldfhyde

Propionaldehyde
TOTALS
Number of Monitoring Stations with Seasonal-Average
Ambient Air Concentration in Listed Season '
Spring
2
1
3
1
3
6
4
20 (19%)
Summer
12
8
7
12
5
8
3
55 (52%)
Fall
1
2
2
2
6
0
6
19 (18%)
Winter
0
4
3
0
1
1
2
11 (11%)
* Results from 14 monitoring stations were considered in this analysis.  The Brattleboro, Vermont
(BRVT) and Davidson, Tennessee (DATN) monitoring stations had incomplete data for at least
one season and were therefore not included in this table.
                                        7-10

-------
                                     Table 7-3
             Annual Trends in Geometric Mean Concentrations Observed at
                 the Camden, New Jersey (CANJ) Monitoring Station
UATMP Program
Year
1989
1990
1991
1992
1993
1994
1995
Geometric Mean Concentration (ppbv)
Benzene
1.46
1.19
0.79
__ •
___ a
0.47
0.46
Ethylbenzene
0.36
0.22
0.32
__ •
	 a
0.24
0.14
Toluene
3.25
2.20
2.45
___ a
a
1.69
1.04
 * In UATMP program years 1992 and 1993, ambient air monitoring was performed by a different
contractor. For reference, the monitoring results for CANJ for these years are available from the
Air Quality Subsystem (AQS) of AIRS.
                                       7-11

-------
Section 8

-------
 8.0    Precision and Accuracy
        This section evaluates the precision and accuracy of the sampling and analytical methods
 used to measure concentrations during the 1995 UATMP. Precision and accuracy indicate the
 reliability of experimental measurements, and both parameters should be considered when
 determining the validity of air monitoring results. Sections 8.1 and 8.2 apply methods commonly
 used to estimate precision and accuracy to the monitoring methods used in this program. In
 Section 8.3, the precision and accuracy of the UATMP ambient air monitoring methods are
 compared to the UATMP data quality objectives (USEPA, 1988). As shown in this section, the
 most precise and accurate measurements were observed for those chemicals consistently found at
 levels exceeding their respective detection limits.

        8.1    Precision
       Precision refers to the mutual agreement between independent measurements performed
 according to identical protocols and procedures. The magnitude of precision is determined by
 random errors inherent in the associated measurements.  Random errors cause measurements to
 be sometimes higher and sometimes lower than the most commonly measured value. Precision
 does not account for systematic errors, which determine the extent to which independent
 measurements accurately represent the "actual" or "true" values (see Section 8.2). In terms of air
 monitoring, precision measures the variability observed upon repeated measurements of ambient
 concentrations.  The following subsections evaluate the precision of both the sampling and
 analytical methods used in the 1995 UATMP.

       8.1.1  Analytical Precision
       Analytical precision measures random errors and variability associated with laboratory
 analysis of environmental monitoring samples. These random errors may result from various
 factors, including random "noise" inherent to analytical instruments and imprecise measurements
made by laboratory analysts.  Monitoring programs typically evaluate analytical precision by
comparing concentrations measured during replicate analysis of ambient air samples. No matter
how samples are collected, the difference between concentrations measured in replicate analyses
                                          8-1

-------
                    f f •.    ..    ~-\ • f    •' •>   < Sf ••"•..
 Treatment of Hen-detects m E^imatittg I^v&aatt
indicates the precision of laboratory analytical methods. Subsection 8.1.2 considers methods for
evaluating precision of the field sampling methods.

