udm_mer_v1.0_tcm4-76493

Model Evaluation Reporton

UDM Version 6.0

Ref. No. SMEDIS/00/9/EVersion 1.0

21 January 2002

Prepared byCambridge Environmental Research Consultants Ltd.

3, Kings ParadeCambridgeCB2 1SJ

UK

UDM, Ver. 6.0 Version 1.0 (21/01/02) ii

ContentsPreface .................................................................................................................................................ivKey to summary information ..............................................................................................................vi0. Evaluation information.....................................................................................................................1

0.1 Protocol ..................................................................................................................................10.2 Evaluator ................................................................................................................................10.3 Date ........................................................................................................................................1

1. General model description................................................................................................................21.1 Name, version number and release date.................................................................................21.2 Short description of model .....................................................................................................21.3 Model type..............................................................................................................................31.4 Route of model into evaluation project ..................................................................................31.5 History of model.....................................................................................................................41.6 Quality assurance standards adopted......................................................................................41.7 Relationship with other models..............................................................................................41.8 Current model usage...............................................................................................................61.9 Hardware and software requirements.....................................................................................61.10 Availability and costs ...........................................................................................................7

2. Scientific basis of model ..................................................................................................................82.1 Specification of the source .....................................................................................................82.2 Specification of the environment .........................................................................................112.3 Model physics and formulation............................................................................................142.4 Solution technique................................................................................................................282.5 Results or output available from model ...............................................................................292.6 Sources of model uncertainty...............................................................................................322.7 Limits of applicability ..........................................................................................................322.8 Special features.....................................................................................................................342.9 Planned scientific developments ..........................................................................................34

3. User-oriented aspects of model ......................................................................................................353.1 User-oriented documentation and help.................................................................................353.2 Installation procedures .........................................................................................................353.3 Description of the user interface ..........................................................................................363.4 Internal databases .................................................................................................................363.5 Guidance in selecting model options....................................................................................373.6 Assistance in the inputting of data .......................................................................................383.7 Error messages and checks on use of model beyond its scope ............................................383.8 Computational costs .............................................................................................................393.9 Clarity and flexibility of output results ................................................................................393.10 Suitability to users and usage .............................................................................................403.11 Possible improvements.......................................................................................................413.12 Planned user-oriented developments..................................................................................41

4. Verification performed ...................................................................................................................424.1 Summary of verification.......................................................................................................424.2 Comments.............................................................................................................................43

5. Validation performed .....................................................................................................................445.1 Validation already performed...............................................................................................44

Contents

UDM, Ver. 6.0 Version 1.0 (21/01/02) iii

6. Conclusions ....................................................................................................................................46General model description..........................................................................................................46Scientific basis of model ............................................................................................................46Limits of applicability ................................................................................................................47User-oriented aspects of model ..................................................................................................47Verification performed ...............................................................................................................48Validation performed .................................................................................................................48Advantages and disadvantages of model....................................................................................48Suitability of protocol for assessment of model.........................................................................48

7. References ......................................................................................................................................49Supplied documents ...................................................................................................................49

Appendix 1: Actively-generated information.....................................................................................50Appendix 2: Comments from model developer .................................................................................51

UDM, Ver. 6.0 Version 1.0 (21/01/02) iv

Preface

Technical models are widely used to inform decisions relating to safety problems. The quality of agiven model and its appropriate and defensible use in simulating a given problem are clearly ofgreat importance, and such issues have provided the primary motivation for the development ofscientific model evaluation, a technique which aims to provide information addressing these aspectsof models and their use.

Scientific model evaluation has been supported by the European Union in a number of ways, one ofthe most recent being the project SMEDIS (Scientific Model Evaluation of Dense Gas DispersionModels). The principal objective of this project is to develop a structured procedure or protocol forthe scientific evaluation of dense gas dispersion models, with particular emphasis on the complexeffects of aerosols, terrain and obstacles, and then to apply this protocol to models used in Europe tosimulate dense gas dispersion problems.

The protocol which has been developed for SMEDIS is consistent with the guidelines issued by theCEC Model Evaluation Group1, in that evaluation comprises the three elements of validation,verification and assessment:

(a) assessment is the examination of a model according to a series of detailed categories,including specific categories for the area of application being highlighted (in this caseaerosols, terrain and obstacles)

(b) validation is the quantitative comparison of experimental observations with modelpredictions

(c) verification is the confirmation that the (computer) implementation is an accurate translationof the model algorithms

The evaluation of each model in SMEDIS focuses on the validation and assessment elements: avalidation exercise is carried out using selected data sets typifying a wide range of release andenvironmental conditions, including a significant number in which complex effects play animportant role; and a scientific assessment is carried out concentrating on the scientific basis of themodel and its user-oriented aspects. Verification is effectively absorbed into the assessment by theinclusion of a section reporting previous verification carried out on the model.

This report represents the output from the assessment element in the scientific evaluation of UDM,Version 6.0. It has been produced by analysing information supplied on this model by a nominatedorganisation (either the model developer or an experienced user). The information has been elicitedby means of questionnaires and supplied mainly in the form of pre-existing documentation on themodel. Only this information has been taken into account in preparing this report2.There are six main sections, each of which is part description and part analysis. Sections 1-3 formthe principal part of the assessment, focusing on the general description, scientific basis and user-oriented aspects of the model, respectively. Each section is divided up into specific subjectheadings, many of which include summary information presented in check box format for rapidoverview of the capabilities and formulation of the model. Sections 4 and 5 summarise previousverification and validation work performed on the model, and Section 6 concludes with a summaryof the findings in the report. References are listed for the documentation utilised in the assessment.In addition two appendices are included containing both the comments from the model developerand a summary of the new actively-generated results from the validation exercise. 1 CEC Model Evaluation Group Model Evaluation Protocol Version 5, May 1994.2 Together with comments from the model developer on the draft version of the report.

Preface

UDM, Ver. 6.0 Version 1.0 (21/01/02) v

This report is intended for use by both the model developer and model users:

for the model developer it represents an independent assessment of their model according to aprotocol which has been applied to a wide range of other dense gas dispersion models. Ithighlights both the strengths and the weaknesses of the model

for the model user the report assists them in deciding whether the model is appropriate to theirintended use. The inclusion of both scientific and user-oriented aspects of the model in thereview helps the user to gauge how well the model can simulate the specific scientificproblem of interest as well as how well the model performs from a practical point of view.

In all cases, the report can form part of the standard documentation accompanying the model in thespecific version.

UDM, Ver. 6.0 Version 1.0 (21/01/02) vi

Key to summary information

In Sections 1-3 summary information is provided on many aspects of the model. This is organisedas follows:

(a) Where there is a mutually exclusive choice of options, the relevant choice is denoted by a circlecontaining a dot , while the remaining choices are accompanied by an empty circle .

(b) A square box containing a tick denotes that a feature is present or an option applies to themodel; if the box contains a cross , this emphasises that the feature is not present or that theoption does not apply.

(c) A character between square brackets can convey one of the following meanings:[U] denotes that the user can specify the details of the feature;[M] denotes that the model specifies the details of the feature;[?] denotes that the status of the feature is uncertain;[n] denotes that the item has the value n.

(d) Any parts which are greyed out are not relevant to the model in question.

UDM, Ver. 6.0 Version 1.0 (21/01/02) 1

0. Evaluation information

0.1 ProtocolThis scientific assessment was carried out using the SMEDIS Model Evaluation Protocol,Version 1.0 (24 June 1997), and the Model Evaluation Report template Version 1.01.

0.2 EvaluatorThe scientific assessment was carried out by:

R.E. Britter, CERC Ltd.

0.3 DateThe date of this scientific assessment is 11 July 2000, comments incorporated 21 January 2002.

UDM, Ver. 6.0 Version 1.0 (21/01/02) 2

1. General model description

1.1 Name, version number and release date

Name: UDM (within PHAST) Unified Dispersion Model

Version number: 6.0

Release date: February 2000 (as part of PHAST6.0 release)

1.2 Short description of modelThe Unified Dispersion Model (UDM) is an integral model to calculate the dispersion following atwo-phase pressurised release or an unpressurised release. It effectively consists of the followinglinked modules:

near-field jet dispersion

non-equilibrium droplet evaporation and rainout, touchdown

pool spread and vaporisation

heavy gas dispersion

far field passive dispersion.In addition to the non-equilibrium droplet thermodynamics model, UDM also allows for a two-phase HF thermodynamics model (including effects of polymerisation). This evaluation documentdoes not address the module for pool spread and vaporisation.

