Mpact Theory - Desbloqueado PDF

DNV Software
MPACT Theory
December 2010
The MPACT model calculates the impact of the release of a toxic or flammable chemical
on the population. It takes the results of the consequence calculations of the toxic and
flammable effects, together with additional data on wind direction, ignition sources, levels
of overpressure-generating obstructions, event location and frequency and superimposes
them on the population to calculate the fatality risk in the surrounding area. The results are
presented in a variety of forms including F-N data for societal risk, individual risk
presented as grid over the calculation area, ranking tables for the contribution of each
event, overall rate of death and other summary societal risk measures as defined by
regulators in the Netherlands and UK. In addition to people risks the model can also be
used to integrate financial consequences and risks.
DNV SOFTWARE
Palace House, 3 Cathedral Street, London SE19DE, UK
http://www.dnv.com/software
Copyright Det Norske Veritas. All Rights Reserved.
No reproduction or broadcast of this material is permitted without the express written consent of DNV.
Contact [email protected] for more information.
- II -
MPACT Theory
ABSTRACT
The MPACT model calculates the impact of the release of a toxic or flammable chemical on the
population. It takes the results of the consequence calculations of the toxic and flammable effects,
together with additional data on wind direction, ignition sources, levels of overpressure-generating
obstructions, event location and frequency and superimposes them on the population to calculate
the fatality risk in the surrounding area. The results are presented in a variety of forms including FN data for societal risk, individual risk presented as grid over the calculation area, ranking tables
for the contribution of each event, overall rate of death and other summary societal risk measures
as defined by regulators in the Netherlands and UK. In addition to people risks the model can also
be used to integrate financial consequences and risks.
Extensive enhancements have been introduced in version 6.6 of Phast Risk. An alternative
approach to modelling explosions has been provided based on either the Multi Energy model or
Baker Strehlow Tang model. The vessel burst model is also included in the risk calculations.
Explosion effects together with frequency are reported enabling exceedence reporting.
The new explosion approach uses a three-dimensional view of the cloud so enables further
modelling enhancements. Ignition source height may be taken into account. There is also an
ignition source shut-down option.
Population may be located in buildings of a specific type. This enables the protective effects of
different types of building construction to be taken into account. The population categories are
now carried through the risk calculations so that it is possible to group the societal results by
population category.
The vulnerability of buildings may be represented by radiation and exposure time criteria rather
than using the existing radiation probit method. This makes the new modelling better applicable to
escalation risk analysis.
Issue Log
Revision
1
2
3
Prepared by
Emerson et al.
Reviewed by
-
David Worthington,
Henk Witlox
David Worthington
Henk Witlox
Issue Date
1982
2001
October
2002
September
2003
April 2004
Oct 2005
David Worthington
Adrian Holt
Henk Witlox
Henk Witlox,
RIVM
Henk Witlox
Nov 2005
David Worthington
Henk Witlox
Dec 2005
David Worthington
Henk Witlox
9
10
Mar 2007
June 2007
David Worthington
David Worthington
Fiona Clements
Henk Witlox
11
12
March 2010
Dec 2010
David Worthington
David Worthington
Yongfu Xu
Yongfu Xu
David Worthington
Comments
SAFETI 5.31 version
SAFETI 6.2 version
SAFETI 6.4 version
Flammable probit upgrade
Update to reflect SAFETI-NL
acceptance version
Update to reflect final SAFETINL version
Update again to reflect final
SAFETI-NL version
Minor Version 6.53 updates
Version 6.54 changes for NL
version
Version 6.6 changes
Version 6.7 changes
Copyright Det Norske Veritas. All Rights Reserved. No reproduction or broadcast of this material is permitted without the express written consent of DNV.
MPACT Theory
- III -
CONTENTS
1.
2.
Introduction.................................................................................................................. 1
1.1
History............................................................................................................................. 1
1.2
Purpose............................................................................................................................ 2
1.3
Layout of the Document ................................................................................................ 2
Model Overview............................................................................................................ 4
2.1
Model Input .................................................................................................................... 4
2.2
Individual Risk Calculations......................................................................................... 7
2.2.1
2.2.2
2.2.3
2.2.4
2.2.5
2.2.6
2.2.7
2.3
2.3.1
2.3.2
3.
Definition of Societal Risk ......................................................................................................14