       Several parameters can be used to estimate method precision from the results of replicate
analyses.  This report uses the relative percent difference (RPD) to estimate the precision of the
UATMP monitoring methods. In the
context of analytical precision, relative
percent difference is calculated as the
absolute value of the difference
between the two concentrations
measured in a replicate analysis
divided by the average of the two
concentrations. Therefore, the RPD
expresses the variability of laboratory
measurements as a percentage of the
measured concentrations. More
precise analytical methods have lower
RPDs than less precise methods.
 result indicates ambient air concentrations are at levels;
 below the sensitivity of the sanding and analytical
 devices. In the summary tables and statistical ..  '
 analyses, non-detect observations were assigned a
 "best estimate1' concentration of one-half the -
 detection Emit. AdpptingiMs approachto precision..
 calculations would result in near perfect predsion ;
 (i.e., identical results) for chemicals aot detected in a ซ
 majority of the samples, even though the sampling and
 analyticalinediod guidelines Abdicate that      -  -"  ^
 concentrations measured at levels near tl^ detect  '
.Kmithavfcthe greatest vajia1afcjy|a$iฃAj 1984s;  -
 1984$. f           ,   "     *   '      ^ ' ~ -
     %            \          "*   '      ' * f        >•.
 Therefore, assigrangnc^iHietects a fixed arbitrary :   ::
 value uitro^uces bias into the pr^^^                '
 To avoid this problem, 01% those chemicals W which
 concentrations were -detected in both xeplkates are-,-
 considered when evaluating analytical precision {see
 Section 8.1.1); similarly, only those chemicals for
 which concentrations were detected in the four
 replicates in a duplicate sample are considered when
 evaluating sampling and analytical precision (see
 Section
       During the 1995 UATMP, 76
VOC and 75 carbonyl samples were
analyzed in replicate. Each replicate
analysis results in two quantified
concentrations (or non-detects) for all
of the chemicals considered.  For each   	
replicate analysis, RPDs were
calculated only for those chemicals for which concentration were measured in both replicates (see
sidebar). Average RPDs were then calculated for each chemical across all of the replicate
     8-2

-------
 analyses. Tables 8-1 and 8-2 summarize the average RPDs calculated for replicate analysis of
 VOCs and carbonyls.

       As shown in Table 8-1, the average RPDs calculated for VOC replicate analyses range
 from 3 percent to 157 percent. Of particular significance, the average RPDs calculated for those
 chemicals detected in more than 20 percent of the replicate analyses were always lower than 40
 percent; the RPDs calculated for chemicals detected in less than 20 percent of the replicate
 analyses exhibited much greater variability. To a first approximation, chemicals detected in less
 than 20 percent of the replicate analyses can be assumed to have concentrations consistently at
 levels near the estimated detection limit. Therefore, the more imprecise measurements were
 observed for chemicals generally found in very low concentrations in urban ambient air. It has
 been widely recognized that most environmental monitoring methods become increasingly
 imprecise when detecting concentrations at levels near estimated detection limits.

       As shown in Table 8-2, the average RPDs calculated for the carbonyl replicate analyses
 range from 3 percent to 21 percent. This data range suggests that analysis of the UATMP
 carbonyl samples is generally more precise than analysis of the UATMP VOC samples.  The
 better analytical precision for carbonyl analysis results in part from both the lower detection limits
 for carbonyl analysis and the higher prevalence of carbonyls in urban ambient air.

       In summary, the RPDs listed in Tables 8-1 and 8-2 suggest that the analytical methods
used in the 1995 UATMP measured concentrations of those chemicals most prevalent in ambient
 air more precisely. The greatest imprecision was observed for VOCs measured at levels near their
 estimated detection limits.  The results for these compounds (see summary tables in Appendix F)
 should be interpreted with caution.  Section 8.3 reconsiders the estimated analytical precision in
 context of the UATMP data quality objectives.

      8.1.2   Sampling and Analytical Precision
       Sampling and analytical precision measures random errors inherent to both the sampling
and analysis of ambient air. These random errors result from the imprecisions introduced by the
                                           8-3

-------
laboratory analytical equipment (see Section 8.1.1) and the field sampling devices. During the
1995 UATMP, random field sampling errors may have been caused by fluctuating flow rates of
ambient air through the sampling canisters and silica gel cartridges or by canister surface
interactions. The sampling and analytical precision of air monitoring methods are typically
evaluated by comparing the concentrations measured in duplicate samples.

       During the 1995 UATMP, duplicate samples were collected roughly once every 10
sampling days. The samples were collected by connecting a second sampling canister or silica gel
cartridge to the field sampling device by means of a tee connection. In so doing, nearly equal
volumes of ambient air passed through both sampling canisters or silica gel cartridges for the same
24-hour sampling period.  The majority of duplicate samples were then analyzed in replicate (Le.,
the two samples were each analyzed twice). Duplicate samples not analyzed in replicate result in
two concentrations measured for each chemical, while duplicate samples analyzed in replicate
result in four concentrations measured for each chemical. Regardless of whether replicate
analyses were performed, the variability among the concentrations analyzed in duplicate samples
indicates the combined imprecision of the sampling and analytical methods.