The UDM allows for continuous instantaneous, constant finite-duration and general time-varyingreleases. The UDM allows for possible plume lift-off when a grounded plume becomes buoyant.

The latest version of the UDM is currently implemented in the consequence-analysis packagePHAST6.0. It is planned to be included in the next version 6.1 of the onshore risk-analysis packageSAFETI and in the next version of the offshore risk-package NEPTUNE (successor to OHRAT).Possible PC-based operating systems are Windows NT, 95, 2000. The programming language forthe model code is Fortran 77, and C++ for the user-interface.

Output is in the form of graphics and/or tabular output. The output is in the form of cloudparameters, e.g. center-line concentration, ground-level concentration, plume height, plume depth,plume width, vapour temperature, liquid temperature, etc.

Category 1: General model description

UDM, Ver. 6.0 Version 1.0 (21/01/02) 3

1.3 Model type Screening tool Integral model Shallow layer model CFD model 1-D 2-D 3-D

1.4 Route of model into evaluation project

1.4.1 Model supplier

Developer Licensee Other

Contact details:

Name: Neil Prophet (sales manager)Address: Risk Management Software

Det Norske VeritasPalace House3 Cathedral StreetLondonSE1 9DEUK

Telephone: +44 20 77166615Fax: +44 20 73577297E-mail/Web: [email protected]

1.4.2 Model developer

As above

Contact details:

Name: Henk W.M. WitloxAddress: Risk Management Software

Det Norske VeritasPalace House3 Cathedral StreetLondonSE1 9DEUK

Telephone: +44 20 77166711Fax: +44 20 73577297E-mail/Web: [email protected]


UDM, Ver. 6.0 Version 1.0 (21/01/02) 4

1.5 History of model

1.5.1 Model ancestors

There is some historical relation to previous models from Technica, specifically the jet modelTECJET and the heavy gas dispersion model within WHAZAN.

The initial work on the UDM model started in 1990.

The original version of UDM was developed by Woodward and Cook as new technology in theearly nineties. The new UDM 6.0 version represents a significant revision and extension to all partsof the model.

1.5.2 Features inherited

Unclear.

1.6 Quality assurance standards adopted

1.6.1 Model development

MEG guidelines Other

Following a thorough internal assessment in the second half of 1997, the UDM documentationhas been thoroughly reviewed and revised. The code has been checked line-by-line to confirmconsistency against the documented theory. Verification and validation, both for each individualmodule in the program and the overall program, has been carried out. Full documentation oftheory, verification and validation is included in the UDM Technical Reference Manual.

1.6.2 Software development

National International Organisation ISO 9000 Other

The software development process at Risk Management Software (Section in DNV Software) isbeing run to the ISO 9000 Tickit Standard.

1.7 Relationship with other models

1.7.1 Status of dispersion model being evaluated

Self-contained Can be used as one part of suite Inextricably bound to other models General-purpose, only one part required Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 5

1.7.2 Extent and demarcation

The latest version of the UDM is currently implemented in the consequence-analysis packagePHAST6.0. It is planned to be included in the next version 6.1 of the onshore risk-analysis packageSAFETI, and in the next version of the off-shore risk-analysis package NEPTUNE (successor toOHRAT).

The UDM model can be clearly defined within the packages PHAST and SAFETI.

1.7.3 Interfacing with other models

Dispersion models are commonly interfaced with, on the one hand, source models, and, on the otherhand, consequence models.

UDM is interfaced with consequence models for jet fires, BLEVEs, pool fires, explosions andtoxicity.

Similarly, from the documentation, it is apparent that UDM is interfaced to release models forsingle and multi-phase materials, pressurised and unpressurised.

This evaluation is essentially for dense gas dispersion models (modules) and not for associatedsource models. Thus the pool spread and vaporisation model is not included within this evaluation.However, it should be noted that the pool spread and vaporisation model has been included withinthe UDM model, partly because it forms a link between rain-out from the dispersion model andsubsequent re-evaporation of the material to enter a further dispersion calculation.

A downside to this is that, from notes in the validation exercise, there is no simple way to use UDMto model a pool evaporation problem directly without preceding it with an artificial jet.

An instantaneous energetic expansion model is included with UDM (though still undergoingdevelopment) and this has been treated as a source model for a subsequent dense gas dispersionmodel and is not considered further here. This said, the portion of UDM that deals with themomentum jet following the release and prior to any grounded dense gas dispersion model isdifficult to separate from the grounded dense gas dispersion model itself. As the linking betweenthese models is often a dominant modelling concern it is probably wise to include the momentumjet aspect of the modelling within this evaluation.

1.7.4 Comments


UDM, Ver. 6.0 Version 1.0 (21/01/02) 6

1.8 Current model usage

1.8.1 Type of user

Background Engineer Consultant Regulator Academic Other

Type of experience Dispersion Fluid dynamics Thermodynamics Numerical methods Programming Consequence modelling Risk analysis Other

Length of experience Hours Days Weeks Months Years

1.8.2 Model distribution

Location Outside model developer

Country of origin Continent of origin Worldwide Other Industry Consultancies Universities Regulatory authorities Other

Numbers 100

1.8.3 Comments

1.9 Hardware and software requirements

1.9.1 List of requirements

Computer platforms PC: 386 PC: 486 PC: Pentium/Pentium II Workstation Main frame Vector/parallel machine Other

Peripheral hardware Monitor Keyboard Mouse CD-ROM drive Printer Plotter Other

Memory 5 Mbytes

Disk space 100 Mbytes

Operating system DOS Windows UNIX VMS Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 7

Additional software Required Optional Not required Compiler Graphics package GIS Other

Graphical device requirements High resolution Colour display Other

Internet Explorer is required for access of the on-line HELP documentation.

1.9.2 Comments on overall requirements

1.10 Availability and costs Proprietary Shareware Public domain Other

Licence Perpetual licence/buy outright Not available Other

Cost not provided.

UDM, Ver. 6.0 Version 1.0 (21/01/02) 8

2. Scientific basis of model

A simple categorisation of dense gas dispersion models is included in the protocol, consisting offour main types:

(a) Simplified empirical screening models or screening tools based on either a generic interpretationof a problem or specific experience in a restricted domain

(b) One-dimensional integral models in which the development of the flow is in one spatialdimension. Variations of, say, the concentration field in the other two dimensions areaccommodated by the assumption of self-similarity in the concentration field.

(c) Shallow-layer models or two-dimensional integral models in which the development of the flowis in two spatial dimensions

(d) Fully three-dimensional models allowing for the development of the flow field in three spatialdimensions and time. This would include computational fluid dynamics (CFD) and the attendantmodels for turbulence.

The present model may be classified as a one-dimensional integral model.

2.1 Specification of the sourceThe inputs for dense gas dispersion models are, typically, provided from a release model e.g. amomentum jet model or a liquid pool model.

2.1.1 Primary origin for source

2.1.1.1 Types of release conditions available directly to dispersion model

Release models available Gaseous jet Liquid jet Two-phase jet Liquid pool

Catastrophic release Other

A liquid pool could be indirectly modeled following 100% rainout. Further improvements areplanned to allow dispersion directly from a pool.

2.1.2 Fluid dynamic properties of source

2.1.2.1 Instantaneous releases

Instantaneous releases

[M] Spatial dimensions [M] Symmetry [U] Velocity[M] Volume [M] Density [U] Mass

Multiple sources [U] Elevation [] Other

Entrained air

Category 2: Scientific basis of model

UDM, Ver. 6.0 Version 1.0 (21/01/02) 9

2.1.2.2 Continuous releases

Continuous releases

[M] Spatial dimensions Orientation [M] Symmetry [U] Velocity[M] Volume flow rate [M] Density [U] Mass flow rate

Multiple sources [U] Elevation [] Other

Entrained air

Model can recalculate source dimensions

UDM allows for initial air mixed in for instantaneous, continuous and time-varying releases. InPHAST, the amount of initial air mixed is calculated by separate source models. PHASTpresently does not allow initial air mixed in for standalone UDM runs.

UDM allows for orientation to be specified for continuous and time-varying releases (releaseangle in vertical plane of wind direction, no upwind or crosswind releases).

