510R04004
\
v Frequency And Extent Of Dispenser
i Releases At Underground Storage
f Tank Facilities In South Carolina
                                  Printed on recycled paper

-------
This document was prepared by ICF, Incorporated and submitted
           in support of Contract No. 68-WO-0065.

                      Lynn DePont
                   EPA Project Officer
           U.S. Environmental Protection Agency
                Washington, D.C. 20460

-------
                             Contents

                                                               Page

Introduction                                                      1

Background                                                      1

Purpose Of This Study                                             3

Methodology                                                      3
     Data Collection                                              3
     Data Screening                                              4
     Data Recording                                              4

Quality Assurance/Quality Control                                  8

Results                                                          9

Conclusions                                                     13

Appendix 1 - South Carolina Dispenser Data Gathering Form
     For First Phase                                             15

Appendix 2 - South Carolina Dispenser Data Gathering Form
     For Second Phase                                           16

Appendix 3 - South Carolina RBCA Site Priority
     Classification System                                        17

Appendix 4 - Summary Of Results                                  19

Appendix 5 - Statistical Analysis                                    22

-------
                             Tables

Number                                                     Page

     1    Summary Of Quality Control Data                       8

     2    Summary of Contamination At Dispenser
               Islands And Facilities                             9

     3    Numbers Of Sampled Facilities, Dispenser
               Islands, And Samples By Pump Type                10

     4    Frequency Of Contamination And Contamination
               Above RBSLs At Dispenser Islands                  11

     5    Frequency Of Contamination Above RBSLs
               At Facilities                                      12

     6    Affected Media And South Carolina RBCA
               Classification Comparison                         13


                             Figures

Number                                                     Page

    1      Contamination At 474 Dispenser Islands                10

    2      Contamination At 370 Facilities                        10

    3      Contamination For Pressure Piping Systems -
               Dispenser Islands                               11

    4      Contamination For Suction Piping Systems -
               Dispenser Islands                               11

    5      Contamination For Pressure Piping Systems - Facilities   12


    6      Contamination For Pressure Piping Systems - Facilities   12

-------
Introduction

       In this document, EPA presents the results of information collected and analyzed from
underground storage tank (UST) closure and assessment reports at sites in South Carolina. The
report discusses the background, purpose, methodology used, quality assurance and quality
control procedures applied, results of the study, and conclusions.  Supporting information is in
the appendices.
Background

       In 1984, Congress responded to the increasing threat to groundwater posed by leaking
underground storage tanks by adding Subtitle I to the Resource Conservation and Recovery Act
(RCRA). This legislation required the U.S. Environmental Protection Agency (EPA) develop a
comprehensive regulatory program for USTs storing petroleum or certain hazardous substances
in order to protect the environment and human health from releases. EPA's Office of
Underground Storage Tanks (OUST) developed regulations for owners and operators of UST
systems covering leak detection, addressing release prevention, and minimizing the potential for
USTs to corrode. (The regulations also cover additional requirements, such as financial
responsibility and corrective action, not discussed in this report.) The regulatory requirements
were phased in over ten years; the final deadline was  December 1998.

       Federal and state regulations helped EPA achieve its mandate. As a result of these
regulatory changes, USTs are now significantly more protective of human health and the
environment than those installed before the requirements were implemented. Despite these
improvements, however, releases from new and upgraded UST systems continue. To respond to
this, OUST initiated an effort to evaluate the performance of UST systems.  The specific
objectives of the effort are:

•     Identify the most common sources and causes of releases.
•     Determine how successful leak detection methods are at detecting releases.
•     Determine what, if any improvements are needed in the UST program.

       EPA has identified releases from UST dispensing systems as a potential concern for the
national UST program and state UST regulatory agencies.  EPA has gathered anecdotal evidence
from representatives of regulatory agencies; the regulated community; UST equipment sales,
repair, maintenance, and consulting companies; and UST remediation firms which repeatedly
cited fuel dispensers as a source of contamination at UST sites.

       Current federal UST leak prevention and leak detection regulations do not address
dispensers. In addition, the federal regulations do not explicitly require assessment under
dispensers at the time of UST system closure. The requirements for assessment at closure are
established in 40 CFR Part 280, Subpart G (280.70-280.74,  Out-of-Service UST Systems and
Closure). This subpart specifies that assessment is required for the places where contamination

-------
is most likely present at an UST site.1 Initially, the most likely sites of release were usually
considered to be under the tank or along the piping system.

       Some states now specify requirements for sampling under UST dispenser systems
because of this evidence of dispenser releases. For example, Florida now requires sampling near
dispensers during site investigations and under dispenser containment (UDC) to contain releases
that do occur. Industry has also recognized the potential for dispenser releases and, as a result,
installing under dispenser containment is becoming increasingly common for new installations.

       South Carolina's Department of Health and Environmental Control (DHEC) first began
requiring dispenser sampling at the time of UST system closure2 in 1990. Additionally, South
Carolina determined sampling under a dispenser may also be required if contamination is
observed during a compliance inspection.  The requirement to sample near or under a dispenser
is specified in DHEC's 2000 guidance document Underground Storage Tank Assessment
Guidelines for Permanent Closure and Change-In-Service. This current guidance specifies:

        All dispenser islands should be sampled. If the dispenser island is located above
       or immediately adjacent (less than five feet) to the UST, the sample for the  island
       can be incorporated into the sample for that UST.  Otherwise, dispenser islands
       should be individually sampled. Samples should be collected approximately two
       feet below the bottom of the associated piping. 4

       EPA  determined that South Carolina's history of sampling  under dispensers provided an
excellent source of data for examining the frequency of under dispenser contamination. In order
to better evaluate the frequency and severity of releases under dispensers, EPA partnered with
DHEC, which has a fairly robust  set of data on petroleum contamination found during UST
system inspections  and closures, including contamination under dispensers.

