PETROLEUM REFINING PROCESS WASTE LISTING DETERMINATION
SUPPLEMENTAL BACKGROUND DOCUMENT

GROUNDWATER PATHWAY RISK ANALYSIS








US Environmental Protection Agency
Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009)Washington, DC

March 1997

                              TABLE OF CONTENTS

                                                       Page

1.0  BACKGROUND                                        1-1
     1.1  GENERAL BACKGROUND                           1-1
          1.1.1  General Overview of Analyses          1-1
     1.2  WASTESTREAMS AND MANAGEMENT SCENARIOS        1-2
     1.3  REPORT ORGANIZATION                          1-2

2.0  GENERAL METHODOLOGY                               2-1
     2.1  MODELING APPROACH                            2-1
          2.1.1  EPACMTP Model                         2-1
          2.1.2  Contaminant Source Term Modeling      2-3
          2.1.3  Processes in Subsurface Fate and 
                 Transport Modeling                    2-4
          2.1.4  Regional Site-Based Modeling Approach 2-5
     2.2  MODELING SCENARIOS AND INPUT DATA            2-6
          2.2.1  Data Sources                          2-10
          2.2.2  Hydroprocessing Waste                 2-11
          2.2.3  Multi Wastestream Codisposal          2-11
          2.2.4  Contingent CSO Sludge Management      2-14
          2.2.5  TC Rule Constrained Maximum TCLP      2-16
          2.2.6  Off-Spec Products and Fines TCLP          
                 Data                                  2-16

3.0  DETERMINISTIC GROUNDWATER PATHWAY ANALYSES        3-1
     3.1  BOUNDING ANALYSES                            3-1
     3.2  TWO HIGH-END PARAMETERS BASELINE ANALYSES    3-1
          3.2.1  1995 Proposal Results                 3-3
          3.2.2  Revised Two High-End Parameters          
                 Analyses                              3-3
          3.2.3  Current Two Parameter Sensitivity        
                 Analysis                              3-4
                 3.2.3.1  Approach                     3-4
                 3.2.3.2  Sensitivity of Parameters    3-5
     3.3  CENTRAL TENDENCY ANALYSIS                    3-5
          3.3.1  1995 (Previous) Analysis              3-5
          3.3.2  1995 (Current) Central Tendency          
                 Analysis                              3-5
     3.4  TWO HIGH-END PARAMETERS ANALYSIS FOR            
          CODISPOSAL                                   3-5

4.0  MONTE CARLO ANALYSES PROCEDURE                    4-1
     4.1  SINGLE-WASTESTREAM BASELINE                  4-1
          4.1.1  On-Site Baseline Monte Carlo Analysis 4-1
                 4.1.1.1  Procedure                    4-4
                 4.1.1.2  Modeled Wastestreams and        
                          Constituents                 4-5
          4.1.2  Off-Site Baseline Monte Carlo            
                 Analysis                              4-5
                 4.1.2.1  Procedure                    4-5
                 4.1.2.2  Modeled Wastestreams and        
                          Waste Constituents           4-7
     4.2  CODISPOSAL MONTE CARLO ANALYSIS              4-7
          4.2.1  On-Site Codisposal Monte Carlo           
                 Analysis Procedure                    4-7
                 4.2.1.1  Procedure                    4-8
          4.2.2  Off-site Codisposal Monte Carlo          
                 Analysis Procedure                    4-10
                 4.2.2.1  Procedure                    4-10

5.0  MODELING RESULTS AND GROUNDWATER RISK             5-1
     5.1  RISK CALCULATION                             5-1
          5.1.1  Direct Groundwater Risk               5-1
          5.1.2  Indirect Groundwater Risk             5-1
          5.1.3  Total Risk                            5-1
     5.2  MODELING RESULTS                             5-2
          5.2.1  Two High-End Parameters Results       5-2
          5.2.2  Central Tendency Results              5-2
          5.2.3  Monte Carlo Results                   5-2
          5.2.4  Contingent CSO Management             5-2
          5.2.5  Codisposal Results                    5-2
          5.2.6  TC Capped Results                     5-2

6.0  REFERENCES                                        6-1


APPENDIX A:  Waste and Leachate Concentrations
APPENDIX B:  Landfill Size and Waste Volume Data
APPENDIX C:  Sensitivity Analyses Data and Results
APPENDIX D:  Calculation of Two High-End Codisposal         
             Parameters
APPENDIX E:  Baseline Modeling Results
                               LIST OF FIGURES
                                                       Page


Figure 2.1  Schematic view of the modeled subsurface      
            pathway                                    2-2

Figure 2.2  Flowchart of EPACMTP for the Regional         
            Site-Based Monte Carlo Approach            2-8











                              LIST OF TABLES
                                                       Page


Table 1.1  Petroleum Refining Listing Wastestreams and    
           Landfill Scenarios                          1-3

Table 1.2  Petroleum Refining Study Wastestreams       1-4

Table 2.1  Petroleum Refining Modeling Analyses           
           Performed in 1995 and 1996                  2-9

Table 2.2a Off-Site CSO Contingent Management Data     2-14

Table 2.2b On-Site CSO Contingent Management Data      2-15

Table 3.1  Summary of 1995 Groundwater Pathway Analysis   
           Results                                     3-2

Table 3.2  Comparison of 1995 and 1996 Modeled Waste      
           Quantities for Three Waste Streams          3-4

Table 3.3  Summary of Two High-End Parameter              
           Sensitivity Analysis                        3-7

Table 3.4  Two High-End Parameter Input Data for          
           Codisposal Scenario                         3-8

Table 4.1  Description of Monte Carlo Input Data and      
           Data Sources                                4-2

Table 4.2  Modeled Wastestreams and Constituents for      
           On-Site Scenarios                           4-5

Table 4.3  Modeled Wastestreams and Constituents for      
           Off-Site Scenarios                          4-7

Table 5.1  Summary of  Maximum Groundwater Risks       5-3

Table 5.2  Deterministic Groundwater Pathway Analysis     
           Results from On-Site Landfills              5-4

Table 5.3  Deterministic Groundwater Pathway Analysis     
           Results from Off-Site Landfills             5-5

Table 5.4  Monte Carlo Groundwater Pathway Analysis       
           Results from On-Site Landfills              5-6

Table 5.5  Monte Carlo Groundwater Pathway Analysis       
           Results from Off-Site Landfills             5-7

Table 5.6  CSO Contingent Management Scenario Results  5-8

Table 5.7  Codisposal Scenario Results                 5-9
Table 5.8  Off-Spec Products and Fines Summary of         
           Analyses Results for Benz(a)anthracene      5-10

Table A.1  Benzene Concentration Data                  A.1

Table A.2  Arsenic Concentration Data                  A.3

Table A.3  Nickel Concentration Data                   A.5

Table A.4  Benz(a)anthracene Concentration Data        A.7

Table A.5  Toluene Concentration Data                  A.9

Table A.6  Benzo(a)pyrene Concentration Data           A.10

Table A.7  Cadmium Concentration Data                  A.11

Table A.8  Vanadium Concentration Data                 A.12

Table A.9  3/4-Methylphenol Concentration Data         A.13

Table B.1  On-Site Landfill Source Input Data          B.1

Table B.2  Off-Site Landfill Waste Volume Data         B.7

Table C.1  Two-High End Parameters Determined from        
            Sensitivity Analyses                       C.1

Table C.2  Parameter Distributions for CSO                
           Sludge/Benzene                              C.2

Table C.3  Sensitivity Analysis CSO Sludge On-Site        
           Landfill/Benzene                            C.3

Table C.4  Sensitivity Analysis CSO Sludge Off-Site       
           Landfill/Benzene                            C.4

Table C.5  Parameter Distributions for Contingent Mgmt.   
           of CSO Sludge/Benzene                       C.5

Table C.6  Sensitivity Analysis for Contingent Mgmt. of   
           CSO Sludge On-Site Landfill/Benzene         C.6

Table C.7  Sensitivity Analysis for Contingent Mgmt. of   
           CSO Sludge Off-Site Landfill/Benzene        C.7

Table C.8  Parameter Distributions for Crude Oil Tank     
           Sludge/Benzene                              C.8

Table C.9  Sensitivity Analysis for Crude Oil Tank Sludge 
           On-Site Landfill/Benzene                    C.9

Table C.10 Sensitivity Analysis for Crude Oil Tank Sludge 
           Off-Site Landfill/Benzene                   C.10

Table C.11 Parameter Distributions for Hydrotreating   
           Catalyst/Benzene                            C.11

Table C.12 Sensitivity Analysis for Hydrotreating          
           Catalyst On-Site Landfill/Benzene           C.12

Table C.13 Sensitivity Analysis for Hydrotreating          
           Catalyst On-Site Landfill/TC Capped Benzene C.13

Table C.14 Sensitivity Analysis for Hydrotreating          
           Catalyst Off-Site Landfill/Benzene          C.14

Table C.15 Sensitivity Analysis for Hydrotreating          
           Catalyst Off-Site Landfill/TC Capped            
           Benzene                                     C.15

Table C.16 Parameter Distributions for Hydrotreating       
           Catalyst/Arsenic                            C.16

Table C.17 Sensitivity Analysis for Hydrotreating          
           Catalyst On-Site Landfill/Arsenic           C.17

Table C.18 Sensitivity Analysis for Hydrotreating          
           Catalyst Off-Site Landfill/Arsenic          C.18

Table C.19 Parameter Distributions for Hydrorefining       
           Catalyst/Benzene                            C.19

Table C.20 Sensitivity Analysis for Hydrorefining          
           Catalyst On-Site Landfill/Benzene           C.20

Table C.21 Sensitivity Analysis for Hydrorefining          
           Catalyst On-Site Landfill/TC Capped Benzene C.21

Table C.22 Sensitivity Analysis for Hydrorefining          
           Catalyst Off-Site Landfill/Benzene          C.22

Table C.23 Sensitivity Analysis for Hydrorefining          
           Catalyst Off-Site Landfill/TC Capped            
           Benzene                                     C.23

Table C.24 Parameter Distributions for Hydrorefining       
           Catalyst/Arsenic                            C.24

Table C.25 Sensitivity Analysis for Hydrorefining          
           Catalyst On-Site Landfill/Arsenic           C.25

Table C.26 Sensitivity Analysis for Hydrorefining          
           Catalyst On-Site Landfill/TC Capped Arsenic C.26

Table C.27 Sensitivity Analysis for Hydrorefining          
           Catalyst Off-Site Landfill/Arsenic          C.27

Table C.28 Sensitivity Analysis for Hydrorefining          
           Catalyst Off-Site Landfill/TC Capped            
           Arsenic                                     C.28

Table C.29 Parameter Distributions for HF Alkylation       
           Sludge/Benzene                              C.29

Table C.30 Sensitivity Analysis for HF Alkylation Sludge   
           On-Site Landfill/Benzene                    C.30

Table C.31 Sensitivity Analysis for HF Alkylation Sludge   
           Off-Site Landfill/Benzene                   C.31

Table C.32 Parameter Distributions for Off-Spec. Products  
           and Fines/Benz(a)anthracene                 C.32

Table C.33 Sensitivity Analysis for On-Site Off-Spec.      
           Products and Fines/Benz(a)anthracene        C.33

Table C.34 Sensitivity Analysis for Off-Site Off-Spec.     
           Products and Fines/Benz(a)anthracene        C.34

Table D.1  Codisposal Data for Two High-End Parameter     
           Analysis, Arsenic in On-Site Landfills      D.1

Table D.2  Codisposal Data for Two High-End Parameter     
           Analysis, Benzene in On-Site Landfills      D.2

Table D.3  Codisposal Data for Two High-End Parameter     
           Analysis, Benzene with TC Cap in On-Site       
           Landfills                                   D.3

