'ASSESSMENT,' OF

     DATA QUALITY IN  THE' 1987

      TOXIC RELEASE INVENTORY:

         SITE VISIT PROGRAM
           March 27, 1990
            Prepared for:

       Mr.  Larry Longanecker
U.S. Environmental Protection Agency
 Office of Toxic  Substances 5(TS-779)
         401 M Street, S.W.
      Washington,  D.C.   204,60
            Prepared by:

         Radian Corporation
    2455  Horsepen  Road,  Suite  250
      Herndon, Virginia  22071
         DRAFT FINAL REPORT

-------
            ASSESSMENT OF

      DATA QUALITY IN THE 1987

      TOXIC RELEASE INVENTORY:

         SITE VISIT PROGRAM
           March 27, 1990
            Prepared for:

        Mr.  Larry Longanecker
U.S. Environmental Protection Agency
 Office, of Toxic  Substances  (TS-779)
         401 M Street, S.W.
       Washington,  D.C.   20460
            Prepared by:

         Radian Corporation
    2455  Horsepen Road,  Suite  250
      Herndon, Virginia  22071
               FINAL REPORT

-------
                               TABLE OF CONTENTS
EXECUTIVE SUMMARY 	   vi
1.0   INTRODUCTION	    1-1
      1.1   EPA's Overall Quality Assurance Program 	    1-1
      1.2   EPA's Site Visit Program	    1-2
      1.3   EPA's Pilot Survey  	    1-3
            1.3.1 Methodology   	    1-4
            1.3.2 Results   	    1-4

2.0   APPROACH	    2-1
      2.1   Survey Instrument 	    2-1
      2.2   Sample Selection  	    2-3
      2.3   Training	    2-4
      2.4   Arranging Site Visits	    2-7
      2.5   Site Visit Methodology	    2-8
      2.6   Data Collection and Calculations	    2-9
            2.6.1 General Information	   2-12
            2.6.2 Threshold Determinations  	   2-15
            2.6.3 Release Estimates 	   2-18
      2.7   Data Management and Data Quality  .  .  .  .	   2-19
            2.7.1 Data Management System	   2-19
            2.7.2 Data Weighting	   2-20
            2.7.3 Survey Bias	   2-22
      2.8   Wrap-up Meeting	   2-25

3.0   RESULTS	    3-1
      3.1   Threshold Determination Results 	    3-2
            3.1.1 Frequency of Errors in Threshold Determinations .  .  .    3-3
            3.1.2 Sources of Errors in Threshold Determinations ....    3-5
            3.1.3 Summary	   3-11
      3.2   Release Estimation Results  	   3-11
            3.2.1 Accuracy of Release Estimation  	   3-12
            3.2.2 Sources of Errors in Release Estimates  	   3-18
            3.2.3 Release Estimation Techniques 	   3-23
            3.2.4 Summary	   3-24
      3.3   Domains of Interest	   3-24

4.0   CONCLUSIONS	    4-1

5.0   RECOMMENDATIONS     	
      5.1   Additional Reporting Guidance 	
      5.2   Additional Areas of Study 	
APPENDIX A. SURVEY INSTRUMENT 	

APPENDIX B.-SAMPLE SELECTION PROCESS (ABT ASSOCIATES)

APPENDIX C. TRAINING MEETING AGENDA 	
                                      ii

-------
                         TABLE OF CONTENTS  (Continued)


                                                                          Page

APPENDIX D. SITE VISIT REQUEST INTRODUCTION LETTER  	  D-l


APPENDIX E. REFERENCES USED BY FACILITIES 	  E-l

APPENDIX F. WRAP-UP MEETING AGENDA AND SUMMARY OF ISSUES  	  F-l

APPENDIX G. SURVEY DATA	-	G-l

APPENDIX H. SUMMARY OF PREFERRED ALTERNATIVE APPROACHES 	  H-l

APPENDIX I. SUMMARY OF THE TOP TWELVE RELEASE ESTIMATE ERRORS FOR
            EACH RELEASE MEDIUM	1-1
                                      iii

-------
                                LIST OF TABLES
                                                                          Page

2-1       TARGET DISTRIBUTION OF FACILITIES AMONG THE FOUR DOMAINS
          IN THE SITE VISIT PROGRAM	      2-5

2-2       TARGET AND ACTUAL DISTRIBUTION OF FACILITIES AMONG THE FOUR
          DOMAINS IN THE SITE VISIT PROGRAM	     2-10

2-3       DISTRIBUTION OF FACILITIES AMONG SIC CODES 20-39  	     2.11

2-4       TYPES OF FACILITY PERSONNEL PRIMARILY RESPONSIBLE FOR
          PREPARING FORM R REPORTS AT THE SURVEYED FACILITIES ....     2-13

2-5       REFERENCES COMMONLY USED BY SURVEYED FACILITIES TO ESTIMATE
          RELEASES AND CONTROL EFFICIENCIES 	     2-14

2-6       APPROACHES COMMONLY USED BY FACILITY PERSONNEL TO MAKE
          THRESHOLD DETERMINATIONS  	     2-17

2-7       TARGET UNIVERSE FOR THE SITE VISIT PROGRAM	     2-21

2-8       EXPECTED PRECISION OF THE SAMPLE SELECTION PROCESS FOR THE
          FOUR DOMAINS OF INTEREST	     2-23

3-1       SUMMARY OF TOXIC CHEMICAL RELEASE QUANTITIES  	     3-13

3-2       FREQUENCY OF ERRORS IN NON-ZERO RELEASE ESTIMATES FOR EACH
          RELEASE MEDIUM	     3-16

3-3       FREQUENCY OF ERRORS IN ZERO RELEASE ESTIMATES FOR EACH
          RELEASE MEDIUM  	     3-17

3-4       TYPES OF ERRORS MADE FOR THE TWELVE LARGEST-QUANTITY ERRORS
          IN EACH RELEASE MEDIUM	     3-22

3-5       THRESHOLD DETERMINATION ERROR RATES WITHIN THE FOUR DOMAINS
          OF INTEREST	     3-25

3-6       RELEASE ESTIMATE ERROR RATES WITHIN THE FOUR DOMAINS OF
          INTEREST	     3-26
                                      iv

-------
                                LIST OF FIGURES


                                                                          Page

2-1       APPROACH USED TO PERFORM THE SARA TITLE III,  SECTION 313
          SITE VISIT PROGRAM	       2-2

3-1       THRESHOLD DETERMINATION RESULTS:   CHEMICAL COVERED UNDER
          SARA TITLE III, SECTION 313	       3-4

3-2       REASONS FOR INCORRECTLY INCLUDING CHEMICALS 	       3-6

3-3       REASONS FOR INCORRECTLY OMITTING  CHEMICALS  	      3-10

-------
                               EXECUTIVE SUMMARY









            As part of a continuing program to assess and improve the quality




of the data contained in the Toxic Release Inventory (TRI) database, the U.S.




Environmental Protection Agency (EPA) conducted a site visit program during




the summer of 1989.  The goals for this program were to identify areas in the




SARA Title III, Section 313 data collection process that could be improved and




to provide a quantitative assessment of the accuracy of the data collected for




calendar year 1987.  The program described in this report was developed




jointly by EPA and its contractors, Abt Associates and Radian Corporation.




The procedures used during the program were tested and refined during a pilot-




scale (25-site) program performed by Radian in the fall of 1988.









            Recognition of a fundamental limitation of the Section 313 report-




ing process is crucial to both the understanding of and the use of the results




and conclusions of the site visit program.  SARA Title III, Section 313




requires facilities  to report releases of listed toxic chemicals into the




environment;  however, the act does not require the facility to carry out new




monitoring and measuring to make estimates of these releases.  Thus, specific




release estimates calculated by facility personnel may differ a great deal (or




very little)  from actual release quantities.  Whether any difference exists




depends on several factors, including the amount and quality of monitoring and




measurement data already available to the facility, the type of estimate




necessary if no monitoring data exist, and the nature of the process involved.




The objectives of this program must be viewed in light of this fundamental




limitation of the release estimation process.  Also, the Form R chemical




report errors discussed in this report should be recognized as differences




from "better" estimates using available data rather than as differences from




                                      vi

-------
actual release quantities.  Likewise, recommendations cited in this report are




oriented toward making better use -of available data and toward improving EPA's



understanding of what data sources may be useful.  However, the quality of




release estimates contained in the TRI database will be limited by the quality




and the amount of data available to the reporting facility personnel.








            To collect information for the 1987 TRI data quality assessment,




Radian engineers and scientists visited 158 industrial facilities, who sub-




mitted a total of 728 Form R chemical reports for 1987.  Participation in this




program by these industrial facilities was voluntary, and they were assured of




anonymity.  These facilities were randomly selected to represent the.entire




population of facilities submitting Form R chemical reports for calendar year




1987.








            The site visits included conducting plant tours and interviewing




facility representatives.  Site surveyors carefully reviewed all documentation



supporting the Form R reports and searched for errors in the threshold deter-




minations and release estimates made by facility personnel.  Radian personnel




examined estimation techniques, checked calculations, and noted the types of




guidance materials used by the facilities.  This information was recorded on a




survey instrument developed specifically for this program.  Radian used a




detailed survey instrument and a surveyor training course to  achieve




facility-to-facility consistency among the survey teams involved in the




program.  The information from each survey instrument was reviewed to assure




quality and consistency and was then compiled in a computerized database.








            As a result of this site visit program, the Agency now has a




database containing information on the quality of 1987 Form R report data.




                                     vii

-------
Because the sample population was carefully drawn to provide a valid




representation of most of the population of reporting facilities, this data-




base can be used for a variety of analyses, such as analysis of data quality




and error rates by SIC code, facility size, and release medium.  However, the




current study was confined to performing an overall assessment of the 1987 TRI




data quality and then making recommendations for reducing error frequency in




future reporting years.








            The most significant finding of the program was that the overall




quality of the TRI database is good.  The data obtained from the site visits,




when compiled and generalized to the TRI database in general, revealed a




slight overall under-reporting of releases and transfers off-site of the total




national release quantity contained in the 1987 TRI data (based only on




reporting facilities).  The most important limitation associated with this




finding was that it applies to aggregate data only, that is, some of the




apparent data accuracy comes from the cancellation effects of under- and over-




reporting of releases.  When the total quantity of under-reported releases is




added to the total quantity of over-reported releases, the combined differen-




tial is less than 20% of the total national release quantity.  This figure




provides a better sense of the accuracy of smaller segments of the TRI data-




base and is also a better indicator of the accuracy of the individual Form R




chemical reports.








            The largest error source, both in frequency and in total impact on




the quantitative release estimate total, was not found in faulty calculations




or in poor estimation techniques, but in the misunderstanding and misinterpre-




tation of instructions and definitions associated with the Form R process.
                                     viii

-------
Data obtained during the site survey program indicated that about one-tenth of

the release quantities contained in the TRI database would have been omitted

had facility personnel correctly applied threshold determinations, activity

categories, and exemptions.  Similarly, the number of missing release

estimates was approximately one-tenth of the total number of release estimates

in the TRI database.  Better and more detailed instructions should reduce

these types of errors in the future.  This project report provides details on

the specific types of mistakes most frequently made on the Form R chemical

reports and provides recommendations for specific areas of the instructions

that warrant revision.  Additional guidance is also recommended for particular

steps in the reporting process.



            The data quality assessment identified five major types of errors

in the Form R data reported by facilities for calendar year 1987:
                  Chemicals reported that need not have been reported
                  (incorrectly-included chemicals);

                  Chemicals that were not reported but should have been
                  reported (incorrectly-omitted chemicals);

                  Chemical releases that need not have been reported;

                  Chemical releases not reported that should have been
                  reported; and

                  Chemical releases reported that should have been reported as
                  a quantity greater than one order-of-magnitude different
                  than the reported quantity.
            The limited testing and environmental monitoring performed by the

facilities made it difficult to assess the quality of the estimation tech-

niques used by these facilities in completing to the Form R chemical report.
                                      ix

-------
According to the survey, 'the vast majority of facilities estimated releases




using a combination of engineering judgment and material balance calculations,




Radian's site surveyors identified preferred alternative approaches for




estimating releases at less than 10% of the visited facilities.

-------
1.0         INTRODUCTION









            Section 313 of Title III of the Superfund Amendment and Reauthori-




zation Act of 1987 (SARA) requires the U.S. Environmental Protection Agency




(EPA) to collect information concerning releases of chemicals to the environ-




ment from manufacturers, processors, and users of listed toxic chemicals.  The




information collected under SARA Title III. Section 313 is useful for many




purposes, but primarily for informing the general public and the communities




surrounding affected facilities of releases of toxic chemicals, assisting in




focusing the Agency's research into the effects and control of toxic




substances, and aiding in the development of regulations, guidelines, and




standards.  At the time the current study was initiated, approximately 68,500




first-year (1987) Form R chemical reports had been submitted to EPA by more




than 18.000 industrial facilities.  When completed. 74,152 chemical reports




had been submitted by 19,278 facilities, and compiled in EPA's Toxic Release




Inventory (TRI) database.









1.1         EPA's Overall Quality Assurance Program









            EPA needed to gain a sense of the accuracy of the TRI database




because of its wide audience and many intended uses.  For this reason, EPA




designed and implemented a program to assess the quality of the data collected




under Section 313 and to identify areas where improved guidance would be




useful for improving the accuracy of future reported data.  In addition to the




site visit program described in this report, EPA sponsored a telephone contact




survey program in which computer algorithms were used to screen for







                                      1-1

-------
technically questionable data.  EPA also conducted a confidential telephone

survey of facilities that may be subject to the requirements of Section 313,

but did not file any reports (the "nonresponders" survey).



1.2         EPA's Site Visit Program



            EPA's site visit program was designed to provide a quantitative

assessment of the accuracy of the data collected for calendar year 1987 by

identifying both the frequency and the magnitude of errors in the Form R data

and the reasons these.errors occurred.  The on-site review of industrial

processes, pollution control technologies,  and documentation supporting the

Form R reports was designed to reveal:
                  Overlooked chemicals and releases that were not likely to be
                  identified in the telephone surveys;

                  Errors in the Form R reports submitted to EPA; and

                  Whether more reasonable release estimation techniques could
                  have been used,  based on information  available to the
                  facilities.
The results from this program can now be used to  improve the Form R reporting

instructions,  and thus improve the quality of information in the TR1 database

in future years.  The quantitative assessment of  the TRI database also pro-

vides a baseline for measuring the extent of data-quality improvement as

guidance changes are implemented.



            Users of the results of the site visit  program,  as well as the TRI

database itself, should be aware of a basic limitation of the Section 313

                                      1-2

-------
reporting process.  Under SARA Title III, Section 313, facilities are not




required to perform any additional monitoring or measurement of the quantity




of toxic chemicals released to the environment to calculate Form R release




estimates.  Therefore, the techniques' selected by facilities to estimate




releases depend on the nature of the data available to facility personnel, and




the quality of these release estimates in turn depends on both the proper




application of the estimation techniques and on the quality of available data.




At facilities where supporting data were available, surveyors carefully




examined the estimation calculations and data sources and then recalculated




the estimates.  In many instances, the site surveyors were able to identify




data sources overlooked by facility personnel, and these new data were used to




recalculate release estimates during the site visits.  Radian's site surveyors




also assessed the quality of the estimation techniques by recalculating




releases using alternative approaches where more accurate estimation tech-




niques were appropriate and where available data warranted.








1.3         EPA's Pilot Survey









            Before conducting the full-scale site visit program, Radian




performed a limited, nonstatistical pilot survey program in the fall of 1988.




The purpose of this pilot survey was to confirm that the approach of perform-




ing site visits would add valuable information for assessing data quality and




would provide feedback that could be used to improve the Form R report. the




reporting instructions, and other guidance materials.  The pilot survey also




provided information about the survey instruments used to collect the quality




assurance program data.  The experience gained in the pilot survey also proved






                                      1-3

-------
useful in planning, scheduling, and costing the full-scale site visit quality




assurance program conducted in the summer of 1989.









1.3.1       Methodology








            For the pilot survey, Radian engineers  and scientists visited 25




of the facilities described in the TRI database.  At each facility, site




surveyors toured the industrial processing operations and storage and waste




treatment areas, reviewed supporting documentation,  and interviewed facility




personnel.  A detailed survey instrument was completed at each facility during




the site visit.  The survey instrument was developed specifically for this




program to ensure consistency in survey procedures  and documentation




materials.  Surveyors assessed whether the facility personnel had reported the




correct chemicals, whether they had overlooked any  uses of the chemical or




release media, whether they had used the best release estimation techniques




available, and whether they had accurately estimated releases.








1.3.2       Results









            During the pilot survey, Radian visited 25 facilities representing




18 of the 20 Standard Industrial Classification (SIC) codes covered by SARA




Title III, Section 313 requirements.  These facilities encompassed a wide




variety of process types, chemicals, waste streams  and calculation methods,




and varying levels of technical skill among facility personnel who prepared




the Form R reports.  A detailed account of the results of this pilot survey




are provided in a Radian report entitled Final Summary Report:  Toxic Release






                                      1-4

-------
Inventory Data Quality Assurance Site Visits (March 1989).  Briefly, this




report concluded that the approach of performing site visits was an extremely




effective way to assess data quality, and that facility tours would provide an




environment for addressing items unlikely to be discussed in a telephone or




paper survey.  The survey instrument was useful for conducting site visits,




although some revisions were recommended and implemented for the full-scale




site visit program.  Also, various errors made by facility personnel in each




section of the Form R report as well as comments regarding data availability,




test and monitoring data, and documentation were summarized in the report.
                                      1-5

-------
2.0         APPROACH



            The approach for performing the site visit program,  shown schema-

tically in Figure 2-1, consisted of the following steps:
            (1)   Developing a Survey Instrument:
            (2)   Selecting facilities to be visited (Sample Selection);
            (3)   Training site surveyors (Training);
            (4)   Arranging Site Visits:
            (5)   Performing site visits (Site Visit Methodology);
            (6)   Completing survey instrument for each site (Data Collection
                  and Calculations);
            (7)   Summarizing, generalizing, and reporting quantitative
                  results (Data Management and Data Quality);  and
            (8)   Reviewing qualitative results (Wrap-up Meeting).
Each of these steps is discussed in the following subsections.



2.1         Survey Instrument



            The survey instrument, shown in Appendix A,  was designed to

facilitate Radian's review of threshold determinations and release estimates.

Each section of the survey instrument focused on identifying specific types of

errors made by facility personnel on their Form R submittals.  The Radian

engineers and scientists who performed the site visits used the survey instru-

ment as a detailed checklist to ensure that all pertinent items were reviewed,

and also as a consistent format for recording both the data collected during

site visits and the errors made by facility personnel on their Form R reports.

In addition to its primary focus on chemical-specific information, 'the survey

instrument contained questions regarding the usefulness of the reporting
                                      2-1

-------
   Develop  the
Survey Instrument
   Develop a
  Computerized
 Data  Management
     System
                              Define the Goals
                               for  the Sample
                             Selection Process
                              Select a Random
                            Group of Facilities
                               for the Sample
                             Obtain Facility's
                          Voluntary Participation
                          and Arrange Site Visits
                             Perform the Site
                            Visits  and Complete
                           the  Survey Instrument
                               for  each  Site
Internally  Check the
 Consistency of the
 Completed Survey
    Instruments
                             Enter and Verify
                              the  Data  in the
                              Data Management
                                  System
                                    Develop a
                                 Training Manual
                                    Train Site
                                    Surveyors
                                   Prepare for
                                 each Site Visit
                                                             Hold an Informal
                                                             Wrap-up  Meeting
                                                               to Discuss
                                                            Qualitative items
                                  Summarize the
                                 Results  of the
                                 Mrap-up  Meeting
                                 in a Memorandum
                          Generalize the Results
                          using Facility Heights
                                    I
                              REPORT RESULTS
      Figure 2-1.   Approach used  to  Perform the SARA
          Title  III,  Section 313 Site Visit  Program
                                    2-2

-------
instructions, the Title III telephone hotline, and the other published




guidance materials.









