United States      Science Advisory Board     EPA-SAB-RAC-97-008
Environmental      Washington, DC       September, 1997
Protection Agency	
  AN SAB REPORT:
  REVIEW OF THE MULTI-
  AGENCY RADIATION
  SURVEY AND SITE
  INVESTIGATION
  MANUAL (MARSSIM)
  PREPARED BY THE
  RADIATION ADVISORY
  COMMITTEE (RAC) OF THE
  SCIENCE ADVISORY BOARD

-------
                             September 30, 1997

EPA-SAB-RAC-97-008

Honorable Carol M. Browner
Administrator
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, DC 20460

Re:   Review of the Multi-Agency Radiation Survey and Site Investigation Manual
      (MARSSIM) draft dated December 1996

Dear Ms. Browner:

      This report was developed by the Radiation Advisory Committee (RAC) of the
Science Advisory Board (SAB) in response to a request to SAB Director, Dr. Donald G.
Barnes from Ms. E. Ramona Trovato, Director of EPA's  Office of Radiation and Indoor
Air (ORIA). Ms. Trovato requested that the Committee review technical aspects of the
Multi-Agency Radiation Survey and Site Investigation Manual (MARSSIM) draft dated
December 1996 and respond to the following questions  in its review:

      a)    Is the overall approach to the planning, data acquisition, data
           assessment, and data interpretation as described in the MARSSIM
           technically acceptable?

      b)   Are the methods and  assumptions for demonstrating compliance with a
           dose- or risk-based regulation technically acceptable?

      c)   Are the hypotheses and statistical tests and their method of application
           appropriate?

      For this review,  the RAC formed the MARSSIM Review Subcommittee (the
Subcommittee).  Because of the  inter-agency nature of MARSSIM, invitations were
extended to advisory committees of other Federal agencies to provide liaisons to the
Subcommittee, as a result of which a member of the U.S. Department of Energy's
Environmental Management Advisory Board agreed to participate in this review. The
Subcommittee met on January 22-23, 1997 and June 17-18, 1997, at which times it
was briefed by and had discussions with members of the multi-agency technical

-------
working group. In addition, the Subcommittee conducted a public teleconference on
July 21, 1997.

      MARSSIM was developed collaboratively by four Federal agencies having
authority for control of radioactive materials: U.S. Department of Defense (DOD), U.S.
Department of Energy (DOE), U.S. Environmental Protection Agency (EPA), and U.S.
Nuclear Regulatory Commission (NRC).  The Subcommittee was very impressed with
the collaboration demonstrated by these agencies, commissions and departments and
commends the agencies and their staff for   this effort. The product of this effort
(MARSSIM) addresses the need for a nationally consistent approach to conducting
radiation surveys of potentially radioactively contaminated sites that are being
considered for release to the public. We strongly encourage the timely completion
and publication of MARSSIM, in concert with the establishment of a formal
procedure for its future revisions, and it is not our intention that our comments
delay its completion.  Most of our comments can be incorporated in the previous
document, while some others may require consideration for future documents. We
strongly encourage these Federal agencies to continue to work together to address
issues related to radiation survey and site investigation.

      The attached report addresses the charge and elaborates upon significant
issues related to technical aspects of MARSSIM.  In particular, this report focuses on
specific technical issues that the Subcommittee felt would be most likely to require the
attention of the Multi-Agency workgroup in order to provide the most comprehensive
and technically-supportable basis for the manual.  Specifically, the Subcommittee
wishes to bring the following major findings and recommendations to your attention:

      a)    In general, the Subcommittee found that MARSSIM is nearly a finished
            product. The multi-agency team is commended for its work in addressing
            the many complex issues involved, resulting in the compilation of an
            exceptionally well-prepared reference which is technically sound and
            which will be a useful tool for guiding final status surveys.  The document
            provides generally consistent and explicit guidance for planning and
            conducting radiation surveys for the decommissioning of radiologically
            contaminated sites.

      b)    MARSSIM should discuss its rationale for limiting its scope to guidance
            for contaminated surface soils and building surfaces. Furthermore, it
            should more clearly state that radioactive contamination of subsurface
            soil, surface water, and ground water are explicitly excluded from its
            coverage. The document should include some discussion  of why these
            particular media were not included, the potential for incorrect decisions if
            they are not evaluated and the plans, if any, to cover them in the future.
            Also, MARSSIM should discuss the extent to which it is necessary to

-------
      evaluate scenarios under which subsurface contamination might be
      expected to contribute to surface contamination in the future, and how this
      affects the decision of whether the site meets release criteria.

c)    Descriptions of field measurement methods, instruments, and operating
      procedures in MARSSIM are technically sound but incomplete.  Some
      additions, clarifications,  and corrections are noted in our report.
      MARSSIM should provide guidance for the development of standardized
      procedures,  including a  list of considerations for designing site-specific
      surface-soil sampling and  preparation methods so as to ensure that
      samples will be representative of the materials of concern in deriving the
      derived concentration guideline levels (DCGLs) for the site.

d)    Descriptions of the selection and operation of radiation detection
      instruments for laboratory  analyses are technically sound and represent
      standard practice but may not be state-of-the-art.  MARSSIM should
      standardize the level of  detail used in its presentation of this material and
      should also provide information on the planned scope and current status
      of plans to prepare a manual on Multi-Agency Radiological Laboratory
      Analytical Protocols (MARLAP). MARLAP may be a more appropriate
      forum than MARSSIM in which to provide more thorough in-depth
      guidance to the user on the selection and operation of laboratory
      instrumentation.

e)    The Subcommittee believes that it is critically important that the
      assumptions and procedures used in MARSSIM to make comparisons
      with the DCGLs match those used in defining the DCGLs. For example, if
      a DCGL for soil is derived  from  a dose limit or risk criterion by assuming
      that a receptor ranges over a certain area on a random basis, then the
      same area should be used for spatial averaging in the MARSSIM
      statistical analyses.  Such averaging is usually performed from the
      standpoint of potential human receptors. The manual should note that
      different spatial and temporal scales of averaging will be necessary if
      dose- and risk-based criteria are applied to components of the ecosystem
      other than humans for derivation of a DCGL. This recommendation
      assumes that the DCGL is derived in a manner appropriate for
      characterizing human and/or ecological exposures likely to occur at the
      site under investigation.

f)     Although  MARSSIM  is applicable to the majority of contaminated sites,
      there appear to be cases that MARSSIM, as currently written, would have
      trouble addressing. These include: 1) cases dealing with the release of
      sites that had been contaminated with  naturally occurring radionuclides

-------
      and in which the DCGL is contained within the ambient (background)
      analyte variability, and 2) cases in which a reference background cannot
      be established. The Subcommittee recommends that future revisions of
      MARSSIM provide guidance to the user regarding appropriate choices
      when such conditions are encountered. For example, the null hypothesis
      might be redefined to be that the distribution of site radioactivity is no
      different from that at the reference site or than ambient radioactivity in
      general.

g)    MARSSIM properly warns the user that the DCGL is not free of error and
      that the uncertainty associated with this quantity may be considerable if
      derived using generic assumptions and parameter values.  However, its
      discussion of this issue is relegated to an appendix. This important
      aspect, together with an expanded discussion of its implications for the
      release decision, needs to be disclosed more prominently in the text of
      the main document.  It is clearly undesirable to design a survey around a
      DCGL that may not be relevant to the actual conditions at a site, such that
      actual exposures, doses, and risks would be largely different than those
      used to derive the generic DCGL. Consequently, MARSSIM should more
      strongly encourage the user to examine critically the assumptions made in
      any model used to derive DCGLs for a site in order to determine whether
      application of site-specific information and parameters would result in
      significant modifications to the proposed  DCGL, or whether development
      of a site-specific model would be warranted in order to obtain a DCGL
      that is more relevant to the human and ecological exposure conditions
      prevailing at the site.

h)    In MARSSIM, the preferred null hypothesis is that a survey unit is not
      ready for release and the information gathered must be sufficient, with a
      high degree of confidence, to accept the alternative hypothesis (i.e., that
      the unit meets the release criteria). Furthermore, MARSSIM discusses in
      detail two non-parametric procedures, the Wilcoxon Rank-Sum test and
      the Sign test, for testing this  hypothesis.  However, MARSSIM allows
      more flexibility  in defining the null hypothesis and in choosing statistical
      analysis methods to test that hypothesis than may be readily apparent to
      most readers.  The existence of this flexibility needs to be more clearly
      stated and the  criteria for selecting among potentially applicable tests
      need to be described.

i)     MARSSIM's  discussion about the mean and median should be revised in
      order to ensure that the correct statistical parameter is used to compare
      concentrations in the survey area to those in the reference area.  The
      target statistic for any exposure assessment should be the arithmetic

-------
            mean concentration for a defined area, together with the uncertainty
            associated with the estimate of the mean.  For a normally distributed
            population, the mean and the median are identical in value.  However,
            when the distribution of sample evidence is moderately to highly skewed,
            then non-parametric statistical techniques cannot be used to determine
            the uncertainty associated with the estimate of the arithmetic mean, and
            the median of such a sample set will underestimate the true arithmetic
            mean of surface contamination.  The majority of soil sampling programs
            usually reveal highly skewed distributions.  Therefore, the Wilcoxon
            Rank-Sum test and the Sign test, which are appropriate for testing
            differences in median concentrations, may not be appropriate to test for
            differences in mean concentrations.

      j)     The guidance provided by MARSSIM may introduce an additional
            measure of conservatism in the process of setting and determining
            compliance with radiation cleanup standards, compounding the
            conservatism already likely to occur in developing default DCGLs.
            Release decisions may be biased correspondingly. MARSSIM should
            include a qualitative summary of any biases that may result from its
            assumptions and policy choices, and recommend that the planning team
            be similarly revealing when developing a site-specific survey design.

Finally, we offer the following comment on an issue that was outside the scope of our
charge but that we felt was important to bring to your attention:

      k)     DCGLs are critical for determining the acceptability of residual levels of
            radioactivity remaining after a site has been remediated.  The
            Subcommittee suggests that the various approaches proposed for
            derivation of DCGLs (not the individual site-specific DCGLs) be reviewed
            and evaluated. This evaluation can be performed by an interagency
            group and by the EPA/SAB.  This evaluation should focus on the
            strengths and weaknesses of current methodologies and opportunities to
            refine generic DCGLs with improved site-specific models and data. This
            review is important but outside the current scope of the SAB/RAC review
            of MARSSIM perse.

      We would like to again commend the multi-agency approach used so
successfully to produce MARSSIM and to encourage the timely completion and
publication of the document. We believe that all of the above recommendations except
for items f) and k) can be incorporated into the document at this time, and the
remaining two items can be addressed in the future. We  strongly encourage the
continuation of this multi-agency approach to plan for future revisions to MARSSIM as
well as for the development of additional radiation survey manuals, such as for

-------
subsurface soils, ground water, and sewers. It would also be beneficial to apply this
successful interagency approach to the preparation of other manuals such as on site
stabilization, decommissioning techniques, and standardized sampling procedures for
various media.

      The RAC and its Subcommittee appreciate the opportunity to provide this report
to you and we hope that it will be helpful.  We look forward to your response to this
report in general, and to the comments and recommendations in this letter in particular.

                              Sincerely,
                                    /signed/
                              Dr.Genevieve M. Matanoski, Chair
                              Science Advisory Board
                                    /signed
                              Dr. James E. Watson, Jr., Chair
                              Radiatio Advisory n Committee and
                              MARSSIM Review Subcommittee
                              Science Advisory Board

-------
                                   NOTICE

      This report has been written as a part of the activities of the Science Advisory
Board, a public advisory group providing extramural scientific information and advice to
the Administrator and other officials of the Environmental Protection Agency. The
Board is structured to provide a balanced, expert assessment of scientific matters
related to problems facing the Agency. This report has not bbeen reviewed for
approval by the Agency; hence, the comments of this report do not necessarily reflect
the views and policies of the Environmental Protection Agency or of other Federal
Agencies.  Any mention of trade names or commercial products does not constitute
endorsement or recommendation for use.

-------
                                ABSTRACT
      The EPA Science Advisory Board's (SAB) Radiation Advisory Committee (RAC)
/ MARSSIM Review Subcommittee (the Subcommittee) reviewed technical aspects of
the Multi-Agency Radiation Survey and Site Investigation Manual (MARSSIM) (12/96).
The reviewed document was developed collaboratively by four Federal agencies,
departments and commissions having authority for control of radioactive materials:
Department of Defense, Department of Energy, Environmental Protection Agency, and
Nuclear Regulatory Commission. MARSSIM addresses the need for a nationally
consistent approach to conducting radiation surveys of potentially radioactively
contaminated sites that are being considered for release to the public.  A condition of
release is a demonstration that residual  radioactivity levels do not exceed a specified
risk or dose level, also known as a release criterion.  MARSSIM provides guidance to
users assessing the survey results for surface soils and building surfaces. The
Subcommittee concluded that MARSSIM needed to more clearly emphasize that its
scope is limited to guidance for surficial  media  and not to radioactive contamination of
any other media. The Subcommittee found that descriptions of field and laboratory
measurement methods, instruments and operating procedures in MARSSIM were
generally technically sound although somewhat incomplete or out-of-date. The
Subcommittee stressed that MARSSIM needed to revise its guidance on the use of the
median in place of the mean to represent the average contaminant level in an area and
to more clearly state the user's flexibility in selecting  statistical methods for evaluating
analytical data against the release criterion. The Subcommittee recommended that
MARSSIM provide a prominent discussion of the uncertainties and level of
conservatism inherent in the setting of the release criterion.
Key Words: Cleanup Standards, Environmental Radiation, Nuclear Facilities,
Environmental Quality, Radionuclide Cleanup, Radiological Characterization of Site

-------
                        SCIENCE ADVISORY BOARD
                  RADIATION ADVISORY COMMITTEE (RAC)
    MULTI-AGENCY RADIATION SURVEY AND SITE INVESTIGATION MANUAL
                   REVIEW SUBCOMMITTEE (MARSSIMRS)

Chair
Dr. James E. Watson, Jr., Professor, Department of Environmental Sciences and
Engineering, University of North Carolina at Chapel Hill, NC

Members
 Dr. William Bair, (Retired) Former Manager, Life Sciences Center, Battelle Pacific
Northwest Laboratory, Richland, WA

Dr. Stephen L. Brown, Director, R2C2 (Risks of Radiation and Chemical Compounds),
Oakland, CA (Statistics Workgroup Coordinator)

Dr. June Fabryka-Martin, Staff Scientist, Chemical Science and Technology Division,
Los Alamos National Laboratory, Los Alamos, NM

Dr. Thomas F. Gesell, Professor of Health Physics and Director, Technical Safety
Office, Idaho State University,  Pocatello, ID

Dr. F. Owen Hoffman, President, SENES Oak Ridge, Inc., Center for Risk Analysis,
Oak Ridge,  TN

Dr. Janet Johnson, Senior Radiation Scientist, Shepherd Miller, Inc., Fort Collins,  CO
(Monitoring  Workgroup Coordinator)

Dr. Bernd Kahn, Professor, School of Nuclear Engineering and Health Physics, and
Director, Environmental Resources Center, Georgia Institute of Technology, Atlanta,
GA

Dr. Ellen Mangione, M.D., M.P.H.,  Director, Disease Control and Environmental
Epidemiology Division, Colorado Department of Health, Denver, CO

Dr. Paul J. Merges, Chief, Bureau of Pesticides and Radiation, Division  of Solid and
Hazardous Materials, New York State Department of Environmental Conservation,
Albany, NY  (Integration Workgroup Coordinator)

Consultants
Dr. Michael E. Ginevan, M.E.  Ginevan & Associates, Silver Spring, MD

-------
Dr. David G. Hoel, Chairman and Professor, Department of Biometry & Epidemiology,
Medical University of South Carolina, Charleston, SC

Dr. David E. McCurdy, Chief Scientist, Yankee Atomic Electric Company, Bolton, MA

Dr. Frank L. Parker, Vanderbilt University, Nashville, TN [Liaison from Environmental
Management Advisory Board (EMAB), U.S. Department of Energy (DOE)]

Science Advisory Board Staff
Dr. K. Jack Kooyoomjian, Designated Federal Official, U.S. EPA, Science Advisory
Board (1400), 401 M Street, SW, Washington, DC 20460

Mrs. Diana L. Pozun, Staff Secretary, U.S. EPA, Science Advisory Board (1400), 401
M Street, SW, Washington, DC 20460
                                     IV

-------
                          TABLE OF CONTENTS

                                                                       Page
1.  Executive Summary	  1
2.  Introduction	  6
   2.1  Overview of the Multi-Agency Radiation Survey and Site Investigation Manual 6
   2.2  Charge to the SAB	  7
   2.3  SAB Review Procedure 	  7
3.  Assessment of MARSSIM's Overall Approach 	  8
   3.1  Response to Charge a)  	  8
   3.2  MARSSIM's Scope of Coverage	  9
       3.2.1  Relationship to Federal Regulations  	  9
       3.2.2 MARSSIM as a Guidance Document	10
       3.2.3 Use and Citations of References  	  11
       3.2.4 Application to Surface Soil and Building Surfaces  	    11
       3.2.5 Application to Subsurface Soil and Other Environmental Media  	  12
   3.3  Planning Guidance	  14
       3.3.1  Overall Approach to Planning the Surveys	  14
       3.3.2 Public Involvement   	  14
4.  Data Acquisition and Assessment	  16
   4.1  Response to Charge b)  	  16
   4.2  Organization of Information on Analytical Instruments in MARSSIM  	  18
   4.3  Guidance on Data Acquisition  	  18
       4.3.1  Scanning Surveys	  18
       4.3.2 Instrument Calibration	  19
       4.3.3 Background Measurements and Adjustments to Measured Values  ...  19
   4.4  Field  and Laboratory Instrumentation	  20
       4.4.1  MARSSIM's Overview of  Instrumentation	  20
       4.4.2 Additional Measurement Techniques 	  22
       4.4.3 Relationship between MARSSIM and MARLAP	  22
   4.5  Quality Assurance and Quality  Control  	  22
       4.5.1  Data Quality Objective  (DQO)  Process as Related to Measurement
            Systems	  22
       4.5.2 National Quality Assurance Standards	  23
       4.5.3 Data Verification and Validation 	  23
       4.5.4 Quality Control Samples and Measurements  	  24
       4.5.5 Use of Spikes 	  24
       4.5.6 Data Quality Indicators	  25
5.  Demonstration of Compliance	  26
   5.1  Response to Charge c)	  26
   5.2  Importance of Appropriate DCGLs	  27
   5.3  Relationship between DCGL  and Reasonable Measurement Requirements .  29