Table D.4  Codisposal Data for Two High-End Parameter     
           Analysis, Arsenic in Off-Site Landfills with   
           HydroCracking Catalyst                      D.4

Table D.5  Codisposal Data for Two High-End Parameter     
           Analysis, Benzene in Off-Site Landfills with   
           HydroCracking Catalyst                      D.5

Table D.6  Codisposal Data for Two High-End Parameter     
           Analysis, TC Capped Benzene in Off-Site        
           Landfills with HydroCracking Catalyst       D.6

Table D.7  Codisposal Data for Two High-End Parameter     
           Analysis, Arsenic in Off-Site Landfills        
           without HydroCracking Catalyst              D.7

Table D.8  Codisposal Data for Two High-End Parameter     
           Analysis, Benzene in Off-Site Landfills        
           without HydroCracking Catalyst              D.8

Table D.9  Codisposal Data for Two High-End Parameter     
           Analysis, TC Capped Benzene in Off-Site        
           Landfills without HydroCracking Catalyst    D.9
Table E.1  Two High-End Parameter Analysis Results of     
           On-Site Landfills                           E.1

Table E.2  Monte Carlo Baseline Modeling Results for      
           On-Site Landfills                           E.2

Table E.3  Two High-End Parameter Analysis Results of     
           Off-Site Landfills                          E.3

Table E.4  Monte Carlo Baseline Modeling Results for      
           Off-Site Landfills                          E.4
1.0   BACKGROUND


1.1  GENERAL BACKGROUND

This document describes groundwater pathway analyses 
performed to support the petroleum refining listing
determination.  These analyses supplement the analyses
carried out earlier as part of the EPA's proposal for the
petroleum refining listing determination (USEPA, 1995a). 
The analyses were designed to determine the potential
exposure, via the groundwater pathway, to human receptors
from petroleum refining wastes, being managed and/or
disposed in land management units.  The exposure is
expressed in terms of the contaminant concentration at a
groundwater extraction well located down-gradient from the
waste-management unit.  

The contaminant concentrations were obtained using the EPA's
Composite Model for Leachate Migration with Transformation
Products (EPACMTP) (USEPA, 1996c ,d ,e; USEPA, 1997d).  
EPACMTP simulates the subsurface fate and transport of waste
constituents leaching from land disposal units. 
Wastestreams and waste constituents of potential concern are
identified by comparing the model predicted exposure
concentrations to health-based numbers (HBNs) for both
carcinogens and non-carcinogens. 

In the current modeling analyses, in addition to the  two
high-end parameter and the central tendency analyses,
Monte-Carlo analyses were also performed to assess the
potential groundwater exposures due to dissolved chemicals
associated with the disposal of petroleum refining wastes in 
waste management units.  The EPACMTP model was selected
because of its capabilities to perform deterministic and
full, Monte Carlo-based, probabilistic exposure assessments.

1.1.1  General Overview of Analyses

EPA received a number of comments related to the groundwater
pathway analyses supporting the November 20, 1995 notice of
proposed rulemaking (60 FR 57747, November 20, 1995). In
responding to these comments, EPA performed a variety of
additional analyses.  A brief overview of what analyses EPA
undertook and the general nature of the comments that
prompted such analyses, are given below. 

Volume Inputs and Waste Fractions.  EPA received comments
regarding the Agency's selection of volume inputs and the
fraction of the landfill waste that was assumed to be the
petroleum waste.  As a result, EPA reexamined the approach
it used and conducted additional analyses to better assess
groundwater risks.   EPA subsequently made revisions to the
deterministic modeling approach, as described in Section
3.0.  In addition, EPA conducted a full,  Monte Carlo based,
probabilistic exposure assessments as a way of confirming
the deterministic modeling analyses.  The Monte Carlo
methodology  is presented in Section 4.0. 

Codisposal of Multiple Wastestreams.  In the proposed rule,
EPA assumed that only one wastestream is managed/disposed of
in a unit at any time.  One of the comments about volume
inputs suggested that EPA should consider co-disposal of the
evaluated wastes  with other refinery wastes. In response,
the Agency evaluated the potential groundwater exposures due
to the impact of codisposal of multiple wastestreams
(codisposal scenarios) as well as the disposal of single
wastestreams (baseline scenario). These results are
described in Sections 3.4, 4.2 and 5.2.5.  

Conditional Exemption and Contingent Management.  In the
proposed rule, EPA raised the option of a conditional
exemption for one waste, Clarified Slurry Oil Storage Tank
Sediment (CSO), if the regulations limited the way the waste
is managed and disposed, i.e, a contingent management
listing.  Commenters indicated that EPA should assess what
impact the conditional exemption might have on the listing
decision, i.e., examine the assumption that unless the waste
is unconditionally listed all of it could potentially be
disposed in Subtitle D landfills. This could substantially
increase the volume of CSO wastes going to the allowed
(Subtitle D) disposal practice (e.g., landfills), since
wastes previously land treated or managed in other ways may
be landfilled  (Subtitle D) to take advantage of the
exemption from Subtitle C regulations.  EPA evaluated the
potential risk of increased landfilling as described in
Sections 2.3.4 and 5.2.4.

Noningestion Risks from Groundwater Contamination. 
Commenters pointed out that the Agency ignored the
noningestion risks arising from the use of contaminated
groundwater.  Such risks might arise from the inhalation of 
contaminants transferred to the air from showers, baths,
etc. In response, EPA examined additional modeling
approaches, as described in the Supplemental Background
Document for Nongroundwater risk assessment pathways (USEPA,
1997c).   The description of the calculation of the combined
risk is presented in Section 5.1.  Details of the models
used to estimate exposure factors for inhalation and dermal
risk are given in Section 6 of the Supplemental Background
Document for Nongroundwater Risk Assessment  (USEPA, 1997c).


Use of TCLP data that exceeds the Toxicity Characteristic. 
Commenters stated that EPA should calculate risks from
Subtitle D landfilling of residuals using only the data that
do not exceed the Toxicity Characteristic level
(40CFR261.24), because such wastes should be handled as
hazardous.  The analyses conducted in response to this
comment can be found in Section 5.2.6, and the input data
are discussed in Section 2.3.

1.2  WASTESTREAMS AND MANAGEMENT SCENARIOS
 
The groundwater pathway analysis involved evaluating land
disposal of a total of 12 listing wastestreams (Table 1.1)
and 11 study wastestreams (Table 1.2).  A complete
description of listing and study wastestreams is presented
in the appropriate background documents (USEPA, 1995a and
USEPA, 1996a).  Both single-wastestream disposal (baseline)
and multi-wastestream disposal (codisposal) scenarios were
evaluated.  Table 1.1 presents a summary of the modeled
listing wastestreams and landfill modeling scenarios.  The
study wastestreams (Table 1.2) were included only in the
codisposal scenario.

1.3  REPORT ORGANIZATION
 
General technical approaches are outlined in Section 2. 
Detailed descriptions for the deterministic and Monte Carlo
scenarios are presented in Sections 3, and 4, respectively. 
Modeling results and exposure risk calculations are
presented in Section 5.


Table 1.1  Petroleum Refining Listing Wastestreams and       
           Landfill Scenarios.

Table 1.2  Petroleum Refining Study Wastestreams.


2.0  GENERAL METHODOLOGY


This section outlines the general methodology used in the
groundwater pathway analysis. It includes: an overview of
the modeling approach, modeling scenarios, and descriptions
of the data sources used in the analysis.

2.1  MODELING APPROACH

The groundwater pathway analysis determines the groundwater
exposure concentrations resulting from the release of waste
constituents from the waste management unit into the
subsurface, and the subsequent fate and transport of the
constituent, through the vadose zone and underlying
saturated zone. The exposure concentration is evaluated at
the intake point of a hypothetical groundwater drinking
water well, located at a specified distance from the
downgradient edge of the waste management unit.  This well
is referred to hereafter as the 'receptor well'.  The
modeled subsurface pathway is depicted schematically in
Figure 2.1.  The modeling approach may be divided into two
major steps: (1) the release of waste constituents into the
subsurface, i.e., the model source term characterization,
and (2) the fate and transport of constituents in the
subsurface(in the vadose zone and the saturated zone).  The
modeling approach for the groundwater pathway analysis
involves aqueous phase migration of waste constituents.
Considering that the wastes associated with petroleum
refining operations may contain a significant amount of
non-aqueous phase liquids, i.e, "oil", the potential may
exist for migration of waste constituents via non-aqueous
phase flow. However, an evaluation conducted for the 1995
Analysis indicated that multiphase flow would not be
expected to be a concern for the wastestreams investigated. 
Laboratory analysis of  petroleum waste samples also shows
that the waste actually contains no or very little free
flowing oil (USEPA, 1997a).  Furthermore, it is very likely
that petroleum waste parcels are buffered by daily soil
cover, and other waste parcels with little or no
free-flowing oil.  For these reasons, the flow and transport
in the vadose and saturated zones are considered significant
in the dissolved phase only, and the EPACMTP model (USEPA,
1996c) is appropriate for the modeling analyses.

In the ensuing subsections, details of the following are
presented:

     - The EPACMTP model;
     - Source-term modeling in the EPACMTP model;
     - Processes in subsurface fate and transport in the     
       EPACMTP model; and
     - Monte Carlo simulation with a regional site-based     
       approach using EPACMTP.

2.1.1  EPACMTP Model

EPACMTP (EPA's Composite Model for Leachate Migration with
Transformation Products) (USEPA, 1995 a, b, c, d) is a
computer simulation model for modeling the subsurface fate
and transport of contaminants leaching from a land disposal
site, e.g., landfill, surface impoundment, waste pile, or
land application unit.  Fate and transport processes
accounted for in the model are: advection, hydrodynamic
dispersion, linear or nonlinear sorption, and chained-decay
reactions. In cases where degradation of a waste constituent
yields daughter products that are also of concern, EPACMTP
has the capability to simulate the fate and transport of up
to six (grand-)daughter products. The composite model
consists of a one-dimensional module that simulates
infiltration and dissolved constituent transport through the
unsaturated zone, and which is coupled to a
three-dimensional saturated zone flow and transport module.
The saturated zone flow module accounts for the processes
affecting the magnitude and direction of groundwater flow
such as leaching from the waste management unit and regional
recharge. The saturated zone transport module accounts for
three-dimensional advection and dispersion, chained-decay
reactions involving up to seven different chemical species
(i.e., parent compound and up to six daughter products), and
linear or nonlinear equilibrium sorption. EPACMTP simulates
steady state flow in both the unsaturated zone and the
saturated zone; contaminant transport can be either steady
state or transient. The steady state modeling option is used 
for continuous source modeling scenarios; the transient
modeling option is used for finite source modeling
scenarios, with optional accounting for source depletion.
EPACMTP predicts the contaminant concentration arriving at a
down gradient groundwater receptor well. This can be either
a steady state concentration value, corresponding to the
continuous source scenario, or a time-dependent
concentration, corresponding to the finite source scenario. 
In the latter case, the model can calculate either the peak
concentration arriving at the well, or a time averaged
concentration, corresponding to a specified exposure
duration, e.g., a nine year average residence time.  EPACMTP
has the capability to perform Monte Carlo simulations to
account for parametric uncertainty or variability. The flow
and transport simulation modules of EPACMTP are linked to a
Monte Carlo driver which permits a probabilistic evaluation
of uncertainty in model input parameters, as described by
specified (joint) probability distributions.