2 . 2         Sample Selection









            Radian's goal in the sample selection activity was to select a




random group of facilities that vould represent the entire population of




facilities included in the TRI database.  Observations made during the pilot




survey indicated that chemical industry facilities (Primary SIC Code 28) used




more complicated estimation techniques than did nonchemical industry facili-




ties, and that facilities reporting large-quantity emissions for chemicals had




different types of problems in estimating releases than did facilities report-




ing smaller quantities of emissions.  Due to these differences, four domains




of interest (chemical industry facilities,  nonchemical industry facilities,




large-quantity emitters,  and small-quantity emitters) were created to assess




trends among these major categories.  Radian elected to over-sample chemical




industry facilities and large-quantity emitters because of their relatively




large contribution to the overall emission quantities for the entire TRI




database.









            A target of 150 facilities was selected as the number of facili-




ties needed to ensure the statistical validity of the data collected during




the site visit program.   Appendix B provides a detailed discussion of the




procedure used by Abt Associates to select the sample jjroup of facilities, and




also provides a description of the weighting system (i.e., the number of




facilities in the TRI database represented by each sample point).  Briefly,







                                      2-3

-------
facilities submitting 151 or fewer Form R reports  were divided into  geographic

clusters on the basis of the first three digits in their zip codes.   This

clustering approach was used to minimize costs by reducing travel costs and

travel time for site survey teams.  A stratified sample of facilities was

drawn randomly from each of the sampled geographic clusters, based on the

desired number of site visits for the two stratifiers (chemical industry

facilities versus nonchemical industry facilities; large-quantity emitters

versus small-quantity emitters) discussed earlier.  Table 2-1 shows  the target

distribution among the four domains of facilities in the site visit  program.



2.3         Training



            Radian's goal in the training program was to ensure consistency

and high-quality work among all site surveyors.  The first step in this

training program involved compiling a comprehensive training manual  that would

be helpful both during and after the training session.  The manual also was

designed as a reference document that would be useful when questions arose

during site visits.   The manual included a copy of the survey instrument and

instructions for its completion, a copy of the pilot survey report,  a discus-

sion of the surveyor's responsibilities on the project, a review of preferred

alternative approaches for completing release estimate calculations, and
     'A site visit at a facility submitting 16 or more  forms would have required
considerable time to complete.   Limiting the visits  to  facilities  with 15 or
fewer forms eliminated only  about 2% of the  facilities  and  22%  of  the Form R
reports from the population  sampled,  thus  including in the survey the maximum
number of facilities within available resources.  Since a major element of the
study was to determine error sources, the larger number of facilities surveyed
gave a greater  range  of  information on how  facility  personnel dealt with the
Form R reporting process and the information available to them.


                                      2-4

-------
                                   TABLE 2-1

           TARGET DISTRIBUTION OF .FACILITIES AMONG THE FOUR DOMAINS
                           IN THE SITE VISIT PROGRAK
                       TARGET:  150 completed site  visits

                       40 chemical  industry  facilities*
                      110 nonchemical industry facilities
                          75 large-quantity emitters8
                          75 small-quantity emitters8
*Chemical facilities were defined as facilities with Primary SIC Code" 28.

BA facility was defined as a large-quantity emitter when its reported release
to any medium for calendar year 1987 was greater than the 90th percentile  for
that medium.  Facilities without  such releases were defined as small-quantity
emitters.
                                      2-5

-------
copies of all guidance documents EPA had published concerning accurate com-




pliance with the Form R report instructions.









            The second step in the training program involved holding a series




of training sessions to familiarize project personnel with program require-




ments.  To minimize travel costs on the site visit program, Radian used




personnel from several of its offices across the country to conduct the site




visits.  Personnel from the following Radian offices were used to arrange and




conduct the site visits:  Austin, Texas; Herndon, Virginia; Los Angeles,




California; Milwaukee, Wisconsin; Research Triangle Park, North Carolina; and




Sacramento, California.  In each of these offices, one person, the office




coordinator, took the lead role in managing the activities from that office,




including selecting site surveyors, training survey teams, scheduling site




visits, supervising site surveying activities,  and performing quality




assurance review of survey instruments.









            Radian central project staff trained office coordinators at a one-




day training session held in Radian's Herndon,  Virginia office (the agenda for




this meeting is provided in Appendix C).  The purposes of this session were to




train the office coordinators in their project duties and to prepare the




office coordinators to train the site surveyors.  Each office coordinator then




trained the survey teams at his/her respective office.  The final step in the.




training program consisted of each office coordinator thoroughly reviewing the




completed survey instruments with his/her site surveyors.  This quality




assurance review allowed the office coordinator to identify and correct any
                                      2-6

-------
problems as the program progressed and Co maintain consistency of approach




among the surveyors.








2.4         Arranging Site Visits








            The program goal in arranging the site visits was to provide each




facility in the sample with an equal chance to participate in the site visit




program, thus ensuring the statistical validity of the program.   Participation




in the site visit program was voluntary;  the facilities were not legally




required to participate.  A key factor encouraging voluntary participation was




the assurance of anonymity to the facilities.  Names,  locations, and all other




facility identification data were shielded from the Agency throughout the site




visit program.








            As a first step, introduction letters (copies of these letters are




provided in Appendix D) were sent to each facility's technical contact, and




where appropriate, to each facility's senior management official.  These




letters explained the purpose of the quality assessment program and the




anticipated burden on and benefits to the facility, and assured the facility




that all facility-specific data would be treated as confidential.  Radian




followed these letters with telephone calls to the technical contacts at the




facilities to solicit their participation, and for those facilities agreeing




to participate in the program, to arrange a date for the site visit.
                                      2-7

-------
2.5         Site Visit Methodology



            The goal of the sice visit methodology was to collect all the

information needed to complete the survey instrument accurately, while mini-

mizing burdens on facility staff.  To complete the site visit program as

efficiently as possible, the engineers and scientists staffing the program

were selected on the basis of their experience in performing environmental

audits of industrial processing facilities, and were required to have a

thorough understanding of chemistry, technical calculations, multimedia

environmental concerns, and pollution control technologies.



            Thorough preparation for each site visit was crucial to meeting

the overall goals for the site visits.  This preparation included gaining a

detailed familiarity with the reports submitted by facilities, the properties

(such as volatility or solubility) and uses of the chemicals to be reviewed,

the types of industrial processing operations to be reviewed, and the expected

efficiencies of the pollution control technologies employed at the facilities.



            On-site survey activities included tours of the facilities, which

focused on material storage areas, industrial processing operations, and

pollution control equipment;  careful review of all readily available documen-

tation; and interviews with appropriate facility employees regarding documen-

tation materials.  The site visits were designed to reveal:
                  Overlooked chemicals and releases that were not likely to be
                  identified in the telephone surveys;

                  Errors in the Form R reports submitted to EPA; and

                                      2-8

-------
                  Whether more reasonable release estimation techniques could
                  have been used, based on information available to the
                  facilities.
Site surveyors recorded these results on the survey instruments and reviewed

them with facility personnel before leaving the sites.  After each site visit,

the appropriate office coordinator reviewed the completed survey instrument

for completeness and technical accuracy.  Where necessary, follow-up phone

calls were made to facility personnel to clarify specific data.



2.6         Data Collection and Calculations



            To obtain the EPA target of 150 facility site visits with the

distribution shown in Table 2-1, and based on the acceptance rates experienced

during the pilot survey, Abt randomly selected 280 facilities from the 1987

TRI database.  Radian contacted the 280 selected facilities to request their

participation.  The overall response rate was higher than expected, and a

total of 188 facilities agreed to participate in the site visit program.  Of

these 188 facilities, 158 sites, distributed among the four domains of

interest as shown in Table 2-2, were visited.  Table 2-3 shows the distribu-

tion among SIC codes of the 156 facilities having usable data.2  Surveyors

visited 26 facilities in the 10-49 employee range; 97 facilities in the 50-499

employee range; and 33 facilities with 500 or more employees.
     2Two of the 158  sites  visited had fewer than 10  employees  and therefore
were not required to submit Form R reports.  -These sites were deleted'from this
study as being nonrepresentative of the population under consideration.


                                      2-9

-------
                                   TABLE 2-2

     TARGET AND ACTUAL DISTRIBUTION OF FACILITIES AMONG THE FOUR DOMAINS
                           IN THE SITE VISIT  PROGRAM
                      TARGET:  ISO completed site visits

                       40 chemical industry facilities*
                      110 nonchemical industry  facilities
                          75 large-quantity emitters8
                          75 small-quantity emitters8
                      ACTUAL:   156 completed site visits0

                       44 chemical  industry  facilities*'
                      112  nonchemical industry facilities

                          78 large-quantity emitters8
                          78 small-quantity emitters8
AChemical  facilities  were defined as facilities with Primary SIC Code 28.

facilities  were  defined as large-quantity emitters when their reported
release to any medium was greater than the 90th percentile for  that medium
reported for calendar year 1987.  Facilities without such releases were
defined as small-quantity emitters.  See page B-3 of Appendix B.

cln addition to the 156 site visits shown, site surveyors visited two other
chemical facilities that reported small-quantity emissions.  These two
facilities had fewer than 10 employees and were deleted from the survey
program results.
                                      2-10

-------
                                   TABLE 2-3

               DISTRIBUTION OF FACILITIES AMONG SIC CODES 20-39

                                                                 Number of
SIC Code    Description                                     Facilities Visited

  20        Food and kindred products                                16
  21        Tobacco products                                          0A
  22        Textile mill* products                                     5
  23        Apparel and other finished products
            made from fabrics and other similar
            materials                                                 1
  24        Lumber and wood products, except
            furniture                                                 2
  25        Furniture and fixtures                                    2
  26        Paper and allied products                                14
  27        Printing, publishing, and allied
            industries                                                3
  28        Chemicals and allied products                            44
  29        Petroleum refining and related
            industries                                                0A
  30        Rubber and miscellaneous plastic
            products                                                  7
  31        Leather and leather products                              0A
  32        Stone, clay, glass, and concrete
            products                                                  2
  33        Primary metal industries                                 16
  34        Fabricated metal products, except
            machinery and transportation
            equipment                                                16
  35        Industrial and commercial machinery
            and computer equipment                                    5
  36        Electronic and other electrical
            equipment and components, except
            computer equipment                                       11
  37        Transportation equipment                                  7
  38        Measuring, analyzing, and
            controlling instruments;
            photographic, medical and
            optical goods; watches and
            clocks                                                    2
  39        Miscellaneous manufacturing
            industries                                              	3

            TOTAL                                                   156B
ANo facilities in SIC Codes 21,  29,  or 31 were visited  for  this  program;  each
of these SIC codes had fewer than 2% of the reporting facilities in the 1987
TRI data.

"Radian visited two additional chemical facilities, but they had fewer than 10
employees, and were deleted from the survey program results.

                                      2-11

-------
            At the 156 surveyed facilities having usable data, site surveyors




reviewed general information pertaining to the Form R report, as well as




specific information related to threshold determinations and release




estimates.  Site surveyors reviewed 728 Form R chemical reports at these 156




facilities.  The following subsections describe the information collected at




these facilities.








2.6.1       General Information








            To identify trends among the surveyed facilities, site surveyors




reviewed general information on the type of personnel primarily responsible




for preparing the Form R reports and on the types of references used by these




personnel.  Table 2-4 lists the six types of facility personnel identified as




primarily responsible for preparing the Form R reports at the surveyed facili-




ties.  Facility staff, both nonenvironmental (47%) and environmental (29%),




were most frequently tasked with preparing Form R reports.   Corporate environ-




mental personnel and outside consultants or contractors accounted for about




10% and 11% of cases, respectively.









            Table 2-5 lists the types of references most commonly used by




personnel at the surveyed facilities to estimate releases and control




efficiencies.  This table provides a summary of the information contained in




Appendix E, which contains details on the number of personnel using specific




references.  The most frequently used references were those published specifi-




cally to support the Section 313 program.  Further, certain industries




frequently used trade association data for reference; for example, the pulp






                                     2-12

-------
                                  TABLE 2-4

            TYPES OF  FACILITY PERSONNEL PRIMARILY RESPONSIBLE FOR
              PREPARING  FORM R REPORTS AT THE  SURVEYED FACILITIES
                                                 Number of Facilities
Report Preparer                                    Preparing Reports


Facility Staff                                               73

Facility Environmental Staff                                 45

Consultants or Contractors                                   17

Corporate Environmental Staff                                16

Company President                                             3

Corporate Safety Department Staff                           	2

TOTAL                                                       156
                                     2-13

-------
                                   TABLE 2-5

          REFERENCES COMMONLY USED-BY SURVEYED FACILITIES TO ESTIMATE
                       RELEASES AND CONTROL EFFICIENCIES
                                                       Number of Facilities
Reference                                                 Using Reference

TRI Reporting Form R and Instructions                           141
EPA 560/4-88-005

The Emergency Planning and Community Right-to Know               40
Act, Section 313 (40 CFR 372)
EPA 560/4-88-001

Estimating Releases and Waste Treatment                          38
Efficiencies for the TRI
EPA 560/4-88-002

Title III Section 313 Release Reporting Guidance,                23
EPA 560/4-88-004 a through q, Estimating Chemical
Releases

Compilation of Air Pollution Emission Factors,                   17
AP-42

Toxic Chemical Release Reporting Proposed Rule,  52               16
FR 21152, June 4, 1987

Industry Trade Association Materials                             15

Privately Sponsored Seminar Materials                            14
                                     2-14

-------
and paper industry frequently used NCASI (National Council of the Paper




Industry for Air and Stream Improvements) emission factor data (9 of 14




facilities in the survey). By contrast, the chemical industry rarely used




trade association data (6 of 44 facilities in the survey).









2.6.2       Threshold Determinations









            Threshold determinations and release estimates were reviewed




separately to identify the frequency, magnitude, and sources of errors in




these two areas.  Site surveyors followed the stepwise approach described in




the Section 313 procedures for completing threshold determinations and release




estimates.  In following the Form R reporting instructions, facilities must




first assess which chemicals are manufactured, processed, or otherwise used in




excess of appropriate thresholds.   Facilities must then estimate and report




all releases of these chemicals exceeding a threshold to the environment, and




all off-site transfers of these chemicals in wastes to other locations.









            To identify errors in the threshold determinations, site surveyors




looked for problems in a facility's documentation and, on the plant tour, site




surveyors looked for evidence of chemicals that were reported but need not




have been reported, and for evidence of chemicals that were not reported but




should have been reported.  Each facility's documentation was reviewed to




track the decision process used to determine whether a chemical should have




been reported.  Furthermore, site surveyors used all available documentation




to recalculate threshold estimates for reported chemicals and for chemicals
                                     2-15

-------
present but not reported to verify the accuracy of facility personnel




calculations.








            As part of the process of collecting data on threshold determina-




tions, site surveyors assessed the approaches commonly used by facility




personnel to make threshold determinations.  Table 2-6 lists the eight




approaches most commonly identified as the mechanism by which facility person-




nel made their threshold determinations.  The most common approach, the




purchase and inventory approach, involves calculating the total amount of a




chemical purchased during the calendar year, taking into account changes in




inventory levels.  This approach usually includes using Material Safety Data




Sheets (MSDS) and other data to determine the concentration of chemicals in




the raw material.








            In general, the other seven approaches, all of which are based on




engineering calculations, were used less frequently than the purchase and




inventory approach.   However, there were certain situations in which these




other seven approaches typically were used.  For example, the emission factor




approach was used frequently by facilities in the pulp and paper industry to




determine the quantity of certain volatile chemicals that are emitted to the




atmosphere immediately upon manufacture.  Also,  facilities commonly used the




mass balance approach to determine the quantity of sodium sulfate manufactured




as a by-product in wastewater treatment systems.
                                     2-16

-------
                                   TABLE 2-6

                     APPROACHES COMMONLY USED BY FACILITY
                  PERSONNEL TO MAKE THRESHOLD DETERMINATIONS*
                                  Approaches

                            Purchase and inventory records
                            Engineering calculations based on:

                                Emission factors
                                Mass balance
                                Process recipes
                                Monitoring data
                                Production data
                                Hazardous waste manifests
                                Assumptions
*Table  G-2  in Appendix G provides tables which define these approaches and
show their frequency of use among the surveyed facilities.
                                     2-17

-------
2.6.3       Release Estimates








            To identify errors in the release estimates,  site surveyors




recalculated the releases during the site visits using the best data avail-




able.  Site surveyors always recalculated releases using the same general




approach as the facility used.  (Guidance materials for Section 313 identify




four approaches for estimating releases:  use of monitoring data, emission




factors, mass balance, and engineering calculations.)   Whenever the site




surveyor's experience and training indicated that a calculation approach




different than that used by the facility personnel was appropriate, the




surveyor attempted to obtain the data needed to calculate releases using the




more appropriate approach.  In many such instances, data were not readily




available during the site visit to recalculate releases using the alternative




approach.  In the few cases where site surveyors were  able to recalculate




releases using alternative approaches, the site surveyors were able to assess




the reasonableness of the estimation techniques used by facility personnel.




Throughout the site survey program, site surveyors only reviewed release




estimates for those chemicals and media where it was appropriate to do so.









            The surveyors quantified all numerical differences between the




facility's release estimates and the recalculated values, even in instances




where surveyors identified only small differences.  As discussed later, these




numerical differences were used to quantitatively assess the accuracy of the




total aggregate releases contained in the 1987 TR1 database.
                                     2-18

-------
            To gain a sense of how frequently facilities made release estimate




errors,  site surveyors examined release estimates using an order-of-magnitude




screen (i.e.,  the percentage of release estimates in which facilities under-




estimated or overestimated releases by more than a factor of ten).   This broad




range (one order-of-magnitude) of difference was selected to identify error




frequency in the first-year Form R reports because the limitations  on




available data and the variety of permissible estimation techniques suggest




that facility release calculations are of limited accuracy.   Reduction in the




frequency of order-of-magnitude release errors in future reporting  data should




provide a good indication of both greater understanding of revised  guidance




materials and improved estimation techniques.








2.7         Data Management and Data Quality









2.7.1       Data Management System








            To manage the sizable volume of data generated by the site survey




effort,  Radian used PC-based SAS™  software  as a  database management  tool




because of its powerful statistical capabilities.








            Since this database has many intended uses, Radian took several




steps to ensure its high quality.  As discussed earlier, the first quality




assurance step was to thoroughly and consistently train the site surveyors




before any site visits were performed.  After each site visit was completed,




the office coordinator performed a quality assurance review of the survey




instrument before submitting the completed forms to the quality assurance  (QA)






                                     2-19

-------
team.  The QA team performed a second review to assure consistency in .the work




performed by the various site survey teams.  The twice-reviewed survey instru-




ment data then were entered into the database.  The final step in the database




QA process was to verify the accuracy of the data entry effort.  Project staff




randomly selected 40 of the 156 survey instruments (25%) for a data entry




verification test.  The results from this test showed an error rate of less




than 0.4%.








2.7.2       Data Weighting








            To allow EPA to assess the impact of the site visit program




results on the TRI database, weighting factors were applied to the site visit




data.  These factors or "weights" represent the number of facilities in the




TRI database corresponding to each of the surveyed facilities.  The weights




compensate for the sample selection probabilities introduced by the sample




selection process, in which chemical industry facilities and large-quantity




emitters were purposely over-sampled.