2.1.2.3 Time-varying releases

Time-varying releases

[M] Spatial dimensions Orientation [M] Symmetry[M] Volume flow rate [M] Density [U] Mass flow rate

Multiple sources Time variation [] Other

Entrained air

Model can recalculate source dimensions

Time-varying releases are treated by segmenting the release rate, tracking the development ofthese segments (taken to be part of a continuous release) and then re-assembling the segmentsincorporating concentration profiles.

2.1.2.4 Other aspects of release types

Guidance provided on choice of dispersion source type Instantaneous Continuous Time-varying Other

2.1.3 Thermodynamic properties of source

[M] Temperature [M] Quality (two-phase) [] Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 10

Thermodynamic properties for input to UDM will typically be supplied by the output of a sourcemodel. Alternatively, the UDM allows the user to specify either the quality (two-phase) or thetemperature (vapour or liquid).

2.1.4 Chemical properties of source Composition

[M] Effective single component [] Components (#) [] Passive tracer [] Other

UDM does not treat multi-component releases, but adopts effective single (or pseudo)compound properties for a user-specified mixture. These effective properties are automaticallygenerated from the PHAST/SAFETI property system.

Dependence of physical properties

[M] Temperature [M] Pressure [M] Composition [] Other

UDM accesses DIPPR thermodynamic database to evaluate physical properties as function oftemperature, pressure and composition. It does perform internal calculations for binarydiffusivity, dynamic vapour viscosity and specific heat of the vapour cloud.

Chemically reactive substances

Reaction with H2O Specific reactions Oligomerisation [] Other

But for HF only.

Radioactive substances

[M] Pre-defined substances (#)[] Flammable [] Toxic [0] Mixtures [] Other

Properties for flammable substances, toxic substances and mixtures are derived from thePHAST/SAFETI property system, using the DIPPR property database. A basic set of 59chemicals (toxic and/or flammable) is presented in the basic database, but access could beprovided to the entire DIPPR database (over 1000 chemicals).

[U] User-defined substances

User-defined substances can be specified by the user using the PHAST/SAFETI property system.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 11

2.2 Specification of the environment

2.2.1 Frame of reference

2.2.1.1 Coordinate system

Cartesian Cylindrical polar Spherical polar Other

2.2.1.2 Relation to environment

Origin at source x-axis downwind Other

PHAST enables the user to import a bitmap, and to indicate the location of the origin on the map.

2.2.2 Atmosphere

2.2.2.1 Mean wind field

Mean wind parameterised

Vertical profile Horizontal field Time-varying

[M] Dependence on stratification [] Other

Vertical velocity profiles used

Logarithmic Other

[U] Velocity at reference height [] Friction velocity specified

Mean wind modelled

[] Vertical profile [] Horizontal field [] Time variation Other

Zero wind allowed

The vertical velocity profile is a power law approximation to the stratification-modifiedlogarithmic law. The approximation is made over the layer from 10m to 100m. Default cut-offvalues of 1m and 200m are used, that is the velocity is uniform at heights below 1m and above200m and is continuous at these two points.

Note also that the logarithmic portion of the velocity profile has been modified from loge(z/z0) tologe[(z+z0)/z0]. This is a common modification to remove the singularity as z becomes small. Itseffect may be significant for large z0 and very dense clouds.A uniform vertical velocity profile may also be selected by the user. This is of little use for theaverage user it may be used by the specialist user (or modeler) to examine the effect of thevertical wind speed gradient on the dispersion predictions.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 12

2.2.2.2 Turbulence

Turbulence parameterised Turbulence modelled

2.2.2.3 Stratification

Stability ranges

[U] neutral [U] stable [U] unstable

Stratification parameterised

Vertical density profile Horizontal density field Time-varying[U] Pasquill-Gifford stability categories used [M] Monin-Obukhov length used

Monin-Obukhov Pasquill-Gifford conversion

Stratification modelled

[] Vertical density profile [] Horizontal density field [] Time-varying Other

Temperature and pressure profiles are internally calculated by the model.

For the temperature profile two forms are available: a constant temperature gradient prescribedas a function of the Pasquill-Gifford stability class or a stability-modified form of a logarithmictemperature profile.

For the pressure profile two forms are available: an atmospheric pressure independent of heightand one where the pressure decreases linearly with height (based on a constant atmosphericdensity).

The atmospheric density is separately calculated from the ideal gas equation using theatmospheric pressure and temperature.

Relative humidity and atmosphere composition are taken to be independent of height.

2.2.2.4 Use of meteorological data

Meteorological data used

[U] Temperature [U] Humidity [] Cloud cover

[] Date/time [] Latitude/longitude [U] Other

The user may specify the solar radiation flux.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 13

2.2.3 Terrain

2.2.3.1 Terrain types available

Non-flat terrain[] Single slope [] Slope segments (#) [] General 2-D [] Other

2.2.3.2 Wind orientation for non-flat terrain

Downslope Upslope General [] Other

2.2.3.3 Surface characteristics

[U] Roughness length

Pre-defined surface types User-defined values

[U] Temperature [] Other

The surface temperature can be specified independently from the ambient atmospheric ground-level temperature (default value of surface temperature equals ground-level ambienttemperature).

2.2.4 Obstacles

2.2.4.1 Obstacle types available

Obstacle types available 2-D fence Cylindrical building Cuboidal building

General shape 1-sided canyon 2-sided canyon Other

2.2.4.2 Obstacle distribution

Distribution of obstacles

[] Max number (#)

[] Positions [] Orientations [] Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 14

2.2.4.3 Obstacle characteristics

Dimensions

[] Horizontal [] Vertical [] Other

Structural characteristics

[] Porous [] Other

2.3 Model physics and formulation

2.3.1 Fundamental equations, initial conditions and boundary conditions

2.3.1.1 Equations used/set up

Volume Mass Momentum Width/radius Enthalpy Temperature

Concentration Species concentration Other

Differential equations are provided for the excess horizontal and vertical components ofmomentum. Differential equations are also provided for the horizontal and vertical position ofan element of a continuous plume and the position of an instantaneous puff; and for the rate ofheat convection and water vapour transfer from the substrate.

2.3.1.2 Dependent variables3

H W R

Uad u v w

c T {ci} Other

2.3.1.3 Independent variables (#)

spatial (#) time Other

For continuous releases the independent variables are spatial while for instantaneous releases theindependent variable is time.

3 H = depth; W = width; R = radius; Uad = advection velocity; = density; (u, v, w) = velocity components;c = concentration; T = temperature; ci = concentration of species i.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 15

2.3.1.4a Model-type-dependent features of equation formulation (screening tool)

Physical quantities for which correlations are available

Concentration Other

Main quantities used to form dimensionless groupings

Length scale Velocity scale Ambient density Other

Main independent variables

Downwind distance Wind speed Release size/rate Other

Physical quantities fixed (or given limited values) for purposes of correlation

Wind speed Atmospheric stability Surface roughness Other

Not applicable.

2.3.1.4b Model-type-dependent features of equation formulation (integral)

Use of similarity profiles in setting up equations

Dependent variables

Concentration Velocity Temperature Other

Profile shapes used

Uniform Gaussian Other

See Section 2.3.6.

2.3.1.4c Model-type-dependent features of equation formulation (shallow layer)

Not applicable.

2.3.1.4d Model-type-dependent features of equation formulation (CFD)

Not applicable.

2.3.1.5 Turbulence modelling

Turbulence models available

[] k- [] Buoyancy-modified k- [] Algebraic stress [] Reynolds stress [] Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 16

2.3.1.6 Initial conditions

Initial conditions specified

Source Atmosphere Terrain Obstacles Other

2.3.1.7 Boundary conditions

Boundary conditions specified

Source Atmosphere Terrain Obstacles

Cloud boundary Other

2.3.2 Dispersion - advection

Advection velocity derived from wind profile

Average over cloud height [] Wind speed at fraction of cloud height Other

Acceleration of stationary cloud from rest calculated

Advection modelled directly

The advection speed of the cloud (puff or element of a plume) is taken to be the cloud speed atthe height of the centroid of the vertical concentration distribution.

The advection speed is calculated from the horizontal momentum equation. This allows for thetreatment of an elevated jet and grounded jet or dense gas plume with the same structure.

