DETERMINATION OF ATMOSPHERIC DILUTION
      FOR EMERGENCY PREPAREDNESS
  A JOINT EPA-DOE TECHNICAL WORKSHOP
                OCTOBER 15-17, 1986
                     Chaired by
    Dr. Francis S. Binkowski, U.S. Environmental Protection Agency
          Dr. Harry Moses, U.S. Department of Energy
                     Edited by
                 Ms. Sharron E. Rogers
                Proceedings Summaries
                    Prepared by


         Research and Evaluation Associates, Inc.

-------
      U.S. ENVIRONMENTAL PROTECTION AGENCY
   ATMOSPHERIC SCIENCES RESEARCH LABORATORY
PEER REVIEW AND WORKSHOP MANAGEMENT SERVICES
                Contract Number 68-02-4129
                      Project Officer
                    Ronald K. Patterson
                       Prepared by
           Research and Evaluation Associates, Inc.
                1030 15th Street, N.W., Suite 750
                   Washington, D.C. 20005
                      (202) 842-2200
                 100 Europa Drive, Suite 590
                   Chapel Hill, N.C. 27514
                      (919)968-4961

-------
                                 DISCLAIMER
     Although the  workshop  results described in this  report  were funded by
the U.S.  Environmental  Protection  Agency through contract Number 68-02-4129
with  Research  and  Evaluation  Associates,  Inc.,  this  report has  not  been
subjected to Agency peer  review.   It does not necessarily reflect the views
of the  Environmental  Protection Agency  or  the Department of Energy,  which
co-chaired  the  workshop,  and  no  official  endorsement should  be inferred.
However, it should be noted that the individual presentations and the papers
included  in the appendices  have been peer  reviewed  by  each contributor's
organization and are reproduced as received.
                                        ii

-------
                                  FOREWORD
     This   report   summarizes   the  Joint   EPA-DOE  Technical   Workshop
Determination'  of  Atmospheric Dilution  for Emergency Preparedness,  held in
Research Triangle  Park,  NC, at  the  Meredith Guest House on  October 15 and
16, 1986.  The individual  contributions  to this document  were prepared from
verbatim transcripts  of the oral  presentations made at the  workshop.   The
transcribed  presentations  included  in  this  report  were  reviewed  by  the
authors  and  their  organizations.    Several   articles  were  prepared  for
presentation at the workshop,  and these  have been included  as appendices in
this document.  In the  opinion of the chairpersons,  the detail contained in
these appendices was important to the overall objectives of the Workshop.

     We would  like to thank the chairpersons and all  who worked to put these
proceedings  together.   Our special thanks go  to  the editor of this report,
Ms. Sharron  Rogers at Research and Evaluation Associates,  Inc.

Ronald K. Patterson
Project Officer
Atmospheric  Sciences Research  Laboratory
U.S. Environmental Protection Agency
                                        iii

-------
                                 ABSTRACT
     The Joint  EPA-DOE Technical  Workshop  on Determination  of Atmospheric
Dilution for Emergency Preparedness was held  in  Research  Triangle Park,  NC,
in October  of 1986.   The  objectives  of  this workshop  were to  review  the
current  methods  of  determining   atmospheric  dilution  for  use  in  hazard
identification, emergency  preparedness  planning,  and  emergency  response, to
provide  recommendations  for choosing among these methods,  and,  finally, to
define  the  role  of the meteorologist  in hazard  identification,  emergency
planning, etc.  Several invited papers were presented, and panel discussions
were held to meet the objectives.  The results are presented and discussed.
                                        IV

-------
                                  CONTENTS
FOREWORD ................................ "H
ABSTRACT ...............................  1V

INTRODUCTION   ............ .................   1

TECHNICAL OVERVIEW
     R. A. Cox .............................   b
THE BHOPAL GAS TRAGEDY
     K. Shankar  Rao
NATIONAL  RESEARCH  NEEDS  FOR  EMERGENCY RESPONSE
    IN THE WAKE  OF  CHERNOBYL
     Joseph  Knox  ............................  Zb

ACCIDENTAL RELEASE SCENARIOS FOR ANALYSIS
     Jane Crum  Bare ...........................  37

SHORT-TERM TOXIC  RELEASES FROM  CHEMICAL
    MANUFACTURING  SITES
     Robert  E.  Rosensteel  .......................  *7
 FATE OF TOXIC RELEASES IN THE ATMOSPHERE-ATMOSPHERIC
    RELEASE ADVISORY CAPABILITY (ARAC)
      Marvin H. Dickerson ........................   53

 SOURCE STRENGTH MODELING
      Jerry M. Schroy   .........................   59

 FUTURE NEEDS FOR DISPERSION MODELS IN  HAZARD EVALUATION,
    EMERGENCY PREPAREDNESS, AND ACCIDENT PREVENTION
      James L. Makris   .........................   65

 ISSUES IN REGULATORY APPLICATIONS OF MODELS
      David E. Lay! and  .........................   71

 COMMUNITY NEEDS FOR HAZARD EVALUATION  TOOLS
      Fred Millar ............................   77

 MATHEMATICAL MODELS FOR ATMOSPHERIC DISPERSION
    OF HAZARDOUS CHEMICAL GAS RELEASES:  AN OVERVIEW
      Jerry A. Havens  ..........................   °'

 DENSE GAS DISPERSION MODELS
      Donald L. Ermak  ..........................   95

-------
                                 CONTENTS (Cont.)
ON THE DEVELOPMENT OF REACTIVE, DENSE GAS MODELS
     Bruce B. Hicks and Will R. Pendergrass	   105
DISPERSION MODELS FOR NEUTRALLY BUOYANT
   AND POSITIVELY BUOYANT GASES
     Thomas E. Pierce	107
LAB-SCALE EXPERIMENTS
     Robert N. Meroney	113
LARGE-SCALE EXPERIMENTS OF THE DOE LIQUEFIED
   FUELS PROGRAM
     Ronald P. Koopman	121
FLUID MODELING OF DENSE GAS DISPERSION OVER A RAMP
     William H. Snyder .  . . .	129
ATMOSPHERIC SCIENCE AND EMERGENCY RESPONSE AT THE
   SAVANNAH RIVER LABORATORY
     Allen H. Weber and R. W. Benjamin	137
EMERGENCY PREPAREDNESS AND RESPONSE  IN THE
   U.S. AIR FORCE
     Captain Lawrence E.  Key	141
POISONOUS GASES FROM LAKES: THE CAMEROON DISASTER
     Daniel A. Livingstone and George Kling  	 151
SUMMARY OF PANEL DISCUSSIONS  	 161
APPENDIX A - The Bhopal Gas Tragedy	A-l
APPENDIX B - On the Development of Reactive, Dense Gas Models	B-l
APPENDIX C - Mathematical Models for Atmospheric Dispersion of
                Hazardous Chemical Gas Releases:  An Overview	C-l
APPENDIX D - Workshop Agenda 	 D-l
APPENDIX E - Workshop Participants   	 E-l
                                        vi

-------

-------
                      KEY TO PHOTOCOLLAGE OF SPEAKERS
Upper row:   (left) Jane Crum Bare, (right) Jerry Schroy.

Middle row:  (left) Tony Cox, (center) K. Shankar Rao, (right) Marvin
             Dickerson.

Lower row:   (left) James Makris,  (right) Robert Rosensteel.
                                       viii

-------
                              INTRODUCTION

     This workshop  was  designed as  a  response to  a  request from U.S.
Environmental  Protection  Agency (EPA)  Headquarters for  a  forum where
the  current  state  of  modeling of  atmospheric  dilution  for  use  in
emergency preparedness  could  be discussed.   It  quickly became obvious
that a joint effort between the EPA and the  U.S.  Department of Energy
(DOE)  was  the most  effective  way  to proceed.   As  the  designated
representatives  of  the  two  agencies,  we  built   an   agenda  for  the
workshop that would meet the following objectives:
     1.   To  review the  current  methods  of  determining the  release
          characteristics,  source  strength,  and rate  of dilution  of
          atmospheric contaminants for use in hazard identification and
          evaluation, emergency preparedness, and  emergency response,
          and to assess the specific strengths and weaknesses for these
          methods and make recommendations for their improvement.
    2.    To  provide recommendations  for choosing among the  current
          modeling  methods  for  immediate  use  in  the identification and
          evaluation  of  potential  hazards,  in  the  preparation  of
          emergency  preparedness  scenarios,  and  in  actual  emergency
          response  situations.
    3.    To define the role of the meteorologist in hazard evaluation,
          emergency preparedness planning, and emergency response.
    The  participants contributed  to these  broad   objectives  from the
perspective  of  their  individual  situations  and  experience.    One
fascinating  aspect of  the  workshop  was the  interaction  between those
who had dealt with  releases of  radioactive material and those who dealt
with  chemical  releases.   The  radioactive releases  have historically
involved  concerns  of   large-scale contamination,  while the  chemical
releases  have involved concerns  of a  more  local   nature.   In view of
this disparity  of  scale, it was surprising to see  how  quickly a common
                                   1

-------
view of  the problems  associated with  potential  and actual  hazardous
releases developed.  One  strong  point of difference was,  however,  the
attitude toward  a source term.   For radioactive  releases,  monitoring
data taken  after the incident will  often  give a good  estimate  of the
source  strength  and  character.    This  was  shown  for the  Chernobyl
accident.
     For  hazardous chemical releases,  similar monitoring data  seldom
exist,   and  sufficient  time   is   seldom  available  to   make  such
measurements.   Only  in cases where routine monitoring  is  done on site
is  this approach  possible  for  chemical  releases.     This  difference
comes  from the  fact that the  hazards of  radioactivity were perceived
many years  ago and appropriate instrumentation developed and installed.
It  is  only recently  that  the  hazards of toxic  chemicals  have been
widely  recognized.   General  agreement has  not been reached on the type
of  on-site  measurements necessary, nor  in some cases, the need for such
measurements.
     Because   this   report   consists  of   transcriptions   of  oral
presentations,  no bibliographic  citations  are given unless provided by
the author.  To obtain reference information  not provided, contact the
author.
     We wish  at  this time  to  express  our  appreciation to  all the
participants  for  their  contributions  and  their  patience   in  the
preparation of  this document.   We  wish  especially to  thank  Sharron
Rogers,  Charlotte Coley,  Brenda White,  Rebecca  Peer,  Laura  Saeger,
Chanya Harris,  and Linda Cooper  of Research  and  Evaluation  Associates,
Inc.,  who were  responsible  for  the  production of  the  workshop  and  for

-------
technical writing,  editing,  and  production of this document.   We also
wish  to  thank Don  Cox  of  Research  and  Evaluation  Associates,  who
designed the cover; Mannie Currin, who made the verbatim transcript and
had to  wrestle with an  unfamiliar  technical  jargon; and  the staff of
the Meredith Guest House for their help and hospitality.
Francis S. Binkowski
Environmental Protection Agency
Harry Moses
Department of Energy

-------
                          TECHNICAL OVERVIEW
                               R. A.  Cox
                            Technica, Ltd.
                            London, England

     In  this technical  overview  for what  I  consider  to  be a  very
important meeting, we need to cover a whole range of topics.  What is a
source?  What is the source of a release of hazardous material into the
environment, and  how  is  it characterized?  How  do  the characteristics
of sources affect the way that the hazardous material is diluted in the
atmosphere?. What are the key issues for research today regarding those
characteristics and the  behavior of hazardous materials?   To what use
do  we  put  this information  when  we have  that  scientific knowledge?
What are we  really  aiming  to  do in the  real world of decision making--
both in regulatory decision making and the sort of decision making that
industry  must  do  in  facility  design  and  siting?    Specifically,
regarding  emergency  response planning,  we  have to  address how our
knowledge of atmospheric dispersion phenomena can be used in helping us
make decisions prior to an emergency and during one.

Examples of Real Sources

     First, what  is a source?   I  will  be  predominantly talking  about
hazardous  chemicals,  although  others will  be  addressing  radioactive
chemicals in this conference as well.  The major difference between the
two  generically  is  that  with  hazardous  chemicals  one typically  is

-------
concerned  only with  relatively  short-term effects,  while radioactive
chemicals typically require both short- and long-term concern.
     A  source that  is  very  typical  of hazardous  releases  into  the
atmosphere,  although  an  extraordinary one in  terms of  magnitude,  is
illustrated  by the rupture of a cross-country  ammonia  pipeline caused
by  a  large  bulldozer   striking,  penetrating,   and  shearing off  the
pipeline.  An  enormous cloud formed. In fact,  a cloud of ammonia vapor,
ammonia  liquid droplets, and condensed  water  vapor was  formed  in  the
atmosphere.  This  source  is not anything like what those of us who work
in  the  field of air   pollution  dispersion  modeling  are  used  to
considering.   The  differences  include:  1)  it  is massive;  2) the  cloud
emerged  under high pressure in a jet-like  stream;  3)  rapid mixing was
induced  near the source,  particularly at the  start  of  the event; and
4)  finally,  the  cloud becomes  denser than  air.   As the cloud sinks and
starts  to  spread  laterally,  the  emission momentum is  lost and  the
effect  of  the wind starts to take  hold  on  the  behavior of the  plume.
We  must  always  consider how complicated  the  phenomena  are near  a
source.  They  can  be  very complicated.
     A  second example  is that of  a puncture in  a  chlorine cylinder.
This  is  the small  end of the  range of events we are considering here.
Consider that  a very  small hole occurs in a 1-ton cylinder, so that the
amount of inventory released is not very large.   The release  occurs as
a  jet-like  plume  with  a  lot  of  initial  mixing.    What  was  the
orientation  of the release?   It  could have  been horizontal * vertically
downwards,   or vertically upwards.    The  hole might  have been  in  the
vapor space of the cylinder, or it  might have  been in the  liquid space.

-------
In contrast,;.the  hole  in  the  cylinder could have been large and, thus,
the releaselof the contents of the cylinder could have been essentially
instantaneous.
     When  the  brakes failed on  a  Mexican train comprised  of chlorine
tank cars,  the cars derailed  at approximately 80  kilometers  per hour
and five or six of the cars were ruptured in the resulting pileup.  The
total  amount  of chlorine released was  about 100 tons.   Compared with
the  previous  two cases  of holes  in  vessels,  this  is an  extremely
complicated  source.    One of  the  tank  cars  was torn  apart, and  the
release of its contents was instantaneous.  While most of the tank cars
were caught in  the  pileup, the head  of one car  failed and the  car
rocketed away  from the scene, spilling  liquid chlorine  under pressure
behind  it.   Trying to  describe this  source of  dispersion, we  must
consider both  instantaneous and elongated releases.
     When  a pipe  failed  in  1974  in  the Flixborough  plant in England, a
flammable  substance, cyclohexane, was released  under pressure of about
10 bars.   The  release  occurred as  two opposing high-pressure jets with
flash  vaporizing  and  aerosol  formation;  in  other  words,  a  very
complicated thing from the  standpoint of estimating rates of dilution.
The result of  this particular failure was total destruction of a rather
large chemical complex.

Characterization  of Sources

     Obviously, there  are a far wider  range of  types of releases than
these examples cover.  A source parameter that  is  really important to

-------
us  is  the amount  of initial  dilution that  occurs  during the  quasi -
spherical expansion phase of  the burst.   In  the  chemical  industry,  we
are dealing  typically with materials  that  are stored  under  pressure,
thus,   having a  vapor  pressure  considerably  higher than  atmospheric
pressure.   These materials will  flash vaporize,  and  there will  be a
simultaneous  process   of  flash  vaporization  and  mixing  with  the
atmosphere.  The phenomena we can expect  to  arise are:   liquid droplet
formation, aerosol  formation,  liquid droplet coalescence,  and rainout
of liquid droplets.  The liquid droplets,  being rather  cold relative to
the ambient  environment, may  hit the  ground and  revaporize or  stay
there as  a  liquid  pool. These phenomena  are  extremely  complicated and
have an  important  effect on the  quantity of material  that ends  up  in
the air  and, therefore,  on the  range of  hazard faced  from such  an
event.
     Transportation   container   bursts  are   obviously   of  a  similar
character.  Another type of rapid, sudden  release is the possibility of
run-away reactions that can lead to bursting of reactor vessels.
     These are  all  relatively large events.  When we  look at smaller-
scale events, one  can list  such  events as:   breakage of pipework, pump
seal leakage, discharges from relief valves, and so on.  These are also
more complex phenomena  that result  in  material  being released into the
atmosphere without any  pressure behind it, such as boil-off from liquid
pools or seepage through the ground or through water.
     From  a  real  process  plant, the  releases  can take on many forms
that seem  almost endless.   We must  describe which of the parameters of
these releases  are of particular importance  to us.   Most important is

                                   8

-------
the quantity of the substance that ends up in the atmosphere.  When you
have a typical chemical release, not all of the chemical ends up in the
atmosphere, and there  may  be  many  factors  that limit the quantity that
actually does enter the  atmosphere.  "These factors include orientation
of the leak and the presence of any secondary containment.
     To consider  the  effect of orientation, if a  liquid were released
that  is  flash vaporizing,  but the orientation  is  vertically downward,
it is possible that the flashing two-phase flow will hit the ground, be
deflected, and  a  portion  of  the  liquid centrifuged out.   That liquid
will  cool  the ground  locally  beneath  the  release point  and  a  pool  of
liquid will  likely  form.   This situation  is even.more  likely in'those
cases where secondary  containment  exists.   Conditions such  as this are
typical of situations where  the  drain valve  of  a pressure  vessel  is
somehow  stuck open.   I submit  to you  that  the  processes  of  liquid
fallout from  a jet of  this sort are  not well  understood at the present
time, and that they are one of the most important  subjects of research.
People  are  currently  planning  and  deliberately  designing  secondary
containment in order to have precisely the effect  described above.
     The quantity of chemical  released into the atmosphere also depends
on  the  amount of  flashing,that is occurring,  and how  high  the vapor
pressure  of  the  material  in  storage  is  relative  to  the  atmospheric
pressure.   If it  is  very high, relatively more  of the material  will
vaporize  and  more will end up in  the  atmosphere.   Flash vaporization
and the  process, pressure  generally both  have  an  influence  on  what is
called initial dilution, either by inducing mixing directly through the
turbulence associated  with the initial velocity or by  the  mixing that

-------
 is  associated with  the expansion  occurring with  flash vaporization.
 Another  important modifying  factor in  actual  chemical  plants  is the
 presence  of  obstacles  and  their  effects   on  initial  dilution.   An
 example would be a leak of a  fully refrigerated liquefied gas spilled
 into  a bounded area; the presence  of  an adjacent large tank structure
 and  the  effects  of the wake of  that structure  can induce a very  large
 degree of initial dilution  of the  vapor boiling off  from the  liquid
 pool.
     All  of  these factors influencing the  amount  of chemical  released
 into  the  atmosphere  are  extremely  complicated.    In  many  cases, the
 effect of  these  source-related  characteristics  on  the  subsequent
 dispersion of the  cloud is very  marked,  particularly the effect of the
 initial pressure.  However, this is perhaps more true of flammable than
 of toxic  substances,  particularly  those  toxic materials that are still
 hazardous  at  concentrations  of  the  order  of  only a  few parts per
 million.

 Dispersion Modeling

     In cases where  releases  occur with  pressure,  dispersion  models
 that  fail  to consider the  pressure  will  seriously overestimate the
 travel distances of  the discharges.   In  cases where  the release occurs
without or with  almost  no pressure,  dispersion models,  such as  typical
dense  gas dispersion  models  or even Gaussian dispersion  models  may be
quite relevant to these events.
                                  10

-------
     Please note that the vast majority  of hazardous materials are, in
fact, in pressurized containment.  Therefore, in order for the momentum
of  such  releases  to be  reduced,  one must  postulate  some  continuous
mechanism for  abstracting the  kinetic  energy from the initial release.
Processes like  seepage through the  ground  or bubbling up through water
are the sorts of processes that one has to postulate in order to ignore
the  effects  of  the initial  momentum  when  the  original   source  is
pressurized.
     In  the  area  of  subsequent  dispersion,   those  of  us  who  are
meteorologists  know  that the subsequent  dispersion is highly dependent
on  the weather conditions and  on factors such as surface roughness and
topography.   Topography  is  particularly  important because  a typical
cloud  is  initially  extremely dense  relative  to the  atmosphere  and,
therefore,  particularly  strongly affected by  topographic  effects.   In
those  situations  with an absence of topographic  effects,  such as flat
plains,  the state of the art  in  dense vapor cloud dispersion modeling
is  very well advanced and does not merit very much additional research.
Research areas  that  do  require additional  attention are the effects of
the  following:    topographic  effects  on  dense  vapor clouds, obstacles
other than  topographic ones, flash vaporization, and rainout.

Hazard Analysis

     We  need  to.look for a  moment  at the uses  of hazard  analysis and
the ways in which  the decisions we must make depend on our knowledge of
the consequences  of  hazardous  chemical releases.   In practice, we find

                                   11

-------
that there are only three uses for hazard analysis for which there is a
real need by real decision makers in the real  world:
     •    Land use planning/site selection,
     •    Plant design, and
     t    Emergency response.
     The  first,  and  most basic,  is  in zoning  and land  use  planning
activities  such  as  selecting a site for a  hazardous  installation.  In
such a  case,  it is obvious  that we are  dealing  with  a hazard  that has
the potential,  if the worst were to happen,  to  have  a physical  impact
over a distance of perhaps several hundred meters.  Thus, siting of the
facility  should  be very carefully considered because  of the potential
risks.    The  siting  decision  must   involve  a  balance  between  the
magnitude  of  the consequences that might occur  and the probability of
the event occurring.   In practice, we find that it is  very difficult to
site typical  industries  far  enough  away from  vulnerable populations or
other vulnerable  entities to  eliminate  the hazard  to  the  people.  In
any case,  the employees  of  the facility will be exposed  to the risk.
Thus, whatever we do about  land  use  planning  and site  selection,  we
must accept some residual risk.
     There  is  an  analogous argument with respect to  plant  design (the
second  of the three  uses  of hazard analysis).    Industries  do  have an
intense  interest  in  designing  their  facilities  to  minimize  major
hazards.   They do  not want to  lose  their  facilities  or  injure their
employees.  They  do  this  not only by  site  selection,  but also by plant
layout  and  by  choice of  process  routes,  level  of  instrumentation
detail,   emergency shutdown  valves,  check  valves,  and other  items of

                                   12

-------
this  nature,   which  are  really  a  matter  of design  detail.   Hazard
analysis  techniques can  be,  and  are being,  used  to  address  these
detailed  matters  related  to  major  hazards,   but   implementing  such
controls for major  hazards  may cost money.   One can conceive of design
features  that  are  impractical   simply  because  they  are  ludicrously
expensive.   Those  things are  not  done  if it  can  be shown  that  the
probability  of the  particular event  occurring  for which  that design
provision was  being considered was  sufficiently low.  Thus,  there is
some acceptance of residual risk.
     The third application of hazard analysis techniques,  for emergency
responses,  is  the  principal  topic  of our  workshop.    I  would  like to
make several points about the way I see emergency response fitting into
the  hazard  analysis framework.   First,  there is one  crucial  decision
that  has to be made,  how large  a  physical  provision  are we  going to
make  for emergency response?   How  much foam are we  going  to provide?
How much  provision  needs  to be made for breathing apparatus and escape
masks?   What will be provided  to the  local  population in anticipation
of  an  emergency situation?  This type of decision  must be essentially
risk-based.    It  definitely  is  impractical  to  provide  a  sufficient
emergency  response capability  to prevent all casualties in  the worst
possible  accident one  can postulate.   Therefore,  we  make provisions
that  fall  short of the worst case,  and the question is:  how far do we
go?    We make  the decision  on  a  risk  basis,  discounting  the worst
scenarios  provided  that  we  can  demonstrate   they   are sufficiently
improbable.
                                   13

-------
     For  effective  emergency response,  we need  good  preplanning,  by
evaluation  of a  representative  range  of  scenarios using  dispersion
models  and  tools  of  that  nature.  By  representative,  I mean  that we
must look at everything from the small  leaks to the most extreme cases.
Even  though  we may  not  be able  to cater  to  the extreme  cases,  they
should  be evaluated  so that deployment  of  existing response capability
will  be the  most  effective possible.   In  preplanning for  emergency
response, the  decision-making algorithm  is  particularly difficult.   It
is not  a  simple thing to decide in  advance whether the  advice given to
local populations should be to  evacuate  or  to  stay indoors.   Depending
on  the  duration  of  exposure and  on the particular  chemical  released,
the  advice might be  different in  different  conditions.   Such  decisions
can  be  very  difficult, but  can  be improved  by  the prior application of
very good models for  consequence evaluation.
     Whatever  emergency  plans   we  select must  be  extremely  practical.
The  decision-making  algorithms  have to  be presented to  the people  who
will give the  orders during the real emergency in such  a way  that they
do not  hesitate about the decisions they must make.

Concluding Remarks -  Future Priorities

     In the  recent past, there  has  been  some misdirection  of  resources
by concentrating too much on the niceties of vapor cloud dispersion  in
the  atmosphere.  There are  in fact  many more phenomena  that are really
involved  in  chemical  hazards  than simply  the atmospheric dispersion
                                   14

-------
effects.   In  the future,  the balance of effort  and  interest  should be
moving towards these other phenomena.
                                   15

-------
                        THE BHOPAL GAS TRAGEDY
                            K. Shankar Rao
            National  Oceanic and  Atmospheric Administration
      Atmospheric Turbulence & Diffusion Division, Oak Ridge, TN
                                  and
                       M. P. Singh and S. Ghosh
                    Center for Atmospheric  Sciences
            Indian Institute of Technology, New Delhi, India
     In the  early  hours of December  3,  1984,  about 40 tons  of  highly
toxic,  volatile,  and reactive  methyl isocyanate  (MIC)  gas  leaked  in
about 90 minutes through a 33-meter-high  atmospheric vent  into the cool
night air of Bhopal  and quickly  spread in  a  fog-like lethal  cloud over
a  large populated  area.    Thus  began  the  world's  worst  industrial
disaster, which killed over 2000 people  and  injured  more  than 200,000.
Here we will sketch the accident scenario and outline the  events  of the
night leading to this catastrophe.
     The meteorological and  topographical  features  of Bhopal and  the
physical,  chemical,  and   toxicological   properties  of  MIC  will   be
described in  the context  of  the cloud  dispersion  and its  effects  on
human  and   biological  life.    A  simple  analytical  dispersion  model
emphasizing  aqueous-phase conversion  and  deposition  of  MIC will  be
presented.      This   model,  based  on   solution  of  the  atmospheric
advection-diffusion  equation,  gives  estimates  of  ground-level  and
vertically   integrated   concentrations.      The  model   estimates
qualitatively correlate with recorded human fatalities and injuries and
the contours of observed  damage  effects  on trees and vegetation in the
affected^areas.
                                   17

-------
     The rehabilitation  efforts and  some preliminary  results of  the
ongoing medical and toxicological studies will be  summarized.   We will
comment briefly on the legacy of Bhopal,  including its implications for
siting  and  operation  of  plants  handling  toxic  chemicals,  worker
training,  risk  assessment  and emergency  preparedness,  occupational
health  and   environmental   regulations   and   enforcement,   and  other
relevant issues.
     The city of Bhopal,  with  a population  of  about 900,000,  is  the
capital of the State  of  Madhya  Pradesh,  India's  largest  state, with an
agriculture-based   economy.    Bhopal's    central   location   (about
600  kilometers  south of  New  Delhi),  communications,   resources,  and
hospitality  attracted  industry  and  people   from all  parts  of  the
country.
     Union  Carbide,   India,  Limited  (UCIL),  which is the  corporation
involved in this accident, is one of the top 20 Indian companies.  They
produce  about  1500  tons  per   year  of  MIC-based  pesticides,  Sevin
(carbaryl) and Temik  (aldicarb).
     The  UCIL  pesticide plant  is   located   in  the  northern  part  of
Bhopal.   To  the east  of  the   plant  is an  industrial  and  warehouse
district.  To  the north  are  mostly  wooded lands  that  are very sparsely
populated.   However,  the plant  is  surrounded by  residential  areas on
three  sides.    Unfortunately,   on  the  night of  the   accident  winds
directed the plume toward the populated part of the city.
     Bhopal  includes  hills  to  a height  of  about 600  meters  and  two
lakes.   It  is likely that  nocturnal  drainage winds and  land breezes
altered the  local  surface wind patterns  on the  night of the accident.

                                  18

-------
Detailed  meteorological  data  are  not  available,  but  various  sources
indicate  that  the winds that  night  were initially  from  the northwest
and subsequently from the north.
     There  are  three MIC storage  tanks on the UCIL  property.   MIC is
the raw material  used  to produce  the pesticides carbaryl  and aldicarb.
At the time of  the release,  tank  610 contained about 41 metric tons of
MIC.  A shift change occurred about 10:45 p.m. on the night of December
2, 1984,  which  was a Sunday.   There  were about 75 people on duty.  The
incoming  shift  was unaware  that a  runaway chemical  reaction was taking
place in  tank 610.
     About  11:00  p.m.,  the  staff noticed  the pressure  in tank 610
increasing  from  3  to   10  pounds  per  square  inch.    They mistakenly
attributed  the  increase to nitrogen pressurization  of  the  tank by the
previous  shift.   About 11:30 p.m., the staff noted eye irritation from
MIC.     Since   minor  leaks  of  MIC  were  not  uncommon,   they  were
unconcerned.   About 12:00 a.m., they  noticed  the  pressure  in tank 610
had  increased  to 30 pounds per square  inch.   This  was reported to the
production  supervisor, who checked it almost immediately and found that
the  tank's  rupture disk (designed to rupture  at 40 pounds per square
inch) had burst,  and the safety valve had popped.
     At  1:00 a.m., MIC  was observed  to be  escaping through a 33-meter
high atmospheric   vent line  into  Bhopal's  cool  night air.  Within a
90-minute period,  about 40  tons  of gaseous  MIC  leaked out.   None of
several  plant  safety systems in place  fully  functioned that night.  A
scrubber,   charged  with  caustic   soda,   contained  only   enough  to
neutralize  about  7 tons of  MIC released.  A  flare  tower did not work

                                   19

-------
because it  was undergoing maintenance.   A  water curtain designed  to
hydrolyze escaping MIC  reached only a  height  of 15 meters,  far below
the release point.  The facility's refrigeration plant was down.
     The public siren was turned on for a few minutes at 1:00 a.m., but
was  not  sounded  again  until   2:00  a.m., to  alert  the  public  to  the
danger.  By then, thousands of people were fleeing.
     About 200  people  died in their beds in the  immediate vicinity of
the  plant where a huge white cloud formed and  moved southward towards
the densely populated areas.   Near the plant, the gas was so thick that
visibility  was  very  low.   These  people  awoke to  a  nightmare  of
suffocation, burning eyes, and panic.  They rushed out into the streets
to  join  thousands of  other  people who  were running for  their lives.
Many  people became  overcome  by  fumes  and  collapsed  along  the  way.
Rescuers converged on the  area around  1:30 a.m.   By  dusk of that first
day,  the  death  toll  mounted  to  about 1000  and   over  100,000  were
injured.   As  the magnitude  of the disaster became  apparent, medical
equipment and  relief were  rushed to the stricken city.
     MIC is  known to  be  a very reactive, toxic,  and  volatile chemical.
It  undergoes  exothermic  and  vigorous  reactions with  a variety  of
compounds  containing  active  hydrogen  atoms.   Analysis  of sludge from
tank 610 after the accident  showed that two  major exothermic reactions
took  place.    One was  MIC reacting with  water; the  products  of MIC
hydrolysis  with water made up 16 to 28  percent  of  the  samples.   The
second,  and  much  faster  reaction,   was  MIC  reacting  with   itself,
catalyzed by iron.  More than 50 percent of the sludge consisted of MIC
trimer resulting  from this second  reaction.   A  small  amount of water
                                   20

-------
with a  trace of contaminants, such as  rust  or salt, could catalyze an
autoreaction.   The  heat  liberated  during an  induction period lasting a
few hours  could generate a reaction of explosive  violence.   It is not
inconceivable  that  a small trace  contamination  could have occurred in
the MIC tank in spite of the best  prevention efforts.
     Once gaseous MIC escaped into the  air, it would  undergo advection,
diffusion, and  chemical transformation, with atmospheric moisture.  The
major  reaction product  would be  methyl amine,  which is  absorbed  and
stays in the soil  until  slowly  broken  down by biological reactions and
weathering.
     We  used an  analytical  dispersion model  based  on  the atmospheric
advection-diffusion  equation.   The source strength  was established at
40  metric  tons in  about  90 minutes.    MIC  vapor is  twice  as  heavy as
air.   This  suggests that density  effects  might  be important for plume
dispersion.   However,  the  vapor  escaped at a very high velocity, which
leads  to  considerable  entrainment.    Also,  the  reaction of  MIC with
moisture  releases about  1.36  x 10^ joules  per kilogram  of heat that
reduces  the  plume  density.   With these considerations,  the density
effects  were not considered, but gravitational spreading was  included.
Post-episodic  reconstruction of the accident by scientists established
that  the  wind was  initially northwesterly  during   the  accident,  and
subsequently northerly.
     According  to our model, most of  the high concentrations occurred
in  the first  kilometer southeast of   the plant.    A region  extending
about 7  kilometers  downwind of  the plant was affected.   Birds, cattle,
cats, dogs,  goats,  and all  other animals were killed over a 65-square-

                                   21

-------
kilometer area  in  the southeast  quadrant  of the plant.   Based on the
model estimates,  we have  divided the affected  area into  four zones.
Zone  1   consists   of  concentrations  >50  parts   per million,  zone  2
consists of  concentrations >15 parts  per  million,  zone  3,  >1.5 parts
per million,  and  zone 4, <1 part per  million.   In  the  first few hours
after the accident,  zone 1 (severely  affected)  out  of  a population of
6173, sustained 360 fatalities.   Initial fatalities  in  zone 2  (with a
much larger  population)  amounted  to 508.   In  zone  3,  there were about
11  initial  fatalities,  and in  zone 4, about 5.  The  correlation with
the fatalities provides some indirect evidence that these concentration
estimates are reasonable,  although the  large  population movement that
night  makes  these  comparisons  somewhat   subjective.    Most  of  the
fatalities,  however, occurred close to their place of residence.
     Perhaps  the   best  concentration  monitors for  the  effects  of the
accident  at  Bhopal  are the plants  and  vegetation  in  the area.   The
Indian  Agricultural  Research  Institute  and  the   Central  Board  for
Control  of  Pollution conducted a vegetation damage  survey that showed
most of the  effects  occurred  in  the  southeast,  southern,  and eastern
directions  from  UCIL's boundaries.   The  same  species  were  affected
differently  depending on location relative to  the plant and the amount
of  moisture  available  in  the   ground  at  the  time  of  exposure.
Vegetation  that had been  irrigated on the previous day was much less
affected.   Some  tall  trees were  completely defoliated.   Much of the
vegetation  in  affected areas  looked as  though it had  been burned.
Additional detailed studies are still  under way  to study  the impacts on
                                   22

-------
the vegetation.  Studies are addressing the histological effects on the
lipids of the leaves and on the microorganisms attached to the foliage.
     Very little is known about MIC's toxicological effects, especially
the  long-term ones.   What  little  data  are available  were  from tests
done  on  rats.   The permissible  MIC  skin exposure  limit,  a voluntary
standard,  is  0.02  parts  per million.   This is  the amount of  MIC to
which  a  person  can be safely  exposed  over  an  8-hour workday.   This
threshold limit  value  (TLV)  has  been  set by the American Conference of
Governmental  Industrial  Hygienists  (ACGIH).    The  lethal  dose  is
estimated  to  be between 15 and  30  parts  per million,  depending on age
and  sex.   Before the  Bhopal  episode,  the effects of MIC had never been
observed  on  such  a large  and  diverse  population.    Isocyanates  were
known  to attack  the respiratory  system, eyes, and skin.
      Laboratory  analyses  have confirmed  that most of the deaths in the
immediate  vicinity  of  the  plant  occurred from  cyanide  poisoning.
Cyanide  was  probably  released because of the high temperatures reached
in  the MIC  tank,  causing the MIC  to break up-into hydrogen  cyanide,
methyl amine,  carbon monoxide, and other organic cyanide products.
      While  it  will  be  some time before  the   studies  at  Bhopal  are
completed, clinical  and toxicological  studies are being coordinated by
the   Indian   Council  of  Medical  Research.    Medical  treatment  has
dramatically  improved the condition of people with eye injuries.  Lung
problems  persist  in  many  victims.   A  large  number  of  victims  with
pulmonary  edema have recovered,  but  many were  left with considerable
damage to  their small  airways  and alveoli.   Children  were the worst
affected,  with  about 39  percent showing moderate to  severe pulmonary

                                   23

-------
disability.  Miscarriages and stillbirth were  frequent in the pregnant
women exposed  to the  gases.   In the  first 20 weeks,  436 spontaneous
abortions occurred out of 2600 pregnancies;  the  normal  rate for Bhopal
was 6 to 10 percent.
     On the night of the accident,  the  doctors of Bhopal, supported by
medical students,  interns,  and  the nursing  staffs,  rallied to respond
to  this  unprecedented, sudden disaster.   Because there  was  so little
known  about  the toxicology  of MIC  and  its  reaction  products,  the
medical personnel  had  to  improvise in the  beginning.   There certainly
was  and is  a real  need  for information  on the  processes,  potential
hazards, biological  effects,  and the medical treatments that should be
made available in any  kind of industry involving toxic chemicals.
     The Union Carbide plant was  located  close  to the  population in
Bhopal.   There  should be  at least a  2-  to  3-kilometer  buffer  zone
between  hazardous  industries and  the  population  centers.   Standards,
enforcement, safety, health, and environment are very important.
     Bhopal  offers  many  lessons  that have to  do with basic  human
concerns for  safety.   If these lessons are  lost,  it is all  too likely
that an accident like  this can happen again.

Note:   A  prepared  paper  by the  authors  on this  topic appears  in
Appendix A.
                                  24

-------
            NATIONAL  RESEARCH  NEEDS  FOR  EMERGENCY  RESPONSE
                        IN THE  WAKE OF  CHERNOBYL
                              Joseph Knox
                 Lawrence Livermore National  Laboratory
                             Livermore, CA
     The material  presented today is a distillation  of materials from
about 3  feet  of shelf space.  We  will  consider  what  actually happened
at Chernobyl,  the  nature of the warning that  the  free world received,
and how  that  warning  came.   The  first tasks that then faced the United
States were  assessment  and  making sense out  of the  inconsistent  and
sometimes conflicting information.
     First,  let  us  briefly  review  the  setting  of  the  Chernobyl
accident.     The  reactor,  in  operation approximately  2 years,  was  a
1000-megawatts  (electric) generating plant with about  3200 megawatts of
thermal  energy.  The  reactor is graphite moderated,  water  cooled,  and
has confinement, but not containment in the United States sense of this
term.    The  reactor  core  is  in  an  interior building  with what  the
Russians  call  a   protective shield  around  it,   built of  steel  and
surrounded by an  inert  gas.   Thus,  there is or was  a barrier between
the  oxygen in  the confinement  building  and  the  material   inside  the
reactor.
     The  accident   sequence has  been  covered  in detail   in  various
publications  and  will   not  be  repeated  here.    During  the  turbine
experiments and the  shutting down of the safety  devices,  there was an
excursion and  a gradually  increasing  instability  of the reactor with  a
large steam  pressure buildup  inside  the interior structure.   In the

                                   25

-------
interior of the reactor, inside of the protective shield, a pressure of
over  1000  pounds  per  square  inch  ripped  off  the  tops  of  some
1650  pressure  tubes  exposing  the graphite  and  fuel  elements  to the
atmosphere.   Shortly  thereafter,  the  graphite, which  was  at  750°C,
ignited.   As a  result of the  preceding explosions, the  plumbing and
instrumentation were probably disrupted.  At that time,  water and steam
could  readily  contact  the zirconium,  and water  and steam  could  also
contact  the hot  graphite,  generating  large amounts  of  hydrogen and
carbon monoxide, both combustible.  The source term at Chernobyl  can be
described  as  a chaotic fire, multiple  explosions, gas  generation, and
various  sources  of  water contacting the  hot graphite.   In  terms  of
initial  distribution  of the  released  inventory,   we   must think  of
burning  gases,  burning graphite, burning carbon  monoxide,  and perhaps
multiple  explosions,  driving  a  plume  up  through  the  atmospheric
boundary  layer.     This is an   "image" of  Chernobyl  at  1:23 a.m.  on
[Saturday,] April 26, 1986.
     Our first warning  came from Scandinavia  late Sunday and on  Monday
morning.   We learned by  telephone  that a suite  of  radionuclides  were
being observed on  particulate  filters  in Scandinavia; radioiodines and
radiocesiums were  present in copious  amounts.    Other  isotopes  (e.g.,
  Co) were  coming  from  neutron activation of  the  cooling system, which
confirmed a disruption and at least partial  destruction  of this system.
The  particles  of  ruthenium  were spherical,  suggesting'that  at least
part  of  the fuel  released  had  seen temperatures of 2500°C.   (Later
interpretations indicate that the Ru particles came from shattered fuel
elements  entrained  into  the plume  or  explosively  injected  into the
                                  26

-------
 along the  route of  cloud  passage,  and  an  estimate of  the curies  by
 isotope that were released into the atmosphere was developed.   Nuclear
 engineers from  both  the  civil  and military communities made estimates
 of the core inventories and the relevant  nuclides of greatest interest.
 At  the   time  of  reactor  shutdown,  80  million  curies  of     I were
 estimated in the core inventory with 6 million curies of     Cs.   During
 the first  24  hours,  it is estimated that all of the  noble gases were
 released when the 1650 pressure tubes  were destroyed.   A high fraction
 of the volatiles were  released.   Our  assumption  is that 40 percent of
 the volatiles were released on the first day into the Zone A cloud and
 10 percent  into  that  zone on each  of the  subsequent 5 days.  The  amount
 of material  going  into  Zones A  and B of the cloud over the 6-day  period
 was developed from experience  based  on  fallout clouds.   However, these
 estimates were checked  by cumulus cloud simulative models that normally
 use a radius of 5  kilometers  for  the  scale of the base of the  cloud.
 The Chernobyl fire  plume was  approximately  20  meters  in  radius—the
 world's  smallest  convective  column; it  was driven  by a  flux   of  62
 megawatts across the  surface of the 20-meter radius.
     Over 100 maps  of material  dispersed  and  deposited  over Europe and
 Asia  were generated  in the  simulations.   One of these maps shows the
 integrated exposures  to an individual standing in  the path of the plume
 at  2  days (48 hours) after the  accident.  The highest  calculated dose
was  133  rem to  the  adult  thyroid  in the  vicinity  of the  plant.  The
exposures at  the distance of the   coast of Sweden were equivalent  to
0.1 rem at the end of 48 hours.   From these simulations, it was  learned
how Kiev  was  fortunate, indeed.   The  plume from the accident rose  up

                                  28

-------
over a  relatively dry warm  front,  so there was not  much  of a problem
from  rainout   of  radioiodine.     There   was,   however,   significant
convective activity in the area of Kiev at the time, although the plume
was  not  significantly  impacted  by  that  activity.    Observations  in
Sweden were significantly affected by the lifting of the plume over the
warm frontal area in the region of the Baltic Sea.
     The  results  of the LLNL simulations were sent  to the President's
Commission  through  the  Department  of  Energy.    These results  are  in
relatively  good agreement  with the  calculations  and  observations  of
Scandinavia  and Eastern and  Southern  Europe,  that is, by a factor of
two with  measured values for deposition  by  isotope.   Agreement was not
good  in Western Europe, possibly due to the gliding  up over the warm
front  (later  calculations  show  better  agreement when meteorological
data  from  all  of  Western  Europe  were  used).   The  calculations  and
measurements showed that there  were no acute health  effects outside of
the Soviet Union.
     To  put  Chernobyl  in perspective,  the  accident was compared  to  a
20-kiloton  nuclear  test and  the  released inventory  of the Three-Mile
Island  (TMI) reactor accident in  the following table.   The  estimated
       COMPARISON* OF  CHERNOBYL WITH OTHER RADIONUCLIDE RELEASES
Nuclide
13l!
137Cs
9°Sr
Noble gases
Chernobyl
10-50
1-6
.001-. 07
100-200
TMI
.00002
None detected
None detected
10
20-kiloton
Nuclear Test
2
.004
.004
5
*Megacuries of radioactivity 3 days after shutdown or test.
                                   29

-------
release  inventories  for Chernobyl  indicate  that radioiodine  releases
were approximately 20 times greater than at a  20-kiloton  nuclear test.
No cesium  was  detected at TMI;  it  stayed  inside the containment.   No
90Sr was detected at TMI, but was at Chernobyl.  While virtually all of
the noble  gases were released at  TMI  and Chernobyl,  the amount was far
smaller  at TMI.   The  differences  in  long-term dose  occur  because the
dose at  TMI  came  mainly  from  the  noble  gases,  while  the long-term dose
at Chernobyl comes mainly  from the  137Cs.  The maximum individual dose
at  TMI  was  11.5  millirems,  while  the maximum  dose at  Chernobyl  was
above  the  lethal dose for half the exposed  population (LD50), delivered
by fission products, not the noble gases.   Rainout was not important at
TMI because  noble gases are not  scavenged.  At  Chernobyl,  the iodines
were scavenged and are of continued importance to the Soviets.   The hot
spots  in Europe  were created by  rainout.  These  rainout  hot spots run
roughly  10  times  above areas subject  to  dry deposition alone.   The
message  from these  findings  is  that  containment,  design,  and safety
devices  are  immensely  important  in  terms of  explaining the differences
between  Chernobyl  and  TMI.  All   of the diagnostic work shown  today in
this  presentation was  completed  in  the  first  14  days  following the
event  and  reported to the President's Commission.
     The New York  Times  reported  that Chernobyl  released much more
radioactivity than all of  the atmospheric  testing by both of the super
powers.   It is  important  that we  understand  that this  was an error.
The correct  number is that Chernobyl's release of 137Cs is 6 percent of
the    Cs released from super power testing.
                                   30