-------
                   TABLE OF CONTENTS (continued)
                                                                       Page
   5.4 Determination and Use of "Background" Measurements	 30
   5.5 Statement of the Null Hypothesis and Statistical Tests 	 31
       5.5.1  Statement of the Null Hypothesis 	 31
       5.5.2 Selecting Appropriate Statistical Tests	 32
       5.5.3 Specified Non-Parametric Methods  	 32
       5.5.4 Importance of Distinguishing between Median and Mean  	 34
       5.5.5 Alternative Statistical Methods 	 35
       5.5.6 Definition  of the Gray Region 	 37
       5.5.7 Specification of Sample Size	 37
       5.5.8 Treatment of Outliers  	 38
6.  Broader Issues  	 39
   6.1  Application of MARSSIM Outside Site Boundaries  	 39
   6.2 Conservatism of MARSSIM  	 39
   6.3 Post MARSSIM	 40
7.  Findings and Recommendations	 41
   7.1  Overall Approach to  Planning Surveys	 42
   7.2 Data Acquisition and Assessment 	 43
   7.3 Demonstrating Compliance  	 44
   7.4 Broader Issues	 46
List of Appendices:
A.   References 	 A-1
B    List of Acronyms  	 B-1
C.   Glossary 	 C-1
                                     VI

-------
                        1.  EXECUTIVE SUMMARY
     The EPA Science Advisory Board's (SAB) Radiation Advisory Committee (RAC) /
MARSSIM Review Subcommittee (the Subcommittee) reviewed technical aspects of the
Multi-Agency Radiation Survey and Site Investigation Manual (MARSSIM) (12/96).  The
reviewed document was developed collaboratively by four Federal agencies,
departments and commissions having authority for control of radioactive materials:
Department of Defense, Department of Energy, Environmental Protection Agency, and
Nuclear Regulatory Commission. MARSSIM addresses the need for a nationally
consistent approach to conducting radiation surveys of potentially radioactively
contaminated sites that are being considered for release to the public. A condition  of
release is a demonstration that residual radioactivity levels do not exceed a specified
risk or dose level, also known as a release criterion. MARSSIM provides guidance to
users performing and assessing the results of such a demonstration for surface soils
and building surfaces.

     The Subcommittee was requested by the Agency's Office of Radiation and  Indoor
Air (ORIA) to respond to the following charge in its review:

     a)    Is the overall approach to the planning, data acquisition, data assessment,
           and data interpretation as described in the MARSSIM technically
           acceptable?

     b)   Are  the methods and assumptions  for demonstrating compliance with a
           dose- or risk-based regulation technically acceptable?

     c)   Are  the hypotheses and statistical tests and their method of application
           appropriate?

     The Subcommittee's full report addresses this charge and elaborates upon
significant issues related to technical aspects of MARSSIM. The report focuses on
specific technical issues that the Subcommittee felt would be  most likely to  require the
attention of the multi-agency workgroup in order to provide the most comprehensive
and technically-supportable basis for the manual. The Subcommittee made the
following major findings and recommendations:

     a)    In general, the Subcommittee found that MARSSIM is nearly a finished
           product.  The multi-agency team is  commended for its work in addressing
          the many complex issues involved, resulting in the compilation of an
          exceptionally well-prepared reference which is technically sound and which
          will be a useful tool for guiding final status surveys.  The document
           provides generally consistent and explicit guidance for planning and

-------
     conducting radiation surveys for the decommissioning of radiologically
     contaminated sites.

b)   MARSSIM should discuss its rationale for limiting its scope to guidance for
     contaminated surface soils and building surfaces.  Furthermore, it should
     more clearly state that radioactive contamination of subsurface soil, surface
     water, and ground water are explicitly excluded from its coverage. The
     document should include some discussion of why these particular media
     were not included, the potential for incorrect decisions if they are not
     evaluated and the plans, if any, to cover them in the future. Also,
     MARSSIM should discuss the extent to which it is necessary to evaluate
     scenarios under which subsurface contamination might be expected to
     contribute to surface contamination in the future, and how this affects the
     decision of whether the site meets release criteria.

c)   Descriptions of field measurement methods, instruments, and operating
     procedures in MARSSIM are technically sound but incomplete. Some
     additions, clarifications, and corrections are noted in our report.   MARSSIM
     should provide guidance for the development of standardized procedures,
     including a list of considerations for designing site-specific surface-soil
     sampling and preparation methods so as to ensure that samples will be
     representative of the materials of concern in deriving the derived
     concentration guideline levels (DCGLs) for the site.

d)   Descriptions of the selection and operation of radiation detection
     instruments for laboratory analyses are technically sound and represent
     standard practice but may not be state-of-the-art. MARSSIM should
     standardize the level of detail used in its presentation of this material  and
     should also provide information on the planned scope and current status of
     plans to prepare a manual on Multi-Agency Radiological Laboratory
     Analytical Protocols (MARLAP). MARLAP may be a more  appropriate
     forum than MARSSIM  in which to provide more thorough in-depth guidance
     to the user on the selection and operation of laboratory instrumentation.

e)   The Subcommittee believes that it is  critically important that the
     assumptions and procedures used in MARSSIM to make comparisons with
     the DCGLs match those used in defining the  DCGLs. For example, if a
     DCGL for soil is derived from a dose limit or risk criterion by assuming that
     a receptor ranges over a certain area on a random basis, then the same
     area should be used for spatial averaging in the MARSSIM statistical
     analyses. Such averaging is usually performed from the standpoint of
     potential human receptors. The manual should note that different spatial
     and temporal scales of averaging will be necessary if dose- and risk-based

-------
     criteria are applied to components of the ecosystem other than humans for
     derivation of a DCGL.  This recommendation assumes that the DCGL is
     derived in a manner appropriate for characterizing human and/or ecological
     exposures likely to occur at the site under investigation.

f)    Although MARSSIM is applicable to the majority of contaminated sites,
     there appear to be cases that MARSSIM, as currently written, would have
     trouble addressing.  These include: 1) cases dealing with the release of
     sites that had been contaminated with naturally occurring radionuclides and
     in which the DCGL is contained within the ambient (background) analyte
     variability, and 2) cases in which a reference background cannot be
     established. The Subcommittee recommends that future revisions of
     MARSSIM provide guidance to the user regarding appropriate choices
     when such conditions are encountered.  For example, the null hypothesis
     might be redefined to be that the distribution of site radioactivity is no
     different from that at the reference site or than ambient radioactivity in
     general.

g)   MARSSIM properly warns the user that the DCGL is not free of error and
     that the uncertainty associated with this quantity may be considerable if
     derived using generic assumptions and parameter values. However,  its
     discussion of this issue is relegated to an appendix.  This important aspect,
     with an  expanded discussion of its implications for the release decision,
     needs to be disclosed more prominently in the text of the main document.  It
     is clearly undesirable to design a survey around a DCGL that may not be
     relevant to the actual conditions at a site, such that actual exposures,
     doses, and risks would be largely different than those used to derive the
     generic DCGL.  Consequently, MARSSIM should more strongly encourage
     the user to examine critically the assumptions made in any model used to
     derive DCGLs for a site in order to determine whether application of site-
     specific information and parameters would result in significant modifications
     to the proposed DCGL, or whether development of a site-specific model
     would be warranted  in order to obtain a DCGL that is more relevant to the
     human and ecological exposure conditions prevailing at  the site.

h)   In MARSSIM, the preferred null hypothesis is that a survey unit is not ready
     for  release and the information gathered must be sufficient, with a high
     degree of confidence, to accept the alternative hypothesis (i.e., that the unit
     meets the release criteria). Furthermore, MARSSIM discusses in detail two
     non-parametric procedures, the Wilcoxon Rank-Sum test and the Sign test,
     for  testing this hypothesis. However, MARSSIM allows more flexibility in
     defining the null hypothesis and in  choosing statistical analysis methods to
     test that hypothesis than may be readily apparent to most readers.  The

-------
           existence of this flexibility needs to be more clearly stated and the criteria
           for selecting among potentially applicable tests need to be described.

     i)     MARSSIM's discussion about the mean and median should be revised in
           order to ensure that the correct statistical parameter is used to compare
           concentrations in the survey area to those in the reference area.  The
           target statistic for any exposure assessment should be the arithmetic mean
           concentration for a defined area, together with the uncertainty associated
           with the estimate of the mean.  For a normally distributed population, the
           mean and the median are identical in value.  However, when the
           distribution of sample evidence is moderately to highly skewed, then
           non-parametric statistical techniques cannot be used to determine the
           uncertainty associated with the estimate of the arithmetic mean, and the
           median of such a sample set will underestimate the true arithmetic mean of
           surface contamination.  The majority of soil sampling programs usually
           reveal highly skewed distributions. Therefore, the Wilcoxon Rank-Sum test
           and  the Sign test, which is appropriate for testing differences  in median
           concentrations, may not be appropriate to test for differences  in mean
           concentrations.

     j)     The guidance provided by MARSSIM may introduce an additional measure
           of conservatism  in the process of setting and determining compliance with
           radiation cleanup standards, compounding the conservatism already likely
           to occur in developing default DCGLs.  Release decisions may be biased
           correspondingly. MARSSIM should include a qualitative summary of any
           biases that may  result from its assumptions and policy choices, and
           recommend that the planning team be similarly revealing when developing
           a site-specific survey design.

Finally, the Subcommittee offered the following comments on issues that were outside
the scope of the charge but that were felt to be important:

     k)    DCGLs are critical for determining the acceptability of residual levels of
           radioactivity remaining after a site has been remediated.  The
           Subcommittee suggested that the various approaches proposed for
           derivation of DCGLs (not the individual site-specific DCGLs)  be reviewed
           and  evaluated. This evaluation can be performed by an interagency group
           and  by the EPA/SAB. This evaluation should focus on the strengths and
           weaknesses of current methodologies and opportunities to refine generic
           DCGLs with improved site-specific models and data.  This review is
           important but outside the current scope of the SAB/RAC review of
           MARSSIM perse.

-------
I)    The Subcommittee strongly encouraged the continuation of this multi-
     agency approach to plan for future revisions to MARSSIM as well as for the
     development of additional radiation survey manuals, such as for subsurface
     soils, ground water, and sewers.  It would also be beneficial to apply this
     successful interagency approach to the preparation of other manuals such
     as on site stabilization, decommissioning techniques, and standardized
     sampling procedures for various media.

-------
                            2.  INTRODUCTION
2.1   Overview of the Multi-Agency Radiation Survey and Site Investigation
     Manual

     A Multi-Agency Radiation Survey and Site Investigation Manual (MARSSIM)
(U.S. EPA et al., 1996) is being developed collaboratively by four Federal agencies,
departments and commissions having authority for control of radioactive materials: U.S.
Department of Defense (DOD), U.S. Department of Energy (DOE), U.S. Environmental
Protection Agency (EPA), and U.S. Nuclear Regulatory Commission (NRC). The
December 1996 draft MARSSIM report reviewed by the EPA's Science Advisory Board
(SAB) was prepared by a multi-agency technical working group composed of
representatives from these four Federal agencies, departments, and commissions.
When finalized,  MARSSIM will be a multi-agency consensus document.

     MARSSIM addresses the need for a nationally consistent approach  to conducting
radiation surveys of potentially radioactively contaminated sites that are being
considered for release to the public. A condition of release is a demonstration to the
responsible Federal or state agency that any residual radioactivity levels do not exceed
a specified risk or dose level, also known as a release criterion, established by the
responsible agency.  MARSSIM assists site personnel and others in performing and
assessing the results of such a demonstration for surface soils and building surfaces.
The guidance provided in MARSSIM is intended to be not only scientifically rigorous
but also sufficiently flexible to be applied to a  diversity of sites at different stages of the
cleanup process.

     Guidance  is  provided in MARSSIM on historical site assessment,  preliminary
survey considerations, survey planning and design, field measurement methods and
instrumentation, sampling and preparation for laboratory measurements, interpretation
of survey results, and quality assurance and quality control measures. Survey types
considered include scoping, characterization, remedial action support, and final status
surveys.  For each type of survey, guidance is provided on survey design, conducting
the survey, evaluating results, and documentation.  Results of the final status survey
are used to determine whether or not the release criterion has been  met.  Statistical
tests are presented for use in the decision-making process.

     MARSSIM notes several areas that are  beyond its scope. These  areas include
translation of dose or risk standards into radionuclide-specific concentrations;
demonstration of compliance with ground water or surface water regulations;
management of vicinity properties not under government or licensee control; surveys of
other contaminated media (such as subsurface soil, building materials, and  ground
water); and the release of contaminated components and equipment.

-------
2.2  Charge to the SAB

     The multi-agency technical working group that prepared the draft MARSSIM
agreed to request review of this draft by the SAB through its Radiation Advisory
Committee (RAC). This request was submitted to Dr. Donald G. Barnes, the SAB
Director, by Ms. E. Ramona Trovato, Director of EPA's Office of Radiation and Indoor
Air (ORIA) in a memo dated July 30, 1996.  Ms. Trovato  requested that the Committee
respond to the following questions in its review:

     a)    Is the overall approach to the planning, data acquisition, data assessment,
           and data interpretation as described in the MARSSIM technically
           acceptable?

     b)    Are the methods and assumptions for demonstrating compliance with a
           dose- or risk-based regulation technically acceptable?

     c)    Are the hypotheses and statistical tests and  their method of application
           appropriate?

     Because of the inter-agency nature of MARSSIM, Ms. Trovato also noted that it
would be appropriate to involve persons from advisory committees of the other Federal
agencies.

2.3  SAB Review Procedure

     The primary review document is the December 1996 draft report, titled
"Multi-Agency Radiation Survey and Site  Investigation Manual (MARSSIM)" (U.S. EPA
et al., 1996). Additional background documentation was provided by ORIA in support
of this review.

     For this review, the RAC formed the MARSSIM Review Subcommittee (the
Subcommittee). Invitations were extended to advisory committees of other Federal
agencies, departments, and commissions to provide liaisons to the Subcommittee, as a
result of which a member of DOE's Environmental Management Advisory Board agreed
to participate in this review. Three workgroups were formed within the Subcommittee to
focus on areas of the report dealing with a) the integration of the MARSSIM process, b)
radiation measurement activities, and c) statistical analyses of the measurement data.
The Subcommittee met on January 22-23, 1997 and June 17-18, 1997, at which times it
was briefed by and had discussions with members of the multi-agency technical
working group. In addition, the Subcommittee conducted a public teleconference
meeting on July 21, 1997 to reach closure on this draft report.

-------
        3.  ASSESSMENT OF MARSSIM's OVERALL APPROACH


3.1   Response to Charge a)

     The charge from ORIA's director regarding MARSSIM's overall approach was:

     Charge a) Is the overall approach to the planning, data acquisition, data
     assessment, and data interpretation as described in the MARSSIM
     technically acceptable?

     MARSSIM was developed collaboratively by four Federal agencies, departments,
and commissions having authority for control of radioactive materials: Department of
Defense (DOD), Department of Energy (DOE), Environmental  Protection Agency
(EPA), and  Nuclear Regulatory Commission (NRC). The Subcommittee was impressed
with the collaboration demonstrated by these agencies, departments, and commissions,
and commends these organizations and their staff for this effort.  The product of this
effort (MARSSIM) addresses the need for a nationally consistent approach to
conducting  radiation surveys of potentially radioactively contaminated sites that are
being considered for release to the public.  With the incorporation of the comments in
our report, the Subcommittee feels that the overall approach to the planning, data
acquisition,  data assessment, and data interpretation as described in MARSSIM will be
technically acceptable and that MARSSIM will fulfill  its stated objectives.

     In its deliberations on this charge, the Subcommittee concluded that MARSSIM
should more clearly emphasize that its scope is limited to guidance for contaminated
surface soils and building surfaces.  Other considerations not addressed in MARSSIM
include the  radioactive contamination of subsurface soil, surface water, and ground
water. The document should include some discussion of why these particular media
were not included and the plans, if any, to cover them  in the future. Also, the
Subcommittee judged it important that MARSSIM at least mention scenarios under
which subsurface contamination might contribute to surface contamination in the future.
Detailed discussion of each of these points is provided below.

     We encourage the timely completion and publication of MARSSIM, and it is not
our intention that our comments delay its completion.  Most of our comments can be
incorporated in the previous document, while some  others will require consideration for
future documents. We encourage these Federal agencies to continue to work together
to address issues related to radiation survey and site investigation.
                                      8

-------
3.2  MARSSIM's Scope of Coverage

3.2.1       Relationship to Federal Regulations

     The MARSSIM brings together, for the first time of which the Subcommittee is
aware, DOE, EPA, NRC, and DOD with a common method of site surveys and
investigations for surficial contamination. Adopting MARSSIM will mean that surveys
done for any of the participating organizations will be immediately transparent to all of
the others.  The information in MARSSIM's Appendix F is particularly useful for
understanding the complex interrelationships between radiological surveys undertaken
following  MARSSIM guidelines and those undertaken as part of CERCLA and RCRA
cleanup actions.  In fact, if the site has significant chemical contamination that might
limit its suitability for unrestricted release, then it may be efficient to coordinate
investigations and sampling with the corresponding CERCLA Remedial
Investigation/Feasibility Study (RI/FS) or closure surveys.