The horizontal momentum equation is, in fact, for the excess horizontal momentum

( )( )cacldxcld

cacldxx

zumumzumII

==2

with mcld being a mass or a mass flux for instantaneous or continuous releases respectively. uxand ua are the velocities of the cloud and the ambient wind, respectively, and zc is the height ofthe cloud centre.

For a stationary instantaneous release Ix is initially set to zero.For a continuous release the initial horizontal cloud speed is derived from the release speed.

The initial excess horizontal momentum is then set as

( )( ).2 Raxcldx zuumI =The equation for excess horizontal momentum has three terms; airborne, impact drag and grounddrag.

The airborne drag has been taken to be zero for both continuous and instantaneous releases.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 17

The impact drag follows the formulation used in HGSYSTEM based on the assumption ofelastic collision is applied to the plume as a whole. The approach is at variance withfundamental laws of fluid mechanics and is, therefore, suspect.

Apart from the effect of local shear stresses at the ground plane the effect of impact on the planemust be to remove all the vertical momentum (flux) and to conserve the horizontal momentum(flux). To this one might postulate the conservation of kinetic energy (flux). This requires thegeneration of lateral velocities (symmetrical to ensure no generation of a net lateral momentumflux).

The ground drag formulation follows the approach used in HGSYSTEM except for amodification to ensure that the drag reduces to zero when the cloud speed equals the wind speed.

Finally, and it may not be of significance, is that the excess horizontal momentum is defined interms the excess over the local ambient velocity. As the cloud descends the local ambientvelocity is changing and so even with no external forces being applied the excess horizontalmomentum is changing; no account for this is taken.

2.3.2.1 Two phase jet model

The grounded dense gas plume dispersion is preceded by a jet model which smoothly changesinto the grounded gas plume model.

The jet model is a conventional model incorporating the usual three entrainment terms

along jet term

across jet term

passive dispersion term.For an elevated jet these three terms are summed (prior to transition to a passive plume) and,after transition, only the last term is used.

The terms for passive dispersion are slightly different before and after transition.

For a grounded jet, the maximum of the sum of the first two and a separate term for dense gasdispersion, is added to the passive dispersion term prior to transition and after transition only apassive term is used.

This is a common and appropriate approach. The coefficients adopted for the first two terms are1 = 0.17 and 2 = 0.35. The authors note literature values for 1 range from 0.11 to 0.28 andfor 2 from 0.16 and 0.60. It is somewhat surprising that such a range of constants are stillprevalent in what should be a reasonably straightforward problem. The UDM values, somewhatcoincidentally are quite central within the range.

For the third, passive entrainment term, the near-field formulation follows that in HGSYSTEMbased on experiments from Disselhorst, while the far-field formulations are based on derivativesof the correlations for the passive dispersion coefficients from McMullen and Hosker asdescribed elsewhere. The latter is appropriate while the former, though somewhat speculative isbased on directly relevant experiments.

The jet model has been extended to instantaneous releases. It is a little unclear what is theintended use of this model extension.

If the intent is to allow consideration of an instantaneous puff with initial momentum etc. thensome reconsideration of all the entrainment terms is necessary. Momentum (or buoyant) puffshave a quite different structure (and consequent entrainment specification) to jets. This is not to


UDM, Ver. 6.0 Version 1.0 (21/01/02) 18

say that the approach adapted in UDM will produce incorrect results, only that some furtherjustification for the approach is probably necessary.

Alternatively, if the intent of the model extension is to provide a model for the acceleration of aninstantaneous release up to the ambient wind speed which is consistent with the jet formulation,then the effect of the formulation may be slight anyway. If the instantaneous release has arisendue to expansion of a catastrophic release of pressurised material then dilution during theexpansion will ensure that the momentum of the puff will be determined by the momentum ofthe entrained air.

It may be helpful to a reader of the documentation to provide some examples of where themodels might be relevant.

2.3.3 Dispersion - gravity spreading

Gravity spreading parameterised

Gravity spreading locations

Plume sides Puff edge Cloud boundary Other

Gravity spreading characterisation

[M] Constant Froude number Other

Upstream/lateral spreading of vapour blanket (continuous releases) Upstream spreading Lateral spreading Other

Gravity spreading modelled directly

The cloud in UDM is characterized by an "equivalent" cloud with effective height Heff, aneffective cloud half-width Weff, a cloud speed Ucld and an equivalent top-hat concentration equalto the centre-line concentration

+==

=

0

0

)(11)(

),,()0,,(

1

xRn

dF

dyxcyxc

H

zv

eff

and, similarly,

+==0

)(11)( xRm

dyyFW yheff

The symbols in these equations are defined in sections 2.3.4.3 and 2.3.6.

An effective cloud velocity is calculated as the integral in the vertical of the concentration-weighted ambient velocity. However UDM does not use this velocity but uses Ucld, which is theambient velocity at the position of the cloud centroid.

Similar approaches are adapted for instantaneous releases.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 19

The model UDM used { } 2/1)1(15.1 deffeff hHgdtdW

+=

where gzgcld

cacld

)(=

The general approach for gravity spreading is a common one and the coefficient typical.

UDM allows its elliptical cloud cross-sectional shape to be retained as it intersects and passesthrough the ground plane.

Only that portion of the cloud above the ground is physical, containing cloud material. Theparameter hd is the fraction of the area in the bottom half of the ellipse which is above ground.While a continuous plume is touching down, the continuous cloud cross-section graduallytransforms from a circle to a semi-ellipse (with increased crosswind spreading during touch-down). While an instantaneous cloud is touching down, the cloud shape gradually transformsfrom a sphere to a semi-ellipsoid.

2.3.4 Dispersion - dilution

2.3.4.1 Dilution by direct turbulent diffusion

Dense gas regime Passive regime

There is no explicit dilution by direct turbulent diffusion in the dense gas regime. There may bean implicit effect of direct turbulent diffusion through the empirical variation of the exponents mand n (see 2.3.3) in the similarity profiles for the concentration profile in the horizontal. Thiswill lead to a concentration profile which is not top-hat and this may be ascribed to turbulentdiffusion. However, it must be noted that the effective width Weff will not be changed, nor willthe centreline concentration.

2.3.4.2 Dilution by air entrainment

Entrainment parameterised

[M] Top surface [M] Side surface [] Other

Entrainment modelled directly

The top entrainment is parameterised with an entrainment velocity )(/ *Rixuu xt =

where

72.147/72.143625.27.1/)8.01(

3625.2010|)|65.01()(

**

*2/1

*

*

*1

**

>=


UDM, Ver. 6.0 Version 1.0 (21/01/02) 20

The side entrainment is parameterised with an entrainment velocity i.e. ud W

dtseff=

with for continuous releasesfor instantaneous releases

==

00 3.

The approach is typical; the former selection is now common. The latter selection forinstantaneous releases is unusual but does reflect that the typical selection of a constant near tounity may just reflect the particular nature and geometry of the Thorney Island experiments; apredominant model validation source.

2.3.4.3 Shear dispersion

Contributions to shear dispersion modeled

[M] Longitudinal Lateral Other

No contribution to shear dispersion is included for instantaneous releases.

For finite duration continuous releases two approaches are offered; a quasi-instantaneous modeland a finite duration correction.

The former is a straightforward change from a finite duration continuous release to an"equivalent" instantaneous release. The drawback to this approach, as the model authors state,is the introduction of an abrupt transition with consequent discontinuities in some outputs. Afurther finite duration adjustment is made to the quasi-instantaneous model but this is anaveraging time correction and will be addressed elsewhere

The second approach is a finite-duration correction. This is a direct application of alongwinddispersion to the finite duration release. An alongwind dispersion coefficient, x, is formed froma contribution from vertical shear and a contribution from alongwind turbulent diffusion (whichis equated with lateral turbulent diffusion).

The centre-line ground level concentration for a constant release with duration tdur is obtainedfrom the steady-state ground-level concentration by multiplying it with a function F where

=

x

durctUF

2/32erf

where Uc is a mean convection velocity of the cloud. It must be emphasised that this provides acorrection only to the centre-line ground-level concentration.

As stated by the authors Strictly speaking the model applies to the following scenario only:

ground-level non-pressurised release

no significant rainout

uniform release of a finite durationIt cannot be used for time-varying release rates.