-------
      In  preparing to model the  upper  cloud from Chernobyl, we  quickly
learned  that we  did  not  need  the whole radioiodine  inventory to  explain
the  readings  in  Scandinavia.   So the  question was, where  was  the  other
half of  the  radioiodine?   It  would  have  been  easy to  say  that  the
models  were no  good,  but  this  was not  the reason.    When the United
States'  sophisticated sampling  aircraft measured  radioiodine at  5500
meters  above  Japan,  we quickly redirected our focus.   Taking  the  long-
range transport  models from nuclear test days and back calculating  from
Japan,   we  found  that  half  the  radioiodine   had  to  be  at  higher
elevations  over  Chernobyl  on the first  day.   Calculating  forward  from
Japan,   it  was  possible to  estimate   how  much  radioiodine  would be
measured off  the coast  of  California  when  the  cloud  began its passage
over the United States.   The models  estimated  that  30  picocuries per
cubic meter would  be found at the California coast.  On cloud arrival,
the   aircraft  measured  12 picocuries  per  cubic   meter.    That  is  a
reasonable  comparison  in this business.  We  now  believe that  an upper
cloud was  created  by   the  early  explosions,  the intense fire,  and
residual  heat from  the  reactor.   A  lower  cloud  was produced  by the
lingering fire that  had  a lower  heat flux.
      From the California coastal estimate, transport was projected over
the   continental  United   States,   assuming  a   reasonable  vertical
distribution,  calculating  a vertical  integral,  and assuming immediate
scavenging  of   100   percent   of  the   vertically  integrated   iodine.
Assuming  this  radioiodine to be deposited  on the  ground  and  eaten by
the  family  cow,  the  maximum amount  of  radioiodine  in  cow's milk in the
United  States  was  estimated to be approximately  900  picocuries per

                                  31

-------
liter.   As  you  will  recall,  guidelines  are  established  at  15,000
picocuries per liter.   Thus,  we estimated that there would  be  no  health
problems in  the  United States as a  result  of radioiodine deposition,
even  in  the  deposition  hot  spots.    In   areas  actually   measuring
radioiodine  concentrations   in milk,  the  maximum  measured  was  600
picocuries per liter in one  sample.   Most of  the measurements  were from
100 to 200 picocuries  per liter or less  across the  United States.   From
these simulations and the many measurements,  we believe that  the  upper
cloud from Chernobyl did exist.
     From all of the northern hemispheric measurements  and  simulations,
we  now  estimate  that  there  were 2.4 million curies of    Cs  released,
which   represents  just  less  than  half   of   the   core   inventory.
Approximately  half  of the 131I inventory of 80  million curies,  or 47
million curies,  was released.   Our  final  estimate is  that half  of the
core  inventory of cesium and  iodine  was  released with 100  percent of
the noble gases,  10 percent  or less  of  the tellurium,  and  1  percent or
less of the  other materials.
     The  experiences  from  Chernobyl  lead   us   to   identify  several
research needs.   One  of the  most  important things  we  can  do  to improve
emergency response  is to improve  the flow of information.   This means
early  warning  by  the  country involved  with the accident;  openness
concerning what  was released,  the nature of  the  accident,  severity and
early  measurements;   in   fact,   everything   known   about   initial
distribution  of  the material released.   In  the United States, we need
to  improve  our  own  flow  of  information.   There  are  nine  Federal
                                  32

-------
agencies involved in emergency  response.   Each one claims a role, does
quality control differently, and perceives a different jurisdiction.
     Additional research  is also needed  in  the area of meteorological
prediction.   Models can  be driven  with  input  from  numerical  weather
prediction  models  that operate  on  the global  scale  down  to mesoscale
models  on  the  smaller  scale,  i.e., state-size,   regional-size,  or
continental  in scale.   Yet, each  serves different  purposes.   On the
larger  scale,  for a massive  event like Chernobyl, the question is where
is  the  released material  going  and what will the concentration be when
it  gets there?  For smaller chemical releases, the prediction serves a
different  purpose.  You may  have a  hazardous  condition to engineer your
way out of and, through  prediction,  one  can select optimal conditions
in  which the material may  not  flash,  or  other risks can be minimized.
Thus,  prediction  serves  different  purposes  depending on the manager's
and decision  maker's  needs.
      We need  to  know the  physics and chemistry  of the  release.   At
Chernobyl, we needed to know the combustion  products, the  heat fluxes,
 and the isotopes  released to the environment, in order to  know what the
 exposure and doses to man would be.   To  say so many curies of activity
 are released is meaningless.   You  must know the number  of isotopes and
 the amount of each isotope in  order  to  know the mode  of  exposure and
 those doses that will  finally be delivered to the  human  community.
      Our  particle-in-cell   techniques  need  to be  expanded to  handle
 complicated and differing sets of isotopes.   It is easy to do particle-
 in-cell  calculation  for  one   to   five  different   isotopes.    To  do
 simulations for a suite  of  15,  16, 40, or 50 and do these calculations

                                    33

-------
quickly, requires improved techniques.  We would  call  this improvement
a hybrid  particle  approach.   In  the  last 5 years, two  proposals have
been developed to address this specific need,  but  have not been funded
(by any Federal agency).
     We need  additional  information about scavenging  properties.   The
radioiodine  from Chernobyl  was  carried  on  a  host  aerosol.    Iodine
exists both in gaseous  and particulate form as  it  moves  downwind.  The
scavenging  is mainly of  the particulate.   It can  be  scavenged  very
efficiently if the iodine is coating the  host  aerosol  particle,  because
iodine  is  a cloud-seeding material and  is  most effective  in the ice-
producing parts of clouds, which probably existed  over Europe and Japan
where the "hot spots" were created by wet deposition.
     We  need  standardization  of   action  levels   of  contamination  in
products to know what actions to take consistently.  In Europe,  part of
the  confusion  was  due  to  the  lack  of  uniformity  of  standards.
Vegetables  were  needlessly destroyed, and populations  were needlessly
disturbed.  The  countries tended  to overreact  because of nonuniformity
of  standards  and extremely  low standards  set  by  people  who wanted to
show what good they  were  doing  for the environment.   When, in fact, it
became  "crisis  time,"  they  created confusion  in  the  population.   The
Soviets claimed  that no damage occurred  outside their country and that
the  Westerners took unnecessary  precautions.    Yet,   some Westerners,
such  as  the  shepherds of Wales,  are seeking  compensation from the
Soviet Union.  Such  compensation  will  not be  possible in international
law  circles  until  there  is  standardization  of action   levels  and
protective  action  guides.   There  needs  to  be  standardization  of

                                   34

-------
measurements,  analysis,  and reporting  procedures.   A specific problem
with reporting occurs with  iodine levels.  When a given iodine level  is
reported  from  a  specific  station,  one  must also  know  how  it was
measured.    It makes  a  difference whether  the  iodine  is  a  gas  or
attached to particulates.
     We  need  to study the biomedical  history of  the 200,000 severely
dosed  people  in  the Soviet Union.   This  is  not the  acute radiation
victims  who saw more  than  a  few hundred rem  in  the first  day  or  so
after  the Chernobyl  accident; they will  certainly  be  studied.   The
200,000  who  saw from 50 down  to  a  few rem should  be medically tracked
and  will constitute  a valuable  database  for  understanding  radiation
effects.
     The  Soviets  proved  to  be ingenious  in minimizing the  impacts of
the  accident.   We  should  learn from how they did it.   They minimized
resuspension  of  the radioactive material  in  the deposition  pattern
around and  up  to  some  distance from the plant.   They controlled runoff
to  the  streams  and  reservoirs by  building  levies  so  as  not  to
contaminate the water resources of the city of Kiev.   They did rainfall
suppression  in the  region  of  Kiev   in  order  to reduce the  problem of
surface runoff in the  neighborhood  of  the  plant.   They also took other
actions too numerous to list here.
     There is one area where the Russians were very lucky.  The initial
plume  from   Chernobyl   went  northwest  to  an   area   of  relatively
uninhabited boreal forest.  As a result, a relatively large fraction of
the  release  inventory is  resting  in  and  on the  pine needles  of the
boreal  forest  some  meters  above  the ground.   As the pine needles die

                                  35

-------
from the  beta radiation,  they will  fall  as fallout  during the  next
several years northwest of  Kiev.   This is an ideal  laboratory  for the
study of long-term ecological  effects of radiation,  which the world has
not seen.
     Thank  you  very  much.    This  discussion  outlines  my  ideas  of
research needs in the wake of Chernobyl.
                                   36

-------
                          ACCIDENTAL RELEASE
                        SCENARIOS FOR ANALYSIS
                            Jane Crum Bare
                 U.S. Environmental Protection Agency
            Air and Energy Engineering Research Laboratory
                      Research Triangle Park, NC
     Accidental releases  are a  new area of  study with  most  activity
occurring in the United States  only since the 1984 accident at Bhopal,
India, where  approximately 2500 people were  killed  and  thousands more
injured.  Until  this accident,  very little recent  effort was  put into
modeling the  dispersion  of  acutely  toxic gas clouds that  could cause
effects due to short-term emissions.  The purpose  of this workshop is
to gather expert dispersion modelers to analyze various scenarios.  The
purpose of this paper  is  to provoke this discussion and present actual
scenarios for  analysis.  The specific subjects discussed were chosen by
those  organizing  the  workshop  who felt that  the  attending dispersion
modelers needed a  background in this area to  which they might not have
had  previous  exposure.  The first section  will discuss  past  releases
that  have  occurred  in the United  States.    The  second  section will
provide  more  of  a  statistical   side   by  presenting  chemicals  most
frequently  involved.  The third  area  will   discuss  some  of  the many
variables that affect  source strength  and dispersion modeling, and the
final  section  will  present three representative scenarios for analysis
at the workshop.    None  of  these  sections  will provide  an exhaustive
account of the topics, but will  simply touch on major points.
                                   37

-------
Case Histories

     Several  questions  were  presented  for   this   section.     These
questions included: Have acutely toxic chemicals similar in toxicity to
methyl  isocyanate  (MIC)  been released in quantities  comparable  to the
quantities  released  in  Bhopal?   Has  anything  equivalent  to  Bhopal
happened  in  the United  States  in the past?   Are releases   occurring
today?   To  find  the  answers  to these  questions case histories  were
reviewed.
     Have  acutely  toxic  chemicals  similar  in  toxicity  to  MIC  been
released in quantities comparable to the quantities released  in Bhopal?
EPA's Acute Hazardous Events Database contains  a computerized  record of
6928  incidents  that have occurred in  the  United States in the  last 6
years.   The  toxic load was calculated for Bhopal  and 300  other events
in  the  database,  where toxic  load  is  defined  to   be  the  amount  of
chemical  released  divided  by the IDLH (immediately  dangerous  to  life
and health)  of  the chemical.  For Bhopal, approximately 90,000 pounds
of MIC  was  released,  where MIC has  an  IDLH value of  43 milligrams per
cubic meter.   This corresponds to  a toxic load  of approximately  2100
pounds  of MIC  per  IDLH value.    IDLH  represents the  maximum  level  at
which  a healthy  male worker could  escape after 30  minutes  exposure
without  irreversible  effects or  loss of life.    Some  chemicals without
established  IDLH  values use calculated IDLH from other health effects
data.    Here  the  IDLH  is  expressed  in units  of milligrams per cubic
meter for consistency in the toxic load calculations.
                                  38

-------
     Six events within the database were determined to have toxic loads

greater than  Bhopal's.   In other words,  if  all  other factors had been

equal  (e.g.,  population  density,   meteorology,   physical  state,  and

emergency  response),  these events  would have represented  at least as

much  potential  for toxic  effects as the event at Bhopal.   In each of

the  six cases,  however,  the factors were  not equal and  not a single

life  was lost.O'2)

      Has anything  equivalent to Bhopal happened in the United States in

the  past?   In doing research,  the following three events seemed not of

equal  magnitude, but  had serious consequences.

      •    Cleveland,  OH,  October  20,   1944.    A  liquid  natural  gas
           storage  tank  ruptured  at  5 pounds per square inch and -250°F
           in  a  residential  and industrial  area.    Liquid  natural  gas
           flowed down the storm sewers,  mixed with  air,  and exploded.
         'A  total  of  128 people  were   killed,  and 200  to  400  were
           injured.

      •    Texas  City, TX,  1947.   Bagged ammonium nitrate fertilizer on
           board  a  ship  caught  fire; 2280 tons of fertilizer burned and
           a nearby ship exploded.    After  the  explosion  and fires,  330
           dwellings,  130  business  buildings,  and 600  automobiles  had
           been destroyed.  A total  of 3000 people were injured, and 552
           lost their  lives.
      •     Brooklyn,  NY,  1944.    Chlorine  was  released  from  a  gas
           cylinder into the subway system.   A total of 208 people were
           injured;  no  one was  killed.^3)

 These  events  happened  approximately  40  years  ago.    Are   releases

 occurring  today?  The  following events from the Acute Hazardous  Events

 Database occurred  within the last  6 years in the United States.

      •     Niagara  Falls, NY.   One  ton  of liquid chlorine  spilled  from  a
           process  vessel  in a  chemical plant.   The plume that  followed
           resulted in  the evacuation of a nearby sports stadium where  a
           game was in  progress.  A total of 76  people were  injured.

      t     Elizabeth,   NJ.     At  the  Newark  Airport,   a  release  of
           isobutane, phosgene, hydrogen chloride,  and hydrogen cyanide
                                   39

-------
          resulted in the evacuation of the entire airport and 15 miles
          of the New Jersey Turnpike.
          San  Francisco,  CA.    A   liquid   and   gaseous  release  of
          polychlorinated  biphenyls   (PCBs),  natural   gas,   and  oil
          resulted in the  evacuation  of  19  buildings containing 30,000
          people at the
Event Statistics

     What are  the  statistics?   What are the chemicals  most frequently
involved in releases?
     The best  source  of information at this point  in time  for data on
accidental releases in  the United States is the  Acute Hazardous Events
Database.   The  database  provides  an  analysis  of  the  frequency  with
which chemicals  are involved in  fatal  and  injurious events, the causes
of  the  events  recorded  in  the  database, and  the  locations of  the
releases.   Even this database  has many  caveats  associated  with  it,
however.   For example, the  deaths and  injuries  may  not  be related
directly to  exposure  to toxic substances,  but may have  resulted  from
collision, fire, explosion, or other related causes.  Also, the method
of  collecting  the  data may  have been  biased  more toward  significant
events, rather than toward less  significant events.   This database may
not  have statistical  significance, but  this  is  the  best source  of
information available at this time.
     What do the data indicate?  For  events  causing death or injury,
chlorine  was  the  chemical  released  in  9.6  percent  of  the  cases,
anhydrous ammonia  was released  in 6.8 percent,  followed  in decreasing
                                  40

-------
frequency  by  sulfuric  acid,  PCBs,  hydrochloric  acid,  nitric  acid,
toluene, methyl alcohol, and sodium hydroxide.0)

Range of Release Characteristics

     What are  the  ranges of parameters that need to be considered when
modeling an  accidental  release?   Some of the major factors in chemical
releases  are:   the  chemicals  involved,   the  physical  state of  the
chemicals,  the conditions  of  release, and meteorological  data.   Most
source   strength   and   dispersion  modelers   are  familiar  with  the
differences  due to different chemicals  and  meteorological data, but may
not  be  familiar  with  some of  the  physical  states  and  conditions of
accidental  releases.
      The physical  state  includes  solids,  liquids, gases, or  two-phase
releases.    Low boiling point  liquids are  known for a slow,   steady
evaporation rate.  Superheated liquids and  refrigerated gases  are known
to  have  an  initial   flashing condition,   and  then  a steady  rate of
evaporation.   After  evaporation takes  place,  liquid  spills  may be
modeled as  light, neutrally  buoyant  or dense  gases,  depending  on the
characteristics  of  the gases formed.   Usually,  the  heavier-than-air
clouds  are considered  the  most  serious because they tend to  stay  near
the   ground  and,  thus,  in   the  breathing  zone  of  the   surrounding
populations.   Sometimes,  however,  due to meteorological   conditions,
lighter-than-air   gases can   act  as   heavier-than-air gases  and  then
warm up to  be neutrally buoyant  and then lighter-than-air gases.
                                   41

-------
     Two-phase releases are mixtures  of gases  and liquids.   These often
result from  a safety valve release  from  a high-pressure vessel  or an
explosion  in a  chemical  plant.    A  boiling liquid  expanding  vapor
explosion  (BLEVE)  is a  subset  of this group.   Solids can  be  divided
into four major groups:  powders,  molten solids,  solids in  solution,
or brick-like solids.  Of these, powders often pose the largest problem
in  terms of  fires  and  explosions,  while brick-like  solids are  very
rarely considered a problem.
     Many  other  parameters  affect  short-term  releases,   including:
unexpected  reactions  (including   products of  combustion,  which  may
themselves be toxic)  and  characteristics  of the  release  (e.g.,  jet vs.
passive  releases).    For  more  information   on  factors  involved  in
accidental  releases  or  equations used to  describe releases,  see the
book by  Frank Lees, Loss Prevention in Process Industries.(3)

Representative Credible Scenarios

     The  final  purpose  of  this  paper  was  to  establish  credible
scenarios  for analysis  at the workshop.   The scenarios  chosen are all
historical  accounts  from  a review  of individual  articles,  the Acute
Hazardous Events Database and the Loss Prevention in Process Industries
by  Frank Lees.   These  cases  represent  some of the  worst situations
involving the chemicals chosen.
     The  first  scenario   is  a  chlorine  release.    Chlorine   is   a
heavier-than-air chemical  involved in 9.6 percent  of the  accidents  in
the Acute  Hazardous  Events  Database  and has been credited with  causing

                                   42

-------
2 deaths and 382 injuries.(*)  Chlorine is stored as a liquefied gas in
amounts up to 450 tons and in larger quantities at low temperatures and
atmospheric pressures.(4)  On  December  10,  1976,  chlorine was released
in Baton  Rouge,  LA,  at a plant  undergoing  start-up  after shutdown for
maintenance.  An explosion overturned the tank, puncturing its sidewall
on a  sharp object.   Over the  next  6 hours, 100 tons of  chlorine were
released.(3)    What  model   would   be   recommended  for  the  emergency
response personnel making evacuation/shelter-in-place recommendations?
     The  second  case study  involves an ammonia release.   Ammonia was
the third highest chemical   in  production  in 1985.(5)  It  is credited
with  271   injuries   and  4  deaths in  the  Acute   Hazardous  Events
Database.U)  Ammonia  is lighter than air but can act as a'heavy gas in
very cold  conditions.(3)  The Mid-America Pipeline system crosses seven
states  covering  720 miles with  a 6-inch  pipe and an 8-inch  pipe.   It
has a  pumping capacity of  1180 tons per day.(^)  Three  releases have
been  reported on this system  in .the last  10 years.U»3)   The latest
release occurred on July 31, 1981,  in the countryside near Willowbrook,
KS.  An 8-inch pipeline of anhydrous ammonia ruptured at high pressures
releasing  700,000 pounds before being stopped by the shut-off valves on
both  ends.   The release  occurred  in  an  unpopulated  portion  of the
country and  resulted in no deaths  or injuries.(l>2,6)   The book, High
Risk  Safety Technology,  states that in  the  event  of a catastrophic
release of liquefied gas under pressure the chemical  may be thrown into
the air as a mixture  of vapor  and  very fine liquid  droplets.  Air may
be entrained  in  the cloud  and evaporate the  liquid, thus  cooling the
cloud.  This  will cause the cloud  to  act as a dense cloud.(7)  Would

                                  43

-------
this have  occurred  In this case?   If so, when  would the cloud  rise?
Would a different model  be used for summer and winter conditions?  What
is  the meteorological  cut-off  point  for  modeling  of this  type  of
release?  Do the models consider ammonia  leaks that initially act as a
dense  cloud release,  then as  a neutrally buoyant cloud, and  then as a
lighter-than-air cloud?
     A  third  scenario  is  the  release   on  January  4,  1986,  at  the
Kerr-McGee Corp. uranium  processing plant at  Gore,  OK.   A  27,000 pound
cylinder  of uranium  hexafluoride  was overfilled by 2000 pounds.   In
response, employees heated the  cylinder to liquify  the  contents and to
remove  the  material.  The cylinder burst, instantaneously  releasing 14
tons of uranium hexafluoride, which reacted  to  form hydrogen fluoride
and  radioactive  uranyl  fluoride  particles  upon   release  into  the
atmosphere.   The hydrogen fluoride then  reacted with the moisture in
the air to form toxic hydrofluoric acid.   Although the media emphasized
the radioactive  aspects of the release,   the  one  person  killed and the
100-plus  injured  were   attributed  to  the  release  of  hydrofluoric
acid.(8)   In  this case,  hydrogen  fluoride has a molecular weight less
than  air  but  may  act as  a  dense gas.   It  may  become  associated and
consist  of  a  mixture  of  hexamers  and  monomers   with  an  effective
molecular weight of 70.   Dissociation takes place as the gas is diluted
with air, but  the heat required to dissociate cools the mixture and can
keep it denser than the surrounding air.(7)  Given this scenario, would
the hydrogen  fluoride form a dense cloud?   What models should be  used
to  predict  the  concentration   of   hydrofluoric   acid  under   these
conditions?  Would the models have predicted  an  18-mile vapor  cloud?(8)

                                   44

-------
Summary



     The  purpose  of  this  paper  was  to  answer  questions  posed  by

identifying  actual  case  histories  and statistics of  releases in  the

United  States  and to  present three  scenarios  for analysis during  the

workshop in which  it  was  presented.   The scenarios chosen  included two

of  the chemicals  most frequently involved  in  accidents resulting  in

death  or injury  (i.e.,  chlorine  and  ammonia), and a  complex  release of

hydrogen fluoride.



                         Table of Conversions
                         1 mile
                         1 Ib
                         1 psi
                         1 inch
                         Tf
1.6093 km
0.454 kg
6.895 kPa
2.54 cm
1.8 x Tc + 32
REFERENCES
1.   U.S.  Environmental  Protection  Agency.   Acute  Hazardous  Events
     Database  Executive Summary,  EPA-560-5-29(a)  U.S.  EPA Office  of
     Toxic Substances, 1985.

2.   Conversation with  Fred Talcott,  EPA, Office of Toxic  Substances,
     Aug. 13, 1986.

3.   Lees, Frank P., Loss Prevention in Process Industries.  Butterworth
     and Company (Publishers) Ltd., 1980.

4.   Kirk-Othmer "Encyclopedia of  Chemical  Technology,"  Third  Edition,
     1980.

5.   "Top  50 Chemicals  Output Declined  3% Last  Year," Chemical  and
     Engineering News, April 21,  1986.

6.   Conversation   with   Max   Young,   Mid-American   Pipeline,
     Sept. 11, 1986.
                                  45

-------
7.   Green, A.  E.,  High Risk Safety Technology,  John Wiley  and  Sons,

     Ltd., 1982.
8.   U.S.  Nuclear Regulatory  Commission    Asse!^F5t ^^tt
     Health Impact from  the  Accidental  Release of UF6  at  the
     Fuels Corporation Facility at Gore,  OK,  Docket No. 40-8027, License

     No. SUB - 1010,  U.S. NRC,  1986.
                                   46

-------
      SHORT-TERM TOXIC RELEASES FROM CHEMICAL MANUFACTURING SITES
                          Robert E.  Rosensteel
             Office  of Air Quality  Planning  and Standards
                 U. S. Environmental Protection Agency
                      Research Triangle Park, NC

     The U.S.  Environmental  Protection Agency's Office of Air Quality
Planning  and  Standards   in  1985  published  in  the  Federal  Register
notices of intent to list eight organic pollutants  under Section 112 of
the  Clean  Air  Act.    Inclusion  on this  list was  based  primarily  on
carcinogenic risk  potential  to  the public and, to  a  lesser extent,  on
other  health effects.   Exposure  estimates  for noncarcinogenic health
effects  for  the  intent  to list  decisions  consisted  of  utilizing
information  on  the  continuous   emission  rates  in  conjunction  with
various  combinations of terrain  and meteorological  conditions.    An
acute  risk  focus  may have  resulted  in  a  different group  of organic
compounds.   The eight  compounds  undergoing evaluation  for short-term
health effects  are:
     1.   Butadiene
     2.   Carbon tetrachloride
     3.   Chloroform
     4.   Ethylene dichloride
     5.   Ethylene oxide
     6.   Methylene chloride
     7.   Perch!oroethylene
     8.   Trichloroethylene
                                   47

-------
     A  program  was  initiated  in  1986  to  gather  information  from
processes producing or using one or more of these chemicals to evaluate
the  potential  for  noncarcinogenic  health  effects  associated  with
releases  of  one or more of  the  eight organics.   The  production and
related processes considered up to the current time are:
     0    Butadiene production
     •    Chlorinated hydrocarbon production
     •    Chlorinated solvent-using processes
     •    Chlorofluorocarbon production
     •    Epichlorohydrin production and user processes
     •    Ethylene  dichloride production
     •    Ethylene  oxide production
     •    Neoprene  production.
     This  paper  summarizes  the   screening  procedure  we designed  to
evaluate  nonroutine  releases  from chemical  manufacturing   and   using
source  categories.   Data  are being  collected  for  several  types  of
releases:
     •    Process  vent emissions  resulting  from start-ups,   shutdowns,
          and control  device  bypasses
     •    Pressure  relief events
     •    Equipment opening losses
     •    Handling  and storage  emissions
     •    Accidental  releases.
Detailed  information to characterize  the  releases was obtained with  a
questionnaire.   The  questionnaires were  fairly extensive,   asking  for
information  to allow us to  pull together the necessary  information  for
submission to our Source Receptor  Analysis  Branch for modeling.
                                   48

-------
     For  process  emissions,  the  questionnaire  addresses information
associated  with start-up,  shutdown,  or  controlled-bypass conditions.
Meeting with the Chemical Manufacturer's Association, we received input
that helped us to identify those  actual  process  events that result in
the  higher emissions.   There is,  of course, a  possibility  that some
events  could  occur  beyond  the  limits   of the  profiles  that  were
established, but we are confident that we captured the vast majority of
events.   Another area where  we  obtained  information concerns pressure
relief events.   There may be  safety relief valves that are designed to
automatically  relieve overpressure process  conditions  that  may result
in relatively  high emission rates for short periods.
     In  the  area  of equipment  openings,  the  questionnaire  inquired
about experience and  procedures,  e.g., in preparing equipment for entry
for  either  maintenance  or  inspection   purposes  and  where  residual
material left  in the  equipment could  be  released  during the  procedure.
We  have  found some  equipment  maintenance  activities that  represent
significant  emission  sources.  Handling and  storage of materials also
represent  a potential source  of release.   For example,  working losses
are associated with such  activities as pumping into a storage tank.   At
some  facilities,  such  transfers  can  involve   fairly large  vessels.
Typical  activities include transferring  of  products  to rail  cars  or
tank  trucks  and  the evaporation  from  wetted  material  that   is  not
properly sealed.
     The   final  category  of nonroutine  releases   included  in  the
questionnaire  involved accidental  releases.   We  have  been  discussing
                                   49

-------
this category throughout this conference.   Examples  include failure of
a pump seal, liquid spills,  and gas or vapor releases.
     The data collected via the questionnaire have been  used as inputs
to  a  screening  procedure to predict ambient concentrations for worst-
case  conditions.    We  are working  with the  Source  Receptor  Analysis
Branch and the Pollutant Assessment Branch  of the Office of Air Quality
Planning and  Standards  to combine the  information we obtain  from the
questionnaires along source category  lines,  as  we typically do in our
standards development work.   For example,  for each  of  the plants that
utilize a carbon tetrachloride  scrubber on  their chlorine  tail  gas, we
are taking  the  events  that  have been reported and combining them into
what we characterize as  a worst-case scenario.    Our modelers  are also
using conservative  meteorological  and  atmospheric stability conditions
for evaluating this worst  case.  If this worst  case analysis  does not
result in a significant  health  concern, then no  further refinements in
that analysis are undertaken.
     Development of modeling  cases incorporating  several  simultaneous
emission events  to evaluate potential ambient exposure  for a  chemical
production category is under way, and some  preliminary modeling results
have  been completed.   We obtain a plot plan from the  source to locate
the emission  points from the process.  Generally,  we  are looking at
averaging times of  15 minutes, 1 hour, and  8 hours, which are typically
used  for the  health effects evaluations being considered.   It is  also
necessary to  compare the relative  contribution  of the various modes of
release, i.e., maintenance  versus  background  equipment leaks.   If  this
type  of evaluation suggests that health effects may be experienced by

                                   50

-------
the  public,   then  a  refined  analysis  will  have  to  be developed  to
evaluate short-term releases from the source category.
                                   51

-------
               FATE OF TOXIC RELEASES IN THE ATMOSPHERE--
             ATMOSPHERIC RELEASE ADVISORY CAPABILITY  (ARAC)
                           Marvin  H.  Dickerson
                 Lawrence Livermore National Laboratory
                              Livermore,  CA
      The  Atmospheric Release Advisory Capability (ARAC)  is  an  emergency
 planning,  response,  and  assessment  service,  developed by the U.  S.
 Departments   of  Energy   and   Defense,   and  focused,  thus  far,   on
 atmospheric  releases of nuclear material.  For  the past 14  years,  ARAC
 has  responded  to over  150 accidents, potential  accidents, and major
 exercises.   The  most notable accident responses  have been the  COSMOS
 954   reentry,  the  Three  Mile  Island (TMI-2)  accident,   the  uranium
 hexafluoride accident  at the  Kerr-McGee Plant  at  Gore,  OK,  and  the
 Chernobyl  nuclear   reactor accident  in  the  Soviet  Union.  At  this
 workshop,  Joe  Knox  presented a paper based on  the ARAC  analysis  of the
 Chernobyl  accident.   I  will  discuss the ARAC  project  in  general.
      I  would like to give you  a  brief history of the  ARAC program.   The
 concept for  the  project was developed about  16  years  ago essentially by
 three  people:    Rudy  Engleman,   now with  the  National  Oceanic   and
 Atmospheric  Administration  (NOAA)  and present  here   today;  Joe Knox,
 still  with  ARAC  and  here today;  and   Todd Crawford,  now  with   the
 Savannah  River  Laboratory.   I have had the good fortune to  be involved
with  the project  for  more  than 14 years.
     ARAC  directly   supports the  Department of Energy (DOE)  and   the
Department of  Defense  (DOD) by  providing real-time  assessments  of  the
consequences that may  result  from atmospheric  release of  radioactive

                                   53

-------
material.  As with any similar program,  there are three key components:
the staff,  the  models,  and the computer systems.  We  have a number of
meteorologists on the team.   Several times during  this workshop I have
heard  comments  that  meteorologically  trained  staff  are the  persons
everyone turns to for advice  during  an accident.   The  core of the ARAC
response team is composed of meteorologists.
     The types of incidents that we  address  are  quite  varied,  and they
have numbered approximately  150.   They  include  accidents/incidents in
transportation,   at  DOE   facilities,  at  DOD   facilities,   at  power
reactors,  extortion  threats,  satellite  reentries,   and  atmospheric
weapon  tests.   We  continue  to be  involved  in  a  number  of  tests  and
exercises  for  our DOE  and DOD facilities and  conduct  exercises  with
states  and  nuclear power  plants.  As a  continuing  link to the research
community, we participate in planned atmospheric tracer releases.
     We  serve  many  masters  depending  on the  type of  accident  that
happens.   We serve  seven DOE  sites  (with the potential  to serve up to
about   15):    Livermore  National  Laboratory  and  Site  300,  Sandia
Livermore,  Savannah  River Laboratory,  Pantex, Mound,  and Rocky Flats.
We directly  serve these facilities by maintaining databases for terrain
and geography and by having meteorology  data coming  into our emergency
response  center in  real   time.   We serve  approximately 40  nuclear-
capable  DOD sites and  their  commands.    During the Three Mile Island
stabilization,  we  worked  for  both  DOE  Headquarters  and  the  EPA.   We
also worked closely with these agencies for the Chernobyl  accident.  We
have   worked   with   several   states   regarding  nuclear   material:
California,  Pennsylvania,  New York,  Virginia, Florida,  and  Texas.   In

                                  54

-------
fact, we have  a  fairly broad customer base and during any one incident
may work with multiple customers.
     While the  source term is very  important  to  modeling efforts, one
of the  first things  we  learned  in  the  emergency response business is
that we will not know the source term at the beginning of an accident.
What we do  is  normalize the  source  term,  which becomes  a useful piece
of information in our early calculations.  Then the people in the field
begin  to  receive  data  to bracket  the  effects  using the  normalized
values as multipliers.
     In recent years,  we have been  involved in  several  toxic chemical
releases:  a 1976  train accident involving uranium hexafluoride (UFs),
a Titan II accident (even though there was a weapon on board it was not
damaged  so   this  was  a  toxic event  because  of  the  missile  fuel),  a
hydrogen sulfide leak during  a material  transfer  at the  Savannah River
Plant  in  1981, the Gore,  OK, UFg accident, and  the  white phosphorous
rail accident  in Miamisburg,  OH.
     The Miamisburg  incident lasted  for 5 days,  and  our  colleagues at
Mound  Laboratory,  near the accident, provided  input  to  the evacuation
planning process during those days.
     The models  that we  maintain  to support our  customers  must cover
widely different scales.   For nuclear weapons tests, we have to cover a
global scale.  As it  turned out, we  also used global-scale modeling for
the  Chernobyl  event.    ARAC's  operational models are highlighted as
fol1ows.
                                   55

-------
                ARAC OPERATIONAL MODELS FOR NUCLEAR MATERIAL
Model
Type
Simple
Intermediate
Complex
Global
Scale
=20,000 km
--
2BPUFF
PATRIC
Synoptic
Scale
=2,000 km
--
2BPUFF
PATRIC
ADPIC
Regional
Scale
=200 km
--
MATHEW/
ADPIC
KDFOC2
MATHEW/
ADPIC
Meso
Scale
=20 km
Gaussian
MATHEW/
ADPIC
KDFOC2
MATHEW/
ADPIC
     MATHEW/ADPIC  is  a three-dimensional, mass-consistent, wind  field
model   coupled   to   a   particle-in-cell   transport,  diffusion,   and
deposition model.     It  is   generally   applied  to  regions   200 by
200  kilometers  or less.   However,  for the  Chernobyl  accident  it  was
expanded to  2000  by  2000 kilometers.   The 2BPUFF  is a two-dimensional
long-range transport and diffusion model,  most often  applied to nuclear
weapons  tests.     PATRIC   is   a  three-dimensional,  particle-in-cell
diffusion model that was specifically  designed to treat continental  and
hemispheric  scales;  it  is  a  simplified  version of the ADPIC  model.
KDFOC2 is a  fallout model used  for  surface or partially  buried nuclear
detonations.
     These are  the models  that  we  have  operational.  For  our on-line
sites, where we have databases, we can produce calculations in about 40
to 45 minutes from the  time  we are  notified  (new computer capabilities
will  reduce this  time  to   15-20  minutes).    This  is  a  full  three-
dimensional calculation.  If a site does  not  have its own capabilities,
                                  56

-------
then  we  provide  them  with  a  system  in  which they  can  get  simple
calculations almost instantly from tower data.
     Our  regional  meteorological   data  come  from  the  established
national meteorological network.  For the DOE and DOD sites, the system
that  remotely  collects the data  is  a DEC PC 350/380.   Some sites are
connected directly to  a  local tower  with the data coming directly into
the system1.  Others have smaller  towers and the data are hand entered.
Still others depend on nearby airport data.   The minicomputers collect
the  data,  display  it  locally,  and  transmit  it  to   the  center  in
Livermore.   We also  have direct links  with Air  Force  Global  Weather
Central  and  a private meteorology company.   We  receive data  from the
National  Weather  Service as  well.    Thus,  we operate  a  large  data
collection,  database,  data  analysis,  and model-execution  facility  at
Livermore.
     We  are  currently  expanding our  capabilities in  several  areas.  We
are acquiring  geographical  information  from the U.S. Geological  Survey
for the  entire United  States.   We are currently experimenting  with our
own local terrain data,  including the overlay of model particles on the
terrain  data.    In  complex terrain,   we   are able  to  overlay  the
geography, the terrain data, and the particle behavior.   Eventually,  we
hope  to  be  able  to accomplish this  in  real  time as  part of our normal
operational response.
     We   think  some   of   the   high   payoff  research   areas   are:
1) development of  a  local-  to  regional-scale  meteorological  forecast
model or  dynamic model,  not for  an  extremely complex terrain  to  begin
with, but  perhaps  for rolling  terrain;  2)  development  of a realistic

                                  57

-------
rainout  module for  both  the regional-  and  global-scale models;  3)
continued  model  evaluation  studies;  and  4)  expansion  into  toxic
chemical  response,   including  addressing  the  issues  of  source  term,
dispersion, and reactivity.  The models that we are using today to deal
with nuclear  releases  will  be applicable to the  toxic chemical events
that are  nonreactive and at ambient density.  We will need to include
new capabilities, however, to deal with the chemicals that are reactive
and  those that are  heavier than  air.   We  believe  these capabilities
should  be integrated into the ARAC  system in  the future  as  part of a
total emergency response capability.
                                   58

-------
                        SOURCE STRENGTH MODELING
                            Jerry M.  Schroy
                                Monsanto
                              St. Louis, MO

     Most  chemical  companies  maintain  up-to-date  material   balances
 around  processes in  newly designed  facilities.   They understand what
                                                                    TM
 the  emissions are and  maintain access to models,  such as  Flowtran   ,
 (which  Monsanto uses)  or  other similar simulation models  to keep the
 data current.
     For the most part,  when we talk  about  process  emissions,  people
 normally  think   of   routine   emissions.     However,   I will discuss
 nonroutine events, such as runaway  reactions  and  fire situations.  They
 will  represent  elevated  sources  for  which  the questions  related  to
 dispersion are a lot  clearer.   Most of the  work I will discuss today
 was   done   in   response  to   the   acrylonitrile  standard  that  the
 Occupational  Safety  and Health  Administration  (OSHA)  developed  in 1979.
 Industry attempted to get a handle on whether or not it could meet the
 objective  and criteria  OSHA  was attempting  to address.   In response,
 Monsanto developed a series of models to assist  in that understanding.
 The  models were developed  in general terms  that allow their use with
 any  chemical  for which  physical  and  chemical  properties are  known.
     In  1979,  Monsanto put its  model  on-line for any company engineer
 to use,  worldwide,  on  a computer  system over which  the developers  of
 the  model  have  no  control.    The  engineers   can  use  it  for whatever
 purpose  they wish.   While the model has been used for many purposes,
most are related to accidents.
                                   59

-------
     Most  of the  accidents  that  I  will  discuss  today,  relating to
releases  of organic  and  inorganic  materials, are  based  on situations
such  as  overfilling  tanks  or  cutting  of small  pipes.    Others  have
discussed bursting tanks and  instantaneous releases today.
     Our  model  best  addresses issues dealing  with releases where there
is  a  release  point,  flashing, liquid,  and  a   pool  formation.   The
potential  presence  of aerosols in  the flashing  material  must also be
considered.   An engineer,  able to  use our company model, can usually
make  a judgment of whether  or  not  an aerosol is  involved.   We do not
know how  to model  the aerosol yet.   The American Institute of Chemical
Engineers  (AIChE)  has  initiated  some  work   in  an attempt  to qualify
aerosol  behavior.    Monsanto's  models  can,  however,  handle questions
related to  flashing,  pool evaporation, and release rate.
     All of the equations  and criteria used  in our model  are presented
in the  paper, "Emissions  from Spills" by John Wu and me (Air Pollution
Control Association   [APCA]  Specialty Conference,  1979).    I  will  not
cover  those details  today,  but will  cover  other issues.   You should
understand  that the  Monsanto emission model   can  handle different tank
configurations,   whether  they are  horizontal  or  vertical  arrangements
and whether the structure is a fixed  tank or a  tank wagon,  on rail or
on truck.    It makes no difference.
     In  dealing  with questions  related  to diking,  the  model  was
designed to deal with the  question  of how fast the surface of the pool
grows.  If  the  release occurs in a  diked arrangement, the surface will
grow out to a maximum area,  and then pool calculations are relatively
straightforward.  If the material  is released into an undiked area, how

                                  60

-------
big will  the pool grow?   Part  of the model and  part  of the equations
presented in our  paper  deal  with  that issue.   The model will calculate
a rate of growth.  When dealing with chemicals that volatilize rapidly,
such as hydrogen  cyanide  (HCN), the  model  will  malfunction, because it
does not  know how to  deal  with  pool  shrinkage.   The model will reach a
point  where  there  is  more volatilization  than  there   is  incoming
material  into the pool,  and  it  will  fail.   However, you can avoid this
problem in most cases.  By cutting study time back to the time interval
that only allows  pool  growth to the  diameter where the amount entering
the  pool  equals  the  amount  evaporating,  you  can define  the critical
area.   Actually  correcting  the  model  is a minor issue  that  simply
requires  time.
     The  question of  model validation comes up  frequently.   One area,
of  which  many of the meteorologists  in  the audience  are aware,  is the
abundance of pan  and  lake evaporation data in the world.  When we have
attempted to match the  data for  big lakes and ponds,  we  have found a
fairly close comparison.
     Our  model   is   dynamic.     It   solves  heat  and  mass  balances
simultaneously  and provides  temperature  responses  of  the  pool.   Some
questions related to  cooling arise  when  dealing  with releases  of a
chemical  like ammonia.
     A  second area,  where  issues related to  source   terms  come  into
play,  occurs  in  cases such  as   spilling material  into  a  sewer  or
attempting  to control  a spill  by washing  it  into a  waste treatment
pond.  Randy Freeman and I and a few others have done some work related
to surface-aerated ponds and systems using diffused air systems.  Short

                                   61

-------
of  a long  explanation of  the models,  this requires  a  steady  state
model.  In this case,  we  are  trying  to solve simultaneous differential
equations for competitive processes, and we  do  not  know how to do that
dynamically yet,  at least  not with our  model  and  not on  a  computer
system we can afford.
     We deal with  the  question of competitive removal  in  the  model  in
terms  of  biological,   chemical   oxidation,  or  chemical  reaction  in
competition with  air  stripping.   The  user must  input a  first  guess
answer to initiate the model.   The better the guess, the lower the cost
to run the  program.   The worse the guess, the more expensive  it  is  to
run  the  program.   So  far, we  have  done fairly well on  closure of the
balance.    When  biological  oxidation  is going  on, you  have to  have
actual information  regarding  pure material  kinetics versus the  amount
stripped.    In  the  absence  of  biological  oxidation,   the model  air
stripping  predictions  are  within  about  3  percent  of  actual  air
stripping rates.
      In those cases where we must use a "lattice of models," as Dr. Cox
discussed  earlier,  we have  the  question  of  what  is   occurring  with
respect to  seepage  into the ground.   We  have  a temperature model that
has  been built  into  our "dioxin"  model.   We refer to the base model  in
terms of  2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) only  because that
is the first  chemical  we modeled with  it.   Now,  quite  a few chemicals
have been studied.
     The temperature model allows variation of temperature with time at
the  surface and  at different depths  within  the  soil.    Behavior  of
chemicals in the ground has always been messy to profile.  The chemical

                                  62

-------
migrates  both  into and out of  the  ground.   Our focus in using a model
for  source strength  is straightforward,  however.    We are  trying to
understand   the  impact   on   people  by   accurately  estimating  the
concentration of the material in the airspace.
     Using dioxin,  the model  was  tested  with 1984 data from a 1972 Air
Force experiment.   The  dioxin started at about 10 centimeters below the
surface  and  moved  up  and  down.    This  is  unsaturated  zone behavior.
Dioxin has a vapor pressure of about  10"9  millimeters.   Additionally,
Dr.  Nash from  the  U.S. Department  of  Agriculture at  Beltsville,  MD,
provided  us  with  data from  experiments  on dieldrin,  heptachlor,  and
lindane  with  which we  tested  the  model.   Lindane, for  example,  has  a
vapor pressure  about  five orders of magnitude higher than  dioxin.   If
we were  looking at benzine or HCN,  the  same migratory behaviors would
be occurring.   The only difference  is that  the  migration would  not be
spread out over 12  to 30 years, it would be over a matter of minutes or
hours.
     From a safety  point of view, the spills model will also model  down
to something equivalent to TCDD,  and we can  use it  for something  that
extreme.    It does  provide  conservative  values,  but  does not  consider
downward migration.   Therefore,  from the people  point  of view,  if you
model only the  evaporation  from  the pool,  you  have a  conservative
approach, and  that is  the  central  question regarding source  terms in
active models.
     Another source term  question   area,  which  we   all  face,   is  the
runaway reaction question.  There you must  go to other models, whether
it be Sapphire or one of the other models available from AIChE or their

