EPA 510-R-92-802
DRAFT FINAL REPORT
BACKGROUND HYDROCARBON
VAPOR CONCENTRATION STUDY
FOR UNDERGROUND FUEL STORAGE TANKS
U.S. EPA CONTRACT NO. 68-03-3409
February 29, 1988
Prepared for:
MR. PHILIP B. DURGIN, PhD
U.S. ENVIRONMENTAL PROTECTION AGENCY
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada
Prepared by:
GEOSCIENCE CONSULTANTS, LTD.
HEADQUARTERS
500 Copper Avenue, NW
Suite 200
Albuquerque, New Mexico 87102
- (505) 842-0001
'FAX (505) 842-0595
WEST COAST REGIONAL OFFICE
5000 Birch Street
West Tower, Suite 3000
Newport Beach. CA 92660
(714) 476-3650
FAX (714) 752-2160
EASTERN REGIONAL OFFICE
1109 Spring Street
Suite 706
Silver Spring, Maryland 20910
(301) 587-2088
FAX (301) 588-0605
Printed on Recycled Paper
-------
-------
DRAFT FINAL REPORT
BACKGROUND HYDROCARBON
VAPOR CONCENTRATION STUDY
FOR UNDERGROUND FUEL STORAGE TANKS
U.S. EPA CONTRACT NO. 68-03-3409
February 29. 1988
Prepared for:
MR. PHILIP B. DURGIN, PhD
U.S. ENVIRONMENTAL PROTECTION AGENCY
Environmental Monitoring Systems Laboratory
Las Vegas. Nevada
-------
-------
. . -- DRAFT FINAL REPORT
BACKGROUND HYDROCARBON
VAPOR CONCENTRATION STUDY
FOR UNDERGROUND FUEL STORAGE TANKS
U.S. EPA CONTRACT NO. 68-03-3409
SUBMITTED
GCU Program Kftna
GCL Project Director d
DATE:
-------
-------
TABLE OF CONTENTS
: se
1.0 EXECUTIVE SUMMARY .......... ,
ป. i
?2.0 PURPOSE OF:STUDY ...... ' - r
- ซซซซซ....... o
C3.JO SITE SELECTION ... . . ... . ' ,
"3.1 LOCATIONS ' ' ' 4
3.2 SERVICE STATIONS . '.'.'.'.'.'.'.'.'.'.I'.'.'.'. 8
-4iO GEOLOGY, HYDROLOGY AND CLIMATE ..... q
r4.a AUSTIN, TEXAS .!!.'.*.'!"'*' q
-4.1.1 Geology and Hydrology ] q
4.1.2 Climate ..... ....... y
4.2 LONG ISLAND SOUND AREA, NEW YORK, RHODE*ISLAND "
AND CONNECTICUT .... ..... IQ
i'H 5eoJฐgy and Hydrology - Long Island, *New*York ! ! ! 10
4.2.2 Geology and Hydrology - Providence, Rhode Island . 11
4.2.3 Geology and Hydrology - Storrs, Connecticut .... 11
4.2.4 Climate n
4.3 SAN DIEGO REGION, CALIFORNIA ....!!!!!!!''*' 12
4.3.1 Geology and Hydrology ...... .....
4.3.2 Climate ........ ....!!! i!!! i i ] {3
5.0 FIELD METHODS 1A
5.1 SAMPLING STRATEGY '.'.'. * * ' ' };
5.2 SAMPLING METHODS iซi
5.3 ANALYTICAL PROCEDURES -1 !!!!!!'!!!!! 16
6.0 QUALITY ASSURANCE AND QUALITY CONTROL ... 17
6.1 QA OBJECTIVES FOR MEASUREMENT DATA (QAPP SECTION 3.1)" .' ' 17
6.1.1 Gas Chromatograph Analyses 17
6.1.2 Soil Moisture Content Analyses .... ' 19
6.2 SAMPLING PROCEDURES (QAPP SECTION 3.2) .... * ' ' ' 19
6.3 SAMPLE CUSTODY (QAPP SECTION 3.3) ..... 21
6.4 CALIBRATION PROCEDURES AND FREQUENCY
(QAPP SECTION 3.4) ?1
6.5 ANALYTICAL PROCEDURES (QAPP SECTION 3.5) . .* 22
6.6 DATA REDUCTION, VALIDATION AND REPORTING
(QAPP SECTION 3.6) . 22
6.7 INTERNAL QUALITY CONTROL CHECKS (QAPP SECTION'3!7) ' ' * ' 23
6.8 PERFORMANCE AND SYSTEM AUDITS (QAPP SECTION 3.8 23
6.9 PREVENTIVE MAINTENANCE (QAPP SECTION 3.9) . 23
6.10 ASSESSMENT OF DATA PRECISION, ACCURACY AND
COMPLETENESS (QAPP SECTION 3.10) ........ 24
6.11 CORRECTIVE ACTIONS (QAPP SECTION 3.11) .... " ' ' ' ?5
6.12 QUALITY ASSURANCE REPORTS TO MANAGEMENT
(QAPP SECTION 3.12) . 25
-------
7.0 REPORTING METHODS 26
7.1 DETERMINATION OF TOTAL HYDROCARBON CONCENTRATIONS
IN MICROGRAHS PER LITER . 26
7.1.1 Gas.Chromatograph and Flame lonization Detector
'Operation 't 27
7.1.2 Calculation of Total Hydrocarbons as Benzene ... 28
"7.1.1:. Calculation of Total Hydrocarbon Concentrations
Using Average Response Factors ........... 29
7.2 DETERMINATION OF TOTAL HYDROCARBON CONCENTRATIONS IN
- PARTS PER MILLION ........ 32
8.0 RESULTS . 36
8.1 SOIL GAS DATA 36
.8.2 CONTAMINATED SITE DATA 36
8.3 EXPANDED AUSTIN STUDY 40
8.4 CHARACTERIZATION OF BACKFILL MATERIAL 46
8.4 U-TUBE SAMPLING ..... 46
8.6 GROUND WATER SAMPLING 51
9.0 UST REGULATIONS.. 53
9.1 AUSTIN, TEXAS 53
9.2 SUFFOLK COUNTY, NEW YORK 53
9.3 SAN DIEGO, CALIFORNIA 54
10.0 TANK TIGHTNESS TESTING RECORDS 56
11.0 DATA ANALYSIS 61
11.1 EMPIRICAL DISTRIBUTION OF TOTAL HYDROCARBON
CONCENTRATIONS (LESS METHANE) FOR NON-CONTAMINATED
SITES 63
11.2 EMPIRICAL DISTRIBUTION OF TOTAL HYDROCARBON
CONCENTRATIONS (INCLUDING METHANE) OF NON-
CONTAMINATED SITES . . 68
11.3 COMPARISON OF TOTAL HYDROCARBON CONCENTRATIONS FOR NON-
CONTAMINATED SITE AND CONTAMINATED SITE DATA SETS .... 71
11.4 NON-PARAMETRIC STATISTICAL TESTING 78
11.4.1 The Risks Associated with Hypothesis Testing ... 79
11.4.2 Comparison of Non-Contaminated Site and
Contaminated Site Data Distributions 81
11.4.3 Non-Parametric Testing for Data Patterns Within
the Non-Contaminated Data 82
11.4.3.1 Location . . ........... . ... 83
H.4.3.2 Sample Depth 84
11.4.3.3 Conclusions from Non-Parametric Tests
Within the Non-Contaminated Data 91
11.5 RESULTS AND CONCLUSIONS OF DATA ANALYSIS .... 92
12.0 CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER STUDY ...... 96
12.1 CONCLUSIONS . . . 96
12.2 RECOMMENDATIONS FOR FURTHER STUDY 96
13.0 REFERENCES CITED ....... .... 98
-------
LIST OF TABLES
TABLE 6-1 RESULTS OF REPLICATE ANALYSES FOR SOIL MOISTURE
CONTENT - j . 20
TABLE 7-1 MAJOR COMPONENTS OF API PS-6 GASOLINE . ">
TABLE 8-1 MAXIMUM CONCENTRATIONS AT AUSTIN, TEXAS .!'*'**' 37
TABLE 8-2 MAXIMUM CONCENTRATIONS LONG ISLAND SOUND COASTAL AREA* ' 38
TABLE 8-3 MAXIMUM CONCENTRATIONS AT SAN DIEGO, CALIFORNIA . . 39
TABLE 8-4 DESCRIPTION OF CONTAMINATED SITES ...... ' 4?
TABLE 8-5 MAXIMUM. CONCENTRATION DATA AT CONTAMINATED ' ' '
SITES 42
TABLE 8-6 MOISTURE RANGES OF SOIL AND BACKFILL SAMPLES .""!!'*' 47
TABLE 8-7 U-TUBE VAPOR SAMPLES . '
SUFFOLK COUNTY, NEW YORK .......;.;. 50
TABLE 8-8 HYDROCARBON CONCENTRATIONS FROM GROUNDWATER '
SAMPLES 52
TABLE 10-1 TANK TIGHTNESS TEST RESULTS 57
TABLE 11-1 DISTRIBUTION OF NON-CONTAMINATED SITE DATA
FOR TOTAL HYDROCARBONS (LESS METHANE) ........ 65
TABLE 11-2 DISTRIBUTION OF NON-CONTAMINATED SITE DATA
FOR TOTAL HYDROCARBONS (LESS METHANE)
(PARTS PER MILLION BY VOLUME) ......... 66
TABLE 11-3 TOTAL HYDROCARBON CONCENTRATIONS LESS METHANE " ' '
GREATER THAN 100,000 MICROGRAMS PER LITER 67
TABLE 11-4 COMPARISON OF TOTAL HYDROCARBONS INCLUDING METHANE
AND LESS METHANE AT NON-CONTAMINATED SITES
(MICROGRAMS PER LITER) 69
TABLE 11-5 COMPARISON OF TOTAL HYDROCARBONS INCLUDING METHANE "
AND LESS METHANE AT NON-CONTAMINATED SITES
(PARTS PER MILLION BY VOLUME) 70
TABLE 11-6 DISTRIBUTION OF CONTAMINATED SITE DATA FOR
TOTAL HYDROCARBONS LESS METHANE 73
TABLE 11-7 COMPARISON OF NON-CONTAMINATED AND CONTAMINATED
SITE DATA DISTRIBUTIONS FOR HYDROCARBONS
LESS METHANE 74
TABLE 11-8 RESULTS OF KRUSKAL-WALLIS TESTS FOR LOCATIONS WITH ' ' '
STEEL TANK SYSTEMS USING NON-CONTAMINATED
DATA 85
TABLE 11-9 RESULTS OF KRUKSAL-WALLIS TESTS FOR LOCATIONS WITH ' ' "
WITH FIBERGLASS TANK SYSTEMS USING NON-CONTAMINATED
DATA ........... 86
TABLE 11-10 RESULTS OF PAGE L FOR DIFFERENCES IN DATA
ACCORDING TO SAMPLE DEPTH 89
TABLE 11-11 RESULTS OF WILCOXON TESTS FOR DIFFERENCES IN
DATA ACCORDING TO SAMPLE DEPTH 90
TABLE 11-12 DESCRIPTIVE STATISTICS FOR TOTAL HYDROCARBON LESS
METHANE CONCENTRATIONS IN STEEL TANK SYSTEMS AT
DIFFERENT LOCATIONS AND SAMPLE DEPTHS (MICROGRAMS
PER LITER) . 93
TABLE 11-13 DESCRIPTIVE STATISTICS FOR TOTAL HYDROCARBON LESS* * ' '
METHANE CONCENTRATIONS IN FIBERGLASS TANK SYSTEMS
AT DIFFERENT DEPTHS (MICROGRAMS PER LITER) 94
-------
LIST OF FIGURES
FIGURE 2-1
FIGURE 7-1
FIGURE 8-1 -
FIGURE 8-2
FIGURE 8-3
FIGURE 11-1
FIGURE 11-2
TYPICAL UST ARRANGEMENT ........; 6
RATIO OF BTEX TO TOTAL HYDROCARBON CONCENTRATIONS
VERSOS CUMULATIVE NUMBER OF SAMPLES 31
.AUSTIN #6 (FRESH SPILL) MEDIAN TOTAL HYDRO-
CARBON DATA ....... 43
AUStlN #6 (FRESH SPILL) MEDIAN C4 - C6 DATA 44
U-TUBE LEAK DETECTION SYSTEM
48
NON-CONTAMINATED SITE DATA DISTRIBUTION 76
CONTAMINATED SITE DATA DISTRIBUTION
77
LIST OF APPENDICES
APPENDIX A TANK SUMMARY
APPENDIX B SUMMARY OF FIELD NOTES AND CONDITIONS
APPENDIX C WEATHER DATA
APPENDIX D SOIL GAS DATA AND SITE MAPS
SOIL MOISTURE AND SIEVE ANALYSIS DATA
INDIVIDUAL GC-FID DATA
QUALITY ASSURANCE PROJECT PLAN
RESPONSE FACTORS AREA COUNTS, AIR ANALYSES AND NITROGEN
BLANKS - DATA FOR QUALITY ASSURANCE
APPENDIX I QA/QC AUDIT - LETTER RESPONSE
APPENDIX J SUPPORTING DOCUMENTATION FOR REPORTING METHODS EVALUATION
APPENDIX K CONTAMINATED SITE DATA
APPENDIX L TANK TESTING RECORDS
APPENDIX M CALCULATIONS FOR STATISTICAL TESTS
APPENDIX E
APPENDIX F
APPENDIX G
APPENDIX H
-------
1.1) EXECUTIVE SUMMARY
The Environmental Monitoring Systems .Laboratory (EMSL) of the USEPA
awarded Contract No: 68-03-3409 to Camp, Dresser and McKee (COM) to
conduct-a study to determine the background hydrocarbon-concentrations In
soil, vapor 1n the backfill of representative underground fuel storage
tank (UST) sites across the country. COM designated Geoscience
Consultants., Ltd. (GCL) to select sampling sites, prepare sampling
strategies, review data collection*, analyze the data and prepare a final
report. Field data oh""clean UST sites were collected from September 14
to December 13, 1987. Data on UST sites with documented releases were
obtained from Tracer Research Corporation files.
Since no database for soil vapor information at non-contaminated under-
ground storage tank 'sites was known to exist, a field sampling program
was undertaken to establish a baseline data set of hydrocarbon vapor
concentrations. Data were collected from 27 gasoline service stations
selected as non-contaminated sites in three diverse geographic regions:
Central Texas (Austin, Texas); areas surrounding Long Island Sound
(Suffolk County, New York; Providence, Rhode Island; Storrs,
Connecticut); and Southern California (San Diego, California). The three
regions were selected for their active underground storage tank
regulatory programs, as well as their differences in geology, hydrology
and climate. The non-contaminated database consists of 279 soil vapor
samples from 25 service stations. At the other two stations, observed
or suspected leaks prevented their data from being used in the non-
contaminated database.
At each location, soil was sampled at varying distances and depths from
UST appurtenances (such as submersible pumps, vents and product flow
lines) to determine 1f a particular pattern of hydrocarbon concentration
existed. Samples were collected by driving a hollow steel probe Into the
ground, and evacuating 5 to 10 liters of soil vapors with a vacuum pump.
Volatile hydrocarbon species were identified and quantified at the site
-------
by utilizing gas chromatograph/flame lonization detection (GC/FID)
equipment. Ten to fifteen samples were collected and analyzed at each
site.
/ -
>
Hydrocarbon "'vapor concentrations from the non-contaminated sites range
from detection limit levels of 0.02 micrograms per liter (ug/1) to
maximum values of 870,000 ug/1 of-methane, 110,000 ug/1 of benzene,
160,000 ug/1 of toluene., 25,000 ug/1 of ethyl benzene, and 110,000 ug/1 of
xylenes. The maximum concentratftuv of total hydrocarbons (less methane)
is 1,000,000 ug/1. Determination of total hydrocarbon concentrations
exclude methane peaks in order to elevate the compounds most
representative of gasoline. Additionally, subtraction of the methane
peaks precludes the inclusion of methane concentrations caused by
naturally-occurring organic decomposition.
The statistical distribution of total hydrocarbons (less methane)
indicates that a majority of the concentration values are in the lower
concentration ranges. The relative frequency distribution shows 53.2
percent of the samples below 1,500 ug/1 (500 ppm by volume) and 93.1
percent below 100,000 ug/1 (27,000 ppm by volme). The median is 800
ug/1 and the mean is 23,300 ug/1.
Contaminated site data were obtained from Tracer Research Corporation's
historical records. The contaminated site data consists of 60 soil vapor
samples taken from nine sites having known contamination from a
petroleum fuel leak or spill. These sites were all active gasoline
service stations or fueling facilities. The contaminated site data also
shows much variability. The statistical distribution of total
hydrocarbons (less methane) shows a majority of sample values to be in
the lower concentration ranges. The relative frequency distribution
shows 35 percent of the samples below 1,500 ug/1 (500 ppm by volume) and
66.7 percent below 100,000 ug/1(27,000 ppm by volume). The median is
9,000 ug/1 and the mean Is 160,000 ug/1.