       Over the course of the 1995 program, 41 duplicate samples were collected for VOCs, and
39 for carbonyls. Based on the results of each duplicate sample, an RPD was calculated as a
measure of the sampling and analytical precision. RPDs were only calculated for chemicals
detected in all concentration measurements of a duplicate sample. For duplicate samples not
analyzed in replicate, RPDs were calculated for the two concentrations measured for each
chemical using the same approach presented in Section 8.1.1. For duplicate samples analyzed in
replicate, the concentrations of each chemical measured in the replicate analysis were first
averaged,  and RPDs were then calculated for the two average concentrations for each chemical.
After calculating RPDs for each duplicate sampling event, average RPDs were then calculated for
each chemical across all of the duplicate samples.  Because the sampling devices used at the
different monitoring stations are virtually identical, averaging RPDs across the samples collected
                                          8-4

-------
 at different locations is justified. Tables 8-3 and 8-4 summarize the results of the duplicate RPD
 calculations for VOCs and carbonyls, respectively.

       As shown in Table 8-3, the average RPDs calculated for VOC duplicate analyses range
 from 11 percent to 73 percent. Like the analytical precision, the sampling and analytical precision
 is best for those chemicals detected in more than 20 percent of the VOC duplicate samples.
 These VOCs generally have RPDs for duplicate samples of less than 40 percent. As shown in
 Table 8-4, the average RPDs calculated for carbonyl duplicate analyses range from 23 percent 49
 percent.  Section 8.3 comments on the observed precision estimates with respect to the
 recommended accuracy bounds specified in the UATMP data  quality objectives.

       The difference between the sampling and analytical precision (Tables 8-3 and 8-4) and the
 analytical precision (Tables 8-1 and 8-2) indicates the extent to which the sampling device
 introduces variability in the measured concentrations. For a given chemical, the RPDs in Tables
 8-3 and 8-4 generally exceed those in Tables 8-1 and 8-2. This observation indicates that both the
 sampling and analysis of ambient air contributes to the variability in the measured concentrations.

       8.2    Accuracy
       Accuracy measures the extent to which experimental measurements represent their
 corresponding "actual" or "true" values. Air sampling and analytical methods said to be "highly
 accurate" measure concentrations in very close agreement to actual ambient levels.  Inaccuracies
resuh from both random and systematic errors. Unlike the random errors discussed hi Section
 8.1, systematic errors in ambient monitoring efforts  cause sampling and analytical methods to
measure air concentrations consistently higher or consistently lower than the actual ambient
values. Some sources of systematic errors include improperly calibrated field sampling
equipment, misinterpreted peaks during laboratory analysis, and contaminated sampling or
analytical devices. The combined impact of random errors, as measured by the sampling and
analytical precision, and systematic errors determines the accuracy of monitoring methods.
                                          8-5

-------
       Systematic errors in ambient air monitoring can be estimated using several different
 external audit approaches.  Several external audit samples were analyzed to evaluate the accuracy
 of the carbonyl sampling, but no external audits were performed during the 1995 UATMP to
 measure the accuracy of the VOC sampling. The results of the carbonyl audits are included in
 Appendix J and indicate that the laboratory consistently analyzed concentrations of formaldehyde
 and acetone to levels within 15 percent of the actual values and concentrations of acetaldehyde to
 within 50 percent of the actual values. Appendix J includes documents suggesting the audit
 results for acetaldehyde may overestimate the percent error inherent in the laboratory analysis. In
 any case, all audit results fell within the UATMP data quality objectives for monitoring accuracy
 (see Section 8.3).

       Without external audits, it is impossible to measure the accuracy of the VOC sampling and
 analytical methods. Nonetheless, Appendix B describes several of the thorough quality assurance
 measures implemented during the 1995 program to prevent systematic errors from influencing the
 monitoring results. Strict adherence to these measures and the Compendium Method TO-14
 quality control guidelines likely minimized the impact of systematic errors on the 1995 UATMP
 results for VOCs.