       As mentioned above, it has become common practice in recent years to install UDC (e.g.
dispenser pans/sumps) at the time an UST system is installed.  Previously, dispenser pans were
very rarely installed at the time of installation, so most older dispensers do not have UDC.
Because the  facilities covered in this study were all closed by 1999, EPA assumed that all of
these systems were installed before it was common practice to install dispenser pans. Therefore,
the results of this study are assumed to reflect systems without UDC, although a good portion of
the current UST population does  have UDC, particularly new facilities.
     1 40 CFR Part 280 Technical Standards and Corrective Action Requirements for Owners and Operators
     of Underground Storage Tanks (USTs), Subpart G  280.70-280.74, Out-of-Service UST Systems and
     Closure. U. S.  Environmental Protection Agency.
     2 Procedures for Abandonment of Underground Storage Tanks. July 25, 1990. Underground Storage
     Tank Program, South Carolina Department of Health and Environmental Control, Columbia, South
     Carolina.
     3 Communication with Mark K. Berenbrok. April 2002. Underground Storage Tank Program, South
     Carolina Department of Health and Environmental Control.
     4 Underground Storage Tank Assessment Guidelines for Permanent Closure and Change-In-Service.
     August 2000. Underground Storage Tank Program, South Carolina Department of Health and
     Environmental Control, Columbia, South Carolina. The term dispenser island is used to refer to the
     raised (usually approximately 6 inches high) concrete platform on which one or more fuel dispensers are
     installed. Dispenser islands are normally 2 to 4 feet wide and may be any length.
                                                                                          2

-------
Purpose Of This Study

       The purpose of this study is to gather and analyze dispenser sampling data5 from South
Carolina DHEC underground storage tank assessment and closure files to determine whether the
data showed any patterns of dispenser releases and to identify the likelihood that releases from
dispensers will contaminate subsurface media (soil and groundwater). These data provide
information about the frequency and impact of releases under dispensers.
Methodology

       This report reflects the results of two separate study phases during which data about
petroleum releases from dispensers associated with USTs was collected and analyzed. The data
came primarily from closure reports, 95 percent of which were received by DHEC between 1996
through 1999.6 South Carolina's regulations require a closure report for the permanent closure
or change in service of an UST system. The closure report must address all components of the
tank system, including tanks, piping, and dispensers. The sites with closure reports on file did
not necessarily have a confirmed release.  Excluding observations made by DHEC personnel, a
release is confirmed when analytical results indicate petroleum constituents are present in excess
of detection limits. Sampling under dispensers was part of site investigations to determine
whether any soil or groundwater contamination existed at the UST facility as a result of a release
from the UST system. These files are maintained by DHEC's Underground Storage Tank
Program.  The first data collection phase was conducted by DHEC in January 2002 to pilot the
data collection techniques and forms. EPA, with contractor assistance, conducted the second
data collection phase in February and March 2002.  Below are more details about the data
collection effort. See page 8 for descriptions of the results.

       Data Collection

       DHEC performed the first phase of data collection in January  2002 to test the
effectiveness of the data collection form, identify unanticipated issues or problems, and gauge
the level of effort required to review the files. The first phase encompassed review of a set of
closure and assessment reports collected in March and April 1996. The reports included:
closure and assessment files which described sites with no contamination; contaminated sites that
had not been actively remediated but were determined to require no further action (NFA); sites
with confirmed releases that had been rehabilitated and determined to require NFA; and
confirmed releases that are still being remediated.
    5 In this study, the terms dispenser sampling data and dispenser data are used to refer collectively to
    measured concentrations of common petroleum components, such as benzene, in samples collected under
    a dispenser.
    6 In a few cases, assessment reports were the source of the site data. An assessment report can be a
    voluntary submittal prepared as a result of a pending property sale or loan application, or may be
    requested if an inspector observes site conditions that indicate a release may have occurred.

-------
       Data Screening

       The first step in collecting the data was to determine whether an UST site report was
valid or invalid for purposes of this study. File information reviewed during the first phase was
determined to be invalid (i.e., not useful for the purposes of this study) for any of the following
reasons:

•      No dispenser samples taken/analysis performed.
•      UST was within five feet of dispenser(s) with sampling/analysis data.
•      Contaminant levels were unknown.
•      No dispensers were present (e.g., UST used to store emergency generator fuel, motor oil,
       or waste oil).

       During the second data collection phase, the following additional criteria were also
applied to identify information that was considered invalid (for purposes of this study):

•      Contamination levels of all chemicals were non-detect, but detection limits for one or
       more samples were greater than state Risk Based Screening Levels (RBSLs).
•      File was incomplete and the reviewer could not determine whether samples were
       collected under the dispenser and/or whether contamination was present.
•      Tank to dispenser distance  was unknown.
•      Sample locations were unclear.
•      Location of dispensers was unknown.

       Sites that only contained emergency generator fuel, motor oil, or waste oil were
considered invalid in the data screening for both phases. For the most part, such sites were
excluded from detailed review because dispensers are generally not used with tanks storing these
substances. If a dispenser was within five feet of an UST at a facility, but samples were
collected at other dispensers located more than five feet from an UST (and otherwise meeting the
above criteria), then the whole facility was considered valid. In such cases, dispensers with valid
data were included in the population of data collected.

       Data Recording

       During the first phase of data collection and as each report was reviewed, the reviewer
completed a data form and recorded selected information about the facility and the release (see
Appendix I on page 13 for a copy  of the blank form). During the second phase, data from
closure reports were entered directly into an electronic form created in an Access database. Data
were entered at DHEC  offices in Columbia, South Carolina.

       More information was gathered about each report during the second phase  of data
collection (see Appendix 2 on page 14) compared to the first data collection. Descriptions of the
data elements collected during both phases are provided below.  Data elements that were only
included in the second phase are identified in the descriptions that follow.

-------
•      Facility ID - Unique site identification number assigned by DHEC.

•      Date Closed - Date the UST(s) on site were closed and/or removed.

•      Number of Dispensers - Total number of dispensers at the site.

•      Invalid File - Box checked if report was invalid, followed  by reason(s) why it was
       invalid.

•      Information on samples taken at the dispenser:

             Dispenser Island ID - If more than one sample was taken on one island, the
             sample was labeled as "la", "Ib," "2a," "2b," etc. If one sample was taken at
             each island, the sample number correlated to the numbers of islands (e.g., sample
             from island #1 would be "1," sample from island #2 would be "2," etc.).
             Piping Type - Second phase only. One of three types of piping product delivery
             methods:  suction (S), pressurized (P), or unknown (U); additional details about
             piping are addressed in the "Note on Piping Delivery Types" section below.
             Contamination - If contamination was present, data recorded as "Yes"; specific
             contamination levels addressed below.
             Above State RBSLs (Risk Based Screening Levels) - Entered as "Yes" or "No"
              in accordance with contamination levels compared to South Carolina PvBSLs.7
             Benzene, Ethylbenzene, Toluene, Xylenes (total), and Naphthalene
             Contamination Levels - If above detectable limits, recorded in ppm (mg/kg); if
             non-detect, recorded as less  than detection limit or non-detect, as appropriate.

•      Contaminants above RBSLs for the dispenser only - Second phase only. If
       contamination was present at levels above RBSLs at the dispenser only (and nowhere
       else at the site), this box was checked. Then EPA obtained a summary report from
       DHEC's corrective action database  and recorded the following information:

              Contaminant Medium - Select whether the contamination reached soil or
              groundwater, both, or unknown.
              South Carolina RBCA Classification - If the site  was classified, enter the South
              Carolina Risk Based Corrective Action Classification (SCRBCA) level (see
             Appendix 3 on page 15 for details about the SCRBCA classification levels).
              Receptor type - Multiple choices, including public well, creek, property
              boundary, etc.