-------
Although much ^variability exists in both the non-contaminated and
contaminated site data, significant differences can be seen between the
two distributions. -A .-ten-fold difference exists between the means and
the medians *of ซeach dat.a set. This ten-fold difference also exists
between -the' numbers of.samples with concentrations above 10,000 ug/1
(3,000 ppmv) for the two data sets." For example; 29.6 percent of the
non-contaminated samples occurred in the range of 10,000 ug/1 to 100,000
ug/1 while 33.3 percent of the contaminated samples concentrations
occurred in the range of 100,000 ug/1 to 1,000,000 ug/1. ' " -
Statistical data patterns associated with site location and sample depth
were delineated using non-parametric statistical methods. Statistically
significant differences were found to exist between the total hydrocarbon
(less methane) vapor'concentrations among the five locations studied for
steel tank systems, whereas these differences were not significant for
fiberglass tank systems. Statistically significant differences also
occurred between the total hydrocarbon (less methane) vapor
concentrations among the sample depths of 2, 6 and 10 feet for both steel
and fiberglass tank systems. Higher concentrations were found at the
lower depths.
A fresh spill at one station in Austin provided an opportunity to add
butane to the list of analytes under study. The butane concentration in
15 soil gas samples taken during the first four days after the spill
occurred ranged from 530 ug/1 to 300,000 ug/1. Butane was also sampled
at sites in Storrs, Connecticut and Providence, Rhode Island both of
which had no evidence of recent leaks or spills. At these two sites,
butane concentrations in 65 soil gas samples ranged from the minimum
detection limit.of 0.02 ug/1 to 930 ug/1. The large difference between
the butane concentrations at the fresh spill site in Austin and the non-
contaminated sites in Connecticut and Rhode Island suggests that butane
may be a good indicator of a fresh spill or leak.
Because .there ,are no standard procedures for calculating and reporting
total hydrocarbon concentration data, GCL evaluated different calculation
-------
methods. It was determined that the best approximation of total
hydrocarbon (less methane) concentrations, based on available calibration
data, was achieved using an average response factor calculated from the
daily response factors of benzene, toluene, elhylbenzene and ortho-
xylene.- -.._--
-------
2.0 PURPOSE OF STUDY
Proposed Federal regulations to monitor ground water contamination around
.underground storage tank (UST) systems require; the development of
-effective external *and internal leak detection 'methods. Soil gas
sampling is an-external detection method which could prove useful in
determining whether an underground storage tank is leaking.
In order to determine the effectiveness of soil gas surveys in leak
detection, a study was designed with the following goals:
- Collection of soil gas data from sites where the tank system
was tested and found to be tight, providing background soil gas
data, and
* Comparison "of these background data to soil gas data from sites
known to be contaminated by spills or leaks in order to
identify a data pattern which may be indicative of a leaking
system.
To fulfill these goals, soil gas surveys were performed at 27 active
gasoline service stations in three diverse geographic regions. Hydrocar-
bon vapor concentrations in the backfill surrounding the underground
storage tanks were sampled and analyzed.
The term "soil, gas" refers to vapors found in the interstitial area
between particles of sand or gravel (pores). "Soil gas" and "soil vapor"
are used interchangeably in this report. These vapors, often loaded with
hydrocarbons when a underground storage tank is leaking, escape into the
gravel or sand which is used to surround the tank during installation.
This surrounding tank medium is called "backfill". Typically pea gravel
is used for bapkfill around fiberglass tanks, and sand around steel
tanks. An overview of a typical UST arrangement is shown in Figure 2-1.
-------
fi
' VENTS
-
' FILLS ^
STEEL TANKS
VENT PIPING ~>y
'^C^r~:~inrir~<
11 . . 1 ||
i . i i
II '
1 ' |!
~2Jfฎ i| ฎ 1 ฎ I **
1 ' H
J "1 " H '
i g II g M g II c
Hr" J ^ - J L-- ป J 1 .
/
^9
^
PTPTMR ^*" = ~ d~ rป
.
_!
i r~*~"
ll
H .
Mlฎ
it
n
ULJj
p
^
"* /
i!
1 ",
!
i i
1 i
j-i
V^r
f
A. ฑ * *.ป* pAvED SURFACE-X
^
' ^
-
&
-------
3.0 SITE SELECTION
.3.1 LOCATIONS e
Soil-gas surveys were conducted-at the following locations:
>
'' . > -ซ
--.. - Austin, Texas
- San Diego, California
- Long Island Sound area *
.: - Suffolk County, New York ' '
- Providence, Rhode Island
- Storrs, Connecticut
Austin, San Diego and Suffolk County, New York were originally selected
as the locations for the study because they were recognized as having
exemplary local underground storage tank regulatory programs, and they
represented different geographical situations. Stations in Providence
and Storrs were added to provide a broader data base from the Long Island
Sound area, and to interact with the underground storage tank evaluation
program at the University of Connecticut.
Active regulatory programs were desired in order to assure that accurate
information would be available for the stations to be studied. Since a
major purpose of the study was to determine background soil vapor levels
at clean, well-managed stations, it was necessary to determine if leaks
or spills had previously occurred at the stations being tested. Records
at Austin, San Diego and Suffolk County were carefully reviewed and all
available information was obtained concerning the specific stations to be
studied.
Different geographical locations were desired for the study in order to
eliminate possible data bias that could occur if sampling were done at
one location. The selected locations represent a wide range of tempera-
ture, humidity, geology and topography. Although soil gas samples were
taken primarily from the backfill areas of the tanks,'the surrounding
geology and climatic conditions can affect the concentration of vapors
existing In the backfill material. *
* 7
-------
3.2 SERVICE STATIONS
Three oil companies cooperated in the study by offering several of their
service stations as candidates for field testing. Twenty-seven stations
t
were selected which represent a variety of. tank ages, tank materials,
products stored, and backfill materials. The stations were selected
according to the following screening criteria:
- The stations were to be clean, well-managed businesses with no
major environmental problems.
* Existing tanks were required to meet the appropriate operation
specifications.
The tanks must have been in the ground and operational for at
least 6 months prior to the site sampling.
The stations were required to have relatively large total
throughputs of product since beginning operation and relatively
large throughputs on a monthly basis.
The stations were required to have good inventory control.
Twenty-seven service stations with ten to fifteen sample points at each
station were selected, providing a broad data base with a variety of
tanks, backfills and field conditions. There were a total of 100 under-
ground storage tanks Involved in this study, of which 63 were made of
steel and 37 of fiberglass. Tank installation dates ranged from 1940 to
1984 for steel tanks, and 1978 to 1984 for fiberglass tanks. A listing
of all of the tanks 1s shown in Appendix A.
8
-------
4.0 GEOLOGY, HYDROLOGY AND CLIMATE
This section briefly describes the geologic, hydrologic and climatic
characteristics wMeh may effect hydrocarbon sol] gas., concentrations
within the three study regions. " '
4.1 AUSTIN, TEXAS
4.1.1 -Geology and Hydrology ~
Bedrock in the Austin area consists dominahtly of limestones, marls, and
shales^ .all of Cretaceous age. ^Terrace deposits and aTluvium locally
overlie the bedrock units in the present valley of.the Colorado River and
on terraces representing older Quaternary drainage levels.
Station sites AU-2, AU-4, AU-5, and AU-6 all lie in outcrop areas of the
Upper Cretaceous Austin Group, which consists of chalk, limestone and
marly limestone. A very thin (less than 5 feet) cover of sand and gravel
terrace deposits may be present at site AU-4. Site AU-5 lies within 100
feet of a fault which exposes Cretaceous clay at the land surface on the
side of the fault opposite the station.
Sites AU-1 and AU-7 are located in areas of alluvial sand and gravel
comprising terrace deposits, but these deposits are probably less than 10
feet thick at both sites. The alluvium Is underlain by Lower Cretaceous
clay of the Del Rio Formation, a pyritic, gypsiferous and calcareous
shale unit which may represent a barrier to ground water or soil gas
movement.
Site AU-3 lies within a small exposure of altered volcanic tuff of
Cretaceous age, In an area consisting dominantly of Austin Group
limestones. A very thin cover of terrace deposits similar to those at
AU-4 may also be present at AU-3. As at site AU-5, a Cretaceous clay
unit crops out within 100 feet of the AU-3 site, on the opposite side of
a fault passing near the station.
The Edwards aquifer underlying the Austin area is contained within
limestones of Cretaceous age. Depth to water in the Edwards aquifer is
-------
highly dependent on topography, ranging from the land surface In river
^valleys to over 250 feet below It In upland areas.
Elevation of the water table varies by as much as 50 feet over time,
depending on recharge and-pumpage. Local zones of perched water occur
above the Edwards aquifer in areas where Impermeable lithologic units are
present. Ground water was encountered at a depth of 7 feet at sites AU-4
and AU-6, at a depth of 9 feet at site AU-7, and at a depth of. 10 feet at
site AU-5. .
4.1.2 Climate
The climate of Austin, Texas 1s humid subtropical with an average
rainfall, of 20 to 40 inches per year which is evenly distributed
throughout the year. During the first sampling period, September 28
through October 2, the weather was partly cloudy to clear with
temperatures ranging from 62'F to 92'F. The barometric pressure during
this period ranged from 29.49 Inches Hg to 30.09 Inches Hg. The second
sampling period was October 26 to October 30. The same weather patterns
were seen with temperatures ranging from 70*F to 96*F and barometric
pressures ranging from 29.84 inches Hg to 30.12 Inches Hg. Appendix B
contains a summary of the actual field conditions.
4.2 LONG ISLAND SOUND AREA, NEW YORK, RHODE ISLAND
AND CONNECTICUT
4.2.1 Geology and Hydrology - Long Island, New York
Long Island consists domlnantly of glacial till and outwash deposits
representing a terminal moraine formed during the Quaternary Period.
Cretaceous and Tertiary rocks crop out locally in western Suffolk County,
but are not areally significant. All station sites examined for this
project are located In areas of glacial till.
Ground water on Long Island 1s contained within the glacial till and
local alluvial deposits of reworked glacial material. Depth to water
varies from about 10 to 100 feet on the Island. At site NY-2, ground
10
-------
water 1s about 22 feet below the surface. At all other Long Island
sites, ground water Is between 60 and 90 feet below the surface.
' >
". ' *
*4.2.2 Geology and Hydrology - Providence, Rhode Island
In the Providence area, Quaternary glacial deposits of varying thickness
-overlie bedrock of Cambrian and Precambrlan age. As on Long Island,
ground water is found at depths up to about 50 feet in the Rhode Island
glacial deposits. Ground-water conditions are not well known in many
areas because most public water supply. 1s derived from surface sources.
The depth to water at the station sites is not known.
4.2.3 Geology and Hydrology - Storrs, Connecticut
In the Storrs area, Quaternary glacial deposits of varying thickness, up
to about 100 feet, overlie crystalline and metamorphlc bedrock of
Cambrian and Ordovlclan age. Limited quantities of ground water are
found in the glacial fill, but water supply wells generally tap more
extensive reserves In fractures of the Paleozoic rocks. Depth to water
at the Connecticut station sites Is 10 feet.
4.2.4 Climate
The three Long Island Sound locations Included in the study have similar
climatic conditions which are Influenced by the continental and oceanic
weather systems. The average rainfall for these locations 1s from 40 to
60 Inches per year. During the sampling period, September 22 to
September 25 in Suffolk County, the temperature ranged from 61'F to 75*F
with the barometric pressure ranging from 29.70 Inches Hg to 29.94 Inches
Hg. During the sampling visit to Storrs, Connecticut from November 11 to
November 13, the temperatures ranged from 29*F to 51 *F with snow and rain
occurring >W1 November 11 and November 12. The barometric pressure during
this time rangad from 29.65 Inches Hg to 29.99 Inches Hg. The sampling
visit to Rhode Island during the period of December 9 to December 11
experienced one day of rain, December 11, with temperatures ranging from
40'F to 58*F and the barometric pressure ranging from 29.32 Inches Hg to
29.83 Inches Hg. Appendix B contains a summary of actual field
conditions at the time of sampling. Appendix C contains general weather
11
-------
data for the Long Island Sound area for the months of September, October,
November-and December 1987. '.
/
4.3 SAN DIEGO REGION, CALIFORNIA
4.3,1 Geology and Hydrology
The San Diego area -of southern 'California contains two distinct
physiographic sections, a coastal plain section and a mountain-valley
section. " The coastal plain section consists of Tertiary marine
sediments, 1n many parts of which wave-cut terraces are apparent, and
'through which alluvial valleys have been cut between Inland watersheds
and the sea. The mountain-valley section Includes alluvium-filled
valleys dissecting mountain ranges which are comprised of a wide variety
of volcanic, sedimentary, and Igneous rocks.
Station sites SD-1, SD-4, and SD-6 are located in Quaternary coastal
sediments overlain by a thin veneer of Recent alluvium. All three of
these sites are at elevations within a few feet above sea level. Water
was encountered 7 feet below the land surface at site SD-1 and 12 feet
below the land surface at site SD-6. Ground water probably exists at a
shallow depth at site SD-4, but was not encountered during the study.
Stations SD-3 and SD-7 are on a terrace of Tertiary sediments elevated
about 200 feet'above sea level, and are located about 3 to 5 miles Inland
from the sea. Depth to water at stations SD-3 and SD-7 Is not known.
Sites SD-2 and SD-9 are located In valleys near the eastern margin of the
coastal plain section. At these locations alluvium of unknown but
probably shallow depth overlies volcanic or metamorphic bedrock. Ground
water was .encountered at a depth of 8 feet at site SD-2. Depth to water
at site SD-9 Is' unknown.
Sites SD-5 and SD-8 are in a broad valley within the mountain-valley
physiographic section. These sites are located on the residuum produced
by in-situ weathering of underlying volcanic bedrock. Based on
12
-------
Information from wells In the vicinity, depth to water at sites SD-5 and
SD-8 1s probably between 10 and .25 feet. :
-4.3.2 Climate ~J
The coastal location of San Diego, California tempers the climate of this
city. Rainfall In San Diego ranges from 10 inches to 20 inches per year,
with 85% of this .precipitation occurring during the months of Novembec
.through March. During the sampling period, September 15 through
September 24, the temperature ranged from 70*F to 86*F with one day of
slight rain (September 22). The barometric pressure during the sampling
period ranged from 29.90 Inches Hg to 30.10 Inches Hg. Appendix B
contains a summary of actual field conditions at the time of sampling.
Appendix C contains general weather data for the San Diego area for the
months of September, October, November and December 1987.
13
-------
5.0 FIELD METHODS
The field Investigation consisted of on-site sampling and analysis of
.soil gas at a total of 27 service stations In the ฃhree regional areas.
Tracer Research Corporation (TRC) performed the soil-gas sampling and the
on-site analysis of the samples. TRC also performed on-site analysis of
backfill 'samples: for each site to determine soil moisture content.
Geoscience Consultants, Ltd. (GCL) was responsible for overall sampling
strategy and'data quality assurance.
The field work began on September 14, 1987 in San Diego, California and
was completed on December 13, 1987 in Rhode Island. The field schedule
was as follows:
San Diego, CA ' 9 Stations Sept 14 - 24, 1987
Suffolk County, NY 5 Stations Sept 21 - 25, 1987
Austin, TX 4 Stations Sept 28 - Oct 2, 1987
3 Stations Oct 26 - Oct 30, 1987
Storrs, CT 2 Stations Nov 10-13, 1987
Providence, RI 4 Stations December 10 - 13, 1987
5.1 SAMPLING STRATEGY
The sampling strategy was designed to determine the range and spatial
distribution of hydrocarbons within the backfill of the underground
storage tanks. The sampling points were very close to the tanks because
excavation and backfill typically extended only one to three feet
laterally from the edges of the tanks.
Soil-gas samples were collected only from the backfill areas of the tank
excavations. The specific sample sites were located at varying distances
from tank fill ports, pump chambers, and product and vent piping, all of
which can iป sources of leaks. A typical sampling grid consisted of four
or five simple1 holes with samples collected at depths of 2, 6, and 10
feet 1n each hole. Typically, ten to fifteen samples were collected at
each service station. The locations of the sample points are Identified
on the site naps of the stations in Appendix D.
14
-------
Soil samples to determine moisture content of the backfill material were
taken from fifty percent of the sample points. These samples were
analyzed on-site by IRC personnel utilizing a portable oven and balance.
Two soil samples were collected at each station by 'GCL personnel. These
samples were sent to an independent certified laboratory, Professional
Service" Industries, Inc., for the determination of moisture content and
particle .size distribution (sieve analysis). The results of these
analyses are included in Appendix E.
Some additional" sampling other than for soil gas was performed at 5
stations where some unusual conditions existed. This consisted of: 1)
vapor sampling from U-Tube monitoring systems at Stations #4 and #6 in
Suffolk County, New York, and 2) water sampling from shallow ground water
at Stations #1 and #2 in Storrs, Connecticut, and Station #6 in Austin,
Texas.
5.2 SAMPLING METHODS
Soil-gas samples were collected by driving a hollow probe into the ground
to an appropriate depth and evacuating a small amount of soil gas (five
to ten liters) using a vacuum pump. A hydraulic hammer was used to
assist in driving probes past cobbles and through unusually hard soil.
Probes consisted of 7-foot lengths of 3/4-inch diameter steel pipe which
were fitted with a detachable drive point. The above ground end of the
sampling probe was fitted with a steel reducer, a silicone rubber tube
and polyethylene tubing leading to the vacuum pump. Samples were
collected in a syringe during evacuation by inserting the syringe needle
into the silicone rubber evacuation line and drawing a sample from the
gas strean. -
A split spoon device was used to collect soil samples of backfill
material utilizing the probe holes that were used to collect the soil gas
samples. The soil samples were stored in sealed plastic bags prior to
analysis.