            As discussed previously, Abt used a two-stage sample selection




procedure, involving geographic clusters and four domains of interest, to




select facilities for the site visit program.  Table 2-7 shows the target




universe for the site visit program distributed among the four domains.  The




universe represents 17,622 of the 18,000 facilities in the TRI database (98%),




corresponding to 53,385 of the 68,500 chemical reports in the TRI database
                                     2-20

-------
                                   TABLE 2-7




                  TARGET UNIVERSE FOR THE SITE VISIT PROGRAM*







Domain                                              Number  of  Facilities




Chemical Industry Facility                                   3,610




Nonchemical Industry Facility                               14.012




TOTAL                                                       17,622
Large-Quantity Emitter                                       2,206




Small-Quantity Emitter                                      15.416




TOTAL                                                       17,622











AThis universe is  further "discussed in Appendix B.
                                     2-21

-------
(78%).3  Inferences can be applied with a known degree  of sampling error to

this universe.  The potential error associated with applying the site visit

program results to the remaining 2% of facilities in the TRI database,

representing 22% of chemicals in the TRI database,  cannot be- quantified from

the data generated by this study.



2.7.3       Survey Bias



            To place boundaries on the accuracy of data collected and analyzed

during the site visit program, project staff analysts reviewed the sources of

bias in the program approach and assessed their implications.   The following

three primary areas in the program approach were  identified as producers of

bias:
                  Sample selection;
                  Surveying; and
                  Data reduction and analysis.
            A sampling bias was introduced to the program results because the

sample for this program was limited to approximately 150 facilities.

Table 2-8 presents the expected precision of results at the 90% confidence

level for the total sample and for the four domains of interest.  As an

example, with 90% confidence, the actual population percentage for large-

quantity emitters is expected to be between 40.3% and 59.7% when the sample
     3At  the time the current study was  initiated, approximately 68,500 chemical
reports had been submitted to EPA by approximately 18,000 industrial facilities.
These data were entered in the 1987 TRI database.  Subsequent to this study, EPA
completed the TRI database, which  now  includes data  from approximately 70,000
chemical reports from approximately 19,000 facilities.

                                     2-22

-------
                                   TABLE 2-8

              EXPECTED PRECISION OF THE SAMPLE SELECTION PROCESS
                       FOR THE FOUR DOMAINS OF INTEREST*
Domain

Chemical Industry Facilities
Nonchemical Industry Facilities

Large-Quantity Emitters
Small-Quantity Emitters

TOTAL SAMPLE
Expected Usable Level of Precision

              ± 13.3%
              ±  8.0%

              ±  9.7%
              ±  9.7%

              ±  6.9%
ADesign precision is discussed further in Appendix B.   The precision levels
shown are at the 90% confidence level.
                                     2-23

-------
percentage equals 50%.  Thus, if 50% of large-quantity emitters report using a

specific calculation approach (such as mass balance),  the odds are 90 in 100

that the population percentage value will be between 40.3% and 59.7%.  Sample

selection precision levels are discussed further in Appendix B.



            A surveying bias was introduced to the program results because

approximately 40 surveyors were used to conduct the site visits.   However,

Radian took several steps to minimize the impact of personal bias in the

surveying process, including using a standardized survey instrument, thorough-

ly and consistently training each site surveyor, and carefully checking the

completed survey instruments before data entry.



            A data reduction and analysis bias also was introduced to the

program results due to the use of four assumptions intended to simplify survey

data reduction and analysis:
                  Incorrectly-omitted releases were quantified using the
                  lowest range suggested by the reporting instructions (1-499
                  pounds) unless sufficient data were available to calculate a
                  larger release;

                  Review of the 22% of the Form R chemical reports (those from
                  facilities submitting more than 15 reports) included in the
                  TRI database, but not addressed by this program, would
                  produce results similar to those documented by the site
                  survey program;

                  A quantitative assessment of the impact of nonresponding
                  facilities on the TRI database would not change the site
                  survey program's conclusions; and

                  The total aggregate releases from incorrectly-included
                  chemicals were equal to those from incorrectly-omitted
                  chemicals.
                                     2-24

-------
            This latter assumption, concerning aggregate releases for chemical

reports that were not submitted by facility personnel, likely underestimates

total release quantities, because release quantities associated with incor-

rectly included chemicals tend to be small, while release quantities

associated with omitted chemicals could be higher.



            In general, these program biases likely do not introduce suffi-

cient bias to substantially change the conclusions discussed later in this

report.  However, a limited analysis of the quantitative impact of these

biases on the program's conclusions should verify whether there is any impact.



2.8         Wrap-up Meeting



            After the site visits were completed, Radian held a debriefing

wrap-up meeting with all office coordinators at its Herndon, Virginia office.

Each office coordinator previously had conducted a debriefing meeting with

his/her site surveyors.  During the wrap-up meeting, attended by  representa-

tives from Abt and EPA's Office of Toxic Substances, office coordinators

discussed the qualitative observations from the site visits and provided input

for improving future efforts.  The agenda for the wrap-up meeting, and a

memorandum summarizing the items discussed, are provided in Appendix F.  The

two key qualitative observations discussed were that:
            (1)   Radian's site visits relied heavily on documentation and
                  insight provided by facility personnel (the one- or two-day
                  site visits provided only enough time to briefly tour
                  facilities and to review readily available records); and
                                     2-25

-------
            (2)   The threshold determination and release estimate results
                  were limited by the quality and the amount of information
                  facility personnel had readily available for review.
Overall, the facilities' documentation was often found to be poor, which made

the task of assessing the quality of the reported release quantities more

difficult.
                                     2-26

-------
3.0         RESULTS



            This section presents the site visit program results and their

implications to the Toxic Release Inventory (TRI) database.  The analysis of

the results focuses on errors in threshold determinations and release

estimates to allow EPA to assess data quality and determine where improved

reporting guidance would be most effective.  This section provides a summary

of the results with regard to the following questions:
            (1)   Threshold Determinations.  Did facilities include the
                  Section 313 chemicals they were required to report?  If
                  facilities made errors in threshold determinations, what
                  were the reasons for the errors?

            (2)   Release Estimates.  For chemicals that were correctly
                  included, what is the impact on the TRI database of the
                  errors facilities made in their release and off-site trans-
                  fer estimates?  If facilities made errors in their release
                  and off-site transfer estimates, what were the reasons for
                  the errors?  Given the data readily available for making
                  release estimates, did facilities use reasonable release
                  estimation techniques?

            (3)   Domains of Interest.  Did the Form R chemical reports
                  submitted by chemical industry facilities contain fewer
                  errors than those submitted by nonchemical industry facili-
                  ties?  Did the Form R chemical reports submitted by facili-
                  ties where large-quantity emissions were reported contain
                  fewer errors than those submitted where small-quantity
                  emissions were reported?
            Because Form R preparers went through two major steps, threshold

determination and release estimates, the analysis of error in Form R chemical

reports for these two areas were reviewed separately.  Four domains of

interest (chemical industry facilities, nonchemical industry facilities,

large-quantity emitters, and small-quantity emitters) were selected to help


                                      3-1

-------
EPA determine where to focus future resources for improving the TRI database.

The survey data upon which these summary results are based are shown on tables

in Appendix G.



3.1         Threshold Determination Results



            Section 313 requires facilities to determine which of their

chemicals are manufactured, processed, or otherwise used in excess of

.appropriate thresholds.  These chemicals must be reported to the Agency using

a Form R chemical report.  The threshold determination actually consists of

three steps:
            (1)   Assessing which activities are appropriate for each
                  chemical;

            (2)   Computing the amount of chemical manufactured, processed, or
                  otherwise used during the calendar year; and

            (3)   Determining if this quantity exceeds the applicable
                  threshold.
A further complexity of Section 313 is that facility personnel may exclude

some or all of a chemical's activity if it meets one of several exemptions.

(The exemptions include de minimis concentrations, laboratory activities,

processing and use of articles and certain use provisions.)  Each of these

exemptions may be incorrectly applied by facility personnel, and may cause

them to either incorrectly include or incorrectly omit a chemical from its

proper reporting.  Tables G-l through G-4 in Appendix G contain the threshold

determination survey results, which are summarized and discussed in the

following section.

                                      3-2

-------
3.1.1       Frequency of Errors in Threshold Determinations



            Radian examined threshold determinations made by the surveyed

facilities to determine whether the facilities reported the Section 313

chemicals they were required to report.   This examination resulted in four

threshold determination categories:  two that represent correct threshold

determinations and two that represent incorrect threshold determinations.

These four threshold determination categories are:
                  Chemical was reported and should have been reported
                  (correctly included);

                  Chemical was reported but need not have been reported
                  (incorrectly included);

                  Chemical was not reported and should not have been reported
                  (correctly omitted); and

                  Chemical was not reported but should have been reported
                  (incorrectly omitted).
            Facilities generally included the Section 313 chemicals they were

required to report and omitted the chemicals they should not have reported.

However, error rates observed in the survey results, when applied to all

reporting facilities, suggest that approximately 6,000 chemical reports were

incorrectly omitted and that approximately 8,000 chemical reports were

incorrectly included in the 1987 TR1 database.4   These overall  findings  are

displayed graphically in Figure 3-1.  As shown in Figure 3-1, less than  40%

of the cases where SARA 313 chemicals were manufactured, processed, or
     4This  finding is exclusive  of the Form  R reports that  should  have been
submitted by nonresponding facilities.

                                      3-3

-------
Incorrectly-Omitted
 Chemical Reports
        3%
                                Corrnctly-Omitted
                                Chemical Reports
                                       59%
Correctly-Included
 Chemical  Reports
       34%
                            Incorrectly-Included
                              Chemical  Reports
                                     4%
                                                           Note:
                                                            The  survey identified
                                                            approximately 60. 000
                                                            correctly-included
                                                            chemicals in the
                                                            1987 TRI database
            Figure 3-1.   Threshold  Determination Results:
        Chemicals Covered Under SARA Title III.  Section 313

-------
otherwise used by facilities submitting Form R reports to EPA were required to




be reported under Section 313.  The remaining 60% of the cases were excluded




through the threshold and exemption process described earlier.








            The survey results also indicate that 94% of the facilities in the




TRI database were correct in submitting at least one of the Form R reports




they submitted.  Five percent of the facilities that did submit at least one




Form R report need not have reported at all under Section 313.  A small group




(1%) need not have reported any of the chemicals they did report, but should




have reported other chemicals they did not report.








3.1.2       Sources of Errors in Threshold Determinations








            The sources for the two types of threshold determination errors




are described below.








Incorrectly-Included Chemical Reports








            The reasons for facilities incorrectly including chemicals on the




Form R reports are listed in Figure 3-2.  The most common reason for incor-




rectly-included chemicals resulted from facility personnel reporting a chemi-




cal despite having calculated a quantity below, an applicable threshold.




Facility personnel probably erred by failing to understand the concept of the




threshold determination for this program, or perhaps by wanting to be




conservative (i.e., reporting a chemical because the calculated quantity was




near the applicable threshold).






                                      3-5

-------
Incorrectly-Omitted
  Chemical Reports
         3X
                              Correctly-Omitted
                              Chemical  Reports
                                     59X
Correctly-Included
 Chemical  Reports
        34X
                                                            Incorrectly-Included
                                                              Chemical  Reports
                                                                      4X
              REASONS FOR  INCORRECTLY  INCLUDING CHEMICALS


             •  Facility personnel  reported  chemical despite
                calculating a quantity below applicable threshold

             •  Facility personnel misinterpreted a  chemical  qualifier

             •  Facility personnel miscalculated a chemical quantity

             •  Facility personnel misinterpreted an exemption clause

             •  Facility personnel misinterpreted the list of chemicals
                covered  by  Section 313

             •  Facility personnel misclassified  a chemical activity

             •  Facility personnel incorrectly assumed thresholds
                were exceeded, without supporting  documentation
                                Figure  3-2
                                        3-6

-------
            Another common error involved misinterpreting chemical qualifiers.

For example, isopropyl alcohol was. frequently reported as an "otherwise used"

chemical although reporting is required for isopropyl alcohol only when it is

manufactured by the strong acid process.   Other instances of misinterpreted

qualifiers involved metals (aluminum, vanadium, and zinc),  where reporting is

required only when the metals are used in a fume or dust form.   Numerous

facilities reported use of such metals but did not handle the material as fume

or dust.  Strengthening the presentation of these qualifiers in the Form R

instructions and guidance documents should help eliminate this problem



            Another frequent type of error facilities made in threshold

determinations was in miscalculating quantities.  This type of error typically

resulted from several sources:
                  Estimating the amount of a constituent chemical in a raw
                  material instead of using composition data available in the
                  Material Safety Data Sheet;

                  Overlooking one area of chemical activity at a facility;

                  Assuming that a chemical was a by-product of processing or
                  an impurity in the raw material without supporting documen-
                  tation; or

                  Calculating threshold amounts based on amounts of chemicals
                  circulated in closed-loop systems instead of only the
                  amounts of makeup chemicals added to the systems during the
                  year.
            The other types of errors listed in Figure 3-2 occurred less

frequently than those discussed above.  Generally, these errors involved

facility personnel misinterpreting a specific item in the reporting

instructions.  such as an exemption clause, misreading the list of chemicals

                                      3-7

-------
reportable under Section 313, or not understanding the definitions of the

activity classifications.  Although activity classifications5-were frequently

misinterpreted (approximately 20% of the activity classifications were

reported incorrectly, especially the manufacturing and processing classifica-

tions) , this error rarely led to a chemical being incorrectly included in the

TRI database (or omitted from the TRI), since the quantity involved usually

exceeded (or was below) the thresholds for both the correct and incorrect

activity classifications.



            In a few instances, facility personnel assumed thresholds were

exceeded and made no calculations.   Some of these chemicals need not .have been

reported.  In these cases, the preparer's assumption regarding the chemical

quantity was simply incorrect.  This was the least common type of error

identified.



Incorrectly-Omitted Chemical Reports



            Application of the site survey results to the TRI database indr-

cated that approximately 6,000 chemical reports were incorrectly omitted from

the TRI database (i.e., chemicals that were not reported-but were manu-

factured, processed, or otherwise used above an applicable threshold).  These
     'Facility  personnel  were  required to  classify  each  reported  chemical
activity as  manufactured,  processed, and/or  otherwise used depending  on its
specific application or applications at the facility.

                                      3-8

-------
chemical reports were incorrectly omitted by 3,000 facilities in the TRI

database.6  The  reasons  for  these omissions  are  shown in  Figure  3-3.



            The most common reason for incorrectly omitting chemicals was

facility personnel overlooking a chemical activity.  The following chemical

activities were overlooked:
                  Manufac tur ing by-produc ts;
                  Wastewater treatment by-products;
                  Wastewater treatment chemicals;
                  Cleaning chemicals; and
                  Miscellaneous chemicals.
Approximately 75% of the incorrectly-omitted chemical reports were omitted

because one or more of these activities was overlooked.



            Other reasons for incorrectly-omitted chemical reports include

miscalculating threshold quantities, misclassifving use activities, and

misinterpreting specific reporting instructions.  Also,  about 10% of the

exemptions claimed by industry were found to be incorrect.  Most of these

exemption errors occurred in the "article" and "certain uses"7  category.   The

"de minimis" and "laboratory" exemptions were generally applied correctly in

calculating thresholds.
     6This  finding does  not  include  nonresponding facilities.


     7The "certain uses" exemption includes use as a structural component of the
facility;  use in  routine janitorial or facility grounds maintenance; personal
use by employees  or other persons;  use of products containing toxic chemicals
for the purpose of maintaining motor vehicles operated by the facility; or use
of toxic chemicals contained in intake  water  (used for processing or noncontact
cooling) or in intake air (used either as compressed air or for combustion).

                                      3-9

-------
Incorrectly-Omitted
 Chemical Reports
        3%
                             Correctly-Omitted
                              Chemical  Reports
                                    59X
                             Correctly-Included
                              Chemical Reports
                                     34X
Incorrectly-Included
  Chemical  Reports
         4X
               REASONS FOR  INCORRECTLY OMITTING CHEMICALS


            •  Facility personnel overlooked a chemical activity

            •  Facility personnel miscalculated a chemical quantity

            •  Facility personnel misclassified a chemical activity

            •  Facility personnel misinterpreted an exemption clause
                              Figure  3-3
                                       3-10

-------
3.1.3       Summary








            The results of the site survey program indicated that industry




reported most of the chemicals they were required to report.  For chemical




reports incorrectly omitted, the most common reason was facility personnel




overlooking a chemical activity.  For chemical reports incorrectly included,




the most common reason was facility personnel reporting a chemical despite




calculating a quantity below an applicable threshold.  Finally, for the most




part, facilities accurately omitted a large number of Section 313 chemical




reports that need not have been reported (due to thresholds and exemptions).








3.2         Release Estimation Results








            Section 313 requires facility personnel to report the total




quantity of each reportable chemical released to the environment and trans-




ferred in wastes off-site during the calendar year.  Releases to the environ-




ment include emissions to the air, discharges to surface waters, and releases




to land and underground injection wells.  Off-site transfers include chemicals




sent to publicly-owned treatment works (POTW) and other locations, such as




off-site landfills.  Section 313 requires facility personnel to use all




readily available data to estimate release quantities as accurately as




possible.








            Site surveyors assessed release estimates by recalculating the




releases using the same approach the facility personnel used, and by using




alternative approaches if they were available.  The database compiled during






                                     3-11

-------
the site visit program contains data for instances where the site surveyor




calculated a release estimate different than the estimate the facility




reported.  'These differences were used to assess the overall TRI database




release accuracy.








            The following subsections focus on the overall accuracy of the TRI




database and on the sources of errors made by facilities in their release




estimates.  Also, the quality of the release estimation techniques used by the




facilities is assessed.  Tables G-5 through G-8 in Appendix G contain the




release estimate survey results, which are summarized and discussed below.








3.2.1       Accuracy of Release Estimation








            The site visit program data showed that the aggregate 1987 TRI




database release estimates reported by industry generally were accurate, but




were slightly underestimated.  Table 3-1 provides a summary of toxic chemical




release quantities by release medium.  This table includes release quantities




as reported in the 1987 TRI and quantities that have been corrected based on




results from the site visit program.








            The data also showed that some release estimates of sizable




quantities were over-reported or under-reported. .As a result,  the accuracy of




the aggregate TRI release estimates is,  to some extent, only "apparent"




accuracy caused by the cancellation effects of summing the overestimated and




underestimated quantities.  The data showed that the total quantity over-




reported and under-reported were less than 10% of the total release quantity,






                                     3-12

-------
                                   TABLE 3-1

                 SUMMARY OF TOXIC CHEMICAL RELEASE QUANTITIES
Release Medium

Fugitive Air

Stack Air

Discharges' to Water

Underground Injection

On-site to Land

POTW

Other Off-site Transfers

TOTAL
                              Quantity of Chemical
                        Quantity of Chemical
                       Released, Adjusted by
                              Released, as Reported   the Results of the Site
in the 1987 TRI
(billion pounds)

     0.8

     1.8

     9.6

     3.2

     2.4

     2.2
    22.5*
 Visit Program
(billion pounds)

      0.8

      1.9

      9.9

      3.2

      2.7

      2.0

      2J.

     23.0*
Quantities shown do not sum to total due to rounding effects.
                                     3-13

-------
and that there were slightly more under-reported release quantities than over-

reported release quantities.



            The site survey program data analysis included a separate assess-

ment of the frequency of errors for zero and non-zero release estimates for

the 62,000 correctly-included chemical reports in the 1987 TRI (i.e., chemi-

cals that were reported and should have been reported).   Non-zero release

estimates are examined separately from zero release estimates because non-zero

release estimate errors give a better indication of the  type of estimation

errors made by facility personnel.  In contrast, facility personnel who

incorrectly omitted a release quantity either overlooked a specific type of

release or misinterpreted a specific instruction, rather than estimating a

quantity incorrectly.



