%EPA
            United States
            Environmental Protection
            Agency
            Solid Waste And
            Emergency Response/
            Research and Development
EPAS30AJST-90/OOS
March 1990
Standard Test Procedures
for  Evaluating Leak
Detection Methods:
Nonvolumetric Tank Tightness
Testing Methods
                 Sj?f$f'
                         / "'-5 7 Vi^^™-* ''Xl35^
                         " ,";,,,„„',;' 'y,,ff,," ,t ,,,,„, ^ ,„,,,,,,}.
                         """<"" "  ' "^t',^',:;;^' ' ' -'
                                      Printed on Recycled Paper

-------

-------
      Standard Test Procedures for
Evaluating Leak Detection Methods:
     Nonvolumetric Tank Tightness
                   Testing Methods
                           Final Report
         U.S. Environmental Protection Agency
          Office of Underground Storage Tanks
                           March 1990

-------

-------
                                 FOREWORD
How to Demonstrate That.Leak Detection Methods Meet EPA's  Performance
Standards

     The Environmental Protection Agency's (EPA's) regulations for
underground storage tanks require owners and operators to  check for leaks
on a routine basis using one of a number of detection methods (40 CFR
Part 280, Subpart D).  In order to ensure the effectiveness of these
methods, EPA set minimum performance standards for equipment used to
comply with the regulations.  For example, after December  22, 1990, all
tank tightness testing methods must be capable of detecting a.0.10 gallon
per hour leak rate with a probability of detection of at least 95% and a
probability of false alarm of no more than 5%,  It is up to tank owners
and bperators to select a method of leak detection that has been shown to
meet the relevant performance standards.                           ,

     Deciding whether a method meets the standards .has not been easy,
however.  Until recently, manufacturers of leak detection methods have
tested their equipment using a wide variety-of approaches, some more
rigorous than others.  Tank owners and operators have been generally
unable to. sort through the conflicting sales claims that are made,based
on the results of these evaluations.  To help protect consumers, some
state agencies have developed mechanisms for approving leak detection
methods.  These approval procedures vary from state to state, making it
difficult for manufacturers to conclusively prove the effectiveness of
their method nationwide.  The purpose of this policy is to describe the
ways that owners and operators can check.that the Teak detection equip-
ment or service they purchase meets the federal regulatory require-
ments.  States may have additional requirements for approving the use of
leak detection methods.

     EPA will not test, certify, or approve specific brands of commercial
leak, detection equipment.  The large number, of commercially available
leak detection methods makes it impossible for the Agency to test all the
equipment or to review all the performance claims.  Instead, the Agency
is describing how equipment should be tested to prove that it meets the
standards.  Conducting this testing is left up to equipment manufacturers
in conjunction with third-party testing organizations.  The manufacturer
will then provide a copy of the report showing that the method meets
EPA's performance standards.  This information should be provided to
customers or regulators as requested.  Tank owners and operators should
keep the evaluation results on file to satisfy EPA's record keeping
requirements.                              •'
                                   ii.i

-------
     •EPA recognizes"three distinct ways to prove that a particular brand
of leak detection equipment meets the federal performance standards:

     1.  Evaluate the method using EPAls standard test procedures for
         leak detection equipment;

     2.  Evaluate the method.using a national voluntary consensus code or
         standard developed by a nationally recognized association or
         independent third-party testing laboratory; or,

     3.  Evaluate the method using a procedure deemed equivalent to an
         EPA procedure by a nationally recognized association or
         independent third-party testing laboratory.

The manufacturer of the leak detection method should prove that the
method meets the regulatory performance standards using one of these
three approaches.   For regulatory enforcement purposes, each of the
approaches is equally satisfactory.  The following sections describe the
ways to prove performance in more detail.


EPA Standard Test Procedures

     EPA has developed a series of standard test procedures that cover
most' of the methods commonly used for underground storage tank leak
detection.  These include:

     1.  "Standard  Test Procedures for Evaluating Leak Detection
         Methods:   Volumetric Tank Tightness Testing Methods"

     2.  "Standard  Test Procedures for Evaluating Leak Detection
         Methods:   Nonvolumetric Tank Tightness.Testing Methods"

     3.  "Standard  Test Procedures for Evaluating Leak Detection
         Methods:   Automatic Tank Gauging' Systems"

     4.  "Standard  Test Procedures for Evaluating Leak Detection
         Methods:   Statistical  inventory Reconciliation Methods"

     5.- "Standard  Test Procedures for Evaluating Leak Detection
         Methods:   Vapor-Phase  Out-of-tank Product Detectors"

     6.  "Standard  Test Procedures for Evaluating Leak Detection
         Methods:   Liquid-Phase Out-of-tank Product Detectors".

     7.  "Standard  Test Procedures for Evaluating Leak Detection
         Methods:   Pipeline  Leak Detection Systems"

Each test  procedure provides an explanation of  how to conduct the test,
how  to perform  the  required  calculations, 'and how to report the
results.   The results from each standard test procedure provide the
                                    iv

-------
information needed by tank .owners and operators to determine if the
method meets the regulatory requirements.

     The EPA standard test procedures may be conducted directly by equip-
ment manufacturers or may be conducted by an independent third party
under contract to the manufacturer.  However, both state agencies and
tank owners typically prefer .that the evaluation be carried out by an
independent third-party in order to-prove compliance with the regula-
tions.  Independent third-parties may include consulting firms*'test
laboratories, -not-for-profit research organizations, or educational
institutions with no organizational conflict of interest.  In general,
EPA believes that evaluations are more likely to be fair and objective
the greater the independence of the evaluating organization.


National Consensus Code or Standard

     A second way for a manufacturer to prove the performance of leak
detection equipment is to evaluate the system following a national volun-
tary consensus code or standard developed by a nationally recognized   -
association (e.g., ASTM, ASME, ANSI, etc.')..  Throughout the technical
regulations for underground storage tanks, EPA has relied on:national
voluntary consensus codes to help tank owners decide which brands of
equipment are acceptable.  Although no such code presently exists for
evaluating leak detection equipment, one is under consideration by the
ASTM D-34 subcommittee.  The Agency will accept the results of evalua-
tions conducted following this or similar codes as soon as they have been
adopted.  Guidelines for developing these standards may be found in the
U.S. Department of Commerce "Procedures for the Development of Voluntary
Product Standards" (FR, Vol. 51, No. 118, June 20, 1986) and OMB Circular
No. A-119.


Alternative Test Procedures Deemed Equivalent to EPA1s

     In some cases, a specific leak detection method,may not be ade-
quately covered by EPA standard test procedures or a national voluntary
consensus code, or the manufacturer may have access to data that makes it
easier to evaluate the system another way.  Manufacturers who wish to
have their equipment.tested according to a different plan (or who have
already done so) must have that plan developed or reviewed by a
nationally recognized, association or independent third-party testing
laboratory (e.g., Factory Mutual, National Sanitation Foundation,
Underwriters Laboratory, etc.).  The results should include an accredita-
tion by the association or laboratory that the conditions under which the
test was, conducted were at least as rigorous as the EPA standard test
procedure.  In general this will require the following:          .

-------
•1.  The  evaluation  tests  the  system both under the no-leak  condition
    and  an  induced-leak condition  with  an induced leak rate as-close
    as" possible to  (or smaller than)  the performance standard.   In
    the  case of tank testing,  for  example,  this will mean testing
    under both 0.0  gallon per hour and  0.10 gallon per hour leak
    rates.   In the  case of ground-water monitoring, this  will mean
    testing with 0.0 and  0.125 inch of  free product.

 2.  The  evaluation  should test the system under at least  as many
    different environmental conditions  as the corresponding EPA test
    procedure.

 3.  The  conditions  under  which the.system is evaluated should be at
    least as rigorous as  the  conditions specified in the  corre-
    sponding EPA test procedure.   For example, in the case  of volu-
    metric  tank tightness testing, the  test should include  a
    temperature difference between the  delivered product  and that
    already present in the tank, as well as the deformation caused
   ' by filling the  tank prior-to testing.                          •

 4.  The  evaluation  results must contain the same information and
    should  be reported following the same general format  as the EPA
    standard results sheet.

 5.  The  evaluation  of the leak detection method must include
    physical testing of a full-sized version of the leak  detection
    equipment, and  a full disclosure must be made of the  experi-
    mental  conditions under which  (1) the evaluation was  performed,
    and  (2) the method was recommended  for use.  An evaluation  based
    solely  on theory or calculation is  not sufficient.
                                vi

-------
                             ACKNOWLEDGMENTS
     This document was written by Jairus D. Flora Jr., Ph.D., Karin M.
Bauer, and H. Kendall Wilcox, Ph.D., for the U.S. Environmental Protec-
tion Agency's Office of Underground Storage Tanks (EPA/OUST) under Con-
tract No. 68-01-7383.  The Work Assignment Manager for EPA/OUST was
Thomas Young and the EPA/OUST Project Officer was Vinay Kumar.  Technical
assistance and review were provided by the following people:

     Russ Brauksieck - New Yqrk Department of Environmental Conservation
     Tom Clark - Minnesota Pollution Control, Agency
     Allen Martinets - Texas Water Commission
     Bill Seiger-Mary land, Department of Environment                  ,
                            *            •' '          '            '
     American Petroleum Institute                     ,
     Leak Detection Technology Association    .
     Petroleum Equipment Institute
                                    vii

-------

-------
                                    CONTENTS
   Foreword	  i i i
   Acknowledgments....i	  vi i

        1.  Introduction.......	    1
                 1.1  Background	    1
                 1.2  Objectives.......	    2
                 1.3  Approach.	    2
                 1.4  Effects of high ground-water  level	    5
                 1.5  Organization of this document.....	    6
        2.  Scope and Applications...	    7
        3.  Summary.............	    9-
        4.  Safety.............	..11
        5.  Apparatus and Materials	   13
                 5.1  Tanks		.	..........   13
                 5.2  Test equipment	   14
                 5.3  Leak simulation equipment	   15
                 5.4  Product..................	   16
                 5.5  Tracers and carriers.,..	   16
                 5.6  Water sensor equipment..	   17
                 5.7  Miscellaneous equipment...........................   17
        6.  Testing Procedure....'	   19
                 6.1  Environmental data records......	   21
                 6.2  Induced leak rates and temperature  differentials..   21
                 6.3  Testing schedule..	   26
                 6.4  Testing problems  and solutions....................   34
                 6.5  Method evaluation protocol  for water detection....   35
        7.  Calculations........................	............'	   37
.                 7.1  Estimation of the method's  performance
                      parameters.	'..	   37
                 7.2  Water detection mode....	   40
                 7.3  Other reported calculations................	   45
                 7.4  Supplemental calculations and data  analyses
                      (optional)	...............r.....	   47
        8.  Interpretation	...51
                 8.1  Basic performance estimates....	*....   51
                 8.2  Limitations		-....,*...„	   52
•v .  •  '•            8.3  Water level detection function....	   52
                 8;4  Minimum water level change  measurement..	*....   53
                 8.5  Additional calculations................	   53
        9,  Reporting of Results......	   55

   Appendices

        A.  Definitions and notatiojial  conventions....	1	  A-l
        B.  Reporting forms	..,.T.................	  B-l
                                       IX

-------

-------
                                 SECTION 1

                                INTRODUCTION
 1.1  BACKGROUND
      The regulations on underground storage tanks (40 CFR Part 280, Sub-
 part D) specify performance standards .for leak detection methods that are
 internal to the tank.  For tank tightness testing, the tests must be
 capable of detecting a leak of 0.10 gallon per hour with a probability of
 (at least) 95%, while operating at a false alarm rate of 5% or less.

      A large number of test devices and methods are reaching the market,
 but little evidence is available to support their performance claims*
 Advertising literature for the methods can be.confusing.  Owners and
 operators need to be able to determine whether a vendor's tank tightness
 test method meets the EPA performance-'standards.  The implementing
 agencies (state and local regulators) need to be able to determine
 whether a tank facility is following the UST regulations, and vendors of
 tank tightness test methods need to know how to evaluate their systems.
                          /       -                  ....               *
     -Bresently, there are two categories of tank tightness testing
 methods on the market:  (a) volumetric testing methods, which measure
 directly the leak rate in gallons per hour, and (b) nonvolumetric testing
 methods, which report only the qualitative assessment of leaking or not
 leaking.*  These two testing methods require different testing and
 statistical analysis procedures to evaluate their performance.  The
 protocol in this document should be followed when the method is a
 nonvolumetric one.  The evaluation of the performance of volumetric tank
 tightness testing methods is treated in a separate protocol.  To simplify
 the terminology throughout this documentj nonvolumetric tank tightness
 testing methods are referred to as tank tightness testing methods.

      The use of tracers for leak detection purposes is one of the
 approaches permitted by the regulations.  While the approach has been
 classified by some as an external (out-of-tank) method, it has several
 characteristics that are common to nonvolumetric internal methods.  In
 particular, the type and amount of data collected and the statistical
 analysis of the data are. nearly identical to those used for other
 honvolumetric methods.  Also, the tracer is internal  to the tank,=r
 although the sensors are external to the tank.  This protocol includes
Conceivably,  a  "nonvolumetric  method"  could  utilize  some  measure  of
volume change, but in a qualitative manner.

-------
procedures for determining whether the performance of a method using
tracers meets the performance requirements for tank tightness testing
1.2  OBJECTIVES

     The objectives of this protocol are twofold.  First, it provides a
procedure to test tank tightness testing methods in a consistent and
rigorous manner.  Secondly, it allows the regulated community and regu-
lators to verify compliance with regulations.

     This protocol provides a standard method that can be used to
estimate the performance of a tank tightness test method.  Tank owners
and operators are required to demonstrate that the method of leak
detection they use meets the EPA performance standards of operating at
(no mor,e than) a 5% false alarm rate while having a probability of
detection of (at least) 95% to detect a leak of 0.10 gallon per hour.
This demonstration, must be made no later than December 22,, 1990.  The
test procedure described in this protocol is one example of how this
level of performance can be proven.  The test procedure presented here j.s
specific, based on reasonable choices for a number of factors.  Informa-
tion about other ways to prove performance is provided in the Foreword of
this document.

     This protocol does not address the issue of safety testing of equip-
ment or operating procedure.  The vendor is responsible for conducting
the testing necessary to ensure that the equipment is safe for use with
the type of product being tested.
1.3  APPROACH

     In general, the protocol calls for using the method on a tight tank
under no-leak conditions and under induced-leak conditions, producing
leak rates of 0.10 gallon per hour or less.  The nonvolumetric test
method being evaluated determines whether the tank is leaking or not
during each test.  This reported result is compared with the actual.con-
dition of the tank during testing to estimate the false alarm rate and
probability of detection.  Once these probabilities have been estimated,
the estimates are compared with the EPA performance standards to deter-
mine whether the method meets the EPA performance standards.

     The companion evaluation protocol for volumetric tank tightness
tests ("Standard Test Procedures for Evaluating Leak Detection Methods:
Volumetric Tank Tightness Testing Methods," March 1990) requires testing
under different conditions that simulate interferences likely to be
encountered in actual test conditions.  For volumetric methods these
include adding product at temperatures different from that of the product
in the tank and filling the tank prior to some of the .tests.,  Such tests
address temperature effects and tank deformation effects that can affect
measurements of level or volume change.  If the nonvolumetric method
being tested uses physical principles that might be affected by

-------
temperature or tank'deformation  effects,  then  the test series  should
account for these.   If the evaluatorrdetermines  that the  physical  princi-
ples of the test are not affected  by these  variables,  then  the tempera-
ture; and tank deformation parameters need not  be varied during the test
series.  Conversely, if the evaluator  determines that  other sources of
interference (e.g.,  background vapor concentrations, external  acoustical
noise) might affect  the performance of the  method,  then conditions to
test for these effects must be included in  the design.  For purposes of
illustration, this protocol assumes that  temperature and  tank  deformation
effects are important, unless the  evaluator determines  otherwise.

  __^ Some nonvblumetric test methods use  more  than  one  approach to
detecting a leak.  In this event,  each approach  must be tested and
evaluated to determine whether or  under what conditions the system meets
rthe EPA performance  standards.   For example, some nonvolumetric methods
rely on detection of water incursion during the  test to detect a leak in
the presence of a high.ground-water level.  If this  is  part of the
standard operating procedure, the  water detection sensor  needs to  be
evaluated as part of the evaluation procedure.   In addition to deter-
mining the performance of the water detection  sensor as a leak indicator,
the performance parameters (minimum detectable water level  and minimum
detectable level change) must be related  to the  size of the test tank to
determine whether the water detector could  sense  water  incursion at the
rate of 0.10 gallon  per hour under the test conditions  with a probability
of at least 95%, while operating at a  false alarm rate  of 5% or less.
That is, each mode of leak detection must be evaluated  and  compared to
the EPA performance  standards.             -

     It is emphasized that testing must include conditions  designed to
test the ability of  the method to  correctly detect a leak of the speci-
fied size (0.10 gallon per hour) in the presence  of  sources of interfer-
ence.  Sources of interference, such as product temperature changes,  that
do not affect the physical principles  of  operation of a method do not
need to be included  in the testing.  However,  the evaluating organization
must consider what alternative sources of interference might affect the
operation of the method and must include tests to determine whether the
method successfully overcomes these sources of interference.  The testing
conditions should be designed to cover the majority of cases; that is,
interference conditions as extreme as would be encountered  in approxi-
mately 75% of real world tests.  Testing need  not include extreme cases
that are rarely encountered.

     This document addresses two general types of nonvolumetric tank
tightness testing methods.  One type is internal  to the tank.  A probe
with sensors is placed in the tank and senses whether some physical
characteristic associated with a leak  is present.  The second type
introduces a tracer material into the tank.   The method then detects.
leaks by monitoring the exterior of the tank for the presence of the
tracer.  Since the only source of the tracer is from the tank,  the
presence or absence of tracer in the external  environment is taken to be
conclusive evidence that the tank  is either-leaking or tight.

-------
     .The technical requirements for the use of tracers are described in
the release detection section of the regulations on vapor monitoring (40
CFR 280.43[e]).  The major requirements which must be considered in
evaluating the tracer method are therefore:

     1.  The backfill where the sampling is conducted must be porous
         enough to readily allow diffusion of vapors to the sensor.

     2.  The tracer must be volatile enough to produce vapor levels which
         are detectable by the monitoring device.

     3.  Ground water, rain, or soil moisture must not interfere with the
         operation of the monitor.

     4.  Background contaminations must not interfere with the detection
         of releases from the tank.  :

     5.  The number and positioning of the monitoring wells must be
         optimized for the detection of leaks from any part of the
         system.                                                       .

Although these requirements are for continuous vapor monitoring devices,
they apply to the use of a tracer technique when it is used as a tank
tightness test.  Accordingly, the present protocol takes these factors
into account when evaluating tracer techniques.
                                r  '      •           ''.'.•       •-,-  i '•',•'
     Two types of tracer techniques have been developed:  those which add
tracer to the fuel and can perform a leak test with product in the tank;
and those which place a gas into an empty tank.  The former typically
uses halogenated hydrocarbons as the tracer material while the latter may
use sulfur hexafluoride or helium as the tracer material.  In both cases,
the tracer 1s placed in the tank and samples are collected outside the
tank.  Depending upon the specific method, or variation thereof, the time
to detect a leak may vary from a few minutes to several days.  Estimates
of"the leak rate can be obtained from methods which add tracer to the
product, for example, by using a spiked sample to produce a known
concentration which can be compared to the observed concentration of
tracer found at a leaking tank.  Methods which use gases in an empty tank
are usually limited to pass/fail conclusions since it is difficult to.
relate the loss of a gas through a hole to an equivalent amount of
product through the same hole.  The tracer techniques may also be used to
test the product lines or any other part of the system which is exposed
to the tracer.

     The application of a single protocol to the various tracer tech-
niques may present some practical problems.  The use of a tracer in an
actual test situation will contaminate the environment with the tracer,
rendering the site unsuitable for replicate testing, at least, for some
period of time.  For methods which rely on halogenated compounds, it may
be possible to use several different tracers at the same site.  For
methods which rely on a single tracer, the tracer must either be removed
from the site using techniques such as forced ventilation, another site  .

-------
must-be selected for the replicate testing tracer, or the replicate tests
must wait until the tracer has dissipated.  Since several replications
are required for satisfactory statistical analysis, the procedures can
prove to be cumbersome,    >                                  ^

     It is recognized that new nonvolumetric methods may be developed
after this document is published.  These new methods could be based on
different physical principles from those employed by currently available
methods.  The detailed test methods described in this document may not be
entirely appropriate for new methods in that they may not" address these
new approaches.  To allow for such contingencies, it will be the respon-
sibility of the evaluating organization to determine whether a new method
,can be evaluated with the current protocol or whether the new method has
aspects that require additional or different testing.  In the latter
case, it is the responsibility of the evaluating organization to devise
an appropriate test series and conduct the testing needed to evaluate the
method in a manner such that its performance can be compared to the EPA
performance standards..  See the Foreword for a description of alternative
approaches.