15
-------
Promptly uponrtcompletlon of the sampling program at each site, all holes
made in the concrete or asphalt apron were patched to restore the
integrity of the-apron. " ; _
>
r.a ANALYTICAL PROCEDURES
Tracer Research Corporation .used a mobile field laboratory which was
equipped with gas chromatographs. and computing integrators. A flame
ionization detector (FID) was used to measure methane, butane, isopen-
tane, benzene, toluene, ethylbenzene, xylenes and total hydrocarbons.
The methane concentrations measured in the. soil,gas represent a total of
Cj to Cs compounds since it was difficult to identify individual peaks
within this range. In instances when butane and isopentane concentra-
tions were reported, a variation in the temperature program in the gas
chromatograph was used to help clarify these peaks. However, some
interference in peaks was still observed.
Typically, three samples were analyzed from each sampling point and
operator judgement was used in the field to determine which of the
various results could be considered as reliable. Mean values were
calculated in the field based upon experienced operator judgement and
these averages were considered to be representative of the actual soil
gas concentration at the individual sample locations. This type of field
judgement is generally used in soil gas surveys because of the
variability of the soil gas analysis technique and the skill required to
achieve reproducible results. Means derived in this manner were used in
this study in order to provide data that is comparable to existing soil
gas data and to data that can be expected to be obtained in future soil
gas surveys. The actual values of each analysis, which may be useful in
further statistical analyses, are provided in Appendix F.
16
-------
6.0 QUALITY ASSURANCE AND QUALITY CONTROL
The quality assurance/quality control (QA/QC) goals and procedures for
;th1s .project-are described 1n the "Quality Assurance Project Plan for
-Background Vapor Value Study" (QAPP) dated August', 1987 (Appendix G).
Tha majority of-the goals for accuracy, completeness and validity of
data, as listed In ths QAPP, were "attained during field sampling and
^analysis. Because some.of the project activities were modified, during
the course of the" field work, to reflect goals slightly different from
those anticipated In the original Work Plan, certain corresponding
modifications were necessary in the QA/QC procedures.
Additionally, a few field procedures were modified because those outlined
1n the QAPP proved unworkable. These modifications to field methods were
discussed with project personnel and approved by the 6CL QA Officer at
the time of the QA Field Audit, which was performed at two sites 1n San
Diego, California on September 17, 1987. All modifications to the
original QAPP are discussed in Sections 6.1 through 6.12 of this report.
6.1 QA OBJECTIVES FOR MEASUREMENT DATA (QAPP SECTION 3.1)
6.1.1 Gas Chromatograph Analyses '
The gas chromatograph (GC) was calibrated dally by measuring the
instrumental area count for each analyte against the known concentration
of that analyte In a standard gas mixture. The gases, which were
traceable to those of the U.S. National Bureau of Standard, were obtained
from Scott Specialty Gases. The calibration procedure is described in
Section 6.4 of this report.
Because calibration was performed directly from the BTEX gas standard,
the independent accuracy check against another standard was not
feasible. Accuracy checks during the field day were performed against
the same gas standard used for initial calibration. These accuracy
checks, generally two or three per field day, were performed at the
discretion of the analyst. They were consistently performed more
frequently than the .goal of once per 20 analyses which was stated in
Section 3.4 of the QAPP. Area counts for all calibration runs and
17
-------
accuracy checks are'tabulated in Appendix H. All response factors
(RF's) determined by the accuracy checks were within ฑ30% of .those
established at the .beginning'of the field day, so no recall brat ions
during .any-field day+were required. RF's used for each" day/s work are
artso-listed In Appendix H.
In order to- assess analytical precision, all analyses performed for this
project were done in triplicate, by injecting three separate aliquots of
the sample into the GC. In a-few cases, where one of the injections
clearly produced anomalous results, additional injections were made as
necessary to yield three valid analytical runs. For each set of three
analyses for each component at each .sample point, Tracer determined a
mean value which 1s presented in Appendix D, and a standard deviation,
which is presented with the three analytical values in Appendix F. The
standard deviation exceeded 25% of the mean value in 58 out of the 950
triplicate analyses 1n which all three results exceeded the detection
limit, or 6.1% of such analyses. This surpassed the goal, stated 1n the
QAPP, that the standard deviation should exceed 25% of the mean in no
more than 15% of the triplicate analyses. At most points where the
standard deviation was more than 25% of the mean concentration
determined at a point, the analyte was present at a relatively low con-
centration, in which case analytical error 1s normally expected to be a
greater percent of the concentration than for samples in which a greater
quantity of the analyte is present.
At sites where low total hydrocarbon and methane concentrations were
encountered, the detection limits for analytes of interest were normally
less than 0.10 ug/1, and in many cases were less than 0.05 ug/1, the goal
stated In the QAPP. As anticipated, detection limits for all analytes
were much higher In locations where high hydrocarbon concentrations were
encountered. Detection limits for all non-detected compounds are
reported in the accompanying data sets.
18
-------
6.1.2 Soil Moisture Content Analyses
Due to sampling and analytical problems encountered in the field, Tracer
reported fewer .soil moisture analytical results thap were anticipated in
-the QAPP. Sample splits,-and in some locations 'the majority of soil
samples, were sealed in air-tight containers and submitted by GCL to
Professional Service Industries, Inc. (PSI) in Albuquerque, New Mexico
for moisture .content analysis. PSI submitted results for 42 soil
samples, and Tracer submitted results for 26 samples. Because of
inconsistent sample identification, particularly in New York and Rhode
Island, 1t was not always possible to identify which Tracer samples were
in fact splits of PSI samples.
Table 6-1 lists and compares all soil moisture replicate analyses
identified in a review of the Tracer and PSI data. In most cases, the
laboratory values agree well with those obtained in the field, but
significant discrepancies exist for the data at sites AU'2 and SD-2.
There is good internal consistency among the values reported for
replicate samples which were both sent to the PSI lab.
6.2 SAMPLING PROCEDURES (QAPP SECTION 3.2)
Soil gas sampling was performed as described 1n the QAPP. At the request
of EPA EMSL, sample points were confined to the area of the backfill
immediately adjacent to the USTs at each site, and in a few cases to soil
just outside the backfill. There were generally no more than 6 sample
points per site, and samples were normally taken from 3 depths at each
point.
A total of 78 soil samples, mostly backfill material, were analyzed for
moisture content. The samples were not uniformly distributed among the
sites because of difficulties encountered in obtaining soil samples at
some locations and the realization that moisture content was of little
utility in others, such as sites where the backfill material consisted of
pea gravel. The values reported In this document represent only samples
that were properly packaged, transported and analyzed. .
19
-------
TABLE 6-1
RESULTS OF REPLICATE ANALYSES FOR SOIL MOISTURE
:-- CONTENT ;
" >
CALL ANALYTICAL VALUES IN PERCENT BY WEIGHT)
SITE
AU-1 .
AU-2
SD-2
NY-2
NY-4
NY-5
NY-6
TRACER
SAMPLE NUMBER
870.9291807
8709300935
8709161636
NY2-SG4-10
NY4-SG4-10
NY5-SG4-10 '
NY6-SG2-10
PS I
- SAMPLE NUMBER
8709301819
8709301825
8709300940
8709300946
8709161637
8709231230
8709241545
8709241600
8709251310
8709251800
8709251830
TRACER
ANALYTICAL
VALUE
14.7
12.4
11.3
t
.10.0
5.0
6.9
5.7
PSI
ANALYTICAL
VALUE
13
11
4
3
20
7
3
,5
8
5
6
!
REMA'RKS
Erroneous date
on samples
delivered to
PSI. Two
replicates to
PSI.
Two replicates
to PSI.
Correlation
uncertain. Two
replicates to
PSI.
Correlation
uncertain.
Two replicates
to PSI.
20
-------
6.3 SAMPLE CUSTODY (QAPP SECTION 3.3)
Chain-of custody procedures described in the QAPP were followed for all
soil samples sent to PSI for moisture content or sleeve analysis. Chain-
.of-custody forms -for these samples ^are on file at the GCL office in
Albuquerque.
6.4 CALIBRATION PROCEDURES AND FREQUENCY
(QAPP SECTION 3.4)
The GC was calibrated daily, using gas standards obtained from Scott
Specialty Gases. These standards are traceable to those of the U.S.
National Bureau of Standards. Two separate three-point calibration
curves were established, as described in Section 3.4 of the QAPP, one for
methane (hydrocarbons Cj-Cs) and one for the aromatic hydrocarbons Cs-Cg.
However, the curve used to quantify hydrocarbons Cs-Cg was established
using the BTEX gas standard rather than an aqueous standard. It was
found that this procedure yielded accurate and replicable results, while
the aqueous standard produced a response factor (RF) that did not
accurately quantify the gaseous BTEX standard. Additional calibration
and accuracy checks were made periodically during each field day, and
RF's were then revised as necessary. Recalculation of RF's during the
field day was not found to be necessary at any site. Area counts and
response factors as reported by Tracer are shown in Appendix H.
Isopentane was not originally Included among the compounds to be
specifically Isolated under the original Work Plan and QAPP. However,
GCL and Tracer were subsequently requested by the EPA to attempt a
determination of Isopentane concentrations at selected locations. Since
no .standard for Isopentane had been provided in the field, Isopentane
values were determined after field work was complete by reanalyzing the
chromatograms to Identify the Isopentane peak. A response factor (RF)
for Isopentane was defined by comparison with the known RF for benzene, a
gas which had been Included among the standards available in the field.
21
-------
To assure the cleanliness of sampling equipment, syringe blanks and
.system blanks (air samples) were taken and analyzed each morning and
periodically during the day, as provided In the QAPP;.
>
6.5 ANALYTICAL PROCEDURES (QAPP SECTION 3.5)
Analytical Procedures are described In Section 5.3 of this report. All
soil gas analyses for benzene, toluene, ethylbenzene and xylenes (BTEX)
and for total hydrocarbons were performed by Tracer personnel In
accordance with the procedures described 1n Section 3.5 and Appendix B of
the QAPP, except for the treatment of samples yielding total hydrocarbon
values greater than 500 ug/1. .Experience during the first day of field
work Indicated that reducing the Injection size for such samples, as
proposed In the QAPP, resulted In obscuration of the chromatogram peaks
for hydrocarbons Cg-Cg (gasoline constituents), while not significantly
Improving the accuracy of methane measurements. Since the use of
smaller Injection sizes resulted in a great loss of data, the practice
was discontinued.
6.6 DATA REDUCTION, VALIDATION AND REPORTING
(QAPP SECTION 3.6)
Data presented to GCL by Tracer were recorded and analyzed as described
in Section 3.6 of the QAPP. The results of the analyses performed are
described elsewhere In this report.
Some extreme values ("outliers") Identified in the original data
recorded on site were discarded from the data set by Tracer because the
on-slte chemist, based on his field experience, believed them not to be
representative of actual hydrocarbon concentrations in the sample
analyzed (see Section 6.10 of this report). Consequently, GCL has made
no attempt, to identify or explain the few outliers remaining 1n the data
set, which would require excessive time and yield little Information.
The data presented 1n this report have been subjected to Tracer's
Internal review*process, and have been spot-checked for accuracy by.GCL
personnel. Although a few minor errors were detected arid corrected
22
-------
during the GCL review, and a few others undoubtedly remain in the large
data set, GCL Is confident that such errors represent a very minor
portion of the total body of data. ;
. > ~
6."7-'INTERNAL QUALITY CONTROL CHECKS (QAPP SECTION 3.7)
GC calibration procedures 'and frequency were described in Section 6.4 of
this report!' As a standard part of .Tracer's analytical procedure, daily
"blanks" consisting ofjpjure. nitrogen, of-a.fr, and of air drawn through a
soil gas.probe and .adapter ("system blank") were analyzed. These blanks
were repeated as necessary during the field day, and specifically after
any event which was suspected may affect analytical results.
Soil gas samples at each point were analyzed in triplicate, as described
1n Section 6.1.1. of this report, and duplicate soil samples for
moisture content analysis were taken at selected points, as described in
Section 6.1.2. Replicability of results was within the goals
established by the QAPP.
6.8 PERFORMANCE AND SYSTEM AUDITS (QAPP SECTION 3.8)
A field system audit and evaluation of operational procedures was
performed in San Diego on September 17, 1987 by the GCL QA Officer.
Minor modifications to field sampling and analytical procedures were
discussed with project field personnel and approved by the QA Officer at
that time. A letter report describing the results of the field audit
was submitted to CDM FPC on September 18, 1987, and is Included in
Appendix I of this report.
-6.9 PREVENTIVE MAINTENANCE (QAPP SECTION 3.9)
All equipment was maintained In operable condition during the field work.
Spare parts and new equipment were obtained as necessary to complete
field work In a timely manner.
23
-------
6.10 ASSESSMENT OF DATA PRECISION, ACCURACY AND
COMPLETENESS (QAPP SECTION 3.10) :
The data presented in :this report are complete in the sense that all
values believed to represent valid analyses have been fncluded. Gas
chTomatographic analysis -.ts a procedure 1s subject to Interpretation by
the GC operator, who -'must evaluate each run, on the basis of his
experience, -to determine Its validity. Volume of sample Injection,
concentrations of the analytes of Interest,- and possible residual effects
of previous sample runs must be considered by the operator in deciding
whether to accept the concentrations Indicated for any given sample
Injection. Concentration values which were clearly. 1n error were rejec-
ted by the GC operator in the field, and are not included in the data
set. Some other values which appear to be "outliers" Inconsistent with
" f .
the rest of the data Set have been included in the tabulated analytical
results (Appendix F), but were not used in determining the mean values
of the triplicate analyses reported in Appendix D. In some of these
cases, the outlying values were excluded by Tracer in calculation of the
mean concentration, but were Included 1n calculation of the standard
deviation. GCL and Tracer have attempted to Indicate such points where
such operator judgment was exercised. These undoubtedly represent far
less than 1% of the total data set.
During the course of the project, Tracer was asked to recalculate the
total hydrocarbon concentrations to show them relative to the BTEX
total, rather than as benzene. Consequently, the mean values used in the
data analysis (Section 11.0) for total hydrocarbons (less methane) differ
from the means of the values shown in Appendix F (Individual GC-FID
injections). The standard deviations for the total hydrocarbon data were
calculated on the basis of the values reported "as benzene", and
consequently should not be applied directly to the total hydrocarbons
(less methane) data calculated from average daily response factors for
BTEX.
24
-------
Concentration values reported in micrograras per liter for analytes of
interest in this report .are normally given to two significant figures if
greater than 10 ug/1, .and to one significant figure ;if less than 10 ug/1.
As illustrated by the standard deviations presented1 with this data set,
and based on Tracer's experience in soil gas analyses, instrumental
precision does not normally justify greater precision in the reporting of
results.
Further information regarding analytical accuracy, precision and
replicability was presented in Section 6.1 of this report.
*6.11 CORRECTIVE ACTIONS (QAPP SECTION 3.11)
During the field system audit, the requirements for proper chain-of-
custody procedures were explained to some site personnel who were not
fully aware of them. Samples previously taken for soil moisture content
analysis had been properly handled, but the QA Officer felt that
additional explanation was necessary to prevent the possibility of future
problems.
No other corrective actions were found to be necessary during field work.
Problems with Tracer's handling procedure of the soil moisture samples
were discovered too late to be remedied by GCL personnel.
6.12 QUALITY ASSURANCE REPORTS TO MANAGEMENT
(QAPP SECTION 3.12)
Monthly quality assurance reports were submitted during the course of
the project, as described in Section 3.12 of the QAPP.
25
-------
7.0 REPORTING METHODS
One of the problems encountered In this study concerned the calculation
.and reporting of the total hydrocarbon concentration data. Different
.practices In calculating .and reporting these data were discovered within
'the environmental Industry and among those who collect and analyze soil
gas data". For example, some leak'detection device's were found to report
total hydrocarbons In parts per million by volume (ppmv) "as hexane", and
others 1n ppmv "as butane" (Radian). Additionally, laboratories using
gas chromatograph, flame ionization detection (GC/FID)-equipment to
analyze soil-gas, report total hydrocarbon concentrations in mlcrograms
per liter (Tracer). The method of determining total hydrocarbon
concentration values using a GC/FID also vary. A GC/FID must use a
response factor based on the calibration of a known gas to determine the
concentration of an unknown gas. This calibration gas, or "gas standard"
may be benzene, toluene, or some other hydrocarbon compound.
Because of these variations, GCL evaluated different calculation methods
to determine the most appropriate method for reporting total hydrocarbon
concentrations. In this method evaluation, both the calculations and
their accuracy were examined. Since these data may be used in developing
threshold limits between non-contaminated and contaminated sites, they
must be comparable to soil gas data determined by different methods.
The evaluation consisted of two parts:
ซ Calculation of total hydrocarbon concentrations In mlcrograms
per liter from the calibration of the GC/FID, reported both "as
benzene" and according to an average response factor, and
ซ Calculation of total hydrocarbon concentrations in parts per
1111oh.
t
7.1 DETERMINATION OF TOTAL HYDROCARBON CONCENTRATIONS
IN HICROGRAHS PER LITER
The field Investigation phase of this study required that soil gas
samples be collected and analyzed at non-contaminated sites. These
samples were analyzed on-site using a portable Gas Chromatograph with a
26
-------
flame ionlzatlon detector (GC/FID). The results of these analyses yield
concentration values in micrograms per liter. Section 7.1.1 contains a
brief discussion on "the function of a GC/FID and^he procedure used to
calculate the total hydrocarbon concentrations from the GC/FID in the
field. _ This procedure uses benzene as the calibration gas. Section
7.1.2 discusses ~a more accurate method used to calculate total
hydrocarbon concentrations in micrograms per liter using data from alj
- the calibration gases. -
7.1.1 Gas Chromatograph and Flame lonization Detector
Operation
A gas chromatograph is (GC) an analytical instrument that can be used to
separate volatile organic compounds for analysis (EPA Methods 8000). A
GC equipped with a flame ionization detector (FID) can be used to
generate a chromatogram that consists of peaks corresponding to different
compounds. The complete analytical system used in the field
Investigation of this study consisted of a chromatograph1c packed column
containing All tech OV101, a hydrogen flame lonization detector, an
integrator-recorder, calibration gases and glass syringes (Tracer).