            Table 3-2 shows the number and percent of non-zero chemical

release estimates (by release medium) in the 1987 TRI that had what is called

a "large magnitude" release estimate error.8  Overall, roughly 84%  of the  non-

zero release estimates in the 1987 TRI were small magnitude errors correct

within an order-of-magnitude. Reducing the screening threshold to a factor of

two only decreased this value from 84% to 77%.  For the  chemical releases in

error by greater than one order of magnitude, Table 3-2  shows higher error

rates for discharges to water, on-site releases to land, and off-site
     'Reference  to  the  frequency of  release estimate  errors  in  this  report
includes only  large magnitude  errors.   Large  magnitude release  errors  were
defined as non-zero release quantities more than one order-of-magnitude different
than the site surveyor's estimated value, or  release quantities that should not
have been reported.   The large magnitude error screen was used  to assess the
general accuracy of non-zero  releases  since  release  calculations usually were
based on limited data that did not lend itself to very precise estimates.

                                     3-14

-------
transfers to POTV than for air releases, discharges to underground injection




wells, or other off-site transfers..  Most of these errors were due to




incorrectly-reported releases, i.e., those that should not have been reported.
            Further analysis of the site survey data showed that almost half




of the facilities represented in the 1987 TRI made at least one large mag-




nitude error in calculating non-zero chemical release estimates.  Acids,




bases, and volatile organic compounds were the chemical groups most frequently




containing such release estimate errors.  The previous finding that most non-




zero release estimates were correct within an order-of-magnitude seems to




contradict this finding concerning number of facilities.  However, the data




showed that facilities reporting multiple chemical releases made only these




large magnitude errors in a small number of their release estimates.








            The finding that almost half of the facilities made at least one




large magnitude error in calculating non-zero chemical release estimates seems




at first to be an apparent discrepancy with the overall finding that the




aggregate 1987 TRI database release estimates reported by industry generally




were accurate.  However, a detailed review of these instances of error




revealed that these large magnitude errors usually involved small quantities




of chemical releases.








            Table 3-3 shows the number and percent of chemical releases (by




release medium) in the 1987 TRI that should have been reported as a non-zero
                                     3-15

-------
                                   TABLE 3-2

   FREQUENCY OF ERRORS IN NON-ZERO RELEASE ESTIMATES  FOR EACH RELEASE MEDIUM
Release Medium
 Number of Non-Zero
Chemical Releases in
      1987 TRI
   Number of Non-Zero
Chemical''Releases With a
 Large Magnitude Error*
Fugitive Air
Stack Air
Discharges to Water
Underground Injection
Ons ite to Land
POTW
Other Off -site Transfers
TOTAL
29,000
33.000
3,500
450
2,200
13,000
22.000
103,000
3
2
1


*
3
16
,800 (13%)
,200 (7%)
,300 (37%)
66 (15%)
790 (36%)
,300 (33%)
.700 ^17%}
,000 (16%)
ALarge magnitude release estimate errors were defined as  non-zero release
quantities more than one order-of-magnitude different than the site surveyor's
estimated value, or release quantities that were reported but should not have
been reported.
                                     3-16

-------
                                   TABLE 3-3

     FREQUENCY OF  ERRORS  IN ZERO RELEASE ESTIMATES  FOR EACH RELEASE MEDIUM*
Release Medium
Number of Chemical
 Releases Omitted
  From 1987 TRI
     Number of
 Chemical Releases
Erroneously Omitted
   From 1987 TRI
Fugitive Air
Stack Air
Discharges to Water
Underground Injection
Ons ite to Land
POTW
Other Off-site Transfers
TOTAL
30,000
25,000
58,000
62,000
60,000
47,000
35.000
317,000
4,200 (14%)
3,900 (16%)
400 (<1%)
0 (0%)
300 (<1%)
1,600 (3%)
4.000 m%>
14,400 (5%)
"Number of zero chemical releases was based on 60,000 correctly-included
chemical reports in the 1987 TRI.
                                      3-17

-------
value but were erroneously omitted from the TRI.8  Overall,  approximately 95%

of the chemical releases omitted from the TRI for correctly-included chemicals

were correctly omitted.  Most of the 5% of the erroneously-omitted chemical

releases pertained to air releases and off-site transfers to POTW and other

locations as seen in Table 3-3.



            As stated above, most release estimate errors were due to incor-

rectly-reported and incorrectly-omitted releases, and resulted from facilities

misinterpreting instructions or overlooking chemical releases.  Most estimates

not affected by these errors were accurate to within one order-of-magnitude,

and most were within a factor of two of the site surveyors'  best estimates.

This finding confirms the previous finding that the aggregate 1987 TRI data-

base release estimates generally were accurate.



3.2.2       Sources of Errors in Release Estimates



            Four general types of errors in the release estimates were iden-

tified in the survey data:


            •     Misinterpretation of reporting instructions;
            •     Use of incomplete information;
            •     Insufficient documentation to support release estimates; and
            •     Calculation and transcription errors.



Misinterpretation of reporting instructions and use of incomplete information

were the most common errors.  Table G-6 in Appendix G provides a summary of
     "These  zero chemical releases include all release media reported as zero or
not applicable (N/A).

                                     3-18

-------
Che frequency of each type of error at the surveyed facilities.  The data

showed that facilities in the following SIC codes had the largest error rates

for release estimates:
                  20-Food and kindred products;
                  22-Textile mill products; and
                  34-Fabricated metal products.
Misinterpretation of Reporting Instructions



            Where misinterpretation of reporting instructions occurred, it

usually concerned errors related to acid/base neutralization, recycle/reuse of

chemicals transferred off-site, metal compound releases, or the appropriate

use of the de minimis exemption.  The acid/base neutralization error occurred

when facilities neglected to include the 100% treatment efficiency for an acid

or base in their release estimates.  This error resulted in over-reporting of

acid/base releases (470 million pounds/year).  If the pH of a liquid release

is maintained between 6 and 9 standard pH units, the reporting instructions

specify that ionic acids and bases (HC1, HN03,  H2S04,  and NaOH)  are completely

(i.e., 100%) neutralized and should be reported as a zero release.



            The recycle/reuse off-site error occurred when 'the surveyed

facilities reported a release for a chemical that was transferred off-site for

recycling or reuse as a fuel.  This error also resulted in over-reporting of

releases (180 million pounds/year).  The reporting instructions specify that

chemicals transferred off-site for recycle or reuse should be reported as a

zero release.


                                     3-19

-------
            The other cases of misinterpretation occurred only rarely.




Facilities misinterpreting the metal compound instructions erred by reporting



the release of a metal compound rather than the parent metal in the




compound(s).  The reporting instructions specify that facilities should




include releases of parent metals only when considering releases for metal




compounds.  This error also resulted in over-reporting of releases (60,000



pounds per year).   Facilities incorrectly interpreting the de minimis




exemption applied this exemption to their waste streams,  contrary to the




reporting instructions.  The de minimis exemption states that if a listed




toxic chemical is present in a mixture, excluding waste streams, at a



concentration below a specified de minimis level, the chemical in that mixture




does not have to be factored into threshold and release determinations.  This




error resulted in under-reporting of releases (28 million pounds per year).








Use of Incomplete Information








            In cases where facility personnel erred in estimating releases,




they typically overlooked three sources of information:  facility data,




chemical property data, and emission factor data.  Facility data often over-




looked included waste manifests, process information, and monitoring data.




Where errors were found, facility personnel also frequently overlooked chemi-




cal property data, such as the volatility, solubility, and fate of chemicals




in certain media.   Only rarely did facilities overlook published emission




factors for a process or treatment technology.  These errors resulted in




under-reporting of releases (220 million pounds per year).






                                     3-20

-------
Other Errors








            In cases where the facility had insufficient documentation to




support a release estimate, surveyors typically could not readily locate




sufficient data to quantify a different release estimate.  However, where



surveyors attempted alternative estimates based on engineering judgment, these



errors were estimated to result in under-reporting of releases (20 million




pounds per year).   The survey also revealed arithmetic calculation errors.  and




transcription errors (facility personnel incorrectly recording release




estimates on the Form R reports).  Overall, arithmetic errors resulted in



under-reporting of releases (832 million pounds per year), while transcription




errors resulted in over-reporting of releases (76 million pounds per year).








            To assess whether certain errors were typical for a release




medium, the types of errors made for the 12 largest-quantity errors in each




release medium were examined.  The results of this review are presented in




Table 3-4.  As this table shows, most air release and on-site land release




errors were due to facility personnel overlooking some type of information,




while water release errors (on site and to POTW) usually were due to facility




personnel misinterpreting acid/base neutralization or the manufacture of




sodium sulfate.   Most other off-site transfer errors were due to facility




personnel misinterpreting items in the reporting instructions or making




transcription errors.
                                     3-21

-------
                                   TABLE  3-4

                         TYPES OF ERRORS MADE FOR THE
            TWELVE LARGEST-QUANTITY ERRORS IN EACH RELEASE MEDIUM*
Misinter-
preted the
Release Reporting
Medium Instructions
Fugitive Air
Stack Air
Discharge to
Water
Underground
Injection
Ons ite to
Land
POTW
Other Off-
0
0
5
2
0

9
4
Used
Incomplete
Information
9
9
1
1
6

2
3
Made
Calculation
Errors
2
2
2
1
3

1
0
Made
Transcription
Errors
1
1
4
1
3

0
5
site Transfer
AAppendix I provides detailed information on the  magnitude of the individual
errors summarized in this table.
                                     3-22

-------
3.2.3       Release Estimation Techniques



            To assess the quality of the release estimation techniques used by

facilities, instances where more accurate release estimation techniques were

identified by Radian surveyors were examined.  As discussed earlier, the four

release estimation techniques outlined under Section 313 are:
            (1)   Monitoring data;
            (2)   Emission factors;
            (3)   Mass balance; and
            (4)   Engineering calculations.
Appendix H lists the few cases (less than 10% of facilities) in which alterna-

tive estimation techniques were identified among the surveyed facilities.



            In most cases, the release estimation techniques used by industry

were reasonable.  A few clear examples of more accurate techniques were found;

however, even using these techniques, the magnitude of error was not substan-

tially different than that experienced for the entire TRI database.  These few

cases showed that the engineering judgment technique may lead to under-

estimated releases and that the emission factor technique may be a more

reasonable estimation approach for these situations.



            As discussed earlier, the site visits were conducted in one or two

days, thus limiting Radian's ability to obtain the additional data necessary

to compute release quantities using alternative estimation techniques.  A more

exhaustive audit, possibly combined with the ability to take some simple

measurements, would likely increase the possibility of locating sufficient

                                     3-23

-------
information to calculate  releases using alternative estimation techniques.



However, with the SARA Title III, -Section 313 condition that additional




measurements and monitoring are not required to prepare Form R reports, the




cases in which a preferred alternative estimation method is available may not




be numerous, even with a more extensive survey or audit.








3.2 .'4       Summary








            Overall, facilities estimated total release quantities with




reasonable accuracy; however, there were numerous areas where better guidance




would improve data quality.  Alternative release estimation techniques were




difficult to apply given the limited amount of readily available information.








3.3         Domains of Interest








            One of the goals for the site survey program was to assess whether




there were trends in data quality based on the type of industry (chemical




versus nonchemical) or the reported size of the emitter (large versus small-




quantity).  To perform this assessment, the chemical-specific error rates for




threshold determinations and release estimates were compared for the four




domains of interest.  The data for this assessment are presented in Tables 3-5




and 3-6.








            As shown in these tables, facilities'in the chemical industry




(primary SIC Code 28), and facilities that reported large-quantity emissions




(large-quantity emitters) consistently made fewer errors than did nonchemical






                                     3-24

-------
                                   TABLE 3-5

                      THRESHOLD DETERMINATION ERROR RATES
                      WITHIN THE FOUR DOMAINS OF INTEREST
                                                        Threshold
                                                      Determination
Domain*-8                                               Error Rate0

Chemical Industry Facilities                              5.1%

Nonchemical Industry Facilities                           8.4%

Large-Quantity Emitters                                   7-1%

Small-Quantity Emitters                                   7.7%



*Chemical industry facilities were defined by Primary SIC Code 28.

BLarge-  and small-quantity emitters were differentiated by the quantity of
emissions reported on the Form R report by the facility.  Large-quantity
emitters reported a release above the 90th percentile for a particular release
medium.  (See page B-3 in Appendix B.)

^Threshold determination errors were defined as incorrectly-included chemicals
and incorrectly-omitted chemicals.
                                     3-25

-------
                                   TABLE 3-6

                         RELEASE ESTIMATE ERROR RATES
                      WITHIN THE FOUR DOMAINS OF INTEREST
                                                         Release
                                                        Estimate
Domain*-8                                               Error Ratec

Chemical Industry Facilities                              23.0%

Nonchemical Industry Facilities                           27.4%

Large-Quantity Emitters                                   19.5%

Small-Quantity Emitters                                   29.3%



AChemical industry facilities were defined by Primary SIC Code 28.

BLarge-  and small-quantity emitters were differentiated by the quantity of
emissions reported on the Form R report by the facility.  Large-quantity
emitters reported a release above the 90th percentile for a particular release
medium.   (See page B-3 in Appendix B.)

GRelease estimate errors were defined as release quantities more than one
order-of-magnitude different than the Radian QA reviewer's estimated value.
                                     3-26

-------
SIC codes and small-quantity emitters, respectively.  This finding was true




for both threshold determinations .and release estimates.









            The better performance by the "large emitter" domain may, to some




extent, be an artifact of the approach taken to the site survey program.  The




use of an order-of-magnitude screen to identify errors means that extremely




large release numbers would have to pass through whatever review process was




employed by facility personnel, before being included in the completed Form R




report.  It seems probable that very large release numbers would be subjected




to more scrutiny, and therefore fewer instances of erroneous reporting would




result.








            As stated above, chemical industry facilities made relatively




fewer errors than nonchemical industry facilities.  However, these relatively




fewer errors made by the chemical industry accounted for a relatively larger




quantity of release errors than the nonchemical industry, because of the




relatively larger release quantities involved.   Further, the data show that




the chemical industry tended to under-report release quantities (especially to




receiving streams),  while the nonchemical industry tended to over-report




release quantities.








            With respect to the other two domains (large- versus small-




quantity emitters) the survey data showed that large-quantity emitters




accounted for relatively fewer errors, but a relatively larger quantity of




release errors than small-quantity emitters.  Again, this result is due to the




relatively larger release quantities involved.   Also, large-quantity emitters






                                     3-27

-------
tended to under-report release quantities,  while small-quantity emitters




tended to over-report release quantities.  These trends are likely due to the




relatively larger quantities of releases associated with chemical industry




facilities and large-quantity emitters.
                                     3-28

-------
4.0         CONCLUSIONS








            The site visit program results revealed that most facilities




included in the 1987 Toxic Release Inventory (TRI) did a good job in perform-




ing threshold determinations and in estimating release quantities.  The 1987




TRI database contains approximately 70,000 Form R chemical reports from




approximately 19,000 facilities, showing a total aggregate release of 22.5




billion pounds of toxic chemicals.  The site visit program results revealed




that approximately 68,000 Form R chemical reports should have been reported by



18,000 facilities, with a total aggregate release of 23.0 billion pounds of




toxic chemicals.   Thus, while the TRI database contains roughly 3% more Form R




chemical reports than it should, and roughly 5% more facilities than it




should, total aggregate releases were slightly under-reported.  These numeri-




cal findings should be considered in light of the various limitations of the




site survey program, described earlier in this report.








            The survey data revealed that facilities incorrectly included and




incorrectly omitted numerous chemicals with respect to threshold determina-




tions.  These problems surfaced in cases where facility personnel misunder-




stood the threshold determination requirements, or where they overlooked




various chemical activities.








            The survey data also revealed that facilities accurately estimated




total releases across the TRI with respect to release estimates, but that they




made 'numerous errors in individual release estimates.  The data showed that




roughly 84% of all release estimates were accurate within one






                                      4-1

-------
order-of-magnitude, given the constraints of the program.  Data analysis




revealed that errors primarily resulted from facility personnel misinterpret-



ing reporting instructions, overlooking releases, using incomplete informa-




tion, poorly documenting release estimates, or making calculation and tran-




scription errors.  Furthermore, the survey data showed that alternative




release estimation techniques were difficult to apply given the limited amount



of readily available data.








            The survey data also revealed that chemical industry facilities




and facilities reporting large-quantity emissions consistently made fewer



errors (relative to the number of chances for making errors) than did non-




chemical industry facilities or small-quantity emitters, respectively.  On the




other hand, the survey results showed that these relatively fewer number of




errors resulted in relatively larger quantities of release errors for both the




chemical industry facilities and the large-quantity emitters.  These trends




probably result from the relatively larger quantities of releases associated




with chemical industry facilities and large-quantity emitters.








            The data indicated that EPA guidance materials specific to the




SARA Title III, Section 313 program, especially the reporting instructions,




were the reference materials most commonly used by facility personnel in




completing Form R chemical reports.  Many types of errors could be eliminated




through additional guidance.  Because they are used so frequently, EPA docu-




ments would be a good vehicle for providing improved guidance.  In some




industries (e.g., pulp and paper), trade association data was used heavily; in
                                      4-2

-------
these cases,  trade associations would be a good vehicle for improving




guidance.
                                     4-3

-------
5.0         RECOMMENDATIONS








            This section presents several recommendations for the SARA




Title III, Section 313 program based on the results and conclusions of the




quality assessment site-visit program.  By implementing these specific recom-




mendations concerning additional reporting guidance and additional areas of




study, the quality of the data contained in future years' Toxic Release




Inventory (TRI) databases would be improved substantially.   The scope of the




site visit program included categorizing reporting errors into discrete types




of errors and misinterpretations to allow EPA to focus improvement efforts on




the areas of highest potential yield.








            Radian's site surveyors identified several specific reasons that




facility personnel made errors in their threshold determinations and release




estimates.  In addition, the survey data revealed that certain facility




characteristics (e.g., chemical versus nonchemical industry facilities, large-




versus small-quantity emitters) and certain release media had distinct types




of errors.  The recommendations presented below result from these findings.









5.1         Additional Reporting Guidance









            (1)   Additional guidance is needed concerning .what constitutes




                  "documentation."  Section 372.10 (recordkeeping) of the




                  regulation cites the need for frequent documentation.  New




                  guidance materials may include a specific definition and/or




                  a number of examples of what constitutes "acceptable"






                                      5-1

-------
      documentation.   Surveyors found poor documentation in both

      threshold determinations and release estimates  at most of

      the surveyed facilities; this generic problem led in turn to

      errors in both threshold determinations and release

      estimates.



(2)    Additional guidance is needed concerning threshold deter-

      minations.   The survey results indicated that approximately

      8,000 chemical reports are incorrectly included in the TRI

      database, mostly because facility personnel did not under-

      stand the concept of making threshold determinations.   The

      results also indicated that approximately 6,000 chemical

      reports are incorrectly omitted from the TRI database,

      mostly because facility personnel overlooked chemical

      activities at their site.  In making both types of threshold

      determination errors,  many facility personnel misinterpreted

      specific instructions provided under Section 313.   Addi-

      tional clarifying guidance should be provided in the

      instructions for the following areas in which particular

      problems were identified with making threshold determination

      calculations:
            Threshold determinations;
            Exemptions;
            Activity classifications;
            Calculations;  and
            Chemical qualifiers.
                         5-2

-------
(3)    Additional guidance is needed concerning release estimates.

      The survey resuLts showed that approximately 14% of the

      release estimates in the TRI database are more than one

      order-of-magnitude different than the Radian surveyor's

      recalculated value.  One of the primary reasons-for making

      release estimate errors was that facility personnel mis-

      interpreted the following items pertaining to release

      estimates in the reporting instructions:
            Acid/base neutralization;
            Off-site recycle and reuse;
            Metal compounds;
            De minimis exemptions;  and
            Fugitive air emissions.
      Additional clarifying guidance should be provided for these

      areas.