2.3.5 Dispersion - concentration fluctuations

Concentration fluctuations considered


UDM, Ver. 6.0 Version 1.0 (21/01/02) 21

2.3.5.1 Fluctuation calculations

[] Derived from data [] Modelled directly Other

2.3.6 Dispersion - concentration profiles

Concentration profiles considered

The similarity form for the concentration profile for steady-state releases is

)()()(),,( 0 yFFxcyxc hv =

where is the co-ordinate normal to the horizontal direction and the trajectory of the plume

=

=

)(

)(

)(exp)(

)(exp)(

xm

yh

xn

zv

xRyyF

xRF

Of particular novelty in this model is that the exponents n(x) and m(x) are prescribed functions.The function m(x), for the horizontal profile of concentration, varies from 2 to 50 essentiallyfrom a Gaussian profile to a top-hat profile. m(x) is a prescribed function of the relative densitydifference of the plume. No evidence is presented to support the empirical functionaldependence or why it should depend upon the relative density of difference, rather than, forexample, a Richardson number.

The function n(x) varies between 2.5 and 1.0, dependent upon the atmospheric stability and theratio of plume depth to Monin-Obukhov length. Under neutral atmospheric stability n(x) reducesto 2.0. The correlation is similar to correlations for atmospheric flux gradients, however noevidence is presented to connect the atmospheric flux gradients with the exponent n(x).The use of empirical functions for m(x) and n(x) does produce useful simplifications for modeldevelopment and there is no evidence that it is physically unwise. However it does appear to beapplied to both an elevated jet flow and for a grounded dense gas plume. This must bequestionable in that, for example, a high pressure gas jet must have the same profiles in bothdirections; a situation apparently not satisfied by the model.

A similar approach is adopted for an instantaneous release. The vertical profile from Fv is thesame as before, however the horizontal profile form Fh, incorporates both x and y, i.e.

yx

m

yxR RRR

yRxyxF =

+

= with exp),(

2/22

The co-ordinate x is relative to the centroid of the puff. The assumption Rx = Ry does not allowfor alongwind shear dispersion.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 22

2.3.6.1 Types of concentration profiles used for dense gas dispersion region

Vertical profile

Uniform Gaussian Exponential decay Other

Lateral profile

Uniform Gaussian Uniform core with erf edges Other

Radial profile

Uniform Gaussian Uniform core with erf edges Other

2.3.6.2 Types of concentration profiles used for passive dispersion region

Concentration profiles applied

Uniform Gaussian Exponential decay Other

2.3.7 Thermodynamics

Thermodynamics considered

The UDM invokes the thermodynamic model while incrementally solving the dispersionequations in the downwind direction. The thermodynamics model calculates the following data:

phase distribution of component and of atmospheric moisture

vapour cloud temperature

cloud density

cloud volume or volume fluxFor ground level dispersion the model may take into account water-vapour and heat transferfrom the substrate to the cloud.

Three types of thermodynamic model are available

equilibrium thermodynamics (no reaction)

non-equilibrium thermodynamics (no reaction)

equilibrium thermodynamics (with reaction)The last model is only available for HF.

Finally, the UDM model also provides for water vapour from an underlying water surface whenthe vapour temperature of the cloud is less than that of the water surface. The approach adoptedfollows that in HGSYSTEM and relates the water vapour pick-up to the rate of heat convectionfrom the water surface.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 23

2.3.7.1 Sources of heat gain/loss for cloud

Ambient air Ground: forced convection Ground: free convection

Insolation Chemical reactions Phase changes [] Other

Chemical reactions option is only available for HF.

The HF model is essentially that contained within HGSYSTEM and is based on the modeldeveloped by Schotte. This is a commonly met approach and is generally thought to beappropriate.

The heat transfer to or from the substrate is the maximum of calculated free and forcedconvection heat fluxes.

Forced convection is calculated from the modified Reynolds analogy

St Cf= 2d iPr 23where St is the Stanton number

Cf is the skin friction coefficientPr is the Prandtl number

The dimensional form is

q c T uu u z mp cld a

==

Pr2

32

10*

max , b gc hThe denominator should use ucld but it must also be noted that *u should be *u under the cloudrather than that in the ambient flow.

Free convection is calculated from

Nu Ra= 0141

3.where Nu is the Nusselt number and Ra is the Rayleigh number. All heat transfer calculationsassume that the ground temperature is unchanged by the passage of the cloud over the surface.

It is worth noting that the symbol Dac is used for two similar but different properties; the binarydiffusivity of the component in air and the thermal diffusivity for use in the heat transfercorrelations.

2.3.7.2 Correlations for thermodynamic properties

Perfect gas law Antoine correlation Other

However most properties of the released material (component) are obtained directly from theDIPPR database. The specific enthalpies of dry air, water vapour, ice and liquid water areinterpolations from tabulated data.

2.3.8 Mass transfer mechanisms

Mass transfer mechanisms considered


UDM, Ver. 6.0 Version 1.0 (21/01/02) 24

2.3.8.1 Sources of mass loss from cloud

Dry deposition Wet deposition Other

2.3.8.2 Sources of mass loss from two-phase clouds

Rain-out (two-phase) Other

A major novel feature of UDM is the incorporation of a rain-out model, rain out being animportant phenomenon for two-phase releases.

The liquid component in the aerosol is considered to consist of spherical droplets, surrounded bya mixture of air and evaporated vapour.

The trajectory of a single drop (with a cloud averaged drop size varying with downstreamdistance) is followed, representing the path of the centre of a cloud of drops. Evaporation andcondensation are treated by either equilibrium or non-equilibrium thermodynamics. Rainout istaken to be centred at the point of grounding of this trajectory.

A quite separate model allows for subsequent evaporation of the rained-out portion of the releaseand this evaporated component is re-incorporated into the dispersion calculation.

The initial droplet diameter is taken to be the smaller of

a mechanical break up diameter based on a critical Weber number of 12.5

a flashing break up diameter for which a correlation between this diameter and the particleexpansion energy is developed and used. This has been based on the CCPS rainoutexperiments.

The trajectory of a droplet of this diameter is determined from horizontal and vertical momentumequations and the evaporation of the droplet.

The general approach is sound and appropriate.

The droplet position/speed should be considered a typical averaged (top hat) velocitycorresponding to the droplet of averaged size.

As the droplets evaporate, the specific volume vcld of the cloud will increase (since vapouroccupies more volume than liquid), and therefore the cloud size will be increased due toevaporation.

2.3.9 Chemical reactions

Chemical reactions considered


UDM, Ver. 6.0 Version 1.0 (21/01/02) 25

2.3.9.1 Substances considered

NH3 HF N2O4 UF6 Other

2.3.9.2 Effects of chemical reactions on cloud

Heat source/sink Species (composition) Other

2.3.9.3 Level of detail in modelling

[M] Multi-phase reactants [M] Chemical kinetics [M] Order of reactions Other

2.3.9.4 Radioactivity

Radioactive decay modelled

2.3.10 Transition to passive dispersion

2.3.10.1 Criteria for transition to passive dispersion

Richardson number small Density difference small

Rate of lateral spreading small Other

UDM acknowledges several aspects of transition to passive dispersion.

Some are relevant to the momentum jet behaviour.

The UDM theory manual provides four criteria for transition to passive behaviour (which ALLneed to be satisfied for transition to passive dispersion):

cloud speed close to wind speed

cloud density close to ambient density

passive entrainment contribution close to total entrainment and

for grounded heavy gas plumes the Richardson number *Ri being less than a critical value.

This is a sound and appropriate approach.

The criteria values selected are


UDM, Ver. 6.0 Version 1.0 (21/01/02) 26

ru = 0.1

r = 0.015rE = 0.3

*Ri = 15

where the first three are criteria based on proportional differences and the last is a criticalRichardson number.

In the Verification Manual 5.2.4 a relevant table lists the criterion defining transition, which isthe last of the criteria to be satisfied. In many cases the jump in the entrainment rate at transitionis quite substantial (up to a factor of 2).

For a grounded dense gas plume a criterion based on the Richardson number *Ri is used, and avalue of 15 is selected for transition. This Richardson number is equivalent to about a halving ofthe top entrainment velocity based on the UDM correlation. Using the DEGADIS correlation itreflects a reduction of about 1/3 in the top entrainment velocity.

The approach is sound though the value could be viewed as quite large, as acknowledged by theauthors. Further improvement of the transition to passive dispersion has been identified by theauthors as an important area for future developments.