       South Carolina's risk-based approach to prioritizing cleanups uses RBSLs established for
two types of soils - sand and clay. In order to determine whether contamination under a
dispenser exceeded South Carolina RBSLs for each of the contaminants examined in this study,
the soil type present at the site had to be identified. If the soil type was not provided in the report
 South Carolina Risk-Based Corrective Action for Petroleum Releases, May 15, 2001. Risk-based screening levels
for groundwater, soil, and air are listed in Appendix B.

-------
and the contamination level reported fell between two South Carolina RBSLs, then EPA
determined the soil type from a map indicating the approximate location of the geologic fall
line.8 Sites to the northwest of the fall line were classified as clays and sites to the southeast
were considered to be sands for the purpose of determining whether the dispenser sample
concentration was above or below the RBSL for each contaminant.

       Collection of information from DHEC's corrective action database was added as part of
the second phase data collection effort to allow additional comparisons of the characteristics of
the facilities with dispenser releases that may require corrective action with a statewide summary
of facility releases available from the corrective action database.  DHEC provided the statewide
summary discussed in the results section on page 8.

       If a report was determined to be  invalid, the "Site ID," "Date Closed," "Invalid File," and
the appropriate "Reason File Invalid" boxes were the only fields completed on the form. Valid
files included sites with at least one soil or groundwater sample (and analytical data) taken at a
dispenser. The analytical results are likely to be representative of contamination, or lack of
contamination, that can reasonably be attributed to the operation of the dispenser itself and not
attributed to a release from other portions of the UST system.

       Note On Dispenser Islands

       An UST facility may have one or more dispenser islands and each dispenser island may
have one or more dispensers. DHEC regulations require that the facility closure process include
collection of at least one soil sample at each dispenser island. However, sampling at each
dispenser is not required. Results from  samples collected for more than one dispenser on a
dispenser island may be submitted voluntarily. If the dispenser island is located above or
immediately  adjacent to (less than five feet from) the UST, then DHEC allows a single sample to
meet both a dispenser island and UST sampling requirement.  DHEC requires that samples be
collected approximately two feet below the bottom of the piping of the dispenser chosen for
sampling and that the dispenser with the highest anticipated level of contamination be sampled.

       Note  On Contaminant Data

       Information from the assessment and  closure reports for soil  contamination levels of
benzene, ethylbenzene, toluene, xylene, and naphthalene were recorded in milligrams per
kilogram (mg/kg) (parts per million [ppm]) for samples taken at the  dispenser(s). If the various
species of xylene were reported separately (i.e., xylene [m, o, and p]), then the total of those
results was recorded as a single value in the database (i.e., concentration of m-xylene +
concentration of o-xylene + concentration of p-xylene). In the case of other contaminants  when
two analytical results were provided, the higher concentration was always used. If no
contamination was present (reported as a "non-detect"), then the result was recorded as less than
the reported laboratory detection limit for each contaminant (e.g., "<0.0005 ppm"). If both
results for a contaminant were non-detect and the detection limits varied, then the higher
detection limit was reported.
 1 The boundary zone between older, resistant rocks and younger Coastal Plain sediments.

-------
       As noted above at sites with detected contamination, EPA used the available data to
determine whether the contamination levels were above or below RBSLs. The data available in
the closure reports are based on the results of standard laboratory analyses for the detection and
quantification of common petroleum constituents. The accuracy (i.e., the closeness of the
measured value to the true value) of the laboratory analysis data depends on a wide range of
factors including the contaminant measured, the concentration level, the type of sample (e.g.,
soil, water), the analytical method, and the equipment used. EPA's "Test Methods for
Evaluating Solid Wastes" (SW-846) reports single-laboratory accuracy for Method 5035
(appropriate for the contaminants included in this study) applied to soil samples that approach
+/- 20 percent. Use of multiple laboratories (as is the case here) introduces further variation in
the measured results. To evaluate the potential effects of these accuracy limitations on the
results of this study, EPA identified the situations where the characterization of a data point as
being "above" or "below" the RBSL was based on a reported concentration that is within 30
percent of the RBSL. EPA found this occurred between 5  and  10 percent of the time. The
percentage of measurements characterized as "above RBSLs" that might actually be below
RBSLs (assuming a +/- 30 percent margin of error)  is similar to the percentage characterized as
"below RBSLs" that actually might be above RBSLs. Therefore, the effect of this inherent
analytical uncertainty on conclusions drawn from the data  is thought to be small.

       There has been increasing concern about contamination of groundwater sources from the
gasoline additive methyl tertiary butyl ether (MTBE). Data on MTBE were not available for this
study because DHEC did not require analysis of MTBE in soil  samples from UST sites.  (Since
1995, DHEC has required groundwater encountered during tank closure or assessment activities
to be sampled for MTBE.) However, the results of this analysis may be useful in trying to
control the release of MTBE - and other constituents found in gasoline - into groundwater.

       Note On Piping Product Delivery Types

       EPA collected data on the types of pumps used to deliver product through the piping at
the sites selected for inclusion in the database.  The  term "Piping Type" was used on the  form in
the second phase of data gathering to indicate the type of pump used to deliver product from the
tank to the dispenser. The piping type (pump) information was collected to determine whether
the presence of the pump in dispensers using suction systems would increase or decrease the
likelihood of a release.  For those sites at which the  pump type  was clearly identified, EPA
recorded the pump type under piping type on the form. For those sites with multiple tanks and
different types of pumps, and where reviewers could not determine which type of pump applied
to the dispenser island sampled, the piping type on the form was marked as unknown.  The pump
type generally  was indicated as unknown for the assessment and closure reports reviewed during
the first phase of the data collection effort because information on pump type was not collected
during that phase. For sites reviewed in the first phase that appeared to have dispenser
contamination, information on pump type was collected during the second phase.

-------
Quality Assurance/Quality Control

       The following Quality Assurance (QA) strategy was used during the second phase data
collection effort, which last two weeks. As part of this strategy, Quality Control (QC)
procedures were developed to ensure the quality of the information gathered.  While on site,
EPA gathered data from DHEC's existing files and recorded the information in an Access
database.

       Because two individuals entered data from DHEC files into the database, each individual
performed a QC check of the data entered by the other. During the first week of data collection,
each individual checked every fifth file entered into the database by the other individual,  for a
review of 20 percent of the data entries. Every field was checked for accuracy on every file
reviewed. During the second week of data collection, EPA assessed this procedure and
determined it to be more extensive than needed because of the experience of the data collectors
and because few errors were detected during the first week. EPA decided that from then  on,
every tenth file would be checked, for a review of 10 percent of the data entries.  This change
was implemented for the second week of the two-week data collection effort. Overall, 15.5
percent of the data entries were reviewed for QC (see Table 1).
                       Table 1 - Summary Of Quality Control Data

            Total number of files reviewed                      1,218
            Number of files double-checked for QC                 189
            Percentage of files double-checked for QC              15.5
            Percentage of files double-checked with errors            3
       Double-checking of the files was performed at the end of each workday or the beginning
 of the next.  Each data collector kept a separate list of the files she double-checked throughout
 the data collection period.