Results for Societal Risk Calculations.....................................................................................16
3.1
Calculation Area .......................................................................................................... 19
3.2
Discretisation of the Calculation Area ....................................................................... 20
3.3
Weather Category and Wind Direction ..................................................................... 20
3.4
Population..................................................................................................................... 22
Introduction .............................................................................................................................22
Population Shapes Conversion to a Grid .................................................................................22
3.5
Risk Ranking Points..................................................................................................... 24
3.6
Conditional Probability Fatality Factors and Building Types................................. 24
3.6.1
3.6.2
3.6.3
3.6.4
3.7
Introduction .............................................................................................................................24
Outdoor and Indoor Methods...................................................................................................25
Indoor and Outdoor Risk Method Using Factors.....................................................................26
Indoor and Outdoor Risk Method Using Specific Indoor Populations ....................................27
Summary Maximum Effect Zones (SMEZ) Reporting ............................................ 27
Toxic Impact Calculations......................................................................................... 29

4.1
Calculation Grid........................................................................................................... 29
4.2
Toxic Effect Zone at the Grid Cell.............................................................................. 30
4.3
Wind Distribution ........................................................................................................ 33
4.4
Individual Risk Calculation ........................................................................................ 35
4.5
Societal Risk Calculation............................................................................................. 36
4.5.1
4.5.2
5.
Societal Risk Calculations ........................................................................................... 14
Calculations Common to Toxics and Flammables................................................... 19
3.4.1
3.4.2
4.
Definition of Individual Risk.....................................................................................................7

Calculation Area ........................................................................................................................8
Wind Distribution and Release Direction ..................................................................................9
Event Trees (toxic or flammable outcomes) ............................................................................10
Ignition Sources and Distribution ............................................................................................11
Materials that are Both Flammable and Toxic.........................................................................12
Results for Individual Risk Calculations .................................................................................12
Calculation of F-N Pairs ..........................................................................................................36

Probabilistic Distribution of F-N Pairs ....................................................................................39
Flammable Impact Calculations ............................................................................... 45

5.1
Introduction.................................................................................................................. 45
5.2
Event Trees ................................................................................................................... 45
MPACT Theory
5.2.1
5.2.2
5.2.3
5.2.4
5.3
- IV -
Flammable Outcomes ..............................................................................................................46

Selection of event tree route and event tree type .....................................................................46
Event Tree Types.....................................................................................................................49
Event Tree Probabilities ..........................................................................................................54
Wind Directional Probabilities ................................................................................... 63
5.3.1
Wind Sector Subdivision .........................................................................................................63
5.4
Flammable Effect Zone Types .................................................................................... 64
5.5
Fireball/BLEVE (Single Zone Method) ..................................................................... 66
5.6
Jet and Pool Fires (Single Zone Method)................................................................... 67
5.7
BLEVE, Jet and Pool Fires (Flammable Probit Method)......................................... 69
5.8
BLEVE, Jet and Pool Fires (Radiation Intensity Method) ....................................... 74
5.9
Vapour-Cloud Flash Fire - Instantaneous Release ................................................... 74
5.10
Vapour-Cloud Flash Fire Continuous, Finite Duration and Time-Varying
Releases ...................................................................................................................................... 76
5.11
Vapour-Cloud Explosion (Wiekema Purple Book method) ..................................... 84
5.12
Vapour-Cloud Explosion (3D Purple Book method)................................................. 86
5.13
Vapour Cloud Explosions (Confined Explosion Method Using 3D Cloud Views) 89
5.13.1
5.13.2
5.13.3
5.13.4
5.14
Overlap of Effect Zone with the MPACT Risk Grid ................................................ 94
5.15
Continuously Variable Pd Type of Effect Zone ......................................................... 96
5.16
Calculation of Individual and Societal Risk .............................................................. 97
5.16.1
5.16.2
5.16.3
5.16.4
5.17
5.18
Options available to specify the immediate ignition probability.......................................103

