DRAFT REPORT
       Submitted by:

 Pope-Reid Associates, Inc.
245 E. 6th Street, Suite 813
     St. Paul, MN 55101

-------
                APPENDIX A
PROBABILITY AND RELEASE VOLUME CALCULATIONS

-------
                               A.O   INTRODUCTION
This appendix consists of two parts:  the derivation of formulas used throughout
the failure analysis, and the documentation of the failure probabilities and
leak rates used for the Individual failure events.  The general derivations are
presented on pages A-2 through A-ll.  The individual failure events are
discussed in the remainder of this Appendix.
                                    A-l

-------
                            A.I  GENERAL DERIVATIONS

The following computations were common to many of the failure events:
    •  Calculation of release rates from underground leaks (fluid bed model);
    •  Calculation of release rates into an air environment (Bernoulli flow);
    t  Conversion from hydraulic head to pressure;
    •  Computation of fluid velocity in pipes;
    •  Determination of pressure of flowing liquid in pipes; and
    •  Calculation of fill/discharge times.
These calculations are described in the following pages.
                                     A-2

-------
A. 1.1.  Underground  Releases  (for leaks that are impeded by soil or backfill)
EQUATIONS:
     •  We calculated underground leak rates using the Ergun equation for the
        pressure drop In beds of mixed particles (Source:  Fluldization
        Engineering, Dalzo Kun11 and Octave Levensplel, p. /u U969)).
        This equation 1s:

       where

          P   «  the pressure drop  across  the  bed

         Em   *  the void  fraction  for the particles

         B    »  viscosity

         u0   *  velocity  of  flow

         4,5    »  sphericity of the  particles

      ..  p    «  density of fluid
        tf
         dp   •  average particle size

         9c   «  1 In metric  units

          L   *  the dispersion length


•  This Is a quadratic equation 1n u0.   Thus,

        „  .   -B + (82 . 4AC)-5
                   2A

         where A-  A-75(1-6.) \   / P
                       - /1.
                         \
                     B .  150(1^)2

                          Em3 ( s
                                   A-3

-------
•  The leak rate is simply
        dQ/dt » u0A
•  L for fluid beds is approximately 4 to 5 bed diameters.  (Source:
   R. H. Perry and C. H. Chi 1 ton, Chemical Engineers' Handbook. 5th
   Ed (1973), pp. 5-49).  The underground leaks studied in this model,
   however. Involve conical dispersion patterns, rather than the cylindri-
   cal fluid beds considered by Perry and Chi 1 ton.  In addition, L will
   also vary-with the dispersion characteristics of the surrounding back-
   fill.  After extensive analysis and sample calculations, we have chosen
   the following dispersion lengths:

                                Backfill Material
   Circular holes
   (usual value)
   (maximum permissible)

   Cracks
                          Clay          Silt          Sand
 2d   ,        7.5d           20d
 2 cm         7.5 cm         20 cm
                           Gravel

                            lOOd "
                            100 cm
   (usual value)           4 w
   (maximum permissible)   4 cm
             20 w
             20 cm
              40 w
              40 cm
               100 w
               100 cm
       where:
              d is the hole diameter, and
              w is the crack width
•  Additional parameters are as follows:
                                Backfill Material
   void fraction
   particle size (mm)
   sphericity
Clay
0.95
0.002
0.075
Silt
0.75
0.064
0.34
Sand
0.53
0.25
0.65
Gravel
 0.50
 9.4
 0.70
                                A-4

-------
USER INPUTS:
        Pressure
        Area of hole
        Density of liquid
        Viscosity of liquid
        Soil characteristics
        Type of hole (circular or crack)
        Diameter or width of hole
                                     A-5

-------
A. 1.2.  Bernoulli Flow (into an air environment)


EQUATIONS:


     t  This calculation applies to leaks that are unimpeded by soil
        resistance.

     •  The loss rate can be calculated by a simple application of
        Bernoulli's equation for a sharp-edged orifice.

     t  dQ/dt « .6A (2gz)-5

        where

              dQ/dt * leak rate

                  A » area of hole

                  g * acceleration due to gravity

                  z • height of liquid


USER INPUTS:
             »-

     0  Hydraulic head
     •  Area of hole
COMMENTS:
        The flow through small holes under relatively low pressure is turbulent,
        as can be verified by calculating a typical Reynold's number.  The
        Reynolds number for loss through a (1/32)" hole for methylene chloride
        under 2 feet of head is 7500.  This is in the upper end of the tran-
        sition zone between turbulent and laminar flow.  Use of Poiseville's
        Equation for laminar flow gives a considerably higher flow rate, so tur-
        bulent flow will be assumed.
                                     A-6

-------
A.1.3.  Conversion from hydraulic head to pressure

EQUATIONS:
     •  Some equations In this Appendix use aP (pressure drop).  Others call  for
        z (hydraulic head).  These two quantities are related by the following
        formula:
                 AP » P(g/gc)z
        where
               p « density of fluid
               g * acceleration due to gravity
              g. » a conversion factor which 1s 1 1n metric units and 32.17 1n
               c   English units

USER INPUTS:
     •  Hydraulic head or pressure drop
     •  Density of fluid
                                     A-7

-------
A.1.4.  Fluid velocity In pipes

EQUATIONS:
     •  For flowing fluid 1n a pipe, the following energy balance equation
        applies (Source:  M. S. Peters and K. 0.  Tiwnerhaus, Plant Design
        and Economics for Chemical Engineers. 3rd Ed (1980) pp.  509-515):
              P9c               'Sc    •
        where
              P!  « Inlet pressure
               p  • density
              g   • acceleration of gravity
              gc  » a conversion factor (1 In metric units,  32.17 1n English unit*
              Vj  * Inlet velocity
              z\  • Inlet elevation above arbitrary base
              ?2  " outlet pressure
              \/2  " outlet velocity
              Z2  * outlet elevation
              hs  » hydraulic head added  by the  pump
              hf  « frlctlonal losses  (measured  as  loss  of hydraulic head)
     •  F Is composed of 3 parts:   losses at entrance to pipe,  losses along
        the pipe, losses at pipe exit.
     •  The following frlctlonal loss  equations  apply to these  losses
        (Source:  Peters and Tlmmerhaus)
        sudden enlargement
             Fe - (^"V2}                                a - 1  (turbulent flow)
                                     A-8

-------
sudden contraction

          KCV22
     Fc * ?fr-
            9


pipes and pipe elbows


     F • 2fv2(le + I)
            gO
                                                        1
                                                        -05 for
                                                        diameter-changes
                                                L


                                                Le


                                                0
                                                        friction factor »
                                                        5 x 10'3 for
                                                        turbulent flow

                                                        length of straight
                                                        pipe

                                                        effective length of
                                                        elbows
                                                        diameter of pipe
•  For systems open to the atmosphere  Pj  * P£ • 1 atm

•  For large Inlet tanks,  Vj » 0

•  Set 21 • 0

•  Then

        V£^ " 2g (-Z2-hs-hf)

        V22 also appears In hf, so

        U:2..,,..hS>/l.l  .1   .  JO-«(L. * L) VI

                         \2g  2g   4g         gD       /
                                  gO
                                A-9

-------
A.1.5.  Pressure  of flowing liquid in pipes



EQUATIONS:
     •  Pressure  In a pipe 1s given by solving the energy balance  equation in
        Section A.1.4 for P£.  Hence:
 P2"P1 ''          *'2 ..... f
                                                   f)
        where V  * V£ » velocity of flow
     •  The pressure difference between the pipe Interior and the pipe
        exterior  1s given by A P " P2*pl«  Thus:
                 P - -/         * 22  * "
     •  At  the midpoint of the pipe,  the friction losses Include  only losses
        from the Inlet and half of the  effective pipe length.   Hence, at the
        midpoint of the pipe:
               hf .  Fc * | F


                  -   -05V2 + 5 x 10-3 y2(Le +
                      —  ~          -
     •  Thus, substituting this equation 1n the preceding equation gives
AP    £.VVi «. ,05V2 + 5 « 10-3 (L. * L)Y» +
                                                              + gh\
               .P --
                                    A-10

-------
A.1.6.  Fill/discharge time
EQUATIONS:
        The length of time necessary for tank filling OP discharge can be calcu-
        lated from the fluid velocity in the pipe, the pipe diameter, and the
        volume to be transferred:

                 T « Q/ irr2v
        where
                 Q • transfer volume
                 r « Inside radius of pipe
                 v * velocity of flow
USER INPUTS:
     t  Transfer volume
     •  Velocity of flow (from Section A.1.4)
     t  Pipe diameter (interior)
                                     A-ll

-------
                            A.2  FAILURE DATA SHEETS

The following data sheets discuss the probabilities and release volumes for
Individual fault tree events.

The events discussed In these data sheets generally represent Individual fault
tree events, but there 1s not always a one-to-one correspondence.  When it
facilitates discussion, we have sometimes combined similar events.  In addition,
we have occasionally divided our discussion of single fault-tree events in order
to focus more clearly on the differences between various tank designs or operat-
ing conditions.  These changes in the fault tree event classifications have been
clearly labeled.

For these reasons, it is not appropriate to read one data sheet without also
becoming familiar with the data sheets for similar events.  Important caveats may
apply to a whole series of events, but in order to focus on the differences among
related events, all similarities are not necessarily re-stated.  Liberal cross-
references have been supplied, but they are not a substitute for a careful
reading of the related data sheets.

The labels for the events described in these data sheets may be confusing to the
first-time reader.  Therefore, the following guide may be useful:

         Label Prefix                    Category of Event Depicted
          Txxxx                    Tank defects (corrosion, rupture, etc.)
          Blx                      Piping, flanges, and gaskets
          B2x - B6X                Secondary-containment devices
          Allllx                   Spills during discharge
          A21x                     Release routes for overflows (corroded .vent
                                   pipe, corroded flanges, open-topped tank,
                                   etc.)
          ANUCAT (I, x)            Catastrophic events (flood, fire, etc.)
          LIFDEF (I, x)            Lifetime defects (Improper Installation,
                                   damage during Installation, etc.)
                                     A-12

-------
                        xxxxxx                   Miscellaneous events (including overfill,
                                                 operator error, alarm failure, level gauge
                                                 failure, etc.)  These are listed in alpha-
                                                 betical order.

              It may also be useful to sunmarize a few Items of commonly-occurring notation:


                       Notation                         Explanation.


                        p                        Binomial probability.

                        N(x,y)                   Normal distribution with mean of x, standard
                                                 deviation of y.

                        pN(xfy)                  Conditional normal distribution,  p is the
                                                 binomial probability that failure occurs;  -
                                                 NXx.y) Is the distribution of times to failure
                                                 given that failure does occur.

                        FNU(x.y)                 A uniform distribution between x and y.

                        Maximum • y              Unless a minimum Is also specified, this
                          ;.                      means FNU(O.y).

                        B(a,b,c)                 Beta distribution with minimum value a,
                                                 mode b, and maximum c.

              These explanations should make the event sheets more accessible to the general
              reader.
                                                   A-13
1

-------
                             HAZARDOUS WASTE TANKS



LABEL:  T1121

FAILURE:  Localized exterior corrosion

SOURCES:  Petroleum Association for Conservation of the Canadian Environment,
          "Underground Tank Systems:  Review of State of the Art and
          Guidelines,"  PACE Report No. 82-3, Ottawa (1983).

CALCULATIONS:

     t  The PACE data are presented and analyzed in Appendix C.

     0  The results of that analysis are cumulative time-to-failure distribu-
        tions for each of 3 categories of Soil Aggressiveness Values  (SAV's).
        (SAV's are discussed in Appendix C).


                       Cumulative Probability of Failure


        Tank
Low SAV
(0-6)
0
0
6.3
24.0
48.3
69.9
Medium SAV
(7-12)
0
11.1
29.1
54.0
67. 3,,
76.62
High SAV
« 13)
0
26.7
49.9
76.5
79.9
83.3
         f'1 "'
         9   "
        14
        19
        24i
        301

     * This age has been chosen as representative of the >. 25 bracket.

     2 A probability for this category could not be calculated from the raw
       data.  In order to complete the table, this number was obtained by
       averaging 69.9 and 83.6.  It appears to be a reasonable value.

PROBABILITY DISTRIBUTION:  Empirical

PROBABILITY:  See above table.

DATE OF INITIAL RELEASE:

     •  SAV must first be calculated as shown in Appendix C.  This will require
        the following inputs:

              -  Soil resistivity (ohm-cm)
              •  Soil pH
              •  Soil moisture


                                     A-14

-------
              -  Presence or absence of sulfides
              -  Differential characteristics

     •  The failure date may now be sampled from the probability table.

     •  A corrosion rate may be obtained by dividing the resulting
        failure date by the average tank wall thickness for the tanks
        in the PACE survey.  This thickness is assumed to be .25
        Inches.

     •  The date of failure will be the date at which the sum of accu-
        mulated local external corrosion plus general Internal corro-
        sion reaches the initial wall thickness of the tank in
        question.  (General Internal corrosion is calculated under
        event T1125).  General exterior corrosion does not enter Into
     •  this analysis, for we assume that It 1s already Included in the
        corrosion rate for localized exterior corrosion.  In addition,
        localized Interior corrosion is not Included in this analysis,
        for it is unlikely that localized Internal and external corro-
        sion will occur at the same point on the tank wall.  Thus, it is
        unlikely that pits from these two causes will "meet 1n the
        middle."


MATERIALS AND CONFIGURATION VARIATIONS:

     •  The pACE-derlved probability distribution is assumed to apply only to
        unprotected steel tanks.

     •  Material changes, use of cathodic protection, and use of coatings will
        alter corrosion probabilities and hence dates.of failure.  These effects
        are summarized 1n the following table:

        Tank design                          Effect on probability of failure

        Carbon steel (underground,           PACE baseline
           unprotected)

        Carbon steel (underground,           Delays onset of corrosion by
           coated)                           N(7,3) years, then corrosion
                                             becomes a certainty.  Time-to-
                                             failure distribution 1s given
                                             below.

        Carbon steel (above-ground,          Corrodes like underground tank
           unprotected)                      with 5X as large a surface area,
                                             located 1n a low-SAV soil.

        Carbon steel (above-ground,          Delays onset of corrosion by
           coated)                           N(9,3).  Then it corrodes like a
                                             low-SAV, underground tank with a
                                             failed coating, and 5X as large a
                                             surface area.

                                     A-15

-------
  Carbon steel (underground,           PACE baseline after mN(10,5) years
     impressed current cathodic
     protection)

  Carbon steel (underground,           Delays onset of corrosion by
     Impressed current cathodic        max CmN(10,5), N(7,3)].  Then it
     protection, coated)               corrodes like a similar coated
                                       tank whose coating has failed.

  Stainless steel                      Corrosion rate of 25% of that
                                       applicable to steel.

  Fiberglass                           No corrosion

  Concrete                             Gradually disintegrates due
                                       to chemical attack, but this
                                       effect has been Included with
                                       ruptures.

• Sources and Explanations:

  Carbon steel (underground, unprotected);  PACE

  Carbon steel (underground, coated):  Corrosion will not begin until the
  coating has failed.  The coating 1s assumed to fail according to a.nor-
  mal distribution, N(7,3).  The 7-year mean was obtained from A.H.
  Roebuck and 6.H. Brevoort, "Coating Work Costs and Estimating,"
  Materials Performance, 22(1): 43-47 (Jan. 1983).  Standard deviations
  were estimated by doubling the reported variations between average
  coating lives 1n differing environments.  In addition, the standard
  deviation was Increased. 50X to account for variations 1n sub-grade
  drainage and the possibility of scratches during Installation.  The
  coating used was 2-coat coal-tar epoxy with an SP10 (near white blast)
  surface preparation.  The mean age to failure Is that reported for
  coatings In a freshwater environment.

  Once the coating has failed, there Is a near certainty of localized
  corrosion at the points of failure, for those points present excellent
  point anodes.  The time-to-failure distribution must therefore be
  adjusted to reflect the certainty of a point anode's existence.  We do
  this by dividing each probability in the PACE baseline distributions by
  the corresponding year-30 failure probabilities.  The following time-to-
  failure distributions therefore apply following coating failures:

   Tank                Low                 Medium             High
   Age                SAV(X)            •  SAV(X)            SAV(X)

     40                    00
     9                  0                   14.5              32.0
    14                 9.0                  38.0              59.9
    19                34.3                  70.9              91.8
    24                69.1                  87.9              95.9
    30               100.0                 100.0             100.0

                               A-16

-------
 Carbon steel (abovegrpund.  uncoated).   We assume .that above-ground tanks
 only experience localized corrosion at their saams and their points of
 contact with cradles.  These areas account for approximately 5% of their
 surfaces.  We assume that air 1s about as corrosive as a low-SAV soil.
 Thus, these tanks corrode like low-SAV, underground tanks of 5% as large
 a surface area.  The effect of area on corrosion rates is discussed
 below under "variations."

 Carbon steel (aboveground.  coated). The coating lifetime was based
 on Roebuck and Brevoort's values for exterior coatings In a moderate
 atmosphere.  The result Is  a coating lifetime at N(9,3).  Once the
 coating has failed, the tanks will have a 100% chance of developing
 point anodes (at the sites  of Initial  coating failure), and will
 corrode like underground, coated tanks of 5X the surface area, in a
 low-SAV environment.

 Carbon steel (underground.  Impressed current cathodic protection).
 We chose impressed current  as the preferred means of cathodic protection
 on the recommendation of the National  Association of Corrosion Engineers
 (NACE) (Houston, phone conversation).  This recommendation was
 given because such a system can be repaired without excavation.
 Furthermore, a crude .check  of Its functioning can be done merely by
 observing the ammeter.

 The components of such a system are crucial.  The wiring Is not subject
 to protection, so 1t Is failure-prone.  The rectifier Is also subject
 to failure, especially 1f much energy  1s dissipated.  In addition,
 catholic protection can fall if the local electrical environment
 changes (e.g. Interference  with another protected system across the
 street, or if someone buries some unprotected metal nearby).

 Impressed-current cathodic  protection  therefore requires an Inexpensive
 check of current distribution every 2-3 months (6 months maximum).
 This can be done by a contractor equipped with a hand-held meter, and
 takes only 2-3 minutes (plus travel).

 With regular maintenance, NACE says such a cathodic protection system
 should last for a long time.  No failure data appears to be available,
 however, and NACE says failures are common among poorly-maintained
 systems.  We therefore assume that cathodic protection systems fail with
 a distribution of (m)N(10,5), where m  1s a random number drawn from (1,
 3) Indicating the stochastic quality of the maintenance effort.  (1
 means no maintenance).  Source:  BEJ following phone conversations with
 NACE.

 As long as cathodic protection 1s functioning, corrosion will be negli-
 gible.
t
 If cathodic protection falls and 1s repaired after a substantial
 Interval, It may be too late to restore complete protection.  Cathodic
 protection does not function well for  creviced surfaces (or pits), for
 it Is difficult to get an adequate charge density.  (Source:  NACE/API,


                              A-17

-------
        Corrosion of Oil-and  Gas-Well Equipment  (1958), p.  71.)  Since belated
        repair will be  insufficient  to halt established localized corrosion,
        we have not modeled belated  repairs of the cathodic protection system.
        Ordinary maintenance, however, is already included  in our time-to-
        failure distriution.

        Carbon steel (underground, impressed current cathodic protection.
        coated).  If cathodic protection and a coating are both used, both the
        cathodic protection system and the coating must fail before corrosion
        will occur.  Once failure occurs, point  anodes will exist at the points
        of coating failure.

        Stainless steel.  Stainless  steel corrodes four times more slowly than
        unprotected carbon steel.  Source:  Peters & Tlmmerhaus, Plant Design
        and Economics for Chemical Engineers. 3rd. Ed., (1980), p. 574.

        Concrete.  The  disintegration of concrete leads to cracking, not to
        localized corrosion holes.   We will discuss the cracking of concrete
        tanks under ruptures, event  T1124, below.
VOLUME OF RELEASE:
        Tank corrosion represents a growing leak.  Eventually the leak rate will
        reach a detectable level.
ASSUMPTIONS:
     •  Corrosion holes start small.  Assume Initial hole size follows a beta
        distribution, with a minimum size of (1/64)", a maximum of  (1/4)", and a
        mode of  (1/32)".

     •  Corrosion holes grow with time.  We use the corrosion rate, r, calcu-
        lated earlier, as the base rate at which corrosion holes grow in radius.
        In addition, once exterior corrosion holes have perforated  the tank,
        they will also grow because of generalized Interior corrosion (see event
        T1125).  We therefore add this corrosion rate to r 1n order to.determine
        the total rate at which the hole grows In radius.

     §  Corrosion holes may occur anywhere on the tank.  Depending  on such fac-
        tors as  water table depth, they may be more likely to occur on the bot-
        tom of the tank.  We have decided to assume that the average corrosion
        hole occurs on the tank bottom.  This estimate could be refined to
        account  for fluctuations In fluid levels and the fact that  leak rates
        are not  linearly dependent on hydraulic head.  Such refinements greatly
        complicate the model, however, without significantly altering the leak
        rates.   Hence, we use simple "average" hydraulic heads throughout the
        model, rather than Integrating loss rates over the cyclical fluctuations
       • In fluid depths.
                                                                  r
     •  For storage tanks, we assume that the average fluid depth Is 50% of the
        tank height.  For treatment tanks we assume that the average fluid
        depth 1s SOX of the tank height.
                                       A-18

-------
     •  For underground tanks, the leak rate is calculated according to the
        underground leak rate formula.  (Section A.1.1).  For;.above-ground tanks
        or tanks in vaults, the Bernoulli  equation (Section A!1.2) applies.
VARIATIONS:
        Tank capacity.  Localized external corrosion Is an electrochemical pro-
        cess which will only occur 1f the tank surface experiences a non-
        uniformity which can function as a locale for electrochemical attack.
        Such non-uniformities are referred to as "point anodes" and may consist
        of scratches in the tank wall, variations In local soil conditions, or
        stones or cinders in contact with the tank wall.

        Logically the probability that point anodes will be present 1s a func-
        tion of tank surface area:  the larger the tank, the higher the probabi-
        lity that a point anode exists.  Similarly, the larger the tank surface
        area, the more likely It 1s that there Is an unusually active point
        anode somewhere on the tank surface.

        The easiest way to account for this factor 1s to adjust .the corrosion
        rate r to account for the surface area of the tank.  According to a pipe
        corrosion model by Rossum (see event 813 for a full discussion), the
        corrosion rate for the deepest pit is proportional to the .16 power of
        the surface area.  Therefore, variations in tank size can be accounted
        for by adjusting r according to the following formula:

                          r • r0 (A/A0)-16

        where

             r  » the adjusted corrosion rate
             r0 • the corrosion rate obtained from the PACE data
             A  » the surface area of the tank under consideration and
             A0'« the average area of the tanks Included 1n the PACE study

        These tanks were all service station tanks, many of which were installed
        prior to 1970; they probably averaged 5000 gallons in capacity.  The
        surface area of such tanks 1s approximately 440 square feet.

        This area adjustment factor applies to coated tanks and stainless steel
        tanks as well as to carbon steel tanks, for in all cases, an Increase in
        tank surface area Increases the probability that there Is an unusually
        active pit.

        Stray currents can accelerate corrosion.  Approximately 10X of tanks are
        subjected to stray currents due to nearby electrical equipment or
        .electrical rail lines.  (Source:  Warren Rogers, Warren Rogers Associates,
        personal communication.)  Stray currents approximately double external
        corrosion rates.  (Source:  Warren Rogers, Warren Rogers Associates, per-
        sonal communication.)  In order to allow greater variability 1n this
        effect, we have assumed that stray currents multiply the corrosion rate


                                     A-19

-------
        r by a  stochastic  factor of  x.  x  is distributed  according  to  a  beta
        distribution with  a minimum  of  1,  a mode of  2,  and  a"maximum of  4.

     •  If the  Teak is detected, the tank  will be repaired  or replaced,  and the
        aging process will start over again, using the  same parameters as those
        used for the original tank.

     •  Cathodic protection will have no effect on hole growth rates;  once the
        cathodic-protection system has  failed, the tank will corrode in  the same ma
        as an unprotected  tank.

     •  Coatings will also have no effect  on hole growth  rates.

     •  Double-walled tanks;  The 2  walls  of such tanks are modeled separately.
        Exterior corrosion attacks the  outside of the outside wall.  Interior
        corrosion attacks  the inside of the inside wall.  The interstitial space
        is not  subject to corrosive  attack.  The interstitial alarm is modeled
        as a leak detector similar to that used in a vault.  See event MOALARM,
        below.

     •  In-ground tanks or above-ground tanks on-grade.   Both the above- and
        below-ground sections of such tanks are subject to  corrosion.  Because
        exterior corrosion is more rapid below-ground than  above-ground,
        however* we have only modeled exterior corrosion  of the below-ground
        segment.  Similarly, since interior corrosion is  most important on the
        bottom of the tank, we have  only modeled interior corrosion of the tank
        bottom.
USER INPUTS:
        Tank surface area
        Corrosion protection methods employed
        Tank material
        Soil characteristics
        Rate of generalized interior corrosion (from event T1125)
        Tank wall thickness
        Tank location (above-, below-, or in-ground)
        Is the tank double-walled?
                                     A-20

-------
                              HAZARDOUS WASTE TANKS
 LABEL:   T1123
 FAILURE:  Localized interior corrosion
 SOURCES:
      •   API Tank and Pipe Leak Survey (referenced in SCS Engineers,  "Assessment
         of the Technical, Environmental,  and Safety Aspects of Storage of
         Hazardous Waste in Underground Tanks," Draft, 1983).
      •   PACE corrosion study, cited under event T1121.
      •   Best engineering judgment.
 ASSUMPTIONS:
      •   Interior corrosion leaks are the  same size and grow at the same rate as
         exterior corrosion holes.
»                                                             *         • •
      •   Volumes are larger, however, for  the hole usually occurs at  the bottom
         of the tank and therefore has more hydraulic head than that  assumed for
         exterior corrosion holes.
 CALCULATIONS:
      t   Using'-SCS Engineers'  analysis of  API survey data, we can construct the
         following tabulation of underground tank leaks:
         Source of Leak       t of Failures    % of Total    Scaled to 77%
         - TANKS
           localized exterior      988           61.6            77.0*
           corrosion
           776 + 212
           interior corrosion      194           12.1            15.Ot
           other                    55            3.4             4.3*
           loose fitting             9            0.6             0.81
           breakage                 17            1.1             1.4*
                                      A-21

-------
   - PIPES & ANCILLARY

     pipe corrosion

     loose pipe fitting

     flex connector

     breakage

     other


•  Explanation;
 353

  64

  38

  43

  54
TSUI
22.0

 4.0

 2.4

 2.7

 3.4
27.5%

 5. OX'

 3. OX

 3.4%

 4.3X
   - We obtained the last column by scaling the first entry to 77%.  We
     then multiplied the remaining entries in this column by the same
     scaling factor.  This scaling factor 1s based on Warren Rogers'
     assertion that 77% of all tanks exhibit significant localized
     exterior corrosion.  This percentage 1s also consistent with the PACE
     data, which show that 76.6% of medium-SAV tanks have leaked by year 30.
     Due to the possibility that a tank system may experience more than one*
     type of failure, the numbers 1n this column do not sum to 100%.