1.4  EFFECTS OF HIGH GROUND-WATER LEVEL

     The ground-water level is a potentially important variable in tank
testing.  Ground-water levels are above the bottom of the tank at approx-
imately 25% of the tank sites nationwide,, with higher proportions in
coastal regions.  Also, tidal effects may cause fluctuations in the
ground-water level during testing in some coastal regions.  If the
ground-water level is above the bottom of the tank, the water pressure on
the exterior of the tank will tend to counteract the product pressure
from the inside of the tank. If the tank has a leak (hole) below the
ground-water level, the leak rate in the presence of the high ground-
water level will be less than it would be with a lower ground-water
level.  In fact, if the ground-water level is high enough, water may
intrude into the tank through the hole*  ••

     The means by which the method deals with the ground-water level must
be documented.  A method that does not take the ground-water level into
account is not adequatei  If the ground-water level is determined to be
above t.he bottom of the tank, a method that tests in this situation must
include a means of compensating for the high ground-water level.  Accept-
able means of compensating are to either ensure that the tank has an out-
ward pressure.throughout or that the groundwater exerts an inward pres-
sure at all levels in the tank.  If an alternative approach to, compensate
ing for ground-water effects is used, the evaluating organization must
perform an engineering evaluation of the approach to ensure that it is
adequate.  If in doubt, the evaluating organization may require tests in
addition to those detailed in this document.

-------
1.5  -ORGANIZATION OF THIS DOCUMENT

     The next section presents the scope and applications of this
protocol.  Section 3 presents an overview of the approach, and Section 4
presents a brief discussion of safety issues.  The apparatus and mate-
rials.needed to conduct the evaluation are discussed iin Section 5.  The
step-by-step procedure, adapted for two existing types of nonvolumetric
test methods, is presented in Section 6.  Section 7 describes the data
analysis and Section 8 provides some interpretation of results.  Sec-
tion 9 describes how the results are to be reported.

  — Two appendices are included in this'document.  Definitions of some
technical terms are provided in Appendix A.  Appendix B presents a com-
pendium of forms:  a standard reporting form for the evaluation results,
a standard form for describing the operation of the method, data report-
ing forms, and an individual test Tog.  Appendix B thus forms the basis
for a standard evaluation report.

-------
                                 SECTION 2

                          SCOPE AND APPLICATIONS
                            /                       .'               _     .

     This document presents a  standard  protocol for evaluating nonvolu-
metric tank  tightness testing  methods*  The  protocol  is designed to
evaluate methods that test a tank  at  a  specific point in time.  The
methods determine a yes or no  answer  to the  question:  "Is the tank  leak-
ing?"  The nonvolumetric methods currently commercially available  use
some physical  result from a leaking tank  to  make this determination.
Some may use more than one characteristic of.a leaking tank in making
their determination.  This protocol is  designed to evaluate the method's
ability to detect a leak of 0.10 gallon per  hour with a probability  of at
least 95%, while operating at  a false alarm  rate of no more than 5%, as
specified in the performance standard in  the UST regulations.       .

     The protocol also provides tests to  determine the minimum water
level that the method can detect.  In addition, the protocol tests the
ability of the water sensor to measure  changes in the water level. These
are evaluated  over a range of  a few inches in the bottom of the tank.   •'.
The minimum  water level and minimum water level change that the method
can detect are converted to gallons using the geometry of the tank.  From
that, the minimum time it would take  the  sensor to detect a 0.ID-gallon
per hour leak  is calculated.   These tests are only performed if the
method uses  a  water sensor to  detect  leaks in situations such as a high
ground-water level.

     The document also presents a  protocol for evaluating tracer methods
at actual tank installations.  The protocol  does not  include laboratory
testing of components such as  vapor sensors.  It is designed to be used
for tracer methods that are applied to  a  tank at a specific point  in
time.          •              .                               '    '    j     '•

     .Subject to the limitations listed  on the Results  of U.S. EPA
Standard Evaluation form (Appendix B),  the results of  this evaluation can
be used to prove that a nonvolumetric tank tightness  testing method meets
the requirements of 40 CFR Part 280,  Subpart D.  The  Results of USEPA
Standard Evaluation form lists the limitations on the  method.  For
example, a minimum time for the test  may  be  required  in order for the
physical characteristic of. a leak  to  be sensed or for  the tracer to reach
the sampling ports.  The performance  results are valid provided the test
,is conducted for at least the  specified time.

-------

-------
                                SECTION  3

                      -           SUWARY


     The evaluation protocol for nonvblumetric test methods calls for
conducting the testing on a tight tank.   The organization performing the
evaluation should have evidence that the  tank used for testing is tight,
independent of the system currently being tested.  The evidence that the
tank is tight may consist of any of. the following:

     1.  At least three automatic tank gauging system (ATGS) records
         within a 3-month period with inventory and test modes indicating
         a tight tank.

     2.  A tank tightness test by another test method in the 6 months
         preceding testing that indicates a tight tank.
    „..'"-'                       .              .                 (

     3.  Continuous vapor or liquid monitoring system installed that
         indicates a tight tank.                          .

Any of;the above, verified by a tight test result on the initial test  .
(trial run) of the method under investigation, constitutes acceptable
evidence.  This information should.be reported on the data report form
(see Appendix B).

     The protocol calls for an initial test (trial run) under stable
conditions to ensure that the equipment is working and that there are no
problems with the tank9 associated piping, and the test equipment.  If
the tank fails the trial run test, however, then testing should not  ,
proceed until the problem is identified and corrected.  Only if the
evaluating organization has strong evidence that the tank is tight,
should testing proceed.

     The tank tightness testing equipment is installed at the tank site
to be tested following the method's standard operating procedure.  A
minimum of 21 independent tests of the tank-under the no-leak condition
are performed.  The results of these tight tank tests will be used to
estimate the false alarm rate, P(FA).  In addition, induced leaks at
rates not to exceed 0.10 gallon per hour  are simulated.  Again, a minimum
of 21 independent tests are performed with these induced leaks.  The.
results of these tests will be used to estimate,the probability of
detecting a leak of the magnitude used, P(D).  The simulation condition
(tight tank or induced leaks) is kept blind to the vendor.

-------
     -If  sources of  interference  are  to  be evaluated, test conditions
including these interferences  are  set up in a balanced  experimental
design.  The conditions that may interfere with the method  are  applied to
both tight and induced leak tests.   The order of the tests  is randomized
to ensure that the  conditions  are  kept  blind to the vendor.  The order of
both the interfering conditions  (if  used) and the leak  conditions are
randomized.  The proportion .of tests under the tight tank condition that
incorrectly indicate a leak is used.to  estimate the probability of a
false alarm, while  the proportion  of induced leak tests correctly iden-
tified is used to estimate the probability of detection.  Thus, each per-
formance parameter, P(FA) and  P(D),  is  estimated based  on at least
21 tests. .        .

     For tracer methods, the protocol calls for the use of the method on
a tank environment  which is representative of a typical UST installa-
tion.  It is not necessary for the tank to be in service to be  acceptable
for the  evaluation  process.  The type of backfill around the tank,
however, should be  known and should  be  either sand, pea gravel, crushed
rock, or other material which  is commonly used as backfill material.  If
the monitoring is conducted in areas other than the backfill, the charv.
acteristies of the  soil at the sampling location should also be known.

     The testing of a nonvolumetric method based on tracer technology
also involves a minimum of 42  tests.  At least 21 tests are done under
the tight tank condition and are used to estimate the probability of a
false alarm.  At least 21 tests are done with an induced or simulated
leak and are used to estimate  the probability of detection.  As before,
if interfering conditions are  to be  incorporated into the experimental
design,  these are established  for tests in a random order.  To estimate
P(FA), the tracer is introduced into the product in the tank.  After
mixing and after the appropriate waiting time determined by the method's
standard operating  procedure has elapsed, the sample ports are sampled to
determine if the tracer is detected.  False alarms could occur if tracer
is accidentally released during the process of adding it to the product
or mixing it with the product.  Consequently, the steps of adding the
tracer and mixing the product  in the tank should be repeated for each
tight tank test.      "   -

     For tracer methods, induced leaks are simulated by spiking the soil
with a sample of nonregulated material containing the tracer.  For
example, a vegetable'oil containing the tracer at the working concentra-
tion (e.g., 10 ppm) could be used to spike the soil  at 0.10 gallon per
hour.  This would be continued for the specified test duration and the
results recorded.  To keep the process blind to the vendor, randomized
samples of spiking  solution, some with and some without; tracer, could be
used and spiking done for each test.
                                    10

-------
                                 SECTION 4

                                   SAFETY


      This discussion does not purport to address all the safety consider-
 ations involved in evaluating leak detection equipment and methods for
 underground storage tanks.  The equipment used should be tested and
 determined to be safe for the products it is designed for.  Each leak
 detection method should have a safety protocol as part of its standard
 operating procedure.  This protocol should specify requirements for safe
 Installation and use of the device or method.  This safety protocol will
 be supplied by the vendor to the personnel Involved in the evaluation.
 In addition, each institution performing an evaluation of a leak detec-
 tion device should have an institutional safety policy and procedure that
.will be supplied to personnel on site and will be followed to ensure the
 safety of those performing the evaluation.

      Since the evaluations are performed on actual  underground storage
 tanks, the area around the tanks should be secured.  As a minimum, the
 following safety equipment should be available at the site:

.  • .   -s-   Two class ABC fire extinguishers
          One eyewash station (portable)
          One container (30 gallons) of spill  absorbent
      •   Two "No Smoking" signs

     .Personnel working at the underground storage tank facility should
 wear safety glasses when working with product and steel-toed shoes when
 handling heavy pipes or covers.   After the safety equipment has been
 placed at the site and before any work can begin, the area should  be
 secured with signs that read "Authorized Personnel  Only"  and "Keep Out."

      All  safety procedures appropriate for the product in the tanks
 should be followed.  In addition,  any safety  procedures required for a
 particular set of test equipment should be followed.

     .This test procedure only addresses the issue of  the  method's  ability
 to detect leaks.   It does not address testing the equipment for safety
 hazards.   The manufacturer needs to arrange for other testing for  con-
 struction standards to ensure that key safety hazards such  as fire^-
 shock, intrinsic  safety, product compatibility, etc.,  are. considered.
 The  evaluating organization should check to see what  safety testing has
 been done before  the equipment is  used for testing  to ensure that  the
 test operation will  be as safe as  possible.
                                    11

-------

-------
                                SECTION  5

                         APPARATUS AND MATERIALS
5.1  TANKS
     The evaluation protocol requires the use of an underground storage
tank known to be tight.  A second tank or a tank truck  is needed to store
product for the cycles of emptying and refilling,  if required.  As dis-
cussed before, the tank should have been tested and shown to be tight by
any of the three methods described in Section 3.   The tank should not
have any history of problems.  In addition, the protocol calls for an
initial trial run with the test equipment under stable  conditions.  This.
test should indicate that the tank is tight; if it does not, there may be
a problem with the tank and/or the test equipment  that  should be resolved
before proceeding with the evaluation*

     The tank facility used for testing is required to  have at least one
monitoring well.  The primary reason for this is to determine the ground-
water level.  The presence of a ground-water level above the bottom of
the tank would affect the leak rate in a real tank, that is, the flow of
product through an orifice.  The flow would be a function of the differ-
ential pressure between the inside and.outside of  the tank.  However, in
a tight tank with leaks induced to a controlled container separate from
the environment, the ground-water level  will not affect the evaluation
testing.  Consequently, it is not necessary to require that testing
against the evaluation protocol be done in a tank  entirely above the
ground-water level.  The monitoring well can also  be used for leak detec-
tion at the. site, either through liquid monitoring (if the ground-water
level is. within 20 feet of the surface)  or for vapor monitoring.

     Volumetric methods that measure volume or level changes of liquid
product that occur as a result of a leak generally have worse performance
as the size of the tank increases.  However, the tank size does not
affect the performance of existing nonvolumetric test methods to the same
extent, since they are based on different physical principles.   Con-
sequently, it is not necessary to restrict the application of these test
results to tanks with a volume equal to, or some arbitrary fraction
larger than, the test tank.  The evaluating organization should determine
the appropriate size limit based on their testing, physical principles
involved, and other available data, and  state the  limit on the  results
form (Appendix B).  For example, tanks larger than 50,000 gallons have a
different construction and geometry than the standard horizontal cylin-
drical tanks used for tanks up to this size.  It may be the tank geometry
and construction that impose limits rather than the size. '•


 ' '         •      '  •   '    •  •  '•'    13       ".     •.•"•••        '

-------
     •The test plan may require some testing with addition of product at a
different temperature from that of the fuel already in the tank.  This
requirement-is to verify that the method can accommodate the range of .
temperature conditions that routinely occur.  The procedure requires that
some tests begin by the tank being filled from about half full to the
test level with fuel that is 5°F warmer than the product in the tank, and
some tests using fuel 5°F cooler than the product in the tank..  This
procedure requires that some method of heating and cooling the fuel be
provided, such as pumping the fuel through a heat exchanger„ or by
placing heating and cooling coils in the supply tank or.tank truck before
the fuel is transferred to the test tank.  In the case of a tracer or.
acoustical method, the evaluating organization may eliminate the tempera-:
ture and filling conditions if they are not relevant.  The total number
of tests to be performed remains the same, however.  The temperature and
filling conditions would obviously be inoperative if a gaseous tracer
were to be used in an empty tank.

     If the protocol or the method requires that the tank be filled or
emptied a number of times, a second tank or a tank truck is needed to
hold reserve product.  A pump and associated hoses or pipes to transfer
the product from the test tank to the reserve product tank pr truck are
also needed.

     For tracer methods, the characteristics of a tank ,are less
important.  However, the test tank must be tight.  The primary purpose of
the tank is to provide an environment which, is representative of typical
tank installations.  The tank is important for testing for false
alarms.  The procedure of adding and mixing tracer to the product is a
potential source of false alarms from inadvertent release of the tracer
into the environment.
5.2  TEST EQUIPMENT

     The equipment for each tank test method will be supplied by the
vendor or manufacturer.  Consequently, it will vary by method.  In
general, the test equipment will consist of some method for monitoring
the tank for the effect used by the method to indicate a leak.  For
tracer methods, the equipment will also include some method for intro-
ducing the tracer(s) into the tank or the backfill.  The test equipment
also typically includes instrumentation for collecting and recording the
data and procedures for using the data to interpret the result as a pass
or fail for the tank.                 ,

     It is recommended that the test equipment for the method being
tested be operated by trained personnel who regularly use the equipment
in commercial tests.  This should ensure that the vendor's equipment is
correctly operated and will eliminate problems that newly trained or
untrained individuals might have with the equipment.  On the other hand,
if the equipment is normally operated by the station owner, then the
evaluating organization should provide personnel to operate the equipment
after the customary training.
                                    14

-------
5.3  LEAK SIMULATION EQUIPMENT

     The protocol calls for inducing leaks in the tank.  The method of
inducing the leaks must be compatible with the leak detection method
being evaluated.  The experimental design in Section 6 gives the nominal
leak rates that are to be used.  These leak rates refer to leak rates
that would occur under normal tank operating conditions.

     For volumetric methods, leak simulation can be accomplished by
removing product from the tank at a constant rate, measuring the amount
of product removed and the time of collection, and calculating the
resulting induced leak rate.  An explosion-proof motor can be used.to
drive a peristaltic pump head.  The sizes of the pump head and tubing are
chosen to provide the desired flow rates.  A variable speed pump head can
be used so that different flow rates can be achieved with the same
equipment.  The flow is directed through a rotameter so that the flow cari
be monitored and kept constant. One end of the tubing is inserted into
the product in the tank.  The other end is placed in a container.

     Although this leak simulation approach may work for some
nonvolumetric methods, most of these methods will require a method of
simulating leaks that is adapted to their specific principle of opera-   .
tion.  Examples of leak simulation methods for two nonvolumetric methods
follow.                              '                               *


5.3.1  Leak Simulation Approach for Acoustical Methods
                            t           '        -~'-'..
     T-wo methods commercially available at the present time are based on
acoustical signals generated when product flows'through an orifice or
when air is drawn through an orifice or hole in the tank that would allow
it to leak.  In order to simulate a leak condition for such a method, an*
orifice must be introduced into the tank so that product or air can flow
through it during the test.  A simulator of this type has been developed
and is in the patent process.  Its principle is described below.  The
size and location in the tank of the orifice must be determined so that
it would represent a leak rate of 0.10 gallon per hour or less if it were
present under norma-l operating conditions in the tank.  One approach is
to insert a pipe into the product in the tank through one of the openings
in the top of the tank.  The pipe has an orifice of the required size,
allowing product to leak from the tank into the pipe, where it can be
removed and measured.  Likewise, if a partial vacuum is applied, air
could be drawn into the tank through the orifice in the pipe.  The
orifice in the pipe can be calibrated by allowing product to flow into
the pipe and measuring the flow rate.
                                    15

-------
5.3.2  Leak Simulation Approach for Tracer Methods

     Two types of leak simulation equipment are required, depending upon
the type of tracer technique in use.  For methods which rely on detecting
the loss from the tank of product containing tracer, the simulation
equipment must be capable of delivering a liquid containing the tracer
into the backfill close to the tank.  The rate of delivery is used to
control the volume of product introduced in the backfill.  For methods
which rely on detecting the loss of gaseous tracer from the tank, the
simulation equipment must be capable of delivering the tracer gas into
the backfill in known quantities so that the ability of the system to
detect the tracer in the backfill can be evaluated.  In either case, the
amount of tracer introduced into the backfill should reflect the amount
that would be released if the tank were leaking at a rate of 0.10 gallon
per hour or less.  To do this, the rate of delivery is used to control
the amount of material introduced into the backfill.  To simulate a zero
leak rate, the tracer material is introduced into the test tank and mixed
with the product as appropriate.  However, a blank spike (without a
tracer) would be introduced into the backfill.

     Other nonvolumetric methods may use principles different from those
of the methods in these examples.  The evaluating organization will need
to develop .a method of leak simulation that is appropriate for a specific
test method.
5.4  PRODUCT

     The most common products in underground storage tanks are motor
fuels, particularly gasoline and diesel fuel.  Analysis of tank test data
based on tanks containing a variety of products has shown no evidence of
difference in test results by type of product, if the same size tank is
considered.  The only exception to this observation is that one tank test
method did produce better results when testing tanks with pure chemicals
(e.g., benzene, toluene, xylene) than when testing gasoline.' This dif-
ference was attributed to better test conditions, longer stabilization
times, and better cooperation from tank owners.                     .
                         ,                  V   ;                     .'
     Any commercial petroleum product of grade number 2 or lighter may be
used for testing, depending on the availability and restrictions of the
test tanks.  The choice of the product used is left to the evaluating
organization, but it must be compatible with the test equipment.


5.5  TRACERS AND CARRIERS

     When testing tracer methods, additional considerations apply.  While
use of petroleum products spiked with tracer would be ideal, the intro-
duction of regulated products into the ground is prohibited in almost all
situations.  Therefore, for test purposes, the carrier used for liquid
tracers should be of some nonregulated liquid such as mineral oil or
vegetable oil.  The concentration of tracer can be elevated in the


                                    16

-------
carrier to reduce the actual volume of material to be  introduced  into  the
ground.                    if         ;                            .  t   •  .

     Direct injection of the tracer gas diluted in air carl be used to
evaluate methods which rely on the loss of tracer gases from the  tank.
The concentrations of tracers injected during the simulation process
should approximate those contained in the tank during  an actual test.


5.6  WATER SENSOR EQUIPMENT

  _:- The equipment to test the water sensor consists of a vertical cylin-
der with an accurately known (to ±0.001 inch) inside diameter.  This
cylinder should be large enough to accommodate the water sensor.  Thus,
it should be approximately 4 Cinches in diameter and 8  or more inches
high.  The probe is mounted so that the water sensor is in the same rela-
tion to the bottom of the cylinder as it would be to the bottom of a
tank.  In addition, a means of repeatedly adding a small measured amount
of water to the cylinder is needed.  This can be accomplished by using a-
pipette.


5.7  MISCELLANEOUS EQUIPMENT

     As noted, the test procedure may require the partial emptying and
filling of the test tank.  One or more fuel pumps of fairly large
capacity will be required to accomplish the filling in a reasonably short
time.  Hoses or .pipes will also be needed for fuel transfer.  Some test
methods require some reserve product for calibration or establishing a
specified product level.  In addition, containers will be necessary to
hold this product as well as that collected from the induced leaks.  A
variety of tools need to be on hand for making the necessary connections
of equipment.                                               •'
                                    17

-------

-------
                                 SECTION  6
                                                                      /

                             TESTING PROCEDURE


      The overall performance of the method is estimated by comparing the
 method's results, leaking or tight tank, to whether a leak was actually
 induced.  Performance is measured over a.variety of realistic conditions
 including temperature changes and filling effects, if applicable.  The
 evaluating organization is responsible for adding any other variables
 that may affect a specific nonvolumetric method.  The range of conditions
 need not represent the most extreme cases that might be encountered,
.because extreme conditions can cause any method to give misleading   .
 results.  If the method performs well under various test conditions, then
 it may be expected to perform well in the field.

