EPA 560/5-85-009
DECEMBER 1987
METHODS FOR ASSESSING EXPOSURE
TO CHEMICAL SUBSTANCES
Volume 9
Methods for Estimating Releases of
Chemical Substances Resulting from
Transportation Accidents
by
Julia Gartseff, Walter C. Crenshaw,
Patricia D. Jennings
EPA Contract No. 68-02-4254
Project Officer
Elizabeth F. Bryan
Exposure Evaluation Division
Office of Toxic Substances
Washington, D.C. 20460
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF PESTICIDES AND TOXIC SUBSTANCES
WASHINGTON, D.C. 20460
-------
DISCLAIMER
This document has been reviewed and approved for publication by the
Office of Toxic Substances, Office of Pesticides and Toxic Substances,
U.S. Environmental Protection Agency. The use of trade names or
commercial products does not constitute Agency endorsement or
recommendation for use.
in
-------
FOREWORD
This document is one of a series of volumes, developed for the U.S.
Environmental Protection Agency (EPA), Office of Toxic Substances (OTS),
that provides methods and information useful for assessing exposure to
chemical substances. The methods described in these volumes have been
identified by EPA-OTS as having utility in exposure assessments on
existing and new chemicals in the OTS program. These methods are not
necessarily the only methods used by OTS, because the state of the art in
exposure assessment is changing rapidly, as is the availability of
methods and tools. There is no single correct approach to performing an
exposure assessment; thus, the methods in these volumes are discussed
only as options to be considered rather than as rigid procedures to be
followed.
Unlike other volumes in this series, this report does not present
exposure calculations based on incident- or source-specific release
scenarios. Instead, it deals with a broad category of source information,
annual releases of chemicals by various modes of transportation.
Exposure assessment methods for individual vehicular accidents involving
chemicals may be addressed in a future volume.
The definition, background, and discussion of planning exposure
assessments are discussed in the introductory volume of the series
(Volume 1). Each subsequent volume addresses only one general exposure
setting. Consult Volume 1 for guidance on the proper use and
interrelations of the various volumes and on the planning and integration
of an entire assessment.
The titles of the nine basic volumes are as follows:
Volume 1 Methods for Assessing Exposure to Chemical Substances
(EPA 560/5-85-001) (PB86-107083)
Volume 2 Methods for Assessing Exposure to Chemical Substances in the
Ambient Environment (EPA 560/5-85-002) (PB86-107067)
Volume 3 Methods for Assessing Exposure from Disposal of Chemical
Substances (EPA 560/5-85-003) (PB86-107059)
Volume 4 Methods for Enumerating and Characterizing Populations Exposed
to Chemical Substances (EPA 560/5-85-004) (PB86-107042)
Volume 5 Methods for Assessing Exposure to Chemical : ubstances in
Drinking Water (EPA 560/5-85-005) (PB86-123; 156)
-------
Volume 6 Methods for Assessing Occupational Exposure to Chemical
Substances (EPA 560/5-85-006) (PB86-157211)
Volume 7 Methods for Assessing Consumer Exposure to Chemical Substances
(EPA 560/5-85-007)
Volume 8 Methods for Assessing Environmental Pathways of Food
Contamination (EPA 560/5-85-008)
Volume 9 Methods for Estimating Releases of Chemical Substances
Resulting from Transportation Accidents (EPA 560/5-85-009).
EPA-OTS intends to issue periodic supplements for Volumes 2 through 9
to describe significant improvements and updates to the existing
information. The Agency also plans to add short monographs to the series
dealing with specific areas of interest. The first four monographs to be
added are as follows:
Volume 10 Methods for Estimating Uncertainties in Exposure Assessments
(EPA 560/5-85-014)
Volume 11 Methods for Estimating the Migration of Chemical Substances
from Solid Matrices (EPA 560/5-85-015)
Volume 12 Methods for Estimating the Concentration of Chemical
Substances in Indoor Air (EPA 560/5-85-016)
Volume 13 Methods for Estimating Retention of Liquids on Hands (EPA
560/5-85-017)
Elizabeth F. Bryan, Chief
Exposure Assessment Branch
Exposure Evaluation Division
(TS-798)
Office of Toxic Substances
-------
ACKNOWLEDGMENTS
This report was prepared by Versar Inc., Springfield, Virginia, for
the EPA Office of Toxic Substances, Exposure Evaluation Division,
Exposure Assessment Branch (EAB), under EPA Contract No. 68-02-4254 (Task
100). The EPA-EAB Task Manager was Greg Schweer; his support and
guidance are gratefully acknowledged.
A number of Versar personnel have contributed to this task over the
period of performance. These individuals are shown below:
Program Management - Gayaneh Contos
Douglas Dixon
Technical Support - Fouad Mohamed
Khadiga Gamgoum
Robert Westin
Gina Dixon
Stephen H. Nacht
Editing - Juliet Crumrine
Barbara Malczak
Secretarial/Clerical - Lynn Maxfield
Kammi Johannsen
vn
-------
TABLE OF CONTENTS
Page No.
FOREWARD
ACKNOWLEDGMENTS
Tabl e of Contents
List of Tables
List of Figures
1 . INTRODUCTION
1 . 1 Purpose and Scope
1 . 2 Structure of the Report
1.3 Sources of Information
2. PATTERNS OF COMMERCIAL DISTRIBUTION OF MANUFACTURED
CHEMICALS
2.1 Distribution of Manufactured Chemicals by
Transportation Mode
2.1.1 Transportation of Manufactured Chemicals by
Rail
2.1.2 Transportation of Manufactured Chemicals by
Truck ,
2.1.3 Transportation of Manufactured Chemicals by
Waterborne Vessel s
2.1.4 Transportation of Manufactured Chemicals by
Air
2.2 Factors Contributing to Transportation Releases
3. METHOD OF CALCULATING THE EXPECTED QUANTITY RELEASED OF
COMMERCIALLY AVAILABLE CHEMICALS
3 . 1 The General Method
3.2 Sample Calculations ,
3.2.1 Expected Releases of DEHP During
Transportation Accidents ,
3.2.2 Expected Releases of Ethylene Oxide During
Railroad Transportation Accidents ,
3.2.3 Expected Releases of Formaldehyde During
Transportation Accidents
4 . REFERENCES
V
vi i
ix
xii
xvii
1
1
1
2
7
7
7
7
9
9
9
13
13
35
35
43
50
59
-------
TABLE OF CONTENTS (continued)
Page No,
APPENDIX A - DEPARTMENT OF TRANSPORTATION HAZARD CLASSES 63
APPENDIX B - STATISTICAL ANALYSIS OF THE TRANSPORTATION-RELATED
RELEASES OF CHEMICAL SUBSTANCES 69
B.I Introduction 70
B.2 The HAZMAT Data Base 71
B.3 Description of Statistical Methods Used in This
Analysis 73
B.3.1 Analysis of Variance (ANOVA) 73
B.3.2 Chi-Square Test of Homogeneity 84
B.4 Other Factors Considered in the Analysis 84
B.4.1 DOT Hazard Class 86
B.4.2 Physical State 87
B.4.3 Mode of Transportation 87
B.5 Analysis of Variance and Summary Statistics for the
Percent of Shipment Released 87
B.6 Analysis of Variance, Summary Statistics, and
Confidence Limits for the Percent of Container
Contents Released 95
B.7 Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Shipment
Released 105
B.8 Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Container
Contents Released 114
B.9 Correlation Between Quantity Released and Shipment
Size 121
B.10 Conclusion 138
APPENDIX C - METHODS FOR ESTIMATING AVERAGE SHIPPING
DISTANCES 141
C.I Steps Common to All the Methods 142
C.I.I Identify the CTS Commodity Code 142
C.I.2 Identify the Geographic Origin of Shipment .... 143
C.I.3 Locate Values for Tons and Ton-Miles Shipped
in CTS Publications 145
C.I.4 Calculate the Average Shipping Distance of
the Chemical 145
-------
TABLE OF CONTENTS (continued)
Page No.
C.2 Selecting a Method 145
C.3 Descriptions of the Methods 146
C.3.1 Method C-l 146
C.3.2 Method C-2 146
C.3.3 Method C-3 147
-------
LIST OF TABLES
Page No.
Table 1. Estimated Transportation of Industrial Inorganic
and Organic Chemicals by Mode in 1977 8
Table 2. Cause of Failure by Mode, 1976-84 10
Table 3. Sources of Information Useful in Determining a
Chemical's Physical State at Standard Conditions of
Temperature and Pressure 18
Table 4. Associations That May Provide Production Volume
Data for Chemicals 23
Table 5. Summary of DOT Hazard Classes 30
Table 6. Values for the Fraction of Container Contents
Released (To Be Used If the Mode of Transportation
Is Known) 31
Table 7. Values for the Fraction of Container Contents
Released (To Be Used If Physical State and Mode of
Transportation Are Known) 32
Table 8. Values for the Fraction of Container Contents
Released (To Be Used if Physical State, DOT Hazard
Class, and Mode of Transportation Are Known) 33
Table 9. Shipping Patterns for STCC 28999 39
Table 10. Locations and Capacities of Ethylene Oxide
Manufacturing Plants, January 1, 1987 44
Table 11. Sample DOT Packaging Requirements Including
Ethylene Oxide 48
Table 12. Shipments of Ethylene Oxide by Railroad Tank Cars
Estimated from ICC Data 49
Table 13. Locations and Capacities of Formaldehyde
Manufacturing Plants, January 1, 1986 51
Table B-l. Types of Data Contained in the HAZMAT Data Base 72
Table B-2. HAZMAT Data Used in the Statistical Analysis 74
Table B-3. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by DOT Hazard Class 75
xii
-------
LIST OF TABLES (continued)
Page No.
Table B-4. The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Physical State 85
Table B-5. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by the Mode of
Transportation 88
Table B-6. Summary Statistics for the Percent of Shipment
Released (SHIPREL) for Each Physical State (Gas,
Liquid, Solid) 89
Table B-7. Summary Statistics for the Percent of Shipment
Released (SHIPREL) by the Mode of
Transportation 89
Table B-8. Summary Statistics for the Percent of Shipment
Released (SHIPREL) by the Physical State and Mode
of Transportation 89
Table B-9. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by the Physical State
(Liquid) 91
Table B-10. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by the Physical State
(Solid) 92
Table B-ll. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by the Physical State
(Gas) 93
Table B-12. Summary Statistics for the Percent of Shipment
Released (SHIPREL) by the Physical State and
Hazard Class 94
Table B-13. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by the Mode of
Transportation for Each Physical State and Each
Commodity Class 96
Table B-14. Summary Statistics for the Percent of Shipment
Released (SHIPREL) by the Commodity Class (DOT
Hazard Class), Physical State, and Mode of
Transportation 97
xiii
-------
LIST OF TABLES (continued)
Page No.
Table B-15. Analysis of Variance Results for the Percent of
Container Contents Released (CONTREL) by the
Commodity Class (DOT Hazard Class) 99
Table B-16. Analysis of Variance Results for the Percent of
Container Contents Released (CONTREL) by the Mode
of Transportation 100
Table B-17. Summary Statistics for the Percent of Container
Contents Released (CONTREL) for Each Physical
State (Solid, Liquid, Gas) 101
Table B-18. Summary Statistics for the Percent of Container
Contents Released (CONTREL) by the Mode of
Transportation 101
Table B-19. Summary Statistics for the Percent of Container
Contents Released (CONTREL) by the Physical State
and Mode of Transportation 101
Table B 20. Analysis of Variance Results for the Percent of
Container Contents Released (CONTREL) by Physical
State (Liquid) for Commodity (DOT Hazard)
Classes 2, 4, 6, 8, 9, 20, 25, and 95 102
Table B-21. Analysis of Variance Results for the Percent of
Container Contents Released (CONTREL) by Physical
State (Solid) for Commodity (DOT Hazard)
Classes 10, 30, 35, and 60 103
Table B-22. Analysis of Variance Results for the Percent of
Container Contents Released (CONTREL) by Physical
State (Gas) for Commodity (DOT Hazard) Classes 45,
50, 55, and 65 104
Table B-23. Summary Statistics for the Percent of Container
Contents Released (CONTREL) by the Commodity
Class (DOT Hazard Class) and Physical State 106
Table B-24. Chi-Square Test of Homogeneity of Results for the
Percent of Container Contents Released (CONTREL)
by Mode of Transportation for each Combination
of Physical State and Commodity Class (DOT Hazard
Class) 107
xiv
-------
LIST OF TABLES (continued)
Page No,
Table B-25.
Table B-26.
Table B-27.
Table B-28.
Table B-29.
Table B-30.
Table B-31.
Table B-32.
Table B-33.
Table B-34.
Summary Statistics for the Percent of Container
Released (CONTREL) by the Mode of Trarsportation
for Each Physical State and Each Commodity Class
(DOT Hazard Class)
The Frequency Distribution and Chi-Square Test
Results for the Percent of Shipment Released
(SHIPREL) by Physical State (Liquid)
The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Physical State (Solid)
The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Physical 5tate (Gas)
The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Mode of Transportation
108
110
112
113
115
Chi-Square Test of Homogeneity Results for the
Percent of Shipment Released (SHIPREL) by Mode of
Transportation for Each Physical State and Each
Commodity Class (DOT Hazard Class)
116
The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Mode of Transportation for
Liquids
The Frequency Distribution and Chi-Square Test
of Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Mode of Transportation for
Solids
The Frequency Distribution and Chi-Square Test
of Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Mode of Transportation for
Gases
Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by the Mode of
Transportation for Each Physical State and Each
Commodity Class
117
118
119
120
xv
-------
LIST OF TABLES (continued)
Page No.
Table B-35. The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Physical State 122
Table B-36. The Frequency Distribution and Chi-Square Test
of Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Physical State
(Liquid) 126
Table B-37. The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Physical State
(Solid) 128
Table B-38. The Frequency Distribution and Chi-Square Test
of Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Physical State
(Gas) 129
Table B-39. The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Mode of
Transportation 130
Table B-40. The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Mode of
Transportation for Liquids 135
Table B-41. The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Mode of
Transportation for Sol ids 136
Table B-42. The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Container
Contents Released (CONTREL) by Mode of
Transportation for Gases 137
Table B-43. Correlation Coefficient Between Quantity Released
and Shipment Size Classified by Physical States
arid Commodity Class 139
Table C-l. 1977 Commodity Transportation Survey Production
Area Descriptions by Division 144
xvi
-------
LIST OF FIGURES
Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.
Figure B-l.
Figure B-2.
Figure B-3.
Figure B-4.
Figure B-5.
Figure B-6.
Figure B-7.
Sample worksheet for predicting the amount of
chemical released because of transportation
accidents
Sample retrieval from the Organic Chemical
Producers' Data Base by CAS registry number ..
Sample worksheet for predicting the amount of
DEHP released annually because of rail, tank
truck, and air transport accidents
Sample worksheet for predicting the amount of
formaldehyde released because of railroad
accidents
Sample worksheet for predicting the amount of
formaldehyde released because of tank truck
accidents
Percentage bar chart for the frequency distribution
of the percent of shipment released for liquids ...
Percentage bar chart for the frequency distribution
of the percent of shipment released for solids ....
Percentage bar chart for the frequency distribution
of the percent of shipment released for gases
Percentage bar chart for the frequency distribution
of the percent of shipment released for the air
mode of transportation
Percentage bar chart for the frequency distribution
of the percent of shipment released for the water
mode of transportation
Percentage bar chart for the frequency distribution
of the percent of shipment released for the rail
mode of transportation
Percentage bar chart for the frequency distribution
of the percent of shipment released for the highway
mode of transportation
Page No.
14
20
36
46
54
77
78
79
80
81
82
83
xvii
-------
LIST OF FIGURES
Figure B-8.
Figure B-9.
Figure B-10.
Figure B-ll.
Figure B-12.
Figure B-13.
Figure B-14.
Percentage bar chart for the frequency distribution
of the percent of container released for liquids ..
Percentage bar chart for the frequency distribution
of the percent of container released for solids ...
Percentage bar chart for the frequency distribution
of the percent of container released for gases
Percentage bar chart for the frequency distribution
of the percent of container released for the air
mode of transportation
Percentage bar chart for the frequency distribution
of the percent of container released for the water
mode of transportation
Page No.
123
124
125
131
132
Percentage bar chart for the frequency distribution
of the percent of container released for the rail
mode of transportation 133
Percentage bar chart for the frequency distribution
of the percent of container released for the highway
mode of transportation 134
xvi 1
-------
1. INTRODUCTION
This report explains how to develop quantitative information on
annual expected releases of manufactured chemicals between the point of
origin (i.e., manufacturing location) and the point of first delivery in
commerce. The data generated (that is, the expected number of releases
and total quantity released annually) are useful in analyses of chemicals
transportation. Examples of appropriate uses of this information are:
(1) to compare expected releases of alternative chemicals that would be
shipped routinely from a manufacturer for a particular use, or (2) to
weigh the effects of potential releases associated with one mode of
transportation with those related to another mode for the same chemical.
1.1 Purpose and Scope
The purpose of this report is to present methods of calculating the
expected annual release of individual chemicals resulting from
transportation-related accidents. The methods are based on historical
patterns of chemical shipments and accidental releases of those
chemicals. Methods of calculation are presented for releases by each of
four major, transportation modes (truck, rail, air, and waterborne
transport ).
The scope of this report and the methods described are limited to
manufactured chemicals when distributed in commerce and to accidental
releases occurring en route, that is, between terminal points. Because
of these limitations, the methods do not include calculations of releases
occurring during loading and unloading, although hazardous material
releases do occur during these activities as well as en route (ICF 1984,
OTA 1986). Limiting this report to manufactured chemicals means that
some hazardous material groups regulated by the U.S. Department of
Transportation (DOT) and the U.S. Environmental Protection Agency (EPA),
such as etiological agents, explosives, crude oil, and hazardous waste,
are not included. In addition, pipelines have not been considered along
with the other modes of transportation.
1.2 Structure of the Report
Because this report focuses on a method of calculation, much of the
text is devoted to explaining how required information on shipping
patterns and accident statistics can be accessed and used in this
method. Section 1.3 discusses the primary sources of data used in the
report and their limitations. A brief overview of manufactured chemical
At present, available data are insufficient to predict chemical
releases from waterborne transport.
-------
shipping patterns in the United States, and the causes of chemical
releases during transportation, are presented in Section 2. The body of
the method is contained in Section 3. This section presents a
step-by-step method of calculating the expected quantity of a chemical
that may be released annually during transportation accidents. Because
some steps of the method cite optional data sources and methods of
application, examples are given to demonstrate those options.
The appendices provide supplementary information that is used in
performing the calculations described in Section 3. A description of the
U.S. Department of Transportation (DOT) hazard classes of chemicals is
contained in Appendix A, and a statistical analysis of historical
transportation-related release data that was performed for this report is
discussed in Appendix B. This statistical analysis focuses on the
significance of the physical state of the chemical, the DOT hazard class,
and the mode of transportation as those factors contributing to the
quantity of a chemical released when an accident occurs. Appendix C
presents methods of estimating average shipping distances of chemicals.
1.3 Sources of Information
An analysis of accidental releases of a chemical distributed in
commerce requires information on commodity shipping patterns, namely, the
quantities of that chemical shipped and the average shipping distances by
various modes of transportation. Historical accident and release data on
the chemical, or a class of chemicals, are also needed to project
frequencies and quantities released.
At this time, U.S. commodity shipping data and historical accident
data are not archived in one system, nor are the available records stored
by one classification scheme. Rather, these data are compiled by various
federal agencies and private industry organizations in distinct data
bases, and the reader will need to consult a number of references in
order to complete the calculations described in Section 3.
To illustrate the variety of information sources needed to establish
shipping and accident patterns, several key sources of information that
were consulted in the preparation of this report are described briefly
below. The use of each of these sources of information is explained in
detail in the description of the methods presented in Section 3 of this
report.
• U.S. Department of Commerce (USDOC), Bureau of the Census, 1977
Census of Transportation, Commodity Transportation Survey (CTS).
The most recent compilation of detailed information on quantities
of chemical commodities shipped in the United States is the 1977
Census of Transportation, Commodity Transportation Survey (USDOC
1981a,b,c). Commodity shipping data are compiled in the CTS
according to the Commodity Classification for Transportation
-------
Statistics (TCC) codes. This numerical system of coding
corresponds closely to that of another key system, the Standard
Transportation Commodity Code (STCC) (STCC 1972). Because of the
similarities of the TCC and the STCC coding systems, the CTS data
can be used to predict the fraction of the total quantity of a
manufactured chemical that is shipped by each major mode of
transportation, and also to identify the average shipping distance
by each mode of transportation.
The age of the data available from the 1977 Census of
Transportation is a major source of uncertainty when calculating
estimated quantities released. It is possible that commodity
shipping patterns presented in the 1977 CTS have changed
substantially over the last ten years. A 1982 Commodity
Transportation Survey intended to update the 1977 CTS was
determined to be unreliable by the Bureau of the Census.
Nevertheless, the unofficial statistics developed from this 1982
transportation census can be obtained from the Bureau of the
Census, Transportation Census Branch, (301) 763-4363. Note:
these data are not to be cited in any published reports and were
not used in these methods.
U.S. International Trade Commission (USITC) annually publishes
Synthetic Organic Chemicals, United States Production and Sales.
This is the preferred source of information for locating annual
production and sales data for specific chemicals. Data are
reported by producers only for those items that exceed minimum
production volumes or annual sales. Chemicals are grouped by
categories, e.g., cyclic intermediates, organic pigments,
plasticizers. These groups are assigned section numbers so that a
specific chemical can be located by referring to the "Alphabetical
Chemical Index" in the appendix of each publication.
Stanford Research Institute Directory of Chemical
Producers-United States (see SRI 1987). This compendium of
information on manufacturers of chemicals is a current source of
annual production capacity data for a limited number of specific
chemicals. The PRODUCTS section of this report is an alphabetical
listing of chemicals and end uses of chemicals that is linl-ed to
information on the manufacturers and their locations (SRI 1987).
Only chemicals produced in commercial quantities (annual
production of 5,000 pounds or $5,000 value) are listed. The
annual production capacity data obtained from this source can be
used in these methods as an upper limit of the quantity of the
chemical that could be distributed in commerce.
U.S. Department of Transportation, Research and Special Programs
Administration, Hazardous Materials Data Base (HAZMAT). A primary
source used in this study for estimating predictive release
-------
factors based on the DOT hazard class of a chemical was the DOT's
HAZMAT Data Base, a primary data base of The Hazardous Materials
Incident Reporting Subsystem (HMIS). This data base is maintained
on the DOT's Digital Electronic Corporation DEC10 computer in
Cambridge, Massachusetts. As of 1986, HAZMAT contained 151,067
records documenting inadvertent releases of hazardous materials.
The data in HAZMAT are provided by carriers on the Hazardous
Materials Incident Report (form DOT F 5800.1) whenever there is an
unintentional hazardous materials release during interstate
commerce. The types of data contained in HAZMAT are listed in
Appendix B, Table B-l.
U.S. Department of Transportation, Associate Administration for
Motor Carriers, Motor Carrier Information Division. This division
maintains a data base on highway vehicle accidents only. It is
derived from accident reports to the Federal Highway
Administration. Currently, computer tapes on the accident data
base are available for the period of 1980-1985 and can be
purchased. Additional information can be obtained by telephoning
the DOT Motor Carrier Information Division at (202) 366-2971.
Note that a similar type of information on railroads is available
through the Systems Support Division (202) 366-2760.
U.S. Department of Commerce, Interstate Commerce Commission
(ICC), Waybill Statistics for Railroad Transportation
Information. The ICC waybill data base is an important source of
information on the volumes and distances that specific STCC-coded
commodities are shipped by rail. While the data are updated
annually, there are restrictions on their use, since parts of the
data are confidential. Nonconfidential data can be purchased
through ALK Associates in New Jersey (609) 683-0220. Government
agencies may access the data at no charge by contacting Mr. Jim
Nash at the ICC (202) 275-6864.
Sources of information on waterborne traffic accidents. A
current limitation in calculating the total expected quantity of a
chemical released from transportation accidents is the lack of
specific accident statistics on waterborne transportation. The
critical data that are missing include the accident rate (i umber
of accidents per mile) of chemical shipments and the probability
of a release given an accident. The U.S. Department of
Transportation, U.S. Coast Guard, Office of Marine Safety,
maintains two data bases on transportation accidents that could
provide some of this information in the future. These data bases
are as follows:
-------
- Marine Investigation Division, Marine Casualty Data Base. This
division of the Office of Marine Safety compiles general data on
waterborne traffic in the U.S. Although no specific statistics
have yet been compiled on hazardous chemical cargo and its
relationship to barge or other waterborne traffic accidents,
such data may be developed in the future. For current
information on the Marine Casualty Data Base, contact LCDR Tom
Purtell at (202) 267-1430.
- Marine Pollution Data Base. Data on pollution incidents
involving releases of chemicals or oil to water are stored in
the data base maintained by the Office of Marine Safety.
Although not directly applicable to the calculation of expected
quantity released, this data base may be useful in conducting a
fate analysis of a chemical accidentally released to waters of
the U.S. Contact Ms. Mary Robey (202) 267-0455.
The Office of Technology Assessment (OTA), in 1986, published
the report, Transportation of Hazardous Materials. The OTA report
presents results of a study of federal and state regulation of the
transport of radioactive materials, munitions, commodities
(manufactured chemicals), and hazardous wastes. An overvie/v of
hazardous materials shipping patterns, which was contained in the
OTA report, is cited in Section 2 of this report.
-------
2. PATTERNS OF COMMERCIAL DISTRIBUTION OF MANUFACTURED CHEMICALS
This section of the report discusses general patterns of chemicals
transportation in the United States. Many of the manufactured chemicals
that are distributed in commerce exhibit dangerous properties, such as
flammability, reactivity, or acute toxicity, which require special
packaging and handling during loading, unloading, and in transit.
Manufactured chemicals with dangerous properties are regulated as
hazardous materials (or hazardous substances) by DOT and by EPA. The
following discussion summarizes available data on the transportation of
manufactured chemicals in the United States for major modes of
transportation excluding pipelines.