Calibration gases were used to generate a chromatogram that formed a
base-line or standard of peaks in the chromatogram. Response factors,
defined as the ratio of the mass of each gas standard to the Integrated
area of the peak produced by that mass, were determined for each gas
standard. Individual hydrocarbon compounds in the soil gas samples were
identified by a comparison of sample chromatograms to the standard
chromatogram. Concentrations of Individual compounds were calculated
from the response factors for the corresponding gas standard.
Concentrations
-------
sample, and consequently, response factors and concentrations are
measured 1n mass units (Tracer).
'
'The calibration gas standards -used were methane^, benzene, toluene,
'ethylbenzene,, and ortho-xylene. Concentrations of each of these
compounds ~1n each- sample were calculated directly using the corresponding
calibration* gas response factor and the sample Injection size. However,
concentrations for total hydrocarbons (less methane) were required to be
approximated.
7.1.2 Calculation of Total Hydrocarbons as Benzene
During the field Investigation, total hydrocarbon (less methane)
concentrations were approximated by using the response factor for benzene
to compute the concentrations. During the data analysis, 1t was
discovered that this approximation yielded a low estimate of total
hydrocarbons (less methane) concentrations. This discovery was made by a
comparison of the combined concentrations of benzene, toluene,
ethyl benzene and xylenes (BTEX) to the total hydrocarbon concentration
(less methane). This comparison shown in Appendix J, Indicates that the
concentration of BTEX was greater than the concentration of total
hydrocarbons (less methane) In 30 percent of the samples.
ซ
A possible cause for the discrepancy between the concentrations of total
hydrocarbons (less methane) and BTEX could have been an erroneous
Interpretation of the chromatogram peaks. However, a re-examination of
the chromatograms showed that no Interpretation errors had occurred.
'The discrepancy was determined to be the result of using the benzene
response factors for the approximation of total hydrocarbon (less
methane) concentrations. By an examination of the response factors for
all of the gas standards (Appendix D), It was found that the benzene
response factor was usually lower when compared to response factors for
toluene, ethylbenzene and ortho-xylene. In theory, response factors for
similar hydrocarbon compounds should be similar. However, in practice,
response factors vary because of chemical and Instrument effects.
28
-------
Because of the discrepancies between the total hydrocarbon (less methane)
concentrations and the combined BTEX concentrations, a better
approximation of total hydrocarbon (less methane) concentrations was
needed. "This-was considered Important because these valued obtained from
nor;-.contaminated sites may affect the development of threshold limits to
be used to distinguish Between contaminated and non-contaminated sites.
7.1.3 Calculation of Total Hydrocarbon Concentrations
Using Average Response Factors
The total hydrocarbon concentration in a soil gas sample is actually the
summation of all the hydrocarbon compounds that can be detected from the
GC/FID analysis. To accurately determine this concentration would
require that a gas standard be analyzed in the GC/FID for every compound
that existed in the soil gas. This comprehensive type of analysis was
considered impractical since an enormous amount of GC/FID calibration
work would have been necessary to quantitatively analyze all.the peaks in
the soil gas samples.
The best approximation, based on the available calibration data, was to
determine total hydrocarbons (less methane) using the average of the
response factors for all the calibration gases (less methane).
Therefore, total hydrocarbon (less methane) concentrations were
calculated from an average of the daily response factors for benzene,
toluene, ethyl benzene and ortho-xylene.
This approximation resulted in new total hydrocarbon (less methane)
concentrations that were generally higher. A comparison of total
hydrocarbon (less methane) concentrations calculated .from average BTEX
response factors and "as benzene" 1s shown below.
' i " .
TOTAL HYDROCARBON (LESS METHANE) PERCENTAGE OF
CONCENTRATIONS SAMPLES
As Benzene > As BTEX Average 8.6%
As Benzene - As BTEX Average 15.1%
As Benzene < As BTEX Average 76.3%
29
-------
In the case where the new values (as BTEX average) were greater than the
old values (as benzene), these new values ranged from 7% to about 100%
higher, ปA-comparison of the old values and new values for^ each sample Is
provided 1n Appendix 0.
The new concentrations also result in values that are larger than the
combined BTEX concentrations which, indicates a more reasonable approxi-
mation of total hydrocarbon, concentration. "A comparison of .the BTEX and
the new. total hydrocarbon (less" methane). concentrations are shown in
Appendix J.
The calculation of total hydrocarbon (less methane) concentrations using
the average BTEX response factors,-was found to be a better approximation
than when using only'benzene, because it accounted for variations in the
response factors. However, 1t 1s understood that some error still exists
1n this method because several peaks in the chromatograms and their
corresponding compounds were not identified and quantified.
To better understand the extent that compounds other than BTEX are
contained in total hydrocarbons, a comparison of the combined BTEX
concentrations to total hydrocarbons (less methane) concentrations
(calculated from average BTEX response factors) was made. These results
are shown In Figure 7-1. The tabular data used to generate this figure
Is included In Appendix J. The percentage of samples where the BTEX
concentrations were less than 50 percent of total hydrocarbons (less
methane) was about 59 percent of the total samples. This means that in
about 59 percent of the samples, compounds other than BTEX make up the
majority of Jthe total hydrocarbon concentrations.
The result that compounds other than BTEX make up the majority of the
total hydrocarbon concentration In most of the samples is not surprising
when the composition of gasoline is considered. A typical gasoline
contains several hundred hydrocarbon compounds, each falling Into one of
four chemical groups: paraffins,, olpfIns, napthenes or aromatlcs (NM
EID). The aromatlcs, which 1nclude.s BTEX, are considered most Important
30
-------
O
I
O
m
I
P
RATIO OF BTEX TO TOTAL HYDROCARBONS VS
CUMULATIVE PERCENT OF SAMPLES
FIGURE 7-1
CUMULATIVE PERCENT OF SAMPLES
100
-------
because they are relatively soluble In water, and therefore, present a
risk of ground-water -untamination. Table 7-1 shows a list of major
components of an API PS-6 Gasoline, some of which can be expected to be
present 1n soil gas. These compounds represent 04-'to CJQ molecules (API-
1985). :-,..--
Some selected sample chromatograms from Suffolk County, NY, San Diego, CA
and Austin, -TX were qualitatively analyzed for a wide range of compounds
where BTEX was found to represent less than 10 percent of the total
hydrocarbon-concentration. These qualitative analyses Identified some
additional compounds: methane, butane, Isopentane, 2-methylhexane, iso-
octane, and octane. These chromatograms are shown in Appendix J.
7.2 DETERMINATION OF TOTAL HYDROCARBON CONCENTRATIONS IN
PARTS PER MILLION
The concentration of extremely dilute solutions are expressed In parts
per million (ppm). Typically, liquid solutions are expressed 1n parts
per million by weight (ppmw) and gaseous solutions are expressed In parts
per million by volume (ppmv). (Himmelblau)
Parts per million by volume (ppmv) Is a measurement unit that 1s commonly
used in the environmental Industry for reporting air pollutant
concentrations '(Hark and Warner). Many leak detection systems report
hydrocarbon contamination in soil gas in ppmv (Radian). Therefore, parts
per million by volume was considered appropriate rather than parts per
million by weight.
Ppmv 1s defined as:
"-.-. : '-..ซ'
'ippmv- 1 volume of gaseous pollutant Equation!
106 volumes of pollutant & air
32
-------
TABLE 7-1
MAJOR COMPONENTS OF API PS-6 GASOLINE
)
PERCENT WEIGHT
2-Methylbutane a 7?
M-Xylene 5 66
2,2,4-Trlmethylpentane 5*22
Toluene 4*73
2-Methylpentane 3*93
N-Butane 333
1,2,4-Trimethy!benzene 3*26
N-Pentane 3*U
2,3,4-Trimethylpentane 2 99
2,3,3-Trlmethylpentane 2 85
3-methylpentane 2*36
0-Xylene 2*27
Ethyl benzene 2*00
Benzene _7 Q*
P-Xylene 17?
2,3-D1methylbutane } 66
N-Hexane J'JQ
1-Methyl, 3-Ethylbenzene 1*54
1-Methyl, 4-Ethylbenzene 1*54
3-Methylhexane 1.30
33
-------
The data in micrograms per liter can be converted to ppmv by the
following equation: . :
ppmv
RT
.P. (Mol Wt)
Equation 2
where:
ppmv
ug/i
R
P
T
Nol Wt
Parts Per Million by Volume
Micrograms Per Liter
Gas Constant - 0.08205. atm liter
Pressure In Atmosphere
Temperature 1n *K
Molecular Weight of Hydrocarbon
This equation was derived from the Ideal gas equation:
PV - nRT
where:
V
n
R
Pressure
Temperature
Volume
Moles
Gas Constant
Equation 3
The derivation is shown in Appendix J (Wark and Warner).
The temperature and pressure used in these calculations represented the
ambient conditions measured in the field at each site.
The assumption of an ideal gas was justified by examining a mean
compressibility factor. The mean compressibility factor is a factor that
is introduced into the ideal gas equation to account for non-Ideal or
real gas relationships. Therefore, the ideal gas equation becomes:
PV
ZmnRT
where:
Equation 4
Zm - mean compressibility factor
34
-------
If calculations can show that Zm Is approximately equal to one for the
soil gas mixtures, then the assumption that the soil gas samples in this
;study can be approximated to an-ideal gas is valid qne. (Himmelblau)
. - . . > %.
Twq_ cases were examined in testing this assumption.. Because the complete
composition of soil gas is not known, Case 1 assumed soil gas contains
80% air and Case 2 . assumed soij.gas contains 20% air. The mean
compressibility factor was determined to be 0.99 for Case. 1 and 0.85 for"
Case 2. Therefore, the ideal gas assumption introduces about 1 to 15
percent error in calculating hydrocarbon concentrations in soil gas.
This small deviation (1 to 15%) from the ideal gas assumption is
reasonable since the pressure conditions are low, and the hydrocarbons
in the mixture are similar In their chemical nature.
The conversion calculations from micrograms per liter to ppmv were made
for each sample and each compound within that sample. The molecular
weight of each compound was used in the conversion calculation. However,
for total hydrocarbons (less methane), an average molecular weight was
used. This average molecular weight was based on the average of the BTEX
concentrations at each sample.
To compute total hydrocarbons (with methane), the methane concentration
was converted to ppmv and then added to total hydrocarbons (less methane)
in ppmv. In these calculations, the detection limits were divided by 2
to approximate the actual concentration. A sample calculation is shown
in Appendix J.
The average of the BTEX concentrations was used to compute the average
molecular weight of each sample since BTEX concentrations were known at
all sample points. It is recognized that some error Is Introduced by
using only BTEX concentrations. However, this 1s considered to be the
best approximation possible from the available data. Reporting
hydrocarbon concentrations in parts per million may be useful for some
purposes. However, reporting them 1n micrograms per liter provides more
accurate values based on fewer assumptions.
35
-------
8.0 RESULTS
8.1 SOIL GAS DATA
The .maximum soil gas concentration values determined in this study are
presented 1n Table 8-1 for the sites In Austin, Table 8-2, for the sites
in the Long Island Sound area and Table 8-3 for those In the San Diego
area.
Average hydrocarbon vapor concentration data and site maps for all 27
gasoline service stations are presented In Appendix D. The average
hydrocarbon vapor concentration data, in most cases represent mean values
for each set of three gas chromatograph/flame ionization detection
(GC/FID) analyses for each sample. These data are presented in two
formats: 1) concentration values listed by sample number and depth, and
2} concentration values listed by depth and sample number. In the second
format, computed average concentrations for all samples at each depth are
shown. Additionally, each site map contains an average total hydrocarbon
concentration computed from concentrations at each depth within each
hole. In computing these average concentrations, the concentrations
reported at detection limits were divided by two to approximate the
actual concentration.
A pipeline was accidentally punctured during the investigations at
Station 16 In Austin, Texas. Data were collected during four consecutive
days at this station to study soil gas migration under dynamic
conditions. These data are also Included in Appendix D.
Data in Appendix D presented both in microgram per liter and parts per
million by volume. ---
8.2 CONTAHINATtD SITE DATA
Soil gas surveys were previously conducted at a number of UST sites in
which product spills were known to have occurred. Data from 27 sites
were examined as candidates. Of these sites, 8 were selected as being
appropriate for comparison purposes because site maps were available and
contamination was known to exist. Data collected from Austin Station #6
36
-------
TABLE 8-1 .
MAXIMUM CONCENTRATIONS AT AUSTIN, TEXAS
(All concentration values in micrograms per liter)
Station 1
Station 2
Station 3
Station 4
Station 5
Station 6
10/27/87
10/28/87
10/29/87
10/30/87
- METHANE
(AS METHANE1
.790.0001'
210,000 '
120,000
870.000
1.500,000
710,000
8,600
13,000
4.800
TOTAL" HYDRO-
~ BENZENE-
7,400
ie,ooo _;
3.300
97.000
24,000
*
110,000
27,000
<250
53,000
TOLUENE ETHYLBENZENE XYLENES
5,300 . <310
. 17,000 . 160
1.700 <63
85.000 <680
26.000 25.000
90.000 <220
83.000 <250
<290 <270
1.600 <20
2.300
21.000
410
83,000
8,200
<240
70.000
<260
<31
CARBONS TANK TIGHTNESS
(LESS METHANE1 TEST RESULTS
21.000
63,000
5.700
210.000
1.100,000
960,000
790.000
690,000
290.000
Tight
Tight
NR
NR
Tight
Station 7 59,000 <42 <48
Notations:
NAZ - Not Analyzed.
NR ซ No records available showing tank tightness results.
Notes:
<50
<58
55.000
Tight
(2) Total hydrocarbons are ealcutated from average response factors for Benzene. Toluene. Ethyibenzene and Orthoxytene.
(4) SpiH occurred at 8:00 AM on 10/27/B7. These data were collected after the spffl.
(5) At stations where C^/Cg are not analyzed, the methane concentration represents CrC5 peaks.
(6) Tight means petroUte test results were < 0.05 gaRons per hour.
37
-------
TABLE 8-2.
MAXIMUM CONCENTRATIONS LONG ISLAND SOUND COASTAL AREA
(All concentration values in micrograms per liter)
METHANE >
TOTAL HYDRO-
Station 1
Station 2
Station 4
Station 5
Station 6
Station 1
Station 2
Station 1
Station 2
Station 3
Station 4
1^-05)
. fAS METHANE)
<40
140
<24
4
15
25,000
11,000
8
72
9
2,800
- BENZENE .TOLUENE
2,700 11.000
<29 420
3.700 1.000
2,300 13,000
<.6 55
<10 840
<6 <6
<.1 110
23 230
<.08 0.8
670 1.400
CARBONS
ETHYLBENZENE XYLENES (LESS METHANE)
12.000 10.000 270,000
130 <41 2.100
<37 <42 . 69.000
2.900 91 110.000
<.7
-------
t
^
V
Station 1
Station 2
Station 3
Station 4
Station 5
Station 6
Station 7
Station 8
Station 9
TABLE 8-3:
MAXIMUM CONCENTRATIONS AT SAN DIEGO, CAUFORNIA
(All concentration values in micrograms per liter)
-..';'.--. >
-" ^METHANE
(AS METHANE^
48.000
110,000
'22
420.000
55.000
33,000
390,000
21,000
280,000
BENZENE
<89
<89
ฐ
<90
<86
<83
<90
<91
<98
''_
TOLUENE
11.000
11.000
17
17.000
2,600
23.000
31.000
22.000
32.000
ETHYLBENZENE XYLENES
<120 4.900
<120 5.100
<.05 .8
<-1 1,800
<-1 1,600
<-1 10,000
<.1 8,800
<.1 8.600
<-1 8,200
TOTAL HYDRO-
CARBONS
(LESS METHANE)
31.000
77,000
62
110.000
7.700
58,000
210,000
120,000
110,000
TANK TIGHTNESS
TEST RESULTS
Leak
Tight
Leak
Tight
Tight
Tight
Tight
Tight
NR
Notations:
NAZ - Not Analyzed.
NR - No records available showing tank tightness results.
Notes;
B>
(4) At stations where C^Cg era not analyzed, the methane concentration represents CrC5 peaks.
(5) Tight means tightness test results were <0.05 gafons per hour.
39
-------
was Included as Site 9 since data from this station represents a fresh
spill.
. , /
Table 8-4 gives a brief description of these 'sites and Table 8-5
presents the maximum concentration data for them. These sites Include
active service stations'"or fueling facilities. Site maps and data.
arepresented in Appendix K. Specific sample locations at these sites
were selected for use in the contaminated site database because of their
close proximity to the tanks or contamination source. It was-desirable
to use sampling points close to the tanks so that the data would be
comparable to the clean site data collected from the tank backfill areas
under this study. A summary of the soil gas data is Included in
Appendix K. Total hydrocarbon values are reported less methane, and "as
benzene*.