Purple Book Relative Flammability of Different Materials ..............................................103
Immediate Ignition for Stationary Vessels........................................................................105
Immediate Ignition for Transport Vessels.........................................................................107
K2+ materials Processed at Elevated Temperatures .........................................................107
Materials that are both Toxic and Flammable...................................................................108
Delayed Ignition ......................................................................................................... 109
5.19.1
5.19.2
5.19.3
5.20
Treatment of Non-Integer Values of N .............................................................................102

Risk Ranking Point Results...............................................................................................102
Immediate and Delayed Ignition Branch Probabilities ............................................ 103
5.18.1
5.18.2
5.18.3
5.18.4
5.18.5
5.18.6
5.19
Single Effect Outcome ........................................................................................................98

Combination of Two Effects in One Outcome..................................................................100
Explosion, flash fire with or without pool fire ..................................................................101
Direct Calculation of Indoor Risk .....................................................................................101
FN Curve..................................................................................................................... 102
5.17.1
5.17.2
Options for the delayed ignition branch of the Event Trees..............................................109

Development of the Cloud with Time...............................................................................109
Probability of Delayed Ignition.........................................................................................111
Ignition Source Modelling ......................................................................................... 112
5.20.1
5.20.2
5.20.3
5.20.4
6.
Introduction.........................................................................................................................89
Overview.............................................................................................................................89
Immediate Ignition Outcomes.............................................................................................91
Delayed Ignition Outcomes ................................................................................................92
Definition of Ignition Sources...........................................................................................112

Multiple Ignition Sources Grid Method .........................................................................116
Multiple Ignition Sources Direct Method for Height-Specific Sources .........................120
Free Field Methods ...........................................................................................................122
Verification and Sensitivity Analysis....................................................................... 124
MPACT Theory
7.
-V-
Future developments................................................................................................ 124