      The test procedures have been designed so that additional statisti-
 cal analyses can be done to determine whether the method's performance is
 affected by the size of the leak or other factors.  These additional
 analyses can only be done if the method makes a substantial number of
 mistakes so that the proportion of errors is between zero and one for
 some subsets of the data.  Thus, they are only relevant if the method
 does not meet the performance standard. ,

      For illustrative purposes, the basic test procedure introduces three
 main factors that may influence the test:  size of leak, temperature
 effects/and tank deformation.  The primary consideration-is the size of
 the leak.  The method is evaluated on its ability to detect leaks of
 specified sizes.  If a method cannot detect a leak rate of 0.10 gallon
 per hour or if the method identifies too many leaks when no leak is
 induced, then its performance is not adequate.

      A second consideration might be the temperature of the product added
 to fill a tank to the level needed for testing.   Three conditions could
 be used:  added product at the same temperature as the in-tank product,
'added product that is wanner than that already in the tank, and added
 product that is cooler.   The temperature difference should be at least
 5°F and should be measured and reported to  the nearest degree F.  For
 some methods, the temperature difference is needed to ensure that the
 method can adequately test under realistic  conditions.   The performance
 under the three temperature conditions can  be compared to  determine.
 whether these temperature conditions have an effect on the method's
 performance.  Note that some nonvolumetric  methods require an empty tank
                                    19

-------
or do not require a "specific product level.  If the principle of the
nonvolumetric method is not affected by product temperature as determined
by the evaluating organization, the test need not include this set of
conditions, although the total number of tests must not be decreased.

     Another consideration might be the tank deformation caused by pres-
sure changes that are associated with product level changes.  This
consideration is addressed by requiring several empty-fill cycles.  One
test is conducted at the minimum time after filling specified by the test
method.  A second test follows without any change in conditions (except,
possibly, leak rate).  Comparison of the order of the test pairs can  •
determine whether the additional time improves the method's perfor-
mance.  Again, if, as determined by the evaluating organization, the
operating procedure of the method is not affected by pressure changes,
this aspect of testing need not be included.

     Nonvolumetric test methods operate on a wide variety of princi-
ples.  Consequently, each method may have a different set of sources of
interference related to its operating principle. .The evaluating organi-
zation should consider possible sources of interference for the method...
being evaluated.  The list of these sources considered and the conclu-
sions reached should be reported.  The considerations do not need to
include the most extreme possible conditions, but should include condi-
tions expected to be encountered in a large majority (e.g., 75%) of the
normal tests cases.

     In addition to varying these factors, environmental data are
recorded to document the test conditions.  These data may help to explain
one or more anomalous test results.                                    .

     The ground-water level is a potentially important variable in tank
testing, and the system's means of dealing with it is to be documented.
A system that does  not determine the ground-water level and take it into
account'is not adequate.  Ground-water  levels are above the bottom of  the
tank at  approximately 25% of  underground  storage tank sites -nationwide,
with higher proportions  in coastal regions.        .

     If  the method  uses water incursion to account for high ground-water
levels,  this protocol evaluates two  aspects of the system's water sensing
function:  the minimum detectable water level and the minimum detectable
change  in  water  level.   Together, these can be used with the dimensions
of the tank to determine the  ability of the system's water  sensing device
to detect  inflows  of water  at various  rates.
                                     20

-------
 6.1 -ENVIRONMENTAL DATA RECORDS

      In general, the evaluation protocol requires that the conditions
 during the evaluation be recorded.  In addition to all the testing condi-
 tions, the foilowing-measurements should be reported (see the Individual
 Test Log forms in Appendix B):

          ambient temperature, monitored hourly throughout each test
          barometric pressure, monitored hourly throughout each test
          weather conditions such as wind speed; sunny, cloudy, or
          partially cloudy sky; rain; snow; etc.
          ground-water level if above bottom of tank
          any special conditions that might influence the results

     When testing tracer methods, the tank environment should also be
 documented as completely as possible.   A detailed site diagram should be
 prepared which identifies the positions of the tanks, piping, and other
 features which are present at the site.  The type of backfill and soil at
 the site should  be verified,  at the minimum, t.o be porous enough to allow
 migration of vapors from the  leak to the sensors.  The evaluation should
 not be run under backfill  conditions outside the range suggested by the
 vendor.

     Both normal,and "unacceptable" test conditions for each  method
 should bedescribed 1n  the, operating manual  for the method arid  should
 provide  a reference against which the  existing test conditions  can be
 compared.   The evaluation  should not be done,under conditions outside the
 vendor's recommended operating conditions.

     Pertaining,to the  tank and the product,  the following items  should
 be  recorded if applicable:

          type  of product  in tank
          type  of tracer(s)  (liquid  or  gas)
          tank  volume
          tank  dimensions and  type
          amount  of water in tank (before  and  after each  test)
          if  applicable, temperature of product  in  tank before filling
          if  applicable, temperature of product  added  each  time the  tank
          is  filled
          if  applicable, temperature of product  in  tank immediately  after
         filling
          if  applicable, temperature of product  in  tank at  start of test
6.2  INDUCED LEAK RATES AND TEMPERATURE DIFFERENTIALS

     Following a trial run in the tight tank, a minimum of 42 tests must
be performed according to an experimental design illustrated in
Table 1.  (As _discussed 1n Section 7, a larger number of tests could be
used.)  For Illustrative purposes, this table presumes that temperature
and tank deflection effects could interfere with the method.
                                    21

-------
  Table 1.  LEAK RATE AND TEMPERATURE DIFFERENTIAL
               TEST SCHEDULE (Example)

Test
No. | Set No.
Tjrjal fiirt \ - --- "' ' - - -
Nominal
Leak Rate
(gal/h)
-0
Nominal
Temperature
Differential *1
(degree, F)
' 0 - '
Empty/Fill cycle *2

1 1
2 1
LR2
LR1
T3
T3
Empty/Fill cycle

32
4 2
LR1
LR1
T2
T2
Empty/RII cycle .

5 3
6 3
LR1
LR3
T1
T1
Empty/Fill cycle

7 4
8 . 4
LR3
LR1
T3
T3 , .
Empty/RII cycle

9 5
10 5
LR4
LR1
T1
T1 "
Empty/RII cycle

11 6
12 6
LR2
LR3
T2
T2 .
Empty/Fill cycle

13 7
14 7
LR4
LR1
in .,
in '
Empty/RII cycle

15 8
16 8
LR3
LR1
T3
T3
Empty/RII cycle
•
17 9
18 9
LR4
LR1
T3
T3
Empty/Fill cycle
'
19 10
20 10
LR1
LR3
T2
T2
Empty/Fill cycle
-
21 11
22 11
LR3
LR1
T1
T1
Note 1:  The temperature differential is calculated as the temperature of
       the product added minus the temperature of the product in the tank.
                        i         ,.        '      .      *     - -
Note 2:  Empty/RII cycles and temperature differentials may not be required.
                                22

-------
  Table 1.  LEAK RATE AND TEMPERATURE DIFFERENTIAL
           TEST SCHEDULE (Example) (Continued)

Test
No. | Set No.
Empty/Fiji cycle *2

23 12
24 12
Nominal
Leak Rate
(gal/h)

LR1
LR2
Nominal
Temperature
Differential *1
(degree F)

T3
T3
Empty/Fill cycle

25 13
26 13
LR2
LR4
T2
T2
Empty/Fill cycle . .

27 14
28 14
LR3
LR1
T3
T3
Empty/Fill cycle

29 15
30 15
LR1
LR2
T1
T1
Empty/Fill cycle

31 16
32 16
LR1
LR1
T2
T2
Empty/Fill cycle

33 1.7
34 17
LR1
LR4
T3
T3
Empty/Fill cycle v

35 '18
36 18
LR1
LR4
T2
T2
Empty/Fill cycle .
* '
37 19
38 19
; LR2
LR1
T1,
T1
Empty/Fill cycle

39 20
40 20.
LR1
LR2
T2
T2
Empty/Fill cycle , ,

41 21
42 21
LR1
LR4
T1
T1
Note 1: The temperature differential is calculated as the temperature of
      the product added minus the temperature of the product in the tank.

Note 2: Empty/Fill cycles and temperature differentials may not be required.
                                23

-------
In Table 1, LR^ denote the nominal leak rates and T^ denote the
temperature differential conditions to be used in the testing.  These
42 tests evaluate the method under a variety of conditions.

     The 42 tests are arranged in 21 sets of two tests each.  Table 1
shows a possible ordering of the 21 sets.  In practice, the evaluating
organization should randomly rearrange the order of the sets so that the
leak rates are blind to the vendor.
Leak Rates

     Of the 42 tests, half will be performed under tight-tank conditions,
that is, at a leak rate of 0.0 gallon per hour.  The remaining 21 tests
will be performed under induced leak conditions with leak rates not
exceeding 0.10 gallon per hour.  Typically, all of these.induced leak
rates would be the same..  Alternatively, different non-zero leak rates
could be used and the results analyzed with a logistic model, as
described in Section 7.4.2.  The test schedule in Table 1 is an example  •
of 21 tests at a 0.0 gallon per hour leak rate (LRj) and 3 groups of
7 tests at non-zero leak rates of LR2, LR3, and LR^, which may all be
equal.                                                                 .

     The most direct evaluation of a nonvolumetric method uses only the
zero and 0.10 gallon per hour leak rates.  This, assuming that the test
results had at most one error at each leak rate, would provide the needed
performance evaluation.  However, a vendor may want to claim that his
method exceeds the EPA performance standards and establish that the prob-
ability of detecting a smaller leak (e.g., 0.01 rather than 0.10 gallon
per hour) is at least 95%.  In that case, two approaches are possible.
One is to use the smaller leak rate as the induced leak rate.  Again,
this is straightforward.  However, if the nominal leak rate selected is
close to or less than the leak rate that the method can actually detect
with 95% reliability, the testing may result in too many detection errors
at that reduced leak rate,  in order to demonstrate that the method meets
the performance standards, the 21 induced leak rate tests would have to
be run again using a nominal leak rate larger than the example of
0.01 gallon per hour (e.g., 0.05 gallon per hour), with additional costs
for the evaluation.

     .Another approach is to induce three non-zero leak rates and estimate
the probability of detection as a function of the leak rate.  In this
case, the method would demonstrate that it meets the EPA performance
standards, provided that the probability of detection at a zero leak rate
(a false alarm) is less than 5%, and the detectable Teak rate that could
be claimed by the method is the leak rate at which the function first
exceeds 95%.  If this option is chosen, a single test series of 42 tests
could demonstrate that the method meets the EPA performance standards at
the smaller leak rate determined by the evaluation.  In order for this
approach to work, the probability of detecting a leak must increase
steadily with the leak rate.  In addition, the non-zero leak rates must
be selected so that the observed results (proportions of tests where a


                                    24

-------
 leak-is detected)  also increase with the induced leak rate.  There must
 be very few detections (zero or one) at zero, some missed detections at
 the smaller leak rates,  and very few at,the larger leak rates.


 Temperature Differentials (1f applicable)

      If temperature differential is important for the test method, three
 nominal temperature differentials between the temperature of the product
 t,o .be  added and  the temperature of the product in the tank during each
 fill cycle should  be used.   These three temperature differentials are
 -5°, 0°, and +5°F  (-2.8°, 0°, and +2.8°C).   The temperature differential
 of 5°F is a minimum.  Larger differences may be used.  If temperature
 differences are  used, the actual differences are to be calculated and
 reported.


 Randomization

     A total of  42 tests consisting of combinations of the four leak
 rates  (U^ = 0.0 gallon  per hour, LR2, LR3,  and LRJ  and  the three
 temperature differentials (Ti,  T2,  and T3) will  be performed.   LR2,  LR3,
 and LR^ may all  be the same, depending on the analysis method  to be
 used.   The 42 tests have been arranged 1n pairs (sets), each pair
 consisting of two  tests  performed at the, same temperature differential.
 However, the leak  rates  within  a pair have been randomly  assigned to the
 first  or second  position in the testing order.   The test  schedule is
 outlined in Table  1.

     A randomization of  the test schedule is required to  ensure that the
 testing is done  blind to the vendor.   The randomization of the tests is
 achieved by the  evaluating  organization by randomly assigning  three
 nominal  leak rates below 0.10 gallon per  hour to LR2,  LR3,  and LRH and by
 randomly assigning the nominal  temperature differentials  of 0°,  -5°, and
-+5°F to Tj,  T2,  and T3,  following the sequence  of 42  tests as  shown  in
 Table  1.  In addition, the  evaluating organization should randomly assign
 the set numbers  (1 through  21)  to the 21  pairs  of tests.   The  results of
 the randomized sequence  should  be kept blind to  the vendor.  That is, the
 vendor should not  know which .induced  leak rate  is used or which  tempera-
 ture condition is  present in advance.   The vendor should  test  for the
 induced leak rate  based  on  his  instrumentation  and standard  operating
 procedure without  knowledge of  the  induced conditions.  /Randomization
 should be done separately for each  method evaluated.

     In summary, each test  set  consists of two tests  performed using two
 induced  leak rates and one  induced  temperature differential  (temperature
 of product to be added - temperature  of product  in the tank).   Each  .set
 indicates the sequence in which the induced  rates are  used  to  remove the
 product volumes  (in gallons per hour)  from the .tank at a  given product
 temperature  differential.   In some  cases, e.g.,  when  a partial  vacuum is
 applied  to the tank,  the simulated  leak will  not actually remove product
 from the tank.   In this  case, the indicated  rates are  those  at which     :


 -."'  •     •    •      '        • ••" •   25    ':•   .    • :'  '" -   '

-------
product would escape" or be removed from the tank  if the:  induced condition
were present under normal tank operating conditions.


Notational Conventions

     The nominal leak rates to be induced are denoted by LRX = 0.0 gallon
per hour, LR2, LR3, and LR,f.  It is clear that the nominal leak rates
selected by the evaluating organization cannot be achieved exactly in the
field. .Rather, these numbers are targets that should be established by a
calibration process.  The maximum must be no more than 1035 greater than
the nominal 0.10 gallon per hour.

     The leak rates actually induced for each of  the 42  tests will be
calibrated for each test series.  They will be denoted by Slt S?,..., ,
St.,.  The results of each test will be denoted by Llt..'.-,L%2» with each
LJ being either "tight" or "leaking."  The L1 may be coded numerically,
e.g., LJ - "0" for tight and "1" for leaking, for convenience.

     The subscripts 1,...,42 correspond to the order in  which the tests
were performed  (see Table 1).  That is, for example, Ss  and L5 correspond
to the test results from the fifth test in the test sequence.


6.3  TESTING SCHEDULE

     The first test to  be done is a trial run.  This test should be done
with a tight tank 1n a  stable condition and this  should  be known to the
vendor.  The results of the trial run will be reported along with the
other data, but.are not explicitly used in the'calculations estimating
the method's.performance.

     There are two purposes to this trial run.  One is to allow the
vendor to check out the tank testing equipment before starting the eval-
uation.  As part of this check,  any faulty equipment should be identified
and repaired.  A second part is  to ensure that there are no problems with
the tank or the test equipment.  Such practical field problems as loose
risers,  leaky valves,  leaks  in plumber's plugs, etc., should be identi-
fied and corrected with this trial run.  The results also provide addi-
tional verification that the tank  is tight and so provide a baseline for
the  induced  leak rates  to  be run in the  later part  of the evaluation.

     The testing will be performed using  a randomized arrangement of
'nominal  leak rates  and  temperature differentials  as illustrated in
Table  1  above,  unless  the.evaluating organization determines that the
filling  and/or  temperature changes are  irrelevant for the particular
nonvolumetric method.   The time  lapse  between'the two tests  in each  set
should  be  kept  as  short as practical.   It  should  not exceed  30 min,  and
preferably should  be  held  to 15  min  or less.  Twenty-one sets  of  two
                                     26

-------
tests each will be carried out.  After each set of two tests, the test
procedure starts anew with emptying the tank to half full, refilling,
stabilizing, etc., as necessary.  The details of the testing schedule are
presented next, in accordance with the example ordering shown in Table 1.

Step 1:   Randomly assign nominal leak rates not to exceed O-.lO gallon
          per hour to LR2, LR3, and LR4.  Note that LRj is identified
          with the zero leak or tight tank condition as 21 trials are run
          in this condition.  Also, randomly assign the temperature
          differentials of 0°, -5°, and +5°F to T,, T2,.and T3.  This
          will be done by the organization performing the evaluation and
          needs to be kept blind to the crew performing the testing.

Step 2:   Follow the vendor's instructions to install the leak simulation
          equipment in the tank if this has not already been done, making
          sure that the leak simulation equipment will not interfere with,
          the test equipment.                        .

Step 3:   Trial run.  Following the test method's standard operating
          procedure, fill the tank to the recommended level, and allow_
          for the stabilization period called for by the method or
          longer.  Any product added should be at the same temperature as
          that of the in-tank product.  Conduct a test on the tight tank
          to check out the system (tank* plumbing, etc.) and/or the
          method.  Perform any necessary repairs or modifications
          identified by the trial run.             .

Step 4:   Empty the tank to half full.  Fill with product at the recom-
          mended temperature. • The temperature differential will be T3
          (Table 1, Test No. 1).  Record the date and time at the comple-
          tion of the fill.  Allow for the recommended stabilization
          period, but not longer.  Induce the appropriate .leak condition.

Step 5:   Continue with the method's standard operating procedure and
          conduct a test on the tank, using the method's recommended test
          duration.  Record the date and time of starting the test.  This
          test will be performed under the first nominal leak rate of the
          first set in Table 1.  This nominal leak rate to be .induced is
          LK2*                                           ,

     When the first test is complete, determine and record the calibrated
induced leak rate, Slt and the method's reported leak condition, Llf  If
possible, also record the data used to determine the leak condition and
the method of calculation.  Save all data sheets, computer printouts, and
calculations.  Record the dates and times at which the test began and
ended.  Also record the length of the stabilization period.  The Individ-
ual Test Log form in Appendix B is provided for the purpose of reporting
these data and the environmental conditions for each test.

    , Record the temperature of the product in the test tank and that of
the product added to fill the test tank (if done; if not, document why
not on the log). 'After the product has been added to fill'the test tank,


       '.         ,''•'••'   27    '           ./•"-.'•

-------
 record the average temperature in the test tank.   Measuring the  tempera-
 ture of the product in the tank is not a trivial  task.   One suggested  way
 to measure the temperature of the product in the  tank is to use  a probe
 with five temperature sensors spaced to cover the diameter of the tank.
 The probe is inserted in the tank (or installed permanently), and the
 temperature readings of those sensors in the liquid are used to  obtain an
 average temperature of the product.  The temperature sensors can be
 spaced to represent equal  volumes or the temperatures can be weighted
 with the volume each represents to obtain an average temperature for the
 tank.

 Step 6:   Change the nominal leak rate to the second in the first set,
           that is LRX (see Table 1).  Repeat Step 5.  Note that  there
           will be an additional period (the time  taken by the first test
           and the set-up time for the second test) during which  the tank
           may have stabilized.  When the second test of the first set  is
           complete,.again record all results (times and dates, induced
          .leak rate and test result, temperatures, calculations, etc.).

 Step 7:   Repeat Step 4.  The temperature differential  will be changed to
          • T2-     '•          '.-.•••                .    •   ;  •        '

 Step 8:   Change the nominal leak rate to the first in the second set.
           In this example, the rate is unchanged  at LRj.  Repeat
           Step 5.  Record all results.

 Step 9:   Change the nominal leak rate to the second in the second set if
           it is different.  In this example the second leak rate is
           LRx.  Repeat Step 6.  Record all results.

 Step 10:  Repeat Step 4.  The temperature differential  will be changed to
           the following one in Table 1.  In this  case,  it will be changed
           to Tx.

•Step 11:  Repeat Steps 5 through 9, using each of the two nominal leak
           rates of the third set, in the order given in Table 1.

      Steps 4 through 9, which correspond to two empty/fill cycles and  two
 sets of two tests, will be repeated .until all 42  tests are performed.

      Normal and "unacceptable" test conditions for each method should  be
 described in the owner operating manual for each  method and should pro-
 vide a reference against which the existing test  conditions are  com-
 pared.  The evaluation should not be done under conditions outside the
 vendor's recommended operating conditions.
                                     28

-------
 6i3.1  Application of the Protocol to Acoustical Methods

      One  class of commercially available nonvolumetric test methods is
 based on  acoustical  principles.   This section describes the application
 of  the protocol to this type of method.   A basic description of the
 method is needed to understand the application of the protocol.

      Acoustical methods use sensitive hydrophones to detect an  acoustical
 signal  from  the tank.   This signal  is recorded and is analyzed  to  iden-
 tify  a specific characteristic associated with a leak.   One such method
 places the tank under a partial  vacuum and investigates the acoustical
 signal  for a characteristic "bubble"  signature .induced  when air bubbles
 are drawn from outside the tank  (in an unobstructed backfill  zone)  into a
 liquid through a hole in the tank.  Leaks in  the ullage are identified  by
 a particular frequency or "whistle" of air ingressing into  the  ullage
 space.  Another approach analyzes the acoustical  signal  for a character-
 istic sound  of fluid flowing out of an orifice in the tank.