      Other reasons for release estimate  errors occurred where

      facility personnel used incomplete  information or less-than-

      preferred estimation techniques for computing release

      quantities.  Additional guidance should be provided on the

      types of information and approaches preferred by EPA in

      given situations.
                         5-3

-------
5.2         Additional Areas of Study



            (1)    The data reduction and analysis simplifying assumptions used

                  in this program should be quantified with respect to their

                  impacts on the program's conclusions.   These assumptions

                  concern:
                        Incorrectly-omitted releases;
                        Incorrectly-omitted chemicals;
                        Form R chemical reports not addressed by this program;
                        and
                        Facilities not included in the  TRI database (i.e.,
                        nonresponders).
            (2)    A comparative evaluation of data quality in the 1988 Toxic

                  Release Inventory should be performed to assess the effec-

                  tiveness of additional guidance materials and other outreach

                  programs targeted to the regulated community.  While the

                  current study showed that chemical industry facilities and

                  large-quantity emitters did a better job than nonchemical

                  industry facilities and small-quantity emitters,  respec-

                  tively, Radian believes that an across-the-board comparison

                  of data quality between the 1987 and 1988 TRI databases

                  would be beneficial.



            (3)    Further study is necessary to identify situations in which

                  the mass balance approach would be reasonable for estimating

                  releases.  The survey showed that facility personnel

                  frequently used the mass balance approach to estimate

                                      5-4

-------
releases (sometimes in combination with engineering calcula-




tions) when monitoring data were not available.  Correlating




the release quantities with the throughput quantities for




each SIC code and other groupings that are appropriate given




the limited size of the database would allow EPA to assess




the usefulness of the mass balance approach.
                    5-5

-------
   APPENDIX A




SURVEY INSTRUMENT

-------
                                 Facility  ID-
       SECTION 313 REPORTING
QUALITY ASSURANCE SURVEY INSTRUMENT
                 A-l

-------
                                                        FaciLicy  ID
 DATE OF VISIT:
 FACILITY  OR  ESTABLISHMENT NAME:
 STREET ADDRESS-  	
 CITY.  	
 STATE:  	
 TELEPHONE NUMBER:  	
 CONTACT:
Radian office performing sice survey:
REVIEWERS:  	    OFFICE COORDINATOR:
                      CODES FOR REPORTED CHEMICALS

               Reported                           Chemical
              Chemical(s)       	CAS »         Code
                                                      B
                                                      H
LIST SECTION 313 CHEMICALS NOT REPORTED BUT EXPECTED TO BE USED BY THIS
INDUSTRY AS IDENTIFIED IN EPA PUBLICATIONS EPA 560/4-88-00&A THROUGH
EPA 560/4-88-004Q AND A REVIEV OF OTHER REPORTS IF THEY ARE AVAILABLE.

Chemical (listed prior to site visit)

o 	    o 	
o 	    o 	
o 	    o 	
o 	    o 	
o 	    o 	
o 	    o 	
o 	    o 	
cml.103
                                      A-2

-------
                                                       Facilicy ID.
                                  SECTION 1.0
                              REPORT PREPARATION
1.1  Who prepared your reports--in particular,  the release estimates?
     (Check all that apply)

          [ ]   Facility Environmental Staff
          [ ]   Company Environmental Staff
          [ ]   Facility Staff
          [ ]   Consultant/Contractor
          [ ]   Other, specify 	
1.2  List all EPA documents and other references used to estimate -releases and
     control efficiencies.  If no references were used,  check here [  ]  and
     continue to question 1.3.  (Use 2-page list of codes.)

                                Reference Used
     o    	    o
     o    	    o
     o    	    o
1.3  If the reporting guideline document was not used, why not?

     Reporting Guidance document was used.  [ ]
     Didn't know it was available.          [ ]
     Other, specify                         [ ]  	
1.4  Did you find the revised 1988 reporting guideline document more useful
     than the previous version?

          N/A - have not done comparison     [ ]
                                     Yes     [ ]
                                     No      [ 1
cml.103
                                      A-3

-------
                                                        FacilUy ID	


1.5  Did you  feel any  section  of  the  instructions.provided with the  Form R vere
     unclear?  yes  [  ]  no  [  ]  If yes,  check  the  appropriate  section below arc.
     briefly  explain the difficulty encountered.

     [  ]   Part I. Facility  Identification  Information
     [  ]   Part II.  Off-site Locations  to which Toxic Chemicals  are transferred
           in  waste
     [  ]   Part III  Chemical specific  information
1.6  Is the facility aware that chemical/material  suppliers are  required  co
     notify customers of all Section  313 chemicals in  their products effective
     in 1989?  yes  [ ]  no [ ]
1.7  Did you call the Emergency Planning and Community Right- to-Know Hotline"1
     yes [ ]  no [ ].  If yes, did you find the operator's response helpful?
     yes [ ]  no [ J.
1.8  How many full-time employees does  the  facility have?
cml.103
                                      A-4

-------
                     Section 2.0
     All  Chemicals  Reported by the  Facility
                                                                  Facility I 0
Chemicals
Reported on
Form R(s)
A
B
C
D
E
F
G
H
1
J
K
L
M
N
0
Reported use
or Activity A
of Chemico™







•







Briefly Describe the Activities in
which the Chemical is Manufactured.
Processed, or Otherwise Used
(NA - Not a Section 313 Chemical)















Was (he Chemical '
Use or Activity
Cuneclly Determined'''
(Yes/No/NAjB















©M - Manufactured ffl
P - Processed II No. document on the following Section 20 worksheet
U - Otherwise. Used









6925E2R
- Not applicable (Reported chemical is not a listed Section 313 Chemical)

-------
 Chemical Code:
                                                        Facility  ID
                              SECTION  2.0 WORKSHEET
 Chemical Code:
Chemical Code:
Chemical Code:   D
Chemical Code:
Chemical Code:
Chemical Code:
Chemical Code:
cml.110
                                      A-6

-------
                                                       Faci.li.cy  ID
                       SECTION 2.0 WORKSHEET (continued)
Chemical Code:
Chemical Code:
Chemical Code:   K
Chemical Code:
Chemical Code:   M
Chemical Code:   N
Chemical Code:   0
cml.110
                                     A-7

-------
           D
                                                   Section 3.0

                         Section 313 Chemicals used but not Reported Because Below

                                          an Applicable Threshold Umft®

                                                Cl»«fc>«r» It d»imrt i r«t ••€!!•• 313 CMBy««toiM»« but »»lf
                                                                                                          FaeWly LO.
                                            D
IBM 10. wtal WM th* UM i
                                        10.

CliMntofll
1
a
3
4
O
0
T
•
•
10

^_ w^^^^ ^^^B ^^^^^^ta^^^B
•MdbvttoPMMMy










Wtal TVM of Aettlvllf
OMHMFaeMly
»p»ripttoU l«r Ikto/^
C^^^te^V IM P IL MAT
*^^^^M^V ^^^ »-f 1^ ••Mf










WMOM
A-iiilullu
CWTMll®
(VM/lto/WA)










WtalwMlMFMMly
••ttMl* •! OM
AMOIM!
MMMfaoturtt^L
Pr«o*>M4 «r
OUM*wto*lto*«T
C»/yr>










toOuppMilM
AModte^A&9
(VM/lto)®











CalculallMM










Waslh*
O«ftl>lfM
•toll*
»»port
C«rr*«IT
(V*a/Ne)










>

oo
                                        70.000 OM
                                        70.000k*

                                  IkJCtf   10.000 OM
                                                      ®MA-»MOJIyOM>*l

                                                      M late. p^f^p^O^^B^O^ A^M ft



                                                    (•)•-•••
                                 3.O wwkatect.
                            313<
           •rt»»p»HiO>ii«M
           r*vt*w «•>(•• 10 «
VDM-
  • -
  o-
  NA-ri
                                                                 •••MM 3.0 wwkvlM*!.
                                                                            wotorti
                                                                                   r r*w4ta. stel*
                                             » •! Ik* r*vl*w*r'a i
                                                                                                                                   i

-------
                                                       Faci.li.cy  ID
                             SECTION 3.0 WORKSHEET
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
cml.110
                                      A-9

-------
                                                  Sectional
                            Section 313 Chemicals used but not reported because
                                             they were overlooked ©
                                                                                             Faculty ID.
D
Ckirti h*r* M Ml
Ckeek k*r* M Ik* MM*** *l •«eUe« 313 CkMBtute •*<
     > Uwy w*r* *«wto«li*tf to •reel** tta*. 1O. H tr**l«r
   i 10. wfcal w*.*lk*l*MM«k*r**lr*p«rl*«*		
nMMno*tfe*y
thMlO.wfc.
•••MM 319®'
•MA A^^^^vft^^
i
•
3
4
8
•
?
•
•
1*

towMehthtaClmriMlto
MM0 Wjf HM ™ ttfl^Ry










What Tya* •! AeMtvMy to
Appropriate Iw IM>^N
CtMododT (fiL p* IU ' ^'^










WlMltotMNcvtoww
ICttMl* •! Ik*
MMMlMtwa<
•»«O««»< «f fi)
(k/yr)










Mould Ihli
ClMMUcal haw*
b*MiH*pOfUd?
(V««/Ne)










                3L*.

                 3131
           >1*i
                                                                     «i Ik* t»iiwto< SMlto* &• w*rk*lM*l.
                                                                                                                       i

-------
                                                       Facilicy ID
                             SECTION 3.1 WORKSHEET
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
cml.110
                                     A-ll

-------
                                                                Section  4.0
                                                                                                                      Facility  ID
           Section  313  Chemicals  used  but  not  Reported  Due  to  an  Exemption  Clause
           a
Check here If not Applicable
Check hero If the number ol Section 313 Chemicals exempted is
greater than  10.  If greater than  10 Section 313 Chemicals
exempted, what was the total number ol Section 313 Chemicals exempted?
Section 313 Chemical
nol Reported
1
2
3
4
5
6
7
8
9
10
Briefly Describe the Activities in which
this Chemical is used by ttie Facility










©
What was the Exemption Basis?










Was the Bosib lor (his Exemption
Correct (Yes/No)®










N)
         © This section does not apply lo those chemicals used but nol
            reported because below an applicable threshold llmH.

         0 If mar* than 10 Section 313 chemicals were not reported due to
            exemption, list the 10 with the greatest amount of use.
                                                            © 0 - Oe-Mlnlmls. Listed chemical Is present In a mixture below 1
                                                                    percent, or 0.1 percent In the cose ol carlnogens.
                                                               A - Article, manufactured  Items formed lo a specific shape during
                                                                    manufacture, which has end use functions dependent on that  shape.
                                                                    and which does not release toxic chemical under normal conditions
                                                                    of processing or uee at the facility
                                                               U - Uses, used as a structural component, routine janitorial or
                                                                    facility groundkeoplng. maintenance,  personal use by employees.
                                                                    maintenance of motor vehicles, or use of toxic chemicals present In
                                                                    process water and non- contact cooling water drawn from the
                                                                    environment or from municipal sources, or air used as
                                                                    compressed air or as part of combustion
                                                               L - Activities In Laboratories, used In a laboratory under a
                                                                    the supervision of a technically qualified Individual.
                                                                                                                                                  S
                                                                            0
                                                                               If no. document on the following Section 4 0 worksheet.

-------
                                                       Facility ID'
                             SECTION 4.0 WORKSHEET
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
Chemical:
cml.110
                                      A-13

-------
Section  5.0
                                                                                              Facility ID
             Threshold  Determinations  for  Reported  Section  313  Chemicals
Section 313
Chemical
Reported
A
B
C
D
E
F
G
H
1
J
K
L
M
N
O
Is Supporting
Doc umentalion
Available?^
(Yes/No)®















What was the Estimate of
(he Amount Manufactured.
Processed, or Otherwise Used?
(lb/yr)P















Reviewer's
Use
Estimate®















Was the Chemical Amount
Used by the Facility
Above the Applicable
Threshold Limit?
(res/No£>















What Approach
Did the Facility
Use to Determine .-.
Threshold Ouanlily'P















i
H—
*-
        © II No. document on the following Section S.O work sheet.


        © On the following Section 5 0 worksheet, present a •ummary al the threshold

          calculations ol both the reviewer and facllHy.
                     A - Proem recipes
                     8 - Purchata and Inventory Records

                     C - Cnalnearlng Judgement

                     0 - Other (specify In Section S 0 worksheet)
                                                                  an
                                                                  o>
CN
a>
to

-------
Chemical Code:
                                                       Faci.Li.cy  ID
                             SECTION 5.0 WORKSHEET
Chemical Code:
Chemical Code:   C
Chemical Code:   D
Chemical Code:   E
Chemical Code:
Chemical Code:   G
Chemical Code:   H
cml.110
                                       A-15

-------
                                                       Facilicy ID
                       SECTION-5.0 WORKSHEET (continued)
Chemical Code.
Chemical Code:
Chemical Code:    K
Chemical Code:
Chemical Code:    M
Chemical Code:   N .
Chemical Code:   0
cml.110
                                     A-16

-------
                                                                                             FceWly I.D.
Section 6.0 Reviewer Checklist
                                                              Reported Section 313 Chemical
                                                             F /Q
          to LflBtf CMI dto)
           to fOTW

    WMteTn
It ««y •»«• w«r« «««Uf«f ••yjirtittiithj •••^•••••rtMfcX      0 m to •vary MM* fw Mckia»wl«« •


-------
                                                                                                         Facility I.O.
     Section 6.0 Reviewer Checklist
3


4

5


6
0


10
   Ararat*
•MtoTCyM/M)'
   Pwi
                lMtMltf>U.V*lto
              Kt *) (g)
   Boto? tbo f o0CNv*0 bttflto •€
   rapwtetf •• HM P«m II U* C. «, O) 0
   Ar* OMf • Mqr <
              iMffw*
                   •r
                   Mrf»K>TW)T
   W«r« n» §•!••••• If•• wa>U
               ilrallHM*!
                           aia
   lpnri«



                                                                     Reported Section 313 Chemical
              iMlltoMtoWkVlMllMir*


                                                                                                                          1

-------
 Chemical Code:
                                                       Facility ID
                             SECTION 6.0 WORKSHEET
 Chemical Code:
Chemical Code:
Chemical Code:   D
Chemical Code:   E
Chemical Code:   F
Chemical Code:   G
Chemical Code:   H
cml.110
                                     A-19

-------
                                                       Facilicy  ID
                       SECTION -6.0 WORKSHEET (continued)
Chemical Code:
Chemical Code:
Chemical Code:   K
Chemical Code:
Chemical Code:   M
Chemical Code:   N
Chemical Code:   0
cml.110                                A_2Q

-------
                                                                       Facility I D.	
                                       Section  7.0

                    Results  of  Detailed  Calculation Review
                  (To be completed for each Reported Section 313 Chemical)
Chemical Code  10:
Release/®
Transfer
Fugitive Air
.Stack Air
Roc. Stream
Land On-site
Underground
Injection
POTW
Other Off-site
Max. On-site
Form
Entry
Pounds








Calculated*
by Same
Method








Reviewer Release®
Estimate








Ratio of "Form®
Entry to Reviewers
Approach"







'//////,
                   Q


                   9
For multiple entries (2 receiving streams), use total amount for that
medium.


Calculated using same approach/data used by facility.


estimate based on reviewer's judgment of a superior estimation method
given tho data available  to the facility .and reviewer.  State method on attached
calculation sheets.

A -  facility underestimated  by  a factor of  1000 or more.
• -  facility underestimated  by  factor of  100 to 999.
C -  facility underestimated  by  a factor of  10  to 99.
0 -  less than factor of  10  difference.
I -  facility overestimated by factor of 10 to 99.
F -  facility overestimated by factor  100 to 999.
6 -  facility overestimated by a factor of over  1000.
H -  facility Incorrectly omitted release quantity (I.e.. rotlo*0)
I  - facility reported a releaase quantity that it should not have.
   (I.e., ratlo»lnflnlty)
                                         A-21

-------
                                                        Facility  ID-
                              SECTION  7.0 WORKSHEET



         Reviewer Release  Estimates of Reported Section 313 Chemicals





Chemical Code:   A
Chemical Code:
Chemical Code:
Chemical Code:   D
Chemical Code:
cml.110
                                       A-22

-------
                                                        Faci.li.cy ID.
                      SECTION  7.0 WORKSHEET   (continued)
Chemical Code:   F
Chemical Code:
Chemical Code:
Chemical Code:
Chemical Coda:
cml.110                               A_23

-------
                                                        Faci.li.cy ID:
                       SECTION 7.0 WORKSHEET  (continued)
Chemical Code:
Chemical Code:
Chemical Code:
Chemical Code:   N
Chemical Code:   0
onl.110                               A_24

-------
       APPENDIX B




SAMPLE SELECTION PROCESS




    (ABT ASSOCIATES)

-------
        Saaple Selection Process

        The pocencial respondent universe for the Evaluation of the Quality
and Accuracy of the Data Submitted by the Eligible Facilities  in Compliance
with Section 313 Requirements consisted of approximately 18,000 responding
facilities.  The 18,000 facilities submitted roughly 68,500 forms (i.e.,
chemical reports) in total, for an average of 3.8 forms per facility.  The
distribution of facilities by number of forms submitted is shown below.

              No. of Forms                   Percent
                    1
                    2
                    3
                    4
                    5
                  5-10
                 11-15
                   16*
Facilities submitting 15 or fewer forms account for 97.8 percent of the total
responding facilities.  Because facilities with 16 or more forms are so few
and would have required a time consuming site visit, they were removed from
the sampling frame.  So, too, were facilities in Alaska and Hawaii where it
would have been costly to conduct site visits due to the travel costs
involved.
        The resulting sampling frame of responding facilities was divided into
geographic clusters.  The clusters were formed on the basis of three-digit zip
code sectional centers.  The distribution of facilities by these section
centers was examined and there were many that contained only a small number
(less than 25) of facilities.  It was therefore necessary to combine adjacent
sectional centers to build up the clusters to a minimum size of 25 facili-
ties.  It vas also necessary to remove remote sectional centers, with five or
fewer facilities, that could not be grouped in a consistent fashion with other
sectional centers.
        The final respondent universe consisted of roughly 17,600 facilities
and about 53,400 chemical report forms.  The final respondent universe
contains 97.8Z of facilities, and 78.02 of forms.  It is to this universe that
inferences can be made with a known degree of sampling error.
                                    B-l

-------
        A two-stage sampling procedure was used to draw the sample of facili-
ties.  At the first stage of sampling 54 zip code sectional centers or
sectional center groups were drawn with probability proportional to size (PPS)
sampling, .the measure of size being the count of facilities in the cluster.
At the second stage of sampling a stratified sample of facilities was drawn
from each of the 54 sample clusters.  Two facility stratifiers were used.
Each defines key analytic domains of interest to the evaluation of data
quality and accuracy.  By using these two variables of interest as
stratifiers, the sample size of site visits could be controlled and as dis-
cussed below, the desired level of precision for each analytic domain could be
met.  The first was a two-category Standard Industrial Classification (SIC)
code variable:

        •     Primary SIC code equals 2800 to 2899.  This covers
              chemical manufacturing.
        •     Primary SIC code does not equal 2800 to 2899, but the
              primary, secondary, or tertiary SIC code is in the 2000
              to 3999 range.  This covers all other manufacturing
              facilities.

The population size and the desired number of facilities with completed site
visits at the time the sample.was drawn in April 1989 is shown below:
                                 Approximate             Desired
                                 Population             Number  of
SIC CODE GROUP               Size of Facilities        Site  Visits
Chemical Manufacturing              3,610                   40
Other Manufacturing                14,012                   110
                                   17,622                   150

The main objective of this stratifier is to allow for the comparison of
chemical manufacturing facilities with'all other manufacturing facilities.
        The second stratifier categorized facilities into large versus small
emitters of chemicals to the environment.  To create this stratifier it was
                                  B-2

-------
necessary to first examine Che distribution of facilities by the number of
pounds of chemicals released to each of six media:


        1.    Fugitive or non-point air emissions and stack or point
              air emissions,

        2.    Releases to receiving streams,

        3.    POTW,

        4.    All other off-site transfer,

        5.    On-site land disposal* and,

        6.    Deep-well injection on-site.


A threshold for defining a large emitter was set for each of the six media
using the 90th percentile of each individual facility distribution among

facilities with non-zero emission to that medium.  The 90th percentile was
selected so that only the largest emitters would fall into the large emitter

stratification category.  A facility was classified as a large emitter if it
fell in the large emitter end of the distribution for any of the six media.