EPACMTP replaces the EPA's Composite Model for Landfills
(EPACML) which was used in 1990 Toxicity Characteristic (TC)
Rule (55FR11798).  EPACMTP extends the capabilities of the
earlier EPACML model. The enhanced capabilities include
accounting for three-dimensional groundwater flow, the
finite source and transformation products options, and
capability to simulate metals transport with nonlinear
sorption isotherms through linkage with the MINTEQ
geochemical speciation model.  EPACMTP has been published in
an international refereed journal (Kool, Huyakorn, Sudicky,
and Saleem, 1994).  It also has been extensively reviewed. 
The SAB (USEPA's Science Advisory Board) commended the
Agency for its significant improvements to the model.  They
also stated that it represents the state of the art for such
analyses.  However, they also encouraged additional
validation studies, especially for the metals (USEPA,
1995c).

2.1.2  Contaminant Source Term Modeling

The release of contaminants into the subsurface constitutes
the source term for the fate and transport model.  The
conceptual differences between a landfill and other waste
management scenarios are reflected in how the model source
term is characterized in different scenarios.  The modeled
subsurface fate and transport processes are the same for
each waste management scenario.  The contaminant source term
for the EPACMTP fate and transport model is defined in terms
of four primary parameters: (1) area of the waste unit, (2)
leachate flux rate emanating from the waste unit, (3)
leachate concentration of each constituent, and (4) duration
of the constituent release. Information on the on-site waste
unit areas was obtained from responses to the 1992 RCRA
3007 Questionnaire of the Petroleum Refining Industry
(1992 RCRA 3007 Survey Database).  The off-site unit areas
were obtained from the USEPA Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009)
Industrial Subtitle D Waste Management Facility Database
(USEPA, 1996c;USEPA , 1997d).  Leachate flux and contaminant
release rates were determined as a function of the design
and operational characteristics of the different waste
management and wastestream characteristics (waste quantities
and waste constituent concentrations).  The modeling
approach related to the design and operational
characteristics for landfills is as follows:

The modeled landfill is a Subtitle D industrial landfill
which has an earthen cover, but no liner or leachate
collection system.  The leachate flux through the landfill
is the result of infiltration of ambient precipitation
through the landfill cover. Leachate flux rates used in the
analysis were determined using the HELP model (USEPA, 1996c
and 1997d). The net infiltration rate is calculated using a
water balance approach, which considers, among other
factors,  precipitation, evapotranspiration, and surface
run-off. The model was used to calculate landfill
infiltration rates for a representative Subtitle D landfill
with 2-foot earthen cover, using climatic data from 97
climatic stations located throughout the United States
(USEPA, 1996c and 1997d).  For the groundwater pathway
analysis it is assumed that the landfill has a 20 year
operational life, and that there are no losses due to
volatilization while the landfill is in operation or due to
mechanisms other than leaching after the landfill has been
closed. The total amount of constituent in the landfill
which is available for leaching is therefore given by the
product of annual waste quantity disposed and constituent
concentration in the waste, times 20 years.  The 20-year
time frame was chosen as representative of the average
operational life for a typical  landfill where petroleum
wastes are disposed.  This time period is based on a median
active life of petroleum waste landfills (USEPA, 1997b) and
on the average active lifetime of Municipal Subtitle D
landfills (USEPA, 1988).  In the groundwater pathway
analysis it is assumed that each constituent initially
leaches at a concentration given by the TCLP concentration
of that constituent in the wastestream being analyzed, with
a gradual decrease in leaching concentration due to source
depletion (USEPA, 1996d).  Assuming that all of the
constituent mass in the waste may eventually leach out, with
linear equilibrium partitioning of the constituent into the
aqueous phase, the leaching concentration follows an
exponential decrease with time according to (USEPA, 1996d):
where:
     CL(t) = Leachate concentration (mg/L) of a constituent
             at time t
     CL0   = Initial leachate concentration (mg/L) of a      
             constituent
     CW    = Total  concentration (mg/kg) of the constituent 
            in the waste
     I     = Infiltration rate (m/y)
     d     = Depth of waste unit
     FW    = Volume fraction of waste in the waste unit
     PW    = Density of the waste (kg/L)
     t     = Time (y)

The volume fraction of waste in the unit, Fw, is given by 20
times annual waste volume divided by the total volume of the
waste unit.

2.1.3  Processes in Subsurface Fate and Transport Modeling

The primary transport mechanisms in the subsurface are
downward movement along with infiltrating water in the
vadose zone, and movement along with (ambient) groundwater
flow in the saturated zone. The advective movement in the
unsaturated zone is one-dimensional, while the saturated
zone module accounts for three-dimensional flow and
transport. The model also considers mixing due to
hydrodynamic dispersion in both the unsaturated and
saturated zones.  In the vadose zone,  the flow is
gravity-driven and prevails in the vertically downward
direction n the vadose zone.  Therefore, it is reasonable to
model flow in the unsaturated zone as one-dimensional in the
vertical direction.  It is also assumed that transverse
dispersion (both mechanical dispersion and molecular
diffusion) is negligible in the vadose zone.  This
assumption is based on the fact that lateral migration due
to transverse dispersion is negligibly small compared with
the horizontal dimensions of waste management units.  In
addition, this assumption is conservative because it allows
the leading front of chemicals to arrive at the water table
relatively sooner, and in the case of finite source, with
greater peak concentration.

Chemical constituents in the subsurface may be subject to a
variety of bio-chemical transformation processes. EPACMTP
accounts for transformations due to hydrolysis with
first-order reaction kinetics. For those waste constituents
which hydrolyze into toxic daughter products, the model also
accounts for the formation and subsequent fate and transport
of hydrolysis products.  However, none of the organic
constituents considered in the present analysis are subject
to breakdown by hydrolysis (Kollig et al., 1993; Washington,
1995). 

Biodegradation may be a significant removal process for some
of the constituents considered in the analysis.  However,
biodegradation data obtained with methods that are
consistent with the USEPA's protocol for measuring
biodegradation rates exhibit inconsistent behaviors (long
lag time, short lag time, no degradation, etc.). 
Furthermore, the data are presently not adequately available
so as to be implemented in the subsurface fate and transport
model.  A preliminary evaluation of all available data and
the documented anaerobic biodegradation studies of benzene
suggest that in-situ anaerobic biodegradation of benzene
rates may be strongly dependent on site-specific conditions. 
These degradation-controlling site-specific conditions are
currently not exactly known, and  may be time-dependent.  In
addition, several field studies and laboratory studies have
revealed that benzene may be recalcitrant to anaerobic
biodegradation (Krumholz et al., 1996; Salinitro, 1993).  In
the report by Krumholz et al. (1996), 12 of the total of 15
field studies showed no biodegradation.  Salinitro (1993)
reported 2 of the total of 26 field studies without
biodegradation.  The necessary conditions for anaerobic
benzene biodegradation are poorly understood.  The absence
of biodegradation could be due to the presence of competing
substrates, such as, toluene, xylenes, and ethylbenzene, as
well as inadequate geochemical conditions and lack of proper
electron acceptors (nitrate, sulfate, iron, etc.). 
Therefore, because of the lack of information to correlate
site-specific controlling factors to biodegradation, the
limited number of field data, and the field and laboratory
evidence that benzene tends to be recalcitrant to anaerobic
biodegradation, biodegradation of benzene is considered to
be negligible in the current groundwater pathway analysis.

The groundwater pathway analysis accounts for (linear)
equilibrium sorption of waste constituents in the soil and
aquifer. For organic constituents the partition coefficient
(kd) is calculated as the product of the
constituent-specific organic carbon partition coefficient
(koc ), and the fraction organic carbon (foc)in the soil and
aquifer. 

Therefore, for a number of the metals analyzed (e.g., lead,
chromium, mercury) the isotherms are highly nonlinear which
adds considerably to the computational effort of the model
simulations.  For metals, the sorption isotherms used in the
Monte Carlo analyses were determined using MINTEQA2
geochemical speciation model (USEPA, 1996e).  These
isotherms reflect varying subsurface geochemical conditions
defined by four major geochemical parameters (pH, leachate
organic matter, natural organic matter, and iron hydroxide
absorbent).  The values assigned to each of these parameters
are allowed to vary over three ranges (low, medium, and
high) to produce sorption isotherms representative of a wide
range of subsurface geochemical conditions.  In the Monte
Carlo analyses of metals, the sorption isotherms used were
based on the distributions of the four geochemical
parameters.  

For bounding and two high-end parameters analyses, partition
coefficients (Kds) were calculated based on metal-specific
empirical relationships between Kd and pH.  The median pH
values from  the nationwide distribution was used to
calculate Kds for metals.

2.1.4  Regional Site-Based Modeling Approach

The regional, site-based, Monte Carlo approach option
available in EPACMTP was implemented for the Petroleum
Refining groundwater pathway analysis.  In general, the
receptor well concentration downgradient from a waste
management site is a function of the following: source
(e.g., area, volume), design (e.g., cover, liner), climate
(e.g. precipitation, evapotranspiration), hydrogeology
(e.g., depth to groundwater, aquifer thickness, hydraulic
conductivity, hydraulic gradient), receptor location (e.g.
depth and downgradient distance and chemical specific
properties (e.g. retardation and decay). The location, area
and volume of the waste management units used in the
site-based approach represent the unit specific
characteristics of an actual waste management site; the
climatic parameters (e.g. precipitation, evaporation) used
for a given waste management unit are given by the climatic
characteristics in the regional vicinity of the unit.  Using
the source characteristics of actual waste management units
when possible and the hydrogeologic characteristics based on
actual hydrogeologic information insures that the
within-site dependence between source terms and between
hydrogeologic characteristics is maintained.  

The regional site-based approach attempts to approximate the
ideal situation where a probability sample of sites has a
complete description of the characteristics needed to
estimate the site's receptor well concentration.  This
approach  uses the site location to place the site within
one of 13 hydrogeologic regions,  (USEPA, 1996c; USEPA,
1997d) to define the site's hydrogeologic parameters and
within one of 97 climatic regions (USEPA, 1996c; USEPA,
1997d), thereby defining the site's climatic parameters.

The following step-by-step procedure is used to implement
the regional site-based approach in EPACMTP (these steps are
discussed in detail for the present petroleum refining
groundwater pathway analysis in Section 4 of this document):


     Step 1:  Waste Site Selection.  This step involves      
     selecting a site at random, from the list of waste      
     management facilities.  The facility information        
     includes location and for on-site landfills, area and   
     volume. For off-site landfills, an area and volume are
     selected at random from the OSW Industrial Subtitle D   
     Waste Management Facility Database (USEPA, 1996c;       
     USEPA, 1997d).

     Step 2:  Generate Recharge and Infiltration for         
     Selected Waste Site.  Given the waste site's location,
     the climatic region in which the site is located can be
     identified and the corresponding precipitation and
     evaporation values specified.  The precipitation and    
     evaporation values for the site are used together with
     the soil (and cover type for landfills) of the unit to
     derive the recharge and infiltration values of the      
     site.  The soil and landfill cover types used at the    
     sites are generated from a national joint probability   
     distribution described in USEPA 1996c and 1997d.

     Step 3:  Generate Hydrogeologic Variables for Selected
     Site in OSW Industrial Subtitle D Facility Database.    
     Given the facility location, the USGS groundwater       
     resources inventory maps (USGS, 1985) are used to       
     identify the aquifer type for the site.  The            
     hydrogeologic parameters for the site are determined by
     using the characteristics of a groundwater
     investigation site selected at random from the          
     Hydrogeologic Database (API, 1989; Newell et al.,1990)
     for the corresponding hydrogeologic environment.  If    
     the selected groundwater parameter set is missing any   
     values, a joint distribution of the parameters derived
     for each region is used to fill in the missing values.

     Step 4:  Generate Remaining Parameters for Selected     
     Waste Site.  The remaining parameters for the waste     
     site (e.g. x, y, and z coordinates of the receptor      
     well) are generated by using nationwide distributions   
     based on data compiled by the USEPA from across the USA
     (USEPA, 1997d).  The median and high end values of x,   
     y, and z are given in Appendix C with the Two high-end
     parameter sensitivity analysis input parameters.  