                                  63

-------
contractors.   The  models available from  AIChE and the  new  ones under
development appear to be relatively realistic in providing results.  At
Monsanto, we  use data from an accelerated rate  calorimeter  to drive a
reaction  kinetics  model for  both  conditions of runaway  reactions and
fire  impingement on  a  vessel  to  determine the rate of  release from a
rupture disc or a relief valve.
     We   have  heard   some   about   routine   emissions   in   terms  of
understanding  background.   We must add fugitive emissions to the list
because  background  makes   a  difference.    If  background  levels  are
considered, pool evaporation  may  be  suppressed.   Fugitive emissions in
the work  place have  been documented.   However,  we  feel  that  EPA's work
is conservative by a factor of at least ten in a lot of cases for small
plants,  but  realistic  for  refineries.    Fugitive emission  estimates
provide an  idea of the  kind of losses that will  occur from sources and
the impacts that they will  have on workers.
     Regarding the conservative nature of EPA's  estimates of emissions
from  small  plants,  we  define  small  plants  as  those  that  have small
piping and use chemicals of lower volatility  than ethylene and propane,
which  are some of  the  lighter  chemicals found  in a gasoline plant.
Comparison  of actual measured mass  emissions  from an overall  process
within a  structure versus  our estimate versus  EPA's estimate will give
you an idea of the kinds of ranges available.
     There are copies of our  paper  available,  and  I would be available
to discuss our work with those interested.
                                   64

-------
        FUTURE  NEEDS  FOR  DISPERSION  MODELS  IN  HAZARD EVALUATION,
            EMERGENCY PREPAREDNESS,  AND ACCIDENT PREVENTION
                            James L. Makrls
                  U.S.  Environmental  Protection  Agency
                            Washington,  DC
     Someone  In  the   audience   observed   that   there  are  about  55
dispersion models available  at this time,  none of  which are great and
some of which are not  any  good at all.   So, the subject we are dealing
with here  is certainly very  timely.   Since this time  last year, many
events have  occurred that have  impact  on  our discussions  today.  The
Chemical  Emergency  Preparedness  Program   Interim  Guidance  has  been
released in  draft form and commented  on extensively by the public.  It
is now being combined with an older Federal Emergency Management Agency
(FEMA)  document  and  will  be  released  as  the  Hazardous  Material
Emergency  Planning  Guide.   Previously referred  to  as  FEMA-10,  the
earlier document dealt with contingency planning for hazardous material
accidents on a state and local level.  The new unified document will be
published  by the  National  Response  Team  and  made  available  through
notification in the  Federal Register by March 17, 1987.
     Recently,  a  couple  of  major  accidents  reminded  us  that  we are
dealing with a serious business.   The Miamisburg  incident caused the
evacuation  of  40,000  people following  a  rail  accident that released
white  phosphorous.    The  Kerr-McGee release  of uranium  hexafluoride
brought an  entire new  group  of facilities into  our  area of interest,
i.e., materials-handling   facilities  from  the  nuclear  industry.   The
Kerr-McGee incident  also alerted  us to  areas  where  we had  thought that

                                   65

-------
all government agencies were  functioning  in  a  unified way.   We learned
that there were still some activities that the Department of Energy and
the Nuclear  Regulatory Commission  had  not  quite  aggregated together.
This has been corrected, and the Federal agencies now meet on a monthly
basis   to   address  issues  of  common   concern   regarding  accidents
specifically.
     This  year,  state and  local  governments,  including the  States  of
Hawaii,  Illinois,  New Jersey,  and  California,  passed legislation aimed
at  preparing to provide  the  public with  information on the  chemical
risks that they face, as well  as requiring the state to move forward in
preparing  for an  accident  before  it  occurs.    We may  even have  new
Federal   legislation   in   the   form  of   the   pending   Superfund
reauthorization  bill.   When  or if the President  signs  the bill,  it
contains an  important right-to-know component in Title III.
     Basically, it  instructs  the governors to  appoint commissions,  who
in  turn will appoint local planning districts and committees.   Those
committees would  undertake immediate contingency  planning  around  the
list of 402  extremely,hazardous chemicals  (originally published  in  the
Chemical Emergency  Preparedness Program Interim Guidance).   Facilities
that exceed the baseline quantity limits for  storage established  by  EPA
must  submit   information  and   participate  in  a  broad-based  community
contingency  planning  effort.    EPA is  currently  rushing to  establish
those levels.
      If EPA  fails  to establish another level, the  new  Superfund bill
requires that a base  level, i.e.,  quantity,  of 2 pounds  be established
as  the  amount  of a  listed  chemical   in  a  facility  requiring  the

                                   66

-------
submission of contingency  planning  information.   Therefore, two pounds
of any one of 402 chemicals will potentially require that a contingency
plan be developed  for the local community.   Certainly,  the base level
for some chemicals will  be increased.   However,  the important thing is
that  this  will  require a  contingency  plan  that  will deal  with  such
issues as  evacuation,  emergency organization, notification, and public
health  protection.   The  local committee  must  include the  specific
elements   of  evacuation,   identification   of   hazardous   chemicals,
identification of the organizations and individuals in charge, and must
in  fact  develop a  plan.   One  of  the  key  issues  is going to  be  what
happens when a chemical release takes  place?  Who  makes judgments on
how far to evacuate  and when?   All  of these issues are going to depend
upon  some  unified  Federal guidance  being   provided  to  that  local
community, in  order to  assure  there  is no division,  or  diversion,  or
confusion  at the  state or  local  levels,  as  they proceed to  comply.
Other  important  aspects of this legislation are that  it  comes  with no
Federal funding  and  that it requires no  Federal  approval.   It  says to
the governor that  it is  the state's responsibility and  lists  several
activities the state must undertake to meet that responsibility.
     This  brings us directly to the issue we are discussing today.   The
states are going to  be  clamoring for specific information on modeling.
It seems  appropriate that the  Federal  government,  in  conjunction  with
public interest  groups,  industrial  associations,  technical  institutes,
scientific groups,  and  other agencies,  can meet  together to  deal  with
this issue and to  provide  state and local governments with information
                                   67

-------
around  which  they  can develop  specific contingency  plans for  those
chemicals that are in their community.
     There was  some  skepticism  earlier about  whether industry  would
really  come forward  with  the information that would be required  to do
this  contingency  planning effectively.   The law  is  going  to  require
that industry not only disclose the existence of the chemical,  but also
how much  they  have,  and for specific chemicals, where they keep  them,
and what protective systems exist.
     Despite all  of  this  planning,  however, a  key issue  will  always
remain.   That is,  what is  going  to happen when  a release occurs in
spite  of all  the  precautions  that precede,  and  all  the  mitigation
planning  that  had   occurred?  When  an  event  occurs,   what  is  the
evacuation route?   What  is  the  dispersion  probability?     What  is the
plume  of the  release and what  are the  effects  going to  be?   It is
around  these  very issues  that this  and other groups must  apply  their
greatest  efforts.    It  is  clear  that  a   lot  of dispersion  models
currently do  not take  into  account  issues  of topography  and  weather.
When  we first  sent  people  to  Cameroon, carbon  dioxide  had  not been
established as the cause of the deaths, although it soon  was.  It was a
shocking thing to learn that people  and all  animal life  were felled in
their  tracks  25 kilometers  away  from  a point source  by  this heavier-
than-air  chemical.    It is hard  to grasp that a chemical  as common as
carbon  dioxide could  disperse for a distance of some 15 miles and still
remain  in such intense concentration to kill.  Obviously,  we have a lot
to learn about the dispersion of chemicals in the atmosphere.
                                   68

-------
     We are involved in policy-in-the-making as we work together trying
to deal with this  problem.   Our solutions will provide state and local
governments with the ability  to respond  quickly and accurately for the
protection  of  public   health,  which  is  clearly  a  state  and  local
responsibility.  Title  III of the Superfund bill is somewhat vague.  We
will  need  information  on  how  to  set thresholds,  add  chemicals,  or
eliminate  chemicals;  that is certainly within  our  prerogative.   There
is,  however,  no mandatory prevention program included  in  Title  III.
While  it  is not legislatively mandated at  this time,  EPA is gathering
information  and reviewing technologies that we think will  be  able  to
lead to an  improved capability of preventing chemical accidents.  While
we  have  a fair amount  of  activity  going  on in the prevention area,  it
is  catalytic,  not  definitive.  Most of  our efforts deal  with prodding
others  to  do  their jobs and  being  sure  that  there  is an  adequate
transfer of technology.
     In this same  vein, we believe  that  dispersion modeling  well  suits
the  kind  of forum we are  having here, where ideas can be shared.   In
February,  1987, we are  having an international  symposium in Washington,
DC, where  the  technology that exists throughout the world dealing  with
the prevention  of  chemical accidents and the  dispersion  issue  will  be
brought  together.     Further,   the  American   Institute   of  Chemical
Engineers  (AIChE)  is  having a major dispersion modeling  conference  in
this country  in November, 1987.   These kinds  of  activities  are  to  be
commended  for  supporting  the coalescence of  information  and guidance
that can  be provided  to state   and  local  governments, as they  try  to
                                   69

-------
make the  determinations  of how to  protect their citizens  from future
chemical accidents.
     We  all  need  to  work together,  and  that  is what  we  are doing.
Everybody has a job.   The  public  sector  has a job of being a catalyst,
a cautious neutral, an informed neutral.   The private sector has a job.
It is their chemical;  their liability.   State and local  government has
a role.  They have been responsible for protection of public health and
safety  on  the local level for  a  long time, and  they  continue  in that
role.   Federal  agencies  like  FEMA, the Department  of Transportation,
and  the Coast Guard all  have  a role.   Agencies like the  Centers for
Disease Control and the Poison Control Centers have responsibilities as
they review these  chemicals and chemical  profiles and reach conclusions
about how  to  deal  with them.   The scientists have a job to continue to
move forward  in these  very complicated technology areas.    The citizen
has  a  job, not only a right  to know, but  a  right  to  become informed.
The  job  requires  active participation  and learning,  not   only  in
assuring  that there  is  a high  level of  safety,  but also  in  passing
appropriate local  ordinances where they best prevent chemical accidents
from having an impact  on people.
     Most  of  all, we all have  a  job  together to  try  to   increase
chemical safety in this  country.   In the event of a chemical accident,
we must  be  sure that  the proper information is available for state and
local officials who have to make key decisions to protect the public.
                                   70

-------
              ISSUES IN REGULATORY APPLICATIONS OF MODELS
                            David  E.  Layland
                  U.S.  Environmental  Protection Agency
                       Research Triangle  Park, NC

     The Source  Receptor Analysis Branch  of  U.S.  EPA's Office  of Air
Quality  Planning  and  Standards  (OAQPS)  is responsible  for supporting
the development  of air regulations  under  the Clean Air Act  and,  more
recently,  the  Resource  Conservation  and  Recovery Act,   as  well  as
supporting  activities  of other  EPA  groups in  the area of dispersion
modeling.   OAQPS views  modeling  needs  and  issues primarily  from the
perspective of  planning for sudden,  unplanned,  routine release events,
rather than catastrophic  releases per  se.  These  issues  and needs will
be discussed from the standpoint of the following:
     t    Regulatory and  technical  background giving  rise  to  the need
          for dispersion modeling,
     •    Various technical issues remaining to be resolved, and
     •    Considerations  in devising a framework to approach meeting of
          those needs.
     Although OAQPS  is mostly concerned  with  more  routine  releases, it
is  worth mentioning  two  aspects  of  planning  for  the  management  of
catastrophic releases.   One is  prevention, and the  other  is  emergency
response planning.   Looking at prevention, specifically as related to
hazard analysis  (the  question  of who gets  hurt), this is really the
place where dispersion models  find their  best use.   Included here are
other modeling  techniques:   source modeling,  health assessment, and
identification of the release mechanism itself.
                                   71

-------
     Section  112 of  the  Clean Air  Act (CAA) deals  with establishing
National  Emission Standards for Hazardous Air Pollutants  (NESHAPS).   It
requires  a listing  of hazardous air pollutants  first and a setting  of
standards  second.   A hazardous air pollutant  is defined  as  one that
could  cause  an  increase  in  mortality or  an  increase  in  serious
irreversible  or  incapacitating  reversible  illness.    Thus,  from  a
regulatory perspective, not just situations that cause immediate danger
to  life,  but  also those that  cause  sublethal acute  and chronic health
effects  are  to  be considered.    Section  112  requires  that  emission
standards  be  set  at  a  level  that  would  provide an ample margin  of
safety to  protect the public health.
     OAQPS is currently  looking  at several  types  of  sudden  releases
from the  standpoint of dispersion  modeling.   These  types  of  releases
may result from:
     •     Process   overpressurization;   e.g.,  pressure  relief  valve
           events;
     •     Transfer  and  handling operations,  e.g., leaks  and  spills  of
           different magnitudes and sizes and  vapor  overflow  from the
           filling of large tanks or vessels;
     •     Malfunction  of  control   devices,   e.g.,  scrubbers   and
           incinerators; and
     •     Maintenance  operations,   i.e.,  typically   smaller  releases,
           tank flushing, and equipment opening.
     These  types  of  releases  have  some  general characteristics  that
must  be  considered.     Duration   of  releases  can   vary  from  quasi -
instantaneous  releases  lasting only  a  few  seconds to  quasi-steady
releases   lasting   several  minutes  or  longer.     There   are  three
depressurization characteristics:  1)  rapid  jetting action  from relief
valves  and from other pressurized  releases, 2) adiabatic  expansion and
                                   72

-------
cooling,  and  3)  two-phase  flow.   Other  characteristics  of  concern
include  formation and  spreading  of liquid pools  and evaporation  from
those  pools,  along with heating  effects  associated with the  substrate
or  the atmosphere.   Density-related effects include the sinking motion
of  negatively  buoyant  elevated   plumes,  the  formation of  secondary
source  clouds with  associated  dampening  of turbulent  motion, and the
gravity  spread of dense gas clouds with terrain contour interaction.
     Technical issues  of a  regulatory  nature  can  be grouped into three
basic  areas:   1) model  selection, 2)  model  application,  and 3) model
verification.   Model  verification  should  probably  be first  on  this
list,  but experience shows it tends to be dealt with last.
     One  of  the  most  important  criteria  in model  selection  is  the
availability  of  the  model  in  the  public  domain.    For  regulatory
purposes, proprietary models must generally be excluded.  Other aspects
related  to   model   acceptability  are   code   standardization   and
maintenance,   and  the  availability  of   users  manuals  and  related
documentation.    The  theoretical  soundness  of  the  model  is  also
extremely  important;  i.e.,  is  the  model  consistent  with  current
knowledge?
     Consideration must  also  be given to  whether a model  is intended
for  screening or whether  it  is  a more refined  model. Typically,  a
screening model is both  simple  and conservative,  while a refined model
is characterized  by  being less  conservative and more data and computer
intensive.    One  must  consider  how conservative  the  screening model
should be and how much better an answer is obtained by the more refined
models for a given application.

                                   73

-------
     Another  important  aspect  of  model   selection  is  the  release
characteristics that were discussed  above,  including buoyancy effects,
momentum  effects,  aerosol  effects,   and  reaction  effects.   A  final
selection characteristic is terrain,  in terms of  a  generalized surface
roughness that affects the rate  of dispersion and  in terms of specific
terrain  features  that  affect  wind flow patterns, e.g.,  upslope  flows,
drainage winds, and ridge effects.  Downslope flows  of dense gas  clouds
may also be an important modeling consideration.
     The  first   issue  with   respect to   model  application  is  the
meteorological conditions to be  used  for the analyses.   OAQPS tends to
perform  two  types of applications.   First  is a  screening application
where  worst  case  meteorological conditions  are used  to establish  a
worst case.   Identification of the worst  case meteorological  conditions
generally requires some sensitivity analysis.   Second, for more refined
applications,  a  time   series  of meteorological  data  is  used  that
reflects  autocorrelation  inherent   in   the   meteorological   data.
Questions here are, "What is a reasonable worst  case?" and "What  is the
joint probability  that  the  release event and the meteorological  event
will occur simultaneously?"
     Another  issue of model  application is averaging time.  It would be
logical  to  look  at  the  duration of the  release   and  call  that  the
averaging time, but  in  modeling, we must be concerned  if the emission
is varying very rapidly with time.   We may  need to focus on the period
of  peak  emission.    Generally,   however,   it  is the occupational  or
community  exposure  standards  or  criteria  that  truly  dictate  the
averaging time.

                                   74

-------
     Integration of  the  source model with the  dispersion  model  is the
final issue  of  model application.   As  the models  and the events to be
modeled become  more  complex,  there  will  be  greater  need  to integrate
these two  aspects.   Of course, there are models available now that do
that.   Two  complex  releases,  previously mentioned,  that  may require
source  modeling in  order to  specify  the initial  conditions for the
dispersion phase are relief valve discharges and liquid spills.
     There  are,  of  course,  a number  of ways of  going  about  model
verification.   One  can   form  a theoretical  evaluation  of  a  model  by
simply  looking  at  the  model  formulation,  evaluating it,  and evaluating
the  assumptions of the model on a theoretical basis.  Validation of the
model code is also  important; that  is, checking how various numerical
techniques in the code are implemented and checking the model code over
a range of conditions of  the parameters.  It is  important to adequately
validate  the model   code  itself before  the  model  is  used.   Validation
against  real data may include both laboratory experiments  and field
trials.    Laboratory experiments  tend to  be  most  useful for  model
development.    For verification purposes, their applicability  to the
real atmosphere may  be suspect.   In the  design of field trials, scale
effects (small  releases versus large releases),  source-related effects,
individual chemical  effects,  and  meteorological conditions must all be
considered.    A  very  large  matrix of  field trial  experiments  is
therefore  possible,  but  only  a very few  are  feasible  to  be actually
carried out.  •
     In deciding  where  to go  from here, we  must ask ourselves, first,
what models  do  we have that  are ready  now for  regulatory applications,

                                   75

-------
to  what  situations  do these  models apply,  and with  what degree  of
confidence?  Then, we  can  ask  what  models  have been formulated but are
incomplete at this point,  to what extent can  they be adapted to actual
situations, and  how  can this  be  accomplished?  We  must  also  identify
those  situations  for  which  no appropriate model  exists,  whether the
absence results  from a lack of  scientific knowledge or  from  simply a
failure to formalize  a mathematical  model  from existing knowledge, and
what can be done to develop the needed models.
     In summary, this has been an overview of dispersion modeling needs
and  issues from the  OAQPS perspective in  the  area  of hazard analysis.
Some of  the issues  of model  selection,  model application, and model
verification were discussed, and  some suggestions  on where we  might go
from here have been made.
                                   76

-------
              COMMUNITY NEEDS FOR HAZARD EVALUATION TOOLS
                              Fred Millar
                     Environmental  Policy  Institute
                            Washington, DC

     The Environmental Policy Institute,  which  has  been  working on the
issues of  major  hazardous accidents for  about  6 years,  believes  that
there are  now  some  new opportunities and some  new  resources  that  need
to come together.   Specifically,  what  can the  public  sector  do within
the next 4 or 5 years to improve chemical  safety?
     A major  public  problem  is  that  people have  not been  told  what
risks they are  being exposed  to now.   There  really  has not  been  a
communication of potential hazard to the people.  Yet,  people  are going
to want to know to what risks and hazards they are exposed.  It is  fair
to  say  that people  in  Bhopal  were not told  of their risks,  nor  were
people  at  Chernobyl.   Certainly,  no  one  in  Europe  was told  that  a
release event  at the Chernobyl facilities could  dump  its radionuclide
inventory  on them.
     Let me  relate  a  story  of  how  a  few  months  ago  I  took my
4  and  1/2-year-old  daughter to  a skating rink  in  Fairfax County, VA.
As we walked  up to the rink,  I  noticed a gigantic petroleum tank farm
looming over the rink.   Also adjacent  to the tank farm are a  shopping
center and  a residential  development.   I  said to myself,  "That doesn't
look like  very smart   zoning."    Fairfax  County   is  one  of  the  10
wealthiest  counties  in the United  States and one  of  the most heavily
developing at this time.  Obviously, there had been encroachment by the
community  on  an  existing tank  farm.    I  decided  to approach  the  local
                                   77

-------
fire chief and inquire about this situation.   He reported that this was
only one of what he referred to as his "target hazards," and offered to
give me what turned out to be a 4-hour tour.
     First, we went to an underground  propane  storage  facility.   Going
into the  facility,  I  knew that this one had  leaked.   Propane has been
found  in  adjacent  creeks.   It  is  good to  go to these  facilities with
the fire chief, because you simply walk in  and talk with the operators.
A residential community is  about 100 yards away from the storage area,
separated  by  only  a small fringe of trees.   The operator related that
when the  facility  is to  be filled,  they bring  in from three to nine
jumbo  tank  cars  simultaneously  and   fill   the cavern.    I  asked  the
operator  what the  community thinks of the  facility.   He  related that
several of the people  in the community had  been  told when  th^ey were
buying  their houses  that the  line  of trees  was   a  park  where  their
children  could play.   In  fact,  the  trees  are  part  of  the  storage
facility.   It  appeared to  me that this was a "Mexico  City" accident
waiting to happen.  Asked why nine tank cars were  on site and unloaded
simultaneously,  the operator  related  that  it  took the same time to
unload  nine  as  to unload  one.  Since  the  people  were paid  the  same
overtime,  the operator simply chooses to take as many rail tank cars as
he  receives at one time.   It appears that this is an opportunity for
some operational improvements to reduce risk.
     The  fire chief took me  to another propane facility,  where there
were  a lot  of storage tanks,  and where  there had been  a  couple of
leaks.  He reported that,  to get water at  the facility, he had to  line
up  fire  trucks with  over  6000  feet  of  hose.    When  he  asked the

                                   78

-------
operators to  get water to  the  facility,  they  refused.   It  is  a pre-
existing facility  and is  protected  by a  grandfather clause  from the
current  safety  regulations.    The  fire  chief  related  several  other
similar problems.   One is a water treatment  plant  built over 10 years
ago with a chlorine facility, but no scrubber.  A similar new plant has
a  scrubber.    It  appears  that a  lot of retrofitting  is  required,  and
even  some  fairly simple  adjustments in operating procedures  would  be
beneficial.
     Part of  the problem  is that people have not been told what risks
they are exposed to.   The people of  Bhopal  and  Chernobyl were not told
of their risks.  There really has not been the  communication of hazard
to people.  In this country,  one can illustrate this by looking at the
U.S. Department  of  Transportation (DOT)  Emergency Response  Guidebook--
the "orange book"--that is  in every  fire truck.   There are  some "Stone
Age," well actually,  more "Model T"  tools  in that  book.  For example,
the  phosgene  entry  says that  in  case  of  a  spill  of phosgene  in
transportation the  fire  chief  is supposed to   evacuate  the  city  5.2
miles  downwind and 3.3  miles wide.    There  is no  fire chief  in  the
country  that   believes  he  or  she could  do  that  kind  of  evacuation.
Further, there are not 10 city council people in the United  States that
have ever been told that this material is coming through their city, or
that  it  could require a  17-square-mile evacuation.    The guidebook is
currently under  revision,  but  that  is  not  the entire  message here.
There is a basic lack of communication of hazard to  people.
     People are  going to  want to know  what risks they are  exposed to,
and, from a community point of view,  they are going  to need some decent

                                   79

-------
models and  some  decent  techniques to evaluate the  risk.   These may be
either paper and pencil  techniques or computerized ones.
     At  a  recent  conference  at  New  York University,  executives  from
Johnson  and Johnson were  talking about  their drug  case, and  Warren
Anderson of Union Carbide was talking about Bhopal,  when someone in the
audience said,  "Look, you're describing these major  industrial  crises
from the perspective of a fisherman,  rather than  the fish."  We need to
try  to  look at  things  the other way  around.   There  is  a need for  a
reversal  of  the  way people look  at  hazard assessment.   It is  fair to
say that most of the hazard assessment done up to now has  been from the
perspective of the facility.  The facilities  have done  a  lot  of hazard
assessment  already,  but  little  has been  done from  the perspective of
the  community.     This   will  be  a  different  perspective,  and  some
simplification will  be required.   This may be a bit  painful  for those
who  have had the  resources  and  the  perspective  of the facility,  and
where you want to know exactly what is going on.
     The  U.S.  Department  of  Housing  and  Urban   Development  (HUD)
discovered, when they started subsidizing low income housing around the
country,  that a lot of communities,  including Columbia,  SC, thought the
ideal place to locate subsidized  housing  for  the  poor was  on what they
called "marginal land,"  meaning right next to the  propane facility or
right next  to the  rail  yard.   The development in Columbia was (to be)
right next  to  (within  150 feet of) 44 propane tanks.   HUD did a quick
analysis and found that  the explosion  of just one  of those tanks would
destroy  60  percent  of  the proposed 1250 housing units.   HUD scrambled
and  established  some "acceptable separation  distances"  between things

                                  80

-------
that explode and  some  radius of fire risk,  but  they still  do not even
consider toxic  clouds.   This  is the kind  of response we  can expect.
Public  safety  officials  are   going  to  establish  various   kinds  of
regulations based on the immediate problems they are faced with.
     The National  Transportation Safety  Board  (NTSB)  did  a  rail  yard
safety study  in  1985.   They discovered that  of  eight  major rail  yards
reviewed  in  the  United  States,  not one  of  them  had  an  emergency
response plan that was coordinated with the local community.
     What kind of  tools  are  available to  the local  community?  You can
take a 7 and 1/2-minute U.S. Geological  Survey (US6S) quadrangle map of
any area of the  United States,  a map  of  the corresponding city, locate
the hazardous  facility of  interest,  and lay  down  the plume  from the
NTSB  on  it.   If we use  the Miamisburg, OH,  chlorine accident  as  an
example, the  chlorine  would be  expected  to go 2 miles downwind  in  10
minutes  at  a toxic  concentration  of 100 parts  per million.   This  is
something you  can work with,   and  these are  the  kind  of tools  that
people are  starting  to work with now.   The city of Denver  is doing a
rail yard risk assessment right  now.  After Miamisburg,  five cities  in
Ohio became interested in what  was coming through their area.
     To  many  of you  who understand  the new  advances in  dense  vapor
cloud technology, this is going to seem like "Model  T" technology.   But
the fact is that we  are going to need something  tomorrow,  not 2  years
or  10  years from  now.   Local  officials  are  going  to need  a workable
system, tomorrow  to  aid  them   in  hazard  assessment,  not  necessarily
quantitative probability  analysis.   Tools will be  needed where normal
                                  81

-------
logic allows you  to identify the chemical and  the  kind of event; then
you can identify the consequences and try to engineer solutions.
     Some  interesting  developments  are occurring in  the United States
following  the  Bhopal  incident.    Union  Carbide  came  forward with  a
statement  that,   "Hey,  we  ship  methyl  isocyanate  around the  United
States  to  five different  places much  safer than  Federal  regulations
require."  They reported that  they  were  using  better tank cars, better
placards,  and  special  routing  to avoid densely populated  areas.   As a
result,  the NTSB  jumped  on  DOT,   and  now  they  have revised  their
regulations for methyl isocyanate (MIC) and other similar materials.
     It is interesting to note that Union Carbide considers the Federal
regulations as minimal.  Thus, the better companies  have been  using far
better  safety  technology  than those  "minimal  Federal  codes."  We can
also lump  National  Fire  Protection  Association  (NFPA)  codes,  consensus
codes,  and other  industry codes together.   Many of  the industries  go
beyond  these,  right?   What has happened  is  a  lot of "grandfathering".
Things  have  been  kept at  a minimum  level.    What we  need  to  do  is
determine  what  is the state of the  art;  what  is the  best that can be
done with  a particular technology?  Then, when opportunities arise, the
regulations can be  ratcheted up,  so that performance is improved.
     In  1985,  EPA conducted an inspection of -an  Allied plant in Baton
Rouge,  LA, that is  using hydrogen fluoride.  .Allied told the inspectors
that  they  are  using  state-of-the-art  handling of  hydrogen  fluoride.
The  inspectors  then go down  the river to the  next facility  and learn
that that  is not  state of the art.   As  a result, Allied must agree to
fix things up,  and  they get a $10,000 fine from the State Department of

                                  82

-------
Environmental Quality  (DEQ).    With  some sort  of public  input,  there
will be  some sort of ratcheting  up  of some of the  standards that are
not as adequate as others.
     The NTSB took the  standard chlorine  plume  and the USGS map of the
Potomac  rail  yard in Washington,  DC,  and overlaid  them.  They learned
that  if the chlorine  plume  moved  east,  it would  take  out National
Airport.   If the  plume  moved  west, it would take out six schools and a
hospital  in  downtown  Alexandria.   In  the  case  of the  plume  moving
north,  it  would  take out  the Crystal  City  business complex  and the
Pentagon.   It would seem to  be a fair question  to ask if the Pentagon
or  the  Crystal   City  office  complex have  a system  of  shutting off
incoming air in case of a serious  release at the rail complex?
     In  Middleport,  NY, there  was a mini-precursor to  Bhopal  when an
FMC facility released some MIC,  and  it  drifted  100 yards to a school.
The students started  gasping  and  choking.   Since the principal did not
have  a way  to  shut off  the  air  ventilation system,  he evacuated the
students outside  into buses that  had been waiting to take children on a
field  trip.   They went  out into the cloud and were evacuated to a fire
station  where 200  children  had  their eyes  flushed because  of methyl
isocyanate poisoning.   Now,  the principal  has a system where pushing a
button shuts off  the  ventilation  system, and there is some possibility
of  sheltering in  place.   This is  the  kind of  simple  solution  we must be
putting  into place.
     What  are our priorities?   What  if the  public  sector is going to
have a role? What do we  need to  do  first?  This  is  where we  need  a lot
of  help from the experts--those  people who  have been working in this

                                   83

-------
area for a long time.  We need help to identify the kinds of situations
that have  the most  potential.   The  National  Oceanic  and  Atmospheric
Administration (NOAA)  developed  a real-time emergency  response system
on  an  Apple Macintosh  computer  and placed  the system at  the  Seattle
Fire Department  for a  year's  trial.    I  am very  skeptical  about  this
working for  emergency  response in real  time.   We may  be  learning how
not to  do  it.   Union Carbide missed the  mark  with their system during
the aldicarb  oxime release at Institute,  WV.   People  are  going to be
developing these real time emergency response systems, whether they are
adequate or not.
     After Bhopal, we are  having a rush  to emergency response planning
on  the part  of  all  the  major  actors.    For  example,  the  Chemical
Manufacturers Association  (CMA) is touting its Community Awareness and
Emergency Response  (CEAR)  program.   CEAR  is being  implemented  in  many
parts  of  the country.    In the Lake Jackson/Brazosport, TX,  area, the
CEAR program  placed  full-page  newspaper advertisements  about emergency
planning.   In  that area a siren program  was developed  that covers the
entire community with sirens that have a range  of one-half mile.  A map
was provided to the  community  showing  the location of the sirens whose
coverage overlapped.  When they were asked how the siren locations were
selected, they  indicated that a map  of the industrial  facilities was
first  developed  and the location of  the  major chemicals  of  concern
noted  (Map A).   Then,  the kinds of plumes that might be expected from
accidents  were  mapped,   and  potentially  affected  community   areas
identified   (Map  B).      From this information,  locations for sirens
(Map C) was established.  When asked how the community  responded to the

                                  84

-------
map of  the potential  plumes and  effects,  the developers of the  system
responded  that  they never  showed  that map to the community.  State  and
local regulators  are  going to want the  whole thing (Map A, Map  6,  and
Map C).  They think they have a right to know this type of  information.
  . 1 We  have to  come  up  with  hazard  assessment  technology  that  is
workable at the local level and not so refined that it can  only be used
by a company with $2 million to put into it.  We have to have some kind
of  a  hierarchy  of hazardous assessment  techniques  and  they have to be
acceptable to the local people.
     Finally, there  is a  bit of  a  double-edged sword  in  the changes
that  hazard assessment  and emergency response  planning are  going  to
precipitate  in  industry.    Union  Carbide  announced  they were  using a
special transportation planning system (similar to one developed  by ALK
Associates  of  Princeton,  NJ)  for  MIC  shipments.     Union  Carbide
indicated  that  special  low  risk routes  are  selected  and population
centers avoided.  However,  they will  not reveal  which areas they do go
through or what routing technology they  use.   They know that revealing
this information will result  in demands  being  placed  on them for other
chemicals to be routed with similarly stringent  rules.   One day before
the Bhopal  incident,  Carbide shipped MIC to five different locations in
the United States.   One year after Bhopal,  there  are no MIC shipments
in the  United  States,  because people have found safer  things  to use.
This raises all  kinds of interesting  questions  about  the potential for
safety improvements.
     At the  AIChE meeting in  New Orleans,  LA, last  year  the  final
session was titled,  "Accident Case Histories and Miscellaneous Topics."

                                  85

-------
At  the conference,  the session  chairman,  Dr.  Ian  Swift, got  up and
shook  his  finger at  a  large audience of engineers  and  told  them that
they  were  going  to  have to  listen  to a whole  lot  of "miscellaneous"
papers because  "not  one accident  case history" had been  submitted.  He
said,  "I guess  your liability lawyers have told you to  shut  up.  This
is a  very  dangerous  trend  in our  industry."   It is understandable, but
it is  dangerous.    It is unlikely that the public will think that such
an industry has  its  act together  on  safety.   Bhopal  gave them a shock.
It is  highly probable that there is going to be even more public sector
involvement.    In  Cincinnati, when  they passed a  right-to-know  law,
companies  submitted  8000 chemicals  and declared that  all  of  them were
proprietary.  The city  appointed  an  advisory  board,  and  now the claims
for proprietary information have plummeted from 8000 to 6.
     We are going to  have  a  lot of  good  dialogue between the  companies
and the public  sector in the next few years.   We need to work out some
workable hazard  assessment  systems.   I am encouraged  by  the  number of
people  from industry  who  agree.    We  need   to get  to  work   on  the
dialogue;  it is inevitable.
                                  86

-------
              MATHEMATICAL MODELS  FOR ATMOSPHERIC DISPERSION
             OF  HAZARDOUS CHEMICAL GAS  RELEASES:  AN OVERVIEW
                             Jerry A.  Havens
                    Department of  Chemical  Engineering
                         University of Arkansas
                             Fayetteville,  AR
      We  need  to  consider our  ability to  make reasonable  dispersion
 predictions for use  in  emergency response.   There are  two  main  issues
 in hazardous chemical  (gas)  consequence analysis that  require the  use
 of atmospheric  dispersion models.   One  is associated  with fire  and
 explosions,   the  other   with   toxicity.     The  flammability   hazard
 associated  with  accidental  release of a gas  usually disappears when  gas
 has  been diluted  (with  air)  to  about  1  percent.   The  toxicity  hazard
 may  extend  to concentrations  four or  five  orders  of magnitude  lower.
 We have spent much more  time working on the development of  dispersion
 models  for  flammable  gas concentration  problems  than  we have   for
 toxicity  problems.
     Analysis of a large number  of  accidents  leads  us to consider  the
 important  phenomenology  in  the  dispersion  process.    In  releases  of
 large quantities of denser-than-air  gases, there will  be a part of the
 process in which the flows are gravity-dominated, i.e., flows driven by
 the density  of  the gas  rather  than the atmospheric wind  field.   This
 process   is   particularly  important  with  flammable   gases  because
considerable dilution sometimes occurs.  If we are interested in
predicting gas dilution  down  to  1 percent  (the flammable  gas  issue),
these effects are  important.   However,  these  processes are  much less
 important  if  we  are  trying  to  predict  distances   to   toxic  gas
                                   87

-------
concentration  levels.    A  gas  that  is  heavier  than  air  is  stably
stratified,  and  vertical mixing is  inhibited.    The  cloud  hugs  the
ground, spreads  laterally,  and moves some distance  downwind.   We  are
trying to  predict  the distance  traveled  before dilution  occurs to  a
given concentration.   In  the  late stages  of the event, the dispersion
is driven  entirely by atmospheric  turbulence.   We  need  a model  that
considers these  different phases of dispersion.   The model  should  be
able to make a gradual  transition starting  from  the gravitational  phase
through  subsequent phases  without the  operator  having  to  make  the
switches.
     We have  tried to  identify  the main  questions requiring study  in
dense gas  dispersion  applications.   Our approach  has been to  isolate
different parts of the problem for study.   One controversial problem we
have addressed occurs  with  a  nearly  instantaneous release of  a  large
volume of  heavy  gas.   We know that the gas  slumps  to the ground  and
flows out laterally, but no one  had  (previously) agreed on whether the
gas is rapidly diluted during  the process.  We  ran many experiments in
the  lab   involving  instantaneous  releases  of   heavy gas   volumes,
measuring the lateral  gravity spreading and dilution.  We  looked at the
scale of the release,  the effect of the initial  density, and  the effect
of the  initial  height-to-diameter ratio of the  release.   Under  these
conditions, a heavy gas cloud behaves in an unusual way.  Upon release,
it  rolls  up  in  what has  been  described as  a doughnut  or  torus
                                                                   /
formation.    Based on  the  laboratory experiments,  we  are  able  to
estimate the gravity spreading and the dilution that  occurs.   Thus, we

-------
can  now  locate the cloud  front  and estimate  peak  concentrations that
will occur as a function of distance from release.
     In the last 10 years, a number of large field-scale experiments on
heavy gas dispersion have  been completed.   At  China Lake,  CA, the U.S.
Department of Energy (DOE) released around 30 cubic meters of liquefied
natural gas (LNG) onto water, forming a cloud (visible due to condensed
moisture) that moved  off downwind.   Shell Research,  Ltd.,  conducted
similar experiments at Maplin  Sands in England.  At  Maplin Sands,  the
material was spilled onto water and the dispersion occurred over water;
whereas  at  China  Lake,  the material  was  spilled on  water and  the
dispersion  occurred   over  land.     The   Thorney  Island   experiments
conducted  by   the  British  Health  and  Safety  Executive  involved  the
instantaneous  release  of  nominal   2000   cubic-meter  gas  volumes  of
heavier-than-air Freon™-air mixtures.  These  experiments  resemble  our
laboratory  experiments;  a  container  is  filled  with  gas  and  the
container (nearly instantaneously) removed.  At Thorney Island, a large
tent-like structure  was filled  with about  2000  cubic meters of test
gas,  and  the  side  walls  of  the  tent   were  dropped.    The  field
experiments showed  responses similar to the lab-scale  ones;  the dense
gas  fell down and rolled up in a doughnut shape.  With results of these
field experiments  corresponding  so closely to the  laboratory ones,  we
gained confidence that we  can  scale our laboratory  results to describe
field-scale events.    We  can  also  take mathematical  models  that  are
applicable  to  a  laboratory scale  and apply  them to  a  field  scale,
further  demonstrating  the   scaling  relationships  derived  in  the
laboratory.

                                  89

-------
     In the  last few years,  several  accidents  involving  catastrophic
releases of  hazardous gases have occurred.  Considering the  liquefied
petroleum gas  (LPG)  terminal disaster  in Mexico  City  in  1984, do  we
have mathematical  models that would  give us  a  reasonable picture  of
what occurred?    I  think the answer  is yes.   Considering the  methyl
isocyanate release  at Bhopal,  would  application  of available  models
draw a  picture  consistent  with  the experience?  I think the  answer  is
yes.   However,  accidents that are  difficult  to  model  (even  after the
                v
fact) have occurred,  and consideration  of these difficulties  leads  to
identification of areas  where additional research is  needed.
     An accidental  release of refrigerated  ammonia, occurred  during  a
river   barge-to-storage   tank   transfer   in  Blair,   NE,   in   1970.
Approximately 150  tons  of  ammonia  overflowed  onto the tank  roof,  ran
down the  tank wall,  and evaporated to  form a  large  cloud.   Reports  of
the  incident indicated  the visible cloud  extended 9000 feet  from the
tank.   Because  of the  prevailing  winds,  the cloud  missed the town  of
Blair.  This is an example of an  accident that  we do  not know  how  to
characterize very well.   The  release  of refrigerated ammonia exhibited
all  of the  characteristics of  a heavy gas  cloud  (the cloud was  very
wide and  shallow and hugged the ground),  and  it occurred  in near-calm
conditions.
     A  high-pressure ammonia pipeline was ruptured by a heavy equipment
operator  near  Enid,  OK,  several  years  ago,  and a wide plume path,
characteristic  of a  heavy gas  cloud was  observed.   These   accidents
raise  basic   questions   for which  we  do  not have   adequate answers.
Ammonia is a material that, because of  its molecular weight, would not

                                   90

-------
be expected to form  a  heavier-than-air  gas  cloud.   Even at the boiling
point,  saturated  ammonia  is  less dense  than  air.    There was  some
physical  mechanism  at  work  in  these  two  accidents  that  caused  the
ammonia  clouds  formed  to   be  heavier  than  air.     It  was  probably
associated  with  aerosol formation and evaporation,  which  cooled  the
mixture and  increased  its  density.   In any case, we  have a situation
where a heavier-than-air cloud  is  formed  under some  release conditions
and not under others.
     There  are  two  main  types of  dispersion  models:   mathematical
models  and  physical  (wind  tunnel)  models.  Mathematical  models  can be
subdivided into those using ordinary or partial differential equations.
Similarity  models  (involving ordinary differential  equations)  are the
mathematical models  that,  in my view, are  most  appropriate for use in
emergency  response applications.   We assume  a  regular  shape  for the
cloud,  talk  about  some integrated  values  that  assign the dimensions of
that  cloud,   and   calculate  how   those   things  develop  during  the
dispersion process.  That is, in fact, what Gaussian models do.  Three-
dimensional   mathematical    models  (involving   partial   differential
equations)  can,   in  principle,   simulate  the  spatial  and  temporal
dispersion  process  without  artificial   separation  of  the  flow  into
separate  regimes and may be able to provide for effects of terrain and
wake  turbulence.   Evaluation of three-dimensional  models is under way,
but most  of the  work is biased strongly toward evaluation against data
for dispersion  in  the absence of  terrain  and  wake turbulence effects,
even  though  description of  these  effects is  a  primary motivation for
their use.