       To verify the information recorded in the database was correct, the individual who had
 not originally entered the data examined each file selected for QC review in detail. Where
 necessary, the South Carolina corrective action database was consulted.  If a discrepancy or error
 was discovered, this was noted and discussed that day with the other data collector.  All
 discrepancies or errors were resolved and any necessary changes were made in the database.
 Overall, approximately 3 percent of the files reviewed were revised in the database as a result of
 this review process.

       While no QA/QC strategy was used during the first phase of the study, 12 of the 77 files
 were reviewed again during the second phase to determine if any files contained samples with
 contamination above RBSLs at the dispenser only.

-------
       EPA provided copies of the draft report to EPA regional offices, states, and industry
experts for peer review. The reviewers identified issues and provided recommendations about
the report. EPA evaluated the comments and incorporated many of the recommendations into
this final version of the report.
Results

       For both phases of this study, a combined total of 1,218 files were reviewed, each
representing an individual closure action (or in a few cases a voluntary site assessment). DHEC
reviewed 77 files in the first phase, of which 54 were invalid; EPA reviewed 1,141 files in the
second phase, of which 794 were invalid. The files reviewed are assumed to be generally
representative of closure actions undertaken in South Carolina during the period of interest. Of
the 1,218 files reviewed, 370 (30.4 percent) provided at least one "valid" (for purposes of this
study) sampling result.  Of the 848 files without a valid sampling result, 269 (32 percent) were
invalid because they lacked dispenser sampling data and 515 (61 percent) were invalid because
dispenser(s) with sampling/analysis data were within five feet of an UST. The other 64 files
were invalid because they did not have dispensers (17); had all contaminants reported non-detect
but with detection limits above RBSLs for one or more contaminants  (16);  or were incomplete in
some way (31) (e.g., sample locations unclear, tank to dispenser distance unknown).  Data for the
370 facilities with at least one valid sample are summarized in Appendix 4 on page 17 and
discussed below.

       The 370 facilities with valid data had a total of 454 dispenser islands with valid data. As
shown in Table 2, contamination was found under dispensers at nearly half of all facilities and
dispenser islands with valid data and dispenser contamination that exceeded RBSLs was found at
just over 25 percent of all facilities and dispenser islands.  The distribution of these data by pump
type is provided in the tables that follow. Of the 454 dispenser islands with valid data, 56 had
data for more than one soil sample.  The data for 40 (71 percent) of these 56 dispenser islands
indicate that contaminant levels in the multiple samples fall in the same contaminant grouping
(i.e., not detected, at least one contaminant detected, or at least one contaminant above RBSL).
Of the 40 dispenser islands with contaminant levels in multiple samples in  the same contaminant
grouping, 27 had no detectable contamination in all samples. For 13 of the remaining 16
dispenser islands from the group of 56 with data for more than one soil sample, at least one
sample had a detected contaminant while one  or more samples had no detected contaminants.
         Table 2 - Summary Of Contamination At Dispenser Islands And Facilities
Total
Number
Dispenser 4$4
Islands
Facilities 370
With At Least One
Contaminant Detected
#
215
179
% (95% Confidence
Limits)
47.4(42.8-51.9)
48.4 (43.3 - 53.5)
With At Least One Contaminant
Above RBSL
#
116
98
% (95% Confidence
Limits)
25.6(21.5-29.6)
26.5(22.0-31.0)

-------
                Figures 1 and 2 visually depict the number of dispenser islands and facilities,
         respectively, where contamination was discovered, and the number of times the contamination
         exceeded RBSLs.
  Figure 1 - Contamination At 454 Dispenser
                   Islands
                                     Figure 2 - Contamination At 370 Facilities
    No
Contamination
    239
Contamination
    215
    No
Contamination
    191
Contamination
    179
                                      Above RBSL
                                         116

                                      Below RBSL
                                         99
                                                                     Above RBSL
                                                                         98
                                                                     Below RBSL
                                                                         81
                Table 3 shows the distribution of the numbers of sampled facilities, dispenser islands, and
         individual samples by pump type.
                        Table 3 - Numbers Of Sampled Facilities, Dispenser Islands,
                                       And Samples By Pump Type
Pump Type
All
Pressure
Suction
Unknown
Facilities
370
135
198
37
Dispenser
Islands
454
183
218
53
Samples
534
211
261
62
                Table 4 shows the frequency of detected contamination and contamination above RBSLs
          for the 454 dispenser islands.  The top portion of Table 4 shows the number of dispenser islands
          with no detected contamination, at least one contaminant detected, and at least one contaminant
          above the RBSL. The bottom portion shows the binomial confidence limits for these
          proportions. As shown, the proportions of pressure and suction piping systems with and without
          detected contamination appear similar. This is generally confirmed by the results of three
          statistical significance tests - Pearson's Chi-Squared, Fisher's Exact Test, and a parametric test
          based on the standard error of the differences in proportions.  The results of these tests, which are
          presented in Appendix 5 on page 25 along with results of other supporting statistical analysis,
          indicate that the small differences in proportions are not significant, with the exception of
                                                                                               10

-------
        dispenser islands without contamination, where the difference between pressure and suction
        piping systems is marginally significant (p values approach 0.05).
                Table 4 - Frequency Of Contamination And Contamination Above RBSLs
                                          At Dispenser Islands
Number Of Islands
Pump Type
All
Pressure
Suction
Unknown

Pump Type
All
Pressure
Suction
Unknown
With Valid
Data
454
183
218
53
Proportion
Percent With
Valid Data
100.0
40.3
48.0
11.7
With No Detected
Contamination
239
84
118
37
Of Islands, Percent (95
Percent With No
Detected
Contamination
52.6 (48.0 - 57.2)
45.9(38.7-53.1)
54.1(47.5-60.7)
69.8(57.4-82.1)
With At Least One
Contaminant
Detected
215
99
100
16
With At Least One
Contaminant
Above RBSL
116
49
59
8
% Confidence Limits)
Percent With At
Least One
Contaminant
Detected
47.4(42.8-51.9)
54.1 (46.9-61.3)
45.9 (39.3 - 52.5)
30.2(17.8-42.5)
Percent With At
Least One
Contaminant
Above RBSL
25.6(21.5-29.6)
26.8 (20.4 - 33.2)
27.1 (21.2-33.0)
15.1 (5.4-24.7)
               Figures 3 and 4 visually depict the number of dispenser islands with pressure and suction
        piping systems, respectively, where contamination was discovered under the dispenser, and the
        number of times the contamination exceeded RBSLs.
     Figure 3 - Contamination For Pressure Piping
              Systems - Dispenser Islands
                              Figure 4 - Contamination For Suction Piping
                                      Systems - Dispenser Islands
    No
Contamination
    84
Contamination
    99
    No
Contamination
    118
Contamination
    100
                                       Above RBSL
                                           49
                                        Below RBSL
                                           50
                                                                Above RBSL
                                                                   59
                                                                Below RBSL
                                                                    41
                                                                                            11