Appendix A. MPACT input Parameters - Event Tree Probabilities ............................................125
Appendix B. MPACT input Parameters Vulnerability Factors ................................................131
Appendix C. MPACT input Parameters Remaining factors.....................................................134
Appendix D. Errors, Warnings and Messages.............................................................................136
Appendix E. Outcome Codes ......................................................................................................143
NOMENCLATURE ............................................................................................................................150
REFERENCES ....................................................................................................................................156
LIST OF FIGURES
Figure 1 Overall Input And Output Data For The MPACT Model ................................... 6
Figure 2 - The MPACT Risk Calculation Grid..................................................................... 8
Figure 3 - Wind Directional Probabilities Shown As A Wind-Rose..................................... 9
Figure 4 - Generalised Event Tree Illustrating The Logic Applied In MPACT ................. 10
Figure 5 - An Example Of The Ignition Probability Function........................................... 11
Figure 6 Example of an Individual Risk Contour Plot .................................................... 13
Figure 7 Example of Individual Risk Ranking Point Results........................................... 14
Figure 8 Example of a FN Curve..................................................................................... 17
Figure 9 Example of Societal Risk Outcome Results...................................................... 18
Figure 10 Calculation of the Limits of the Risk Zone...................................................... 19
Figure 11 Wind Rose Showing Directional Probabilities Provided as Input to MPACT 21
Figure 12 - Population Object Types Imposed on the MPACT Grid ................................. 23
Figure 13 Lower and Upper Wind Directions to Affect a Point...................................... 29
Figure 14 - The Geometry of the Toxic Cloud.................................................................... 32
Figure 15 Definition Of Variables Used To Describe Each Wind Sector ....................... 33
Figure 16 Integration of Directional Probability, P|w and Probability of Death, Pcl..... 35
Figure 17 Variance Function for a Range of Pc|o, Pcl and Nx,y=10 ................................. 40
Figure 18 F-N distribution for Nx,y=10, Pc|o =0.5 and Pcl=1 .......................................... 42
Figure 19 Selection of event tree route and event tree type ............................................ 48
Figure 20 Event Tree Type B Continuous, Not Free Jet, No Rainout .......................... 50
Figure 21 Event Tree Type C Instantaneous, No Rainout ............................................ 51
Figure 22 Event Tree Type D Continuous with Rainout............................................... 52
Figure 23 Event Type E - Instantaneous with Rainout, No Vapourising Pool Left Behind
..................................................................................................................................... 53
Figure 24 Event Tree Type F Instantaneous with Rainout and Vapourising Pool Left
Behind ......................................................................................................................... 54
Figure 25 Wind Sector Subdivision Parameter = 2 ......................................................... 63
Figure 26 - The Fatal Effect-Zone for a Fireball or BLEVE.............................................. 66
Figure 27 - The Fatal Effect-Zone for a Jet Fire................................................................ 68
Figure 28 - The Fatal Effect-Zone for a Pool Fire............................................................. 68
Figure 29 (Figure 5.4 from the Purple Book) Calculation of the probability of death, PE,
where the respective fractions of the population dying indoors and outdoors are FE,in
and FE,out, for exposure to a BLEVE, pool fire and jet fire. The probit function for
heat radiation is f(Q,t). ............................................................................................... 69
Figure 30 Probability of Death as a Function of Radiation and Exposure Time (Based on
the Purple Book probit constants a= -36.38, b=2.56, n=4/3)...................................... 70
Figure 31 Representation of the Radiation Hazard Zone Using Flammable Probits....... 71
MPACT Theory
- VI -
Figure 32 (Figure 5.A.1 from the Purple Book) Example calculation of a BLEVE of 100
tonne propane (burst pressure 15 bar). The exposure time is set at the maximum
value of 20 s. Indicated are heat radiation, Q, probability of death, PE, fraction of
people dying outdoors, FE,out, and fraction of people dying indoors, FE,in. ................ 72
Figure 33 Comparison of different interpolation methods with lethality curve .............. 73
Figure 34 - The Flammable Zone for an Instantaneous Release........................................ 75
Figure 35 Example of a UDM Cloud represented by a Half-Ellipse .............................. 76
Figure 36 - Example of a UDM Cloud represented by a Full-Ellipse................................ 77
Figure 37 - The Truncation of a Short-Duration Release to Define the Flash Fire Effect
Zone............................................................................................................................. 79
Figure 38 Behaviour of Front and Back of the Cloud with Time .................................... 80
Figure 39 - The Cloud Shape for Multiple Release-Rate Segments at a Given Time......... 81
Figure 40 Flash Fire Zone for a Two Segment Cloud at a Given Time .......................... 82
Figure 41 - The Displaced Effect-Zone for an Elevated Release ....................................... 83
Figure 42 - The Effect-Zones for a Vapour Cloud Explosion............................................. 84
Figure 43 Use of Multi Energy Curve 10 to obtain radii R1 and R2............................... 87
Figure 44 Overview of a Vapour Cloud Explosion from an Obstructed Region............. 90
Figure 45 Main Stages to calculate vapour cloud explosions using the obstructed region
explosion modelling .................................................................................................... 92
Figure 46 - An Effect Ellipse Superimposed on the Calculation Grid ............................... 94