                                   91

-------
     DEGADIS  (a  similarity  model  we  developed for  the Coast  Guard)
distills  what we  consider  the best  features  of several  heavy  gas
dispersion  models.    It  describes  the  dispersion of  denser-than-air
gases on  flat terrain and  accounts  for  steady or  transient  releases.
The DEGADIS model, which  was  based  on  laboratory-scale  experiments and
theoretical considerations,  has been applied to data from all  of the
major field-scale  tests:    the  DOE series,  the LNG  series, the Shell
Maplin  Sands  series,  and  the  Thorney   Island series.    With  these
results, a statistical analysis has shown a confidence level  with which
we  can  predict the  results of these  field tests.   For all  of these
field tests,  the ratios  of observed-to-predicted  maximum distance  to
the 1.0 percent  concentration level  ranged from 0.88 to  1.33  for a  99
percent confidence interval.   In  my judgment,  this degree of  accuracy
is  good enough for  most of the applications  we will  find in  emergency
response.
                                                 \
     DEGADIS  and  similar simple models  do have shortcomings  however.
They  cannot adequately provide for situations involving  obstacles  in
the  flow   field,  chemical  reaction  effects, or  phase change effects
(such as  formation of  aerosols, evaporation,  and rainout).   There have
been some  field tests conducted to address some of these questions.  In
the Thorney Island  tests,  some experiments were  conducted  with  (gas)
barriers  erected  around  the  release site.   This  particular  test was
aimed at addressing  the situation where a line  of trees forms  a  barrier
at  the  boundary of  a  hazardous material installation.   Simple models
also are  inadequate  in cases where a reactive gas is  released.   In such
cases, one may need  to consider the importance  of reaction effects.  An

                                   92

-------
example  is  the  series  of  field  test  releases  of  nitrogen tetroxide
conducted by the Air Force.
     For modeling large gas releases on flat terrain, when no important
chemical  reaction  or  phase-change  effects  are  involved,  we  have  a
number of techniques that  are  adequate  for risk  analysis and emergency
response  application.     Future  research  should  be  directed  to  the
verification of models that provide for the effects of terrain and wake
turbulence and  for  description of  jetting releases,  aerosol  formation,
chemical reaction, and deposition.
Note:   This  is  a summary  of the verbatim  transcript of  Dr.  Havens'
remarks at the workshop.  A complete copy of a paper of the same name,
which  was  presented  at  the  AIChE  CCPS  International  Symposium  on
Preventing Major Chemical Accidents,  February 3-5, 1987, in Washington,
DC, appears as Appendix D.
                                   93

-------
                      DENSE GAS DISPERSION MODELS
                            Donald L. Ermak
                 Lawrence Livermore National  Laboratory
                            Livermore, CA
     There  have  been  numerous  efforts  to model  the dispersion of heavy
gas releases and most of these have been reviewed  in at least one model
comparison  study  or  another.   Models  have been proposed  that  vary
considerably  in physical  completeness  and numerical  complexity,  and
generally,  these  two  model  characteristics  increase  (or  decrease)
together.
     The  models  that provide  the  most physically complete description
of  dense  gas  dispersion  are  those  that  are  based on the  three-
dimensional,  time-dependent  conservation equations.   Examples  of this
type  of  model  include FEM3,  SIGNET,  MARIAH,   and  ZEPHYR.    At  the
intermediate  level  of completeness and  complexity are the similarity-
type  models.   These  models use simplified forms of  the conservation
equations that are obtained  by averaging the cloud properties over the
crosswind  plane.   Quasi-three-dimensional   solutions  are  obtained  by
using similarity profiles, that is, by assuming a crosswind profile for
the concentration and other cloud properties.  Examples of this type of
model include  SLAB,  HEGADAS, and  DEGADIS.   At the  simplest  level  are
the modified Gaussian  plume  models.   These  models are usually used to
simulate continuous  releases and employ a  variety of  modifications to
include the effects  of dense gas dispersion within  the Gaussian plume
model for  trace  gas  releases.    Top-hat  models, which  are used  to
simulate instantaneous  releases,  fall  into  either the intermediate or
                                   95

-------
simple  category  depending upon the  complexity of the  model  regarding
the number of conservation equations to be solved.
     In  this  presentation,  a  brief  description of  the representative
model types  is  given with emphasis  on the FEM3 and SLAB  models being
developed  at  the Lawrence Livermore  National  Laboratory  (LLNL).   The
details  of the  various heavy  gas  dispersion  models will  not  be given
here except to point out the main differences  between them.
     The LLNL has been conducting research for a variety of sponsors in
the field of dense gas dispersion in the atmosphere.   This  research has
centered about a number of field-scale test series  that were  performed
in order to gather data  to  develop  and validate dense gas dispersion
models.  LLNL performed liquefied natural gas  (LNG)  field tests for the
U.S.  Department  of  Energy  in 1978  and 1980,  and  again in  1981  with
additional  sponsorship  by the Gas Research  Institute.   In  1983,  LLNL
performed  ammonia  dispersion  tests  for the U.S. Coast Guard  and The
Fertilizer  Institute  and  nitrogen  tetroxide spill experiments  for the
U.S.  Air  Force.    During this same  time period,  LLNL developed two
state-of-the-art atmospheric dispersion models called FEM3  (Chan, 1983;
Gresho  et  al.,  1984)  and SLAB (Morgan  et  al., 1983;   Ermak  and Chan,
1985).
     Both   of   these   computer   models  incorporate   mathematical
descriptions of  the physics  of heavy  gas  dispersion.    These  include:
1) gravity spread,   2)  effect of  density  stratification  on  turbulent
mixing,  and 3)  ground heating into  the cloud and its effects on density
stratification  and  turbulence.  Of  the two  codes,  FEM3  provides the
                                  96

-------
more detailed and complete description of the physics Involved In dense
gas flows.
     FEM3  simulates the dispersion  of a released  gas by  solving the
time-dependent,  three-dimensional,  conservation  equations  of  mass,
momentum,  energy,   and  species  along  with  the  Ideal  gas law  for the
equation  of  state.   In  addition,  it  can  treat  flow  over  variable
terrain   and  around  obstructions   such   as   cylinders   and   cubes.
Turbulence is treated by using  a  K-theory submodel.   Since it is fully
three-dimensional,  FEM3 can  simulate complicated cloud structures such
as:   1)  the vortices  that  are typical  of dense  gas flows, 2)  cloud
bifurcation  that has  been  observed  during heavy  gas releases  under
stable,  low  wind speed conditions,  and  3)  cloud deflection caused  by
sloping terrain.                                       \
     The  SLAB model solves  the  crosswind-averaged  equations for  the
conservation  of mass,  species,  momentum,   and   energy  along  with
additional equations for cloud  width  and the  ideal  gas law equation of
state.   The  current version of  SLAB also  includes the  steady-state
assumption for continuous  releases.   Thus, the code is one-dimensional
with downwind distance being the  independent  variable.   However, since
cloud width  and cloud height are also calculated, the model is,  in this
sense,    quasi-three-dimensional.      The   crosswind   concentration
distribution  is  determined  by  using  similarity profiles based  on  the
calculated crosswind height  and width.  Mixing of the cloud with the
ambient atmosphere  is  treated  by  using  the  entrapment concept.   The
main advantage  of  the SLAB  code  is  its low  computing cost.   Typical
                                  97

-------
simulations require only a few seconds on  a  CDC 7600 computer or a few
minutes on an IBM microcomputer.
     The various types of models  differ  considerably in their approach
to  simulating  the atmospheric  dispersion  of  a dense  gas  release.
Perhaps the most obvious differences are related to the degree to which
each type of model  incorporates the  basic  conservation  laws and three-
dimensional effects.  The modified Gaussian plume model  is based on the
single  conservation  of species equation and either  neglects  momentum
and energy transfer or attempts to  include them in some ad hoc manner.
On the  other  hand, the SLAB model includes  the conservation equations
of mass, momentum, and energy*  in  addition to the species equation, but
only  in  an  average  way.    Variations  in  the  crosswind  plane  are
neglected,  and  all  properties  of  the  vapor  cloud  are  expressed  as
crosswind  averages that  vary  in  the downwind direction  only.    The
conservation equation model  FEM3  includes the most complete description
of  the  conservation  laws  by   treating  them  explicitly  in  three
dimensions.
     A  unique  feature of  the SLAB  model is  that it  calculates  only
crosswind-averaged properties and characterizes the  cloud shape by the
height, h, and half-width, B.  The parameters B and h do not correspond
to any  particular concentration  level;  rather, they  can  be considered
to describe a surface that encloses  the  bulk of the cloud, for example
90 percent.   Consequently, the crosswind concentration distribution is
not  uniquely  defined,  making  it  difficult to compare  the predicted
cloud shape from  this model with  the contour  plots  obtained from the
experiments.   To  overcome   this  difficulty,  one generally  assumes   a

                                  98

-------
distribution  (such  as  Gaussian,  exponential,  or  quadratic)  for  the
crosswind horizontal and vertical vapor cloud concentration.
     There  are other  important  differences  related  to the  manner in
which  each  model  treats  the  effects  of gravity  and turbulence.   As
indicated above, the modified Gaussian plume models use ad hoc formulas
with  "empirical"  coefficients  to  describe the   gravity  spread  and
turbulent dispersion  of the cloud.  In  contrast to this,  the SLAB and
FEM3  models  use  conservation  principles   to  treat  the  effects  of
gravity.  This  is  done  in  the FEM3 model by solving the three momentum
conservation  equations,  including  the  buoyancy  term  and  variable
density,  while  the  SLAB  model  solves two layer-averaged  momentum
equations and uses the  hydrostatic approximation.
     These  two  models  differ   considerably  in   their  approach  to
turbulence.   The  SLAB  model  uses  the  somewhat artificial concept of
entrainment across the  cloud-air interface and essentially neglects any
explicit  treatment  of  turbulence  within  the  vapor  cloud.    Air  is
entrained  into the  cloud  at the  surface and  then is assumed  to mix
rapidly  in  the  cloud,  creating a nearly uniform layer in the crosswind
plane.   Thus,  there  are  two  separate  regions:   the  cloud  and  the
ambient  atmosphere.   Mixing  between the two is  assumed to occur at the
interface  and  is governed  by an entrainment velocity  that depends on
the local  properties of both  the cloud and the surrounding atmosphere.
The FEM3 model  assumes  that turbulence can be described as a diffusion
process 'and uses a continuous diffusion  coefficient that depends on the
local  properties of  the  dense  gas  cloud.    While  the entrainment and
diffusion   concepts  are   peculiar  to  the  SLAB  and  FEM3  models

                                   99

-------
respectively,  the  choice  of  a  particular entrainment  or  diffusion
submodel is  not  an  essential  aspect of the models.   Several  submodels
have been proposed in the literature and could  be used without changing
the whole model.
     Finally,    a  three-dimensional,   time-dependent,   conservation
equation model is required in order to describe the distribution of gas
concentration  in space  and  time  from  a  heavy gas  release  into  an
atmospheric boundary layer with speed and directional wind shear in the
presence of  complex terrain and man-made structures.   The versatility
of  the  three-dimensional  conservation equation  model  in  treating more
realistic  situations  and  providing a more  detailed  description of the
flow  is somewhat balanced by  the  increased computer costs  in  running
these  models.    Conversely,  similarity  models, while  giving  up  some
degree  of  realism  and  detail,  are much faster to run  on  computers.
Consequently,   at  LLNL  we   tend   to   view   the   three-dimensional
conservation  equation model  more  as  a  research  tool  that helps  us
discover new things about the flow, and the intermediate,  or similarity
model,  more  as  an  operational  model for  situations where  computing
costs and time are of the essence.
     Over  the  past  few years, the predictions from both  the FEM3 and
SLAB models have been compared with the data obtained from a variety of
field-scale experiments.  These  include the Burro  and  Coyote series of
LNG  dispersion  experiments  and   the  nitrogen  tetroxide spill  tests
conducted  by  LLNL and the refrigerated liquid propane tests conducted
by  Shell  Research,  Ltd., at  Maplin  Sands,  England.  The models have
been observed to perform quite satisfactorily within the limits of each

                                   100

-------
model and have been very helpful in Identifying the important phenomena
associated with dense gas dispersion in the atmosphere.
     Basically, three  major effects can be observed  in  the dispersion
of dense  gas clouds that  are  not observed in the  dispersion  of trace
emissions.   The first  effect is  a reduction  of turbulent mixing within
the  vapor cloud due to stable stratification of  the  dense layer.   The
second is the generation of gravity-spreading and self-induced vortices
due  to density  gradients  in   the  horizontal  direction.    The  third
effect,  cloud  lingering,   occurs  when  the  dense  gas  cloud  travels
downwind at  a  slower rate  than  the  ambient wind speed.  It is due to a
lack of mixing between  the dense gas  layer and the  ambient atmosphere.
All of these effects are most pronounced when the ambient wind speed is
low  and the  atmospheric stability conditions are stable.
     Perhaps the most dramatic example of these effects was observed in
the  Burro  8  LNG   experiment.    In  this  test,   the   dispersion  cloud
developed a  distinct bifurcated  shape  with two  lobes  on  either side of
the  cloud  center!ine.    This  bifurcated  structure   was   due  to  the
combined  effects  of gravity  spreading  of the dense  gas   cloud  and a
reduction  in  turbulence  allowing  this  structure  to develop  without
damping  out  the  bifurcated  shape.    In addition,   the   vapor  cloud
lingered over the source region for more than 2 minutes after the spill
ended.   In  all of  the other  experiments conducted  under higher wind
speeds and less stable  conditions, the cloud  drifted  downwind from the
source within  10  to 15 seconds  after the spill  had  ended.   The FEM3
simulation of  the Burro 8 experiment  simulated both  of  these effects,
namely  cloud  bifurcation  and   cloud  lingering,  quite   accurately.

                                  101

-------
Similarity models have been able to reasonably simulate the average and
peak concentration versus downwind distance for the Burro 8 experiment;
however, prediction of the cloud structure is beyond the scope of these
models because cloud structure is a basic assumption of the models.
     One  final  comment  is  important regarding  the use  of dense gas
models  in  operational  settings.  When  an accident occurs,  the  actual
amount of material released and the release rate are often unknown.  In
order  to  make  dispersion  predictions,   a  unit  source simulation  is
generally performed, and the resultant concentrations are then modified
when actual source  strength values are  obtained.   For  dense gas  flows,
the unit source simulation technique is generally inappropriate because
of nonlinear effects that are  occurring.   One cannot simply double the
predicted  concentration  because the  measured source strength was found
to be double that used in the original calculation.  Consequently, this
complication  of  dense  gas  flows  will  require  a different  way  of
planning for potential accidents.

REFERENCES
Chan, S. T.  FEM3 -- A Finite Element Model for the Simulation of Heavy
     Gas  Dispersion and  Incompressible  Flow:    User's  Manual.  UCRL-
     53397,    Lawrence   Livermore   National    Laboratory,   Livermore,
     California, 1983.  83 pp.
Ermak,  D.  L.,  and  S.  T. Chan.   A Study  of  Heavy Gas  Effects  on the
     Atmospheric  Dispersion  of  Dense   Gases.  UCRL-92494,  Lawrence
     Livermore  National  Laboratory,  Livermore,  California.    Paper
     presented  at the 15th NATO/CCMS International Technical Meeting on
     Air Pollution  Modeling  and  Its Applications, St. Louis, Missouri.
     April 15-19, 1985.   23 pp.
Gresho,  P.  M., S.  T.  Chan,  C.  Upson, and R. Lee.   A Modified Finite
     Element  Method  for  Solving  the  Time-Dependent,   Incompressible
                                  102

-------
     Navier-Stokes Equations:  Part  1  -  Theory;  Part 2 - Applications.
     Int. J. Num. Meth. Fluids, 4:557-598 and 619-640, 1984.

Morgan,  D.  L.,  Jr.,  L.  K. Morris,  and  D.  L. Ermak.   SLAB:   A Time-
     Dependent  Computer  Model  for  the  Dispersion  of  Heavy  Gases
     Released   in   the  Atmosphere.   UCRL-53383,   Lawrence  Livermore
     National Laboratory, Livermore, California,  1983.  15 pp.
                                  103

-------
           ON THE DEVELOPMENT OF REACTIVE, DENSE GAS MODELS

                Bruce B. Hicks and Will R. Pendergrass

            National Oceanic and Atmospheric Administration
             Atmospheric Turbulence and Diffusion Division
                            Oak Ridge, TN



     A summary  of  this  presentation was  not prepared at  the authors'

request.   The  abstract  of a formal paper is given  below.   The text of

this paper can be found in Appendix C.
          Chemical  reactions  between   released   pollutants  and
     atmospheric  constituents  can   be   affected  or  mechanical
     turbulence   damped.      If   the  reaction   is   sufficiently
     exothermic,  plume rise  can  be  increased;  if  sufficiently
     endothermic,  then plume rise  can  be  suppressed.     It  is
     necessary  to  consider  the  potential   influence  of  such
     reactions  on  the mixing rates  characteristic of the ambient
     air.  Reaction rates  are also a critical  issue.   If reactions
     are  completed  rapidly,  then  in  concept  it  is  possible  to
     assimilate the  consequences  of  the reactions in  the  source
     term initialization or in the  first time  step  of a numerical
     model.   On  the  other hand,  if reaction  rates  are slow  in
     comparison with the time step  of  a  simple model, then  a more
     complicated and detailed modeling effort  may be required.
                                  105

-------
               DISPERSION MODELS FOR NEUTRALLY BUOYANT
                     AND POSITIVELY BUOYANT GASES
                           Thomas E. Pierce
         U.S. National Oceanic and Atmospheric Administration
               and U.S. Environmental Protection Agency
                      Research Triangle Park, NC
     It is my  good  fortune to be involved with  buoyant plume modeling
which  seems  to  be  much more  manageable than  reactive and  dense  gas
modeling.     I   am  assigned  as  a  National  Oceanic  and  Atmospheric
Administration  (NOAA)  meteorologist  to  the Environmental  Operations
Branch of EPA.   Our branch helps to provide a technology transfer link
between the  research side  and the regulatory side of  EPA.   One of our
jobs is to develop,  adapt, and evaluate atmospheric  dispersion models.
These models tend to be  operational  and not research-grade models like
those  presented by others  at this workshop.   We  make  these models
available  through  the  User's Network  for  Applied  Modeling of  Air
Pollution, which is  commonly known by its acronym,  UNAMAP.
     Since its  beginning in  1973,  hundreds of people  have used UNAMAP
either through  EPA's UNIVAC  computer or through  the  National Technical
Information Service  (NTIS), where computer tapes can be purchased.  The
most recent version  of UNAMAP, Version 6, was released in August, 1986,
with more than 23 models and programs.
     Today, I would  like to focus on three programs on UNAMAP that were
developed by the Environmental Operations Branch:
     •    TUPOS - Dispersion estimates using on-site turbulence data
     •    MPDA -  Meteorological processor for dispersion analysis
     •    INPUFF -  Integrated puff model.
                                  107

-------
     TUPOS  represents  one  of  the   first   operational   models  that
incorporated  on-site  turbulence  data.   It  operates  like  a  routine
Gaussian  plume  model   in  its  steady-state  assumptions.     TUPOS,
therefore, deals only with continuously emitting sources.
     In  addition  to  handling on-site  turbulence, TUPOS  has  several
other special features.   It  incorporates layer-by-layer plume  rise and
partial penetration of a  plume into an  elevated temperature  inversion.
It  also  incorporates  recent  ideas  by  Briggs  on  the modeling of .a
buoyant plume in a convective boundary layer.
     The modeling of partial penetration is particularly  important for
some sources.   Current  routinely  used Gaussian  plume models assume an
all-or-nothing approach to plume penetration.  If  the  centerline of the
plume  is  predicted  to remain  below the elevated inversion, then  100
percent  of  the plume  is  estimated  to  remain  in  the  mixed  layer.
However,  if  the centerline  is  predicted to  be above the mixed layer
(even by  1 meter),  then  the entire plume is  assumed to loft above the
mixed layer.   Clearly,  partial penetration is  an improvement  that we
feel should be included in Gaussian plume models.
     Another  important  development   is  the  modeling of  hesitant  or
bumping plumes.  Willis  and Deardorf noted that  when  a buoyant plume,
which  is  in  a convective boundary   layer,  encounters  an  elevated
inversion,  that it  can  bump against  the  inversion and  spread  out
laterally.   In  the  Journal of Climate  and Applied Meteorology. Briggs
discussed a  method  for predicting diffusion  for  this situation.   The
most recent version  of TUPOS, Version 2, includes  this methodology.
                                  108

-------
      It  is  hoped that with models like TUPOS we can do  a  better  job  of
modeling  atmospheric pollutants,  and certainly there is plenty of  room
for  improvement  in  this  area.    If  we  are  to use  more complicated
models,  however,  we  should  have  a  means  for  incorporating on-site
meteorological data  and estimating similarity parameters.
      Based  on  ideas  of  Irwin,  we  have  developed  a  meteorological
processor for dispersion analysis (MPDA).  MPDA allows  input of three
types  of meteorological  data:    upper-air  data from  National Weather
Service  stations, surface  observations  from National  Weather Service
stations,  and on-site data.   MPDA  processes  all  of these  data  and
produces  a  merged or  clean data  set.   The processor also determines
several  parameters,   such  as  the  solar  angle, friction  velocities,
Monin-Obukhov  length,  turbulence  profiles,  and vertical  variation of
wind  speed  and wind direction.   A great deal  of  flexibility has been
programmed  into  this  model.    For  many  of  the  parameters,  several
options  are  available.    For example,  there  are  three   options  for
specifying  the  sensible  heat  flux.    They  are  the  Irwin-Binkowski
scheme,  the Holtzlag-van  Ulden  scheme,  and direct  determination from
on-site turbulence measurements.
      INPUFF,  developed  by  Peterson,  resulted  from  EPA's  need  to
evaluate dispersion  from  incinerator ships.   The problem with modeling
incinerator ships was  that a steady-state Gaussian  model  could not be
used.  Therefore, the integrated puff model was developed.   In recent
years, demand  for INPUFF  has grown  as people  have  sought programs to
model toxic gas releases.
                                  109

-------
     INPUFF  simulates  individual   puffs   from  one  or  more  sources.
Although many  sources  can  be included, the model's  speed  decreases as
the number of time steps increase.
     Briefly, INPUFF works in the following manner.  At  the first time
step,  a puff  is  released  and  disperses  according to  the  specified
meteorology  for  that  time  step.   At  the next  step,  another  puff is
released so that it and the previous puff disperse according to the new
meteorological conditions.   INPUFF  proceeds  in this  step-wise manner
and is able to calculate contributions at specified receptors from each
puff for each time step.  Hence its name,  the  integrated  puff model.
     Inputs  to  INPUFF  include  miscellaneous  options,  modeling region
specification,  and  source  characteristics.    Source characteristics
include emission  rate, time of emission rate,  and  how the  emissions
vary with time.   Receptor  points and meteorological data  are included
as inputs.  Dispersion  algorithms included in the  model  are provisions
for initial dispersion, short travel-time  dispersion,  and  long travel -
time dispersion.
     Currently, the model  is designed so that meteorological  data are
entered for  every time step  for a  user-designed  meteorological  grid.
Eventually we hope that a processor such  as MATHEW can be  front-loaded
into INPUFF to allow a three-dimensional  wind  field projection.
     In reviewing  these three  programs  that  are  available  on UNAMAP,
hopefully,  I  have presented  some  ideas  that may be  useful  to those
involved in the modeling of toxic gas releases.
                                  110

-------
REFERENCE
Briggs,  Gary  A.    Analytical   Parameterization  of  Diffusion:    The
     Convective  Boundary  Layer.    J.  Climate Appl. Meteorol.  24:1167-
     1186, 1985.
                                  Ill

-------
                         LAB-SCALE EXPERIMENTS
                           Robert N.  Meroney
                 Fluid Mechanics and Wind Engineering
                     Civil Engineering Department
                       Colorado State University
                           Fort Collins, CO
     As a physical modeler in a room full  of numerical  modelers, I must
confess I feel like a chicken in a fox house.  So,  I will  do the best I
can, and you  must decide whether or not you are convinced.   I  want to
talk about the capabilities  of  physical fluid  modeling  with respect to
meeting the needs of the hazardous materials community.   There are many
experiments that can be classified as basic fluid mechanics experiments
associated with  the mixing  process.   There  are tests that  have been
performed  for the  meteorological  community associated with  pre-field
test  planning  experiments.    There  are   concept  testing  experiments
where,  for example,  a  certain  mitigation  device  can  be tested  for
feasibility.   There are  validation  experiments that can  be  performed
specifically  to  determine  whether  physical  modeling can provide  a
viable approach to solving some particular problem.   Finally,  there are
experiments that  have  been  directly  applied  to hazard analysis.   The
goals of all of these experiments is to test, calibrate, and validate a
numerical code.
     One might  note that  if a code  is not capable of predicting the
behavior of an  idealized laboratory-controlled  experiment,  one should
not feel the  right  or  the ability exists  to predict a far more chaotic
field experiment.   As  a proponent, I will  say a couple of words about
the advantages of fluid  modeling.   Wind/water  facilities  are in effect
                                  113

-------
analog computers,  and  they have special  attributes.   They  have  near-
infinitesimal resolution.  Grid sizes are not a concern.   The transport
processes,  one could  argue,  go down  to the  molecular  level.    The
facilities have near  infinite memory.   They also have the  ability to
look  at  very  large three-dimensional  grid  regions.   Fluid  modeling
incorporates real  fluids,  not models of fluids.  We,  therefore,  start
out using the  right stuff  in  the right  place,  not someone's  concept of
how the atmosphere or fluid  behaves.  Implicitly, this analog  computer
is  nonhydrostatic,  non-Bosinnesque,  capable  of compressible  effects,
thermal effects,  and  includes variable  property.   It includes a  non-
slip  boundary  condition,  effects  of dissipation,  and many  nonlinear
processes.   Also, it  inherently includes full  conservation  equations
without truncation.
     On the down  side, there  are some limitations.   At smaller scales,
one  must  recognize  that  some  similarity   is  lost  in  the  mixing
processes.   As speeds drop  to handle stratified fluids,  the  Reynolds
numbers decrease,  and it is  possible that  the  Reynolds number can drop
below  some  critical  value in  different  phases of the mixing  process.
Depending on the  phenomena,  this  may  result  in  a minor or  a  major
error.   When  running  experiments  at  very low  speeds,  one reaches  a
point, perhaps associated  with  the  fluid number dominance, where the
ratio described as the  Peclet number over the  Richardson  number first
proposed by Colenbrander and Puttock from Shell Research,  Ltd., is less
than  some  critical   value.    This  means,  basically,  that  you  are
operating at  such  a  low  rate  of  mixing  that  the molecular mixing
exceeds the scaled turbulent mixing.  This is  the only phenomenon that

                                  114

-------
tends  to give a  nonconservative  result in physical  modeling.  Almost
all other errors  tend to give conservative results.
     As  smaller  scales are  used  in the laboratory,  the separation we
are familiar with  in the atmosphere between the integral scales, Taylor
scales,  and Kolmogorof  scales  of  turbulence,  will  bunch up.   Depending
on  the  kind  of  mixing  process under  study,  this  may or may  not be
important.
     There are a  number of basic  fluid mechanics  experiments one might
wish to  consider.   Anyone  who  has produced a numerical model that uses
an  entrainment  rate at  a  box-model or  a  slab-model level or who has
worked  with  various  K-theory-type models   has   probably  drawn  from
physical  modeling  for  basic  turbulence  coefficients.    Some  of the
earliest work by  Lofquist,  associated with overflow of fresh water over
saltwater, generated  the  information  we use today  on  the entrainment
variation of Richardson numbers.   Basic  experiments  by Kantha, where a
surface  plate was dragged  around and  around on top  of a  circular
channel   filled   with   stratified  salt  water,  the  merry-go-round
experiments, have provided us with additional information.  Some errors
were   later   found  in  these  experiments,   and   meteorologists  did
experiments  to  improve  this work (Willis  and  Deardorff  and,  later,
Lindberg).   In  England,  McQuaid  has  done  some  basic  experiments on
transport  through dense shear  layers  associated  with  carbon dioxide
(C02)  releases.    Jerry  Havens   and   I  also  have  done  some  basic
experiments.
     We should consider  some simple idealized cases.   Yang,  a student
of mine  in 1972,  and  I  did some laboratory puff  model  experiments,

                                  115

-------
where we released  thousands  of puffs of gas  in  a  boundary layer shear
flow and were able to develop  models for a  transport  similarity that
included  probability  density  distribution   and  various  coefficients
associated with  lateral,  vertical,  and longitudinal  variation.   These
models and the coefficients very closely agree with models we are using
today.     In   1980-81,  Lohmeyer  and  I   did  some  experiments  with
instantaneous volumes of dense gases, releasing them at the wall  of the
boundary layer.  At that time, there were no field experiments to guide
us, but we identified the basic characteristics of the dense gas  cloud.
The  behavior  of  the  arrival  time,  the  departure  time,  and  the
statistical deviations  within  a  multiple sum  average  condition of  a
torus-shaped cloud, I believe,  are  very  important.   These are critical
points  for  issues of  flammability  and  toxicity.    We  cannot  make
decisions  based  on  average  conditions,   but  must  know  about  the
statistical range of conditions that  exist.   A gas cloud is not  set on
fire by  the average  conditions that exist.   It  is  the instantaneous
concentration that sets the fire.
     There  have been  several  somewhat complex  studies  done  in  the
laboratory, including  both dense gas cloud  effects  and  heat transfer
effects.   The  best  known  of  these is  the tank  experiments  on  the
convective boundary layer concept by  Willis  and  Deardorff.   These were
heated water  experiments  that have  had a  revolutionary  effect on both
numerical  modeling  and  our  field  understanding   of  the  convective
boundary layer.  One  of the  key points  from these experiments is that
we now know that clouds released through  plumes at the ground tend, in
a short distance in a convective boundary layer, to rise.  Also,  clouds

                                  116

-------
released at elevated points  in  the  convective boundary layer can fall.
Maybe on average this point is not so important, but it can make a very
big difference in how you  should  be calculating the results of a toxic
or flammable cloud passage.
     What kind of pre-field planning experiments are possible?  We have
performed  pre-field  tests  studying  the  effects  of field  terrain,
windward  variation,  and  stability.    We  have  looked  at  instrument
placement and cloud extent.
     Post-field test experiments also can be  useful.   Recently in some
DOE and Gas Research Institute-funded work preceding the 1987 LGF vapor
barrier tests, we found  some  unexpected  things  occurred.   In the vapor
barrier-contained region,  we found  that the  gases sloshed  up against
the  end,  hit the  barriers like a  wave, were  caught  in the  air flow
passing  over  the  barriers,  and  were  transported downwind  at  fair
heights.   Thus,  we found  bursts  of  higher  concentrations at  higher
elevations than were observed in the numerical experiments.
     Physical  fluid modeling also  can play a  very  useful  role  in
testing of concepts.  When the  gas  industry was looking at alternative
secondary  containment  schemes  for  large  tank storage,  there  were
questions about berm  heights  for  large volumes of cryogenic chemicals.
There  were  questions  also  about  whether soil  surfaces or  insulated
concrete  surfaces   were  better  inside  the   berms   and   about  the
comparative  effects of  these  choices   on the  eventual dispersion  of
resulting gases.   With  physical  fluid  modeling, we were  able to test
many options  and  learn  not only the  answer to those questions, but at
                                  117

-------
that time due to the early  nature  of those experiments (mid 1970s), we
were able to learn new fluid flow fundamentals as well.
     In  some work  with  Factory  Mutual,  funded by  the  Gas  Research
Institute  (GRI),   we   looked  at  water  spray  curtains   and  how  they
mitigate cloud dispersion.  In this case, simultaneous field tests were
run  that  were very   confusing  due  to  the  unexpected   results  that
occurred.   Our work on nozzle sizes,  varying water pressure,  and wind
speeds  helped  to  explain the nature and  extent  of  mitigation  that did
occur in these experiments.
     As  mentioned  earlier,   idealized  experiments  have  been  used
extensively  to calibrate  modules  of  various  models.   Some  numerical
models  have  been  calibrated against  both laboratory and field  results.
Physical  fluid modeling  can  serve as  a useful method  for evaluating
Federal  regulatory-specified  accident  scenarios at  existing or planned
facilities.   For  example,  we were  able to model  such  a scenario,  a
guillotine  of  a pipeline with 10 minutes  of  spill  at the maximum flow
rate, for  the  Brooklyn Union  gas  storage facility on Long Island.  The
experiment  showed a  positive result,  that no  effects  would  be  seen
beyond  the  facility  boundary in  the  event of that accident.   Since
there were resources  remaining  in the  contract,  they had us  look at
larger,  even  less  likely  potential  accidents,  such as  spilling the
entire   tank.    This  time  the  results  were  potentially  far  more
catastrophic, even though unlikely.
     Wind   tunnel   measurements  can   be  very  helpful  during  risk
assessment,  licensing,  or the  regulatory process.    DOT regulations
currently  require  the use  of an  extremely over-conservative numerical

                                   118

-------
algorithm.    The  model  does  not  account  for  roughness,  obstacles,
terrain, or mitigation devices.   To prepare for licensing hearings for
a  liquefied  natural  gas (LNG) peak-shaving  facility on Staten Island,
we  examined  spills   at  about  a  1:250 scale  model.   We  introduced
conservatism  into the experiment by  simulating larger, heavier spills
in  an  environment with reduced mixing.  The facility appeared to meet
DOT  requirements even when  significantly  more  extreme conditions were
considered than  required by DOT regulations.
     Recently,   field/laboratory   validation   experiments   have   been
completed  for both instantaneous and continuous  releases  of dense and
cryogenic gases.   In  work  for the GRI (Meroney, 1986), I examined some
26  field/laboratory  data sets and  found that the laboratory-predicted
distances  to  lower flammability  limit  (LFL) on the  average to within
0.4  percent of actual  values with a standard deviation of ±22 percent.
Pattern  comparison plots  of concentration  isopleths could always  be
matched  by  appropriate fluid modeling techniques with  less  than  a 15°
shift  in surface patterns.   The  British  Maritime Technology  group  in
the  United  Kingdom scaled  the recent Thorney  Island  field spills  of
     TM
Freon  -air mixtures.  They  found no  apparent lower  limit  for Reynolds
number  or  Peclet/Richardson  number-scaling criteria  for  collapsing
dense clouds.
     It  is  now  apparent that  fluid modeling can  faithfully reproduce
the physics of transient dense gas cloud entrainment and motions within
the  inherent   variability  of   the   process  for   many   interesting
situations.   Fluid modeling can  contribute  valuable input information
                                  119

-------
for  future  siting  and  risk  analysis models  for  the  chemical  and

petroleum industries.



REFERENCE*
Meroney, Robert  N.   Guideline for Fluid Modeling  of Liquefied Natural
     Gas  Clouds, Volume  II:   Technical  Support  Doc^ent.   Colorado
     State  University,  Civil  Engineering  Report  CER84-85RNM-50b,GRI
     86/0102.2.    Gas  Research  Institute,  Chicago,  Illinois,  1986.
     266 pp.
*Author's/Editor's  Note:    All  of  the  work  by  other  researchers
mentioned  in this presentation are referenced in the above citation.
                                   120

-------
       LARGE-SCALE  EXPERIMENTS OF THE DOE LIQUEFIED  FUELS  PROGRAM
                           Ronald P. Koopman
                 Lawrence  Livermore  National Laboratory
                             Livermore, CA

     The  accidental  release of  hazardous gases  into the  environment
often results in what has come to  be  known as dense gas behavior, even
when  the  gases  themselves may   be   nominally  less  dense  than  the
atmosphere into  which they  are  released.   This dense gas behavior can
dominate  the  consequences of the  accidental  release making  it  either
worse or  better  than might  otherwise  be expected  and making prediction
of  the  consequences of the  release difficult.   There are  a  number of
contributing  causes  of  dense  gas  behavior,   including  large-scale
release,  low temperature,   flashing,  chemical  reactions,  evaporative
cooling,  and  molecular  weight.   Cloud behavior differences are  due to
density  or gravity-induced  effects (turbulence  damping);  thermodynamic
effects   (aerosol   formation  and   flash   vaporization   upon   release,
evaporative cooling,  or heat transfer from the ground  modifying cloud
buoyancy  and turbulence);  and chemical  reaction  effects  (hydrolysis
with  atmospheric water  vapor,  polymerization, or decomposition, which
also  affects  cloud  density).  These effects are very important close to
the release  point,  but  also can have  dramatic effects  on the cloud as
it disperses  downwind.
      The  study of  atmospheric dispersion  of dense gases began as a new
scientific field in the  1970s.   Interest was prompted by  a  series of
accidents:    Flixborough  in  1974,  Seveso in  1976,  and  the  Houston
ammonia  accident  also  in  1976.   The  interest in  this field has  always
                                  121

-------
been more  intense in Europe  than  in the United States.   As a result,
much of the research has been done in Europe.  Early work in the United
States centered around liquefied natural gas (LNG) safety, and this was
when  Lawrence  Livermore  National  Laboratory  (LLNL)  got  involved,
starting in 1976.  Many people at this conference got involved at about
this same time with the LNG safety program.
     Work   on   dense  gas   dispersion   continues  to   involve   three
complementary areas:
     •    Mathematical modeling based on physical laws
     •    Near full-scale field experiments to:
          1)   Discover unknown and important effects in scaling laws
          2)   Simulate accidents or evaluate mitigation equipment
          3)   Validate models
     •    Scaled-down simulations  in the  laboratory,  wind  tunnel,  or
          water flume.
We  have  discussed  all  of  these  at this  workshop except  large-scale
experiments on which I will focus my remarks.
     There  have  been a number of  recent large-scale field  tests  with
dense  gases.     I   will   focus  on   the  most  recent  and  the  best
instrumented.   These are  the tests that  are currently  being  used in
developing  databases for model comparisons  and model  validation.   Most
of  the  good  quality  data   have   been  obtained  since  1980  when,
essentially  simultaneously,  Shell/NMI  conducted a series  of LNG  and
liquefied petroleum  gas  (LPG) trials at Maplin Sands  in  England,  and
DOE/LLNL conducted  the  Burro  series of LNG  tests at  China Lake,  CA.
These  tests  resulted  in   the discovery   of  previously  unknown  and
important effects.   The Maplin Sands tests  involved  spills  of about 5
to 25 cubic meters of LNG  with spill  rates between instantaneous and 1
                                  122

-------
to 5 cubic  meters per minute.  DOE's  Burro series,  conducted at about

the same  time,  focused on dense gas dispersion  and  involved spills of

24 to 39  cubic  meters  at  rates  of 12 to 18 cubic meters per minute.  A

follow-up series  of DOE  tests conducted  in  1981  was called  Coyote.

These tests were  designed to  get  information  on  three different areas:

combustion,  dispersion, and rapid-phase transition explosions.

     There  are  a  number of reasons for doing  field-scale experiments.

Many of the effects  that  occur upon accidental  release of dense gases

are poorly  understood.   In  some cases, the effects  have  been observed

in the  laboratory,  but the extent  of  what will  occur or  what  will  be

important or dominant in accidental  releases is unknown.   The important

physical effects  in dense gas dispersion include:

     • Gravity spreading      Low cloud
                              Displaced ambient wind field
                              Self-induced vortices
                              Increased surface area
                              Increased wind shear
                              Increased heat transfer

     a Turbulence damping     Stable stratification
                              Impenetrable interface
                              Negative turbulent mass flux
                              Evaporating  droplets

     t Heat  transfer          Increased ground contact
                              Decreased air entrainment
                              Enhanced evaporation and condensation
                                for cryogenics spilled on  water
                              Chemical  reactions with water vapor or
                                other species

     • Turbulent entrainment  Turbulence generated by the  release
                                (internal)
                              Atmospheric  turbulence  excluded
                              Heat transfer generates convective cells
                              Jet  releases  entrain air
                              Instabilities
                                  123

-------
     Additionally, much of the physics dominant in dense gas dispersion
is highly nonlinear  and,  therefore,  not easily scaled  to  larger size.
Very  stable  atmospheric  situations  and  nonisothermal  (i.e.,  heat
transfer, condensation,  etc.) effects  can  often  be important  and  are
also  not easily  scaled.    In many cases,  dominant behavior  at large
scale was not even observed  in  small-scale  tests.   For instance, dense
gas  behavior is  not observed  for  small  releases  in   the  atmosphere;
turbulent dispersion dominates.  Thus,  two of the major  reasons  for
performing  these tests  are  the discovery  of unknown and  important
effects  and  determination   of  their  scaling   laws.     These  well-
instrumented tests allow quantitative measurement of these  effects  so
that  correct descriptions can  be incorporated  into the  mathematical
dispersion models.  This is necessary if accurate  predictions are to  be
made  for circumstances  different  than  those  under which tests  were
performed.    As  you  can readily  see,  dense  gas  dispersion   is  very
complicated.
     Other  reasons   for  conducting   field  tests  include   accident
simulation or evaluation of mitigation equipment such as water or steam
curtains.   These can  be situations  that are  simply too  complicated,
with  too many   unknown  contributions  to  yield  to  mathematical   or
physical  model   simulation.     Other  complicating  effects  that  are
difficult to model or scale include  chemical  reactions  and  certain non-
reversible thermodynamic effects, such as  flashing, two-phase flow.
     Perhaps the most common  reason to  conduct large field experiments
has been to obtain   basic data  for  dispersion  model validation.   This
requires extensive,  carefully verified,  quantitative  data  from well-

                                  124

-------
documented and  well-instrumented experiments.   The  need  for such data
was demonstrated by McQuaid of the British Health and Safety Executive,
in 1983 when he  invited  the  mathematical  modeling community to predict
in advance the  results of an instantaneous  release of 200 cubic meters
of gas  with  an initial   density two  times  that  of air.   Variations
between models  of two orders  of magnitude  were  present  even  for this
relatively small, isothermal  release onto flat terrain without chemical
reactions, thermodynamic  effects,  or  other  complications.   Models have
improved  since  that time, but the tasks we  are giving them  are much
more difficult  and  include many  more  complicated  effects.   Thus,  model
validation continues  to  be  one  of the major reasons  to  conduct  field
experiments.  In spite of the number of tests that have been conducted,
data for model validation are still in short supply.
     There are  many difficulties associated with  the conduct of large-
scale field  tests.   They involve  huge instrument  arrays  with  hundreds
of   individual   instruments   and   their  attendant   maintenance  and
calibration.   These extensive arrays  are necessary because gas  cloud
behavior  is  highly variable  in  space  and time and  complicated by the
presence  of the  many simultaneous and  interactive effects  mentioned
earlier.
     The  large-scale, well-instrumented  spill  tests  that have been
conducted since the 1970s have contributed much to our understanding of
dense  gas  dispersion.    Dense  gas  characteristics  are  particularly
important  for   large  releases,  low  wind   speed  releases,   stable
atmospheric  conditions,   low-level inversions,  and  proximity to  the
release point.   Proximity to  the  release point  is  very  important for

                                   125

-------
dense  gas releases.    In our  ammonia release  tests,  we were  still
measuring dense gas  effects  at  a distance of 3  kilometers.   Dense gas
behavior  is  caused  by large size, low temperature,  flashing  two-phase
flow,  chemical  reactions, evaporative cooling,  and molecular  weight.
Many of these effects are nonlinear,  and,  therefore,  it is important to
base calculations and predictions on  size.
     Not  all of  the  things that people do can be  scaled,  depending on
which  dominant  effects are  of  interest.   Dense gas  releases  exclude
ambient  atmospheric  processes.    Gravity  spreading   and  turbulence
damping   decrease  dispersion,   while  heat  transfer   and   turbulent
entrainment  can  increase  it.   Dense  gas  effects  do  have  profound
influence  on our  ability to  predict the  consequences of  accidental
releases.
     In  summary, there are a number  of  reasons for doing field-scale
experiments.   Many  of the effects that  occur upon  accidental  release
are  poorly understood.   In some cases, they have  been observed in the
laboratory, but the extent of what will occur or what will be important
or dominant  is  unknown,.   In addition, much  of the physics dominant in
dense  gas dispersion  is  highly nonlinear  and,  therefore, not easily
scaled to  larger size.   Thus, two of the major  reasons for performing
these  tests  are the discovery  of  unknown  and  important effects  and
determination  of  their scaling  laws.   Well-instrumented tests  allow
quantitative measurement  of  these effects so that  correct descriptions
can  be incorporated   into  the mathematical  dispersion models.   Models
are necessary if accurate  predictions are to be  made for circumstances
different than those  under which tests were performed.   Other reasons

                                  126

-------
for conducting field tests include accident simulation or evaluation of
mitigation  equipment  such as  water or  steam  curtains.   These  can be
situations  that  are   simply  too  complicated,  with  too may  unknown
contributions, to  yield to mathematical or  physical  model  simulation.
Other complicating effects that are difficult to model or scale include
chemical reactions and certain nonreversible thermodynamic effects such
as flashing two-phased flow.  Perhaps the most common reason to conduct
field experiments  has been to  obtain  basic data  for dispersion  model
validation.   Today's  sophisticated models  require extensive,  carefully
verified,  quantitative  data  from well-documented and well-instrumented
experiments.
                                   127

-------
           FLUID MODELING OF DENSE GAS DISPERSION OVER A RAMP
                           William H. Snyder
                  U.S.  Environmental  Protection  Agency
          and National Oceanic and Atmospheric Administration
                       Research  Triangle Park, NC

     Recently,  we completed our  first  foray  into dense gas dispersion
at the  U.S.  EPA Fluid  Modeling  Facility at Research Triangle Park, NC.
Rex Britter  from the University  of Cambridge  drew up the plans for the
first research  study.   Unfortunately, he  could  not  be here to present
the work today.
     We  have  drawn upon  our  previous  experiences  in modeling  and
simulating dispersion  in  the neutral  atmospheric boundary layer.   This
work has focused mostly on ground-level and elevated point sources.  We
have  also  done  numerous  studies  in  the  wind  tunnel  on  the  flow
structure and  dispersion around  isolated  hills  of simple geometry and
studies  in  the  towing  tank of  flow  structure  over  and  dispersion  of
plumes within stratified flow over isolated hills.  The new addition to
these programs  is our subject today, dense gas.
     One of  the biggest  problems we have encountered was  alluded  to
today by Bob Meroney.   That is  the fact that  when you simulate a plume
that has a density  different from that  of its environment,  you need to
do  Froude  number  scaling,   which means you  need to reduce the  w.ind
speed.  By reducing both the scale and the wind speed, you have reduced
the Reynolds number, which  means  that what would have been a turbulent
flow in the wind  tunnel  may now  become  a laminar flow, irrespective of
the density difference.   In  addition,  if you  are modeling a positively

                                  129

-------
 buoyant  plume,  i.e.,  a destabilizing  influence,  you  may be  able  to
 model  that at a smaller-than-normal Reynolds  number.   However, in  the
 case  of a dense gas  plume,  the  dense  gas has a stabilizing  influence.
 This  stabilizing  influence  of  the  density gradient plus  the reduced
 wind  speed due to  the Reynolds number makes the simulation of dense  gas
 dispersion in the  wind  tunnel doubly difficult.
      Our   first   exercise  in  the  wind   tunnel   was   thus  a  flow-
 visualization study.  We used carbon dioxide  (C02) as the effluent in  a
 wind  tunnel boundary layer, which  had previously been  used  for point
 source  studies.   Me fixed the mass flux of C02 at the rate planned  for
 later quantitative studies.   Smoke was injected in order to  be able to
 see  the plume.  We visualized the  plume  in the tunnel  at various wind
 speeds,  watched  the  plume  to  see that  it  remained   turbulent,   and
 progressively reduced wind speeds  until  the  flow became  laminar.   We
 found that the plume  remained  fully turbulent to wind speeds as low as
 1.0  meters per  second.   Thus,  we  selected  this  wind  speed  for   the
 remaining  quantitative  tests.    We  have prepared a videotape of these
 flow  visualization tests.
      Our  source  configuration  consisted  of  a metal  can filled  with
 gravel  up  to  the  level  of the tunnel   floor  to maintain a homogeneous
 flat  surface   across  the  source.    The  CO*  was  emitted through  the
 gravel,  and  was thus  emitted  in  a  circular area  through the  flat
 surface.   We  placed a honeycomb  at  the entrance of the test section of
the wind  tunnel  to stabilize the  flow.  A straight fence  served  as  a
trip to generate a simulated neutral atmospheric  boundary layer.   From
previous  experiments,  we  knew  that   this  was  a  reasonably  well-

                                  130

-------
developed  boundary  layer*    The  source  was  placed  about  8  meters
downstream of the fence.
     For an initial series of tests, we emitted C02 and/or air from the
source and then measured the detailed structure of the plume downstream
in  longitudinal,  vertical,  and  crosswind  or lateral  concentration
profiles.   Later  a  ramp was  included  in one  of  two positions either
1800   millimeters   downstream  from  the  source  or  600  millimeters
downstream from the  source.
     Looking  at  the boundary  layer  structure of these experiments, we
found  that  for mean  velocity  profiles  between  1  and  4  meters per
second,   the  flow  is   virtually  independent  of  Reynolds   number.
Longitudinal  turbulence intensity profiles also show that structure to
be virtually Reynolds-number independent.
      For the quantitative portion of the  study, about 3  percent ethane
was added  to the source,  so that  concentrations  could  be measured
downstream.   These measurements,  again,  show  that  at  1  and 4  meters per
 second the concentration and  dispersion  field in  the wind tunnel were
 independent  of  Reynolds number.    Much  additional  data on  the  flow
 structure are available due to this and  previous  work.   There  are hot-
wire  anemometer  measurements of  mean  velocity profiles,  flow angle,
 longitudinal fluctuating velocities, and other parameters.
      Streamline  patterns  derived  from   mean  velocity  profiles  show
divergence of the flow from upstream to  the  base  of  the  ramp,  followed
 by convergence  of the  flow  from the  base  to the  top, and a slight
divergence downstream from there.
                                   131

-------
     Using  C02  (52  percent  heavier  than air)  with  smoke added  for
visibility, we have demonstrated that for flat terrain with free stream
wind  speeds  of  2  and  4  meters   per  second,   plume  buoyancy  is
unimportant.   However,  at lower wind speeds,  strong  lateral  spreading
of the  dense  plumes was observed.  Thus,  we  conclude that under these
conditions, a significant buoyancy effect exists.
     Repeating   the  above  experiments   with  a  ramp  placed  1800
millimeters downstream of  the source,  a significant widening  of the
plume  is seen both  in the neutral plume case and in the dense plume
case.    When  the wind  speed was  reduced to  0.75 meters  per  second,
laminarization  of the  plume was  observed.   At 0.5  meters per second
wind  speed,   obvious  streaks,   indicating   laminar   behavior,  were
observed.  We do not believe that  a valid model  of a  full-scale  problem
 is achieved when laminarization occurs.
      To compare  the dense gas plume with  the  neutral  plume, we  plotted
 the concentration normalized by the source concentration  as a function
 of downstream distance.  The decay of the centerline  concentration with
 downstream distance in the dense plume is essentially identical  to that
 of  the  neutral  plume.   The model  scale ratio used  was  approximately
 1000 to 1.  Far downstream,  the slope  is about  -1.5,  typical of  the
 downstream behavior of a point source in a neutral atmospheric boundary
 layer.   Closer  to  the source, the slope  is  -1.0,  which has  to do with
 the finite width or area source.   Thus,  in  spite  of the differences in
 the plumes observed visually,  the centerline ground-level  concentration
 patterns changed very  little.   Therefore, this is not a good indicator
                                   132

-------
in  field measurements of  whether a dense  gas plume or  a neutral gas
plume has been released.
     In  our  studies,  lateral  concentration  profiles of the plumes were
measured  at  300,  600,  1800,  5000,  and 8500  millimeters downstream.
Gaussian  distributions  were  extremely  well   fitted  to the  lateral
concentration  profiles  of  the  neutrally buoyant  plumes.   Dense gas
crosswind   lateral   profiles  are  not   Gaussian   at  all;   they  are
essentially  flat topped.
     The  centerline  concentrations for  the  dense gas  and neutral gas
plumes  are  essentially  identical.    However,  vertical  concentration
profiles  differ dramatically.    At the  50-millimeter elevation,  the
concentration  for  the neutral gas plume  at  300 millimeters  downstream
is  about 50  percent  of  the  surface concentration.   In  the  dense gas
plume,  the  concentration at  the 50-millimeter elevation  is  more than
two  orders   of  magnitude lower  than  the surface  concentration  value.
Farther  downstream, the  dense plume tends more toward  neutral  or well
mixed at the surface.
     In  our  experiments, the dense gas  plume  starts  out  substantially
wider than   the  neutral  plume due  to lateral  spreading  right at the
source.   The dense gas issuing from the source actually creeps upstream
approximately  one  half  of  a  source diameter.   A very short distance
downstream,   the  lateral   growth   rates  of the  neutral  and   dense  gas
plumes are essentially identical, i.e., equal slopes of the curves.
     Comparing the centroids of the two distributions, a measure of the
vertical widths  of the  plumes,   the  negative  buoyancy  appears  to  be
significant as far downstream as  you choose  to compare.