-------
                Table 5 is organized identically to Table 4, but shows the frequency of detected
         contamination and contamination above RBSLs for the 370 facilities.  The results of the three
         statistical significance tests, which are presented in Appendix 5 on page 25 along with results of
         other supporting statistical analysis, indicate that the small differences in proportions are not
         significant.
           Table 5 - Frequency Of Contamination And Contamination Above RBSLs At Facilities
Numbers Of Facilities
Pump Type
All
Pressure
Suction
Unknown

Pump Type
All
Pressure
Suction
Unknown
With Valid
Data
370
135
198
37
Proportions
Percent With
Valid Data
100.0
36.5
53.5
10.0
With No Detected
Contamination
191
65
103
23
Of Facilities, Percent
Percent With No
Detected
Contamination
51.6(46.5-56.7)
48.1(39.6-56.2)
52.0 (45.0 - 58.9)
62.2 (46.6 - 77.8)
With At Least One
Contaminant
Detected
179
70
95
14
With At Least One
Contaminant Above
RBSL
98
34
57
7
(95% Confidence Limits)
Percent With At
Least One
Contaminant
Detected
48.4 (43.3 - 53.5)
51.9(43.4-60.3)
48.0(41.0-54.3)
37.8 (22.2 - 53.5)
Percent With At
Least One
Contaminant Above
RBSL
26.5(22.0-31.0)
25.2(17.9-32.5)
28.8(22.5-35.1)
18.9(6.2-31.5)
                Figures 5 and 6 visually depict the number of facilities with pressure and suction piping
         systems, respectively, where contamination was discovered under the dispenser, and the number
         of times the contamination exceeded RBSLs.
     Figure 5 - Contamination For Pressure Piping
                  Systems - Facilities
    No
Contamination
    65
Contamination
    70
                                       Above RBSL
                                           34
                                        Below RBSL
                                            36
                                Figure 6 - Contamination For Suction Piping
                                            Systems - Facilities
    No
Contamination
    103
Contamination
    95
                                                                   Above RBSL
                                                                       57
                                                                   Below RBSL
                                                                       38
                                                                                               12

-------
       Table 6 provides a comparison of the media affected and the South Carolina RBCA site
priority classification (see Appendix 3 on page 15) for the 19 sites found to have contamination
above RBSLs only at the dispenser and data for all releases in South Carolina for the period
January 1996 through June 2002.  As shown, the proportions of facilities that affected
groundwater (approximately three-quarters) or only soil (approximately one-quarter) are similar
for the facilities with dispenser-only contamination and for all facilities with releases. Although
there are some differences with respect to the South Carolina RBCA classifications, the
relatively small number of facilities with dispenser only contamination makes it difficult to
conclude that these differences are significant.

       As discussed on page 3 in the methodology section, information on the type of receptor
potentially affected by contamination was included in the data collection effort.  Because these
data do not indicate whether receptors have actually been affected or not, EPA later decided
comparison  of these data for the facilities studied with other facilities in the state would not yield
meaningful results.
                   Table 6 - Affected Media And South Carolina RBCA
                                Classification Comparison

                        Facilities With Dispenser-      All Facilities With Release From
                          Only Contamination	January 1996 Through June 2002
Number
Total # Of Facilities
Affected Media
Groundwater
Soil Only
South Carolina RBCA**
1
2B
3B
4B
5
Other
19

14
5

1
2
6
1
9
0
Percent*


74
26

5
10
32
5
47
0
Number
1581

1103
478

39
309
433
61
639
100
Percent*


70
30

3
20
27
4
40
6
 * Totals may not add to 100 due to rounding.
 ** See Appendix 3 (page 15) for descriptions of these categories.
 Conclusions

       For closed facilities in South Carolina, which EPA assumes had no under dispenser
 containment, releases to the environment near dispensers appear to be quite common; nearly one-
 half (48 percent) of all facilities included in the study show contamination detected under one or
 more dispenser islands. This result is likely somewhat lower than actual contamination rates
 because data was often not available for every dispenser or even every dispenser island at a
 facility. The frequency of these releases appears to be affected little, if at all, by the type of
                                                                                       13

-------
pump delivery system (pressure or suction) used at the facility. At about 10 percent (19) of the
facilities with contamination above RBSLs detected under a dispenser island, the dispenser
appears to be the only source of release.  Based on a comparison with corrective action data for
all releases in South Carolina, the consequences of these dispenser releases in terms of the media
affected and the need for corrective action (as measured by the South Carolina RBCA
classification) appear to be comparable to those from other sources of release, such as tanks and
piping.
                                                                                       14

-------
                         Appendix 1 - South Carolina Dispenser Data Gathering Form For First Phase
                              D INVALID FILE - Check here & indicate the reason below:
                                     n No dispenser samples taken/analysis performed.
                                     a UST was within five (5) feet of the dispenser.
                                     n Contaminant levels are unknown.
                              n Other:
Facility ID:
Date Closed:	
Number Dispensers:
Assessment Report / Closure Report  (Circle One)
Dispenser
Island ID




Contamination
Above State RBSLs
(Yes/No)








Benzene
Ethylbenzene
Toluene
Xylenes
Naphthalene
(mg/kg)




















Instructions:
1. Dispenser Island ID - If more than 1 sample is taken at an island, ID the islands as la and Ib.
2. If more than one constituent is reported [i.e., for xylenes - xylene (m,p) and zylene (o)], add those numbers together for reporting purposes.
3. If value is not reported, indicate with N/A.
4. If value below detection level, indicate the detection limit by using "<" followed by the detection limit value.