     On  the other hand, because of its close tie to various regulatory programs,
MARSSIM also has the challenge and  responsibility to provide information that is
consistent with the current regulatory picture. Because this aspect is somewhat of a
moving target, we recommend that references not be made to regulations in draft form,
and that such references in MARSSIM instead be revised to refer to the regulatory
responsibilities of the agencies.   For example, the discussion on page 1 -1, lines 8-14,
could be  revised as follows (proposed changes are underlined):

     "The Environmental Protection Agency (EPA), the Nuclear Regulatory
     Commission (NRC), and the Department of Energy (DOE) are charged with the
     preparation of regulations for the release of certain categories of  radioactively
     contaminated sites following such cleanup.... Some states may prepare similar
     rules that will apply to sites under their control."

     It needs to be emphasized early in MARSSIM that several critical aspects of the
cleanup process are not within its scope. For example, MARSSIM provides guidance
on verifying cleanup but is not intended to provide guidance on methods for cleanup
activities  nor on methods for setting cleanup standards. Furthermore, MARSSIM
should note for its users that compliance with the Derived Concentration Guideline
Level (DCGL) is only one of the considerations for release of a site for unrestricted use.
Other considerations not addressed in  MARSSIM include the radioactive contamination
of subsurface soil, surface water, or ground water. Furthermore, if the release criterion
is supposed to be ultimately risk-based, how will the risks of residual chemicals be
factored into the decision about release criteria for radionuclides (see MARSSIM page
3-5)?  If a site has mixed hazardous and radiological contamination, their combined risk
will need  to be considered.

-------
     MARSSIM should discuss the extent to which it is necessary to evaluate
scenarios under which subsurface contamination might be expected to contribute to
surface contamination in the future, and how this affects the decision of whether the
site meets release criteria.  For example, cleanup under 40 CFR 192 (Uranium Mill
Tailing Radiation Control Act, UMTRCA) standards allows a higher concentration of
radionuclides to be left below the surface, which may well be exposed following a large
flooding or erosion event.

3.2.2 MARSSIM as a Guidance Document

     A major concern of the Subcommittee is the use to which MARSSIM will be put,
i.e., is there a risk that this document will be used by regulatory agencies as de facto
regulation rather than guidance, discouraging flexibility in  approaches and precluding
use of creative and reasonable site-specific measurement methodology? This concern
is particularly applicable to sites where the radiological  constituents may be present at
levels not distinguishable from background.  Consequently, the Subcommittee suggests
that Section 2.6 in MARSSIM either begins with a paragraph, or else adds a new
section (e.g., a new Section 2.6.5), that emphasizes the function of the document for
guidance and that offers the opportunity for the surveyor and investigator to provide an
alternative plan to the regulatory agency.  This plan has to meet the agency's criteria
for acceptable and pertinent information, but can be adapted to the specific
circumstances of the site.

     While in the realm of policy options, the Subcommittee noted a related concern
with the question of sufficiency, i.e., will use of methods described in MARSSIM be
considered sufficient to show compliance with regulatory requirements?  As an
extension of this issue, if a scientific peer review has concluded that the  methods in
MARSSIM are sufficient on technical grounds to decide whether or not a particular site
meets its DCGL criterion, then to what extent will a regulatory agency be bound to
accept findings based on surveys performed strictly according to MARSSIM guidance?

     The document is confusing by the extent to which various terms and
requirements have legal or regulatory connotations, as opposed to common
convention. For example, MARSSIM should explain where the concepts of
DCGL—and its subsets DCGLW and DCGLEMC—come from. References for the
classification of areas as Class 1, Class 2 and Class 3, the use of a graded approach,
and the use of the surrogate approach or the unity rule  should also be provided.
MARSSIM should also clearly distinguish between actions that are merely good
practice and those that are necessary to ensure defensible data.  Examples of the latter
case are maintenance of documentation  on sample chain  of custody and use of
calibrated instrumentation.
                                      10

-------
3.2.3 Use and Citations of References

     The reader's confidence that the document presents up-to-date information is
eroded by the extent to which supporting references are outdated, inappropriate, or not
cited at all. Although MARSSIM may well be the best compilation of cleanup
documents available in the U.S., its citations and references on industry standards and
instrumentation could be improved.  In particular, references to publications by the
American National Standards Institute (ANSI) and the American Society of Testing and
Materials (ASTM) should be updated because these are revised on a five-year cycle.
Also, many of the references cited in the report do not appear in the list of references,
and vice versa as well. Numerous examples of these shortcomings are listed in our
detailed comments submitted separately to ORIA (Kooyoomjian, 1997).

     There is also a heavy reliance on the unpublished literature to support the
information in MARSSIM, including contractor reports, reports at meetings, and even
draft documents. In many other instances, information  is provided without identifying
any reference.  In some cases there may be no choice; and we recognize that
government or contractor reports, while not always easily accessible, may nonetheless
be the best or most appropriate information sources.  In other cases, however,
equivalent or better sources can be found in the conventionally published literature.
For example, there are several sections on instrumentation in the MARSSIM draft but
no citation of any of the standard nuclear instrumentation texts except for a citation to
Knoll (1979) regarding propagation of error,  for which a statistics book might have been
a better source. Even then, Knoll is not listed in the reference section nor is 1979 the
most recent edition (it is 1989). The credibility of the MARSSIM report would be greatly
enhanced by a strong reference section that includes recognized, readily accessible
literature  wherever possible.

3.2.4 Application to Surface Soil and Building Surfaces

     In the view of the Subcommittee, MARSSIM is largely intended to guide the user
in the acquisition and assessment of radiological data relevant to the decision about
releasing the site. Of course, if the site fails, remediation is probably indicated, and the
survey can help identify the areas needing such work. But the survey should not be
designed to tell a site manager whether or not 10 pCi/g of radium in sewer sludge is
acceptable because such a decision would require developing a specific scenario for
exposure to the sludge so that dose or risk can be calculated.  The same may be true
for more dispersed radioactivity, such as a layer of higher activity material in soil at
some depth beneath the surface.  Although it may be important to warn MARSSIM
users about such possibilities,  MARSSIM should not attempt to define methods for
every such contingency.
                                      11

-------
      In support of the Subcommittee's impression of a limited scope of coverage,
Chapter 1 of MARSSIM clearly establishes its scope as providing guidance for
contaminated surface soils and building surfaces.  In this regard, the title of the
document is misleading; it would be more appropriately called MARSSSIM—Multi-
Agency Radiological Surficial Survey and Site Investigation Manual— because it only
addresses "surficial" characterization of the top 15 cm of soils (see MARSSIM p. 2-31,
lines 663-664 defining "site boundary") and of building surfaces. The contents of
MARSSIM should be consistent with its stated scope of coverage throughout, such that
sections not clearly falling within  the scope should be omitted or moved to an appendix.
Examples include guidance on sampling surface water, ground water, sediment,
vegetation, aerosols, and subsurface soils. If these sections are left in the report in any
form,  then they raise the question as to whether or not the  DCGLs established for
surface soils and building surfaces also apply to these media. It would be useful and
appropriate to add an explanation in Chapter 1 as to why MARSSIM is limited to
surface soils and buildings, and what the user should do for other types of
contaminated media.

      MARSSIM should clarify what constitutes "surface soil" by citing a regulatory-
based definition (e.g., the appropriate section in 40 CFR Part 192).   In one section of
MARSSIM as well as the glossary, surface soil is described as being the top 15 cm
without indicating whether or not this particular depth is an arbitrary decision or based
on regulatory requirements. In addition,  MARSSIM never makes clear the extent to
which the user should be attempting to characterize contamination over this entire
depth. Numerous consequences would result if this is so.  For example, the quantity of
sample collected in some cases would be orders of magnitude larger than needed for
an analytical measurement. Should it matter that contamination limited to the top 1-cm
would possibly be diluted to below the DCGL if it were mixed to a  depth of 15 cm during
the sampling process? Should pavement samples also be collected to a depth of 15
cm? How should cobbly or skeletal soils be sampled?  The importance of
homogenizing samples before removing an aliquot for analysis should be discussed,
particularly if replicate analyses are part of the standard quality control protocol for
judging data reproducibility (see  Sections 3.5.4 and 3.5.5 of this report). It is also
critical to examine how the definition of surface soil relates to how the DCGL is
established from pathway modeling (see Section 4.2 of this report).  MARSSIM should
provide guidance to the user on the development of standardized methods for field
measurements and for sample collection and processing in order to address these
areas of concern (see Section 3.3).

3.2.5  Application to Subsurface Soil and Other Environmental Media

      In addition to more explicit  discussion of topics that are included in the present
scope of MARSSIM (see Section 3.2.4 above), the manual should also be more explicit
in an early section in its identification of topics that are  outside its scope.  Furthermore,
                                      12

-------
it should ensure that the manual's contents stay consistent with these limits.  In this
proposed section, the document should include discussion of the rationale for
excluding coverage of these specific topics and the plans, if any, to cover them in the
future, or else where to find guidance on them now.  Examples of environmental media
outside the scope of MARSSIM include subsurface soils (presumably meaning soils
more than 15 cm beneath the surface),  ground water, and buildings (beyond surface
contamination).   As presented by the MARSSIM co-authors in their discussions with
the Subcommittee, the rationale for excluding these media includes such sound
reasons as:

     a)   contamination is limited to surface soil in the majority of the sites (80-90%)
          to which the MARSSIM approach would be applied;

     b)   as a consequence of this fact, existing computer models for dose
          assessment generally only consider surface soils;

     c)   MARSSIM was written in support of cleanup rulemaking for which most of
          the supporting analyses have been limited to contaminated surface soils
          and building surfaces; and

     d)   a limited scope was necessary in order to ensure the development and
          issuance of a useful product—with multi-agency consensus—within a
          reasonable time.

We recommend that MARSSIM contain wording similar to that above in its introductory
material.

     MARSSIM should also warn the user that its application may be sufficient only in
those cases  where the surface soil contains the majority of the overall inventory  of
contamination.  If contamination of other environmental media is a possibility, then
following MARSSIM's guidance for surveying a site may be inadequate as a basis for
determining the site's suitability for release.  Consequently, the document should more
clearly state  that its domain for application is for cases where the surface soil is the
dominant source of human and ecological exposure, dose, and risk.  But even in this
case, it may  be appropriate to advise  the user that some subsurface sampling may be
required in order to prove that the  contamination was limited to the surface and that any
subsurface radiological contamination will not be deposited on the surface at some
future date (e.g., through being exposed via erosion or excavation).
                                      13

-------
3.3  Planning Guidance

3.3.1 Overall Approach to Planning the Surveys

     The Road Map provided in MARSSIM following the appendices is very useful for
understanding the overall MARSSIM approach to site characterization. It should be
moved to Chapter 2, and the reader should be strongly urged to review it carefully as
an introduction and integration of the MARSSIM process. Provision of an abstract and
Executive Summary would also help the reader understand the overall approach.

     MARSSIM guidance is unclear as to when the results from the scoping and
characterization scans or surveys may also be applicable to the final status survey.
MARSSIM specifically states that in order to use scoping and characterization data,
they must be of adequate quality to meet the DQOs. (For example, MARSSIM
references as to the applicability of characterization and scoping survey data to the
final status survey occur on page 2-25,  lines 487-490;  page 5-3, lines  88-91 and 99-
101; and page 5-10, lines 287-289). Further clarification of the issue of data
applicability, including some examples,  might be useful to the reader.  Also, as  noted
during presentations to the Subcommittee at its January 22, 1997 meeting, MARSSIM
is intended to apply to the final site status survey and not to scoping or characterization
surveys undertaken for the specific purpose of planning remedial action.  The latter
objective could involve different guidelines for sampling strategies and analyses. This
fact should be strongly emphasized in the discussion of MARSSIM's scope.

3.3.2 Public Involvement

     The MARSSIM recommends including stakeholder group representatives on the
planning team. Thus, the public will be involved  in the early planning stages of any
initial site evaluations and surveys. How the planning team would be constituted and
how it would arrive at consensus on these and other survey design issues are less
clear, especially the role of the public stakeholders. In addition, several key aspects of
the MARSSIM process are not straightforward and require thoughtful evaluation by the
planning team of the various alternatives and their implications for the release decision.
For example,  the document discusses the potential difficulties in identifying suitable
areas for establishing background radiological conditions against which the site's
conditions will be compared.  It also recognizes that use of default DCGLs and
standardized values for Type I and Type II decision errors (a and p, respectively) may
not be appropriate for all sites and consequently allows for comparisons with
site-specific DCGLs using decision criteria selected by the planning team.

     Although the Subcommittee understands that such issues cannot be fully
discussed in a technical document like MARSSIM, they are important for the public
acceptability of its  results. We therefore recommend that the document acknowledge
                                      14

-------
the issue of public participation and, to the extent possible, reference activities within
the organizations that deal with public involvement in the design and implementation of
radiation surveys for site release decisions (e.g., in its section 3.2).
                                        15

-------
               4. DATA ACQUISITION AND ASSESSMENT


4.1   Response to Charge b)

     The charge from ORIA's director regarding data acquisition and assessment was:

     Charge b) Are the methods and assumptions for demonstrating compliance
     with a dose- or risk-based regulation technically acceptable?

     Measurement methods applicable to and appropriate for use in demonstrating
compliance with the DCGLs are discussed in Chapters 6 and 7 and Appendix H of
MARSSIM.  Chapter 6 describes  methods and instrumentation for collecting data in the
field, i.e., direct measurement of radionuclides in surface soils and on structure
surfaces and scanning surveys. Chapter 7 addresses issues related to sampling
methods and laboratory instrumentation used for analysis of the samples.  Appendix H
is a comprehensive compilation of descriptions of instruments available for use in
demonstrating compliance with the DCGLs. Chapter 9 also deals with an integral part
of data acquisition in that it covers quality assurance and quality control measures.

     The Subcommittee finds the description of data acquisition methods to be
technically acceptable and concludes that MARSSIM is a very good compilation of
methods and instruments and will be a useful tool for regulators and the regulated
community. However, treatment of field and laboratory operations in MARSSIM could
be improved in the following areas:

     a)   In MARSSIM Section 6.4.1, Direct Measurement Sensitivity, there appears
          to be an inconsistency between the presentation (and terminology) of
          Figure 6.2 and the subsequent (pp 6-20 and 6-21) derivation and
          definitions of the critical level Lc and the a priori detection level LD. The
          classical approach developed by L. Currie (Currie, 1968) for these two
          parameters used figures and derivations based on a net signal response
          (us = 0) distribution together with the standard deviation (o0) of the net
          signal  response distribution. It is recommended that Section 6.4.1 be
          revised to show consistency between the included figure and the derivation
          of the subsequent instrument detection parameters.

     b)   Sample collection protocols and analytical techniques are discussed in
          MARSSIM Chapters 6 and 7 and in Appendix H.  The redundancy in
          coverage  between Chapters 6 and 7 should be minimized by careful editing
          and  cross-referencing, by redefining the scope of each chapter, or by
          combining the two chapters into a single one, resulting in a more concise,


                                     16

-------
           user-friendly, and internally consistent document. Standardized
           nomenclature should be used throughout MARSSIM in references to field
           and laboratory equipment in order to avoid confusion or ambiguity.

     c)    In general, MARSSIM contains technically sound descriptions of field
           measurement methods, instruments, and operating procedures.  Some
           additions, clarifications, and corrections are noted in our report.  For
           example, some sampling methods described in MARSSIM are incomplete.
           MARSSIM should provide guidance for the development of standardized
           sampling procedures for surface soils.  Standard procedures should be
           referenced, such as ASTM C998-83, "Standard Method for Sampling
           Surface Soil for Radionuclides," and ASTM C999-83, "Standard Method for
           Soil Sample Preparation for the Determination of Radionuclides." However,
           the user should also be encouraged to design site-specific surface-soil
           sampling procedures in order to ensure representative samples of the
           surface materials that are of concern in deriving the DCGLs for the site.
           MARSSIM should provide a list of considerations for designing these site-
           specific surface soil sampling methods. For example, the depth to which a
           sample is taken may affect the measured concentration if the radionuclide
           is deposited in the top few centimeters. Under some circumstances,
           averaging over the top 15 cm is appropriate if the exposure pathway of
           concern is ingestion of food raised in the area, but may underestimate the
           potential dose if the exposure pathway of concern is soil ingestion or
           inhalation of resuspended dust.  Other considerations might include the
           size fraction to be collected and/or analyzed and whether vegetation, large
           gravels or cobbles, or debris  should be removed in the field or prior to
           laboratory analysis.

     d)    Descriptions of the selection and operation of radiation detection
           instruments for laboratory analyses  are technically sound and represent
           standard practice but may not be state-of-the-art. MARSSIM should
           standardize the level of detail used in  its presentation of this material and
           should also provide information on the planned scope and current status of
           plans to prepare a manual on Multi-Agency Radiological Laboratory
           Analytical Protocols (MARLAP), which may be a more appropriate forum in
           which to provide more thorough in-depth guidance to the user on the
           selection and operation of laboratory instrumentation.

     Specific findings and recommendations of the Subcommittee with regard to these
aspects of MARSSIM are given in the following sections.
                                      17

-------
4.2  Organization of Information on Analytical Instruments in MARSSIM

     Sample collection protocols and analytical techniques are discussed in
MARSSIM Chapters 6 and 7 and in Appendix H.  The distinction between the scopes of
Chapters 6 and 7 is not altogether clear, and there is some overlap between the two,
particularly with respect to radon analyses (MARSSIM Sections 6.6 and 7.4.5).  The
excessive redundancy in coverage could be minimized by careful editing and cross-
referencing, by redefining the scope of each chapter, or by combining the two chapters
into a single one, resulting in a more concise, user-friendly, and internally-consistent
document. An example of the redundancy can be found in sections on radon
measurements (MARSSIM Sections 6.6.1 and 7.4.5), and other examples are
mentioned in detailed comments submitted to ORIA  separately from this report
(Kooyoomjian, 1997).