If it did not cross the threshold for any of the six media then it was

classified as a non-large emitter.  The threshold values for defining a large
emitter by mediums are shown below:


                                                     Threshold
        Release to!                                 (Ibs/year)

        Water*                                       > 100,000
        POTWs*                                       > 200,000
        Air                                          > 300,000
        Off-Site Locations
          Other than POTWs                           > 300,000
        Land                                         > 500,000
        Underground Injection                     > 25,000,000

        *Excludes releases of Sodium Sulfate
                                   B-3

-------
        The desired sampLe size of facilities in each of these two categories
is shown below.  The population size of facilities in each stratum, at the
time the sample was drawn, is also shown:

                                Approximate             Desired
        Type of                Population  Size           Number  of
        Emitter                 of  Facilities          Site Visits
        Large                      2,206                  75
        Non-Large                 15.416                  75
                                  17,622                 ISO

The main objective of this stratifier is to allow for the comparison of large
versus non-large emitters.
        Other potential stratifiers such as number of forms submitted, number
of employees, total release amount, type of emission, type of chemical, and
presence of basic data errors in the forms submitted were also given con-
sideration in the design.  The two selected stratifiers  discussed above were
ranked highest (given a total sample size of 150 facilities) based on the
analytic needs of the study.
        After consideration of the cost of conducting a  single site visit, the
cost of travel to a zip code sectional center,  and the expected low degree of
homogeneity of facilities within section centers, it was determined that a
reasonable initial sample size of facilities per section center was around
eight.  With an expected response rate of around 60 percent,  the typical
sectional center would yield about 4.8 completed site visits, and therefore, a
total of approximately 32 out of the 54 sample  sectional centers would have to
be travelled to attain roughly 150 total site visits. The procedure of
selecting approximately 32 out of 54 zip code sectional  centers is discussed
below.
        It was very important to come as close  as possible to the target of
ISO completed site visits, because of the considerable cost per site visit
compared to the typical cost per interview in a household survey.  Although a
target response rate of 60 percent had been set, the actual response rate
might have turned out somewhat higher or lower  than this, and furthermore, the
response rate might also vary across the four strata formed by the two
category SIC code variable and the two-category level of emissions

                                   B-4

-------
scratifier.  Keeping in mind the need to have a national probability sample of
facilities, a first-stage sample of 54 zip code sectional centers was drawn.
Before any facility sampling occurred, the sectional centers were sorted into
zip code order, a proxy for geographic ordering, and randomly subsampled into
nine replicates, each containing six zip code sectional centers.  In the first
five replicates, containing 30 sample sectional centers in total, the
selection of a stratified sample of eight facilities per sectional center was
carried out.
        The section of sample facilities involved three steps.  First, the
desired marginal sample sizes were input into an iterative proportional
fitting algorithm along with the following population table:
                                 Non-Large
Large
3,093
12,323
517
1,689
                 Chemical
                 Other
A set of expected cell frequencies was then derived under the constraint that
the cell frequencies satisfied the desired marginal sample sizes.  The desired
cell (stratum) sample sizes are shown below:
                                 Non-Large
Large
Chemical
Other
18
57
22
53
75 75
40
110
150
These stratum sample sizes were next allocated to the sample zip code
sectional centers with the objective of having a roughly self-weighting sample
of facilities for each of the four strata.  The sample sizes actually
allocated to the zip code sectional centers were inflated to allow for an
expected 60Z response rate.  The third step involved drawing a stratified
random sample of facilities without replacement from each sample zip code
sectional center.
        The sample facilities in the 30 zip code sectional centers were then
contacted for participation in the site visits.  The results of the first five
                                   B-5

-------
replicates were used  Co  judge  Che. overall response race (with a factor  for
subsequent cancellations  built in), and Che response rates in the four
strata.  It turned out noc  to  be necessary co sample facilities in any
additional replicates beyond Che fifth.
        Because site  visits are expensive co conduce, Che minimum sample  size
needed Co meet Che analytic and estimation needs of the study was determined
Co be around ISO.  As noted above,  estimates of percent of facilities with
various types of data' errors was to be computed for Che encire sample and for
Che four domains of interest.   The  sample was therefore allocated so that the
separate domain escimates chat are  derived will all have a useable Level  of
precision at the 90 percent confidence level as indicated below.1

        Total Sample  (n  » 150)                       ±6.91
        Chemical Manufacturers (n » 40)              i 13.3Z
        Other Manufacturers (n = 110.)                i  8.0Z
        Large Emitters (n = 75)                      i  9.7Z
        Small Emitters (n = 75)                      i  9.7Z

The sample size allocated to chemical manufacturers is the smallest of  the
domain sample sizes.  One hundred ten of Che 150 facilities were allocated to
Che non-chemical manufacturing domain* because Che general expectation  was
chat chemical manufacturers made fewer reporting errors than other types  of
manufacturers.
        The actual number of completed sice visits is very close to the
desired CargeC sample size  as  shown below:

                                  Non-Large	Large
Chemical
Other
19
59
25
53
78 78
44
112
156
A cocal of  156 useable sice visits were completed.
       1 The calculation uses P » SOS sine* the variance (P(IOO-P) is largest for this value.
The 90 percent confidence lleits equal / 1.04  / 1.65  / 50(100-50)/n, where the design effect
due to clustering is assuned to be 1.04

                                     B-6

-------
        1C is necessary co use weighcs when drawing inferences to each of the
four domains, Che entire target universe, or other subclasses of interest.
The weighcs compensate for the unequal selection probabilities arising from
Che disproportionate allocation of the sample to the four strata.
        The weighting methodology involved several steps.  First, the
reciprocal of Che selection probability of each of Che 30 sample zip code
sectional centers in the first five replicates was computed.  This was divided
by 5/9 to reflect the fact that only 5 out of Che 9 replicates were used.
        This resulting zip code weight was multiplied times Che reciprocal of
che wichin zip code sectional center selection probability of each facility
where a site visit was completed.  This quantity equalled Nnj/nhl» wnere Nhi
is the population count of facilities in che h-th stratum for che i-th zip
code sectional center and n^j is the corresponding count of sample facilities
where a sice visit was conducted.  The produce equals Che facility weight.
        The distribution of the 156 facility weights was examined, and values
above the 90th percencile of che discribucion were Cruneated Co che weight
value at the 90th percentile.  The truncation procedure reduces increases in
sampling variance from unequal weighting.  All facilities with a truncated
weight value are in che Other/Non-Large Emitter stratum, which was
undersampled.
        The weights after truncation were then ratio-adjusted so that the
weighted discribucion by che four strata equalled the target universe
discribucion shown below.

        Chemical/Non-Large
        Chemical/Large
        Ocher/Non-Large
        Ocher/Large
The total weighted counc of reported chemicals equals 53,393,
                                    B-7

-------
       APPENDIX C




TRAINING MEETING AGENDA

-------
             AGENDA FOR MAY  15,  1989 OFFICE  COORDINATOR TRAINING



1.     Opening Remarks	      TM

2.     Introduce Training Program  	      RW

3.     Overview of SARA Title III, and Purpose of the Quality Assurance
      Program	      TM

4.     Results of the Pilot Survey	      KB

5     Office Coordinator Responsibilities 	      RW

LUNCH BREAK

6.     The Survey Instrument and Instructions  	      GO

7.     Helpful Hints in Conducting Surveys 	      KB

8.     Preferred Alternative Approaches  	      GO

9     Case Studies	      RW

10.    Wrap-up	      TM
                                     C-l

-------
              APPENDIX D




SITE VISIT REQUEST INTRODUCTION LETTER

-------
2O
*
Anniversary
                                                        1359S Dulles Technology Or.
eo mmom mi o M                                              Herndon. VA 22071
                                                                   (703)834-1500
April 28, 1989
Dear

Your facility has been randomly selected by the U.S. Environmental Protection
Agency (EPA) to participate in a quality assurance survey of facilities that
submitted first year reports (Form R) for the Toxic Chemical Release Inventory
(TRI) under Section 313 of SARA Title III.  The purpose of the survey is to
evaluate the quality of the data submitted in the Form R's and to provide
feedback to EPA that can be used to improve the reporting instructions,
guidance, or the reporting form for the future.  EPA has contracted with
Radian Corporation to perform the survey by conducting visits to some
150 facilities selected from the TRI database.

Participation in this program will assist you in determining reporting
thresholds and release estimates for your facility.  This should be valuable
information for completion of future Form R's.

As the contractor for EPA on this effort, we will visit your facility sometime
in May, June,  or July 1989.  The visit will last 1 to 2 days depending on che
number of Form R's you submitted.  We will spend the majority of that time
reviewing the calculations and data used to determine reporting thresholds and
release estimates.  During the site visit, we will need your assistance to
tour the plant extensively enough to allow us to understand your overall
process and identify potential release points.  Your assistance will also be
needed in filling out a questionnaire, which is designed to address areas
where we believe data quality problems are most likely to occur.  We will make
every effort to minimize disruptions to your schedule while we are on site.

tfe will contact you by telephone within the next week to arrange a date for
the site visit.  I would like to emphasize that the name and location of your
facility will not be released to EPA in order to ensure confidentiality of
your facility's information.  The EPA Project Officer for this program is
Larry Longanecker of the Office of Toxic Substances.  Enclosed is a letter
from him requesting your participation.

Thank you for your cooperation.  If you have any questions, please feel free
to call Glenn Osmond at (919) 541-9100 or me at (703) 834-1500.

Sincerely,
Tim McAdams
Project Director

Enclosure

                                    D-l

-------
2O
*
Anniversary
                                                        13595 Dulles Technology Dr.
                                                              Herndon. VA 22071
                                                                   (7031834-1500
April 28, 1989

                                                  id \
Dear

Your facilicy has bean randomly selected by tha U.S.  Bnvlrotinancal Protection
Agancy (EPA) Co participate in a quality aaauranca aurvay of facilities  that
submitted first year reports (Font R) for the Toxic Chemical Release
Inventory (TRI) under Section 313 of SARA Title III.   The purpose of the
survey is to evaluate the quality of the data submitted in the Form R's  and
to provide feedback to EPA that can be used to improve the reporting
instructions, guidance, or the reporting fora for the future.  EPA has
contracted with Radian Corporation to perform the survey by conducting visits
to some 150 facilities selected from the TRI database.

Participation in this program will assist you in determining reporting
thresholds and release estimates for your facility.  This should be valuable
information for completion of future Form R's.

A letter detailing what to expect during the site visit, which will occur
sometime in Hay, June, or July 1989, has been sent to Karen Jandar, who  was
given as the technical contact on your Form R(s).  We will contact Ms. Jander
by telephone within the next week to arrange a date for the site visit.

I would like to emphasize that the name and location of your facility will
not be released to EPA in order to ensure confidentiality of your facility's
information.

The EPA Project Officer for this program is Larry Longanecker of the Office
of Toxic Substances.  Enclosed is a letter from him requesting your
participation.

Thank you for your cooperation.  If you have any questions, please feel  free
to call Glenn Osmond at (919) 541-9100 or me at (703) 834-1500.

Sincerely,
Tim McAdams
Project Director

Enclosure

cc:  Karen Jander



                                   D-2

-------
         APPENDIX E




REFERENCES USED BY FACILITIES

-------
                                   TABLE E-l

          REFERENCES COMMONLY USED.BY SURVEYED FACILITIES TO  ESTIMATE
                       RELEASES AND CONTROL EFFICIENCIES
                                                       Number of  Facilities
Reference                                                 Using Reference
TRI Reporting Form R and Instructions
EPA 560/4-88-005

The Emergency Planning and Community Right-to Know
Act, Section 313 (40 CFR 372)
EPA 560/4-88-001

Estimating Releases and Waste Treatment
Efficiencies for the TRI
EPA 560/4-88-002

Title III Section 313 Release Reporting Guidance,
EPA 560/4-88-004 a through q, Estimating Chemical
Releases:

   a. Monofilament fiber manufacturing
   b. Printing operations
   c. Electrodeposition of organic coatings
   d. Spray application of organic coatings
   f. Formulating aqueous solutions
   g. Electroplating operations
   h. Textile dyeing operation
   i. Press wood and laminated wood products
      manufacturing
   j. Roller, knife, and gravure coating
      operations
   k. Paper and paperboard products
   q. Rubber production and compounding

Compilation of Air Pollution Emission Factors,
AP-42

Toxic Chemical Release Reporting Proposed Rule,  52
IE 21152, June 4, 1987

Privately Sponsored Seminar Materials

Government Sponsored Seminar Materials

National Council of the Paper Industry for Air and
Stream Improvements (NCASI) Memorandum

CMA Guidance for Estimating Fugitive Emissions
1
1
1
3
1
7
3
1
2
1
         141
          40
          38
          23
          17


          16


          14

           9

           9
                                   E-l

-------
                             TABLE E-l (Continued)

              REFERENCES USED BY SURVEYED FACILITIES TO ESTIMATE
                       RELEASES  AND CONTROL EFFICIENCIES
                                                       Number  of  Facilities
Reference                                                 Using Reference

Toxic Chemical Release Inventory Questions and                   6
Answers
EPA 560/4-89-002

Vendor Brochures                                                 4

Common Synonyms for Section 313 Chemicals                        4
OTS-ETD-001

Corporate Guidance Manual                                        4

Comprehensive List of Chemicals Subject to                       3
Reporting Under the Act (Title III List of Lists)
EPA 560/4-88-003

Understanding the Small Quantity Generator                       2
Hazardous Waste Rules:  A Handbook for Small
Business
EPA 530/SW-86-019

Chemical Hazard Communication Guidebook                          2

"313 Advisor" Computer Program, duPont                           2

Assessment of Atmospheric Emission from Petroleum                2
Refining, EPA 600/2-80-075 C

Fugitive Emission Sources of Organic Compounds -                 2
Additional Information on Emissions, Emission
Reductions, and Costs
EPA 450/3-82-010

Draft Protocols for Generating Unit Specific                     2
Emission Estimates for Equipment Leaks of VOC and
VHAP, EPA, April 1987

"Fugitive Emissions Calculation Guidelines,"  J.W.               2
Davenport and L. J. McGregor, Paper presented at
80th Annual Meeting of APCA

"Community Right-to-Know Manual," Thompson                       1
Publishing Group
                                   E-2

-------
                             TABLE E-l .(Continued)

              REFERENCES USED BY SURVEYED FACILITIES TO ESTIMATE
                       RELEASES  AND CONTROL EFFICIENCIES
                                                       Number of Facilities
Reference                                                 Using Reference


Locating and Estimating Air Emissions From                        1
Sources of Manganese
EPA 450/4-84-007h

Electric Arc Furnace Dust Disposal, Recycle and                   1
Recovery, Center for Metals Production

Chemical.Engineers Handbook, Perry and Chilton,                 	1
McGraw-Hill

TOTAL REFERENCES USED                                           352
                                    E-3

-------
        APPENDIX F




WRAP-UP MEETING AGENDA AND




    SUMMARY OF ISSUES

-------
           AGENDA FOR SARA 313 SITE VISIT PROGRAM WRAP-UP MEETING
                     SEPTEMBER 7, 1989; HERNDON, VIRGINIA
ATTENDEES
Larry Longanecker, U.S. EPA, Office of Toxic Substances
Kathy Franklin, U.S. EPA, Office of Toxic Substances
Richard Wells, Abt Associates
Roger Christman, Radian Corporation, Program Manager
Tim McAdams, Radian Corporation, Project Director
Kevin Fisher, Radian Corporation, Office Coordinator
Glenn Osmond, Radian Corporation, Office Coordinator
Tim Rimpo, Radian Corporation, Office Coordinator
Larry Sontoski, Radian Corporation, Office Coordinator
Richard Weisman, Radian Corporation, Office Coordinator
Vic Yamada, Radian Corporation, Office Coordinator
Deb Matthews, Radian Corporation, Site Surveyor
Sandy Howard, Radian Corporation, Research Assistant

DISCUSSION ITEMS

1.   Overall Impressions

     How do you feel about the information obtained?

     How do you feel about the effort put forth by industry?

2.   Survey Instrument

     What specific questions in the survey instrument should be changed, and
     how should they be changed?

     Please comment on each section of the survey instrument.  Are there
     poorly phrased questions?  Are you not sure of a question's intent?

3.   Site Selection

     Did you encounter any problems in the site selection process (e.g., plant
     locations vor groupings)?

4.   Arrange Site Visits

     How could the process used to contact facilities and arrange site visits
     be improved?

     How could the training for the telephone callers be improved?
NRJ-50
0821-03.nrj.2
                                    F-l

-------
     What problems did you encounter when either rescheduling or cancelling
     site visits?

     Surveyor Training

     How could the training for the site surveyors be improved?  What elements
     could be added or improved?  What are the elements of a reasonable model
     training program?

     Were the training materials useful?  Which were most useful?  Which were
     least useful?

     Was the training budget adequate?  What would be adequate?

     Surveyor Qualifications

     What qualifications (education and experience) are necessary for a person
     to successfully perform a quality assurance site visit?

     Site Visit Preparation

     How much preparation is needed for a site visit?  What preparation do you
     think is necessary before making a site visit?  What about the facility
     and the industry should be reviewed prior to visiting a site?

     Facility Tour

     How is a facility tour a useful component of a quality assurance site
     visit?

     How detailed should the facility tour be to obtain information needed to
     complete the site visit?

     Did the facility personnel appear to be candid during the tour?

     Do you think the operational status of the facility affected the quality
     of the information you obtained?

     Did you tour any facilities that were not in operation?

     For those facilities toured, did you see things that were overlooked?

     For those non-operating facilities toured, do you think you missed
     anything?  For non-operating facilities who did not use a mass-balance
     approach, how can you tell if you missed anything?

     Documentation

     How much confidence should be placed on the documentation facilities
     offered for review?
NRJ- 50
0821-03.nrj.3
                                     F-2

-------
     Did Che facility personnel appear to be candid in providing documentation
     for review and when answering questions about the documentation?

10.  Facility Concerns

     Did the facility personnel express concern about public perceptions of
     the release estimates they reported?

     Did the facilities take steps to improve their release estimates or to
     reduce their releases of Section 313 chemicals?

11.  Range Reporting

     Did the facility personnel find the range reporting option useful?

12.  Causes of Release Estimate Errors

     For situations where the facility personnel estimated releases higher
     than we calculated, what are the primary reasons why the facility's
     release estimates were over reported?

     For situations where the facility personnel estimated releases lower than
     we calculated, what are the primary reasons why the facility's release
     estimates were under reported?

13.  Hotline

     Did you find the EPA/EPCRA Hotline helpful?  Did the facility personnel
     find the Hotline helpful?

     What problems with the Hotline were encountered?

     Did the Hotline provide accurate, consistent answers to specific ques-
     tions?

14.  "Toxic Chemical Release Inventory Reporting Form R and Instructions"

     What specific items in the instructions did you find unclear?

     What specific items in the instructions did the facility personnel find
     unclear?

     How could these items be improved?

15.  Survey Instrument Quality Assurance

     How could the survey instrument quality assurance (both in-office and
     inter-office) be improved?
NRJ-50
0821-03.nrj.4

                                       F-3

-------
16.  Future Activities

     Future activities planned for the EPA quality assurance program are an
     audit manual, an examination of large emitters, and an assessment of
     companies with large changes from year-to-year.  What are your thoughts
     on these activities?  What other activities would you suggest?
NRJ-50
0821-03.nrj.5
                                      F-4

-------
 CORPORATION
MEMORANDUM
2455 Horsepen Road, Suite 250
         Herndon, VA 22071
             (703)834-1500
TO:       Larry Longanecker and Kathy Franklin,
          U.S. EPA, Office of Toxic Substances

FROM:     Tim McAdams and Richard Weisman,
          Radian Corporation

DATE:     24 October 1989

SUBJECT:  Summary of Issues Discussed at Wrap-up Meeting
This memorandum briefly summarizes the issues discussed during the SARA Title
III, Section 313 Site Visit Program wrap-up meeting held in our office on
September 7, 1989.  This memorandum includes the comments and recommendations
made by the meeting attendees on each of the discussion items on the agenda  (a
copy of the agenda is appended to this memorandum),  as well as on the three
additional discussion items raised by Richard Wells.