   • Line 1 Includes 212 tanks: that were not reported as falling by loca-
     lized exterior corrosion.  We included these tanks because we assumed
     that localized exterior corrosion 1s a slower process than the other
     four forms of tank failure.  Thus, we assume that 77% of the 275 tanks
     that failed by these other mechanisms were undergoing undiscovered
     localized exterior corrosion at the time they failed.  We make no
     similar adjustment to account for tanks that might have been
     undergoing localized exterior corrosion at the time of piping
     failures, because we assume that tank and piping systems are repaired
     independently.  Thus, the API survey data already account for systems
     which suffer both tank and piping failures.

•  We can read the following results off the above table:
     Failure Mechanism

     exterior tank corrosion

     Interior tank corrosion

     seam leaks (the most
       likely form of "other")

     tank rupture (breakage)
    Probability     Conditional Distribution
      0.77

      0.15

      0.043


      0.014
        See event T1121

        See below

        See event T1124


        See event T1124
     pipe corrosion
      0.28
        See event B13
                                A-22

-------
          pipe rupture (breakage)      0.034            See event 811

          loose fittings               0.05             See, event B121


PROBABILITY DISTRIBUTION:  Conditional normal


PROBABILITY:

     •  ,15c where c is a material-dependent coefficient given by the following
        table:

                Tank material                   c
                steel                           1
                stainless steel                 1
                fiberglass                      0
                concrete                        0

        Coefficient c merely Indicates that fiberglass does not corrode and that
       . Instead of treating the aging of concrete as corrosion, we will treat it
        as contributing to rupture under event T1124, below.


TIME-TO-FAILURE:

     •  We. estimate that localized Interior corrosion should occur more quickly
        than;-localized exterior corrosion.  Since the PACE data show that the
        conditional mean date of localized exterior corrosion failure for tanks
        in medium-SAV soils Is approximately 16 years, we assume that localized
        interior corrosion proceeds twice as fast.  Thus, for quarter-inch
        steel, we have assigned a localized interior corrosion time-to-failure
        distribution of N(8,5).  In choosing this distribution, we have used a
        relatively large standard deviation In order to cover unknown factors,
        Including variations 1n yearly flow rates.

     •  Let x be the failure date sampled from N(8,5).  Since the typical ser-
        vice station tank is .25" in thickness, the corrosion rate (r) for those
        tanks which experience localized Interior corrosion 1s given by:

                            r « .25
                                 x

        In deriving this equation, we have .assumed that localized interior
        corrosion was the primary corrosion mechanism for those tanks listed as
        falling by Interior corrosion.

     »  We can now compute the failure date by using the combination of loca-
        lized interior corrosion and generalized exterior corrosion to keep
        track of remaining wall-thickness.  Because generalized exterior corro-
        sion 1s usually much slower than localized Interior corrosion it usually


                                     A-23

-------
        has only a minor effect on the date of failure.  For this reason, we did
        not account for it in calculating our corrosion rate r, above.          i
        Generalized exterior corrosion is only important in those cases when it
        occurs unusually quickly.
VOLUME:
        Interior corrosion holes begin with the same size distribution as
        exterior corrosion holes, and grow 1n radius at the rate r calculated
        above.

        In addition, corrosion holes will also grow due to the effects of
        generalized exterior corrosion.  See event T1126, below.

        Interior corrosion leaks tend to occur at the bottom of the tank.
        For storage tanks, we assume that the tank 1s on average 50* full; for
        treatment tanks, we assume that 1t Is 80% full.

        For underground tanks, we calculated leak rates according to the
        underground leak rate equation (Section A.1.1).  For above-ground tanks,
        tanks In vaults, or the above-ground sections of In-ground or on-grade
        tanks, the Bernoulli equation (Section A.1.2) applies.
VARIATIONS:
     •  Coated, steel.  We assume that the coating 1s 3-coat epoxy.  Then the
        distribution of coating failure times Is N(7,3).  Source:  Roebuck and
        Brevoort (1983).  Corrosion will not begin until the coating fails.  We
        assume that the failure of Interior coatings does not Influence either
        the probability of corrosion or the corrosion rate for Interior corro-
        sion.  Thus tanks with failed Interior coatings experience Interior
        corrosion in the same manner as new, uncoated tanks.

     •  Cathodic protection.  Cathodlc protection has the same effects and the
        same time-to-failure distribution for Interior corrosion as 1t has for
        exterior corrosion.  We have assumed that when the exterior cathodic
        protection system falls, the Interior cathodic protection system also
        falls.

     •  Double-walled tanks.  See event T1121.

     •  Tank capacity.  Unlike localized external corrosion, localized Interior
        corrosion 1s not likely to be-correlated with tank capacity.  Localized
        Internal corrosion 1s most likely to occur beneath the fill tube, and   '
        the affected area Is likely to be Influenced by factors such as method
       .of filling or pumping rate, rather than tank size.

     •  Stainless steel tanks.  Stainless steel tanks corrode at 25X of the rate
        applicable to carbon steel.
                                     A-24

-------
USER INPUTS:
     •  Rate of generalized exterior corrosion
     •  Tank material
     •  Tank wall thickness
     •  Type of corrosion protection system
                                    A-25

-------
                             HAZARDOUS WASTE TANKS

LABEL:  T1124
FAILURE:  Tank Rupture
ASSUMPTIONS:
     •  Tank rupture can occur from Inadequate component strength, settling, or
        exterior force.
     •  Tank rupture Includes both large cracks and seam leaks.

SOURCES:  See T1123 (Breakage + "other")
CALCULATIONS:
     t  SCS/API data (see event T1123) Indicates a tank rupture rate of 1.4%.
     •  SCS/API data Indicates a seam leak rate of 4.3%.
     •  SCS/API data Indicates an average tank age of 10.8 years.
     •  Assuming ruptures are approximately uniformly distributed over the 10.8
        years of tank lifetime, the average annual probability of ruptures is
        5.7X/10.8 yrs - 5.3 x 10'3/yr.
PROBABILITY:  5.3 x 10'3/yr
PROBABILITY DISTRIBUTION:  Binomial
VOLUME:
     t  Ruptures may take 2 forms:  large cracks and seam leaks.  The data from
        event B13 Indicate that 75X of ruptures are seam leaks, while 25% are
        large cracks.
     •  Seam leaks are assumed to range 1n length from 0 to 5 ft, and 1n width
        from 0 to (1/16)".
     4  Large cracks are assumed to range 1n length from 3" to 36" and in width
        from 0 to 3".
                                     A-26

-------
VARIATIONS:
        Aboveground. in-qroundr or vaulted tanks.  Above-grade ruptures occur
        because of different types of accidents than occur for below-ground
        tanks.  Vehicle collisions, collisions with fork-lifts, and inadequate
        component strength are major causes of ruptures.  We assume that the
        probability of rupture is still 5.3 x 10~3.  Volume of loss will be
        limited by the amount of fluid in the tank.

        Fiberglass tanks.  Fiberglass tanks are approximately twice as likely to
        rupture as steel.  The probability of rupture is therefore 1 x 10'2/yr.
        Source:  Brown Minneapolis Tank Co., personal communication 4/30/85.
        Brown Tank Company said that 2% of fiberglass tanks collapse in the
        first couple of years.  We obtained our scaling factor of 2 by comparing
        that percentage to the .5X annual probability of steel tank rupture.

        Double-walled tanks.  We assume that since the Inner wall of a double-
        walled tank 1s subject to fewer stresses than 1s the wall of a single-
        walled tank, it 1s only SOX as likely to rupture.  The outer wall of
        such a tank, though, 1s likely to be just as vulnerable to rupture as
        the wall of a single-walled tank.  We assume, however, that there is
        only a 50% chance-that a rupture of the outer wall will also breach the-
        1nner wall.  Since ruptures of the Inner wall do not involve large
        exterior forces (such as vehicle collisions or settling of the
        backfill), we assume that the number of times that Interior ruptures
        also breach the outer wall 1s small enough to be Ignored.

        When a rupture of a double-walled tank occurs, Infiltrating soil
        moisture or leaking hazardous waste will trigger the Interstitial alarm.
        There 1s a 10X probability that the alarm will fall to function.  See
        MOALARM, below.
                                     A-27

-------
                             HAZARDOUS WASTE TANKS

LABEL:  T1125

FAILURE:  Thickness loss due to generalized interior corrosion.

ASSUMPTIONS:
     •  This event will cause a loss when remaining wall-thickness reaches zero.
     •  This event will also reduce the time it takes for localized exterior
        corrosion holes (if any) to penetrate the tank.

SOURCES:  Perry and Chi 1 ton (1973).

CALCULATIONS:
                 »                                            .
     •  According to Perry and Chi 1 ton, generalized interior corrosion rates of
        .002"/yr to .02Vyr are reasonable.  We expect that the lower corrosion
        rates are the most likely.
VOLUME:  See event T1123.
         •   tr
PROBABILITY DISTRIBUTION:  Uniform

PROBABILITY:
            Cumulative Probability           Corrosion Rate
                 0.00 to 0.65                 0.002"/yr
                 0.65 to 0.90                 FNU(.002,.OD
                 0.90 to 1.00                 FNU(.01,.02)

VARIATIONS:
     •  Coatings.  See event T1123.
     •  Cathodic protection.  See event T1123.
                                     A-28

-------
                             HAZARDOUS WASTE TANKS


LABEL:  T1127


FAILURE:  Thickness loss due to generalized exterior corrosion.


ASSUMPTIONS:

     •  This event will reduce the time It takes for interior corrosion holes
        (if any) to penetrate the tank, as well as Increasing the rate of
        Interior hole growth once a hole occurs.


SOURCES:

     t  "Prediction of Leaks 1n Unprotected Storage Tanks,"  Warren Rogers
        Associates, Inc., Included in correspondence package from Betsy Tarn,
        EPA, to Chris Lough, PRA. .   _                        •  .
                                     «
     •  W. H. A1lor, Atmospheric Corrosion (John Wiley and Sons:  New York)
        1982, p. 33.


CALCULATIONS:

     t  According to Ailor, generalized corrosion in an air environment occurs
        at approximately 1.4 m1ls/yr.

     •  According to Warren Rogers Associates, generalized corrosion 1s unlikely
        to cause failure (for a quarter-inch tank) within the 40-year time-
        horizon of their data.

     t  Generalized external corrosion in a soil environment 1s unlikely to
        occur more slowly than In air.  1.4 mils/yr 1s therefore a minimum value
        for the generalized external corrosion rate.

     a  5 mils per year Is a suitable upper bound for the generalized corrosion
        rate.  This choice correspondes to a 50-year minimum time-to-failure for
        quarter-Inch steel tanks in non-extreme environments.

     t  SAV will have some effect on generalized corrosion.  Assuming the effect
        Is roughly linear, and noting that according to the Canadian data the
        average SAV Is 10, we account for SAV by a multiplying the generalized
        exterior corrosion rate by SAV/10.  Under no circumstances, however, do
      . we use a generalized corrosion rate of less than 1.4 m1ls/yr.
                                     A-29

-------
CORROSION RATE:
     where

          rgen, ext * tne generalized external corrosion rate.


VARIATIONS:

     •  Direct currents win have the same effect on generalized exterior corro-
        sion as they have on localized exterior corrosion.

     0  Stainless steel tanks will corrode at one-fourth the rate applicable to
        steel tanks.  See event T1121.

     •  Above-ground tanks.  The atmospheric corrosion rate of 1.4 m1ls/yr
        applies (or one-fourth of that for stainless steel tanks).

     •  Cathodlc protection or coatings.  Significant generalized exterior
        corrosion Mill not begin until corrosion protection fails.  Since
        generalized exterior corrosion. 1s not Influenced by the presence of
        point anodes, we assume that once the coating has failed, generalized
        exterior corrosion proceeds in the same manner for coated and untreated
        tanks.
                                     A-30

-------
                              HAZARDOUS  WASTE  TANKS


LABEL:  T1128


FAILURE:  Tank  falls  due  to  generalized corrosion.


ASSUMPTIONS:

     •  tank corrosion may occur generally, due to a combination of rapid
        generalized exterior  and generalized  Interior corrosion.


CALCULATION:

     •  Tank wall-thickness must be computed even for tanks without localized
        corrosion.  In a few  cases, this will produce corrosion failure.


VOLUME:

     •  We determine the sizes and the growth rates for these holes in the same
        manner as we determine the sizes and growth rates for localized corro-
        sion holes.
                                     A-31

-------
                             HAZARDOUS WASTE TANKS


LABEL:  Bll


FAILURE:  Pipe Rupture


SOURCES:  SCS Engineers' analysis of API data (see event T1123).


CALCULATIONS:

     •  According to the API data, 3.4* of service station tanks experienced
        pipe rupture.  (See the table entry labeled "breakage" 1n event T1123)'.

     •  These tanks had been 1n use an average of 10.8 years.

     •  The annual rupture probability 1s therefore 3.4X/10.8 « 3.15 x 10'3/yr.


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  3.2 x 10"3/yr
VOLUME:
     •  We assume minimum and maximum crack sizes of 1" x .1" and 10" x (1/4)",
        respectively.

     •  We assume that rupture dimensions are uniformly distributed over the
        range of possible values.  Therefore:

                  width - FNU(.1,.25) Inches
                  length • FNUU.10) Inches

     •  Vie determine hydraulic heads by applying the pipe-pressure formula
        (Section A.1.5) to the geometry of the system design under con-
        sideration.  We assume that the rupture occurs at the midpoint of the
        pipe.

     •  Loss v<111 only occur during filling or discharging.  We calculate
        fill/discharge duration according to Section A.1.6.
                                     A-32

-------
VARIATIONS:

     •  Pipe length and system geometry will vary with system design.

     •  Above-ground pipes or pipes in vaults.  We assume that the rupture pro-
        bability for above-ground pipes is the same as that for underground
        pipes.  We also assume that the hole size distributions are the same for
        both types of pipes.

     t  Fiberglass pipes.  The probability of rupture is doubled.  See event
        T1124.

     •  Multiple pipes.  We evaluate each pipe separately.  We assume that the
        rupture of one pipe has no influence on the probability that other
        pipes will  also rupture.

USER INPUTS:

     •  System geometry

     •  Pipe diameter

     •  Pipe length
                                    A-33

-------
                             HAZARDOUS WASTE TANKS


LABEL:  8121


FAILURE:  Welded Flange Leak


SOURCES:

     t  Nuclear Regulatory Commission, Reactor Safety Study (1975).

     t  Henley and Kumamota (1981).

  .   •  SCS Engineers (1984).


ASSUMPTIONS:

     •  A flange leak can only occur when fluid is in contact with the.flange. -
        If the flange is located at the top of the tank, fluid only contacts it
        during filling, discharging, or overflow.  If the flange is at the bot-
        tom of the tank, fluid will always be in contact with it.  The location
        of the flange is dependent on system design.


CALCULATIONSr

     •  3 x 10'7/hr » welded flange leak rate (Reactor Safety Study).
        3 x 10-7/hr - .003/yr

     •  10'4 to 10-} per 10,000 hrs - welded flange leak rateJHenley & Kumamoto),
       . lO'4 to 10'1 per 10,000 hrs - 8.8 x 10'5 to 8.8 x 10'2/yr.  The
        geometric mean of this range of values * .0028/yr.

     •  API service station data gives a loose fitting rate of 5.OX (SCS
        Engineers).  The average system age is 10.8 yrs.  The failure rate is
        therefore 5.OX/10.8 - 4.6 x 10'3/yr.

     •  We therefore use a conservative flange leak probability of 5 x 10~3/yr.


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  5 x 10'3/yr
                                     A-34

-------
VOLUME:

     •  The flange is at the point of attachment of the pipe to the tank.

     •  The pressure drop due to friction losses therefore includes the entire
        length of the pipe.

     •  The elevation change can be calculated from the system geometry.

     t  We assume a 50% weld failure as upper bound, 0 as lower bound.  I.e., we
        assume that the length of the breach is distributed according to
        FNU(0, 502 of flange circumference).  We assume that the width of the
       'breach is given by FNU(0,l/8) inches.

     0  We assume a 300 to 400 Ib flange.  The exterior flange diameter is 6.5".
        (Source:  Perry and Chi 1 ton, Chemical Engineers'  Handbook, 5th Ed.
        (1973), pp. 6-66 and 6-67, using a 2"-diameter pipe).  The maximum weld
        failure is therefore 10" long.

     0  For underground flanges, we calculate the leak rate by using these
        values in the underground leak rate formula (Section A.1.1).  For above--
        ground flanges (or flanges in,an air environment such as a vault) we use
        the Bernoulli equation (Section A.1.2).

     t  We calculate the loss per day by multiplying the leak rate by the length
        of time that the fluid is in contact with the flange.

   *       *. •  •
VARIATIONS:   "

     •  Flange location*  A flange may be at the top of the tank or near the
        bottom of the tank.  Its location will influence the hydraulic head of
        the fluid and will determine the fraction of the time that the flange is
        in contact with.fluid.  The flange location will  be determined by the
        choice of system design.

     •  Multiple flanges.  We evaluate each flange independently.  Thus, we
        assume that the failure of one flange has no influence on the probabi-
        lity that other flanges will leak.


USER INPUTS:

     •  System design
                                     A-35

-------
                             HAZARDOUS WASTE TANKS
LABEL:  B122
FAILURE:  Gasket Falls
ASSUMPTIONS:

     •  We model gasket deterioration as an erosion-like process.  We assume
       ^that under Ideal conditions, 1t will occur relatively slowly, at a rate
        of less than 50 mils per year.


SOURCE:

     •  Best engineering judgment based on Nuclear Regulatory Commission,
        Reactor Safety Study (1975), and EPA, Case Studies 1-23 (1984) (case
        study #4).
PROBABILITY DISTRIBUTION:  Empirical
PROBABILITIES (baseline):

        Cumulative Probability

            0.00 to 0.77
            0.77 to 0.83
            0.83 to 1.00
Deterioration Rate (m1ls/yr)

        FNU(0,12,5)
        FNU(12.5,25)
        FNU(25,50)
DATE OF FAILURE:

     t  For a 2" pipe, the standard gasket is a flat disk with an inner radius
        of 1" and an outer radius of 2.0625".  Since disintegration occurs from
        the inside outward, 1.0625" of gasket must be dissolved or eroded before
        the fluid can escape.  A gasket for a 4" pipe Is slightly thicker
        (1.4375").  (Source:  Perry and Chllton, 1973).
VOLUME:  See event B121 (welded flange leaks).
COMMENTS:

     o  Under Ideal conditions, we have assumed that a gasket will deteriorate
        very slowly.  Waste/gasket Interactions will speed up the deterioration
        process, however.  To model such interactions we use a multiplicative
                                     A-36

-------
   factor selected from FNU(1,20).  Gasket failure can therefore occur as
   early as year 2.  This is consistent with the facts of EPA case study
   number 4 (Biocraft site).
                                                          i-

•  The Reactor Safety Study quotes a failure rate for containment-quality
   gaskets of 3 x 10-°/hr = 2.6%/yr.  This may be too low for a conven-
   tional hazardous waste tank system, but it serves as an approximate
   check for the results of our gasket disintegration model.
                                A-37

-------
                             HAZARDOUS WASTE  TANKS

LABEL:  B13

FAILURE:  Pipe Corrosion

SOURCES:
     •  John R. Rossum, "Prediction of Pitting Rates in Ferrous Metals from
      " Soil Parameters, Jour. AWWA, 1969, pp. 305-310.
     •  PACE (see event T1121).
     •  SCS Engineer's analysis of API survey  data (See event T1123).

PROBABILITY DISTRIBUTION:  Cumulative Empirical

CALCULATIONS (underground pipes):
     •  Localized exterior corrosion
        Rossum gives the following formula for maximum exterior pit depth for
        burijed steel pipes:
             p . l.Q6Knr(10-pH)t1nA.16
                       L   P   J
        where
              p • the depth of the deepest pit (in mils)
             Kn * 170 for soils of good aeration
                » 222 for soils of fair aeration
                » 355 for soils of poor aeration
             pH » the soil pH (must be between 5 and 9)
              n « 1/6 for soils of good aeration
                • 1/3 for soils of fair aeration
                » 1/2 for soils of poor aeration
                • the soil resistivity (ohm-cm)
              t « the number of years since the pipe was burled
              A • the surface area of the pipe (in square feet)
                                     A-38

-------
   We have added a fourth, aeration category, moderate aeration.   We
   obtained «n for this category by averaging the values for good aeration
   and fair aeration.  We obtained the value of n by taking the  geometric
   mean of the values for good and fair aeration.  Thus: '•

        KM = 196 for soils of moderate aeration
         n
         n » .236 for soils of moderate aeration

0  Multiple failures.  Unlike our tank-corrosion model,  which only predicts
   the date of first failure, the Rossum pipe-corrosion  model can be used
   to determine the occurrence of subsequent corrosion holes. The number
  *of leaks (L) at time t 1s given by:

                                              n/.16
              L - A
                    |-1.06Knl '•«  f
                    L~i~J    -   L
   We make this formula stochastic by converting fractional  values  of L
   Into probabilities.  Thus, we have assumed that the first leak occurs
   between L » 0 and L » 2.  At L * 1, there Is a 50% chance of failure.
   Similarly, the second leak occurs between the times when  L « 1 and L  »
   3, the third leak between L » 2 and L » 4, etc.  This means, for
   example, that a value of L equal to 3.2 means that 2 leaks have  occurred
   with certainty; there Is a 60% chance that a 3rd leak has occurred, and
   if the third leak has occurred, there Is a 10% chance that the 4th leak
   has' also begua.

   Note that the value of t 1s set to zero whenever detection/repair
   occurs.

   Caveat:  The Rossum model applies only for soil pH between 5 and 9.
   Outside that range, other corrosion mechanisms come Into  play.   These
   are not Included in our model.

•  Generalized Interior corrosion.  When fluid 1s 1n contact with the pipe,
   we assume that generalized Interior corrosion occurs under the same pro-
   bability distribution as applies to tanks.  Because these are indepen-
   dent events, however, we determine the corrosion rates separately for
   each tank and each pipe.  When the pipe Is in contact with air,  we use a
   corrosion rate of 1.4 mils per year (A1lor,. 1982).  When  we combine
   corrosion from the fluid and corrosion from the air, we obtain a genera-
   lized Interior corrosion of:

                          1.4(l-f) •»• f(r')

   where
 *

        f * the fraction of time the pipe 1s 1n contact with the fluid

       r' « the corrosion rate obtained for those times when fluid  1s
            present


                                A-39

-------
   Caveat:  The Rossum model applies only for soil pH between 5 and 9.
   Outside that range, other corrosion mechanisms come  into play.  These
   are not included in our model.

•  Localized interior corrosion.  Since our tank-corrosion model shows that
   localized interior corrosion 1s 19* as likely as localized exterior
   corrosion, we assume that the same ratio applies to  pipes.  We cannot
   use this Information directly, however, because our  localized exterior
   pipe-corrosion model only applies 1f soil conditions are known.  We
  , therefore must use API service station data to determine the probability
   that pipes undergo localized Interior corrosion.  From this, we can then
   obtain the probability of localized Interior pipe corrosion.
                                                                          >.
   According to the API data, 2B% of service station tanks experience pipe
   corrosion.  Since serv.ice station tanks generally have two pipes, we
   convert this value to a per-pipe probability by assuming that corrosion
   Is equally likely in either set of piping.  Thus, letting "a" be the
   probability of localized corrosion for a single pipe, 1t must be the
   case that:

                        .28 * 1 - (l-a)E

   or, a « .15.  Thus, 1f localized Interior pipe corrosion 1s 19X as com-
   mon as localized exterior pipe corrosion, and 1f other forms of corro-
   sion failure are uncommon, then approximately 2.4X of the API survey
   pipes;,failed by localized interior corrosion, while  12.6% failed by
   localized exterior corrosion.

   We assume that the time-to-failure distribution for  localized interior.
   corrosion is the same for pipes and tanks.  Thus, the time-to-fallure
   distribution for those pipes which exhibit localized Interior corrosion
   Is N(8,5).

   We adjust for area by multiplying both the probability of localized
   corrosion and the corrosion rate by (A/10)*1*.  See the discussion of
   above-ground pipes, below, for an explanation of this area correction
   factor.

•  Generalized exterior corrosion.  Generalized exterior corrosion reduces
   the time to pipe failure by localized Interior corrosion in the same way
   that It reduces the time to Interior corrosion failure for tanks.  We
   assume that generalized exterior corrosion rates for pipes follow the
   same probability distribution as applies to tanks.  See event T1127.

•  Erosion.  Because pipes carry moving fluids, they are subject to erosion
  . by suspended solids 1n the waste.  Erosion rates are dependent on the
   concentration of suspended solids.  Since our baseline corrosion values
   are derived for pipes carrying a non-erosive fluid (gasoline), we assume
   that they do not already account for erosion.  We therefore add the
   following erosion rates to both our localized Interior and generalized
   Interior corrosion rates:
                                 A-40

-------
                  Fraction of suspended               Erosion
                    solids in the waste                 rate
                  _ (ppm) _                (mils/yr)'

                        <1,000                           0
                      1,000-10,000                    FNU(O.S)
                        >10,000                       FNU(5,10)


VOLUME OF RELEASE:

     •- We assume that localized exterior corrosion holes follow the same size
        distribution for pipes as they do for tanks.  We calculate their growth
        rates 1n the same manner as we calculate growth rates for exterior
        corrosion holes In tanks.