Figure 47 - Sub-Dividing the Square to Calculate Frx,y, o................................................. 95
Figure 48 Cell Corner Locations and Potential Subdivision ........................................... 97
Figure 49 - Ignition Probability Function Example ......................................................... 115
Figure 50 Illustration of the Interaction of a Cloud and Cells containing an Ignition
Source ....................................................................................................................... 117
Figure 51 Ignition probabaility vs time for a shut down time of 5s and cooling times of 1,
10, 20 and 60s ........................................................................................................... 122
LIST OF TABLES
Table 1 Event Tree Types ................................................................................................ 49
Table 2 Outcome Probability Po for each Type of Event Tree ........................................ 59
Table 3 Conditional probability of explosion based on obstructed cloud volume.............. 62
Table 4 Relevance of Wind speed and Stability to Flammable Effects........................... 65
Table 5 Treatment of the Explosion Combined with the Flash Fire and Pool Fire Effects
..................................................................................................................................... 88
Table 6 Fractions used for Calculation Squares with Partial Overlap ............................. 96
Table 7 - (Table 4.7 of Purple Book) Reactivity of a number of substances ([CPR14]).
If little or no information is available, substances are classified as high reactive.
These substances are indicated with an *................................................................. 104
Table 8 Classes of flammable materials ........................................................................... 104
Table 9 - Probability of direct ignition for stationary installations................................... 105
Table 10 - Probability of direct ignition for transport units in establishment................... 107
Table 11 - K2+ materials above flame point temperature ................................................ 107
MPACT Theory
1. Introduction
1.1
History
The risk-analysis program SAFETI was conceived during the study of the chemical
industry in the Rijnmond area of Holland between 1976 and 1979. This was a study
invoked by a public enquiry into concerns raised by the public living in this area.
One of the conclusions was that it was worthwhile to pursue the issue of streamlining the
risk methodology developed using automated techniques. Two directions for development
were identified: the Parametric Correlation Method, in which individual risk levels and
effect distances would be coupled to basic properties of an installation by correlation
diagrams and a more sophisticated approach the Simplified Classical Method, which was
a risk analysis using standardised failure events: Equivalent Discrete Failures.1
SAFETI was commissioned by the Dutch government as the tool to automate the
Simplified Classical Method. The project was awarded after a competitive tender to
Technica Ltd in 1981.
There was strong demand from industry for this automated tool because it provided for the
first time a framework for the risk analysis and the productivity benefits of a computerised
system. The decision was made between Technica and the Dutch government to make the
tool available under licence and it has been available from Technica (now part of the DNV
organisation) since then. It is now used around the globe to analyse toxic and flammable
risks both by regulators and industry itself.
Over the years SAFETI has been developed extensively. Initially the techniques employed
were developed in a close relationship with the Dutch Government and many of the
features were designed to meet the needs of the legislation in that country. As other
organisations took licences for the software their requirements for further development
have driven the product forward through the SAFETI User Group. Some organisations
have provided direct funding for specific developments and all users contribute to
development through annual maintenance payments. The investment in the tool over the
years has been considerable. In recent years the software has been renamed Phast Risk and
the SAFETI name has been used to refer to DNVs range of risk software.
There are many facets to Phast Risk because it draws on a wide range of input data types.
Dealing with Event Tree input is quite different to assembling the necessary population
data. However, there is a common core to the product and this is the risk integration
model, MPACT. Within MPACT the different causes of fatality are transformed into a
common format and brought together to provide a total picture of the risk. MPACT is and
has been the centre of Phast Risk since the beginning.
In recent years the MPACT model has been improved incrementally. Some changes are
general, for instance more powerful and flexible grid sizing. Most recently changes
support the published QRA guidance in the Dutch Purple Book9 and the BEVI law. These
included event tree modifications and use of the flammable probit method to calculate
radiation hazards.
MPACT Theory
Extensive enhancements have been introduced in version 6.6 of Phast Risk. An alternative
approach to modelling explosions has been provided based on either the Multi Energy model or
Baker Strehlow Tang model. The vessel burst model is also included in the risk calculations.
Explosion effects together with frequency are reported enabling exceedence reporting.
The new explosion approach uses a three-dimensional view of the cloud so enables further
modelling enhancements. Ignition source height may be taken into account. There is also an
ignition source shut-down option.
Population may be located in buildings of a specific type. This enables the protective effects of
different types of building construction to be taken into account. The population categories are
now carried through the risk calculations so that it is possible to group the societal results by
population category.
The vulnerability of buildings may be represented by radiation and exposure time criteria rather
than using the existing radiation probit method. This makes the new modelling better applicable to
escalation risk analysis.
1.2
Purpose
This theory document is intended to provide the reader with a complete description of the
MPACT model.
This description is provided to help the user understand
what final results are produced by the model