DISCUSSION ITEMS

1.    Overall Impressions

     There were limited data available for industry to make threshold determi-
     nations and to estimate releases.

     Calculations usually went beyond data obtained/maintained to satisfy
     permit requirements.

     Smaller companies need more help than larger companies.

     The electroplating and furniture industries had more trouble than others
     with reporting.

     Electroplaters had problems with chemistry-oriented issues.

     Facilities had difficulty incorporating existing computer programs for
     inventory tracking with SARA Title III, Section 313 reporting require-
     ments .

     Facilities with environmental staffs generally did a better job of
     reporting than facilities without environmental staff.
NRJ-50
0922-07.bjg.l
                                      F-5

-------
  RADIAN
  CORPORATION

Tim McAdams and Richard Weisman
24 October 1989
Page 2
     Richard Wells' Discussion Items

     The supplier notification was not very helpful to industry because of the
     large number of MSDSs received by industry and the irregular systems used
     for tracking toxic chemicals.

     Additional technical guidance documents would be helpful to industry -
     especially for electroplaters and furniture makers.

     The current series of industry-specific guidance documents were found to
     be overly simplistic and, as a result, of limited usefulness

     Site Selection

     Problems were encountered with facilities entered in the TRI database
     incorrectly.

     The computer printout occasionally missed reported chemicals,  especially
     when amended Form R's were submitted after July 1, 1988.

     Is the facility name and address typed in once or separately with each
     chemical entered?

     Arrange Site Visits

     We could have been better prepared to respond where  facilities requested
     confidentiality agreements.

     Surveyor Training

     More detailed and realistic  guidance documents,  including an expanded
     question and answer document, would aid surveyors in determining how the
     Form R's should be completed where industries and chemicals are complex.

     An entire day is needed to allow the surveyors to read all the training
     materials that were provided.

     Surveyor Qualifications

     Qualified surveyors need to  have an understanding of chemistry, indus-
     trial processing operations, SARA Title III,  Section 313 regulations, and
     people management skills.

     Site Visit Preparation

     Obtain detailed information  over the phone on what major processes each
     facility operates.
NRJ-50
0922-07.bjg.2

                                      F-6

-------
Tim McAdams and-Richard Weisman
24 October 1989
Page 3


     Cross-reference 3-digit SIC codes with reported chemicals to help prepare
     for visits.

8.   Facility Tour

     Tours were found to have many advantages - they provide a -direct mecha-
     nism to address facility contact's concerns,  let you view storage and
     waste treatment areas, and provide a clear way to ask meaningful ques-
     tions .

     'Tours were not as useful for facilities where a mass balance approach was
     used to estimate releases.

9.   Documentation

     The quality of documentation reviewed varied significantly, and sugges-
     tions were made that EPA should publish guidance on what is good.documen-
     tation.

10.  Facility Concerns

     Facilities -generally had limited feedback from the public on reported
     releases.

     Since July 1988, many facilities have taken steps to improve reporting
     .accuracy and reduce their releases.

11.  Range Reporting

     EPA should add a lower range, for example, 1-10 pounds per year.

12.  Causes of Release Estimate Errors

     Facilities frequently wanted to report, "just to be complete", even if
     they were below applicable thresholds.

     Other errors were due to reporting releases of metal compounds, reporting
     releases of all material purchased, without considering inventory
     changes, on-site treatment, or incorporation into products, failing to
     consider recycle/reuse practices, fraction of.a compound in a material,
     neutralization of acids and bases, and volatile releases of nonvolatile
     compounds.

     Facilities were found to have performed limited QA on their release
     calculations, and to sometimes have poor internal communication at large
     facilities with multiple groups using the same chemical.
NRJ-50
0922-07.bjg.3
                                      F-7

-------
 CORPORAT

Tim McAdams and Richard Weisman
24 October 1989
Page 4


13.  Hotline

     The hotline was usually a good information source for our surveyors when
     visiting facilities.

     The hotline sometimes gave contradictory answers to questions asked by
     different people.

     The hotline often could not cite the written source for an answer.
14.  Reporting Instructions

     Several items in the Revised '88 Instructions were found to be unclear:
     the definition of an "article," de minimus concentrations,  and metal
     fumes or dust; how to address chemicals that are manufactured, processed,
     and used in the same operation; when to consider metals different than
     metal compounds; how to determine waste treatment efficiencies;  and how
     to consider materials brought to a facility for reclamation purposes.

15.  Survey Instrument Quality Assurance

     The QA protocol should be published at the beginning of the program
     Internal QA should be performed on a more rapid basis.

16.  Future Activities

     A manual on how to perform QA surveys would be very helpful.

     A detailed checklist, with questions similar to the survey instrument,
     should be included in the manual.
NRJ- 50
0922-07.bjg.4
                                       F-8

-------
APPENDIX G




SURVEY DATA

-------
                            TABLE G-l
              THRESHOLD DETERMINATION SURVEY RESULTS
Category

Correctly-Included
Chemical Reports

Incorrectly-Included
Chemical Reports

Correctly-Omitted
Chemical Reports

Incorrectly-Omitted
Chemical Reports
Category

Correctly-Included
Chemical Reports

Incorrectly-Included
Chemical Reports

Correctly-Omitted
Chemical Reports

Incorrectly-Omitted
Chemical Reports
UNWEIGHTED RESULTS

Number of Chemicals

         657


          71


         746


          48



 WEIGHTED RESULTS

Number of Chemicals

      62,000


       8,000


     110,000


       6,000
Number of Facilities

          149


           46


          135


           33
Number of Facilities

       17,000


        4,000


       15,000


        3,000
                             G-l

-------
                                  TABLE G-2

             THRESHOLD DETERMINATION APPROACHES USED BY SURVEYED
        FACILITIES TO  DETERMINE AMOUNT OF CORRECTLY-INCLUDED CHEMICALS
                  MANUFACTURED,  PROCESSED,  OR OTHERWISE USED
Threshold Determination Approach
Percent of Chemical Reports
  Correctly Included by the
     Surveyed Facilities
Purchase/Inventory Records
Emission Factors*
Mass Balance8
Facility Assumed Threshold Exceeded0
Process Recipes0
Monitoring DataE
Production DataF
Hazardous Waste Manifests0
TOTAL
85.3
3.7
3.2
2.4
2.7
1.4
1.1
0^2
100.0
AEmission factors  published by industry  trade associations were used  to
compute quantities of chemicals manufactured.

BMass  balance  calculations were based on the quantity of chemical  in  raw
materials, products, and wastes, and included complex calculations involving
chemical reactions and stoichiometry.

cFacility personnel assumed thresholds were exceeded  for a chemical based on
knowledge of facility operations, usually without support  documentation.

°Process recipes were used to calculate  the quantity  of chemical manufactured
or processed based on the quantity of raw material purchased.

EData  characterizing wastewater discharges and  solid waste materials  were used
to compute the quantity of disposed chemical.   The quantity  of chemical used
was assumed to be equal to or greater than the quantity of disposed chemical.

FThe quantity  of product manufactured multiplied by the concentration of  the
chemical in that product was used to determine the quantity  of chemical
manufactured.

°Waste quantity data from hazardous waste manifests and corresponding
characterization data were used to compute the quantity of disposed chemical.
The quantity of chemical used was assumed to be equal to  or  greater than the
quantity of disposed chemical.
                                     G-2

-------
                                   TABLE G-3

              REASONS CHEMICAL REPORTS WERE INCORRECTLY INCLUDED
                      IN THRESHOLD DETERMINATION RESULTS
Reason
Percent of Chemical Reports
    Incorrectly Included
 Among Surveyed Facilities
Qualifier Misinterpreted

Quantity Less Than Threshold Limit

Facility Assumptions Not Supported
by Records

Activity Misclassified

Quantity Miscalculated

Chemical Subject to Exemption

Not a Section 313 Listed Chemical

TOTAL
            19.7

            45.1

             5.6


             1.4

            15.5

             8.5
           100.0
                                     G-3

-------
                                                         i-A
                    TABLE G-4

REASONS CHEMICAL REPORTS WERE INCORRECTLY OMITTED
        IN THRESHOLD DETERMINATION RESULTS
                                                       Number of Facilities
                                                       Incorrectly Omitting
                          Number of Chemical Reports       At Least One
Reason                        Incorrectly Omitted*         Chemical Report*

Overlooked Chemical                         36                   26
Activity

  -  Manufacturing By-             4
    Products
  -  Wastewater Treatment         11
    By-Products
  -  Wastewater Treatment          7
    Chemicals
  -  Cleaning Chemicals            4
  -  Miscellaneous
    Process or Otherwise         10
    Used Chemicals

Misclassified Activity8                    5                       4

Miscalculated Quantity0                    4                       4

Other0                                     3

TOTAL                                     48                      33E


*The number of chemical reports and facilities presented  are  based on  data
collected at 156 surveyed facilities.

BThese errors occurred when facilities used more  than 10 ,'000  Ibs/yr  of a
chemical, but less than 75,000 Ibs/yr, and misclassified the activity as
manufacturing or processing.  Radian surveyors found the correct activity for
these chemicals to be "otherwise used."

cRadian's surveyors recalculated the quantity of  chemical manufactured,
processed, or otherwise used, and determined a quantity greater than the
applicable threshold.

°0ne facility incorrectly interpreted requirements  to mean that  repackaging
did not need to'be reported.  One facility followed a manufacturer's advice
that a chemical did not need to be reported.  One facility incorrectly
calculated the quantity of pure metal used instead of the metal compound.

ESome of the 32 facilities incorrectly omitting chemicals did so for more than
one reason.  Therefore, the total number (32) does not equal the sum of the
reason-specific numbers.

                                    G-4

-------
                                   TABLE G-5

                   FREQUENCY OF  ERROR  IN RELEASE  ESTIMATES*
                          AT THE  SURVEYED FACILITIES:
                               UNWEIGHTED DATA
Magnitude of Facility's Error
 Percent  of Release  Estimates
Applicable to a Release Medium
Incorrectly-Omitted Release Quantity

Underestimated by three or more
orders-of-magnitude

Underestimated by two orders-of-
magnitude

Underestimated by one order-of-
magnitude

Not an error--within one order-of-
magnitude

Overestimated by one order-of-
magnitude

Overestimated by two orders-of-
magnitude

Overestimated by three or more
orders-of-magnitude

Incorrectly-Reported Release
Quantity

TOTAL
              8.7

              0


              0.5


              1.4


             77.9


              1.4


              0.5


              0.1
             100.0
ARelease  estimate  errors were defined as release quantities more than one
order-of-magnitude different than the Radian QA reviewer's estimated value.
                                    G-5

-------
                                  TABLE G-6

                    SOURCES OF ERROR IN RELEASES  ESTIMATES*.
                         AT THE  SURVEYED  FACILITIES:
                               UNWEIGHTED DATA
Type of Error
Number of Instances
   Error was Made
Number  of
Facilities
1. Misinterpreted the reporting
instructions:

   Acid/Base Neutralization                   49
   Recycle/Reuse Off-Site                     24
   Metal Compound Reported                     2
   De Minimus Exemption                        6

2. Used incomplete information:

   Overlooked facility data                   70
   Overlooked chemical properties             48
   Overlooked emission factors                15

3. Documentation did not support release      51
estimates

4. Other errors made by facility:

   Calculation errors                         11
   Transcription errors                       33
                        23
                        15
                         1
                         1
                        26
                        25
                         9

                        26
                         8
                        14
AThis  table summarizes instances where the QA reviewer's  release  estimate
differed from the amount reported by the facility by more than one order-of-
magnitude.
                                   G-6

-------
      TABLE G-7 FREQUENCY OF ERRORS (3) IN RELEASE ESTIMATES-
                           UNWEIGHTED SURVEY DATA

Type of
Industry (1)
Chemical
Non- Chemical
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
165
215
380


A
0
0
0


B
0
1
1


C
5
4
9
Fugiti

D
127
173
300
ve Ai

E
6
2
8
r

F
1
0
1


G
0
0
0


H
11
24
35


I
15
11
26
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
295
85
380
Fugitive Air

A
0
0
0

B
1
0
1

C
5
4
9

D
239
61
300

E
7
1
8

F
1
0
1

G
0
0
0

H
22
13
35

I
20
6
26
Notes:  1 -  Chemical industry is defined here by Primary SIC Code 28,
       2 -  Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 -  Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A = Facility underestimated by a factor of 1000 or more
           B = Facility underestimated by a factor of 100 to 999
           C •= Facility underestimated by a factor of 10 to 99.
           D - Not an error; within one order of magnitude.
           E - Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G = Facility overestimated by a factor of over 1000.
           H = Facility incorrectly omitted release quantity.
           I - Facility reported a release quantity that it
               should not have.
       4 -  Analysis applicable only for chemical reports correctly
           included.
                                     G-7

-------
         TABLE G-7 FREQUENCY OF ERfcORS(3) IN RELEASE ESTIMATES -
                           UNWEIGHTED SURVEY DATA
Total Number
Of Releases
Estimates
Type of Applicable to
Indus try (1) Medium (4)
Chemical
Non- Chemical
Total
163
225
388
A
0
0
0
B
1
1
2
C
1
3
4
Stack Air
D
137
202
339
E
6
1
7
F
1
0
1
G
0
0
0
H
9
14
23
I
8
4
12
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
285
103
388


A
0
0
0


B
2
0
2


C
3
1
4
Stack

D
255
84
339
Air

E
4
3
7


F
1
0
1


G
0
0
0


H
11
12
23


I
9
3
12
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A = Facility underestimated by a factor of 1000 or more.
           B - Facility underestimated by a factor of 100 to 999."
           C = Facility underestimated by a factor of 10 to 99.
           D - Not an error; within one order of magnitude.
           E = Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G - Facility overestimated by a factor of over 1000.
           H = Facility incorrectly omitted release quantity.
           I - Facility reported a release quantity that it
               should not,have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                        G-8

-------
    TABLE G-7 FREQUENCY OF ERRORS(3) IN RELEASE ESTIMATES-
                     UNWEIGHTED SURVEY DATA

Type of
Indus try (1)
Chemical
Non- Chemical
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
36
63
99


A
0
0
0


B
0
0
0
I

C
1
1
2
)ischa

D
28
35
63
rge t

E
0
1
1
o Wat

F
0
0
0
er

G
0
0
0


H
0
6
6


I
7
20
27
Reported
Size of
Emitter (2)
Large
Small
Total
Total Number
Of Releases
Estimates
Discharge to Water
Applicable to
Medium (4)
93
6
99
A
0
0
0
B
0
0
0
C
2
0
2
D
61
2
63
E
1
0
1
F
0
0
0
G
0
0
0
H
5
1
6
I
24
3
27
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order, of magnitude different
           than the site surveyor's recalculated value.
           A " Facility underestimated by a factor of 1000 or more.
           B - Facility underestimated by a factor of 100 to 999.
           C - Facility underestimated by a factor of 10 to 99.
           D = Not an error; within one order of magnitude.
           E = Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G = Facility overestimated by a factor of over 1000.
           H = Facility incorrectly omitted release quantity.
           I - Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                    G-9

-------
        TABLE G-7 FREQUENCY OF ERRORS(3) IN RELEASE ESTIMATES-
                           UNWEIGHTED SURVEY DATA


Type of
Industry(l)
Chemical
Non- Chemical
Total
Total Number
Of* Releases
Estimates
Applicable to
Medium (4)
20
34
54





On-Site Release
A
0
0
0
B
0
0
0
C
0
2
2
D
15
24
39
E
0
0
0




to Land
F
0
0
0
G
0
0
0
H
1
4
5
I
4
4
8
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
49
5
54
On-Site Release

A
0
0
0

B
0
0
0

C
2
0
2

D
38
1
39

E
0
0
0
to Land

F
0
0
0

G
0
0
0

H
4
1
5

I
5
3
8
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A = Facility underestimated by a factor of 1000 or more.
           B - Facility underestimated by a factor of 100 to 999.
           C = Facility underestimated by a factor of 10 to 99.
           D - Not an error; within one order of magnitude.'
           E = Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G = Facility overestimated by a factor of over 1000.
           H « Facility incorrectly omitted release quantity.
           I = Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                     G-10

-------
         TABLE G-7 FREQUENCY OF EKRORS(3) IN RELEASE ESTIMATES-
                           UNWEIGHTED SURVEY DATA
Total Number
Of Releases
Estimates
Type of Applicable to
Indus try (1) Medium (4)
Chemical
Non- Chemical
Total
20
1
21
A
0
0
0
B
0
0
0
Underground
C D E
0
0
0
18
0
18
0
0
0
Injection
F G
0
0
0
0
0
0
H
0
0
0
I
2
1
3
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
21
0
21
Underground

A
0
0
0

B
0
0
0

C
0
0
0

D
18
0
18

E
0
0
0
Injection

F
0
0
0

G
0
0
0

H
0
0
0

I
3
0
3
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more .than one order of magnitude different
           than the site surveyor's recalculated value.
           A - Facility underestimated by a factor of 1000 or more.
           B - Facility underestimated by a factor of 100 to 999.
           C = Facility underestimated by a factor of 10'to 99.
           D = Not an error; within one order"of magnitude.
           E - Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G - Facility overestimated by a factor of over 1000.
           H = Facility incorrectly omitted release quantity.
           I = Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                    G-ll

-------
          TABLE G-7 FREQUENCY OF ERRORS(3) IN RELEASE ESTIMATES-
                           UNWEIGHTED SURVEY DATA

Type of
Industry (1)
Chemical
Non- Chemical
Total
Total Number
Of Releases
Estimates
Discharge to POTW
Applicable to
Medium (4)
62
103
165
A
0
0
0
B
0
3
3
C
0
0
0
D
44
65
109
E
0
0
0
F
1
1
2
G
0
0
0
H
13
7
20
I
4
27
31
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
132
33
165
Discharge to POTW

A
0
0
0

B
1
2
3

C
0
0
0

D
93
16
109

E
0
0
0

F
1
1
2

G
0
0
0

H
17
3
20

I
20
11
31
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28,
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A - Facility underestimated by a factor of 1000 or more.
           B = Facility underestimated by a factor of 100 to 999.
           C = Facility underestimated by a factor of 10 to 99.
           D = Not an error; within one order of magnitude.
           E « Facility overestimated by a factor of 10 to 99.
           F = Facility overestimated by a factor of 100 to 999.
           G = Facility overestimated by a factor of over 1000.
           H - Facility incorrectly omitted release quantity.
           I - Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                    G-12

-------
          TABLE G-7 FREQUENCY OF ERRORS(3) IN RELEASE ESTIMATES -
                           UNWEIGHTED SURVEY DATA


Type of
Indus try (1)
Chemical
Non-Chemical
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
124
172
296



Other
A
0
0
0
B
1
0
1
C
0
2
2






Off-Site Transfer
D
104
122
226
E
2
1
3
F
1
2
3
G
1
1
2
H
10
23
33
I
5
21
26
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
228
68
296
Other

A
0
0
0

B
1
0
1

C
2
0
2
Off-Site Transfer

D
181
45
226

E
2
I
3

F
2
1
3

G
2
0
2

H
21
12
33

I
17
9
26
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A — Facility underestimated by a factor of 1000 or more
           B = Facility underestimated by a factor of 100 to 999
           C = Facility underestimated by a factor of 10 to 99.
           D = Not an error; within one order of magnitude.
           E - Facility overestimated by a factor of 10 to 99.
           F = Facility overestimated by a factor of 100 to 999.
           G = Facility overestimated by a factor of over 1000.
           H - Facility incorrectly omitted release quantity.
           I - Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                   G-13