SC RBSLs in Clays (mg/kg):  Benzene 0.008; Ethylbenzene 6.168; Toluene 1.167; Xylenes 22.495;  Naphthalene 0.069
SC RBSLs in Sands (mg/kg):  Benzene 0.007; Ethylbenzene 1.150; Toluene 1.450; Xylenes 14.500;  Naphthalene 0.036
                                                                                                                            15

-------
                           Appendix 2 - South Carolina Dispenser Data Gathering Form For Second Phase
    Site ID:
Date Closed:
Number of Dispensers:
D INVALID FILE - Check here & indicate the reason below:
       n No dispenser samples taken/analysis performed.
       n UST was within five (5) feet of the dispenser.
       n Contaminant levels are unknown.
D Other:
                            D Contaminants are above RBSLs for the dispenser only
                            (If yes, print out a corrective action summary report using the
                            site ID)
                                   Contaminant Medium	
                                   SC RBCA Classification	
                                   Receptor Type	
Dispenser
Island ID




Piping Type
(S) = Suction
(P) — Pressurized
(U) = Unknown




Contamination
Above State
RBSLs
(Yes/No)








Benzene
Ethylbenzene
Toluene
Xylenes
Naphthalene
(mg/kg)




















    Additional Comments:

    SC RBSLs in Clays (mg/kg):  Benzene 0.008; Ethylbenzene 6.168; Toluene 1.167; Xylenes 22.495; Naphthalene 0.069
    SC RBSLs in Sands (mg/kg):  Benzene 0.007; Ethylbenzene 1.150; Toluene 1.450; Xylenes 14.500; Naphthalene 0.036
                                                                                                                          16

-------
      Appendix 3 - South Carolina RBCA Site Priority Classification System

1.  Sites are placed in Classification 1 if:

   1A    an emergency situation exists
   IB    a fire or explosion hazard exists
   1C    vapors or free product exists in a structure or utility
   ID    concentrations of chemical of concern (CoC) have been detected in a potable
          water supply or surface water supply intake
   IE    free product exists on surface waters
   IF    CoC exists in surface water

2.  Sites are placed in Classification 2 if:

   Classification 2a:

   2AA  a significant near term (0 to 1 year) threat to human health, safety, or sensitive
          environmental receptors exists
   2AB  potable supply wells or surface water supply intakes are located < 1 year ground
          water travel distance downgradient of the source area

   Classification 2b:

   2BA  free product exists in a monitoring well at a measured thickness > 1 foot
   2BB  potable supply wells or surface water supply intakes are located < 1000 feet
          downgradient of the source area (where groundwater velocity data is not
          available)

3. Sites are placed  in Classification 3 if:

   Classification 3a:

   3AA  a short-term (1 to 2 years) threat to human health, safety, or sensitive
          environmental receptors exists
   3AB  potable supply wells or surface water supply intakes are located > 1 year and < 2
          years groundwater travel distance downgradient of the source area
   3AC  sensitive habitats or surface water exist < 1 year groundwater travel distance
          downgradient of the source area and the groundwater discharges to the sensitive
          habitat or surface water

   Classification 3b:

   3BA  free product exists in a monitoring well at  a measured thickness > 0.01 foot
   3BB  concentrations of CoC above the RBSL have been detected in a non-potable water
          supply well
   3BC  hydrocarbon containing surface soil (< 3 feet below grade) exists in areas that are
          not paved

                                                                                   17

-------
   3BD   sensitive habitats or surface water used for contact recreation exist < 500 feet
          downgradient of the source area (where groundwater velocity and discharge
          location data are not available)
   3BE   the site is located in a sensitive hydrogeologic setting, determined based on the
          presence of fractured or carbonate bedrock hydraulically connected to the
          impacted aquifer
   3BF   groundwater is encountered < 15 feet below grade and the site geology is
          predominantly sand or gravel

4. Sites are placed  in Classification 4 if:

   Classification 4a:

   4AA  a long term (> 2 years) threat to human health, safety, or sensitive environmental
          receptors exists
   4AB   potable supply wells or surface water supply intakes are located > 2 years and < 5
          years groundwater travel distance downgradient of the source area
   4AC   non-potable supply wells are located < 1 year ground water travel distance
          downgradient of the source area

   Classification 4b:

   4BA  free product exists as a sheen in any monitoring wells
   4BB   non-potable supply wells are located < 1000 feet downgradient of the source area
          (where groundwater velocity is not available)
   4BC   the ground water is encountered < 15 feet and the site geology is predominantly
          silt or clay

5. Sites are placed  in Classification 5 if:

   5A    there is no demonstrable threat, but additional data are needed to show that there
          are no unacceptable risks posed by the site
          assessment data for the site indicate concentrations in some samples are above the
          RBSL or Site-Specific Target Level (SSTL), as appropriate, and further
          assessment is needed
   5B    assessment data for the site indicate concentrations in samples are below the
          RBSL or SSTL, as appropriate, but the samples are determined not to be
          representative; therefore, further assessment is  needed
                                                                                   18

-------
Appendix 4 - Summary Of Results
Total number of facilities reviewed
Number of facilities with valid data
Number of dispenser islands with valid data
Number of facilities with invalid data
Number of facilities with contamination above RBSLs at dispenser only
Number of facilities with detectable contamination under dispensers*
Number of facilities with detectable contamination above RBSLs
Number of facilities with detectable contamination below RBSLs
Number of facilities with no detectable contamination under dispensers
Number of dispenser islands with detectable contamination under
dispensers*
Number of dispenser islands with detectable contamination above
RBSLs
Dumber of dispenser islands with detectable contamination below
RBSLs
Number of dispenser islands with no detectable contamination under
dispensers
1,218
370
454
848
19
179
98
81
191
215
116
99
239
Pressurized Pumping System Data
Number of facilities with pressurized pumping
Number of facilities with pressurized pumping systems and
contamination under dispensers*
Number of facilities with pressurized pumping systems and
contamination under dispensers above RBSLs
Number of facilities with pressurized pumping systems and
contamination under dispensers below RBSLs
Number of facilities with pressurized pumping systems and no
detectable contamination under dispensers
lumber of dispenser islands with pressurized pumping
lumber of dispenser islands with pressurized pumping systems and
contamination under dispensers*
Number of dispenser islands with pressurized pumping systems and
contamination under dispensers above RBSLs
Number of dispenser islands with pressurized pumping systems and
contamination under dispensers below RBSLs
Number of dispenser islands with pressurized pumping systems and no
detectable contamination under dispensers
135
70
34
36
65
183
99
49
50
84