2.1 Distribution of Manufactured Chemicals by Transportation Mode
Four modes of transportation are used to carry most manufactured
chemicals in the United States: (1) truck, (2) rail, (3) water, and
(4) air. Table 1 presents estimates of the tons of manufactured
chemicals transported by each of these modes (USDOC 1981a). The table
shows that rail and truck transport were the modes of transportation by
which the largest total quantities of manufactured chemicals were shipped
in 1977. Waterborne transport ranks third in quantities shipped and
ton-miles accumulated, and air transport was the least used mode of
transportation for manufactured chemicals. Patterns of transport by each
of these modes of transportation are discussed below.
2.1.1 Transportation of Manufactured Chemicals by Rail
In 1977, railroads hauled 65.9 million tons of manufactured
chemicals. Rail shipments of manufactured chemicals are usually made by
tank car. When ranked by tonnage, rail transportation accounts for
33.4 percent of all industrial inorganic and organic chemicals shipped.
Some chemicals are also carried by hopper cars and intermodal flat cars
i.e., flat cars carrying intermodal tanks (OTA 1986). In 1977, the
average distance of a rail shipment of manufactured chemicals was
approximately 500 miles (USDOC 1981a).
2.1.2 Transportation of Manufactured Chemicals by Truck
As shown in Table 1, truck transport was the mode carrying the
greatest tonnage of manufactured chemicals in 1977, although trucks
ranked second after rail transport in total ton-miles shipped. According
to the USDOC Bureau of Census 1977 Commodity Transportation Survey (CTS),
trucks transported 77 million tons of chemicals in 1977, with an average
shipping distance of 175 miles (USDOC 1981a,b,c).
-------
Table 1. tbtimdted Iianspoitat ion of Industrial Inorganic.
and Organic Chemicals by Mode in 1977
Mode
Rail
FruCK
Water-
Air
Other
Total
Tons transported
(thousands)
65,930
77,038
15,386
12
38,985
197,j51
Percent
of totala
33 4
39 0
7 8
0 006
19 8
100
Ton-mi les
(mi 1 lion)
32,834
14,252
8,546
9
1,464
57,105
Percent
of totala
57.5
25 0
15.0
0.02
2.5
100
dTotals may not equal 100 because of rounding
Source USDOC (1981d)
-------
2.1.3 Transportation of Manufactured Chemicals by Waterborne Vessels
Waterborne vessels rank third in ton-miles and third in tonnage of
manufactured chemicals shipped in 1977. In 1977, an average shipping
distance for manufactured chemicals was approximately 550 miles (USDOC
1981a, OTA 1986). In its evaluation of hazardous materials shipments by
water, OTA (1986) noted a trend toward declining numbers of bulk
shipments of some chemicals classified as hazardous materials. The total
tonnage of waterborne shipments of chemicals dropped 13 percent between
1977 and 1982 (OTA 1986).
2.1.4 Transportation of Manufactured Chemicals by Air
According to the CTS (USDOC 1981a), only 12,000 tons of manufactured
chemicals were transported by air in 1977. This accounted for less thar
1 percent (0.006 percent) of the total tonnage shipped in 1977.
Although quantities of manufactured chemicals carried by air are
small, the distances shipped may be large. Manufactured chemicals
including cosmetics, Pharmaceuticals, and agricultural chemicals account
for 80 percent of hazardous materials shipped by air in 1977 (OTA 1986).
2.2 Factors Contributing to Transportation Releases
OTA (1986) reviewed the causes of transportation-related failures
reported to the DOT Hazardous Materials Information System (HMIS) between
1976 and 1984. These data, summarized in Table 2, indicate the nu-nber of
times each type of failure was reported for various modes of
transportation.
Although the frequent causes of failures cited in Table 2 vary by
mode of transportation, it can be seen that external punctures and loose
and defective fittings were frequent causes of releases reported to the
HMIS in the eight-year period studied. From the data, OTA concluded that
such failures are typical of loading and unloading operations or of cargo
shifts during transport (OTA 1986). It should be noted, however, that
not all the failure codes are mutually exclusive. For example, OTA could
not determine with certainty whether an external puncture occurred
because of a vehicle accident or because other cargo shifted/fell during
loading and unloading.
-------
o en ex) r-. -H
CM CO CD (NJ
O u3 CM O CO
CT> o CO to CM
•—i o en
OO CD CO CD
-C 13
en i_
•- t)
C\J --H —«
co ro oj
—« Co OJ
ro co ^r •—<
CO «-H CO tO
OJ CO
U~> OJ CO OJ *—I .-H
•—t en oj
—• ro
^-i r-^ to
O U~) OJ LO CD CO
*J-
ro
OJ o r--
CM co •—i
CO CO
Oo uT)
cri r--. co oj co ^r —i *g- o
to <—i co •—<
.
oo en
CO CNJ OJ
u~> ro CM
.-* LT>
c
01
Q)
O)
-JZ
c
0)
X
— :
in
QJ
Q-
'0
C
QJ
C
t-
o
t:
0
1/1
o
o
-M
+-J
QJ
U
QJ
Q)
i/i
CT
C
-M
OJ
O
O
L-
O)
c
c
0)
1 —
ft)
QJ
=3
t)
4-
E
O
o
r^-
QJ
"O
I/)
>^
-o
O
QJ
~o
QJ
~>
OJ
fO
QJ
E
.C
O
-*-)
T>
C
o
u
Q)
-t-»
zt
_Q
0)
W
O
Q.
O
t"
~n
o
^.
*D
OJ
_C
C/l
0
Oi
(0
^
m
0
u
OJ
Q-
0
!_
CL
E
0)
"O
l_)
u
•t)
QJ
O
QJ
E
4-
M-
O
C in
4-J QJ
OJ Q)
C
O
o
u
"
l_
4-J
u
QJ
QJ
O
_t_)
+J
OJ
rs
(C
fC
-f-1
QJ
^-
10
-------
A recent statistical analysis of selected failure codes reported in
the HAZMAT data from 1975 to 1986 indicates that more releasing incidents
occur at terminals than during accidents en route, but that the average
quantity released per incident is greater for vehicular accidents than at
terminal points. For rail, truck, and waterborne transport the number of
releases from "loading-unloading," "dropped in handling," or "hose burst"
were three to four times greater than the number of releases from
"vehicle accident." The mean quantity released from incidents in the
last category was an order of magnitude greater than any of the first
three categories, which are related to handling at terminal points
(Versar 1987).
Previous studies have shown that the probability of a hazardous
materials transportation release is somewhat related to traffic density
and physical obstructions. French and Richards (1973) found that the
highest percentage of barge casualties on the West Gulf Intracoastal
Waterway occurred at locations with den:.e traffic or obstructions such as
bridges or pilings. Similarly, ICF (1984) analyzed 1980-1982 truck
accident and volume data from Texas, California, New Jersey, and
Massachusetts. Combining the state analyses with an analysis of DOT's
HAZMAT data base, ICF estimated that the truck accident rate (for
accidents in which there is a release of hazardous materials) is highest
in urban areas (7.3 x 10"' releasing accidents per mile), lower on U.S.
or state highways (4.5 x 10"' releasing accidents per mile), and lowest
on U.S. interstates (1.3 x 10"7 releasing accidents per mile) (ICF
1984). Another analysis of 1980, 1981, and January, February, and March
1984 data from the HAZMAT data base, combined with en route vehicular
accident rates and collision data provided by DOT's Bureau of Motor
Carrier Safety, confirmed this range of releasing accidents for tank
trucks. It was calculated that, on the average, tank trucks are involved
in 3.5 x 10"' releasing accidents per mile (USEPA 1985).
11
-------
3. METHOD OF CALCULATING THE EXPECTED QUANTITY RELEASED OF
COMMERCIALLY AVAILABLE CHEMICALS
The amount of a chemical expected to be released because of
transportation accidents can be calculated using several types of
information about shipment of the chemical. The information includes:
(1) the quantities that will be shipped, (2) the mode(s) of shipment, and
(3) historical accident data. Engineering judgment is required when su
-------
i— ro Q-
cn Hi
QJ -O
ra ^
n i_)
o O
<-. o
H- X
O •-
O T>
U~> C
ID tU
CL
« e
OJ Q.
•— (O "O
E
QJ
14
-------
.
O C -—
•— U Q.
o $
•- o
O
in
01
o ^ ^
o o o
OJ 0) ED
TD T3 T3
O O O
E E E
OJ
-C
i
OJ 0) CD CU
I- 1/5 en to
CO t— t—
15
-------
applying the method to a specific chemical, one should record informs*ion
on values obtained for the required parameters on a copy of the sample
worksheet.
For some of the steps, there are optional sources of information. In
particular, the availability of Interstate Commerce Commission (ICC)
waybill data on railroad shipments allows for several "shortcuts" in
calculating releases by rail only. These options are noted where
applicable.
Note also that the information required to calculate releases during
waterborne transportation is incomplete at present. However, because
potential sources of needed information (e.g., accident statistics) may
become available in the near future, these sources are described in this
method, and spaces for calculating waterborne transportation releases are
included in the worksheet.
Step 1. Provide information that characterizes the chemical in
question. The information obtained in this step (DOT hazard
class, physical state, STCC code, CAS registry number, and
quantity shipped) is used in subsequent steps to determine
the average annual release of a chemical because of
transportation-related accidents. To obtain the necessary
information, the following actions should be taken:
• For a given chemical, determine its DOT hazard class.
Classes of DOT regulated chemicals are listed in 49 CFR
172.101 (USDOT 1986b). Descriptions of each class are
also presented in Appendix A of this report, and examples
are included in the sample calculations in Section 3.2
(note that the identification of the DOT hazard class is
helpful but not essential to completing the calculations
described in this method).
• If the chemical in question is a newly manufactured
chemical (e.g., a PMN chemical), then an appropriate DOT
hazard class may be assigned by using the definitions in
Appendix A.
• Ascertain the Standard Transportation Commodity Code
(STCC) of the chemical. STCC codes are derived from the
Standard Transportation Commodity Code Tariff, No. 1-A
(STCC 1972). A complete listing of STCC codes can be
purchased from the Western Truck Line Company, 222 South
Riverside Plaza, Chicago, Illinois 60606. The telephone
16
-------
number is (312) 648-7849. Individual STCC codes may be
obtained by contacting Mr. Gordon Anderson at the National
Motor Freight Classification Board (703) 838-1811.
• Determine the physical state of the chemical (i.e.,
whether the chemical is a solid, liquid, or gas at the
standard conditions of 25°C and 1 atmosphere
pressure). Sources of information that may be helpful in
determining a chemical's physical state are listed in
Table 3. Note that although compressed gases are
transported as liquids, they are considered gases for the
purpose of this method because they volatilize readily upon
release from a shipping container.
• Determine the CAS registry number of the chemical. CAS
registry numbers can be found in the CRC Handbook of Data
on Organic Compounds (CRC 1985) or in the USEPA TSCA
Inventory (USEPA 1986). Online computerized data bases
that can be accessed for CAS registry numbers include the
National Library of Medicine's Chemline and the Chemical
Abstracts Registry File, which is part of the DIALOG online
system.
Step 2. Estimate the annual quantity shipped using one of the
following options. Option 1 is preferred. Option 4 can be
used only if ICC information is available on rail transport.
Option 1: The U.S. International Trade Commission
publication, Synthetic Organic Chemicals--United
States Production and Sales, can be used to find
the amount of the chemical sold; this amount is
then assumed to be the quantity shipped. The
limitation in using this source is that sales
information is often preserted for categories of
chemicals rather than for individual chemicals.
An advantage of this data
-------
Table 3. Sources of Information Useful in Determining a Chemical's
Physical State at Standard Conditions of Temperature and Pressure
Title Comment
Chemical Engineer's Handbook, 6th ed (1984) See Section 3 of handbook
CRC Handbook of Data on Organic Compounds, Vols I and II (1985) See alphabetical listing of chemicals.
CRC Handbook of Physics and Chemistry, 6th ed. (1986) See Sections C-42 through C-553.
Handbook of Toxic and Hazardous Chemicals and Carcinogens (1985) See alphabetical listing of chemicals
The Merck Index (1983) See alphabetical listing of chemicals.
18
-------
Also, many facilities use some or all of the
chemicals produced in onsite processes.
Therefore, some knowledge of the industry may be
required to make an educated estimate of the
amount of chemical shipped based on production
capacity data alone.
Option 3: A third source of annual production volume data is
the Chemical Producers' Data Base. This system
consists of three files: organic chemicals,
inorganic chemicals, and dyes and pigments. A
sample printout from the data base is presented in
Figure 2.
Information on the Organic Chemical Producers'
Data Base can be obtained by contacting th^ U.S.
Environmental Protection Agency, Office of
Research and Development, Hazardous Waste
Engineering Laboratory, 26 Saint Clair Street,
Cincinnati, Ohio 45268; contact Mr. Jerry •Jaterman
(513) 569-7214. Note that much of the ini armation
in the Organic Chemical Producers' Data Base is
ten years old, and, according to Mr. Waterman,
there are no plans to update it.
Option 4: Another source of information on chemical
production is the 1977 EPA TSCA inventory data,
available online as the TSCAPP subsystem of
Chemical Information Systems (CIS).
Nonconfidential business data included in TSCAPP
are: (1) names of reported chemicals,
(2) production volume range, and (3) manufacturing
plant location. CIS plans to supplement TSCAPP
with information from the EPA 1986-87 upda ,e of
the TSCA Inventory. If these chemical production
data are used, some knowledge of the particular
industry may be needed to estimate the quantity
shipped versus the quantity used onsite.
Codes for production volume range in TSCAPC are as
follows:
19
-------
oouo
00200
OOlOO
00*0*
oosoo
00600
00)00
00600
00400
OIUOO
01100
0)200
01)00
01400
0|SOO
0|6oO
01700
oi«oo
0|900
02000
02100
02200
02100
02*00
OJSoO
02600
02700
02800
02<>00
05000
05100
05200
05100
05100
05500
05600
057«0
05000
05?00
0*000
0*100
OqiOO
Ou5oO
0«400
CAS »95ao7
1350
SVNONtMS
NLN
NI03H NUMBER
lUtlCUT
»TOLUENt-2,«-OI»HIN|
2,4-OIAMlNOTULUEhE
»-Mt!HYL-H-PHEML£NEDIAMlNE
N-IOLU£N£OIAM|NE
2,«-IOL»L£HED|AMINt
1, }-b£n/ENEOIAhlh£, 4-MEIHYL-
lu*
id CI D
liOlU^O J
0^050201
Oil noioa
Oo080<;06
ObU70
-------
Code Volume
0 Less than 1,000 Ib
1 1,000 to 10,000 Ib
2 10,000 to 100,000 Ib
3 100,000 to 1,000,000 Ib
4 1,000,000 to 10,000,000 Ib
5 10,000,000 to 50,000,000 Ib
6 50,000,000 to 100,000,000 Ib
7 100,000,000 to 500,000,000 Ib
U No Report
To obtain additional information on accessing
TSCAPP, one should contact:
Ms. Laurie Donaldson
Chemical Information Systems, Inc.
7215 York Road
Baltimore, MD 21212
1-(800) 247-8737 or
(301) 321-8440
Option 5: FOR RAIL ONLY: The Interstate Commerce Commission
(ICC) maintains a waybill file of annual
quantities of all commodities shipped by rail.
Shipping information on individual commodities can
be searched in this file by STCC code (see Step
1). USEPA or other government agency personnel
can obtain current waybill data on a specific
chemical by contacting Mr. Jim Nash of the ICC at
(202) 275-6864.
Because the ICC waybill data are based on
1 percent (or 6 percent for 1986 or later) of the
actual quantities shipped, multiply the value
given for quantity shipped in the waybill file by
the appropriate factor (100 for 1 percent waybill
data, 16.7 for 6 percent waybill data) in order to
estimate the actual quantity shipped by rail.
Then, because the ICC quantities are reported in
tons, divide by 1.1 to convert to kkg.
If ICC data are used and only rail transport is
being considered, enter the value calculated from
waybill data for quantity shipped (in kkg) on the
line marked "S" on the worksheet, and then skip to
Step 4, Option 2.
21
-------
If the annual quantity shipped cannot be estimated using an;/
of these options, it may be helpful to contact trade
associations or professional organizations for shipping
information on a specific chemical. Potential contacts for
this information are listed in Table 4.
Enter the value obtained for annual quantity shipped on the
worksheet (Step 2). Convert this value to metric tons (kkg),
and enter the corrected value on the line designated by "S"
on the worksheet.
Step 3. Calculate the fraction of the total annual quantity of
chemical shipped that is transported by each mode of
transportation. This calculation requires data from the
Bureau of the Census Commodity Transportation Survey Summary
(CTS Summary) for 1977 (USDOC 1981a). Commodities included
in the CTS Summary are classified using the Commodity
Classification for Transportation Statistics (TCC) codes.
The system of numbering within the TCC codes closely
parallels that of the STCC codes (see Step 1). Therefore,
for the purposes of this method, the data on commodity
shipments in the CTS Summary are searched by matching the
STCC code, obtained for the specific chemical in Step 1, with
the most closely related TCC code listed in Table 2 of the
CTS Summary. It is assumed that the fraction of the STCC (or
TCC) commodity group that is shipped by each mode of
transportation is representative of the shipping pattern for
each chemical within that commodity group. For calculating
the fraction that is shipped, the following procedures are
used:
(a) Using the STCC code of the chemical (from Step 1), find
the values for tons shipped in the CTS Summary, Table 2,
Column B. The total quantity of the TCC commodity group
shipped by all modes of transportation is listed first,
followed by values for tons shipped by different modes of
transportation: rail, motor carrier (ICC and non-ICC),
private truck, air, water, parcel delivery, and other
and nknown. For truck transport, sum the values for
tons shipped for motor carriers (total of ICC and
non-ICC), and private truck categories.
(b) Calculate the fraction of the STCC commodity group that
is shipped by each mode of transportation by dividing the
value for quantity shipped by each mode by the
corresponding value for total quantity shipped.
22
-------
Table 4. Associations That May Provide Production Volume Data for Chemicals
Association
Address
Telephone number
American Chemical Society
Chemical Marketing Research Association
Chemical Specialties Manufacturers Association
National Association of Chemical Distributors
Synthetic Organic Chemical Manufacturers Association
1155 16th St. NW 202-872-4600
Washington, DC 20036
139 Chestnut Ave. 212-727-0550
Staten Island, NY 10305
1001 Connecticut Ave. NW 202-872-8100
Washington, DC 20031
1110 Vermont Ave. NW, Suite 1150 202-296-9200
Washington, DC 20005
1075 Central Park Ave. 914-725-1492
Scarsdale, NY 105B3
23
-------
(c) Enter the calculated values for fraction shipped by each
mode of transportation on the worksheet on the line
designated by "F."
Alternative approach: Use this method if truck and rail are
the primary modes of transport and ICC (rail) data are
available to a level of STCC detail greater than the CIS
data. In this case, subtract the ICC (rail) quantity from
the total annual quantity shipped (Line 5) to estimate the
quantity transported by truck. For example, this approach
can be helpful if ICC data are available for STCC 2818144
but CIS data are available only for STCC 2818.
Step 4. Estimate the quantities shipped annually for each mode of
transportation.
Option 1: This option makes use of information developed in
Steps 2 and 3. Multiply the total estimated
quantity shipped (parameter "S" from Step 2) by
the fraction shipped by each mode of
transportation (parameter "F" from Step 3). Enter
the results on the line identified by "W" on the
worksheet.
Option 2: FOR RAIL ONLY: If estimating for rail, one can
use the ICC waybill data. The total quantity
shipped by rail can be obtained directly, as
described in Step 2, Option 4.
Enter the value for quantity shipped by rail (in kkg) on the
worksheet on the line marked "W."
Step 5. Estimate average quantity per shipment for each mode of
transportation.
Option 1: Information on standard volumes of liquids or
compressed gases shipped by tank truck or rail is
available in a recent article by Genereaux et al.,
"Transportation and Storage of Fluids," in Perry's
Chemical Engineers Handbook, 1984. The following
standard volumes can be used to estimate the
average quantity per shipment for tank trucks and
rail:
- Tank trucks: 5,000 to 7,000 gallons (Genereaux
et al. 1984); and
24
-------
- Rail cars: approximately 20,000 gallons of
liquid chemicals, or 30,000 to 33,000 gallons of
liquified compressed gases (e.g., propane, vinyl
chloride, or butadiene) (Genereaux et al. 1984).
The quantity per shipment of all modes of
transportation can be estimated in two steps:
(1) determine the TCC code that best describes the
chemical in question, and (2) locate the median
value for quantity shipped in Table 4 of the CTS
(USDOC 1981a) of the specific mode of
transportation. Use this quantity as the average
shipment size.
Additional information on specifications of containers used
to carry hazardous materials can be found in DOT
regulations, 49 CFR 173 and 178. Part 173 of the regulation
deals with container and packaging requirements for specific
hazard classes of chemicals (USDOT 1986c). Part 178
describes specifications of various types of containers:
metal barrels, drums, and kegs (USDOT 1986d); portable tanks
(USDOT 1985); and containers for motor vehicles (USDOT
1986e).
Container volume data must be converted to kkg before a
value is entered on the worksheet. This is done by
multiplying the volume of the container by the density of
the solid, liquid, or liquified compressed gas and
appropriate conversion factors (e.g., kg/L x 3.785 L/gal x
0.001 kkg/kg). Densities of specific chemicals can be found
in the CRC Handbook of Chemistry and Physics (CRC 1986).
Option 2: FOR RAIL ONLY: If estimating rail releases, one
can use the ICC waybill data as a source of the
average quantity per shipment (i.e., rail car).
These data are organized by STCC code (Step 1).
The average quantity shipped per rail car is given
in tons, which should be converted to kkg by
dividing by 1.1.
Enter the value (in kkg) for average quantity per shipment
for each mode of transportation on the line marked "V" on
the worksheet.
Step 6. Estimate the average distance that the chemical is
transported by each mode of transportation.
25
-------
Option 1: Information in the 1977 Commodity Transportation
Survey Summary (USDOC 1981a) can be used with one
of the methods found in Appendix C to estimate the
average distance a chemical is shipped. The
method of choice depends on the availability of
information on the quantity, origin, and
destination of shipments from manufacturing
facilities as follows:
Method
C-l
C-2
C-3
Option 2:
Average
quantity/
shipment
Unknown
Unknown
Known
Origin of
shipments
Unknown
Known
Unknown
Destination of
shipments
Unknown
Unknown
Unknown
FOR RAIL ONLY: When rail releases are estimated,
the average shipment distance can be obtained from
the ICC waybill data. This can be calculated by
dividing the total car miles by the number of
cars. An example of this calculation is presented
in Section 3.2.2, Part (2), Step 6 of this report.
Enter the values for average distance shipped by each mode
of transportation on the line designated by "M" on the
worksheet.
Step 7. Estimate the annual number of shipments.
Option 1: Calculate the annual number of shipments by each
mode of transportation using data obtained in
Steps 4 and 6. The annual number of shipments
(designated here by "Y") is equal to the quantity
shipped annually (W) divided by the average
quantity per shipment (V):
Y = W/V = (quantity shipped annually)/(quantity
per shipment).
Option 2: FOR RAIL ONLY: If estimates are for rail, the
number of shipments per year can be taken directly
from ICC waybill data. Because the data are based
on a 1 or 6 percent sample, multiply the number of
cars by 100 (1 percent sample) or 16.7 (6 percent
sample) to get the actual number of shipments.
26
-------
Enter the value for annual number of shipments for each mode
of transportation on line "Y" of the worksheet.
Step 8. Select the average accident rate for each mode of
transportation. In this step, statistics for the number of
accidents per mile are factored into the release
calculation. The factor varies by mode of transportation.
The values for truck and rail transportation accident rates
have been entered on line "A" of the sample worksheet
(Figure 1). These values are as follows:
Truck: 1.2 x 10^ accidents/mile (USEPA 1985);
Rail: 6.0 x 10"° accidents/mile (USDOT 1986a); and
Air: 5.0 x 10"9 accidents/mile (USDOT 1987).*
Accident rates for barges and other forms of waterborne
transportation are not available at present. To obtain
information regarding the development of a data base that
can provide this information in the near future, contact
Mr. Theo Moniz, USDOT, U.S. Coast Guard Office of Marine
Safety, Marine Investigation Division, (202) 267-1430.
Step 9. Select the appropriate probability of a release given an
accident for each mode of transportation. The values
available for this parameter vary with the mode of
transportation, and, for trucks, with the type of
container. Probability of a release (P), given an accident,
is as follows:
Truck: For tank trucks, P is 0.29 releases/accident
(USEPA 1985). For trucks transporting
containers, i.e., other than tank trucks, P is
0.26, with the estimate based on data from ICF
(1984).
Rail: The probability of a release, given an accident
involving rail transport, is 0.130 (USDOT 1986i).
Air: Because the aircraft accident rate is based on
accidents that involve fatalities, it is assumed
that every accident is severe enough to damage
containers and release chemicals. Under these
conditions, the probability of a release, given
an accident, is 1.
This accident rate was developed from statistics in USDOT (1987) for
the number of fatal accidents per million miles.
27
-------
Waterborne: Presently, data are not available for
estimating probable release factors for
accidents during water transport in the United
States. However, other data on incidents
involving releases of chemicals to water are
stored in the Marine Pollution Data Base,
maintained by the U.S. Coast Guard Office of
Marine Safety. For more information on this
data base, contact Ms. Mary Robey at (202)
267-0455.
Enter the values for probability of release, given an
accident, for each of the appropriate modes of
transportation on the worksheet on the line designated "P."
Step 10. Calculate the expected annual number of releases for each
mode of transportation (N). This value is obtained
according to the following equation:
N = MxYxAxP
where
M = Average shipment distance (Step 6)
Y = Annual number of shipments (Step 7)
A = Accident rate (accidents/mile, Step 8)
P = Probability of a release, given an accident (Step 9).