     •  Localized Interior corrosion holes also follow the same size distribu-
        tion and grow in the same manner as do the corresponding holes in tanks.

     •  If the pipe falls due -to generalized corrosion, the holes will be larger.
        than they will be for localized corrosion.  We assume that such holes
        begin with the dimensions of 'a pipe rupture (see event BID, and double
        In length and width -each year until they are detected.  In order to
        account for the physical limitations Imposed by the dimensions of the
        pipe, we do not allow such corrosion holes to become more than 3" wide
        or longer than the length of the pipe.
                              *

     •  We calculate fluid velocity, transfer time, and pressure 1n the same
        manner as we used for underground pipe ruptures.  For both corrosion
        holes and ruptures, we make the simplifying assumption that the leak
        occurs at the midpoint of the pipe.

     •  We use the underground leak rate equation (Section A. 1.1) to calculate
        loss rates.  Multiplying this leak rate by the fill or discharge time
        (Section A. 1.6) then gives the total volume of loss per filling or
        discharging.

ABOVE-GROUND PIPES:

     •  The Rossum Model does not apply to above-ground pipes.

     •  Localized exterior corrosion.  We obtain the localized exterior corrosion
        rate for above-ground pipes from the following conditional normal
        distribution:

                               fN(16(10/.05A)16. 6.6(10/.05A)-16]
                               l               '                 J
        where
                                      A-41

-------
     A = the surface area of the pipe in square feet. *

Interpretation.  The first term of this formula, .1H.05A/10)-16 1|gthe
probability that localized exterior corrosion occurs.  The (A/10)'
factor is the same area adjustment factor that we use for tanks, except
that for pipes, the baseline area 1s 10 square feet.  This area
corresponds to 20 feet of 2" pipe or 10 feet of 4" pipe.  These pipe
dimensions are appropriate for service station fill and discharge pipes,
respectively.
    .11 coefficient In this expression 1s the baseline probability of
localized exterior corrosion.  We derived this probability from a com-
bination of the PACE tank data and the API pipe-corrosion data.
According to the PACE data, 70X of low-SAV tanks fail by localized
exterior corrosion within 30 years, while 77% of medium-SAV tanks fail
by that mechanism 1n the same period.  According to the API data, 12.6%
of service station pipes fall by localized exterior corrosion.  If we
assume that this percentage applies to medium-SAV soils and that the
same proportionality factor applies to low- and medium-SAV pipes as
applies to low- and medium-SAV tanks, then (70/77) x 12. 6X of low-SAV
service station pipes fall by localized exterior corrosion.  To two
significant figures, this percentage Is lit.  If we assume that above-
ground pipes corrode like low-SAV underground pipes (an assumption s~imi--
lar to that which we made for above-ground tanks) then we can apply this
percentage to above-ground pipes.

The second part'of our localized exterior pipe-corrosion formula,
N(16(10/A)-16, 6.6(10/A)-16), gives the conditional date of failure for
pipes that experience localized Interior corrosion.  The factor
(10/A)*16 reduces the time to failure according to the Inverse of the
area adjustment factor.  Thus, when a corrosion rate 1s calculated from
the sampled time-to-failure, that corrosion rate will be higher than
baseline by a factor of (A/10)-16.  .

Our baseline time-to-failure distribution in this expression 1s
N(16,6.6).  We derived this distribution in the same manner as we
derived the lit baseline probability of failure.  According to the API
survey data, service station pipes show the following distribution of
failure dates:

           Year of             Percentage of reported
           failue                  pipe failures

             0-1                        2.1
             2-5                        6.4
             6-10                      32.8
            11-15                      33.1
            16-20                      18.1
            21-25                       5.0
            26-30                       2.3
            30*                         0.2
                              A-42

-------
We can approximate this distribution as N(12,5).  Assuming that this
distribution applies to medium-SAV pipes, we can obtain"a low-SAV
time-to-failure distribution by comparing the PACE distributions for
medium- and low-SAV tanks.  These distributions show conditional mean
dates of failure of 21 and 16 years, respectively.  Adjusting both
parameters of the API distribution by 21/16 gives us a low.-SAV time-to-
failure distribution of N(16,6.6).

The final element of our above-ground localized corrosion formula 1s the
.05 factor used to reduce the effective surface area of the pipe.  We
use this factor because we assume that like above-ground tanks, above-
ground pipes are only vulnerable to localized exterior corrosion at
their seams and points of support.  These regions account for approxima-
tely 5X of their surface areas.

In our pipe-corrosion distribution, we have applied the area adjustment
factor to both the baseline probability and the baseline corrosion rate.
This 1s different from the way that we adjust for area 1n our tank
corrosion model, for in that model, we only applied our area adjustment
factor to corrosion rates.     •                        ,

We have modeled pipes and tanks differently because we believe that
three factors control the onset of localized corrosion.  One of these 1s
the surface area of the burled metal.  The others are the corrosivity of
the soil and the care with which the component 1s Installed.  We have
assumed that surface area 1s an Important factor for localized corrosion
event's with low probabilities.  In other words, we have assumed that the
probability of pipe corrosion 1s relatively low because a pipe is small
enough that It Is relatively unlikely to experience a point anode.  For
tanks, however, the baseline probabilities of localized corrosion are on
the order of 70-85%.  We assume therefore, that the principal factors
Influencing the onset of tank corrosion are the corrosivity of the soil
and the care with which the tank 1s Installed.  Surface area will be
Important, but Its primary effect will be to determine the number of
point anodes and thus the depth of the deepest pit.

We have generalized the preceding discussion to obtain the following
rule of thumb:  whenever the baseline probability of a corrosion event
1s less than SOX, we have assumed that component surface area will
Influence the probability of the onset of corrosion; whenever the base-
line corrosion probability 1s over SOX, we assume that the component 1s
already large enough that surface area has little Influence on probabi-
lity.  Since all of our relevant baseline probabilltes are either
greater than 70X or less than 12X, we never had to elaborate this rule
of thumb by developing a model to deal with Intermediate cases.  In
addition, since our area adjustment factor requires an area more than
SOOO times larger than baseline to increase a 12X probability to SOX, we
did not need to modify that factor to assure that our area adjustments
do not Increase the probability of failure to a value greater than SOX.
Finally, our model 1s Insensitive to any choice of cut-off probabilities
between 251 and 70X, so it Is unnecessary for us to be precise In our
determination of what value In that range Is the theoretically best
choice.

                              A-43

-------
     •  Generalized exterior corrosion.  Generalized exterior corrosion for
        above-ground pipe segments occurs at the atmospheric corrosion rate of
        1.4 mils/yr.

     •  Localized interior corrosion and generalized interior corrosion.  The
        corrosion rates for these types of corrosion are the same for above- and
        below-ground pipes.

     •  Combinations of corrosion mechanisms.  We take combinations of corrosion
        mechanisms into account in the same manner for pipes as we used for
        tanks.  Thus, to determine the date of interior corrosion failure, we
       "compute the remaining wall-thickness for the combination of localized
        interior corrosion and generalized exterior corrosion.  To determine the
        date of exterior corrosion failure, we calculate the remaining wall-
        thickness from the combination of localized exterior corrosion and
        generalized Interior corrosion.  To determine the date of generalized
        Interior corrosion, we calculate the remaining wall-thickness from the
        combination of the two generalized corrosion mechanisms.  We do not com-
        bine generalized exterior and localized exterior corrosion (or genera-
        lized Interior and localized Interior) because we assume that our
        localized exterior corrosion rates already Include both forms of
        exterior corrosion.

     •  Pipe thickness.  Pipe thickness 1s directly Included as a parameter in
        the Rossum localized exterior corrosion model.  For our PACE- and
        API-derived corrosion rate formulas, we use a baseline pipe thickness of
        190 mils, which 1s the average of the thicknesses generally used for
        2" and 4" pipe (Peters and Timmerhaus (1980)).

     •  Hole sizes.  We use the same hole-size distribution for above- and
        below-ground pipes.  We calculate hole growth rates in the same manner
        for both types of pipes.


VARIATIONS (above- and below-ground):

     •  Coated pipes.  When a coating falls, there Is a 100X chance of point
        anodes developing at the sites of coating failure.  Thus, once the
        coating falls, localized exterior corrosion begins with certainty.
        Because the Rossum model does not take this factor Into account, we use
        the following baseline time-to-failure distributions for low-, medium-,
        and high-SAV pipes:

                                    Piping t1me-to-fallure distribution
                Soil type                following coating failure

                Low SAV                          NU6.6.6)
                Medium SAV                       N(12,5)
                High SAV                         N(9,4)

        These distributions are derived from the PACE and API data sets.  The
        derivation process for high-SAV pipes is the same as that described
        earlier for medium- and low-SAV pipes.

                                      A-44

-------
        These distributions apply to 190-mil pipes.  We model different pipe
        thicknesses using the same approach that we used to model variations  in
        tank wall-thicknesses.

      • Cathodic protection.  We assumed that the entire tank facility uses a
        cathodic protection system powered by a single power supply.  Thus, when
        that system fails and repairs are not undertaken within a reasonable
        time, cathodic protection falls for both the tank and the pipes.
        Cathodic protection failure Is discussed under events T1121.  As long as
        it 1s functioning, cathodic protection will prevent both Interior and
       "exterior corrosion.

     t  Loss rates from above-ground pipes or below-ground pipes with secondary
        containment.  Leak rates in such circumstances will be controlled by the!
        Bernoulli equation (see Section A.1.2).

     0  Stray currents.  If stray currents exist, they will affect pipes and
        tanks similarly.  See event T1121.  Because stray currents are likely to
        be equally severe for all elements of a tank system, we only sample the
        stray current event once for .the entire facility.

     •  Stainless steel.  Stainless steel reduces all corrosion.rates by a fac-
        tor of .25.

     •  Fiberglass.  Fiberglass pipes do not corrode.
USER INPUTS:
        Pipe material
        System design
        Corrosion protection system
        Pipe thickness
        Surface area of pipe
        Soil characteristics
                                      A-45

-------
                             HAZARDOUS WASTE TANKS


LABEL:  Alllll


FAILURE:  Strainer drain left open after maintenance (or during pump-out for
          storage or accumulation tanks).


ASSUMPTIONS:
       i

     •  We assume that the minimum operator response time is 15 seconds.

     •  Because the operator is likely to be in the vicinity, we assume a maxi-
        mum response time of 3 minutes.  We therefore assume that the response
        time is distributed FNUC.25,3) minutes.

     •  Strainer maintenance occurs once per month.


SOURCES:

     •  JRB Associates (1982)

     •  Nuclear Regulatory Commission, Reactor Safety Study (1975)
             '-
                                          *
CALCULATIONS:

     •  1.7 x 10'3 per operation » operator failure rate for operations
        embedded in a procedure (JRB).

     •  This is a relatively infrequent procedure (once per month) with no imme-
        diate feedback.  Therefore, 1 x 10"' (general failure rate for opera-
        tions with no status display) is probably a better figure (Reactor
        Safety Study).


PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY:  1 x IQ"2/month


VOLUME:

     •_ We assume that the leak rate is equal to the pumping rate.


VARIATIONS:

     t  Storage and accumulation tanks.  For many of these tanks, the strainer
        will be included as part of the pump-out truck.  In such cases, error in

                                     A-46

-------
        strainer maintenance will be detected and corrected at the first pump-
        out following pump maintenance.

     t  We assume that the pump-out truck visits 40 tanks per month.

     t  Let n be the number of pump-outs per month for the modeled tank (n may
        be less than 1).  Then the probability 1s n/40 that the facility is the
        first one to be visited after any given strainer maintenance.  The
        annual probability of a spill due to faulty maintenance of the pump-out
        truck's strainer Is therefore given by:

                   1 x 10'2 x (n/40) per month » 3n x 10'3/yr.


USER INPUTS:

     •  Number of pump-outs per month (n)
                                     A-47

-------
                             HAZARDOUS WASTE TANKS

LABEL:  A11112

FAILURE:  Pump drain left open during pumping.

ASSUMPTIONS:
     •  This can only happen following pump maintenance.
       ^
     0  Pump maintenance occurs annually.
     •  The response time In the event of a spill is FNU(.25,3) minutes.

SOURCES:
     t  Nuclear Regulatory Commission, Reactor Safety Study (1975)
     •  JRB Associates (1982)
     •  Event Alllll (strainer drain left open)
CALCULATIONS:
     •  See event Alllll

PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY:  10'2/yr

VOLUME:
     •  The spill rate Is equal to the pumping rate.

VARIATIONS:
     •  Storage and accumulation tanks.  For many of these tanks, the pump-out
        pump Is Included as part of the truck.  An error In pump maintenance
        w11rtherefore be detected and corrected at the first pump-out
       •following pump maintenance.
     •  We assume that the truck visits 40 sites/month.
                                     A-48

-------
     0  Let n be the number of pump-outs per month at the facility being
        modeled.  Then the probability that the facility being modeled  is  the
        first one to be visited after pump maintenance is n/4Q^ ^-and the
        probability that it is the first one visited after faulty pump
        maintenance is given by:

             (1 x 10-2)(n/40) per year - 2.5n x 10'4/yr


USER INPUTS:

     •  Number of truck visits per month (n)
                                     A-49

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A1112


FAILURE:  Hose ruptures (above ground) during pump-out (storage or accumulation
          tanks only).


ASSUMPTIONS:

     0 -Hose ruptures inside the pump-out pipe will be inconsequential.  Leakage
        will return to the tank in all but extraordinary circumstances.


SOURCES:

     t  JRB Associates (1982)


CALCULATIONS:

     •  Hose rupture probability » 1 x 10"4/hr (JRB)

     •  Annual probability of hose rupture « 1 x 10"4/hr x T hr/pump-out
        x n pump-outs/week x 52 weeks/yr » 5(nT) x 10"3/yr

             tf
PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  5(nT) x 10'3/yr


VOLUME:

     •  The maximum leak rate will be the pumping rate for the pump-out pump.
        The minimum is approximately 0.

     •  We assume that the leak rate is uniformly distributed between these two
        exremes.

     t  The maximum detection/response time is the entire discharge time (see
        Section A.1.6).  The minimum detection/response time 1s about 15 seconds.


USER INPUTS:
      V
     •  Number of discharges per week (n)
     0  Time required for pump-out (from Section A.1.6)
                                      A-50

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A1113, A1114


FAILURE:  Pump or strainer ruptures during pump-out (storage or accumulation
          tanks only).


SOURCES:

     •  U.S. Coast Guard (1978)
     •  MRS Associates (1982)


CALCULATIONS

     •  1 x 10'8/hr « strainer rupture rate (JRB)

     •  1 x 10'8/hr • pump rupture rate (JRB)  -


COMMENTS:

     •  These are very low probability events producing spills which will pro-
        bably be contained by an above-ground pad.  These events are
        therefore not included in our model.
                                     A-51

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A1115


FAILURE:  Loose flexible hose connection during pump-out (storage or accumula-
          tion tanks only).


SOURCES:

     9* Nuclear Regulatory Commission, Reactor Safety Study (1975)


CALCULATIONS:

     •  1 x 10"2/demand * general rate of human error (Reactor Safety Study)

     0  1 x 10'2/demand « m x 10"2/month where m 1s the number of pump-outs per
        month (m will generally be less than 1).


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:

     tlx lt)'2/demand » m x 10"2/month where m 1s the number of pump-outs per
        year.


VOLUME:

     •  Assume that the maximum loss rate 1s the hose flow rate.  So leak rate
        1s FNU(0, hose flow rate).


USER INPUTS:

     t  Number of pump-outs/year (m)
     •  Tank capacity (Q)
                                    .A-52

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A2113, A2114


FAILURE:  Strainer or pump rupture during tank filling.


SOURCES:

     t  U.S. Coast Guard (1978)
     • , JRB Associates (1982)


CALCULATIONS:

     •  1 x 10"8/hr » strainer rupture rate
     •  1 x 10*8/hr « pump rupture rate


COMMENTS:

     •  These are very low probability events producing spills which will
        probably be contained by an above-ground pad.  These events are
        therefore not included in the computer model.
                                     A-53

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A2116


EVENT:  Pump Corrodes


ASSUMPTIONS:

     t  Pump corrosion 1s composed of:
        *
        -  generalized Interior corrosion
        -  localized Interior corrosion
        -  generalized exterior corrosion
        -  loca-lized exterior corrosion
        -  erosion


CALCULATIONS:

     •  Our pump-corrosion model Is very similar to our pipe-corrosion model.

     •  Localized interior corrosion." We model localized Interior corrosion in
        the same way for pumps and pipes.  Because a pump's complex shape makes
        1t more vulnerable to localized Interior corrosion than 1s a pipe, we
        use-.the same probability distributions for localized interior corrosion
        of pumps and pipes.  See event B13, above.

     •  Generalized interior corrosion.  Generalized Interior corrosion 1s also
        the same for pumps and pipes.

     •  Generalized exterior corrosion.  Because pumps are exposed to the atmos-
     -  phere, we use a generalized exterior corrosion rate of 1.4 mils per
        year.

     •  Localized exterior corrosion.  Because of the pump's small surface area
        and above-ground setting, localized exterior corrosion is highly un-
        likely and we have not Included it in our pump-corrosion model.

     •  Erosion.  Because of the pump's complex shape, it will be more subject
        to erosion than 1s a pipe.  We assume the following dependence of ero-
        sion rate on fraction of suspended solids:

                         Fraction of             Erosion rate
                      Suspended Solids            (mils per
                           (ppm)	             year)

                         0-10,000                 FNU(O.IO)
                          10,000                  FNUU0.20)
                                       A-54

-------
VOLUME:

     •  We use the same hole sizes for pumps as we used for pipes.


COMMENTS:

     •  In some cases, pumps may be located inside the tank.  In these cases,
        pump corrosion cannot produce a release of fluid.
                                     A-55

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A210


FAILURE:  Fluid flows over the top of an open-topped tank during overfill
          events.


ASSUMPTIONS:

     •  .Open-topped tanks have a ready overflow route over the top of the tank,


VOLUME:

     •  The leak rate will be equal to the rate at which fluid is pumped into
        the system.

     •  The overflow may be detected and remedied immediately, or it may con-
        tinue through the entire batch.

     •  We assume that the overflow duration is uniformly distributed between
        zero and the entire fill time for the tank.  Thus, the overflow volume
        is given by FNU(0, volume of entire batch).
                                     A-56

-------
                             HAZARDOUS WASTE TANKS

LABEL:  A211

FAILURE:  Fluid flows out the vent of an above-ground tank during overfill
          events.

ASSUMPTIONS:
     •  Above-ground tanks have a ready overflow route through their vents.

VOLUME:"
     •  The leak rate will be equal to the rate at which fluid 1s pumped into
        the system.
     •  The overflow may be detected and remedied Immediately, or 1t may con-
        tinue through the entire batch.
     t  We assume that the overflow duration is uniformly distributed between
        zero and the entire fill time for the tank.  Thus, the overflow volume
        1s.given by FNU(0, volume of entire batch).
                                     A-57

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A213


EVENT:  Pump-out pipe rupture leads to loss during overfill events.


SOURCES:  Bll


PROBABILITY DISTRIBUTION:
        *
     •  Non-stochastic.  This event will occur if the pump-out pipe has ruptured
        under event Bll.


PROBABILITY:  See event Bll (pipe rupture)


VOLUME:

     •  If there is a pump, we assume that It shuts off automatically.  Then, the
        pressure at the point of the rupture will be determined by the static
        hydraulic head of the backed-up fluid.

     •  We assume that the average rupture occurs at the midpoint of the pump-
        out 'pipe.  Then the hydraulic head can be determined from the system
        1 ayout''.

     •  We assume an overfill detection/response time of FNU(.25,60) minutes
        (the operator notices that the fluid is not flowing).


USER INPUTS:

     •  System design


COMMENTS:

     t  This event is only important for systems for which pump-out is by
        flexible hose.  For these tanks, this event is the only source of loss
        through a ruptured pump-out pipe.  For other tanks, losses by this
        mechanism will be overshadowed by losses through the same holes during
        normal discharge operations.
                                     A-58

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A214


FAILURE:  Pump-out pipe corrosion results in leak during overfill events.


SOURCES:  B13


PROBABILITY DISTRIBUTION:

     t  Non-stochastic.  This event will occur if the pump-out pipe has corroded
        under event B13.


PROBABILITY:  See event B13 (pipe corrosion)


VOLUME:

     •  If there is a pump, we assume that it shuts off automatically,  then, the
        pressure at the corrosion hole will be determined by the static
        hydraulic head of the baclced-up fluid.

     •  As in event B13, we assume that the average corrosion hole occurs at the
        midpoint of the pump-out pipe.  Then the hydraulic head can be deter-
        mined 'from the system layout.

     •  We assume an overfill detection/response time of FNU(.25,60) minutes
        (the operator notices that the fluid is not flowing).


USER INPUTS:

     t  System design
COMMENTS:
        This event 1s only Important for systems for which pump-out is by
        flexible hose.  For these tanks, this event 1s the only source of loss
        through a corroded pump-out pipe.  For other tanks, losses by this
        mechanism will be overshadowed by losses through the same holes during
        normal discharge operations.
                                     A-59

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A215


FAILURE:  Outlet flange or gasket leak produces losses during overfill events.


SOURCES:  See B121 or B122


PROBABILITY DISTRIBUTION:

     0  Non-stochastic.  This event will occur if the flange or gasket is
        leaking under events B121 or B122.


PROBABILITY:  See event B121 (flange leak) or B122 (gasket leak).


VOLUME:                              .

     •  If there Is a pump, we assume that 1t shuts off automatically.-  Then, the
        pressure at the point of"the rupture will be determined by the static
        hydraulic head of the backed-up fluid. .This will be determined by the
        system design.

     t  We assume an overfill detection/response time of FNU(.25,60) minutes
        (the operator notices that the fluid 1s not flowing).


USER INPUTS:

     •  System design


COMMENTS:

     •  This event Is only Important for systems for which pump-out is by
        flexible hose.  For these tanks, this event 1s the only source of loss
        through a leaking outlet flange or gasket.  For other tanks, losses by
        this mechanism will be overshadowed by losses through the same hole
        during normal discharge operations.
                                    A-60

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A216


FAILURE:  Inlet pipe rupture produces leaks during overfill events.


PROBABILITY DISTRIBUTION:

     •  Non-stochastic.  This event will only occur if the inlet pipe has-
        already ruptured under event Bll.


PROBABILITY:  See event Bll (pipe rupture).


VOLUME:

     •  The leak-rate calculations are similar to those used for event A213
        (pump-out pipe rupture).


COMMENTS:                                                r

     •  Leakage from this pipe will also produce losses during normal filling.
        Cumulative losses from that mechanism will generally be much larger
        than losses occurring during overfill events.
                                     A-61

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A217


FAILURE:  Fill pipe corrosion produces losses during overflow events.


PROBABILITY DISTRIBUTION:

     •  Non-stochastic.  This event will only occur if the fill -pipe  is already
       , corroded leaking under event B13.


PROBABILITY:  See event B13 (pipe corrosion).


VOLUME:

     t    The leak rate can be obtained from the geometry of event B13,. using
          a static hydraulic head.  The size of the corrosion hole is determined
          under event B13.


COMMENTS:

     •  Leakage from this hole will also produce losses during normal filling.
        Cumulative losses from that mechanism will generally be much  larger than
        losses occurring during overfill events.
                                     A-62

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A218


FAILURE:  Inlet flange or gasket leaks during overfill events.


PROBABILITY DISTRIBUTION:

     •  Non-stochastic.  This event will only occur if the inlet flange or
        ,gasket is already leaking under events B121 or B122.


PROBABILITY:  See event B121 (flange leak) or B122 (gasket leak).
VOLUME:
     t  The leakage calculations are identical to those used for event A215
        (outlet flange or gasket leak).                      .  .        ..
COMMENTS:
        Leakage from this flange or gasket Mill also produce losses during nor-
        mal .filling.  The leak rate will be much higher during overflow,
        however, because the pressure will be much higher under the conditions
        of static hydraulic head that occur during overflow than it will be when
        fluid is in motion during normal filling.  During normal filling, the
        pressure at this location is very low.
                                    A-63

-------
                             HAZARDOUS WASTE TANKS





LABEL:  A219





FAILURE:  Vent pipe rupture produces losses during overfill events.





CALCULATIONS:  See event Bll (pipe rupture).





PROBABILITY:   See event Bll.
        n.



VOLUME:



     •  The volume can be obtained from the system layout, using static

        hydraulic heads and sizing the rupture according to the method

        used for other pipe ruptures.  See event Bll.
                                     A-64

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A220


FAILURE:  Vent pipe corrosion produces losses during overfill events.


CALCULATIONS:  See event B13 (pipe corrosion).


PROBABILITY:   See event B13.


VOLUME:

     •  The volume can be obtained from the system layout, using static
        hydraulic heads and sizing the corrosion hole according to the method
        used for other pipe ruptures.  See event 613.
                                     A-65

-------
                             HAZARDOUS WASTE TANKS


LABEL:  A221


FAILURE:  Vent pipe flange leaks during overfill  events.


CALCULATIONS:  See event B121 (welded flange leaks).


PROBABILITY:  See event B121.


VOLUME:

     •  The volume can be obtained from the system layout,  using static
        hydraulic heads and sizing the flange leak according to the method used
        for other flange leaks.  See event B13.
                                    A-66

-------
                             HAZARDOUS WASTE TANKS


LABEL:  B21


FAILURE:  Asphalt Pad Breached


SOURCES:  Conversations with local asphalt contractors.


ASSUMPTIONS:
        •V
     •  Breakage 1n a 2" pad with 6" class-5 base (a mixture of gravel,
        sand, and clay) will become general In an average of 8-12 years,
        depending on maintenance.  It could occur as early as 3-5 years.

     •  A 3-4* pad with a crushed limestone base should last 15 years
        before generalized break-up begins.

     •  Break-up will occur earlier If the pad is not properly maintained.

     •  The entire spill volume 1s lost 1f the pad has broken up.


PROBABILITY DISTRIBUTION:  Beta .


PROBABILITY:'  ;,

     •  For 2" pad, no maintenance, we use a beta distribution with a
        minimum value of 2.5 years, a mode of 8 years, and a maximum of
        12 years.

     •  For a 2" pad, with maintenance, (or a 3-4" pad without maintenance)
        we use a beta distribution with parameters (4, 12, 15).