                                                  19

-------
Suction Pumping System Data
Number of facilities with suction pumping
Number of facilities with suction pumping systems and contamination
under dispensers*
lumber of facilities with suction pumping systems and contamination
under dispensers above RBSLs
lumber of facilities with suction pumping systems and contamination
under dispensers below RBSLs
lumber of facilities with suction pumping systems and no detectable
contamination under dispensers
Number of dispenser islands with suction pumping
Number of dispenser islands with suction pumping systems and
contamination under dispensers*
Number of dispenser islands with suction pumping systems and
contamination under dispensers above RBSLs
Number of dispenser islands with suction pumping systems and
contamination under dispensers below RBSLs
Number of dispenser islands with suction pumping systems and no
detectable contamination under dispensers
198
95
57
38
103
218
100
59
41
118
Unknown Pumping System Data
Number of facilities with unknown pumping
Number of facilities with unknown pumping systems and
contamination under dispensers*
lumber of facilities with unknown pumping systems and
contamination under dispensers above RBSLs
Number of facilities with unknown pumping systems and
contamination under dispensers below RBSLs
Number of facilities with unknown pumping systems and no detectable
contamination under dispensers
Number of dispenser islands with unknown pumping
Number of dispenser islands with unknown pumping systems and
contamination under dispensers*
Number of dispenser islands with unknown pumping systems and
contamination under dispensers above RBSLs
Number of dispenser islands with unknown pumping systems and
contamination under dispensers below RBSLs
Number of dispenser islands with unknown pumping systems and no
detectable contamination under dispensers
37
14
7
7
23
53
16
8
8
37
* Number in shaded box indicates total of two subsets of data below, which include contamination detected above
and below state RBSLs.
                                                                                                  20

-------
Dispenser Data
Number of dispensers at facilities with valid data
Average number of dispensers per facility
1,251
3.38
Values within 30% of RBSL value
Number of facilities with contamination above RBSLs within 30% of
sand RBSL value
Number of facilities with contamination below RBSLs within 30% of
sand RBSL value
Number of dispenser islands with contamination above RBSLs within
30% of sand RBSL value
Number of dispenser islands with contamination below RBSLs within
30% of sand RBSL value
5*
8
5* •
9
* An additional 6 facilities/dispenser islands were within 30% of the clay RBSL only
         Data For Facilities With Contamination Above RBSLs At Dispenser Only
Number Of
Records
14
5
Contaminant Medium
Groundwater
Soil
Number Of
Records
1
1
1
1
1
4
1
3
6
South Carolina RBCA Classification
ID
2BA
2BB
3BA
3BE
3BF
4BC
5A
5B
Chemicals detected in water
Free product >1 ft
Water supply wells <1000 ft downgrade
Free product >0.01 ft thick
Sensitive hydrologic setting
GW < 15 ft in sand or gravel
GW< 15 ft in silt of clay
No pending threat, additional data
Assessment data NOT conclusive
                                                                                       21

-------
                             Appendix 5 - Statistical Analysis

       This appendix presents the results of statistical analyses done to investigate whether
variation in facility closure date, size, or other characteristics might affect contamination patterns
and, thus, the conclusions that can be drawn from the data collected.

Characteristics Of Closures And Closed Facilities

       Valid data files included closures that took place between 1982 and 1999.  The primary
focus of the data collection effort was on closures that occurred between 1995 and the present,
but some earlier closures were also included (see Table A).  As shown, almost half of the
closures (162, or 44 percent) occurred in 1998. (The regulatory deadline for replacing or
upgrading tank systems was December  1998.)  The table also shows that the relative proportions
of closures at facilities with different piping delivery types changed somewhat over the time
period covered in the database, with the proportion of pressure (P) closures generally increasing
and the proportion of suction (S) facilities decreasing.
               Table A - Distributions Of Closures By Year And Pump Type

       ,.,            T, ,  , ^,,           Pressurized     Suction Facilities     Unknown
       Year        Total Closures   _   .....   _.    ,       _,.    ,       _  .....   „,    ,
                                      Facilities Closed       Closed       Facilities Closed
Pre-1996
1996
1997
1998
1999
All
20
52
80
162
56
370
1(5)
14(27)
21(26)
75(46)
24(43)
135(37)
11(55)
31 (60)
53 (66)
79 (49)
24 (43)
198 (54)
8(40)
7(13)
6(8)
8(5)
8(14)
37(10)
Note: Values outside of parentheses represent total number of facilities in the specific category; values inside
parentheses represent the percent of facilities in the specific category.
       The size of closed facilities, as indicated by the numbers of dispensers present, also
varied (Table B).1 The bulk of the closures took place at small facilities; slightly fewer than 30
percent of the facilities had only one dispenser, and about 84 percent had four dispensers or
fewer. The largest facility had 50 dispensers. There was a slight tendency for the largest
facilities that were closed (e.g., those with more than six dispensers) to have pressure piping.
Facilities with three or fewer dispensers were also more likely to have suction piping. There was
no significant correlation between closure date and size of facility.
 1 The data collection effort did not include collection of data on the number of dispenser islands at the closed
 facilities.
                                                                                          22

-------
         Table B - Size Distribution Of Facilities Closed (Number Of Dispensers)

    _.              _,  . , _,           Pressurized    Suction Facilities     Unknown
    Dispensers     Total Closures   _  .....   „,    ,       ^,   ,      „  .....   „,    ,
       F                             Facilities Closed       Closed      Facilities Closed
1 109 (29)
2 84 (23)
3 74 (20)
4 45 (12)
5 13 (4)
6-10 31 (8)
>10 14(4)
All 370(100)
44 (40)
24 (29)
26 (35)
15(33)
3(23)
16(52)
7(50)
135(37)
55(51)
50 (60)
43 (58)
24(53)
8(62)
13 (42)
5(36)
198 (54)
10(9)
10(12)
5(7)
6(13)
2(15)
2(7)
2(14)
37(10)
Note: Values outside of parentheses represent total number of facilities in the specific category; values inside
parentheses represent the percent of facilities in the specific category.
Analyses of Soil Contamination Data

       Valid data collected included 534 samples. Of the 370 facilities in the data set, 273 (74
percent) included only one sample result. A total of 454 dispenser islands were sampled at least
once. Multiple dispenser islands were sampled at 57 facilities, including 41 facilities with results
from two islands. The bulk of dispenser islands (398, or 88 percent) were sampled only once; 56
islands were sampled two, three, or four times each.  As evident from these data, the number of
dispensers at the facility is not necessarily representative of the number of dispenser islands
sampled. It is possible there were some islands at a given facility that were not sampled, or
which were not recorded on the data form; but for each  facility, at least one island had a valid
sample.

       Analytical results were reported for benzene, toluene, ethylbenzene, xylenes, and
naphthalene (Table C). Individual analytes were reported above detection limits in between 15.7
percent (benzene) and 34.6 percent (xylenes) of the chemical analyses. Detection limit values
were not reported for 28 samples (5 percent) of the samples.  Among the analytes, only
benzene's RBSLs are close to the typical analytical detection limits.