      While these methods have been  called "acoustical,"  they  typically
 have  additional modes  of detecting  leaks  that are used  in conditions of a
 high  ground-water level.   Generally they  rely on identification  of water
 ingress to detect leaks in the presence of a  high ground-water  level.
 The evaluation must  test all  modes  of leak detection  used by  the method
 to  "detect leaks from  any portion of  the  tank that normally contains
 product."  Section 6.5 contains  a protocol  to evaluate  a water  sensor
 used  to detect inflow  of water during a test  period.

      Acoustical  methods can be used with  a fairly wide range  of product
 levelsHn  the  tank.  The deformation  caused by filling the tank would not
 affect  these methods,  nor would  the temperature-'of the product  in the
 tank.   Consequently, the  sequence of  temperature  and filling  conditions
 does  not need  to be  considered with these  tests.   The tank should be
 filled  to  a  level  in the  range specified  by the method.

      To induce a leak  for the  acoustical methods,  it is necessary to use
 a device that  will create  the  same  signal  that a  real leak would cre-
 ate.  One way  to do  this  is to use  an orifice-type leak simulator,   this
consists of a  pipe inserted  into  the tank  through one of the tank open-
 ings.  The pipe  is sealed  to the  tank.  The bottom of the pipe is fitted
with  a  cap that  contains  a calibrated orifice to  allow product to leak
 into  the pipe  at  the desired  leak rate under a standard head.  This
 simulator will work  for either type of acoustical signal.  Flow of  liquid
through the orifice would produce the signal typical of liquid'flow.  If
the tank ,1s under partial  vacuum, air will be'drawn into the tank through
the orifice below the  liquid level  and will produce bubbles.  % means of
closing the orifice  is needed so that a zero leak rate can be induced and
kept  blind to the vendor.                                        —

     Since neither temperature differential nor tank deformation should
affect the acoustical methods, the approach discussed earlier in this
                                    29

-------
subsection is simplified as follows.  The steps refer to Table 1, with
the understanding that there are no differences among Tlt T2, T3, and the
partial emptying and refilling is not necessary.

Step 1:  Decide whether one or three nonzero leak rates will be used.
         (The use of three may allow one to fit a model relating prob-
         ability of detection to leak rate, but if this is not important
         to the vendor, it is sufficient to use a single non-zero leak
         rate (less than or equal to 0.10 gallon per hour), which may be
         the preferred approach.)

Step 2:  Decide what leak rates will be used.  If only a single non-zero
         leak rate is used, it can be selected between zero and 0.10 gal-
         lon per hour.  If the vendor wants to establish a smaller
         detectable leak rate, a value of less than 0.10 gallon per hour
         may be used.  (The risk of doing this is that if the system does
         not pass, more.testing with larger leak rates below 0.10 gallon
         per hour may be needed.)

Step 3:  If only two leak rates (0 and one other) are used, randomly
         assign one of them to LRt and the other to all cases where LR2,
         LR3, or LRn are listed.  If four leak rates are to be used,
         assign LRX to zero and randomly assign the other three to LR2,
         LR3, and LR^.

Step 4:  Randomly rearrange the order of the 21 pairs of tests listed in
         Table 1.  (This allows for additional randomization and provides
         better control on keeping the induced leak rates blind to the
         vendor.)

Step 5:  Have the'vendor install the test equipment in the tank.

Step 6:  Trial run.  Following the test method's standaird operating
         procedure, fill the tank to the recommended level.  Have the
         vendor conduct a test with a known zero leak rate and verify
         that the equipment has been installed and is functioning cor-
         rectly.  This also provides confirmation that the tank is still
         tight and is compatible with the test method.

Step 7:  Induce the leak rate called for in the randomization developed
     .    above.  Have the vendor test the tank with this induced leak
         rate and report the results.  Record the calibrated induced leak
         rate and the vendor's results (tight or leaking).  Record the
         environmental conditions data and other ancillary data on the
         test logs (see Appendix B).
                                    30

-------
 Step 8:   When the first test is completed, change the leak rate to estab-
     ,    lish the second leak rate called for in the randomized series
          (Table 1).   When this induced rate has been established, have
          the vendor  test the tank.  Record the environmental conditions
          data.  When the vendor has completed the test, record his
          reported result and the induced leak rate.

 Step 9:   Repeat step 8 until all 42 tests have been completed.

      As  will  be described in Section 7,  the system can produce no more
 than one false alarm and still  pass.  Thus, if a second, false alarm
 occurs in the test series,  the system will not pass,  and testing could be
 terminated.   Similarly, if  only one non-zero,leak rate is used,  and if a
 second mistake is made with that non-zero leak rate,  the system  will  not
 pass.  At the point  where the evaluating organization determines that the
 system will  not pass,  it might  be desirable to conclude testing.   The
 series could  be completed to provide added information to the vendor.   If
 a  leak rate  of less  than 0.10 gallon per hour was used,  starting the  test
 series again  with a  leak rate closer to  0.10 gallon per hour might be
 done since the method  might pass at that rate  but not at the smaller  leak
 rate.  If no  errors  have occurred when 20 tight tank  or 20 induced leak
 tests have been done,  the system will  pass.  Since only one more  test  is
 needed,  it probably  would not effect much savings to  stop at this  point.


 6.3.2  Application of  the Protocol  to Tracer Methods

     There are many  variables present  in external  monitoring that  are
 difficult to  predict or control.   These  include  the nature  of the  back-
 fill material,  moisture content  of the soil, size of  the  excavation, type
 of soil  surrounding  the excavation,  the  ground-water  level,  position of a
 leak relative to  the sampling locations,  and whether  the method is aspi-,
 rated or passive.  In  general, some minimum threshold concentration of
 tracer must be  reached before a  signal is  generated.  The  lower the
 threshold, the  more  sensitive the method,  but the  more susceptible it
 will be  to false  alarms.

     For test methods  that  involve  the loss of product from the tank, the
 induced  leak .rates should be  designed to  introduce the amount of tracer
 materialinto the  soil  that would be released by  leak rates of the speci-
 fied size over  the test period.  Methods that add  liquid tracer to the
 product  specify a  concentration of the tracer in the product.  Using this
 concentration  (e.g., 10 ppm), a  leak rate  (e.g., 0.10 gallon per hour)
 and a test and waiting time after introducing the tracer into the tank
 (e.g., 24 hours), one  can calculate the amount of tracer that would be
 released.  This is the amount that should be released during the leak
 simulation.  A suggested way to accomplish this is to make up sampTes of
 a carrier that can be  introduced into the environment, say vegetable oil/
with tracer added  in the appropriate concentrations.  These samples can
be used to spike the ground at small rates, giving the same amount of
tracer that would be released by the specified leak rates.
                                    31

-------
     .If the method uses gas tracers, they can be introduced into the
ground to simulate leaks by using a flowmeter to allow the gas to flow at
the rate that would occur under the testing conditions,, e.g., in a tank
at 2 PSI and through a small orifice, representing a hole that would leak
liquid product at the designated leak rates (less than 0.10 gallon per
hour).

     Note that once a tracer, gas or liquid, has been introduced into the
soil in a test, the tracer must be eliminated before the next test.
Forced air may be used to disperse the tracer to levels that will not be
detected and interfere with the method; the next test may be conducted
with a different tracer; or a different site may be used.

     The following steps assume that multiple tracers are available, one
of which is used in the tank to investigate the false alarm possibili-
ties, and others that are used in leak simulations.

     Neither the temperature conditioning nor tank stabilization is an
issue with tracer methods.  Consequently, it is not necessary to change
fuel temperatures and fill and empty the tank frequently as part of the .
evaluation.  At least 21 tests of the tank in the no-leak condition are
required, as are at least 21 tests using the induced leaks.

Step 1:   Decide whether a .single non-zero leak or three non-zero leak
          rates will be used and select these leak rates.

Step 2:   Identify the zero leak rate with LRl in Table 1.  Randomly
          assign the other leak rate(s) to LR2, LR3, and LR%..

Step 3:   Randomly rearrange the order of the 21 pairs of tests  in
          Table 1'that result from the assignment of the leak rates.

Step 4:   Determine the rate of introducing tracer (if a gas) or liquid
          carrier and tracer  (if a liquid) into the backfill to  simulate
          the  selected leak rates.   If a liquid tracer is used,  prepare
          samples with the carrier and tracer in the needed concentra-
          tions, label these"with the randomized test sequence,  and
          provide them to the test crew.  The crew should not know
          whether or  in what concentration the tracer is in the  leak
         . simulation  samples*

Step 5:   Prepare the tank.   If a  liquid tracer is used, have the vendor
          introduce  it at the desired concentration into the test tank
          and  fill the tank to the desired level following normal oper-
          ating procedures  for the method.   If a gas tracer  is used,
          empty the  tank  and  have  the vendor  introduce the gas to the
          tank.  The tank thus prepared will  serve to provide the data on
          the  zero  leak rates.
                                     32

-------
Step.6:   Have the vendor  locate  the  sampling  ports.   Also  locate  a spik-
          ing port for  leak  simulation  as far  from  the sampling  ports  and
          as close to the  tank  as possible.  Be  careful  not to damage  the
          tank in installing the  ports  in the  backfill.

Step 7:   Conduct the trial  run.  For tracer methods,  the trial  run will
          be of a different  nature than for other methods.   The  trial  run
          for a. tracer  usually  consists of verifying that the site  condi-
          tions allow the  use of  a tracer method.   A compound is intro-
          duced at the  spiking  port.  The test locations are sampled to
          determine whether  the compound is detected at the sampling •
  —      locations.  The  trial run accomplishes two purposes.   First, it
          verifies that the  soil  or backfill conditions are such that the
          tracer can migrate from the tank to  the sensors.   Second,  it
          determines the time needed  for the migration and  so establishes
          a test time.

Step 8:   Have the vendor  conduct a test of the tank (zero  leak  rate).

Step9:   Begin testing using the first non-zero leak rate.  Have the
          vendor conduct a test.  Note:  If two different tracers are
          used, it may  be  possible for  the vendor to conduct the test on
          the tank (zero leak rate) and the induced leak test at the same
          time.

Step 10:  When the test in step 8 and/or 9 is completed, record the .
          induced leak  rate, the  vendor's determination (tight or leak-
          ing), and the environmental conditions data on the test log
          (see Appendix B).   '                    ,

Step 11:  Ensure that the  test site can be used for a second leak test
          (by removing  the current tracer or using a different one).
          Start the next induced  leak rate as in steps 8 and 9 and have
          the vendor conduct another test..  Record all results.

Step 12:  Repeat step 11 until the test series is completed.

     It should be possible.for the vendor to conduct tests on the tank
containing the tracer repeatedly  for the zero leak rate tests.   In con-
ducting.the repeated tests on the tight tank to estimate the false alarm
rate, the steps of adding  tracer to the product and mixing the tracer in
the product should be repeated.   The process of adding and mixing tracer
is a likely cause of false alarms as it could lead to inadvertent release
of tracer into the environment that could be mistaken for a leak.  It
should be possible to simulate the addition and mixing of the tracer,by
using tracer-containing product and handling it in the same manner as the
tracer solution.                                                     ;

     Assuming that at least two tracers are available, the tight tank
tests and the simulated leak tests can be run simultaneously.  For each
test, the carrier sample is introduced in the spiking port.   The con-
tainers, of carrier,'are made up in advance and coded.  Half of them  .

-------
contain tracer and half do not.  Each test would consist of introducing
one tracer (say type A) into the tank and another sample (either a blank
'or containing tracer type B) into the spiking port.  The testing company
samples the soil gas and reports on the presence of any detected
tracer.  A finding of tracer A would be a false alarm.  A finding of
tracer B (when it was spiked) would be a correct detection.  If
additional distinct tracer compounds are used, this process could
continue spiking tracer C, etc.  A finding of both tracer B (from a
previous spike) and tracer C from the current spike would be a correct
detection.    .     .

     As will be described in Section 7, the system can record only one
false alarm and still pass.  Thus, if a second false alarm occurs in the
test series, the system will not pass, and the evaluating organization
may recommend to the vendor that testing might be terminated.  Similarly,
if only one non-zero leak rate is used, and if a second mistake is made
with that non-zero, leak .rate, the system will not pass.  At the point
where the evaluating organization determines that the system will not
pass, it-might be desirable to conclude testing.  If a leak rate of less
than 0.10 gallon per hour was used, starting the test series again with a
leak rate closer to 0.10 gallon per hour might be done since the method
might pass at that rate but not at the smaller leak rate.   .


6.4  TESTING PROBLEMS AND SOLUTIONS

     Inevitably, some test runs will be inconclusive due to broken equip-
ment, spilling of product used to measure the induced leak rate, or other
events that have interrupted the testing procedure.  It is assumed that,
in practice, the field personnel would be able to judge whether a test
result is valid.  Should a run be judged invalid during testing, then the
following rules should apply.

Rule No. 1   The total number of tests must be at least 42.  That is, if
             a test is invalid, it needs to be rerun.  Report the test
             results as invalid together with the reason and repeat the
             test.

Rule No. 2    If equipment fails during the first run (first test of a set
             of two) and if the time needed for fixing the problem(s) is
              less than 4 hours, then repeat that run.  Otherwise, repeat
             the empty/fill cycle, the stabilization period, etc.  Record
             .all time periods.               •

             Note:  The average stabilization time or average time after
             introducing the tracer will be reported on the Results of
             U.S. EPA Standard Evaluation form in Appendix B.  If the
             delay would increase this time noticeably,, then the test
             sequence should be redone.

Rule No. 3    If equipment fails during the second run (after the first
             run in a set has been completed successfully), and  if the


                                    34

-------
                   •        time needed for fixing the problem(s) is less than 4 hours,
                           then repeat the second run.  Otherwise, repeat the whole
                           sequence of empty/fill cycle, stabilization, and test at the
                           given conditions.

                   Rule numbers 2 and 3 are only applicable if the testing schedule
             . requires temperature conditioning and tank deformation effects.  Other-
              wise, the time between tests is not an important limitation.

                 .  Note that an acceptable alternative to conducting the tests in pairs
              is to set up the tank conditions (as required) for each test.  Thus,
              while the protocol allows for the tests to be run in pairs for economy,
              they may all be run individually.


              6.5  METHOD EVALUATION PROTOCOL FOR WATER DETECTION

                   Some, methods rely on detection of water incursion to identify leaks
             .in the presence of a high ground-water level.  These often use a water
              sensor installed at the bottom of the tank.  A standpipe device to test
              the function of the water sensor consists of a cylinder with an accu-
              rately known (to ±0.001 inch) inside diameter attached to the bottom of
              a pipe of 4- to 6-inch diameter pipe.  The probe is mounted so that the
              sensor is in the same relation to the bottom of the cylinder as to the
              bottom of a tank when installed in the field.  Enough product is put into
f^  '•         the cylinder and pipe so that the product level  sensor is high enough so
              as not to interfere with the water sensor.  A measured amount of water is
              then added to the cylinder until the water sensor detects it, at which
              time the water level is calculated and recorded.  Additional  measured
              amounts of water are added to produce calculated level  changes.   The
              amount of water added, the calculated level change, and the level  change
              measured by the method are recorded.  This is done over the range of the
              water sensor or 4 inches, whichever is less.   When testing is complete,
              the product and water are removed, separated, and the process is
              repeated.  The testing procedure is given in  detail next.

              Step 1:   Install the probe temporarily in a  test standpipe*,   The bottom
                        section of about 1 foot should have an accurately known (to
                        ±0.001 inch) inside diameter.  The  diameter must be large
                       - enough to accommodate the probe and must be known accurately so
                        that the volume of water added can  be  used  to calculate the
                        water level.

              Step 2:   Fill the bottom section of the standpipe with the product
                        (typically this will  require a gallon  or less).  Enough product
                        needs to be added so that the product  level sensor is  high
                        enough not to interfere with the water sensor.             .
                                                 35

-------
Step .3:   Add water'to the cylinder .with, a pipette until the sensor
          detects the presence of the water.  Record the volume of water
••   .     added and the sensor reading at each increment.  The sensor
          reading will be zero until the first sensor response.  At that
          point, total the water increments and calculate the correspond-
          ing level, Xls. of water detected.  Record all data on page l.of
          the Reporting Form for Water Sensor Evaluation Data in
          Appendix B.

Step 4:.  Add water to the cylinder with a pipette in increments to
          produce a height increment, h, of approximately l/20th inch.
          At each increment, record the volume of water added and the
          water height (denoted by W1 ^ in Table 2 of Section 7.2)
          measured by the sensor.  Use1 pages 2 to 4 as necessary of the
          Reporting Form for Water Sensor Evaluation Data in Appen-
          dix B.  Repeat the incremental.addition of water 60 times until
          a total height of about 3 inches (or the range limit of the
          sensor, if less) has been reached.

Step 5:   Empty the product and water from the standpipe, refill with
          product (the same product can be used after separating the
          water) and repeat Steps 2 through 4 20 times to obtain
          20 replications.   .

Record all data using the Reporting Form for Water Sensor Evaluation Data
in'Appendix B.  The 20 minimum detectable.water levels are denoted by Xj,
j=l,...,20.  The sensor reading at the itn increment of the 3   test isj
denoted by W1 -j as described in Section 7.2 and Table 2.
                                    36

-------

                                 SECTION 7

                               CALCULATIONS


      From the results obtained after  all  testing  is  completed,  a  series
 of  calculations will  be performed  to  evaluate  the method's performance.
 If  the method has more than one  mode  of .leak .detection, then the,  perfor-
 mance of the  method must be evaluated and the  results reported  for each
 testing mode  separately.   If the performance is different for different
 modes, this may limit the conditions  under which  the method can be used
 and these should be reported under the limitations section of the results
 form.  •

      The evaluation of the nonvolumetric  test  method is presented
 first.  A separate section (7.2) presents the  calculations to estimate
 the minimum water level  and the  minimum water  level change that the water
 sensor can detect*  Section 7.2  is only needed if the method measures or
 detects  water incursion as one mode of its leak detection.

     The performance  of the nonvolumetric test method is judged on the
 basis- of the  percentage of false alarms and the percentage of correctly
 identified leaks.  The performance standards specify that the false alarm
 rate must be  no more  than 5% and. that the probability of detecting a leak
 rate of  0.10  gallon per hour must  be  at least 95%.  The test procedure
 includes ,21 tests  of  the  tank in the  no-leak condition and 21 tests of.,
 the tank with leaks induced at rates  of 0.10 gallon per hour or less.
 These  data are  used to estimate  the probability of false alarm  and
 probability of  detection  directly.


 7.1 ESTIMATION OF THE METHOD'S  PERFORMANCE PARAMETERS

     After ail  tests  are  performed according to the schedule outlined in
 Section  6, a  total of at  least 42  test  results will be available.  Of
 these, 21  will  have been  obtained  under tight tank conditions,  and 21
 under  induced leak conditions.  The probability of false alarm, P(FA),
 and  the  probability of detection,  P(D), are' calculated next.


 7.1.1  False Alarm Rate,  P(FA)

     The results obtained  from the tests performed under tight  tank
 conditions will  be used to calculate  P(FA).  Let Nj denote the  number of
these tests,  normally 21.   (Note:  This number must be at least 21, but
could be larger  if more tests are called for in the experimental  plan set


     -:...   .       !             .   37        :'       '  "    '    '     .

-------
up at the beginning of the testing.)   Let m denote the  number of  cases
where the method indicated a leak.   If the test results,  U9  are coded  as
zero when no leak is Indicated and  1  when.a leak is indicated,  then
                              TL'=
where the sum is taken over the
estimated by the ratio
tests at zero leak rate.   The P(FA)  is
                              P(FA) =
 In  order  for  the  system to meet the performance standards, the estimated
 P(FA)  must  be less  than or equal to,5%.  Thus, in order for the system to
 meet the  performance  standards, 11i must be no more than 1 if the  .
 standard  21 tests are performed.'           ;                .

      If the method  did not identify the tank to be leaking when it was
 tight, that is, TL, = 0,  then  the  proportion of false alarms becomes
 0%.  However, this  does not  mean that  the method is perfect.  The
 observed  P(FA) of 0%  is an estimate of the  false alarm rate based on the
 evaluation  test results and  the given  test  conditions.

      One  can  calculate an upper confidence  limit for P(FA)  in the case of
 no mistakes.   Let Na  be the  number of  tests performed under the tight
 tank condition.  Choose a confidence coefficient,  (1 - a),  say 95% or
 90%.  Then the upper  confidence .limit, UL,  for P(FA) is calculated as:

                          UL for  P(FA) =l-a
 In the case of 0 false alarms out of 21 tests, the upper limit to P(FA)
 becomes 0.133 or 13.3% with a 95% confidence coefficient.  That is, P(FA)
 is estimated at 0%, and with a confidence of 95%, P(FA) is less than or
 equal to 13.3%.  In general the confidence interval for P(FA) can be
 calculated from the binomial distribution with Nx trials.  The 95%
 confidence interval must be calculated and reported on the results form
 in Appendix B (see page 48).
 7.1.2  Probability of Detecting a Leak, P(D)

      The probability of detection, P(D), is calculated for a'specific
 size of leak.  The size of  leak that can be detected with this proba-
 bility is  also to be reported.  Normally this will be 0.10 gallon per
 hour, as required by the  performance standards.  The exception to this

                                     38

-------
 would occur if . a method is tested using  induced  leak  rates  smaller than
 0.10 gallon per hour,  for example,  0;05  gallon per 'hour.  Report  the
 probability of detection, P(D),  together with the maximum leak  rate used
 in the evaluation testing.  The  leak  rate corresponding to  the  P(D) will-
 be 0,10 gallon per hour or less.