     Step 5:  Calculate the Receptor Well Concentration      
     Value for Selected Waste Site.  Given the site data,    
     hydrogeologic characteristics, and receptor well        
     location generated in the previous four steps, and the
     chemical-specific characteristics, the groundwater      
     transport model is used to compute the receptor well    
     concentration value for the site.

     Step 6:  Repeat Steps 1-5 N (10,000) Times (N is a      
     predetermined number of Monte Carlo iterations or       
     parameter realizations based on a convergence test) and
     Estimate the National Distribution of Receptor Well     
     Concentrations.  After Step 5, the receptor well        
     concentration value for a specific realization is       
     obtained.  The process is repeated to yield N (10,000)
     values which represent the nationwide distribution of   
     drinking water exposure concentrations.  Given this     
     distribution and the drinking water standards for the   
     waste, the groundwater exposure risk can be calculated
     for any level of protection (e.g. 85-th percentile,     
     90-th percentile, 95-th percentile).

A flow chart of the above procedure is shown in Figure 2.2.

2.2  MODELING SCENARIOS AND INPUT DATA

All of the modeling scenarios performed for the Petroleum
Refining Groundwater Pathway Analysis, including the 1995
analyses are summarized in Table 2.1.  Complete descriptions
of each current analysis are presented in Sections 3 and 4
and the results are presented in Section 5.  The 1995
Petroleum Refining Groundwater Pathway Analysis Background
Document includes complete descriptions of the previous
analyses conducted in 1995 (USEPA, 1995d).

Additional modeling analysis have been performed to further
enhance the results of the 1995 Petroleum Refining
Groundwater Pathway Analysis.  The analyses included
additional deterministic high-end analyses (Section 3) and
several Monte Carlo scenarios (Section 4).  Prior to
performing the Monte Carlo simulations , waste constituents
were screened based on the 1995 bounding analysis (USEPA,
1995d).  Then, Monte Carlo analyses were performed for waste
constituents not eliminated in the bounding analysis.  Both
the deterministic and Monte Carlo analyses include both 
single-wastestream baseline scenarios and multi-wastestream
codisposal scenarios.  For the multi-wastestream codisposal
scenario, discussed further in Section 2.2.3, total
codisposed 20-year petroleum waste quantities at each
landfill were modeled based on the median active lifetime of
20 years for Subtitle D landfills (USEPA, 1997b).  Volume
weighted waste and leachate concentrations were calculated
for both  benzene and arsenic to estimate a representative
effective leachate rate and an effective waste concentration
for the landfills.  Volume weighting of the input parameters
was determined to be the most representative method for
combining the concentration data. Other additional modeling
considerations included a contingent CSO management scenario
(Section 2.2.4) and TC Rule limitations on waste leachate
concentrations (in all scenarios; Section 2.2.5).  The
source term input parameters (e.g. waste quantity, landfill
size, waste concentration, and leachate concentration) used
for modeling were obtained from responses to the RCRA 3007
Survey (1992 RCRA 3007 Survey Database) and the USEPA
Record Sampling (USEPA, 1995a) and are provided in
Appendices A and B.  The parameters obtained from the OSW
hydrogeological modeling database are listed in the EPACMTP
Background Document (USEPA, 1996c).                         Table 2.1  Petroleum Refining Modeling Analyses Performed in
           1995 and 1996
The  following are described in subsections:

     Data sources;
     Hydroprocessing waste data;
     Multi Waste Stream Codisposal; 
     Contingent CSO sludge management data; and
     Maximum TCLP data for TC-Rule constrained waste         
        management.
     Off-Spec Products and Fines TCLP Data

2.2.1  Data Sources

Model input data pertaining to waste quantities and density,
waste management scenarios and waste unit information (e.g.,
area, volume; Appendix B) based on responses to the RCRA
3007 Survey of the Petroleum Refining Industry (1992 RCRA
3007; USEPA, 1995a).  Waste quantities managed by
petroleum refineries in on- and  off-site landfills in 1992
and on-site waste unit information were provided in 
responses to the survey.  For all wastestreams except
hydrotreating catalysts, hydrorefining catalysts, and
clarified slurry oil sediment (contingent management) the
modeled waste quantities were based on annual 1992 Subtitle
D landfill management (see sections 2.2.2 and 2.2.4 below
for a discussion of the modeled waste quantities for those
waste streams).

For waste managed in off-site landfills the following
information provided in the RCRA 3007 Survey Database was
used for the modeling analysis: the off-site landfill
location, waste quantity managed in the off-site landfill in
1992 and a limited number of waste density values.  For
waste managed in on-site landfills, waste unit area and
waste height were also available in addition to the 1992
waste quantity and density values (1992 RCRA 3007). 
Again, off-site waste unit areas and depths were obtained
from the OSW database of industrial Subtitle D landfills
(USEPA 1996c; USEPA, 1997d).  Waste density values were
available for only a limited number of locations and since
density is relatively non-varying compared to other
parameters in the analysis, average on-site and average
off-site density values were used in the analysis.  On-site
waste unit information, 20-year waste quantity, and average
on-site density values are listed for each modeled
wastestream in Table B.1, Appendix B.  Off-site 20-year
waste volumes and average off-site densities are listed in
Table B.2, Appendix B.

Information on wastestream composition, including waste
concentrations and TCLP leaching concentrations, were
obtained from record sampling data collected by the USEPA
from each of the wastestreams investigated, as part of the
Listing Determination (USEPA, 1995a).  Leachate
concentrations for all constituents in landfills are
presented in Appendix A.

The 1992 RCRA 3007 Survey responses provided limited data
on hydrogeological setting at the waste management unit
locations (1992 RCRA 3007).  Available information for
on-site waste management units was restricted to waste unit
size (area and depth), geographical location of the unit,
depth to groundwater and distance to the nearest drinking
water well.  However, the depth to groundwater and nearest
drinking water well distance data available in the RCRA
3007 Database were not used in this modeling analysis
because the hydrogeological data set was not complete.  Well
distances  were not reported for approximately one-half of
the refinery sites and the well distances that were provided
in the 3007 responses were much larger than the median
value in the OSW  Industrial Subtitle D Waste Management
Facility Database (e.g. an average of 2,600 meters versus a
median value of 430 meters in the OSW database; 1992  RCRA
3007; USEPA, 1997d).  In a number of cases the unsaturated
zone and saturated zone thicknesses were given as a range of
values indicating that the value was estimated rather than
directly measured (e.g. 0 - 40 feet).  For these reasons the
OSW Database (USEPA, 1997d) and API Hydrogeological Database
(API, 1989; Newell et al., 1990; USEPA, 1997d)) were deemed
to be more reliable sources of data for the hydrogeological
parameters.  The hydrogeological parameters from the API
database are not selected from arbitrary nationwide
distributions rather they are site based and correlated to
the aquifer type located at the site.

For off-site units, the survey provided only the waste
management unit location.  The Monte Carlo OSW
hydrogeological modeling database which is incorporated into
EPACMTP (USEPA, 1996c and 1997d) was therefore used to
obtain values for missing parameters.  Data on
constituent-specific koc values, as well as metals sorption
isotherms, were also obtained from the OSW data base (USEPA,
1996c and 1997d).

2.2.2     Hydroprocessing Waste

Information provided by catalyst reclaimers indicates that
because of the depressed metals markets, landfilling of
catalysts is becoming significantly more cost-effective than
reclamation.  EPA predicts that if the risk assessment
modeling shows no basis for listing hydroprocessing
catalysts, landfilling will continue to increase in the
future.  Because of this trend, the Agency made the decision
to also perform the modeling analysis using the entire
distribution of waste volumes reported in 1992 (excluding
volumes reported to go to Subtitle C management) for
hydroprocessing wastes (hydrotreating, hydrorefining, and
SCOT catalysts) rather than just the quantities reported to
be managed in Subtitle D units (USEPA, 1995a).  The total
quantity of hydroprocessing waste generated by  each
refinery was compiled including recycled, reclaimed and
stored waste (Subtitle C landfill waste quantities were
excluded). For the on-site landfill analysis, each
refinery's total hydroprocessing waste was modeled as
disposed in the on-site landfill.  Only refineries reporting
on-site landfills were included in the data set to be
modeled.  In the case of off-site landfills, each refinery's
total hydroprocessing waste was modeled as disposed at
either the nearest off-site landfill or at an off-site
landfill known to have received waste from that refinery
based on other wastestream data.  The 20-year non-Subtitle C
landfill waste quantities modeled for hydroprocessing wastes
are listed in Tables B.1 and B.2, Appendix B (1992 RCRA
3007 Database).

2.2.3     Multi Waste Stream Codisposal

In response to comments, the Agency has evaluated the
potential groundwater exposures due to the impact of
codisposal of multiple wastestreams. For the
multi-wastestream codisposal scenario, total codisposed
20-year petroleum waste quantities at each landfill were
modeled based on the median active lifetime of 20 years for
Subtitle D landfills (USEPA, 1997b).  Volume weighted waste
and leachate concentrations were calculated for both 
benzene and arsenic to estimate a representative effective
leachate rate and an effective waste concentration for the
landfills.  Volume weighting of the concentration data was
determined to be the most representative method for
combining the concentration data of the individual
codisposed wastestreams.  The rationale is presented below.

To determine an effective constituent-specific leaching rate
for codisposed wastestreams, it was conservatively assumed
that the wastestreams codisposed in a landfill are
configured in such a way that each wastestream occupies the
landfill depth which is assumed to be uniform throughout the
landfill, but occupies an area that is determined by the
ratio of wastestream volume to total landfill volume.  This
configuration allows leachates with relatively high leaching
concentrations  to migrate unimpeded chemically by other
wastestreams and/or wastestream solid matrices.  With this
configuration, the relative contribution of a wastestream to
the effective leaching rate of the codisposed wastestreams
at the base of a landfill is therefore dependent on the
relative infiltration flux through the areal portion of the
landfill attributed to that wastestream.  The estimation of
the effective leachate concentration of codisposed
wastestreams  is based on the consideration of conservation
of the total leachate flux, thus:
     Equation 2
     Equation 3
Note that d and Fh are not related to the flux calculation. 
They are added to both sides of  equation (2) to obtain the
expression in equation (3).

Following the above wastestream configuration assumption,
the effective constituent-specific waste concentration may
be determined based on the mass conservation consideration:
     Equation 4

The waste solid density and porosity are not exactly known;
however, the product is very close to the reported wet waste
density which does not vary significantly.  Therefore, it
may be inferred that the density- and porosity-related ratio
is close to unity whereas the volume ratio may vary over a
few orders of magnitude.  Based on this reasoning, the
effective waste concentration may be approximated by:
    Equation 5

Two high-end parameter (Section 3) and Monte Carlo
simulations (Section 4) were performed for the Codisposal
scenario.  The results are presented in Section 5.