     Standardized nomenclature should be used throughout MARSSIM  in references
to field and laboratory equipment in order to avoid confusion or ambiguity.  All
techniques mentioned in the main body of MARSSIM should have corresponding
detailed descriptions in Appendix H, but it is difficult  to check the extent to which this
recommendation has already been met because of the different terminologies that have
been used (MARSSIM Tables 6.1 to 6.3, Chapter 6 and 7 text, Table 7.2, Appendix H).

     The level of detail provided for the various techniques is uneven, ranging from no
discussion (e.g., low-energy x-rays are mentioned in MARSSIM Chapter 6, line 103, but
are not subsequently discussed as have been gamma, alpha, and beta radiation) to
overly detailed descriptions (e.g., it is inappropriate in MARSSIM Chapter 6 to discuss
the energy response of the PIC (line 149); the need  for a site-specific calibration curve
for Nal(TI) detectors (line 152); or n and p type configurations (line 391)).

4.3  Guidance on Data Acquisition

4.3.1 Scanning Surveys

     Statements should be made in MARSSIM Section 6.4.2.1 (Scanning for Beta and
Gamma Emitters) and in Section 6.4.2.2 (Scanning for Alpha Emitters) as to the
applicability of a "scanning survey" as a "final status survey." Scanning surveys are
performed to locate radiation  levels or radioactivity above an investigation level (i.e.,
"hot spots").  The terminology "scanning survey" may be confused with the MARSSIM
terminology of "scoping survey", "characterization survey," and "final status survey."
Maybe the term "instrument scan" rather than "scanning survey" would be a better
choice  of words since the former is then similar to the terminology "direct
measurement." Both types of measurements (direct and instrument scan) can be used
to characterize the site (MARSSIM Sections 6.2 and 6.2.2).
                                      18

-------
     The statistical treatment of "human factors" in scanning field measurements is
inadequate. The section focused on the "Poisson Observer" (p. 6-24, line 221) is
technically weak. In particular, the concept is discussed to a far greater extent than
appears suitable without adequate documentation and references.  The problem should
be stated in one or two paragraphs, and a human factors efficiency should be
proposed. Information concerning the magnitude of this value and the basis of
Equation 6-6 should be referenced. Other approaches for dealing with loss of
information during a scanning survey  may include analogy to manner in which one
deals with the loss of information by an instrument as it moves from one location to
another, e.g., by time constant or coincidence counting by paired monitors.

4.3.2 Instrument Calibration

     MARSSIM should include a  recommendation that calibrations of survey and
laboratory instruments be linked to a national standard traceable to the National
Institute of Standards and Technology (NIST).  Such linkage can be established
through a NIST secondary laboratory  accreditation program for survey meters (e.g.,
Eisenhower, 1991), or a program sponsored by the Health Physics Society (HPS, no
date), or through one of the national measurement assurance programs (MAP) that is
linked to NIST for laboratory analyses (e.g., ANSI, 1996).

4.3.3 Background  Measurements and Adjustments to Measured Values

     A clear distinction needs to be made among instrumental background, method or
process blank background,  field blank, and environmental background wherever the
term "background" is used.  For example, the "s+b" subscript for Equation 6-1 is
confusing because this summation refers to total count rate in just the survey area and
has nothing to do with the count rate in the reference (background) area, while the
subscript "b" in the same equation refers specifically to the count rate in the reference
area.  (MARSSIM Sections  6.2.5, 6.2.7.1, 9.3.4)

     The described approach for  the use of surrogates should not be presented as
one of deriving a correction factor, but rather an application of a surrogate indicator for
the radionuclide of interest.  A more realistic example (such as Co-60 and Ni-63 from
MARSSIM's Section 4.3.2)  should be provided for illustrating this approach (in
MARSSIM Section 6.2.7.1)  because it does not make sense to expect Co-60  and H-3 to
have or to maintain relatively fixed ratios due to their dramatically different chemical
behaviors. In addition, guidance should be provided on methods to be used to
characterize the fixed ratio and its variability (e.g., number of samples needed).

     The derivation of the alpha-scanning equation is provided in Appendix J of
MARSSIM.  These equations and  calculations have been verified by the Subcommittee.
However, the basis for the average number of counts expected is confusing.  In
                                      19

-------
Equations 6-7 (page 6-34) and J-3 (page J-2), the average number of counts expected
is represented by the term "G E d / 60 v."  Although the source activity (G) is defined as
the number of decays per minute (dpm) measured over the effective detector area as
calculated by the equation on line 48 of Appendix J, the efficiency (E) is not specified
as applying to a particular area.  The parameter E, which is embedded in the definition
of G, should likewise be determined for activity distributed over the detector area.

     A related concern is that the derivation of Equation J-5 seems to assume that the
dwell time for the entire area of the detector will be the same. This works for
rectangular or square detectors but not for circular ones.  If the user assumes that the
width of the detector in the direction of the scan is equal to the diameter of a circular
detector,  the true probability of getting a single count at a given level of contamination
for a specified scan time will be overestimated.

4.4  Field and Laboratory Instrumentation

4.4.1 MARSSIM's Overview of Instrumentation

     Appendix H in MARSSIM provides a useful and  informative listing of alternative
instrumental techniques for radiation surveys,  presented in the format of summary
tables and more detailed thumbnail sketches.  This appendix is likely to become a
frequently consulted reference for decision-makers and planners interested in the
advantages and disadvantages of the different techniques. The following suggestions
are made to improve the usability of MARSSIM's Appendix H for the intended audience.
Subcommittee comments submitted separately to ORIA provide details on specific
technical and editorial corrections (Kooyoomjian, 1997).

     a)    Add a description in the  introduction of the appendix that describes each
           heading used in the detailed descriptions.  For example, what factors
           determine whether or not an instrument is used in the laboratory or field?
           What does "Secondary Radiation Detected" mean; what is the basis for
           specifying primary versus secondary? What is the basis of the cost
           estimates per measurement (e.g.,  typical quotes from commercial
           laboratories, vs. inhouse labor rates)? A  range of estimates for  all costs
           would be more appropriate than the single value  that is given for some of
           the estimates.

     b)    The introduction to the appendix would be the place to warn the reader that
           analytical laboratories need to be contacted for actual cost schedules, and
           that costs are highly variable depending upon the matrix, turnaround time,
           sample preparation requirements, the potential for cost savings for large
           batches, required detection limit, and required level of documentation of
           quality assurance (QA) or quality control (QC) measures.
                                      20

-------
c)   Consider adding a new heading category, "Advantages and
     disadvantages," for the detailed description of each instrumental technique.
     Such a discussion should be worded in such a way as not to construe
     endorsement of a particular manufacturer's product.

d)   The detailed descriptions and summary tables could be better organized to
     provide a more useful thumbnail guide to decision-makers who are
     planning a survey. The present approach of having multiple sections and
     multiple summary tables with duplicative information and an uneven level of
     detail results in much flipping through the pages.  A more convenient
     organization of this information would be to combine these into a single
     section and a single table, with instruments simply listed alphabetically, and
     to revise the summary table column headings  to focus on those aspects of
     greatest interest to the survey planner (an example table has been included
     in the comments transmitted under separate cover to ORIA). The column
     labeled 'Remarks' should be omitted from the  summary table because this
     level of detail is more appropriately discussed in the detailed descriptions.

e)   Add descriptions for the additional methods listed on page H-55
     (fluorimetry, passivated ion implanted detectors, Cerenkov counter,
     PERALS scintillation counter, Cd-Zn-Te)

f)    There is no specific  listing of equipment in the summary tables for
     measuring low-energy X-radiation, although this application is mentioned
     on page H-19.

g)   General inconsistencies  noted between table  entries and the detailed
     descriptions should be corrected.  These include differences in system
     names, systems missing from the tables in which they belong, and
     differences in cost estimates.

h)   There is a lack of correspondence between the equipment mentioned in
     Chapters 6 and 7 with those in Appendix H. It is recommended that there
     be correspondence between the equipment mentioned in Chapters 6 and 7,
     the sheets in Appendix H, and Tables H.1 through H.5. Each sheet in
     Appendix H should be cross-referenced from the mentions of the equipment
     in the main body of the text.
                                 21

-------
4.4.2 Additional Measurement Techniques

     Detailed descriptions should be added of the Frisch-Grid alpha detector for both
field and laboratory use, and of the Pulsed-Laser Phosphorimetry (also known as
Kinetic Phosphorescence Analysis, or KPA) for analysis of total uranium concentrations
in the laboratory.

     Appendix H should also address the advantages of isotopic analyses of some
samples.  In some cases, determination of isotopic ratios may be useful or even
necessary to distinguish contamination from background, e.g., if the nuclide of concern
is uranium and the contaminant is either depleted or enriched uranium with an isotopic
signature readily distinguishable from that of natural uranium.

4.4.3 Relationship Between MARSSIM and MARLAP

     The RAC was informed that ORIA is also currently involved in preparation of
another multi-agency manual, the Multi-Agency Radiological Laboratory Analytical
Protocols (MARLAP).   In view of its expected overlap with  MARSSIM's scope of
coverage (especially Chapters 6, 7, and 9 and Appendix H), MARSSIM should discuss
the relationship between MARSSIM and MARLAP and provide a short description of
the scope of MARLAP in Chapter 1.

4.5  Quality Assurance and Quality Control

4.5.1 Data Quality Objective (DQO) Process as Related to Measurement Systems

     Some danger exists that the Quality Assurance Program Plan (QAPP) will be too
prescriptive and bureaucratic for a reasoned decision. The guidance given  in
MARSSIM Chapter 9 seems very specific and detailed, which may hinder development
of reasonable site-specific plans. Hence, the document should encourage modification
of the sample QAPP to fit site-specific conditions.

     MARSSIM provides excellent guidance to showing how the Data Quality
Objective (DQO) process is used to determine the number of measurements that must
be performed—implicit in the selection of kg and kp probabilities (for detecting Type I
and II errors). An equivalent level of detailed guidance should be provided for the
establishment of specific criteria (e.g., bias and precision) for determining the quality or
acceptability of analytical data that link the DCGL, measurement detection level, and
analyte concentration or surface activity.  This recommendation is consistent with
guidance presented in MARSSIM Section 9.2.3, which states "The type and quality of
environmental data needed for each project should be defined and documented using
the DQO Process." For example, for cases  in which the radionuclide of concern is  not
present in the background and in which the DCGL is one or two orders of magnitude


                                     22

-------
higher than the analyte concentration and detection capability of the instruments, the
data quality objectives in terms of bias and precision may be less stringent than when
the analyte concentration is near the DCGL. In the latter case, much more restrictive
bias and precision criteria would be required because a measurement error of a factor
of two in analyte concentration could overlap the DCGL.

4.5.2 National Quality Assurance Standards

     Industry guidance on the acceptable practices of a radioassay laboratory can be
found in ANSI (1996). A description of several available government and industrial
(interlaboratory) measurement assurance programs (MAPs) is presented in Appendix B
of this ANSI standard. The ANSI standard should be mentioned during the discussion
on the review of a laboratory's qualifications (MARSSIM, page 7-3, lines 77 through
114).

4.5.3 Data Verification and Validation

     In the introductory material presented in Section 2, the term "Quality Assurance
Project Plan (QAPP)" is used. More extensive discussion on the QAPP process is
presented in Section 9.  QAPP is an EPA term and concept dating to 1980 (U.S. EPA,
1980).  Within the EPA QAPP guidance,  detailed data verification and validation (Data
V & V) specifications are delineated and  required for data assessment. Although
mentioned in Section 9.2.4 and Appendix E, MARSSIM  does not discuss in any detail
the Data V & V requirements or the extensive measurement documentation associated
with the Data V & V process.  (MARSSIM Chapter 9, line 114, refers the reader to
Chapter 8, but there is no discussion of the topic in Chapter 8.)  Furthermore, the
glossary definitions of these terms are unintelligible for the  average reader.

     In the Introductory Section 9.1, it is stated that "The QAPP is  a formal document
describing in comprehensive detail the necessary QA, QC,  and other technical
activities that should be  implemented to ensure that the results satisfy the stated
objectives and to produce  legally defensible data (EPA  1994c)." A writing group of
ANSI was formed in 1995 to develop a consensus industry  standard that addresses
Radiation Data V & V.  If MARSSIM is  going to be a living document with routine
updates, guidance from  this ANSI standard can be referenced in the future.

     Some of the Data V & V process has been incorporated into Section 9.4 but in a
different format and intertwined throughout the six subsections of this section.  Data V
& V terminology and its data review cycle should be used explicitly and transparently in
this section. MARSSIM should clarify whether the guidance it provides is sufficient to
make the measurement documentation legally defensible in the absence of a formal
Data V & V process.
                                      23

-------
4.5.4 Quality Control Samples and Measurements

     Section 9.3, which deals with quality control (QC) samples and measurements,
should be better organized to delineate between field QC measurements and samples,
and laboratory QC samples. At present, due to the attempt to define and provide
guidance on the types of QC samples, the reader may be confused as to the type and
number of QC samples that are needed for field applications (or for an external
laboratory QC program) compared to the requirements for an internal QC program at
the laboratory.

     Well-characterized performance evaluation samples are QC samples used by the
field survey contractor to evaluate the performance of the laboratory to meet
performance criteria for measurement quality as specified in the contractor's Statement
of Work, e.g., detection limits, precision and bias specifications. These can be matrix
spikes (including natural matrix spikes), duplicate samples (well-prepared split
samples), or blanks (matrix sample without analyte).  These samples are submitted as
either double-or single-blind QC samples along with the field samples.  Typically, the
frequency with which performance evaluation samples are submitted is one per batch
or one per twenty samples. The unique sample identifier for each performance
evaluation sample is cross referenced or indexed with the samples in the same batch.

4.5.5 Use of Spikes

     During the past five years, the operational QC programs for many national
laboratory and uranium mining/milling site remediation projects  have used  site-specific
natural matrix soil QC materials to ensure proper laboratory processing or  gauge the
quality of the very difficult radiochemical analyses. These natural matrix QC samples
are used  instead of matrix spikes (MARSSIM Section 9.3.2.1) wherein a soil matrix is
spiked with  an analyte as an ionic species in solution. In many cases the spiked matrix
sample will  be an ineffective QC material for monitoring certain  laboratory
radiochemical processes if the analyte is incorporated into the soil matrix of a survey
sample rather than surficially adsorbed. Since the analyte in the natural matrix QC
material is in the same chemical/physical form as the survey samples, these QC
samples would be more effective in monitoring the laboratory's capabilities for the
proper radioassay of the survey samples.

     For natural matrix QC materials, a large amount of soil is collected from an area
of known contamination and processed for the purpose of making laboratory
performance evaluation (PE) material or samples.  The contaminated soil  material is
homogenized by blending, or in some cases is dried, pulverized to micron size, and
then blended. The blended soil is then split uniformly into many samples of a given
weight. A statistically determined  number of samples is selected to characterize the
material for analyte concentration  and other parameters. In some cases, several
                                      24

-------
analyte concentration levels of the PE material can be prepared by blending a higher
activity soil PE material with a blank soil material that has gone through the same
preparation process and has the same particle-size distribution. The acceptable
degree of sample analyte homogeneity and analyte distribution in the PE material for
use in applying MARSSIM would be related to several factors: the DQOs of the project,
the differential between the DCGL, measurement detection limit and expected residual
analyte concentration at the survey site, and laboratory quality performance
specifications. These PE samples are considered as true "natural matrix" spikes. The
PE samples are included as double or single blind matrix spikes with each batch of
samples submitted to the laboratory.  This practice is consistent with the "Monitoring
Laboratory" concept presented in ANSI (1996).

4.5.6 Data Quality Indicators

     Under the current EPA Data V & V process, data qualifiers are assigned to each
datum after the data package deliverables have been verified for compliance with the
contractor's Statement of Work and validated against qualitative and quantitative
criteria.  If MARSSIM suggests similar guidance, it is recommended that the Data V & V
process and the final assignment of the data qualifiers be linked to reasonable data
quality objectives that reflect the  relationship between the DCGL, measurement
detection limit, and the analyte concentration or surface activity (MARSSIM Section
9.4.6).
                                      25

-------
                 5. DEMONSTRATION OF COMPLIANCE


5.1   Response to Charge c)

     The charge from ORIA's director regarding statistical and related issues was:

     Charge c) Are the methods and assumptions for demonstrating compliance
     with a dose- or risk-based regulation technically acceptable? (b) Are the
     hypothesis and statistical tests and their method of application appropriate?

     MARSSIM begins with the assumption that appropriate Derived Concentration
Guideline Levels (DCGLs) have been defined that meet the dose- or risk-based release
criteria for contamination of surface soil and building surfaces by radionuclides. It then
proposes methods for dividing a site into survey units, surveying and sampling those
units to develop distributions of radionuclide concentrations, and comparing the results
to the DCGLs to decide whether the units pass the release criteria.

     The Subcommittee made the following overall findings on this approach and the
statistical issues involved in carrying it out:

     a)   Although the proposed approach is generally appropriate for demonstrating
          compliance with a dose- or risk-based regulation, it must be carefully
          applied to assure that it provides a technically acceptable solution to the
          intent of the regulations.  While the designers of the MARSSIM are clearly
          aware of this issue and have prepared a document that allows the user to
          conduct an appropriate survey and analysis, more can  be done to assure
          that they do so. In particular, MARSSIM allows the user to make
          appropriate choices for the hypothesis to be tested with the results of the
          survey and for the methods of conducting the statistical tests, but does not
          always provide sufficient guidance to make the best choice.

     b)   Although MARSSIM is applicable to the majority of contaminated sites,
          there appear to be cases that MARSSIM, as currently written, would have
          trouble addressing. These include: 1) cases dealing with the release of
          sites that had been contaminated with naturally occurring radionuclides and
          in which the DCGL is contained within the ambient (background) analyte
          variability, and 2) cases in which a reference background cannot be
          established. The Subcommittee recommends that future revisions of
          MARSSIM provide guidance to the user regarding appropriate choices
          when such conditions are encountered.  For example, the null hypothesis
          might be redefined to be that the distribution of site radioactivity is no
          different from the reference site or than ambient radioactivity in general.