     0  For a 3-4" pad with maintenance, we use a beta distribution with
        parameters (5, 15, 18).

VOLUME:  Entire volume of the spill.
                                     A-67

-------
                             HAZARDOUS WASTE TANKS



LABEL:  B31



FAILURE:  Concrete Pad Breached



SOURCES:


     •  Telephone conversations with concrete contractors and officials 1n the
        Minnesota Department of Transportation.
       •L


PROBABILITY DISTRIBUTION:  Normal



PROBABILITY:  N(30,5)



VOLUME:  Entire volume of spill.
                                     A-68

-------
                             HAZARDOUS WASTE TANKS

LABEL:  833

FAILURE:  Breach of Concrete Berm

ASSUMPTIONS:
     t  Concrete berms will age similarly to concrete pads (see event B31).

SOURCES:  'Event B31

PROBABILITY DISTRIBUTION:  Normal
Probability:  N(30,5)

VOLUME:  Total spill volume.
                                     A-69

-------
                              HAZARDOUS WSTE TANKS


LABEL:  B41


FAILURE:  Concrete Vault Fails


SOURCES:

     0  Best engineering judgment after converstatlons with concrete contrac-
        tors. .
       •s.

PROBABILITY DISTRIBUTION:  Normal


PROBABILITY:  N(35,10)


VOLUME:  Total volume of spill.
                                     A-70

-------
                              HAZARDOUS WASTE TANKS

LABEL:  851

FAILURE:  Synthetic Liner Fails

SOURCES:
     0  EPA, Liner Location Report (1984-).

PROBABILITY DISTRIBUTION:  Normal

PROBABILITY:  N(35,10)
VOLUME:  Total volume of spill.
                                     A-71

-------
                             HAZARDOUS WASTE TANKS

LABEL:  ANUCAT(I.l)

FAILURE:  Vehicle Crash

ASSUMPTIONS:
     •  This event 1s Included 1n tank or pipe rupture (events T1124 and 811),
                                     A-72

-------
                             HAZARDOUS WASTE TANKS


LABEL:  ANUCAT  (1,2)  (aboveground or In-ground tanks only)


FAILURE:  Vandalism of tank system resulting in total system loss.


ASSUMPTIONS:

     •  We assume that the probability of catastrophic release due to vandalism
        is of the same order of magnitude as the probability of catastrophic
        release due to a vehicle crash.
       s.


SOURCES:

     •  SCS Engineers (1983), Figure 4-18, p. 4-29.

     •  JRB Associates (1982), Exhibit 3-5.


CALCULATIONS:

     •  < 1 x 10-10/hr - failure rate for vehicle crash (JRB).

     0  10-10/hr - (1 x 10'10/hr)(24 hr/day)(365 day/yr) - 10-6/yr.
                                                        «

PROBABILITY: " 1 x 10'6/yr


PROBABILITY DISTRIBUTION:  Binomial


VOLUME:  Entire tank contents.
                                     A-73

-------
                             HAZARDOUS WASTE TANKS


LABEL:  ANUCAT (1,3)


FAILURE:  Tornado/hurricane (I.e. high wind storm event) resulting in total
          system loss (above-ground or in-ground tanks only).


ASSUMPTIONS:

     t ^The facility is in coastal area subject to periodic hurricanes or in a
        tornado-prone area.

     0  The annual probability of a great hurricane (winds exceeding 125 mph)
        for 50-mile segments along the U. S. coastline ranges from 1% to 7X
        (Petak and Atkisson).

     •  Assume 20X damage for hurricanes with wind speeds of 125 mph (Petak and
        Atkisson).

     0  Assume an average of 2.5 tornado strikes per 10,000 square miles for the
        continental U.S.  (U.S. Weather Bureau).

     0  Approximately 35X of all tornadoes have a Fujita classification of
        F2 or above (135 - 290 mph) (Petak and Atkisson).
                                                  *
     0  Assume approximately 50% damage for structures affected by tornadoes
        with a Fujita classification of F2 or above (Petak and Atkisson).

     0  Assume a facility has an area of approximately 10 acres.

     0  Assume that the average tornado strike is 2 mi x 300 yards * 200 acres.
        (U.S. Weather Bureau.)


SOURCES:

     0  Petak, William J. and Arthur A. Atkisson, Natural Hazard Risk Assessment
        and Public Policy; Anticipating the Unexpected. Sprlnqer-Verlag, New
        York (1982).

     0  United States Weather Bureau, Minneapolis Office, Personal
        Communication.


CALCULATIONS:
       •
     0  7% annual probability of hurricanes x 20% chance of damage » 1.4X chance
        of a damage due to a hurricane.
                                       A-74

-------
     •  (2.5/10,000) tornado strikes per square mile x (200 acres/strike) x
        (Imi2/640 acres) x 10 acres/facility x (.35 x .5) probability of damage
        =» 1.5 x 10'4/yr per facility.


PROBABILITY:

     t  0.014/yr for hurricanes

     •  1.5 x 10'Vy'r for tornadoes


PROBABILITY DISTRIBUTION:  Binomial
        ^


VOLUME:  Entire tank contents


USER INPUTS:

     •  Is facility 1n a tornado zone?
     •  Is facility 1n a hurricane zone?


COMMENTS:  This event only applies for above-ground or In-ground facilities.
                                     A-75

-------
                             HAZARDOUS WASTE TANKS


LABEL:  ANUCAT (1,4)


FAILURE:  Earthquake causes total system loss.


ASSUMPTIONS:

     t  The facility Is located in a seismically active area.  Recurrence
        intervals for damaging earthquakes in Los Angeles and San Francisco
        areas are between 100 and 125 years.
       \
     •  Assume .5-15% damage to commercial structures (built in California
        after 1933) in response to an earthquake with intensity of 7 or above.


SOURCES:

     •  California Institute of Technology (personal communication).


CALCULATIONS:               .

     •  (1/125) earthquakes per year x 10* average probability of damage
         » 8 x 10'* damaging earthquakes per year.

          \  • '
PROBABILITY:'' .0008/yr


PROBABILITY DISTRIBUTIONS:  Binomial


VOLUME:  Entire tank contents


USER INPUTS:

     •  Is facility 1n a seismically active area?
                                     A-76

-------
                             HAZARDOUS WASTE TANKS


LABEL:  ANUCAT (1,5)


FAILURE:  Flood causes total system loss (above- or In-ground tanks only).


ASSUMPTIONS:

     •  The facility 1s located in a flood prone area.

     0  The facility 1s designed to withstand up to a 100-year flood.


SOURCES:

     •  Thomas Dunne and Luna Leopold, Water In Environmental Planning.
        W. H. Freeman and Company, San Francisco (1978).

     t  William J. Petak and Arthur A, Atklsson, Natural Hazard Risk Assessment
        and Public Policy: Anticipating the Unexpected, Springer-Verlag, New
        York (1982).


CALCULATIONS:

     •  According to Petak and Atklsson there Is a 50% chance that a flood will
        result in damage to an above-ground tank.

     •  There 1s a 1% chance per year of a 100-year flood.

     t  IX x 50* » .5%


PROBABILITY:  .005/yr


PROBABILITY DISTRIBUTION:  Binomial


VOLUME:  Entire contents of tank.


USER INPUT:
        •

         Is facility located in a flood-prone area?


COMMENTS:  This event applies only for above- or In-ground facilities.
                                     A-77

-------
                             HAZARDOUS WASTE TANKS

LABEL:  ANUCAT  (1,6)

FAILURE:  Ignition source available to Ignite waste in tank system (all
          systems).

ASSUMPTIONS:
     •  The tank 1s properly grounded.
     •  The operator 1s reasonably cautious In handling the waste.
PROBABILITY:  1 x 10*6/yr

PROBABILITY DISTRIBUTION:  Binomial

VOLUME:  Entire tank contents

COMMENTS:
     •  This event applies 1f the waste Itself Is the source of the fire.  If the
        tank 1s ruptured by a nearby fire or explosion, ANUCAT (1,7) applies.
     •  This event applies only if the waste 1s flammable.
                                     A-78

-------
                             HAZARDOUS WASTE TANKS

LABEL:  ANUCAT (1,7)

FAILURE:  Nearby fire or explosion causes complete system loss.

ASSUMPTIONS:
     •  A nearby fire or explosion is 1/3 as likely to damage an underground
       .tank as it is to damage an above-ground tank.
SOURCES:
     •  JRB Associates (1982), Exhibit 3-5
CALCULATIONS:
     t  3 x 10'7/hr - probability of nearby fire (JRBj
     •  (3 x 10-7/hr)(24 hr/day)(365 days/yr) » 2.6 x  10'3/yr

PROBABILITY?
     0  3 x 10'3/yr (above-ground or in-ground) tank
     •  1 x lO'Vyr (below-ground tank)

PROBABILITY DISTRIBUTION:  Binomial

VOLUME:  Entire tank contents
                                     A-79

-------
                             HAZARDOUS WASTE TANKS




LABEL:  LIFDEF (1,1) and LIFDEF (1,2)




FAILURE:  Vibrational/tortlonal stress causes rupture due to Inadequate

          support or due to a construction defect.




ASSUMPTIONS:


     •  The only part of the system subject to vibration is the pump.
        i

     •  We assume that these losses are included in pipe rupture (event BID,




PROBABILITY:  Zero.  This loss mechanism is Included- in event A1113.
                                    A-80

-------
                             HAZARDOUS WASTE TANKS


LABEL:  LIFDEF (1,3)


FAILURE:  Inspection falls to detect installation  damage  or  fabrication  errors.


ASSUMPTIONS:

     • 'Assume four levels of inspection/testing:
        1)  none
        2)  low—visual  inspection
        3)  medium—visual inspection and  weld  testing
        4)  high—visual  inspection, weld  testing,  and  tightness  testing


SOURCES:

     0  Best engineering  judgment based on human error  probabilities listed in
        Nuclear Regulatory Commission, Reactor  Safety Study  (1975)


PROBABILITY (by inspection level):

     None   -1.00
     Low    -"0.50
     Medium - 0.25
     High   - 0.05


PROBABILITY DISTRIBUTION:  Binomial


USER INPUTS:

    •  Level of inspection or testing.
                                     A-81

-------
                             HAZARDOUS WASTE TANKS


LABEL:  LIFDEF (1,4)


FAILURE:  Off-spec materials used In construction.


ASSUMPTIONS:

     t  Use of poor-grade materials would accelerate the onset of
        .various leaks and ruptures.  Since our probability distributions
        for these events are based on empirical data (the API/SCS
        survey) we assume that these probability distributions already
        account for off-spec materials.
                                     A-82

-------
                             HAZARDOUS WASTE TANKS


LABEL:  LIFDEF (I,5)A


FAILURE:  Tank damaged during installation.


SOURCES:

     •  Best engineering judgment

        *
PROBABILITY:  2 x 10'2


PROBABILITY DISTRIBUTION:  Binomial


VOLUME:

     t  We assume that the damage is similar to a seam leak.

     t  The volume will therefore be identical  to volume loss from seam leaks
        (see event T1124).  Leakage will begin in year 1.


VARIATIONS:'.  ,
              tf
     •  Concrete and Stainless Steel Tanks.  Since steel and stainless steel are
        approximately the same strength, we assume that stainless steel is just
        as vulnerable to Installation damage as is steel.  Based on conver-
        sations with concrete contractors, we assume that concrete also has a 2%
        chance of cracking due to improper installation.

     t  Fiberglass tanks.  Fiberglass tanks are twice as likely to rupture as
        are steel tanks (see event T1124).  We therefore assume that they are
        also twice as vulnerable to installation damage.
COMMENTS:
        Due to a transcription error, we used a value of .03 for the installa-
        tion damage probability for steel, sta-inless steel, and concrete.  This
        error did not substantially alter our results.  It will be corrected in
        subsequent versions of the model.
                                      A-83

-------
                             HAZARDOUS WASTE TANKS


LABEL:  LIFDEF (1,5)8


FAILURE:  Underground piping damaged during Installation.


SOURCES:  Best engineering judgment


PROBABILITY:  1 x 10'2


PROBABILITY DISTRIBUTION:  Binomial


VOLUME:

     •  We assume that the damage is similar to that from a pipe rupture.

     t  The leak rate will therefore be Identical to that from-event Bll.  Loss-
        will begin in year 1.


VARIATIONS:

     0  Stainless steel pipes.  Since steel and stainless steel  are approxima-
        tely the same strength, we assume that they are equally vulnerable to
        Installation damage.
     •  Fiberglass pipes.  Fiberglass 1s twice as likely to rupture as 1s steel
                   T]
        Installation-damage probability of 2 x 10'
(see event T1124).   We therefore assume that fiberglass pipes have an
                                      A-84

-------
                             HAZARDOUS WASTE TANKS


LABEL:  LIFOEF (I,5)C


FAILURE:  Above-ground piping damaged during installation.


SOURCES:  Best engineering judgment


PROBABILITY:  1 x 10'2
        **

PROBABILITY DISTRIBUTION:  Binomial


VOLUME:

     •  The leak rate is the same as that from an above-ground pipe rupture
        (see event BID.

     •  Leakage will begin in year 1. •


VARIATIONS:

     t  Stainless steel  pipes.  Since steel  and stainless steel  are approxima-
        tely the same strength, we assume that they are equally vulnerable to
        installation damage.

     •  Fiberglass pipes.  Fiberglass is twice as likely to rupture as  is  steel
        (see event T1124).  We therefore assuem that fiberglass pipes have an
        installation-damage probability of 2 x 10'2.
                                      A-85

-------
                             HAZARDOUS WASTE TANKS

LABEL:  LIFDEF (1,5)0

FAILURE:  Welded flange damaged during installation.

SOURCES:
     •  ^Best engineering judgment

PROBABILITY:  2 x 10'2

PROBABILITY DISTRIBUTION:  Binomial

VOLUME:
     •  We assume that the leak rate is the same as that for a welded flange
        leak (event B121).
     t  Leakage will begin in year 1.
                                     A-86

-------
                             HAZARDOUS WASTE TANKS

LABEL:  LIFDEF (I,5)E

FAILURE:  Gasket damaged during installation (or improperly installed).

SOURCES:
     0  Best engineering judgment

PROBABILITY:  1.5 x 10'2

PROBABILITY DISTRIBUTION:  Binomial

VOLUME:
     •  We assume that the leak rate is the same as that for a gasket failure
        (event B122).
     •  Leakage begins In year 1.
                                     A-87

-------
                             HAZARDOUS WASTE TANKS





LABEL:  LIFDEF (1,6) and (1,7)





FAILURE:  Stresses due to settling.





ASSUMPTIONS:  This event is already included in tank and piping ruptures.
                                     A-88

-------
                             HAZARDOUS WASTE TANKS

LABEL:  FLVCNl

FAILURE:  Automatic level controller fails.

ASSUMPTIONS:
     t  Assume that the failure of the controller, the controller settings,  and
        the Impulse lines may each cause controller malfunction

SOURCES:
     •  Anyakora, Engel and Lees, Table V, p. 400

CALCULATIONS:
     •  0.29/yr « 7.9 x 10'Vdy. » controller failure rate (Anyakora,  Erigel  and
        Lees).
     •  0.14/yr « 3.8 x 10~4/dy » controller settings failure rate (Anyakora,
        Engel and Lees).
     •  0.77/yr » 2.1 x 10'3/dy « impulse lines failure rate (Anyakora,  Engel
        and Le^s).
     t  7.9 x lO'4 + 3.8 x lO'4 + 2.1 x 10'3 - 3.3 x 10'3/dy
     t  3.3 x 10-3/dy » 9.4 x 10-2/mo

PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY:  9.4 x 10'2/mo
                                     A-89

-------
                             HAZARDOUS WASTE TANKS


LABEL:  FLVCN2


FAILURE:  Emergency shut-off level controller fails to function.


ASSUMPTIONS:

     •  We assume that the emergency shut-off controller Is inspected monthly.
       A

SOURCES:  See event FLVCN1


CALCULATIONS:

     •  According to event FLVCN1, the probability of automatic level controller
        failure is 9.4 x 10"2/mo.

     •  If the level controller is inspected monthly, then the average failure
        lasts half a month.  The probability that the controller is in a
        failed state at any given time is therefore .5 x 9.4 x 10'S which is
        approximately 5 x 10'2.


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  .OS/Demand
                                    A-90

-------
                             HAZARDOUS WASTE TANKS


LABEL:  LEVIN2


FAILURE:  Emergency shut-off level sensor falls to function.


ASSUMPTIONS:

     0  We assume that the published failure rates include failures in the
        meter, the sensor, and the impulse lines.
       i

     •  We assume that the emergency shut-off level sensor is inspected monthly.


SOURCES:

     •  Anyakora, Engel, and Lees (1971), Table V, p. 400.


CALCULATIONS:

     t  .2/yr « failure rate for a capacitance-type level transducer (Anyakora,
        Engel and Lees).

     •  .2/yr - 5.5 x 10'4/day » 1.6 x 10'2/mo.
           »  .                                         »
     •  If the level controller is Inspected monthly, then the average failure
        lasts half a month.  The probability that the controller is in a
        failed state at any given time is therefore .5 x 1.6 x 10'2, which is
        8 x lO'3.


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  8 x 10'3 .


COMMENTS:

     •  Due to a transcription error, this event's probability was set to the
        probability calculated for event FLVCU2.  The result was a conservative
        error but the number will be changed to the value calculated above in
        future versions of the model.
                                      A-91

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOALARM


FAILURE:  High liquid level alarm system failure.


ASSUMPTIONS:

     •  We assume that the high level alarm Is tested once per year.  Thus, the
       ^average alarm failure will persist for 6 months.


SOURCES:

     •  Lawley (1974), p. 54, note 6.


CALCULATIONS:

     t  0.2/yr » frequency of dangerous high level alarm failures (LawTey).

     •  6 months » duration of average undetected failure.

     •  0.2 (6/12) « 0.1 » fractional dead time for high level alarm.
           i                                    «

PROBABILITY: ".I/demand


PROBABILITY DISTRIBUTION:  Binomial


COMMENTS:

     •  The same failure probability applies to leak detectors in vaulted tanks
        or interstitial alarms in double-walled tanks or pipes.

     •  We also model Interstitial alarms according to this probability distri-
        bution.  Many of these alarms, however, have status lights which can be
        checked at any desired frequency.  Thus,  if status is checked conscien-
        tiously, the per-demand failure probability may be considerably lower.
        Our value of 102 is therefore a conservative estimate.
                                     A-92

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOFILL


FAILURE:  Tank 1s to be filled nearly to capacity.


ASSUMPTIONS:

     •  Treatment tanks are always filled to their operating capacities.
        ^

     0  We assume that pump-out schedules for storage or accumulation tanks
        generally allow sufficient margin for error that the tank is not filled
        completely to capacity unless something Interferes with the normal pump-
        out schedule or there is an unexpected upsurge In the generation of
        waste.  We conservatively assume that this happens once per year.


CALCULATIONS:

     •  1 per year » .0027 per day « 1 - X1-.0027)30 • .079 per month.


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  '•

     t  1.00 for treatment tanks.

     t  .079/mo. for storage or accumulation tanks.
                                     A-93

-------
                             HAZARDOUS WASTE TANKS

LABEL:  MOLEVIN

FAILURE:  Level indicator malfunction results in attempted overfill.

ASSUMPTIONS:
     •  We assume that the published failure rates include failures in the
        meter, the sensor, and the impulse lines.
      *
     •  We assume that a fault is detected and repaired after one faulty
        transfer.

SOURCES:
     •  Anyakora, Engel, and Lees (1971), Table V, p. 400.

CALCULATIONS:
     •  .22/yr » failure rate for a capacitance-type level transducer (Anyakara,
        Engel, and Lees).
     •  .22 x .5 » ,1/yr « rate of overfill events due to level transducer
        failure.
     •  .2/yr - 5.5 x 10'4/day « l-(l-5:5 x 10-4)30/mo.

PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY: .16/mo.

VOLUME:
     •  Assume that the overflow consists of between 0 and 100X of a batch.
        Thus Q - FNU(0, volume of one batch).

USER INPUTS:
     0  Volume transferred per batch
      •
COMMENT:
     •  Due to a round-off error our model uses a probability of .15/mo.  The
        difference between this and the correct value is not significant.
                                     A-94

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOPMCE


FAILURE:  Outlet pump fails to start on demand (extreme environment)


ASSUMPTIONS:

     t  75% of pump failures are failures to start.  The remaining 25% are
       ^failures to run under event MOPMOE


SOURCES:

     t  Nuclear Regulatory Commission, Reactor Safety Study (1975),
        Table III 2-3.

     •  Southwest Research Institute (1982).

     •  Henley and Kumamoto (1981), Figure 6.7, p. 278.


CALCULATIONS:

     0  1 x, 10*4 to 1 x 10*3 per operating hour » probability of pump failure
        in. extreme environment (Reactor Safety Study)

     •  The geometric mean of this range of values « 3 x 10~4 per operating
        hour.

     •  (3 x 10-4/hr)(.75) « 2.25 x 10'4 per operating hour » probability that
        pump fails to run.

     •  We convert this per-hour probability into a per-demand probability.  We
        do this by noting that the pump must have started properly when the pre-
        vious batch drained.  Otherwise, the operator would have noticed the
        failure and repaired it.  Thus, if the pump fails, it does so during the
        fill time for the present batch.  The probability of this is given by:

                          (2 x 10-4 x fill time)


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  (2 x 10'4 x fill time per batch) per demand^
                                     A-95

-------
USER INPUTS:
     •  The per-batch fill time

COMMENTS:
     •  For continuous systems, the fill time 1s the entire operating day.
     •  Due to a round-off error our model uses a probability of 3 x 10-* x
        fill time per batch.  The difference between this and the correct value
        Is not significant.
                                    A-96

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOPMCN


FAILURE:  Outlet pump fails to start on demand (normal environment)


ASSUMPTIONS:

     t  75X of pump failures are failure to start.  The remaining 25% are
        failures to run under event MOPMON.


SOURCES:

     0  Nuclear Regulatory Commission, Reactor Safety Study (1975),
        Table III 2-3.

     •  Southwest Research Institute (1982).

     •  Henley and Kumamato (1981), Figure 6.7, p. 278.


CALCULATIONS:

     •  1 x 10'* to 1 x 10~5 per operating hour « probability of pump failure
        in normal environment (Reactor Safety Study).

     •  The geometric mean of this range of values » 3 x 10"^ per operating
        hour.

     t  (3 x 10'6/hr)(.75) « 2.25 x 10"6 per operating hour « probability that
        pump fails to run.

     •  We convert this per-hour probability into a per-demand probability.
        We do this by noting that the pump must have started properly when the
        previous batch drained.  Otherwise, the operator would have noticed the
        failure and repaired it.  Thus, if the pump fails, it does so during
        the fill time for the present batch.  The probability of this is given
        by:

                         (2 x 10'6 x fill time)


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  (2 x 10'6 x fill time per batch) per demand.
                                     A-97

-------
USER INPUTS:

     •  The per-batch fill time


COMMENTS:

     •  For continuous systems, the fill time Is the entire operating day.

     •  Due to a round-off error our model uses a probability of 3 x 10-6 x f
        time p.er batch.  The difference between this and the correct value is
       -not significant.
                                    A-98

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOPMOE


FAILURE:  Pump fails 1n "on" position, preventing emergency shut-off (extreme
          environment).


ASSUMPTIONS:

     • " Assume that failure of the solenoid or the controller can cause a pump to
        fail in the open position.  We assume that these subcomponent failures
        make up approximately 2556 of all pump malfunctions.


SOURCES:

     •  Nuclear Regulatory Commission, Reactor Safety Study (1975).

     •  Southwest Research Institute.(1982).

     •  Henley and Kumamoto (1981), Figure 6.7, p. 278-.
CALCULATIONS:

        1 x 1
        extreme environment (Reactor Safety Study).

        The g
        hour.
•  1 x 10"  to 1 x 10   per operating hour « probability of pump failure in


•  The geometric mean of this range of values • 3 x 10   per operating
     •  (3 x 10"4/hr) (.25) - 7.5 x 10"5 per operating hour » probability that
        the pump fails in the "on" position.

     •  We need to convert this per-hour probability into a per-demand probabi-
        lity.  We do this by noting that the pump must have shut off properly
        after the previous batch finished filling.  Otherwise, the operator
        would have noticed the failure and repaired it.  Thus, if the pump fails
        It does so during the fill time for the present batch.  The probability
        of this 1s given by:

                           (7.5 x 10"5 x fill time)
PROBABILITY DISTRIBUTION:  Binomial
      V


PROBABILITY:  (7.5 x 10"5 x fill time per batch) per demand
                                     A-99

-------
USER INPUTS:
     t  The per-batch fill time

COMMENTS:
     •  For continuous systems, the fill time is the entire operating day.
                                    A-100

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOPMON


FAILURE:  Pump fails 1n "on" position, preventing emergency shut-off (normal
          environment).


ASSUMPTIONS:

     •  Assume that failure of the solenoid or the controller can cause a pump
        to fail in the open position.  We assume that these subcomponent
        failures make up approximately 25X of all pump malfunctions.


SOURCES:

     •  Nuclear Regulatory Commission, Reactor Safety Study (1975).

     t  Southwest Research Institute (1982).

     •  Henley and Komamoto (1981), Figure 6.7, p. 278.


CALCULATIONS:

     •  1 x. 10;"7 to 1 x 10"^ per operating hour » probability of pump failure
        in normal environment (Reactor Safety Study).

     •  The geometric mean of this range of values » 3 x 10~6 per operating
        hour.

     t  (3 x W6/hr)(.25) - 7.5 x 10'7 per operating hour « probability that
        the pump fails In the "on" position.

     0  We convert this per-hour probability Into a per-demand probability in
        the same way that we do for pumps in an extreme environment (event
        MOPMOE).  Thus, the probability 1s given by:

                        (7.5 x 10-7 x fill time)


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  (7.5 x 10*7 x fill time per batch) per demand
                                     A-101

-------
USER INPUTS:
     0  The per-batch fill time

COMMENTS:
     0  For continuous system, the fill time is the entire operating day.
                                    A-102--

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOVLCE


FAILURE:  Outlet valve fails in the closed position, preventing emergency shut-
          off (extreme environment).


ASSUMPTIONS:

     0  We assume that 50% of valve failures occur in the closed position.


SOURCES:

     •  Anyakora, Engel, and Lees (1971), Table V.