       Concentrations were highly variable, with relatively few high results (> 1 mg/kg).
Representative concentrations of toluene (geometric mean 0.085 mg/kg) and benzene (0.088
mg/kg) were the lowest, followed by ethylbenzene (0.159 mg/kg), with xylenes (0.294 mg/kg)
and naphthalene (0.352 mg/kg) having the highest geometric mean concentrations in samples
where contamination was detected.
                                                                                       23

-------
                        Table C - Summary Of Analytical Results
                                 Representative
I
Analyte

Benzene
Ethylbenzene
Toluene
Xylenes
Naphthalene
' requency Of
Detection
(percent)

15.7
24.9
29.2
34.6
31.8
^s VllA V-C11 U MllUlfl^
Geometric
Mean/Standard
Deviation
(mg/kg)
0.088/2.87
0.159/3.05
0.085 / 3.46
0.294/3.65
0.352/3.11
Representative
Detection Limit
(mg/kg)

0.001 -
0.001 -
0.001 -
0.001 -
0.002 -

0.005
0.005
0.005
0.015
0.005
RBSLs For
Sand/Clay
(mg/kg)

0.007 / 0.008
1.15/6.17
1.45/1.17
14.5/22.5
0.036 / 0.069
       Preliminary analysis suggested that average contaminant concentrations and frequencies
of detection were higher in samples from large facilities (i.e., those with larger number of
dispensers), and that contaminant concentrations were generally lower in samples from more
recent closures. EPA found the latter trend to be driven primarily by higher concentrations from
a relatively few early closures and there were no significant trends in contaminant concentrations
in closures after 1995. Results of the preliminary analysis prompted more detailed evaluation of
detection frequencies and exceedences of RBSLs discussed below.

Analysis Of Frequency Of Detection And Frequency Of Occurrence Above RBSLs

       The distribution of the numbers of sampled facilities, dispenser islands, and individual
samples, as well as the frequency of contaminant detection and the frequency of occurrence
above RBSLs are presented in Tables 2 through 4 in Section VI of this report.

       ICF used three significance tests to evaluate the differences in proportions of facilities
with detected contamination and with contamination above RBSLs: Pearson's Chi-Squared,
Fisher's Exact Test, and a parametric test based on the standard error of the difference in
proportions.2 All give results that are generally consistent, but vary slightly depending on the
size of the data set and the magnitude of the difference in proportions.

       Both the Fisher's exact test and Chi-Squared procedure assume that individual
observations are not correlated (are independent). Fisher's Exact Test is the most robust and
reliable test for small sample sizes, while the Chi-Squared generally is considered to be reliable
as long as the number of observations in the smallest group is not too small. The results of these
 : Sheskin, DJ, Parametric and Nonparametric Statistical Procedures, Chapman and Hall, New York, 200, p.381.
                                                                                      24

-------
two tests may differ slightly because the Chi-Squared test uses an approximation of the
distribution of difference between proportions, while Fisher's exact test calculates the difference
between proportions explicitly.  The parametric test assumes that the difference between the
proportions is normally distributed, and may therefore be somewhat less robust than the other
two procedures. EPA included this test as a contingency in case the data set resulted in Fisher's
calculations that exceeded the capability of a typical computer (which did not occur).

       Tables D and E present the frequencies of having at least one sample with detected
contamination or one sample with contamination above RBSLs at dispenser islands and
facilities, respectively, with pressure (P) and suction (S) piping systems. The modest differences
in the proportions of P and S islands with contamination are marginally significant (p - values
approach 0.05) by two of the three significance tests (Table D). The very small difference in the
proportions of P and S islands that have contamination above RBSLs is, as expected, not
significant. Similarly, the small differences in the frequencies of contamination seen at the P and
S facilities also are not significant by any of the statistical tests.
  Table D - Significance Test Results For Difference In Proportions Of Dispenser Islands
                                  With Contamination

                                  «,.A-         Fisher's Exact        Diff. in
                                                    Test          Proportions
      Proportions of Islands With, Without Detected Contamination	
      	PvsS	NS     	P-0.05	p ~ 0.05
      Proportions of Islands with Contamination > RBSLs	
              PvsS               NS               NS               NS
      Table £ - Significance Test Results For Differences In Proportions Of Facilities
                                  With Contamination
                                 ™ .A.         Fisher's Exact        Diff. in
                                                    Test          Proportions
       Proportions of Facilities With, Without Detected Contamination	
              PvsS               NS1               NS               NS
       Proportions of Facilities with Contamination Above one or More RBSLs
              PvsS               NS               NS               NS
       1.   NS = not significant at p = 0.05
 Frequency Of Contamination Above RBSLs As A Function Of Facility Size

       Table F shows the patterns of contaminant occurrence (detection and presence above
 RBSLs) in facilities of different sizes. When analyzed as a whole, the data generally indicate
 weak, insignificant, positive relationships between the frequency of contamination and the
                                                                                    25

-------
number of dispensers. This means the frequency of detected contamination increases with the
number of dispensers. However, this relationship is inconsistent. The smaller facilities (one
dispenser) have more frequent contamination than the middle-size facilities (four dispensers).
To the extent there is a positive trend, it appears to be limited to the largest facilities (>10
dispensers). The high frequency of contamination at these facilities (71.4 percent) and their
dispenser islands (64.9 percent) are driven by a relatively high frequency of contamination in
individual samples (61.9 percent).  The reasons for the apparent higher frequency of
contamination in samples from the largest facilities are not clear; they may reflect real
differences, or may be a statistical artifact associated with the relatively small number of samples
(42) from this size range.
    Table F - Frequency Of Contamination As A Function Of Facility Size (In Percent)
Number Of
Dispensers
1
2
3
4
5
6-10
>10
Facilities
Detected
51.4
45.2
40.5
46.7
53.8
56.7
71.4
>RBSL
28.4
19.0
24.3
22.2
30.8
41.9
42.9
Dispenser
Detected
51.8
44.3
42.4
42.2
46.7
45.5
64.9
Islands
>RBSL
28.2
18.2
28.2
15.6
26.7
34.5
32.4
Samples

Detected >RBSL
51.3
42.9
38.7
37.5
44.4
43.1
61.9
27.4
16.3 .
26.1
13.8
22.2
34.7
31.0
       The proportion of facilities with contamination above RBSLs also increases with
increasing facility size (Table F). However, in this case, there is no corresponding increase in
contamination above RBSLs at individual dispenser islands or in individual samples. Thus, the
increase at the facility level may be a consequence of more samples and/or more dispensers per
facility at larger sizes. In such a case, a constant proportion of independent samples above
RBSLs would naturally translate into larger numbers of dispenser islands and facilities with at
least one sample above RBSLs.  (For example, if the probability of any sample being above an
RBSL were 50 percent, a facility with two samples would have a 75 percent chance of having at
least one sample above an RBSL.)  The fact that the proportion of dispenser islands and facilities
with values above RBSLs is only slightly higher than the proportion of samples above RBSLs
illustrates that sampling results from the same islands and/or facilities are highly correlated.
                                                                                      26

-------
United States             Solid Waste And        EPA 510-R-04-004
Environmental Protection   Emergency Response    September 2004
Agency                  5401G                 www.epa.gov/oust/pubs

-------

-------