Enter the calculated value for the expected number of
releases per year for each mode of transportation on line
"N" of the worksheet.
Step 11. Estimate the fraction of the container contents released for
an accident involving a release. This step incorporates the
results of a statistical analysis of the DOT HAZMAT data
base on the history of chemical releases during
transportation. This analysis is described in detail in
Appendix B of this report. If data are not available to
determine the types of containers in which a chemical is
transported, or if the chemical is carried as part of a
larger shipment, then the fraction of shipment released
should be obtained from Tables B-7, B-8, and B-14 in
Appendix B and used in the calculations. (NOTE: Data in
Tables B-7, B-8, and B-14 are presented as percentages (vs.
fractions) of container contents released. The data from
these tables should be converted before they are used in
these calculations.)
28
-------
The results of the analysis indicate that the fraction of
the contents in a shipping container that is expected to
be released during a transportation accident will vary
depending on the mode of transportation, the physical
state of the chemical, and the DOT hazard class
(identified in Step 1 of this method).
Table 5 lists the DOT hazard classes and the corresponding
physical states and commodity codes used in the
statistical analysis. Depending upon the availability of
information for the specific chemical in question, mean
values for the fraction of container contents released
during an accident can be obtained from Tables 6, 7, and
8, as follows:
Mode of DOT
transportation Physical state hazard class Table
Known Unknown Unknown 6
Known Known Unknown 7
Known Known Known 8
Because the data in the HAZMAT data base were not normally
distributed (see Appendix B), three options are available
when choosing a value for fraction of container contents
released from Tables 6, 7, and 8:
(1) Select the 90th percentile value for "worst-case"
estimates; or
(2) Select the "mean" value for a conservative
estimate (because the data are not normally
distributed, the use of the "mean" value may cause
an overestimation of the quantity of chemicals
released); or
(3) Select the "median" value.
For each mode of transportation, enter the value obtained
from the tables for fraction of container contents
released during an accident on line "R" of the worksheet.
Step 12. Estimate the quantity of chemical released annually for each
mode of transportation (Q). This value is calculated
according to the following equation:
Q = V x N x R
29
-------
Table 5. buninary of DOT Hazard Classes
Physical
state
Liquid
Solid
Gas
Commodity class
(code for DOT hazard class)
2
4
6
8
9
20
25
95
10
30
35
bO
45
50
55
65
DOT hazard class
Other regulated material Class A
Other regulated material Class B
Other regulated material Class C
Other regulated material Class D
Other regulated material Class E
Combustible liquid
Flammable liquid
Corrosive material
Organic peroxide
Flammable solid
Oxidizer
Poison, Class B
Nonflammable compressed gas
Flammable compressed gas
Poison, Class A
Irritating mater ial
(ORMA)
(ORMB)
(ORMC)
(ORMD)
(ORME)
Note, bee Appendix A for definitions of each 001 hazard class
30
-------
Table 6. Values for the Fraction of Container Contents Released
(To Be Used If the Mode of Trdnsportation Is known)
Mode of
transportat ion
Air
Barge
Rail
Truck
Number
of data
records
(N)
589
110
6120
45738
Mean
2867
.2842
1118
3256
Standard
deviat ion
.3706
.3702
.2847
.3925
Upper 90%
confidence
limit
3118
3421
.1177
3286
Median
0700
073b
0001
0935
90th
Percent i le
1 000
1.000
0 510
1.000
Source Statistical analysis of the HAZMAT data base 1986. (See Appendix B for more details )
31
-------
(able / Values for the Fraction of Container Contents Released (lo Be Used
If Physital State and Mode of Transportation Are known)
Physical
btjte
uas
das
Gas
Gas
I iqu id
1 iquui
L iquid
L iquid
sol id
Sol id
Solid
Sol id
Mode of
transportat ion
Air
Barge
Rai 1
Truck
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
Number
of data
records
(N)
9
6
1043
622
534
83
4606
40257
46
21
469
4859
Mean
8197
5519
0518
4907
2761
.2828
1C72
3,739
3059
2135
2899
.3179
Standard
deviat ion
3689
4981
2070
.4333
.3620
3678
.2776
.3913
3997
3217
.4071
.3919
Upper 90%
confidence
1 unit
1 0000
8853
0623
.5192
3018
.3490
.1139
.3272
4026
.3287
o207
3271
Median
1 0000
6273
.0000
4685
.0560
0817
.0002
.0909
1125
.0376
0182
0909
90th
Percent i le
1.0000
1.0000
0176
1.0000
1 0000
1 0000
.5000
1.0000
1 0000
9333
1.0000
1.0000
Source Statistical analyses of the HAZMAF data bnse i986 (See Appendix B for more details/
32
-------
Table 8 Values for the Fraction of Container Contents Released (To Be Us,ed It
Physical State, DOT Hazard Class, and Mode of [ran>portation Aie known)
Phyb u ,-Tl
bt dte
ijTiii
U ,1 h
i.j „)
UJb
1] It.
r..is
das
Gas
Gas
bas
b,is
Gas
L iquid
L iquid
L iquid
Liquid
1 iquid
i iquii)
L quid
i iqu id
iqind
i HI, i,i
i H>nd
, 'quid
i iquid
i q ' r d
Vlld
,juld
• i:-!j
JMI>
11 -t-
2^^.'
"- -^ •
Upper 90%
confidence
limit
1 0000
.9831
0694
6086
1.0000
--
0648
4«24
0880
5472
--
--
5030
5484
5411
5138
7 679
Ot-24
2737
i . OOOu
r,b2!
- ~> i 2
^ j 0 •'
'Jll
040
c 89
1 . j r-
414
Median
1.0000
.6273
.0001
6667
1.0000
1 0000
.0000
.3636
0005
.0744
--
0022
C909
5000
.1069
.3636
8000
.0012
0523
4-.. 7 3
2/tiO
M "JO
i GCOJ
0. 86
•I/'-, I
i.'OO
'.' ,iO
r c 1 5
f i'0;'
0 ,33
90th
Percent i le
1.0000
1 0000
.0099
1.0000
1.0000
1.0000
0196
1.0000
.1000
1 0000
_-
0022
1 0000
1 0000
1 0000
1 0000
1 0000
0708
/4'5
i'.-i4i
1 0000
1 0000
1 0000
02ht
1 0000
i OU.i
1.0000
.-Oi J
64 -r
!?'.• -.4
33
-------
Tnhle H (cont mued)
Physical
btate
1 iquid
L iquui
1 iquio
L iquid
L iqu id
1 iquid
L iquid
L iquid
Sol Hi
•solid
Sol id
solid
Solid
Sol id
sol id
Sol id
sol id
Sol id
Solid
Sol id
Sol ni
Sohd
solid
Commodity
i. Kiss
(code for
DOT hazard
class)3
25
25
25
25
95
95
95
95
10
10
10
30
30
30
30
35
35
35
35
60
60
60
60
Mode ot
t ransportat ion
Air
Bjrge
R-iil
Truck
Air
Barge
Rail
Truck
Air
Rail
1 1 uck
Air
Barye
Rail
T ruck
Air
Bat ye
Rail
Truck
Ai r
Qdt y e
RL, i 1
1 ruck
Number
of data
rtJLOr ds
(N)
377
43
1665
17494
93
32
2511
2021«
1
7
328
4
3
79
353
9
3
207
1238
3c
15
176
2940
MUMP
.1996
2990
1274
27-!"-,
4303
3178
0689
.3747
1000
.2167
.44-17
5428
0405
.1431
2166
5604
1967
3967
32"<
20S5
°S1C
2330
3l08
standard
deviat ion
3114
3653
2979
3629
4130
3995
2574
.4121
36C2
4110
5? 21
0099
33 5B
3521
4517
32'.7
44Go
3958
3331
3523
3650
3889
Upper 90>.
confidence
I imit
.2259
3904
1394
.2788
.5006
4336
0973
.3795
._
.4437
.4819
9791
.1067
.2051
24^3
8273
5041
4469
.3484
3021
400"
2782
32 2c
Median
.0350
.0849
.0002
.0727
.2500
.0955
.0001
.1600
1000
OOOb
.2500
.5833
.0003
.0001
.0227
.7500
.0182
1515
1000
.0165
.0659
0182
.0886
90th
Percent i le
.8000
1 0000
6667
1 0000
1.0000
1 0000
.2727
1.0000
1000
1 0000
1 0000
1.0000
.1212
1.0000
1 0000
1.0000
.5714
1 0000
1 0000
1.0000
1 0000
1 0000
1 0000
dReter to Table 5 for the coiresponding DOT hazard cla-b
Source. Statistical Analysis of the HAZMA1 Data Base, 1986
(see Appendix B for mcie details )
34
-------
where
V = Average quantity of each shipment (Step 5)
N = Expected number of releases per year (Step 10)
R = Expected fraction of container
contents released in an accident (Step 11).
Enter the product of the calculation for each mode of
transportation on the line designated by "Q" on the
worksheet.
Step 13. Estimate the total quantity of chemical released annually by
all modes of transportation. By summing the values for
quantities of chemical released annually by each mode of
transportation (Step 12), one can calculate the total
expected quantity released.
Qtotal = Struck + Qrail + Qwaterborne
Spaces for this calculation are provided in the worksheet
under Step 13. This step completes the general method.
3.2 Sample Calculations
In this section of the report, the general method de:cribed in
Section 3.1 is applied using available information on th< transportation
of three chemicals, di-(2 ethylhexyl) phthalate (DEHP), tthylene oxide,
and formaldehyde. Section 3.2.1 includes calculations ol releases of
DEHP from tank trucks, railroad tank cars, and air transport.
Section 3.2.2 presents an example of the use of ICC waybill data for
calculation of releases of ethylene oxide by rail only. In addition,
releases of formaldehyde from tank trucks and steel drums are calculated
in Section 3.2.3. Each of these calculations is ac:ompanied by a copy of
the worksheet (Figure 1) completed using data speci ric to the chenical
and the mode of transportation considered. The technique of predicting
releases of the formaldehyde from steel drums is explained without an
accompanying worksheet.
3.2.1 Expected Releases of DEHP During Transportation Accidents
The following example demonstrates the use of available information
to calculate the expected quantity of di-(2-ethylhexyl )phthalate (DEHP)
that would be accidentally released from railroad tank cars, trucks, and
air transport over a one-year period. The example is presented in steps
corresponding to the general method discussed in Section 3.1. Figure 3
is a sample worksheet that has been completed using data specific to
transportation of DEHP by railroad tank cars, tank trucks, and air
transport.
35
-------
E;
0)
o •-
"O
O)
cx
TD
CJ1
c
"O •—
C '-
-------
c
0>
TD
10
c/i
Ol
0)
-4->
E
01
1
0)
-M
C
fO
r> TD
cr tu
n
.— 03
' —
O 0)
1— (-
II
un
CO 1 O OO
.— I X L-O
oo
C\J
O)
c
t_
o
_Q
* s-
U •— Q>
3 •- -M U
l_ 05 ro .-
- ct: ^
_Q
r—
>
(O
>»
1
c:
OJ
l_
(_
ri
I- L)
o
+-J -M
U 0
re c
M~
OJ
to u
00 to
0)
i~
O to
•- T)
t/i
C (U
QJ O>
E (-
i- fO
< — \ pp
fO _Q
in
to
ro
0
tJ
N
fO
-C
h—
O
Q
TD
d
n3
QJ
+j
(O
4-)
in
, —
fO
U
CO
>>
-C
CL
c
5
o
c
t/)
c
o
(D
(-
O
Q.
t;
i-
-i->
i-t-
o
QJ
O
E
c
QJ
_C
2
T3
OJ
13
I/)
0}
JD
fO
O
in
3
JD
c
5
o
c
_*r
Q)
fD
OJ
4-J
OJ
-4-1
(/)
U
in
>-,
x:
Q.
T3
C
CO
C
O
*~
(O
t-
o
Q.
in
C.
-M
M-
O
QJ
O
E
d
QJ
_C
s
-o
QJ
cn
Z3
cn
r^.
0>
_D
n3
1 —
C
^L
(O
QJ
fO
in
in
(O
o
TD
t-
fO
NJ
03
.c
1 —
o
o
TU
C
OJ
QJ
-t-»
fO
+J
03
O
tn
>,
_C
O-
c
0
•"
ro
+->
l_
O
Q.
C
O3
t_
-*-J
4-
O
O>
0
E
c
QJ
_c
i
T3
QJ
13
tn
oo
a;
-Q
iO
1 —
37
-------
Step 1. DEHP is not regulated by the DOT. Therefore, there is no
DOT hazard class designation for DEHP. The STCC code for
DEHP is 2899991 (STCC 1972), and its physical state at
standard conditions is liquid. The CAS registry number is
117-81-7 (USEPA 1986).
Step 2. Currently USITC does not list quantities of DEHP produced or
sold, but incorporates these data with all dioctylphthalate
data. In other words, DEHP production and sales data are
not listed separately. In order to estimate the quantities
produced and sold, the ratios of DEHP (produced and sold) to
the total quantity of all dioctylphthalates (produced and
sold) were derived for 1979, 1980, 1981, and 1982. These
ratios were averaged and then multiplied by the dioctyl-
phthalate volumes reported in USITC 1986. This seems a
reasonable approach, as plasticizer production has remained
relatively constant over the past five years. Based on
these estimates, 260,245,000 pounds of DEHP were sold and
therefore assumed to have been shipped in 1985.
Step 3. The quantities of DEHP that are shipped by each mode of
transportation can be estimated using data from the CTS
Summary for 1977 (USDOC 1981a). The STCC code for DEHP,
2899991, corresponds to TCC code 28999 in the 1977 CTS
Summary (see Table 9 of this report). The total quantity of
STCC 28999 commodities shipped in 1977 was 5,253,000 tons.
Of this quantity, 3,503,000 tons (67 percent) were shipped
by truck (including private trucks and both ICC and non-ICC
motor carriers). Another 1,080,000 tons (21 percent) were
transported by rail, and 398,000 tons (8 percent) were
transported by waterborne transportation. An additional
1,000 tons (0.02 percent) were carried by air, and the
remaining quantity was transported by other modes of
transportation.
Step 4. Using the data obtained in Steps 2 and 3 above, one can
calculate that in 1985, 79,256 kkg (118,293 kkg x 0.67) of
DEHP was transported by truck, and 24,842 kkg (118,293 kkg x
0.21) were transported by rail. Additionally, 24 kkg
(118,293 kkg x 0.0002) were carried by air. (This does not
account for 9,463 kkg (118,293 x 0.08) transported by
waterborne vessels and the remaining quantity transported by
other modes of transportation.)
-------
Table 9 Shipping Patterns for RTCC 28999
Chemical products, nee,
Rail
Motor cdrner
Motor carrier, ICC
Motor carrier, non-lCC
Private truck
Air
Water
Pipe 1 me
Parcel delivery
Other and unknown
Value
(mi 11 ion
dollars)
3 , 039
475
1 , 382
1,341
41
652
7
216
58
51
198
Tons
(thousand)
5,253
1,080
1,417
1,363
54
2,086
1
398
225
2
43
Ton-
mi les
(mi 1 lion)
1,804
672
452
433
20
305
1
370
2
2
(2)
Source. USDOC 1981a.
39
-------
Step 5. The capacity of rail tank cars carrying DEHP, a liquid
chemical, is assumed to be 20,000 gallons (Genereaux et al.
1984). Because the density of DEHP is approximately 1 kg/L,
each tank car would hold 75,700 kilograms (75.7 kkg) DEHP
(20,000 gal/car x 3.785 L/gal x 1 kg/L).
An average tank truck capacity of 6,000 gallons (22.7 kkg
DEHP) is assumed (per Genereaux et al. 1984).
The average quantity shipped by air was 0.01 kkg. This was
determined from Table 4 of USDOC (1981a).
Step 6. The average shipping distances of DEHP transported by rail
and tank truck can be estimated using Method C-l from
Appendix C of this report and data from Table 2 of the 1977
CTS Summary, USDOC 1981a (see Table 9 of this report). CTS
Summary data for TCC code 28999 are used to represent DEHP,
as discussed in Step 3 above.
For shipments of DEHP by rail, the average shipping distance
would be 622 miles (672,000,000 ton-miles/1,080,000 tons
shipped).
The value for shipping distance by truck is calculated using
a weighted average of shipping distances calculated for the
two major truck categories listed in the CTS Summary: motor
carriers (ICC and non-ICC combined) and private trucks. The
average shipping distance for motor carriers is 318 miles
(452,000,000 ton-miles/1,417,000 tons shipped). Motor
carriers account for 40 percent of the total tons of TCC
category 28999 shipped by truck. The average shipping
distance for private trucks (60 percent of the total tons of
TCC category 28999 shipped by truck) is 146 miles
(305,000,000 ton-miles/2,086,000 tons shipped).
The weighted average shipping distances for trucks carrying
TCC 28999 commodities would be 215 miles [(318 miles x 0.40)
+ (146 miles x 0.60)].
For air, the average shipping distance of 1,000 miles was
determined by using Method C-3 in Appendix C.
Step 7. The annual number of rail shipments of DEHP would be 328
(24,842 kkg/75.7 kkg/shipment).
40
-------
For tank trucks, the annual number of shipments is 3,491
(79,256 kkg/22.7 kkg/shipment).
For air transport, the annual number of shipments is 2,400
(24 kkg/0.01 kkg/shipment).
Step 8. The accident rate for rail is 6.0 x 10"° accidents/mile
(USDOT 1986a); for trucks, it is 1.2 x 10'° accidents/mile
(USEPA 1985); and for air, it is 5.0 x 10"9 accidents/mile
(USDOT 1987).
Step 9. For rail transport, the probability of a release, given an
accident, is 0.130 release/accident (USDOT 1986a). For taik
trucks, the probability of a release, given an accident, i;
0.29 release/accident (USEPA 1985). For air, the
probability of a release, given an accident, is assumed to
be 1.0.
Step 10. The estimated annual number of rail releases is:
622 miles/shipment x 6.0 x 10~6 accidents/mile
x 0.13 releases/accident x 328 shipments/year
= 0.16 releases/year.
The annual number of predicted releases of DEHP from tank
trucks is:
215 miles/shipment x 1.2 x 10'6 accidents/mile
x 0.29 releases/accident x 3,491 shipments/year
= 0.26 releases/year.
The annual number of predicted releases of DEHP from air
transport is:
1,000 miles/shipment x 5.0 x 10~9 accidents/mile
x 1.0 release/accident x 2,400 shipments/year
= 0.01 releases/year.
Step 11. Because the physical state of DEHP and the applicable modes
of transportation are known but no DOT hazard class applies
to DEHP, the correct source of data on fractions of
container released is Table 7. For rail, the mean value of
fraction of container contents released is 0.107, and for
truck, it is 0.324. For air transport, the mean value of
fraction of container contents released is 0.276.
41
-------
Step 12. The predicted quantity of DEHP released because of rail
accidents is:
0.16 release/year x 0.107 fraction of container released
x 75.7 kkg/container = 1.3 kkg/yr.
The predicted quantity of DEHP released because of tank
truck accidents is:
0.26 release/year x 0.324 fraction of container contents
released x 22.7 kkg/shipment = 1.9 kkg/yr.
The predicted quantity of DEHP released because of air
accidents is:
0.01 release/year X 0.28 fraction of container contents
released x 0.01 kkg/shipment = 2.8 x 10"5 kkg/yr.
Step 13. Summing the calculated values for rail and tank truck
releases, the total amount of DEHP released annually is
3.2 kkg (1.3 kkg + 1.9 kkg). Note that the quantity of DEHP
released because of air accidents is 5 orders of magnitude
less than the quantities released by truck and rail
transport. It is therefore not summed with these two modes
of transportation.
Alternatively, if it is assumed that all DEHP shipped by
truck is shipped in 55-gallon drums, each drum would contain
208 kilograms (55 gal x 3.785 L/gal x 1 kg/L), or 0.21 kkg.
If it is assumed that the steel drums are transported in
20-cubic yard trucks, the total capacity of each truck would
be 4,039 gallons (20 yd3 x 27 ft3/yd3 x 7.48 gal/ft3).
The average quantity per shipment would be 15,288 kilograms
(4,039 gal x 3.785 L/gal x 1 kg/L) or 15.3 kkg. This would
be equivalent to a capacity of 73 55-gallon drums. The
annual number of shipments would be 5,180 (79,256 kkg
shipped by truck/15.3 kkg per shipment).
In Section 3.1, Step 9, it was estimated that, given an
accident, the probability of a release from a steel drum
(container) being transported by truck is 0.26. For the
purposes of this method, it is assumed that all steel drums
on an individual truck shipment would be equally susceptible
to damage during an accident. Also, the accident rate for
trucks is 1.2 x 10~6 accidents per mile (USEPA 1985). If
an average shipping distance of 215 miles (see Step 6) is
assumed, the annual number of releases per year from trucks
carrying steel drums would be:
42
-------
5,180 shipments/year x 215 miles/shipment x 1.2 x 10"
accidents/mile x 0.26 release/accident = 0.3 release per
year.
Since the fraction of container contents released (Table 7)
is 0.324, then the predicted amount of DEHP released from
drums is 0.3 release/year x 0.21 kkg/drum x
73 drums/shipment x 0.324 drum/release = 1.52 kkg.
If we compare the total estimated quantity of DEHP released by rail
tank cars and tank trucks (3.2 kkg =1.3 kkg + 1.9 kkg) with the total
estimated quantity released by rail tank cars and steel drums in trucks
(2.8 kkg =1.3 kkg +1.5 kkg), we can predict a probable range of 2.8 to
3.2 kkg of DEHP, released annually because of combined releases from rail
and truck accidents. These values may underestimate the expected total
releases of DEHP inasmuch as releases during waterborne transportation
and "other" modes of transportation were not included because of a lack
of information.
3.2.2 Expected Releases of Ethylene Oxide During Railroad
Transportation Accidents
This section illustrates the use of ICC waybill data and other
sources of information for calculating expected releases of chemicals by
rail only. Ethylene oxide is used as an example because shipping data
for this chemical are available in the nonconfidential files of the ICC
waybill data base.
(1) Background. Ethylene oxide is a colorless, flammable gas at
ordinary room temperature and pressure. Also called Oxirane and
Anprolene, it is used as a fumigant for foodstuffs and textiles.
Ethylene oxide is used as a sterilant for surgical instruments and an
agricultural fungicide. It is a precursor in ethylene glycol synthesis
and a starting material for the production of acrylonitrile and non-ion c
surfactants. According to USITC (1986), 5,430,359,000 pounds
(24,468,427 kkg) of ethylene oxide were produced in 1985. Sales of
ethylene oxide accounted for only 615,170,000 pounds (279,623 kkg).
Presumably, this is the quantity of ethylene oxide that was shipped. SRI
(1987) lists 12 manufacturers of ethylene oxide. These manufacturers,
along with their locations and annual capacities, are listed in Tible 10.
Because ethylene oxide boils at 10.7°C (Windholz 1983), it is
transported in pressurized containers, i.e., railroad tank cars, tank
trucks, and pressurized cylinders. Estimated releases of ethylena oxide
from railroad tank cars are presented below.
43
-------
fable 10 Locations and Capacities of Ethylene Oxide ManufdCtur ing Plants, January 1, 1987
Plant
Local ion
Annual capacity
(mi 11 ions of pounds)
BASF Corporation
ChemicaIs Divis ion
Industrial & Performance Chemicals Group
Industrial Organics Business
Celanese Corporation
Celanese Chemical Company, Inc
Dow Chemical U S.A
Eastman kodak Company
Eastman Chemical Products, Inc , subsidiary
Texas Eastman Company
ICI American Holdings Inc,
1C I Americas Inc
ICI Specialty Chemicals Group
ICI Specialty Products
National Distillers and Chemical Corpora'inn
Chemicals Division
IIS I Chemicals Company, division
PD Glycol
She 11 Oil Company
Shell Chemical Company, division
SunClin Chemical Company
Texaco Inc
Texaco Chemical Company, subsidiary
Union Carbide Corporation
Industrial Chemicals Division
Total
Geismar, Louisiana
Clear Lake, Texas
Plaquemine, Louisiana
Longview, Texas
Bayport, Texas
Morr is, 1 1 I mo is
Beaumont, Texas
Geismar, Louisiana
C laymont, Delaware
Port Neches, Tevas
495
450
465
200
500
Tdft, Louisiana
800
100
700
b u
1..-.5
Source. SRI 19b7
44
-------
(2) Estimated releases of ethylene oxide from railroad tank cars
where ICC waybill data were used. The following example demonstrates tne
use of ICC waybill and other data to calculate the expected quantity of
ethylene oxide that would be accidentally released from railroad tank
cars over a one-year period. The example is presented in steps
corresponding to those options in the general method (Section 3.1) in
which ICC waybill data are used. Figure 4 is a sample worksheet that has
been completed using data specific to transportation of ethylene oxide by
railroad tank cars.
Step 1. Ethylene oxide is classified as a flammable liquid by the
DOT (Table 11, USDOT 1986b). Its STCC code is 2818239 and
its physical state at standard conditions is gas (b.p.
10.7°C). The CAS registry number is 75-21-8 (USEPA
1986).
Step 2. Based on the 1985 ICC 1 percent waybill data presented in
Table 12, there were an estimated 6,880 railroad tank car
shipments of ethylene oxide during 1985 and the average car
contained 71.3 kkg (78.4 tons (1.1 ton/metric tons))
ethylene oxide. Therefore, the estimated annual quantity
shipped by rail is 71.3 kkg/car x 6,880 cars = 490,509 kkg.
Note that this quantity is almost twice the amount reported
as sold in USITC (1986). This may indicate that
(1) multiple counting occurs if a volume of ethylene oxide
is hauled by more than one rail carrie~ from its point of
origin to its final destination, and (2) companies ship
ethylene oxide by rail to facilities under the same
ownership for further processing.
Steps 3 Because the quantity shipped annually by rail is available
and 4. directly from ICC waybill data, Step 3 of the method can be
omitted, and the value from Step 2 for quantity shipped by
rail (490,509 kkg) is entered on the worksheet which is
Step 4.