       8.3    Comparison to UATMP Data Quality  Objectives
      According to the UATMP data quality objectives, concentrations measured to within 100
percent of the actual ambient levels are considered sufficient for evaluating the nature and
magnitude of urban air pollution (USEPA, 1988). Provided the systematic errors were effectively
minimized (as the carbonyl audit samples indicate), the 1995 UATMP program accuracy can be
assumed to be determined primarily by the random errors inherent to the corresponding sampling
and analytical methods. As shown in subsections 8.1.1 and 8.1.2, concentrations measured for
most VOCs and all carbonyls did not have variabilities exceeding 100 percent. Therefore,
assuming the absence of significant systematic errors, the concentrations measured for most
chemicals during  the 1995 UATMP meet the program data quality objectives. As noted earlier,
                                          8-6

-------
results for chemicals detected near the estimated detection limit should be interpreted with
caution.
Note:  Other environmental monitoring applications, such as those used for risk assessment,
       may have stricter data quality objectives. The precision data listed in Tables 8-1 through
       8-4 should indicate whether the UATMP monitoring data can also be used for other
       applications.
                                          8-7

-------
                                              Table 8-1
                                     VOC Analytical Precision
                           (Based on Replicate Analysis of 76 Samples)
Chemical

Benzene

BrQmodichloromet"flne
Bfomoloras '".

1,3-Bntadiene
Carbon tetrachloride
CnloTobenzene:
Chloroethane
Cldorofonn
Chloromethane
vwuOHHwmc
Dibromochloromethane
m-DicMofobenzene
o-Dichlorobenzene
j^lHchlorobenzene
1,1-Dicbloroethane
1^2-Dicblofoethane
/ron?-l,2-Dicbloroetnylene
i^-XftcbloropTOpane
cw-l,3-Dichloropropylene
^ans-1 J-PidiloroDropvlene

Ethylbenzene
Methyleae chloride
it-Octane
Propylene
Styrene
i^l^^-TetracMoToethane
Tetrachloroethylene
TAtnene
Number of Observations *

76

V

12
X*ป
'' 32
55
4
0
11
46
ปฃ•
3
8
6
25
0
4
3
0
0
0

71
44
56
63
52
3
29
73
Average RPD Observed
in Replicate Analysis

, „*_ g:,^^,,, ^
'' ' ' " '"T5t '•"" " "•'• " "" "' '
	 "l36 	 ^

43 	 " % — -- -
^J
19
ป/ ,„ J Jv,, !
—
% 23 v;w; ;, :\ :
, ^
t *
68
<>5 ' -^
83 ' "
4& f \- ^ ^ \
— b
6 - i
106
ซซ.*" '' ', -i
	 b
!h < :-

18
^ *1^ ^ iv :
25
15 ' ' '':
25
3 1
38
t4
* Replicate analyses resulting in non-detects were not considered in evaluating precision.
b Precision can not be evaluated for chemicals not detected in the replicate samples.

Note:   Chemicals detected in more than 20 percent of the replicate analyses are shown in bold font As discussed in
        Section 8.1.1, concentrations measured for these chemicals consistently had analytical precision better than 40
        percent
                                                 8-8

-------
                                       Table 8-1 (Continued)
                                     VOC Analytical Precision
                           (Based on Replicate Analysis of 76 Samples)
Chemical
1,1,1-Trichloroethane
1,1,2-Trichloroethane
Trichloroethylene
yioyl chloride
Number of Observations *
69
3
12
0
73
Average RPD Observed
in Replicate Analysis
36
.. , , ..^ ^ , ;„.,
56
/ :•. •:
16
* \ ซ'•><"—


' "' x
, -A *S
\ '• <
' Replicate analyses resulting in non-detects were not considered in evaluating precision.
b Precision can not be evaluated for chemicals not detected in the replicate samples.

Note:   Chemicals detected in more than 20 percent of the replicate analyses are shown in bold font  As discussed in
        Section 8.1.1, concentrations measured for these chemicals consistently had analytical precision better than 40
        percent
                                                 8-9