8.3 EXPANDED AUSTIN STUDY
A four-day study was conducted at Austin Station #6 to take advantage of
a spill that occurred when a product line was punctured during the field
Investigations. Soil gas samples were taken from the same holes each day
and the results are included in Appendix D. Figure 8-1 shows the
concentration of total hydrocarbons for each of the four days at 2-foot
and 6-foot depths, and Figure 8-2 shows the corresponding concentrations
of C4-C6 components.
This Intensified study provided the following basic Information:
Total hydrocarbon concentrations Increased Initially to
>100,000 ug/1 near the spill site and higher concentrations
migrated Into the entire backfill area.
Total 'hydrocarbon concentrations decreased after peaking one
day after the spill.
High concentrations of C4-C6 components were found to parallel
the total hydrocarbon concentrations.
Since high concentrations of C4-C6 components were not usually
encountered in the field sampling at clean stations, it may be
40
-------
TABLE 8-4
.DESCRIPTION OF CONTAMINATED SITES
">
tSite 1 New Service Station. Tanks were tested
tight, but found floating product in
-ground water. Ground-water depth - 8'.
Site 2 Active Service Station.
Site 3 Active Service Station. Floating product
in ground water. Ground-water depth -
15' - 20'.
Site 4 Active Fueling Facility. Pipeline leak.
No ground-water contamination. Ground-
water depth = >20'.
Site 5 Active Fueling Facility. Ground-water
depth - 12'.
Site 6 Active Service Station. No ground-water
contamination. Ground-water depth - 15'.
Site 7 Active Fueling Facility.
Site 8 Active Service Station. Floating product
on ground water. Ground-water depth -
25' - 35'.
Site 9 Active Service Station (Austin #6). Spill
resulting from product like puncture.
f!fiig: I1??56 /1tes were selected from Tracer Research Corporation
files to develop database of hydrocarbon vapor concentrations
for sites with known hydrocarbon contaminated.
41
-------
TABLE 8-5 ;
MAXIMUM CONCENTRATIONS AT CONTAMINATED SITES
(All concentration values in micrograms per liter)
Station
Station
Station
Station
Station
Station
Station
Station
1
2
3
4
5
6
7
8
METHANE
"-
1,200,000
NAZ .
NAZ :
NAZ
NAZ
NAZ
NAZ
100,000
BENZENE
.100,000
: <1ฐ; : -'
:. NAZ
780
28,000
<230-
<55
60.000
TOLUENE
68.000
1,200
31.000
620
11.000
4.000
1,700
40.000
ETHYLBENZENE
61.000
120
NAZ -
50
-------
TOTAL HYDROCARBON CONG (ug/I)
(Thousands)
\
NJ
d
-n
ซ
tn
ya
m
oo
>
o
o
to
o
o
u
o
o
V
NJ
I
OJ
..!._.
01
o
o
_L_
o>
o
o
/
V * v * V \
.. ^
S/777-.77',
d
CO
ZJ
r
en
m
H
m
:r
o
o
CU
O
O
I
-------
frfr
N>
n
i
at
I
oo
ro
/
q>
o
C4-C6 CONG (ug/I)
(Thousands)
o
o
NJ
o
o
u
o
o
g
N)
I
04
R\5^
f$$,, ,,.,,,,,
'//y/////y//^/'/^///
o
i
cn
o
o
//
^fflUH
at
o
o
^
O
d
CO
-H
m
o
2:
o
|
g o^
rn
n:
a
0
o
QJ
o
--P
-------
possible to use C4-C6 concentrations, as compared to those of total
hydrocarbons, to detect fresh leaking conditions. More study is reauired
:to confirm this preliminary indication.
45
-------
8.4 CHARACTERIZATION OF BACKFILL MATERIAL
Soil moisture and particle size of the backfill materials impacts
hydrocarbon vapor concentrations because of liquid/vapor partitioning and
porosity effects. Consequently, soil moistures and sieve analyses were
performed 011 soil samples collected from the backfill of the non-
contaminated sites. A summary of the results of these sample analyses
are presented in Table 8-6. Laboratory analyses are included in
Appendix E. ' -
a* - *
Backfill' soil material at steel tank installations included fine, medium
and sllty sands while the backfill at fiberglass tank installations were
of fine gravel, gravelly sand and coarse sand mixed with gravel.
Moisture contents were higher in the sands than in the gravels and the
porosities of the sands were less than those of the gravels.
Because gravel 1s more porous and less moist, hydrocarbons will likely
move more quickly through gravel backfill than through sand. Also,
moisture will tend to Inhibit the movement of hydrocarbons and will
absorb hydrocarbons through liquid/vapor partitioning.
8.4 U-TUBE SAMPLING
Leak detection methods are classified into four groups: Volumetric,
Nonvolumetric, Inventory Control, and Leak Effects methods (EPA).
Methods within the Leak Effects classification are those that identify
leaks by examining the environmental effects of the leak. Those methods
usually require the Installation of monitor wells and chemical analysis.
Since soil gas contamination 1s an environmental effect that can result
from a leaking 1)ST system, then soil gas sampling, as performed in the
field Investigation of this study, would be classified as a Leak Effects
method.
Another method for monitoring leaks within the Leak Effects
classification utilizes a U-Tube device. The U-Tube consists of a four-
Inch diameter, schedule 40, PVC pipe installed as shown in Figure 8-3.
46
-------
TABLE 8-6
MOISTURE RANGES OF SOIL AND BACKFILL SAMPLES
(Values in percent by weight. Moisture content analyzed by PSI,
'Albuquerque, NM)
TANK
LOCATION/5TATIQM TYPE
AUSTIN/TEXAS
AU1 Steel
AU2 - 'Steel
AU3 FRP
AU4 FRP
AU5 steel
AU6 FRP
AU7 FRP
STORRS, CONNECTICUT
CONN1 steel
CONN2 steel
PROVIDENCE, RHODE ISLAND
RH Steel
RI2 steel
RI3 steel
Rlซ Steel
SUFFOLK COUNTY. NEW YORK
NY1 FRP
NY2 ' steel
NY4 FRP
NY5 steel
NY6 FRP
SAN DIEGO COUNTY. CA
SD1 steel
S02 steel
SD3 - FRp
SD4 .. steel
SD5 FRP
SD8 FRP
SD7 steel
S06 steel
SD9 steel
MOISTURE CONTENT
SAND
11-13
3-4
4-13
GRAVEL
NATIVE SOIL
10
11
79 '
5
1-15
15
10
4
4
8
54
15-17
7-8
8-7
3-10
11
SIEVE
ANALYSIS RESULTS
Sitty sand
Sandy gravel
Gravefly sand
Medium sand
Fine gravel
Fine sand
Medium sand with silt
Fine sand
Medium to fine sand
Fine sand
Fine sand with silt
Fine sand with silt
Crs sand with gravel
Medium sand with silt
Medhm sand with sift
SWy sand
NOTE: All Sieve Analysis results from backfill samples.
^
* Native Soil Sample taken from saturated zone in bottom of monitor well.
47
-------
FINISHED
-GRADE
OVERFILL
PREVENTION
DEVICE WITH EXTRACTABLE
TEE TO GRADE
\
OBSERVATION WELLS: WATERPROOF CAPS
CAPABLE OF BEING
SEALED
EXTENSION OF
MANWAY TO GRADE
(OPTIONAL) -
/
4'TEE-*-
SEALED
CAP
NOTE: ALL PIPING
TO BE 4'
SCHEDULE
--90'SWEEP
Sourca: EPA
4' DIAMETER HALF SLOTTED PIPE
WRAPPED WITH FILTER MATERIAL1/4"
FOOT PITCH TOWARDS SUMP.
SLOT SIZE .060
PER
SPACING AND FILL TO BE IN ACCORDANCE TO
TANK MANUFACTURER SPECIFICATIONS
FIGURE 8-3
U-TUBE LEAK DETECTION SYSTEM
48
-------
These tubes were Installed under each tank within the backfall material
at Stations 14 and #6 1n Suffolk County, New York.
A comprehensive comparison of leak detection methods was not within the
scope of this project. However, two stations with U-Tubes were Included
in the study in order to make a comparison of hydrocarbon vapor
concentrations from U-Tubes versus hydrocarbon vapor concentrations in
soil gas.
The method of collecting soil gas samples from the backfill areas was
presented ,1n Section 5.2. Briefly, soil gas samples were collected by
Inserting a hollow probe into the backfill and evacuating a soil gas
sample using a vacuum pump. Vapor samples from the U-Tubes were also
collected by Inserting a hollow probe to the desired depth 1n the U-7ube
and evacuating a sample using a vacuum pump. Samples were collected near
the bottom of the U-Tubes to minimize the effects of dilution from the
outside air.
Since vapor samples from the U-Tubes were collected near the bottom of
the U-Tubes, these data were compared to soil gas samples collected from
the backfill at the 10-foot depth. The U-Tube samples and soil gas
samples (at 10 feet) are shown in Table 8-7.
At Station #4 In Suffolk County, New York, the U-Tube sample contained
90,000 micrograms per liter of total hydrocarbons (less methane) while
the soil gas samples ranged from 42,000 to 69,000 micrograms per liter of
total hydrocarbons (less methane). Benzene and toluene were found in
both the U-Tube and soil gas samples while methane, ethylbenzene and the
xylenes were not found at detection limits for either the U-Tubes or soil
gas samples, i
At Station 16 1n Suffolk County, New York, the U-Tube sample contained 47
micrograms per liter of total hydrocarbons (less methane) while the soil
gas sample contained 1,500 micrograms per liter of total hydrocarbons
49
-------
Station *4
U-Tubft-11*
TABLE 8-7 :
U-TUBE VAPOR SAMPLES
SUFFOLK COUNTY, NEW YORK
(Micrograms Per Liter)
METHANE
BENZENE TOLUENE ETHYLBENZENE XYLENES
<24
2800
950
<37
<42
TOTAL HYDRO-
CARBONS
(LESS METHANE)
90,000
TANK TIGHTNESS
TEST RESULTS
SQ1-10'
SQ2-10'
SQ3-10'
SQ4-10'
Station *S
U-Tube-14'
SQ2-10'
<24
<24
<24
<24
<0.02
<0.4
?730"
880
3300
1800
<0.03
<0.8
120
300 .
1000
930
2
55
<37
<37
<37
<37
<0.04
<0.7
<42
<42
<42
<42
<0.04
<0.8
42.000
42.000
69,000
58,000
47
1500
Notations;
NAZ Not Analyzed.
NB ซ No records avalable showing tank tightness results.
jMoios:
(1) Total hydrocalbons are cateuteted from the average response factors for BTX
(2) <24 indications that the concentration is less than the detection In* of 24 mJcrogmms per Her.
50
-------
(less methane). Only toluene was Identified 1n both the U-Tube and soil
gas samples.
These results Indicate that the composition of hydrocarbon vapors found
:Tn_U-Tubes:_are .similar to the vapors found in soil gas. However, the
magnitude of the vapor concentrations may differ. Thase conclusions are
preliminary, since more sample data-is required to accurately delineate
these differences.- -
.816, GROUND WATER SAMPLING
Shallow ground water was encountered at several locations which prevented
soil gas samples from being taken at the 10-foot levels. In these cases,
samples of the ground water were taken and analyzed by the GC/FID using
the same procedures "as were used for the soil gas. These results are
shown in Table 8-8.
51
-------
STATION SAMPLE NUMBER
TABLE 8-8
HYDROCARBON CONCENTRATIONS FROM
GROUNDWATER SAMPLES
DEPTH (FT) METHANE BUTANE ISOPENTANE BENZENE TOLUENE ETHTt-BENZENE XYLENES TOTAL HYOROCARBQ
AUS
AUS
AUS
AUS
AUS
AUS
AUS
AUS
AUS
AUS .
AUS
AUS
AUS
COHH1
COHH2
COHH2
CONN2
HW/H28 _ -
W/H20/P
HW/H20
HV/H20/S
KW/H20
SG4/H20
SG5/H20
SG2/H20
HW/H20 '
HW/H20/P
HW/H20/S
HW/H20
HV/H20
GV-04
GV-04
GU-03
GU-05
10/29' "'
10/29'
10/30
10/29;.
10/i9
10/28
10/28
ID/28
10/29
10/29
10/29
10/30
10/28
11/12
11/13
11/13
11/13
7.
8.
8.
8.
9."
10.
10.
10.
11.
11.
11.
11.
NA
10.
6.
10.
10.
4000
5400"'
6700
. 6600
4200
' .2100...
' 4700
1800
9300
10000
13000
4200
8600
62
18
18
4400
5700.
5000.
8900.
6200.
4900.
4300.
'2400.
2100.
5700.
1000.
690.
2400.
8500.
<7.
<4.
<4.
1700.
NA
-NAT
NA
NA
~NA
NA
-. 'NA '
NA
NA
NA,
NA
NA
NA
<6
<4
<4
<6
77000.
52000.
' 50000.
71000.
67000.
27000.
. 5600.
5600.
67000'.
7300.
7500.
4500.
10000.
<6.
<6.
<6.
<30.
150000.
130000.
16000.
18000.
120000.
83000.
10000.
15000.
160000. .
15000.
15000.
1300.
25000.
<8.
<6.
<6.
-31.
<140.
<140.
<49.
<140.
<140.
<25.
<12.
<49.
<140.
<140.
<140.
<49.
<250.
<4.
<7.
<8.
<37.
80000.
110000.
<79.
110000.
51000. -
70000.
12000.
17000.
93000.
17000.
<130.
<79.
21000.
<8
<10.
-------
9.0 UST REGULATIONS
?9.1 AUSTIN, TEXAS
Underground storage tanks iat existing facilities i* Austin must have a
jpermit to operate and ,are required to be tested or'monitored for leaks on
*a regular basis. If tank testing 1s conducted, a,precision tank test, as
defined in the -NFPA National Fires Codes, Section 329, is performed on
each tank according to the following schedule:
Tank Age ......
(A$ of 6/18/85) Test Frequency
0 to 5 years 0
6 to 10 years Within 12 months of 6/18/85 and then
every 2 years until over 10 years old.
over 10 years Annually, beginning within 12 months
of 6/18/85.
The Department of Environmental Protection (DEP) assumed the underground
tank responsibility from the fire department on January 14, 1987. At the
present time, the DEP has approved seven tests for tank tightness
testing: Petro-tite (Kent-Moore), Hunter, Horner, Acutest, Massney,
Tanty-Tech, and Tank Auditor. Companies who perform these tests are
registered by the DEP.
Monitoring wells may be used as an alternative to precision tank testing
for leak detection of underground storage tanks. For existing facili-
ties, leak detection monitoring by surface geophysical methods such as
ground penetrating radar, electromagnetic induction, resistivity,
magnetometers, and X-ray fluorescence or by tracer analysis may be
permitted-only by approval from the DEP.
**:-,.
9.2 SUFFOLK COUNTY, NEW YORK
Suffolk County began regulating underground storage tanks In 1980 when a
law was passed stating that all new tank installations except underground
petroleum tanks had to be double-walled with leak detection between the
walls. The law further stated that all tanks had to be replaced with
53
-------
double-walled tanks by 1990. Underground petroleum tanks could remain
single-walled up to 1985 1n critical aquifer recharge areas at which time
they had to be replaced with double-walled tanks; with leak detection
between walls. The main aquifer recharge area 1s Inland and encompasses
75/4 of the Island. 'The coastal areas do not affect the recharge of the
aquifer and ..tanks 1n this area can remain single-walled with external
leak detection.
Testing of underground storage tanks . 1s performed by county licensed
testing companies. Tests are performed every two years on older tanks
and every 5 years on newer tanks (since 1975). The only test recognized
by the county Is the Petro-Tlte Tank Tester (formerly Kent-Moore) system.
9.3 SAN DIEGO, CALIFORNIA
California state law regarding the monitoring and testing of underground
storage tanks allows for Implementation of these regulations to be
carried out at the local level. Counties Implement the regulations
through the Issuance of permits to underground storage tank owners. A
city may, by ordinance, assume such responsibilities within Us boun-
daries.
All owners of existing underground storage tanks are required to Imple-
ment a visual' monitoring or alternative monitoring system. Visual
monitoring should be used as the principal leak detection monitoring
method, where feasible. When visual monitoring Is not possible, an
alternative method should be Implemented. The alternative methods are:
Underground Storage Tank Testing,
"
r Other Vadose Zone Monitoring and Ground Water Monitor-
Ing with Soil Sampling,
Vadose Zone Monitoring, Soil Sampling, and Underground Storage
Tank Testing,
Ground Water and Soil Testing,
Inventory Reconciliation, Underground Storage Tank Testing, and
Pipeline Leak Detectors,
54
-------
Inventory Reconciliation, Underground Storage Tank Testing
Pipeline Leak Detectors, Vadose Zone, or Ground Water Monitor-
ing and Soil Testing,
ป Underground Storage Tank Gauging and Testing, and
*ป Interim Monitoring.
Most tank owners select the first alternative - underground storage tank
testing method. In the past, Initial testing was required on all tanks
within 12 months but subsequent testing on non-leaking tanks less than 10
years old was authorized to be done in 30 months rather than annually.
Following the expiration of the 30 month period, all underground storage
tanks operating under the option will require annual testing. The
specific test is not designated, but it must comply with the NFPA
National Fire Codes,'Section 329.
55
-------
10.0 TANK TIGHTNESS TESTING RECORDS
Tank tightness test records were available for most of the study sites.