Step 5. Based on ICC data from Table 12, the average quantity of
ethylene oxide shipped per rail car is 71.3 kkg
(79.0 tons/1.1 ton/kkg).
Step 6. Using ICC data from Table 12, the average shipment distance
by rail is 946 miles (6,508,700 car miles/6,880 tank cars).
Step 7. The annual number of rail shipments reported in ICC waybill
data (Table 12) is 6,880.
Step 8. The 1985 average accident rate for rail transportation is
6.0 x 10"5 accidents per mile (USDOT 1986a).
45
-------
0)
fO i—
l~ -Q
CO
S- t—
o
M- (O
(O OO
_
cn 0)
r- -
tu
LJ r- r-
CO Q) If}
o o •—
QJ
Q-
Q,
-------
U O •—
o
.a
o •- •-
QJ 0> OJ
TD
O
E
_c
I
O) OJ O) OJ
OJ OJ CU OJ
47
-------
Table 11. Sample DOT Packaging Requirements Including Ethylene Oxide
§172.101 Hazardous Materials Table—Contd.
in
w
A/
W
-
(2)
ikippml umet
Ethyl hontt
Ethyl butyl mile
•thyl butyl ether
Elhyl burynldchyde
ithyl bulyrtte
Ethyl chloride
Ethyl chloroiccuie
Elhyl chtarofbrmile /cUonautamtul
Ethyl chlorothiofanntle
Elhyl crotontie
Ethyl dxhlarailine
Ethykne or Ethykne. compreued
Elhykne cblorohydnr'
Ethykne. rtfnienled liquid (cryogenic /«nuoV
Elhyknedumtne
Etiivleit* aiamtme aipetctiionit
Ethykne dibronude
Ethylene dichlonde
Eihvkne Uycol diethyl ether idtnkyl
•Ce/too/nr;
£VAv/mr fiycol eimtrare
rOttuwWy
Ethykne jiycol monoelhvl ether tceute
CCelloaol* acetate I
Elhykne (Jycol monomethvl ether ImeiM
'Celloulve 1
Elhvlene ilycol monomethyl ether Mettle
Imetktf Cellcaot* acetate 1
Ethykne uninc. inhibited
Elhykne otide
Ethyl ether
Elhvl ronntte
Elhylheuldehyoe
Elfivl kyanfjeraniae Itiptodet aoore 100 drr Cl
Ethvl Itcute
Ethyl merctpttn
Ethvl methyl ether
Elhyl methyl kelonc
Elhyl minue 'mine eikert
Ethvl mime (nitroui ether)
Ettivt pcrtltlorate
(3)
Htnrd
cuat
FUmmtbk
liquid
^ombujttble
Mind
Fltmmibk
iquid
Rtmmibk
liquid
FltmtMOk
liquid
FUmmtok
liquid
r«nbunible
liquid
Fltnunibk
liquid
Cormnve
mtuml
Flinaublc
tqoM.
FUoMubt*
tquid
Fluuubk
f*
Poem B
Fltmmtble |tf
Corrative
mtientl
Pooon B
FUmratok
liquid
FUmmtbk
liquid
Forbidden
Combuinbk
Uquid
GxMxmiok
liquid
Cambuiubk
liquid
Combuttibk
liquid
FUmmtbk
liquid
FUmmtbk
liquid
Fltmmtok
liquid
Fltmmtbk
liquid
Combuuibk
liquid
FoRndden
Combuilibk
liquid
iFlunmiMe
1 liquid
FUmmtbk
liquid
FUmmtble
liquid
FUmmtbk
liquid
FUmnitbk
liquid
Forbidden
I3A)
lanti-
ncmtnn
number
UNI 176
UN1IT7
UNI 179
UNI 171
UNIIIO
UN 1037
UNI 111
UNI 112
UN2I26
UN 1162
UNI 113
UNI962
UNII1S
UNI01I
UNI604
UNI60S
UNIIM
UNII51
UNII7I
UNII72
UNI III
UNIII9
UNIII5
UNIOM
LN1DS
LN1I90
UNII9I
|L'SII92
1
| LN2363
1
UM019
LNII9!
NAI993
UNIIM
(4)
UoeKil
required
Wnot
euxpted)
FUmniabtt
(quid
None
Uquid
Fltmmtok
liquid
FUmmtok
bqmd
lammibk
bquid
None
FUmnuok bmitd
utt Pom
Corrotivc
FUmmtbk
hautd
FUmmtbk
liquid
liamuUc
Pooon
ItmmtHt |M
lofrative
Poem
nimmtbk
liquid
Fltmmibk
None
None
Note
None
FUmmtbk liquid
ind Potion
Fltmnubk
iiquid
FUmmtbk
liquid
FUmmtbk
liquid
None
None
FUmmtok
liquid
Fltmmtok
liquid
Fltmmtok
liquid
FUmmtok
l«nud
FUmmtbk
liquid
(3)
Pick.
(I)
EicnttoM
73 III
71 lib
73 III
7} III
73.111
None
73 lib
None
73244
73.111
None
7) We,
7)345
None
73244
73)45
173111
71111
171 lib
173 lib
17] lib
17) lib
None
None
None
17! Ill
17) lib
17) lib
None
None
171 111
173 III
None
fill
a»
Spnafe
rtqmre-
73119
None
73119
71.119
71119
71123
None
71.211
7324]
732491
73119
73133
71304
71346
7! It
71111
71)19
171243
171)46
17)119
171119
None
None
None
None
17] IJ«
17) 124
17)119
17! 119
None
None
173141
17) 119
171119
171119
17)119
W
matt ptckife
(!)
Puenpr
ctrrymi
tmnA or
mlcnr
train
No lira
qun
qun
qun
Forbidden
No lira
Forbidden
qun
qun
Fortndden
Fortndden
qun
Forbidden
qun
qun
1 qun
1 qun
No limit
No limit
No limn
No Itmil
Fort-dde.
Forbidden
Forbidden
1 qun
No limit
No limn
Forbiddm
Forbidden
1 qun
Forbtddea
Forbidden
a
u
u
u
u
u
a
u
a
u
a .
u
a
i
u
u
u
u
u
u
tj
1.2
3 pan 1,2
See 173 124 ' ; 2
I
10 plloni
10 puom
.J
!J
No limn ! ,2
1
No Umu .:
10 pUom ' 2
10 plloni 1 U
10 pitorn
Forhtddcn
Forbidden
!_2
U
U
fkt
P»
xod
1
U
'
1
U
'
u
'
1
1
1
4
1
i
u
u
1
u
u
u
a
1.2
i
i
5
4
1.2
1 2
1
1
1
1
!
(c)
Kt<*dry
3T •llllllllll pM
Siov **iv from •*-
inf juflrlen.
*fr?mi°0 ***"*,.
4OW nt»WtV fttMl •*•
ing ainmnerv
riov i.wiv from •*-
int ,Hi.mrten
Serpea.»iiOB SUM u 1
far 'jonnuoie tmet \
Seijre^Daa u.ar •
Source: USDOT 1986b.
48
-------
Table 12 Shipments of Fthylene Oxide by Railroad
lank Cars Estimated from ICC Data
Number of tank cars 6,8bO
Total tons lading 539,560
Total car-miles 6,506,560
Average tons/car 78 4
Average haul car-miles 946
Source ICC Waybill Sample (ICC 1985).
49
-------
Step 9. The probability of a release given an accident is 0.13
(USDOT 1986a).
Step 10. The annual number of expected releases of ethylene oxide
during rail transportation is:
946 miles/shipment x 6.0 x 10"^ accidents/mile
x 0.13 release/accident x 6,880 shipments/year =
5.0 releases/year.
Step 11. Because the physical state (liquid), mode of transportation
(rail), and DOT hazard class (flammable liquid commodity
class number 25 on Table 5) are known, then the fraction of
the container contents released during an accident can be
found in Table 8. This value is 0.127.
Step 12. The estimated quantity of ethylene oxide released annually
because of rail accidents is:
5.0 releases/year x 71.3 kkg/container
x 0.127 container/release = 45.3 kkg/yr.
3.2.3 Expected Releases of Formaldehyde During Transportation Accidents
(1) Background. Commercial formaldehyde is produced and shipped as
an aqueous solution containing 37 percent formaldehyde and up to
10 percent methanol. Formaldehyde in aqueous solutions rapidly hydrates
to form methylene glycol and a series of low molecular weight polymeric
polyoxymethylene glycols. The methanol is added to ptevent the
formaldehyde from polymerizing. The concentration of formaldehyde as the
aldehyde in aqueous solutions has been found to be well under 0.1 percent
(Walker 1975).
The most recent estimate available of annual U.S. production of
37 percent formaldehyde solution is 5,606,140,000 pounds (2,548,245 kkg)
(USITC 1986). This amount, which represents 1985 production, is
equivalent to 66 percent of the January 1, 1986, production capacity of
8,584,000,000 pounds (3,901,818 kkg) reported by SRI (SRI 1986). At the
beginning of 1986, production capacity for formaldehyde was distributed
among 15 manufacturers and 47 facility locations, as summarized in
Table 13 (SRI 1986). The production capacity represented in Table 13 was
geographically concentrated in the southeastern and southwestern states.
It is not known, however, how actual production and sales were
distributed among these facilities.
The quantity of 37 percent formaldehyde solution sold and presumably
shipped in 1985 was 1,742,409,000 pounds (792,004 kkg) (USITC 1986).
This quantity sold represents 31 percent of reported production (per
50
-------
Table 13. Locations and Capacities of Formaldehyde Manufacturing Plants, January 1, 1986
Plant name
Location
Annual capacity
(thousand
metric tons)
Borden Inc.
Borden Chemical Division
Adhesives and Chemicals Division
Petrochemicals Division
Demopolis, Alabama
Diboll, Texas
Fayetteville, North Carolina
Fremont, California
Kent, Washington
La Grande, Oregon
Louisville, Kentucky
Missoula, Montana
Sheboygan, Wisconsin
Springfield, Oregon
Geismar, Louisiana
43
36
107
102
36
30
36
41
59
109
114
BTL of Arkansas, Inc.
Malvern, Arkansas
50
Celanese Corporation
Celanese Chemical Company
Celanese Specialty Operation
Celanese Engineering Resins Division
Newark, New Jersey
Bishop, Texas
53
818
Chembond Inc
Andalusia, Alabama
Moncure, North Carolina
Springfield, Oregon
Winnfield, Louisiana
32
55
64
32
E I. DuPont de Nemours & Company, Inc.
Chemicals and Pigments Department
Belle, West Virginia 227
Grasseli, New Jersey 73
Healing Springs, North Carolina 100
La Porte, Texas 145
Toledo, Ohio 123
GAF Corporation
Chemical Products
Calvert City, Kentucky
Texas City, Texas
45
45
51
-------
Table 13 (continued)
Plant name
Location
Annual capacity
(thousand
metric tons)
Georgia-Pacific Corporation
Chemical Division
Hercules Incorporated
Operations Division
International Minerals & cheiT.ic.al Corporation
IMC Chemical Group
Industrial Chemicals Division
Monsanto Company
Monsanto Chemical Company
Nuodex INC
Perkins Industries, Inc
Reichhold Chemicals, Ini.
Rogue Valley Polymers, Inc
Wi ight Chemical Corporation
TOTAL
Albany, Oregon
Columbus, Ohio
Conway, North Carolina
Crossett, Arkansas
Lufkin, Texas
RusselIvi1le, South Carolina
Taylorsv11le, Mississippi
Vienna, Georgia
Louisiana, Missouri
beiple, Pennsylvania
Addyston, Ohio
Chocolate Bayou, Te\as
Springfield, Massachusetts
Fords, New Jersey
Vicksburg, Mississippi
Hampton, South Carolina
Houston, Texas
Kansas City, Kansas
T'jbca loosa, Alabama
White City, Oregon
Acme , North Care. 1 ina
55
75
48
75
48
99
55
48
79
61
52
82
134
64
25
23
105
23
33
91
36
3,902
Source: SRI 1986
52
-------
USITC 1986) and approximately 20 percent of production capacity (per SRI
1986).
(2) Estimating releases of formaldehyde from tanks trucks. In
addition to being shipped by rail, formaldehyde solution is also
transported by tank truck. The following example describes the
application of the general method (Section 3.1) to predicting the annual
release of formaldehyde resulting from tank truck accidents. Figure 5 i;
a worksheet that has been filled out using data on formaldehyde transport
by tank truck. Following Step 13, an alternative calculation of releases
from trucks carrying steel drums is presented.
Step 1. The DOT hazard class is combustible liquid, the STCC number
is 2818144 (STCC 1972), and the physical state is liquid.
The CAS registry number is 50-00-0 (USEPA 1986).
Step 2. The USITC (1986) reports that 1,742,409,000 pounds of
formaldehyde were sold in 1985. Assuming the amount sold
was the amount shipped, and converting to metric tons, the
estimated quantity of formaldehyde shipped in 1985 was
792,004 kkg.
Step 3. The STCC code for formaldehyde, 2818144, corresponds most
closely to TCC code 2818, Miscellaneous Organic Chemicals,
in the CTS Summary for 1977 (USDOC 1981a). Because the
quantity shipped is known, the CTS Summary data (USDOC
1981a) can be used to estimate the quantity shipped by
truck. According to Table 2 of the CTS Summary for 1977,
10,273,000 tons of commodity code TCC 2818 were transported
by truck (the quantity carried by motor carriers, plus the
quantity carried by private truck). That amount was
equivalent to 32 percent of the total quantity
(32,324,000 tons) of this TCC category that was transported
in 1977.
Step 4. It is assumed that 32 percent of the total formaldehyde
solution shipped was shipped by truck; this is equivalent to
253,441 kkg (0.32 x 792,004 kkg) of formaldehyde solution
transported by truck.
Step 5. The density of a 37 percent solution of formaldehyde is
1.083 kilograms/liter (Aldrich 1983). A 6,000-gallon truck
(average capacity, per Genereaux et al. 1984) would contain
24,595 kilograms formaldehyde solution (6,000 gal x
3.785 L/gal x 1.083 kg/L). This is equivalent to 24.6 kkg
per truck shipment.
53
-------
U *-i
CJ O •—
- 03
*— OJ Z5
ct to c.
(./I 05 C
CO fO
O >>
tJ ^~_
o ^r
•^t r-- 01
GO • r—
cr>
1.- rtl
•— tJ CX
n co
en
QJ —i
0 O
-— Q
CO 00
c •-
0) -Q
Q- >>
OJ 0)
g. *•
|
T3
- O
4J S
54
-------
0)
XI
•o
O O" Cr cy
0) O
ro -D •- '—
OJ QJ
"D "O
O O
E E
OJ
•— co tn
tz •*->•-
o ro
— T3 CO
Qj OJ QJ
-------
Step 6. The average shipping distance of formaldehyde transported by
truck can be estimated using Method C-l from Appendix C of
this report and data from Table 2 of the 1977 CTS Summary
(USDOC 1981a). CTS Summary data for TCC code 2818
(Miscellaneous Organic Chemicals) are used to represent
shipping patterns of formaldehyde, as discussed in Step 3
above.
The value for shipping distance by truck is calculated using
a weighted average of shipping distances calculated for the
two major truck categories listed in the CTS Summary: motor
carriers (ICC and non-ICC) and private truck. The average
shipping for motor carriers is 320 miles
(2,338,000,000 ton-miles/7,302,000/tons shipped). Motor
carriers account for 71 percent of the total tons of TCC
category 2818 transported by truck. The average shipping
distance for private truck (29 percent of the total tons of
TCC category 2818 shipped by truck) is 283 miles
(841,000,000 ton-miles/2,971,000 tons shipped). The
weighted average shipping distance for trucks carrying TCC
2818 commodities would be 309 miles ((320 x 0.71) + (283 x
0.29)).
Step 7. If each tank truck contains 24.6 kkc, of solution, then
10,302 shipments would be needed to transport 253,441 kkg of
formaldehyde solution each year.
Step 8. The average accident rate for trucks is 1.2 x 10"^
accidents per mile (USEPA 1985).
Step 9. The probability of a release, given an accident, for a tank
truck is 0.29 (USEPA 1985).
Step 10. The expected number of releases each year would be:
309 miles/shipment x 1.2 x 10"^ accidents/mile
x 0.29 release/accident x 10,302 shipments/year
=1.1 releases/year.
Step 11. Since the DOT hazard class (commodity class number 20 on
Table 5), mode of transportation, and physical state are
known, the percent of container released is found in
Table 8. This value is 0.226.
Step 12. The predicted quantity of formaldehyde solution released
annually because of tank truck accidents is:
i.l releases/year x 0.226 container/released
x 24.6 kkg/container =6.1 kkg.
56
-------
(3) Estimating releases of formaldehyde solution from truck
transport in steel drums. Not all formaldehyde truck shipments are made
in tank trucks; glass carboys and stainless steel drums are also used as
containers. If it is assumed that the formaldehyde solution is
transported by steel drums and that tank trucks are not used, then the
estimate of releases is calculated as follows: According to Table 11
(USDOT 1986b), formaldehyde solution transported in containers of
110-gallon capacity or less is regulated under DOT hazard class, Other
Regulated Material-A (ORM-A). If 110-gallon drums of 37 percent
formaldehyde are transported in 20 cubic yard trucks, the total capacity
of each truck would be 4,039 gallons (20 yd3 x 27 ft3/yd3 x
7.48 gal/ft3). This would be equivalent to a capacity of 37 110-gallon
drums. The average quantity per shipment would be 16,556 kilograms, or
16.6 kkg (4,039 gal x 3.785 L/gal x 1.083 kg/L). The annual number of
shipments this size would be 15,268 (253,441 kkg shipped by
truck/16.6 kkg per shipment).
The probability of a release, given an accident, for trucks
transporting steel drums is 0.26 (ICF 1984). For the purposes of this
calculation, it is assumed that, given a nontrivial accident, all of the
drums carried by a truck during an individual shipment would be equally
subject to damage and potential release. If an average shipping distance
of 309 miles and a truck accident rate of 1.2 x 10"° accidents/mile
(USEPA 1985) are assumed, the expected annual number of releases is:
309 miles/shipment x 1.2 x 10"6 accidents/mile
x 15,946 shipments/year x 0.26 release/accident
= 1.5 releases/year.
Because formaldehyde solution is classified under ORM-A (commodity
Class 2 from Table 5), and is transported as a liquid by truck, Table 8
indicates that, given a release, 47.3 percent of the container contents
will be lost during an accident involving a release. If there are
1.5 releases per year, then the average quantity of formaldehyde released
per year is:
1.5 releases/year x 16.6 kkg/shipment
x 0.473 fraction released
= 11.8 kkg.
(4) Summary. These results and those obtained for tank trucks
indicate that the expected annual quantity of formaldehyde solution
released as a result of truck accidents would range from 6.1 kkg (tank
trucks) to 11.8 kkg (steel drums). This estimate does not include
waterborne or other modes of transportation.
57
-------
4. REFERENCES
Aldrich. 1983. Aldrich catalog/handbook of fine :hemicals. Milwaukee,
WI: Aldrich Chemical Company.
CRC. 1985. Handbook of data on organic compounds, Vols. I and II.
Weast RC, Astle MJ, eds. Boca Raton, FL: CRC Press.
CRC. 1986. CRC handbook of chemistry and physics, 66th ed. Weast RC,
Astle MJ, Beyer WH, eds. Boca Raton, FL: CRC Press Inc.
French D, Richards HA. 1973. Hazardous materials flow in intercoastal
waterways — a case study of risk-exposure factors for the future of a
specific area. In: National Academy of Sciences. 1973. Bulk
transportation of hazardous materials by water in the future-proceedings
[microfiche]. Washington, DC: Committee on Hazardous Materials,
National Academy of Sciences. NTIS AD-776 618/1.
Genereaux RP, Mitchell CJB, Hempstead CA, Curran SP. 1984.
Transportation and storage of fluids. In: Perry RH, Green D, eds.
Perry's chemical engineers handbook, 6th ed. New York, NY: McGraw-Hill
Book Co.
ICC. 1985. Interstate Commerce Commission. Computer printout:
Railroad tank car shipments of formaldehyde (STCC Code 2818144) during
1979 from 1 percent waybill sample. Retrieved June 8, 1981. Washington,
DC: Interstate Commerce Commission.
ICF. 1984. Assessing the releases and costs associated with truck
transport of hazardous wastes. Washington, DC: U.S. Environmental
Protection Agency. PB84-224468.
OTA. 1986. Office of Technology Assessment. Transportation of
hazardous materials. Washington, DC: Office of Technology Assessment.
OTA-SET-304.
Perry DH, Green D, eds. 1984. Perry's chemical engineer's handbook, 6th
ed. New York, NY: McGraw-Hill Book Co.
SRI. 1978. Stanford Research Institute. Chemical economics handbook--
formaldehyde. Menlo Park, CA: Stanford Research Institute.
SRI. 1986. Stanford Research Institute. 1986 Directory of chemical
producers, United States of America. Menlo Park, CA: Stanford Research
Institute.
SRI. 1987. Stanford Research Institute. 1987 Directory of chemical
producers, United States of America. Menlo Park, CA: Stanford Research
Institute.
59
-------
STCC. 1972. Standard Transportation Commodity Code Tariff No. l-*\.
USDOC. 1981a. United States Department of Commerce. 1977 Census of
transportation, commodity transportation survey summary. Washington,
DC: Bureau of Census.
USDOC. 1981b.
transportation,
Washington, DC:
United States Department of Commerce. 1977 Census of
commodity transportation survey, geographic area series.
Bureau of Census [microfiche].
USDOC. 1981c. United States Department of Commerce. 1977 Census of
transportation, commodity transportation survey, commodity series.
Washington, DC: Bureau of the Census [microfiche].
USDOC. 1985. U.S. Department of Commerce. 1982 Census of
transportation, truck inventory and use survey, United States.
Washington, DC: U.S. Government Printing Office.
USDOT. 1985. U.S. Department of Transportation. Regulations for
shipping containers; specifications for portable tanks. 40 CFR 178 and
amendments through March 19, 1985. 50 FR 11048.
USDOT. 1986a. U.S. Department of Transportation. Accident/Incident
Bulletin No. 154. Washington, DC: Federal Railroad Administration
Office of Safety.
USDOT. 1986b. U.S. Department of Transportation.
amendments through June 25, 1986. 51 FR 23075.
49 CFR 172 and
USDOT. 1986c. U.S. Department of Transportation. Regulations for
shippers--general requirements for shipments and packagings. 49 CFR 173
and amendments through February 18, 1986. 51 FR 5968.
USDOT. 1986d. U.S. Department of Transportation. Regulations for
shipping container specifications; specifications for metal barrels,
drums, kegs, cases, trunks, and boxes. 40 CFR 178 and amendments through
February 18, 1986. 51 FR 5968.
USDOT. 1986e. U.S. Department of Transportation. Regulations for
shipping container specifications; specifications for containers for
motor vehicle transportation. 49 CFR 178 and amendments through February
18, 1986. 51 FR 5968.
USDOT. 1987.
transportation
fatalities and
Board.
U.S. Department of Transportation. National
statistics. Update 6/12/87 for air carrier accidents,
rates. Washington, DC: National Transportation Safety
60
-------
USEPA. 1985. U.S. Environmental Protection Agency. Office of Policy,
Planning and Evaluation. Assessment of incineration as a treatment
method for liquid organic hazardous waste. Background report IV:
Comparison of risks from land-based and ocean-based incineration (EPA
230-02-86-002 through 008) Washington, DC: U.S. Environmental Protection
Agency.
USEPA. 1986. U.S. Environmental Protection Agency. Office of Toxic
Substances. Toxic Substances Control Act chemical substance inventory
(1985 ed.--EPA 560/7-85-002). Washington, DC: U.S. Environmental
Protection Agency.
USITC. 1986. U.S. International Trade Commission. Synthetic organic
chemicals--United States production and sales, 1985. USITC publication
1745. Washington, DC: U.S. Government Printing Office.
Walker JF. 1975. Formaldehyde. 3rd ed. New York: Robert E. Krieger
Publishing Co.
Windholz M. ed. 1983. The Merck index. An encyclopedia of chemicals
and drugs. 10th ed. Rahway, NJ: Merck and Co., Inc.
Versar. 1987. Versar Inc. Computer printout: univariate statistics
for HAZMAT data base failure codes 1, 18, 19, 22, and 28. Retrieved
September 18, 1987. Springfield, VA: Versar Inc.
61
-------
UJ
Q-
E
TO
X
l_U
t_
O
•M
E
4~
QJ
a
t/>
1.1
TO
U
ID
'D
M
TO
x:
OJ
L:
0
4-J
QJ
L)
TO
QJ
L-
O
to
TO
Cn
O
_c
0
u
TO
>!
x:
*~*
LJJ
v/>
^-J
to
OJ
•M
V,
.a
-D
QJ
C
E
t_
QJ
-t-J
Ql
-D
(A
TO
O
tJJ
3
O
Q)
_Q
C~
0
Q.
-C
to
ro
if-
fO
CD
C
>
TO
£1
~D
CT
>i
tr
,
-C
-t->
OJ
E
ID
Q)
t"
QJ
N
r
QJ
-Q
cr.
Li_
LJ
Cn
«3-
C
T3
QJ
-t-J
CO
QJ
U
TO
to
c
o
f-'
CL
a>
u
X
U_J
~o
ul
ro
r--.
cr:
u_
CJ
CT,
T
C TO
ui
"O •— 1
QJ .— 1
to ro
O
Q)
-3
>•*•
Q)
C
O
-*-J
CD
^
E
TO
•^1
-d
a)
E
j*
c
Li-
O
O
CXJ
3
0
ID
_a
T)
c
TO
O
O
CD
•>
O
XI
fD
k__
O
4^,
re
-t-J
cr
0
a
_c
to
TO
4-
TO
a>
c:
>
TO
T3
rs
cr
r
t£
"O
3
CT
QJ
XI
4-J
D
F
O
QJ
(~
TO
to
c
o
-t-J
O.
OJ
u
X
I-U
. — .