2.2.4  Contingent CSO Sludge Management

The disposal of Clarified Slurry Oil (CSO) Sludge was
evaluated assuming two potential disposal scenarios:  (1)
only waste quantities disposed in Subtitle D Landfills in
1992 were modeled and (2) all quantities of CSO Sludge
generated in 1992 were modeled with the exclusion of waste
recycled on-site in refinery process units, waste
transferred for use as fuel, and waste used in an on-site
industrial furnace.    The assumption is that under
contingent management, refinery wastes going to other
non-hazardous disposal scenarios (such as land application
units) would become hazardous and that therefore, facilities
would choose to landfill those wastes.  For the on-site
landfill analysis, each refinery's total CSO sludge waste
was modeled as disposed in the on-site landfill.  Only
refineries reporting on-site landfills were included in the
on-site disposal scenario.  In the case of off-site
landfills, each refinery's total CSO sludge waste was
modeled as disposed at either the nearest off-site landfill
or at an off-site landfill known to have received waste from
that refinery (based on other wastestream data).  In some
cases other wastestream information indicated that offsite
landfills received waste from more than one refinery; and in
some cases, the data indicated that several refineries sent
their waste to more than one off-site landfill.  The
modeling scenario was constructed to reflect these
possibilities; therefore, the Contingent CSO Sludge
quantities presented in Table 2.2a (off-site volumes),
unlike the on-site quantities, do not have a one-to-one
relationship with the original refinery Contingent CSO data. 
Furthermore, refineries included in the on-site landfill
disposal scenario were also included in the off-site
disposal scenario under the assumption that refineries could
elect to either dispose of their waste on-  or off-site. 
The 20-year non-recycled/non-reused site-specific waste
quantities modeled for CSO sludge are listed in Tables 2.2a
and 2.2b for off-site and on-site disposal, respectively. 
Two high-end parameter data for modeling Contingent
Management of CSO Sludge are listed in Appendix C with the
results of the sensitivity analyses for contingent CSO
Sludge. 

Table 2.2(a) Off-Site CSO Contingent Management Data (1992
RCRA 3007 Database)

Table 2.2(b) On-Site CSO Contingent Management Data (1992
RCRA 3007 Database)


Note * The landfill code number is used to uniquely identify
a landfill in the current analysis.  Owing to the CBI
restrictions, details relating these landfills cannot be
disclosed.
2.2.5  TC Rule Constrained Maximum TCLP

The Toxicity Characteristic (TC) Regulation requires  all
waste shown to exceed a specified leaching (TCLP)
concentration to be managed  as hazardous waste.  Therefore,
all petroleum waste shown to exceed the TC concentration
limit will not legally be managed in Subtitle D landfills. 
As a consequence the TC Rule sets an upper limit on the
leaching concentration for waste going to Subtitle D
landfills.   To reflect that maximum leachate concentration
(TCLP) limitation, the baseline and codisposal groundwater
pathway modeling analyses were also conducted with an upper
limit set on the source leaching concentration for benzene
and arsenic.    The upper concentration limits in the
modeling runs were set to the TC Rule value for the
constituent, 0.5 mg/l for benzene and 5.0 mg/l for arsenic.
TCLP data values less than the TC Rule limit were not
changed.  The "TC Rule Constrained" or "Capped" results are
presented in Section 5.

TCLP data for all modeled constituents are listed in Table
A.1, Appendix A.

2.2.6  Off-Spec Products and Fines TCLP Data

The Off-Spec Products and Fines TCLP data for
benz(a)anthracene and benzo(a)pyrene consisted of one "J"
value and five non-detect readings for benz(a)anthracene and
four non-detect readings for benzo(a)pyrene.  A "J" value is
defined as an estimated concentration in the case where mass
spectral data indicate the presence of a compound that meets
the criteria for which the result is less than the
laboratory quantitation limit, but greater than zero (EPA,
1995a).  The estimated "J" values were 0.013 mg/L for
benz(a)anthracene and 0.01 mg/L for benzo(a)pyrene in one
sample and the « quantitation limit reported for the
non-detect samples was .05 mg/L. Because of the large
uncertainty in the TCLP data for these compounds, the
high-end parameter and Monte Carlo analyses were conducted
with the data treated in several different ways.  The Monte
Carlo Analysis results and the previous (1995) Two High-End
Parameter Analyses for  benzo(a)pyrene showed insignificant
risk; therefore, additional Monte Carlo and Two High-End
Parameter analyses were not performed for this chemical. 
Additional analyses were performed for benz(a)anthracene
only.  

Benz(a)anthracene and benzo(a)pyrene have strong tendencies
to readily get sorbed to the soils because they have high
organic carbon partition coefficients (Koc=219,000 ml/g and
Koc= 631,000 ml/g, respectively).  These high Koc values
tend to retard their movement in soils significantly. 
Therefore, they tend not  to show up in ground water or have
very low concentrations in groundwater. 

2.3.5.1   Monte Carlo Distributed Data

For the Monte Carlo Analysis, distributions consisting of 
individual data values are typically used.  However, the
TCLP data were used in two different ways for these wastes. 

1.  The "J" value was assumed to be the only reliable data
point and it was used to represent a fixed value for the
TCLP concentration of benz(a)anthracene.  Other parameters
were varied as before in the Monte Carlo Analyses.  

2.  The "J" value of 13 ppb was used with the five «
quantitation limit values of 50 ppb to construct a leachate
distribution.

2.3.5.2  High-End Parameter Data

The TCLP data were handled in two different ways for
conducting the high-end parameter sensitivity analysis:

1.  In the first case the "J" value was assumed to be the
only reliable data point and it was assumed to represent the
mean or expected value of the TCLP for benz(a)anthracene in
off-spec products and fines.  The sensitivity analysis was
therefore conducted to determine the two high-end parameters
for off-spec products and fines.

2.  In the second case the "J" value was assumed to be  the
maximum or high-end TCLP value and the sensitivity analysis
was therefore conducted to determine one other high-end
parameter for the two high-end parameter analyses.


The results for the Off-Spec Products and Fines analysis are
presented in Table 5.8.3.0  DETERMINISTIC GROUNDWATER PATHWAY ANALYSES


3.1  BOUNDING ANALYSES

A bounding analysis was performed in the 1995 groundwater
pathway modeling analysis to eliminate from further
consideration those wastestreams and/or waste constituents
which even under worst-case conditions do not show risk or
HBN exceedance in groundwater (USEPA, 1995d).  According to
EPA's Guidance on Risk Characterization (USEPA, 1992),
bounding estimates are intended to purposely overestimate
the exposure or dose in an actual population for the purpose
of developing a statement that, the risk is "not greater
than the results of the analysis".  Therefore those
constituents not posing risk in the bounding analysis can be
eliminated from further consideration.

The bounding analysis consisted of deterministic EPACMTP
groundwater transport runs with  sensitive modeling
parameters set to high end values and the remaining
parameters set to median values.  Assignment of parameters
in the bounding analysis was as follows: the leachate
concentration of each waste constituent was set to its
maximum TCLP value; the waste unit area, infiltration rate,
waste quantity, landfill waste fraction and waste
concentration were set to their 90-th percentile values, and
the receptor well was placed on the plume centerline, at the
10-th percentile of the down gradient distance and depth. 
Other parameters were set to their median values.  In
addition, the source depletion option was not used in the
bounding analysis.  Instead, the release of the waste
constituents were modeled as a pulse with constant
concentration equal to the maximum measured TCLP for the
waste constituent.  The peak receptor well concentration was
used for both carcinogenic and non-carcinogenic
constituents.

Single-wastestream baseline analyses were then performed in
the 1996 Monte Carlo Groundwater Pathway Analyses for all
wastestreams and waste constituents that showed risk in the
1995 bounding analysis, i.e., if the HBN of a waste
constituent was exceeded in the bounding analysis, the
single-wastestream baseline analysis was performed for that
waste constituent (see Section 5.1).

Qualitative results of the bounding analysis are shown in
Table 3.1.  The quantitative results are given in detail in
the 1995 Groundwater Pathway Background Document, Appendix C
(USEPA, 1995d). 

3.2  TWO HIGH-END PARAMETERS BASELINE ANALYSES

According to EPA's Guidance on Risk Characterization, "The
high end risk descriptor is a plausible estimate of the
individual risk for those persons at the upper end of the
risk distribution.  The intent of this descriptor is to
convey an estimate of risk in the upper range of the
distribution, but to avoid estimates which are beyond the
true distribution.  Conceptually, high end risk means risks
above the 90-th percentile of the population distribution,
but not higher than the individual in the population who has
the highest risk" (USEPA, 1992).  If limited information on
the distribution of exposure or dose factors is available,
the high-end can be estimated by identifying the most
sensitive parameters and using maximum or near-maximum
values for one or two of these variables, leaving other at
their mean values. (USEPA, 1992).  In order to identify the
most sensitive parameters for performing a high-end analysis
in the 1995 groundwater pathway analysis, a sensitivity
analysis was conducted on crude oil tank sediment (USEPA,
1995d) to determine the most sensitive waste source and
receptor well parameters. This analysis was conducted by
individually varying the source and well location parameters
from 50-th to 90-th percentile values, and ranking them in
terms of the corresponding change in predicted receptor well
concentration.  For the subsequent 1995 two high-end
parameter groundwater impact analysis, the two most
sensitive parameters were set to their 90-th percentile
values, while the remaining model parameters were all kept
at median values. Table 3.1  Summary of Previous Groundwater Pathway Analysis
           Results (USEPA, 1995d)

3.2.1  1995 Proposal Results

The modeling results presented in the 1995 petroleum
refining waste listing determination rule proposal were the
results of the 1995 two high-end parameter modeling
analysis.  For the two high-end parameter analysis, the two
most sensitive parameters were set to their respective 90-th
percentile values.  The selection of the two high-end
parameters was as follows for the landfill and land
treatment scenarios:

     Landfill:  Waste Unit Area 
                Distance to Well

     Land treatment unit:  Waste Unit Area 
                           Waste Quantity 

It should be noted that in the case of the distance-to-well,
the high-end value corresponds to the 10-th, rather than the
90-th percentile of the distribution, i.e., a closer well
will result in a more conservative assessment.  The waste
fraction that was used in the modeling analysis was
calculated as the ratio of the median 20-year wastestream
volume and the mean landfill volume.  The mean landfill
volume was chosen because its value was generally
intermediate between the 50-th and 90-th percentile values,
yielding a corresponding intermediate value for the waste
fraction which was kept constant.  The consequence of the
parameter values used is that the modeled waste volume was
derived from the pre-determined waste fraction. For most
wastestreams, the modeled waste quantity used in the
high-end analysis exceeded the actual median value of the
reported wastestream volume, but under-estimated the median
wastestream volume in the central tendency case.

In the single parameter high-end case, the single most
sensitive variable was assigned to its 90-th percentile
value, while all other model parameters were set to median
values.  This variable was the waste unit area for both
landfills and land treatment units.  In the high-end
analysis the source depletion option was used for the
landfill scenario.  The receptor well was placed at half-way
between the plume centerline and the lateral extent of the
contaminant plume.  In the evaluation of groundwater
exposure concentrations, the 9-year maximum average
concentration was used for carcinogens, and the peak
concentration for non-carcinogens.  Cancer risks and
health-based number (HBN) exceedances for the high-end
analysis are shown in Section 5 (Table 5.1 Summary of
Groundwater Pathway Analysis Results) along with the revised
two high-end parameter analysis results (see Section 3.2.2). 
 

3.2.2  Revised Two High-End Parameters Analyses

In the 1995 Two High-End Parameters Analysis, the waste
fraction was defined as the ratio between the median 20-year
wastestream volume and the mean landfill volume, while  the
20-year waste volume was treated as a derived parameter in
the modeling (USEPA, 1995d).  Consequently, as mentioned
above, for most wastestreams the modeled waste volume was
greater than actual 50th percentile waste volume.  In
response to comments on the 1995 Proposal most of the
wastestreams were reanalyzed using the reported 1992  waste
volumes in detailed sensitivity analyses (Section 3.2.3). 
In addition, the 1995 Two-High End Parameter Scenario was
reanalyzed for two listing wastestreams, for which a
detailed Two Parameter sensitivity was not performed (Sulfur
Complex Sludge and Unleaded Gasoline Tank Sludge) using the
reported waste volumes from the 1992 RCRA 3007 Survey
(USEPA, 1995a). Table 3.2 shows a comparison between the
waste volumes derived in the 1995 Proposal Groundwater
Analysis and the reported waste volumes used in the revised
Two High-End Parameters analysis. In the case of On-Site
Unleaded Gasoline Tank Sludge the modeled waste quantity was
less than the 1992 RCRA 3007 reported waste quantity
(USEPA, 1995a).  For the Off-Site Unleaded Gasoline Tank
Sludge and Off-Site Sulfur Complex Sludge the modeled waste
quantities were greater than the reported values (USEPA,
1995a: Table 3.2).  A comparison between the 1995
groundwater risk results and the revised results are shown
in Section 5 (Tables 5.1, 5.2, and 5.3).