                                      26

-------
     c)    In MARS SIM, the preferred null hypothesis is that a survey unit is not ready
           for release and the information gathered must be sufficient, with a high
           degree of confidence, to accept the alternative hypothesis (i.e., that the unit
           meets the release criteria). Furthermore, MARSSIM discusses in detail two
           non-parametric procedures, the Wilcoxon Rank-Sum test and the Sign test,
           for testing this hypothesis. However, MARSSIM allows more flexibility in
           defining the null hypothesis and in choosing statistical analysis methods to
           test that hypothesis than may be readily apparent to most readers. This
           latitude needs to be more  clearly stated and the criteria for selecting among
           potentially applicable tests need to be described.

     In the sections that follow, the Subcommittee  presents its detailed findings and
recommendations regarding the above concerns. Additional comments have been
transmitted under separate cover to ORIA.

5.2  Importance of Appropriate DCGLs

     The MARSSIM team has decided that the best method of determining compliance
with a dose- or risk-based standard is to define DCGLs—concentrations in soil or on
surfaces that are unlikely to cause unacceptable doses and risks—and then test a
survey area to see if those DCGLs are met. The Subcommittee notes that other
approaches could have been proposed (e.g., using  the measured concentrations to
calculate dose or risk and then comparing the calculated value with the dose or risk
criterion) but does not object to the method proposed. However, the Subcommittee
believes that it is critically important for the assumptions and procedures used in
MARSSIM to make comparisons with the DCGLs match those used in defining the
DCGLs.  For example, if a DCGL for soil is derived from a risk criterion by  assuming
that a receptor ranges over a certain area on a random basis, then the  same area
should be used for spatial averaging in the MARSSIM statistical analyses.  Such
averaging is usually performed from the standpoint  of potential human receptors.  The
manuscript should note that different spatial and temporal scales of averaging will be
necessary if dose- and risk-based criteria are applied to components of the ecosystem
other than humans for derivation of a DCGL.

     The three main steps in the MARSSIM process are: 1) determining the DCGLs
from an agreed-upon release dose  limit; 2) making accurate measurements of site
contamination; and 3) determining whether the sampling data prove or disprove the
hypothesis that the site contamination meets the release limits. (The Subcommittee
recommends that in MARSSIM on page 1-3, lines 39-41, this process should be stated
more neutrally as just done.) Nowhere, in the opinion of the Subcommittee, is the
critical importance of the  DCGLs given.

     Furthermore, the document does not give any indication of how difficult the


                                     27

-------
DCGLs are to derive nor what the likely range of uncertainty would be under the most
likely conditions.  More prominent discussion should be given to the uncertainty
associated with the derivation of the DCGLs.  Often the DCGL is obtained using a
computer model such as RESRAD (Yu et al., 1993) to back-calculate a value from a
predefined risk or dose limit. There is considerable uncertainty associated with the
results of this calculation. The effects of this uncertainty should be discussed.  It is
likely that the uncertainty associated with the mathematical derivation of a
concentration guideline will greatly exceed variability of data from field samples.
Information about the uncertainty in DCGLs is necessary so that the discussion of
"Measure" and "Decide" would provide some idea of the accuracy needed in the
derivation of DCGLs to be similar to that in the "Translate" portion of the study
(MARSSIM, page 1-3, lines 33-41, and page 1-2, Figure 1.1).   It would be useful to also
give some idea of the techniques and models that can be used to determine the
"Translate" portion of the study. Examples might include such models as MEPAS and
RESRAD or the critique of the methodology by EPA or the NAS/NRC (Bucketal., 1995;
Yuetal., 1993).

     The use of  any model to derive the  DCGLs will induce uncertainty about how it
relates to the dose or risk goal. Contrary to what is stated on  MARSSIM page D-22,
lines 514-527, this uncertainty can (and should) be quantified. Currently available
models for characterizing sites with radioactive materials will not necessarily produce
conservative estimates of risk unless conservative parameter choices are employed.
DCGLs might best be derived in an iterative process, first using conservative screening
models and parameter values and then, if the cost of sampling is prohibitively large,
using site-specific models, parameter values, and uncertainty estimates to refine the
uncertainty estimates for the DCGLs.  Procedures for doing so can be found in NCRP
Commentary No.  14 (NCRP, 1996).

     The lack of specificity as to how DCGLs will be derived  limits the reviewer's
ability to comment on the propriety of the  methods proposed in MARSSIM, because the
assumptions and procedures used there must be consistent with those used in
MARSSIM to ensure that the latter's methods are appropriate. For example, if the
DCGL is based on uniform distribution in  a semi-infinite slab,  then measurement only in
the top 15 cm may lead to incorrect release decisions (MARSSIM, page 1-4).

     The issue of the derivation of the wide-area DCGL, DCGL^,, is critical.  The
DCGLW will change depending on the  assumptions used to define the appropriate area
for averaging of contamination to assess  exposures to human and/or ecological
receptors. The uncertainty  in the derivation of the DCGLW may well exceed one order
of magnitude. Consistency in the values  of DCGLW is likely to change dramatically from
site to site depending upon level of stakeholder involvement and the level to which
stakeholders are  risk-tolerant versus risk-averse.
                                      28

-------
     The survey objectives seem to include the need to find radionuclides even in
areas where they do not currently pose much risk (e.g., alpha-emitters under paint;
beta-emitters behind walls; underground pockets of material). The DCGLs appear to
be designed for materials that are readily available on wall surfaces and in surface
soils. The possibility for a mismatch between the DCGL models and the MARSSIM
survey  methods seems significant.  How would more appropriate scenarios for
developing site-specific DCGLs be defined? (Pages 4-5, 4-21, 4-22).

     If the maximum areas for the survey units are of the order of a large room or a
medium-sized residential lot for building surface and soil contamination, respectively,
are these survey areas consistent with assumptions underlying the calculations of
DCGLs from the cleanup standard? For example, do the DCGLs always assume an
infinite  horizontal planar source?  Would the DCGL be different if there were only one
survey  unit than it would be if there were several contiguous survey areas (MARSSIM
page 4-13)

5.3  Relationship Between DCGL and Reasonable Measurement Requirements

     There appear to be cases that MARSSIM, as currently written, would have
trouble addressing. These cases include: a) cases dealing with the release of sites
having  been contaminated with naturally occurring radionuclides and in which the
DCGL  is nested in the ambient (background) analyte variability,  and b) cases where no
reference background for certain  or old construction materials can be established. The
Subcommittee recommends that future revisions of MARSSIM provide guidance to the
user regarding appropriate choices when such  conditions are encountered.  For
example, the null hypothesis might be redefined to be that the distribution of site
radioactivity  is no different from that at the reference site or than ambient radioactivity
in general.

     At the January 23, 1997 MARSSIM's SAB meeting, it was explained that for
certain  cases where MARSSIM may be difficult to apply, site surveys may be performed
based on the "As Low As Reasonably Achievable (ALARA)" concept with concurrence
of the regulator. The inference from such a philosophy is to set the Type I and Type II
decision errors so that the probability of releasing the site is consistent with public
policy goals. As stated in MARSSIM Section 2.5.4, "MARSSIM does not recommend
values  for any of these parameters (Type I or II decision errors, os, oror A), although
some guidance is provided. A prospective power curve (see MARSSIM Appendix D)
that considers the effects of these parameters can be very helpful in designing a survey
and alternative values for these parameters, and is highly recommended." This ALARA
concept should be presented in MARSSIM Sections 1 (page 1-3, line 43; page 1-4, line
67) and 2 (Section 2.5.4).  If a survey plan using the ALARA approach is acceptable
for such cases, then a specifically designed ALARA example of such an application
should  be included within the document.
                                     29

-------
5.4  Determination and Use of "Background" Measurements

     MARSSIM provides good guidance on how to obtain background measurements
from surveys of reference areas and how to use such measurements in the statistical
analysis of results from a survey unit in order to determine its acceptability for release.
How the reference areas are to be defined, located, and defended is less clear. The
revised document should provide more guidance on the following issues:

     a)    If radionuclides are present in a survey unit as a result of human activities
           not related to the site in question, are they counted as background or site-
           related? (Pages 2-27, 4-11,5-11)

     b)    MARSSIM does not make it clear whether background reference areas
           must always be on site, must always be off-site, or may be either. (Page 4-
           10)

     c)    It should be noted that representative reference areas cannot be found for
           many disturbed sites, such as former landfills.  MARSSIM should provide at
           least a reference to identify acceptable characterization approaches for
           such sites.

     d)    MARSSIM should discuss whether subsurface background measurements
           should be taken at a "reference site". If so, MARSSIM, or the approved
           subsurface reference manual, should discuss how subsurface background
           measurements should be taken at a "reference site".

     e)    MARSSIM specifies that background reference areas must be similar to the
           survey areas with respect to radiological properties (MARSSIM  pages 4-11
           through 4-14).  At the outset,  this requirement seems to be an oxymoron.
           The document needs to provide more explanation regarding what is
           intended here.

     f)     Especially if a hostile party is on the decision team, agreement on what
           constitutes an acceptable background reference area may be difficult to
           achieve. For example, in the risk assessment  performed for the cleanup of
           the Cotter uranium mill in Canon City, Colorado (ENVIRON Corp., 1991),
           definition of background was  a key issue that limited the ability to make a
           widely shared decision. (Pages 4-2, 4-11)

     MARSSIM defines "background radiation" to include fallout from  nuclear weapons
testing and  nuclear power plant incidents in addition to radiation arising from natural
sources.  The Subcommittee notes that such a broad  definition is widely—but not
universally—accepted and the most common baseline used for evaluating levels of
                                      30

-------
radioactive contamination.  However, other guidance provided by various Federal
agencies,  including the EPA, restricts background radiation to include only natural
sources, which may introduce some ambiguity for MARSSIM users (e.g., U.S. DOE,
1980; U.S. EPA, 1996; Mettler et al., 1990; U.S. Enrichment Corporation, 1993).  The
Subcommittee recommends that MARSSIM note the prevalence of other definitions of
background, and that it discuss the scientific basis for its own broad definition.
Wording similar to that used by the NRC in its recent decommissioning rule would be
appropriate,  in which the NRC defined background as follows (U.S. NRC, 1997):

     "Background radiation means radiation from cosmic sources; naturally occurring
     radioactive material, including radon (except as  a decay product of source or
     special nuclear material); and global fallout  as it exists in the environment from
     the testing of nuclear explosive devices or from past nuclear accidents such as
     Chernobyl that contribute to background radiation and are not under the control
     of the licensee.  'Background radiation' does not include radiation from source,
     byproduct, or special nuclear materials regulated by the Commission."

Defending this broad definition in its response to comments, the NRC stated that (NRC,
1997):

      "the Commission continues to believe that the inclusion in background of global
     fallout from weapons testing and accidents such as Chernobyl  is appropriate.  No
     compelling reason was presented that would indicate  that remediation should
     include material over that the licensee has no control  and that is present at
     comparable  levels in the environment both on and offsite."

5.5  Statement of the Null Hypothesis and Statistical Tests

5.5.1 Statement of the Null Hypothesis

     In MARSSIM, the preferred null hypothesis  is that a survey unit is not ready for
release. The information gathered must be sufficient, with a high degree of confidence,
to accept the alternative hypothesis (i.e., that the  unit meets the release criteria). The
Subcommittee supports MARSSIM's choice of this particular null hypothesis because it
minimizes  the potential for release of a survey unit for which insufficient information  has
been generated.  Because the cost of surveying and sampling a site  is
usually—although not always—small in comparison to the costs of error in site
classification, it is reasonable to adopt a strategy  that encourages more rather than
less investigation.

     However, the Subcommittee can also support the use of a null  hypothesis that
presumes  that a survey unit is no different from a reference unit where only background
levels of radionuclides are present.  With that hypothesis, the survey can be designed
                                      31

-------
to reject the site for release only if the data show with a reasonable degree of
confidence that residual radiological contamination at the site exceeds background
levels plus the release criterion.  Based on its conversations with the MARSSIM team,
the Subcommittee believes that the document allows this type of hypothesis structure
as a valid option  if the site survey planning team agrees that it is a better approach for
the specific situation. However,  the document is less than explicit about this possibility,
and it deserves more prominence, with appropriate cautions regarding its applicability.

5.5.2 Selecting Appropriate Statistical Tests

     As with the selection of a hypothesis structure, MARSSIM also allows more
flexibility in choosing statistical analysis methods than may be readily apparent to most
readers.  The document can easily be misinterpreted to confine the universe of
permissible statistical tests to two non-parametric procedures, the Wilcoxon Rank-Sum
test and the Sign test.  Through more careful reading and conversations with the
MARSSIM team, the Subcommittee was able  to determine that other statistical tests
could be more appropriate in specific situations and that the MARSSIM user was not
restricted to the two specified tests. This latitude also needs to be more clearly stated,
and the criteria for selecting among potentially applicable tests need to be described.
Specifically, the non-parametric tests are designed to distinguish between distributions
of concentrations that have different medians, and so are most applicable to
reasonably symmetric distributions in which the median is likely to be a good
approximation of the mean. This feature is important because the methods that relate
dose or risk to concentration in the derivation  of DCGLs use the average or mean
concentration, not the median. Therefore, decidedly asymmetric distributions are
probably better analyzed with methods other than the two featured in MARSSIM.

      In the following sections, the Subcommittee details its comments and
recommendations on the specified non-parametric tests, offers some suggestions
regarding alternative methods, and expands on the issue of mean vs. median in
statistical testing. Additional comments are also provided regarding three statistical
design issues-the definition of the gray region, specification of sample size, and
treatment of outliers.

5.5.3 Specified Non-Parametric Methods

     While many users may not like their guidance to focus on statistics, MARSSIM
may go a little too far in the direction  of simplification.  A reader may be led to believe
that the Wilcoxon Rank Sum (WRS) test and the Sign test are all that one needs to do
credible environmental statistics. These tests are useful tools, but hardly the last word
in characterization statistics, and they will not  perform well with markedly asymmetric
(skewed) distributions.
                                       32

-------
     In some cases of severely skewed distributions, the arithmetic mean can equal
the 90th percentile of the distribution. In such a case, a sample of size 10 has about a
35% probability of having all observations below the true population mean. Thus one
has an a priori 35% chance of declaring that a site whose mean slightly exceeds the
DCGL is clean.  Discussions with the MARSSIM team revealed the assumption that, for
Class 1  areas, the need to detect areas of elevated concentration ("hot spots") would
alleviate this problem. However, neither Table 5.8 nor Roadmap-9 states that hot spot
detection must take place.  If real hot spots exist, they will probably be detected on the
preliminary scan, and the need to characterize these hot spots will increase the sample
size, but how and when? The rationale for this assertion needs to be articulated better.

     The Sign test presents further difficulties.  Figure 1 shows a histogram of 1000
samples from a moderately skewed log  normal distribution that, in the Subcommittee's
experience, could easily be observed for environmental contamination.  Its geometric
mean (which is also the median in this case) and its geometric standard deviation are
both equal to 2.72.  Its arithmetic mean  is  about 4.5, approximately 1.6 times higher
than the geometric mean and median and correspoonding roughly to the 70th
percentile. If the DCGL were in the range 3 to 4, it would be lower than the arithmetic
mean, and the site should not be released, but it would be higher than the geometric
mean and median.

     Suppose a sample of 11 observations is taken from this distribution. First,  there
is about a 2% (0.7011) chance of drawing a sample where aN observations are less than
the mean. This rate is much lower than the 35% discussed above.  However, the
critical value for the Sign test is two or fewer values above the DCGL (probability
P-0.03). That is, if 2 or fewer of the observations exceed the DCGL, the site is
considered "clean" (assuming no "hot" values).  For our test distribution (which should
fail meeting the release  criterion), the probability of a "clean" result is about 0.3
because the probability of seeing a value above the mean is only about 30%.  That is,
our dirty site example would be released three times in ten, an unacceptable rate.
                                      33

-------
                                     Histogram
       210-
       140-
    i
    I
                    6    9    12   15    8   21    24   27   30   33   36
                                                                      39
                     Figure 1. A hypothetical contamination distribution.


5.5.4 Importance of Distinguishing between Median and Mean

     The target statistic for any exposure assessment should be the arithmetic mean
concentration for a defined area, and the uncertainty associated with the estimate of
the mean, due to all sources of potential error (variability of samples, analytical error,
compromises in experimental design, and uncertainty due to differences in judgment
amongst analysts). When the distribution of sample evidence is moderately to highly
skewed, then non-parametric statistical techniques cannot be used to determine the
uncertainty associated with the estimate of the arithmetic mean. These techniques are
more appropriate for estimating uncertainty at specified quantiles (percentiles) of the
sample distribution, including the 50th percent quantile (or median). As noted in the
document, when the data are highly skewed, the median of the sample will
underestimate the true arithmetic mean of surface contamination.

     Model assumptions that use a uniformly distributed source term do so as a
surrogate for the arithmetic mean of a heterogeneously distributed contaminant. If one
hypothetically homogenized a heterogeneously contaminated area to produce a
uniform contamination, the value of the uniform contamination would be equal to the
arithmetic mean of the heterogeneously contaminated system.
                                      34

-------
     The Sign test provides an indirect test of the mean only when the distribution of
the sample is symmetrical in arithmetic space. In the experience of the Subcommittee,
this is almost never the case. The majority of soil sampling programs usually reveals
highly skewed distributions.  Therefore, the Sign test, which is appropriate for testing
differences in median concentrations, may not be appropriate to test for differences in
mean concentrations.

     The discussion about the mean and median on MARSSIM page D-9, lines
210-226, should be revised. For the purposes of limiting exposure as well as for the
purposes of estimating exposure from  a defined area, the target statistic should always
be the arithmetic mean, regardless of whether the underlying distribution is symmetrical
or skewed.