      The results obtained from the  tests performed under induced  leak
 conditions  (leak rates less than  or equal  to 0.10 gallop per  hour)  will
 be used to  calculate P(D).  Let  N2  be the number of such tests.   Typi-
 cally, N2 will  also be 21, but could  be  larger if the evaluation  was
 initially "set up to include more  tests.   Let TL2 be the number of cases
 where the method indicated a leak.  As before, the test results,  L1§ are
 coded as zero when the tank is declared  to be tight and 1 when the  tank
 •is declared to be leaking.  Thus, TL2 is calculated as
 where  the sum is  taken. over  the N2 tests with induced leaks.  The P(D) is
 then estimated  by the  ratio


    '.'.".                        P(P) = TL2/N2.                    ;.
                    '     ' •

^The estimated P(D) must  be at  least 95% for the system to meet the per-
 formarice  standards.  Thus, TL2 must be either 20 or 21 (out of 21 tests)
 for the estimated probability  of detection to be at least 95%.

     If the method identified  the tank to be leaking in all tests where a
 leak was  simulated, then the proportion detected becomes 100%.  However,
 this does not mean that  the method is perfect.  The P(D) of 100% is an
 estimate  of the probability of detection, based on the evaluation test
 results and the given  test conditions.

     One  can  calculate a lower confidence limit for P(D) in the case of
 no mistakes.   Let N2 be  the number of tests performed under the induced
 leak conditions.   Choose a confidence coefficient, (1 - a), say 95% or
 90%.  Theh the  lower confidence limit, LL, for P(D) is calculated as:

                                          1/N2
                           LL  for P(D) = a  .     .


     In the case  of correct identification of 21 tests performed under
 leak conditions,  the lower limit to P(D) becomes 0.867 or 86.7% with a
 95% confidence  coefficient.  That 'is i P(D) is estimated at 100%, ancT-wlth
 a confidence  of 95%, P(D) is greater than or equal to 86.7%.  The 95%
 confidence interval for P(D) must be calculated based on the binomial
 distribution  with N2 trials and reported on the results form in
Appendix  B (see page 48).


                               '    39     .   '      '   •    -  -     '         '

-------
7.2  WATER DETECTION MODE

     This section is only applicable if the method being evaluated uses
detection of water incursion as a leak detection mode.

     Two .parameters will be estimated for the water detection, sensor:
the minimum detectable water level or threshold that the sensor, can
determine, and the smallest change in water level that the device can
record.  These results- will also be reported on the Results of U.S. EPA
Standard Evaluation form in Appendix B.  These parameter estimates will
then be used to calculate the minimum time needed to detect water
incursion at 0.10 gallon per hour for various tank sizes.
                     .               ' f                        ' •

7.2.1  Minimum Detectable Water Level

     The data obtained consist of 20 replications of a determination of
the minimum detectable water level (see test schedule, Section 6.5).

These data, denoted by X,-, j=l,...20, are used to estimate the minimum
                        j                ' ,      '       r    •      '
water level, or threshold, that can be detected reliably.   •

Step 1:  Calculate the mean, X, of the 20 observations:
                                   20
X =
                                       X../20
Step 2:  Calculate the standard .deviation, SD, of the 20 observations:
                         SD =
                                20
                                           o
                                     .  -X)
                                   20-1
                1/2
Step 3:  From'a table of tolerance coefficients, K, for one-sided normal
       *  tolerance intervals with a 95% probability level and a 95%
         coverage, obtain K for a sample size of 20.  This coefficient is
         K = 2.396.  (Reference:  Lieberman, Gerald F.  1958.  "Tables
         for One-Seided Statistical Tolerance Limits."  Industrial Quality
         Control.   Vol.  XIV,  No.  10.)                           -
                                    40

-------
Step 4:  Calculate the upper tolerance limit, TL, for 95% coverage with
         .tolerance coefficient 95%:   f
                              TL = X + K SD9
                                   °r        •'          •'.•-•.
                              TL = X + 2.396 SD

     TL estimates the minimum level of water that the sensor can
detect.  That is, with 95% confidence, the method should detect water at
least 95% of the time when the water depth in the tank reaches TL.

7.2.2  Minimum Water Level Change
     This following statistical procedure provides a means of estimating
the minimum water level change that the water sensor can detect, based on
the schedule outlined in Section 6.5.
     Denote by W^j the sensor reading (in inches) at the jth replicate
(j=l,...,20) and the ith increment (i=l,...,n,-, with n,- being 60 in each
              .       •       ...            M        W
replicate).  Note that the number of steps in each replicate need not be
the same, so the sample sizes are denoted by n.».  Denote by X,- the water
level detected for the first time by the sensor at the jth replicate.
     Denote by h the level change induced at each increment.  The level
change, h, should be chosen to be consistent with the system's claimed
resolution.  That is, the increments should be about half (or less) of
the method's claimed resolution.
                                   »                        •
Step 1:  Calculate the differences between consecutive sensor readings.
         The first increment will be Wlti-X]_ for the first replicate
         (j=l); more generally, W^j-X-j, for the jth replicate.   The
         second increment will be W2ji-Wljl for the first replicate; more;
         generally, W2 i-W1 ,- for the jth replicate, etc.
                      *vA*J           ,              ^              •  ,

Step 2:  Calculate the difference, at each incremental step, between h,
         the level change induced during testing, and the difference
         obtained in Step 1.  Denote these differences by dn- .,,  where i
                                                           1 > J
         and j represent increment and replicate numbers, respectively.
         Table 2 below summarizes the notations.
                                   41

-------
          Table 2.  NOTATION SUMMARY FOR WATER SENSOR READINGS

                          At THE jth REPLICATE
Increment
No.
1
2
3
•
Calculated
1 evel
change
(inch)
A
+ h
+ h
+ h
o
0
0
o
o
•
+ h
Measured
Sensor sensor
reading increment
(inch) (inch)
B C
W2,j W2,o"wl,j
W3 , W3 ,-W2 ,
O,J O,J t,J
« •
• • - •
W W j-W
11^,3 n4.,j n.— 1,
o j j
Increment
difference
calculated-meas.
(inch)
C-A
* . -
.*
'•
     _X,- is the water level  (Inches)  detected  for the first time
       j                                                '

      by the sensor during the jth replication of the test.
Note that using the first sensor reading,  Xit may vary from replicate to
                                           J

replicate, so that the number of differences d^  j will also vary.  Let n


be the number of increments necessary during replicate j.





Step 3:  Calculate the average,  Dj,  of the differences d^ j, -i-l,...,hj,


       .  separately for each replicate j,  j=l,...,20.
                                   o


                            •j-'2
                                   42

-------
Step -4:   Calculate  the variance of the differences d,  ,,  i=l ..... TI .
          separately for each  replicate j,  j=l,..., 20.' J           J
Step 5:  Calculate  the pooled  variance,  Var_,  of  the  20  Variances
                 '».,-• _ 
-------
7.2.3  Time to Detect an Increase In Water Level

     The minimum detectable water level and the minimum detectable change
can be used to estimate the minimum time needed to detect water incursion
into the tank at a specified rate.  This time is specific to each tank
size and geometry and depends on specific assumptions.  The calculations
are illustrated for an 8,000-gallon steel tank that is 96-inch diameter
and 256 inches long.

     Suppose.there are x inches of water in the tank.  The tank is made
of quarter-inch steel, so the inside diameter is 95.5 inches, giving a
radJus, r, of 47.75 inches and a length of 255.5 inches.  The water sur-
face will be 2d wide, where d, in inches, is calculated as
                           d =   r2 - (r - x)2,
where x is the water depth.  The area of the water surface at depth of
x inches of water is then given by 255.5 x 2d inch2.  Multiplying, this by
the minimum level change and dividing the result by 231 inch3 per gallon
gives approximately the volume change in gallons that the sensor can
detect reliably.  This differs with the level of water in the tank.

     For these calculations, the following as'sumptions are used.  The
probe is assumed to be inserted at the midpoint of the tank length and to
rest on a striker plate the top of which is 0.63 inch above the bottom of
the tank.  The initial water depth is taken as the minimum .depth the
sensor can detect with 95% probability plus the striker plate depth of
0.63 inch, rounded up to the next quarter inch.  The tank is assumed
level.  (Calculations show that if the tank is tilted, the cross-
sectional area of the water surface will be slightly less for the same
water depth at this location, so these calculations slightly 'overestimate
the volume.)

     To determine how long the method will take to detect a water incur-
sion at the rate of 0.10 gallon per hour, divide the minimum volume
change that the water sensor can detect by O.JLO gallon per hour.  As a
numerical example, suppose the minimum depth of the water detectable is
0.3 inch and the minimum detectable change is 0.02 inch.  This. gives
x ~ 0.95 inch (0.3 •*• 0.625 rounded up).  In an 8,000-gallon tank with
inside diameter 95.5 inches and length 255.5 inches, the water surface
width, d, is calculated as
                  d =^(47.75)2 - (46.8)2 = 9.43 inches

The volume, in gallons, corresponding to a 0.02-inch increase is

                     V = 2(9.48)  x 255.5 x  (0.02)/231   -

                                    or

                             V = 0.42 gallon

                                   '44

-------
",
•
The time that the  sensor will  take to detect  water incursions at the rate
of 0.10 gallon per hour will  be     ;


           time =  0.42 gallon/0.10 gallon per hour = 4.2 hours


Thus,, the sensor would detect  water coming in at  the rate of 0.10 gallon
per/hour after about 4 hours  15 minutes.   The incursionvof the water into
the tank should be obvious under  these conditions if the test is run for
at least 4 hours 15 minutes.                    •     .

     The minimum amount of water  in a tank that can  be detected by a
sensor depends on  the placement ;of the sensor, any tilt, of the tank,  the
tank size, and the sensor threshold.  This, minimum amount varies from
about 2 gallons to 10 or 15 gallons,  depending on the combination of .
these factors.  If water enters at a  rate  of  0.10 gallon per hour, it
would require anywhere from a  day  to  a week for enough water to be
detected, starting with no water  in the tank.


7.3  OTHER REPORTED CALCULATIONS

    .This section  describes other  calculations needed to  complete the
Results of U.S. EPA Standard Evaluation form  (Appendix B).   Most of these
calculations are straightforward and  are described. here  to provide com-
plete instructions  for the use  of  the results form.
      •  "       '-••'.       -'       —       "       '      •'-••'
    -These sections are only required if they are  applicable  to the
particular nonvolumetric method being evaluated.    If a section  is not
applicable, skip the calculations  and report  "not applicable" on the
results form.


Size of Tank                                \

     The evaluation results are applicable to tanks up to at most 50%
larger capacity than the. test tank and to all  smaller tanks.  Multiply
the volume of the test tank by  1.50.  Round this  number to the nearest
100 gallons and report the result on page 2 of the results form.
                                   45

-------
Maximum Allowable Temperature Difference

     This section is only applicable if temperature conditioning was
needed and used as part of the evaluation procedure.  If temperature does
not affect-the operation of the method, ignore this section and indicate
"not applicable" on the results form.

     Calculate the Standard deviation of the 21 temperaiture differences
actually achieved during testing.  Multiply this number by the factor
± 1.5 and report the result as the temperature range on the limitations
section of the results form.

     The nominal temperature difference of 5°F used in the design was  ,
obtained from data collected on the national survey (Flora, J. D., Jr.,
and J. E. Pelkey, "Typical Tank Testing Conditionsj" EPA Contract
No. 68-01-7383, Work Assignment 22, Task 13, Final Report, December
1988).  This difference was approximately the standard deviation of the
temperature differences observed in the tank tests conducted during the
national survey.  The factor 1.5 is a combination of two effec.ts.  One
effect results'from scaling up the standard deviation of the design
temperature differences to 5°F.  The second effect results from using the
rule that about 80% of the temperature differences on tank tests are    ;
expected to be within ± 1.282 times the standard deviation.


Average Waiting Time After Filling

     Calculate the average of the time intervals between the end of the
filling cycle and the start of the test for the 21 tests that started
immediately after, the specified waiting time.  (Note:  If more than
21 tests are dbnfe immediately after the filling, use all such tests.
However, do not use the time to the start of the second test in a pair as
this would give a misleading waiting time.)  Report this average-time as
the waiting time after adding product on the .results form.  Note:  The   ,
median may be used as the average instead of the mean if there are
atypical waiting times.

     For tracer methods, the average waiting time may more appropriately  -
be the time from adding the tracer to the tank until the completion of
the test.   "                                            .


Average Waiting After "Topping Off"

     If the method fills the tank up into the fill pipe, calculate the
average time interval between the time when the final topping off was
completed and the start of the test.  Calculate this average  using data
from all tests when this step was performed.  Report the result  on the
results form as the waiting time after "topping off" to the final testing.
level.  If this step is not performed  (e.g., for  a test with  the  tank  at
95% of capacity), enter NA  (not  applicable) in the appropriate  space on
                                    46

-------
the results form.  Note:  The median may be used instead of the mean' if
there are some atypical waiting times;                              .


Average Data Collection Time Per Test   _

     Use the duration of the .data collection phase of the tests to
calculate the average data collection time for the total number (at  least
42) of tests*  Report this time as the average data collection time  per
test.         •
Product Level                                                   ;

     If all tests are done at the same product level, report that level
on the results form.  If testing was done at different levels, report the
applicable product level as the acceptable range (e.g., from 6Q% to 90%
full) used in the testing.


Minimum Total Testing Time

     Finally, calculate an average total test time from the test data.
This is the time it would, take from the time the test crew arrives at the
site until a test is completed, the equipment dismantled, and the tank
returned to service.  Typically, it will be the time from initial  setup
of equipment through the first test data collection, plus the time
required to dismantle the equipment.  Report this, total time lapse on the
results form as the minimum time that the tank can be expected to be out
of service for a test of this type.

     The intent of this is to provide an estimate of the time that the
testing will interfere with normal operation of the tank.  The nohvolu-
metric methods will differ in those parts of their operation that require
the tank to be out of service.  Consequently, the time that should be
reported here is the estimated time for which testing with this method
will interfere with the use of the tank by requiring that it be out of
service.                                      ,


7.4  SUPPLEMENTAL CALCULATIONS AND DATA ANALYSES (OPTIONAL)

     This section discusses some additional data analyses that may be
possible with the data, depending on the actual  results.  It also pro-
vides some rationale for the sample size selection.
                                    47

-------
7.4.1  One-Sided Confidence Limits on P(FA) and P(D)
                         «--  -           -             '       j
     It is-possible to estimate the false alarm rate and probability of
detection directly as done in Section 7.1 with any sample size.  However,
for fewer than 20 tests, the estimate of P(FA) will be zero or will
exceed 5%, depending on whether any false alarms are found.  Similarly,
P(D) will be 100% or less than 95% for sample sizes less than 20 depend-
ing on whether any leaks are missed .or not.  Thus, the sample size of 20
is the smallest that allows for one mistake in each case and still pro-
vides estimated performance meeting the EPA standards.  The sample size
of 21 was chosen from experimental design considerations to balance the
different conditions.

     Confidence limits for P(FA) and P(D) can be calculated based on the
observed results and the sample sizes.  The formulas for perfect scores
were given in Section 7.1.1 for P(FA) and in Section 7.1.2 for P(D).
These also depend on the selected confidence coefficient.  Table 3 below
gives 90% and 95% one-sided confidence limits for P(FA) and P(D) based on
samples-of 21 tests for the case of no mistakes and one mistake, the two
conditions under which the method meets the EPA performance standards if
evaluated with the minimum 21 tests.
                  Table  3.   ONE-SIDED CONFIDENCE  LIMITS
                             FOR P(FA) AND P(D)
Field test
results
0
1
Error
Error
out
out
of
of
21
21
Confidence

P(FA)
P(FA)
90%
< 0.
< 0.
coefficient
95%
104
173
P(FA)
P(FA)
<
<
0
0
.133
.207
0
1
Error
Error
out
out
of
of
21
21
P(D)
P(D)
> 0
> 0
.896
.827
P(D) i
P(D) ;
> 0
> 0
.867-
.793
Table  3 shows  that the confidence limits  start  to  become  fairly  large for
high .confidence with even one error.   Using  a larger  sample  size would
improve the confidence limits, but would  add significantly to the  cost of
testing.  The  sample sizes were selected  as  a compromise  to  provide
reasonable estimates while not requiring  excessively  expensive testing.
 7.4.2  Alternative Statistical Model

      If the evaluation uses three non-zero leak rates and if the method
 fails to detect some of the induced leaks, an alternative statistical
 analysis may be possible.  This alternative'statistical method fits a
 logistic model  to the data, assuming that the probability-of'detecting a
                                    48

-------
 JSJTT8"?! w1th"tJe Si2e of ««: !«*•  I* one assumes that the logis-
 tic model with parameters A and B holds, then the probability of detect-
 ing a leak can be expressed as:                                  ucueut


        PfDetecting a leak  given a leak of size  S]  =  l/[l+exp(A+BS)]


 That is, the probability that the test method will indicate a leak when
 there is an actual induced leak rate,  S, is given by the logistic func-
 I ™H RTh| ftta fr01?.a11 ^ests' Can  be used to es*1mate the parameters
 ?hS*iS ™;H?*MqUatl0n-  T?1S reqUlres an 1terative estimationV technique
 * a£ 1S ayailab.l£ ln several  commercial statistical  software packages
 such as SAS, BMDP, or SYSTAT.   The estimation will not converge  if no
 mistakes are made, and it may not converge if only a few mistakes are
 I?!??,**  I J   es|imates do converge, then the function with  the estimated
 values of A and B can be used  to estimate the P(FA)  of the method bv sub-
 stituting S - - 0.   The P(D).can be estimated for any  leak rate  S by sub-
 stituting^S into  the equation.   Specifically, S =  0.10 gallon, per hour
  anKbK-?^Stltuted to con>Pare  with the EPA performance standards  for
 probability of detection.'                     '


 7.;4.3  Estimation of Temperature Effect

   ' .  JI  the temperature and stabilization time  variables  influence the
 operat on of thetest and  testing  is done according to the full set of
 conditions in Table  1,  the logistic model can also be used to test
 whether  the additional  variables did have a significant effect on the
S5J«™ nf ;»« 9 1  fiwh,e!her th}! 1s Possible depends on the number and
pattern of the actual data results.  The approach is to add one or more
indicator variables to the logistic model to estimate the effect of thl
additional factor.  The model would become


    P[Detecting a leak given a leak of size S] = l/[l+exp(A+BS+C-T-)]


where the three temperature conditions were identified by T,  and coded
appropriately.  This modeling becomes rather involved.  The evaluating
organization should involve statistical support if these additional  cal-
culations are warranted.  Note that this modeling will generally hot be
p??!?b]VLn?VyStem Performs so well that the direct estimates of
P(FA) and P(D) described in Section 7.1 meet the EPA performance stan-
dards. Thus,  this approach is supplemental  to provide information for a
vendor to use in improving a method by identifying factors that  sianifi-
cantly affect the system's performance.                         »i-gm.n
                                   49

-------

-------
                                 SECTION 8

                               INTERPRETATION


      The results reported are valid for the experimental conditions dur-
 ing the evaluation, which have been chosen to represent situations com-
 monly encountered in the field. • These should be typical of most tank
 testing conditions, but extreme conditions can occur and might adversely
 affect the performance of the method.   It should be emphasized that the
 performance estimates are based on average results obtained in the
 tests.  An individual test may not do  as'well.  Some individual tests may
 do better. .
 8.1  BASIC PERFORMANCE ESTIMATES

    .The relevant performance measures  for proving  that  a tightness  test
 method meets EPA standards are the P(FA)  and  P(D) for  a  leak  rate  of
 O.lO^gallon per hour.   The estimated  P(FA)  can  be compared with  the  EPA
 standard of P(FA) not  to exceed 5%.   In general, a  lower P(FA) is
 preferable, since it implies  that the chance  of mistakenly indicating  a
 leak on a tight tank is less.   For a  concern  with many tanks, there  will
 be  fewer false.alarms.   However,, reducing the false alarm rate may also
 reduce the chance of detecting a leak.  The probability  of detection
 generally increases  with the  size of  the  leak.  The EPA  standard .spec-
 ifies that P(D)  be at  least 95% for a leak  of 0.10  gallon per hour.  A
 higher estimated P(D-) means that there  is less  chance  of  missing a small
 leak.                                          ,

      The  discrete nature of the data  implies  that only a  few values  of
 P(FA)  or  P(D)  are possible.  With the standard 21 tests for each test
 condition (tight or .leaking tank), the  possible values are 0, 1/21,  2/21
 etc.   Consequently,  the reported  estimates  are only precise to about
 5%.   The  confidence  limits  reported in  the  case of a perfect score
 indicate  the range in which the true  P(FA)  or P(D)   is  expected to  be.
 For  example, a method that  achieved zero  false alarms  out of 21 would not
 be expected to have a zero false  alarm  rate.  Instead,'its false alarm
 rate should_be less than  10.4$  with 95% confidence.