Two commercially available systems were used to test the tanks - the
Petro-Tlte Tester -(formerly Kent-Moore) and the Hunter Leak Lokater. The
Petro-T1te- Tester has been ~a recognized standard for accurate tank
-testing within the Industry for many years. This system works on the
principle of applying a-hydraulic pressure head to the tank by an
externally connected, graduated standplpe which 1s filled with product to
approximately four feet above ground level. Product level in the
standplpe Is monitored for rise and fall and measured amounts of product
are added or removed. Readings are taken every fifteen minutes for six
hours.
*
The Hunter Leak Lokater measures tank leakage by sensing weight changes
1n a sensor which 1s suspended in the liquid of the tank. Changes in
weight are transmitted to a recorder that registers these changes as
leaks In or out. The only station In this study to use the Hunter Leak
Lokater was RI-4.
The manufacturers of the Petro-Tite Tank Tester and the Hunter Leak
Lokater both report that these systems can detect leaks as low as 0.05
gallons per hour (gph) In tanks and pipes. The accuracy of these tests
1s currently be"1ng examined in other EPA-related studies. Both tests do
not have the capability of detecting spills.
Some records of tank tightness tests were obtained from the oil companies
who owned the various sites. In addition, San Diego County provided test
results for several of the San Diego sites (SD-1 and SD-3 through SD-7).
A government agency provided tightness data for Conn-1. All records
which were obtained are .Included In Appendix L. These records have been
modified to protect the confidentiality of the site locations and
operators.
Table 10-1 presents the Tank Tightness Test Results of the study sites.
Tanks with absolute leak rates of less than 0.05 gph are labeled "TIGHT".
56
-------
TABLE 10-1
TANK TIGHTNESS TEST
-RESULTS
SITE/STATION
AU-1
AU-2
AU-3
AU-4
AU-5
AU-6
AU-7
NY-1
NY-2
NY-4
NY-5
NY-6
TANK '
MATERIAL
Steel
FRP -
Steel
FRP
FRP
Steel'
FRP
FRP
FRP
Steel
FRP
Steel
FRP
NUMBER
OF
TANKS
3
--.: i
3
-4
4
3
4
4
3
6
3
3
3
TANK
INSTALLATION
DATE
- 1961
. 1981
1973
1984
1981
1984
1984
1984
1982
1968
1980
1972
1980
DATE OF
TEST
4/9/86
4/9/86
5/1/86
NT
NTl
4/15/86
NT*
NT
T
12/30/85
NT
NA
NT
TEST
RESULTS
TIGHT
TIGHT
TIGHT
TIGHT
TIGHT
NA
FRP - Fiberglass Reinforced Plastic
NA - Not Available
NT - Tank Tightness Tests Not Required
1 if80'19?7 "aintenance records Indicate station had several small soills in
dispensing areasr, and possibly some pipeline spills. P
2 Spill occurred 'from product line during testing. Corrective action was taken.
57
-------
TABLE 10-1 (CONTINUED)
TANK TIGHTNESS TEST
RESULTS
ff
SITE/STATION
RI-1
RI-2
RI-3
RI-4
CONN-1
CONN-2
-.- .TANK
MATERIAL -
Steel
Steel
Steel "'
Steel
Steel
Steel
Steel'
Steel
FRP
Steel
Steel
Steel
Steel
Steel
Steel
Steel
NUMBER
OF
TANKS
3
,3 *
6 ,
1
1
1
1
1
1
.
1
1
I
I
I
I
2
TANK
INSTALLATION
DATE
1973
1976
1965
. 1966
1966
1966
1966
1966
1984
1984
1966
1978
1966
1966
1985
1940
-
DATE OF
JISJ
NA
9/25/87
NA
1/22/86
(Hunter)
1/22/86
(Hunter)
1/22/86
(Hunter)
1/22/86
(Hunter)
1/22/86
(Hunter)
1/22/86
(Hunter)
1/22/87
1/21/87
1/21/87
1/21/87
1/21/87
NA
NA
TEST
. RESULTS
NA
TIGHT
NA
LEAK1
TIGHT
A
LEAK2
TIGHT
TIGHT
TIGHT
TIGHT
TIGHT
TIGHT
LEAK3
TIGHT
NA4
NA
FRP - Fiberglass Reinforced Plastic
NA - Not Available
1 Failed tightness test on 1/22/86 due to a leak in system line. No records on
further testing.
2 Failed tightness* test on 1/22/86. No records on further testing.
3 Failed tightness test on 1/21/87 due to leak in suction piping under pump.
Tank has been out of service since 1/87.
4 Hฃ0 was discovered in super unleaded tank in 1/85. Tank was excavated
and replaced with new steel tank.
58
-------
TABLE 10-1 (CONCLUDED)
TANK TIGHTNESS TEST
RESULTS
SITE/STATION
SD-1
SD-2.
SD-3
SD-4
SD-5
SD-6
SD-7
SD-8
SD-9
"TANK
MATERIAL
Steel
Steel
FRP
Steel
FRP
FRP
Steel
FRP
FRP
Steel
Steel'
Steel
Steel
Steel
NUMBER
OF
TANKS
2
1
1 ,
3
2
1
4
3
3
1
1
1
4
3
TANK '
INSTALLATION
DATE
1971
1971
1978
1972
1982
1982
1965
1983
1983
1972
1965
1965
1965
1967
DATE OF
TEST
11/11/86
11/21/86
11/21/86
6/17/87
12/10/86
12/22/86
11/5/86
5/7/86
5/18/87
4/16/86
4/16/86
4/17/86
1/21/86
.NA
TEST
RESULTS
TIGHT
TIGHT!
TIGHT2
TIGHT
TIGHT
TIGHT3
TIGHT
TIGHT
TIGHT
TIGHT
TIGHT
TIGHT
TIGHT
NA
FRP - Fiberglass Reinforced Plastic
NA - Not Available
1 Failed tightness test on 11/11/86 due to a leak in diesel vent line.
on 11/21/86 and passed.
2 Failed tightness test on 11/11/86 due to tank leak of -0.5 gph.
on 11/21/86 and passed.
3 Failed tightness test on 12/10/86 due to leak in the vapor line.
Retested on 12/22/86 and passed.
Retested
Retested
59
-------
Tanks with leak rates greater than 0.05 gph are labeled "LEAK" and an
explanation,of the leak and the surrounding circumstances Is provided In
-the accompanying .footnote. Several sites had no available records or had
not been tested due to recent tank Installations and are labeled "MA" and
"NT", respectively, In'the table...
There 'are ปa total of 100 underground storage tanks at the 27 gasoline
stations that were 'studied. 0,f this total, 63 tanks are fabricated from
steel and were Installed between 1940 and 1984. The remaining 37 are
made of fiberglass reinforced plastic (FRP) and were .Installed between
1978 and 1984.
Of the 63 steel tanks, 42 were determined tight In recent tests. Three
steel tanks, two from RI-4 and one from CONN-1, were found to be leaking.
No further records are available to Indicate repair and/or subsequent
testing of these tanks. No tank tightness test records are available on
the remaining 18 steel tanks.
Tank tightness tests were conducted on 12 of the FRP tanks; all tested
tight. Tests on the remaining 25 were not required by the regulating
government agency due to the relatively new age of the tanks.
Seven gas stations had histories of leaks: AU-4 & 6; RI-4; CONN-1 & 2;
and SD-1 & 3. Maintenance records from AU-4 for the period of 1980 to
1987 Indicate that numerous surface spills occurred from vandalized split
hoses and dispensers. Records also exist of low or slow flow which might
Indicate pipeline leaks. AU-4 was removed from the database as a clean
site because of Its history of high maintenance and Its unusually high
soil gas concentrations. AU-6 was also removed from the database because
of a known spill that occurred from a product line break. The five other
stations remained 1n the database as background data because the soil gas
concentrations were not excessive.
60
-------
11.0 DATA ANALYSIS
Geoscience Consultants, Ltd. investigated hydrocarbon vapor concentra-
tions in the backfill of underground storage tanks ;(UST) In two phases:
^a field investigation phase and a data analysis phase.
Since no.database for soil gas information in non-contaminated UST sites
was known to exist, it was necessary to conduct field investigations to
establish a baseline of hydrocarbon vapor concentrations. Data were
collected from twenty-seven gasoline service stations selected as non-
contaminated sites. Selection criteria (Section 3.0) were used to
develop a data set which Included a variety of tank ages, tank
materials, stored products and backfill materials. The underground
storage tanks selected were believed to be non-leaking, or "tight." UST
systems were considered to be tight if:
Tightness testing Within the previous two years Indicated the
system to be without leaks, or
In cases where test records were not available, the environ-
mental and maintenance personnel of the oil company had no
knowledge of contamination due to leakage at the site.
Two stations sampled (Stations #6 and #4 In Austin, Texas) were deter-
mined to be inappropriate as non-contaminated sites, and their data were
not Included in the data set. Station #6 had a fresh gasoline spill from
a product line puncture that occurred during the field investigation.
Station #4 had a history of frequent product line and dispenser problems,
according to maintenance records, and no test records were available.
The non-contaminated site data, therefore, consisted of 279 soil gas
samples taken from twenty-five service stations.
Contaminated site data were obtained from Tracer Research Corporation
historical records. The contaminated site data was selected from sixty
soil gas samples taken from nine sites having known contamination from a
petroleum fuel leak or spill. These sites were all active gasoline
service stations or fueling facilities.
61
-------
The strategy for data analysis was determined by the fact that no usable
data for non-contaminated sites were known to exist. Therefore, analyses
were employed ปwhieH~ could delineate patterns In; the data, if they
existed, and -which could prove useful in establishing contamination
thresholds." ~-
Data analysis was broken down Into three parts:
- Analysis of total hydrocarbon concentrations ("less*methane"
and'"Including methane") in soil gas at non-contaminated sites
with the objective of establishing a descriptive statistical
baseline.
Comparison of the non-contaminated site baseline information to
data from sites where petroleum fuel contamination was known to
exist. This comparison examined the appropriateness of
establishing an upper limit for total hydrocarbon (less
methane) vapor concentrations at non-contaminated sites that
could provide a "threshold" concentration value between non-
contaminated and contaminated sites.
Non-parametric statistical testing of each data set (non-
contaminated and contaminated) in order to substantiate
observed differences and Identify significant trends among
total hydrocarbon vapor concentrations, sample depth, location,
'backfill materials, tank age and tank material.
Analyses focused on concentrations of total hydrocarbons (less methane)
in soil gas, as the presence of total hydrocarbons 1s Indicative of
contamination from a petroleum leak or spill. Methane was excluded from
the reported concentrations In order to present a profile of compounds
similar to that of gasoline, and to exclude methane concentrations which
may have been present due to naturally-occurring decomposition of organic
matter.
The use of total hydrocarbon concentrations in soil gas as a
contamination index is consistent with current EPA ground water and soil
monitoring proposals. An analysis of total hydrocarbon data (including
methane) is presented (Section 11.2) to show how these data are
62
-------
distributed as compared to total hydrocarbon concentrations (less
methane). This comparison may be useful in evaluating total hydrocarbon
concentrations from leak detection devices which include methane.
Accuracy In the data analysis was essential because the results may be
used to provide "direction for future leak detection methods. Towards
-this goal,-the soil gas data were-reported 1n mlcrograms per liter (ug/l)_
because this provided ,a better approximation of the total hydrocarbon
vapor concentrations than parts "per million by volume (ppmv) (Section
7.0). 'Also, three gas chromatograph/flame ionization detection (GC/FID)
analyses were generally performed on each sample, and the arithmetic mean
of the usable samples, as judged by the GC/FID operator, was used in the
analyses. The repllcablHty of analytical results were within 25
percent of the average concentration value for each sample.
11.1 EMPIRICAL DISTRIBUTION OF TOTAL HYDROCARBON
CONCENTRATIONS (LESS METHANE) FOR NON-CONTAMINATED
SITES
An empirical distribution of the total hydrocarbon (less methane) vapor
concentrations 1n soil gas surrounding non-contaminated UST systems 1s
useful for two reasons:
It shows what concentrations can be considered as "background"
concentrations 1n a UST system, and
The distribution can be compared to similar concentration
distributions from contaminated sites.
Even at sites with no known contamination, a level of total hydrocarbon
vapor concentrations is present resulting from surface spills or small
undetected leaks of .petroleum fuels. These concentrations are defined as
the total hydrocarbon background level of the soil gas at the site.
The best way to describe the distribution of total hydrocarbon
concentration data 1s by using the relative frequency distribution.
The relative frequency distribution's obtained by grouping the data Into
63
-------
concentration classes and determining the proportion of samples in each
of the classes. This distribution -for total hydrocarbon (less methane)
concentrations Is shown In Table 11-1 in microgramsj per liter (ug/1) and
In Table 11-2 1n parts per milIon by volume (ppmv).
The classes in these distributions were chosen to show the overall
distribution of samples, as well as the percentage of samples below 1500
ug/1 (approximately 500 ppmv). The 1500 ug/1 concentration class was
chosen because proposed EPA regulations concerning leaking UST systems
have considered 500 ppmv as a possible threshold value to differentiate
non-contaminated from contaminated sites. The. relative frequency
distribution shows that 53.2 percent of the samples were below 1500 ug/1.
The overall distribution shows that 93.1 percent of the samples were less
than 100,000 ug/1.
There are nineteen samples (6.8 percent of the total) that have average
concentration values greater than 100,000 ug/1. Site and sample data
were examined to explore causes for these high values. Table 11-3 shows
the site and sample location of the data points. The nineteen samples
came from seven service stations studied. Tightness test results showed
the UST systems at four of these stations to be tight, while no test
records were available for the other three.
A possible source for the high total hydrocarbon (less methane)
concentrations at the seven sites Is from surface spills. Interviews
with the participating oil companies revealed that underground fuel
storage tanks are occasionally overfilled by the transporter. Since
there is no system for monitoring these surface spills, the frequency of
this event is unknown., . ,. ... - .
t
Another possible source for the high concentrations could be related to
the age of the tanks. Six of the stations contained steel tanks
Installed between the years 1965 and 1971. One station contained a
64
-------
TABLE li-1
DISTRIBUTION OF NON-CONTAMINATED SITE DATA
FOR TOTAL HYDROCARBONS LESS METHANE
CUMULATIVE
Not Detected - 65 '?* 2
< 150ฐ " 84 - * ]?) o ll'l
1501 - 5000 -,- - 17 ? n - *3'2
5000 - 10,000 ';12 43 5?'2
10,000 - 50,000 56 ^ 20 0 S3'5
50,000 -. 100,000 - 27 96 ?3'5
100,000-270,000 18 *A 93'1
l.JOO.OOO _ , ซ"J ^
Mean 23,300
Median 800
Upper
Quartile 33,000
65
-------
; . TABLE 11-12
DISTRIBUTION OF NON-CONTAMINATED SITE DATA
FOR TOTAL HYDROCARBONS LESS METHANE
(Parts Per Million by Volume)
, % - . CUMULATIVE
vPPnป) --- RELATIVE
CONCENTRATION RANGES NUMBER OF ' RELATIVE FREQUENCY FREQUENCY (%)
(Hicroorams Per Liter SAMPLES DISTRIBUTION t%) DISTRIBUTION
Not Detected " " 65 23.2 23.2
< 500 88 31.4 - 54.6
501 - 1,350 14 . 5.0 59.6
1,351 - 2,700 11 3.9 63.5
2,701 - 13,500 57 20.4 83.9
13,501 - 27,000 27 9.6 93.5
27,001 - 72,900 17 6.1 99.6
> 72,900 1 JL 100.0
280 100.0
Mean 7,200
Median 220
Upper
Quartile 9,200
66
-------
TABLE 11-3
1TOTAL HYDROCARBON CONCENTRATIONS LESS METHANE
GREATER THAN 100,000 MICROGRAMS PER LITER
STATION
Austin. Tx
Station #5
Suffolk County. NY
Station *1 (1)
Station t5
San Dleao. C/\
Station 4>4
Station #7 (1)
Station #8 (2)
Station *9 (1)
TANKAGE
AND
MATERIAL
1971-Stee)
PETROTITE
TEST RESULTS
Tight
1 9 ' 8 2
Fiberglass
1972-Sted
1965-Steel
1965-Steel
1965-Steel
1967-Stee)
NR
NR
Tight
Tight
Tight
NR
SAMPLE NUMBER-nFPTH
SQ1 -2
SQ1 -6
SQ1 -10
SQ2-6
SQ3-2
SG4-2
SQ2-2
SG2-6
SG2-8
SG4-10
SQ4-2
SG1 -10
SQ2-2
SQ2-6
SQ2-10
SQ2-10
SQ3-10
SQ4-10
SG2-6
TOTAL HYDROCARBONS
CONCENTRATION LESS
METHANE
(Mlcroorams Per Liter)
150,000
110,000
1.100,000
120.000
190,000
140,000
170.000
210,000
270,000
110,000
110.000
120.000
120,000
130,000
210.000
110.000
104,000
120.000
110.000
NOTES: (1) SQ2 is located near a tank ซ cap.
(2) Station ซ to an inactive service station.
67
-------
fiber-glass tank Installed 1n 1982. The possibility of undetected leaks
could be greater 1n older tanks.
11.2 EMPIRICAL DISTRIBUTION OF TOTAL HYDROCARBON
. CONCENTRATIONS (INCLUDING METHANE) OF NON-
CONTAMINATED SITES-
It may be useful to report total hydrocarbons as
two reasons:
'Including methane" for
- Methane can also occur by the natural decomposition of
petroleum fuel In soil, and
Some UST leak detection methods are based on detection equip-
ment that 1s sensitive to any hydrocarbon compound. Therefore,
. these detection devices will detect the presence of methane in
soil gas in addition to other hydrocarbon compounds.