                                  133

-------
     Looking at ground-level  concentrations of the neutral  gas plume as
a  function  of  downstream distance  on the  center!ine with  the  ramp
installed, the  ramp  induces  a ground-level reduction  of  approximately
20 percent.   This is not a major difference.   With the dense gas plume,
the ramp  induced  perhaps  a 20- or 30-percent  decrease in  ground-level
concentration.
     Anticipating that  the buoyancy would be  more  important  closer to
the source, we  moved  the  ramp  closer to the source.  In that case, the
ramp  induces perhaps  a  30-  to  40-percent  reduction  in  ground-level
concentration  for the neutral  plume.   With the  dense plume,  the ramp
induces about a 50 percent reduction in concentration at ground level.
     An extensive experimental data  set on the detailed structure of a
dense gas plume resulted  from these studies.  The model depicts a dense
gas   plume   that  has   significant   buoyancy  effects,    but   which,
nevertheless,  will  be  insignificantly affected  by  molecular effects.
Data also were  collected  for the  neutral plume in the  same experimental
setting to  support comparative analysis.
     In  summary,  the longitudinal centerline concentration profiles of
the  dense gas  plume  were strikingly  similar to those of the neutral
plume.   However, plume shapes were dramatically different;  the dense
plume  being  much wider  in  the  lateral  direction  and much narrower
vertically.   Lateral  distributions  of the dense gas plume were top-hat
 shaped,  essentially uniform all  the  way  to  the  edges.   In contrast,
 lateral  distributions for the  neutral plume were essentially  Gaussian
 in shape.   Introduction of a ramp into the experiment induced  a  slight
 reduction (20  to  30  percent) in  the  ground-level  concentration  when the

                                   134

-------
ramp  was  far  downstream.    When the  ramp  was  located  close to  the
source, the  reductions were  somewhat  greater.   Reductions were  also
somewhat greater for the dense plumes than  for the  neutral  plumes,  but
not significantly so.
                                 135

-------
              ATMOSPHERIC  SCIENCE AND  EMERGENCY  RESPONSE
                   AT THE SAVANNAH RIVER LABORATORY
                   Allen H.  Weber and  R.  W.  Benjamin
                 E. I. du Pont de Nemours and Company
                       Savannah River  Laboratory
                               Aiken,  SC
     The Savannah  River Plant (SRP)  is  a nuclear production  facility
owned by the  U.S.  government and operated by du Pont.   SRP is located
in South  Carolina on  the  Georgia border  near  Aiken,  SC.   It  is  the
third largest nuclear materials  production  site  of the  Department of
Energy  (DOE)   and  includes  nuclear  fuel   production,  fabrication,
irradiation,  and  reprocessing.   A program for  responding to unplanned
releases  into the  atmospheric  or  aqueous  environment  of radioactive
toxic  chemicals  is  essential,  and  one  has been  in  place  since  the
facility  began  operation   in  1953.   There has  been  an atmospheric
dispersion  research   program  at  SRP  for   almost  15  years.    The
consolidation  of  emergency response  and  atmospheric   research  has
strengthened  both  programs.
     The  SRP  encompasses   315  square miles of mostly wooded,  gently
rolling terrain.   Four operational reactors, two chemical  reprocessing
plants  with  high-level waste  tank  farms,   a  tritium processing  and
handling  facility, and  a  fuel  fabrication  facility  are operating.   A
waste  vitrification  facility and a naval  fuels  fabrication facility are
under   construction.   Hazardous  effluents  that  accidentally could  be
released  into  the  atmosphere from  these  facilities include  tritium
 (elemental  and  oxide  forms),  gamma-emitting  aerosols,  radioiodine,
                                   137

-------
transuranic aerosols,  radioactive  noble gases, and chlorine  gas.   The
Savannah River runs across the southern border of the plant.
     The   Environmental   Transport   Group   (ETG)   is  charged   with
characterizing  and   calculating   the  transport  and   dispersion  of
atmospheric and  aqueous releases  from SRP facilities.   The  group  is
composed   of   14  professionals   and  8   technicians,  including   5
meteorologists,  an  oceanographer,  2  health  physicists,   a  nuclear
chemist, and  a computer systems  manager.   At  least  two meteorologists
and a supervisor are on call for emergencies.   Pertinent information  is
reported to DOE  and  du Pont managers  for  final  action.   Environmental
impact and regulatory document matters are assigned to a separate  group
although ETG offers expert assistance as needed.
     The primary  aim  of  emergency  response  activities at  SRP is  to
quickly identify  and  characterize  any accidental  release of  materials
and to  provide  SRP decision makers  with  the  information necessary  to
respond  wisely.     The  emergency  response  system   has  two  major
components, the  WIND  system  and  the TRAC  mobile   laboratory.    The
Weather  Information  and  Display   (WIND)  system  is  used  for   data
acquisition,  archiving,  and model  calculations.   Data enter  the  WIND
system from the Remote Environmental  Monitoring System that  provides:
     •    Real-time wind  and  turbulence data  collected  from  8 single-
          instrument,  60-meter  on-site towers  and a  nearby  300-meter
          television tower instrumented at 7 levels;
     •    Regional  meteorological    data   from  the   Automated  Field
          Operations Services (AFOS)  system;
     •    Data from  temperature  monitors  in  on-piant  streams  and'the
          Savannah River;
     §    Real-time source term information from stack monitors; and
                                 r
                                  138

-------
     •    Real-time  plant  boundary  radioactivity  measurements  from
          perimeter monitors.
The WIND  system utilizes  VAX 780  and  750 computers to  process these
data,  run  dispersion  models,  and  support  remote  terminals  in  each
operating area.  The TRAC  mobile  laboratory  is used for plume tracking
and real-time radioactive monitoring.  These systems are operational 24
hours per day, 365 days per year.
     The  research  programs  at  SRP emphasize  mesoscale  atmospheric
dispersion.  The present field program,  Mesoscale Atmospheric Transport
Studies (MATS), is designed to determine the accuracy of dispersion and
transport models  and  to .measure horizontal plume width.   The downwind
distance  range   of   greatest  current   interest  is  approximately 30
kilometers,   which  coincides  with  the  major   population  centers
surrounding SRP, Augusta,  GA,  and  Aiken  and  Barnwell,  SC.  Since 1983,
this program has been concerned primarily with measurements in unstable
atmospheric  conditions,  but  this  year  the  emphasis  has shifted  to
measurements  in stable  conditions.   The  experiments simulate  short (15
minutes), intense  puff  releases.   A sulphur-hexafluoride (SF-6) tracer
is  measured  downwind   from  a  60-meter   stack  with fixed,  sequential
samplers  and  a mobile,  continuous  sampler  in  the TRAC  vehicle.   The
fixed  samplers  are deployed  by many of the same team  workers who are
called during actual emergencies.  Sampling in the TRAC vehicle  is done
to measure  the  instantaneous  horizontal  puff width during passage over
the 30-kilometer  arc.   To date,  26 daytime experiments and 3 nighttime
experiments have  been completed in the  MATS program.  An additional 25
nighttime and several more daytime experiments will be completed in the
coming year.
                                   139

-------
     A  spin-off  from  the  MATS  program  is  the  SJable  Atmospheric
Boundary layer  Experiment  (STABLE).   The  objective of  STABLE  is  to
develop  an understanding  of  turbulence  and  dispersion  at  mesoscale
range  in  the  stable  (nocturnal)  boundary  layer  and  to  improve  the
models for dispersion  under these conditions.   It  is expected that the
largest offsite doses  would occur from a  release  in  stable conditions.
STABLE   will    include  study  of   a   5-year  turbulence   database,
inter!aboratory field  experiments,  and extensive model evaluation and
development.   The fact  that  both the  Bhopal  and Chernobyl  accidents
occurred  during  stable  nighttime  conditions  has   given  additional
impetus to this program.
     The consolidation of operational  emergency response  and research
programs  in  one  medium-sized  group with  a single  group  manager  has
greatly  benefitted   both   programs.     Researchers   can   utilize  the
laboratory  and  technical  abilities   of carefully  trained  emergency
response specialists.   On the  other hand,  emergency  response personnel
learn  the latest observational  procedures  while gathering data  for
research  applications.  Moreover, dispersion  and transport  codes are
tested and improved  using results  from research.   Finally, operational
feasibility of research results can be tested during emergency response
exercises.
     The information contained in this summary was developed during the
course  of work  under Contract No.  DE-AC09-76SR00001  with the  U.  S.
Department of Energy.
                                  140

-------
       EMERGENCY PREPAREDNESS AND RESPONSE IN THE U.S. AIR FORCE
                        Captain Lawrence E. Key
             U.S.  Air Force Engineering and Services Center
                       Tyndall  Air Force Base,  FL

     The  U.S.  Air  Force  Engineering  and  Services  Center, where  I am
 assigned,  is designated  as  the focal  point  within the  Air  Force for
 environmental quality  research  and  development  (R and D).  While we do
 not  do all  of  the  R  and D work,  we do  try  to keep up  with  what is
 happening  in this  area.   When  we recognize a need  for research, we are
 responsible  for initiating a  project  to fill  that  need.   We also are
 responsible  for coordination of research activities with other agencies
 within   the  Air  Force  and  the   Department   of   Defense,   such  as
 Headquarters, Air Weather Service,  and Air Force Geophysics Laboratory.
 We  identify  needs  in the  field and try to develop tools that people in
 the  operational  areas can  use to  prevent  an  adverse  impact  on the
 environment.
     With  regard  to emergency  response,  the Air Force  is interested,
 basically, in two  kinds of events—hot spills  and cold spills.  In more
 technical  terms,   a hot spill  refers  to  any event  involving  a fire or
 an explosion, while  a  cold spill does not.  That terminology comes from
 the people actually  working in the field in missile range safety and at
 various missile  sites.   A fire or explosion  generally  involves  some
 type of  missile  explosion,  either  the vehicle  itself  or the  stored
 propellents  used  to fuel that  vehicle.   Cold  spills  may  involve a
 release of liquid missile propel1 ant or a release of aircraft fuel, but
do not involve fire  or explosion.
                                  141

-------
     Focusing on hot spills, we are primarily concerned  with  two types
of propellants.  Both the liquid and solid  propellants  are toxic.  The
primary liquid propellants  are nitrogen tetroxide  (N204)  and  Aerozine-
50 (ABO),  a  50:50 mixture, by weight,  of hydrazine  and  unsymmetrical
dimethylhydrazine (UDMH).   Both  of these are  very toxic  and,  because
they  are  stored  in  liquid form,  are  mobile  in  the environment when
spilled.  The typical solid propellant in the Air  Force  is composed of
ammonium perchlorate  and aluminum.   The primary  combustion  material
resulting  from  the ammonium  perchlorate-aluminum  mixture  is  hydrogen
chloride, which,  when  compared  with N204 and  A50, is not  very toxic.
However, we must be concerned with local  and climatological effects of
a large hot spill of the  solid propellants.  We must also consider the
situation of a hot spill  with both types  of propellants.   For example,
when the Titan 34-D blew up shortly after liftoff  at Vandenberg earlier
this year,  both liquid and solid propellants were  involved.
     A  typical explosion  involves a fireball that  lifts,  stabilizes at
a certain height, and then disperses.   At the two  largest launch sites,
Vandenberg Air Force Base and Kennedy Space  Center, meteorological data
available at the time of the accident are extensive.  There are a large
number  of  towers.    Vandenberg  currently   has   26  towers  on  the
reservation.   These towers are  instrumented at various  heights, have
instruments for mean wind  speed and direction  and  temperature, and can
measure temperature  differences  at various  heights  to get an idea of
atmospheric stability.  Vie also directly measure the standard deviation
of  wind  direction  at  various  levels.    Prior   to launch   at  these
locations,  a rawinsonde run is made to gather further information about

                                  142

-------
the stability characteristics of the atmosphere, particularly at higher
altitudes.
     At  other  sites,  primarily  Intercontinental  Ballistic  Missile
(ICBM) sites, instruments are much less numerous.  At these ICBM sites,
we  have  the  capability to  get  mean  wind  speed and  direction  at
approximately 10 meters above the ground.  Also, we have the capability
to estimate  the standard deviation  in  wind direction  and  determine a
value  for the  vertical  change  in  temperature.    That value  for  the
vertical change in temperature is  the  primary  means that the Air Force
currently uses to characterize atmospheric stability.
     The tools we have  for  modeling  a  hot spill  are currently limited.
The   Hypergolic  Accidental   Release   Model   (HARM)   was   developed
specifically  for the Titan  II  missile  system  that   is  fueled  with
hypergolic liquid propel1 ants.   The  Rocket  Exhaust Effluent Dispersion
Model  (REEDM) was designed  to estimate the  ground-level concentrations
of gaseous  hydrogen  chloride resulting ,from  a Space  Shuttle  launch.
REEDM  is  used  primarily  for  planning  prior  to  a   Shuttle  launch;
however,  it  does  have a module to handle a launch abort or explosion,
which would release fairly  large quantities of hydrogen chloride.
     The primary meteorological input to HARM is rawinsonde data.  HARM
has a preprocessor that  calculates the  source  term taking into account
the  heat  released   by  varying  amounts  of hypergolic liquids.    It
calculates an initial  fireball  size  and  cloud  rise, and the resulting
dispersion  is  modeled   using a  Gaussian  puff model.    There is  an
algorithm that estimates  the  amount  of deposition  due  to rainout.   The
                                  143

-------
primary outputs  of the  model  are  ground-level  concentrations of  the
combustion products of interest.
     The role of the meteorologist in these hot spills depends on where
the  accident  occurs.   At  Kennedy  and  Vandenberg,  a meteorologist  is
heavily involved in planning and in responding to accidents.  The REEDM
model  is  run by  the  meteorologists and  is used extensively  for risk
assessment.   Risk assessment  is  generally done  well  in advance of a
launch and  involves  a large number of  computer  runs  using  climatology
to determine  a  worst case  scenario.   REEDM is  used  for planning just
prior  to  a launch using forecast  winds and temperatures.   It is used
for accident response as required; but there really is limited time and
many more things to worry  about than  running  a model  immediately after
an accident.
     At other  locations, such as  the ICBM sites,  the  meteorologist's
role is  less well-defined  and tends  to be up to the  local commander.
For  risk  assessment  and   for planning  and  response,  the  heaviest
involvement  is  with  the  safety  and operations personnel.   This  is
principally a condition  of geography.   The  meteorologist is not at the
missile site that  has a  problem.    He or  she  is  at  a  central location,
and  it is the on-site people  that must do the  planning and response.
However, the meteorologist  is involved in post-accident analyses.
     There  are  two broad  areas of interest with  respect  to  the  Air
Force's  response  to  cold  spills;   1)  a  spill  involving Air  Force
materials and occurring  on Air Force property or  2)  a spill involving
industrial  chemicals  occurring off of Air  Force property.   The latter
category is included because the Air Force  is obligated to  assist if an

                                  144

-------
accident  occurs   near  an  Air  Force   installation   and  the  local
authorities  asked  for  assistance.    As  a  result,  the  Air  Force
meteorologist   occasionally  gets  involved  with   calculating  toxic
corridors for non-Air Force materials and non-Air Force locations.
     On  Air  Force  property,  our  primary  concerns  are the  missile
propel!ants,  N^O.  and A50,  and  other   types  of  hydrazines  used  in
various  aircraft.    There are  also  relatively  small   quantities  of
typical  industrial  chemicals  on  Air  Force  installations,   such  as
chlorine, which is  used in water treatment plants.
     Cold  spills  do  sometimes  occur  in  missile operations.    For
example,  in  1978,  a Titan  II  site  in  Kansas  experienced a N204 spill.
In  such  cases, the meteorological  information  available depends upon
the  site.   Kennedy  and  Vandenberg  are,   again,  well  instrumented.   At
the missile sites,  such as Titan II and Miniteman sites in the Midwest,
meteorological  information is relatively limited,  but  they  will  have,
at  a  minimum, wind  speed,  direction,   and   the  vertical  change  in
temperature.  At a  typical Air Force base, not supporting missiles, but
experiencing  a  hydrazine  spill  from  an  F-16  aircraft,   the  only
meteorological  information available  is  usually mean  wind  speed  and
direction, temperature  at a height of about  6  feet  above the ground,
and a  very  rough  means of  estimating  the vertical  temperature profile
and the  standard  deviation of wind direction.   If  a spill occurs near
Air  Force  property,  but  not  on  it,   the  amount  and  quality  of
information available  is  highly  variable  and, typically, very limited.
Off base, the  information available typically depends on the proximity
to the base or to a National Weather Service site.

                                  145

-------
     For cold  spills,  there are a  few models that we  have available.
One is  the  well  known Ocean  Breeze/Dry Gulch (OB/DG)  equation,  which
has  been available  since  the  mid-sixties.   This  is  an  empirical
equation  derived from  a  large number  of  tracer  releases.   It  was
developed by  the Air Force Geophysics  Laboratory.   There  are  several
other  similar equations  that  are  not as  well   known,   such  as  the
Mountain Iron  and Sudden Ranch  equations,  which are  used  at Vandenberg
primarily.   These are based on  a  series of releases of  material  at a
specific location and are used only at that location.   Vandenberg is an
interesting  and  complex place  meteorologically.   Vandenberg is  in a
coastal  situation approximately 150  miles  northwest  of  Los  Angeles;
therefore,  it has coastal  meteorology over very  rough terrain.   The
southern  half of the base happens to  be the most  complex,  and  that is
where the launch facilities are located.  The northern half of the base
is  less complex.   That is the  reason  why  there are  two equations used
at  Vandenberg.
     For cold  spills, we also  use  an  Air Force-modified version of the
                                               TM
Complex Hazardous  Air Release Model  (CHARM)    developed by  Radian
                    TM
Corporation.   CHARM   is  used specifically  for the current program of
                                                                     TM
deactivation  of  the Titan  II  missile system.   The modified  CHARM
model   was   added  to  our   pool   of  models  following   a  thorough
investigation  of a 1980 hot spill accident of a Titan II near Damascus,
AR.  As a part of this investigation,  a thorough  review  of the safety
procedures  for  the  entire Titan  II  program  was made.    One  of the
recommendations  from that  review  was  the need to  replace  the  OB/DG
                                                                     TM
equation.   OB/DG does not  handle  dense gases, such  as N«0..   CHARM

                                  146

-------
does include  a  dense gas algorithm and has  been  shown  to be useful in
the  field  for  estimating  a  toxic  corridor that  would result  from a
release of NgO^.
     The OB/DG  equation is  fairly simple.   It  is  based on  a  linear
regression of the data gathered during a large series of test releases.
As  a meteorologist,  it  is  a little disturbing that  the  equation does
not  consider  wind  speed.    This  basic  equation  is  even  further
simplified for the typical  Air Force base by adjusting the coefficients
                                                TM
and  exponents  accordingly.    We  adapted  CHARM    for Air  Force  use
specifically  for the  Titan  II  deactivation.   It  is a  Gaussian puff
model that  includes  a  dense gas  algorithm  based on Eidsnik's work.  It
has  a source  strength module,  while OB/DG does  not.    Thus,  CHARM
gives  us  an  improved  capability  in estimating  the  source  strength
resulting from  a spill.
     At  the  time  of  a cold spill,  by  regulation,  the Air  Weather
Service  meteorologist  on   site  calculates  a  toxic  corridor.    In
practice, the response varies.  The  on-site commander  who responds to
the  spill  decides  whether  the corridor  is calculated.   This  does not
mean that  no  action is taken  to prevent loss  of  life or  injury if the
modeling is not conducted.   Typically, what  happens,  in my experience,
is  that  the commander responding to  the spill  directs  that the entire
downwind area be evacuated.   This is a very conservative,  but effective
approach.
     Just   as  a  point   of  interest,   there   are   several   weather
restrictions  on handling materials,  and,  frankly, one  is a  point of
contention with some of the maintenance personnel.  This limitation was
                                  147

-------
particularly troublesome during the Titan II deactivation that involved
quite a bit of transfer of propellants.  Essentially, there could be no
transfer of propellants during calm winds and/or stable conditions, the
stable conditions being determined by the vertical temperature profile.
This  limitation  created a lot of  arguments  between  meteorological and
maintenance staffs.
      The  Air  Force  has  identified several  research needs.   We need,
like  most others in  this  field,  improved source  strength  models.  We
need  a  better  means  of  estimating   the  surface  temperature  in our
models.   We are concerned about mixtures  of chemicals released to the
environment mixing  with other chemicals:   water,  other pollutants, and
naturally occurring materials.  We do not  really have a good model to
handle fires,  particularly with solid  propellants.
      We  are looking  for  an  improved  dispersion model  for  general use
and are  still looking  for a  better replacement  for OB/D6, which will
consider dense gas effects and incorporate  a terrain wind flow model,
particularly  at Vandenberg.   Finally, we  need better model validation
techniques.   The commander on site needs to  have  some reliable means of
quantifying the uncertainty associated with the toxic corridor  that  is
given to him.   Frankly, right now we  do not have a  good means of  doing
that.
      We  have  several activities  under  way  to meet these research needs.
The Air Force Geophysics Lab  is working on  a new dispersion model.  It
 will  be  a relatively  simple,  Gaussian  puff  model  based  on  the  Shell
 spills model, although  we are making several  modifications.   Most  of
 this  effort  involves  looking at  the source  strengths involved with

                                   148

-------
spills of the various chemicals  about  which  the Air Force is routinely
concerned.  We  are planning to  include a dense gas algorithm.   We are
also looking at various other  models,  including the DEGADIS model that
was discussed earlier in this  conference.  A wind flow model, based on
work done by the Army, is also being evaluated.
     Also,  we  have  initiated  a  study at Lawrence  Livermore National
Laboratory  to  recommend a  methodology that  could  be used  by  the Air
Force when  we  are faced with  the need to do an evaluation of whatever
model  we  finally  select.    This  validation  technique  will  involve
compiling the dense  gas data that are  available in  a form suitable for
model evaluation  and,  then, recommending certain techniques to use in
comparing  our  model calculations to the actual  field data.   We also
have  a  small  effort under  way to look at quantifying the uncertainty
associated  with  toxic  corridor  calculations.    Additional   hot  spill
models are  being  considered to fill  the current gaps in our capability
in that area.
     This   discussion   presents  a   brief   overview   of   the  current
capabilities of the  Air Force  and of the  research we have under way to
improve our emergency response modeling support capabilities.
                                  149

-------
         POISONOUS GASES FROM LAKES: THE CAMEROON DISASTER
                Daniel A.  Livingstone and George Kling
                           Duke University
                              Durham, NC

     In  1960,  I became  interested  in  poisonous  gases  because  of  an
experience while on  vacation with  my  family  in  the  crater lake  district
of  Ruwenzori  Park  in western  Uganda.   We  smelled  rotten  eggs  and
tracked the smell about five miles overland  with a  Landrover,  ending up
at  a  small lake  that was purple with sulfur  bacteria.   There was a
whitish deposit of  salts  around the  shore.   Upwind,  the stench was not
too bad.   We stood there  and  saw no dead elephants  along  the edge of
the lake, but  then,  elephants  are  not  usually attracted  to  drinking
water  that is supersaturated with halite or trona.   It seemed to us the
better part  of valor  not  to  venture down into the  crater  to  look for
smaller dead  animals.  Elephants  might steer clear;  limnologists might
not.
     This incident  stayed in the back of my mind for 10  years.  When
one of my students,  John  Melack, went to the  same lake in 1971, it no
longer stank  of rotten eggs.  The climate of Uganda  had  changed, with  a
number of wet  years  occurring  in succession.  When Melack got to  the
lake,   it  was saturated  with  respect  to  sodium chloride  in  the  deep
water, but there was a  meter and  a  half  of rainwater  lying  on  the
surface.  There was  no evident evolution of gas.   On our  first visit,
we had  not observed  any bubbles either, but  we were probably  too  far
away  to see  any but catastrophically large  ones.
                                   151

-------
      I have  since spent  several  months on  Lake Bosumtwl  In  Ghana,  a
 meteorltic explosion crater,  where there are many  traditional reports
 of large bubbles  of  gas  coming up out of the water.  There are even  a
 few reports in the scientific literature of such degassing.   Although
 few people have  been killed  by these  bubbles  of gas,  accidents  have
 happened  with sufficient frequency to generate  an interesting  folklore
 regarding the  lake  and  the  proper  way  to propitiate  its dangerous
 resident  deity.
      The  bubbles  of gas tend to rise  in  August.   I was there in August,
 1966,  when the  lake was  circulating freely.   There was  no smell of
 rotten eggs,  but  the lake had an odor  to  suggest  that  some  form of
 reduced sulfur,  perhaps mercaptan,  was percolating slowly from  it.  We
 saw no bubbles of gas burst from the  lake.
      Last year,  Kling and I  went  to Cameroon  with  Curt  Stager, Jean
 Maley,  and  support  from National  Science  Foundation   (NSF),  Centre
 National  de  la   Recherche  Scientifique  (CNRS),  and  Office   de  la
 Recherche Scientifique  et Technique  d'Outre-Mer (ORSTOM)  to study the
 history of Cameroonian lakes  and  make a limnological survey of them.
 Ndoni  Paul  was  our Cameroonian  assistant.   We  arrived  shortly after a
 very big  bubble of gas had burst from Monoun, killing 37 people.  That
 happened  on August 15, 1984, at  approximately 11:30 p.m.   On August 21,
 1986,  at  9:30 p.m.,  a bubble of  gas came  out of  Lake  Nyos  in  the
western part of Cameroon  killing almost  1700 people  and  a great number
of  cattle.    This  was a major disaster,  comparable   to  Bhopal  and
Chernobyl.
                                  152

-------
      When the first reports of  a  gas discharge came  from  Lakes  Monoim
 and  Nyos, it was  regarded  as a mystery.   Various explanations,  natural,
 artificial,  and  supernatural,  were  advanced  to  explain  the  mystery.
 Most of these explanations were less satisfactory to  us than they were
 to  others.   Now, after the  post-event observations  of  Kling  and the
 other  members  of  the  Agency  for  International  Development  (AID)-
 supported Foreign Disaster  Team,  there  is little  mystery left.   We
 think we  understand the phenomenon quite well.  We can probably predict
 it in a broad sense, and we can  almost certainly control it.
      These natural  events differ  from most catastrophes we  have been
 discussing at this workshop.  Meteorological  processes dissipate this
 problem,   but  meteorological   processes   probably  also  trigger  the
 catastrophe.     They  must  also  be   considered   in   preventing  its
 recurrence.
      In a tropical  lake,  there  can  be  many layers.   This layering can
 be very  stable.   The first layer  is the  air-water interface.   Next,
 there is   a  layer of  surface  water,  usually  fairly  low  in  dissolved
 salts,  that  circulates  freely  during  the  season of  minimal  thermal
 stratification  in the  lake.   Underneath that  layer, there  are  usually
 one  or  more  layers that are  stabilized by dissolved substances.  These
 deep  layers  are not stabilized thermally,  and  may  even display  reverse
 thermal  stratification, but are  chemically so  stabilized  that they may
 be perennial.   During  the season  when  the  surface  water is  at  its
coolest,  when  the thermal  stratification  is  weakest, a strong  wind,
low-pressure  system,  landslide,  subterranean  slump,  or  anything else
                                  153

-------
that will disturb the free water surface produces  standing waves at all
the deep discontinuities.
     In  a  small  lake,  such as  Singletary  Lake in North  Carolina,  the
surface  rises  in  the downwind direction on even a fairly still  day by
something  like  5  centimeters.  The  lakes  in Cameroon are  also  small,
with a similar surface setup;  but deep in a stratified crater lake,  the
next  discontinuity  is  set up much  more than  5 centimeters  in  the
opposite direction.  These two  slopes,  that  of the air-water interface
and  that of  the  first water-water  boundary,  are not the same.   The
surface  slope  involves air with a density  close to zero  and water with
density  close  to  one,  while  the internal  slope involves  two layers of
water  with   density  differences  about   1   percent   of  the  density
difference  at  the  surface.    So, when a  wind   sets  up  a lake,  the
internal  slope  is some hundred times that of the air-water interface.
The vertical displacement will be 5 meters or more.
     If  a  deep  layer of  the lake,  for whatever reason, is charged with
a  very  soluble  gas,  Carbon   dioxide  (COp),  hydrogen sulfide  (HpS),
methane, or ammonia  to the point of saturation at  the ambient pressure,
and  the  layer is displaced by a large  set up, then  the  deep water can
effervesce violently.   When   this happens,  we can have  a tragedy like
that in  Nyos or Monoun.
     Kling has detailed  stratification data from Barombi  Mbo,  a lake
near Kumba in  southwestern Cameroon,  which fortunately  does not have
much gas dissolved in its deep waters.   Because  of  the  annual  pattern
of cloudiness, southwest Cameroonian air temperature reaches its annual
minimum  in August.   The surface water  of  Barombi Mbo also reaches its

                                  154

-------
annual minimum  around  August.   This kind  of  thermal  regime is typical
of  a  tropical  lake, which typically  has  a season of  maximum thermal
stability  with  warm   surface   waters.     Stratification  involves  a
temperature range where water density changes quickly with temperature,
so  small  temperature   differences  are  sufficient to  produce  stable
stratification.   The upper part  of  the lake is thermally stable during
most  of  the  year,  and the deep  waters are perennially  stabilized  by
dissolved salts even to the point of  allowing  a thermal inversion.   In
June, July,  and  August,  the  shallow  stability  is largely removed  by
cooling  of  the surface water.   The surface  layer and the next  layer
develop  very  similar densities.   When something sets up  the  air-water
interface, the  discontinuity  between  the  surface  and deep  waters  will
be  set up enough  to generate large-scale effervescence of dissolved gas
if  the  deep  waters  are close to  saturation  at the  ambient  pressure.
That  was apparently the   situation at Lake Nyos  when  it  killed  1700
people.
     The discharge  of  gas at Lake  Nyos occurred near an 80-meter-high
promontory.   The  initial  wave rose  over the top of the promontory;  and
stripped  the  shallow  layer of  soil and  vegetation from  its  granitic
rocks.   That  initial   displacement generated  a surface  seiche.    The
amplitude of  the surface  seiche was  about 6  meters at the far end of
the lake.  This is  obviously  a  very serious kind of event.  It takes a
great deal of force to move such an amount of even fizzing water to so
great a height.
     Maps of  the area  are poor,  and  the  distribution  of dead animals
and people  have not been  plotted with a high level  of accuracy.   The

                                  155

-------
village of Old Nyos, which is 100 meters or more above the lake was not
affected.  Observers in Old Nyos, who were awake at 9:30 p.m., claim to
have seen  the burst of gas  and  water from the lake, but no one died
there.    On the other  side  of the lake  at  Sawe,  which is not  so high
above the  lake,  no deaths occurred  either.   However,  downstream from
the  lake the  village  of New Nyos was  wiped  out.   The fairly  heavy
settlements along the  road following the  river, as  far as Su Bum, were
also wiped out.   At  that  point, the  river takes  a turn  and  passes
through  an area that is lightly inhabited, until it reaches  the village
of Mashi, where again there were very heavy deaths.  The distance along
the course of the river from Lake Nyos to Mashi  is about 34 kilometers.
The width of the valley is not well known, but from the map contours we
estimated   it   to   be  about  2  kilometers  wide.  If the valley were
2  kilometers wide  and if  it were filled  with gas  to  a depth  of 10
                                                         o
meters,  then  we are talking about something  like  7 x 10  cubic meters
of gas.
     Autopsy  reports  indicated that the  people died  of C0«  poisoning.
There  was no  indication  of carbon monoxide  (CO)  or  hydrogen cyanide
(HCN),  nor was there any indication of  HpS  in  the victims.   They died
of COp  poisoning.
     If the lake,  which  is  very deep for  its  size  (1  kilometer by 2
kilometers,  with  a maximum  depth of 220 meters), were  fully charged
with  C02,  it  would  hold about  10*°  cubic  meters of  C02-    This is
sufficient  to fill  the valley  more than 10  times.  An  amount  of gas
that  would  fill   the  valley,   if expanded  from  22  atmospheres  to
atmospheric pressure,  could  do a lot of  work.   It could easily lift a
                                  156

-------
destructive  wave of water to a  height of  80  meters.   Remember,  the
entire volume of the lake  is not being lifted, but only a small portion
of one side.  It struck the promontory with enough force to splash over
the top,  even though the  water  surface generally was not  raised that
high.
     What was  the  source of the gas?   The limnologist thinks first of
detritus,  the corpses  and feces  of lake-dwelling plants  and animals
that decompose as they settle to the bottom.  This decomposing material
generates C02  by aerobic respiration and by fermentation together with
methane  when  the  oxygen  is  depleted.    Ammonia  and ^S   are  also
generated under  anoxic conditions.   It turns out that the COg  in  the
lake   cannot  come  solely  from  biological  processes.     There   is
proportionately  too much CO* and  not enough of  the  other  products of
.decomposition  in the gases  of Nyos  and  Monoun.   There is not  enough
methane,  FLS,  or ammonia.   If the gases in the deep waters of the lake
contained  appropriate amounts of biologically derived  ammonia,  water
samples  from the lake would bring tears  to your  eyes.  While there is
ammonia  in  Lake  Nyos waters, there is not enough to support a  theory of
biological  origin for the  COg.
     The isotopic  ratios  of  gases   in  the  lake  waters also provide
clues.   The first  U.S.  team to sample Lake Monoun last year found that
the  C02  in the water  had a  radiocarbon  age  of 18,000 years.   This
suggests that  perhaps  one-eighth  of  the C02  came  from biological
processes.    The remaining  seven-eighths  came  from  some  other, non-
biological  source.   These ratios  are consistent  with various mixtures
of  primeval- CO-, i.e.,  COg  coming  from  magma.   Overall, the evidence

                                   157

-------
indicates that  approximately one-tenth of  the C02  was  coming  from a
biological source,  and  the remaining nine-tenths, probably,  from deep
springs.     Looking  at   Cameroon's   springs,   we  find   that   most,
particularly those of West Cameroon,  are effervescent and that most are
a good source of C02«
     In each of the papers we have heard at this conference, we learned
about opportunities  to  select sites, to  design plants,  or  to develop
emergency  response  plans  as  an  effective  means   of   dealing  with
environmental emergencies.  Such plans are not helpful  in Cameroon.  If
you are  ever caught in  a Cameroonian CO^-release emergency,  the only
thing you can do is hold your breath and run uphill.   You are very much
better  off   suffocating  from  lack  of  air  than   breathing   dense
concentrations of CO*.   Breathing the C02 causes  severe  acidosis, your
brain shuts down in  seconds, your glottis goes  into  spasm,  and you die
quickly.
     What can you  do in a situation where you  cannot  select your site
by moving away  and  raising  your  crops  elsewhere?  You may  be able to
modify the site.  It is possible to take a pipe, of perhaps 6 inches in
diameter,  put it  down  into  the  deep  water of  a lake  that  is fully
charged   with  CO-,  and  pump  the   water up  with  a  pump  until  it
effervesces.  Then, you  can  shut off the pump and  sit  back and watch
the  lake  degas.    It would  take about  10  years  by this  method to
discharge a  lake the size of Nyos  to  saturation at surface pressure.
But do  not  forget  meteorology while degassing the  lake.   First, you
must  know where your  discharge plume  will  travel.    Second, you must
take  care how  fast  the  lake  will  degas:   too  fast,  and  you might

                                  158

-------
produce your  own small  catastrophe;  too slowly,  and the  springs  can
keep up with  you,  and you will never  get  rid of the C02.   There also
must be a limnological portion to your plan.  When the pipe is running,
you must  dispose of the waste  water  carefully.   If  it  is  returned to
the lake  in  a turbulent fashion and mixes  freely with  the upper layer
of  the  lake,  you may  generate  a catastrophe worse  than any happening
naturally.
                                   159

-------
                      SUMMARY OF  PANEL  DISCUSSIONS

     Panel discussions were organized  to  directly  address the goals of
the workshop  and  to provide a forum for  free discussion of the issues
raised  by  the  goals.    The  panelists  and  participating  workshop
attenders  sometimes followed  an  independent  agenda,  but the resulting
discussions were very fruitful.
Panel 1:   Summary  of Current  Status of Source Strength and Dispersion
           Models,  Their  Strengths and Weaknesses,  and Recommendations
           for Improvement
Panel 2:   Recommendations  for  Selection  from  Among  Current Models for
           Immediate  Use   in   Hazard  Identification  and  Evaluation,
           Preparation  of  Emergency   Preparedness  Scenarios,  and  in
           Response to Emergencies
Panel 3:   The  Role of  the  Meteorologist  in  Hazard  Evaluation  and
           Emergency Response
     One   major  accomplishment  of  these  panel  discussions  was  the
fostering of  strong  interactions  between  two technical  communities
present  at the workshop.   The people involved  with chemical  releases
seldom  have  interacted  directly with those involved  in radioactive
releases.   The  difference  in  viewpoint was profound.  For  example, Dr.
Knox  related his  experience with Chernobyl  and showed  how the source
term  could  be  inferred  from measurement  of the  radioactive debris.
This  procedure took several  days,  but   as  his major  concern was the
impact  upon the United States,  this time interval  was not  important.
Even  if the concern was for  European  exposure,  there was  time to  take
measurements  and estimate  the source term.  On the other hand,  Dr.  Gait
showed  that when one is dealing  with  an  overturned vehicle,  immediate
                                   161

-------
estimates  of  the  source  term  and  the relevant  species  are  required
immediately,  especially  if  evacuation  of  people  is  contemplated.
Between these two  extremes was  a  continuous  spectrum of concerns.  The
panels allowed for a  free expression  of these  concerns  and thus proved
useful.
Panel 1:  Summary of  Current Status of Source  Strength  and Dispersion
          Models, Their  Strengths and Weaknesses,  and  Recommendations
          for Improvement
Chairperson: Frank Schiermeier
Reporter:    Steve Perry
Panel Members:   Gary  Briggs,  Tony Cox,  Don Ermak,  Dan McNaughton,
                 Bob Meroney, Jerry Schroy
       This  panel  summarized  the  current  status of source strength and
dispersion  models,  especially  their  strengths  and  weaknesses,  and
recommended  necessary  improvements.   Four  major areas were addressed:
source  term models,  dispersion models,  evaluation and  validation of
models, and  practical application of models.
     Knowing  how  much  material  is  emitted   in  a  given  emergency
situation  is of primary  importance.   Most often  the  emission rate is
estimated by means of  a model calculation.   The panel  reported on  five
source  term model categories:  process emissions,  contained emergency
releases,   spills,   burning  sources,  and  density   impacts.  Process
emissions  are  usually  point  sources  from  established  stacks  and are
thus  well  documented.     Contained  emergency  releases  result   from
components  designed  to  release  material   under  certain  conditions.
Examples of  these  are  discharges  from relief valves and rupture disks.