      If testing  is done  at an  induced lak rate less  than 0.10 gallon per
 hour, the  P(D) may be reported  at the smaller leak  rate actually used.
The  standard test, using an induced leak rate of 0.10 gallon per hour
would report P(D) for the rate of 0.10 gallon per hour.  In general, a
                                    51

-------
method that can detect a smaller leak with high probability ,i's to be
preferred because it will identify a potential problem earlier.  This may
reduce the amount of pollution and the cost of remedial action.
8.2  LIMITATIONS              .

     Nonvolumetric tank tightness testing methods that are based on
different operating principles will have different factors that can
interfere with their performance.  Consequently, the limitations on the
applicability of-the performance estimates will also vary with the
method.  If a factor, for example temperature, does not affect the
principle of operation, it should not be reported as a limitation.
However, there may be interfering factors other than those listed in the
experimental plan that affect a particular test method.  If so, those
additional factors might limit the applicability of the method.  The
•reporting form provides a place to identify other sources of interference
and to  state the test conditions for them.

     Some nonvolumetric test methods use more than one mode of
operation.   If so, different limitations may apply to each mode of leak
detection.   It is possible that one mode of operation may be unaffected,
by size of tank, but that another may depend strongly on tank size.  For
example, a water sensor may be used to .test for  leaks in the presence of
a high  ground-water  level.  It may do so by sensing water incursion, in
which case  it must be  able to detect water incursion at the rate  of
0.10 gallon  per  hour.  Since the time required for the water level to be_
.detectable  at  a  fixed  rate of  incursion will be  a function of.the size of
the tank, this mode  of  leak detection is dependent on tank size.
 8.3  HATER LEVEL DETECTION FUNCTION

      If the .system uses a water level sensor,  the following results are
 reported.                    .       ,"•••',

      The minimum water level detected by the sensor is;  estimated from_the
 average threshold of detection, and the variability of  the water level
 threshold is estimated' b'y the standard deviation of the test data.  The
 minimum water level that will be detected at least 95%  of the time is the
 level to be reported.  Statistically, this is a one-sided tolerance
 limit.    .                            •

      The'tolerance limit calculated  in Section 7.2.1 estimates the
 minimum water level that the sensor  can detect above the bottom of the
 probe.  If the installation of the sensor leaves the probe at a specified
 distance above the bottom of the tank (for example, 1 inch), then this
 minimum distance needs to'be added to the reported minimum detectable
 water level.            ,
                                     52

-------
 8.4  MINIMUM WATER LEVEL CHANGE MEASUREMENT
                •            -' c         -: •          - .
 _    The water sensor may be  used to test for leaks in the event of a
 h]9h ground-water level.  If  the ground-water level is above the bottom
 of  the tank, there will be an inward pressure when  the product level is
 sufficiently low, and if there is a hole in the tank, water will flow
 into the tank under these conditions.  Based on the'ability of the water
 sensor to detect a change in  the level of water in  the product, one can
 determine how much water must enter the tank in order for an increase in
 the water level to be detected.   From this information, in turn, one can
 determine the size of a leak  of  water into the tank that the system can
 detect at a given time.

      The standard deviation of the differences between the change in
 water level  measured by the sensor and the change induced during the
 tests is used to determine the ability of the water level  sensor to
 detect changes in the water level.   A two-sided 95% tolerance interval  is
 then calculated for this detection ability (Section 7.2.2).

      The minimum change  in water level  that can be  detected  is  used to
 compute a minimum change in water volume  in the tank.   This  conversion  is
 specific to  the tank size.  Using the minimum change  in water volume that
 the sensor can detect, the time  needed  for the method  to detect  an  incur^
 sion of water at the rate of 0.10 gallon  per hour is calculated  (Sec-
 tion 7.2.3).   This calculation indicates  the minimum time  needed for the
 water detector to identify an  inflow of water at the minimum leak rate
 and to alert  the test operator that  the water level has increased.


 8.5  ADDITIONAL CALCULATIONS

    .If  the performance  estimates do  not meet  the performance require-
ments, the vendor may want  to  investigate the  conditions under which
errors occurred.   Calculating  the percent of errors by size of leak  by
temperature condition, and  by  length of stabilization time as applicable
may  suggest .ways  to  improve the method.  This may be as straightforward
as  identifying conditions that lead to popr performance and revising the
operating procedure to avoid those, or  it may require redesign of the
method.      (.        .        .

     The relationship of performance to test conditions is primarily of
interest when the method does not meet the EPA performance standards.
Developing these relationships is part of the optional or supplementary
data analysis that may be useful  to the vendor, but  not to many tank
owners or operators.
                                   53

-------

-------
                                SECTION  9

                           REPORTING OF RESULTS


    .Appendix B is designed to be the framework for a standard report.
There are five parts to Appendix B, each  of which is preceded by instruc-
tions for completion.  The first part is  the Results of U.S. EPA Standard
Evaluation form.  This is basically an executive summary of the find-
ings.  It is designed to be used as a form that would be provided to each
tank owner/operator that Uses this system of leak detection.  Conse-
quently, it is quite succinct.  The report should be structured so that
this results form can be easily reproduced for wide distribution.

     A method that uses more than one mode of leak detection may achieve
different performance results for the different modes of operation.  The
results form is structured to allow for reporting the P(FA) and P(D)
separately for different modes of leak detection.  The method meets the
EPA performance requirements only if all  modes of leak detection meet
those requirements.                  .      .        :

     Suppose that a method had two modes  of testing,  a basic one and an
ancillary one for testing in the presence of a high ground-water level.
Suppose that the test method when evaluated in the case of high ground-
water level did not meet the EPA performance requirements, but the basic
one did.  Then a report could be issued,  stating that the method meets
the EPA performance requirements, but cannot test when the ground-water
level.is above the bottom of the tank. -

     The statement of compliance with the EPA performance standards must
be consistent with stated limitations on  the form and also with the
standard operation of the method as described on the  Description form.

     The second part of the standard report consists  of the Description
of the method.  A description form is included in Appendix B and should
be completed by the evaluating organization assisted  by the vendor.

     The third part of the standard report contains a Reporting Form for
Leak Test Results, also described in Appendix B.  This table summarizes
the test results and contains the information on starting dates and
times,  test duration, leak test results, etc.                       .
                                    55

-------
     •The fourth part of Appendix B contains a blank Individual Test
Log.  While the Individual Test Log has been designed to be flexible, it
may need modifications for some test methods.  This form should be repro-
duced and used to record data in the field.  Copies of the completed
daily test logs are to be included in the standard report.  These serve
as the backup data to document the performance estimates reported.

     The fifth part of Appendix B provides a form to record the test
results when evaluating the system's water sensor.  The data to be   .,
recorded follow the testing protocol (in Section 6.5) to determine the
minimum level of water and the minimum water level change that the system
can detect.  This part is only applicable if the system uses a water
sensor.

     If the optional calculations described in Section 7.4 are performed,
they should be reported to the vendor.  It is suggested that these
results be reported in-a separate section of the report, distinct from
the standard report.  This would allow a user to identify the parts of
the standard report quickly while still having the supplemental informa-
tion available if needed.

     The limitations on the results of the evaluation tire to be reported
on the Results of U.S. EPA Standard Evaluation form.  The intent is to
document that the results are valid under conditions represented by the ,
test conditions.  Section 7.3 describes the summary of the test condi-
tions that should be reported as limitations on the results form.  These
items are also discussed below.  The test conditions have been chosen to
represent the majority of testing situations, but do not include the most
extreme conditions under which testing could be done.  The test condi-
tions were also selected to be practical and not impose an undue burden
for evaluation on the test companies.

     One practical limitation of the results is the size of the tank.
Tests based on volumetric changes generally perform less well as the size
of the tank increases.  However, for some nonvolumetric test methods,
size is not such a restriction.  The evaluating organization must deter-
mine the extent to which tank size affects performance and report a size
limitation here.

     A second potential limitation on the results is the temperature
differential between the product added to the tank and that of the
product already in the tank.  Testing during the EPA national survey
(Flora, J. D., Jr., and J. E. Pelkey, "Typical Tank Testing Conditions,"
EPA Contract No. 68-01-7383, Work Assignment 22, Task 13, Final Report,
December 1988) found that temperature differentials were no more than 5°F
for at least 60% of the tests.  However, it is clear that larger differ-
ences could exist.  If temperature affects the method, then the tempera-
ture differences used in the evaluation must be reported.  If the physi-
cal principle of the method is not affected by temperature, then report
that the method is not limited by temperature and the basis for this
conclusion.  The evaluation testing may be done using larger temperature
                                    56

-------
differentials, reporting those  actually  used.  The  results  cannot  be
guaranteed for temperature differentials larger than those  used  in the
evaluation.

     A third  limitation on the  results is the time  heeded by the method
for its operation.  For example, tracer  methods require some time  for  the
tracer to move through the backfill to the sensors.  The Individual Test
Logs call for recording the actual time  used in the testing.  The  average
time is to be reported and the  results should be valid for  times at least
this long.  It may be the case  for some  nonvolumetric methods that the
time for preparation does not require taking the tank out of service.  If
so, this should be noted.

     The duration of the data collecting phase of the test  is another
limitation of the method.  If a test shortens the data collection  time
and so collects less data, this may adversely affect the method's  perfor-
mance.  As a  consequence, the results do not apply  if the data collection
time is shortened.  This is primarily of concern in documenting that a
tank is tight.  If results clearly indicate a leak, this may sometimes be
ascertained in less time than needed to document a tight tank, particu-
larly if the  leak rate is large.  Thus, while the false alarm rate may be
larger if the test time is shortened, this is not usually a problem in
that if test'results indicate a leak, efforts are usually made to  iden-
tify and correct the source of  the leak.

     If the method uses a water detector as part of its operation, the
minimum depth of water that the sensor can detect is reported.  In addi-
tion, the minimum change in water level that the sensor can detect is
reported.  From this minimum detectable change in water level, a,minimum
volume change can be calculated based on the tank size and depth of the
water.  A minimum time for detection is calculated and reported as the
time needed for water flowing into the tank at the rate of 0.10 gallon
per hour to increase the water volume enough to be detected by the
sensor.
                                                              /• -
     It is expected that nonvolumetric methods may require some
modification of the forms.  It  is hoped that the forms supplied will be
flexible enough to provide for most of the data recording needs.   How-
ever, if modifications are needed to accommodate a particular method,  the
evaluating organization should make the required modifications and use
the resulting forms.   The conditions during the evaluation tests  are to
be recorded.  The factors that affect the performance of the method being
evaluated must be recorded.   The performance results are limited  by the
test conditions actually used and reported.
                                   57

-------

-------
              APPENDIX A





DEFINITIONS AND NOTATIONAL CONVENTIONS
                 A-l

-------
     •In this, protocol leaks are viewed as product lost from the tank.  As
a convention, leak rates are positive numbers, representing the amount of
product loss per unit time.  Thus a larger leak represents a greater
product loss.  Parts of the leak detection industry report volume changes
per unit time with the sign indicating whether product is lost from the
tank (negative sign) or is coming into the tank (positive sign).  We
emphasize that here, leaks refer to the direction out of the tank and the
rate to the magnitude of the flow.

     The performance of a leak detection method is expressed in terms of
the false alarm rate, P(FA), and the .probability of detecting a leak of
specified size, P(D(R)), where R is the leak rate.  In order to under-
stand these concepts, some explanation is helpful.     ,

     Nonvolumetric test methods make a determination of whether a tank is
leaking or not.  The false alarm rate is the proportion of times that the
method would incorrectly indicate that a tight tank is leaking.  The
probability of detection is the probability that the method will cor-
rectly identify a leak of specified size, R.  Usually, the larger the
leak rate, the more likely the method is to detect it, so the probability
of detection must specify the leak rate to be detected.  In evaluating
nonvolumetric methods, the performance measures are generally estimated
directly from the test results.  The false alarm rate is estimated by
conducting a number of trials on a tight tank and calculating the pro-
portion of those during which the method incorrectly indicates a leak.
The probability of detection is estimated by conducting a series of
trials with an induced leak rate, R, and calculating the proportion of
those trials during which the method correctly identifies the tank as
leaking.                                                     :

    . Definitions of some of the terms used throughout the protocol are
presented ne'xt.

Nominal Leak Rate:        The set or target leak rate to be achieved as
                          closely as possible during testing.  It is a
                          positive number in gallon per hour.

Induced Leak Rate:        The actual leak rate, in .gallon per hour, used
                          during testing, 'against which the results from
                          a given test device will be compared.

False Alarm:              Declaring that a tank is leaking when in fact,
                          it is tight.
                     '                             '             '
Probability of            The probability of declaring a tank leaking
False Alarm, P(FA):       -when it is tight.  In statistical terms, this
                          is also called the Type I error and is denoted
                          by alpha (o)..  It is usually expressed in
                          percent, say, 5%.
                                   A-2

-------
Probability of      '      The probability of detecting a leak rate of a
Detection, P(D(R)):       given sizei R gallon per hour.  In statistical
                          terms, it is the power of the test method and
                          is calculated,as one minus beta (e), where beta
                          is the probability of not detecting (missing) a
                          leak rate R.  Commonly the power of a test is
                          expressed in percent, say, 95%.

Resolution:               The resolution of a measurement system is the
                          least change in the quantity being measured
                          which the system is capable of detecting.
                                  A-3

-------


-------
  APPENDIX  B





REPORTING FORMS
  B-l

-------

-------
 Appendix B provides'five sets of blank forms.  Once filled out, these
 forms-will provide the framework for a standard report.  They consist of
 the  following:                                                    .     ,

 1.    Results of U.S.  EPA Standard Evaluation—.Nonvolumetric Tank Tight-
      ness Testing Method (four pages)

 2.    Description—Nonvolumetric Tank Tightness Testing Method (six  pages)

 3.    Reporting  Form for Leak  Test Results—Nonvolumetrtc Tank Tightness
      Testing Method (three  pages)

 4.    Individual  Test  Log—Nonvolumetric Tank Tightness Testing Method
      (five pages)       .,

 5.    Reporting  Form for Water Sensor Evaluation Data—Nonvolumetric Tank
      Tightness  Testing Method (four  pages)

 Each set of forms  is  preceded by instructions on how the  forms  are  to be
 filled  out and  by  whom.   The  following  is an overview  on  various
 responsibilities.

 Who  1s  responsible for filling out which form?

 1.    Results of  UoS.  EPA Standard Evaluation.   The evaluating organiza-
      tion is responsible  for  completing this  form at the  end of the
      evaluation.

 2.    Description of Nonvolumetric Tank Tightness Testing  Method.  The
      evaluating  organization  assisted by the  vendor will  complete this
      form by the end  of the evaluation.

 3.    Reporting Form for  Leak Test Results.  This form  is to be completed
      by  the  evaluating organization.  In general, the  statistician
      analyzing the  data will complete this form.  A blank form can be
      developed on a personal computer, the data base for a given
      evaluation  generated, and the two merged on the computer.  The form
      can  also be filled out manually. The input for that form will
      consist of  the field test results recorded by the evaluating
      organization's field crew on the Individual Test Logs (below) and
      the vendor's test results.
                       ' '                   •     -      -          l .
4.    Individual Test Logs.  These, forms are to be used and completed by
     the evaluating organization's field crew.  These forms ne.ed to be
      kept blind to the vendor during testing.  It is recommended that the
     evaluating organization reproduce a sufficient number (at least 42
     copies) of the blank form provided in this appendix and produce a
     bound notebook for the complete test period.

      It is expected that nonvolumetric methods may require some modifica-
     tion of the test  log.  The form provided in this appendix was
     designed from a volumetric test log.  It is the responsibility of


                               B-2

-------
     the evaluating organization to design the appropriate forms with the
     vendor's input.  It is important to include in the test logs all
     parameters relevant to the evaluation of a specific method:  In
     particulars it is necessary to document the inducted leaks.

5.   Reporting Form for Water Sensor Evaluation Data.  These forms pro-
     vide a template for the water sensor evaluation data if the method
     includes such a leak detection.mode.  The forms are to be used,and
     completed by the evaluating organization's field crew.  It is
     recommended that the evaluating organization reproduce a sufficient
     number (at least 20 copies) of the blank form provided in this
     appendix and produce a bound notebook to be used in the field.

At the completion of the evaluation, the evaluating organization will
collate all the forms into a single Standard Report in the order listed
above.  In those cases where the evaluating organization performed addi-
tional, optional calculations (see Section 7.4 of the protocol), these
results may be attacin-J to the standard report. There is no reporting
requirement for these calculations, however.                 :

Distribution of the Evaluation Test Results

The organization performing the evaluation will prepare a report for the
vendor describing the results of the evaluation.  This report consists
primarily of the forms in Appendix B.  The first form reports the results
of the evaluation.  This four-page form is designed to be distributed
widely.  A copy of this four-page form will be supplied to each tank
owner/bperator who uses this method of leak detection.  The owner/
operator must retain a copy of this form as part of his record keeping
requirements.  The owner/operator must also retain copies of each tank
test performed at his facility to document that the tarik(s) passed the
tightness test.  This four-page form will also be distributed to regula-
tors who must approve leak detection methods for use in their jurisdic-
tion.

The complete report, including all the forms in Appendix B, will be
submitted by the evaluating organization to the vendor of the leak detec-
tion method.  The vendor may distribute the complete report to regulators
who wish to see the data collected during the evaluation.  It may also be
distributed to customers of the leak detection method who want to see the
additional information before deciding to select a particular leak detec-
tion method.                  .
                                B-3

-------
The optional part of the calculations (Section 7,4), if done, would be
reported by the evaluating organization to the vendor of the leak detec-
tion method.  This is intended primarily for the vendor's use in under-
standing the details of the performance, and perhaps suggesting how to
improve the method. It is left to the vendor whether to distribute this
form, and if so, to whom.

The evaluating organization of the leak detection method provides the
report to the vendor.  Distribution of the results to tank owner/
operators and tp regulators is the responsibility of the vendor.

The forms, each preceded by its instructions for completion, are
presented next.
                               B-4

-------

-------
                  Results pf U.S. EPA Standard Evaluation
                Nonvolumetric Tank Tightness Testing Method

                    Instructions  for  completingi the  form

 This 3-page form is to be filled out by the evaluating organization upon
 completion of the evaluation of the  method.  This form will  contain the
 most important information relative, to the method evaluation.   All  items
 are to be filled out and the appropriate boxes  checked.   If  a  question is
 not applicable to the method, write  "NA" in the  appropriate  space.

 This form consists of six main parts.   These are:

 1.    Method Description                                        ,
 2.    Evaluation Results                                               .
 ,3.    Test Conditions During Evaluation             .
 4.    Limitations on the Results
 5.    Certification of Results
 6.    Additional Evaluation Results (if applicable)

 Method Description                                          ,          "

 Indicate the commercial name of  the  method,  the  version, and the name,
 address, and telephone number of the vendor.  Some  vendors might use
 different versions  of their method when using it  with  different products
 or  tank sizes.   If  so, indicate  the  version  used  in the evaluation.   If
 the vendor is not the party responsible for  the development and use of
 the method,  then indicate the home office  name and  address of the
 responsible  party.

 Evaluation Results

 The evaluation  results must be reported  separately  for each detection
 mode if the  method  operates in different detection  modes depending on
 field  conditions.   Describe the  mode of detection for which the results
 are applicable.                               .

 P(FA)  is  the probability  of false alarm as calculated  in Section 7.1.1.

 Report  the number of  false  alarms and  the  number of tight tank tests, and
 report  the 95%  confidence interval based on the binomial distribution
 with NH  tests.  Some  values  are  tabled on  page 48;

 The  leak  rate .used  in.the evaluation, is to be inserted in the blank.
 This is the  leak rate  corresponding to the reported P(D) below.

 P(D) is the  probability of  detecting a leak of the  size induced (no more
than 0.10 gallon per  hour)  as calculated in Section 7.1.2.

Report the number of correct detections and the number of simulated leak
tests, and report the  95% confidence interval based on the binomial
distribution with N2 tests.  Some values are tabled on page 48.


  '..-.-•             . B.5   -   :  •       .  •   '••- ,       .  •''   -••  '•

-------
 If the calculated P(FA)  is  5% or  less  and  if the calculated P(D)  is 95%
 or more,  then check the  "does"  box.  Otherwise, check the  "does not1'
'box.   Note:   the P(FA) and  P(D) requirements apply to.each leak detection
 mode  used by the method.

 Indicate  whether this method  operates  under mbre than one mode of detec-
 tion.  Check the appropriate  box  and complete page 4  (Additional  Evalua-
 tion  Results) if applicable.        ,

 Test  Conditions During Evaluation

 Insert the information in the blanks provided.  The nominal volume of  the
 tank  in gallons is requested  as is the tank material, steel, or fiber-
 glass.  Also report the  backfill  material  in the tank excavation, e.g.,
 clean sand or pea gravel.  Give the tank diameter and length in inches.
 Report the product used  in  the testing.  Give the range of temperature
 differences actually measured as  well  as the standard deviation of the
 observed  temperature .differences. Report  the ground water level  for .the
 test  tank in inches above the bottom of the tank.  Report zero for ground.
 water at or below the bottom  of the tank.