     •  Southwest Research Institute (1982), Table 2, p. 32.

     •  Henley and Kumamoto (1981), Figure 6.7, p. 278.

     •  Nuclear Regulatory Commission, Reactor Safety Study, (1975), Table III-
        2-3.


CALCULATIONS':

     0  0.60/yr * expected number of control valve failures in a normal
        environment.  (Anyakora, Engel and Lees (1971).

     0  Valve failures are 10 times more common in extreme than normal environ-
        ments.  (Source, best engineering judgment based on Henley and Kumamoto
        (1981) and Reactor Safety Study). Therefore, 6.0/yr » the expected
        number of control valve failures in an extreme environment.

     0  If SOX of these failures occur in the closed position, then 3.0/yr = the
        expected number of control valves failing in the closed position.

     0  3.0/yr « 3.4 x 10'4/hr

     0  Since the valve must have been functional at the time the previous batch
        drained, this event can only occur If failure occurs during the time
        when the tank 1s being filled.  Thus, the per demand failure probability
        1s given by:

                      (3.4 x lO'4) x (fill time per batch)

PROBABILITY DISTRIBUTION:  Binomial (per demand)
                                      A-103

-------
PROBABILITY:  (3.4 x 10"4) x (fill time per batch) per demand
USER INPUTS:
     •  Fill time per batch
     •  Number of operating hours per day

COMMENTS:
     •  For continuous systems, the fill time is the entire operating day.
                                      A-104

-------
                             HAZARDOUS WASTE TANKS

LABEL:  MOVLCN

FAILURE:  Outlet valve fails in the closed position, preventing emergency shut-
          off (normal environment).

ASSUMPTIONS:
     •  Failures are only 10X as likely in normal as extreme environments
        (see sources cited under event MOVLCE).
SOURCES:  See event MOVLCE

CALCULATIONS:
     •  The per demand probability of failure in an extreme environment Is
        3.4 x 10'4 x (fill time per batch).. See event MOVLCE.
PROBABILITY DISTRIBUTION:  Binomial (per demand)

PROBABILITY: "(3.4 x 10"4) x (fill time per batch)

USER INPUTS:
    _ •  Fill time per batch
     •  Number of operating hours per day

COMMENTS:
     t  For continuous systems, the fill time is the entire operating day.
                                      A-lOi

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOVLOE


FAILURE:  Inlet valve falls 1n the open position, preventing emergency shut-off
          (extreme environment).


ASSUMPTIONS:
       *
     t  Assume that SOX of valve failures occur in the open position.


SOURCES:

     •  Anyakora, Engel, and Lees (1971), Table V.

     •  Southwest Research Institute (1982), Table 2, p. 32.

     •  Henley and Kumamoto (1981), Figure 6.7, p. 278.

     •  Nuclear Regulatory Commission. Reactor Safety Study (1975) Table III-
        2-3.


CALCULATIONS:''

     •  0.60/yr « expected number of control valve failures in a normal
        environment (Anyakora, Engel and Lees (1971)).

     •  Valve failures are 10 times more common In extreme than normal environ-
        ments.  (Source, best engineering judgment based on Henley and Kumamoto
        (1981) and Reactor Safety Study).  Therefore, 6.0/yr « the expected
        number of control valve failures in an extreme environment.

     •  If SOX of these failures occur 1n the open position, then 3.0/yr » the
        expected number of control valves failing in the open position.

     •  3.0/yr - 3.4 x 10'4/hr.

     •  Since the valve must have been functional at the time the batch began to
        fill, this event can only occur if failure occurs during the time when
        the tank is being filled.  Thus, the per demand failure probability 1s
        (3.4 x 10'4) x (fill time per batch).

PROBABILITY DISTRIBUTION:  Binomial (per demand)


PROBABILITY:  (3.4 x 10"4) x (fill time per batch)
                                      A-106

-------
USER INPUTS:

     0  Fill time per batch

     •  Number of operating hours per day


COMMENTS:

     t ,For continuous systems, the fill time 1s the entire operating day.

     •  This event can also cause an attempted overfill.  We only model this
        form of attempted overfill for gravity-fed systems, however, because
        for pump-fed systems, the failure can be remedied by shutting off the
        pump.  In theory, the pump could also fail 1n the "on" position, causing
        an attempted overfill even for pump-fed systems, but such simultaneous
        failure 1s extremely unlikely, and 1s overshadowed by the other types
        of failure (e.g. operator error) which are more likely to cause attempt-
        ed overflows.
                                      A-107

-------
                             HAZARDOUS WASTE TANKS


LABEL:  MOVLON


FAILURE:  Inlet valve falls 1n the open position, preventing emergency shut-off
          (normal environment).


ASSUMPTIONS:

     •  Failures are only 10X as likely 1n normal as extreme environments
        (see sources cited under event MOVLOE).


SOURCES:  See event MOVLOE


CALCULATIONS:

     •  The per demand probability of failure 1n an extreme environment 1s
        (3.4 x 10'4) x (fill time per batch).  See event MOVLOE.  	

PROBABILITY DISTRIBUTION:  Binomial (per demand)


PROBABILITY:'' (3.4 x 10"5) x (fill time per batch)


USER INPUTS:

     •  Fill time per batch

     •  Number of operating hours per day


COMMENTS:

     0  For continuous systems, the fill time 1s the entire operating day.

     •  This event can also cause an attempted overfill.  We only model this
        form of attempted overfill for gravity-fed systems, however, because
        for pump-fed systems, the failure can be remedied by shutting off the
        pump.  In theory, the pump could also fall 1n the "on" position,
        causing an attempted overfill even for pump-fed systems, but such
     .  simultaneous failure Is extremely unlikely, and 1s overshadowed by the
     .   other types of failure (e.g. operator error) which are more likely to
        cause attempted overflows.
                                      A-108

-------
                             HAZARDOUS WASTE TANKS

LABEL:  OFTROM

FAILURE:  Operator falls to respond to high level alarm.

ASSUMPTIONS:
     • "Failure may be due to failure to hear alarm, failure to take corrective
        action, or inability to take corrective action.
SOURCES:
     •  Lawley (1974), p. 54, note 7.

PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY:   3 x 10~2/demand
                                      A-109

-------
                             HAZARDOUS WASTE TANKS


LABEL:  OFTRCOM


FAILURE:  Operator erroneously responds to high level alarm.


ASSUMPTIONS:

     •* At the time when the alarm first sounds, the operator feels no sense
        of panic.  He responds 1n a routine manner.


SOURCES:

     •  Nuclear Regulatory Commission, Reactor Safety Study (1975).


CALCULATIONS:

     •  3 x 10'3/demand « probability of human error of commission (selecting
        wrong switch, etc.),  Reactor Safety Study. Table III 6-1.


PROBABILITY DISTRIBUTION:  Binomial


PROBABILITY:  3 x 10'3/demand


COMMENTS:

     •  If the operator panics, the probability of error will be much higher.
        The Reactor Safety Study gives an error rate of 20-30X for trained
        personnel under high stress levels where dangerous activities are
        occurring rapidly.
                                     A-110

-------
                             HAZARDOUS WASTE TANKS

LABEL:  OPCOMM

FAILURE:  Operator error in batch start-up leads to attempted overfill.

ASSUMPTIONS:
     •  Operate action is required to Initiate the transfer of fluid at the
        start of each batch.  Mistakes may result in an attempt to transfer
        *too much fluid.
     •  Operator action is also necessary whenever a continuous process is
        started up.  We assume that this occurs once per operating day.
SOURCES:
     •  Nuclear Regulatory Commission (1975)

CALCULATIONS:
     •  3 x 10"3/demand « estimated rate of human errors of commission (e.g.
        selecting a wrong switch).  Source:  Nuclear Regulatory Commission.
     •  Let n-be the number of batches per day.  (Let n « 1 for continuous
        systems).
     t  Then the probability that the operator makes no errors is 1-3 x 10~3
        per batch, or
                                (l-.003)30n
        per month.  The probability of 1 or more errors is
                               l-(l-.003)30n
        per month.
     •  If n » 1 this value is .086/mo.

PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY:
     •  1 - .99730n/mo for batch systems
                                     A-lll

-------
     •  .086/mo for continuous systems

USER INPUTS:
     0  Number of batches per day (n)
                                     A-112

-------
                             HAZARDOUS WASTE TANKS

LABEL:  OPVLOE

FAILURE:  Inlet valve falls to close, causing overflow (extreme/environment).

ASSUMPTIONS:
     •  This event only applies for automatic valves.  Manual valves are very
        unlikely to fall since they have no automated components.

SOURCES:  See event MOVLOE.

CALCULATIONS:
    . •  According to event MOVLOE, the probability of such an event Is
        3.4 x 10'4/hr.

     t  The probability of failure during any given batch 1s therefore
        (Tb).(3.4 x 10-4) where Tb 1s the fill time per batch.

     0  The probability of failure during any given month 1s

                             l-[l-Tb<3.4 x 10-4)]nbm

        where nb 1s the number of batches per day, and m is the number of
        operating days per month.  For continuous systems, nb 1s 1 and Tb Is the
       • entire operating day.

PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY:  1- fl - Tb(3.4 x 10'4)] nbm

USER INPUTS:
     0* Fill-time per batch (Tb)
     0  Number of batches per day (nb)
     0  Number of operating day per month (m)
                                      A-113

-------
                             HAZARDOUS WASTE TANKS

LABEL:  OPVLON

FAILURE:  Inlet valve falls to close, causing overflow (normal environment)

ASSUMPTIONS:
     •  Yhls event only applies for automatic valves.  Manual valves are very
        unlikely to fall since they have no automated components.

SOURCES:  See event MOVLON

CALCULATIONS:
     •  According to event MOVLON, the probability of such an event 1s 3.4 x
        10-5/hr.

     0  The probability of failure during any given failure 1s therefore
        (Tb>(3.4 x 10-5) where *Tb Is the fill time per batch.

     0 The probability of failure during any given month 1s

                               1- [l-Tb(3.4 x 10-4)]nbm

        where nb 1s the number of batches per day, and m 1s the number of
        operating days per month.  For continuous systems, nb 1s 1 and Tb 1s the
        entire operating day.

PROBABILITY DISTRIBUTION:  Binomial

PROBABILITY:  1- [l-Tb(3.4 x 10'5)] nbm

USER INPUTS:
     0 'Fill-time per batch (Tb)
     0  Number of batches per day (nb)
     0  Number of operating days per month (m)
                                      A-114

-------
                             HAZARDOUS WASTE TANKS


LABEL:  PADINSF, VLTINSF, CRBINSF


FAILURE:  Visual Inspection falls to detect secondary containment failure
          (pad, yault, curb)

        \
ASSUMPTIONS:

     •  We assume that this visual Inspection 1s a passive "walk-around."

     •  We assume that the visual Inspection 1s Infrequent enough that 1t does
        not become monotonously routine to the operator.


SOURCES:

     t  Best engineering judgment based on Nuclear Regulatory Commission,
        Reactor Safety Study (1975), and Lawley (1974).


PROBABILITY'. DISTRIBUTION:  Binomial
             f,.

PROBABILITY:  .1


COMMENTS:

     t  For simplicity, we assume that all secondary-containment failures occur
        at the beginning of the month and that they are repaired Immediately
        after detection.  Thus, the secondary-containment system 1s 1n a failed
        state for a minimum of one complete month.  A* similar result would be
        obtained by assuming that cracks occur 1n the middle of the Inspection
        cycle, and that repair takes two weeks.

     •  We assume that all secondary-containment Inspections are statistically
        Independent events.  Thus a failure to detect a breach 1n a vault does
        not Influence the probability that the Inspector will also fall to
        detect-a breach in a pad or curb.  Similarly, a failure to detect a
        fault 1n one month does not change the probability that 1t will be
        detected during the next Inspection-.
                                      A-115

-------
         APPENDIX B
Statistical Analysis of PACE
    Tank Corrosion Data

-------
                                  INTRODUCTION

A group of Canadian oil companies, working through the Petroleum Association for
Conservation of the Canadian Environment (PACE) have compiled data on 300 under-
ground gasoline storage tanks.*  It is unclear what sampling techniques were
used to select the 300 tanks, but it appears that the intent was to obtain a
"snapshot" of the contemporary situation.  It is not clear whether the data dis-
tinguishes interior and exterior corrosion, but since the intent was to deter-
mine the effect of soil variations on tank leakage, it must be assumed that the -
survey focused on exterior corrosion.  The Canadian survey therefore represents
raw data distinct from the API tank leak survey? and independent from either
Warren Rogers'3 preliminary or revised statistical model.

In raw form, these data are pr.esentedMn Figures 1 and 2.*  They consist of
scatter diagrams of tank age and "Soil Aggressiveness Values" (SAV) for 108
leaking tanks and 192 non-leaking tanks.  Each dot on the scatter diagrams
represents one or more tanks, with overlapping points tallied by the small
numerals adjacent to the relevant dots.  Numerical listings of all 300 points
are presented in Tables 1, 2, 3, and 4.

SAV is calculated according to the formula depicted in Figure 3.  It is designed
to Incorporate soil resistivity, pH, moisture content, and sulfldes, as well as
the effect of variations in resistivity and pH over the tank installation site.
The resultant numerical index 1s designed to present a cardinal ranking of soil
^This data 1s discussed in PACE, "Underground Tank Systems:  Review of State of
the Art and Guidelines," PACE report No. 82-3. Ottawa (1983).
2American Petroleum Institute, Tank and Piping Leak Survey, 1977 to 1980.
^Warren Rogers Associates, Inc., "Prediction of Leaks 1n Unprotected Steel
Storage.Tanks," Included in correspondence package from Betsy Tarn,  EPA to Chris
Lough, PRA.
4These data were provided by Esso Petroleum, Canada.  Esso Canada participated in
the PACE study.
                                    B-l

-------
00
O£

s
O H-



M 00

5^ O
O


LU
          40
          35
          30
          25
   ~     20
          15
          10
                                     MliUKt  1




                                   LEAKING TANK CHART




                              AGE  VS.  SOIL  AGGRESSIVENESS











































-



















































,












































,































































2




t


































'




t







































3




, ,




,3







































2
i

































L










k i



I





















,











,2




»-j_
I i
,4,
I J



^






































5,
2


<
,3
2


h










"
























,2,


*3
'
>g

... J

































>,




21



,2

»
1












__














M








2
»,_


2


'








k
































4
(2





>, ,


f
































»*




























'










5









,2
'

2
4







•






















































•












'

























j
1
1
,
1
(






















































































1'-
!












































i
*
i
i
i































•




"



I i
I

1

I
,
1



1


j

1 * '


1
i
1


i
i
1
!

I

i
i
1
i
i


i
i
•
i
t
i
t
'
. .j._
t
;
' 1
•
t
                                      10 B-2      15   .


                                SOIL AGGRESSIVENESS VALUE
20
25

-------
                      FIGURE 2
  TASK FORCE  WON-LEAKING  TANK  CHART
   35
   30
CO
c:
<
LJ
 »

sx
U.
o

UJ
o
25
   20
15
    10

1
1 .
I
3



•
•


4
' 2<
2 •

3 2 '
• •
• •
. .2 •

2,
•?-1
• *•
•
"*
• • • • 4
• 4
2
<

•


2-

5
*
<
2
••-id
1 • . > '
• •" j I •
' 4
3

, *3
* r 2
*
•
•
* 9 2
t
<
»
r • • .
* . .2-
»2 • • <
>2 *2

> • -t






•
8
3 »
2
6:

3 •
• i
•

• •
3
* * 2
•3 •

3»3 3
- .4I.
^


i ^^
0
,







4
*
•
•
; 4


> 4^
2


3

•

•
i










•
• .

'

•
' • ; .


•










. *


.



             5       10        15     .  20

              SOIL AGGRESSIVENESS VALUE

                      B-3
                                          25
                                              November 15

-------
                                                               TABLE 1
Boil Aggressiveness Values (SAV) and ages for leaking tanks.
          Leaking tanks
                           Leaking tanks
    BAV
Ag«
Age»SAV
8AV
Age
Age*BAV
s
3
1O
9
13
4
4'
6
6
6
16
10
14
14
3
6
9
10
1O
7
7
10
10
10
6
6 '
6
9
a
14
9
12
10
9
9
9
9
11
18
15
10
9
9
17
11
11
11
9
12
23 •
7
19
6
2O
20
19
19
19
6
11
B
a
39
2O
19
14
14
20
20
13
19
19
29
29
29
3O
19
11
18
14
17
19
19
19
19
16
10
12
18
20
20
II
17
17
17
21
60
69
70
79
78
80
BO
9O
90
9O
96
110
112
112
117
120
139
140
140
140
140
190
130
19O
190
190
190
ISO
192
194
162
168
17O
171
171
171
171
176
180
1BO
ISO
ISO
18O
187
187
187
187
189
     Leaking tanks

BAV      Age    Age*SAV
9
10
10
16
16
12
12
11
9
10
1O
10
10
10
14
11
11 •
10
12
13
13
9
9
14
14
16
16
12
12
10
a
14
14
14
14
IS
IS
IS
13
16
13
13
12
10
21
IS
IS
12
21
19
19
12
12
16
16
18
22
20
20
20
20
20
IS
20
20
22
19
IB
IB
26
26
17
17
15
IS
20
2O
24
30
IB
18
IB
18
17
17
17
20
17
21
21
23
29
14
2O
20
25
189
19O
190
192
192
192
192
198
198
200
2OO
2OO
200
200
21O
220
220
220
228
234
234
234
234
238
238
240
240
240
24O
240
24O
232
232
252
252
25S
255
255
260
272
273
273
276
29O
294
3OO
30O
.3OO
                                                                      10
                                                                      16
                                                                      12
                                                                      16
                                                                       9
                                                                      16
                                                                      16
                                                                      16
                                                                      16
                                                                      16
                                                                      16
                                                                      14
                                                                 30
                                                                 2O
                                                                 2B
                                                                 22
                                                                 40
                                                                 29
                                                                 29
                                                                 29
                                                                 29
                                                                 29
                                                                 26
                                                                 30
                                                                  3OO
                                                                  32O
                                                                  336
                                                                  352
                                                                  360
                                                                  400
                                                                  4OO
                                                                  4OO
                                                                  4OO
                                                                  400
                                                                  416
                                                                  420

-------
                                                                  TABLE 1
   Soil Aggressiveness Values  (SAV) and ages  for  leaking tanks.
             Leaking tank*
                           Leaking tank*
        BAV
CO
i
ui
Agi
Age»SAV
BAV
s
3
to
5
13
4
4
6
6
6
16
to
14
14
3
6
9
10
10
7
7
10
10
10
6
6
6
S
a
14
9
12
10
9
9
9
9
11
18
IS
10
9
9
17
11
11
11
9
12
23
7
IS
6
2O
2O
IS
IS
IS
6
11
a
a
39
20
IS
14
14
20
2O
IS
IS
IS
23
23
23
30
19
11
18
14
17
19
19
19
19
16
10
12
18
2O
2O
11
17
17
17
21
60
69
70
73
78
80
BO
9O
9O
90
96
110
112
112
117
120
133
140
140
140
140
130
130
150
ISO
ISO
ISO
ISO
132
IS4
162
168
I7O
171
171
171
171
176
ISO
IBO
IBO
18O
ISO
187
187
187
187
189
Age
Age*BAV
     Leaking tanks

BAV      Age    Age»6AV
9
10
to
16
16
12
12
11
9
10
10
10
to
10
14
11
11
10
12
13
13
9
9
14
14
16
16
12
12
10
a
14
14
14
14
13
IS
IS
13
16
13
13
12
10
21
IS
IS
12
21
19
19
12
12
16
16
18
22
20
20
20
20
20
IS
20
2O
22
19
. 18
IB
26
26
17
17
IS
IS
20
2O
. 24
3O
18
IB
IB
18
17
17
17
2O
17
21
21
23
29
14
20
20
25
189
190
19O
192
192
192
192
198
198
20O
200
2OO
200
200
210
22O
220
22O
228
234
234
234
234
238
238
240
240
240
240
240
240
232
232
232
2S2
235
233
233
26O
272
273
273
276
29O
294
300
300
3OO
                                                                          10
                                                                          16
                                                                          12
                                                                          16
                                                                          9
                                                                          16
                                                                          16
                                                                          16
                                                                          16
                                                                          16
                                                                          16
                                                                          14
                                                                  30
                                                                  20
                                                                  28
                                                                  22
                                                                  40
                                                                  23
                                                                  23
                                                                  23
                                                                  23
                                                                  23
                                                                  26
                                                                  30
                                                                   3OO
                                                                   32O
                                                                   336
                                                                   332
                                                                  •360
                                                                   400
                                                                   4OO
                                                                   4OO
                                                                   4OO
                                                                   400
                                                                   416
                                                                   420

-------
  Soil Aggressivmnmmm Value* 
-------
Boil Aggre«*ivenese Values  and ages for leaking tank..
                                                            TABLE 3
        Leaking tank*
                 Leaking tank*
                                                                 Leaking  tank*
   BAV
            Age
Age«8AV
SAV
         Agi
                                               Age*SAV
                                                            BAV
13
16
1O
14
14
18
to
14
17
9 •
15
16
16
1O
1O
12
21
9
6
6
6
9
1O
1O
10
14
16
16
11
12
12
1O
11
11
11
14
14
19
19
19
16
9
1O
11
13
13
14
14
6
6
7
a
8
10
11
11
11
12
12
12
12
14
14
14
14
19
19
19
19
19
19
19
19
19
19
19
16
16
16
17
17
17
17
17
17
17
17
17
17
18
18
IB
IB
18
18
18
78
96
70
112
112
14
14
a
9
9
18O 9
11O 9
194 10
187 1O
6O 12
180 4
192 4
192 6
14O 7
140 7
168 9
294 9
79 1O
9O 1O
90 10
9O 1O
139
190
19O
190
21O
240
24O
176
192
192
170
187
187
187
238
10
11
11
12
12
13
13
19
16
9
9
13
13
9
1O
238 16
239 3
299 12
299 10
272 6
162 6
180 6
198 12
234 16
234 16
232 16
232 16
18
IB
19
19
19
19
19
19
19
19
20
2O
20
2O
20
20
20
20
20
20
2O
20
20
2O
20
2O
20
20
2O
20
21
21
21
21
22
22
22
23
23
24
23
23
23
23'
23
25
25
25
252
252
152
171
171
171
171
19O
19O
228
80
BO
120
140
140
180
ISO
2OO
2OO
2OO
2OO
2OO
220
22O
240
24O
26O
3OO
3OO
32O
189
189
273
273
198
22O
332
69
276
24O
13O
ISO
ISO
3OO
4OO
400
4OO
4OO
                                                                            Age«SAV
16
9
9
16
12
TO
3
8
10
14
3
9
23
26
26
26
28
29
3O
3O
30
3O
39
40
400
234
234
416
336
290
ISO
240
3OO
420
117
360

-------
     Boil  Aggresciveness  Value*  
-------
 Boil Aggr«.«lvene.« Value. (SAV) and age< for non-leaking tanks.
     8AV
Ag.
I
to
         9
         9
         2
         2
         9
         9
         7
         7
        13
         2
         9
         9
         7
         7
        10
        11
         9
         9
         3
        *
        10
        12
        14
        16
        16
         9
         9
         9
        13
        13
        13
        13
        14
        14
        i4
         9
         9
         7
         7
        It
        II
        11
        16
    2
    2
    3
    3
    3
    3
    3
    3
    3
   7
   7
   7
   7
   8
   8
   a
   a
   a
   a
   a
   a
   a
   a
   .9
   9
   9
   9
   9
   9
   9
   9
SAV*Age

     IO
     18
      6
      6
     19
     19
     21
     21
     39
      a
     20
     20
     28
     28
     4O
     44
     29
     29
     18
     24
     24
     24
     72
     28
     28
     70
     84
     98
   112
   112
     4O
    40
    4O
   104
   104
   104
   IO4
   112
   112
   112
    49
    49
    63
    63
    99
   99
    99
   144
                                     SAV
                                              Age    SAVftAge
                                                                      SAV
                                                                               Agi
                                                                       6AV«Age
SAV
2
2
10
to
10
3
9
12
12
12
14
14
17
3
4
9
9
12
13
13
13
16
16
16
2
2
2
4
4
6
9
6
6
6
IO
12
2
2
6
6
6
IO
12
12
12
»?
16
16
10
10
IO
10
IO








12
12
12
12
12
12
12
12
12
12
12
13
13
13
13
13
13
14
14
14
14
14
14
19
19
19
19
19
19
15
15
IS
IS
15
IS
20
20
100
1OO
too
33
99
132
132
132
154
194
187
36
48
108
108
144
156
196
156
192
192
192
26
26
26
92
92
78
7O
84
84
84
140
168
3O
30
90
90
90
150
ISO
ISO
ISO
ISO
24O
24O
3
12
5
. 9
12 '.
12
12
12
19
16
6
8
8
a
9
9
9
9
IO
10
10
14
14 •
14
2
2
2
2
9
9
9
9
9
9
2
2
2
9
9
7
10
IO
10
IO :.
IO
10
IO
10
16
16
17
17
17
17
17
17
17
17
18
IB
18
18
18
18
18
IB
18
IB
18
18
18
18
19
19
19
19
19
19
19
19
19
19
20
20
20
2O
20
20
20
20
20
20
20
20
2O
2O
48
192
85
85
2O4
2O4
204
2O4
255
272
108
144
144
144
162
162
162
162
180
ISO
ISO
292
252
292
38
38
38
38
99
171
171
171
171
171
4O
40
4O
1OO
1OO
140
2OO
20O
• 2OO
20O
2OO
2OO
20O
2OO
                                                                                                             Age    SAV*Age
IO
IO
11
13
13
13
13
13
13
17
17
17
17
9
9
13
13
3
IO
IO
IO
12
16
12
12
12
12
12
12
12
12
9
9
9
9
9
16
16
16
16
6
12
2
2
2
7
8
B
2O
2O
20
2O
20
20
20
2O
20
20
20
2O
20
22
22
22
22
24
24
24
24
24
24
25
25
25
25
25
25
25
25
26
26
26
26
26
26
26
26
26
28
28
30
30
30
30
3O
30
2OO
200
220
26O
26O
26O
26O
26O
26O
340
34O
340
340
198
198
286
286
72
240
240
240
288
384
3OO
3OO
3OO
3OO
3OO
3OO
30O
3OO
234
234
234
234
234
416
416
416
416
168
336
6O
6O
60
210
24O
240