     MARSSIM may allow the user to select a statistical test other than the WRS  or
Sign tests when a  skewed, non-symmetric distribution is observed, but it does not make
clear what is expected of the survey team when that occurs.  The document should be
explicit about whether other tests are to be considered at this juncture and how they
should be selected.  (Page 8-3, line 80; page 8-9)

5.5.5 Alternative Statistical Methods

     The Subcommittee urges the MARSSIM team to consider, and perhaps
encourage, alternative statistical methods for analyzing the survey data, especially
when skewed distributions are encountered.  Two techniques to be considered are
discussed below: bootstrapping and Bayesian analysis.

     Bootstrap estimator. The Sign test could be replaced by using a "resampling" or
"— bootstrap" estimator for the distribution of the arithmetic mean (Efron and
Tibshirani, 1993).  Bootstrapping is a process that generates a series of estimates for
the mean of a distribution by repeatedly resampling from the  actual set of measured
values, and then analyzes those means with standard statistical techniques.  Such an
approach is straightforward; we simply perform a large number K (e.g., K=1000) of
iterations in which we resample, with replacement, from the original  N sample values
(sample values can occur more than once in each resampling), and calculate the mean
of each iteration.  For example, if N=10, the  original sample might be
2,5,3,2,6,4,6,3,4,4, which has a  mean of 3.9. The first resampling might yield
5,4,6,3,3,5,2,2,2,4 for a mean of 3.6.  Additional resamplings would yield other means,
both above and below 3.9. Depending on the skew of the original data and the number
of iterations, the grand mean of the resampling might be higher or lower than 3.9, and
we could also obtain an estimate of the uncertainty about that grand mean. The 50th
largest mean value from our 1000 alternative realizations (of  a sample of size N=10)
would be equal to  the upper 95th  percentile of the true but unknown arithmetic mean.
This 95th percentile value on the arithmetic mean must be less than the DCGL in order
                                      35

-------
to declare the site safe for release. Using the bootstrap approach does not necessarily
require one to generate a K of 1000 mean values; however, fewer mean values reduce
the confidence of the final estimate of the mean. For example, if one had generated a
K of only 45 mean values, then this data set K would produce a statistical tolerance
limit that says that there is 90% confidence that there is at least a 95% chance that the
true but unknown mean is less than that value.  In this manner, the bootstrap method
can be combined with the tools of non-parametric statistics to estimate uncertainty on
the arithmetic mean of a sample (Conover, 1980).  The foregoing is not to suggest that
the non-parametric tests be abandoned, but rather that it might be a good idea for
MARSSIM to discuss some possible alternatives.

     Bayesian analysis. In some cases, the contamination data will not represent a
truly random sample of the environment (e.g., data for hot spot samples). Such
information can still be useful, but prior information about the sample's properties is
needed, leading to a Bayesian view of hypothesis testing.  When data are only partially
representative of a remediated site, due to the fact that they are not taken from a
randomized design or that they do not conform precisely to the same spatial and
temporal scales as those upon which the DCGL is based, then classical statistical
techniques are of limited use in determining the uncertainty about the  true but unknown
arithmetic mean concentration for that site.  Under these circumstances, approaches
based on Bayesian statistics may be  advantageous (Carlin and Louis, 1996; Gelman et
al., 1995). Bayesian statistics permit the explicit use of expert judgment to account for
the inherent possibility of flaws and biases in the data.  The result is that a credibility
(or subjective confidence) interval can be obtained about the arithmetic mean (or any
desired quantile) of the true but unknown distribution of soil concentration for both the
remediated site and any reference site. These credibility intervals form the  basis upon
which subsequent decisions are made.

     Nonparametric statistical tests are indeed superior to tests that are based on the
assumption of an underlying normal distribution. However, the results of such
nonparametric tests require that the samples be taken from a  random  or stratified
random design. Nonparametric statistics cannot be used to estimate the uncertainty of
the arithmetic mean of the sample. The target objective, however, should be a
subjective confidence interval, within  which the true but unknown arithmetic mean is,
with a high degree of belief, contained.  The proposed section in MARSSIM that
discusses alternate statistical methods should include references to Bayesian
approaches that will allow expert judgment to be combined with availability of imperfect
sample evidence.

     The use of professional judgment will be essential when data  sets cannot be
reliably evaluated using classical statistical procedures.  Professional  judgment is very
useful in combination with Bayesian approaches so that the transition  from imperfect
                                      36

-------
sample evidence to subjective statements of confidence about the nature and extent of
site contamination is transparent.

     On Pg. 8-7 of MARS SIM, reference could be made to the use of Bayesian
techniques to permit expert judgment to be used with sample evidence to determine the
uncertainty of the mean value. If the variability among samples is very high and the
number of samples are few (on the order of 6 to 10), it is possible for the maximum
value to be less than the true mean for the system.

5.5.6 Definition of the Gray Region

     The definitions of the  terms, 'gray region,' 'LBGR (Lower Bound of the Gray
Region), and 'relative shift in MARSSIM are less clear than they should be and seem to
differ among various parts of the document. It should be made clearer that the LBGR is
a design choice that influences the sample size but does not enter directly into the
release determination once the samples are taken. An analogy to the power
specification in an epidemiological investigation might be useful. Moreover, it could be
explained better that selecting the LBGR involves  a tradeoff between spending too
much on the survey and taking too large a chance that the site will fail even when the
DCGL criterion is met. A small relative shift (LBGR close to DCGL) is appropriate
when the decision is too close to call from a priori information. If the site is expected to
pass easily, a lower LBGR (larger relative shift) may be appropriate.  The sampling
should be planned so that there is an acceptably small probability of rejecting a site
that is, in fact, clean, given the cost of remediation.

     It is not clear in MARSSIM how the gray area boundary conditions relate to the
Type I and II decision errors, i.e., ka and kp probability set points.  How would one
relate the gray region boundary values to two sample distributions (reference
background and actual analyte sample concentrations with associated standard
deviations, or and os, respectively) separated by 4or, where an assumed equivalent
standard deviation of o= os (such  as the distributions represented in MARSSIM Figure
6.2)? Guidance is also needed on whether or not there needs to be consistency of
setting the Type I and Type II decision errors for different survey areas or survey type
for the same site survey plan. For example, can the selection of the Type I and Type II
decision errors ( ka and kp probability values) be different for building surface areas, in
situ field measurements, field sampling, core samples, etc.?

5.5.7 Specification of Sample Size

     The Subcommittee agrees with the MARSSIM assumption that the desired bias
should always be towards demanding an adequate number of data points.  The cost of
additional sampling and analysis will usually be small with respect to the cost of
cleanup.  In those cases where the cost of cleanup is small with respect to the cost of
                                      37

-------
sample analysis, then it may be cost effective to clean up using conservative estimates
of a DCGL to ensure that contamination has been effectively removed.

     If we want to be sure of sampling the w upper percentile (say 95th) of the
distribution with some large probability P (e.g., P=0.95), the relevant equation for the
required sample size N is:

                                                                             [1]
The result of [1] with P and w = 0.95 is 59 if we round up to the nearest integer.
Therefore, a sample size of about 60 will nearly always be sufficient to characterize a
survey unit.

5.5.8 Treatment of Outliers

     It is inevitable that some investigations will yield one or more results that seem
inconsistent with the remaining data and other information. There is no discussion in
MARSSIM of statistical tests for identifying these "outliers." Although they cannot be
rejected a priori, they should be viewed with caution, especially when they are
sufficiently high to shift the test result from "pass" to "fail." The MARSSIM document is
not as  clear as it should be regarding the treatment of outliers. Although it guides the
user toward further investigation when a value exceeds an "investigation level," it is not
explicit about what investigations are appropriate or how any resampling results should
be combined with the original value. Whether or not the value of the measurement
exceeds an investigation level, what responses  are appropriate for an outlier,
especially if it is well above the typical range? How should outliers be treated when
there is no obvious failure in application of quality assurance or quality control
practices? How are remeasurements weighted when they differ from the original
measurements? MARSSIM should provide guidance to the user as to how to identify
and treat outliers, or a justification for including such data in subsequent statistical
calculations.
                                       38

-------
                           6.  BROADER ISSUES
     The following comments, while not directly in response to the charge, may help to
improve the MARSSIM document and its use.  Included are comments on application of
MARSSIM outside site boundaries, the conservatism of MARSSIM, and the future of
MARSSIM after issuance of the final document.

6.1   Application of MARSSIM Outside Site Boundaries

     Many considerations enter the decision whether or not a site can or should be
released. One issue is that of the radiological condition of off-site areas, such as those
in which contamination may have been introduced in the past (or during remedial
activities) by surface runoff or windblown material. MARSSIM does not make it clear
how areas not within the site boundaries but potentially impacted by site activities
would be treated.  When are they subject to survey? Are off-site survey areas subject to
the  same rules as on-site ones? While such areas may need cleanup to reach risk- or
dose-based criteria, they may already be unrestricted in use, so "release" is not, strictly
speaking, an issue. However, access to potentially impacted lands outside site
boundaries is often a significant problem.

     The Subcommittee recognizes that application of the MARSSIM approach to
investigating radioactive contamination in off-site areas is a policy decision to be made
by the governing Federal agency.   However, it would be prudent for MARSSIM to urge
the  user to be vigilant during the Historical Site Assessment in identifying "vicinity
properties" that may be contaminated as a result of contamination from, and waste
disposal practices at, the primary or secondary sites.  Ancillary sites that may need to
be checked include landfills, sewage treatment plants (STP), STP sludge disposal
sites, dredge spoils disposal sites, site operations spoils disposal sites, and sewer
maintenance disposal sites.

6.2   Conservatism of MARSSIM

     If default DCGLs are derived using risk assessment methods similar to those
used in the EPA Draft Technical Support Document for the Development of
Radionuclide Cleanup Levels for Soil (U.S. EPA, 1994), they will be conservative in the
sense that they will more often than not  be lower than necessary to achieve the risk or
dose goals for site release. MARSSIM also appears to be conservative, in the sense
that it seems more likely to reject a site that meets the DCGLs than to release a site
that does not. Its principal conservatism lies in its selection of the recommended null
hypothesis: "The residual radioactivity in the survey unit exceeds the release criterion."
The degree to which that choice will bias release decisions will depend on the values
specified for Type I and Type II decision errors (a and p, respectively).


                                     39

-------
     Therefore, MARSSIM has the potential to compound the conservatism likely to be
present in the DCGLs if they are not made site-specific. Although this extra margin of
safety may be prudent public policy, the decision-makers ultimately responsible for site
release need to understand its existence and, to the extent possible, its magnitude.
Furthermore, any conservatism in translating a dose or risk criterion into DCGLs and in
determining compliance with those DCGLs through MARSSIM was probably not
included in the cost/benefit analysis for the cleanup standard, which could result in
lower cleanup benefits or greater cleanup costs than expected by those who will
eventually set the cleanup standards.

     Although not calling for a quantitative analysis of the effects of compounded
conservatism, the Subcommittee recommends that MARSSIM include a qualitative
summary of its assumptions  and policy choices, showing the likely direction of any
biases  introduced and, if possible, the relative magnitude of such biases. It should also
recommend that the planning team reveal its own assumptions and choices when a
site-specific survey design is developed.

6.3  Post MARSSIM

     A process should be established before MARSSIM is finalized to provide for
future revisions that will reflect changes in regulations and agency policies as well as
improvements as experience is gained in applying MARSSIM to real sites. For
example, once MARSSIM is  complete, will each of the federal agencies make
subsequent revisions to it on their own, or will MARSSIM continue as a multi-agency
collaborative effort in the future? Each revision should be reflected in the NUREG,
EPA, and NTIS publication numbers.   Clearly, it is  preferable to keep  MARSSIM a
multi-agency document after it is finalized.

     As needs arise, this successful interagency approach to survey surficial
contamination could be applied to radiological surveys of other media, such as
groundwater, subsurface soils and sewer contamination. In addition, the interagency
approach could be applied to activities such as site stabilization,  decommissioning
techniques, and standardized sampling procedures for various media.
                                      40

-------
                7.  FINDINGS AND RECOMMENDATIONS
7.1   Overall Approach to Planning Surveys

7.1.1 Value ofMARSSIM.  The MARSSIM document brings together DOE, EPA, NRC,
     and DOD with a common method of site surveys and investigations.  Adopting
     MARSSIM will mean that surveys done for any of the agencies will be
     immediately transparent to all.  In general, the Subcommittee found that
     MARSSIM is nearly a finished product. The multi-agency team is commended for
     its work in addressing the many complex issues involved, resulting in the
     compilation of an exceptionally well-prepared reference which is technically
     sound and which will be a useful tool for guiding final status surveys.  The
     document provides generally consistent and explicit guidance for planning and
     conducting radiation surveys for the decommissioning of radiologically
     contaminated sites. A major value of MARSSIM is providing managers with the
     quantitative tools necessary for determining how much to budget for sampling
     and analytical efforts.

7.1.2 Use of references.

     The credibility of the MARSSIM report would be greatly enhanced by an improved
reference section that includes recognized, readily accessible, up-to-date literature.

7.1.3 Consistency with federal regulations.

     MARSSIM has the challenge and responsibility to provide information that is
consistent with the current regulatory picture.  Because this aspect is admittedly
somewhat of a moving target, the Subcommittee recommends that MARSSIM avoid
referencing regulations presently  in draft form and instead refer to the regulatory
responsibilities of the agencies in these cases.

7.1.4 Additional considerations for site release.

     It needs to be emphasized early in MARSSIM that compliance with the DCGL is
only one of the considerations for release of a site for unrestricted use. Other
considerations not addressed  in MARSSIM include the radioactive contamination of
subsurface soil, surface water, or ground water.   Furthermore, if the release criterion is
supposed to  be ultimately risk-based, risks of residual chemicals may also need to be
factored into the decision about release criteria for radionuclides.
                                     41

-------
7.1.5 Scope limited to contaminated surfaces.

     MARSSIM should discuss its rationale for limiting its scope to guidance for
contaminated surface soils and building surfaces. Furthermore, it should more clearly
state that radioactive contamination of subsurface soil, surface water, and ground water
are explicitly excluded from its coverage. The document should include some
discussion of why these particular media were not included and the plans, if any, to
cover them in the future.  The contents of MARSSIM should then be made consistent
with its stated scope of coverage throughout, such that sections not clearly falling
within the scope should be omitted or moved to an appendix. If these sections are left
in the report in any form, then they raise the question as to whether or not the  DCGLs
established for surface soils and building surfaces also apply to these media.  Also,
MARSSIM should discuss the extent to which it is necessary to evaluate scenarios
under which subsurface contamination might be expected to contribute to surface
contamination in the future, and how this affects the decision of whether the site meets
release criteria.  For example, contaminated subsurface soil could be exposed via
erosion or excavation, and contaminated groundwater could result in surface
contamination via seepage and direct pumping of the groundwater.

7.1.6 Consistency between sampling methods and assumptions underlying
     derivation of DCGL.

     It is critical to verify the consistency between the definition of surface soil and the
method by which the DCGL is established from pathway modeling.  For example, it is
not entirely clear whether a DCGL applies to soil at the surface, at all depths, or only to
a certain depth (e.g., 15 cm).  Conversely, it is critical that the user ensure that the
manner in which samples  are collected and analyzed is consistent with the underlying
assumptions about distribution of radioactive contaminants on which the DCGL is
based.

7.1.7 Public involvement.

     The MARSSIM document is not very clear about the composition or functioning of
the planning team, especially the issue of public participation.  This issue should be
raised in the document and, to the extent possible, activities in the agencies regarding
public involvement in site survey design and implementation should be referenced
there.
                                      42

-------
7.2  Data Acquisition and Assessment

7.2.1 Organization of information concerning data acquisition.

     Sample collection protocols and analytical techniques are discussed in
MARSSIM Chapters 6 and 7 and in Appendix H.  The redundancy in coverage between
Chapters 6 and 7 should be minimized by careful editing and cross-referencing, by
redefining the scope of each chapter, or by combining the two chapters into a single
one, resulting in a more concise, user-friendly, and internally consistent document.
Standardized nomenclature should be used throughout MARSSIM in references to field
and laboratory equipment in order to avoid  confusion or ambiguity.  All techniques
mentioned in the main body of MARSSIM should have corresponding detailed
descriptions  in Appendix H.

7.2.2 Field Measurement Methods.

     Descriptions of field measurement methods, instruments, and operating
procedures in MARSSIM are technically sound but incomplete.  Some additions,
clarifications, and corrections are noted in our report.  MARSSIM should provide
guidance for the development of standardized procedures including  a list of
considerations for designing site-specific surface-soil sampling and preparation
methods so as to ensure that samples will be representative of the materials of concern
in deriving the DCGLs for the site.

7.2.3 Field and laboratory instrumentation.

     Descriptions of the selection and operation of radiation detection instruments for
laboratory analyses are technically sound and represent standard practice but may not
be state-of-the-art.  MARSSIM should standardize the level of detail used in its
presentation of this material and should also provide information on the planned scope
and current status of plans to prepare a manual on Multi-Agency Radiological
Laboratory Analytical  Protocols (MARLAP), which may be a more appropriate forum in
which to provide more thorough in-depth  guidance to the user on the selection and
operation of  laboratory instrumentation.

7.2.4 Sample collection and preparation for analysis.

     Surface-soil sample collection and preparation methods described in the manual
are technically sound  but incomplete. MARSSIM should provide guidance for the
development of standardized methods, referencing relevant ASTM standards.
MARSSIM should also provide a list of additional considerations for designing site-
specific surface-soil sampling and preparation methods so as to ensure that samples
will be representative  of the materials of concern in deriving the DCGLs for the site.