The empirical distribution of average total hydrocarbon vapor
concentrations (including methane) is compared to the distribution of
average total hydrocarbon vapor concentrations (less methane) in
micrograms per liter in Table 11-4, and in parts per million by volume in
Table 11-5.
The distribution of total hydrocarbons including methane are similar to
total hydrocarbons less methane in two class ranges: 5,001 - 10,000 ug/1
and 50,001 - 100,000 ug/1. However, differences exist In the other class
ranges. These differences can best be shown by summarizing the
distributions into two classes as follows:
CONCENTRATION RANGES
(Mlcroarams Per Liter)
i
ฃ 100,000
> 100,000
RELATIVE FREQUENCY PERCENT
Less Methane Including Methane
93.2
6.8
100.0
73.8
26.2
100.0
68
-------
TABLE 11-4 "
MCTurF TOTAL HYDROCARBONS INCLUDING
METHANE AND LESS METHANE AT NON-CONTAMINATED SITES
(Micrograms Per Liter) .;
CONCENTRATION RANGES RELAE ^QUENCY DISTRIBUTION (PERCENT)
(Mlcroqrqms Per HtPr Less Methane Includina
< 5,000^* 59 2 AO o
5,001 - 10,000 43 4?'f
10,001 - 50,000 20.*0 n'n
50,001 - 100,000 96 'H
100,001 - 400,000 64 >? o
400,000 - 1,000,000 - Z\'*
1,100,000 05
1,250,000 . "
. . 100.0
* Includes non-detected values.
69
-------
TABLE 11-5
COMPARISON OF TOTAL HYDROCARBONS INCLUDING
METHANE AND LESS METHANE AND LESS METHANE AT
NON-CONTAMINATED SITES
(Parts Per Mi 11 i on by Vol ume)>
CONCENTRATION RANGES
(Mlcroorams Per Liter
RELATIVE FREQUENCY DISTRIBUTION (PERCENT)
Less Methane Including Methane
< 500 .*.
501 - 1,350
1,351 - 2,700
2r701
13,501
27,001
72,901
250,001 -
> 600,000
13,500
27,000
72,900
250,000
600,000
54.6
5.0
3.9
20.4
9.6
6.1
0.4
100.0
45
2.1
2.5
8.9
5.0
11.1
15.0
6.4
100.0
Include non-detected values.
70
-------
The effect of Including methane In the total hydrocarbon concentration is
to lower the percentage of samples with concentrations equal to or less
than lOO.OOOug/1 Kor 30,000 ppmv) by 21 percenj. This effect was
-expected since the ;soil gas data showed high concentrations of methane at
many of the sites. This was probably due to naturally-occurring methane
as well as .methane which occurs from the decomposition of hydrocarbon
compounds.
11.3 COMPARISON OF TOTAL HYDROCARBON CONCENTRATIONS FOR NON-
CONTAMINATED SITE AND CONTAMINATED SITE DATA SETS
The data distribution in Section 11.1 has shown that a wide range of
background hydrocarbon vapor concentrations exist in the soil gas in
backfill at non-contaminated UST sites. These concentrations ranged from
the lower detection limits of 0.02 micrograms per liter (ug/1) to
1,100,000 ug/1 for total hydrocarbons (less methane). Although much
variability exists in these data, a comparison of these data to data
from known contaminated sites 1s required to determine if background
vapor concentrations differ from vapor concentrations at sites with known
contamination. If statistically significant differences exist between
these data distributions, then the results of this comparison could be
useful to UST regulators, service station owners and others who must
interpret soil gas data to determine if contamination exists at a UST
site.
An evaluation of these differences could also determine the
appropriateness of establishing a threshold concentration for total
hydrocarbons (less methane). Statistical testing was performed (Section
11.4) to determine If observed differences concluded from the descriptive
statistics are significant differences.
i
In order for the data sets to be comparable, the data in each set must be
collected In a similar fashion. Since the contaminated site data set was
obtained from historical records, data for this set were selectively
chosen to be iconslstent with the samples taken at non-contaminated sites
during the field Investigation.
71
-------
The sampling "Strategy for non-contaminated sites, as outlined In the
Field Methods (Section 5.0) was to collect samples from the backfill of
the tanks and -at depths of 2, 6 and 10 feet. ; Although samples at
contaminated.sites-were .usually not in backfill, data were chosen that
>wefe within approximately 50 feet of the USTs, and at 2, 6, and 10-foot
depths. The method of sampling was similar for both data sets since soil
gas samples were collected by Tracer Research Corporation (TRC) using-
slmllar procedures. . r-r .
In this comparison, total hydrocarbons are reported "less methane" and in
micrograms per liter for both data sets. - The total hydrocarbon (less
methane) concentrations in the non-contaminated data set were calculated
from average response factors for benzene, toluene, ethyl benzene and
xylenes (BTEX). However, in the contaminated data set, total hydrocarbon
concentrations (less methane) were calculated from the response factor
for benzene. Therefore, contaminated site data could be as much as 50 to
100 percent higher If it were reported on the basis of an average BTEX
response factor. A comparison of calculation methods and their effects
on total hydrocarbon concentrations was presented in Section 7.0.
The sample size for the non-contaminated data set was 279 samples from 25
sites. The sample size for the contaminated data set was 60 samples from
9 sites.
The descriptive statistics used to compare the non-contaminated and
contaminated data sets were: mean, median, upper quartile and the
relative frequency distribution percentages. These statistics are useful
because they show the distribution of each data set and these
distributions cSn be compared even though the sample sizes in each data
set are different. The descriptive statistics for the non-contaminated
sites were shown in Table 11-1 and those for the contaminant sites are
shown In Table 11-6. A comparison of these descriptive statistics are
shown in Table 11-7 in micrograms per liter for total hydrocarbons (less
methane). The relative frequency distribution for the hon-contaminated
72
-------
TABLE 11-6
DISTRIBUTION OF CONTAMINATED SITE DATA FOR TOTAL
HYDROCARBONS LESS METHANE
CONCENTRATION RANGES-
(MICROGRAMS PFR I JTFR)
Not Detected ..
50,000- - 100,000
100,000 - 270,000
270,000 - 1,100,000
>1'000'000
NUMBER OF
'
1
6
13
FREQUENCY
DlimSrinil
33%
100.0%
CUMULATIVE
RELATIVE
w
100.0
Mean
Median
Upper Quart He
160,000
9,000
22,000
73
-------
TABLTir-7 -
COMPARISON OF NON-CONTAMINATED AND CONTAMINATED
SITE DATA DISTRIBUTIONS FOR
HYDROCARBONS LESS METHANE
--- ' --: RELATIVE RELATIVE
. - - FREQUENCY FREQUENCY
CONCENTRATION RANGES PERCENT PERCENT
fHICROGRAMS JER LITER). CONTAMINATED NON-CONTAMINATED
Not Detected -3.3 23.2
< 1500 31:7 30.0
1501 - 5000 10.0 6.0
5001 - 10,000 10.0 4.3
10,001 - 50,000 10.0 20.0
50,001 - 100,000 1.7 9.6
100,001 - 270,000 10.0 . 6.4
270,001 - 1,100,000 * . 21.6 0.4
2,200,000 1.7 0.0
100.0 100.0
Mean 160,000 23,300
Median 9,000 800
Upper Quartlle 220,000 33,000
74
-------
site data was shown in Figure 11-1 and that for the contaminated site
data 1s shown in Figure 11-2.
/
The relative frequency distributions show much variability in both data
sets. Nine concentration ranges were selected to show this variability.
An evaluation of the means and medians gives additional information about
these data sets. The mean is an arithmetic average that is computed by
summing the concentration values and dividing by the total number of
samples. The median is defined as the middle value after the samples
have been arranged in order of magnitude (Hoel 1967).
In both data sets, ,the medians are much lower than the means. These
differences show that both data distributions are skewed to the right
with a majority of samples 1n the lower concentration ranges. The high
mean values show the effect of a few high concentration values that exist
in both data distributions.
Although similarities exist in the distribution of these data sets, some
differences can also be seen. An order of magnitude difference exists
between the mean of each data set, and between the medians of each data
set. This suggests that although similarities exist in how these data
sets are skewed, that an order of magnitude difference exists for much of
the data.
The order of magnitude can best be seen 1n the concentration ranges above
10,000 ug/1. The relative frequency percentages from Table 11-7 are
summarized below for concentrations above 10,000 ug/1, or about 3000
parts per nillion by volume.
i '
CONCENTRATION RANGES RELATIVE FREQUENCY PERCENT
(Hicroarams Per Liter) Non-Contaminated Contaminated
10,000 - 100,000 29.6 13 4
100,000 - 2,200,000 6.9 33[3
36.5 46.7
; ~ ' ' 75
-------
RELATIVE FREQUENCY DISTRIBUTION (%)
01
O
01
O
at
O
o -
01
o
o
s
0
ฐ
o
o H
o
o
O
I
to
O
o
o
\
Oi
o
o
o
ฐ
8
0
to
vl
O _
o
o
o
o
o ~
o
o
o
g
o -
o
o
o
g
T
o
o
m
8 o
CO
00
-H
O
-------
CONTAMINATED SiTE DATA DISTRIBUTION
TOTAL HYDROCARBONS LESS METHANE
iuu -
90 -
& 80 -
Z
o
g 70 -
m
1 60 "I
Q
RELATIVE FREQUENCY
\
:> S. 8 ป- ฃ g
U- 1 - I 1. _.l L_
' " ' -- ' . ' ' '- ' i '" '
. ..
///
% ^
-r-^n Y//A fySA ^/// ^//A V// V//
777\ \/// \//A \/// '/// (/// y///
ill i
0 1500 5000 10000 50000 100000 270000 1100000 220onnn
MAXIMUM CONCENTRATION (ug/!)
FIGURE 11-2
-------
Most of the non-contaminated samples occur in the 10,000 to 100,000 ug/1
range, while most of the contaminated samples occur above 100,000 ug/1.
y M ;
i
The order of .-magnitude difference between the data sets can also be seen
by comparing'-the upper quartiles of each data set. The definition of
upper quartile is that 75% of the samples occur below the upper quartile
(Hoel 1967).
The upper quartile for the non-contaminated and contaminated data sets
are 33,000 ug/1 and 220,000 ug/1, respectively.
The observed conclusions from these descriptive, statistics is that both
data sets contain much variability and both are skewed to the right. An
order of magnitude difference exists between the data sets for
concentrations above 10,000 ug/1. Statistical testing is Section 11.4
confirms the significance of these differences between the data sets.
11.4 NON-PARAMETRIC STATISTICAL TESTING
The purpose of statistical methods is to describe data quantitatively,
and to draw inferences for decision-making (Kilpatrick 1987). The
descriptive statistics have been examined in the previous sections, and
these described the means, medians, upper quartiles and relative
frequency distributions for the data sets.
In this section, statistical methods are employed to determine what
Inferences can be made about the non-contaminated site and contaminated
site data sets.
The statistical testing in this data analysis served two purposes:
i
The testing determined the significance of the observed
statistical differences between the data sets (non-contaminated
and contaminated) noted in the descriptive statistics, and
The testing delineated data patterns that existed among such
parameters as location of.site, depth of sample, tank material,
tank age and backfill material.
78
-------
The types of statistical tests chosen were dictated by the characteris-
tics of the data set distributions. These distributions, as described
previously, did not appear to correspond to apy known statistical
distribution :such *s ,a Normal distribution. Non-parametric statistical
metiiods were usfld since these methods did not require that the sample
data correspond to a kiipwri statistical distribution (Harval).
These statistical" methods also introduce the element of probability as
related to the drawing of conclusions. Probability was considered
Important 1n developing conclusions about these data sets because these
data sets do not contain complete Information about the entire data set
of underground storage tanks that exist. Therefore, a probability must
be attached to any conclusions made about the data sets. A discussion of
the risks associated with statistical testing, and how these risks were
controlled is given in Section 11.4.1.
11.4.1 The Risks Associated with Hypothesis Testing
There 1s always the possibility of making an Incorrect decision when
testing a hypothesis. This is because inferences about a particular
distribution are based upon random samples from that distribution. A
statistical hypothesis 1s simply an assumption or statement, which may or
may not be true, concerning one or more populations.
There are two types of error or risk associated with the testing of any
hypothesis. Type 1 error Is the probability of rejecting a true null
hypothesis, while Type 2 error 1s the probability of rejecting a true
alternative hypothesis. A null hypothesis Indicates that no differences
exist between distributions. An alternate hypothesis indicates that
differences do exist between distributions.
t
Type 1 error Is usually controlled by setting the significance level of
the test to a small value. This significance level, designated as "p",
numerically describes the probability that a particular hypothesis is
true. Typically this value Is set at 0.05. This corresponds to a
79
-------
confidence level (probability) of 95 percent. The significance level
becomes a specification of the Type 1 error rate of probability. :
-' ' i
> --.
'Type-2 error Is usually-con trolled by taking a properly-si zed sample.
Th~is'study did not consider the control of Type 2 error as a criteria for
determining sample size*. However, when large discrepancies exist between
the Information contained in the .samples and the specification of the
null hypothesis with respect to the sample's, then the Type 2 error will
generally be small. :; " . .
When testing more than one hypothesis, the Type .1 error rate must be
controlled. A simple example -will demonstrate what happens to the Type 1
error rate when testing several hypotheses.
Suppose that each of 10 independent hypotheses are to be tested at a
significance level of 0.05. If the null hypothesis is true in all 10
cases, the probability of detecting this is only 0.60. Therefore, the
Type 1 error rate Is 0.40, which is totally unacceptable. One way to
control the Type 1 error rate when testing several hypotheses is to test
each hypothesis at a reduced significance level. A good conservative
procedure for determining the significance level in a multiple testing
situation 1s the Bonferroni procedure. This procedure 1s described
below.
If an overall Type 1 error rate of 0.05 is to be attained, the
significance level for each hypothesis tested is computed by dividing
0.05 by the number of hypotheses to be tested.
In the example -above, the significance level of each hypothesis should
be: '
0.05 / 10 - 0.005
Thus, 1f each hypothesis 1s tested at a Type 1 error rate of 0.005, then
an overall Type 1 error rate of 0.05 will be maintained.
80
-------
There were 16 statistical tests performed In this study. Therefore, in
-order to maintain an overall Type 1 error rate of 0.05 for this study,
-each hypothesis was be tested at;a Type 1 error rate of 0.003.
11.4.2 Comparison of Non-Contaminated,Site and
Contaminated Site Data Distributions
The descriptive statistics showed some similarities in how the non-
contaminated and contaminated isite data were distributed. The
distribution of both data sets were skewed to the right with a majority
of samples in the lower concentration ranges. However, an order of
magnitude difference existed in the data above 10,000 ug/1. This
difference was seen by a comparison of the means, medians and upper
quartiles of each data set. In this section of the report, a non-
parametric test is used to compare these data sets. This test will
determine if the distributions of these data sets are significantly
different.
The non-parametric test used for this comparison 1s the Two-Sample
Wilcoxon Rank Sum Procedure (Siege! 1956). This test Is designed to
determine if two independent samples are from different distributions.
Since the sample values within each data set contain much variability,
the question is whether the differences observed between the data sets
signify genuine differences in distributions or whether they represent
differences that can be expected between two random samples from the same
distribution.
The Wilcoxon technique tests the null hypothesis that two independent
samples come from Identical distributions. This is called a null
hypothesis because It assumes that there is no difference between
distributions.ป If the outcome of the test rejects the null hypothesis
(that Is, p < 0.003), then 1t can be concluded that the samples came from
two different distributions.
This test was computed using a computer software package called
Statgraph. Results of the test are Included in Appendix M. In most
81
-------
cases, the data -used In this test represent the mean of three GC-FID
Injections for each sample. The concentrations at non-detection levels
-were approximated by-dividing the detection limit iri half.
The outcome of "this test is show below.
DISTRIBUTION-
Non Contaminated
Contaminated- -'-
-SAMPLE SIZE
279
60
AVERAGE RANK
160
215
LEVEL OF .
SIGNIFICANCE
0.00008
This test result shows that there is a significant difference (p < 0.003)
between the distributions of the non-contaminated and contaminated site
data. This test 'result confirms that the distributions of non-
contaminated and contaminated data, as shown in Table 11-7, actually
represent two different distributions.
11.4.3 Non-Parametric Testing for Data Patterns Within
the Non-Contaminated Data
Non-parametric techniques can be 'used to Identify patterns in the non-
contaminated data set if they exist. The results of non-parametric
testing can be used to draw Inferences about the data.
The purpose of this testing was to examine the effects that different
parameters had on the data. These parameters Included site location
sample depth, tank material, tank age and backfill material. The testing
was designed so that Independent effects from each parameter could be
seen. However, Insufficient data were available to delineate the
Individual effect of tank material, tank age and backfill material.
i
The determination of Insufficient data was made from observations about
the data at a time when further data could not be collected (I.e., the
field Investigation had been completed). Two observations were made:
82
-------
-All the fiberglass tanks used pea gravel backfill and
corresponded to newer tank ages (1978 to 1984), and
-All the "steel tanks used sand backfill/ and corresponded to
.older tank ages (1940 to 1984). '
The data co.uld not be separated to distinguish between tank materials,
tank age and backfill material. In this analysis, these three parameters
.are combined and referred to *s either a steel tank system or *a
fiberglass tank system. The presentation of test results are organized
according to he parameters of location, sample, .depth and steel or
fiberglass tank systems. Test results that involve fiberglass tank
systems are only shown for the locations of Austin, Texas, Suffolk
County, New York and San Diego, California since no fiberglass tank
systems were sampled'in Providence, Rhode Island or Storrs, Connecticut.