-------
                                                             TABLE  4
Boil Aggre»*ivene«« Value* (SAV)  and age* for non-leaking tanks.
     BAV
       Agi
I
»-•
o
  3
  9
  2
  2
  9
  9
  7
  7
 13
  2
  9
  9
  7
  7
 1O
 11
  9
  9
  3
  4
  4
  4
 12
  4
  4
 1O
 12
 14
 16
 16
 9
 9
 9
13
13
13
13
14
14
14
 9
 9
 7
 7
11
11
11
16
 2
 2
 3
 3
 3
 3
 3
 3
 3
 4
 4
 4
 4
 4
 4
 4
 9
 9
 6
 6
 6
 6
 6
 7
 7
 7
 7
 7
 7
 7
 B
 0
 8
 a
 a
 a
 a
a
a
a
BAV«Age

     10
     18
      6
      6
     19
     19
     21
     21
     39
      a
     20
     20
     28
     28
     40
     44
     29
     29
     18
     24
     24
     24
     72
     28
     28
     70
     84
     98-
    112
    112
     40
     40
     4O
    104
    104
    104
    104
    112
    112
    112
     49
     49
     63
     63
     99
     99
     99
    144
                                     SAV
                                              Agi
                                              SAVttAga
                                                                      BAV
                                                                               Agi
                                                                              SAV«Age
                                                                                   SAV
                                                                                            Age
                                                                                                                     SAV*Age
2
2
1O
10
10
3
9
12
12
12
14
14
17
3
4
9
9
12
13
13
13
16
16
16
2
2
2
4
4
6
9
6
6
6
10
12
2
2
6
6
6
1O
12
12
12
12
16
16
to
10
10
10
10
11
II
11
11
11
11
11
11
12
12
12
12
12
12
12
12
12
12
12
13
13
13
13
13
13
14
14
14
14
14
14
19
19
19
19
19
13
IS
19
13
13
IS
IS
20
• 20
IOO
100
100
33
99
132
132
132
134
134
187
36
48
108
108
144
136
136
136
192
192
192
26
26
26
92
92
78
7O
84
84
84
140
168
30
3O
9O
90
90
ISO
ISO
ISO
180
ISO
240
240
3
12
3
12
12
12
12
. IS
16
6 '
a
8
8
9
9 •
9
9
IO
10
IO
14
14
14
2
2 '
2
2
3 •
9
9
9
9
9
2
2
2
9
3
7
IO
10
10
10 .
10
10
10
10
16
16
17
17
17
17
17
17
17
17
18
18
18
18
IB
IB
18
18
18
18
18
18
IB
IB
19
19
19
19
19
19
19
19
19
19
2O
20
20
20
20
2O
2O
2O
20
20
20
20
20
20
48
192
as
as
2O4
204
204
204
253
272
1O8
144
144
144
162
162
162
162
ISO
ISO
ISO
232
232
232
38
38
38
38
93
171
171
171
171
171
40
40
40
100
IOO
• 14O
200
20O
2OO.
2OO
2OO
2OO
2OO
2OO
IO
10
11
13
13
13
13
13
13
17
17
17
17
9
9
13
13
3
10
10
10
12
16
12
12
12
12
12
12
12
12
9
9
9
9
9
16
16
16
16
6
12
2
2
2
7
a
8
20
20
20
2O
20
20
20
2O
20
20
20
20
20
22
22
22
22
24
24
24
24
24
24
23
25
23
23
23
23
25
23
26
26
26
26
26
26
26
26
26
28
28
30
3O
30
30
30
3O
200
2OO
2 2O
26O
26O
26O
26O
26O
260
340
340
340
340
198
198
286
286
72
240
240
24O
288
384
300
3OO
3OO
30O
3OO
3OO
300
300
234
234
234
234
234
416
416
416
416
168
336
6O
6O
60
210
24O
24O

-------
                         FIGURE 3:   COMPUTATION OF SAV
  I  BASIC CHARACTERISTICS
     o  Soil  Resistivity
     o  Soil  pH
     o  Soil  Moisture
          . <300
   300 -  1,000
 1,000 -  2,000
 2,000 -  5,000
 5,000 - 10,000
10,000 • 25,000
        >25,000

3 -
5 -
6.5 -
7.5 -

Saturated
Damp
Dry
<3
5
6.5
7.5
9
>9



                 POINTS

                   12
                   10
                    8  .
                    6
                    3
                    1
                    0

                    8
                    6
                    4
                    2
                    1
                    0

                    3
                    2
                    0
 II   DIFFERENTIAL CHARACTERISTICS

     o  Resistivity
        (ratio of extremes)
        Soil pH
        (Difference 1n
        pH Value)
1.5
  0
              5
              3
              3
              3
              3
              1.5
3
2
1
0
2
1
0
III  SULFIDES
                                                Positive
                                                Negative
                            4
                            0
                                    B-ll

-------
corrosivities; that is, a soil with an SAV of 2x is expected  to be  (on  average)
twice as corrosive as a soil with an SAV of x.

On first consideration, the Canadian data reveals one important fact:   there is
a lot of scatter.  But it 1s also clear that non-leaking tanks are clustered
somehwat closer to the lower left-hand corner of the diagram  than are leaking
tanks.  Thus, age, SAV, or the combination of the two does appear to have some
predictive effect on the probability of leakage.

This tentative conclusion can be verified by simple statistical analyses.  Using
chi-squared techniques, age, SAV and SAV x age can be tested  for statistically
significant effects on the probability of leakage.  In addition, SAV can be
tested to determine if 1t has any effect Independent from the effect of age.
The following section will describe each of these analyses.

                               STATISTICAL TESTS

1.  SAV x Age

PACE asserts,that the product of SAV and tank age 1s an appropriate predictor
for tank leakage.  This assertion makes Intuitive sense, for  1t would appear to
allow for the continuing effects of various types of soils.   It 1s also born out
by the data, as 1s Indicated by the following contingency table:
                                    B-12

-------
                                    TABLE 5
    SAV x Age


      0-49

     50 ~  99

    100 - 149

    ISO - 199

    200 • 249

    250 - 299

    300 - 399

       >  400
    Number
      of
Leaking Tanks
                               108
  Number
  of Non-
Leakinq Tanks
0 / 16.92
11 / 13.68
10 / 13.32
36 / 24.12
22 / 18.00
14 / 10. 08
8 / 7.92
7 / 3.96
47 / 30.08
27 / 24.32
27 / 23.68
31 / 42.88
28 / 32.00
14 / 17.92
14 / 14.08
4 / 7.04
                                  192
                    47

                    38

                    37

                    67

                    50

                    28

                    22

                    11

                   300
Each cell of this table contains two numbers.  The first represents the observed
observed number of tanks 1n each category; these numbers were obtained from
Tables 1 and 2.  The second number 1n each cell represents the expected number
of tanks falling Into that category 1f SAV x Age had no effect on the probabi-
lity of leakage.  These numbers are computed by multiplying the row total by the
column total and dividing by the grand total (-300).

One of the assumptions underlying the ch1-squared test Is that the sampled size
1s large enough to allow a large-sample approximation.  Often, this assumption
1s expressed as a requirement that there be at least 5 observations In each
cell, but more rigorously, the assumption may be stated as a requirement that
                                    B-13

-------
the expected values be greater than 5 in at least 30% of the 'sample cells.5
This assumption is clearly met.

The test statistic (T) is simply the summation of (x^j - E7-j)2/Ej, over all
cells, where x^j is the number of observations in cell,j and EJJ is the corres-
ponding expected number of observations.  For Table 3, T « 45.11, With (r-1) •
(c-1) • 7 degrees of freedom (r and c are the numbers of rows and columns, res-
pectively).  This 1s highly significant, Indicating that there 1s far less than
a .IX chance that the difference between the SAV x Age distributions of leaking'
and non-leaking tanks 1s random.  SAV x Age 1s therefore a statistically signi-
ficant factor in the differentiation of leaking and non-leaking tanks.  In par-
ticular, 1t appears from Table 5 that SAV x Age has a trichotomous effect.  For
very low values of SAV x Age, there were no observed leaking tanks (the raw data
Indicates that for all leaking tanks SAV x Age > 59).  For Intermediate values
(50 < SAV x Age < 150), approximately 28X were leaking, and for high values (SAV
x Age > 149), approximately 49X were leaking.
                              •            •
2.  Tank Age

The conventional wisdom 1s that age 1s a very poor predictor of .tank leakage.
This, however, 1s an overstatement, as Is evident from the following contingency
table:
5W. J. Conover, Practical NonparametHc Statistics (John Wiley & Sons:  New
York), 1971, p. 15?:
                                    B-14

-------
      Tank
       Age
      (yrs)
      0-4


      5 -  9


     10 - 14


     15 - 19


     20 - 24


     25-29


      > 29  '
                                    TABLE 6
   Number
     of
Leaking Tanks
                                  108
 Number of
Non-Leaking
   Tanks
0 / 5.76
5 / 13.32
12 / 17.28
41 / 31.32
30 / 24.12
14 / 11.88
6 / 4.32
16 / 10.24
32 7.23.68
36 / 30.72
46 / 55.68
37 / 42.88
19 / 21.12
6 / 7.68
                                192
                    16


                    37


                    48


                    87


                    67


                    33


                    12


                   300
For this table, T * 28.17, with 6 degrees of freedom.  This 1s significant at
something 1n excess of the 99.9X level.  Age therefore 1^ a statistically signi-
ficant determinant of the probability of leakage.  It 1s clear from Table 1 that

while some tanks are leaking at ages 5 to 14, they represent a fairly small
fraction (17X) of the entire sample.  After age 15, however, the percentage of

leakers Increases to 46%.

Another Interesting observation also emerges from Table 6i  1f a new contingency
table 1s constructed only for tanks of age 15 or higher, there 1s no statisti-

cally significant effect of age upon the probability of leakage:
                                   B-15

-------
     Tank
      Age
    15 - 19
         •\

    20 - 24



    25 • 29


      > 29
                                    TABLE 7
  Number of
Leaking Tanks
    Number of
Non-Leaking Tanks
41 / 39.78
30 / 30.68
14 / 15.09
6 / 5.49
46 / 47.22
37 / 36.36
19 / 17.91
6 / 6.51
                                  91
                               108
                       87"


                       67


                       33


                       12


                      199
For this table, T .» .33, with-3 degrees of freedom.  This 1s not significant,
even at the 75* level.  Thus, age seems to have an effect only for tanks younger
than 15 years; above that age, the probability of leakage Is apparently con-
stant.


3.  SAV

A contingency table can also be set up to test the effect of SAV upon the proba-
bility of leakage:
                                     B-16

-------
       SAV



      0-4


      5-9


     10 - 14


       > 14
                                    TABLE 8
   Number of
 Leaking Tanks
                       Number of
                   Non-Leaking Tanks
4 / 13.32
28 / 31.68
53 / 47.52
23 / 15.48
33 / 23.6?
60 / 56.32
79 / 84.48
20 / 27.52
                                                  37
                               108
                                192
                                         132


                                          43


                                         300
For this table, T * 17.55 with 3 degrees of freedom.  This Is significant at the

99.9X level.                                               .



4.  Interaction of SAV and Age


Unfortunately, Age and SAV are not Independent variables, as 1s shown by Table

9:

                                    TABLE 9
     Age of
      Tank
 Low
(0-6)
Soil Aggressiveness Value

        Medium             High
        (7-12)             O12)
     0-14 years


    15 - 20 years
        w


     > 20 years
37 / 24.91
26 / 33.05
11 / 16.03
32 / 47.47
72 / 62.98
37 / 30.55
32 / 28.62
36 / 37.97
17 / 18; 42
                            74
                141
                            85
                                                 101
                                                 134
                                                  65
300
                                    B-17

-------
The T statistic for this table  is  17.26 with 4 degrees  of  freedom.   This  indi-
cates that there is less than a .5% chance  that SAV  and Age  are  uncorrelated.
An examination of Table 9 shows that  the  strongest correlation appears  to  occur
for low and medium SAV's.  High-SAV tanks are fairly randomly distributed  across
all three age groups.

This correlation would be easy  to  explain if older tanks were more  likely  common
to be found In low-SAV soils; In that case, the relationship between Age and SAV
would simp-ly be due to a survival  factor  (a disproportionate number  of  older
tanks in aggressive soils would already have been replaced long  before  the sur-
vey was taken).  Such, however, 1s not the case.  Instead, younger tanks are
more likely to be found in low-SAV soils.  This cannot be a survival effect.
Instead, It probably represents a  shift in Installation practices in favor of
the less corrosive soils.

The correlation between tank age and SAV  makes it difficult to determine which
Is the dominant variable.  It Is possible, for example, that the observed
effects of SAV and SAV x Age are actually the effects of Age, transmitted
through the linkages among these variables.
              tf
This hypothesis can be tested by constructing contingency tables examining the
effects of SAV upon tank leakage for each of the tank age-groups.  This will
reveal whether SAV has any effect  Independent from Age.
                                     B-18

-------
                                TABLE 10
Low     SAV


Medium  SAV


High    SAV
                                     Young Tanks
                           (T » 12.34, 99.5X significance)
                        Number of
                      Leaking Tanks
    Number of
Non-Leaking Tanks
1 / 6.23
5 / 5.39
11 / 5.39
36 / 30.77
27 / 26.61
21 / 26.61
                          17
     84
                        37
                        32
                        32
101
Low     SAV


Medium  SAV


High    SAV
                                   Medium-age Tanks
                            (T - 4.67, 90X significance)
                        Number of  .
                      Leaking Tanks
    Number of
Non-Leaking Tanks
7 / 11.84
35 / 32.78
19 / 16.39
19 / 14.16
37 / 39.22
17 / 19.61
                           61
      73
                        26
                        72
                        36
134
                                B-19

-------
    Low     SAV


    Medium  SAV


    High    SAV
                                          Old Tanks
                                  (T * 2.44, Insignificant)
                            Number of
                          Leaking Tanks
                               30
    Number of
Non-Leaking Tanks
6 / 5.08
14 / 17.08
10 / 7.85
5 / 5.92
• 23 / 19.92
7 / 9.15
         35
11


37


17


65
These tables Indicate that SAV and Age have some Independent effect, but. only
for the younger tanks.  Combining the results of Tables 10 and 7, it appears
that neither Age nor SAV have much effect for tanks older than 20 years.
                                     B-20

-------
                                 DATA ANALYSIS

Proving that there are statistically significant linkages between leakage rates
and SAV, Age, and SAV x Age does not conclude the analysis, however, for it is
also necessary to determine the correct interpretation of these linkages.  The
ultimate goal of such an interpretation is to deduce a cumulative probability
distribution for tank failure over a range of SAV x Age categories.
Unfortunately, this 1s not a straightforward task.
          i
One approach to this problem is to assume that the data actually represent the
desired cumulative distributions.  This assumption would be correct if leaking
tanks were never repaired or replaced, for in that case, the number of leaking
tanks in any SAV x Age bracket would include both new and pre-existing leaks.
Under this simple assumption, the Canadian data yields the following distribu-
tion (obtained from the numbers in Table 5):

                                    TABLE 11
                                                    Cumulative
                                                 Leak Probability
       SAV x Age              .                          (X)
         0-49                                            0
        50-99                                           28.9,
       100-149                                          27.0
       150-199                                          53.7
       200-249                                          44.0
       250-299                                          50.0
       300-399                                          36.4
         > 400                                          63.6

Since a cumulative probability distribution Is by definition non-decreasing, the
fluctuation In 300-399 category must be assumed to be anomolous.  It can be
reduced by combining the two highest brackets: *
                                    B-21

-------
                                    TABLE 12
                                                    Cumulative
                                                 Leak Probability
       SAV x Age                                 	X	
         0-49                                            0
        50-99                                           28.9
       100-149 .                                         27.0
       150-199                                          53.7
       200-249                                         .44.0
       250-299                                          50.0
         > 300                                          45.0
This distribution reveals that tank failure occurs in two spurts:  one at SAV x
Age between 50 and 99, and the other at SAV x Age between 150 and 199.  Other
fluctuations in failure rates are statistically insignificant, as-can be
verified by constructing the appropriate contingency tables.-

This distribution has the advantage that It conforms to the expected sigmoidal
pattern, with most of the failures occurring during the middle brackets and with
some  tanks which are effectively immortal even in highly aggressive soils, but
the numbers in Table 12 do not seem appropriate.  Leakage should  not occur at
such tightly defined intervals; i.e. there are too few leakers In the 100-149
category.  Even more Importantly, there are too many Immortals.   It is very
unlikely that half of the tanks would still survive after 40 years In a soil of
SAV « 10 (or 20 years with SAV » 20).  Yet that is what this distribution seems
to indicate.

The problem with the distribution in Table 12 is simple:  some leaking tanks
will have been replaced relatively soon after they began to leak.  Thus, in the
upper brackets the survey self-selects for norr-leaking tanks (since they are
more likely to still be in use), and the cumulative percentages are too low.

One way.to cure this problem would be to convert the survey data  Into a cumula-
tive distribution by computing the number of missing tanks.  This could be done
by determining the relative numbers of tanks burled in each year  and assigning
these to SAV brackets according to the distributions in Table 10.6  To the

*>Th1s apportionment assumes that the missing tanks in each age group followed the

                                  B-22

-------
extent that Table 11 underrepresents certain SAV x Age groups, it can then be
assumed that the missing tanks are those which have previously leaked, and the
cumulative probabilities can be adjusted accordingly.  In order to normalize
these calculations, 1t must be assumed that one SAV x Age-bracket Is fully
represented.  Presumably, this would be bracket 0-49.

Unfortunately, this analysis Is fraught with difficulties.  Not only are the
calculations complex (and somewhat recursive), but the necessary tank burial
data 1s not available.  Instead, the best alternative are three data sources on
service station construction, and even these are difficult to obtain for Canada
(U.S. data are presented in Table 13).  Furthermore, once the data are
assembled, It appears that the Canadian survey underrepresents younger tanks.
This Indicates either serious problems in the application of U.S. service sta-
tion data to Canadian tank burials, or .it Indicates that the Canadian survey was
not random, but Instead favored older tanks.  In either event, tank attrition
cannot be computed, and another approach .must be used to obtain a more reaso-
nable cumulative distribution.

A simpler approach to the problem of self-selection of older non-leaking tanks
may be found by varying the assumptions underlying Table 12.  Instead of
assuming that leaks are never detected before the survey, it can be assumed
instead that all leaks are detected and the tanks replaced before enough time
has passed for the tank to move Into the next SAV x Age category.  This assump-
tion requires the detection period to be Inversely proportional to SAV, but that
requirement would be sensible If monitoring 1s better for tanks known to be in
more aggressive soils.

Under this assumption, the failure rates 1n Table 12 become elements of a proba-
bility density function and the corresponding cumulative distribution may be
calculated.
same SAV distributions as their surviving kin.  Such an assumption 1s probably
not accurate, but it 1s better than nothing.

                                    B-23

-------
                                                                TABLE  131
ro
Date
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972 |
Hunter of
Service
Stations2
•
179,647
•»
180,347
180,697
181,040
181.390
181.747*
188.100
194,600
200,100
206.755
207,800
208,800
209,700
210,600
211,473*
212.600
213.550
214.500 ,
216,059*
219,100
222.200
222.000
220,000
fc 226, 459*
Change In API Reports3
Nunter of of New
Service Service
Stations Stations
350
350
350
350
343
350
357
6353
6500
5500
6655
1045
. 1000
900
900
873
1127
950
950
1559
3041 * 3740
3100
(200) 2508
(2000) 2068
6459 1689
Ratio of
Rehabilitations
API Building* to new
Deacti vat Ions3 Permits Constructions5


9.021
9.826
10,615
5,391
7,801
8.050



6,080
6,150
6,500
6,275
6,606
4554 6000-7000
6,200
3586
3630
34*


24.30
0.55
0.63
0.00
0.17
6.70



5.96
4.46
5.84
5.60
3.24
1.14
1.00



Number of
Rehabilitations6


8664
3473
4115
(109)
1146
7005



5207
5023
5550
5325
5047
3500
3100




-------
                                                        TABLE 13  (Continued)
    Date
Number of
Service
Stations
Change in
Number of
Service
Stations
API Reports
of New
Service
Stations
API
Deactivations
Building
Permits
Ratio of
Rehabilitations
to new
Constructions
   Number of
Rehabilitations
ro
in
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
215.880
196.130
189.480
186.340
176.400*
172,300
164,790
158,540
151,250
147.000
(579)
(19.750)
(6.650)
(3.140)
(9.940)
(4.100)
(7.510)
(6.250)
(7.290)
(4.250)
1172
206
206
319
'284
353
286
169
297

9342
6041
4127
5676
5683
5138
3724
3380
4273

                                                                                            9.0
                                                                                            3.8
                                                                                            3.03
                                                                                            6.16

                                                                                            8.34
                                                                                            9.68
     1854
      783
      967
     1749

     2385
     1636
     * Source:  National Petroleum News,  annual  Fact Book of  petroleum statistics.
     2 Obtained by National Petroleum News  (NPN)  from U.S.  government data  and  NPN
      estimates.  Numbers marked  by  a star (*)  are census  totals.
     3 Obtained by NPN from American  Petroleum Institute totals for  a somewhat  varying number of responding companies.
     ^ Obtained by NPN from the Department  of  Labor, Bureau of Labor Statistics.
     5 From 1974 through 1980, these  are  obtained by NPN from an API survey.  For 1954 through 1969, these numbers are
      obtained from columns 3 and 8.                                      .    .
     6 For 1954 through 1969, these numbers are  the difference between the  number of building permits Issued and the change
      in the total number of service stations.   For 1974 through 1980,  these numbers are obtained from columns 4 and 7.

-------
                                    TABLE 14


                                 Probability of              Cumulative Failure
   SAV x Age                     New Failure (X)              Probability (X)

     0-49                              0                             0
    50-99                             28.9                         .28.9
   100-149                            27.0                          46.7
   150-199                            53.7                          75.3
   200-249                            44.0                          86.2
   250-299                            50.0                          93.1
   300-399>                           36.4                          95.6
     > 400"                           63.6                          98.4


Unfortunately, these cumulative probabilities seem to be too high.  Anecdotal
data would seem to Indicate that a larger percentage of tanks have extremely
long lifetimes, even 1n very aggressive soils.  Furthermore, 1t 1s likely that
some leaks remain undetected over very, long periods of time.  For both of these

reasons, the calculated failure rates are probably too high, though they pro-
bably can safely be used as upper bounds on the actual probabilities.?
                                             «
              tf

A third assumption may be used 1n an effort to split the difference between the
two polar cases discussed above:  1t may be assumed that SOX of the leaks

detected 1n each bracket are new, while the remaining leaks are ones that have
been continuing since a previous bracket.  Under this assumption,.the first two

brackets are unaltered,** but for the other brackets, the previously leaking



7There 1s another, more theoretical problem with Table 14:  the cumulative proba-
bility distribution 1s highly dependent on the width of the SAV x Age brackets
used 1n Its computation.  Decreasing their width Increases the number of cate-
gories (without significantly changing the second column of the table), thereby
causing the cumulative probability to converge, upon 1.0 at a considerably more
rapid rate.  Broadening the categories has the reverse effect.

In non-mathematical terms, this problem Is related to the detection-period
problem discussed earlier.  If leaks are detected rapidly, then the data Indica-
tes a high rate of new leak formation.  If leaks are detected slowly, then the
cumulative probability approaches the no-detection assumption depicted 1n Table
11.  Since these problems relate to the proper Interpretation of real-world
data, they are not the same as the scenarios ultimately to be studied 1n the
computer model.  Instead, the goal Is to determine just how conservative the
Canadian oil companies1 detection/repair policies actually were prior to 1977.
The use of 50-po1nt brackets seem to be a reasonable assumption for the tech-
nology then 1n use.

8The first bracket Is unaltered because there are no leaks.  The second bracket


                                   B-26

-------
tanks must be removed from both the leaking tank category and the bracket

totals.  With these modifications, the data becomes:
1s unaltered because to alter 1t would be contrary to the observation that there
are no failures 1n the first bracket.
                                    B-27

-------
                                    TABLE 15
 SAV x Age

    0-49 ,
   50-99
  100-149
  150-199
  200-249
  250-299
  300-399
    > 400
  Number of
Leaking Tanks

      0
     11
      5
     18
     11
      7
      4
      3.5
  Number of
Previously Non-
 Leaking Tanks

     47
     38-
   -  32
     49
     39
     21
     18
      7.5
Percent
of New
 Leaks
 0
28.9
15.6
36.
28,
33,
22.2
50.0
    ,7
    ,2
    ,3
                                                    Cumulative
                                                       Leak
                                                   Probability
                                                       (X)

                                                        0
                                                       28.9
                                                       40.0
                                                       62.0
                                                       72.7
                                                       81.8
                                                       86.0
                                                       93.0
These cumulative probabilities from these three assumptions can be combined in
a single table:


                                    TABLE 16
  SAV x Age

    0-49
   50-99
  100-149
  150-199
  200-249
  250-299
  300-399
    > 400
    Lower
    Bound
(Assumption 1)

      0
     28.9
     27.0
     53.7
     44.0
     50.0
     45.0
     45.0
                       Upper
                       Bound
                   (Assumption 2)
                         0
                        28,
                        46,
                        75,
                        86,
                        93,
                        95.6
                        98.4
                   Assumption
                       3

                       0
                      28.9
                      40.0
                      62.0
                      72.7
                      81.8
                      86.0
                      93.0
            Average of
            Assumption
               1-3
                0
               28,
               37,
               63.
               67.6
               75.0
               75.5
               78.8
The last column 1s the one which will be used fn the Monte Carlo model, though
It 1s subject to revisions as better data become available.
                                    CAVEATS


The cumulative probability distribution presented 1n Table 16 must be used with
caution, for unfortunately, the Canadian data set does not represent a random
                                   B-28

-------
survey of existing tanks.  Instead, the data were collected in 3 ways:9

     o  Over a 6-month period In 1977, all PACE member companies were requested
        to report leak Incidents.  Soil samples were taken at the leaking tanks'
        sites.
     o  During this same time period, PACE member companies were requested to
        report tank decommissioning*.  Decommissioned tanks were then tested for
        leaks, and soil samples were taken.
     o  Other tanks on the same site as a leaking or decommissioned tank were
        also tested.
Thus, the survey is biased both toward leaking tanks and toward older tanks.
(The latter bias occurs because older tanks are more likely to be
decommissioned).  The age bias Is relatively unimportant.  The bias toward
leaking tanks, however, means that the resulting data present a worst-case
portrait of the existing tank situation.  This bias may not be overly severe,
however, for the fact that only 36X of the sample tanks were leaking Indicates
that the other two sampling methods may have predominated.  Furthermore, this
bias may be offset by the fact that the second and third sampling techniques
tend to self-select for non-leaking tanks.  Nevertheless, an unknown net bias
probably results, and the data must be viewed as only an approximation of the
results of a truly random survey.
                            AN ALTERNATIVE APPROACH

Instead of complex calculations based on Age x SAV, it may be more appropriate,
given the data biases discussed above, to attempt a simpler model.  With this
in mind, the data can be grouped Into high, middle, and low SAV soils, and
failure rate versus age may be calculated for each soil group.  The results are
presented 1n Table 17.
^PACE, "Underground Tank Systems," supra ru-4?-p;-43, and personal communication
with J.R. Clendenlng, Esso Petroleum.  Canada, June 1985.