 Other sources of interference may affect non-volumetric methods.  Report
 any sources of interference specific to the method on the  lines pro-
 vided.  Also report the  range of  test  conditions for the indicated
 interference source.  If no additional sources of interference were
 identified, check "None."

 Limitations on the Results

 The size (gallons) of the largest tank to  which these results can be
.applied may be calculated as  1.50 times the size  (gallons) of the test
 tank.                                      ,

 The temperature differential, the waiting  time after  adding product until
 testing,-and the total data collection time should be completed using  the
 results from calculations in  Section 7.1.4.  Alternately,  if the
 principle of operation of the method  is riot affected  by product
 temperature changes, check the box indicating that temperature is not  a
 limiting factor and give' the  justification.

 Certification of Results

 Here, the responsible pers.on at the  evaluating organization  indicates    .
 which test procedure was followed and  provides his/her  name  and  signa-
 ture, and the name, address,  and  telephone number of  the  organization.,

 Additional Evaluation Results (if applicable)

 If the "yes" box relating to other leak detection modes on page  1 was
 checked,  then provide the necessary  information  for  the P(FA)  and P(D)
 for the additional leak  detection mode.  These probabilities  will have
 been calculated as described in Sections  7.1.1  and 7*1.2,  based  on the
 evaluation results obtained in jthat detection mode.               ,

                                B-6                      ..-..••

-------
     out this section as described on page B-5.

If the method includes a water sensor, then complete the results for that
sensor.

The minimum detectable water level and the minimum detectable level
change that the sensor can detect will have been obtained from the
calculations in Sections 7.2.1 and 7.2.2.

The minimum time for the water sensor to detect a leak of 0.10 gallon per
hour by detecting an increase in the water level in the tank will have
been obtained from the calculations in Section 7.2.3.  This time is
calculated based on a water depth equal to the striker plate height plus
the minimum detectable water level (above the striker .plate).  It assumes
a level tank and that the sensor is located midway along the tank length.
The minimum detectable increase is used to calculate the volume change
needed.  This volume is divided by 0.10 gallon per hour to get the time
reported. .Indicate the size of the tank on which this t/ime calculation
is based.
                               B-7

-------
  I
*

-------
                     Results of U.S. EPA Standard Evaluation

         Nonvolumetric Tank Tightness Testing Method

  This form tells whether- the tank tightness testing method described below complies with the
  performance requirements of the federal underground storage tank regulation. The evaluation was
  conducted by the equipment manufacturer or a consultant to the manufacturer according to the
  U.S. EPA's "Standard Test Procedure for Evaluating Leak Detection Methods: Nonvolumetric
  Tank Tightness Testing Methods." The full evaluation report also includes a form describing the
  method and a form summarizing the test data.      •

  Tank owners using this leak detection system should keep this form on file to prove compliance
  with the federal regulations. Tank owners should check with State and local agencies to make sure
  this form satisfies their requirements.             .


  Method Description                   '   , ',      -x   ,       -
 .Name
  Version

  Vendor
                      (street address)
                                  (state)    -   (zip)        •         (phone)
 Evaluation Results
 This method, which declares a tank to be leaking when
 has an estimated probability of false alarms [P(FA)] of	% based on the test
 results of _	false alarms out of.	 tests. A 95% confidence interval for P/FAV
 is from                                                        .,.•'-
            to           %.
 The corresponding probability of detection [P(D)] of a '•          gallon per hour leak is
 __;	% based on the test results of	 detections out of       '  ,
 simulated leak.-tests. A 95% confidence interval for P(D) is from ______ to   	%.

    Does this method use additional modes of leak detection? D Yes D No. If .Yes, complete
    additional evaluation results on page 3 of this form.

 Based on the results above, and on page 3 if applicable, this method  D does  ED does not  '
 meet the federal performance standards established by the U.S. Environmental Protection
 Agency (0.10 gallon per hour at P(D) of 95% and P(FA) of 5%).

 Test Conditions During Evaluation                 "'~        ""

The evaluation testing was conducted in a __^_____ -gallon  D steel D fiberglass tank
 that was	___^—_ inches in diameter and	 inches long, installed in
,—	:	      • -  '    •	 backfill.   —'

 The ground-water level was	inches above the bottom of the tank.

 Nonvplumetric TTT Method - Results Form                                        " Page i of 3

-------
Nonvolumetric
Version	
Method
Test Conditions During Evaluation (continued)
The tests were conducted with the tank        ' '   percent full.
The temperature difference between product added to fill the tank and product
already in the tank ranged from	°F  to             °F.
with a standard deviation of _____	°F.
The product used jn the evaluation was	____.
This method may be affected by other sources of interference. List these interferences below
and give the ranges of conditions under which the evaluation was done: (Check None if not'
applicable.)  ~~
   None
            Interferences
                                 Range of Test Conditions
Limitations on the Results
The performance estimates above are only valid when:
• The method has not been substantially changed.       ...
• The vendor's instructions for using the method are followed.
• The tank contains a product identified on the method description form.-
• The tank capacity is  -         gallons or smaller.
* The difference between added and in-tank product temperatures
   is no greater than+• or -	degrees Fahrenheit.
   CU Check if applicable:
   Temperature  is not a factor because  	.                	
 • The waiting time between the end of filling the test tank and the start of the test data collec-
   tion is at least _________ hours.            .            .•'.•,.
 • The waiting time between the end of "topping off" to final testing level and the start of
   the test data collection is at least	hours.
 • The total data collection time for the test is at least            hours.
 • The product volume in the tank during testing is ______% full.
 • This method Dean EH cannot be used if the ground-water level is above the bottom of
   the tank.                      '
Other limitations specified by the vendor or determined during testing:
Nonvolumetric TTT Method - Results Form
                                                          Page 2 of 3

-------
  Nonvolumetric TTT Method,
  Version
   >  Safety disclaimer: This test procedure only addresses the issue of the method's
     ability to detect leaks. It does not test the equipment for safety hazards.


  Additional Evaluation Results (if applicable)        "~   ~    "~~"                ~~~
  This method, which declares a tank to be leaking when	._	
  has an estimated probability of false alarms [P(FA)] of -   '	% based on the test
  results of	 false alarms out of _____ tests. Note: A perfect score during testing
  does not mean that the method is perfect/Based on the observed results  a 95% confidence
  interval for P(FA) is from 0 to         %.

  The corresponding probability of detection [P(D)] of a	 gallon per hour leak is
          % based on the test results of  .-        detections out of      -    simulated
.  leak tests. Note: A perfect score during testing does not mean that the method is perfect
  Based on the observed results,  a 95% confidence interval for P(D) is from 0 to ____%.


  >  Water detection mode (if applicable)

  Using a false a'latm rate of 5%, the minimum water level that the water sensor can detect
  with a 95% probability of detection is        •    inches.

•  Using a false alarm rate of 5%, the minimum change in water level that the water sensor
  can detect with _a 95% probability of detection is _;	inches.

  Based on the minimum water level and change in water-level that the water sensor can
  detect with a false alarm rate of 5% and a 95% probability of detection, the minimum time for
  the system to detect an increase in water level at an incursion rate of 0.10 gallon per hour is
  	;	: minutes in a          -gallon tank.                                  .


  Certification of Results                *                    '  ' - '        •    *  -  .

  I certify that the nonvolumetric tank tightness testing method was installed and operated
  according to the vendor's instructions. I also certify that the evaluation was performed
  according to the standard EPA test procedure for nonvolumetric tank tightness testing
  methods and that the results presented above are those obtained during the evaluation.
 (printed name)   ~~  ~~              !         (organization performing evaluation)
 (signature)                ,    •                 (city, state, zip)
 (date)                                         (phone number)
 Nonvolumetric TTT Method - Results Form                           .''..-     Page 3 of 3

-------

-------
         Description of Nonvolumetric Tank Tightness Testing Method

                    Instructions for completing the form

 This 6-page form is to be filled out by the evaluating organization with
 assistance from the vendor, as part of the evaluation of the method.   '
 This form.provides supporting information on the principles behind the
 system or on how the equipment works.

 To minimize the time to complete this form, the most frequently expected
 answers to the questions have been provided.  For those answers that are
 dependent on site conditions, please give answers that apply in "typical"
-conditions.  Please write in any additional information about the testinq
 method that you believe is important.

 There are seven parts to this form.  These are:

 1.   Method Name and Version
 2., .  Product
      > Product type
 .     > Product level                                           ,
 3.   Principle of Operation
 4.   Temperature Measurement
 5.   Data Acquisition
 6.   Procedure Information
      > Waiting times                       .        .
      > Test duration
      > Total  time
      > Other important elements  of the procedure  or method
      > Identifying  and correcting  for interfering factors
      > Interpreting test results
 7.   Exceptions

 Indicate the, commercial  name and the  version of the method  in the first
 part.

 NOTE:   The  version  is  provided for methods that use different versions
        of the  equipment -for different products or tank sizes.

 For the six remaining  parts,  check  all appropriate boxes for each
 question. Check more than  one box  per question if  it applies.  If a box
 "Other"  is checked,  please complete the space provided to specify or
 briefly describe  the matter.  If necessary, use all the white space next
 to a question for a description.

 The section "> Other important elements of the procedure or method"
 should  be completed carefully.   List, here any other important elements of
 the procedure or method that  could affect its performance.  For example:

 -  If-the pressure in the ullage space is different from atmospheric
  during testing, indicate whether a negative or positive pressure was
  applied.  Report that pressure and its units.


                   -           B-ll

-------
 If the method  used  is  a  tracer method,  clearly  document the process of
.adding the1 tracer to the tank and  in  the  spiking  port.

 If a tracer  is added to  the product in  the  tank,  provide  information on
 the following  items:
 * type of  tracer(s)              '                                  ,
 * tracer concentration in  the product            '•.;..
 * type of  carri er                                         :
 * time between spiking and starting the test
 * .type of  sampling, e.g.,  whether  sampling  is active  or passive  (in
  other words, how  does  the tracer reach  the sampling ports?   by
  natural  diffusion process?  is the  process enhanced by  adding  forced
  air? etc.)
 * other relevant items

 When sampling  ports are  installed  for tracer methods, measure ,the
 distances  between any  part of the  tank  to its nearest sampling port.
 Report the largest  of  these distances.
                              B-.12

-------
                                 Description
       Nonvoiumetric Tank Tightness Testing Method
 This section describes briefly the important aspects of the nonvolumetric tank tightness testine
 method. It is not intended to provide a thorough description of the principles behind the
 Ttifitnnn nr nnw tnp Am-iin-me-nt m/-«rlre                          .
method or how the equipment works.
 Method Name and Version
 Product
- > Product type
 For what products can this method be used? (check all applicable)
    ED gasoline     /
    O diesel
    D aviation fuel
    D fuel oil #4
    [H fuel oil #6           ,
    IZ1 solvents
    ^	                   *        '  ' ;  -4     -   -
    EH waste oil
    D other (list)     ''  .  • •    	'   ' ,
 >  Product level
What product level is required to conduct a test?
    EH above grade
    [H within the fill pipe                         •  .
    D greater than 90% full .
 .   CH greater than 50% full
    D empty  .
    CH other (specify)	'______^_
Nonvolumetric TTT Method - Description                                    Page 1 of 6'

-------
Principle of Operation
What principle or principles are used to identify a leak?
    D .acoustical signal characteristic of a leak
    CU identification of a tracer chemical outside the tank system
    D changes in product level or volume
    CD detection of water inflow
    HH other (describe briefly)    '	
Temperature Measurement
If product temperature is measured during a test, how many temperature sensors
 are used?                                        •     .                  .
    CU single sensor, without circulation
    EH single sensor, with circulation
    D 2-4 sensors                                            '                   •
    D 5 or more sensors                                       .       .    '   '    .
    D temperature-averaging probe
If product temperature is measured during a test, what type of temperature sensor is used?
    EU resistance temperature detector (RTD)
    C3 bimetallic strip   ,
    D quartz crystal                                                       .    ,  •
    D thermistor   .                                 .                      .'.".-
    D other (describe briefly) _ _ _____ _ •   •         -•  • •
If product temperature is not measured during a test, why not?               ,
    D the factor measured for change in level or volume is independent of temperature
       (e.g., mass)  .                                                         ;
    D the factor measured for change in level or volume self-compensates for changes in
       temperature            ,                                    .          .
    CD other (explain briefly) _ _. _ _ __ .  ' •  - _
Data Acquisition
How are the test data acquired and recorded?
    D manually  ,
    CD by strip chart
    D by computer        •
Nonvolumetric TTT Method - Description                                ,       Page 2 of 6

-------
  Procedure Information
  > Waiting times
  What is the minimum waiting period between adding a large volume of product to bririq the
  level to testrequirements and the beginninQ of the test (e.g., from 50% to 95% capacity)?
     D not applicable
     D no waiting period
     D less than 3 hours
     D 3-6 hours     ,                                              ,
     D 7-12 hours          /   ,                              .
  -   D more than 12 hours
     D variable, depending on tank size, amount added, operator discretion, etc,
  > Test duration
 What is the minimum time for collecting data?
    D less than 1 hour.                                              .         .
     D 1 hour              •  ',      ,          .                 •    •  •
    -D 2 hours           _.
 -  D 3 hours                                               .  •             :
    Q 4 hours                                   .   -                .
    lU 5-10 hours                                                      ._-'•
    D more than 10 hours
    D variable
 >  Total time                                             -
 What is the total time needed to test with this method?
 (setup time plus waiting time plus testing time plus time to return tank to service)
 	hours	      minutes       :
 >  Other important elements of the procedure or method
 List here any other elements that could affect the performance of the procedure or method
 (e.g., positive or negative ullage pressure, tracer concentration,  distance between tank and
 sampling ports,  etc.)                .
Nonvolumetric TTT Method - Description           '  ,      ,             ,      Page 3 Of 6

-------
> Identifying and correcting for interfering factors
How does the method determine the presence and level of the ground water above the
bottom of the tank?           .                      ..
   D observation well near tank.                      '       .,
   . D information from USGS, etc.
   D information from personnel on-site
   C3 presence of water in the tank
   D other (describe briefly)	[	       .	•_	           ,
  •  D Jevel of .ground water above bottom of the tank not determined
How does the method correct for the interference due to the presence of ground water
above the bottom of the tank?
    D head pressure increased by raising the level of the product       ,
    D different head pressures tested and leak rates compared
    D tests for changes in water levej in tank
    D other (describe briefly)   •	.          	' ' '  .•  	_j	•  • "'  •
    D no action                           .
Does the method measure inflow of water as well as loss, of product (gallon per hour)?
    Dyes                                 .
    Dno  _               ,              .                        I
Does the method detect the presence of water in the bottom of the tank?
    Cl yes    .                                     .             ,
'    Dnq .
How does the method identify the presence of vapor pockets?
    CD erratic temperature, level, or temperature-compensated volume readings
    D sudden large changes in readings
    IH statistical analysis of variability of readings
    [H other (describe briefly)  •  	•	. '  ':	
    D not identified
    D not applicable; underfilled test method used
Nonvolumetric TTT Method - Description                                '      Page 4 of 6

-------
  How does the method correct for the presence of vapor pockets?
     D bleed off vapor and start .test over
     D identify periods of pocket movement and discount data from analysis      -:
     D other (describe briefly)	..  ,   ..    .  .   ;  •  •--.     •
   ..Q not corrected
     D not applicable; underfilled test method used
  How does the test method determine when tank deformation has stopped followina
  delivery of product?  .                               .           •            a
   .                        .  •       "                  :        .          i ,
     D wait a specified period of time before beginning test
     Q watch the data trends and begin test when decrease in product level has stopped
     D other (describe briefly)   .             .
     D no procedure                   :
     LJ not applicable, does not affect principle of operation
 Are the method's sensors calibrated before each test?                            -
     D yes                                                      .  -            .
-."  D'no  -
                       5  • , . .      ...        -.'     '       * .         '
 If not, how often are the sensors calibrated?           .        .       '
  .   D weekly •           .                •          >
     CH monthly                    .  ' - ;
     EH yearly or less frequently
  \;  D never
 > Interpreting  test results
 What effect is used to declare the tank to be7leaking? (List all modes used  by the.method.)
 If a change in volume is used to detect leaks, what threshold value for product volume
 change (gallon per hour) is used to declare that a tank is leaking?
    D 0.05 gallon per hour                                                 •
    C] 0.10." gallon per hour
    D 0:20 gallon per hour                           .              ,   .
  -  D other    '•'•'.'.-'   '                ;  ..   '   -  .  '       <_

Nonvolumetric TTT Method - Description                                      Page 5 of 6

-------
Under what conditions are test results considered inconclusive?
    CD ground:water level above bottom of tank
    EH presence of vapor pockets
    D too much variability in the data (standard deviation beyond a given value)
    D unexplained product volume increase
    D other (describe briefly)   '        •  ' .•'    '    '	j	
Exceptions
Are there any conditions under which a test should not be conducted?
    D ground-water level above bottom of tank
    D presence of vapor pockets
    D large difference between ground temperature and delivered product temperature
    D extremely high or low ambient temperature
    D invalid for some products (specify)	-'••	•  •     	
    D soil not sufficiently porous                                  ......
    D other (describe briefly)   '      	-   •                	•  •
What are acceptable deviations from the standard testing protocol?
    HU none •
    D lengthen the duration of test
    ID other (describe briefly)	•  •        	'
What elements of the test procedure are left to the discretion of the testing personnel
on-site?
    D watting period between filling tank and beginning test              •
    EH length of test
    D determination of presence of vapor pockets
    D determination that tank deformation has subsided
    D determination of "outlier" data that may be discarded
    CD other (describe briefly)	•'	.- :    •
       none
Nonvolumetric TTT Method - Description
Page 6 of 6

-------
                   Reporting Form for Leak Test Results
                Nonvolumetric Tank Tightness Testing Method

                   Instructions for completing the form

This  3-page  form  is  to  be  filled  out by the evaluating organization  upon
completion of  the evaluation  of the method  in each  of its  leak  detection
modes.  This form provides for 60  test  results, although the minimum
number  of tests required in the protocol  is  42.  Use as many pages as
necessary to summarize  all  of  the  tests attempted.   Report the  results
for each leak  detection mode on separate  forms.

Indicate the commercial name and the version of the  method and  the period
of evaluation  above  the table.  The version  is provided for methods  that
might use different  versions of the equipment for different products or
tank  sizes.  Also, indicate the leak detection mode  for which these
results were obtained.

In general,  the statistician analyzing  the data will complete this
form.   A blank form  can be  developed on a personal computer,, the data
base  for a given  evaluation generated,  and the two merged  on the com-
puter.  The  form  can also  be filled out manually. The input for that form
will  consist of the  field  test results  recorded by the evaluating
organization's field crew on the Individual Test Logs and  the Vendor's
test  results.

The table consists of 10 columns.  One  line is provided for each test
performed during  evaluation of the method.  If a test was  invalid or was
aborted, the test should be listed with the appropriate notation (e.g
invalid) on  the line.                          •

The Test Number in the first column refers to the test number from the
randomization design determined according to the.instructions in Sec-
tion 6.2 of the protocol.   Since some changes to the design might occur
during the course of the field testing, the test numbers  might  not always
be in sequential order.

Note that the results from  the trial run need to be reported here as
well.

The following list matches the column input required with its source, for
each column in the table.
                               B-18

-------

-------
                   Reporting Form for Leak Test Results
                Nonvolumetric Tank Tightness  Testing Method

                   Instructions for completing the form

This 3-page form  is to be filled out by the  evaluating organization  upon
completion of the evaluation of the method in each  of its  leak detection
modes.  This form provides for 60 test results, although the minimum
number of tests required in the protocol  is  42.  Use as many pages as
necessary to summarize all of the tests attempted.   Report the results .
for each leak detection mode on separate  forms.

Indicate the commercial name,and the version of the  method and the period
of evaluation above the table.  The version  is provided for methods  that
might use different versions of the equipment for different products or
tank sizes.  Also, indicate the leak detection mode  for which these
results were obtained. .

In general, the statistician analyzing the data will complete .this
form.  A blank form can be developed on a personal computer, the data  -
base for a given  evaluation generated, and the two merged  on the com-
puter.  The .form  can also be filled out manually. The input for that form
wfll consist of the field test results recorded by the evaluating
organization's field crew on the Individual  Test Logs and  the vendor's
test results.

The table consists of 10 columns.  One line-is provided for each test
performed during  evaluation of the method,.   If a test was  invalid or was
aborted, the test should be listed with the  appropriate notation (e.g.,
invalid) on the line.

The Test Number in the first column refers to the test number from the
randomization design determined according, to the instructions in Sec-
tion 6.2 of the protocol.  Since some changes to the design might occur
during the course of the field testing, the test numbers might not always
be in sequential order.