                                     43

-------
7.3  Demonstrating Compliance

7.3.1 Consistency between sampling data and the basis for definition ofDCGLs.
     The Subcommittee believes that it is critically important that the assumptions and
procedures used in MARSSIM to make comparisons with the DCGLs match those used
in defining the DCGLs. For example, if a DCGL for soil is derived from a dose limit or
risk criterion by assuming that a receptor ranges over a certain area on a random
basis, then the same area should be used for spatial averaging in the MARSSIM
statistical analyses. Such averaging is usually performed from the standpoint of
potential human receptors.  The manual should note that different spatial and temporal
scales of averaging will be necessary if dose and risk based criteria are applied to
components of the ecosystem other than humans for derivation of a DCGL.  This
recommendation assumes that the DCGL is derived in a manner appropriate for
characterizing human and/or ecological exposures likely to occur at the site under
investigation.

7.3.2 Cases for which the MARSSIM approach may not be able to demonstrate
     compliance.

     There appear to be cases that MARSSIM, as currently written, would have
trouble addressing. These cases include: 1) cases dealing with the release of sites
having been contaminated with naturally occurring radionuclides and in which the
DCGL is contained within the ambient (background) analyte variability, and 2) cases
where no reference background for certain or old construction materials can be
established. The Subcommittee recommends that future revisions of MARSSIM
provide guidance to the user regarding appropriate choices when such conditions are
encountered.  For example, the null hypothesis might be redefined to be that the
distribution of site radioactivity is no different from that at the reference site or than
ambient radioactivity in general.

7.3.3 Uncertainty in DCGL values.

     MARSSIM properly warns the user that the DCGL is not free of error and that the
uncertainty associated with this quantity may be considerable if derived using generic
assumptions and parameter values. However, its discussion of this issue is relegated
to an appendix.  This important aspect, with an expanded discussion of its implications
for the release decision, need to be disclosed more prominently in the text of the main

                                     44

-------
document.  It is clearly undesirable to design a survey around a DCGL that may not be
relevant to the actual conditions at a site, such that actual exposures, doses, and risks
would be largely different than those used to derive the generic DCGL. Consequently,
MARSSIM should more strongly encourage the user to examine critically the
assumptions made in any model used to derive DCGLs for a site in order to determine
whether application of site-specific information and parameters would result in large
modifications to the proposed DCGL, or whether development of a site-specific model
would be warranted in order to obtain a DCGL that is more relevant to the human and
ecological exposure conditions prevailing at the site.

7.3.4 Reference areas for background measurements.

     MARSSIM provides good guidance on how to obtain background measurements
from surveys of reference areas and how to use such measurements in the statistical
analysis of results from  a survey unit in order to determine its acceptability for release.
How the reference areas are to be defined, located, and defended is less clear.  For
example, if radionuclides are present in a survey unit as a result of human activities not
related to the site  in question, are they counted as background or site-related? Must
background reference areas always be on- site, off-site, or either? What should be
done if a representative reference area does not exist, e.g., for a disturbed site such as
a former landfill?

7.3.5 Flexibility in statistical procedures.

     In MARSSIM, the preferred null hypothesis is that a survey unit is not ready for
release, and the information gathered must be sufficient, with a high degree of
confidence, to accept the alternative hypothesis (i.e.,  that the unit meets the release
criteria).  Furthermore,  MARSSIM discusses in detail  two non-parametric procedures,
the Wilcoxon Rank-Sum test and the Sign test, for testing this hypothesis.  However,
MARSSIM allows  more flexibility in defining the null hypothesis and in choosing
statistical analysis methods to test that hypothesis than may be readily apparent to
most readers.  This latitude needs to be more clearly stated and the criteria for
selecting among potentially applicable tests need to be described.

7.3.6 Distinction between median and mean.

     MARSSIM's discussion about the mean and median should be revised in order to
ensure that the correct statistical parameter is used to compare concentrations in the
survey area to those in the reference area. The target statistic for any exposure
assessment should be the arithmetic mean concentration for a defined area, together
with the uncertainty associated with the estimate of the mean. For a normally
distributed population, the mean and median are identical in values.  However, when
the distribution of sample evidence is moderately to highly skewed, then


                                      45

-------
non-parametric statistical techniques cannot be used to determine the uncertainty
associated with the estimate of the arithmetic mean, and the median of such a sample
set will underestimate the true arithmetic mean of surface contamination. The majority
of soil sampling programs usually reveal highly skewed distributions. Therefore, the
Sign test, which is appropriate for testing differences in median concentrations, may not
be appropriate to test for differences in mean concentrations.

7.3.7 Alternative statistical methods.

     The Subcommittee urges the MARSSIM team to consider, and perhaps
encourage, alternative statistical methods for analyzing the survey data, especially
when skewed distributions are encountered.  Two techniques suggested for
consideration are bootstrapping and Bayesian analysis.

7.3.8 Treatment of outliers.

     It is inevitable that some investigations will yield one or more results that seem
inconsistent with the remaining data and other information.  MARSSIM does not
discuss statistical tests for identifying these "outliers." Although they cannot be rejected
a priori, they should be viewed with caution, especially when they are sufficiently high
to shift the test result from "pass" to "fail."

7.4  Broader Issues

7.4.1 Conservatism of MARSSIM.

     The guidance provided by MARSSIM may introduce an additional measure of
conservatism in the process of setting and determining compliance with radiation
cleanup standards, compounding the conservatism already likely to occur in developing
default DCGLs. Release decisions may be biased correspondingly.  MARSSIM should
include a qualitative summary of any biases that may result from its assumptions and
policy choices, and recommend that the planning team be similarly revealing when
developing a site-specific survey design.

7.4.2 Future revisions of MARSSIM.

     Before MARSSIM is finalized, a process should be established to provide for
future revisions that will reflect changes in regulations and agency policies as well as
improvements as experience is gained in applying MARSSIM to real sites.
                                      46

-------
7.4.3 Extension of MARSSIM and development of other multi-agency guidance
     manuals.

     As needs arise, this successful interagency approach to survey surficial
contamination could  be applied to radiological surveys of other media, such as
groundwater, subsurface soils and sewer contamination.  In addition, the interagency
approach could be applied to activities such as site stabilization, decommissioning
techniques, and standardized sampling procedures for various media.

7.4.4 Evaluation of methods for deriving DCGLs.

     DCGLs are critical for determining the acceptability of residual levels of
radioactivity remaining after a site has been remediated.  The Subcommittee suggests
that the various approaches proposed for derivation of DCGLs be reviewed and
evaluated.  This evaluation can be performed by an interagency group and by the
EPA/SAB. This evaluation should focus on the strengths and weaknesses of current
methodologies and opportunities to refine generic DCGLs with improved site-specific
models and data. This review is important but outside the scope of MARSSIM per se.
                                     47

-------
                       APPENDIX A- REFERENCES
American National Standards Institute (ANSI), 1996.  "Measurement and Associated
Instrument Quality Assurance for Radioassay Laboratories," ANSI N42.23 -1996.

American Society of Testing and Materials (ASTM), 1983. ASTM C998-83, "Standard
Method for Sampling Surface Soil for Radionuclides."

American Society of Testing and Materials (ASTM), 1983. ASTM C999-83,  "Standard
Method for Soil Sample Preparation for the Determination of Radionuclides."

Buck, J.W., G. Whelan, J.G. Droppo, D. L.  Strenge, KJ. Castleton, J.P. McDonald, C. Sato, and
G.P. Streile, 1995.  Multimedia Environmental Pollutant Assessment System (MEPAS)
Application Guidance, Guidelines for Evaluating MEPAS Input Parameters for Version 3.1,
Report PNL-10395, Pacific Northwest Laboratory (Richland, Washington).

Carlin, B.P., and Louis, T.A. 1996. Bayes and Empirical Bayes Methods for Data Analysis.
Monographs on Statistics and Applied Probability  69. Chapman & Hall, St.  Edmundsbury Press,
Bury St. Edmunds.

Conover, W. J., 1980. Practical nonparameteric statistics. Wiley & Sons, NY.

Currie, L.A., 1968. Limits for Qualitative Detection and Quantitative Determination. Analytical
Chemistry, 40(3):586-693.

Efron, B., and R.J. Tibshirani, 1993. An Introduction  to the Bootstrap. Chapman and
Hall, New York.

Eisenhower, E.H., 1991. Criteria for the Operation of Federally-Owned Secondary
Calibration Laboratories, National Institute of Standards  and Technology (NIST)
Special Publication 812,  October 1991.

ENVIRON Corporation, 1991.  Health Risk Assessment of the  Cotter Uranium Mill Site,
Canon City, Colorado. October 29.

Federal Register. Vol. 62, No.  139, Monday, July 21,  1997, p.  39087 (Reference on
Background  Radiation Definition.)

Gelman, A., Carlin, J.B., Stern, H.S., and  Rubin, D.B. 1995.  Bayesian Data Analysis.
Chapman & Hall, St. Edmundsbury Press  Ltd., Bury St. Edmunds.
                                     A-1

-------
Health Physics Society (HPS), no date. Calibration Laboratory Accreditation Program,
Criteria for Accreditation of Calibration Laboratories, Laboratory Accreditation
Assessment Committee of HPS.

Kooyoomjian, J., 1997. Memorandum to L.G. Weinstock, M.E. Clark, M.P. Doehnert, B.
Littleton, and C. Petullo concerning "Transmittal of Detailed Comments on the Multi-
Agency Radiation Survey and Site Investigation Manual (MARSSIM)," dated July 24,
1997.

Mettler, F., et al., 1990. Medical Management of Radiation Accidents. CRC Press.
Definition of "background" on page 374 of the Glossary.

NCRP (National Council on Radiation Protection and Measurements), 1996. A Guide
for Uncertainty Analysis in Dose and  Risk Assessments Related to Environmental
Contamination. NCRP Commentary No. 14. NCRP, Bethesda, MD.

U.S. DOE (U.S. Department of Energy), 1980.  A Background Report for Formerly
Utilized Manhattan Engineer District/Atomic Energy Commission Sites Program.
Report DOE/EV-0097A. September 1980. Definition of "background radiation" on
page 173 of the Glossary.

U.S. Enrichment Corporation, 1993.   Environmental Assessment for the Purchase of
Russian Low Enriched Uranium Derived from Dismantlement of Nuclear Weapons in
the  Countries of the Former Soviet Union.  Draft dated  November 25, 1993. Definition
of "background radiation" on page 10-1 of the Glossary.

U.S. EPA (U.S. Environmental Protection Agency), 1980.  Interim Guidelines and
Specifications for Preparing Quality Assurance Program Plans.  QAMS-005/80, EPA,
Washington D.C.

U.S. EPA (U.S. Environmental Protection Agency), 1994.  Technical Support Document
for the Development of Radionuclide  Cleanup Levels for Soil. Review draft dated
September 1994.

U.S. EPA (U.S. Environmental Protection Agency), 1996.  Technology Screening Guide
for Radioactively Contaminated Sites. Report prepared under Contract No. 68-D2-
0156 for the Office of Radiation and Indoor Air (ORIA)  and the Office of Solid Waste
and Emergency Response (OSWER). Report EPA 402-R-96-017. November 1996.
Definition of "background radiation" on page A-6.
                                    A-2

-------
U.S. EPA (U.S. Environmental Protection Agency), U.S. Department of Energy (DOE),
U.S. Department of Defense (DOD), and U.S. Nuclear Regulatory Commission (NRC),
1996.  Multi-Agency Radiation Survey and Site Investigation Manual (MARSSIM). Draft
for public comment, Decembers, 1996.  EPA402-R-96-018, NUREG-1575, NTIS-
PB97-117659.

U.S. NRC (U.S. Nuclear Regulatory Commission), 1997. NRC Final Rule on
Radiological Criteria for License Termination, 10 CFR Part 20 and Conforming
Amendments in 10 CFR Parts 30, 40, 50 and 51, 62 Federal Register (FR), 39058, July
21, 1997.

Yu, C., A.J. Zielen, J.-J. Cheng,  Y.C. Yuan, L.G. Jones, D.J. LePoire, Y.Y.  Wang, C.O.
Loureiro, E. Gnanapragasam, E. Faillace, A. Wallo III, W.A. Williams,  and  H. Peterson,
1993.  Manual for Implementing Residual Radioactive Material Guidelines Using
RESRAD Version 5.0. ANL/EAD/LD-2, Argonne National Laboratory.
                                   A-3

-------
                  APPENDIX B - LIST OF ACRONYMS
a         Type I decision error
P         Type II decision error
A         relative shift
or         one-sigma uncertainty of the mean background concentration
os         one-sigma uncertainty of the mean sample concentration
ALARA    As Low as Reasonably Achievable
ANSI      American National Standards Institute
ARAR     Applicable and Relevant or Appropriate Requirements
ASTM     American Society of Testing Materials
Bq        Bequerel (unit of radioactivity)
Ci         Curie (unit of radioactivity)
Cd        cadmium
CERCLA  Comprehensive Environmental Response, Compensation, and Liability Act
CFR      Code of Federal Regulations
cm        centimeter
Co-60     cobalt-60, a radioactive isotope of cobalt
DCGL     Derived Concentration Guideline Level
DCGLEMC  Elevated Measurement Comparison DCGL
DCGLW    Wide-area DCGL
DOD      Department of Defense
DOE      Department of Energy
dpm      decays per minute
DQO      Data Quality Objective
EPA      Environmental Protection Agency (U.S. EPA)
h         hour
H-3       hydrogen-3 (tritium), a radioactive isotope of hydrogen
ka         probability of a Type I decision error
kp         probability of a Type II decision error
kg        kilogram
LBGR     Lower Bound of Gray Region
m         meter
MAP      measurement assurance program
MARLAP  Multi-Agency Radiation  Laboratory Analytical Procedures (Manual)
MARSSIM Multi-Agency Radiation  Survey and Site Investigation Manual
MDC      Minimum Detectable Concentration
mSv      milli-Sievert
Nal       sodium iodide
NAS      National Academy of Science
NCRP     National Council on Radiation Protection and Measurements
                                   B-1

-------
Ni-63     nickel-63, a radioactive isotope of nickel
NIST     National Institute of Standards and Technology
NRC      Nuclear Regulatory Commission
NTIS     National Technical Information Service document
NUREG   Nuclear Regulatory Commission document
ORIA     Office of Radiation and Indoor Air
PE        performance evaluation
PERALS  Photon Electron Rejecting Alpha Liquid Scintillator
PIC       Pressurized lonization Chamber
QA       quality assurance
QAPP     Quality Assurance Project Plan
QC       quality control
R         Roentgen
RAC      Radiation Advisory Committee (U.S. EPA/SAB/RAC)
RCRA     Resource Conservation and Recovery Act
RESRAD  Residual Radioactive Material (transport code, which includes dosimetry
          and other components)
RI/FS     Remedial Investigation/Feasibility Study
SAB      Science Advisory Board (U.S. EPA/SAB)
SI         Le Systeme Internationale de Unites (International System of Units)
SOW     statement of work
STP      sewage treatment plant
Te        tellurium
UMTRCA  Uranium Mill Tailings Radiation Control Act
V&V      verification  and validation
WRS     Wilcoxon Rank Sum (statistical test)
Zn        zinc
                                    B-2

-------
                        APPENDIX C - GLOSSARY
accuracy - Closeness to the true value; the extent to which a given measurement
agrees with the standard value for that measurement: the degree of agreement of the
observed value with the true value of the quantity being measured.

action level - The numerical value that causes the decision maker to choose one of the
alternative actions. It may be a regulatory threshold standard (e.g., Maximum
Contaminant Level for drinking water); a dose- or risk-based concentration (e.g.,
DCGL); or a reference-based standard.

ALARA (As Low as Reasonably Achievable) - A basic concept of radiation protection
which specifies that exposure to ionizing radiation and releases of radioactive  materials
should be managed to reduce collective doses as far below regulatory limits as is
reasonably achievable considering economic, technological, and societal factors,
among others. Reducing exposure at a site to ALARA strikes a balance between what
is possible through additional planning and management, remediation, and the use of
additional resources to achieve a lower collective dose level.  A determination, and the
use of additional resources to achieve a lower collective dose level.  A determination of
ALARA is a site-specific analysis that is open to interpretation, because it depends on
approaches or circumstances that may differ between regulatory agencies. An ALARA
recommendation should not be interpreted as a set limit or level, and this system of
dose limitation which is based on keeping exposures "as low as reasonably
achievable," also takes economic and social factors into account.

alpha particle - Two neutrons and two protons bound  as a single particle that is
emitted from the nucleus of certain radioactive isotopes in the process of decay or
disintegration.

alpha-scanning equations - Equations used to determine the probability of detecting a
particular level of surface contamination using an alpha survey meter or to  calculate a
scan rate which will assure that a particular level of contamination can be detected with
a specific level of confidence.

ARAR (Applicable and Relevant or Appropriate Requirements) - A guidance value
(often a concentration level) taken from another regulatory context (such as maximum
contaminant levels for drinking water), but applied as a criterion for cleanup of
hazardous waste [e.g., Superfund, Resource Conservation and Recovery  Act (RCRA)
and other] sites.

area - A general term that refers to any portion of a site,  up to and including the entire
site.

                                     C-1

-------
arithmetic mean - The average value obtained when the sum of individual values is
divided by the number of values.

background (instrumental, method or process blank, field blank, environmental) -
The normal level observed, above which the phenomenon under investigation must be
detected.

background radiation - Ambient signal response recorded by measurement
instruments that is independent of radioactivity contributed by the radionuclides being
measured in the person or sample.  It includes radiation from cosmic sources; naturally
occurring radioactive material, including radon (except as a decay product of source or
special nuclear  material); and global fallout as it exists in the environment from the
testing of nuclear explosive devices or from nuclear accidents like Chernobyl which
contribute to background radiation and are not under the control of the cognizant
organization.  Background radiation does not include radiation from source, byproduct,
or special  nuclear materials regulated  by the cognizant Federal or State agency.

Bayesian  analysis - An approach to data analysis that uses both the information
contained  in a data set and prior information in the form of a probability distribution,
concerning the likely value of the quantity of interest. For example, a Bayesian
estimate of the sample mean would use both the sample data, and a prior distribution
for the value of the sample mean. The prior distributions can be based on historical
data or on a "personal" probability distribution assumed by the data analyst.