-------
                                           Table 8-2
                                 Carbonyl Analytical Precision
                         (Based on Replicate Analysis of 75 Samples)
Chemical
Acctaldhyde ^
Acetone
|U3Kd*3B ... \ ,, - \J
Bcnzaldehyde
$#%ฎsA$raM^ ' , 1 J
Crotonaldehyde
2,5-DimetnyIbenzaldehyde
Fonnaldehyde
Isovaleraldebyde
Tolualdebydes
Number of Observations '
, 74 _ _,.
70
i . ' , 44
56
*V W.V ^ ^/ -.-. -. •. "*^ S f S % -. ,
61
; '- ' 44
74
65 ' v
20
56
51
Average RPD Observed
in Replicate Analysis
'- -4 < '*• " ^ - '
•- -. v^ % "-S •• •.}< A^ > *-.\ %%W r S
3
*12 ™"_ ^"""
"• ••' 'i/i ^. . 5%% ^v^Vsv1- ^*^
s *" s -J '
" *""*"' ' ^jj"-"™"
% •. % ^* % ^,vปA> s -V v^ s S
4
.- ^ "" % "*.• "• f
17
, 14 ^J" ' [
* Replicate analyses resulting in non-detects were not considered in average RPDs.

Note:   All carbonyls were detected in more than 20 percent of the replicate analyses.
                                              8-10

-------
                                             Table 8-3
                             VOC Sampling and Analytical Precision
                                 (Based on 41 Duplicate Samples)
Chemical
i__4_ป 	
Acetylene
Benzene
Brwnochloromethane
Bromodichloromethane
':Jif.jXffB(OniCI|l!lff •• •.
oroinftf|^(^ h^f]ฃ
1.3-Birta<ฃene -
Carbon tetrachloride
Cniorobenzene
Chloroethane
Chbroform
Chloromethane
f*tmt
vAuOCOpItJUC
Dibromochloromethane
OT-DicMoroDenzene
0-Dichlorobenzene
j^DicUorobenzene
1,1-Dichloroethane
1,2-Dicnloroethane
trans- 1,2-Dichloroethylene
i;Z-I>icMoropropane
cis~ 1,3-Dichloropropylene
frans-l^'BicbioropropyJene
Ethylbenzene
Meibyiene chloride
it-Octane
Propytene
Styrene
1,1^,2-Tetrachloroethane
Trtrflchlnrnrthvlene
Number of Observations '
38
41
^
0
6
' 12 -
27
0
0
10
21
i
2
3
1
13
0
2
2
0
0
0
37
21
27
31
26
0
13
Average RPD Observed
in Analysis of Duplicates
13 , ^ ,';
11 % ^ 	 '
' ' ' ;.n-|-i '-. "" '< * ? v^^ •• ^'5
" " 	 b " %
68

11^
/ ' " " — v', " " , J
—
23
13
40
73
'.33 ""\- " ^:
17
31
	 b
-23" "
19
>— * , :
	 b
•ซ-*.* / ' •
20
&
28
13
30
— * '
25
* Duplicate samples resulting in non-detects were not considered in average RPDs.
k Precision can not be evaluated for chemicals not detected in any of the duplicate samples.

Note:   Chemicals detected in more than 20 percent of the duplicate samples are shown in bold font As discussed in
        Section 8.1.2, concentrations measured for these chemicals generally had sampling and analytical precision
        better than 40 percent.
                                                8-11

-------
                                     Table 8-3 (Continued)
                                    VOC Sampling Precision
                                (Based on 41 Duplicate Samples)
            Chemical
Number of Observations *
Average RPD Observed
in Analysis of Duplicates
  1,1,1-Trichloroethane

  Trichloroethylene
  m,/>-Xylene
          39
          35
           C
           5
          ^
          38
          37
          29
          20
" Duplicate samples resulting in non-detects were not considered in average RPDs.
b Precision can not be evaluated for chemicals not detected in any of the duplicate samples.

Note:    Chemicals detected in more than 20 percent of the duplicate samples are shown in bold font As discussed in
        Section 8.1.2, concentrations measured for these chemicals generally had sampling and analytical precision
        better than 40 percent.
                                               8-12

-------
                                           Table 8-4
                         Carbonyl Sampling and Analytical Precision
                               (Based on 39 Duplicate Samples)
Chemical
Acetaldhyde
Acetone
Aoroiem
Benzaldehvde
. ............ .y— ป
Botyj/feobiityraldeliyde
Crotonaldehyde
2^-Dinw%Iben2al(iehyde
Formaldehyde
HcxwnaJdehvde
• .-wrปnr ™™lrl y ^*~
Isovaleraldehyde
Ptoptonaldfihyde
Tolualdefaydcs
Valeralderiivd&
Number of Observations "
35
32
19
24
* '- .B ' :
28
iซ
35
30
8
. 25 ' -
24
^23
Average RPD Observed
in Analysis of Duplicates
(%)
* <•• 39 •• ........ ^ - :
34
/" .. ",'*— :^%"\1
30
27
- -47 :: * " !
39
30^ * 1
33
! ^ 23 - ' j
35
3ฃ ;
* Duplicate samples resulting in nan-detects or invalid samples were not considered in average RPDs.