Note that the results from the trial run need to be reported here as
well.        .-                               •

The following list matches the column input required with  its source, for
each column in the table.
                               B-19

-------
Column No.     Input                          ,

   1       •    Test number or trial run.
   2           Date at completion of last fill
                 (if applicable)
   3           Time at completion of last fill
                 (if applicable)
   4           Date test began
   5           Time test began
   ,6           Time test ended
   7           Product temperature differential
  _^             (if applicable)
   8   ,        Nominal leak rate
   9           Induced leak rate
  10           Leak test result
                                                    Source !             ,

                                                    Randomization design
                                                    Individual Test Log

                                                    Individual Test Log

                                                    Individual Test Log
                                                    Individual Test Log
                                                    Individual Test Log
                                                    Individual Test Log

                                                    Randomization design
                                                    Individual Test Log
                                                    Vendor's test result
Note:  the product temperature differential (column 7) is the difference
between the temperature of the product added and that of the product in
the tank each time the tank is filled.  This temperature 'differential .is
the actual differential achieved in the field and not the nominal
temperature differential.
                               B-20

-------
                                               m
                                              n(^>r
         Reporting?%rm for Leak Test Results
 Nonvolumetric Tank Tightness Testing Method
 Method Name and Version:
                                                                  Leak Detection Mode:
 Evaluation Period: from
to
.(Dates)

Test No.
Trial Run
...'../......',. 	 .'..
1
2
3
4
5
6
7
8
I 9 I
10
11
12
13
14
15
16
17
18
19
20
If applicable
Date at
Completion
of Last Fill
(m/d/y)

•\














-


-


If applicable
Time at
Completion
ofLastRii
(military)
i
( , .
Date Test
Began
(m/d/y)

Time Test
Began
(military)

Time Test
Ended
(military)

If applicable
Product
Temperature
Differential
(deg F)
0

Nominal
Leak Rate
(gal/h)
0
Induced
Leak Rate
(gal/h)
0
11 < , '* MS''' , : •• s s ^










- ,

















.






















'































































-










-







- -





Tank Tight?
(Yes, No, or
Test Invalid)

<.''•'













.,






Nonvolumetric TTT-Data Reporting Form
                                                                                                   Page 1 of 3

-------
                                        ReportinSBrrm for Leak Test Results
                                Nonvolumetric Tank Tightness Testing Method
 Method Name and Version:
 Evaluation Period: from
                                    Leak Detection Mode:
to
. (Dates)

Test No.
21
22
23
24
25
. 26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
!f applicable
Date at
Completion
of Last Fill
(m/d/y)


















i
-
If applicable
Time at
Completion
of Last Fill
(military)












_








Date Test
Began
(m/d/y)










. .









Time Test
Began
(military)













.






Time Test
Ended
(military)




















If applicable
Product
Temperature
Differential
(deg F)




















. • - 	 - - ' - . - -
Nominal
Leak Rate
(gal/h)

















•


induced
Leak Rate
(gal/h)




















Tank Tight?
(Yes, No, or
Test Invalid)











•








Nonvolumetric TTT-Data Reporting Form
                                                                      Page 2 of 3

-------
                                        Reportirrtp^rm for Leak Test Results
                                 Nonvolumetric Tank Tightness Testing Method
 Method Name and Version:
                                                                  Leak Detection Mode:
Evaluation Period: from
                              .to.
. (Dates)

Test No.
•41
42
43
44
45
46
47-
48
49
. 50
51
52
53
54
55
56
57 .
58
59
60
If applicable
Date at
Completion
of Last Fill
(m/d/y)




















If applicable
Time at
Completion
of Last Fill
(military)





••















Date Test
Began
(m/d/y)






- -i













Time Test
Began
(military)


-

















Time Test
Ended
(military)'




















If applicable
Product
Temperature
Differential
(deg F)





-














• • ' • . ' . •
Nominal
Leak Rate
(gal/h)

•


















Induced
Leak Rate
(gal/h)









. i,










Tank Tight?
(Yes, No, or
Test Invalid)




'-•



..,-. -
,,
-









Nonvolumetric TTT-Data Reporting Form
                                                         Page 3 of 3

-------

-------
:*'
                            Individual  Test Log
                Nonvolumetric Tank Tightness Testing Method

                   Instructions for completing the form

 This  5-page  test  Tog  form  is to be filled  out  by  the field crew of the
 evaluating organization.   A  separate form  is to be  filled out for each
 individual test including  the trial run  (at least 43.)  The information
 on these  forms  is  to  be kept blind to  the  vendor  during the period of
 evaluation of. the  method.  Adaptations of  the  form may be made as needed
 to document  the evaluation data.

 The form  consists  of  nine  parts.   These  are:

  1.   Header  information
  2.   General background information                           -
  3.   Conditions before testing                                        :
  4.   Topping off records (if  applicable)
  5.   For  tracer methods only
  6.   Conditions at beginning  of test
  7.   Conditions at completion of Resting  .            .     .     .
  8.   Leak rate data                                                   .
  9.   Additional comments,  if needed
 10.  -Data sheet for leak simulation for tracer methods
 11.   Data sheet for induced  leak rate calibration

All items are to.be filled out and the appropriate boxes checked.   If a
question  is not applicable, then indicate so as "NA".  The following
provides  guidance  on the use of this form.

Header Information

The header information is to be repeated on all five pages,  if used.   If
a page is not,used, cross it out and initial it.   The field  operator from
the evaluating organization needs to print-and sign his/her  name and note
the date  of the test on top of each sheet.

The test  number is the number obtained from the randomization  design. It
is not the sequential running test number.  If a test needs  to be rerun,
indicate .the test  number of the test being rerun and indicate  that  on the
test  log  (e.g., Test No. 5 repeat).

General Background Information              .

Indicate the commercial name of the method.  Include a version identifi-
cation if the method uses different versions for different products or
tank sizes.   The vendor's recommended  stabilization  period (if appli-
cable) has to be obtained from the vendor prior to testing.  This is
important since it will impact on the  scheduling  of  the evaluation.   All
other items  in this section refer to the test tank and product.   Indicate
the ground-water level at the time of  the test.
                                            B-24

-------
*

-------
 Theoretically, this information would remain unchanged for the whole
 evaluation period.   However, weather conditions could change and affect
 the ground-water level.   Also, the evaluating organization could change
 the test tank. ,                  v  -    ..

 Conditions Before Testing

 Fill in all  the.blanks.   If the information is obtained by calculation
 (for.example the amount  of water,in the tank is obtained from the stick
 reading and  then converted to volume), this can be done after the test is
 completed. Indicate the  unit of all temperature measurements by checking
 the appropriate box.

 Note that the term  "conditioning"  refers to all activities undertaken by
 the evaluating field crew to prepare for a test.  As such, the term
 refers to emptying  or filling the  tank, heating or cooling product, and
 changing the leak rate.   In some .cases, all of the above is performed, in
 others, only one parameter might be changed.  For tracers, "conditioning"
 refers to preparation of the tank  for testing.  It includes the determin-
 ation of the time to wait between  spiking and testing.

 Topping Off  Records (if  applicable)

 If this step is performed, fill  in the'appropriate blanks.       •

 For Tracer Methods  Only

 Fill in the  appropriate  information.   Follow the instructions and
 complete the form on page 4.

 Conditions at Beginning  of Test

 The, evaluation organization's field crew will  have calibrated the leak
 simulation equipment prior to the  test.   All leak rate  calibration  data
 need to be documented using the  form on pages  4 or 5, as appropriate.
 Refer to previous calibration if this has  been done.  Adapt the form as
 necessary.

 Once the evaluating"organization's  field crew  is ready  with the induced
.leak rate simulation,  and the vendor starts the actual  testing, record
 the date and time.that the vendor's test data  collection starts.  Also,
 indicate the product temperature at that time.  .Fill out the weather
 condition section of the form.   Indicate the nominal leak rate which is
 obtained from the randomization  design.

 Conditions at Completion of Testing

 Indicate date and time when the  test  is  completed.

 Again,  stick the  tank  and record the  readings  and the amount of water  in
 the tank.  Record all  weather conditions as requested.        '
                               B-25

-------
 Leak Rate Data.

 This section is  to be filled out by the evaluating organization's       ^
 statistician or  analyst performing the .calculations.   This section can
 therefore be filled out as the evaluation proceeds or at the end of the
'evaluation.                                                 ,

 The nominal  leak rate is obtained from page 2 (Conditions at Beginning of
 Test).   It should be checked against the nominal  leak rate in the
 randomization design by matching test numbers. ..

 The induced  leak rate is obtained from the simulation data reported by
 the evaluating field crew on page 4 or 5 of this  form.

 The test result  is that obtained by the vendor for that test.

 Give the mode being investigated on the line following the test  answer if
 the method uses  more than one mode of leak detection.

 Additional Comments (if needed)

 Use this page for any comments (e.g., adverse weather conditions,
 equipment failure, reason for invalid test, etc.)  pertaining  to  that
 test.

 Leak Simulation  Form for Tracer Methods (page 4)

 For tracer methods, use the form on page 4 to document and measure
 delivery of the  carrier with the appropriate concentration of the tracer
 to the  spiking ports.  Indicate the tracer used and the concentration  of
 tracer  in the carrier in the appropriate spaces.   Report the  distances
 between spiking  port and all sampling ports. Record  the time and amount
 of material  released in the spiking port to document  the leak simulation
 for. tracer methods.  Use as many pages as needed.

 Induced Leak Rate Calibration Form (page 5)

 For acoustical methods, the form on page 5 may be used to calibrate the
 liquid  flow through the simulator under a standard set of conditions.
 The induced  leak rate is the rate at which the liquid will flow  at a
 specified head or depth of product.  This rate is determined  by
 calibration and  used as the leak rate for detection.   The calibration
 will have to be  done at a different time, preferably  before)  than the
 testing.  A calibration is needed for each distinct leak rate.  Once the
 calibrations have been done, document on each daily_test log the simula-
 tion conditions'and reference the appropriate calibration data sheets,
 which should be  attached to the daily test log that first uses the given
 induced leak rate.
                                B-26

-------
 Name of Field Operator	     . ;.	          •  ,'.	
 Signature of Field Operator    	       .       Test No.
 Date of Test                          .     ;1         .
                            individual Test Log
     Nonvolumetric Tank Tightness Testing Method
 Instructions:                                                       .
 Use one log for each test.
 Fill in the blanks and check the boxes, as appropriate.
 Keep test log even if test is inconclusive.

 General Background Information      "~~    """""""~"~   -•  '         "
 Method Name and Version               	'••.'•	- ,
 Product Type    •'  '         	-.'•.•.•'    -'            ,   '
 Type of Tank    •  •   -    ,  .	-   -   .  •
Tank Dimensions (nominal)                            ..•/..
   Diameter          inches
   Length     ^_	__inches
   Volume    	gallons                            .,. .
Ground-water feyel         inches above bottom of tank
Recommended stabilization period .before test (per vendor SOP)                 . -  -
   	hours        minutes

Conditions Before Testing
Date          and military time          at start of conditioning test tank
Stick reading before partial emptying of tank
   Product        inches         gallons                       s
                              , •        _ /     ,               -           -  -
   Water          inches  	gallons                    .
Temperature of product in tank before partial emptying           °F D or °C D .
Stick reading after partial emptying of tank      '      •
   Product	inches	gallons
Amount of product removed from tank (by subtraction)   	^gallons
Stick reading after filling tank to test level          _
   Product        inches         gallons
   Water	inches         gallons                              '
Amount of product added to fill tank (by subtraction)         gallons

Nonvolumetric'TTT Method - Test Log                                      Page 1 of 5

-------
Name of Field Operator	.     	r___	"•'
Signature of Field Operator	   '                  Test No._	'   .  '   •'
DateofTest__.	

Conditions Before Testing (continued)
Temperature of product added to fill tank   	°F D  qr °C D
Temperature of product in tank immediately after filling        °F EH or °G CD
Date           and military time           at completion of fill

Topping Off Records (if applicable)
Date .          and military time__	at completion of topping off
Approximate amount of product added      " .   :    gallons
If tank overfilled, height of product above tank       inches

For Tracer Methods Only
Date	"     and military time           tracer(s) is added to product in test tank _
Tracer used	;	                                   ,
Amount of tracer used	
Amount of product in test tank          . gallons
                           -iui—r-nr-        ,                            ^      ,
> Complete the Tracer Leak Simulation form (use page 4)
Date	and military time    	at start of test                      -
Date	and military time	at conclusion of test

Conditions at Beginning of Test
Date	1 and military time           vendor began setting up test equipment
> Document induced leak rate determination (use page 5)
Date	and military time           at start of vendor's test data collection
Temperature of product in tank at start of test          °FD or °clZl  .
Weather Conditions
   Temperature	°FD or°cD         '     -'    '-...'
   Barometric pressure	jnm Hg D or	in.  Hg d
   Wind            NoneD     Light D    Moderate D     Strong Q
   Precipitation       None CD     Light HH    Moderate D     Heavy D
   Sunny D         Partly Cloudy D         Cloudy D                •  • ...
Nominal leak rate        gallon per hour
Nonvolumetric TTT Method - Test Log   .   .        •                         Page 2 of 5

-------
 Narne-of Field Operator            .    	•   •   '-_..'
 Signature of Field Operator	        jest No.
 Conditions at Completion of Testing
 Pate          and military time_	at completion of test data collection
 Stick reading at completion of test data collection
    Product          inches          gallons                         .
    Water            inches          gallons
 Date of Test_v	

 Conditions at Completion of Testing (continued)           .     ""~"~~
 Weather Conditions                                                  [  •  '  '
    Temperature	°F CD or °C D         .           .
    Barometric pressure         mm HgCD or         in. HgD ,
    Wind             NoneD  ,   LightD     Moderate CD     Strong CD
    Precipitation       None CD     Light EJ     Moderate D     Heavy D
    Sunny D          Partly Cloudy D     '     Cloudy CD
 Date_	 and military time	   •  test equipment is disassembled (if done
 for this test) and tank is ready for service                                   .
 Leak Rate  Data       "~   "  ""      "     :       ~            ~~
 Leak detection mode__	     ••''•-./    '     •  •    •;••.'
 Nominal leak rate	     • gal/h
 Induced leak rate	gal/h                .
       "*"'        •      '            '-                                     * '

 Findings for Tracer Methods
                          r        .
   CD No tracer found  DTracer(s) found    •
 If tracer(s) found, list	' •
Test answer          CD leaking      CD tight     CD inconclusive

Additional Comments (Use back of page if needed")"
NonvolumetricTTT Method-Test Log              -                         Page 3 of 5

-------
Name of Field Operator    -
Signature of Field Operatpr.

Date of test	
                           Test Nci.
              Leak Simulation Form for Tracer Method
                            (Reproduce form if needed)
Tracer used.
Carrier	
Concentration of tracer in carrier.
Distance from spiking port to:
 Sampling port 1	
 Sampling port 2	
 Sampling port 3      ,
 Sampling port 4	
Sampling port 5.
Sampling port 6.
Sampling port 7.
Sampling port 8,.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Time
(military)










(



. *





Carrier amount
released in
spiking port





'














Comments



' .•













, -' , . . " ~
,'•/*'

    Indicate all measurement units!
Nonvolumetric TTT Method-Test Log
                                                    Page 4 of 5

-------
 Name of Field Operator	
 Signature of Field Operator.

 Date of test      •  -
Test No.
               induced Leak Rate Calibration Form
                         (Reproduce form if needed)

1
2
3
4
5
6
7
8
9
10
11
12
13
14.
15
16
17
18
19
20
21
22
23
24
Time
(military)











'












Amount *


-












.-








Comments , :

' • - ' -
- •

' •' . ^
t - •
- •_ 	 '

" - . '. - . •
, . - • •.'."••


. ^
• " ' ' , ' - '

- • • '••..'




-. '.•..,•'



    * Indicate all measurement units!
Nonvolumetric TTT Method-Test Log
                    Page 5 of 5

-------

-------
             Reporting Form for Water Sensor Evaluation Data
               Nonvolumetrtc Tank Tightness Testing Method

This 4-page form is to be filled out by..the field crew of the evaluating
organization when evaluating the performance of the method's water
sensor, if applicable.  A separate form is to be filled out for each
individual test replicate (at least 20).  The form provides a template to
record the data and consists of three parts.  These are:

1.   Header information
2.   Template for recording the data obtained to determine the minimum
     water level that the sensor can detect in each-replicate (page 1)
3.   Template for recording the data obtained when determining the
     minimum water level change that the sensor can detect in each
     replicate (pages 2-4).

Header Information

The header information is to be repeated on all four pages, if used. ' If
a page is not used, cross it out and initial it.

Indicate the commercial name of the method.  Include a version identifi-
cation if the method uses different versions for different products or
tank-sizes.  Complete the date of test'and product type information.
Indicate the test (replicate) number on each sheet for each test. .

The field operator from the evaluating organization needs to print and
sign his/her,name and note the date of the test on top of each sheet.

Minimum Detectable Water Level Data

Follow the test protocol described in .Section 6.5 arid record all data on
page 1 of the form.  When the sensor first detects the water, stop test-
ing for this replicate.  The minimum detected water level is calculated .
from the total amount of water added until the first sensor response and
the geometry of the probe and the cylinder.  This calculation can be done
after all testing is completed and is generally performed by the statis-
tician or other person responsible for data analysis.

Minimum Detectable Water Level Change

After the first sensor response, continue with the test protocol as
described in Section 6.5.  Record all amounts of water added and the
sensor readings at each increment using pages 2 to 4 as necessary.  The
data to be entered in the third, fifth,, and sixth columns on pages, 2, 3,
and 4 of the form will be calculated once all testing is completed.
Again, the person responsible for the data analysis will generally~~
compute these data and enter the calculated minimum water level  detected
in that replicate run.
                               B-32

-------

-------
               Reporting Form for Water Sensor Evaluation Data
            Nonvolumetric Tank Tightness Testing Method
 Method Name and Version:
 Date of Test:
 Product Type:
                                  Name of Field Operator: _

                               Signature of Field Operator:
Increment
No.
1
2
3
4
5
6
7
8
' 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Total
Volume
(mL)
Volume of
Water Added
(mL)








.

















Sensor
Reading
(inch)












, •

•>










-
                                                                     Test No.
                                                          Calculated Minimum
                                                       Detectable Water Level (inches)
 NOTE:
This form provides a template for data reporting. Since the number of
increments is not known from the start, the length of the report form
will vary from test to test.
Nonvolumetric TTT-Water Sensor
                                                                               Page 1 of 4

-------
              Reporting Form for Wafer Sensor Evaluation Data
           Nonvolumetric Tank Tightness Testing  Method
Method Name and Version:

Date of Test:	

Product Type: 	•
  Name of Reid Operator: _

Signature of Field Operator:.
                                                                   Test No.

Increment
No.
A
Volume of
Water Added
(mL)
B
Calculated
Water Height
Increment, h
(in)
C
Sensor
Reading
(in)
- D
Measured
Sensor
Increment
(in)
E
Increment
Difference
Caic.-Meas.
(in)
C-E
Minimum water level detected, X: inches (from page 1)
1
• 2
3
4
5
6
7
8
9
10
11
'12
13
14
15
16
17
18.
19
20
21
22
23
24
25 .

































































•.






























,

i '"






' I
' m
4
1







f








NOTE:    This form provides a template for data reporting.
         Use as many pages as necessary.
Nonvolumetric TTT-Water Sensor
                                      Page 2 of 4

-------
               Reporting Form for Water Sensor Evaluation Data
            Nonvolumetric Tank Tightness Testing Method
 Method Name and Version:
 Date of Test:     '	
 Product Type:    --
  Name of Reid Operator:.
Signature of Field Operator:
                                                                    Test No.

Increment
No.
A
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
Volume of
Water Added
(mL)
B

	 , 	 P —







.







;







Calculated
Water Height
increment, h
(in)
C
















•








Sensor .
Reading
On)
D









',. ,















Measured
Sensor
Increment
(in)
E




















*
' . ' ,



Increment
Difference
Caic.-Meas.
(in)
C-E
























/
NOTE:    This form provides a template for data reporting.
         Use as many pages as necessary.,
Nonvolumetric TTT-A/Vater Sensor
                                                                             Page 3 of 4

-------
              Reporting Form for Water Sensor Evaluation Data
           Nonvolumetric Tank Tightness Testing Method
 Method Name and Version:
 Date of Test: 	
 Product Type: 	;
  Name of Field Operator: •„,
Signature of Field Operator:.
                                                                    Test No.

Increment
No.
A
51
52
53
54 ,
• 55
56
57
58 .
59
60
61
62
63-
64
65
66
67
68
69
70
71
72
73
74
75
Volume of
Water Added
(mL)
B

























Calculated
Water Height
Increment, h
(in)
C












.












Sensor
Reading
(in)
D








. " > • !











• • . ,-.




Measured
Sensor
. Increment
(in)
E
' ;
























Increment
.Difference
Calc.-Meas.
(in)
C-E



>
'



it
H
: • • •








_• .





NOTE:    This form provides a template for data reporting.
         Use as many pages as necessary.
Nonvolumetric TTT-Water Sensor
                                      Page 4 of 4

-------

-------

-------