3.2.3  Current Two Parameter Sensitivity Analysis

A sensitivity analysis was conducted in the current analysis
to determine the two model input parameters which have the
greatest impact on  the receptor well concentration.  These
two parameters were set to their 90-th percentile when
performing the two high-end parameters analysis. 

The analysis focused on source-related parameters (area,
waste and leachate concentration, infiltration, and waste
volume) and the location of the receptor well (x-, y-, and
z- coordinates).  All other variables were assigned median
values.  Sensitivity analysis was performed on all the
wastestreams that showed a significant risk  in the 1995
analysis (USEPA, 1995d) and the 1996 Monte Carlo analysis. 
Table 3.3  shows a list of wastestreams and constituents for
which sensitivity analysis was performed.  Tables in
Appendix C show the input parameter distributions used in
the modeling analysis for each wastestream.  In the case of
constituent waste and leachate concentrations, instead of a
statistical distribution, the average concentration measured
across the samples was used for the 50th percentile and the
maximum concentration was used for the high end. The
relatively few waste and leachate samples available
warranted this approach.  The distributions of the source
input parameters vary from wastestream to wastestream and
constituent to constituent.   Therefore, the combination of
the two most sensitive parameters also varies and is
dependent on  wastestream and  constituents.
Table 3.2  Comparison of 1995 and 1996 Modeled Waste Volumes
3.2.3.1  Approach

The analysis was carried out in two steps.  First, the model
was run for a base case scenario.  A base case scenario
represents a simulation in which all the input parameters
are set to their respective median values.  Then, a series
of simulations were performed by setting a combinations of
two parameters, at a time, at their 90-th percentile values. 
In total, 28 simulations  were performed for each
constituent in a wastestream.  In the case of TC rule capped
scenarios the leachate concentration distribution above the
TC rule regulatory values was truncated.  The sensitivity of
each combination of  two parameters, set at their 90-th
percentile values, was determined based on the relative
change in receptor well concentration from its base case. 
In the case of carcinogens, 9 year average receptor well
concentrations were used to estimate the relative
concentrations. 

3.2.3.2  Sensitivity of Parameters

The relative sensitivities of the combination of two
parameters set to their high-end values, for each
constituent and wastestream are presented in Appendix C.  A
summary of the two most sensitive parameters for each
constituent and wastestream is presented in Table 3.3.  As
discussed in the previous sections, it can be seen that the
most sensitive parameters vary from wastestream to
wastestream and also from constituent to constituent.  Since
the sensitivity analysis was performed by varying a
combination of two parameters at a time, the synergism
between the two parameters and their impacts on the receptor
well concentration are clearly demonstrated (see Appendix
C).  All data values listed in Appendix C are from the 1995
Listing Background Document ( USEPA, 1995a) with the
exception of the offsite landfill dimensions.  The offsite
landfill areas and depths were obtained from USEPA RCRA
Confidential Business Information Documents (RCRA CBI,
1995).
     
3.3  CENTRAL TENDENCY ANALYSIS

Central tendency analyses were performed in both the
previous (1995) and current (1996) Groundwater Pathway
Analyses to reflect the central estimate of exposure or
dose.  In these analyses, all parameters were set to their
median (50th percentile) value except TCLP and waste
concentration which were set to their mean values due to the
limited number of data points.

3.3.1  1995 Analysis

Details and results of the previous (1995) central tendency
analysis are presented in the 1995 Groundwater Pathway
Analysis Background Document, Section 4 (USEPA, 1995d). 

3.3.2  Revised Central Tendency Analysis

The central tendency analysis performed in 1996 (current
analysis) was similar to the previous (1995) analysis.  The
one difference was that in the current analysis the waste
fraction was treated as a derived parameter and the reported
waste volume (1992 RCRA 3007 Survey; USEPA, 1995a) was
used in the modeling. The results of the current central
tendency analysis are presented in Appendix E.  The median
value parameters used in the current central tendency
analysis are presented in Appendix C.  The central tendency
parameters are listed in the top row of each sensitivity
table (Base Case).

3.4  TWO HIGH-END PARAMETERS ANALYSIS FOR CODISPOSAL
 
A two high-end parameters analysis was performed to
investigate the impacts of waste codisposal on risks
associated with benzene and arsenic in 20 wastestreams
managed in on-site and off-site Subtitle D landfills. 
Tables 1.1 and 1.2 list  wastestreams included in the
codisposal analyses.  As indicated by Table 1.1,
Hydrotreating and Hydrorefining Catalyst were not included
in the codisposal analysis because these wastestreams are
recommended to be listed, and would, therefore, not be
codisposed with nonlisted wastestreams in Subtitle D
landfills.  Risks were also evaluated with and without
Hydrocracking Catalyst to evaluate the relative contribution
of Hydrocracking Catalyst to benzene and arsenic risk in the
codisposal scenario. 

All model runs were conducted in deterministic mode by
setting two  parameters ( waste unit area and x-well
distance) to their high-end values i.e. 90-th percentile for
landfill area and 10-th percentile for the well distance. 
The two high-end parameters were selected based on the
sensitivity analysis conducted in 1995 (USEPA, 1995d). 
Although a  sensitivity analysis was not performed to
determine the two most sensitive parameters for the
codisposal scenario, the analysis results (Table 5.7)
indicate that the selection of Area and X-Well as the two
high-end parameters produced receptor well concentrations
between the 95th and 99th percentile on the Monte Carlo
distribution.  The remaining input parameters of the model
were assigned to their median values obtained from parameter
specific distributions (USEPA, 1996c and 1997d).  Codisposed
values for each source parameter were calculated from the
data values from the individual wastestreams as described
below.  

The on-site and off-site codisposed waste volumes were
determined by adding up the 50-th percentile of the 20- year
waste volumes for individual wastestreams.  For other waste
characteristics such as source leachate concentration, waste
concentration, and density, the volume-weighted average
values were used in the analyses for both  on-site and
off-site modeling scenarios.  The waste management unit area
used was the 90-th percentile of the 90-th percentile
landfill areas The waste unit depth was obtained similarly
using the 50-th percentile of the 50-th percentile depth
values.  A summary of model input parameters used in the
analysis are in Table 3.4.  Tables showing the individual
wastestream data that went into the calculation of the
codisposal input parameters are shown in Appendix D.

As with the single wastestream scenarios, the codisposal
scenario was evaluated both with and without a TC limitation
(TC Cap) on the source leaching rate.  The TC Cap scenario
was evaluated by restricting the source leaching rate to be
less than or equal to the toxicity characteristic (TC) value
, 0.5 mg/L  and 5.0 mg/L for benzene and arsenic,
respectively.  Consequently, whenever a constituent TCLP
exceeded the TC Rule limit of a constituent, it was capped
at the TC Rule limit.  For the case of no TC Rule limit,
risks were evaluated using source leachate concentrations
(TCLP) collected from site record sampling (USEPA, 1995a). 
The two high-end parameters codisposal risks for benzene and
arsenic are presented with the Monte Carlo risk results in
Section 5 (Table 5.7).
Table 3.4  Two-Parameter High-End Input Data for Codisposal  
           Scenario

4.0  MONTE CARLO ANALYSES PROCEDURE

Monte Carlo simulations were performed to generate
probability distributions of downgradient receptor well
concentrations for each waste constituent of concern, e.g.
benzene for all wastestreams which were not eliminated in
the bounding analysis.  The Monte Carlo simulations
consisted of 10,000 groundwater chemical transport
simulations, each with a different set of input parameters. 
A particular combination of parameters is  referred to as a
realization.  In this analysis, for each Monte Carlo
realization, a receptor well concentration was calculated by
randomly selecting one landfill along with its correlated,
site specific parameters and site-specific parameter
databases.  Each parameter with a statistical distribution
was then selected from its distribution database.  These
distributed parameters were either correlated to the site,
e.g. hydraulic conductivity or hydraulic gradient, or they
were independent of the site, e.g. leachate concentration. 
Generic and site-specific data are further described below. 
Simulations were performed for each realization using the
selected parameters.  Receptor well concentrations
calculated for all of the realizations were sorted to
determine the concentration probability distribution.

Site-specific parameters are those modeling parameters which
were correlated to the waste unit site location. 
Site-specific parameters were available either from
responses to the RCRA 3007 Survey (1992 RCRA 3007
Database)  or from a reference source such as the API
hydrogeologic parameter data base (API, 1989; Newell et al.,
1990: USEPA 1997d).  Site-specific parameters were
correlated to the refinery location.  Site-specific
parameters selected from a source other than the RCRA 3007
survey were correlated to the site based on the geographic
location of the site through the assignment of a groundwater
region code or a climatic region code.  Generic parameters
are those modeling parameters for which limited data or no
reliable data exist for each landfill location. 
Consequently, generic parameters were randomly selected from
distribution databases independently of the site selection. 
For example, a limited number of waste concentration and
TCLP concentration data points were available for each
wastestream.  These data are used as source concentration
input values in the groundwater model.  Therefore, a generic
source concentration distribution consisting of two to six
paired concentration data values per wastestream was used
for all sites.  For each Monte Carlo realization, paired
source concentration values were randomly selected from the
concentration distribution.  Site-specific and generic
modeling parameters and their sources are summarized in
Table 4.1. 

4.1  SINGLE-WASTESTREAM BASELINE

The petroleum refining waste data were categorized into
waste disposed at on-site refinery landfills and waste
disposed at off-site industrial Subtitle D landfills.  More
detailed waste unit information was available for on-site
landfills than for off-site landfills; therefore, different
Monte Carlo procedures have been developed for on- and
off-site landfills.  In addition, separate Monte Carlo
procedures have been developed to simulate multi-wastestream
codisposal at on- and off-site landfills.  The Monte Carlo
procedures for each of the four scenarios are described in
the following sections.

Single-wastestream baseline Monte Carlo analyses were
performed for all wastestreams and waste constituents that
showed risk in the bounding analysis, i.e. if the HBN of a
waste constituent was exceeded in the bounding analysis, the
single-wastestream baseline analysis was performed for that
waste constituent.

4.1.1  On-Site Baseline Monte Carlo Analysis

Waste unit area and depth values and wastestream volume data
are available for each on-site landfill.  Therefore,
single-valued site-specific parameters for on-site landfills
include:  waste unit area and depth, wastestream volumes,
and groundwater temperature.  Site-correlated distribution
information for each simulation includes infiltration rates,
groundwater recharge rates, and aquifer flow and transport
properties.  Generic parameters are not correlated to site
location and include wastestream and leachate concentrations
(see explanation above), waste densities, and receptor well
location.Table 4.1  Description of Monte Carlo Input Data and Data    
           Sources4.1.1.1  Procedure

The step-by-step Monte Carlo procedure for the baseline
on-site landfill scenario is outlined below: 

     STEP 1.  A Wastestream is selected,

     STEP 2.  A Constituent is selected then, 

     For the I-th Monte Carlo realization:

     STEP 3.  An on-site landfill is selected which includes
     the following information:

          - landfill area, waste unit depth and volume       
            (information reported in responses to the RCRA   
            Section 3007 Survey; 1992 RCRA Section 3007      
            Survey Database),

          - wastestream volume (information reported in      
            responses to the RCRA Section 3007 Survey *20    
            year;1992 RCRA Section 3007 Survey Database),
            (In a limited number of cases the landfill       
            volumes reported in the RCRA Section 3007 Survey
            were less than the reported 20 year waste        
            volumes; in those cases the landfill waste       
            fraction will be fixed at 1.00, thereby          
            implicitly increasing the depth of the landfill;
            the calculated volume of a landfill (= total     
            thickness (height +depth) x area) may be         
            different from the reported volume due to the    
            slopes of the sides of the landfill (assuming no
            other sources of errors)

          - hydrogeologic region (and hydrogeologic          
            parameters; API, 1989 and USEPA, 1997d),

          - climatic region (and climatic-related            
            parameters; USEPA, 1997d and

          - groundwater temperature.         