Note:   All carbonyls were detected in more than 20 percent of the duplicate samples.
                                             8-13

-------
Section 9

-------
9.0    Conclusions and Recommendations

       As indicated throughout this report, the results of this extensive ambient air monitoring

study offer a wealth of information for evaluating trends and patterns in urban air pollution. The

data collected under the UATMP program should ultimately help a wide range of audiences put

both UATMP and other urban air monitoring data into proper context. The following discussion

summarizes the main conclusions of this report and presents recommendations for ongoing urban

air monitoring efforts.


       9.1    Conclusions
       The most significant conclusions of this report are as follows:
      Summary of Ambient Air Monitoring Data (Section 3). Tables 3-1 through 3-6
      summarize the prevalence, highest concentrations, and geometric mean concentrations of
      chemicals commonly found in urban ambient air.  The summary tables indicate that, with
      some exceptions, most VOCs and carbonyls considered in this study have geometric mean
      concentrations consistently lower than 1 ppbv.  Only acetylene, acetaldehyde, acetone, and
      formaldehyde have geometric mean concentrations greater than 2 ppbv at the majority of
      monitoring stations participating in the 1995 UATMP. Although the summary data
      represent a broad cross-section of urban environments, the data ranges observed in this
      study do not necessarily characterize all urban settings.

      Statistical Analysis (Section 4).  Statistical analysis of the 1995 UATMP data suggests
      that concentrations measured during this program are neither perfectly normally or
      perfectly lognormally distributed. Evidence does suggest that the concentrations are
      better represented by lognormal distributions. Therefore, geometric mean concentrations
      are more representative of annual average concentrations than either arithmetic mean or
      median values.

      The statistical analysis also shows that very few significant correlations exist between
      chemicals measured in different urban locations. This suggests that each urban
      environment has unique emission sources and local meteorological conditions which
      influence the composition and magnitude of air pollution.  The data correlations did,
      however, identify significant trends among certain aromatic hydrocarbons. As shown in
      Section 6, these correlations likely result from motor vehicle exhaust, an emission source
      common to urban environments.

      Interpreting Ambient Air Monitoring Data (Section 5). The number and complexity of
      industrial and motor vehicle emission sources complicates efforts to analyze ambient air

                                        9-1

-------
 concentrations measured in urban environments. Using 1,3-butadiene as an example,
 however, this report does present a methodology for considering appropriate emission
 sources and atmospheric fate and transport mechanisms when evaluating the composition
 and magnitude of urban air pollution.

 Spatial Variations (Section 6).  As expected, the concentrations observed during the 1995
 UATMP are highest at sites that are densely populated, highly industrial, or near major
 roadways.  Consequently, concentrations observed at the Vermont monitoring stations,
 which are mostly removed from major industrial emission sources, are significantly lower
 than the concentrations observed at monitoring stations along the coast of the Gulf of
 Mexico, where many chemical manufacturing facilities are located.

 The spatial variations observed for VOCs indicate that industrial emissions data in the
 Toxic Release Inventory do not correlate with ambient air concentrations for most of the
 chemicals considered. Similarities among VOC concentration profiles, particularly for
 BTEX compounds, are observed across all monitoring stations participating in the 1995
 program. These highly correlated profiles suggest that emissions from motor vehicles
 contribute significantly to the VOC concentrations of aromatic hydrocarbons measured
 during the program.

 The spatial variations observed for carbonyls cannot be readily explained by either TRI
 emissions data or motor vehicle emissions trends.  Correlations between formaldehyde and
 acetaldehyde were identified across monitoring stations located within a given geographic
 region (i.e., the Vermont stations), but the correlations did not apply to all urban areas.
 The absence of significant correlations suggests that carbonyls found in urban ambient air
 may arise from other sources, most likely as products of photochemical reactions. Many
 atmospheric chemistry references confirm this conclusion.