"D
LT>
i — i
.-o
a;
u
«~J
Cn
*£f
C
~o
QJ
_(_» ^--,
to _Q
r- U1
to ,-i
-«-»
to -o
QJ f-
-t_> ^-i
>> ct
.a tj_
LJ
-c
OJ cn
C *T
E c
I- •-
OJ
-*-> 13
a> QJ
TJ -M
t-T
fC ' —
to
u
(-.
o
-C
Q,
to
O
jr
CL
.
ri
M-
GJ
10
0
u
QJ
U
0
i-
4-)
*
CO
QJ
4-
OJ
CO
CJ
O
•M
QJ
OJ
O
a_
X
QJ
c;
TO
C
^~
-M
QJ
O
U
QJ
E
O
c:
T3
C^
CT
OJ
_c
TO
TO
Lu
fD
0
CJ
I-
tJ
-c:
o
c:
to
CO
QJ
CJ
O
CL
L-
O
01
c
u
TO
r
e
E
O
-*->
TO
CD
"D
CD
cr
QJ
n
o
H-
rn
"3
o
r
tr
O
4-J
03
O
CL
to
C
TO
to
O
QJ
t/)
Dl
C
»—
-*->
QJ
O
fD
QJ
u
CD
OJ
n
E-
fD
U
6
t
o
CD
. — i
ro
o
cr.
r—
C
QJ
to
to
QJ
CL
>-,
lO
o
o
CT
t
13
OJ
13
n
QJ
-D
TO
M
ai
-C
c
a>
Dl
O
l_
-a
-C
QJ
-M
fD
L_
£1
E:
3
(/:
ft)
+->
O
Q.
fD
O
QJ
"O
X
o
QJ
U
fO
cn
o
c
QJ
C
CJi
fD
E
OJ
CL
QJ
-M
6
x:
to
ai
-IL
CJ
L/l
QJ
r.
TO
t/1
1_
OJ
M
~O
X
CD
ID
U
TO
U
fc"
0
k-
x:
CJ
L:
0
4~J
Z3
O
10
QJ
T)
X
O
U
QJ
Q_
4_
O
C
o
to
.u
g
CJ
.c:
a)
TO
ZJ
E
i/i
vJ
>,
"O — -
QJ kO
I— T~I
tl O")
QJ P^~
Cn .—i
x cr:
O Li^
O
O) -.-
QJ
TO -M
OJ 0
TO C
U TO
M CTt
i- O
Q)
-o
X
0
t
QJ
0.
i —
>i
O
Nl
cr
QJ
_Q
OJ
T)
X
o
1-
QJ
Q.
TO
Q)
\-
ZD
fO
4_j
"O
c
TO
CD
I-
Z3
CJ
-M
CO
1
O
1
c
TO
JD
QJ
JT
a>
c
c
TO
C
O
u
Z)
o
CL
E
o
u
re
cr,
O
C
C
O
o.
QJ
L)
X
Li-J
CJ
t
CJ
x;
LJ
o
L
TJ
>-,
-C
QJ
E
TO
T3
O
00
QJ
c:
E
O
t-
CD
OJ
XI
to
QJ
CD
l_
O
C
o
u
ZJ
t-T
Ql
T3
OJ
to
to
OJ
CT
13
U
-t->
~D
0
CO
O
TD
_1
C
a>
t ~/
0
i~
u
0
l_J
c
o
13
0
CT
QJ
"O
X
O
(-
TU
>^
x:
b
Z3
~a
0
tO
-a
LJ
TO
to
"D
13
D
I
+_j
-M
C
O
CJ
0
QJ
-M
to
QJ
X
TO
QJ
13
to
lO
r—
C
I/1
c
TO
E
d
to
c:
O
TO
QJ
TO
O
C\J
ro
r-.
, — (
U.
CT-
QJ
TJ
CJ
tr
QJ
TO
QJ
QJ-
-M
to
QJ
O
O
CJ
QJ
QJ
QJ
to
TO
-T
4-j
TJ
"J
4-
LJ>
tz
•4-J
U-
TO
+-J
C/J
CD
C
cn
c
OJ
QJ
C
0)
T3
TO
+-J
Z3
CO
C
TO
4->
(_
QJ
CJ
CO
+-J
4->
rO
x:
•M
.
TO
CD
0
ro
rO
cr:
u.
o
CD
•*r
c
"O
QJ
c:
QJ
-Q
to
TO
LO
ro
cn
T3
QJ
t/)
CT
QJ
t-
CL
E
0
u
<£
LO
TO
cn
QJ
XI
TO
TO
LJ_
fO
E
QJ
O
i_
4-J
QJ
CL
-a
Q)
4-
QJ
ZJ
cr
C
OJ
cn
O
t_
"D
>>
_C
, — ,
X3
CJ
CD
CO
rO
r-
cc:
u_
u
cn
•T
LO
-M
C
QJ
E
QJ
J
cr
QJ
1_
+-J
XI
TO
TO
4-
C/l
n
o
u
~C)
>^
x:
c
TO
C
0
0)
c
c:
o
c
QJ
X
QJ
C
t-
o
-C
U)
to
TO
Dl
UJ
-Q
TO
E~
ro
4-
TO
c:
TO
x:
-M
L.
QJ
x:
-M
O
to
TO
cn
~O
QJ
to
to
QJ
i~
CL
E
O
L)
«C
to
rO
D
OJ
XI
TO
TO
4-
C
O
~z.
64
-------
TJ
Q>
3
L;
_^j
o
u
2
4-J
u
(D
4->
C
0
u
c
o
c:
o
-C
5
o
c:
rQ
4J
10
-Q
3
to
TD
o
CO
u
O
TD
CT
t_
L-
r—
>,
QJ
CO
c: —
QJ -— t
4-> OO
c m
ro
i- r^
O -H
10 rf
3 U_
0 0
t-
QJ cn
cn
CO
en ' —
'O
-- QJ
fD 4-*
1C
o e
•*j
TD "*
QJ C
o in
Q- •-
X O
CD CL
01
C
O
4-J
OJ
L)
03
O
e oj
o c
t- 0)
J2 CT
t/>
- O
TJ -C
r- Q.
LJ
•D
QJ
L) T>
C X
fO O
>1
u o
o •-
k- l_
TD 4->
>i '-
-a c
TD
01
X
in —
in E
o *n
TD
CO •-
13
-C CT
0
4-' OJ
LO
cn QJ
C" _C
0 +J — -
QJ 4- CXI
-Q O rO
1/5 in ro
TD i- r-
O .— i
3 CL
CT u_
r- o
u o cn
o -3-
m - — -
i/) OJ
QJ iO Q)
in fO 4-
fO CTl r-
cn i—
QJ
c to o
O QJ 4-J
-- 4-j in
D :D
a. 4- o
O i-
to at
3 CO CD
O 4-1 C"
QJ 13 TD
Di O
C E QJ
X3 fD
ai E -~
GJ
I- >i JT
*-- t- -M
X 0) •-
14J ^> 5
4-J
(T3 l/>
U Q)
fD CL
Q.
~O
- QJ
QJ to
C rO
>~ £}
r— |
•- >1
C t-
fD ^3
O O CO
i- t- +->
+-> QJ C
- E "3
C -M
U
- QJ 0)
o •- c
C C •-
QJ (D tO
£:>!•-
Q_ O T3
- QJ r~
• — - C~l O
CO «-
TD O X
•- -*-> o
i — +j
O C
co 3 C
r- O fD
E C E
QJ _* 3
QJ
TD t- C
C fO O
in CD 4->
QJ _C fD
4J +-> TD
tn
fD QJ
Q- in 4-J
C r- CJ
TD QJ TD
T3 4-* fO
U £ 4-
C O
— C QJ
-- u
TD (O QJ
•- 4-1 in
r— •- XI
o u to
- QJ
O U. +-J
o
c
r~ C
=5 O
o in in
>- •- c
• — O fD
tx E
CO C -C
O) .O
u cr o
tl 4-> 4^
co QJ
t.n o -*
03
c^
o
fl
c
a.
o
^-i
u
•- od
X LL.
O c_j
cn
QJ *=r
fO
CO
QJ 0
-C •-
4-^ 4->
4- TD
— C
O
in u
tl
03 TD
E QJ
O 0
4-* Q)
Q.
U CO
x t-
O QJ
4J TD
C
QJ 3
_Q
O Q)
4-> cn
O
TD Q-
OJ X
E Q)
c/i in
Q) •—
(- fO
Q- E
QJ d
t- (t)
>-,
•- o • — -
i~ u ro
QJ O
4-» -Q ro
."0 fD r-~
£--,-,
E
D
TD CO
QJ
u - a.
4-J t/1 >•>
in ^j 4-J
O k- O
r— r- t.
O > QJ
to
- O
QJ .- r—
(O f— -—
i- O 03
QJ Q.
O - fD
-C E •—
U 13 "—
O - C
I- 13 E
-Q 4-» —
- 0 fO
=> .Q LO
ro
QJ
-C CM
+-> r^.
>> o:
QJ -Q Li-
CO o
13 TD
(O QJ CSj
U +-> ^3-
CO
>> —
fD ' — CD
E QO
LO ro
t- 4-1
O C ro
QJ r-*.
cn CD — <
QJ fO
in Q:
3 0 U.
fO 4-> O
o
TD CO
4-» QJ ^3-
fO 4-* '
_c •-
4-» E t/>
QJ
. ,-- (J
c •-
- QJ >
X i- l~
O fO QJ
4-J t/">
tn
10 i— C
4-» fO (O
t- 13
(- QJ 31
O 4-J
fO TD
- E C
E tO
CO Q)
"• CO _C
C QJ 4->
fD -C r—
cn t— ro
(_ QJ
o ai
O
u a) w-
o in o
— (O
E a> -M
cn c
QJ •- OJ
' — TD E
J2 4->
fD C (-
> E CL
^ QJ
^c. -£-- cn
C)
tj
c
QJ
CD
fO
o
cn
O
o
4-J
.U
E
3
C
fD
u
13
QJ
4-1
fD
l_
4->
c
E
rj
L_
0
.c
i—
fD
cn
fD
-C
O
4-*
fD
O
cn
c
N
C
0
tn
4->
E
QJ
13
O
QJ
C
C
o
a.
CO
4-J
13
_cr
F
fO
^
QJ
E
fO
i-
QJ
4->
fO
E
QJ
>
4-J
u
03
O
TD
fO
cn
QJ
TD
U
O
Z3
4-
,
t.
J- O
OJ CD
D. >
fD C 4-1
a> o •-
CT 4-> •-
>^ O O
•- '-4- O
- CO TD
4-* CO (O
O fO t-
O 4-
U 0
4- QJ CO
<- -C •—
U 4J OJ
QJ U- >
Q- 3 QJ
i/l Li_ "~
TO
c:
C 13
O O
E Q-
E E
O O
L) O
1- JZ 0
O O U")
13
>> W CO
*D W «— *
E to
•~ (1) Ct
Q- C i_J
QJ cn
o
QJ in QJ
0 O -
> CL r-
QJ x in
TD QJ in
1- >i ' —
O -Q O
- C <1J
Q) O tO
=3 40 5
4-1 O f—
x c: QJ
•- 3 -C
E M- +->
o
- o
TD 4~» in
r •-
13 CO
O •- QJ
D- U
E -C -
o u >
LJ — QJ
_C TD
ro u
LJ 4- O
O
QJ QJ QJ
-C CO (-
0 O ^
Q- 4->
>! 1- X
<£ Q. E
i)
>
in
o
Ci-
X
U_J
t-
QJ
4-*
in
O
O
JD
Q)
>
CO
O
Q-
X
QJ
4-J
C
3
4->
in
^
t-
.C
4-1
4-*
OJ
•-D
^— ,
rO
m
ro
r*-.
ry
U_
O
cn
^r
in cn
QJ QJ
to >
tn •-
rO cn
r— O
O ' —
-Q Q-
^3 X
in 0)
QJ CD
a> c
t_ t~
-SZ- 4->
4-J fO
c
o o
4-) 4_>
C QJ
•- ~O
TD Q)
Q) l-
> in
*- Q)
TD >
QJ in
t- o
d
CO X
QJ QJ
>
,- <;
CO
O LO
i — tO
O. fD
UJ tJ
65
-------
+J
(/)
Z3
-Q
E
0
u
TJ
Cl
03
i-
>.
JD
C
n
u
C
-J
4—
>i
to
0
Q_
X
QJ
03
in
in
03
o
"D
C
(13
CO
CO
.•o
r--
a;
U-
o
CD
vr
C
o
+J
fd
C
0
-t-J
OJ
"O
.a
t/1
t/)
fO
_c
u
Z3
i/l
to
QJ
O
l~
fU
~o
QJ
t_
Z3
-«->
L)
fO
4-
Z3
CZ
rO
E
OJ
U
TJ
t-T
QJ
>
1/1
0
Q-
X
QJ
<—)
l/i
ro
^3
i/)
(/»
it!
LJ
O
t/i
oO
OJ
ra
-t-J
O
U
-t-J
ro
.C
+J
C
o
_»J
cr
1
cn
«3-
to
^.
0
a
QJ
t_
r—
4-
W-
o
to
QJ
CL
>,
4-1
C
tl
L
Q)
U
"D
C
ro
in
QJ
>
to
O
CL
X
0)
in
t/>
i/i
fO
0
OJ
-C
-t->
CP
>
U~)
C
QJ
to
C
O
l/l
3
J3
cn
c.
-*-j
to
j5
JD
t_-
O
T?
QJ
C
cn
co
QJ
-o
05
(_
QJ
*s:
T
+->
i-
o
tn
1/1
05
o~>
[Z
[—
ZJ
~D
t
O
C
at
4-
0
"*»
-O
03
0
Q-
QJ
4-»
^-
t-
QJ
0) C
E -
^- 1/1 cn
2^ X
Cf. O
O -M
-C a.
u
- ^o
_c c
i ro
M "D
n cu
^ U CL
2: 0)
o: a
o -
^ u .a
CL
-D
O O
Q. O
t/5 U^
66
-------
0.
E
o
.r
o
13
O -Q -M ^-<
67
-------
B. 1 Introduction
This appendix describes an analysis of historical data on transpor-
tation-related releases of chemical substances. The data used in the
analysis are part of the HAZMAT data base operated by the U.S. Department
of Transportation. A complete tape of the data in the HAZMAT data base
was obtained from DOT in August 1986, .tnd the data were studied using the
Statistical Analysis System (SAS) on the EPA mainframe computer.
The purpose of this analysis was to determine whether the physical
and chemical properties of a given substance can be correlated to the
quantity of that substance released during domestic transportation. It
had been ascertained in another study of the HAZMAT data base by ICF
(1984) that the accident rate (number of accidents per mile) for trucks
carrying chemicals is independent of the type of cargo. Also, because
the HAZMAT data concern histories of releases only--not general
transportation data--no information is available from HAZMAT on the
probability of a release for a given accident. Therefore, this study
focused on the quantity of substance released during a given release.
Specifically, the percent of shipment released and the percent of
container released were calculated for different groups of substances.
The percent of shipment released (SHIPREL) and the percent of
container released (CONTREL) were calculated using data from various
fields of the HAZMAT data base as follows:
Percent of shipment released (SHIPREL) = (B-l)
Quantity Released (RQUAN)
Number of the Shipment's Containers (NSH1) x Container's Capacity (CAP1)
x 100
Percent of Container Released (CONTREL) = (B-2)
Quantity Released (RQUAN)
Number of Failed Containers (NFL1) x Container's Capacity (CAP1)
x 100.
One problem encountered in performing these calculations was that
some of the HAZMAT data records used differeit units to report the
quantity released (RQUAN) and the container's capacity (CAP1).
Therefore, units had to be converted to the smallest possible unit within
the measuring scale available for each DOT hazard class. Summary
statistics (mean, standard deviation, median, 95th percentile, and 95
percent confidence limit), frequency tables, frequency histograms, and
analysis of variance (ANOVA) were prepared for the SHIPREL and CONTREL.
A series of ANOVAs was performed on each of the SHIPREL and CONTREL to
70
-------
determine sources of variation within the sample. When a source of
variation proved to be significant, the SHIPREL and CONTREL for that
sample were analyzed separately and summary statistics were determined.
A Chi-Square test was performed on the frequency tables of the SHIPREL
and CONTREL. The Chi-Square test for the homogeneity of the distribution
of each percentage among the levels of the factors was considered in the
analysis. The correlation between the quantity released and the shipment
size is presented in this appendix. The shipment size is calculated as
the number of containers per shipment (NSHI) x the container's capacity
(CAP!).
An overview of the HAZMAT data base is presented in Section B.2. A
discussion of the analysis of variance (ANOVA) method is found in Section
B.3.1 and a review of the Chi-Square method is given in Section B.3.2.
Section B.4 defines the factors considered in the analysis and their
levels. Analysis of variance and summary statistics results are provided
in Section B.5 for the percent of shipment released. Similarly, analysis
of variance and summary statistics results for the percent of container
released are presented in Section B.6. Frequency distribution and the
Chi-Square test of homogeneity results are provided in Section B.7 for
the percent of shipment released and in Section B.8 for the percent of
container released. Correlation coefficients between the quantity
released and the shipment size are found in Section B.9. Section B.10
discusses the conclusions derived from these analyses.
B.2 The HAZMAT Data Base
The primary data source used in estimating predictive release factors
for each hazard class was the DOT's Hazardous Material Incident File
(HAZMAT). This data base is maintained on the DOT's Digital Elec;ronic
Corporation DEC10 computer in Cambridge, Massachusetts. As of 1936,
HAZMAT contained 151,067 records documenting inadvertent releases of
hazardous materials. The data in HAZMAT are provided by carriers on the
Hazardous Materials Incident Report (form DOT F 5800.1) whenever there is
an unintentional hazardous substance release. The types of data
contained in HAZMAT are listed in Table B-l.
The data in the HAZMAT data base were manipulated using the
Statistical Analysis System (SAS) on the EPA mainframe computer. This
was done in order to calculate the relative frequency distribution of
the percent of shipment released and the percent of container contents
released for a given hazard class carried by each mode of transportation.
Phone conversation with Sadie Willoughby, USDOT, September 25, 1986.
71
-------
Table 6-1 Types ot Data Contained in the HA/MAT Data Base
Report number
Multiple code
Date ot incident
Incident city
Incident state
Mode
Carr .er's ID
Carrier's name
Shipper's ID
Shipper's name
Oi ig m city
Origin state
Dest mat inn c ity
Destination state
Major injuries
Minor miuries
Deaths
Damages
Damage code
(1 = Damage unknown,
0 = Damage as shown)
Quant ity
released
Units, of quantity
released
Commodity code
Commodity name
Commodity class
Failure code 1
Container 1
Fa i lure code 2
Container 1
Container 1
Capacity of
Container 1
Capacity units
Container 1
Number of fai led
Lontainers
Number of containers
in shipment
Gauge of
Container 1
Manul d(. turer ' s ID
Tank car ID No
Labe I or
placard
Registration
exempt ion no.
Inspection date
General cause
of incident
Result of
release
Miscel laneous
info 1
Miscellaneous
info 2
Container 2
code
Date added to
,iata base
Date of last
change
Source DOT Research and Special Programs Administration,
Washington, DC No date
72
-------
The types of HAZMAT data included in this statistical analysis are
presented in Table B-2. Four DOT hazard classes w>re excluded from the
statistical analysis. They were (1) blasting agents, (2) radioactive
materials, (3) explosives (A, B, and C), and (4) er.iological agents.
The modes of transportation covered in HAZMAT
-------
Table B-2. HAZMAT Ddtd Used in the Statistical Arid lysis
Multiple code
Mode
Quantity released
Commodity code
Commodity name
Commodity class
Fai lure code
Capacity of container
Capacity units of container
Number of failed containers
Number of containers in shipment
74
-------
-------
mode of transportation). The second source of variation is called the
error or residual, which is the part of the total variation caused by
other factors not investigated.
For each source of variation, the degrees of freedom (DF), sum of
squares, mean squares, F-value, and significance of the F-value (or the
P-value) are calculated (see Table B-3).
The number of degrees of freedom for the model source is equal to
the number of independent comparisons between the averages of the
levels of that factor and the grand average of the factor.
Therefore, the degrees of freedom of a model source equal the
number of levels of that source minus one (e.g., for physical
states, the number of levels is limited to three, that is, gas,
1iquid, and solid).
Sum of squares of the model source is the sum of the square of the
mean deviations of the source (e.g., chemical classes) from the
grand mean of the data. Therefore, the sum of squares of the
model source tends to te large if the individual means vary
considerably around the grand mean. The corrected total sum of
squares (SST) is then equal to the sum of the squares of the data
from the grand mean. The error sum of squares (SSE) is the
difference between the total sum of squares and the model sum of
squares.
The mean squares are obtained by dividing the sums of squares by
the corresponding degrees of freedom. Squares can be considered
as the average of the sum of squares.
The F-value of the model is obtained by dividing the model mean
square by the error mean square. This ratio follows a probabili ,y
distribution known as the F-distribution.
The P-value corresponds to the area to the right of the F-value
under the probability curve of the F-distribution. Therefore, the
P-value of a source of variation is the probability that the
contribution of that source to the total variation is not
significant. Accordingly, if the P-value is small, there is a
high probability (1-P) that the contribution is significant. The
P-value is considered small if it does not exceed a pre-assigned
level known as the significance level. The significance level
assigned in this study is 0.10.
The relative frequency histograms presented in Figures B-l through
B-7 are skewed and U-shaped; while under the assumption of normality of
76
-------
PERCENTAGE
80
70
60
50
30 +
20 +
10 +
*****
*****
*****
*****
KXMMM
MMM M M
MMM M M
*****
*****
*****
*****
*****
*****
*****
10
30 50
SHIPREL MIDPOINT
90
Figure B-l. Percentage bar chart for the frequency distribution
of the percent of shipment released for liquids.
77
-------
PERCENTAGE
80
70
60
50
30
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
10
*****
30 50
SHIPREL MIDPOINT
70
90
Figure B-2. Percentage bar chart for the frequency distribution
of the percent of shipment released for solids.
78
-------
PERCENTAGE
80
70
60
50
40
30
20
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** ***** *****
30 50 70
SHIPREL MIDPOINT
*****
*****
*****
10
90
Figure B-3. Percentage bar chart for the frequency distribution
of the percent of shipment released for gases.
79
-------
PERCENTAGE
50
40
30
20
10
1 *****
+ *****
*****
*****
*****
*****
*****
**-***
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** *****
***** *****
***** ***» * *****
**•<** ***** *****
***** »**» * *****
***** ***** *****
***** ***** ***** *****
***** *»*„» ^^ ^^
10 30 SO 70
S IIPREL MIDPOINT
*****
*****
*****
*****
*****
*****
*****
*****
90
Figure B-4. Percentage bar ^hart for the frequency distribution of the percent
of shipment released for the air mode of transportation.
80
-------
PERCENTAGE
80
70
60
50
30
20
10
*****
*****
*****
*****
tt K M N K
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
10
30 50
SHIPREL MIDPOINT
70
90
Figure B-5. Percentage bar chart for the frequency distribution of the percent
of shipment released for the water mode of transportation.
-------
PERCENTAGE
I
90 +
SO
70
60
50
30
20
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** *****
30 50
SHIPREl MIDPOINT
10
70
90
Figure B-6. Percentage bar chart for the frequency distribution of the percent
of shipment released for the rail mode of transportation.
82
-------
PERCENTAGE
| *****
80 + *****
70
60
50
&Q
30
20
10
*****
**ȣ
*****
*****
*****
if if If if it
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
****^
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*^**t
*****
***** ***** *****
***** ***** ***** *****
***** ***** ***** ***** *****
10
30 50
SHIPREL MIDPOINT
70
90
Figure B-7. Percentage bar chart for the frequency distribution of the percent
of shipment released for the highway mode of transportation.
83
-------
the data, these histograms would be similar in shape to that of a "bell
shaped" curve (symmetric). In most of the applications of ANOVA to
similar data (skewed), the histograms of the log-transformed data are
symmetric, and the results of the application of ANOVA on the original
data and the log-transformed data agree. This conclusion is known in
statistical theory as the "robustness" of the ANOVA to the assumption of
normality (symmetry). This robustness is due to the "monotonicity" of
the log-transformation. The purpose of using ANOVA is to justify the
nonpooling of the data when estimating the percentages of shipment and
container released. The significance (if any) of the statistical
differences was confirmed by the distribution-free Chi-Square test of
homogeneity.
B.3.2 Chi-Square Test of Homogeneity
The Chi-Square tests provide a basis for judging whether the
frequency distributions for each level of a factor can be considered to
be equal (analysis of variance [ANOVA] techniques test whether the means
for these levels of a factor can be considered equal). A frequency
distribution of a set of data and other related statistics are usually
displayed in a two-dimensional cross-classification table, as shown in
Table B-4. The rows in the table represent groupings of the data (e.g.,
groups of SHIPREL: Group 1 = 0 to 20 percent, Group 2 = 20 to
40 percent, Group 3 = 40 to 80 percent, and Group 4 = 80 to
100 percent). The columns represent the levels of the factor considered
(e.g., physical state = liquid, solid, and gas).
The Chi-Square statistic is a measure of the deviations between the
observed and expected frequencies of the data. The expected frequencies
of the data are obtained under the assumption of homogeneity. If Che
assumption of homogeneity is true, then the data in each column of the
frequency table are combined and the percentages in these columns are
used to separate the data in each row into row groups. If the assumption
of homogeneity is not true, then the observed frequencies will tend to
depart from the expected frequencies and the Chi-Square statistic value
will be large. The P-value of the Chi-Square test is the probability
that the frequency distributions of the different levels of a factor are
equal. Accordingly, if the P-value is small (<.10), there is a high
probability (>.90) that the frequency distributions are not equal.