          According to the Bayesian view, all quantities are of two kinds; those known
to the person making the inference and those  unknown to the person.  The former are
described  by their known values, the uncertainty surrounding the latter being described
by a joint probability function for them all. Bayesian statistics permit the explicit use of
expert judgement to account for the inherent possibility of flaws and biases in the data.
The result is that a credibility (or subjective confidence) interval can be obtained about
the arithmetic mean (or any desired quartile) of the true, but known, distribution and
any reference site.

          In an analysis in which Bayes' theorem is used to derive posterior
probabilities from assumed prior knowledge together with observational data, for
example, biological information on the relationship between species and hazardous
substances can be combined with data on interspecies dose response to calculate the
response of human populations.

bias - A type of measurement error where each observation is mismeasured by a
constant amount.  For  example, if we measure objects with a "12 inch" ruler that is
really  11 inches long, all measurements will be too short by about 8.4%.  This is a
systematic or persistent distortion of a measurement process which causes errors in


                                     C-2

-------
one direction; the systematic or persistent distortion of a statistic may be as a result of
sampling procedure or other anomaly.

          In ANSI N13.30-1989D, bias is defined as "(a) The deviation of the expected
value of a random variable from a corresponding correct value,  (b) A fixed deviation
from the true value that remains constant over replicated measurements within the
statistical precision of the measurement. (Synonyms: deterministic error, fixed error,
systematic error.)

bootstrap estimator - An estimate of the variability of some statistic of interest such as
the sample mean, which is obtained by repeatedly sampling with replacement, a set of
measured values.  For example, if one has a set of N observations and is interested in
the variability of the sample mean, one would take, say, 1000 resamples of N from the
sample data (because we sample with replacement, some values will  occur more than
once in a given sample) and examine the distribution of the means of these resamples.
Thus, the bootstrap estimator is an estimate of some statistic of a distribution (such as
its mean) derived by repeatedly sampling from a set of measured values from that
distribution.

characterization survey - A type of survey that includes facility or site sampling,
monitoring, and analysis activities to determine the extent and nature  of contamination
and to provide the basis for acquiring necessary technical  information to develop,
analyze, and select appropriate cleanup techniques.

cleanup - Actions taken to prevent, minimize, or mitigate damage to the public health
or welfare or to the environment, which may otherwise result from a release or threat of
release of a hazardous  substance to the environment. Cleanup is sometimes used
interchangeably with the terms remedial action, response action, or corrective action.

coincidence counting  - Recording the occurrence of counts in two or more detectors
simultaneously or within an assignable time interval. In the Radiological Health
Handbook-1970 and  ANSI N1.1-1976), coincidence counting is defined as ..."the
occurrence of counts in two or more detectors simultaneously or within an assigned
time interval.  A true coincidence is one that is due to the incidence of a single particle
or of several genetically related particles. An accidental, chance or random
coincidence is one that  is due to the accidental occurrence of unrelated counts in the
separate detectors. An anticoincidence is the occurrence of a count in a specified
detector unaccompanied simultaneously or in an assignable time interval by a count in
other specified detectors. A delayed coincidence is the occurrence of a count in one
detector at a short, but measurable, time after a count in another detector. The two
counts are due to a genetically related occurrence such as successive events in the
same nucleus."
                                     C-3

-------
compliance - conformity or in accordance, as in compliance with rules, orders or
guidance

confidence interval - A range of values for which there is a specified probability (e.g.,
80%, 90%, 95%) that this set contains the true value of an estimated parameter.

data life cycle (DLC) - The process of planning, implementing, and assessing the
survey plan and assessing the survey results prior to making a decision.

data qualifiers - qualitative designators (such as the letters U, J, R) used in the data
verification and validation process that reflect the quality of the analytical data obtained
during a remediation project.  The data qualifiers reflect, in a qualitative manner, how
the data meet the project objectives defined by the quality indicators. The quality
indicators, when verifying and validating analytical data, are defined as:  Measurable
attributes of the attainment of the necessary quality for a particular environmental
decision.  Indicators of quality include precision, bias, completeness,
representativeness, reproducibility, comparability and statistical confidence.

data quality assessment (DQA) - The process of assessing the survey results,
determining that the quality of the data satisfies the objectives  of the survey, and
interpreting the survey results as they apply to the decision being made. This  process
focuses on the scientific and statistical evaluation of data to determine if the data are of
the right type, quality,  and quantity to support their intended use.

data quality objective (DQO) - A process to ensure that the survey results are of
sufficient quality and quantity to support the final decision.  It includes both  qualitative
and quantitative statements that clarify study objectives, define the appropriate type of
data, and specifies  levels of  potential decision errors that will be used as the basis for
establishing the quality and quantity of data needed to support decisions.

data validation - To substantiate, confirm, or approve the observed rule. Confirmation
by examination and provision of objective evidence that the particular requirements for
a specific intended use are fulfilled.  In design and development, validation concerns
the process of examining a product or result to determine conformance to user needs.

data verification - To ascertain the true value, authenticate, or establish as the correct
value. Confirmation by examination and provision of objective evidence that specified
requirements have been fulfilled. In design and development,  verification concerns the
process of examining a result of a given activity to determine conformance to the stated
requirements for that activity.
                                      C-4

-------
DCGLW - The wide-area DCGL, which is based on pathway modeling. It is the uniform
residual radioactivity concentration level within a survey unit that corresponds to the
release criterion (e.g.,  regulatory limit in terms of dose or risk). This is the area used
for the averaging of contamination to assess exposures to human and/or ecological
receptors.

decommissioning - The process of removing a facility or site from operation; to retire
(as in a facility) from active service.  Specifically, in reference to MARS SIM, it is the
process of removing a site safely from service, reducing residual radioactivity through
remediation (decontamination) to a level that permits release of the property and
termination of the license or other authorization for site operation. The objective of
decommissioning is to reduce the residual radioactivity in structures,  materials, soils,
groundwater, and other media at the site so that the concentration of each radionuclide
that contributes to residual radioactivity is indistinguishable from the background
radiation concentration for that radionuclide.

derived concentration guideline level (DCGL) - Based on pathway modeling, this is
the uniform residual radioactivity concentration level within a survey unit that
corresponds to the release criterion (e.g., regulatory limit in terms of dose or risk).
Again, based on pathway modeling, it is a radionuclide-specific predicted concentration
or surface area concentration of specific nuclides that could result in a dose (TEDE -
Total Effective Dose-Equivalent, or CEDE - Committed Effective Dose Equivalent)
equal to the release criterion.  In the MARSSIM, such concentration is termed the
DCGL.

direct measurement - A type of measurement of radioactivity or radiation exposure
which gives an immediate and direct response from which the quantity of radioactivity
or radiation exposure can be inferred eiither through calibration of the instrument
against a standard or by calculation.

dose - the quantity of radiation absorbed by a given mass of material or tissue. The
unit of dose is  the rad or gray (Gy).

double blind samples - A sampling and analysis procedure in which neither the
sample-labeler nor the analytical chemist knows the source of the sample.

duplicate - One of two independent samples collected in such a manner that they are
equally representative of the parameter(s) of interest at a given point in space and
time.

dwell time - The duration over which a survey instrument is kept in one place before
recording a radiation level. The shorter the dwell time, the less sensitive the
measurement (i.e., higher detection limit).


                                     C-5

-------
elevated measurement - A measurement that exceeds a specified value DCGLEMC.

elevated measurement comparison (EMC) - This comparison is used in conjunction
with the Wilcoxon test to determine if there are any measurements that exceed a
specified value, DCGLEMC.

exposure - The state of being exposed, such as exposure to radiation, or to the
elements. This is usually determined as a measure of x- or gamma-radiation at a
certain place, based on its ability to  produce ionization in air. The unit of exposure is
the roentgen (R).

final status survey - Measurements and  sampling to describe the radiological
conditions of a site, following completion of decontamination activities (if any) and in
preparation for release.

historical site assessment (HSA) - A detailed investigation to collect existing
information, primarily historical information, on a site and its surroundings.

hot spot - A strong or intense level  of radioactivity, above an investigation level, in a
particular location on a site.

human factors efficiency - Of or pertaining to study of human proficiency in
performance and the ability to accomplish a job with a minimum expenditure of effort;
the ratio of work done or energy expended to accomplish a given task.

hypothesis - An assumption about a property or characteristic of a set of data under
study.  Tentative conclusion about the distribution function of a random variable. The
goal of a statistical inference is to decide which of two complementary hypotheses is
likely to be true.  The null hypothesis describes what is assumed to be the true state of
nature and the alternative hypothesis describes the opposite situation.

impacted area - Any area that is not classified as non-impacted.  Areas with a
possibility of containing residual radioactivity in excess of natural background or fallout
areas.

instrument time constant - A measure of the response time of an instrument; it is the
product of the resistance and capacitance of the electrical circuit.
                                     C-6

-------
investigation level - A derived media-specific, radionuclide-specific concentration or
activity level of radioactivity that is based on the release criterion that triggers some
response, such as a further investigation or cleanup, if the level is exceeded. May be
used early in decommissioning to identify areas requiring further investigation, and may
also be used as a screening tool during compliance demonstration to identify potential
problem areas.  A DCGL is an example of a specific investigation level.  See also
action level.

Less-than data - Measurements that are less than the minimum detectable
concentration.

lower bound of gray region (LBGR) - The minimum value of the gray region.  The
width of the gray region (DCGL-LBGR) is also referred to as the shift, A.

matrix spike - A known amount of a specific  radionuclide [e.g., an  aliquot of sample
with a known concentration of target analyte(s)] added to a sample of the matrix being
analyzed (soil,  water, filter, etc.), prior to sample preparation and analysis,  to determine
the fraction of the radionuclide present in the  sample that  is recovered by chemical
separations or other sample handling processes. A matrix spike is used to document
the bias of a method in a given sample composition. It may be appropriate to direct the
laboratory to use specific samples as matrix spikes, and it may be necessary to provide
additional samples for this purpose.

mean - a quantity having a value intermediate between the values  of other quantities;
an average which is the sum of the values of  the observations divided  by the number of
observations.

median - the middle number in a given sequence, or the average of the two middle
numbers when the sequence has an even number of numbers (e.g., 4 in the median  of
1,3,4,8,9).

minimum detectable concentration (MDC)  - The  a-priori activity level that a specific
instrument and technique can be expected to detect 95% of the time. When stating the
detection capability of an instrument, this value should be used.  The MDC is the
detection limit,  Ld, multiplied by an appropriate conversion factor to give units of
activity.

non-parametric test - A test based on relatively few assumptions about the exact form
of the underlying probability distributions of the measurements. As a consequence,
nonparametric tests are generally valid for a fairly broad class of distributions. The
Wilcoxon Rank Sum test and the Sign test are examples of nonparametric tests.
                                     C-7

-------
null hypothesis - An assumption that the distribution, value, or set of observations or
measurements are not of the same distribution or source. The hypothesis that the
distribution function is the expected one, and the discrepancy between expected and
observed is due to chance.

outlier - A value which lies outside the range of other values in a set of observations;
measurements that are unusually large relative to the bulk of the measurements in the
data set.

performance evaluation - To carry through or execute in the proper or established
manner in order to fulfill a command,  promise or undertaking, such as the
decommissioning of a site,  and  to determine or set the value or amount or to appraise
and/or ascertain the numerical value of a function or relation, such as the effectiveness
of a decommissioning action..

Poisson observation - An observation from a Poisson distribution. The Poisson
distribution often describes the distribution of the number of events or objects on a fixed
interval of time (e.g., counts per minute) or space (e.g., defects per square meter). It
has a single parameter which corresponds to both the mean and variance of the
distribution of counts.

power curve  - The relationship between the number of samples to be taken and the
ability to detect a significant difference between the sampled population  and the
comparison population.

precision - The extent to which a set of repeated measurements agree with one
another; the degree of mutual agreement characteristic of independent measurements
as the result of repeated application of the process under specific conditions.  A set of
perfectly precise measurements of the same quantity will all have the same value. It is
concerned with the closeness of results. Note that precise measurements are not
necessarily accurate; see the definition for bias.

quality assurance (QA) - An integrated system of management activities involving
planning, implementation, assessment, reporting, and quality improvement to ensure
that a process, item, or service is of the type and quality needed and expected by the
client.

quality assurance/quality control (QA/QC) - Procedures and levels of  documentation
which are performed during implementation of a survey plan to collect information
necessary to evaluate the survey results.
                                     C-8

-------
quality assurance project plan (QAPP) - The process which documents how quality
assurance and quality control (QA/QC) are applied to obtain results that are of the type
and quality needed and expected.  This is usually manifested into a formal document
describing in comprehensive detail the necessary QA, QC, and other technical
activities that must be implemented to ensure that the results of the work performed
satisfies the stated performance criteria.

quality control (QC) - The overall  system of technical activities that measures the
attributes and performance of a process, item, or service against defined standards to
verify that they meet the stated requirements established by the client.  The system
includes operational techniques and activities that are oriented toward  and used to
fulfill requirements for quality.

quantile - (In a frequency distribution) - One of the values of a variable that divides the
distribution of a variable into groups having equal frequencies (e.g., as a quartile which
would divide the variable into four groups having equal frequencies).

receptor - an organ or group of organs exposed to stimulating agents (e.g.,
radionuclides).

reference (background) unit  - A level which is considered  as normal,  ambient or
natural, beyond which any addition is considered added by other means (e.g.,
anthropogenic  - man-made means).

relative shift (A/a) - A divided by  a, the standard deviation of the measurements.

release criterion  - a regulatory limit expressed in terms  of dose (mSv/y or mrem/y) or
risk (cancer incidence or cancer mortality). The terms, release limit or  cleanup
standard have  also been used to describe this term.

remediation -. The process and associated activities resulting in removal of
contamination from a site. Remediation is sometimes used  interchangeably with the
terms remedial action, response action, or decontamination..

risk - The hazard  or chance (degree of probability) of loss, exposure or injury.

roentgen (R) - The unit of radiation equal to the amount of x- or gamma-radiation that
will produce ions in air containing a quantity of positive or negative electricity  equal to
one electrostatic unit in 0.001293 gram of air.  See also  exposure.

sample distribution function - Frequency of each value obtained in a random sample.
                                     C-9

-------
scanning - An evaluation technique performed by moving a detection device over a
surface at a specified speed and distance above the surface to detect radiation.

scanning survey - The process of identifying contaminants (e.g., radionuclides) of
concern and their relative abundances on a site or tract of land, building, structure, by
scanning the surface with a detection device at a specified speed and distance above
the surface to detect radiation. Note that in most multi-radionuclide contaminations,
only sampling and lab analyses can (and not always "will") identify the radionuclides of
concern and their relative abundances.

scoping survey - A type of survey that is conducted to identify (1) radionuclide
contaminants, (2) relative radionuclide ratios, and (3) general levels and extent of
contamination.

Sievert (Sv) - The special name for the International System (SI) unit of dose
equivalent [1 Sv = 100 rem = 1 Joule per kilogram].

sign test - A nonparametric statistical test used to determine compliance with the
release criterion when the radionuclide of interest is not present in background and  the
distribution of data is not symmetric. See also Wilcoxon Rank Sum test.

single blind sample - A sampling and analysis procedure in which the analytical
chemist does not know the source of the sample.

site - Any installation, facility,  or discrete,  physically separate parcel of land, or any
building or structure or portion thereof, that is being  considered for survey and
investigation.

skewed distribution - (Statistics) - A statistical distribution with a majority of the
observations above or below the arithmetic mean. Such measurements or
observations of values have an oblique course or distorted form; asymmetric, oblique
direction or position about an  axis.

surface soil - The outer face  or exterior boundary of the soil, generally confined to the
top six inches (15 centimeters).

survey unit - A geographic area consisting of structures and/or land areas of specified
size and shape  at a  remediated site for which a separate decision will be made whether
the unit attains the site-specific reference-based cleanup standard for the designated
pollution parameter. Survey units are generally formed by grouping contiguous site
areas with a similar use history and the same classification of contamination potential.
Survey units are established to facilitate the survey process and the statistical analysis
of survey data.

                                     C-10

-------
symmetric distribution - A statistical distribution with approximately equal numbers of
observations above and below the arithmetic mean.  Such sets of observations typically
consist of pairs of points exhibiting symmetry about an axis, or a structure that exhibits
a regular repeated pattern or symmetry of the component parts and a structure which is
divisible into two similar parts by more than one plane passing through the center.

Type I decision  error - The probability of rejecting the null hypothesis when it is true,
or accepting the  alternative hypothesis when it is false.

Type II decision error - The probability of accepting the null hypothesis when it is
false.

uniform sample space - One for which all outcomes are equally likely.

vicinity property - An area or region adjacent to or near a place being examined,
evaluated, remediated, decommissioned  or otherwise surveyed.  Vicinity properties
may have been contaminated by any manner of transport mechanisms, including wind,
water erosion,  and groundwater contamination.  For instance, in the case of inactive
mill sites, it represents locations away from the mill sites where uranium  mill tailings
were used for construction or were transported off-site by wind or water erosion.

Wilcoxon Rank  Sum (WRS) test - A nonparametric statistical test used to determine
compliance with  the release criterion when the radionuclide of concern is present in
background. See also Sign test.
                                     C-11

-------
                             DISTRIBUTION LIST
Deputy Administrator
Assistant Administrators
EPA Regional Administrators
EPA Laboratory Directors
Deputy Assistant Administrator for Office of Policy,  Planning and Evaluation
          Director, Office of Strategic Planning and Environmental Data
          Director, Office of Policy Analysis
          Director, Office of Regulatory Management and Evaluation
Deputy Assistant Administrator for Air and Radiation
          Director, Office of Radiation and Indoor Air
          Director, Office of Radiation Programs
EPA Headquarters Libraries
EPA Regional Libraries
EPA Laboratory Libraries
Library of Congress
National Technical Information Service
Congressional Research Service

-------