                                  162

-------
The  material  released could be from a runaway  reaction,  which  requires
careful  engineering  judgment.    If  the  material   is  accidentally  or
deliberately   ignited,  there  would  be  a  burning  source,  which  is
discussed  further below.   Finally,  if the  material  is  stored under high
pressure,  the release would be a  jet  that might produce a  multiphased
flow.   This  last case is not well understood at  all  and requires  a
great deal  of careful research.     Spills  of pure components and binary
mixtures that undergo, simple flashing and  pool  formation  are reasonably
well understood.   Spills  that  involve  multiphase flows are, again, not
well understood.   Steady-state spill models  are useful for  situations
that are most nearly  steady (obviously).   As the conditions  depart  from
steady-state   conditions,  such  as  occurs  with  pressure   drops,  the
steady-state  models  decrease  in  usefulness.    There is  a  very strong'
coupling  between  source  models  and dispersion models  for  spills  that
involve aerosols? and evaporation from these aerosols.   In  fact, it is
virtually  impossible  to separate the two for some conditions.
     Burning   sources  are another  category  where  there  are  major
unknowns  and  much research  to be  done.   A  relatively simple burning
source  would  be   a   flare that  resulted  from  igniting  a contained
emergency  release.   The input products are  known;  so presumably would
be the combustion products.  This is not always true because combustion
may  be  strongly influenced   by   the   ambient  meteorology  and  any
mitigation techniques that might alter the combustion temperature.  The
reaction products  often are controlled by the  combustion temperature.
Thus, even a  flare is not well  enough understood.  With an open burning
source^ the?unknowns  multiply very quickly.  The combustion temperature

                                  163

-------
is unknown, the product species are unknown, and the effect on material
that is not burning but adjacent  to  the  fire is unknown.  Furthermore,
attempts to extinguish the fire may  make  the situation worse by making
the species emitted  more  toxic than the  original  combustion products.
Major research is needed here.
     The density impact source category is a general category for which
parameters   like   aerosol   formation   from  condensation,   partial
vaporization,   chemical    reaction,   initial   dilution,   and   source
temperature are  important.    The single  most  important  point  about
density impact sources is the very strong nonlinear effect  between  the
size  of  the  source  release  and  the   subsequent  behavior  of  the
dispersing plume.  Here again,  there is  a  very  close  coupling  between
the source term  and  appropriate  dispersion  model.   The  panel  reported
three major dispersion  modeling  categories  related to  plume buoyancy:
nonbuoyant, positively  buoyant,   and  negatively buoyant  plumes.    The
nonbuoyant or passive plume category  applies to all  plumes  after enough
dilution.  The  character  of  the planetary  boundary layer  (PBL)  is  the
most  important  variable  here.    If  the  PBL  is  unstable,   i.e.,  large
convective  eddies  driven  by  the  surface  heat   flux  dominate  the
turbulent  exchange,  the  dispersion  of  passive plumes  is  quite  well
understood.  The models developed for this case do  much better than the
Gaussian  models.   This   is  rarely  a  worst case (where  very  high
concentrations occur) since dilution  is  quite rapid.   For  the neutral
case where the  turbulent  exchange is dominated by  eddies  generated by
the  vertical  wind shear,  the Pasquill-Gifford type Gaussian model does
an adequate job.   But, this  is  only true  for the  near-ideal  cases in

                                  164

-------
smooth flat  or  gently rolling terrain.   This,  also,  is seldom a worst
case, because there is often adequate mixing and dilution.
     The passive  plume in  stable conditions  is  often the  worst case
scenario,  as  well   as   the   most   difficult  case  to  model.    The
Pasquill-Gifford method only  works  for smooth terrain and unrestricted
mixing height  (an  unlikely occurrence).   The passive  plume in stable
conditions  is  the  hardest  case  to   generalize   because   of  surface
roughness effects,  limited  vertical  mixing,  flow meander, and drainage
flows in sloping terrain with the possibility of channeling.  This case
requires an extensive research effort.
     Buoyant  plumes  from  stacks  are  reasonably well  modeled  with
current  methods.  Surface  releases are  very  difficult  to  model  because
of surface deposition, high turbulence  levels, and entrainment effects.
If the plume  is  buoyant enough,  it  will lift from the ground.  Buoyant
plumes  interact with  the PBL  and  will  reach greater heights during
convective  conditions and  will  be restricted  in stable  conditions,
especially  if the  buoyancy  is insufficient to completely  penetrate the
nocturnal stable layer.
     A special type of buoyant plume that has not been well studied is
the  flare.   Often,  a  mitigation  technique for a release is to burn the
material  in a  flare.   An  example  is  a release  of  hydrogen sulfide,
which  would be  converted  to sulfur * dioxide after  combustion.    The
detailed  consequences  of  such  an  action  under  all  possible  PBL
conditions are unknown at present.
     Dense  gas  plumes are  currently  the subject of  intense interest
because of their potential  for immediate health  consequences.  A dense

                                  165

-------
plume  is  one  that   is   negatively   buoyant   under  all   atmospheric
conditions.    Such  plumes  often  modify  the  atmospheric  conditions
locally that can inhibit dilution.   The current view is that in sloping
terrain, with slopes as small  as 0.1  percent,  these plumes will travel
downslope.  With an  adverse  wind,  this might not  be the case, however.
In complex terrain,  such  as  canyons with steep walls,  a dense gas can
travel  great  distances as the event  in Cameroon  showed.    Much  work
needs  to  be  done  to   characterize  dense  plume  behavior  in  complex
terrain, especially with different  wind fields.  In  terms  of accidents,
however, the most important questions concern  the  direction a dense gas
cloud will travel in real  terrain with real  wind speeds and directions,
and what the concentration field will  look like.
     One technique that should  be  explored  for all  possible source and
dispersion  situations   is the  use  of  virtual   sources  that  can  be
nondimensionalized.  One  mathematical model then could be used  for a
variety  of releases.   Some  problems,  of course,  cannot be treated in
this  manner because of their  inherent  nonlinearity,  e.g.,  a burning
source.
     Dispersion models  cannot  deal satisfactorily with complex terrain
and  winds  of  very  low  speed  to  calm,  nor  can  they  deal   with the
transition  from  near source density-dominated  dispersion  to far-field
passive  dispersion.   There  has been  little effort  to include spatial
and temporal variability  into current dispersion models.
     Introduction  of  new or   improved  models into  the  inventory of
operational models requires  a demonstration  that  the new models are
indeed  improvements.    For cases  where there  is no  current model,  a

                                   166

-------
rigorous comparison of model  results  with data is most important.  For
cases where there  are current models,  the new models must show a clear
improvement.     Further,   all  new  models  should  undergo  extensive
sensitivity testing.  In addition to the sensitivity analyses, a set of
cases where previous models have failed should be developed and the new
model tested on these cases to ensure correct model performance.
     The  panel  concluded  that  in  selecting  a suitable  model  for
application,   several    issues  must   be  considered.      In   hazard
identification,  the emphasis  is  on source characteristics,  i.e.,  the
possible ways  in which  the source  could  emit  hazardous  material.   The
models  required  for this phase  of  emergency  response concern  the
different types  of accidental spills that can  occur  along with a good
engineering estimate of the probability of such an accident.  Emergency
action planning requires the  results of the hazard identification study
followed by consideration  of  all possible ways in which the release can
be  influenced  by weather conditions.   The models required must account
for  the transport  and  dilution  of the  released effluent.   Chemical
transformation of  some species must also be considered.
     A  family  of  concentration  isopleths  for  a  variety  of  release
scenarios and environmental conditions are developed.  This information
can  be  viewed together  with  appropriate demographic  data to estimate
human exposures  to the  hypothetical  releases.   This  last step should
result  in  a  plan  for  responding to  accidental  releases.   Emergency
response requires  immediate  answers  for  alleviating  the risk  to  the
public.  The  models used  must be able to be  run  on  very short notice
with unambiguous results.   If hazard  identification  and preparedness

                                  167

-------
planning processes are  done  correctly,  then there  should  be a menu of
recommended  actions   from which  to  choose.    If enough   cases  were
examined In the planning stage, then there should be a close analogy to
the actual accident.   Appropriate action can then be taken.
Panel 2:  Recommendations for Selection from Among  Current Models  for
          Immediate  Use  in  Hazard   Identification  and  Evaluation,
          Preparation  of  Emergency  Preparedness  Scenarios,  and  in
          Response to Emergencies
Chairperson: Ray Hosker
Reporter:    Tom Pierce
Panel Members:  Harry Allen, Jerry Gait, Jerry Havens,  Reed  Hodgin,
                Gene Runkle, Bruce Turner
     The  second  panel   considered  recommendations  for  selection  of
models  for  immediate  use  in  hazard  identification  and  evaluation,
preparation of  emergency  preparedness scenarios, and for  response to
actual emergencies.   The panel  did not recommend  specific  models, but
rather agreed  to a list  of  features and technical  criteria for model
choice.
     The  first  general  feature   was  scientific  defensibility.   The
scientific  basis of a  model  for  a particular  application  is of great
importance.   The model  should  be  operated  only  for  those situations
where scientific credibility has been established.  The model should be
widely available with full  documentation on  both  the  scientific basis
and on test results showing  how the model  performs over the full range
of applicability.   The  documentation should  also include all technical
criticism  and  responses  to  this  criticism.    Further,  the  model
structure  should be  organized  into modules  that  allow updates to be
                                   168

-------
easily installed.   Each model  should have been  subjected to rigorous
peer  review  and  all  results  of  the  review  available  to  the  user
community.
     The  panel  recommended  a number  of technical  points that models
should deal with:
     1. Near calm  to very light winds
     2. Wind shear in the vertical  (speed and direction)
     3. All stability conditions
     4. Site-specific meteorological data
     5. Local terrain and local flow conditions
     6. Local obstructions
     7. Dense gas  effects
     8. Effects of hydrometeors (rain, snow, etc.)
     9. Momentum and buoyancy effects  at the source
     10. Time-varying release rate
     11. Chemical reaction of effluent
     12. Real time  on-site concentration measurements.
As  is  obvious  from this list, the panel  determined  that the problem of
application of  models  was  very much dependent upon  the individual case
and that  no global recommendation  of  a specific model for  all  cases was
possible.   Rather,  a  suite  of  models,  each carefully  matched to the
others in  the suite, was the  best  approach.
     The  panel  next  recommended that  potential  model users examine the
suitability of  a given  model  with  the following guidelines.   First, for
application  to  toxic  releases,   the  effects  of the  toxicity  and   a
probabilistic  statement of associated risk should be incorporated  into

                                   169

-------
the model output so that  the  user is  able to relate this output to the
problem at hand without going  through  another  step.   Second, the model
should be capable  of  producing results commensurate with  the level  of
accuracy actually required in the application.
     The panel  considered the  issue  of possible categories  for model
use.   The  first  use  is for  hazard  identification.    Does  a  given
situation pose a hazard to the public?  The best approach here is to do
an  analysis  of possible  release  scenarios,  then use  the model  most
applicable for  that  case.  Since rapid turnaround of  results  are not
necessary for  this situation  and realistic estimates are  desired, the
model  should  be as complete  a representation  of the  situation  as  is
scientifically  possible.    The  second  possible  use  identified  was
preplanning  for the  possible  occurrence  of the  releases that  posed
hazards.   As  before,  the  model should  be  as complete and  realistic as
possible.  Response plans should be developed using these results.  The
third model  use category occurs when and if an  actual  release occurs; a
simple model  with quick  turnaround is most appropriate.    Of  course,
this quick turnaround model should  have been run in  the planning stage
so that a record of its performance relative to the most complete model
is  available.   Thus,  a check on  the  uncertainty of  the simple model's
results are available to  the  user,  and safety  factors are incorporated
into  response plans  that are  to  be  activated.   The  final  model use
category is for post-event analysis.  The most complete model should be
used along with simple  model  results to  assess  the  performance of the
entire emergency response system.
                                  170

-------
   '  The  panel   recommends   that  all   emergency  response  plans  be
thoroughly  tested.   These  plans can be tested  by  holding simulation
exercises to ensure familiarity with all components.
Panel 3:  The Role of the Meteorologist in Hazard  Evaluation Emergency
          Response
Chairperson: Lew Nagler
Reporter:    Jack Shannon
Panel Members:  Marvin Dickerson, Allen Weber
     The  role of the meteorologist in hazard identification, emergency
preparedness  planning, and  in emergency response  is multifaceted but
must  remain  closely  connected  with  his  or  her  expertise.    This
expertise should  include  some  knowledge   of  allied  fields  such  as
industrial  hygiene  and  health physics because  the advice sought most
often  is  related to  these  fields.
      In   hazard  identification, the  role  of  the  meteorologist  is to
advise  on how  a given  potential  hazard might  be  affected by weather
conditions.   For example,  in analyzing the  risk  for a release from some
storage   or  processing  facility,  the  wind  field  and   atmospheric
stability determine  the rate  of  dilution  for  the  released material.
The details  of  the  source  configuration as  well  as  any  proposed
mitigation   techniques  are   very   likely  affected  by  meteorological
conditions.   Thus,  a  meteorologist  should be  consulted  at  the very
start of  a  hazard identification exercise.
      In  terms of emergency  preparedness, the meteorologist  will  select
the appropriate models  for  the problem  and will  be able to advise  on

                                   171

-------
the most  probable exposures arising  from  a given release.   He or she
will also select the models to be used in the event of a real emergency
and demonstrate  how the operational model  handles  the same situations
simulated  by the  more  sophisticated  planning models.    Finally,  the
meteorologist will  prepare weather  scenarios  for  initial  training in
emergency response and will also provide real data for further training
exercises.  The meteorologist will  also  advise  on the most appropriate
placement of on-site instrumentation (both weather sensors and chemical
or radiological).
     For  actual  emergency response situations, the  meteorologist will
have to  provide  the weather data and model  predictions  for use by the
appropriate  authorities.   If  the  meteorologist  has been  included in
this process from  the  beginning,  then  this  last phase, which  is the
most difficult, will be  that much more reliably done.
                                  172

-------
      APPENDIX A
The Bhopal Gas Tragedy
         A-l

-------
                   THE BHOPAL GAS TRAGEDY
                       K. SHANKAR RAO
        Atmospheric Turbulence and Diffusion Division
        NOAA/ARL, P. 0. Box-2456, Oak Ridge, TN 37831
                            and
                  M. P. SINGH and S. GHOSH
              Center for Atmospheric Sciences
        Indian Institute of Technology, New Delhi-110016
                          Abstract
    In the early hours of December 3, 1984, about 40 tons of highly
toxic, volatile, and reactive methyl isocyanate (NIC) and other
poisonous gases leaked from a storage tank at a pesticide plant in
Bhopal, India.  The gases escaped in about 90 rain through a 33 m high
atmospheric vent into the cool night air, and quickly spread in a
foglike lethal cloud over a large populated area.  Thus began the
world's worst industrial disaster involving toxic chemicals which
killed over 2,200 people and injured more than 200,000.

    In this paper, we sketch the accident scenario and outline the
events in the aftermath of this catastrophe.  The topographical
features and meteorology of Bhopal, and the physical, chemical, and
toxicological properties of MIC and other gases in the leak are
described in the context of the dispersion of the cloud and its
effects on human and biological life.  A simple atmospheric dispersion
model, emphasizing aqueous phase conversion and deposition of NIC, is
presented.  This model, based on an analytical solution of the 3-D
advection-diffusion equation, gives estimates of ground-level
concentrations and deposition fluxes of MIC.  The model estimates are
qualitatively correlated with recorded human fatalities and injuries,
and the observed damage effects on trees and vegetation in the
affected areas.

    In the epilogue, the medical care, relief and rehabilitation
efforts, and some preliminary results of the ongoing medical and
toxicological studies are summarized.  The legacy of the Bhopal
Tragedy including its implications for information, siting and
operation of plants that handle toxic chemicals, safety, health, and
environmental standards and enforcement, emergency response programs,
and other relevant issues, are briefly discussed.
Presented at the Special Joint Session of the Ocean and Atmospheric
Sciences on "Paths and Fates of Toxic Pollutants in the Atmosphere and
Oceans," American Geophysical Union Fall Meeting, Dec. 7-12, 1986, San
Francisco, CA.
                                     A-3

-------
INTRODUCTION

    On the night of December 2-3, 1984, at around midnight, methyl
isocyanate (MIC) and other poisonous gases, escaping from the Union
Carbide India Ltd. (UCIL) pesticide plant in Bhopal, silently spread
out into a huge cloud that enveloped a large populated area of the
city.  Hundreds, mostly children and old people, died in their beds.
Thousands more awoke to a Kafkaesque nightmare of suffocation,
blindness, and chaos, and stumbled into the streets, joining human
stampedes fleeing the gas cloud.  In the confusion, some ran in the
direction of the wind and died along the way.  Many others died later
in hospitals.  Thousands more were injured due to exposure to the
toxic gases.  In all, over 2,200 people died, and more than 200,000
were injured.  Thus, "Bhopal" came to be associated with the world's
worst industrial disaster involving toxic chemicals.

TOPOGRAPHY AND METEOROLOGY

    Bhopal was an unlikely setting for this tragedy.  A city of about 1
million people, 600 km south of New Delhi, Bhopal (lat. 23°17'N) is the
capital of Madhya Pradesh, India's largest state, with an agriculture-
based economy.  Because of its central location, resources, and hospi-
tality, the state traditionally attracted people from all parts of the
country to work in its 8000 industrial plants of various sizes.  Bhopal
was once a center of India's mogul past, and the city still reflects
this history despite a skyline of modern state office buildings,
colleges and hospitals built around two ("Upper" and "Lower") lakes.

    The UCIL pesticide plant is located in the northern parts of the
city (Figure 1).  North of this facility, there are hillocks, woods,
and stretches of residential areas extending out to 24 km from the
plant.  The Indian Agricultural Research Insitute is situated about
4 km north of the UCIL factory.  The latter is surrounded by residen-
tial areas on the other three sides.  The approximate distances and  •
locations, with respect to (w.r.t.) UCIL, of some Bhopal landmarks
shown in Fig. 1 are as follows: Electrical and Mechanical Engineers
(EME) Center (1.5 km H), airport (6 km N), Hamidia Hospital (3 km SW),
Sultania Hospital (3 km S), the State Secretariat (Vidhan Sabha) and
Police Headquarters  (PHQ) (4 km S), and the Railway Station (1.5 km
SB).  An industrial and warehouse area extends out to 15-20 km on the
plains ESE of the plant along the railroad.           {

    Table 1 shows the elevations and locations (w.r.t. UCIL) of some
of the important topographical features of Bhopal.  Some of these
features could have influenced the local wind and dispersion patterns
on the night of the accident.

    Detailed meteorological data are not available regarding the wind
speed and direction, and the vertical temperature structure at the
site on the night of the accident.  According to Ekalavya  (1985), a
private non-profit organization based in Bhopal, the winds that night
were initially from the northwest and subsequently from the north.
Post-episodic wind data indicate that the most probable wind speeds
were of the order of 3 m/s.  The temperature was reported to be 14°C.
                                   A-4

-------
Table 1.  Elevations and Locations of  Bhopal's Topographical  Features
                       Elevation  (m)       Distance  (km) & Direction
                       above MSL                 (w.r.t. UCIL)
Singarcholi Hillock
Idgah Hill
Ahnadabad Hill
Dharampuri Hill
Birla Temple Hillock
Shahpur Hill
Upper Lake
Lower Lake
625
575
513
601
562
525
-
-
4.5
2.0
2.5
5,5
4.5
8.0
4.0
4.0
W
NSW
WSW
SW
SSE
SE
SW
S
This is consistent with the minimum winter temperature of about 10°C
in Delhi.  Since both cities are located  in the Indo-Gangetic plains,
a mixing height (H) of 200 m, close to its climatological-mean value
in Delhi (Kumari, 1985) for a December night, was assumed for Bhopal.

    The gas-affected areas in Bhopal are  characterized by a gently
rolling terrain surrounded by hillocks and lakes.  It was likely that
nocturnal drainage winds and land breezes altered the local surface
wind patterns on the night of the gas leak.  In the absence of data,
the best way to decide on the wind direction is by charting the course
of the catastrophic imprint of the poisonous gases over the city.
Interviews with survivors and rescue personnel confirmed that the wind
direction varied during the night.  Based on this information, Singh
and Ghosh (1985) adopted the following sequence of wind directions,
apportioned appropriately for the duration of the episode: (i) 285°
for 45 min, (ii) 340° for 30 min, and (iii) 45° for 15 min.  This
scheme gave realistic model simulation results, as described later.

THE ACCIDENT AND ITS AFTERMATH

    The UCIL factory in Bhopal employed 950 people to manufacture and
market Sevin (carbaryl) and Temik (aldicarb), two MIC-based pesticides.
The accident began on the night of Sunday, December 2, 1984, when about
75 workers were on duty.  As reported in the biweekly India Today.
which investigated the accident and reconstructed it with approximate
timings, the sequence of events (in local time) is as follows:
10:45 pm -  A shift change occurred at the factory.  Unknown to the
incoming shift, a runaway chemical reaction was already under way in
storage tank 610, which contained 41 metric tons of liquid NIC.
11:00 pm -  Pressure in tank 610 was observed to increase from 3 to 10
psi.  The new staff mistakenly thought this was due to nitrogen pres-
surization of the tank by the previous shift.
11:30 pm -  Operating staff noted eye irritation but ignored it, because
tiny leaks of NIC in the utility area were not unusual.
00:15 am -  MIC control room staff reported the high pressure (30 psi)
in tank 610 to the Production Assistant; he checked and found that the
tank's rupture disk (designed to rupture at 40 psi) had burst, and the
safety valve had popped.
01:00 am -  Untreated MIC vapor was seen escaping through the 33 m
high atmospheric vent line.  Within a period of 90 min, about 40 tons
of MIC leaked out in gaseous form.
                                   A-5

-------
    The  escaping gases  overwhelmed  the scrubber  designed to  detoxify
process  vent  gases.   Other elaborate safety systems that failed
included the  flare  tower, the refrigeration system, and  the  water
curtain  (which  was  designed to shoot water to a  height of only 15 m),
and some of the other systems which had been out of operation for
repairs.  The public  warning siren  was turned on for a few minutes
around 1:00 am,  but was not sounded again until  2:00 am.  The city
police control  room was alerted. Just after 1:00 am, by  a town
inspector on  patrol near the factory.  The police, in turn,  alerted
the area hospitals, the army, and the state medical director at 1:30
am.

    The  poisonous gases emerging from the vent stack formed  a huge
white cloud that moved  south towards the densely populated areas. In
the vicinity  of the plant, according to survivors' reports,  the gas
was so thick  that visibility was very low.  More than 200 people died
in their beds in the  area immediately south of the plant.  Others
awoke to a  feeling  of suffocation and burning eyes, and  rushed out of
their homes In  panic.   In the narrow streets, they joined tens of
thousands of  half-blinded people running for their lives.  It must
have been a terrifying  spectacle.   Hundreds were overcome as they
fled.  Many collapsed on the roads, only to be removed to the
hospitals later.  The cloud engulfed Bhopal's railway station as it
spread southeast from the plant.  Passengers and station personnel
collapsed.  As  humans died, so did  animals.  The carcasses of dogs,
cats, chickens,  birds,  goats, cows, and other animals were later found
strewn over an  area of  65 km2 in the quadrant SE of the  plant.

    Starting  about  2:00 am, the first victims began arriving at
Bhopal's  hospitals.   Buses, army trucks, ambulances, and private
vehicles  were pressed into service.  All seven of the city's hospitals
were beseiged by thousands of gas victims.  Doctors, joined by
interns,  nurses,  and  medical students, began a seemingly unending
battle to save  lives.   As the hospitals became jammed, doctors,  relief
agencies, and private groups set up emergency clinics on sidewalks,  in
stores, and wherever  they could find space.  Bhopal's medical
community rose  to the occasion to handle the emergency resulting from
the unprecedented disaster and performed heroically in saving many
lives.

    A house-to-house  search of the  affected areas turned up hundreds
of victims.   By  late  afternoon, the city morgue  was overflowing with
bodies, many  of  them  unidentified.  There were mass burials and
cremations  for the dead.  As dusk fell the death toll mounted to over
1000 and  the  number of  injured rose to more than 100,000.  As the
magnitude of  the  tragedy became apparent, additional medical
personnel, equipment, and relief supplies were rushed to the city.
                                   A-6

-------
CHEMISTRY AND TOXICOLOGY

    Methyl isocyanate  (CH3N=C=0) is a reactive, toxic, volatile and
flammable compound which is commercially used to make carbamate pesti-
cides.  MIC, an ester  of isocyanic acid, is industrially produced by
the reaction of phosgene (COC12) with the base methylamine  (CH3NH2):

    COC12  +  CH3NH2   	>  CH3N=C=0  +  2HC1                    (1)

It should be noted that MIC is stored with a purity of 99.5%, and
contains up to 0.1% phosgene to inhibit MIC polymerizations and
reaction with water.   Thus, 40 tons of MIC (in tank 610) could also have
contained up to 40 kg  of phosgene.

    MIC undergoes exothermic and vigorous reactions with a
variety of compounds containing active hydrogen atoms.  The cumulative
effect of the adjacent double bonds adds to the instability of the
compound.  From a practical point of view, only a limited number of
those reactions could  conceivably have taken place in the plant at
Bhopal (depending on the presence of specific reactants during the
episode).  These possible reactions are shown in Table 2.  Post-
episodic analysis of sludge samples from tank 610 showed that two
major reactions took place in the tank — that of MIC reacting with
water, and that of MIC reacting with Itself, catalyzed by iron, which
is a much faster reaction.  More than 50% of the sludge was found to
consist of MIC trimer  resulting from the latter reaction, which
liberates about 1.26 x 106 joules of heat per kg of MIC.  About 16 to
28% of the sludge was  made up of compounds produced by MIC reacting
with water, which starts off slowly at room temperature, producing
about 1.36 x 106 joules of heat per kg of MIC.  Just a small amount of
water with a trace of  contaminants such as rust and salt, for example,
could catalyze an autoreaction.  The heat liberated during an induction
period lasting a few hours could generate a reaction of explosive
violence.  It is not hard to imagine situations in which such trace
contamination could occur, despite best efforts at prevention.

    Though the initial mix of poisonous gases that escaped from the
factory into Bhopal's  air primarily consisted of MIC. it apparently
also included smaller  amounts of phosgene, hydrogen cyanide, methyl-
amine, and other gases such as carbon monoxide.  Though officials
initially denied any possibility of cyanide presence, scientists
detected cyanide near  the MIC storage tank, and 50 m downwind, three
days after the accident.  Further, autopsies confirmed symptoms of
cyanide poisoning among some of the victims.  During the runaway
reaction in tank 610,  it is possible that temperatures exceeding 200°C
were attained.  At these high temperatures, MIC can break down into
hydrogen cyanide, carbon monoxide, »ono-methylamine, and some organic
cyanides.  The physico-chemical, and toxicological properties of
various poisonous gases in the Bhopal tragedy are given in Table 3.

    The behavior and fate of a MIC plume in the environment  is schema-
tically shown in Figure 2.  In the atmosphere, MIC reacts with
moisture to yield methylamine (MA) and C02:

    CH3N-C»0  +  H20   	>  CH3NH2  +  C02                       (2)
                                   A-7

-------
                           Table 2.  Some Important Chemical Reactions of MIC.
>
oo
                           1 .  Reaction with alcohols and phenols
                                                          o
ROH
                                                        ROCNHCHg
                                                         An N-methylcarbamate
                           2.  Reaction with water
                             (a)CH3N = C = O  + H2O (excess)
                             (b) CH3N = C a O (excess) + H2O



                            3.  Reaction with itself


                                                 catalyst
                              3CH3 N = C = O
                       CH3NHCNHCH3 +  C02
                         1,3-Dimethylurea


                           O  CH3 O
                           II  I    II
                      CH3NHC-N —CNHCH3 H
                         1,3,5-Trimethylbiuret
                                                                                        CO
                                                                        I
                                                                       CH3

                                                                 Trimethyl isocyanurale

-------
Table 3.  Physico-chemical and Toxicological Properties of NIC and
          Other Poisonous Gases In the Bhopal Tragedy.

Chemical
formula
Description
Nol. Height
Boiling
Point (°C)
MIC
CH3N=C=0
colorless
liquid with
powerful
odor
57.05
39.1
Phosgene
coci2
colorless
gas or
volatile
liquid
98.92
7.6
Hydrogen
Cyanide
HCN
colorless
liquid with
faint odor
of bitter
almonds
27.04
25.7
Methylamine
CH3NH2
colorless
liquid or
gas with
strong odor
of ammonia
31.06
-6.3
Liquid Density   0.96
(w.r.t. water)
(i 200/4eC)
1.38
0.69
0.66
Vapor Pressure
(mm of Hg)
Vapor Density
(w.r.t. air)
LC50 (ppm)
(rats)
MSO
(humans)
TLV-TWA
(ppm)
References r Sax
348
(§ 20°C)
2.2
5
(4 hr)
>20*
ppm
0.02

(1979), Weast
1180
(§ 20°C)
3.4
1482
(1 min)
3200
mg/m3
0.1

(1982).
400
(@ 9.8°C)
0.93
544
(5 min)
200**
mg/m3
10



1.07
-
-
10


LC50: The concentration lethal to 50k of a specified population.

TLV-TWA: Threshold Limit Value-Time Weighted Average.
         (The average concentration for an 8-hour workday to which
         workers may be repeatedly exposed without adverse effect).

* Estimated as 1000 times TLV-TWA value.

** Lowest published lethal concentration for a 10 min inhalation.
                                  A-9

-------
Similarly,  phosgene  reacts with moisture  to  yield  hydrogen  chloride
and C02.   In the  presence of sunlight, MIC degrades  into polymers  and
gases  such as  CO  and N2.  In soil and water, methylamine is the major
product.   It is held tenaciously by  soil  particles until decomposition
is accomplished by weathering and biological reactions.  In plants,
MIC could  have competed with C02 in  the process of photosynthesis.   It
could  have been transformed into methylamine and other  carbamylated
derivatives of the plant constituents.

    One  of the disturbing aspects of the  chemical  disaster at Bhopal
is that  scientists know relatively little about the  toxic effects,
especially the long  term ones, of MIC.  Good toxicological data on MIC
were non-existent in the open literature.  The permissible MIC skin
exposure limit is 0.02 ppm averaged  over  8 hrs.  This threshold limit
value  (TLV) has been set by the American  Conference  of  Governmental
Industrial Hygienists (ACGIH).  It represents the  time  weighted
average  concentration (for a normal  8 hour day or  40 hour work week)
to which nearly all  workers may be repeatedly exposed,  day after day,
without  adverse effect.  Isocyanates long have been known to attack
the respiratory system, eyes, and skin.  They can  injure the lungs and
bronchial  airways, and cause permanent eye damage.  But until the
Bhopal episode, the  effects of MIC had never been  observed on such a
large  and  diverse population.  Most  of the deaths  have  been attributed
to various forms  of  respiratory distress.

    Indian scientists have initiated a variety of  studies to inves-
tigate the long-term consequences of exposure to MIC.   Many of these
research efforts  are focused on damage to specific organs, such as the
eye or lung.   Others address possible genetic and  carcinogenic
effects.   The  value  of all these clinical and laboratory studies,
which  may  take up to 3 to 5 years, will largely depend  on building a
data base  of high-quality epidemiological information.  For this
purpose, researchers need to answer  at least three key  inter-connected
questions:
(i) Where  were the victims when they were exposed  to the gas?
(ii) What  dose did the victims encounter?
(iii)  What  were the  health effects from exposure to the gas?
The most difficult part of this investigation will be to estimate  the
concentrations of MIC that individuals encountered.  One indirect way
to estimate the MIC  concentrations is by charting  the degree of damage
to sensitive trees and plants, in different parts of the city; this
will enable one to determine the level of MIC to which  people in those
areas  were  exposed.   Scientists from the Indian Agricultural Research
Institute  and  the Central Board for Prevention and Control of
Pollution  studied the trees and crops in Bhopal immediately after the
disaster.

    Another indirect  way to estimate the concentrations is through the
use of a suitably formulated atmospheric dispersion model.  Singh  and
Ghosh  (1985, 1987) have developed a simple analytical dispersion model
including the  aqueous phase transformation and deposition of MIC,  as
described below,  for  use in preliminary assessments.
                                     A-10

-------
DISPERSION MODEL, RESULTS, AND CORRELATION WITH OBSERVED EFFECTS

    The density of MIC vapor is about twice that of air.  This
suggests that density effects might be  important for plume dispersion.
However, the velocity of ejection of the gas was very high and, hence,
there was probably considerable entrainment and mixing.  Exothermicity
of the MIC-moisture reaction suggests that a great deal of heat was
generated in the process, and this could further reduce the density of
the plume.  As a first approximation, therefore, the density effects
were not explicitly considered in the model.  The plume dispersion
was assumed to consist of spreading due to passive diffusion as well
as spreading due to gravitational effects; this would imply larger
a values than in the case of a purely passive diffusion.  Singh and
Ghosh (1985) found that Briggs1 (1973)  urban dispersion parameters for
stable conditions (E-F) give concentration estimates that were
consistent with the observations.  Other assumptions used in their
model formulation were as follows: (i)  the time-dependent emissions
are specified by a uniform emission rate over the duration of the
episode, (ii) the terrain is flat and spatially homogeneous, and
(ill) removal by dry deposition and chemical transformation is a
first-order irreversible process.

    The time-dependent three-dimensional atmospheric advection-
diffusion equation is given by


               3c + uac = D  a!c + D   afc_kc
               at     3x    y 3y2    z  8z2    *
where C(x,y,z,t) is the pollutant (MIC) concentration, U is the
constant average wind speed along the x-direction, k^ is the total
removal rate, and Dy and Dz are the constant eddy diffusivities in the
crosswind and vertical directions, respectively.  The source condition
is described by
               C(0,y,z,t) =   p  5(z-hs) 6(y)                       (4)
                   i
                      Q     ,    0 < t < t
where    Q(t)       '   °               =  e
                      0
In Eq. (4), Qo is the MIC emission rate, hs is the effective release
height, and te = 5400 s is the duration of the release.  The initial
and boundary conditions are given by

         C(x, y, z, 0)    =0
                ' z=0,H

         C(x ,± oo, z, t)  =   0




                                  A-ll

-------
    The total removal rate  (kt) of MIC is given by
                      kt = kc * kd

where kc is the chemical reaction rate and k
-------
 mated by the power-law profile from the wind speed (3  m/s)  at 10  m
 height,  using an  exponent  value of  p =  0.35 for  stable conditions.
 The wind direction and mixing height H  were specified  as  discussed  in
 Section  2.   The spreads, ay and ffz,  were given by  Briggs1(1973) urban
 curves for  stability classes E-F.   The  stack inside diameter  was  0.2 m,
 and the  estimated gas exit-velocity was 88  m/s.  The effective release
 height hs was estimated to be about 40  m from simple plume  rise
 calculations (e.g.,  Holland,  1953)  under stable  conditions.

     Figure  3 shows the variation of the ground-level concentrations
 (GLC) as a  function of downwind distance along the plume  centerline
 after 45 min of time integration.   The  model simulation results are
 given in Table 4,  which lists the total time-integrated GLC,  surface
 deposition  fluxes,  and the approximate  times of  initial impact (i.e.,
 plume front arrival) at the various affected sites.  These  sites  are
 marked in Figure  4 which shows the  concentration contours demarcating
 various  zones of  the affected areas based on the estimated  total  GLC.
 In  general,  the estimated  GLC distribution  and the plume  arrival  times
 seem to  correlate well with the extent  of affliction at the various
 sites and information obtained from interviews with survivors.

     It is clear (from Fig.  4) that  within zone I,  encompassing sites
 1-3.  extremely high concentrations  of the gas prevailed.  The calcu-
 lated maximum GLC were greater than 50  ppm.   According to local
 residents,  these  colonies  were engulfed by  copious dense white fumes
 with a bitter sweet smell;  visibility was very low.  These were also
 the worst affected areas,  with total mortalities adding up to 360 (out
 of  a population of 6,173)  by the time the smoky  pall lifted at 5:30
 am.   The major medical symptoms  in  the  survivors were  severe  eye  irri-
 tation,  coughing,  vomiting,  and excruciating chest pain.  Zone II,
 which includes sites 4-6.  was also  a very badly  affected  zone.  Here
 the estimated GLC  were above  15 ppm.  About  508  casualties were
 reported within this area  (the toll was higher in  this  zone since it
 includes many more colonies  with a  much larger population than zone
 I).   About  87% of  the initially recorded casualties  were within these
 two zones.   Model  results  (Table 4)  suggest  that the plume  front
 arrived  at  these  sites within the first 6 min of the episode.

     Zone III,  the  "zone of moderate effects",  encompassed sites 7-11,
 14,  22,  and 23.  The immediate fatalities were 11,  much lower than
 in  zones I  and II.   People experienced  severe eye  irritation  and
 breathing difficulties.  The  estimated  gas concentrations in  this
 zone  were between  15 and 1.5  ppm.   Zone IV,  encompassing  sites 12, 13,
 15-21, and  24-26,   is the "zone of marginal effects"  with  only five
 fatalities.   The estimated concentrations here were less  than I ppm.
 Some  residents experienced slight eye and throat irritation,  though no
 traces of the  fumes  were visible.  As described  earlier,  many people
 fled  their homes during the episode and,  hence,  there  is  an inherent
 subjectivity involved in this correlation study.   However, many of the
 casualties occured close to their normal  place of  residence.  The
 effects  of the gas faded significantly  with  increasing downwind
distance.  Though humans and  animals inhabiting  the  region extending
out to 6  km  downwind of the plant were  affected  by the gas, the
heaviest  toll occured  within  the first  200 m to  1  km distance.
                                A-13

-------
   Table 4.   Model Estimates of GLC and Deposition Fluxes  of NIC.

      Site      GLC    Surface  Deposition   Time  of Initial
               (ppm)    Flux (|*g/m2/s)        Impact (min)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
93.5
74.2
57.4
25.9
18.5
15.5
8.0
3.4
2.5
8.2
8.6
0.1
0.2
1.7
0.9
1.0
0.4
1.0
0.3
0.6
0.2
3.4
1.8
1.0
0.8
0.7
4133
3280
2537
1145
818
685
354
150
111
362
380
5
9
76
39
43
18
44
13
26
9
150
80
44
35
30
1.2
1.6
1.6
3.9
5.4
5.8
8.6
10.9
13.7
49.7
50.4
51.6
57.4
57.4
58.6
60.1
61.3
61.7
61.7
63.6
63.6
78.9
80.8
82.8
82.8
85.2
This agrees with the calculated downwind variation of the GLC (see
Fig. 3).  The  total of early fatalities listed above is rather low by
present estimates, since many of the deaths were not recorded.

    The best concentration monitors available in the Bhopal gas
disaster were  the affected trees, plants, crops, and soil, which were
studied by Indian scientific organizations.  The imprint of the toxic
gas could be discerned on all types of vegetation including tall trees,
shrubs, and ground-level crops.  Vegetable and spice crops were most
affected in the eastern, southeastern, and southern directions from
the factory.   Some vegetation looked as if it had been burned.  It was
observed that  the same species of plants, which were badly damaged by
the gas in other areas, were mostly unaffected near the water. Also,
the same species of crops showed relatively less damage in fields
which were irrigated on the evening preceding the accident.  This
could be due to the great affinity of NIC to water and its rapid
hydrolysis resulting from the increased moisture content in the soil,
vegetation, and surface air.  Some of the large trees were totally
defoliated, while others showed acute to mild effects, depending on
the species and location relative to the factory.  Table 5 shows the
extent of the vegetation damage in terms of area.  These vegetation
                                     A-14

-------
Table 5.  Classification of Vegetation Damage Due to Exposure  to  the
          Toxic Gases.
Zone
A
B
C
D
Classification
Severely affected
Badly affected
Moderately affected
Mildly affected
Area
3.5
10
6
5
(km2)




damage zones, shown in Figure 5, indicate that trees and vegetation
within zones A, B and C could indeed be affected by exposure to the
spreading gas, with estimated concentrations ranging from very high to
moderate values.  Zones A and B correspond reasonably well to zones I,
II, and III shown in Fig. 4.  The vegetation-damage contours are also
tilted towards the SE, similar to the concentration contours.

MEDICAL CARE,*RELIEF, AND REHABILITATION

    Inhalation of MIC, mixed initially with unknown amounts of other
poisonous gases (see Table 3) from the leak, is considered to have
resulted in the large number of deaths in humans and animals.  The
initial symptoms displayed by the survivors were burning sensation in
eyes, nose and throat, breathing difficulty, coughing, vomiting,
headache, lethargy, and disorientation.  The main target organs were
the eyes and the lungs.  Lung damage.was the cause of death in most of
the casualties.

    While vision and condition of the injured eyes have dramatically
improved following medical treatment, lung problems persist in many
victims.  Pulmonary edema resulting from the exposure to MIC gas was
treated with steroids, bronchodilators, oxygen, and antibiotics; a
large number of patients have recovered.  However, many of the sur-
vivors were left with considerable damage to their small airways and
alveoli.  The worst affected among this group were children.  About
39* of the patients showed moderate to severe pulmonary disability.
Those who display normal functional status may still have significant
lung pathology.  Open lung biopsy performed in some of the severely
affected patients has led to the detection of bronchiotitis obliterans
and interstitial fibrosis.  Attempts are being made to Identify
lesions in victims in the early stages, and treat them using drugs and
physiotherapy.

    Miscarriages and stillbirths were frequent to the approximately
3000 pregnant women who were exposed to the gases.  In the first 20
weeks, 436 spontaneous abortions occurred out of 2600 pregnancies; the
normal rate for Bhopal was 6 to 10%.  Gynecological disorders in
various forms were reported in a large percentage of the women exposed
to the gases.
                                    A-15

-------
    Cyanide  poisoning  in  the survivors  resulted  from  direct exposure
to HCN gas,  during the first hours  in the areas  closest to the plant,
as well  as indirect  exposure owing  to unusual generation of toxic
amounts  of cyanide in  the body after exposure to NIC.  Some experts
believe  that the  "cyanogen pool" resulting  from  the latter can persist
in the human body for  a very long time.  The Indian Council of Medical
Research (ICMR) studied this controversial  issue, and recommended
continued treatment  of symptomatic  patients with sodium thlosulfate,
the antidote for  cyanide  poisoning, under careful medical supervision.

    The  state government  has opened 16  dispensaries,  a new 30-bed
hospital with modern equipment, and 25  mobile clinics in the affected
areas, in order to provide effective medical care to  the Bhopal gas
victims.  A  60-bed ward and MIC clinic, equipped with a computerized
blood gas analyzer,  was established in  the Medical College Hospital to
undertake various sophisticated investigations and develop the best
methods  of treatment.  In addition, 100 more beds were reserved
exclusively  for the  gas victims in  other hospitals in Bhopal.
Disaster relief measures  undertaken by  the government include
financial assistance to affected families, cash  grants to low income
group victims, free  rations of essential commodities, and supplies of
nutritious food to pregnant women,  lactating mothers, and infants.  A
comprehensive rehabilitation plan consisting of  housing and urban
development,  vocational training and employment  schemes, and
establishment of  work  centers and community facilities, is being
implemented  in the affected areas.  Many voluntary organizations are
supplementing the government efforts in medical  care and
rehabilitation.

    Numerous  studies on the Bhopal  gas disaster  have been carried out
in India, and many others are still in progress  at various government
laboratories, universities, medical colleges, and hospitals.   The
biomedical and toxicological studies on the long-term effects of MIC
gas on the survivors are  being coordinated by the ICMR.  The Council
of Scientific and Industrial Research is coordinating the scientific
and technical projects.   Various issues raised by the Bhopal tragedy
(such as development policy, Industrial hazards  and public safety,
occupational  health, environmental  protection, legislation, and other
topics)  are  being studied and addressed by a number of government
agencies, and by  scientific and public interest  groups.  The Central
Insecticides  Board has banned the use of 25 highly toxic pesticides,
including the two MIC-based products, Sevin and  Temik.

CONCLUSIONS

    The most  important legacy of the Bhopal tragedy is its enormous
worldwide impact  in raising the level of public  awareness and concern
over the hazards of chemical plants to the general public, especially
in those plants involving toxic chemicals.  This will inevitably
generate change in safety procedures, equipment  to prevent and contain
toxic releases,  and the kind of Information given to  the public about
chemical hazards.   Most chemical companies have  perceived the need to
reassure the public of the safety of their operations after Bhopal.
                                    A-16

-------
    The  Bhopal  disaster  bears  many important lessons of immediate
global relevance, which  can  be briefly summarized as follows:

(1) Information:  Industries  with potentially dangerous  processes
should work  together with  state and local  officials  to  maintain a
current  understanding  of what  they are doing,  giving details of the
toxic substances, potential  hazards,  toxicological effects, medical
treatment, and  other pertinent information.

(2) Industrial  siting  and  urban planning:  Hazardous  industries should
be  located far  away from population centers  with  due regard to poten-
tial impacts on the total  environment.   Existing  chemical plants  in
urban areas  should  be  relocated,  if possible,  or  isolated from dense
residential  areas by a 2-3 km  radius  green buffer zone  planted with
sensitive tree  and  vegetation  species to act as chemical monitors.

(3) Standards and enforcement:  Industrial  safety, worker and community
health,  and  environmental  regulatory  standards appropriate for each
industry should be  established and strictly  enforced.   Regulatory
agencies should carry  out  frequent inspections of hazardous plants to
examine  the  safety  of  processes and equipment, and to verify compliance
with regulations.

(4) Emergency response:  This consists of several  components.
The scientific  program should  address atmospheric behavior of toxic
chemicals including transport,  dispersion, transformation, and
deposition,  ultilizing appropriate models  and  monitoring.  The medical
response should require  that hospitals  and physicians be knowledgeable
about the toxicology and treatment for  exposure to locally used toxic
chemicals, and  should  be prepared  to  handle  any emergency.  The state
and local authorities, in  cooperation with the concerned plant, should
develop  effective communications,  evacuation plans,  and community
education programs.

(5) Management: The management  of  chemical companies at all levels,
in view  of their responsibilities  for industrial  and public safety,
should play  a leading  and  constructive  role  in most  of the areas
listed above, especially in  providing information and developing
appropriate  emergency  response  programs.   Procedures involving the
manufacture,  storage,  transportation, and  disposal of toxic chemicals
should be periodically reviewed to  minimize  potential hazards to
public health and safety.  Worker  training and plant safety audit
programs should be  intensified.

     Bhopal  holds out  lessons  that  have  to do  with social
responsibilities of science  and technology,  and basic human concerns
for safety,  for standards, and  for  good  common sense.   If these
lessons are  lost, then it  is all too  possible  that accidents such as
this may be  repeated in  another factory  in another city in another
way.  A fitting memorial to  the victims  of the Bhopal Gas Tragedy
would be to  make sure that this shall not  happen  again.
                                   A-17

-------
                       ACKNOWLEDGEMENTS

    This work was performed under an agreement between the National
Oceanic and Atmospheric Administration and the U.S. Department of
Energy.  The literature search was accomplished while the first author
was visiting the Center for Atmospheric Sciences, New Delhi, during
the summer of 1985 under the ATDD/IIT Collaborative Research Program
in Air Pollution, which is part of a multi-institutional project
supported by the U. S. National Science Foundation.  Many Individuals
and publications including technical books and reports, scientific
Journals, news papers and magazines were consulted in the course of
this work.  It is impossible to list all of these articles and their
authors; only the key references and general sources are given here.
                           DISCLAIMER

     This paper has been reviewed by the Air Resources Laboratory
(ARL), National Oceanic and Atmospheric Administration (NOAA), and
approved for publication.  Approval does not signify that the contents
necessarily reflect the views and policies of NOAA or any other U.S.
Government agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use by NOAA/ARL.
Use of information from this publication for publicity and advertising
purposes is not authorized.
                                    A-18

-------
                           REFERENCES

Briggs, G. A.,  1973: Diffusion  estimation  for  small emissions.  ATDL
    Annual Report,  NOAA,  Oak  Ridge, TN,  83-146.

Ekalavya, 1985:  Bhopal  -  A reconstruction  of the gas tragedy, its
    background  and  aftermath, from press reports and local  informa-
    tion.  Ekalavya Report-El/208, Arera Colony, Bhopal.

Hicks,  B. B., 1985: Personal  communication.

Holland, J.  Z.,  1953: USAEC Report ORO-99, Washington, D.C., 584 pp.

Kumari, M.,  1985: Diurnal variation of mean mixing depths in different
    months at Delhi.  Mausam  36, 71-74.