B.4 Other Factors Considered in the Analysis
In order to investigate the possible significant contributions in the
total variation of the percent of shipment released and the variation of
the percent of container released, the following variables were used:
84
-------
Table B-4 The Frequency Distribution and Chi-: _iuare Test
of Homogeneity Results for the Perctnt of
Shipment Released (SHIPREL) by Phys cal State
Class
Frequency
percent
(row PCT)
Gas
Liquid
Solid
Total
1
1707
3 25
93.53
41060
78 07
90.09
4811
9.15
92.61
47578
90.46
2
47
0.09
2 58
2470
4.70
5.42
223
0.42
4.29
2740
5.21
Group
3
45
0 09
2.47
1533
2.91
3 36
131
0.25
2.52
1709
3.25
4
26
0.05
1 42
512
0.97
1.12
30
0 06
0.58
568
1.08
Total
1825
3.47
45575
86.65
5195
9.88
52595
100.00
Frequency Missing = 2,701
Statistics for Table of Class by Grcjp
Statist ic
Chi-Square
Likelihoid Ratio Chi-Square
Mantel- Hdenszel Chi-Square
PHI
Contingency Coefficient
Cramer's V
Degrees of
freedom
(DF) Value
6 70 488
6 79 868
1 11 821
0 037
0 037
0 026
Probabi 1 ity
(P-value)
0 000
0 000
0 001
Effective Sample Size = 52,595.
35
-------
(1) DOT hazard
solid, liquid,
class (or commodity class), (2) physical state (i
or gaseous), and (3) mode of transportation.
B.4.1 DOT Hazard Class
Differences among DOT hazard classes of chemicals are viewed as a
possible factor for variation in the HAZMAT data. Three physical states
are considered: (1) liquid, (2) solid, and (3) gas. Data on physical
state are not provided in individual incident records contained in the
HAZMAT data base. Therefore, data records were classified according to
the type of physical state described for each hazard class in the HAZMAT
data base. The following classification is used in the statistical
analysis:
Physical
state
Liquid
Solid
Gas
Commodity class
(CMCL1*
2
4
6
8
9
20
25
95
10
30
35
60
45
50
55
65
DOT hazard class
Other Regulated Material Class
Other Regulated Material Class
Other Regulated Material Class
Other Regulated Material Class
Other Regulated Material Class
Combustible Liquid
Flammable Liquid
Corrosive Material
Organic Peroxide
Flammable Sol id
Oxidizer
Pois.on, Class I!
Nonflammable Compressed Gas
Flamnable Compressed Gas
Pois >n, Class A
Irri .ating Material
Some DOT hazard classes include materials of more than one ph>sical
state. For example Poisons, Class B, includes both liquids and sclids,
and Poisons, Class A, includes both liquids and gases. In these cases,
the physical state most representative of the hazard class was selected,
Alternative classification by physical state was investigated, anc the
Commodity class (CMCL) is a numerical code corresponding to . DOT
hazard class. Commodity class is used as a field in the DOT HAZMAT
data base.
86
-------
ANOVA results and Chi-Square results for alternative physical state
classification were not significantly different from those results for
the physical state selected.
B.4.2 Physical State
Differences among physical states of the chemicals for which HAZMAT
release records are available are considered a possible factor for
variation in the HAZMAT data. The physical states assigned to each
commodity class (i.e., DOT hazard class) are listed above in Section
B.4.1.
B.4.3 Mode of Transportation
Differences among modes of transportation are considered another
possible reason for variation in the HAZMAT data. The modes of
transportation included in this study are air, barge (waterborne), rail,
and truck.
B.5 Analysis of Variance and Summary Statistics for the Percent of
Shipment Released
The first ANOVA was performed on the SHIPREL data to investigate the
significance of the DOT hazard class as a source of variation in the
HAZMAT data. The results of this ANOVA were presented in Table B-3; they
show that the DOT hazard class has a significant effect on the variations
in the percent of shipment released (P-value < 0.10). This means that
the mean percent of shipment released (SHIPREL) is significantly
different for DOT hazard classes. Therefore, these data for shipment
releases may not be pooled or combined for further analysis.
The second ANOVA was performed on the percent of shipment released
data to investigate the significance of the mode of transportation as a
source of variation. These results are presented in Table B-5; they
indicate that mode of transportation has a significant effect on the
variation in SHIPREL (P-value < .10). The significance of the mode of
transportation implied that the SHIPREL data for different modes of
transportation should not be combined for further analysis.
Summary statistics (number of data records used; mean and standard
deviation, 95 percent upper confidence limit, median, and 90th
percentile) for the SHIPREL are presented in Table B-6 (by physical
state), Table B-7 (by mode of transportation), and Table B-8 (by physical
state and mode of transportation).
The overall average (that is, for all DOT hazard classes and all
modes of transportation) of the percent of shipment released (SHIPREL) is
87
-------
Ill
"D
O
z;
o ~~*
T>
O
2!
-------
Table B-6. Summary Statistics for the Percent of Shipment Released
(SHIPREL) for Each Physical State (Gas, Liquid, Solid)
Physical
state
Gas
Liquid
Solid
All
Number
of data
records
(N)
1,697
45,904
5,484
Mean
13.2099
11 6824
9.3291
11.26
Standard
deviation
29.4428
24.2284
22 3211
Upper 90%
confidence
limit
14.3820
11 8679
9.8234
Median
0.03333
1.25000
0.62500
90th
Percent i le
70.1667
40.0000
25.0000
Table B-7. Summary Statistics for the Percent of Shipment Released
(SHIPREL) by the Mode of Transportation
Physical
state
Air
Barge
Rail
Truck
All
Number
of data
records
(N)
594
110
6,130
46,251
Mean
16.7608
10 8142
7,0591
12.0090
11.26
Standard
deviation
28.4926
24 7357
22 7003
24.3063
Upper 90%
confidence
limit
18.6781
14.6820
7 5346
12 1944
Median
1 96154
0 54710
0.01238
1 49254
90th
Percent i le
52.8977
33.2885
10.0000
40.0000
Table B-8 Summary Statistics for the Percent of Shipment Released
(SHIPREL) by the Physical State and Mode of Transportation
Physical
state
Gas
Gas
Gas
Gas
L iquid
Liquid
L iquid
Liquid
Solid
Solid
Solid
Solid
Mode of
transportation
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
Number
of data
records
(N)
9
6
1,043
639
538
63
4,616
40,667
47
21
471
4,945
Mean
57 7914
39.9056
5 0815
25 5988
16 4369
10 7439
6 9636
12.1571
12 6114
2.7800
12.3741
9 0357
Standard
deviation
46.8759
47 3122
20 4939
35 7337
27 9920
24.2523
22.4186
24.3102
24 2501
6 4653
28 5742
21 6341
Upper 9 OX
confidence
1 imit
83 4169
71 5825
6 1222
27 9171
18 4161
15 1097
7.5047
12 3548
18 4125
5.0938
14.5333
9 5402
Median
63 3333
16.8950
0 0042
4.7225
1 9615
0.5594
0 0152
1.6667
0 7000
0 3030
0.2353
0.6494
90th
Percent i le
100.000
100.000
1.634
ICO. 000
JO 000
23.739
3.982
4 1 . 667
EL 000
11 123
49.999
25.000
89
-------
11.26 percent. This means that when an accident involving a release
occurs, the average loss of cargo will be 11 percent.
The results displayed in Table B-6 indicate that the SHIPREL for
gaseous chemicals had a higher mean and the SHIPREL for solid chemicals
had a lower mean than the overall mean.
It can be seen from the results presented in Table B-7 that HAZMAT
records for air transportation had a higher average percent of shipment
releases, and rail transportation had a lower average percent shipment
releases than the overall mean. The results also show that the averages
for barge and truck transportation are significantly different from the
overall mean. The results in Table B-7 reveal statistically significant
differences among the means of percent of shipment released (SHIPREL) for
the modes of transportation for each physical state.
The means of the percent of shipment released (SHIPREL) presented in
Tables B-6, B-7, and B-8 can be used as estimates or predictions for the
average percent of shipment released. Upper confidence limits are
obtained for the average of percent of shipment released at the
95 percent confidence level. The relative frequency histograms presented
in Figures B-l through B-7 are skewed and U-shaped. For skewed
distributions,the sample mean tends to overestimate the central tendency
of the distribution. The sample median (the value that 50 percent of the
data are less than) is a preferred estimate of the central tendency of
these distributions. The median and the 95th percentile (the value that
95 percent of the data are less than) are presented in Tables B-6, B-7,
and B-8. The 95 percent upper confidence limit and the 95th percentile
represent very conservative estimates of the percentage of shipment and
container releases. The 95 percent upper confidence limit is computed as
the sample mean (1.64 x standard error of the mean). The standard error
of the mean equals the standard deviation divided by the square root of
the sample size. This computation is justified by the "Central L mit
Theorem" for large sample sizes (>30). For small samples (<30), ,he
confidence limit is not justified and the 95th percentile represe its a
nonparametric conservative estimate.
The third analysis performed was a sequence of ANOVAs of the lercent
of shipment released (SHIPREL) to investigate the signficance of he
physical states as a source of variation within each DOT hazard c ass.
The results listed in Tables B-9, B-10, and B-ll show that significant
differences exist among hazard classes having the same assigned physical
state (P-values < .10). Summary statistics for the percent of shipment
released (SHIPREL) classified by DOT hazard class and physical state are
cited in Table B-12.
90
-------
O- CJ
—1
-------
92
-------
«r
_j
CJ
01
>
01
93
-------
Table B-12 Summary Statistics tor the Percent of Shipment Released
(SHIPREL) by the Physical State and Hazard Class
Phys ica 1
state
Gas
Gas
Gas
Gas
L iqu id
Liquid
Liqu id
L iqu id
Liquid
Liquid
L iquid
Liqu id
Solid
Solid
Solid
Solid
Hazard
class
46
50
55
65
2
4
6
8
9
20
25
95
10
30
J5
bO
Number
of data
records
(N)
715
935
21
26
335
52
27
31
213
2407
19G70
23169
349
452
1490
3193
Mean
14 9162
11 4638
24 2708
20.1427
23.7631
23.4281
32.7767
6.7674
8 1421
16 5154
9 9732
12 4449
9 8715
6 1609
U 1130
8 4192
Standard
deviation
31 4173
27 3097
39.3608
34.1419
33 2710
33.4280
37.6030
14.8483
20 1145
28.7261
22.4397
24.8456
23.1605
19.6327
25.3C16
20 9371
Upper 90X.
cont idence
limit
16.8431
12 9285
38 3572
31.1238
26.7442
31.0305
44.6449
11 1410
10 4024
17 4756
10.2356
12 7126
11 9047
7 6754
13.1880
9 0268
,
Median
0 0760
0 0165
1.6667
2.7652
5 0000
4.3182
11.4286
1.4933
0 5195
1 3043
0 9091
1 6667
1.1111
0 1230
1 0000
0 5882
90th
Percent i le
89.216
52 680
100.000
100.000
100.000
94 . 000
100.000
23.610
: 2 . ooo
(6.690
. 2.727
43,804
25.000
12.100
.1.570
.'5.000
94
-------
The fourth analysis performed on a sequence of the percent of
shipment released (SHIPREL) investigated the significance of the mode of
transportation for each physical state. The P-values of the ANOVA
results are shown in Table B-13 and indicate that mode of transportation
is a significant factor for some combinations of physical state and
commodity (hazard) class (uses with P-value < .10). Table B-14 presents
summary statistics for the fraction of shipment released by commodity
(hazard) class, physical state, and mode of transportation. It should be
noted that a small number of data records were used for the computation
of the summary statistics for some of the cases in Table B-14 (e.g.,
first line: physical state = gas, CMCL = 45, and mode = air). Results
based on number of data records (N) less than ten are unreliable and
should not be considered representative of the population from which they
were drawn.
B.6 Analysis of Variance. Summary Statistics, and Confidence Limits
for the Percent of Container Contents Released (CONTREL)
Four sequences of ANOVAs were performed on the percent of container
contents released (CONTREL). The first ANOVA was performed to
investigate the significance of the commodity class (DOT hazard class).
The results are listed in Table B-15 and indicate that commodity class
(DOT hazard class) is a significant factor. The second ANOVA was
performed to investigate the significance of the mode of transportation.
These results are contained in Table B-16 and indicate that mode of
transportation is also a significant factor.
Summary statistics for the percent of container contents released are
listed in Table B-17 (by physical state), Table B-18 (by mode of
transportation), and Table B-19 (by physical state and mode of
transportation). The overall average of the percent of container
contents released is 30 percent. The results in Table B-17 show that the
mean values for percent of container contents released for liquids and
solids do not differ significantly from the overall mean. These results
also demonstrate that chemicals shipped as a gas have a lower mean of
percent of container contents released than does the overall mean. The
results in Table B-18 show that the mean percent of container contents
released for air, barge, and truck did not differ significantly from the
overall mean, but that rail had a lower mean percent of container
contents released than did the overall mean. The results in Table B-19
reveal that the mean percent of container contents released classified by
mode of transportation and commodity class differs from the overall mean.
The third sequence of ANOVAs was performed to investigate the
significance of the physical states within each commodity (DOT hazard)
class. The results are listed in Tables B-20, B-21, and B-22 and show
95
-------
Taole B-13. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) by the Mode of
Transportation for Each Physical State and Each
Commodity Class
Physical state
L iqu id
Sol id
Gas
Commodity class
(CMCL)
2
4
6
8
9
20
25
95
10
30
35
60
45
50
55
65
DOT hazard class
ORM-A
ORM-B
ORM-C
ORM-D
ORM-E
Combustible Liquid
Flammable Liquid
Corrosive Material
Organic Peroxide
Flammable Solid
Oxidizer
Poiion B
Nonf lammab le/Compr^ssed
Gas
Flammable Compress -d
Gas
Poison A
Irr i tat ing Mater la 1
P-value
.5784
0001
.7640
.0048
.6876
.0011
.0032
.0001
.7986
.0930
.0001
.1061
.0001
0001
2624
.5625
96
-------
Table B-14 Summary Statistics for the Percent of Shipment Released (ShIPREL) by the Commodity
Class (DOT Hazard Class), Physical State, and Mode of Transportation
Pllys K'jl
state
bdS
Gas
ucTJ
uas
uj;.
UdS
GdS
Odb
Gdo
GdS
bdS
Gas
UflS
L iquid
L iquid
Liquid
L iquid
L iquid
L iquid
L iqu id
Liquid
L iquid
L iquid
t iquid
L iquid
L iquid
Liquid
L iqu id
1 iqu id
L iqu id
CMCL
45
45
45
45
50
50
50
50
55
55
65
65
65
2
2
2
4
4
4
6
6
8
8
9
9
9
20
20
20
20
Mode of
transportat ion
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
Rail
Truck
Air
Barge
Truck
Air
Rail
Truck
Air
Rail
Truck
Rail
Truck
Air
Truck
Barge
Rail
Truck
Air
Barge
Rail
Truck
Number
of data
records
(N)
3
4
431
277
6
1
609
319
3
18
0
1
25
38
18
279
17
3
32
2
25
4
27
1
21
191
6
7
389
2005
Mean
100.000
34 804
5 276
28 708
36.687
100.000
4.968
23 113
0 169
28 288
__.
0.216
20 940
26 487
29.207
23 041
49 863
2 402
11.356
24.873
33 409
15 534
5.469
0 014
11.383
7 828
17.486
5 894
11 464
17 530
Standard
deviation
0.0000
44 3530
21.0567
37.6940
43.7304
---
20.1511
33 6233
0.2869
41.2682
___
---
34.5982
33 5358
43 3406
32.5874
41,1131
4 0504
19.7094
35.1183
38 4032
23 2395
13.3604
---
23 7671
19 7579
15 3460
11.3715
28 2030
28 7988
Upper 90%
confidence
1 imit
100.000
71.174
6.939
32 422
65 966
---
6.307
26.200
0 440
44 240
__.
---
32 288
35 409
45 960
26 240
66.216
6 237
17.070
65.598
46 005
34 590
9 685
.--
19 889
10 173
27.760
12.943
13 810
18 584
Median
100.000
16.895
0 009
6.857
16 192
100 000
0 003
4.000
0.003
5 435
—
0 216
3.030
12 500
0.656
5.000
33 333
0.124
2.947
24.873
11.429
5 373
1.250
0.014
0.519
0 519
18 750
0.132
0 024
2 062
90th
Percent lie
100.000
100.000
0.992
100.000
100 000
100.000
1.861
86 364
0 500
100.000
—
0.216
100.000
91.000
100.000
100 000
100 000
7 078
47.000
49.705
100 000
50.000
12 121
0.014
36 922
26 818
40 000
30 000
62 500
68 290
97
-------
Tdble b-14. (continued)
PhvsKdl
^tate
L iquul
I iqu id
I H|U id
1 if|u id
. iquid
i i.iu id
1 ;qu id
1 iquid
~.n] n1
jo I id
"50 1 id
Sol id
So i id
bond
-ol ill
bol id
:,olid
jol id
bo! id
bo 1 1 d
foi id
bol id
iolu)
CMCL
25
25
25
25
95
95
95
95
10
10
10
30
30
30
30
35
35
35
35
60
60
eo
60
Mode of
transportat ion
Air
Barge
Rdi 1
Trjrk
Air
Barge
Rail
Truck
Air
Rail
Truck
Air
Barge
Rai 1
Truck
Air
Barge
Rail
l>u,k
Air
Barge
Rail
Truck
Number
of data
records
(N)
379
43
1667
17581
94
32
2516
20527
1
7
341
4
3
79
366
9
3
208
12/0
33
15
177
2968
Mean
11 577
12.323
8.125
10 108
25 897
10 018
5 294
13 264
3 333
14 480
9 796
14 174
1 010
9 789
5.332
30 392
5 672
19 529
10 /84
7 654
2 555
5 037
8 657
Standard
deviation
23 1801
25.3477
24 2840
22 2259
33.7606
25.4415
19.5042
25 2236
___
37.7125
22 8771
12 7232
1.7494
28.5600
17 2089
35 2536
9 5221
33 6799
23 3429
20.1050
6.6166
17.8456
21 1479
Upper 90%
confidence
limit
13.529
18.662
9.100
10.383
31 608
17 394
5 931
13 552
---
37 856
11 828
24 607
2.667
15 059
6 808
49 664
i4 6«9
,.3 358
11.858
13 594
5 358
7 237
9 293
Median
1.120
0.58B
0.020
1.091
10.000
0.851
0.012
2 000
3.333
0.061
1 224
15.625
0 000
0.007
0 180
16 667
0 303
1.551
0 y46
0 313
0 395
0 120
0 625
90th
Percent i le
50.000
56.970
20.000
33 333
100.000
55 909
5.609
50 000
3 333
100 000
25 000
25.000
3.030
44 118
11 111
100.000
16 667
100.000
33.333
34 000
12 812
6 317
25 000
98
-------
d
•4
o
L4_
O
o
o
o
--* ro
99
-------
5:
g
o
Ll_
O
100
-------
Tdhle B-17. Summary Statistics for the Percent of Container Contents
Released (CONTREL) for Each Physical State (Solid, Liquid, Gas)
Physical
state
Number
of data
records
(N)
Mean
Upper 90%
Standard confidence
deviation limit
Median
90th
Percentile
Gas
Liquid
Solid
1,680
45,462
5,395
22.0186
30.1369
31.4928
38 0087
38.6531
39.3124
23 5394
30.4341
32.3705
0.04962
7.27273
9.09091
100
100
100
Table B-18 Summary Statistics for the Percent of Container Contents Released
(CONTREL) by the Mode of Transportation
Physical
state
Number
of data
records
(N)
Mean
Standard
deviation
Upper 90%
confidence
limit
Median
90th
Percenti le
Air 594 16.7608 28.4926 18.6781
Barge 110 10 8142 24 7357 14.6820
Rail 6,130 7 0591 22 7003 7.5346
Truck 46,251 12.0090 24.3063 12 1944
All 11 26
1.96154
0 54710
0 01238
1.49254
52 8977
33.2885
10.0000
40.0000
Table B-19 Summary Statistics for the Percent of Container Contents Released
(CONTREL) by the Physical State and Mode of Transportation
Phys ical
state
Gas
Gas
Gas
Gas
Liquid
Liquid
L iquid
L iquid
Solid
Solid
Sol id
Solid
Number
of data
records
(N)
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
Air
B<;rge
Rail
Truck
Mean
9
6
1043
622
534
83
4608
40257
46
21
469
4659
Standard
deviat ion
81 9729
55 1834
5 1787
49 0691
27.6085
28.2776
10.7230
32 3964
30 5908
21 3520
28 9877
31 7869
Upper 90%
conf idence
1 imit
36.8916
49 8122
20.7024
43.3285
36 1999
36.7816
27 7564
39 1327
39 9684
32.1740
40.7142
39 1929
Median
57 3785
14 5119
3.8966
45.5945
24 4754
20 2030
9 9052
32 0064
18.8047
7 3101
25 2276
30 6624
90th
Percent i le
106.567
95 855
6 461
52.544
30 742
36.352
11.541
32.787
42.377
35.394
32.748
32.911
101
-------
-------
<:
2:
cr:
O
a: _ J r—
o
Ll_
O
CD --H
--t CD
103
-------
g
CD t_J U.
104
-------
significant differences between the physical states within each commodity
(DOT hazard) (P-values < .10).
Summary statistics for the percent of container contents released
classified by commodity (DOT hazard) class and physical state are listed
in Table B-23.
The fourth sequence of ANOVAs was performed on the percent of
container contents released to investigate the significance of mode of
transportation within each physical state sorted by commodity class. The
P-values of the ANOVA results, which are found in Table B-24, show that
for some of the physical state/commodity class combinations (cases with
P-value < .10), the mode of transportation was significant. Summary
statistics for the percent of container contents released classified by
commodity class, physical state, and mode of transportation are displayed
in Table B-25. Note, however, that results obtained from the small
number of data records (< 10 records) are unreliable.
B.7 Frequency Distribution and Chi-Square Test of Homogeneity
Results for the Percent of Shipment Released
The records of percent of shipment released were classified into five
groups (intervals) defined as follows:
Group Percent of shipment released
1 0 < SHIPREL < 20%
2 20 < SHIPREL < 40%
3 40 < SHIPREL < 60%
4 60 < SHIPREL < 80%
5 80 < SHIPREL <100%
The first Chi-Square test of homogeneity was performed on the percent
of shipment released compared to the frequency distributions for the
three physical states. The frequency distribution for each physical
state and the Chi-Square test results were presented in Table B-4. Also,
percentage frequency histograms (percentage bar charts) for physical
states are shown in Figures B-l, B-2, and B-3. These figures illustrate
that the frequency distributions of the SHIPREL for the three physical
states are different. The Chi-Square test results confirmed this
observation (P-value < .10).
The second Chi-Square test of homogeneity was performed on the
percent of shipment released for each commodity class separately to
compare the frequency distributions for the physical states. The results
are presented in Tables B-26 (liquids), B-27 (solids), and B-28 (gas);
105
-------
Summary Statistics for the Percent of Container
Contents Released (CONTREL) by the Commodity
CldSb (DOT Hazard Class) and Physical State
Physical
state
Gas
Gas
Gas
Gas
L iquid
Liquid
Liquid
L iquid
L iquid
Liquid
L iquid
L iquid
Solid
Solid
Sol id
Solid
CMCL
45
50
55
65
2
4
6
a
9
20
25
95
10
30
35
60
Number
of data
records
(N)
708
925
21
26
332
52
27
29
213
2396
19579
22854
336
439
1457
3163
Mean
25 5146
18.7533
32.0308
34.8997
46 3385
31 5434
43.5444
73 2245
25 8910
20.8682
26.0440
34 3455
43.8933
20 5134
34 0679
30 5132
Standard
deviat ion
40.5842
35.3042
44.4022
42 4623
41 7415
38.2472
40.9765
35.8201
36.4302
32 5887
35.9281
40.7964
41.0719
35 1684
43 3417
33.7279
Upper 90%
confidence
1 unit
28 0160
20.6570
47.9214
48.5568
50 0956
40 2419
56 4773
84 1332
29 9846
21.9601
26.4651
34 7880
47 5680
23 2661
35 8011
31 6425
Median
0 093
0.019
6.489
9 091
36.364
10.330
27.600
100.000
5.000
2.000
5.455
10 000
25 000
1 818
11.111
7.273
90th
Percent lie
100.000
100.000
100.000
100.000
100.000
100.000
100.000
100.000
86 541
100.000
100.000
100.000
100.0' 0
100.0 0
100. OuO
106
-------
Table B-24 Chi-Square Test of Homogeneity of Results for the Percent
of Container Contents Released (CONTREL) b/ Mode of
Transportation for Each Combination of Physical State and
Commodity Class (DOT Hazard Class)
Physical state
L iquid
Solid
Commodity class
(CMCL)
2
4
6
8
9
20
25
95
10
30
35
60
DOT hazard class
ORM-A
ORM-B
ORM-C
ORM-D
ORM-E
Combustible Liquid
Flammable Liquid
Corrosive Material
Organic Peroxide
Flammable Solid
Ox i d i i e r
Poison B
P-value
329
454
.094
.020
.768
.000
.000
000
.997
.000
.000
.002
Gas
45
50
55
65
Nonflammable Compressed .000
Gas
Flammable Compressed .000
Gas
Poison A .848
Irritating Material .998
107
-------
Table B-25 Summary Stdtistics for the Percent of Container Released
(CONTREL) by the Mode ot Transportation foi Each Physical
State and Each Commodity Class (DOT Hazard Class)
Physical
state
Gas
Gas
Gas
bdS
Gas
GaS
bas
u,iS
Gas
Gas
Gas
Gas
baS
L iqu id
L iquid
L iqind
I iquid
t iqu id
Liquid
Liquid
L iquid
t iquul
L iqu id
1 iqu id
L iquic!