                                    B-29

-------
                                              TABLE 17
                          SAV                      Medium SAV                  High SAV
                          6>                         (7-12)                      (> 13)
Tank Age         Leakers   Non Leakers         Leakers   Non Leakers
0-4
5-9
10-14
15-19
20-24
> 25

* 0
0
1
4
4
_5
14
8
13
15
14
6
_.* . ~
60
0
1
4
22
19
'••• _8
54
7
8
12
25
18
17
87
0
4
7
15
7 .
_7
40
1
11
9
7
13
" JL:-
45
                                         B-30

-------
Table 17 may be  Interpreted under the same 3 assumptions that were used in
Tables 11, 12, and 14-16.  The resulting probability distributions are presented
in Tables 18-20.

The same approach can also be used employing only two SAV categories.  If soils
are classified as benign when SAV 1s 9 or less and aggressive when SAV 1s 10 or
greater, then the data can be summarized in Table 21.

Table 21 can be  used to calculate cumulative probability distributions as in
Tables 18-20.  The results, using the same three assumptions, are presented in
Tables 22-23.

The 2-part and 3-part SAV distinctions have certain similarities.  In both
cases, there are clear differences between soils of different aggresslvttles.
These differences can most readily be appreciated by presenting the results in a
single table, as is done in Table 24.

These distributions can be plotted graphically, as can the probability of tank
failure versus.-SAV x Age (from Table 16).  This 1s done 1n Figures 4-9.  For
Interpretive purposes, these graphs have converted the cumulative distributions
reported in the  tables Into the underlying probability densities.  Thus, these
histograms represent the probability that the tank failure will originate in
each of the designated Intervals.

These graphs Indicate that SAV x Age. 1s probably not the best measure of tank
deterioration.   The reason for this conclusion 1s the bimodal nature of the SAV
x Age probability density.  While such bimodal1ty might possibly be an accurate
reflection of the real world, 1t Is more likely that the bimodal distribution
results from Improperly aggregating unlike distributions.  This latter explana-
tion appears particularly appropriate In the present situation.  As Figures 4-6
Indicate, the probability distributions are differently-shaped for low-SAV and
Mgh-SAV soils.  Low-SAV soils produce a relatively steady failure rate for all
years after year 9, while Mgher-SAV soils produce much higher failure rates in
the lower years, but declining failure rates 1n later years.W  Combining these
10The reason for the low failure rates after year 19 1s simply that by that year,
a large fraction of h1gh-SAV tanks have already failed.

                                    B-31

-------
                       TABLE 18.  LOW SAV SOILS (SAV < 6)


                     Cumulative Probability of Leakage (X)


           Lower Bound         Upper Bound         Assunption 31
Tank    (Detection within     (No detection       (75* of survey
Age          5 years)        prior to survey)     leaks are new)        Average

 0-4           0                    0                   0                  0
 5-90                    0                   0                  0
10-14    "     6.3                  6.3                 6.3                6.3
15-19        22.2                 27.1                22.8               24.0
20-24        40.0                 56.3                48.5               48.3
 > 25        55.6                 80.6                73.6               69.9
1 Assumption 3 in this table has been adjusted from that used In Tables 15 and
  16 in order to be consistent with the lower leak rates for low-SAV tanks.
                       TABLE 19.  MEDIAN SAV SOILS (7-12)


                     Cumulative Probability of Leakage (%)


           Lower Bound         Upper Bound         Assumption 3*
Tank    (Detection within     (No detection       (SOX of survey
Age     	5 years)        prior to survey)     leaks are new         Average

 0-40                   0                    00
 5-9          11.1                11.1                 11.1              11.1
10-14         25.0                33.3                 28.91             29.1
15-19         46.8                64.5                 50.6              54.0
20-24         51.4                82.8                 67.7              67.3,
 > 25         32.0                88.3                 73.8               -  2
                             •

1 This calculation uses a 75X rate of new leak development 1n order to be
  consistent with the lower-bound estimates in the previous column.

2 No number Is calculated for this range, for the anomalous decline'In pro-
  bability for the "lower bound" would produce an equally anomalous fluctuation
  in the average.
                                   B-32

-------
                      TABLE 20.  HIGH SAV SOILS (SAV > 13)


                     Cumulative Probability of Leakage (X)


           Lower Bound         Upper Bound         Assumption 3
Tank    (Detection within     (No detection       (SOX of survey
Age         5 years)	    prior to survey       leaks are new        Average


 0-40                    0                    0                 0
 5-9         26.7                 26.7                 26.7              26.7
10-14    '   43.8                 58.8                 47.2              49.9
15-19        68.2                 86.9                 74.5              76.5
20-24        35.0                 91.5                 79.9              79.91
 > 25        63.6                 96.9                 89.3              83.3


1 Obtained by Interpolation between the values for ages 15-19 and > 25.  An
  average of Assumptions 1, 2, and 3 is dominated by the anomalous~~value for
  Assumption 1.                                        .
                                    TABLE 21

               Benign Soils (SAV < 9)              Aggressive Soils (SAV > 10)

            Number of        Number of            Number of       Number of Uon-
 Age      Leaking Tanks   Non-leaking Tanks     Leaking Tanks     Leaking Tanks

 0-40               13                    03
 5-9            0               15                    5                17
10-14       .1               18                   11                18
15-19          11               26                   30                20
20-24          11                9                   19                28
 > 25           9               12                   11                13
                                    8-33

-------
                       TABLE 22.  BENIGN SOILS (SAV < 9)


                     Cumulative Probability of Leakage (X)


                                Upper Bound           Assumption       Average
         Lower Bound           (Assuming all          3 (67X of          of
         (Assuming no          leaks detected         observed       Assumptions
Tank    detection prior      and repaired within      leaks are       1, 2, and
Age       to survey          5-year age bracket)        new)              3

 0-40                       0                   00
 5-90                       0                   00
10-14        5.3                     5.3                 5.3,            5.3
15-19       29.7                    33.4                32.4*           31.8
20-24       55.0                    70.0                62.8            62.6
 > 25       42.9                    82.9                75.2            67.0

1 Calculated under the assumption that only one of the observed leakages was
  pre-existing, in order to be consistent with the preceding bracket's low-leak
  rate.                               •
                     TABLE 23.  AGGRESSIVE SOILS (SAV > 10)


                     Cumulative Probability of Leakage (X)


                                Upper Bound           Assumption
         Lower Bound           (Assuming all          3 (50% of
         (Assuming no          leaks repaired         observed
Tank    detection prior        within 5-year          leaks are
Age       to survey             age bracket)            new)            Average

 0-40                      0                    00
 5-9        22.7                   22.7                 22.7,            22.7
10-14       37.9                   52.0                 48.I*            45.0
15-19       60.0                   80.8                 70.3             70.4
20-24       40.4                   88.6                 77.8             72.62
 > 25       45.8                   93.8                 84.4             74.7

* Calculated under the assumption that SOX of the observed leaks are new, in
  order to be consistent with the previous bracket's observed low leak rate.

2 Calculated by Interpolation between the preceding and following brackets, in
  order to prevent the decline In the value for assumption 1 from causing
  anomalous results.
                                   B-34

-------
                                 TABLE 24.
                   Cumulative Probability of Leakage  (X)
Tank
Age1
4
9
14
19
24
> 25
Low SAV
(0 to 6)
0
0
, 6.3
24.0
48.3
69.9
Medium SAV
(7 to 12)
0
11.1
29.1
54.0
67.3
-
High SAV
(> 13)
0
26.7
49.9
76.5
79.9
83.3
Benign Soil
(0-9)
0
0
5.3
31.8
62.6
67.0
Aggressive
Soil
(> 10)
0
22.7
45.0 '
70.4
72.6
74.7
Tank ages have been changed from age brackets to the age corresponding to the
top of each bracket.
                                B-35

-------
            25 T
            20 •
  FAILURE
PROBABLITY
                                     FIGURE 4
                               LOV SAV SOLS (0-6)
                  0-4     3-9     10-14    15-19   20-24

-------
  FAILURE
PROBABILITY
                                     FIGURES
                              MEDUMSAV SOLS (7-12)
                 0-4     5-9   10-14  15-19  20-24   >24
                                    TANK ACE
                                    B-37

-------
         '   20
  FAILURE
PROBABILITY
                                     FIGURES
                               HIGH SAV SOLS (13-21)
                                  10-14    13-19
                                     TANKAGE
                                    B-38

-------
  FAILURE
PROBABILITY
             10
                                     FIGURE 7
                                 BENIGN $OLS (0-9)
                   0-4     5-9     10-14   15-19    20-24    >24
                                      B-39

-------
  FAILURE
PR08ABLITY
                                     FIGURE 8

                             AGGRESSIVE SOLS(> 10)
                          5-9     10-14   15-19   20-24     >24
                                     B-40

-------
                                    FIGURE 9
                       FAIURE PROBABLITY USING SAV X ACE
  FAILURE
PROBABLITY
            30
            20
13
            10
                0-49  50-99  100-   ISO-   200-   250-   300-   XOO
                              149    199   249   299    399
                                    B-41

-------
two dissimilar age distributions could easily produce the bimodal distribution
depicted in Figure 9, even after the data have been converted from age cate-
gories to SAV x Age categories.  It therefore appears that the data has greater
usefulness if SAV and Age are both used as separated variables than it
does if they are combined Into the single variable of SAV x Age.
          •„
This analysis, however, does not Indie-ate whether two or three SAV categories
are preferable.  This decision can be facilitated, though, by a re-examination
of the raw data in Figures 1 and 2.  These scatter diagrams indicate that a
substantial percentage of the tanks, particularly leaking ones, are to be found
between SAV»9 and SAV»11.  There 1s no theoretical reason for dividing
aggressive soils from benign soils at an SAV of either 9, 10, or 11, yet because
of the clustering of the data, this arbitrary division can significantly .alter
the probability density functions when only two SAV categories are used.  Thus,
the natural clustering of the data favors the use of three SAV categories, and
therefore three such categories will be used in the computer model.

                               COMPUTER MODELING

In order to carry out the Monte Carlo simulation on a year-by-year basis, it is
necessary to calculate failure probabilities for each year between 1 and 20.
This can most conveniently be done by straight-line Interpolation between the
age brackets used in Table 24.  The results are presented in Tables 25 and 26.
(Table 25 presents cumulative probabilities, while Table 26 presents probabili-
ty densities).  Once SAV 1s determined, these tables can then be used to deter-
mine annual probabilities of failure.

SAV can be determined 1n one of two ways:  1t can either be postulated as an
exogenous parameter, or 1t can be determined stochastically.  The deterministic
approach.1s the simplest, and 1s to be preferred for the Initial simulations,
but the stochastic approach may be useful for modeling more complex'scenarios.

The PACE data can be used to obtain a distribution of SAV's for the 300 tanks
covered by the survey.  This distribution 1s presented in Table 27.

While there Is no guarantee that this distribution 1s representative of U.S.
soils, 1t 1s probably a reasonable approximation, and it can be used to calcu-
                                    B-42

-------
TABLE 25.  CUMULATIVE PROBABILITIES OF FAILURE  IN LOW-,
           MEDIUM-, AND HIGH-SAV SOILS
                     Cumulative Failure Probability  (%)
Tank
Age
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
IS
19
20
Low
SAV
(0-6)
0
0
0
0
0
0
0
0
0
1.26
2.52
3.78
5.04
6.30
9.84
13.38
16.92
20.46
24.00
28.86
Medium
SAV
(7-12)
0
0
0
0
2.22
4.44
6.66
8.88
11.10
14.70
18.30
21.90
25.50
29.10
34.08
39.06
44.04
49.02
54.00
56.66
                                                            High
                                                            SAV
                                                           (> 13)
                                                             0
                                                             0
                                                             0
                                                             0
                                                            5.34
                                                           10.68
                                                           16.02
                                                           21.36
                                                           26.70
                                                           31.34
                                                           35.98
                                                           40.62
                                                           45.26
                                                           49.90
                                                           55.22
                                                           60.54
                                                           65.86
                                                           71.18-
                                                           76.50
                                                           77.18
                      B-43

-------
TABLE 26.  PROBABILITY DENSITIES FOR FAILURE IN LOW-,
           MEDIUM-, AND HIGH-SAV SOILS

                         Probability Density (X)
Tank
Age
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15.
16
17
IS
19
20
Low SAV
(0-6)
0
0
0
0
0
0
0
0
0
1.26
1.26
1.26
1.26
1.26
3.54
3.54
3.54
3.54
3.54
4.86
Medium SAV
(7-12)
0
0
0
0
2.22
2.22
2.22
2.22
2.22
3.60
3.60
3.60
3.60
3.60
4.98
4.98
4.98
4.98
4.98
2.66
High SAV
(> 13)
0
0
0
0
5.34
5.34
5.34
5.34
5.34
4.64
4.64
4.64
4.64
4.64
5.32
5.32
5.32
5.32
5.32
0.68
                     B-44

-------
            TABLE 27.  DISTRIBUTION OF SOIL AGGRESSIVENESS VALUES
AV          Number of Tanks
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
0
. 20
7
10
21
16
10
7
34
43
12
35
22
20
7
28
6
1
0
0
1
Probability
Density
0
6.7
. 2.3
3.3
7.0
5.3
3.3
2.3
11,3
. 14.3
4.0
11.7
7.3
6.7
2.3
9.3
2.0
.3
0
0
.3
Cumulative
Probability «i
0
6.7
9.0
12.3
19.3
24.7
28.0
30.3
41.7
56.0 .
60.0
71.7
79.0
85.7
88.0
97.3
99.3
99.7
99.7
99.7
100.0
                               B-45

-------
          TABLE 28.  CONDITIONAL SAV DISTRIBUTION UNDER A REQUIREMENT
                     THAT SAV NOT EXCEED 10

                         Probability Density             Cumulative Probability
SAV                      	(X)                     	(XJ	

  1                               0                                 0
  2                             11.9                               11.9
  3                              4.2                               16.1
  4                              5.9                               22.0
  5                             12.5                               34.5
  6                              9.5                               44.0
  7                              5.9    .                           50.0
  8                              4.2                           '    54.1
  9                             20.2                               74.4
 10                             25.6                              100.0
                                    B-46

-------
late the probability that any given tank falls into each of the three SAV cate-
gories used in Tables 25 and 26.  In addition, the SAV distribution can be used
to obtain the conditional SAV-distribution under various regulatory scenarios.
Consider, for example, a regulation requiring that SAV not exceed 10.  If this
regulation has no effect on the distribution of acceptable SAV's, then the con-
ditional SAV distribution can be obtained simply by dividing the numbers in
Table 27 by 56X (the unconditional probability that SAV < 10).  The results are
presented in Table 28.

Similar computations could be undertaken for any other SAV cut-off.  These
results could then be used to determine the probability that the tank falls in
each of the three SAV-categories used-to predict the probability of failure.

More complex regulatory scenarios could also be modeled under this approach.
For example, a proposed regulation might make the use of cathodic protection or
secondary containment dependent on the aggressiveness of the soil in question.
This could be modeled by first sampling a value for SAV and then using that
value to determine other system parameters.  Such a scenario is considerably
more complicated than the scenarios that have been modeled to date, but if it is
desired that such composite scenarios be studied, they are well within the capa-
bilities of the model.
                                   ' B-47

-------
        APPENDIX C
TANK FAILURE CASE STUDIES

-------
                           TANK FAILURE CASE STUDY #1


SITE:  Transformer manufacturing plant of Federal Pioneer ltd


LOCATION;  Regina, Saskatchewan, Canada


RELEASE MECHANISM;  Pipe rupture


DATA SOURCE
        *x
     •  "A Case Study of a Spill of Industrial Chemicals:  Polychlorinated
        Biphenyls and Chlorinated Benzenes," National Research Council Canada,
        NRCC No. 17586, 1980.


DESCRIPTION OF RELEASE

     •  In mid-1976 an underground pipe carrying PCB's from a 31,000 liter tank
        ruptured.


CAUSE OF RELEASE;  Underground pipe ruptured




RELEASE MATERIALS

     •  How Detected:
     •  Material Types:  PCB's (70%)t chlorobenzenes (30X)
     0  Quantities Released:  6800 - 21,000 liters


RELEASE DURATION



RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:  land
     0  Description of contamination:  underground


COMMENTS

     0  Additional Information on this site can be found In:

        •  Roberts, Russell J., John A. Cherry, Franklin W. Schwartz, "A Case
           Study of a Chemical Spill:  Polychlorinated Blphenyl, (PCB's)—1.
           History,  Distribution, and Surface TransTocation," In Water
           Resources Research, Vol. 18, No. 3,   pp. 525-534, June 1982.


                                   C-l

-------
Roberts, Russell J., John A. Cherry, Franklin W. Schwartz  "A Case
Study of a Chemical Spill:  Polychlorinated Biphenyls (PCB's)--2
Hydrological Conditions and Containment Migration," in Water
Resource Research. Vol. 18, No. 3, pp. 535-545, June 195?:	
                         C-2

-------
                             TANK  FAILURE CASE  STUDY  #2


SITE;  Unknown


LOCATION;  Unknown


RELEASE MECHANISM;  Catastrohlc release


DATA SOURCE

     •  Dartnell Jr., R.C.,  T.A.  Ventrone, "Explosion of a Para-Nltro-Meta-
        Cresol Unit," Chemical Engineering Progress, Vol. 67, No. 6, pp. 58-61."


DESCRIPTION OF RELEASE

     t  A temperature Indicator on the feed-tank Indicated a temperature of 154°
        C for the entire holding  period up to  the time of explosion.  This was ~
        also the temperature of the product leaving  the process step Immediately
        upstream.  Prior to  the explosion, the pressure on the feed tank
        Increased from 40 to 10056.  No product was being fed.


CAUSE OF RELEASE;  Explosion



RELEASE MATERIALS

     •  How Detected:  pressure sensor
     •  Material Types:  para-nltro-meta-cresol (PNMC)
     •  Quantities Released:  1500 gallons


RELEASE DURATION



TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage
     •  Equipment Material:  stainless steel
                                   C-3

-------
                            TANK FAILURE CASE STUDY  #3


SITE;  Unknown


LOCATION;  Unknown


RELEASE MECHANISM;  Unknown, probably tank corrosion
        •v

DATA SOURCE

     0  Eagen Jr., H.B., et al. "Removal of Hazardous Fluid from the Groundwater'.
        1n a Congested Area—A Case History," Control of Hazardous Material
        Spills, Proceedings of 1976 National Conference on Controls of Hazardous
        Material Spills.


DESCRIPTION OF RELEASE               .

     •  Hydrocarbon migrated through the top of a shallow water table.  It
        seeped at the land surface 1n low lying areas discharging 200 gallons
        per day Into a perennial stream.  Domestic wells were abandoned and
        product seeped Into sewer lines.


CAUSE OF RELEASE;  Unknown



RELEASE MATERIALS

     t  How Detected:
     0  Material Types:  Hydrocarbon product (80% gasoline)
     0  Quantities Released:   500,000 gallons


RELEASE DURATION   "long period of time"


RELEASE ENVIRONMENT

     0  Land, Water, A1r, Unknown:  water
     0  Description of contamination:  underground Into groundwater and from
        groundwater Into surface water


COMMENTS

     0  This paper deals mostly with recovery operations and doesn't describe
        the failure event very well.
                                  C-4

-------
                            TANK FAILURE CASE STUDY #5


SITE;  Bulk Terminals, tank storage farm


LOCATION  Calumet Harbor Area, Chicago, Illinois


RELEASE MECHANISM;  Pipe rupture


DATA SOURCE

     •  Hampson, T.R. "Chemical Leak at a Bulk Terminals Tank Farm," Control of
        Hazardous Material Spills. Proceedings of 1976 National Conference on
        Control of Hazardous Material Spills


DESCRIPTION OF RELEASE

     •  Silicon tetrachloride leaked from a pipe rupture, forming.an add cloud
        with the moist air.  A rain storm worsened the situation, causing such
        dense fumes, that electrical.lines and transformers corroded and failed:


CAUSE OF RELEASE

     •  A block''valve on an Inlet line and a pressure relief valve were inadver-
        tently closed.  Pressure in the line began to build up.  At about 12:30
        p.m. on April 26, 1974, a flexible coupling on the Inlet line burst
        under the pressure.  The entire piping system shifted and a second line
        also cracked.


RELEASE MATERIALS

     •  How Detected:  fumes
     •  Material Types:  silicon tetrachloride
     t  Quantities Released:  plume contained 40 ppm of HC1; 284,000 gallons
        were leaked; Initially the acid cloud was about .25 miles wide, 1000
        to 1500 feet high, and 1 mile in length, but due to the storm, it grew
        to 9 miles In length.
                •

RELEASE DURATION

     •  It was 2.5 to 3 days before leak was sealed.  However, 7 days passed
        before there was no threat of additional releases.
                                   C-5

-------
TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage, steel tank with dry air or nitrogen.padding equip-
        ment, capacity of 1,500,000 gallons.  The tank contained 750,000
        gallons of fluid.
     •  Tank/Treatment Components, Ancillary Equipment:  extra -automatic
        pressure vents, special valves, and closed transfer pumps.
     •  Equipment That Failed:  flexible coupling, piping system, tank
     •  Dikes/Berms:  present, they contained the liquid spill.


RELEASE ENVIRONMENT
        "V
     •  Land, Water, Air, Unknown:  air and land
     •  Description of contamination:  silicon tetrachlorlde poured out of tank
        Into diked area and reacted vigorously with water In the air and rain-
        fall to form HC1 vapor.  The enormous acid cloud spread over the far
        south side of the city.


COMMENTS

     •  Additional information on this site 1s located under:

        -  Hoyle, W.C. and Melvln, 6.L. "A Toxic Substance Leak in Retrospect:
           Prevention and Response."  Control of Hazardous Material Spills,
           Proceedings of 1976 National Conference on Control of Hazardous
           Material Spills.
                                  C-6

-------
                           TANK FAILURE CASE STUDY #6


SITE;  Service station


LOCATION;  Cresskill, New Jersey


RELEASE MECHANISM;  Tank corrosion


DATE SOURCE

     •  Kramer, William H. "Ground-water Pollution from Gasoline," GWMR.
        Spring 1982, pp. 18-22.


CAUSE OF RELEASE;  Leaks in four 4000-gallon steel tanks due to corrosion.



RELEASE MATERIALS

     •  How Detected:  routine inventory check
     •  Material Types:  gasoline
     •  Quantities Released:   1200 gallons


RELEASE DURATION  two or three days


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  gasoline storage for service station
     •  Equipment That Failed:  tank
     •  Equipment Material:  steel
     •  Equipment Age:  17 years


RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:  land
     •  Description of contamination:  Underground


COMMENTS
       •                                                                   .
     t  The article 1s very detailed; It Indicates how the gasoline was reco-
        vered and the cost of recovery.
                                   C-7

-------
                            TANK FAILURE CASE STUDY #7


SITE:  Unknown


LOCATION;  Unknown


RELEASE MECHANISM;  Interior tank corrosion

        n.
DATA SOURCE

     •  Bosich, Joseph F. Corrosion Prevention for Practicing Engineers. Barnes
        and Noble, Inc.  New York, 1970, p. 186.


DESCRIPTION OF RELEASE


CAUSE OF RELEASE

     •  A workman accidentally dropped a 1" diameter hexagon-shaped nut to the
        bottom of the tank, causing localized Interior corrosion.


RELEASE MATERIALS

     •  How Detected:  visual detection
     •  Material Types:  concentrated sulfurlc acid
     •  Quantities Released:  Unknown


RELEASE DURATION


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage
     •  Equipment Material:  steel

-------
                            TANK FAILURE CASE STUDY  #8


SITE:  Unknown


LOCATION;  Sussex, Wisconsin


RELEASE MECHANISM;  Unknown, probably tank corrosion
       ^

DATA SOURCE                •      .

     •  Llndoff, David E., Keros Cartwrlght, Groundwater Contaminat1 on;
        Problems and Remedial Actions, Environmental Geology Notes, Illinois
        State geological Survey, May 1977, No. 81, Case history 88, p. 50.


DESCRIPTION OF RELEASE



CAUSE OF RELEASE


RELEASE MATERIALS

     •  How Detected:  complaints that water from some wells tasted and smelled
        like petroleum
     •  Material Types:  petroleum products
     •  Quantities Released:


RELEASE DURATION


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use: storage


RELEASE ENVIRONMENT

     •  Description of contamination:  surflclal material, surface water, and
        probably groundwater
                                   C-9

-------
                            TANK FAILURE CASE STUDY  #9


SITE:  Unknown


LOCATION;  Spring Mills, Pennsylvania


RELEASE MECHANISM;  Unknown, probably corrosion or rupture


DATA SOURCE

     •  Llndoff, David E., Keros Cartwrlght, Ground-Water Contamination;
        Problems and Remedial Actions, Environmental Geology Notes, Illinois
        State Geological Survey, May 1977, No. 81, Case History 24, p. 30.