     STEP 4.  Remaining parameters are selected such as well
     location and dispersivity, from nationwide              
     distributions (USEPA, 1997d).

     STEP 5.  The receptor well concentration for the above
     selected parameters is calculated.

     For each wastestream, and a constituent of that         
     wastestream, the process (Steps 3 to 5) is repeated     
     10,000 times. The receptor well concentrations are then
     sorted and reported in terms of the 85-th, 90-th, 95-th
     and 99-th percentile exceedances (and/or Cancer Risks   
     or Health Quotients).

     Steps 2 to 5 are repeated for all constituents of the   
     selected wastestream.  Steps  1 to 5 are repeated for   
     all wastestreams of interest.
     
4.1.1.2  Modeled Wastestreams and Constituents

The wastestreams and waste constituents shown in Table 4.2
were included in the single-wastestream baseline modeling
analysis.  In the previous bounding analyses (USEPA, 1995d),
these wastestream constituents had been found to exceed
their respective HBN values.


Table 4.2  Modeled Wastestreams and Constituents for On-Site
           Scenarios

4.1.2  Off-Site Baseline Monte Carlo Analysis

Monte Carlo simulations of the off-site landfill scenario
similar to the on-site scenario were conducted.  Locations
were available for each off-site landfill.  Therefore,
hydrogeologic information were selected from an API database
(API, 1989; Newell et al., 1990; USEPA 1997d) based on its
geographic location.  Those parameters were correlated with
the landfill site along with site-specific parameters such
as waste volume, and groundwater temperature.  As with the
on-site scenario, the wastestream and leachate
concentrations as well as receptor well locations, are
generic and were randomly selected from distributions
consisting of industry-wide site sampling data.  Receptor
well locations were selected from a uniform distribution of
x and y well coordinates.  Climate information (i.e.,
infiltration and recharge) was selected from a nationwide
database for each landfill site based on its location. 
Waste unit depths and areas were not available for off-site
landfills; therefore, paired values were randomly selected
for each realization from the HWIR Subtitle D landfill
database (USEPA, 1996c and USEPA, 1997d).  A check was
implemented to ensure that the selected landfill volume was
greater than the wastestream volume; otherwise, the landfill
was rejected and another area-depth pair was selected from
the HWIR distribution.

4.1.2.1  Procedure
     
The Monte Carlo procedure for the single-wastestream
baseline off-site landfill scenario is similar to the
on-site baseline procedure except that after an off-site
landfill and its hydrogeologic parameters are selected.  The
following steps are taken : 

     STEP 1.  A Wastestream is selected,

     STEP 2.  A  Constituent is selected then, 

     For the I-th Monte Carlo realization:

     STEP 3.  An off-site landfill is selected which         
     includes the following information:

          - wastestream volume (information reported in      
            responses to the RCRA Section 3007 Survey *20    
            years; 1992 RCRA Section 3007 Survey Database).

          - hydrogeologic region (and hydrogeologic          
            parameters; API, 1989; USEPA, 1997d),

          - climatic region (and climatic-related            
            parameters; USEPA, 1997d), and

          - groundwater temperature.         

     STEP 4.  Paired landfill area and depth values are      
     selected from the HWIR distribution of Subtitle D       
     industrial landfills.   

     STEP 5.  The volume of the selected landfill is         
     computed and compared with the codisposed waste volume
     (see paragraph below).  If the codisposed waste volume
     is greater than or equal to the landfill volume, the    
     landfill area and depth pair are rejected and a new     
     area-depth pair is selected until a pair is selected    
     that meets the criteria of waste unit volume being      
     greater than waste volume. 
     
     Waste management facility volumes are compared against
     codisposed wastestream volumes in both the              
     single-wastestream (baseline) and codisposal scenario   
     to ensure consistency between the scenarios and to      
     ensure that unrealistically small landfill areas are    
     not selected for the baseline analysis.

     STEP 6.  Remaining parameters are selected such as well
     location and dispersivity, from nationwide              
     distributions (USEPA, 1997d).

     STEP 7.  The receptor well concentration for the above
     selected parameters is calculated.

     For each wastestream, and a constituent of that         
     wastestream, the process (Steps 3 to 7) is repeated     
     10,000 times. The receptor well concentrations are then
     sorted and reported in terms of the 85-th, 90-th, 95-th
     and 99-th percentile exceedances (and/or Cancer Risks   
     or Health Quotients).

     Steps 2 to 5 are repeated for all constituents of the   
     selected wastestream.  Steps  1 to 5 are repeated for   
     all wastestreams of interest.

4.1.2.2  Modeled Wastestreams and Waste Constituents

The wastestreams and constituents shown in Table 4.3 were
included in the baseline modeling analysis.

Table 4.3  Modeled Wastestreams and Waste Constituents for   
           Off-Site Scenarios


4.2  CODISPOSAL MONTE CARLO ANALYSIS

Concerns raised in the public comments concerning the impact
of codisposal of wastestreams in landfills were addressed by
performing multi-wastestream codisposal modeling analyses
similar to the single-wastestream baseline scenarios
described in Section 3.  However, instead of considering the
wastestreams individually, information on codisposal at each
landfill was used to model total petroleum waste volumes and
volume averaged waste and leachate concentrations.  The  on-
and off-site codisposal scenarios were simulated for benzene
and arsenic transport using the entire set of either on-site
or off-site landfills.  Since hydrorefining and
hydrotreating wastes are recommended to be listed and will
not be codisposed with non-listed wastestreams in Subtitle D
landfills, those wastes were excluded from the codisposal
scenario.  In the case of off-site landfills, analyses were
performed both with and without hydrocracking waste to
evaluate the impact of listing that wastestream.
 
4.2.1  On-Site Codisposal Monte Carlo Analysis Procedure

Monte Carlo simulations were performed to determine a
probability distribution of receptor well concentrations for
each constituent of concern.  Again site-specific parameters
were correlated to each refinery and generic parameters were
randomly selected from distribution databases.  The general
Monte Carlo procedure is similar to that for on-site
baseline analyses (Section 4.1.1).  In this procedure,
however, wastestream and leachate concentrations were
volume-weighted averages of all wastestreams disposed of  at
a landfill.  Volume-averaged TCLP and waste concentration
values for each constituent of concern were calculated for
the landfills as follows:
          
Concentration values are selected at random from the TCLP,
(Cij), and waste concentration, (Cwij), data sets for each
wastestream(I) disposed at an on-site landfill.
     
The volume averaged TCLP (leachate) concentration and waste
concentration values are calculated using the formula below,




Where the subscript, I, refers to the wastestream, j refers
to the value selected at random from the concentration data
set for wastestream I with m values, and Vi is the volume of
wastestream I.

If the TCLP data for a wastestream showed no indication of
the presence of a particular constituent, then the
wastestream is included in the codisposal with the TCLP of
that constituent set to 0.0.

Site-specific parameters for codisposal at on-site landfills
include waste unit information, total wastestream volume,
and some aquifer parameters.  Generic information for each
simulation includes volume-weighted average densities, well
location and some aquifer parameters.

4.2.1.1  Procedure

The Monte Carlo procedure for the on-site landfill
codisposal scenario is outlined below: 
     
     STEP 1.  A Wastestream is selected

     STEP 2.  A Waste Constituent is selected, then

     For the I-th Monte Carlo realization:

     STEP 3.  An on-site landfill is selected which includes
     the following information:

          - landfill area, waste unit depth and volume       
            (information reported in responses to the RCRA   
            Section 3007 Survey; 1992 RCRA Section 3007      
            Survey Database).

          - wastestream volume for each wastestream          
            codisposed in the landfill (information reported
            in RCRA Section 3007 Survey *20 years) (In a     
            limited number of cases the landfill volumes     
            reported in the RCRA 3007 survey are less than   
            the reported 20 year waste volumes; in those     
            cases the landfill waste fraction will be fixed
            at 1.00; see on-site baseline for additional     
            explanation, Section 4.1.1)

          - hydrogeologic region (and hydrogeologic          
            parameters; API, 1989; Newell et al., 1990;      
            USEPA, 1997d),

          - climatic region (and climatic-related            
            parameters; USEPA, 1997d), and

          - groundwater temperature.

     STEP 4.  A 20-year codisposed waste volume for each     
     landfill is calculated by summing the reported 1992     
     annual volumes of all wastestreams (including study     
     wastestreams) disposed in that landfill and multiplying
     the volume by 20 years.

     STEP 5.  Volume averaged TCLP and waste concentration   
     values for each constituent of concern are calculated   
     for the landfills as follows:

        5.1  Concentration values are selected at random     
        from the TCLP and waste concentration data sets for
        each wastestream(I) disposed at the selected         
        landfill.

        5.2  Volume-averaged TCLP (leachate) concentration   
        and waste concentration  values are calculated using
        Equations 1 and 2. 

        5.3  If the TCLP data for a wastestream showed no    
        indication of the presence of a particular           
        constituent, then the wastestream is included in the
        codisposal  with the TCLP of that constituent set to
        0.0.

     STEP 6.  Other parameters such as well location and     
     dispersivity from nationwide distributions are          
     selected.

     STEP 7.  Receptor well concentration for the above      
     selected parameters is calculated.

     The process (Steps 3 to 7) is repeated 10,000 times.    
     The receptor well concentrations are reported as the    
     90-th, 95-th and 99-th percentile exceedances (and/or   
     Cancer Risks or HQs). 

     Steps 2 to 7 are repeated for all constituents of the   
     selected wastestream.  Steps 1 to 7 are repeated for    
     all wastestreams of interest.

4.2.2  Off-site Codisposal Monte Carlo Analysis Procedure

Off-site codisposal is simulated much the same as the
off-site single wastestream disposal except that
volume-weighted average concentrations and densities along
with total wastestream volume were correlated with the
landfill location. 
     
4.2.2.1  Procedure

The Monte Carlo procedure for the on-site landfill
codisposal scenario is outlined below: 
     
     STEP 1.  Select a wastestream

     STEP 2.  Select a Waste Constituent, then

     For the I-th Monte Carlo realization:

     STEP 3.  Select an off-site landfill which includes the
     following information

          - wastestream volumes for each wastestream         
            codisposed in the landfill (information reported
            in responses to the RCRA  3007 Survey * 20     
            years; 1992 RCRA Section 3007 Survey Database)

          - hydrogeologic region (and hydrogeologic          
            parameters; API, 1989; Newell et al., 1990;
            USEPA 1997d),

          - climatic region (and climatic-related            
            parameters; USEPA, 1997d), and

          - groundwater temperature          

     STEP 4.  A 20 year codisposed waste volume for each     
     landfill is calculated by summing the reported 1992     
     annual volumes of all wastestreams disposed in that     
     landfill and multiplying the volume by 20 years.

     STEP 5.  Volume averaged TCLP and waste concentration   
     values for each constituent of concern is calculated    
     for the landfills as follows:

        5.1  Concentration values are selected at random     
        from the TCLP and waste concentration data sets for
        each wastestream(I) disposed at the selected         
        landfill.

        5.2  Volume averaged TCLP (leachate) concentration   
        and waste concentration values are calculated using
        Equations 1 and 2. 