L iqu id
Commodi ty
c lass
(code for
DOT hazard
class)3
45
45
45
45
50
50
bO
bO
55
55
65
65
65
2
2
2
4
4
4
6
6
8
8
9
9
9
Mode of
transportat ion
Air
Barge
Rail
Truck
Air
Barge
Rail
T ruck
Rail
Truck
Air
Barge
Truck
Air
Rail
Truck
Air
Rail
Truck
Rail
Truck
Air
Truck
Barge
Rail
Truck
Number
of data
records
(N)
3
4
431
270
6
1
609
309
3
IB
0
1
25
37
18
277
17
3
32
2
25
4
25
1
21
191
Mean
100.000
57.721
5.278
56.514
72 959
100.000
5.118
44 312
3 351
36.811
--
0.216
36.287
44.125
36 738
47.2b8
59 955
2 402
19.182
49 725
43 050
53 753
76 343
2 657
21 602
26 043
Standard
deviat ion
0.0000
49.4996
21 0563
43.5303
43 4170
--
20.5240
42.1445
5.7583
46.3274
--
--
42 7323
39.7553
44 9297
41 8516
42.3318
4 0504
28.2501
70 2651
40.1229
34 9949
35 6427
--
34 9612
36.7334
Upper 90%
confidence
1 imit
100.000
98.311
6.941
60.859
100 000
--
6.481
48.244
8.803
54.719
--
--
50.303
54 643
54 105
51 382
76 793
6 237
27 372
100 000
56 210
82.429
88 034
--
36 114
30 402
Median
100.000
62.728
0.009
66.667
100.000
100.000
0 003
36.364
0.050
7.444
--
0.216
9.091
50.000
10 694
36.364
80 000
0.124
5 227
49.725
27 600
50 000
100 000
2 8b7
7.514
5.000
90th
Percent i le
100 000
100.000
0.992
100 000
100 000
100 000
1.975
100 000
10.000
100.000
--
0.216
100.000
100.000
100 000
100.000
100 000
7 078
74 750
99.410
100 000
100 000
100 000
2 857
100.000
100 000
108
-------
Table B-25 (continued)
Commodity
class
(code for
Physical DOT hazard
state class)3
Liquid
L iqu id
Liquid
L iquid
L iquid
L iquid
L iquid
L iquid
L iquid
L iquid
L iquid
L iqu id
Solid
Solid
Solid
Solid
Solid
Solid
Solid
Solid
Solid
Solid
Sol id
Solid
Solid
Solid
Solid
20
20
20
20
25
25
25
25
95
95
95
95
10
10
10
30
30
30
30
35
35
35
35
60
60
60
60
Number
of data
Mode of records
transportation (N)
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck.
Air
Rail
Truck
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
Air
Barge
Rail
Truck
6
7
388
1995
377
43
1665
17494
93
32
2511
20218
1
7
328
4
3
79
353
9
3
207
1238
32
15
176
2940
Mean
58 336
5.913
11.785
22 575
19 958
29.903
12.739
27.432
43.032
31 780
8.890
37.471
10 000
21 667
44.471
54 278
4.049
14 312
21.658
58 039
19.670
39.669
32.992
20.553
25 149
23 303
31 081
Standard
deviation
38.1618
11.3605
28 3146
33.0452
31.1363
36.5267
29.7910
36.2851
41.3037
39.9485
25.7385
41.2103
--
36.6178
41.0993
53 2086
6.9904
33.5832
35 2059
45.1660
32.4649
44.0629
39 5760
33.3119
35.2286
36 4973
38 8852
Upper 90%
confidence
limit
83.886
12.955
14 143
23.788
22.588
39.038
13.936
27.882
50.056
43.362
9 732
37.946
--
44.365
48.193
97.909
10.668
20.509
24.732
82.730
50.409
44.692
34.837
30.211
40.066
27 815
32 257
Median
55.000
0.152
0 024
3.333
3.500
8.485
0 020
7.273
25.000
9.545
0.012
16.000
10 000
0.061
25.000
58.333
0.026
0.008
2.273
75.000
1 818
15.152
10 000
1.653
6.591
1 818
8.864
90th
Percent i le
100.000
30 000
64.316
89 336
80.000
100.000
66.667
100.000
100.000
100.000
27.273
100 000
10.000
100.000
100.000
100 000
12 121
100.000
100.000
100.000
57 143
100 000
100.000
100.000
100.000
100 000
100 000
Refer to Table 3-3 for the corresponding DO! hazard class.
Source Statistical analysis of the HAZMAT data base, 1986. (See Appendix B for more details
109
-------
Table B-26 The Frequency Distribution and Chi-Square
Test Results for the Percent of Shipment
Released (SH1PREL) by Physical State (Liquid)
Class
Frequency
percent
(row PCT)
2
4
6
8
9
20
25
95
Total
1
236
0.52
77.63
44
0.10
80 00
21
0.05
72 41
30
0.07
96. 77
223
0.49
92 92
1919
4 21
83 84
17518
38 44
91 37
21069
46.23
89 83
41060
90 09
2
26
0.06
8.55
5
0 01
9.09
2
0.00
6 90
0
0.00
0.00
13
0 03
5 42
164
0.36
7 16
870
1.91
4,54
1390
3.05
5 93
2470
5.42
Group
3
30
0 07
9 87
2
0.00
3.64
2
0.00
6 90
1
0.00
3 23
3
0.01
1.25
126
0 28
5 50
552
1.21
2 88
817
1 79
3.48
1533
3 36
4
12
0.03
3 95
4
0.01
7 27
4
0.01
13.79
0
0 00
0 00
1
0 00
0 42
80
0.18
3 49
232
0 51
1 21
179
0 39
0.76
512
1 12
Total
304
0.67
55
0.12
29
0.06
31
0 07
240
0.53
2269
5 02
19172
42 07
23455
51 46
45575
100 00
Frequency Missing = 2,317
110
-------
Table B-26 (continued)
Statistics for Table of Class by Group
Degrees of
f reedom
Statistic (DF)
Chi Square 21
Likelihood Ratio Chi-Square 21
Mantel -Hderisiel Chi-Square 1
PHI
Continyency Coefficient
Cramer's V
Value
394 998
294.348
5 212
0.093
0 093
0.054
Probabi 1 ity
(P-value)
0 000
0 000
0 022
Effective Sample Size - 45,575.
Ill
-------
Tdble B-27 The Frequency Distribution and Chi-Sqiare Test
of Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Physical State (Solid)
Class
Frequency
percent
(row PCf)
10
30
35
60
Total
1
299
5.76
92.86
414
7.97
96.28
1248
24.02
90 11
2850
54 86
93 20
4«11
92 61
2
14
0.27
4 35
10
0.19
2 33
82
1.58
5 92
117
2 25
3 83
223
4 ?9
Group
3
6
0 12
1 86
5
0.10
1 16
47
0 90
3.39
73
1 41
2 39
131
2 52
4
3
0.06
0 93
1
0.02
0 23
8
0.15
0.58
18
0 35
0.59
39
0 58
Total
322
6 20
430
8.28
1385
26.66
3058
56.86
5195
100.00
hrequency Missing = 234
Statistics for Table of Class by Group
Degrees of
f reedom
statistic (OF)
Cm-Square 9
Likelihood Ritio Chi-Square 9
Mantel -Harnsze 1 Chi-Square 1
PHI
Contingency Coefficient
Cramer's V
Value
25 346
25 669
0 674
0 070
0 070
0 040
Pi'obabi 1 ity
(P-Vdlue)
0 003
0 002
0 412
Effective Sample Size - 5,195
112
-------
Table B-28 The Frequency Distribution and Chi-Square Test of
Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Physical State (Gas)
Class
Frequency
percent
row PCT 1
45 722
39.56
92 92
50 935
51.23
94.16
55 21
1.15
91.30
65 29
1 59
90 63
Total 1707
93.53
2
24
1.32
3 09
21
1.15
2 11
1
0.05
4.35
1
0.05
3 13
47
2 58
Group
3
20
1.10
2 57
23
1.26
2 32
1
0.05
4.35
1
0.05
3.13
45
2 47
4
11
0.60
1 42
14
0.77
1.41
0
00.00
0 00
1
0 05
3.13
26
1.42
Total
777
42.58
993
54.41
23
1.26
32
1 75
1825
100.00
Frequency Missing = 150
Statistics for Table of Class by Group
Statistic
Chi-Square
Likelihood Ratio Chi-Square
Mante!-Haenszel Chi-Square
PHI
Contingency Coefficient
Cramer's V
Degrees of
freedom
(OF) Value
9 3.538
9 3 . 589
1 0.011
0 044
0 044
0.0250
Probabi 1 ity
(P-value)
0 939
0 936
0 915
Effective Sample Size - 1,825.
113
-------
they show that the frequency distributions for the physical state within
each commodity (DOT hazard) class are significantly different.
The third Chi-Square test was performed on the percent of shipment
released to compare the frequency distributions fo^ modes of transporta-
tion. The results, provided in Table B-29, show tiat the frequency
distributions for modes of transportation are significantly different.
The frequency histograms for each mode are presented in Figures B-4, B-5,
B-6, and B-7. Table B-30 lists the resulting P-values (observed
significance level) from this test.
The fourth Chi-Square test was performed on the percent of shipment
released for each commodity class separately to compare the frequency
distributions of modes of transportation. The results are presented in
Tables B-31, B-32, and B-33 and show that the frequency distributions of
the modes of transportation for liquid chemicals and gas chemicals are
significantly different. The results also indicate that the frequency
distributions of the modes of transportation carrying solid chemicals are
not significantly different (P-value > 0.1). This result implies that
the values of percent of shipment released (SHIPREL) for each mode of
transportation used to carry solid chemicals are similarly distributed.
The fifth Chi-Square test was performed on the percent of shipment
released for each physical state within each DOT hazard class to compare
distributions for modes of transportation. The P-values of the
Chi-Square tests are listed in Table B-34 and show that the frequency
distributions for modes of transportation are significantly different for
some of the physical states (cases with P-value < .1).
B.8 Frequency Distribution and Chi-Square Test of Homogeneity
Results for the Percent of Container Contents Released
The values for percent fractions of container contents released
(CONTROL) were classified into five groups (intervals) defined as follows:
Group Percent of shipment released
1 0 < CONTREL < 20%
2 20 < CONTREL < 40%
3 40 < CONTREL < 60%
4 60 < CONTREL < 80%
5 80 < CONTREL £100%
The first Chi-Square test of homogeneity was performed on the percent
of container contents released to compare the frequency distributions of
the commodity classes. The frequency distributions of the commodity
114
-------
Table B-29. The Frequency Distribution and Chi-Square Test
of Homogeneity Results for the Percent of Shipment
Released (SHIPREL) by Mode of Transportation
Class
Frequency
percent
(row PCT)
Air-
Barge
Rail
Truck
Total
1
499
0.95
63.31
105
0 20
90 52
6115
11 63
96 85
40859
77.69
89 67
475/8
90.46
2
50
0 10
8 35
7
0 01
6 03
89
0 17
1 41
2594
4 93
5 69
2740
5 21
Group
3
39
0.07
6.51
1
0 00
0.86
68
0.13
1.08
1601
3 04
3 51
1709
3 25
4
11
0 02
1 84
3
0.01
2 59
42
0 08
0.67
512
0 97
1 12
568
1 08
Total
599
1 14
116
0.22
6314
12 00
45566
86 64
52595
100.00
Frequency Missing = 2,701.
Statisticb for Table of Class by Group
Statistic
Chi-Square
Likelihood Ratio Chi- Square
Mantel-Haenszel Chi-Square
PHI
Contingency Coefficient
Cramer's V
Degrees of
freedom
(DF) Value
9 381 539
9 479.704
1 51 041
0 085
0.085
0.049
Probabi 1 ity
(P-value)
0 000
0 000
0 001
Effective Sample Size = 52,595
115
-------
Table B-30. Chi-Square lebt of Homogeneity Results for the Percent of
Shipment Released (SHIPREL) by Mode of Transportation for
Each Physical State and Each Commodity Class (DOT Hazard
Class)
Physical state
Liquid
Solid
Commodity class
(CMCL)
2
4
6
8
9
20
25
95
10
30
35
60
DOT hazard class
ORM^A
ORM-B
ORM-C
ORM-D
ORM-E
Combust ible L iquid
Flammable Liquid
Corrosive Material
Organic Peroxide
Flammable Sol id
Oxidizer
Poison B
P-value
170
033
.252
074
936
.000
.000
.000
825
575
042
.200
45
50
55
65
Nonf lamiiwljle Compressed 000
Gas
Flammable Compressed .000
Gas
Poison A .554
Irritating Material .913
116
-------
Table B-il. The Frequency Distribution and Chi-Square Test
of Homogeneity Results for the Percent ot Shipment
Released (SHIPREL) by Mode of Transportation for
Liquids
Mode
Frequency
percent
(row PCT)
Aii-
Barge
Rail
Truck
Total
1
449
O.S9
82.99
80
0.18
89.89
4598
10.09
96.64
35933
78 84
89 41
41060
90.09
2
46
0 10
8 55
5
0 01
5.62
69
0 15
1 45
2350
5 16
5 85
2470
5 42
Group
3
35
0.08
6 47
1
0 00
1 12
55
0 12
1.16
1442
3.16
3.59
1533
3 36
4
11
0.02
2.03
3
0 01
3.37
36
0.08
0.76
462
1.01
1.15
512
1.12
Total
541
1.19
89
0.20
4758
10 44
40187
88 18
45575
100.00
Frequency Missing = 2,371
Statistics for Table of Mode by Group
Statistic
Chi-Square
Likelihood Ratio Chi-Square
Mantel-Haenszel Chi-Square
PHI
Contingency Coefficient
Cramer's V
Degrees of
freedom
(DF) Value
9 294.850
9 369 332
1 26 387
0.080
0.080
0 046
Probabi lity
(P-value)
0 000
0.000
0.001
Effective Sample Size = 45,575
117
-------
Tdblt: B-3,'. The Frequency Distribution and Chi-Square les:
of Homoqeneity Results for the Percent of
Shipment Released (SHlPREL) by Mode of
Transportation for Solids
Mode
broup
Frequency
percent
(row PCI)
Air
Barge
Rail
Truck
Total
1
36
0 73
84 44
20
0.38
95 24
405
7.60
93 97
4348
83 70
92 55
4811
92.61
2
3
0 06
6 67
1
0.02
4 76
12
0 23
2 78
.07
3.98
4 41
223
4.29
3
4
0 08
8 89
0
0.00
0 00
9
0.17
2 09
115
2.27
2 51
131
2.52
4
0
0 00
0.00
0
0.00
0.00
5
0.10
1 16
25
0 48
0.53
30
0 58
Total
45
0 87
21
0.40
431
8.30
4698
90 43
5195
100 00
Trequenc/ Missing = 234
Statistics for Table of Mode by Group
Stat 1sttc
DF
Value
Prob
Chi -Square 9
Likelihood Ratio Chi-Square 9
Mantel-Haenszel Chi -Square 1
PHI
Contingency Coefficient
Cramer's V
14 624
12 342
1.036
0 053
0 053
0 031
0.102
0 195
0 309
Effective Sample Size = 5,195
118
-------
Table B 33 1 he Frequency Distribution and Chi Square Test
of Homogeneity Results for the Percent of
Shipment Released (SHIPREL) by Mode of
1ranbpoitat ion for Gases
Mode
F requency
percent
(row PCT)
Air
barge
Rail
Truck
Total
1
12
0 66
92.31
5
0 27
83 33
1112
60 93
98 64
5/8
31 67
84 88
1707
93 53
2
1
0 Ob
7 69
1
0 05
16 67
8
0 44
0 71
37
2 03
5 43
47
2.58
Group
3
0
0 00
0 00
0
0 00
0 00
4
0 22
0.36
41
2.25
6 02
45
2.47
4
0
0.00
0.00
0
0.00
0.00
1
0 05
0 09
25
1.37
3.67
26
1 42
Total
13
0 71
6
0 33
1125
61.64
681
37.32
1825
100 00
Frequency Missing - 150
Statistics for Table of Mode by Group
Statistic
Chi-Sqjare
Likelihood Ratio Chi-Square
Mantel-Haens.?el Chi-Square
PHI
Contingency Coefficient
Cramer's V
OF Value
9 145 814
9 149.100
1 108 084
0 283
0 272
0 163
Prob
0 000
0 000
0 000
Erfective Sample Size -
119
-------
Table B 34. Analysis of Variance Results for the Percent of
Shipment Released (SHIPREL) ay the Mode of
Transportat ion for Each Physical State and Each
Commodity Class
Commodity class
Physical state (CMCL)
Liquid 2
4
6
8
9
20
25
95
Solid 10
30
35
60
DOT hazard class
ORM-A
ORM-B
ORM-C
ORM-D
ORM-E
Combustible Liquid
Flammable Liquid
Corros we Materia 1
Organic Peroxide
F lammable Sol id
Ox idizer
Poison B
P-value
5527
.0003
.8295
.2482
.8183
.0001
.0001
.0001
.2477
.0660
0377
0275
Gas
45
50
55
65
Nonflammable Compressed .0001
Gas
Flammable Compressed 0001
Gas
Poison A .2362
Irritating Material .4160
120
-------
classes and the Chi-Square test results are listed in Table B-35 and show
that the frequency distributions for each of the physical states are
significantly different. The percentage frequency histograms (percentage
bar charts) for the physical states are presented in Figures B-8
(liquids), B-9 (solids), and B-10 (gases).
The second Chi-Square test was performed on the percent of container
contents released for each commodity class separately to compare the
frequency distributions for the physical states. The results are
provided in Tables B-36, B-37, and B-38 and show that the frequency
distributions for the physical state within each class are significantly
different.
The third Chi-Square test was performed on the percent of container
contents released to compare the frequency distributions for modes of
transportation. The results, presented in Table B-39, show that the
frequency distributions for modes of transportation are significantly
different. The percentage histograms for each mode are listed in
Figures B-ll, B-12, B-13, and B-14.
The fourth Chi-Square test was performed on the percent of container
contents released for each commodity class separately to compare the
frequency distributions of modes of transportation. The results, which
are listed in Tables B-40, B-41, and B-42, show that the frequency
distributions for modes of transportation are different for liquid and
gas chemicals (P-values <.l), but are not significantly different for
solid chemicals (P>.1).
The fifth Chi-Square test was performed on the percent of container
contents released for each physical state separately to compare the
frequency distributions for modes of transportation. The P-values of the
tests show that for some of the physical states, the frequency
distributions for modes of transportation are significantly different
(cases with P-value <0.1).
B.9 Correlation Between Quantity Released and Shipment Size
Correlation measures the closeness of a linear relationship between
two variables. If one variable can be expressed as a linear function of
another variable, then the correlation i? J or -1, depending on whether
the two variables are directly or inversely related. A correlation of 0
between two variables means that each variable has no linear predictive
ability for the other. The correlation between two variables can be
estimated using the sample correlation coefficient. The sample
correlations presented in this analysis are known as "Pearson's
Correlation Coefficient."
121
-------
Table B-3_>. The Frequency Distribution and Chi-Squaie Tu:,t
of Homogeneity Results for the Percent ot
Container Contents Released (CONTREL) by Physical
State
Table of Class by Group ?
Clas,
Frequency
percent
(row PCD 1 2
Gas 1542 51
2.79 0.09
76.08 I 58
Liquid 31107 3431
56 26 6.20
64 95 7 16
Solid 336-1 427
6 08 0.7/
61 96 7 67
lotdl 36013 390H
65.13 7 07
Stat ist ics for
Stat ist ic
Chi-Square
Likelihood Ratio Chi -Square
Mantel-Haenszel Chi-Square
PHI
Contingency Coefficient
Cramer's V
Group 2
345 Total
55 32 295 19/5
0 10 0.06 0 53 3.57
2 78 1.62 14.94
2864 1288 9202 47892
5. 1» 2.33 16 64 86 61
5 9t< 2 69 19 21
35: 124 1162 5429
0 61 0.22 2 10 9.82
6 4J 2 28 21 40
3271 1444 10659 5^29o
59! 2 61 19.28 100 00
Table of Class by Group 2
Degrees of
freedom Probability
(OF) Value (P-value)
U 200 069 0 000
ii 225 655 0 000
l 69 386 0 000
0 060
0 060
0 043
Sample size = 55,29t
122
-------
PERCENT At
60 <
50
40
30
20 i
10 i
;E
EE
*****
*****
*****
*****
*****
*****
*****
*****
*****
> *****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** *****
***** ***** *****
***** ***** *****
***** ***** *****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** *****
***** *****
10
30 50
CONTREL MIDPOINT
70
90
Figure B-8. Percentage bar chart for the frequency distribution of the percent
of container released for liquids.
123
-------
PERCENTAGE
50
40
30
20 -i
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
**^I
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** *****
***** *****
***** *****
***** *****
***** *****
10 30
*****
XXX XX
*****
*****
*****
*****
*****
*****
*****
***** *****
***** *****
***** ***** *****
50 70 90
CONTREL MIDPOINT
Figure B-9. Percentage bar chart for the frequency distribution
of the percent of container released for solids.
124
-------
PERCENTAGE
70
60
50
30
20
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
70
10
30 50
CONTREL MIDPOINT
90
Figure B-10. Percentage bar chart for the frequency distribution
of the percent of container released for gases.
125
-------
Table B-36 The Frequency Distribution and Chi-Square Test
of Homoyeneity Results tor the Percent of Container
Contents Released (CONTREL) by Physical State (Liquid)
Physical State = Liquid
Table of Commodity Class by Group 2
Commodity
c lass
Frequency
percent
(row PCT)
2
4
6
8
9
20
25
95
Total
Group 2
1
156
0.33
45.61
39
0.08
65 00
18
0 04
54.55
7
0 01
22 58
179
0 37
72.47
1793
3 74
72 50
13645
28.49
68 25
15270
31 88
61 79
31107
64 95
2
29
0 06
8 48
7
0.01
11.67
3
0 01
9.09
1
0 00
3.23
21
0 04
8 50
171
0 36
6.91
1453
3 03
7 27
1746
3 65
7 07
3431
7 16
3
30
0 06
8.77
0
0 00
0.00
1
0.00
3 03
3
0.01
9 68
8
0 02
3.24
144
0.30
5.82
1206
2 52
6 03
1472
3 07
5 96
2864
^ 98
4
17
0.04
4.97
5
0 01
8.33
4
0 01
12 12
2
0 00
6 45
2
0.00
0.81
98
0 20
3 96
592
1 24
2 96
568
1.19
2.30
1288
2 69
5
110
0 23
32 16
9
0 02
15.00
7
0.01
21 21
18
0 04
58 06
37
0 08
14 98
367
0 56
10.80
3097
6 47
15 49
5657
11 81
22 89
9202
19 n
Total
342
0.71
60
0 13
33
0.07
31
0.06
247
0 52
2473
5 16
19993
41 75
24713
51 60
47892
100 00
126
-------
Table B-36 (continued)
Statistics for Table of Class by Group 2
Statist ic
Degrees of
freedom
(OF)
Value
Probabi1ity
(P-value)
Chi-Square
Likelihood Ratio Chi-Square
Mantel-Haenszel Chi-Square
PHI
Contingency Coefficient
Cramer's V
28
28
1
672.003
674.392
342.281
0.118
0.118
0 059
0.000
0.000
0.000
Sample size = 47,892.
127
-------
Table B-37. The Frequency Distribution and Chi-Square Test
of Homogeneity Results for the Percent of
Container Contents Released (CONTREL) by Physical
State (Solid)
Physical State = Solid
Table of Commodity Class by Group 2
Commodity
class
Frequency
percent
(row PCD
10
30
35
60
Total
Statistic
Chi-Square
L ikel ihood
1 2
149 44
2.74 0.81
44.21 13.06
343 18
6 3? 0.33
77.25 4 05
873 122
16 08 2.25
59 23 8.28
1999 243
36 82 4 48
62 98 7 66
3364 427
61 96 7.87
Statistics for
Rat 10 Chi-Square
Mantel-Haenszel Chi-Square
PHI
Contingency
Cramer's V
Coeff ic lent
Group 2
345
33 6 105
0.61 0.11 1.93
9 79 1.78 31.16
10 9 64
0 18 0.17 1.18
2.25 2 03 14 41
94 30 355
1 73 0 55 6 54
6.38 2.04 24 08
215 79 638
3 96 1 46 11 75
6 77 2 49 20 10
352 124 1162
6 48 2.28 21 40
Table of Class by Group 2
Degrees of
freedom
(DF) Value
12 108.888
12 112.189
1 12 133
0 142
0.140
0 082
Total
337
6.21
444
8.18
1474
27 15
3174
58 46
5429
100 00
Probabi lity
(P-value)
0 000
0.000
0.000
Sample size = 5,429
128
-------
Table B-38 The frequency Distribution and Chi-b ;uare Test
of Homogeneity Results for the Perce it of
Container Contents Released (CONTRFL) by Physical
State (Gds)
Physical State = Gas
Table of Comnodity Class by Group 2
Commodity
class
Frequency
percent
(row PCT)
45
50
55
65
Total
Group 2
i
644
32.61
75 59
854
43.24
80.49
20
1 01
74.07
24
1 22
68.57
1542
78.08
2
24
1 22
2.82
23
1.16
2.17
1
0.05
3.70
3
0 15
8.57
51
2.58
3
20
1 01
2 35
34
1 72
3.20
0
0 00
0.00
1
0 05
2.86
55
2 78
4
13
0.66
1.53
19
0.96
1.79
0
0 00
0.00
0
0 00
0 00
32
1.62
5
151
7.65
17 72
131
6.63
12.35
6
0.30
22.22
7
0 35
20.00
295
14.94
Total
852
43.14
1061
53 72
27
1.37
35
1 77
1975
100.00
Statistics for Table of Class by Group 2
Statistic
Chi-Square
Likelihood Ratio Chi-Square
Mantel-Haenszel Chi Square
PHI
Contingency Coefficient
Cramer's V
Degrees of
freedom
(DF)
12
12
1
Value
21.813
21 614
1 750
0.105
0.105
0 061
Probabi 1 i ty
(P-value)
0.040
0 042
0 186
Sample size = 1,975
129
-------
Fable B-39 he Frequency Distribution and Chi-Square Ttst
of Homogeneity Results for the Percent of
Container Contents Released (CONTREL) by Mode of
Transportation
Table of Mode by Group 2
Mode
Frequency
percent
(row PCT)
Air
Barge
Rail
Truck
Total
Statistic
Chi Square
1 2
425 43
0.77 0 08
66.20 6.70
76 14
0.14 0.03
63.93 11.48
5793 146
10.48 0.26
87 52 2.21
29717 3706
53 74 6 70
62 02 7 73
36013 3909
65 13 7.07
Statistics for
Likelihood Ratio Cm-Square
Mante 1-Haenszel Chi-Square
PHI
Cont ingency
Cramer's V
Coeff ic lent
Group 2
345
56 16 102
0 10 0 03 0.18
8.72 2.49 15 89
6 5 19
0.01 0.01 0.03
4.92 4 10 15.57
107 62 511
0.19 0 11 0 93
1 62 0 94 7 72
3102 1361 10027
5 61 2 46 18 13
6 47 2.84 20 93
3271 1444 10659
5.92 2.61 19 28
Table of Mode by Group 2
Degrees of
freedom
(DF) Value
12 1689 . 04
12 1961. ' 51
1 682.. -48
0 175
0 172
0.101
Total
642
1.16
122
0.22
6619
11.97
47913
86 65
55296
100.00
Probatn 1 ity
(P-value)
0 000
0.000
0 000
Sample size = 55,296.