Of particular interest is that in the example shown in the Verification Manual the Richardsonnumber is rarely the defining criterion, reflecting that the large value set for this criterion isfrequently being overridden by the other criteria.

2.3.10.2 Treatment of passive dispersion

Dispersion parameters (s) used Other

Dispersion parameter dependencies

Downwind distance Roughness length Atmospheric stability Other

The passive cross-wind dispersion coefficients are taken from McMullen (1975) - not a well-known reference.

The passive vertical dispersion coefficients are taken from Hosker (1973); an obscure conferencepaper.

However in the UDM Dispersion Verification Manual it is shown that the UDM dispersionformulas (base on the above papers) are very similar to the more widely-accepted TNOdispersion formulas (= power-law of HGSYSTEM).

While not criticizing these choices more recent review references might be more appropriate andreassuring.

2.3.11 Complex effects: aerosols

Aerosols considered


UDM, Ver. 6.0 Version 1.0 (21/01/02) 27

2.3.11.1 Type of model

Homogeneous equilibrium Explicit droplet Other

A choice of models is available. The homogenous equilibrium model is conventional andappropriate. The non-equilibrium model is conventional structure and typical of such models.

2.3.11.2 Effects incorporated in aerosol model

Mass transfer between phases Heat transfer between phases

Interaction with atmospheric water Other

2.3.11.3 Cloud variables affected by aerosol

Temperature Density Velocity Concentration Other

2.3.12 Complex effects: terrain

Terrain considered

2.3.12.1 Physical processes modified in formulation of terrain effects

Advection Entrainment Gravity spreading Other

2.3.12.2 Modification of ambient flow by terrain

Mean flow Turbulence Other

2.3.13 Complex effects: obstacles

Obstacles considered

2.3.13.1 Level of detail

Net effect Local details Other

2.3.13.2 Cloud variables affected by obstacles

Depth Width Velocity Concentration Other

Puff time-of-arrival


UDM, Ver. 6.0 Version 1.0 (21/01/02) 28

2.3.13.3 Physical processes affected by obstacles

Advection Entrainment Gravity spreading Other

2.3.13.4 Modification of ambient flow by obstacles

Modification

Mean flow Turbulence Other

2.4 Solution technique

2.4.1 Equation types

2.4.1.1 Type of main equations

Algebraic Ordinary Differential Partial Differential Other

Coupled equations

2.4.2 Analytical solution methods

Significant analytical solution methods used Methods developed as part of model Existing methods used Other

2.4.3 Numerical solution methods

Significant numerical solution methods used

Methods developed as part of model Existing methods used Other

Numerical integration is by a Runge-Kutta-Milne integration system, which steps forward indownwind distance (continuous releases) or time (instantaneous releases). The step size isgoverned by the accuracy with which the set of differential equations is solved. The user can setthe accuracy threshold value, by adjusting the accuracy for cloud integration parameter.According to this accuracy, the step size may be increased (doubled) or reduced (halved).


UDM, Ver. 6.0 Version 1.0 (21/01/02) 29

2.4.3.1 Computational grid

[?] Structured Unstructured Multi-block Other

Grid size/arrangement

See above.

2.4.3.2 Discretisation methods

[?] Spatial Explicit Implicit Other

[?] Temporal Explicit Implicit Other

See above.

2.4.3.3 Convergence

[?] Features to enhance convergence [?] Convergence criteria Other

Standard Runge-Kutta-Milne algorithm is adopted (see above).

2.4.3.4 Accuracy

Desired accuracy specified Other

2.5 Results or output available from model

2.5.1 Concentration-related output for steady situations

2.5.1.1 Plume centreline

Plume centreline trajectory

2.5.1.2 Pointwise concentration

Pointwise concentration distributions

Centreline Longitudinal Lateral Vertical Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 30

The PHAST6.0 graphical interface allows for a wide range of tabular and graphical output. Thisincludes centerline, longitudinal, lateral and vertical concentrations, as well as contour plots suchas cloud footprints and cloud side views.

2.5.1.3 Derived concentration data

Concentration information derived

Contours Flammable inventory

Lateral distance to given concentration Other

2.5.1.4 Concentration fluctuations

Concentration fluctuations

Averaging time PDF Other

An effect of averaging time on the output concentration is incorporated. The effect is feltthrough a multiplicative factor on the lateral plume dispersion for passive (non-dense gas)atmospheric dispersion. That is no effects of averaging time are felt for the dense gas dispersionphase. The multiplicative factor is of a conventional form

tav600

0 2FHG

IKJ

.

where the value of 600 reflects the averaging time used for the original dispersion correlations.

A further effect of averaging time occurs when the release is not a continuous steady release andthe averaging time is larger than the duration of steady concentration at a point in space.

UDM allows calculation of this for finite duration, steady releases and for the maximumconcentration on the centreline.

2.5.2 Concentration-related output for time-varying situations

2.5.2.1 Cloud position

Position of cloud

Centroid Time-of-arrival Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 31

2.5.2.2 Pointwise concentration

Pointwise concentration distributions

Centreline Longitudinal Lateral Vertical Other

2.5.2.3 Derived concentration data

Concentration information - derived

Contours Flammable inventory

Dose Toxic load Lateral distance to given dose

Concentration time history at a point Maximum concentrations Other

Locus of maximum concentration on centreline is provided.

2.5.2.4 Concentration fluctuations

Concentration fluctuations

Averaging time PDF Other

See 2.5.1.4.

2.5.3 Other information available

2.5.3.1 Temperature

Mean temperature Fluctuations

2.5.3.2 Further variables

Other variables available in output

Radius/width Height/depth Density Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 32

2.6 Sources of model uncertainty

2.6.1 Stochastic processesThe model contains no explicit consideration of concentration fluctuations and so does not addressthis aspect of uncertainty.

2.6.2 Modelling assumptionsThe principal source of model uncertainty will be the appropriateness of the model physicsassumptions. Many of them are based on limited experimental data reflecting the modeldevelopment period of the mid to late 1980s, e.g. the vertical entrainment relation, the buoyancy-driven lateral growth rate correlation.

The near field passive entrainment formulation is different to the far field formulation. This maylead to uncertainty.

The impact drag force (sec 2.3.2) arguments based on elastic collisions are at variance withfundamental laws of fluid mechanics.

The model does not acknowledge that entrainment coefficients for puffs are quite different to thosefor jets.

The model does not include shear dispersion for a time-varying release (as distinct from a finiteduration (continuous) release).

2.6.3 Numerical methodSince there is no information concerning the appropriateness of the numerical method used tointegrate the main equations, a definitive statement cannot be made on the uncertainty in thesolution associated with the numerical method. However, since the equations are ordinarydifferential equations, it is unlikely that significant uncertainty will arise.Note that in the UDM Verification Manual, exact analytical solutions are derived and comparedwith the numerical UDM results for a range of situations, e.g. passive dispersion, horizontal jets and2-D heavy gas dispersion, while for a large number of other cases UDM results are compared withcomparable HGSYSTEM runs. This does confirm the expectation that the equations are solvedaccurately. This has also been confirmed by varying the accuracy adopted for solving the numericalequations in the UDM.

2.6.4 Sensitivity to inputDetailed sensitivity studies have been carried out to examine the sensitivity to input of the model, aswell as to confirm the robustness of the model. The results of these sensitivity studies have not beenreported as part of the UDM Technical Reference Manual.

2.7 Limits of applicabilityThe Unified Dispersion Model (UDM) models the dispersion following a ground-level or elevatedtwo-phase unpressurised or pressurised release. It allows for continuous, instantaneous, constantfinite-duration and general time-varying releases. It includes a unified model for jet, heavy andpassive two-phase dispersion including possible droplet rainout, pool spreading and re-evaporation.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 33

It calculates the phase distribution and cloud temperature using either a non-equilibriumthermodynamics model, a non-reactive equilibrium model or an equilibrium model specific for HF(including effects of polymerization). The release direction is in the vertical plane of the winddirection (downwind releases). The dispersion takes place over flat terrain with a uniform surfaceroughness.

2.7.1 Sources

2.7.1.1 Primary origin Release models are required to provide the input to the dense gas dispersion model.

A momentum jet model is linked directly to the dense gas dispersion model.The jet model allows single and two-phase. The two-phase model includes an explicit dropletrain out model.An evaporating pool model PVAP is also available.