-------
          TABLE G-8 FREQUENCY OF ERRORS(3) IN RELEASE ESTIMATES-
                          WEIGHTED SURVEY DATA
Type of
Industry(1)
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
        Fugitive Air

 B    C     D    E    F

 0  444  6944  213   35

24  217 11698  255    0

24  661 18642  468   35
G    H    I

0  713  938

0 2465  762

0 3178 1700
Chemical

Non-Chemical

Total
       9287    0

      15421    0

      24708    0
Reported
Size of
Emitter(2)
Total Number
Of Releases
Estimates
Applicable to
Medium (4)     A
        Fugitive Air

 B    C     D    E    F

24  189  6556  243   35

 0  472 12087  224    0

24  661 18643  467   35
G    H    I

0  667  572

0 2511 1127

0 3178 1699
Large

Small

Total
       8286    0

      16421    0

      24707    0
Notes: 1 • Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities, more than one order of magnitude different
           than the site surveyor's recalculated value.
           A = Facility .underestimated by a factor of 1000 or more.
           B » Facility underestimated by a factor of 100 to 999.
           C = Facility underestimated by a. factor of 10 to .99.
           D - Not an error; within one order of magnitude.
           E - Facility overestimated by a factor of 10 to 99.
           F = Facility overestimated by a factor of 100 to 999.
           G - Facility overestimated by a factor of over 1000.
           H = Facility"incorrectly omitted release quantity.
           I = Facility reported a release quantity..that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                    G-14

-------
        TABLE G-8 FREQUENCY OF ERRORS(3) IN RELEASE ESTIMATES -
                          WEIGHTED SURVEY DATA
Type of
Industry(1)
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
        Stack Air

 B    C     D    E    F

22   71  7902  378   22

31  227 15638  173    0

53  298 23540  551   22
G    H    I

0  980  234

0 1979  551

0 2959  785
Chemical

Non-Chemical

Total
       9609    0

      18599    0

      28208    0
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
8460
19748
28208


A
0
0
0


B
53
0
53


C
124
173
297
Stack

D
7568
15972
23540
Air

E
129
423
552


F
22
0
22


G
0
0
0


H
274
2685
2959


I
290
495
785
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A - Facility underestimated by a factor of 1000 or more
           B = Facility underestimated by a factor of 100 to 999.
           C - Facility underestimated by a factor of 10 to 99.
           D - Not an error; within one order of magnitude.
           E = Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G - Facility overestimated by a factor of over 1000.
           H - Facility incorrectly omitted release quantity.
           I = Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                  G-15

-------
          TABLE G-8 FREQUENCY OF ERRORS(3) IN RELEASE ESTIMATES -
                           WEIGHTED SURVEY DATA
Type of
Total Number
Of Releases
Estimates
Applicable to
                                          Discharge to Water
Industry (1)
Chemical
Non-Chemical
Total
Medium (4)
846
2114
2960
A
0
0
0
B
0
0
0
C
11
27
38
D
536
1124
1660
E
0
23
23
F
0
0
0
G
0
0
0
H
0
289
289
I
299
651
950
Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
2210
751
2961


A
0
0
0


B
0
0
0


C
38
0
38
Dischc

D
1537
124
1661
irge t

E
23
0
23
o Wat

F
0
0
0
er

G
0
0
0


H
116
173
289


I
496
454
950
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A = Facility underestimated by a factor of 1000 or more.
           B = Facility underestimated by a factor of 100 to 999.
           C = Facility underestimated by a factor of 10 to 99.
           D - Not an error; within one order of magnitude.
           E = Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G - Facility overestimated by a factor of over 1000.
           H = Facility incorrectly omitted release quantity.
           I = Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                    G-16

-------
        TABLE G-8 FREQUENCY OF ERRORS IN RELEASE ESTIMATES -
                          WEIGHTED SURVEY DATA
Total Number
Of Releases
Estimates
Type of Applicable to
Indus try(l) Medium (4)
Chemical
Non- Chemical
Total
974
920
1894
A
0
0
0
On- Site Release
B
0
0
0
C
0
55
55
D
418
651
1069
E
0
0
0
to Land
F
0
0
0
G
0
0
0
H
108
120
228
I
448
94
542

Reported
Size of
Emitter(2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
1128
767
1895

A,
0
0
0



On- Site Release
B C D E
0
0
0
55 842
0 227
55 1069
0
0
0


to Land
F G
0
0
0
0
0
0

H
120
108
228

I
111
432
543
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are -defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A = Facility underestimated by a factor of 1000 or' more.
           B = Facility underestimated by a factor of 100 to 999.
           C - Facility,underestimated by a factor of 10 to 99.
           D - Not an error; within one'order of magnitude.
           E - Facility overestimated by a factor of 10 to 99.
           F = Facility overestimated by a factor of 100 to 999.
           G = Facility overestimated by a factor of over 1000.
           H - Facility incorrectly omitted release quantity.
           I = Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                   G-17

-------
         TABLE G-8 FREQUENCY OF ERfcORS(3) IN RELEASE ESTIMATES-
                           WEIGHTED SURVEY DATA
Total Number
Of Releases
Estimates
Type of Applicable to
Industry (1) Medium (4)
Chemical
Non- Chemical
Total
386
8
394
A
0
0
0
Underground Injection
B C D E F G
0
0
0
0
0
0
341
0
341
0
0
0
0
0
0
0
0
0
H
0
0
0
I
45
8
53
Reported
Size of
Emitter (2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
394
0
394
Underground

A
0
0
0

B
0
0
0

C
0
0
0

D
341
0
341
Injection

E
0
0
0

F
0
0
0

G
0
0
0

H
0
0
0

I
53
0
53
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A *• Facility underestimated by a factor of 1000 or more.
           B - Facility underestimated by a factor of 100 to 999.
           C = Facility underestimated by a factor of 10 to 99.
           D - Not an error; within one order of magnitude.
           E - Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G - Facility overestimated by a factor of over 1000.
           H - Facility incorrectly omitted release quantity.
           I - Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                   G-18

-------
           TABLE G-8 FREQUENCY OF'ERRORS(3) IN RELEASE ESTIMATES -
                           WEIGHTED SURVEY DATA
Type of
Indus try (1)
Total Number
Of Releases
Estimates
Applicable to
Medium (4)     A
              B

 2867    0    0

 8430    0  390

11297    0  390
  Discharge to POTW

C    D    E    F    G
     H
Chemical

Non-Chemical

Total
                         0 2374

                         0 4377

                         0 6751
          0    8

          0  150

          0  158
0  428   57

0  785 2728

0 1213 2785
Reported
Size of
Emitter (2)
Large
Small
Total
Total Number
Of Releases
Estimates
Discharge
to POTW
Applicable to
Medium (4)
4280
7016
11296
A
0
0
0
B
56
334
390
C
0
0
0
D
3197
3554
6751
E
0
0
0
F
8
150
158
G
0
0
0
H
476
737
1213
I
543
2241
2784
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on- the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A = Facility underestimated by a factor of 1000 or more.
           B - Facility underestimated by a factor of 100 to 999.
           C - Facility underestimated by a factor of 10 to 99.
           D - Not ah error; within one-order of magnitude.
           E - Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of .100 to 999.
           G - Facility overestimated by a factor of over 1000.
           H - Facility incorrectly omitted release quantity.
           I - Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                     G-19

-------
        TABLE G-8 FREQUENCY OF ERR'ORS(3) IN RELEASE ESTIMATES-
                          WEIGHTED SURVEY DATA
Type of
Industry(1)
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
 B

71

 0

71
   Other Off-Site Transfer

 C     D    E    F    G    H
Chemical

Non-Chemical

Total
       7066    0

      13266    0

      20332    0
 0  5373  212  190   22  943  255

61  9164    8   46   24 2085 1878

61 14537  220  236   46 3028 2133


Reported
Size of
Emitter (2)
Large
Small
Total
Total Number
Of Releases
Estimates
Applicable to
Medium (4)
6739
13593
20332




A
0
0
0




B
71
0
71




C
61
0
61


Other

D
5305
9232
14537




Off-Site

E
30
190
220







Transfer

F G
46
190
236
46
0
46

H
551
2477
3028

I
629
1504
2133
Notes: 1 - Chemical industry is defined here by Primary SIC Code 28.
       2 - Large and small emitters are differentiated by the
           quantity of releases reported on the Form R.
       3 - Release estimate errors are defined here as reported
           quantities more than one order of magnitude different
           than the site surveyor's recalculated value.
           A - Facility underestimated by a factor of 1000 or more.
           B - Facility underestimated by a factor of 100 to 999.
           C = Facility underestimated by a factor of 10 to 99.
           D - Not an error; within one order of magnitude.
           E = Facility overestimated by a factor of 10 to 99.
           F - Facility overestimated by a factor of 100 to 999.
           G « Facility overestimated by a factor of over 1000.
           H - Facility incorrectly omitted release quantity.
           I = Facility reported a release quantity that it
               should not have.
       4 - Analysis applicable only for chemical reports correctly
           included.
                                   G-20

-------
                APPENDIX H




SUMMARY OF PREFERRED ALTERNATIVE APPROACHES

-------
                                   TABLE H-l

      SUMMARY OF RELEASES AND  CALCULATION APPROACHES FOR- CHEMICALS AT
SURVEYED FACILITIES WHERE RADIAN IDENTIFIED A PREFERRED ALTERNATIVE APPROACH
Facility
1
1
2
2
3
3
A
5
5
5
5
5
6
7
8
9
10
10
10
10
10
10
10
10
10
11
12
Chemical
Code
A
B
A
B
A
B
A
A
B
C
D
E
A
A
A
A
A
B
C
D
E
F
G
H
I
A
A
Facility's
Estimation
Approach*
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
M
M
M
M
M
M
M
0
0
Facility's
Reported
Release
Estimate
(lbs/yr)B
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
10,000-20,000
10,000-20,000
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
0-1,000
10,000-30,000
10,000-30,000
10,000-30,000
10,000-30,000
10,000-30,000
1,000-10,000
1,000-10,000
1,000-10,000
0-1,000
Radian's
Preferred
Approach*
E
E
C
C
E
E
M
E
E
E
M
M
E
C
E
E
E
E
E
E
E
E
E
E
E
M
E
Radian's
Release
Estimate
(Ibs/yr)
50,000
50,000
80,000
40 , 000
. 3,000
800,000
100
10,000
2,000
1,000
0
0
100
80 , 000
20,000
0
100
100
2,000
2,000
500
500
100
500
100
0
3,000
  C «= Mass Balance
  E = Emission factors
  M - Monitoring data
  0 = Engineering judgment

  Facility releases were reported as ranges to  protect confidentiality.
                                    H-l

-------
                    APPENDIX I




SUMMARY OF THE TOP TWELVE RELEASE ESTIMATE ERRORS




              FOR  EACH RELEASE MEDIUM

-------
  Facility Release
               TABLE 1-1

   SUMMARY OF THE TOP TWELVE RELEASE
ESTIMATE ERRORS FOR EACH RELEASE MEDIUM

              Fugitive Air

   Radian QA  Reviewer's
     Release Estimate
Estimate (Ibs/yr)*-9
700,000
40 , 000
300,000
300,000
10,000
2,000
0
200,000
0-1,000
0-1,000
0-1,000
(Ibs/yr)*-8
10,000
200,000°
200,000
200,000
100,000
90,000
80 , 000C
100,000
50,000C
50 . 000C
40 , 000C
Source of Error8
Transcription error
Overlooked emission factors
Overlooked % of chemical
Overlooked % of chemical
Overlooked emission factors
Overlooked air releases from
treatment lagoons
Overlooked chemical properties
(volatility)
Miscalculated release estimate
Overlooked emission factors
Overlooked emission factors
Overlooked chemical properties
     100,000
         100,000
                                              (volatility)
Calculation error
ARelease estimates reported by  facilities on the Form R  chemical  reports are
shown with only one significant  figure  for confidentiality reasons.  Therefore,
some values shown here may not appear to be errors in release estimates.

BThe errors summarized in this table were made by 39  facilities.

CA preferred alternative method was identified and used to calculate  this release
estimate.   Otherwise,  Radian's  estimate  was computed  using  the facility's
approach.

                                   1-1

-------
                             TABLE 1-1 (Continued)
                       SUMMARY OF THE TOP TWELVE RELEASE
                    ESTIMATE ERRORS FOR EACH RELEASE MEDIUM

                                   Stack Air
Facility Release
Estimate (lbs/yr)*iB
Radian QA Reviewer's
  Release  Estimate
      (Ibs/yr)*-9
Source of Error8
     100,000

       3,000


      80,000


       3,000

     0-1,000
           0

       6,000

      20,000


      10,000


      60,000

     200,000
     800,000

     400,000


     200,000


      80,000

      60,000


      60,000


      50,000

      40,000C

      50,000


      40,000


      40,000

     200.000
Overlooked emission factors

Overlooked air releases from
process vents and stacks

Overlooked air release from one
process unit

Miscalculated release estimate

Overlooked air releases from
process vents

Overlooked air releases
completely

Overlooked emission factors

Overlooked facility data

Incorrectly split air emission
between fugitive and stack

Overlooked air release from
process source

Transcription error

Calculation error
ARelease  estimates reported by  facilities on the  Form  R chemical reports are
shown with only one significant  figure for confidentiality reasons.  Therefore,
some values shown here may not appear to be errors in release estimates.

BThe  errors summarized in this table were made by 39  facilities.

CA preferred alternative method was identified  and used to calculate this release
estimate.   Otherwise,  Radian's  estimate  was  computed using  the facility's
approach.
                                      1-2

-------
                             TABLE 1-1 (Continued)

                       SUMMARY OF THE TOP TWELVE RELEASE-
                    ESTIMATE ERRORS FOR EACH RELEASE MEDIUM

                              Discharge  to Water
Facility Release
Estimate (lbs/yr)*-B
Radian QA Reviewer's
  Release  Estimate
      (lbs/yr)*-B
Source of Error8
3,000,000
6,000,000
1,000,000
1,000,000
500,000
2,000,000
400,000
200,000
100 , 000
100,000
100,000
100 , 000
40,000,000
0
0
0
0
3,000,000
0
0
0
0
0
2,000
Miscalculated sodium sulfate
production
Transcription error
Transcription error
Transcription error
Misinterpreted acid/base
neutralization instructions
Miscalculated sodium sulfate
production
Transcription error
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Overlooked chemical properties
AR.elease estimates reported  by facilities on the  Form R chemical reports  are
shown with only one significant figure for confidentiality reasons.  Therefore,
some values shown here may not appear  to be errors in  release estimates.

BThe errors summarized in this table were made by 39 facilities.

CA preferred alternative method was identified and used to calculate this release
estimate.   Otherwise,  Radian's  estimate  was  computed using  the facility's
approach.

                                      1-3

-------
                             TABLE 1-1 (Continued)

                       SUMMARY OF THE TOP TWELVE RELEASE
                    ESTIMATE ERRORS FOR EACH RELEASE MEDIUM

                             Underground Injection
Facility Release
Estimate (lbs/yr)AlB
Radian QA Reviewer's
  Release  Estimate
      (Ibs/yr)*-8
Source of Error8
      50,000

      30,000


     0-1,000


     0-1,000


     0-1,000
        40,000

        40,000


             0


             0
Double-counted,a process source

Incorrectly classed one source
of release as an air emission

Misinterpreted acid/base
neutralization instructions

Misinterpreted acid/base
neutralization instructions

Transcription error
ARelease estimates reported  by  facilities on the  Form- R chemical reports 'are
shown with only one significant  figure for confidentiality reasons..  Therefore,
some values shown here may not appear  to be errors in release estimates.

BThe errors summarized in this table were made by 39 facilities..

CA preferred alternative method was identified and used to calculate this  release
estimate.   Otherwise,  Radian's  estimate was  computed using  the facility's
approach.

                                  1-4

-------
Facility Release
         TABLE 1-1 (Continued)

   SUMMARY OF THE TOP TWELVE RELEASE
ESTIMATE ERRORS FOR EACH RELEASE MEDIUM

            On-Site to Land

   Radian QA Reviewer's
     Release Estimate
Estimate (Ibs/yr)*-8
400,000
0-1,000
200,000
7,000
6,000
5,000
2,000
0-1,000
2,000
0-1,000
0
0
(Ibs/yr)*-8
1,000,000°
300,000
0
70,000
50,000
0
0
2,000
3,000C
0-1,000
0-1,000
0-1,000
Source of Error8
Overlooked monitoring data
Miscalculated release estimate
Overlooked chemical properties
Miscalculated release estimate
Miscalculated release estimate
Incorrectly reported spill which
was neutralized/transcription
error
Transcription error
Overlooked additional source of
release
Overlooked emission factors
Overlooked facility data
Transcription error
Overlooked release from
                                               treatment  system
'ARelease  estimates  reported by facilities  on the Form R chemical  reports  are
 shown with only one significant figure for confidentiality reasons.   Therefore,
 some  values  shown here may not appear to be errors  in release estimates.

 8The errors summarized in   his table were made by 39  facilities.

 GA'preferred alternative method was identified and used to calculate this release
 estimate.    Otherwise,  Radian's  estimate  was  computed  using the  facility's
 approach.

                                    1-5

-------
Facility Release
         TABLE 1-1 (Continued)

   SUMMARY OF THE TOP TWELVE RELEASE
ESTIMATE ERRORS FOR EACH RELEASE MEDIUM

        Other Off-Site Transfer

   Radian QA Reviewer's
     Release Estimate
Estimate (Ibs/yr)*-8
0
0
1,000,000
10,000
800.000
0
400,000
0
0
400 , 000
0
300,000
(Ibs/yr)*-8
5,000,000
2,000,000
200,000
800 , 000C
70,000
500,000
0
400,000
400 , 000
0-1,000
300,000
0-1,000
Source of Error8
Transcription error
Transcription error
Misinterpreted recycle/reuse
instructions
Overlooked emission factors
Overlooked % of chemical in
release estimate
Misinterpreted de minimus rule
Misinterpreted acid/base
neutralization instructions
Overlooked chemical properties
Transcription error
Transcription error
Transcription error
Misinterpreted recycle/reuse
                                              instructions
ARelease estimates reported by  facilities on the Form R  chemical  reports are
shown with only one significant  figure for confidentiality reasons.  Therefore,
some values shown here may not appear to be errors in release estimates.
     errors summarized in this table were made by 39 facilities.

CA preferred alternative method was identified and used to calculate this release
estimate.   Otherwise,  Radian's  estimate  was computed using the  facility's
approach .

                                      1-6

-------
                             TABLE 1-1 (Continued)

                       SUMMARY OF THE TOP TWELVE RELEASE
                    ESTIMATE ERRORS FOR EACH RELEASE MEDIUM

                                      POTW
Facility Release
Estimate (Ibs/yr)*8
Radian QA Reviewer's
  Release  Estimate
     (Ibs/yr )*•"
     100,000


     100,000
Source of Error8
2,000,000
500,000
500,000
400,000
400,000
200,000
200,000
400,000
0-1,000
100,000
3,000,000
0
0
0
0
0
0
600,000
200,000
0
Overlooked amount of chemical
purchased and released
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Misinterpreted acid/base
neutralization instructions
Miscalculated amount of chemical
in solution
Overlooked chemical properties
Misinterpreted acid/base
                       neutralization instructions

                       Misinterpreted acid/base
                       neutralization instructions

                       Misinterpreted acid/base
                       neutralization instructions
ARelease estimates reported  by facilities on the  Form R  chemical reports are
shown with only one significant figure for confidentiality reasons.  Therefore,
some values shown here may not appear  to  be errors in release estimates

^The errors summarized in this table were made by 39  facilities.

CA preferred alternative method was identified and used to calculate this release
estimate.   Otherwise,  Radian's  estimate was computed  using  the facility's
approach.

                                     1-7

-------