 Temporal Variations (Section 7). Most VOCs and carbonyls considered in this program
 have elevated ambient air concentrations during warmer months, as opposed to colder.
 These trends are stronger for carbonyls. The notable tendency for ambient air
 concentrations of carbonyls to increase in the summer (along with the results of Section 6)
 strongly suggest that carbonyls  in urban ambient air generally are formed as products of
 photochemical reactions.

 Monitoring data collected at the Camden, New Jersey (CAN!) location were compared to
 those collected during four previous UATMP program years.  The concentrations for
 selected aromatic hydrocarbons measured at CANJ steadily declined over the long term.
 With only five years of data, however, it is not possible to determine (or predict) the
 nature and extent of concentration patterns in the CANJ area over the longer term.

Precision and Accuracy (Section 8). The UATMP data quality objectives specify that
 concentrations measurements under this program should fall within ฑ100 percent of the
 actual ambient air concentrations (USEPA, 1988).  The concentrations measured during
the 1995 program generally fall well within these guidelines. Measurements for chemicals
 at ambient levels  significantly exceeding the estimated detection limits are most reliable.

                                    9-2

-------
       9.2    Recommendations

       In light of the 1995 UATMP program, a number of recommendations for future ambient

air monitoring are warranted:
       Ambient air monitoring studies use several different parameters, including the median,
       arithmetic mean, and geometric mean, to estimate the central tendency of chemical
       concentration distributions. The lack of a consistent approach makes comparisons
       between different air monitoring studies difficult. Although the statistical analysis in this
       report strongly suggests that ambient air concentrations are lognormally distributed,
       further research should be conducted to determine whether this trend applies to
       concentration distributions in other urban environments.

       As noted throughout this report, the treatment of non-detects has a significant impact on
       the conclusions drawn from ambient air monitoring data; therefore, meaningful and
       consistent methods should be developed for treating non-detects in future ah* monitoring
       efforts. At the very least, the frequencies of detection should be reported in all ambient air
       monitoring studies to put the measured concentrations in proper context.

       Estimated detection limits do not always represent the instrument sensitivity for ambient
       air monitoring methods. Without representative detection limits, non-detect results
       (typically assigned values of one-half the detection limit) may be evaluated out of context
       when analyzing urban ambient air monitoring data.  Accordingly, procedures for
       estimating detection limits should be researched and implemented.

       Of the VOCs considered in this study, nearly half were detected in less than 50 percent of
       the samples collected. For these chemicals, the limited number of detections does not
       allow for accurate estimate of many data summary parameters. To prevent such
       limitations in future air monitoring projects, research efforts should continue to focus on
       developing more sensitive sampling and analytical technologies to measure concentrations
       of urban air pollutants at trace levels in the atmosphere.

       Ongoing ambient air monitoring at fixed locations should provide useful information both
       on long term trends in urban air quality and on the potential for urban air pollution to
       cause adverse health effects among the general population.  To meet these goals, state and
       local agencies should be strongly encouraged to develop either their own ambient ah*
       monitoring programs or participate in future UATMP monitoring efforts.
                                          9-3

-------
Section 10

-------
 10.0  References

 ATSDR, 1992. "lexicological Profile for 1,3-Butadiene." U.S. Department of Health & Human
       Services, Agency for Toxic Substances and Disease Registry.  July, 1992.

 Brimblecombe, 1986. "Air Composition & Chemistry." Peter Brimblecombe.  Cambridge
       University Press. 1986.

 Conner, Lonneman, Seila, 1995.  "Transportation-Related Volatile Hydrocarbon Source Profiles
       Measured in Atlanta." Ten L. Conner, William A Lonneman, Robert L. Seila.  Journal of
       the Air and Waste Management Association, 45: 383-394. 1995.

 ES, 1987. "Urban Air Toxics Monitoring Program." Engineering-Science, Lie. Prepared for
       U.S. Environmental Protection Agency, Office of Air and Radiation, Office of Air Quality
       Planning and Standards. September, 1987.

 FR, 1984. "Definition and Procedure for the Determination of the Method Detection Limit -
       Revision 1.1" Federal Register, Volume 49, Number 209. October 26,1984.

 Gilbert, 1987. "Statistical Methods for Environmental Pollution Monitoring." Richard 0. Gilbert.
       Van Nostrand Reinhold Publishers. 1987.