        5.3  If the TCLP data for a wastestream showed no    
        indication of the presence of a particular           
        constituent, then the wastestream is included in the
        codisposal with the TCLP of that constituent set to  
        0.0.

     STEP 6.  Select paired landfill area and depth values   
     from the HWIR distribution of Subtitle D industrial
     landfills. The volume of the selected landfill is       
     compared against the codisposed waste volume of the     
     landfill selected in 3 above. If the codisposed waste   
     volume is greater than or equal to the landfill volume,
     the landfill area and depth pair is rejected and a new
     area-depth pair is selected until a pair is selected    
     that meets the criteria of waste unit volume being      
     greater than waste volume. 

     Waste management facility volumes are  compared against 
     codisposed wastestream volumes in both the              
     single-wastestream (baseline) and codisposal scenarios
     to ensure consistency between the scenarios and to      
     ensure that unrealistically small landfill areas are    
     not selected for the baseline analysis (if the landfill
     volume is too small, that value of volume is rejected   
     and a new volume is selected).

     STEP 7.  Select other parameters such as well location
     and dispersivity from nationwide distributions.

     STEP 8.  Calculate a receptor well concentration for    
     the above selected parameters.

     Repeat the process (Steps 3 to 8) 10,000 times.  The    
     receptor well concentrations are reported as the 90-th,
     95-th and 99-th percentile exceedances (and/or Cancer   
     Risks or HQs).

     Steps 2 to 7 are repeated for all constituents of the   
     selected wastestream.  Steps 1 to 7 are repeated for    
     all wastestreams of interest.
5.0  MODELING RESULTS AND GROUNDWATER RISK

After the modeling results were processed, receptor well
concentrations were used to assess the potential for adverse
health effects from both direct and indirect exposures to
groundwater.  Calculated exposure levels were compared to
EPA's health-based numbers (HBN's).  For carcinogens, where
risk is expressed as a probability of getting cancer the
HBN's were set at the exposure level where the risk is   1 x
10-6.  For non-carcinogens, HBN's were EPA's reference dose
(RFD's) levels.  RFD's are values at or below which adverse
health effects are not expected to occur.

5.1  RISK CALCULATION

5.1.1  Direct Groundwater Risk

The groundwater pathway modeling analysis results consist of
modeled receptor well concentrations.  In the case of the
two high-end parameter deterministic analyses, a single well
concentration was calculated for each waste constituent. 
For the Monte Carlo analysis, distributions of receptor well
concentrations were generated for each waste constituent. 
To complete the risk analysis, well concentrations were
compared with the health-based numbers (HBN's). The HBN
values were calculated  from standard equations for drinking
water ingestion from the USEPA Risk Assessment Guidance for
Superfund Volume I, Human Health Evaluation Manual (USEPA,
1989), using an exposure duration of 9 years, ingestion rate
of 1.4 L/day and adult body weight of 70 kg.  Ratios of the
well concentration to the HBN's for the various model
scenarios are summarized in Tables 5.1 -  5.5.  The 85-th,
90-th, 95-th and 99-th percentile model results are
presented for the Monte Carlo Analysis.

5.1.2  Indirect Groundwater Risk

An additional indirect risk due to inhalation during
showering was added to the Benzene direct risk estimates to
calculate a total estimated groundwater risk.  The showering
dose-response for Benzene was calculated as 6.06 x 10-5 risk
per 1.0 mg/l of Benzene in the showering water (USEPA,
1996b).  Showering risks for other less volatile modeled
constituents were determined to be negligible.

5.1.3  Total Risk

The indirect inhalation risk from showering combined with
the direct ingestion risk produces a total dose-response of
1.606   10-4 risk per 1.0 mg/l of benzene.  The estimated
total groundwater risk for benzene was calculated from the
product of the EPACMTP calculated well concentration
(high-end parameter, 90-th, 95-th, or 99-th percentile) and
the total dose response for benzene.  The modeling results
(risk estimates) listed in Tables 5.1 - 5.5 are presented as
multiples of 1.0   10-6 risk or risk at 1.0   10-6.  The
total risk for the other less volatile modeled constituents
is equal to the direct risk; showering risks were not added.

5.2  MODELING RESULTS

Results for the two high-end parameter and Monte Carlo
analyses are shown in Tables 5.1 through 5.7.  The Monte
Carlo results are presented as calculated 85-th, 90-th,
95-th, and 99-th percentile risk estimates.  Both direct
(ingestion) and total (ingestion plus showering inhalation)
risks are presented for benzene and direct risk is presented
for all other waste constituents.  Risk estimates which
exceed 1.0 are shown in bold.  This corresponds to a hazard
quotient exceeding 1.0 in the case of non-carcinogens and a
cancer probability greater than 1 in a million (1 x 10-6
risk) in the case of carcinogens.  The maximum 95-th
percentile and two high-end parameter groundwater risks for
each waste stream are summarized in Table 5.1.

5.2.1  Two High-End Parameters Results

The two high-end parameter results are presented for the
baseline on-site landfill scenario in Table 5.2 and for the
baseline off-site landfill scenario in Table 5.3.

5.2.2  Central Tendency Results

The central tendency  results are presented for the baseline
on-site landfill scenario in Table 5.2 and for the baseline
off-site landfill scenario in Table 5.3.

     
5.2.3  Monte Carlo Results

The baseline on-site landfill Monte Carlo results are give
in Table 5.4 and the off-site landfill Monte Carlo results
are presented in Table 5.5.

5.2.4  Contingent CSO Management

Results for both Monte Carlo and two high-end parameter
contingent CSO management analyses are presented in Table
5.6 along with the non-contingent CSO (only 1992 annual
Subtitle D Landfill waste quantities). 

5.2.5  Codisposal Results

On- and off-site codisposal results are shown in Table 5.7. 

5.2.6  TC Capped Results

TC-Capped results are presented in Tables 5.1 to 5.5.6.0   REFERENCES
     

1992 RCRA Section 3007 Survey of the Petroleum Refining
Industry Database.

API, 1989.  Hydrogeologic Database for Groundwater Modeling,
API Publication No. 4476, American Petroleum Institute.

Chappelle, F.M., 1993. Groundwater Microbiology and
Geochemistry.  John Wiley and Sons, New York.

Kool, J.B., P.S. Huyakorn, E.A. Sudicky, and Z.A. Saleem,
1994. A composite modeling approach for subsurface transport
of degrading contaminants from land disposal sites. Journal
of Contaminant Hydrology, vol.17 pp. 69-90.

Kollig, H.P., J.J. Ellington, E.J. Weber, and N.L. Wolfe. 
Pathway analysis of chemical hydrolysis for 14 RCRA
chemicals.  U.S. EPA Environmental Research Brief,
EPA/600/M-89/009.

Krumholtz, L., M.E. Caldwell and J.M Suflita. 1996. 
Biodegradation of BTEX Hydrocarbons Under Anaerobic
conditions.  Chapter in "Bioremediation: Principals and
Applications" R. Crawford & D. Crawford, eds.  Cambridge
University Press.

Newell, C.J., L. P. Hopkins, and P. B. Bedient, 1990.  A
Hydrogeologic Database for Ground-Water Modeling. 
Groundwater, Vol.  28, No. 5, pp. 703-714.

RCRA  CBI Documents, 1995.  Document Control Numbers
BP9500077 and  BP9500086.  (On and Offsite Petroleum
Refining Waste Landfill Statistics). 

Salinitro, J.P., 1993.  The Role of Bioattenuation in the
Management of Aromatic Hydrocarbon Plumes in Aquifers. 
Groundwater Monitoring and Remediation, p. 1-29.  Batelle
Press.  Columbus, Ohio.

USEPA, 1988.  Draft National Survey of Solid Waste
(Municipal) Landfill Facilities, EPA/530-SW-88-034, U.S.
Environmental Protection Agency, Washington D.C. 

USEPA, 1989.  USEPA Risk Assessment Guidance for Superfund,
Volume I: Human Health Evaluation Manual.  Washington D.C.,
20460.

USEPA, 1992.  Guidance on Risk Characterization for Risk
Managers, USEPA Memorandum, Washington, D.C., 20460.
 
USEPA, 1995a.  Listing Background Document for the 1992-1996
Petroleum Refining Determination.  U.S. EPA Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009), Washington, D.C., 20460.

USEPA, 1995b.   Assessments of Risks from the Management of
Petroleum Refining Waste:  Background Document.  U.S. EPA
Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009), Washington, D.C., 20460.

USEPA, 1995c.  Review of EPA's Composite Model for Leachate
Migration with Transformation Products-EPACMTP.  Science
Advisory Board Exposure Model Subcommittee of the
Environmental Engineering Committee.  Report
EPA-SAB-EEC-95-O10, August.  U.S. EPA Science Advisory
Board.

USEPA, 1995d.  Background Document for Groundwater Pathway
Analysis.  U.S. EPA Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009), Washington, D.C.,
20460.

USEPA, 1996a.  Study of Selected Petroleum Refining
Residuals, Industry Study.  U.S. EPA Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009),
Washington, D.C., 20460.

USEPA, 1996b.  Assessments of Risks from the Management of
Petroleum Refining Waste:  Background Document.  U.S. EPA
Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009), Washington, D.C., 20460.

USEPA, 1996c.  EPA's Composite Model for Leachate Migration
with Transformation Products.  Background Document, Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009), Washington, D.C., 20460.

USEPA, 1996d.  Background document for EPACMTP:  Finite
Source Methodology for Degrading Chemicals with
Transformation Products.  U.S. EPA, Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009),
Washington, D.C., 20460.

USEPA, 1996e.  Background  Document for Metals.  EPA's
Composite Model for Leachate Migration with Transformation
Products (EPACMTP).  Volume I:  Methodology.  U.S. EPA,
Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009), Washington, D.C., 20460.

USEPA, 1997a.  Docket Report Petroleum Waste Listing Notice
of Data Availability: Analyses Regarding Leaching of Oily
Waste. U.S. EPA, Washington D.C.

USEPA, 1997b.  Docket Report Petroleum Waste Listing Notice
of Data Availability: Active Lives of Landfills Used for the
Disposal of Petroleum Refining Wastes. U.S. EPA, Washington
D.C.

USEPA, 1997c.  Supplemental Background Document;
Nongroundwater Pathway Risk Assessment, Petroleum Process
Waste Listing Determination, U.S. EPA, Washington D.C.

USEPA, 1997d.  EPA's Composite Model for Leachate Migration
with Transformation Products.  User's Guide, Office of Solid Waste, OSW (Note:  Office of Solid Waste, OSW, was renamed Office of Resource Conservation and Recovery, ORCR, on January 18, 2009), Washington, D.C., 20460.

USGS, 1985.  National Water Summary, 1984, United States
Geological Survey Water-Supply Paper 2275.

Washington, J.W., 1995.  Hydrolysis Rates of Dissolved
Volatile Organic Compounds, Principles, Temperature Effects
and Literature Review.  Groundwater 33(3): 415-424.
                           APPENDIX A
                                
                WASTE AND LEACHATE CONCENTRATIONS                            APPENDIX B

               LANDFILL SIZE AND WASTE VOLUME DATA                           APPENDIX C
                                
            SENSITIVITY ANALYSES:  DATA AND RESULTS
(Data presented in this Tables C.2 through C.36 are from the
1995 Listing Background Document (USEPA, 1995a) and from
RCRA CBI Documents (RCRA CBI, 1995) )                            APPENDIX D

        CALCULATION OF TWO HIGH-END CODISPOSAL PARAMETERS
(Data presented in Tables D.1 through D.9 are from the 1995
Listing Background Document (USEPA, 1995a,) Study of Select
Petroleum Petroleum Refining Residuals (USEPA, 1996a), and
RCRA CBI Documents (RCRA CBI, 1995) )                            APPENDIX E

                    BASELINE MODELING RESULTS