Reid, R. C,, and T. K.  Sherwood, 1958: The Properties of Gases and
    Liquids.  McGraw Hill Book  Company,  Inc., New York, NY.

Sax, N. I.,  1979: Dangerous Properties of  Industrial Materials.  Van
    Nostrand Reinhold Co.,  New  York, NY, 1118 pp.

Singh,  M. P., and S. Ghosh, 1985: Perspectives in air pollution model-
    ing with a  special  reference to the  Bhopal Gas Tragedy.  CAS Report,
    Indian Institute of Technology, New  Delhi.

Singh,  H. P., and S. Ghosh, 1987: Bhopal Gas Tragedy - Model
    simulation  of the diffusion scenario.  Accepted for publication in
    J.  Hazardous Materials.

Schwartz, S. E., 1984:  Acid Precipitation  Series. Vol. 3, Chapter 4.
    J.  I. Teasley,  Series Ed.,  Butterworth Publishers, Stoneham, MA.

Weast,  R. C., 1982: CRC Handbook of Chemistry and Physics. 63rd Edi-
    tion, CRC Press, Inc.,  Boca Raton, PL.

              MAGAZINES AND NEWS PAPERS  (Unreferenced)

Chemical and Engineering  News,  Washington, D.C., February 11, April 8,
    and December 2, 1985.

India Today, New Delhi, December 31, 1984.

The New York Times, New York, December 10, 1984.

The Illustrated Weekly  of India, Bombay, September 1, and September 15,
    1985.

The Statesman, New Delhi, March 28 and May 4, 1985.

The Sunday, New Delhi,  April 13, 1985.

The Times of India,  Bombay, December 16, 1984, March 25 and April 17,
    1985.
                                   A-19

-------
 Bhopal
            HAMIDlAv'.. .'i
                         : '-"SULTANIA
                         ^-="== LOWER
                            ZFLAKE:
::::--..::::::....:::::—:. HAM ID IA V
        ""—^COLLEGE
  MINISTERS'
  BUNGALOWS
                                                          SCALE
                                                         Kilometers
Figure 1.  A partial Map of Bhopal  showing the location of the DCIL
          pesticide plant and its  environs.  The dotted area shows the
          localities affected by the  toxic gases.
                                     A-20

-------
                                                                                       ATDL-M86/624
'ro
                                               MOISTURE
                    ADVECTION BY    - .-:,^-w:i»rJl^ •• CHEMICAL •...•:.;   - -. .
                        ^"*   ^^   ^^   * ***** *"••*• * "• • * ^^^^^^jBH^^   i™ £3 AM O r***/^ Q ft M A ~r" i /*x K i "
                    WIND; BENT-0/ER >LUMEy;.;'^J//.'.^^^nMATION;:.;.. -
                                                                                     DEPOSITION
                                                                  WATER,
Soil and water:
    MIC	
                                                       Adsorption
                                                       in soil surface
                                       Humans and Animals	+• MA (major  metabolite)
                                Figure 2. A »che»atic diagraa ahcmine the atmospheric behavior and fate of
                                        MIC plu»e in the environment.   Methylanine (MA) is the major
                                        product of MIC reactions in toil and water.

-------
                                                                                             ATDL-M 86/622
ro
ro
            E
            ex
            ex
            Z
               60
               40
cc

£  20
UJ
o
Z
o
o
                 0
                 200
                                                      (a]
              400     600     800

                        x (m)
1000
                                                 30 i	r
                              Figure 3.
                                               -
                                       45 .in of ti«e integration: (a) near field,  (b) far field.

-------
   Bhopal
                         FIRDAUS
                         NAGAR
   E.M.E. CENTER
                                                      CHOLA
                                                      KENCHI
               SHAHAJAHA^
               NABAD   _ __
                     /  22
                                J.P. NAGAR
                                      2
                                   KAZI CAMP
                                                 ILWAY KHAJANCHI
                                              COLONY  BAGHV  1