11.4.3.1 Location
The first parameter examined was site location. The Kruskal-Wallis One-
Way Analysis of Variance by Ranks (Siege! 1956) was chosen to test the
null hypothesis that samples from different locations come from the same
distribution.
This testing was again accomplished by the use of the Statgraph computer
software package. In order to test only for the effect of location, the
data set was broken down into subsets corresponding to sample depth and
the combined group of tank material, tank age and backfill material.
The above breakdown yields six subsets as follows:
fiberglass tank systems at sample depths of 2, 6 and 10 feet
and
steel, tank systems at sample depths of 2, 6 and 10 feet.
The mean concentrations for each sample were used as data. The
concentrations below detection limits were set to positive values at the
detection limits to represent the worst case for concentrations at these
sample points.
83
-------
The results of these tests are shown In Table 11-8 for the steel tank
systems and Table 11-9 for the fiberglass tank systems, and are also
Included In Appendix M. . j
The subsets consisting of steel tank systems at 2, -6 and 10 foot sample
depths show significance at p < 0.003. The interpretation of these
results 1s that the null hypothesis, which states that these subset
samples are from the same distribution set, must be rejected. I.t is
concluded that significant differences do exist among the total
hydrocarbon (less methane) vapor concentrations from the five locations
studied for steel tank systems; The differences were significant at all
three sample depths (2, 6 and 10 feet).
The average rank 1s an Indication of how these concentrations were
ranked. The total hydrocarbon concentrations in Austin, Texas and San
Diego, California were greater than in Providence, Rhode Island, Suffolk
County, New York and Storrs, Connecticut.
The subsets consisting of fiberglass tank systems at each of the 2, 6 and
10 foot sample depths do -not show significance ( p > 0.003) at any of
the "sample depths. The Interpretation 1s that the null hypothesis, which
states that these subset samples are from the same distribution, is
accepted. It Is concluded that no significant differences exist among
the total hydrocarbons (less methane) vapor concentrations from the three
locations studied for fiberglass tank systems. This conclusion can also
be seen by examining the average ranks. The value of these ranks are
similar within each sample depth subset.
11.4.3.2 Sample Depth
The second parameter examined was sample depth. The analysis was
designed to determine if differences existed among samples taken at
different depths. This analysis Is based on the assumption that samples
taken from different depths within a hole are related, and the tests
determine if data at different sample depths have been drawn from the
same distribution.
84
-------
TABLE 11-8
RESULTS OF KRUSKAL-WALLIS TESTS FOR LOCATIONS
WITH STEEL TANK SYSTEMS USING .NON-CONTAMINATED DATA
STEEL TANK SYSTFMS ;
Sample Depth - 2 Foot
Sample Depth - 6 Toot
Sample Depth - 10 Foot
SAMPLE ' AVERA'&E
LOCATION SIZE RANK
Austin, TX 14 51
San Diego.CA 29 49
Providence, RI - 14 30
Suffolk County, NY 8 20
:Stores,. CT 10 15
San Diego, CA 28 48
Austin, TX 13 43
Suffolk County, NY 6 28
Providence, RI 15 22
.Storrs, CT 9 17
S*n Diego, CA 17 33
Austin, TX n 27
Suffolk County, NY 5 18
Providence, RI n 14
Storrs, CT 37
SIGNIFICANCE
LEVEL
0.000003
0.00002
0.0006
85
-------
TABLE 11-9 :
RESULTS OF KRUSKAL-WALLIS TESTS FOR LOCATIONS
WITH FIBERGLASS TANK SYSTEMS USING NON-CONTAMINATED DATA
FIBERGLASS'TANK SYSTEMS
Sample Depth - 2 Foot
Sample Depth - 6 .Foot
Sample Depth - 10 Foot
SAMPLE AVERAGE
LOCATION SIZE RANK
Suffolk County, NY 10 21
Austin, TX - ' 9 20
San Diego, CA -14 12
Suffolk County,~ NY 11 18
Austin, TX 8 . 14
San Diego, CA 11 14
San Diego, CA 8 13
Suffolk County, NY 9 12
Austin, TX 5 9
SIGNIFICANCE
LEVEL
0.06
0.4
0.5
86
-------
Two non-parametric tests were chosen. These were the Page L Test for
Ordered Alternatives based on Friedman Rank Sums, and the Wilcoxon
Matched-Pairs Signed-Ranks Test (Siege! 1956). .
. '.-..- . ' '
The'Page L~'Test-was chosen to test the null hypothesis that data at
different sample depths have been drawn from the same distribution. If
differences do exist, this test also reveals how these data are ordered.
Specifically, this test will determine if one of the following trends
exist for total hydrocarbon (less methane) vapor concentrations taken
from non-contaminated sites:
2' < 6' .- 10'"
2' - 6' < 10'
2' < 6' < 10'
2' - 10' < 6'
If test results show a level of significance ( p < 0.003) then the null
hypothesis is rejected and one of these conditions exist.
In cases where these test results showed a level of significance for a
particular data subset, the Wilcoxon Matched-Pairs Signed-Ranks Test was
employed to further test the following hypotheses for total hydrocarbon
(less methane) vapor concentrations at non-contaminated sites:
2' < 6'
6' < 10'
i 2' < 10'
-.1 - ,.
A separate calculation was required to test for each of these conditions.
The benefits in using the Wilcoxon Test as a supplement to the Page L
test are not only to determine exactly how the data at different depths
are ordered, but also to utilize more data from the non-contaminated data
87
-------
set. There were service stations In San Diego and Austin in which
shallow perched water .zones were encountered that precluded taking
samples ปat 10 fe.et.-^ Therefore, soil gas samples were only collected at 2
and 6 foot depths. By using the Wilcoxon Test, these data could also be
utilized. 'The computations for.both techniques (Page L and Wilcoxbn)
were d~one. by hand, under the direction of a qualified statistician and
are shown 1n Appendix M.
The results of the Page L Tests and the Wilcoxon Tests are shown in
Tables ll-10'and 11-11, respectively. Calculations .for these tests are
Included In Appendix M. These test results show variations in signifi-
cance levels at individual locations in both the steel and fiberglass
tank systems. A summary of the significant test results is given below.
1) Two significant test results were shown from the Page L Test
for the overall data. The significant differences were among
total hydrocarbon (less methane) vapor concentrations at the
different sample depths (2, 6 and 10 feet) for both steel and
fiberglass tank systems. The overall test represents data that
are combined from the different locations.
2) Significant test results were also shown from the Page L Test
for Individual locations. There were significant differences
among total hydrocarbon (less methane) vapor concentrations at
the different sample depths (2, 6 and 10 feet) for steel tank
systems in San Diego, CA and for fiberglass tank systems in San
Diego, CA and Suffolk County, NY.
3) One significant test result was shown from the Wilcoxon Test
for San Diego, CA. The significant difference was shown in the
test of 2'<6'. Therefore, total hydrocarbon (less methane)
concentrations are greater at 6 feet than at 2 feet for the
steel tank system In San Diego, California.
The variations 4n significance at the different locations could be due to
two factors: il) The differences in the locations, such as geology,
hydrology, backfill material, etc., and 2) insufficient data to detect
significant differences using the statistical methods.
-------
TABLE 11-ia
RESULTS OF PAGE L TEST FOR DIFFERENCES
IN DATA ACCORDING TO SAMPLE DEPTH
STEEL TANK- SYSTEMS
FIBERGLASS TANK SYSTEMS
LOCATION
Austtn, TX
Suffolk Co, NY
San Diego, CA
Providence, RI
Overal1
Austin, TX
Suffolk Co, NY
San Diego, CA
Overall
'SAMPLE
SIZE
II
3
15
5
34
6
7
8
21
SIGNIFICANCE
LEVEL
< 0.05
> 0.05
< 0.001
> 0.05
< 0.0002
< 0.05
< 0.001
< 0.001
< 0.0002
89
-------
; . TABLE 11-11
RESULTS OF-WILCOXON TESTS FOR DIFFERENCES
IN DATA ACCORDING TO SAMPLE DEPTH
SAMPLE SIGNIFICANCE
STEEL TANK SYSTEMS LOCATION TEST SIZE LEVEL
San Diego, CA - 2'<6' 24 <0.001
San Diego, CA 6'<10' 16 0.004
San Diego, CA 2'<10' 11 0.0012-
90
-------
Unfortunately, the paired-sample Wilcoxon test is not as sensitive as the
Page 1 test for detecting significant differences. This is due to the
nature of the hull distribution of the paired-sample Wilcoxon test for
small samples. Thus, even though the Page L test may'have detected
significant differences in total hydrocarbon concentrations between the
three sample depths, "the paired-sample ^Wilcoxon may not uncover the
nature of these differences. Also,.the W11coxon could only be applied in
cases where the sample:size was greater than nine samples.
Each of the paired-sample-Wilcoxon tests were tested at individual
significance levels of 0.0015. This was derived by dividing 0.003 by
two, since two independent test cases (2'<6' and 6'<10') were performed.
11.4.3.3 Conclusions from Non-Parametric Tests
Within the Non-Contaminated Data
The data patterns associated with site location and sample depth were
delineated by the use of Kruskal-Wallis, Page L and Wilcoxon non-
parametric statistical methods. The Kruskal-WalUs method, used to
delineate patterns according to location, revealed that significant
differences in total hydrocarbon (less methane) vapor concentrations
among the five locations studied for steel tank systems. The differences
were significant at all three sample depths (2, 6 and 10 feet). There
were no significant differences between the total hydrocar-bon (less
methane) vapor concentrations at the three locations studied for
fiberglass tank systems.
The Page L method, used to delineate patterns according to sample
depths, revealed that ^significant differences exist between the total
hydrocarbon (less methane) vapor concentrations among the different
sample depths .{2, 6 and 10 feet) for both steel and fiberglass tank
systems. ......'
The results of these tests Indicate that data from steel tank systems at
different locations and sample depths represent significantly different
91
-------
data distributions. Also, data from fiberglass tank systems from all
locations, but ,at different .sample depths represent significantly
.-different distributions. ;
>
'The means, medians, lower rand upper quartiles are shown in Table 11-12
for the steel tank systems .and Table 11-13 for the fiberglass tank
systems for total hydrocarbon (less methane) vapor concentrations in
mlcrograms per liter.
The difference in total hydrocarbon (less methane) vapor concentrations
at different sample depths can be seen In these tables. The steel tank
systems in Austin, TX, San Diego, CA and Suffolk County, NY show
Increasing concentrations with depths in the means, medians, and lower
and upper quartlles. The differences In concentrations at the different
locations can also be seen.
11.5 RESULTS AND CONCLUSIONS OF DATA ANALYSIS
The distribution of total hydrocarbon (less methane) vapor concentrations
was skewed to the right with a majority of samples In the lower
concentration ranges. The relative frequency distribution showed 53.2
percent of the samples below 1,500 ug/1 and 93.1 percent below 100,000
ug/1. The median was 800 ug/1 and the mean was 23,300 ug/1. The
difference between the mean and the median 1s because of a few high
concentration values.
The distribution of total hydrocarbon (Including methane) vapor
concentrations showed that 21 percent more samples existed above 100,000
ug/1 as compared to total hydrocarbons (less methane). High
concentrations of methane were seen at many of the sites. These
concentrations i are probably due to decomposition of the background
hydrocarbons as well as naturally occurring methane.
Although much variability existed in both the non-contaminated and
contaminated data, significant differences could be seen between the two
distributions. Both distributions were skewed to the right with a
92
-------
TABLE 11-12 :.
DESCRIPTIVE STATISTICS FOR TOTAL HYDROCARBON
LESS METHANE CONCENTRATIONS IN STEEL TANK
.SYSTEMS AT DIFFERENT LOCATIONS AND SAMPLE DEPTHS
(Micrograms Per Liter) -'
Austin, TX
-Mean .
Median
Lower Quartile
Upper Quartile
Providence, RI
Mean
Median
Lower Quartile
Upper Quartile
San Diego, CA
Mean
Median
Lower Quartile
Upper Quartile
Storrs, CT
Mean
Median
Lower Quartile
Upper Quartile
Suffolk County, NY
Mean
Median
Lower Quartile
Upper Quartile
2 Foot - -
41000
15000'-
570
36000
1700
1
Detection Limit
0.1
30000
27000
5100
37000
270
Detection Limit
Detection Limit
1.0
5300
1.6
Detection Limit
2100
SAMPLE DEPTH
6 Foot
24000
16500
. 380
35000
10 Foot
120000
12000
160
36000
1200
0,3
Detection Limit
450
44000
41000
2400
70000
5300
0.3
Detection Limit
11.0
16000
1100
Detection Limit
39000
1300
0.1
Detection Limit
350
72000
71000
39000
104000
1.0
0,06
Detection Limit
3.0
27000
110
Detection Limit
36000
93
-------
TABLE 11-13 :
DESCRIPTIVE STATISTICS FOR TOTAL HYDROCARBON
LESS METHANE CONCENTRATIONS IN FIBERGLASS TANK
SYSTEMS AT DIFFERENT DEPTHS
(Mlcrograms Per Liter) 1
Mean
Median
Lower Quartile
Upper Quartile
2 FOOT
16142.9
28
.1
21000
SAMPLE DEPTH
6 FOOT
21689.1
780
2
38500
10 FOOT
49132.7
5850
27
58000
94
-------
majority of samples 1n the lower concentration ranges. However, an
order of magnitude difference existed between the mean of each data set,
and between the -median of each data set. The order of magnitude was best
seen 1n concentrations above 10,000 ug/1. Of; the non-contaminated
sampler, 29,6 percent occurred In the range of 10,000 to 100,000 ug/1
while 33.3 percent of the contaminated samples occurred 1n the range
above 100,000 ug/1.
95
-------
-------
12.0 CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER STUDY
12.1 CONCLUSIONS
The following conclusions are derived from the results of this study:
~)
' Underground storage tank sites evaluated in this study where
total hydrocarbon (less methane) concentrations In soil vapor
exceeded 100,000 ug/1 (27,000 ppmv) were generally considered
contaminated, whereas sites that exhibited vapor values less
than 100,000 ug/1 typically had not had a release and were
considered non-contaminated. This apparent threshold value of
100,000 ug/1 (27,000 ppmv) of total hydrocarbon (less methane)
vapors may be used to help differentiate between non-
. contaminated and contaminated sites.
-* Calculation of total hydrocarbon values "as BTEX" based on the
average of the response factors for benzene, toluene,
ethyl benzene and ortho-xylene provides a more accurate
representation than when calculated "as benzene".
Because of the regional variability of the data collected in
this study, any soil vapor concentration limits that are to be
utilized to differentiate between contaminated and non-
contaminated sites may best be established on a regional or
local basis.
Soil gas techniques can effectively be used to evaluate the
backfill areas of underground gasoline storage tanks to
determine if significant leaks exist, especially if appropriate
regional or local threshold levels are established.
Limited analysis of butane vapor concentrations indicates, that
butane analysis may be useful in detecting recent leaks or
spills.
12.2 RECOMMENDATIONS FOR FURTHER STUDY
Analysis of the data collected 1n this study revealed several areas where
additional study would be useful in developing a more complete
understanding of the occurrence and characteristics of soil gas at both
clean and cqntamlnated underground gasoline storage tank sites.
Recommendations for further study are:
Develop a standardized method for reporting soil gas
concentrations In the backfill areas of underground storage
tanks. This can be done by a more thorough analysis of soil
gas in each of the three geographical areas used in this study.
The objectives would be to measure the concentrations, develop
96
-------
simplified calculations to be used 1n reporting the
concentration values and determine the appropriate assumptions
.and approximations.
* /Determine the minimum amount of data required to decide If a
.site 1sr contaminated by a leak. The objectives would be to
determine the required number and locations of sampling points,
: the number of samples above a specified threshold limit that
would be acceptable, and whether butane concentrations can be
-used to distinguish between a leak and a spill.
Determine the effects of geology, backfill material, tank age
and tank material on soil gas concentrations. A sufficient
amount of data was not collected in this study to determine the
effects of these .parameters.
Examine the dispersion and decomposition of contamination by
additional sampling at Austin 16, taking advantage of the
recent documented spill.
r
Determine the effects of a leaking pipeline on an underground
storage tank system as compared to the effects of only a
leaking tank.
97
-------
13.0 REFERENCES CITED
American Petroleum Institute, Publication No. 4395, Auqust 1985
Laboratory Study on Solubilities of Petroleum, Hydrocarbons in Ground
ซwater, August" 1985 '
Harvel, Chuck, Statistician Consultant, Personal Communication
Himnielblau, David M., Basic Principles and Calculations 1n Chemical
Engineering, Third Edition, Prentice-Hall, IncV, New Jersey, 1974?
"^'in?"! New YEorek,eni96r7 Stat1st1cs' Second Edition, John Wiley & Sons,
UtUS 1-2'3' John
Radian Corporation, Personal Communication
for the Behaviorai
Tracer Research Corporation, Personal Communication
U*S> Molht^cT 1tacl fr0tlcl1on. ^9^nc^ป Underground Tank Leak Detection
Methods: A State-of-the-Art Review, 1986
Pฐllut1onป STS Ori91n and Control> "^Per and Row,
UST\FINAL.RPT
98
-------
.1
-------
NOTICE
This report is an external draft for review
purposes only and does not constitute agency
policy. Mention of trade names or commercial
products does not constitute endorsement or
recommendation for use.
-------
-------
-------
------- |