2.7.1.2 Release type The external expansion to ambient pressure for high-pressure releases is not part of the UDM

model. These calculations are carried out in PHAST immediately prior to the UDM model. For a low momentum release problem the dense gas model assumes that the initial plume

width is the same as the source width. There is no allowance for a near source vapour blanketas included in HGSYSTEM, DEGADIS, GASTAR. This will also be a significant omissionfor evaporating cryogens.

The jet model is only suitable for jet releases that are in the vertical plane in the winddirection and through the source position.

Single sources only are modelled.

2.7.1.3 Thermodynamic properties The jet model is an incompressible jet model and for gas releases there may be confusion

between temperatures and stagnation temperature.

2.7.1.4 Chemical properties No mixtures or multi-components, unless expressed as an effective pure (single component)

substance No chemically reactive substances (unless acceptable to ignore chemical reactions) other than

HF No radioactive substances (unless acceptable to ignore radioactivity)

2.7.2 Environments

2.7.2.1 Atmosphere For continuous releases a non-zero wind speed is required. No special treatment is applied for

low wind speeds, and as a result the model may be less accurate for these conditions.

2.7.2.2 Terrain Terrain features or sloping ground cannot be modelled. No advice is given as to when terrain features might be ignored.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 34

2.7.2.3 Obstacles No significant obstacles must be present, i.e. features which vary on a length-scale short

compared with the variation in the cloud, such as buildings and fences No advice is given as to when obstacles might or might not be significant.

2.7.3 Targets/output

2.7.3.1 General situations No information on concentration fluctuations is available from the model

2.7.3.2 Steady situations

2.7.3.3 Time-dependent situations

2.8 Special features

2.8.1 CapabilitiesThe inclusion of a continuous change of concentration profile from passive to dense is novel.However no evidence is offered to support the approach adopted.

Although not part of this evaluation, the inclusion of a source model which allows a two-phase jetincluding droplet rainout is an important capability.

2.8.2 FormulationNone.

2.8.3 Mathematical aspectsNone.

2.9 Planned scientific developmentsThe following are considered to be major items for further work

Remove rather arbitrary passive transition zone, and possibly improve passive formulation. Neaten modular code, particularly for droplet modeling, link between pool and dispersion

model, and time-dependent releases (introduce along-wind diffusion). Assessment and possible improvement for pressurized instantaneous expansion, lift-off and

mixing layer logic. Multi-compound dispersion and solid components. More detailed investigation and validation of pool model PVAP. Improved flash calculations.

UDM, Ver. 6.0 Version 1.0 (21/01/02) 35

3. User-oriented aspects of model

3.1 User-oriented documentation and help

3.1.1 Written documentation

User Manual Other

User Manual not supplied for evaluation.

3.1.2 On-screen help and documentation

Context-sensitive help User Manual online Other

3.1.3 User support

Telephone support Training courses Other

3.2 Installation procedures

3.2.1 Medium

Diskette CD-ROM Tape Internet download Other

3.2.2 Procedure

Copy files manually Installation program/script Other

Installation by means of installation diskettes (installation program).

3.2.3 User-friendliness

Description of procedure Help available during installation

Category 3: User-oriented aspects of model

UDM, Ver. 6.0 Version 1.0 (21/01/02) 36

3.3 Description of the user interface

3.3.1 General properties

Interactive Batch Other

Graphical Textual Other

3.3.2 Provision of input

Edit files directly Guided input Enter data on graphical forms Other

Edit existing files Edit default input Other

3.3.3 Information when model running

Numerical values Error/warning messages Status of calculation Other

3.3.4 Examining output

Graphical display of output

Integral graphical display facilities Separate graphical display program Other

Select data with graphical forms Guided input Other

Examining numerical values

Integral numerical display facilities Separate numerical display facility Edit output files Other

3.4 Internal databases Internal databases available


UDM, Ver. 6.0 Version 1.0 (21/01/02) 37

3.4.1 Databases available

Material properties Scenarios Other

3.4.2 Access by user

Access from model Access outside model Access by model itself Other

3.4.3 Modification

General users Specific users No users Other

Expert users can modify model parameters (defining, for example, entrainment coefficients) inaddition to usual input data. Also material properties may be modified in PHAST. In some casesthese changes can only be applied by specific users (i.e. PHAST6.0 program administrators).

3.5 Guidance in selecting model options

3.5.1 Main choices required

Primary origin details Source configuration

Substance released Properties of substance released

Atmospheric conditions Terrain Obstacles

Output required Other

3.5.2 Guidance in choices availableSources

User Manual Within interface Other

Type Pre-set lists of values Defaults Other


UDM, Ver. 6.0 Version 1.0 (21/01/02) 38

3.6 Assistance in the inputting of data

3.6.1 Facilities available

Checks Valid range Valid type Entry has been made Other

Filing system Choose input file location Choose input file name

3.6.2 Comments

3.7 Error messages and checks on use of model beyond its scope

3.7.1 Facilities to trap inappropriate use

Facilities available

Checks on intermediate results Other

Warning messages given Other action taken

3.7.2 Error/warning Messages

Occurrence During input During model run During output examination Other

Type Self-explanatory Look up online Look up in documentation Other

3.7.3 Comments


UDM, Ver. 6.0 Version 1.0 (21/01/02) 39

3.8 Computational costs

3.8.1 Execution times for specified problems

Seconds Minutes Hours Days

3.8.2 Comments

3.9 Clarity and flexibility of output results

3.9.1 Summary of model output

The output is in the form of cloud parameters e.g. centreline concentration, ground levelconcentration, plume height, plume depth, plume width, vapour temperature, liquid temperature,cloud density.

Output is provided at a base averaging time (18.75 seconds; no time averaging) and at anadditional range of averaging times (toxic and/or flammable averaging time, user-specifiedaveraging time). Output is usually given through Crystal Reports exportable to Microsoft Word,Excel, etc.

Graphs that may be produced include cloud footprints, cloud cross-section, cloud side view, andthe concentration at a given location.

3.9.2 Post-processing facilities

Graphical formats Line (x-y) plots Contour plots Vector plots Display on GIS Animated display Virtual reality Other

Numerical formats Tabulated output User interrogation Other

Hard copy facilities available directly from program Numerical output Graphical output Other

Additional software Necessary Optional Not required

Required for Line (x-y) plots Contour plots Vector plots Display on GIS Animated display Virtual reality All Other

Direct incorporation of directives/guidelines


UDM, Ver. 6.0 Version 1.0 (21/01/02) 40

3.9.3 CommentsOutput result capability appears to be adequate.

3.10 Suitability to users and usage

3.10.1 Suitability with respect to type of user

Background Engineer Consultant Regulator Academic Other

Type of experience Dispersion Fluid dynamics Thermodynamics Numerical methods Statistics Programming Consequence modelling Risk analysis

Length of experience Hours Days Weeks Months Years

Frequency of use

Occasional users Constant users

3.10.2 Suitability with respect to type of usage

Intensity Specific incidents Risk assessment Other

Range of problems Narrow Medium Wide

Integration with other models Standalone Knowledge base Other

Aspects of the jet model are strongly linked with the dense gas dispersion model.


UDM, Ver. 6.0 Version 1.0 (21/01/02) 41

3.11 Possible improvements

Written documentation On-line help and documentation Installation

User interface Selection/entry of input data

Information and checks while running

Output results

3.12 Planned user-oriented developments Real-time emergency response version of PHAST

3D output representation of dispersion plumes with and without certain pathway obstacles

Links to popular GIS programs (including Intergraph and GIS products)

UDM, Ver. 6.0 Version 1.0 (21/01/02) 42

4. Verification performed

4.1 Summary of verificationThe model authors write

Each of the modules in the UDM has been investigated and verified in detail in conjunctionwith a literature review and a sensitivity analysis. The modules have been corrected wherenecessary, validated where possible, and been compared with similar third-party softwareapplications. This has been carried out in extensive detail for the entire basic continuous model[phases of dispersion (passive, jet, heavy), equilibrium thermodynamics without rainout] and tosome extent also the instantaneous model [far-field passive dispersion, ground-level heavy-gasdispersion]. Following this work the UDM comparison against large-scale experiments improvedconsiderably, despite the elimination of tuning coefficients.

A detailed assessment and limited corrections have been carried out for the transition to passive,finite-duration releases, HF thermodynamics, and heat/water transfer from the substrate.