DESCRIPTION OF RELEASE


CAUSE OF RELEASE:  Storage tank leak


RELEASE MATERIALS

     •  How. Detected:  explosion
     •  Material Types:   gasoline
     •  Quantities Released:  200-250 gallons


RELEASE DURATION;  2 weeks


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:   storage


RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:   water
     •  Description of contamination:  groundwater
                                    C-10

-------
                            TANK FAILURE CASE STUDY #10


SITE;  Unknown


LOCATION;  Southeastern Pennsylvania


RELEASE MECHANISM:  Unknown, probably tank corrosion

        •i
DATA SOURCE

     •  Llndoff, David E., Keros Cartwrlght, Ground-Water Contamination;
        Probems and Remedial Actions, Environmental Geology Notes, Illinois
        State Geology Survey, May 1977, No. 81, p. 35, Case History 19.


DESCRIPTION OF RELEASE


CAUSE OF RELEASE;  Leak 1n burled 10,000-gallon -tank


RELEASE MATERIALS

     •  How Detected:  appeared in a stream
     t  Material Types:  fuel oil
     •  Quantities Released: 60,000 gallons


RELEASE DURATION


TANK DESIGN AND OPERATING CHARACTERISTICS

     t  Tank Use:  storage


RELEASE ENVIRONMENT

     •  Land, Water, A1r, Unknown:  water
     •  Description of contamination:  surface and groundwater
                                   c-n

-------
                            TANK FAILURE CASE STUDY  #11


SITE;  Unknown


LOCATION;  Unknown


RELEASE MECHANISM;  Unknown, probably tank corrosion or rupture

        "t.
DATA SOURCE

     •  Undorff, David E., CartwMght, Keros; Ground-Water Contamination;
        Problems and Remedial Actions, Illinois State Geological Survey
        Environmental Geology Notes, May 1977, No. 81, case history 5, p. 32.


DESCRIPTION OF RELEASE

     •  A leak was discovered 1n a gasoline storage tank at a service stast1on.~~
        Further Investigation Indicated that several thousand gallons of gaso-
        line had been lost over a period of three weeks	


CAUSE OF RELEASE;  Leak 1n storage tank


RELEASE MATERIALS

     •  How Detected:  fumes in nearby houses
     •  Material Types:  gasolines
     •  Quantities Released:  several thousand gallons


RELEASE DURATION   3 weeks


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage


RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:  groundwater, explosive concentrations of fumes
       •In four houses
     t  Description of contamination:
                                   C-12

-------
                            TANK FAILURE CASE STUDY #12


SITE;  Essex  Industrial Chemicals, Inc. chemical processing plant


LOCATION;  Baltimore, Maryland


RELEASE MECHANISM;  Tank rupture


DATA SOURCE

     •  "News In Brief:  Hazardous Materials,"  Hazardous Materials
        Intelligence Report, 30 December, 1983, pp. 3-4.


DESCRIPTION OF RELEASE;  An outdoor storage tank burst


CAUSE OF RELEASE


RELEASE MATERIALS

     t  How Detected:  storage tank burst
     0  Material Types:  sulfuric acid
     •  Quantises Released:  485,000 gallons


RELEASE DURATION


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage


RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:
     •  Description of contamination:  approximately 388,000 gallons traveled
        over the frozen ground at the plant and spilled Into the Cabin Branch
        waterway, which leads Into Curtis Creek and eventually Into the
        Chesapeake Bay.
                                   C-13

-------
                            TANK FAILURE CASE STUDY #13


SITE;  Allied Chemical Corporation


LOCATION;  Louisiana


RELEASE MECHANISM;  Tank rupture


DATA SOURCE

     •  Shields, Edward, Dessert, W.J., "Learning a Lesson from a Sulfuric Acid
        Tank Failure," Pollution Engineering, December 1981, pp. 39-40.


DESCRIPTION OF RELEASE;  Into the ground


CAUSE OF RELEASE                                       "

     •  An Inlet nozzle for the addition of add to the tank was located too
        close to the tank wall.-  The cast Iron Inlet pipe broke due to corrosion
        and the high velocity of the Incoming add stripped the protective
        coating on the sides of the tank.  Eventually the tank ruptured.

             »„
RELEASE MATERIALS

     •  How Detected:  rupture In tank
     •  Material Types:  93X sulfurlc add
     •  Quantities Released:  2500 tons


RELEASE DURATION;  625 minutes (10.4 hours)


TANK DESIGN AND OPERATING CHARACTERISTICS

        Tank Use:  storage
        Tank/Treatment Components, Ancillary Equipment:  outlet fittings,  inlet
        nozzle manhole
        Equipment That Failed:  vertical weld
        Equipment Material:  A-283 grade C steel plate
        Equipment Age:  Unknown
        Corrosion Protection:  ferrous sulfate film
                                  C-14

-------
RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:
     •  Description of contamination:  effluent discharge pumps to Mississippi
        River were turned off and the rain water drainage pipe from the tank
        farm Impoundment area was sealed.
COMMENTS
        A 3000-ton sulfrlc acid tank In southern Canada also failed because
        the inlet nozzle was too close ta the tank wall.
                                 C-15

-------
                            TANK FAILURE CASE STUDY #14  .


SITE;  Unknown


LOCATION;  Unknown


RELEASE MECHANISM;  Pipe rupture

        A

DATA SOURCE

     §  Vervalln, Charles H., "Learn from HPI plant fires," Hydrocarbon
        Processing


DESCRIPTION OF RELEASE

     •  About 1000 gallons of the hydrocarbon mixture flowed through a 3/8" pipe
        opening 1n the pump housing from which the pipe plug had fallen.  Most
        was absorbed by the ground, however some flowed about 50 feet fr-om the
        tank through a shallow ditch.


CAUSE OF RELEASE

     •  A pipe plug had fallen or had been blown.  The plug was non-metal He.


RELEASE MATERIALS

     •  How Detected:  fire
     a  Material Types:  hydrocarbon mixture of cyclohexane and n-heptane
     0  Quantities Released:  1000 gallons


RELEASE DURATION


TANK DESIGN AND OPERATING CHARACTERISTICS    -

     a  Tank Use:  storage


RELEASE ENVIRONMENT

     •  Land, Water, A1r, Unknown:  land
     •  Description of contamination:  The leaking fluid was absorbed by the
        ground.  Some of 1t flowed about 50 feet from the tank to a shallow
        ditch.
                                     C-16

-------
                            TANK FAILURE CASE STUDY #15


SITE;  Tank farm of General American Transportation Co.


LOCATION;  San Pedro, California


RELEASE MECHANISM;  Pipe rupture

        •a
DATA SOURCE

     §  Vervalin, Charles H., "Learn from HPI Plant Fires," Hydrocarbon
        Processing, December 1972, pp. 49-50.


DESCRIPTION OF RELEASE

     •  A tank truck collided with a pipe, and fire engulfed the tank and truck.
        The fire spread to nearby trucks.


CAUSE OF RELEASE

     •  A tank, truck apparently sheared a pipe leading to a 30,000 gallon tank.


RELEASE MATERIALS

     •  How Detected:  fire
     0  Material Types:  vinyl acetate
     •  Quantities Released:  30,000 gallons


RELEASE DURATION


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage
     •  Equipment That Failed:  pipe
     •  Dikes/Berms:  diked area around tank farm


RELEASE ENVIRONMENT
       V
     t  Land, Water, Air, Unknown:  land
     •  Description of contamination:
                                   C-17

-------
                            TANK FAILURE CASE STUDY #16


SITE;  011 processing and reclamation facility, Bridgeport Rental and Oil
        Services

LOCATION;  Southern New Jersey


RELEASE MECHANISM;  Tank corrosion

        •i
DATA SOURCE

     •  Whittaker, Kenneth T., Goltz, Robert, "Cost Effective Management of an
        Abandoned Hazardous Waste Site by a Staged Clean-up Approach,"
        Management of Uncontrolled Hazardous Waste Sites, 1982, pp. 262.


DESCRIPTION OF RELEASE


CAUSE OF RELEASE;  Tank corrosion


RELEASE MATERIALS

     •  How Detected:
     •  Material Types:  uncharacterized hydrocarbons, various benzene and
        phenolic polyaromatic hydrocarbons (phenanthene, napthalene) lagoon
        surface — high concentration of solvents and PCBs
     0  Quantities Released: 5 of the 88 on-site 300,000-gallon tanks were empty


RELEASE DURATION


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage


RELEASE ENVIRONMENT

     t  Land, Water, Air, Unknown:  water (surface and groundwater)
     •  Description of contamination:
                                    C-18

-------
                            TANK FAILURE CASE STUDY #17


SITE:   An electronic components manufacturing plant


LOCATION;  A suburban area adjacent to a major city.  The site 1s surrounded by
        residential neighborhoods and small farms.


RELEASE MECHANISM;  Unknown, presumably tank corrosion


DATA SOURCE

     e  Assessment of the Technical, Environmental and Safety Aspects of Storage
        of Hazardous Waste In Underground Tanks, Vol. I, SCS Engineers. Reston,
        Virginia, August 1983, pp. 3-36 - 3-46.


DESCRIPTION OF RELEASE.

     •  Lack of Inventory and/or environmental monitoring, tank Inspection, or
        tank testing programs at this site allowed a waste solvent storage tank
       . leak to go undetected for approximately H years.  The leak material
        contaminated soil and ground water.  As a result of the duration and
        size of the leak and the hydrogeology of the site, transport of the con-
        tamination Into-three aquifers and over an area of about 1/3 square mile
        occurred.


CAUSE OF RELEASE


RELEASE MATERIALS

     t  How Detected:  a mass balance analysis on the solvents entering and
        exiting the plant disclosed a leaking tank
     •  Material Types:  solvents - Acetone, l-l-D1chloroethylene, Freon 113,
        Isopropyl alcohol, 1.1.1-Trlchloroethane, and Xylene
     •  Quantities Released:  58,000 gallons


RELEASE DURATION   1* years


TANK DESIGN AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage


RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:  land
     t  Description of contamination:

                                    C-19

-------
                            TANK FAILURE CASE STUDY #18


SITE;  A manufacturing plant that produces electronic computing equipment,
        semi-conductors and related devices.


LOCATION;  A suburban area adjacent to a major city and Is surrounded by resi-
        dential neighborhoods, small farms, a hospital, and a golf course.


RELEASE.MECHANISM;  Unknown, but Includes a corroded pipe


DATA SOURCE

     t  Assessment of the Technical, Environmental and Safety Aspects of Storage
        of Hazardous Waste 1n Underground Tanks, Vol. I, SCS Engineers. Res ton,
        Virginia, August 1983, pp. 3-47 - 3-64.


DESCRIPTION OF RELEASE


CAUSE OF RELEASE

     •  Some of the probable causes are Improper disposal of the chemicals,
        past operational problems, and a corroded drainline, but most of the
        causes are unknown.


RELEASE MATERIALS

     •  How Detected:  unknown
     •  Material Types:  solvent (acetone; ethyl amyl ketone; Freon 113; Isopro-
        pyl alcohol; 1,1,1-tHchloroethane; 1,1,1-trlchloroethylene; or Xylene)
     t  Quantities Released:


RELEASE DURATION


TANK DESI6N AND OPERATING CHARACTERISTICS

     •  Tank Use:  storage, treatment, sumps
     •  Equipment Material:  concrete, fiberglass, carbon steel, stainless steel

      •
RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:  land
     •  Description of contamination:
                                   C-20

-------
COMMENTS
     a  Lack of inventory and/or environmental monitoring, tank inspection or
        tank testing programs at this site allowed many leaks to go undetected
        for as long as 11 years before detection.  The source of pollution has
        been determined for only one of the three areas found to have soil and
        groundwater contamination.  Transport of the released chemicals into
        three aquifers for over a mile away from the site resulted from the
        duration and size of the leaks and the hydrogeology of the area.

     •  This appears to be the same site as the one discussed in the preceding.
      , case study, but this incident apparently involves a separate set of
        failures.
                                  C-21

-------
                          TANK FAILURE CASE STUDY #19


SITE:  Agricultural chemical manufacturing plant


LOCATION;  Northern California


DATE SOURCE

     • -^United States Environmental Protection Agency, Case Studies 1-23:
        Remedial Response at Hazardous Waste Sites. EPA-540/2-84-0026, March
        1984 (Case Study #2).


RELEASE MECHANISM

     0  Pipe corrosion or rupture, accidental pills, and tank corrosion or
        rupture.
                •

DESCRIPTION OF RELEASE

     •  The tanks in question were part of the treatment system for rainfall
        run-off and rinsewater from the plant's chemical handling areas.  15,000
        gaMons leaked from an underground "skimmer tank."  There were also a
        number of" small-scale chemical spills, and leakage from 2 joints in a
        300' chemical drain used to connect parts of the system.


CAUSES OF RELEASE;  Unknown, but probably include corrosion or rupture of both
        the tank and the drain.


RELEASE MATERIALS

     0  How Detected:  "Foul taste" it) nearby drinking wells.
     •  Material Types:  Toluene and various herbicides.
     a  Quantities Released:  In excess of 15,000 gallons.


RELEASE DURATION

     •  Unknown.  The system was constructed In 1971; the leak was discovered in
        1979.

       •
RELEASE ENVIRONMENT


     •  Land, Water, Air, Unknown:  land
     •  Description of contamination:  There was contamination of shallow
        groundwater.  Nearby drinking wells had a "foul taste," but no detec-
        table chemicals.


                                   C-22

-------
                          .TANK FAILURE CASE STUDY #20


SITE:  Blocraft Laboratories


LOCATION;  Waldwick, NJ


RELEASE MECHANISM;  Gasket leak


DATA SOURCE;

     •  United States Environmental Protection Agency, Case Studies 1-23:
        Remedial Response at Hazardous Waste Sites. EPA-540/2-84-0026, March
        1984 (Case Study 14).


DESCRIPTION OF RELEASE

     •  A gasket in a fill pipe for an underground storage tank disintegrated
        due to incompatibility with the waste.


CAUSE OF RELEASE

     t  Incompatibility between gasket and waste


RELEASE MATERIALS

     •  How Detected:  Groundwater testing
     •  Material Types:  Methylene chloride, N-butyl alcohol, dimethyl aniline,
        acetone, and a variety of trace organics.
     t  Quantities released:  Uncertain—possibly as much as 360,000 Ibs.


RELEASE DURATION:  Probably 3 years


TANK DESIGN AND OPERATING CHARACTERISTICS

     0  Tank use:  storage
     t  Equipment that failed:  gasket
     0  Tank design:  steel tank with no secondary containment and apparently no
        'corrosion protection
     •  Age of system:  new (constructed In 1972, failure detected 1n 1975)
                                    C-23

-------
RELEASE ENVIRONMENT;

     •  Land, Water, Air, Unknown:  land
     •  Extent of contamination:  Groundwater was contaminated.   The release was
        also the probable cause of a fish kill in a nearby stream in 1973.
                                 C-24

-------
                           TANK  FAILURE  CASE  STUDY  #21


SITE;  General  Electric  transformer  manufacturing  and  repair facility


LOCATION;   Oakland,  CA


RELEASE MECHANISM;   Accidental  spills,  overflows,  and  tank rupture

          r
DATA SOURCE;

     •  United  States Environmental  Protection Agency, Case Studies 1-23:
        Remedial Response  at Hazardous  Waste Sites. EPA-540/,2-84-0026, March
        1984  (Case Study f9).


DESCRIPTION OF  RELEASE

     •  Over  the operating history of the plant, a number of spills had occurred
        (a) at  an above-ground  tank  farm used for  a petroleum-based thinner and
        oil;  (b) near two  above-ground  5,000-gallon tanks used for Pyranol
        (contains PCS);  (c) 1o  the area where rail tank cars were unloaded by
        pumping; (d) possibly due to minor leakage from oil-warming operations
        Inside  the building; and (e) from a mobile filtering unit that would
        occasionally "blow" from too much pressure.

        Additional contamination came from trench  burial of liquid PCB's and
        contaminated solids such as  d1alectrie paper,  and from continued
        discharges from  a lab sink following the collapse of a septic tank (date
        of  collapse  Is unknown.  Discharges continued  until the mid 1960's).


CAUSES OF RELEASE

     •  Above-ground spills
     0  Overflows
     0  Tank rupture


RELEASE MATERIALS

     0  How detected:
     0  Material types:  Hydrocarbon products and  PCB-contaminated oils
     0  Quantities released:  20,000 gallons
        *

RELEASE DURATION;  Miscellaneous spills between 1927 and 1975.
                                   C-25

-------
RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:  land                      ^  :
     0  Description of contamination:  surface spills and underground leaks,
        There was widespread contamination of on-site soils and groundwater,
                                  C-26

-------
                          TANK FAILURE CASE STUDY #22


SITE;  Houston Chemical Company


LOCATION;  Houston, MO.


RELEASE MECHANISM;  Tank rupture and pipe design error


DATE SOURCE

     •  United States Environmental Protection Agency» Case Studies 1-23;
        Remedial Response at Hazardous Waste Sites, EPA-540/2-84-0026, March
        1984 (Case Study fi3>.


DESCRIPTION OF RELEASE

     •  The tank was a 21,000-galIon,.steel, horizontal, above-ground storage  •
        tank.  There was no containment system.  It collapsed for several
        reasons:

           • The saddle support blocks were not sufficient either in spacing or
             number.

           - there were weaknesses due to corrosion and previous abuse.

           » The saddle support blocks were not engineered to fit the curvature
             of the tank.

           • A drain pipe and valve were Installed on the underside of tank.
             When the tank collapsed, the drain control valve and piping
             sheared off.  This was a design error.

        An overflow pit contained 5X of the spill.  The remainder bypassed the
        pit due to the absence of suitable dikes.  The oil flowed Into a road-
        side ditch, under a culvert and Into a catch basin where 1t remained
        temporarily.  It then infiltrated the ground, reappearing 125' downgra-
        dlent.  Eventually it flowed into a farm pond where the water level was
        low enough that It remained.

      . The Initial report was not received by EPA until 4 days after the spill.
        That report falsely stated that the spill had been contained by a dike
        and that clean-up was under way. .The volume of the spill was initially
        reported as 10,000 gal.
                                   C-27

-------
RELEASE MATERIALS

     •  How Detected:
     0  Material Types:  5X solution of PCP in diesel oil
     •  Quantities released:  15,000 gal.
RELEASE DURATION:  Release occurred on June 14, 1979.
TANK DESIGN AW OPERATING CHARACTERISTICS

     •  Tank Use:  Storage
     •  Tank Design:  21,000-gallon, steel, horizontal, above-ground, on
        cradles.
     •  Equipment that failed:  cradles
     •  Secondary containment:  present, but Inadequate.


RELEASE ENVIRONMENT

     t  Land, Water, A1r, Unknown:  land and water
     •  Description of Contamination:  There was a total fish kill  1n the farm
        pond and a threat of overflow Into a navlgatable river known as a
        valuable wildlife habitat.  Soil along the path of flow was con-
        taminated.
                                   C-28

-------
                           TANK  FAILURE  CASE  STUDY  #23


NAME;  Howe, Inc.


LOCATION:  Brooklyn Center, MN


RELEASE MECHANISM;  Hazardous materials carried off-site by smoke from fire and
        run-off water from fire-fighting efforts.
        *

DATE SOURCE

     •  United States Environmental Protection Agency, Case Studies 1-23;
        Remedial Response  at Hazardous Haste Sites. EPA-540/2-84-0026, March
        1984 (Case Study fI4j.


DESCRIPTION OF RELEASE

     •  In January, 1979.,  a fire occurred at a warehouse site containing 100
        different pesticides totaling 80 tons of active Ingredients.  Water
        used to fight the  fire  flowed off-site, carrying with 1t dissolved
        pesticides and herbicides.  Several additional dangers were Involved:

           .'-;.A1r pollution from combustion of organic solvents.  Pigeons
              flying through the plume died Immediately.  Eleven fire fighters
              became 111.

            - Fallout from the  plume.

            - Contaminated building debris.

            - Run-off from contaminated soils.


CAUSE OF RELEASE;  Faulty  acetylene torch


RELEASE MATERIALS

     0  How detected:  Immediate visual detection
     •  Material types:  Pesticide- and herbicide-contaminated water; fumes.
     e  Quantities released:  500,000 gallons of contaminated water; unknown
        amount of. fumes.
       •

RELEASE DURATION;  several hours
                                     C-29

-------
RELEASE ENVIRONMENTS

     0  Land, Water, Air, Unknown:  land and air

     •  Description of contamination:  A1r and land surface.  Because of cold
        temperatures, the contaminated water froze on the ground surface.
                                   C-30

-------
                           TANK  FAILURE CASE STUDY  124


NAME;  N. W. Mauthe,  Inc.


LOCATION:  Appleton,  Wisconsin


RELEASE MECHANISM;  Accidental  spills; cracks 1n concrete floor.

       *
DATA SOURCE

     •  United States Environmental Protection Agency, Case Studies 1-23;
        Remedial Response  at Hazardous Waste Sites, EPA-540/2-84-0026, March
        1984 (Case Study 116).


DESCRIPTION OF RELEASE

     •  A blower vent for  a chrome-plating tank discharged chromium-laden m1st~"
        to the outside;

     •  Drippings from chromatlng tanks were channeled to a sanitary sewer by a
        trough 1n the floor.  Cracks 1n the trough and the concrete flooring led
        to -seepage Into underlying soil.
             »„

CAUSE OF RELEASE

     •  Poor design and aging of concrete floor


RELEASE MATERIALS

     •  How Detected:  In March, 1982, yellow puddles were observed on adjacent
        property.
     •  Material Types:  Chromium-contaminated water.
     •  Quantities released:


RELEASE DURATION

     •  The shop operated from  1966 through 1976.  The releases were probably
        ongoing through much or all of that time.
                                    C-31

-------
RELEASE ENVIRONMENT

     •  Land, Water, Air, Unknown:  land
     •  Description of contamination:  Continuing spills contaminated soils at
        the site.  The contaminants migrated off-site, where they were disco-
        vered as "yellow puddles."  There was a threat to nearby residences and
        schools and a threat of run-off to storm sewers leading to the Fox
        River.
                                        C-32

-------
                          TANK FAILURE CASE STUDY #25


NAME;  Quanta Resources


LOCATION;  Queens, NY


RELEASE MECHANISM;  Vandalism

        ^
DATA SOURCE

     •  United States Environmental Protection Agency, Case Studyes 1-23;
        Remedial Response at Hazardous Waste Sites. EPA - 540/2-84-0026,
        March 1984 (Case Study fl9).


DESCRIPTION OF INCIDENT

     •  The facility was a processing facility containing about 500,000 gal. of
        miscellaneous wastes, Including PCB-contam1nated oils, cyanides, heavy
        metals, and low flash-point chlorinated solvents.  Bankruptcy of the
        owner left the facility without security against arson or vandalism.

            *t • •
CAUSE OF RELEASE

     •  None—no release actually occurred.  However, the unguarded state of
        the facility posed a substantial risk of arson or vandalism.
                                   C-33

-------
                           TANK  FAILURE  CASE  STUDY  #26


NAME;  White's Septic  Tank Services


LOCATION;   OuPage  County,  IL


RELEASE MECHANISM;   Operator  error


DATA SOURCE

     •  Landfilling  of Special  and Hazardous Waste 1n Illinois;  A Report to
        the Illinois General  Assembly.  Illinois Legislative Investigating
        Commission,  19//.


DESCRIPTION OF INCIDENT

     •  White  operated a land treatment facility on his 30-acre farm.  He was  *
        permitted to accept only domestic septage.   He could spread It only In
        good weather and not  near the river.  Never- theless, he accepted.com-
        mercial wastes of  unknown nature, mixing them with the septage In his
        trucks.  He  professed Ignorance of his permit requirements and was
        "cordial.11   His record-keeping was virtually non- existent.  According
        to  the  report, he  was "totally  lacking in the skill, equipment,
        knowledge, and desire necessary [for] toxic waste disposal...--blythely
        spread  Cingj Industrial wastes over farm land."
COMMENTS
        White had also been 1n the same business at different times at other
        Illinois sites.

        This case study does not Involve hazardous waste tanks.  It Is never.
        theless relevant, because tank operators can be just as untrained as
        landfarm operators.
                                   C-34

-------
                          TANK FAILURE CASE STUDY 127


NAME;  White's Septic Tank Services


LOCATION;  DuPage County, IL


RELEASE MECHANISM;  Inspection error


DATA SOORCE

     •  Landfill1ng of Special and Hazardous Waste In Illinois;  A Report to
        the Illinois General Assembly. Illinois Legislative Investigating
        Commission, 19/7.


DESCRIPTION OF INCIDENT (see preceding case study)

     •  An Illinois Department of Public Health Inspector made a routine visit
        to White's landfarm on 7/6/78, accompanied by an Illinois Legislative
        Investigation Commission observer.  The visit Involved an applica-
        tion for license renewal,  the IDPH Inspector:

        -  Was uncertain of what the law required;

        •  Was''unaware of the danger of leachate contaminating river;

        -  Took no water samples;

        -  Placed a checkmark next to "Inspection of servicing equipment"
           without ever going near the one truck that made a dump while he was
           on the site; and    .

        •  Did not seem particularly knowledgeable about the operating require-
           ments.


COMMENTS

     •  This case study does not Involve hazardous waste tanks.  It Is never-
        theless relevant, because similar Inspection errors could occur at any
        type of hazardous waste facility.
                                   C-35

-------
                          TANK FAILURE CASE STUDY #28


NAME;  Oestructol Carol awn


LOCATION:  Kernersvllle, NC


DATA SOURCE;  EPA/SCS Remedial Action Cases, Slte'D

       ^
RELEASE MECHANISM;  Catastrophic release (vandalism)


DESCRIPTION OF RELEASE

     •  On June 3, 1977, vandals opened the valves on six storage tanks at a
        commercial hazardous waste Incinerator.  There were no locks on the
        valves.  There was no secondary containment.


RELEASE MATERIALS:

     t  How Detected:  Immediate visual observation
     •  Material Types:
     •  Quantities Released:  30,000 gal.
         •   tf

RELEASE ENVIRONMENT;

     •  Land, Water, Air, Unknown:  land and water
     t  Description of contamination: There was a 90-99* fish kill  in a nearby
        50-acre reservoir.  200 local residents were temporarily evacuated.
        Temporary water rationing and Industrial layoffs resulted, as the muni-
        cipality sought an alternative water source.  So far,  there has been
        permanent loss of the reservoir as a drinking water source (the state
        refused to approve the reservoir as a drinking water source as long as
        the threat of future contamination remained).
                                  C-36

-------