13Q
-------
PERCENTAGE
40
35
30
25
20
15
10
5
•
h
•
*****
*****
*****
*****
*****
*****
*****
h *****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
M**M»
*****
*****
*****
*****
MMM M*
*****
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
30
*****
*****
*****
*wl*»
*****
*****
**t*t
MUMMM
*****
*****
*****
*****
*****
MMMMM
*****
*****
MMMMM
*****
*****
*****
*****
*****
*****
*****
*****
*****
MMKMM
*****
*****
***** *****
***** *****
***** *****
***** *****
***** *****
***** *****
***** ***** *****
***** ***** *****
***** ***** *****
***** ***** *****
***** ***** *****
50 70 90
CONTREL MIDPOINT
Figure B-ll. Percentage bar chart for the frequency distribution of the percent
of container released for the air mode of transportation.
131
-------
PERCENTAGE
60 + *****
50
40
30
20
10
JHUWWJ
sMWMWt
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** *****
***** *****
***** *****
***** ***** *****
***** ***** ***** *****
***** ***** ***** *****
*****
*****
*****
*****
*****
*****
*****
*****
*****
10
30 50
CONTREL MIDPOINT
70
90
Figure B-12. Percentage bar chart for the frequency distribution of the oercent
of container released for the water mode of transportation.
132
-------
PERCENTAGE
80
70
60
50
30
20
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
30 50
CONTREL MIDPOIHT
*****
*****
*****
10
70
90
Figure B-13. Percentage bar chart for the frequency distribution of the percent
of container released for the rail mode of transportation.
133
-------
PERCENTAGE
50
30
20
10
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
***** *****
***** *****
***** ***** *****
***** ***** *****
***** ***** *****
*****
*****'
„„„...»
Jill MUM
*****
Hit H )Ht
*****
***** *****
***** *****
10
30 50
CONTREL MIDPOINT
70
90
Figure B-14. Percentage bar cfart for the frequency distribution of the percent
of container released for the highway mode of transportation.
134
-------
Table B-40 The Frequency Distribution and Cht-Square Test
of Homogeneity Results for the Percent of Container
Contents Released (CONTRE1) by Mode of Transportation
for Liquids
Physical State = Liquid
Table of Mode by Group 2
Mode
Group 2
Frequency
percent
(row PCT)
Air
Barge
Rai 1
Truck
Total
0
66
0
64
1
384
80
.55
60
13
.52
4370
9
87
12
80
26293
54
62
90
.24
31107
64
96
2
39
0 08
6.76
10
0 02
10 75
111
0.23
2 23
3271
6 83
7 /4
1-431
7 16
0
9
0
5
0
1
3
54
11
36
5
01
38
87
.18
75
2718
C
J
6
68
.43
2864
5
.98
4
0
2.
0.
4.
0
1.
12
2
2
12
2
15
03
60
4
01
30
51
11
02
18
54
88
88
69
0
14
0
15
0
7
5
85
18
73
14
.03
05
358
75
.19
8/45
18
20
.26
70
9202
10
21
Total
577
1.20
93
0 19
4977
10 39
42245
88 21
47892
100 00
Statistics for Table of Mode by Group 2
Statistic
Chi-Square
Likelihood Ratio Chi-Square
Mantel -Haenszel Chi-Square
PHI
Contingency Coefficient
Cramer's V
Degrees of
freedom
(DF)
12
12
1
Value
1301.456
1514 345
513 124
0 165
0.163
O.OH5
Probabi 1 ity
(P-value)
0 000
0.000
0.000
Sample size = 47,892
135
-------
fable B-41. Thi. Frequency Distribution and Chi Square lest
of Homogeneity Results for the Percent of Container
Contents Released (CONTREI.) by Mode of Transportation
for Solids
Physical State = Solids
Table of Mode by Group 2
Mode
Frequency
percent
(row PCT) 1 2
Air 31 3
0 57 0.06
65 96 6 38
Barge 14 3
0 26 0.06
66 67 14 29
Rail 312 27
5.75 0 50
66.24 5 73
Truck 3007 394
55 39 7 26
61 49 8 06
Total 3364 427
61.96 87
Statistics for
Stat ist ic
Chi~Square
Likelihood Ratio Chi-bquare
Mantel-Haenszel Chi-Squure
PHI
Contingency Coefficient
Cramer's V
Group 2
3
2
0.04
4.26
1
0 02
4.76
16
0 29
3 40
333
6 13
6 81
352
6.48
Table of Mode
Degrees of
freedom
(DF)
12
12
1
4 5 Total
1 10 47
0 02 0 18 0 87
2 13 21 28
1 2 21
0.02 0 04 0 39
4 76 9 52
10 106 471
0.18 1 95 8.68
2.12 22 51
112 1044 4890
2.06 19 23 90 07
2.29 Zl 35
124 1162 5429
2.28 21.40 100.00
by Group 2
Probahi 1 1 ty
Value (P-value)
16 409 0 173
16.196 0 110
0.875 0 34rf
0 055
0 055
0 032
Sample size = 5,429
136
-------
Table B-42 The Frequency Distiibution and Chi Squjre Te: t
of Homogeneity ResiIts for the Percent of Container
Contents Released (CONTREL) by Mode of Transportation
for Gases
Phybical State = Gas
Table of Mode by Group 2
Mode
Frequency
percent
(row PCT) 1 2
Air 10 1
0.51 0.05
55 56 5.56
Barge 4 1
0.20 0.05
50.00 12.50
Rail 1111 8
56.25 0.41
94.88 0 68
Truck 417 41
21.11 ? OS
53.60 5.27
Total 1542 51
78.08 2.58
Statistics for
Stat lit ic
Chi-Square
Likelihood Ratio Chi-Square
Mantel-Haenszel Chi-bquare
PHI
Contingency Coefficient
Cramer's V
Group 2
345
0 0 7
0.00 0.00 0 35
0.00 0.00 38.89
0 0 3
0 00 0.00 0.15
0 00 0.00 37 50
4 1 47
0.20 0 05 2.38
0 34 0.09 4 01
51 31 238
2.58 1.57 12 05
6.56 3 98 30 59
55 32 295
2.78 1.62 14 94
Table of Mode by Group 2
Degrees of
freedom
(DF) Value
12 486 446
12 507.943
1 270.767
0 496
0 445
0 267
Total
18
0.91
8
0.41
1171
59.29
778
39.39
1975
100.00
Probabi 1 i ty
(P-value)
0 000
0 000
0.000
Sample size = 1,975
137
-------
The correlation discussed in this study is between the quantity
released and the shipment size. The correlation coefficients for the
data, classified by physical states and commodity classes, are presented
in Table B-43.
The correlation coefficients presented in Table B-43 showed ,
positive correlation between the shipment size and the quantity
released. These coefficients are relatively high for commodity classes
2, 6, and 50. A statistical test of the significance of the correlation
coefficients in Table B-43 was not performed because the assumption that
the data were normally distributed was not valid.
B.10 Conclusion
The results presented in this report indicate that the percent of
shipment released (SHIPREL) has different statistical characteristics for
each of the three physical states (liquid, solid, and gas), for each mode
of transportation, and for each commodity (DOT hazard) class. The
analysis of variance test results show that the means of the percent of
shipment released (SHIPREL) for the different physical states are
significantly different for each factor considered (e.g., DOT hazard
class). The Chi-Square test of homogeneity results shows that the
frequency distributions of percent of shipment released (SHIPREL) for the
various levels of each factor are significantly different.
The analysis of variance method and the Chi-Square technique were
also performed on the fraction of container released and show that the
means of the percent of container contents released (CONTREL) for the
different levels of each factor are significantly different. The
analyses also showed that the frequency distributions of the percent of
container released (CONTREL) for the various levels of each factor differ
significantly as well.
Other factors that were not considered in this study but which could
be investigated in the future are the type of container used, the
distance traveled, and the location of the incident. The interaction of
some of tne factors and regression analysis of the quantity of chemical
released on the distance traveled should reveal relationships among the
various factors.
138
-------
Table B 43 Correlation Coefficient Bttvveen Quantity
Released and Shipment Size Classified by
Physical States and Commodity Class
Physical state Commodity class Correlation coefficient
(CMCL)
Liquid 2 .782
4 .244
6 .999
8 .231
9 .206
20 .196
25 .163
95 186
Solid 10 031
30 .186
35 .049
60 259
Gas 45 082
50 .543
55 .150
65 019
139
-------
This appendix presents three alternative methods for estimating
average distances over which chemicals are shipped during distribution in
commerce. These methods rely on data on shipping patterns in the 1977
Bureau of the Census Commodity Transportation Survey (CTS) publications,
including the following:
• Commodity Transportation Survey Summary (USDOC 1981a);
• Commodity Transportation Survey, Geographic Area Series (USDOC
1981b); and
• Commodity Transportation Survey, Commodity Series (USDOC 1981c).
Information on the locations of chemical manufacturers is also
useful. One source of such information is the Stanford Research
Institute Directory of Chemical Producers, published annually (see SRI
1987).
The appendix is divided into three sections. Sec ion C.I describes
the steps common to all methods, as well as general information on use of
the CTS publications. Section C.2 presents criteria or selection of a
method. Section C.3 describes each method in detail.
C. 1 Steps Common to All Methods
Although each method differs in the type of source information used,
four steps are common to all methods. The general steps are as fallows:
• Identify the CTS commodity code most closely related to the
specific chemical for which shipping data are required.
• Identify the geographic origin of shipments.
• Locate values for tons and ton-miles shipped for the selected
commodity code (STCC) and geographic specificity in the CTS
publications.
• Calculate the average shipping distances of the chemical for each
mode of transportation.
Each of these steps and the sources of information used to compete them
are described below.
C.I.I Identify the CTS Commodity Code
Commodities included in the CTS publications are classified using the
Commodity Classification for Transportation Statistics (TCC) codes. The
system of numbering within the TCC codes closely parallels that of the
142
-------
Standard Transportation Commodity Code (STCC 1972, USDOC 1981a).
Therefore, for the purposes of this method, the data on commodity
shipments in the CTS will be searched by first matching the STCC code
obtained in Step 1 of the general method (Section 3.1 of this report)
with the most closely related TCC code listed in tables of the CTS
publications; the more digits in the TCC code, the more specific the
commodity classification.
For example, the STCC code for malathion is 2879978 (STCC 1972). The
available TCC commodity codes to match this STCC in the CTS publications
are:
28 Chemicals and Allied Products
287 Agricultural Chemicals
28799 Agricultural Chemicals, NEC (USDOC 1981a).
The best match would be TCC 28799. However, as will be explained in the
following discussion, sufficient data on commodity shipments are not
always available at the greatest level of specificity.
C.I.2 Identify the Geographic Origin of Shipment
For some chemicals, it will be possible to identify the location(s)
of manufacture. This information can be used to obtain data from the CTS
publications that closely correspond to the actual shipping patterns.
Areas of origin and destination of commodities vary with respect to
geographic level of detail depending upon the CTS report used. Data in
the CTS Summary (USDOC 1981a) and the CTS Commodity Series (USDOC 1981c)
are summarized for the entire United States. In the CTS Geographic Area
Series (1981b), data are presented by state of origin and by production
area of origin.
Geographic levels of detail included in the CTS, in order from least
to most detailed, are census division, state, and production area.
Census divisions of the United States include New England, Middle
Atlantic, East North Central, West North Central, South Atlantic, East
South Central, West South Central, Mountain, and Pacific. Production
areas consist of large Standard Metropolitan Statistical Areas (SMSAs) or
clusters of SMSAs that represent a single geographic industrial unit
having 900 or more manufacturing establishments. Forty-nine SMSAs are
defined in the 1977 CTS publications (USDOC 1981a). Table C-l presents
production areas by census division, with descriptions of SMSAs included
in each production area.
An up-to-date source of information on plant locations of
manufacturers of specific chemicals is the SRI Directory of Chemical
Producers (see SRI 1987).
143
-------
/1S1H
Table C-i . 1977 Coirmodity Transportation Survey
Production Area Descriptions by Division
(Standard metropolitan statistical araes included in each production arta)
NEW ENGLAND
PA 1-1 (Previously PA 1) Bonon. Worcester. Providenoe-
Wenvlek-Piwiucket. Brockton. Lawranca-Heverhill,
Lowell
PA 1-2 (Pravioutly PA 2) Hartford, Ntw Britain, Meriden.
Watarbury. New Haven-Wen Haven, aridgaport.
Springfietd-CNcopee-Holyoke
MIDDLE ATLANTIC
PA 2-1 (Previously PA 3) New York. Nassau-Suffolk.
Norwalk, Stamford •
PA 2-2 (Previously PA 4) Newark, Jersey City. Pittanon-
Cllfton-Passaic. New Brunswick-Perth Amboy-
Sayrevilla, Long BrandvAtbury Pirk
PA 2-3 (Previously PA 6} Philadelphia. Wilmington. Trtnton
PA 2-4 (Previously PA 8) Hamsburg. Lancaster. York
PA 2-6 (Prevtously PA 7) Allentown-Sethlehem-Easton.
Reading
PA 2-4 (Pr»viou»ly Market Area 311 Northeast Pennsylvania.
Blnghamton. Elmira
PA 2-7 (Previously PA 9) Syncute. Utlca-flome. Albany
Schenectady-Troy
PA 28 (Previously PA 101 Buffalo, Rochester
PA 2-9 (Previously PA 12) Pittsburgh. s'teubenville-Weirton,
Wheeling
EAST NORTH CENTRAL
PA 3-1 (Previously PA 11) Cleveland, Akron. Canton, Loram-
Elyna, Youngnown- Warren, En a
PA 3-2 (Previously Market Area 34) Columbus. Springfield
PA 3-3 (Previoutly PA 14) Cincinnati. Dayton. Hamilton-
Middletovm
PA 3-4 (Previously PA 13) Detroit, Flint. Toledo. Ann Arbor
PA 3-6 LantinfEaat Unung, Kalamazoo-Portage, Jaduon.
Bartia Creek
PA 34 (Previoutlv Market Area 36) Grand Rapids. Muskagon-
Norton Shorea-Muskegon Heio)in
PA 3-7 (Previously PA 20) Indianapolis. Anderton, Muncie
PA 3-8 (Previouary PA 16) Chicago, Gary-Hammond-Eut
Chicago
PA 3-9 (Previously PA 16) Milwaukee. Kanosha. Racine
WEST NORTH CENTRAL
PA 4-1 (Previously PA 18) Si Loon
PA 4-2 (Previously PA 27) Kamaa City. Lawrence. St.
Joseph, Topeka
PA 4-3 (Previoutly PA 17) Mlnneacolis-SL Paul
SOUTH ATLANTIC
PA 5-1 (Previously PA 6) Baltimore
PA 5-2 (Previously Market Area 32) Washington. O.C. Md.-
Va.
SOUTH ATLANTIC-Contmuod
PA S3 (Previously Market Aree 33) Norfolk-Virginia Beech-
Portsmouth, Newport Newt-Hampton, Petersourg-
Colonial Helghts-Hopevwll, Richmond
PA 5-4 Greeniboro-Wlmton Salam-High Point, Surlir gton,
Raleigh-Ourham
PA 5-6 Charlotte-Gastonia, Greenville-Spertanburg
PA 5-6 (Previously PA 19) Atlanta
PA 5-7 (Previously Market Area 43) Davtona Beach.,
Melbourne-Titusville-Cocoa, Orlando, Lakeland-
Winter Haven, Tampa-St. Petersburg
PA 5-3 (Previously Market Area 41) Miami, Ft. Laudardale-
Hollywood, Wen Palm Beech-Boca Raton
EAST SOUTH CENTRAL
PA 8-1 (Previously Market Area 37) Louisville
PA 3-2 (Previously Market Area 38) Nashville-Davidson.
Clartuvitle-Hopkimville
PA 8-3 (Previously Market Area 39) Memphis
PA 6-4 (Previously Market Area 42) Birmingham, Tusealoosa,
Anniston, Gadsdan
WEST SOUTH CENTRAL
PA 7-1 (Previously Market Areas 44 and 45) Baton Rouge,
New Orleans. Biloxi-Gulfport, Piscagoula-Moss Point.
Mobile. Pensacola
PA 7-2 (Previously PA 21) Houston, Beaumont-Port Arthur-
Orange, Galveston-TtMas City
PA 7-3 (Previously Market Area 49) Austin, San Antonio
PA 7-4 (Previously PA 20) Dallas-Fort Worth
PA 7-5 (Previously Market Area 48) Tulsa, Oklahoma City
MOUNTAIN
PA 3-1 (Previously PA 22) Oenvar-aoulder, Colorado Springs
PA 3-2 (Previously Merket Area 50) Salt Lake City-Ogden.
Provo-Omm
PA 8-3 (Prwiouily Market Area 51) Phoenix. Tucson
PACIFIC
PA 9-1 (Previously PA 23) Seartle-cverert. Tacoma
PA 9-2 (Previously Market Area 52) Portland. Salem
PA 9-3 (Previously PA 241 Sen Francisco-Oakland. Vallejo-
Fairheld-Napa. San Jose, Santa Rosa. Santa Cruz
PA 3-4 (Previously Market Area 53) Sacramento. Stockton,
Modesto
PA 9-5 (Previously PA 25) Los Angeles-Long Seacn,
Anaheim-Santa Ana-Garden Grove, Rivenide-San
Bernardino-Ontario, Oxnard-Simi Valley-Ventura
PA 9-6 (Previously Market Area 55) Sui Oiego
Source USDOC 1981a.
144
-------
C.I.3 Locate Values for Tons and Ton-Miles Shipped in CTS Publications
Once the commodity type and required geographic specificity of the
search are established, specific values for tons and ton-miles shipped
must be located in the appropriate tables of the CTS reports. The source
of this information differs with each method. However, a concern common
to the three methods is that the values obtained for tons and ton-miles
shipped must be significant. These data are reported in thousands of
tons shipped and millions of ton-miles shipped (USDOC 1981a). If the
reported quantity for tons or ton-miles for a specific combination of TCC
code and geographic specificity is less than one-half of the unit of
measure (i.e., less than 500 tons shipped or less than 500,000 ton-miles),
those values are considered insignificant. In such cases, a more general
TCC code (one with fewer digits, see Section C.I.I) or a larger
geographic area should be selected.
C.I.4 Calculate the Average Shipping Distance of the Chemical
For each mode of transportation, the average shipping distance is
computed by dividing the value obtained from the CTS tables for ton-miles
shipped by the value for tons shipped:
Average distance shipped (in miles) =
ton-miles shipped
tons shipped
For shipments by truck, a weighted average shipping distance can be
calculated from the values obtained for the two major truck categories
included in the CTS reports, that is, motor carriers (ICC and non-ICC)
and private truck. In order to calculate the average shipping distance
for trucks, multiply the average shipping distance (ton-miles/tons
shipped) for each truck category by the fraction of the total quantity
shipped by truck that is represented by that category. The sum of the
products for the two categories is the weighted average shipping distance
for truck shipments.
C.2
Selecting a Method
Of the three methods presented in
select the method appropriate for the
shipment of the chemical, as follows:
this appendix, the reader should
level of information available on
Method
C-l
C-2
C-3
Average
quantity/
shipment
Unknown
Unknown
Known
Origin of
shipments
Unknown
Known
Unknown
Destination of
shipments
Unknown
Unknown
Unknown
145
-------
The more detailed the available information, the more accurately the
average shipping distance can be estimated.
C.3 Descriptions of the Methods
This section of the appendix describes each of the methods available
for estimating shipping distance of a chemical.
C.3.1 Method C-l. Estimation of the average shipping distance of a
chemical when the average quantity per shipment, the origin of
shipment, and destination of shipments are all unknown. The
average shipping distance is derived in 4 steps, as follows:
(a) Match the STCC code for the chemical (determined in Step 1 of
the general method, Section 3.1 of this report) with the most
closely related TCC code available in Table 2 of the CTS Summary
(USDOC 1981a).
(b) The origin of shipments is unknown, and therefore the shipping
data in the CTS Summary (USDOC 1981a) are used.
(c) For each mode of transportation, identify from Table 2 of the
CTS Summary the values for tons shipped (Table 2, Column B) and
ton-miles (Table 2, Column C).
(d) Divide the value for ton-miles for each mode of transporcation
by the corresponding value for tons shipped. The quotient is
the average shipping distance of the chemical by that mode of
transportation.
C.3.2 Method C-2. Estimation of the average shipping distance of a
chemical when the origin is known but the average quantity per
shipment and the destination of shipments are unknown.
If manufacture of a chemical is restricted to a particular geographic
region of the country, the Commodity Transportation Survey, Geographic
Area Series (USDOC 19815), can be used as a source of information for
estimating the average shipment distance of the chemical. The steps of
the method are as follows:
(a) Match the STCC code for the chemical (determined in Step 1 of
the general method, Section 3.1 of this report) with the most
closely related TCC code availa51e in Ta51e 1 of the Commodity
Transportation Survey, Geographic Area Series (USDOC 19815).
(b) Identify the manufacturing location using the PRODUCTS section
of the SRI Directory of Chemical Producers (see SRI 1986).
Identify the most specific geographic area of origin available
in the CTS Geographic Area Series (USDOC 19815) that corresponds
to the manufacturing location.
146
-------
(c) For the geographic area and TCC code selected, identify in
Table 1 of the CIS Geographic Area Series the values for tons
shipped and ton-miles shipped. Ascertain that these values are
significant (see Section C.I.3).
(d) For each mode of transportation, divide the value for ton-miles
shipped by the corresponding value for tons shipped. The
quotient is the average shipping distance of the chemical for
that mode of transportation.
C.3.3 Method C-3. Estimation of the average shipping distance of a
chemical when the average quantity per shipment is known but the
origin and destination of shipments are unknown.
This method allows greater specificity in the calculation of average
shipping distance by using data available for specific weight intervals
of commodity shipments that are presented in the CTS Commodity Series
(USDOC 1981c). It does not require information on the origin of
shipments.
(a) Match the STCC code for the chemical (determined in Step 1 of
the general method, Section 3.1 of this report) with the most
closely related TCC code available in Table 3 of the CTS
Commodity Series (USDOC 1981c).
(b) The origin of shipments is unknown; therefore, the U.S. summary
data in the CTS Commodity Series (USDOC 1981c) are used.
(c) Identify the average quantity per shipment determined in Step I
of the general method, Section 3.1 of this report. Then, locate
a corresponding weight interval of shipments listed for the
selected TCC commodity code in Table 7 of the CTS Commodity
Series. For each mode of transportation, locate the values for
tons shipped and ton-miles shipped for that weight interval.
(d) Divide the value for ton-miles shipped by the corresponding
value for tons shipped for each mode of transportation. The
quotient is the average shipping distance of the chemical for
the selected weight interval and mode of transportation.
147
-------
502X2 -101
REPORT DOCUMENTATION
PAGE
l._REPORT NO.
EPA 560/5-85-009
3. Recipient's Accession No.
4 Title and Subtitle
Methods for Estimating Releases of Chemical Substances
Resulting from Transportation Accidents-Volume 9: Methods
for Assessing Exposure to Chemical Substances
5. Report Date
December 1987
7. Author(s)
Julie Gartseff, U.C. Crenshaw, Patricia Jennings
8. Performing Organization Rept. No.
9 Performing Organization Name and Address
Versar Inc.
6850 Versar Center
Springfield, VA 22151
10. Project/Task/Work Unit No.
Task 100
11. Contract(C) or Grant(Q) No.
10 68-02-4254
(G)
12. Sponsoring Organization Name and Address
United States Environmental Protection Agency
Office of Toxic Substances
Exposure Evaluation Division
Washington. DC 20460
13. Type of Report & Period Covered
Final Report
14.
15. Supplementary Notes
The EPA Project Officer was Elizabeth Bryan; EPA Task Manager was Greg Schweer.
1C. Abstract (Limit: 200 words)
This report oresents methods for calculating expected annual releases of
manufactured chemicals resulting from transportation accidents. The scope of the report
is limited to releases en route rather than leaks and other releases at transportation
terminals. A step-by-step method of calculating annual quantity released per mode of
transportation is presented, and sources and limitations of the supporting data are
discussed in detail. This method is suitable for comparing estimates of annual releases
of several chemicals or for comparing releases by various modes of transportation for
one chemical.
A statistical analysis of the Department of Transportation (DOT) HAZMAT data base is
included as an appendix to the report. The analysis focuses on differences in the
expected fraction of shipment released or fraction of container released based on mode
of transportation and type of chemical.
17. Document Analysis a. Descriptors
b. Identifiers/Open-Ended Terms
Annual Releases/Manufactured Chemicals
Transportation Accidents
c. COSATI Field/Group
18. Availability Statement
Distribution Unlimited
19. Security Class (This Report)
Unclassified
20. Security Clan (This Page)
Unclassified
21. No. of Pages
148
22. Price
(See ANSI-Z39.18)
See Instructions on Reverse
149
OPTIONAL FORM 272 (4-77)
(Formerly NTIS-35)
Department of Commerce
------- |