A brief assessment has been carried out for droplet modeling, pool spreading/evaporation, thelink between pool and dispersion model, pressurised instantaneous expansion, lift-off andmixing-layer logic, and time-dependent releases. A more detailed assessment is to be carried outfor these areas.

4.1.1 Parts of implementation verified

The verification manual for the dispersion model provides separate chapters for

passive dispersion

jet dispersion


transitions

unpressurised instantaneous dispersion

finite-duration releases

link between dispersion and pool modeland for the thermodynamics model

non-reactive equilibrium model: heat/water transfer from substrate

equilibrium model for HF

droplet model

Category 4: Verification performed

UDM, Ver. 6.0 Version 1.0 (21/01/02) 43

4.1.2 Verification undertaken

For the dispersion model the only verifications against analytical solutions practically possibleare those for passive dispersion and for a horizontal release; these were undertaken.

In other cases comparisons were made with relevant published correlations (though, of course,this includes some element of validation).

Much use is also made of comparisons with the results from other well-established models withsimilar parameter settings e.g. HGSYSTEM. The latter point is particularly the case for thethermodynamics model where several aspects are similar to HGSYSTEM. For the droplet modelcomparisons are made of the model predictions with calculations based directly on thecorrelations used.

In general terms the authors have gone to considerable lengths, using whatever means andtechniques available, to verify the coding of their model.

Additionally and simultaneously they have undertaken extensive sensitivity studies in order toensure that output trends are consistent with expectations.

4.1.3 Quality assurance

The authors state

The software development process at RMS is being run to ISO 9000 Tickit standard. Thequality assurance of the software development is carried out by means of a Quality ManagementSystem; documentation available on request.

4.2 Comments

UDM, Ver. 6.0 Version 1.0 (21/01/02) 44

5. Validation performed

5.1 Validation already performedAn extensive validation document has been assembled for UDM. The document includes adescription of each validation experiment, the details of the assumptions made for the UDMsimulation plus a detailed discussion of the results obtained from a statistical and graphicalcomparison against the field data.

The validation comparisons are made exclusively with field experiments. The experimentsselected are wisely chosen to cover a wide spread of phenomena.

The input data were taken from Hanna et al. (1991), data sheets provided from the SMEDISproject and from McFarlane et. al. for the Goldfish experiments.

It is not clear what is implied by PEAK centre-line concentrations; it appears to be just thecentre-line concentration for the averaging time of the experiments. There is an important pointhere. It is not specifically stated in the theory or validation document what the model-predictedconcentration is actually intended to be; and to what it should be compared when undertakingvalidation. For example is the model predicting a PEAK value or an ensemble value? There aremany similar questions which are rarely asked, let alone answered prior to model validation.

Comparisons are made with peak centreline concentration and cloud widths.

Two of the validation experiments (Maplin Sands and Burro) involve dispersion from anevaporating pool. The UDM does not allow for user-defined source conditions for dispersionfrom a pool. These releases were modelled as a ground level jet with immediate rainout of theliquid. The pool spread and vaporisation model provided the source conditions as a series ofconstant segments.

5.1.1 Validation exercises

Validation comparisons were made with:

Instantaneous releases Thorney Island

Continuous releases Prairie Grass 7-9, 13, 15, 17, 34, 41, 50, 58 Desert Tortoise 1, 2, 3, 4 EEC 360, 500 FLADIS 9, 16, 24 Goldfish 1, 2, 3 Burro 7, 9 Maplin Sands LNG 27, 29, 34, 35

Other releases

Category 5: Validation performed

UDM, Ver. 6.0 Version 1.0 (21/01/02) 45

5.1.2 Conclusions

Statistical comparison measures are presented for each experiment; a mean and a variance. Theresults range from reasonable to good; typical of the better dispersion models.

Considering the concentration predictions, and just taking an average value of mean and variancefor each experiment, and then averaging all the experiments together (that is weighting eachexperiment equally) produces an overall mean of 1.3 and an overall variance of 2.6. Alternativeways of grouping performance values are obviously possible.

In the broadest sense the results could be interpreted by stating that the majority of individualpredictions are likely to fall within a factor of 2 of experimental results.

The authors critically interrogate the results to locate aspects of the model which might beimproved.

UDM, Ver. 6.0 Version 1.0 (21/01/02) 46

6. Conclusions

General model descriptionThe Unified Dispersion Model (UDM) is an integral model to calculate the dispersion following atwo-phase pressurised release or an unpressurised release. It effectively consists of the followinglinked modules:

near-field jet dispersion

non-equilibrium droplet evaporation and rainout, touchdown

pool spread and vaporisation


far field passive dispersion.In addition to the non-equilibrium droplet thermodynamics model, UDM also allows for a two-phase HF thermodynamics model (including effects of polymerisation). This evaluation documentdoes not address the module for pool spread and vaporisation.The UDM allows for continuous instantaneous, constant finite-duration and general time-varyingreleases. The UDM allows for possible plume lift-off when a grounded plume becomes buoyant.

It is not capable of treating dispersion over complex terrain/slopes or dispersion near buildings orobstacles.

The latest version of the UDM is implemented in the consequence-analysis package PHAST,version 6.0. It is planned to be included in the onshore risk-analysis package SAFETI, and in theoffshore risk-package NEPTUNE (successor to OHRAT).

The model runs on a PC and requires little in the way of memory or disk space. It is a self-containedmodel requiring only a short time for a knowledgeable user to set up and run the model.

Scientific basis of modelUDM is a typical one-dimensional integral dense gas dispersion model, with respect to its maincapabilities and model physics. The model is interfaced directly with release models for amomentum jet or an evaporating pool.

For the dense gas dispersion model, releases may be instantaneous or continuous. Finite duration(constant release rate) releases are treated as continuous with either an analytic correction for thefinite duration or a (less preferred) abrupt transition to an instantaneous model. Time-varyingreleases are treated by combining segments of continuous releases. In this case no allowance forlongitudinal diffusion or shear dispersion is introduced.

Equilibrium thermodynamics and heat transfer physics are incorporated.

Chemical reactions are included for HF but for no other chemicals.

Category 6: Conclusions

UDM, Ver. 6.0 Version 1.0 (21/01/02) 47

The two-phase momentum jet model is novel for operational models in that it includes both avapour jet and a droplet trajectory model that is, the important problem of liquid rainout is directlyaddressed in the model.

The atmospheric environment is characterized in the typical integral model fashion. Like mostsimilar models the atmospheric environment is assumed stationary for the time of the release thatis, it is essentially a short range/short time scale model.

Concentration profiles are based on assumed similarity shapes. A novel feature is the incorporationof consistent similarity shapes throughout the evolution of the release. There is no confirmationthat the assumed similarity shapes are correct but the approach ensures that solution discontinuitiesdo not occur.

Cloud advection speed, spreading and dilution are treated in a standard way. Concentrationfluctuations are not considered. The transition from dense gas behaviour to passive behaviouroccurs at a Richardson number *Ri of 15, which is substantially larger (a factor of 10) than used inother similar operational models.

The passive dispersion parameterisation is from what might be regarded as an obscure source, but isshown to be very close to more commonly-accepted sources.

The governing equations are coupled ordinary differential equations, which are straightforward tosolve and lead to short run times.

Limits of applicability

Source: Only single component releases may be treated, though these can be two-phase.Environment: Low wind speeds cannot be treated. The atmospheric conditions are taken to be

steady which will limit the downwind extent that can be modeled. No terrain orslope can be modelled. The effect of buildings and obstacles on flow anddispersion cannot be modelled. For the momentum jet release the direction of thejet must be in a vertical plane in the direction of the wind passing through thesource. Many other source orientations are likely and these may lead to largerdownwind concentrations. Results from upwind directed jets will be suspect.

Targets/output: No particular limitations.

User-oriented aspects of modelNo specific information on user-friendliness is available however as a commercially viable modelthis will be an important consideration of the vendor.

The technical documentation is extensive and comprehensive. The suppliers have gone to greatlengths to document their model well. And provide an audit trail for changes and modification.They have been refreshingly frank in the technical documentation to point out the limitations of themodel, areas where there is uncertainty, and areas they are in the process of improving. Thedocumentation indicated that attention has been paid to providing appropriate warnings and help tothe user.

Categor

udm_mer_v1.0_tcm4-76493

Documents

route of model

model physics

model type

history of model

general model description

short description of

current model usage

scientific basis of