 Graedel, 1978.  "Chemical Compounds in the Atmosphere." T.E. Graedel. Academic Press.
       1978.

 Hamilton, 1990. "Modern Data Analysis." Lawrence C. Hamilton, Brooks/Cole Publishing
       Company. 1990.

 Harnett, 1982. "Statistical Methods." Donald L. Harnett, Addison-Wesley Publishing Company,
       Third Edition. 1982.

 Radian, 1990.  "1989 Urban Air Toxics Monitoring Program -Final Report." Radian
       Corporation. Prepared for U.S. Environmental Protection Agency, Office of Air Quality
       Planning and Standards. September, 1990.

Radian, 1991.  "1990 Urban Air Toxics Monitoring Program-Final Report." Radian
       Corporation. Prepared for U.S. Environmental Protection Agency, Office of Air Quality
       Planning and Standards. June, 1991.

Radian, 1996.  "1994 Urban Air Toxics Monitoring Program - Final Report." Radian
       Corporation. Prepared for U.S. Environmental Protection Agency, Office of Air Quality
       Planning and Standards. June, 1996.
                                         10-1

-------
Rhodes, 1991. "Much Ado About Nothing, Or What to do with Measurements Below the
       Detection Limit." Environmetrics 81: Selected Papers, SIAM SIMS Conference Series
       No. 8, Society for Industrial and Applied Mathematics, Philadelphia, pp. 157-62. 1991.

Seinfeld, 1986. "Atmospheric Chemistry and Physics of Air Pollution." John H. Seinfeld.  John
       Wiley & Sons. 1986.

Scheff, 1993. "Receptor Modelling of Volatile Organic Compounds. 1. Emission Inventory and
       Validation." Peter A. Scheflf and Richard A Wadden. Environmental Science and
       Technology, 27: 617-625. 1993.

Shapiro and Wilk, 1965. "An Analysis of Variance Test for Normality (complete samples)."
       Shapiro, S.S. and M.B. Wilk. Biometrika, 52: 591-611.  1965.

Snedecor and Cochran, 1980. "Statistical Methods." Snedecor, George W. and William G,
       Cochran. Ames, Iowa: The Iowa State University Press, Seventh Edition. 1980.

TRI, 1994. 1994 Toxic Release Inventory Reporting. U.S. Environmental Protection Agency,
       Office of Pollution Prevention and Toxics.  (Data downloaded from the TOXNET system
       of the National Library of Medicine.) 1994.

USDOC, 1993. "1990 Census of Population and Housing." U.S. Department of Commerce,
       Bureau of the Census. July, 1993.

USEPA, 1984a. "Determination of Volatile Organic Compounds (VOCs) in Ambient Air Using
       Summa* Passivated Canister Sampling and Gas Chromatographic Analysis." U.S.
       Environmental Protection Agency, Quality Assurance Division, Environmental Moniotring
       Systems Laboratory.  June, 1984.

USEPA, 1984b. "Determination of Formaldehyde in Ambient Air Using Adsorbent Cartridge
       Followed by High Performance Liquid Chromatography." U.S. Environmental Protection
       Agency, Quality Assurance Division, Environmental Moniotring Systems Laboratory.
       June, 1984.

USEPA, 1986. "Guideline on Air Quality Models." U.S. Environmental Protection Agency,
       Office of Air and Radiation, Office of Air Quality Planning and Standards. July, 1986.

USEPA, 1988. "Data Quality Objectives for the Urban Air Toxics Monitoring Program (Stages I
       and D)." U.S. Environmental Protection Agency, Office of Air and Radiation, Office of
       Air Quality Planning and Standards. June, 1988.
                                        10-2

-------
USEPA, 1989. "Risk Assessment Guidance for Superfund. Volume I: Human Health Evaluation
       Manual (Part A)." Interim Final. U.S. Environmental Protection Agency, Office of
       Emergency and Remedial Response. December, 1989.

USEPA, 1995. "Toxic Chemical Release Inventory Reporting Form R and Instructions."
       Revised 1994 Version.  U.S. Environmental Protection Agency, Office of Pollution
       Prevention and Toxics.  March, 1995.

Wayne, 1991. "Chemistry of Atmospheres."  Richard P. Wayne.  Clarendon Press.  1991.
                                       10-3

-------