                                                       6     I

                                                    II    II  /
                                                    CHANDBAD/


                                             RAILWAY-
                                             STATION' *
                       SINDHI
                       COLONY
                               STRAW
                               PRODUCT
 OLD
 SECRETARIAT J
                                        BUS
                                      • STAND
              HAMtOIA
              HOSPITAL
                                                AISH BAG
                                                STADIUM
                              SULTANIA
                              HOSPITAL
                                          .CENTRAL
                                         IL1BRARY
                            JAMA  11
                            MASJID
~~~~~~  ~~LAKE I'.'."."""
                                                JAHANGIRAi?
                                                BAD
                              ..- _~ LOWER r
                                 "LAKE
         ::~~HAMIDIA
          --—COLLEGE
                                                            CENTRAL
                                                            SCHOOLX
                                         LAL PARADE
                                         GROUNDS
               CHAR
               BUNGALOW
                                         BIRLA  X
                                         MANDIR
                                                             LEGEND
                                                          ZONE I  > 50ppm
                                                          ZONE II > 15ppm
                                                          ZONE III > I.Sppm
                                                          ZONE IV < 1 ppm
MINISTERS'
BUNGALOWS
                                                               SCALE
                                                              Kilometers
   Figure 4.  Modeled concentration contours demarcating the various zones of
              the gas affected areas based on the estimated 6LC.  The numbers
              1-26 denote the localities in Bhopal affected by the gas leak,
              where the 6LC are specifically calculated and given in Table 4.
                                       A-23

-------
 Bhopal
                          ™^ LOWER_=
                           ""=_" LAKE:
 MINISTERS-
 BUNGALOWS
       LEGEND
SEVERELY AFFECTED (3.5 km2) °
BADLY AFFECTED (10 km2)
MODERATELY AFFECTED (
MILDLY AFFECTED (5 km2)

           SCALE
                                                          Kilometers
Figure 5.  Vegetation danage contours demarcating the various  zones of
          injury to plants by the  toxic gases.  These contours are baaed
          on a detailed survey of  vegetation and crops in Bhopal conducted
          by the Indian Agricultural Research Institute and the Central
          Board for Prevention and Control of Pollution, following the gas
          leak.
                                       A-24

-------
        APPENDIX B

  On  the Development of
Reactive,  Dense Gas Models
            B-l

-------
                          On the Development of
                       Reactive, Dense Gas Models*
                               B. B. Hicks
                                   and
                            W. R. Pendergrass

                       NOAA, Atmospheric Turbulence
                          and Diffusion Division
                   P. 0. Box 2456, Oak Ridge, TN  37831
                                          ATDD Contribution No. 86/26
March 31, 1987
ODR/16
* For presentation at the Joint EPA/DOE Technical Workshop on
Determination of Atmospheric Dilution for Emergency Preparedness, Research
Triangle Park, NC, 27711, and for inclusion in the proceedings.
                                    B-3

-------
ABSTRACT



     Chemical reactions between released pollutants and atmospheric



constituents can influence atmospheric dispersion if the heats of reaction



are such that buoyancy can be affected or mechanical turbulence damped.



If the reaction is sufficiently exothermic, plume rise can be increased;



if sufficiently endothermic, then plume rise can be suppressed.  It is



necessary to consider the potential influence of such reactions on the



mixing rates characteristic of the ambient air.  Reaction rates are also a



critical issue.. If reactions are completed rapidly, then in concept it is



possible to assimilate the consequences of the reactions in the source



term initialization or in the first time step of a numerical model.  On



the other hand, if reaction rates are slow in comparison with the time



step of a simple model, then a more complicated and detailed modeling



effort may be required.
                                     B-4

-------
1. INTRODUCTION




     The wide variety of needs for numerical models of atmospheric




dispersion imposes requirements for models of a wide range of




complexities.  It is clear that simple models giving conservative




estimates are most suitable for purposes of real-time emergency response,




whereas more sophisticated models can be used for assessment of risk, for




planning response strategies, or for evaluation of the consequences of



previous accidents.  The range of modeling capabilities must also be




adequate to.encompass the spectrum of circumstances to be addressed,




involving terrain complexity, different meteorological conditions, and




source configurations.






     In general, the complexity of the model best suited in some special




application is increased by increasing complexity of the circumstance and




limited by the time interval available' in which to run the model.




Considerable additional complexity can arise if the pollutant involved is




dense, or if it reacts chemically after emission.  The case of dense gas




dispersion has received extensive attention, largely as a result of




problems associated with the transport and storage of liquified gaseous




fuels.  In essence, the density interface is a stabilizing influence,




tending to isolate the underlying "bubble" of dense gas from the




turbulence of the ambient atmosphere.  The volume of dense gas itself can




then be free (if dense enough) to respond to its own dynamical forces,




largely independent of the atmosphere passing over it.  Spreading may be




rapid, but dilution may be correspondingly slow.






     Endothermic chemical reactions impose a further set of complications,




that can serve to make a pool of emitted trace gas act as if it were a
                                    B-5

-------
genuine dense gas, even though quite dilute.  On the other hand, some

chemical reactions are exothermic, and will tend to increase vertical

mixing between the trace gas and the ambient air.


     The present purpose is to explore the options available to modelers

faced with the need to consider the consequences of exothermic and

endothermic reactions.  We start with the assumption that a suitable

modeling framework is available, including dense-gas formulations

appropriate to the case under consideration, with which chemically-induced

dynamic factors must then be combined.


2.  SOME THEORETICAL CONSIDERATIONS

     Consider a trace gas released into the air and reacting with some

atmospheric constituent, such as atmospheric water vapor.  The chemical

reaction is then of the general type

                              kc
                   K! + H20  	>  X2 + X3 -i- J                       (1)

where species X^ reacts with H20 (in this example) to form species X2 and

X3 with rate constant kc (s~l) and releasing thermal energy J (Joules

mole"1).  If J is positive, then the reaction enhances the mixing due to

buoyancy.  In this case, it is informative to consider the role of the"

additional thermal energy in the context of related features of the

ambient air.


     In a convective atmosphere with sensible heat flux H, the rate of

generation of turbulent energy associated with the heat flux is


                    HB = (H/cp).(g/e)


where cp is the specific heat of air at constant pressure, g is the
                                    B-6

-------
acceleration due to gravity, and 0  is potential  temperature  (absolute).



The equivalent mechanical term is






                     HM = -T«(3u/3z),                                  (3)






where T is the surface momentum flux and  3u/3z is  the  local  wind  gradient




at the height in question.  The negative  sign is a consequence of the  sign




convention, with positive T being directed  away  from the  surface  and 8u/3z




positive when wind increases with height.






     In classical micrometeorology, the importance of  buoyancy relative  to




shear-produced mechanical diffusion is quantified  by the  ratio of (2)  to



(3):






                   C = -(gH/cp0)/(T-3u/3z)



                                        <|>m-l




                                                                       (4)
Here,  standard  micrometeorological  relations  have  been  invoked  to  relate



the  local wind  gradient  and  shear stress to the  friction velocity, u,f



involving air density  p  and  height  z,  and  introducing the  von Karman



constant k  and  the  stability-dependent dimensionless  wind  shear m.  The



quantity L  is the Monin-Obukhov length scale  of  turbulence,  initially



derived from dimensional arguments.  The quantity  C is  thus  an  index



of dynamic  instability associated with buoyancy, much like the  familiar



index  z/L.





     In this context,  it is  useful  to  consider a second buoyancy term,




like !IB( DUt representing the  consequences of an exothermic  chemical




reaction.
                                    B-7

-------
                      nc = kcJpv/Mv



where pv is the partial density of the species Xj in air (kg m~3) and Mv



(kg mole"1) is the molecular weight of the species Xj.






     It is then apparent that if the ratio




               CB - nc/nB



                  - kcJpvcpe/(HgMv)                                   (6)



is large (with respect to unity, then the reaction is sufficiently



exothermic to modify buoyancy significantly, in unstable conditions.





     In unstable stratification, an endothermic reaction will tend to



reduce the rate of buoyant mixing.  In general, therefore, a modified



index of the net effective instability can be postulated:



              -?' =  (HB + nc)/HM                                      (7)



where nc is positive for an exothermic reaction (positive J) and negative



for endothermic.





     Although the modified stability index C1 has been developed here for



unstable conditions, its generality is not so constrained.  Just as  z/L



and Ri are stability indices that  extend across the range of stable  and



unstable stratification, so does C' provide  an equivalent mechanism  for



modifying  such standard quantities in the cases of exothermic and



endothermic chemical reactions.





     Equations  (6) and (7) provide a basis for "screening"  chemical



reactions  for potential concerns related  to  modifying atmospheric  mixing.



Consideration of dense-gas effects can be included in the same  general



framework.  In this  case, consideration of  relevant  time scales provides



useful physical insight.  At a  density interface, characterized by a
                                     B-8

-------
difference Ap, the restoring force  (per unit  volume)  associated  with  the



displacement of a unit volume of denser fluid into  the  (upper) less-dense



medium is



                         Fr = g.Ap                                     (8)





A time scale  (T
-------
density variations must be used to correct ambient stratification



calculations when considering local stability=   Such matters have been



dealt with extensively elsewhere.







3.  DISCUSSION



     Inspection of the relations involving J reveals a few hidden



difficulties.  In particular, the chemical reaction rate kc is not usually



a constant, but depends on variables which might include temperature,



pressure, and solar radiation.  Furthermore, the reaction may be



equivalent to a gas-phase titration, in which the reaction is controlled



by the rate of delivery of one gaseous reactant or the other.  Thus, from



the present viewpoint the specification of kc is far from trivial.





     Inspection of equations  (6) and (7) reveals several intriguing



conclusions.



    (a)  Plume rise enhancement/suppression



         •  The practical effect of the release of heat of reaction  is



            likely to be greatest in near-neutral conditions (when H is



            small).  The diurnal cycle of H is large, such that H varies


                               —2                              -2
            from -10 to -20 W m   at night to more than 200 W m   at midday,



            typically.  The influence of J is therefore likely to be



            greatest near dawn and dusk.



    (b)  Mechanical mixing with the ambient air



         •  The suppression of dilution by an endothermic reaction  is a



             strong function of the friction velocity, suggesting an



             inverse cubic dependence on wind speed.
                                    B-10

-------
         •   The  roughness  of the surrounding surface is important,  insofar




            as  it controls u,,, in given wind speed conditions.   Minimum




            boundary dilution rates will occur in light winds,  over smooth




            terrain.




    (c)   The role of the reaction rate




         •   The  reaction rate enters as a first-order factor in both




            CB  and CM-   The appropriate reaction rate is the effective



            value in the conditions of interest, which will generally be



            less than the rate based on chemical consideration alone.  The




            problem that arises is associated with turbulent mixing;




            chemical reaction consumes material available in any specific




            volume, and hence the rate of resupply of reactions must be




            considered.  Thus, if the rates kc in equations (6) and (7)




            are based on chemistry alone, then the properties Cn and
                                                               o



            C« should be considered as indices of the potential
             M


            importance of the heat of reaction, rather than an indication




            of actual importance in any specific instance.








4.  A PRACTICAL APPROACH



    The need ,to consider the special characteristics of reactive gases in




dispersion models  leads immediately to two fundamental questions:






    (a)  Does the reaction  cause the dynamic behavior of the atmosphere to




         be modified?



    (b)  Is the reaction completed fast enough that the consequences  can




         be accommodated in the source term of a relatively standard




         dispersion code, or in its first time step?
                                     B-ll

-------
    The answers to those questions will determine the complexity of the




model which must then be used.  In some instances, such as when a reaction




is strongly exothermic yet quite slow, a complicated model is likely to be




required in all situations, and the need for rapid computation as for




real-time emergency response are unlikely to be realized.






    An objective approach is to use a detailed simulation of the chemistry




to derive the necessary answers concerning reaction rates and




exothermicity in natural conditions.  If the results of these explorations




are satisfactory, then perhaps simpler models can be designed without loss




of either generality or applicability.






    This general philosophy is presently being tested in the development of




a model of dispersion of uranium hexafluoride.  This material is used in




uranium processing and isotopic enrichment.  Upon release into the




atmosphere, it reacts exothermically with water vapor, generating




particulate uranyl fluoride and gaseous hydrogen fluoride.  The amount of




heat released and the rapidity of the reaction are clearly dependent on




ambient humidity.  In this instance, consideration of scaling properties




such as were discussed above is not completely satisfying, since factors




not included in the development of the scaling properties might be critical.






     The chemical reaction is as follows:






                    UFg + 2H20 	> U02F2 + 4HF + J.              (14)






The quantity J is a function of temperature and pressure, but in typical




conditions involving gaseous UFQ and water vapor, J * 5.86 x 104 kJ/kg



based on chemical considerations alone (see summary by Just et al.. 1985).
                                    B-12

-------
    A sophisticated numerical model of UF6 reactions  in air was developed




earlier (Just, et al., 1985).  This includes a detailed description of the




turbulent resupply of reactants as material is consumed (Varma. 1982), and




provides a convenient tool for investigating the rapidity of UF6 reactions




in air.  Figures 1 and 2  illustrate the results.   In  Figure 1, the reaction




rate is seen  to be short  in comparison to the time step of typical




puff-dispersion models (usually about five minutes),  except for very dry




conditions.   Figure 2 shows that there is little need to consider




background atmospheric stability as a controlling  factor, since in even




the most stable case  considered the reaction appears  to be completed in 30




seconds.  Hence, in this  particular case  it appears adequate to assimilate




the chemistry and  its consequences in the first time  step of a standard




puff model,  and to use the more detailed  chemical  model to initialize the




chemically-simpler scheme.








5.  CONCLUSIONS




    Two major questions  arise when adding chemical reactions into an




existing dispersion  scheme  (either of possible trace  gases or dense gas).




First, the rapidity of the reactions must be considered.  If the




reactions are completed  quickly in comparison with the  time step of the




model  under  consideration, then the relevant chemical reactions can be




accommodated within  the  initial time step,  or perhaps assimilated  in the




specification of the  source  term.






    Second,  consideration must be given  to  the  consequences of  the  heat  of




reaction.  If exothermic, the reaction is essentially a destabilizing




factor, working  to increase  plume rise and  to enhance buoyant  mixing  (and




subsequent dilution  rates).
                                     13-13

-------
     If endothermic, the reaction is essentially a stabilizing influence,




increasing local stratification, and hence decreasing mixing (and




dilution) rates.  If all other factors are equal, a bubble of endothermic




material must maintain its integrity for a longer time than a bubble of




exothermic material.






    An examination of these factors can be based either on first




principles, by consideration of the relative magnitudes of the dynamic




effects of the heat of reaction, or on the use of a detailed model.   In




practice, if the reaction involves a trace gas constituent of the ambient




atmosphere (e.g., H20 or C02), then the use of a detailed chemical model



appears especially beneficial.  In such cases, the reaction can proceed as




a gas-phase titration in which the rates involved and the thermodynamic




consequences are controlled by the rates at which reactant gases are fed to



the reacting volume.
                                    B-14

-------
6.   REFERENCES

Just, R. A. and  W. R. Williams.  1985:  A computer program for
    simulating the atmospheric dispersion of UF6 and other
    reactive gases having positive, neutral, or negative
    buoyancy.   Martin Marietta Energy Systems. Inc., Engineering
    document No. K/D-5694.  Oak Ridge, TN.

Varma, A. K. ,  1982:  Development of models for the analysis of
    atmospheric Releases of Pressurized Uranium Hexafluoride,"
    ARAP Report No. 482. Aeronautical Research Associates of
    Princeton, Inc., Princeton, NJ.
                                    B-15

-------
                                               ATDL-M86/647
           100
         V)
            10
                                                   i  i  i  i LL
                                     10                   100
                      WATER  MIXING  RATIO   (g/kg)
Figure 1.  The humidity dependence of the exponential time scale of
     reaction of UFg released into air, as computed by a sophisticated
     chemistry/turbulence mixing model (Just and Williams, 1985).  Time
     scales correspond to (1-1/e) completion of the reaction.
                                  B-16

-------
   CO
   CO
   o

   H
   _J
   CD

   CO
         0
•—•
                    1
                                        ATDL-M86/648
                                         T
                                        •—•
  10          20        30
    TRAVEL TIME  (s)
40
Figure 2  The dependence on stability class of the exponential
     characteristic time scale, as plotted in Figure 1.
                                 B-17

-------
                              APPENDIX C

                       Mathematical Models for
                       Atmospheric Dispersion of
                   Hazardous Chemical Gas Releases:
                             An Overview*
     *Presented  at  the  AIChE  Center  for  Chemical  Process  Safety,
International   Symposium  on  Preventing   Major   Chemical   Accidents,
February 3-5, 1987, Washington, DC.
                                  C-l

-------
              MATHEMATICAL MODELS FOR
             ATMOSPHERIC DISPERSION OF
          HAZARDOUS CHEMICAL GAS RELEASES:
                    AN OVERVIEW
                    Jerry Havens

         Department of Chemical Engineering
               University of Arkansas
               Fayetteville, Arkansas
Mathematical models are required for prediction of
atmospheric dispersion of heavier-than-air gases to
assess hazards of unconfined vapor cloud combustion
and toxicity.  The dispersion process involves three
more-or-less distinct regimes of fluid flow:
buoyancy-dominated, stably stratified, and passive
dispersion.  For releases on uniform terrain,
similarity models are available for the description
of all three regimes with sufficient accuracy "for
most risk assessment and emergency response
requirements.  3-D mathematical models can, in
principle, simulate the spatial and temporal
dispersion process without artificial separation of
the flow into separate regimes and may be able to
provide for effects of terrain and wake turbulence.
Evaluation of 3-D models is underway, but most of the.
work is biased strongly toward evaluation against
data for dispersion in the absence of terrain and
wake turbulence effects, even though description of
these effects is a primary motivation for their use.
Future research should be directed to the
verification of models which provide for the effects
of terrain and wake turbulence, and for description
of jetting releases, aerosol formation, chemical
reaction, and deposition.
                          C-3

-------
                         INTRODUCTION

    The recent accidents at Bhopal and Mexico City illustrate
the potential dangers of acutely toxic or combustible gas
clouds.  These catastrophes have intensified pressures on
Government to "do something" about major industrial accidents.
The European Community's Seveso Directive calls for tighter
regulatory controls to reduce the likelihood and consequences
of "major hazards", and the response has been to implement
tighter regulations of activities using hazardous chemicals.
The regulations require provision of evidence that hazards to
man and the environment have been identified and that suitable
precautions have been taken to minimize occurrence and
potential consequences.  Many chemical manufacturing,  storage,
and transportation activities with large inventories of
hazardous materials have had specific requirements imposed,
including
    • submission of safety assessments
    • preparation of on-site emergency response plans
    • preparation, in cooperation with the local authorities,
        of off-site emergency plans
    • provision of information to people in the vicinity who
        might be affected by a major accident
Pressures for similar regulatory controls are increasing in the
United States.
    The concept of separation distances, by which the
consequences of major accidents are ameliorated, is implicit in
regulations being considered world-wide.  Methods are required
for prediction of atmospheric dispersion of hazardous gas
releases, of which the main hazards are vapor cloud combustion
and acute toxicity, to aid in the determination of appropriate
separation distances.  Separation distances are required for
siting fixed hazardous chemicals operations, and for the
                               C-4

-------
preparation of emergency response plans.  Methods are also
required for predicting atmospheric gas dispersion in support
of emergency response to transportation accidents.
    Prediction of separation distances for use in risk
assessment and emergency response is controversial.  The
inclusion of this review of mathematical models for atmospheric
dispersion of hazardous chemical gas releases, and lesser
consideration of modeling methods applicable to predicting
other accident phenomena such as fire/explosion damage or
toxicity effects, indicates the importance attached to this
question by the chemical industry and regulatory parties.  The
Environmental Protection Agency and Department of Energy
jointly sponsored a "Workshop on Determination of Atmospheric
Dilution for Emergency Preparedness" in October 1986 with the
following principal objectives.
    • to review current methods for determining release
        characteristics, source strength, and dispersion
        of hazardous gases released into the atmosphere for use
        in hazard evaluation and emergency response
    • to provide recommendations for choosing among the current
        methods for immediate use
    • to assess the specific strengths and weaknesses of
        available methods and make recommendations for their
        improvement.
The controversy can be traced to predictions in the mid-70's of
the extent of the hazard zone around an accidental release of
liquefied natural gas (1).   A postulated accident of concern
was the rupture of one or more 25,000 cubic meter LNG ship
cargo tanks.  Since the vapor cloud which would be formed
following the spillage of such a large quantity of LNG onto
water would be cold, it would be expected to be significantly
heavier than air.  Air pollutant dispersion models (primarily
Gaussian similarity models with ad hoc provision for gravity
spreading of the heavy gas) were applied, with widely disparate
results, to predict the extent of flammable gas-air mixtures
                               C-5

-------
from these "worst credible"  releases  (2).   The  disparity of
such dispersion predictions  made more uncertain the  risks
associated with LNG importation, and  fostered an extensive
research program aimed at better understanding  the atmospheric
dispersion of large quantities of heavy gases which  are
accidentally released.
    Since van Ulden's proposal (3) of a simple  mathematical
model to describe the gravity spread  and subsequent  dispersion
of a large volume of heavy gas instantaneously  released  into
the atmosphere, there have been numerous other  models proposed
and extensive laboratory and field test experimental programs
to provide data for model validation  have been  completed.
Several recent reviews which describe the mathematical models
are available (4,5,6).  The field test programs completed at
China Lake, California (7),  Maplin Sands, England (8), Thorney
Island, England (9),  and Frenchman Flat, Nevada (10) to  provide
data for dispersion model validation have also  been reported.
    For dispersion scenarios where the gas cloud can be
represented as having a regular shape, similarity mathematical
models  (such as the familiar Gaussian models in widespread use
for passive atmospheric pollutant dispersion)  are applicable.
Similarity approximations are also justified for modeling
accidental releases of large quantities of heavy chemical gases
under certain conditions.
    Complex (three-dimensional) mathematical models can, in
principle, be used to simulate  the spatial and temporal
dispersion processes  without the  "artificial"  separation of the
flow into separate regimes required by the similarity models.
Predictions can me made of dispersion effects due to nonuniform
terrain and wake  turbulence.  It may also be possible to
provide useful estimates of concentration fluctuations  around
the predicted mean (time average) values.
    This paper reviews the mathematical modeling methods which
have been developed for predicting atmospheric dispersion  of
hazardous chemical gas releases.  The emphasis is on
                               C-6

-------
mathematical models for dispersion of heavier-than-air gases.
Such emphasis is justified because the majority of hazardous
chemical gases form heavy gas clouds following accidental
release, and because such clouds remain near the ground for
longer times and to greater distances.
             PHENOMENOLOGY OF HEAVY GAS  DISPERSION
                    AND.MODEL APPLICABILITY
     The typical heavy gas dispersion process involves three
more-or-less distinct regimes of fluid flow.  Immediately
following rapid release of a large quantity of heavy gas, a
cloud having similar vertical and horizontal dimensions may
form.   Initial slumping and lateral spreading motion continues
until the kinetic energy of the buoyancy driven flow is
dissipated.  As the dispersion proceeds, stable stratification
due  to  the negative vertical density gradient decreases until
the  process can be represented as a neutrally buoyant plume (or
cloud)  embedded in the wind flow.
     For heavy gas releases at ground level on uniform terrain
with unobstructed wind each of the three flow regimes
(buoyancy-dominated, stably stratified, and passive dispersion)
can  be  approximated as a cloud (or plume) having a regular
shape,  and can be described with similarity models.

                       Similarity Models
     When the cloud formation time tf is very small compared to
the  time of cloud travel to the maximum distance exposed to the
concentration of interest tt, an "instantaneous puff"
representation is applicable; if tf »  tt, a stationary  "plume"
representation of the cloud is indicated.
     Figure 1 depicts the cloud shape which has been most
frequently used in similarity heavy gas dispersion models  for
instantaneous releases.  An initial volume of gas, usually
                                C-7

-------
                      Wind Direction
          Source
    FIGURE  1.   Cloud  shape  for  instantaneous release.


represented as a vertically oriented cylinder,  is  placed in the
flow field.  The cloud is moved downwind with a velocity
determined from the wind vertical profile.   Figure 2 depicts
the cloud shapes which have been most frequently used in models
for continuous releases.  The principal dimensions of the puff
or plume change as a result of gravity spreading (assumed to
occur crosswind only in the plume representation)  and
entrainment of air across the top and/or side boundaries.
    Buovancv-Dominated Flow Regime.  For heavy gas releases
with initially similar vertical and horizontal dimensions there
is conclusive evidence that the rapid gravity-driven flow
results in large-scale turbulent structures which  effect rapid
dilution of the cloud (11,12,13,14,15).  Since this initial
turbulent motion can result in a ten-fold to one hundred-fold
dilution, it must be accounted for in heavy gas dispersion
predictions.  The lateral spreading that follows such releases
is most frequently modeled as a gravity current with a frontal
velocity calculated from the relation

    uf - CE (g H)1/2                                        (1)

Equation (1), which reflects the assumption of quasi-steady
exchange of cloud potential and kinetic energy, indicates a
                               C-8

-------
                   C(x,y,j
.y >«>
                             C(x,y,i) » CA(X) exp - (~rT
• |y|*«>
                       \           '(
      • ISO CONCENTRATION \         S
              CONTOURS  \     /

       FOR C«CU            V
                           \l
FIGURE  2.   Cloud  shapes for  continuous  releases.
                               C-9

-------
 step change  of the  front velocity  to  its maximum value at  the
 instant  of release.   The heavy gas must accelerate  from rest,
 and van  Ulden (16,17)  and Meroney  and Lohmeyer  (18) have
 proposed methods  for modeling the  acceleration  phase.  Air
 entrainment  at the  spreading cloud front during the buoyancy-
 dominated flow is often modeled by specifying an entrainment
 velocity proportional to the front velocity:
     ue - GI  uf                                             (2)

 Havens and Spicer (19)  reported laboratory  instantaneous
 releases,  in calm air,  of right-circular cylinders of Freon-12
 with initial volumes 0.035 and 0.51 m3.  Figure 3 summarizes
 the measured cloud  front position  (radius/V1/3) vs. time (t/T).
 The cloud front position is  well represented, for t > « 20 T,
 by the solution of  Equation  (1) with  Cg - 1.16.  Figure 3 also
 indicates  the predicted cloud frontal position vs. time
 obtained using van Ulden's model for  the acceleration phase.
 Figure 4 shows ground level,  peak-measured concentration as a
 function of  distance from the release center for the
 instantaneous  Freon-12  releases.  The volume-averaged
 concentrations of the clouds,  determined by spatial integration
 of vertical  and horizontal cloud concentration profiles,
 indicate a value for C± in Equation (2) of about 0.6.
     Stablv Stratified Flow Regime.   An intermediate phase of
 the  typical heavy gas dispersion process (between the buoyancy-
 dominated flow regime and passive dispersion) is similar to a
 stably stratified plume embedded in a mean flow.  This regime
 is characterized by  the persistence of a lateral (crosswind)
 gravity-driven flow  and vertical density stratification which
 damps turbulent mixing.  The  lateral gravity spread can be
modeled using  Equation  (1).   The vertical mixing is usually
modeled with a vertical entrainment velocity which is a
 function of the friction velocity of the flow and the
                               C-10

-------
      100-
       10-
A- 0.035 m
O- 0.054
D- 0.135
A- 0.530
                     model fit
                     (Havens and
                     Spicer, 19E5)
                                     t/ g&
                                     ~^/6"
               l   I  I

                             t*
FIGURE  3.   Cloud front position vs.  time, Freon-12,
            H/D = 1.0,  instantaneous  release.
              100-
             X
              
-------
stabilizing effect of the density gradient.  The stabilizing
effect of the density gradient is determined from a bulk
Richardson number for the flow:
    ve -
                                                           O)
The function ^ in Equation (3) is chosen to agree with
laboratory experimental measurements of mixing in density-
stratified flows.  Figure 5 shows vertical entrainment velocity
data vs. the bulk Richardson number of the flow from the
experiments reported by McQuaid (20), Kantha, Phillips, and
Azad (21), and Lofquist (22).  The plotted line represents a
curve fit of the three data sets, which cover a Richardson
number range from near zero to about 105.  This range
encompasses heavy gas dispersion scenarios of interest.  It is
noted that questions have been repeatedly raised about the
interpretation of both KFA's data and McQuaid' s experiments,
and there exist data reported by Ellison and Turner (23) and
more recently by Deardorf f (24) , Kranenberg  (25) , and Stretch
(26) which may justify some modification of the entrainment
velocity specification shown in Figure 5.  However, it is
anticipated that such modifications would favor an increase in
the entrainment velocity, and Figure 5 should reflect
conservatism.  Furthermore, modifications which may be
necessary appear to be for higher Richardson numbers than are
likely  to be experienced in heavy gas dispersion scenarios of
interest  (27).
                                C-12

-------
         .s
        10-'.
        10-*.
        io-J
        »er«
                                 10*
• McQuaid (20)
I K»nth« et «1. (21)
* Lefquist (22)
      IB4
                                                       Rl,
    FIGURE 5.  Correlation of vertical entrainment velocity
               with bulk Richardson number.
    Passive Dispersion Regime.  Vertical passive dispersion
from ground level sources  is  conventionally modeled as a
gradient transfer process  by  application of similarity
principles developed by Monin (28)  and Batchelor (29) .  The
velocity profile in a stratified  flow against a rough wall
boundary is determined from
    du    u.
                                                            (4)
where the function ^ has been determined from experimental
measurements of vertical momentum transfer by Businger (30) .
For the limiting case  of neutral stratification, \4 - 1, and
Equation (4) indicates a logarithmic velocity profile with
roughness height zr.   The  corresponding vertical diffusivity,
defined as the ratio of momentum flux to the mean velocity
gradient, is given by
    K -
        k u* z
                  (5)
                               C-13

-------
and (invoking the Reynolds analogy),  the equivalent vertical
entrainment velocity is
    v  -*                                                 (6)
     e   z
For neutral stratification, $ - 1 and ve/u* - k,  the von Karman
constant, which is about 0.4.  This result is consistent with
the extrapolation of data summarized in Figure 5  to zero
Richardson number.
    A General Purpose Similarity Model.  Havens and Spicer (31)
have developed a general purpose heavy gas dispersion
similarity model for incorporation in the U.S. Coast Guard
Hazard Assessment Computer System.  The model is  designed for
simulating dispersion from ground level sources on water or
level, unobstructed terrain.  The DEGADIS (DEnse  GAs
Dispersion) model is an adaptation of the Shell HEGADAS model
described by Colenbrander  (32,33).  The buoyancy-dominated flow
regime is simulated using a box model to predict a "secondary"
heavy gas source which is  input to the downwind dispersion
model.  The box model of the buoyancy-dominated flow regime
incorporates air entrainment at-the gravity-spreading front
based on the data correlation shown in Figures 3 and 4.  The
downwind dispersion phase  of the calculation assumes a power
law concentration distribution in the vertical and a modified
Gaussian profile in the horizontal direction, with a power law
specification for the wind profile.  Vertical mixing
(entrainment) is modeled using the data correlation shown in
Figure 5.  Horizontal dispersion1in the stably stratified flow
regime and the ensuing passive dispersion regime  (a smooth
transition, based on the vertical mixing data of  Figure  5,  is
effected by the model) is  forced  to reflect  experimental data
on horizontal dispersion  of  passive plumes from point  sources,
such as  the power law correlation of av developed by Pasquill
(34).  DEGADIS provides for  treatment  of transient (including
instantaneous) releases as a series of pseudo-steady state
releases.  The model also  provides for heat  transfer from the
                               C-14

-------
underlying surface to the cloud, as well as enhancement of
vertical mixing by the unstable temperature gradient which
results from heat transfer to the cloud.  The convective
turbulence is modeled using an approach adapted from Zeman and
Tennekes (35).
    DEGADIS has been used to simulate a wide range of field
heavy gas releases, including small to intermediate LPG (0.1 to
1 kg/s) and LNG releases (1 to 100 kg/s) on land,  large-scale
releases (10 to 150 kg/s) of LPG and LNG on water, and
instantaneous releases of approximately 5000 kg Freon/air
mixtures on land.  A comparison between the predicted and
"observed" distance to the 5%, 2-1/2%, and 1% concentration
levels for the Burro/Coyote, Maplin Sands, and Thorney Island
field tests has been made.  The observed values were determined
from reported maximum concentrations for each experiment by
drawing a visual best-fit straight line through the reported
points in the concentration range of interest; all of the
measurements used were made at heights at or below 1 m.  The
predicted distance to a given concentration level  was based on
the ground level centerline concentration calculated by
DEGADIS; for the concentrations and conditions of  interest, the
predicted concentration level is essentially constant for
heights below 1 m.  From these values, a ratio of  the observed
to the predicted distance for each experiment was  calculated.
Table I summarizes ratios of the observed to predicted
distances for the concentration levels of interest.  As well, a
90% confidence interval of these ratios is included for each
test series and for all of the experiments together.  (For
instance, the Maplin Sands comparisons indicate the ratio of
the observed to the predicted distance to the 2-1/2%
concentration level would be between 0.91 and 1.20 in nine out
of ten realizations.)  For all of the comparisons in Table I,
the ratio of observed distance  (DBS) to predicted distance
(PRE) for a given ensemble-averaged concentration level ranged
from 0.73 to 0.96 for the 5% level nine out of ten times  (i.e.
                               C-15

-------
   TABLE I.   Comparison between "Observed" and  DEGAD1S-
               Predicted  Maximum Distance  to Gas  Concentrations
               in the Flammable Concentration Range.


Test Series
Thorney Island 7
8
9
11
13
15
Thorney Island
90X Confidence
Interval 0.71 S
Haplin 22
27
29
34
35
39
43
46
47
49
50
54
56
Haplin 90X
Confidence
Interval 0.80 s
Burro 3
7
8
9
Coyote 5
6
Burro 90X
Confidence
Interval 0.46 s
Ratio of
Distance
v « 5X
1.17
1.09
1.00
0.55
0.70
1.05


(06S/PRE)5 s 1.11
0.47
0.95
0.89
1.2S
1.28
0.46
0.73
1.09
0.77
1.27
0.92
1.12
0.93


5 S 1.03   0.95 S (OBS/PRE), S 1.24

Sunroary 99X
Confidence
Interval    0.64 s (OBS/PRE)j £ 1.03    0.78 s 
-------
 90% confidence interval);  for the 2.5% level,  (OBS/PRE)  ranged
 from 0.82 to 1.03 for a 90% confidence interval;  for the 1%
 level,  (OBS/PRE) ranged from 0.95 to 1.24 for  a 90% confidence
 interval.
                    3-D Mathematical, Models
     The Gas Research Institute in the  United States  is
 sponsoring a research project (36)  to  evaluate  the SIGMET-N
 (37),  ZEPHYR (38),  MARIAH-II (39),  and FEM3  (40)  3-D
 mathematical models for LNG vapor dispersion.   3-D mathematical
 models have been also reported by Schreurs  (41),  Rioux  (42),
 Betts  et al. (43),  and Deaves (44).  These models have  had
 limited use, partly because of hesitation by regulatory
 authorities to allow their use without a  thorough technical
 evaluation.
     Comparison simulations were made in the  GRI evaluation
 project with SIGMET-N,  ZEPHYR,  MARIAH-II, and FEM3 of the Burro
 9 LNG  spill test (7).   Figure 6 shows  the centerline vertical
 and one meter elevation horizontal  5%  gas concentration
 contours predicted  at 80  seconds from  spill  start by the four
 models.   (Also shown  are  the grid cell  sizes used, the  total
 number of cells computed,  and the VAX  11-730 cpu  time required
 for the  models.)  Important differences were observed in the
 predicted maximum downwind extent as well as the maximum
 lateral  extent of the LFL contour.  These differences were
 attributed primarily to two  factors:
     • numerical diffusion  (truncation)  errors
     • turbulent mixing (closure) models
The model  evaluation program has  focused on these two model
characteristics.
                              C-17

-------
               SICXET-N
                          MAKIAB-II
                                     TOO
*
»
S 10
S
•
4J
X


--•N ['
d* V^ ,j ' \'N^'rV-'->' i
• 0 100 200 300 400 500
K
e
"100
*J
•
«4
o

t • 80 •






600
t - 80 •
* 0 100 200 300 400 500
Dovnvind Distance, n
LTL Diet. • ix. « iy, » t«- t
ZEPHYR >440 10 .67 6
S1CMEI-H 215 10 .67 6
MUUAH-IX 310 10 .67 6
TOO 375 * * *

600
Cells CPU, h
11,250 173
11,250 79
11,250 6.*
3,680 -60
                •variable, expanding grid

    FIGURE 6.  Comparison predictions of vertical centerline
               and one meter height horizontal 5% gas
               contours—Simulation of Burro 9 experiment
               with four models.
    Numerical Diffusion Properties.  The numerical solution of
advection-diffusion equations (the momentum, mass, and energy
balance equations) using finite differences or elements results
in truncation errors which cause a "numerical diffusion"
component in the solution (45).  The numerical diffusion errors
must be maintained small compared to the contributions from the
actual diffusion terms in the balance equations.  The severity
of numerical diffusion depends on the finite difference scheme
used and the discretization  (spatial step  size).
    The SIGHET-N model uses  first-order, forward-time, centered
space finite differences for the advection terms  of  all balance
equations.  The method used  is equivalent  to using  "upwind"
differencing, and test calculations for representative heavy
gas dispersion scenarios demonstrate the severity of numerical
diffusion (Havens and Schreurs, 46).
                                C-18

-------
    ZEPHYR uses second-order, centered-space finite differences
for the advection terms in the momentum balance equations.  The
contaminant gas balance equation is simulated with a Lagrangian
model in which particles representing the contaminant gas are
transported in the finite difference grid by advection at the
local (interpolated) flow velocity and diffused via a random
walk prescription.  The particle treatment of the gas
contaminant balance equation effectively avoids the problem of
numerical diffusion.  However, the random walk method used to
simulate the turbulent mixing (diffusion) of the gas is not
suitable for application to the strongly varying diffusivity
field which is expected in an LNG vapor cloud (Schreurs
et al.  (47)).
    MARIAH-II uses a second-order Crowley difference
approximation for the advection term in the gas contaminant
balance equation.  This method substantially reduces numerical
diffusion errors but can result in oscillatory behavior in
regions of high concentration gradients.  In MARIAH-II these
oscillations are identified and locally damped using Chapman's
FRAM (filtering remedy and methodology) method (48).
    FEM3 uses finite elements to approximate the spatial
variation of the problem variables.  With the first-order
function approximations which have been used in applications to
LNG vapor dispersion, FEM3 can be compared to the finite
difference methods.
    An extensive computational exercise to evaluate the
numerical diffusion properties of the four models being
evaluated in the GRI project has been reported by Havens and
Schreurs (49).  Based on the results, it was determined that
satisfactory control of numerical diffusion errors in LNG vapor
dispersion predictions was not practical with the SIGMET-N
model,  and it has not been evaluated further.  The numerical
diffusion properties of the ZEPHYR model and the development
efforts which would be required to adapt (and verify) the
Langrangian particle treatment of turbulent gas mixing for the
                               C-19

-------
typical heavy gas dispersion scenario also led to its
elimination from the evaluation project.   The evaluation of
MARIAH-II and FEM3 has continued,  with emphasis on testing by
comparison with wind tunnel test data which reflect strong
density stratification effects and for which the turbulent
mixing can be quantified.  Particular use has been made of two
wind tunnel heavy gas release data sets:
    • two-dimensional, continuous, heavy gas releases in a
        wind tunnel
    • three-dimensional, continuous, heavy gas releases in a
        wind tunnel.  In the experiments selected for analysis,
        the dispersion appears to have been confined to a zone
        adjacent to the wind tunnel floor where turbulence was
        damped and the mixing was essentially molecular (50,51).
    McQuaid (20) reported measurements of the vertical
dispersion of carbon dioxide (p/pa - 1.52) introduced as a
floor level, near-line source in a fully developed wind tunnel
flow of width 0.3 m, height 0.9 m, and working length 5.5 m.
The mean velocity, ua, in the tunnel (=» 0.8 u^^) ranged from
0.82 to 3.52 m/s.  The friction velocity (in the absence of any
dense gas) was reported as u* — 0.0412 ua.  The Reynolds number
(based on the mean velocity and the tunnel hydraulic diameter)
ranged correspondingly from 2.5 x 10^ to 1.1 x 10 .  Three of
McQuaid's experiments have been selected for simulation.  The
conditions for the three experiments, which were chosen to
represent near-passive dispersion to stably stratified flow
dispersion, are given in Table II.
                                C-20

-------
    TABLE II.  Summary of Experimental Conditions--McQuaid (20)

Average velocity over
channel (m/s)
Friction velocity (m/s)
Source strength (kg/m s)
Source exit (vertical)
velocity (m/s)
Richardson number1 range
Gas density (kg/m3)
Source dimensions width,
m - length, m
Roughness length (m)
(calculated from
velocity profile)

1

3.6
0.144
0.0113

0.124
0.5-0.9
1.83

0.3-0.05


1 x lO'5
Experiment
2

1.83
0.073
0.0142

0.155
4.0-8.0
1.83

0.3-0.05


1 x 10'5

3

1.28
0.05
0.0226

0.248
28.0-44.0
1.83

0.3-0.05


1 x 10'5
          .  H  --
   L  -  v  ££  eff
   *    &  p     2
          a  u,
               *
    McQuaid's experiments have been used to evaluate numerical
diffusion in the MARIAH-II and FEM3 models in an indirect way.
McQuaid has shown that the vertical gas concentration
distributions in the three experiments chosen for evaluation
are well represented as Gaussian.  The measured gas
concentration distribution has been used to estimate the
vertical air entrainment velocity and a diffusivity profile
consistent with the measurements specified.  This approximate
knowledge of the "correct" turbulent eddy diffusivity for these
experiments allows comparison calculations with either zero
(MARIAH-II) or very small (FEM3) diffusivities to determine the
relative importance of numerical vs. "physical" diffusion.
These computations have also been reported (49).   For the three
experiments chosen for analysis, which represent near-passive
to strongly stratified flow dispersion, the slope of the
                               C-21

-------
maximum (ground level) concentration vs. distance which is
attributable to numerical diffusion is satisfactorily small
compared to the experimentally observed concentration decay.
    Kothari and Meroney (KM) (50) and Meroney and Neff (KN)
(51) reported measurements of the lateral gravity spread and
dispersion of isothermal heavy gases released continuously from
floor level area sources in an unobstructed wind tunnel flow.
Two experiments have been carefully studied.  Table III lists
the experimental conditions of KM 75 and MN 42.  It appears

that in these two experiments the heavy gas dispersion
processes were confined to the zone adjacent to the wind tunnel
floor where mixing was molecular.  Accordingly, both
experiments were simulated using the molecular diffusivities

for the gases involved (C02 and Argon, respectively) for the
diffusion calculation.

    Figure 7 compares the measured and predicted ground level
centerline gas concentration and the ground level horizontal
concentration distributions at four downwind distances for KM
75 using MARIAH-II and for MN 42 using MARIAH-II and FEM3.
Since these three-dimensional experiments involve gas flows
    TABLE III.  Summary of Experimental Conditions--
                Kothari-Meroney (50) and Meroney-Neff (51)
                                 KM 75
                        KM 42
Tunnel velocity (m/s)
Friction velocity (m/s)
Source strength (kg/s)
Source exit velocity (m/s)
Gas density (kg/m3)
Source diameter (m)
Roughness length (m)
Gas molecular diffusivity
  (m2/s)
0.24 @ 0.4 ra
    0.017
 2.68 x 10
 2.11 x 10
-4
-3
  .82 (C02)
     0.3
   0.00015
 1.5 x 10
         -5
0.20 @ 2.1 cm
    0.015
 2.77 x 10'4
  9.6 x ID'3
1.63 (Argon)
     0.15
    0.0001
           1.5 x 10
                   -5
                               C-22

-------
         Kothari-Meroney  75
       Diffusivity Sensitivity
                                          Meroney-Keff 42
   20-
I   0
                       Predicted
                        MARIAH-H
                           data
         3v
          .5    1.0   1.5   2.0   2.5 a
                                   S
                                                  0  Exp.      ,
                                                     Pre. MARIA?-II
                                         1.0    2.0
                   3.0    4.0
                                                        x=.3 m
20


101


 0
  \
  0
                        x=.8 a
                    x-1.2 m
                                     20
                                   .3 10!
                                   g  °
                                   0)
                                   o
                                     io)
                                                         X-.6 m
                 x-1.2 m

 FEM3 same as MARIAH-II
    20
    10
                        x-2 n
            .2    .4    .6     .8
         Crosswind Distance, o
                                     20i
                                 10
                 x=2.4 m
FEM3 not  at steady state
here, but appears same as
MARIAH-II
°ooonnnn 0-0-0-^^-0—
                                  0      .2     .4    .6     .8
                                   Distance from Centerline,  m
    FIGURE 7.   Comparison of  measured and  predicted gas
                 concentrations for Kothari-Meroney  75 and
                 Meroney-Neff 42.
                                    C-23

-------
 with a strong lateral gravity- driven component,  and since the
 molecular diffusion coefficient is well known,  the agreement of
 these simulations with the experimental data provides important
 assurance of the general applicability of the model equations
 used in MARIAH-II and FEM3 as well as the adequacy of the
 numerical methods used.
     Turbulent Mixing (Closured Models.  The four  models being
 evaluated in the GRI project assume the gradient transfer
 hypothesis with ad hoc specification of eddy diffusivities.
 The primary specification is of eddy diffusivity for transfer
 in the vertical direction;  horizontal diffusivities are
 obtained from prescribed ratios with the vertical diffusivity.
     SIGMET-N, ZEPHYR,  and MARIAH-II use the diffusivity
 specification method incorporated in the original SIGMET  model
 (52).   The vertical diffusivity is specified as  a function of
 height,  "local stability" (based on temperature  or density
 gradient), and local velocity as shown in Figure 8.   The
 selection of a "local" Pasquill stability category is based  on
 the correlation of stability with vertical temperature gradient
 specified in the USAEC Safety Guide 23  (53).  The Pasquill
 stability categories can be correlated with a bulk Richardson
 number for the atmospheric  flow (54),  and the vertical
 diffusivity coefficient  can be expressed in terms of a vertical
 entrainment velocity,  v' , as
          k u*
Figure  9  shows representative values of nondimensionalized
vertical  entrainment velocity with these models, based on
Equation  (7), for bulk Richardson numbers greater than 0.1.
Figure  9  also shows nondimensionalized vertical entrainment
velocity  as a function of bulk Richardson number (data points)
obtained  from analysis of the laboratory stratified fluid
mixing experiments of McQuaid (20); Kantha, Phillips, and Azad
(21); and Lofquist (22).  The solid line curve plotted on
                               C-24

-------
                 10
                x.l
                 .01
                    1          10        100

                            Height, m


FIGURE 8.  Vertical diffusivity specification in MARIAH-II,

            SIGMET-N, and ZEPHYR.
      0.5
     10
       -1
                       ... . McQuaid (1976)
            SIGMET Turbulence
               Closure
                                     KPA (1977)
                                                 B,y».01-10a
                                              Ravens and
                                              Spicer
                                              DEGADIS (1985)
FIGURE  9.   Vertical entrainment as a  function of bulk
            Richardson number.
                             C-25

-------
Figure 9 is a curve fit (of the laboratory experimental data)
which has been incorporated in the DEGADIS model.   The DEGADIS
correlation of nondimensional entrainment velocity with bulk
Richardson number is also similar to that used in FEM3.
    The vertical diffusivity calculated in a simulation of LNG
vapor dispersion with the MARIAH-II, SIGMET-N, and ZEPHYR
models will have a lower limit corresponding to "G" stability
when the bulk Richardson number is greater than about 0.1, and
there is no provision for further damping of turbulence which
would be indicated for the stronger temperature (and density)
gradients expected in an LNG cloud.  Bulk Richardson numbers,
based on release conditions, can also be estimated to
characterize laboratory and field experiments.  Figure 9
indicates the Richardson numbers estimated for three wind
tunnel heavy gas releases reported by McQuaid (20), the Burro 9
field experiment, and continuous LNG releases of 160, 980, and
5250 kg/s  (representing maximum source rates from instantaneous
LNG spills onto water of 10, 100, and 1000 m3 respectively) in
a  5 m/s wind.  There is a clear indication that the turbulent
mixing model used in MARIAH-II, SIGMET-N, and ZEPHYR will tend
to give too low vertical diffusivities for Richardson numbers
between about 0.1 and about 20, and too high vertical
diffusivities for Richardson numbers greater  than about 20.   It
is also significant that the turbulence model may fortuitously
represent  vertical mixing reasonably well for Rig around  30,
which  is probably appropriate  for  characterizing the Burro  9
experiment.  Based on these observations, the turbulence
closure model used in MARIAH-II,  SIGMET-N, and  ZEPHYR  is  not
expected  to  scale adequately for  LNG releases which have  widely
different  ratios of vertical shear (high  friction velocity) and
density stratification.  It follows that vertical  turbulent
mixing will be underestimated  for releases with low bulk
Richardson numbers and  overestimated for  releases with large
bulk Richardson numbers.  The  former cases  include many small-
scale wind tunnel experiments,  and the  latter include  large LNG
                               C-26

-------
releases in low-wind conditions, such as Burro 8.  The
underestimation of vertical diffusivities for the small-scale
McQuaid 2-D wind tunnel heavy gas experiments has been
demonstrated by Havens and Schreurs (46).
    A simplified second-order turbulence model has been
developed by Freeman (47) for simulation of LNG vapor cloud
dispersion.  The principal simplifying assumptions invoked in
the model are:
    • The theory is local: all turbulent quantities are assumed
        point functions depending only on the state of the mean
        field.
    • The mean field employed in the turbulence terms is
        simplified to retain the dominant terms describing a
        stratified boundary layer; turbulent kinetic energy is
        dependent on the magnitude of the shear of the flow
        field

      S - [(3u/5y)2 + (aw/ay)2]1/2
      and the gradient Richardson number
    MARIAH-II with the turbulence closure method developed by
Freeman and modified to provide for turbulence kinetic energy
diffusion has been used to simulate HcQuaid's experiments.
Figure 10 shows the experimental vs. MARIAH-II-predicted ground
level gas concentration decay with distance.  The agreement is
now reasonably good for all three experiments.
    MARIAH-II, with Freeman's turbulence closure model, has
also been used to simulate Havens' and Spicer's laboratory
instantaneous releases of heavy gas.  Figure 11 shows a MARIAH-
II simulation of a 55 -liter Freon release with an initial
height- to -diameter ratio of 1.0.  The cloud profile (half of
the radially symmetric section of the cloud is shown) , defined
by the 1% volume fraction, is shown at zero time and at 0.5, 1,
2, 3, 4, 5, and 6 seconds after release.  Figure 12 shows the
                               C-27

-------
             100
g
          2
          O
          B
             10
           McQuaid,
           -   measured
                                     No. 3
                            lo. 2
                                     No. 1
                     -MARIAH-II, prediction
               0.1           1           10
                       Distance, m

FIGURE  10.   McQuaid 2-D wind tunnel experiments—MARIAH-II
             predictions with improved air entrainment
             model.
  0
 40

  0

 40

  0

 40

.  0
, 40

  0

 40

  0
 40

  0

 40

  0
                1
       t « 0
                     t • 0.5 8
                        t " 1 8
                              t " 2 8
                                   t - 3 8
                                       t " 4 8
                                 t - 5 8
                                   t - 6 8
                40  BO  120 160  200 240 280 320
                    Radial Distance, cm
FIGURE 11.   Laboratory, calm-air,  55-liter Freon release,
             (H/D)i = 1.0—MARIAH-II prediction of cloud
             boundary vs. time.
                          C-28

-------
20


10
    r-l -rl
     «
     B!
              	 Measured
               '  *MARIAH-II
                           10
                                     100   200
FIGURE  12.   Laboratory, calm-air, 55-liter Freon release,
             (H/D).  -  1.0—MARIAH-II  prediction of cloud
             radial  extent.
          100.
           10
                  0
-------
predicted vs. measured radial cloud extent for the same  55-
liter Freon release.  Figure 13 shows predicted and measured
maximum values of the cloud gas concentration and the cloud
spatial average gas concentration as a function of
dimensionless time for the 55-liter Freon release.  The
predicted cloud average gas concentrations were obtained by
spatial averaging the MARIAH-II prediction using cloud boundary
concentration limits of 1% and 2.5%.  The DEGADIS model  cloud
average concentration prediction is shown for comparison (31).
    MARIAH-II and FEM3 are now being tested by comparison with
selected data from the Burro/Coyote, Maplin Sands, and Thorney
Island experiments.

                            CLOSURE

    Similarity models are available for heavy gas releases at
ground level on uniform terrain with unobstructed atmospheric
flow.  Some models have been verified for the prediction of
dispersion to flammable concentration levels  (order 1%)  and  are
sufficiently accurate for most risk assessment and emergency
response  applications.
    Most  field heavy gas dispersion tests have provided data on
cloud dispersion  to  fuel flammable-limit concentration levels;
such  data are not sufficient  to verify models  for application
to toxic  gas dispersion to  parts per million concentration
levels.   It  is expected that  models which have been verified
for flammable fuel dispersion prediction  and also incorporate
an appropriate description  of the  passive dispersion regime
should be useful  for prediction of dispersion of dense  gases to
toxicity  levels.
    It appears likely  that  3-D mathematical models  can  be
verified  for flammable gas  dispersion prediction, but current
work  is biased strongly toward evaluation using data without
terrain or wake turbulence  effects.  Since  the primary
                              C-30

-------
motivation for using these models is to describe such complex
effects, research should be directed to their validation for
such use.
    3-D mathematical models are not likely to prove useful for
predicting heavy gas dispersion to very low concentration
levels because of difficulties in managing numerical diffusion
in calculations with several orders of magnitude change in
concentration.
    Application of dense gas dispersion models to toxic gas
dispersion is expected to require additional research to
provide for description of jetting releases, aerosol formation
and evaporation, chemical reaction, and deposition.
    The participants in the recent joint EPA-DOE workshop
agreed that source specification in an accident may be a
greater concern than the accuracy of a dispersion model.  A
technology transfer problem was identified; many of the
improvements in dispersion models have not been transferred to
the potential user.  There was a consensus that atmospheric
dispersion predictions should be done to plan emergency
response whenever possible.  No selection or recommendation of
models was made.
                        LIST OF SYMBOLS

Cg       constant in density intrusion (spreading) relation
C-^       constant
                                     f\
g        acceleration of gravity  (m/s'')
H        height or depth of density intrusion or cloud (m)
H£pp     effective cloud depth  (m)
K        vertical turbulent diffusivity  (m2/s)
k        von Karman's constant, 0.35
                             C-31

-------
Ri*      Richardson number associated with density differences

T        characteristic time scale, V1/6/7g(P - /»a>//>a

tf       characteristics time of cloud formation (s)

tt       cloud travel time to the maximum distance exposed to
           a given concentration (s)

u        velocity along x-direction, (m/s)

ua       ambient average velocity,  (m/s)

ue       horizontal or frontal entrainment velocity, (m/s)

uj       cloud front velocity, (m/s)

Ujj       wind velocity, along x-direction, (m/s)

umax     maximum velocity

u*       friction velocity,  (m/s)

V        cloud volume, (nr)

ve       vertical entrainment velocity associated with HT,
           (m/s)

w        velocity along z-direction, (m/s)

x,y,z    Cartesian coordinates, (m)

a        constant in power law wind profile

A        Monin-Obukhov length (m)

P        density (kg/m3)

fla       ambient density (kg/m3)

        logarithmic velocity profile correction function
                              C-32

-------
                          REFERENCES
 1.  General Accounting Office (GAO),  "Liquefied Energy Gases
     Safety," EMD-78-29, July 31, 1978.
 2.  Havens, J. A., "Predictability of LNG Vapor Dispersion
     from Catastrophic Spills onto Water: An Assessment,"
     Department of Transportation--Coast Guard Report CG-M-09-
     77, April, 1977.
 3.  van Ulden, A. P., "On the Spreading of a Heavy Gas
     Released Near the Ground," 1st International Loss
     Symposium, The Hague, Netherlands, 1974.
 4.  Havens, J. A., "A Review of Mathematical Models for
     Prediction of Heavy Gas Atmospheric Dispersion," Institute
     of Chemical Engineers Symposium Series No. 71, 1982.
 5.  Havens, J. A., "Evaluation of 3-D Hydrodynamic Computer
     Models for Prediction of LNG Vapor Dispersion in the
     Atmosphere," Gas Research Institute Contract No. 5083-252-
     0788 with the University of Arkansas, March, 1983 -
     September, 1986.
 6.  Wheatley, C. J. and D. M. Webber, "Aspects of the
     Dispersion of Denser-than-air Vapors Relevant to Gas Cloud
     Explosions,"  Safety and Reliability Directorate, United
     Kingdom Atomic Energy Authority,  Warrington, England,
     1984.
 7.  Koopman, R. P. et al., "Burro Series Data Reports,
     LLNL/NWC 1980 LNG Spill Tests," Lawrence Livermore
     National Laboratories Report UCID-19075, December, 1982.
 8.  Colenbrander, G. W., A. E. Evans, andJ. S. Puttock,
     "Spill Tests of LNG and Refrigerated Liquid Propane on the
     Sea, Maplin Sands, 1980: Dispersion Data Digests," Shell
     Thornton Research Center, May, 1984 (confidential).
 9.  HSE--British Health and Safety Executive, Research and
     Laboratory Services Division, Red Hill, Sheffield, UK--
     Heavy Gas Dispersion Trials, Thorney Island 1982-83, Data
     Digests.
10.  McRae, T. G., R. T. Cederwell, H. C. Goldwire, Jr., D. L.
     Hippie, G. W. Johnson, R. P. Koopman, J. W. McClure, and
     L. K. Morris, "Eagle Series Data Report: 1983 Nitrogen
     Tetroxide Spills," Lawrence Livermore National
     Laboratories Report UCID-20063, June, 1984.
11.  Hall, D. J., "Experiments on a Model of an Escape of Heavy
     Gas," Warren Springs Laboratory,  UK, Reports LR 217 (AP),
     1976, and LR 312 (AP), 1979.
12.  Hall, D. J. et al., "A Wind Tunnel Model of the Porton
     Dense Gas Spill Field Trials," LR 394 (AP), Warren Spring
     Laboratory, Department of Industry, Stevenage, UK, 1982.
                             C-33

-------
13.  Havens, J. A. and T. 0.  Spicer,  "Gravity Spreading and Air
     Entrainment by Heavy Gases Instantaneously Released in a
     Calm Atmosphere," Proceedings I.U.T.A.M. Symposium on
     Atmospheric Dispersion of Heavy  Gases and Small Particles,
     Delft University of Technology,  The Netherlands,  August
     29-September 2, 1983.
14.  Picknett, R. F.,  "Dispersion of  Dense Gas Puffs Released
     in the Atmosphere at Ground Level," Atmospheric
     Environment r 15.  1981.
15.  Spicer, T. 0. and J. A.  Havens,  "Modeling the Phase I
     Thorney Island Experiments," Symposium on the Thorney
     Island Heavy Gas Trials, sponsored by the British Health
     and Safety Executive, Sheffield,  UK,  April,  1984.
16.  van Ulden, A. P., "The Unsteady  Gravity Spread of a Dense
     Cloud in a Calm Environment," 10th International Technical
     Meeting on Air Pollution Modeling and its Applications,"
     NATO-CCMS, Rome,  Italy,  October,  1979.
17.  van Ulden, A. P., "A New Bulk Model for Dense Gas
     Dispersion: Two-Dimensional Spread in Still Air,"
     I.U.T.A.M. Symposium on Atmospheric Dispersion of Heavy
     Gases and Small Particles, Delft University of Technology,
     The Netherlands,  August 29-September 2, 1983.
18.  Meroney, R. N. and A. Lohmeyer,  "Gravity Spreading and
     Dispersion of Dense Gas Clouds Released Suddenly into a
     Turbulent Boundary Layer," Draft Report CEF82-8RNM-AL-7 to
     Gas Research Institute,  Chicago,  Illinois, August, 1982.
19.  Havens, J. A. and T. 0.  Spicer,  "Development of an
     Atmospheric Dispersion Model for Heavier-than-Air Gas
     Mixtures," Volume II, "Laboratory Calm Air Heavy Gas
     Dispersion Experiments," Report  No. CG-D-23-85, U.S. Coast
     Guard, Final Report, May, 1985.
20.  McQuaid, James, "Some Experiments on the Structure of
     Stably Stratified Shear Flows,"  Technical Paper P21,
     Safety in Mines Research Establishment, Sheffield, UK,
     1976.
21.  Kantha, L. H., 0. M. Phillips, and R. S. Azad, "On
     Turbulent Entrainment at a Stable Density Interface,"
     Journal of Fluid Mechanics. 7_9_,  1977. pp. 753-768.
22.  Lofquist, Karl, "Flow and Stress Near an Interface Between
     Stratified Liquids, Phvsics of Fluids. 3, No. 2, March-
     April, 1960.
23.  Ellison, T. and J. S. Turner, "Turbulent Entrainment in
     Stratified Flows," Journal of Fluid Mechanics. 6, 1959.
24.  Deardorff, J. W., 1983,  "A Multi-Limit Mixed-Layer
     Entrainment Formulation," Journal of Physical
     Oceanography. 13. pp. 988-1002.
25.  Kranenberg, C., "Wind-Induced Entrainment in a Stably
     Stratified Fluid," Journal of Fluid Mechanics. 145. 1984,
     253-273.
                             C-34

-------
26.  Stretch,  D.  D.,  R.  E. Britter, and J. C.  R. Hunt,  "The
     Dispersion of Slightly Dense Contaminants," I.U.T.A.M.
     Symposium on Atmospheric Dispersion of Heavy Gases and
     Small Particles, Delft, University of Technology,  The
     Netherlands, September, 1983.
27.  Britter,  R.  E. ,  "A Review of Mixing Experiments Relevant -
     to Dense Gas Dispersion," The Institute of Mathematics and
     Its Applications Conference on Stably Stratified Flow and
     Dense Gas Dispersion, Chester, England, April,  1986.
28.  Monin, A. S., "Smoke Propagation in the Surface Layer of
     the Atmosphere," in Atmospheric Diffusion and Air
     Pollution, ed.,  F.  N. Frenkiel, Academic Press, 1959.
29.  Batchelor, G. K.,  An Introduction to Fluid Mechanics.
     Cambridge University Press, Cambridge, UK, 1967.
30.  Businger, J. A., J. C. Wyngaard, Y. Izumi, and E.  F.
     Bradley,  Flux-Profile Relationships in the Atmospheric
     Surface Layer,"  Journal of the Atmospheric Sciences.  28.
     March, 1971.
31.  Havens, J. A. and T. 0. Spicer, "Development of an
     Atmospheric Dispersion Model for Heavier-than-Air Gas
     Mixtures," Volume I, Report No. CG-D-23-85, U.S. Coast
     Guard, Final Report, May, 1985.
32.  Colenbrander, G. W., "A Mathematical Model for the
     Transient Behavior of Dense Vapor Clouds," 3rd
     International Symposium on Loss Prevention and Safety
     Promotion in the Process Industries, Basel, Switzerland,
     1980.
33.  Colenbrander, G. W. and J. S. Puttock, "Dense Gas
     Dispersion Behavior: Experimental Observations and Model
     Developments," International Symposium on Loss Prevention
     and Safety Promotion in the Process Industries, Harrogate,
     England,  September, 1983.
34.  Pasquill, F., Atmospheric Diffusion. 3rd ed., Halstead
     Press, New York, 1983.
35.  Zeman, 0. and H. Tennekes, "Parameterization of the
     Turbulent Energy Budget at the Top of the Daytime
     Atmospheric Boundary Layer," Journal of the Atmospheric
     Sciences. January,   1977.
36.  Havens, J. A., "Evaluation of 3-D Hydrodynamic Computer
     Models for Prediction of LNG Vapor Dispersion in the
     Atmosphere," GRI Contract No. 5083-252-0788, Annual Report
     (March 1984-February 1985), University of Arkansas.
37.  Su, F. Y. and P. C. Patniak, "SIGMET-N: Near-Field
     Solution of Contaminant Dispersion Phenomena; Model
     Description and User's Guide," Science Applications, Inc.,
     July, 1981  (proprietary).
38.  Hertel, J. and L.  Teuscher, "Advances in Heavier-than-Air
     Vapor Cloud Dispersion Modeling," Proceedings of the AGA
     Transmission Conference, Chicago, Illinois, May, 1982.
                             C-35

-------
39.   Taft,  J.  R.,  M. S. Ryne,  and D.  A.  Weston,  "MARIAH:  A
     Dispersion Model for Evaluating  Realistic Heavy Gas  Spill
     Scenarios," Proceedings of the AGA Transmission
     Conference,  Seattle, Washington, May,  1983.
40.   Chan,  S.  T.,  "FEM3--A Finite Element Model  for the
     Simulation of Heavy Gas Dispersion and Incompressible
     Flow--User's Manual," Lawrence Livermore Laboratory, UCRL-
     53397, February 1983.
41.   Schreurs, P.  J., "Mathematical Modeling of  the Dispersion
     of Accidental Releases of Heavy  Gases  at Ground Level in
     an Industrial Environment," Catholic University,  Leuven,
     Belgium,  1983.
42.   Riou,  Y., "The Use of a Three-Dimensional Model in
     Simulating Thorney Island Field  Trials," The Institute of
     Mathematics and Its Applications Conference on Stably
     Stratified Flow and Dense Gas Dispersion, April,  1986,
     Chester College, Chester, England.
43.   Betts, P. L.  et al., "Finite Element Calculations of
     Transient Dense Gas Dispersion," The Institute of
     Mathematics and Its Applications Conference on Stably
     Stratified Flow and Dense Gas Dispersion, April,  1986,
     Chester College, Chester, England.
44.   Deaves, D. M., "Application of Advanced Turbulence Models
     in Determining the Structure and Dispersion of Heavy Gas
     Clouds,"  I.U.T.A.M. Symposium on Atmospheric Dispersion of
     Heavy Gases and Small Particles, Delft University of
     Technology,  The Netherlands, August 29-September 2,  1983.
45.   Roache, P. J., Computational Fluid Dynamics.  Hermosa
     Publishers,  Albuquerque,  NM, 1976.
46.   Havens, J. A. and P. J. Schreurs,  "Evaluation of 3-D
     Hydrodynamnic Computer Models for Prediction of LNG  Vapor
     Dispersion in the Atmosphere," Proceedings,  Eighth
     International Conference on Liquefied  Natural Gas, Los
     Angeles,  California, June, 1986.
47.   Schreurs, J.  A. et al., "A Lagrangian  Particle Model for
     Atmospheric Dispersion of Heavy  Gases," submitted to
     Atmospheric Environment.  Spring, 1986.
48.   Chapman,  M.,  "FRAM--Nonlinear Damping  Algorithms for the
     Continuity Equation," Journal of Computational Physics.
     44, 1981.
49.   Havens, J. A. and P. J. Schreurs,  "Evaluation of 3-D
     Hydrodynamic Computer Models for Prediction of LNG Vapor
     Dispersion in the Atmosphere," Annual  Report, Contract No.
     5083-252-0788, (March 1984-February 1985),  with the
     University of Arkansas.
50.   Kothari,  K.  M. and R. N.  Meroney,  "Accelerated Dilution of
     Liquefied Natural Gas Plumes at  the Source," Final Report
     to Gas Research Institute, Contract No. 5014-352-0203,
     1982.
                             C-36

-------
51.  Meroney, R. M.  and D.  E.  Neff,  "The Behavior of LNG Vapor
     Clouds: Wind-Tunnel Tests on the Modeling of Heavy Plume
     Dispersion," Final Report to Gas Research Institute,
     Contract No. 5014-352-0203, 1982.
52.  Havens, J., "A Description and Assessment of the SIGMET
     LNG Vapor Dispersion Model," U.S. Coast Guard Report No.
     CG-M-3-79  (available through NTIS), 1979.
53.  U.S. Atomic Energy Commission Safety Guide 23, 1972.
54.  Randerson, D., Editor, "Atmospheric Science and Power
     Production," U.S. Department of Energy, Report DOE-TIC-
     27601,  1984.
                              C-37

-------
  APPENDIX D
Workshop Agenda
      D-l

-------
                                       AGENDA
             Determination of Atmospheric Dilution for Emergency Preparedness
                            A Joint EPA-DOE Technical  Workshop
                                   October  15-17,  1986
         TIME
       TOPIC
     SPEAKER
Wednesday. October 15.  1986


SESSION 1

8:00 - 8:10 a. m.



8:10 - 8:40 a. m.

Two Recent Catastrophic Events

8:40 - 9:00 a. m.

9:00 - 9:20 a. m.
Meredith Guest House
Research Triangle Park,  NC
Wei come
Technical Overview
Bhopal

National Research Needs for
Emergency Response in the
Wake of Chernobyl
SESSION  2
Characteristics  of Sources of Releases  of Toxic  Material
Chairman:  Joe Knox,  Lawrence Livermore Laboratory

 9:20  -  9:40  a.  m.
 9:40  -  10:00  a.  m.


 10:00  -  10:20  a.  m.


 10:20  -  10:40  a.  m.
Historical Accounts and
Credible Scenarios
Short Term Toxic Releases from
Industrial Sites

Fate of Toxic Releases in the
Atmosphere - ARAC

Source Strength Modeling
SESSION 3
Reaulatorv and  Operational  Needs for Models
Chairman:  Bob  Hangebrauck,  USEPA
Frank Binkowski, USEPA &
NOAA  and  Harry  Moses,
USDOE

R. A. Cox, Technica
Shankar Rao, NOAA/ATDD

Joe Knox, Lawrence
Livermore Laboratory
Jane Crum, Bob
Hangebrauck, and
Bill Rhodes, USEPA

Robert Rosensteel, USEPA
Marvin Dickerson, LLL
Jerry Schroy, Monsanto
10:40 - 11:00  a. m.
Future Needs for Dispersion
Models in Hazard Evaluation,
                                                                   Jim Makris,  USEPA
                                        D-3

-------
         TIME
       TOPIC
    SPEAKER
SESSION 3 (Cont'dl




11:00 - 11:20 a. m.


11:20 - 11:40 a. ra.



11:40 - 1:00 p. m.
Emergency Preparedness, and
Accident Prevention

Issues in Regulatory
Applications of Models

Community Needs for Hazard
Evaluation Tools
Lunch
Dave Lay!and, USEPA
Fred Millar,
Environmental Policy
Institute
SESSION 4
Dispersion Models
Chairman:  Dick Benjamin, Savannah River Laboratory
1:00 - 1:30 p. m.


1:30 - 1:50 p. m.

1:50 - 2:10 p. m.


2:10 - 2:30 p. m.



2:30 - 2:45 p. m.
Keynote Talk


Dense Gas Dispersion Model

Reactive Dense Gas Models
Jerry Havens, University
of Arkansas

Don Ermak, LLL

Bruce Hicks and Will
Pendergrass, NOAA/ATTD
Dispersion Models for Neutrally   Tom Pierce, USEPA & NOAA
Buoyant and Positively Buoyant
Gases

Break
SESSION 5
Results from Dispersion Experiments
Chairman:  Ray Hosker, NOAA/Atmospheric Turbulence & Diffusion Division
2:45 - 3:05 p. m.


3:05 - 3:25 p. m.


3:25 - 3:45 p. m.
Laboratory Scale Experiments
Experimental Work at the EPA
Fluid Modeling Facility

Large Scale Experiments of
the DOE Liquefied Gaseous
Fuels Program
Bob Meroney, Colorado
State University

Bill Snyder,
USEPA & NOAA

Ron Koopman, LLL
                                           D-4

-------
	TIME

3:45 - 4:05 p. m.



4:05 - 4:42 p. m.
	IQPIC	

 Atmospheric Science and
 Emergency Response at the
 Savannah River Plant

 Emergency Preparedness and
 Response In the U. S. Air
 Force
    SPEAKER	

Alan Weber, Savannah
River Laboratory
Larry Key, USAF
Thursday.  October 16.  1986

SESSION  6
Panel  Discussions

8:00  - 11:30 a.  m.
(Break 9:30 - 9:45 a.  m.)
 1:00  -  3:30 p. m.
 3:30 - 3:45 p. m.

 3:45 - 5:00 p. m.
 Summary of Current Status
 of Source Strength and
 Dispersion Models, Their
 Strengths and Weaknesses,
 and Recommendations for
 Improvement

 Recommendations for Selection
 from Among Current Models for
 Immediate Use in Hazard Identi-
 fication and Evaluation, Prepara-
 tion of Emergency Preparedness
 Scenarios, and in Response to
 Emergencies

 Break

 The Role of the Meteorologist
 in Hazard Evaluation, Emergency
 Response
Panel 1
Chairman:  Frank
Schiermeier
Reporter:  Steve Perry
Panel 2
Chairman:  Ray Hosker
Reporter:  Tom Pierce
Chairman:
Reporter:
Lew Nagler
Jack Shannon
 Friday. October 17. 1986

 8:00 - 11:30 a. m.
 (Break 9:30 - 9:45 a. m.)
  Presentation of Results of Panel
  Discussions
                                         D-5

-------
      APPENDIX E
Workshop Participants
         E-l

-------
          DETERMINATION OF ATMOSPHERIC DILUTION FOR EMERGENCY  PREPAREDNESS
                         A JOINT EPA-DOE TECHNICAL  WORKSHOP
                               OCTOBER  15 - 17, 1986
Allen, Dr. Harry L.
USEPA-ERT
Edison, NJ  08837
(201) 321-6747

Bach, Dr. Walter D., Jr.
Meteorologist
Army Research Office
P. 0. Box 12211
Research Triangle  Park, NC   27709
(919) 549-0641
FTS 629-3890

Barr, Dr. Sumner
Atmos, Sci. G-8, Mail  Stop  D 466
Los Alamos National Lab
P. 0. Box 1663
Los Alamos, NM  87545
(505) 667-2636
FTS 843-2636

Benjamin, Dr. Richard  W.
Room 1015A, Bldg.  773A
E. I. Du Pont
Savannah River  Laboratory
Aiken, SC  29808
(803) 725-3325

Binkowski, Dr.  Francis S.
USEPA (MD-80)
Research Triangle  Park, NC   27711
(919) 541-2460

Blaunstein, Dr. Robert P.
EPG, RCD, EH-33, USDOE
Washington, DC  20545
(301) 353-5849

Bresnick, Dr. Gerald  I.
Amoco Corp., MC 4903
200 E. Randolph Drive
Chicago, IL  60601
(312) 856-7198
Briggs, Dr. Gary A.
USEPA (MD-80)
Research Triangle Park, NC
(919) 541-2606
27711
Catalano, Joseph A.
Aerocomp, Inc., Technical Director
3303 Harbor Blvd.
Cost Mesa, CA  92626
(714) 957^-6596

Catalano, Jude J.
Connecticut Dept. of Environmental
   Protection
165 Capitol Avenue, Room 146
Hartford, CT  06106
(203) 566-2690

Chaplin, Anton
Manager, Environmental Meteorology
Unocal Corporation
1201 W. 5th Street
Los Angeles, CA  90017
(213) 977-5202

Compton, Harry
USEPA-ERT
Edison, NJ  08837
(201) 321-6751

Cox, Dr. R. A.
Technica, Ltd.
Lynton House
7/12 Tavistock Square
London WC1H9LT
England
011-44-1-388-2684

Crum, Jane
USEPA, MD-62B
Research Triangle Park, NC  27711
(919) 541-1528
                                   E-3

-------
Cupitt, Dr. Larry
USEPA (MD-84)
Research Triangle Park, NC  27711
(919) 541-2878

Cushmac, Dr. George E.
RSPA, USDOT (DHM-21)
Washington, DC  20590
(202) 366-4493

De Wolf, Glenn B.
Radian Corp.
P. 0. Box 9948
Austin, TX  78766
(512) 454-4797

Dicker-son, Dr. Marvin H.
Associate Division Leader
Atmospheric & Geophysical Science
P. 0. Box 808
Livermore, CA  94550
(415) 422-1806
FTS 532-1806

Doelp, Louis C.
Air Products and Chemicals, Inc.
P. 0. Box 538
Allentown, PA  18105
(215) 481-6136

Eheman, Christie
Health Physicist
1600 Clifton Rd., NE, CMB-28N
Atlanta, GA  30333
(404) 452-4161

Elderkin, Dr. Charles
Atmospheric Science Dept., PNL
P. 0. Box 999
Richland, WA  99352
(509) 376-8639

Eltgroth, Mark
Radian Corp.
P. 0. Box 9948
Austin, TX  78766

Englemann, Dr. Rudolph J.
NOAA
11701 Karen Drive
Potomac, MD  20854
(301) 443-8721
Ermak, Dr. Donald L.
Modeling Group Leader
Liquefied Gaseous Fuels Program
Lawrence Livermore Laboratory
P. 0. Box 808, L-451
Livermore, CA  94550
(415) 423-0146
FTS 469-0146
Fairobent, James
Office of Environmental
Department of Energy
Washington, DC  20585
(202) 252-4852
Guidance, EH-23
Foster, Jerry R.
Union Carbide Corporation
P. 0. Box 8361
South Charleston, WV  15303
(304) 747-3615

Gait, Jerry A.
Hazardous Materials Response
  Branch-NOAA
7600 Sand Point Way
Seattle, WA  98115
(206) 526-6317
FTS 392-6317

Gelinas, Gary
Safer Emergency Systems, Inc.
5700 Corsa Avenue
Westlake Village, CA  91362
(818) 707-2777

Hall, Loren
USEPA, Office of Toxic
  Substances (TS-798)
Washington, DC  20460
(202) 382-3931

Hangebrauck, Robert
USEPA (MD-62B)
Research Triangle Park, NC  27711
(919) 541-4134

Havens, Dr. Jerry A.
Chemical Engineering Dept.
University of Arkansas
Fayettville, AR  72701
(501) 575-4951
                                   E-4

-------
Hicks, Bruce
ATDD, P. 0. Box 2456
Oak Ridge, TN  37831
(615) 576-1232

Hodgin, C. Reed
Rockwell International
Rocky Flats Plant
P. 0. Box 464
Golden, CO  80402-0464
(303) 966-7084

Hosker, R. P., Jr.
NOAA-Atmospheric Turbulence  &
  Diffusion Division
2456 S. Illinois Avenue
Oak Ridge, TN  37831
(615) 576-1248
FTS 626-1248

Jann, Paul R.
E. I. Du Pont
Engineering Dept.  (L-13W40)
Wilmington, DE   19898
(302) 366-3219

Jensen, Andreas  V.
Motech  Services  Co.
220 South River, Suite A
Eaton Rapids, MI   48827
(517) 663-7502

Jersey, Gilbert
Mobil Research & Development Corp.
Billingsport Road
Paulsboro, NJ  08066
(609) 423-1040

Keith,  William
USEPA  (RD-680)
Washington, DC   20460
(202) 382-5716

Key, Capt. Lawrence E.
Research Meteorologist
Headquarters, Air  Force  Engineering
  & Services Center (HQ  FECS/RDV)
Tyndall AFB, FL  32403
(904) 283-4234
Knox, Dr. Joseph B.
Atmospheric & Geophysical Science
Lawrence Livermore Laboratory
University of California
P. 0. Box 808
Livermore, CA  94550
(415) 422-1818

Koopman, Dr. Ronald P.
Lawrence Livermore Laboratory
P. 0. Box 808, L-450
Livermore, CA  94550
(415) 422-7381

Koretzky, Herman
IBM Corporation, D/728, 4A-07
44 S. Broadway
White Plains, NY  10601
(914) 686-1822

Kornasiewicz, Dr. Robert
Office of Nuclear Regulatory Research
Mail Stop NL - 007
U. S. Nuclear Regulatory Commission
Washington, DC  20555

Lake, Robin
Standard Oil Research & Development
4440 Warrensville Center Road
Cleveland, OH  44128
(216) 581-5976

Lantzy, Dr. Ronald
Rohm & Haas Company
P. 0. Box 584
Bristol, PA  19007
(215) 785-7456

Layland, David E.
Monitoring   &   Data  Analysis  Division
USEPA (MD-14)
Research Triangle Park, NC  27711
(919) 541-5690

Lee, Dr. Gene K.
Corporate Engineering Department
Air Products & Chemical
Box 538
Allentown,  PA  18105
(215) 481-6424
                                   E-5

-------
 Livingstone, Dr. Daniel
 Dept.  of Zoology
 Duke University
 Durham,  NC  27706
 (919)  684-3264

 Machta,  Dr.  Lester
 NOAA,  ARL, 8060 13th St.
 Silver Spring, MD  20910
 (301)  427-7684

 Makris,  James L.
 Director of Preparedness
 Office of Solid Waste &
   Emergency Response
 USEPA  (WH-548A)
 Washington,  DC  20460
 (202)  475-8600

 Matthiessen,  Craig R.
 USEPA  (TS-779)
 401  M  Street,  SW
 Washington,  DC  20460
 (202)  382-3694

 Mayer, Alice R.
 Chemical  Manufacturers Assoc.
 2501 M Street, NW
 Washington,  DC  20037
 (202)  887-1176

 McNaughton,  D. J.
 TRC  Environmental  Consultants
 800  Connecticut Blvd.
 East Hartford,'CT  06108
 (203)  299-8631

 Meier, Gerald  E.
 Manager,  Environmental  Services
 Bureau of Explosives
 Association  of American Railroads
 50 F Street, NW
 Washington,  DC  20001
 (202) 639-2136

Meroney,  Dr. Robert
 Fluid Mechanics  &
  Wind Engineering Program
Engineering Research Center
Colorado State University
Ft. Collins, CO  80523
(303) 491-8574
 Millar,  Dr.  Fred
 Environmental  Policy Institute
 218 D Street,  SE
 Washington,  DC  20003
 (202) 544-2600

 Moser, Dr. James H.
 Environmental  Engineering  Dept.
 Shell  Development Company
 Westhollow Research  Center
 P.  0.  Box 1380
 Houston, TX  77001
 (713)  493-7941

 Moses, Dr. Harry
 PTRD,  ER-74, GTN, USDOE
 Washington,  DC   20545
 (301)  353-5572

 Nagler, Lewis
 Air  Programs Branch, EPA
 AP&TMD, 345  Courtland Street, NE
 Atlanta, GA  30365
 (404)  347-4253

 Papal ski, Ray
 New  Jersey Dept.  of  Environmental
  Protection
 401  East State Street
 P. 0.  Box CN027
 Trenton, NJ  08625
 (609)  633-1142

 Parnarouski, Dr.  Michael
 U. S. Coast Guard (G-MTH-1)
 2100 2nd St., SW
 Washington, DC   20593
 (202) 267-1577

 Patrinos, Dr. A. A.  N.
 Department of Energy, ER-74
 Washington, DC   20545
 (301) 353-3764

 Pearson,  Johnnie  L., Chief
Model Application Section
USEPA (MD-14)
Research Triangle Park, NC  27211
 (919) 541-5690
                                   E-6

-------
Pendergrass, William R.
ATDD, P. 0. Box 2456
Oak Ridge, TN  37831

Perry, Dr. Steve
USEPA/NOAA (MD-80)
Research Triangle Park, NC  27711
(919) 541-1375

Peterson, W. B.  -
USEPA (MD-80)   f
Research TriangWPark, NC  27711
(919) 541-4564

Pierce, Thomas E.
NOAA Meteorologist
USEPA (MD-80)
Research Triangle Park, NC  27711
(919) 541-1375
FTS 629-1375

Pritchett, Thomas H.
USEPA-ERT
GSA Raritan Depot, Bldg. 10
Edison, NJ  08837
(201) 321-6738

Randerson, Dr. Darryl
National Weather Service
Nuclear Support Office, NOAA
P. 0. Box 14985
Las Vegas, NV  89114
(702) 598-3234

Rao, Dr. K. Shankar
ATDD, P. 0. Box 2456
Oak Ridge, TN  37831
(615) 576-1238
FTS 626-1238

Rao, S. Trivikr
Bureau of Air Research
New York State Dept. of
  Environmental Conservation
Division of Air Resources
50 Wolf Road
Albany, NY  12233
(518) 457-3200
Rhodes, William J.
USEPA  (MD-62B)
Research Triangle Park, NC  27711
(919)  541-2853
FTS 629-2853

Rodak, Victoria A.
ICF Technology
1850 K Street, NW, Suite 950
Washington, DC  20006

Rosensteel, Robert E.
USEPA  (MD-13)
Research Triangle Park, NC  27711
(919) 541-5605
FTS 629-5605

Runkle, Gene E.
USDOE/AL, ESHD
P. 0. Box 5400
Albuquerque, NM  87115
(505) 846-2046

Sakenas, Cheryl
Office of Inspection & Enforcement
NRC, MS-MNBB-3302
Washington, DC  20555

Schiermeier, Francis A.
Meteorology Division (MD-80)
USEPA
Research Triangle Park, NC  27711
(919) 541-4541

Schroy, Jerry M.
Monsanto Company
Mail Code CS7H
800 N. Lindbergh
St. Louis, MO  63167
(314) 694-6174

Sedefian, Leon
New York Dept. of Environmental
  Conservation
Division of Air
50 Wolf Road, Room 115
Albany, NY  12233
(518) 457-7605
                                   E-7

-------
Shannon, Dr. Jack
Atmospheric Physics
ER Building 181
Argonne National Laboratory
Argonne, IL  60439
(312) 972-5807

Shephard, Cathleen
USEPA
OSWER/PS, WH-548 A
401 M Street, SW
Washington, DC  20460
(202) 475-8247

Snyder, Dr. William H.
USEPA (MD-81)
Research Triangle Park, NC  27711
(919) 541-2811

Steinberg, Ken W.
Exxon Research & Engineering Company
P. 0. Box 101
180 Park Avenue
Florham Park, NJ  07932
(201) 765-1209

Swank, Dr. Robert R., Jr.
Director of Research
Athens Environmental Research Lab
USEPA
College Station Road
Athens,.GA  30613
(404) 546-3128
FTS 250-3128

Topoleski, Gary
USEPA (TS-779)
401 M Street, SW
Washington, DC  20460
(202) 382-2251

Touma, Joe
USEPA, MD-14
Research Triangle Park, NC  27711
(919) 541-5681

Turner, D. Bruce
USEPA (MD-80)
Research Triangle Park, NC  27711
(919) 541-4564
Vaughan, Dr. William M.
Aero Viroment, Inc.
8515 Delmar Blvd., Suite 215
St. Louis, MO  63124
(314) 993-0543

Weber, Dr. Alan H.
Bldg. A773, Room A1012
E. I. Du Pont
Savannah River Laboratory
Aiken, SC  29808
(803) 725-3717
FTS 239-3717

Wiser, Dr. Herbert L.
Office of Air & Radiation
USEPA (ANR-443)
Washington, DC  20460
(202) 382-7747

Wratt, David S.
New Zealand Meteorology Service
P. 0. Box 722
Wellington, New Zealand

Yatnada, Tetsuji
MS F665
Los Alamos National Laboratory
P. 0. Box 1663
Los Alamos, NM  87545
(505) 667-8353
FTS 843-8353
                                   E-8

-------