what MPACT does to calculate the final results
how MPACT uses the various forms of input data it requires
how to interpret the results according to the way the model works
The intention is also to develop a baseline for improvements to the model. Hopefully users
who find this description useful will be able to make better overall use of the model and
contribute ideas towards its future improvement.
1.3
Layout of the Document
This document is arranged in a logical order, starting with an overview of the model in
Chapter 2 which is a useful introduction to the general approach taken by MPACT. This
could be read on its own and/or used to present the theory in a summary fashion.
The subsequent Chapters 3,4,5 include a detailed description of the MPACT theory. In
Chapter 3 calculations common to toxics and flammables are described. In Chapter 4
calculations specific to toxic impact calculations are described. In Chapter 5 calculations
specific to flammable impact calculations are described. The use of event trees by MPACT
is explained in Chapter 5.2. This is of particular importance for flammable effects where
the method chosen depends on the categorisation of the event by the prior consequence
modelling.
MPACT Theory
Verification work and future development needs are addressed in other documents.
Chapter 6 and Chapter 7 are placeholders for any cross references.
MPACT Theory
2. Model Overview
2.1
Model Input
MPACT is a mathematical model that calculates a number of results relating to the impact on
the land and population potentially affected by the various toxic and/or flammable eventsi
that are defined as inputs by the user.
MPACT employs the assumption that each event can be treated independently from other
events. This assumption can be made because the frequency of accidental releases in the
chemical industry is mathematically very low in comparison with the duration of the
hazardous effects themselves. The model does not model escalationii of effects to other
vessels automatically but the risk ranking point feature can be used to predict the risk of
reaching threshold effect levels that could lead to escalation events. The user may then enter
these events in order to include them in the risk.
The integration of the risk from all possible events can, therefore, be built up event by event
and this is one of the principles underlying the MPACT algorithm.
While each event can be treated separately this does not necessarily mean a discrete
outcome. A given release event can give rise to different outcomes. Releases of flammable
materials in particular can give rise to a range of hazardous phenomena depending on a
number of factors. The time of ignition, the level of confinement, the behaviour of any liquid
in the release can all influence the range of possible outcomes.
In Phast Risk the consequence modelling generates the relevant consequences effects for a
given event. These therefore form one of the inputs required by MPACT.
Since MPACT is concerned with risk it also needs frequencies for each possible event and
outcome. These are calculated through the use of event trees. Within MPACT the frequency
of each outcome is calculated from the frequency of the event itself and the probabilities
defined on each branch of the event tree. The event tree probabilities and event frequencies
therefore form another important input to MPACT.
For the flammable events the timing of the ignition affects the probabilities of the delayed
ignition outcomes on the event trees. MPACT includes an ignition model to predict these
probabilities. This model depends on the definition of ignition sources so these again form
another important input to MPACT.
For some outcomes the wind direction, speed and possibly stability class are important
factors. The wind directional probability needs to be taken into account so that the risk is
correctly distributed within the region at risk from the collection of events. Wind directional
probabilities by speed and stability therefore form another important input.
Finally, to enable fine adjustment of the risk calculations various model input parameters and
probability of death modifiers are also defined as input to MPACT. In addition these
i
ii
The term Events and Equivalent Discrete Failure (EDF) are used here interchangeably.
Also referred to as knock-on or domino effects
MPACT Theory
parameters can be used to represent damage to different types of object so plant equipment
damage may be taken into account. This facility allows the model to be applied to the
calculation of financial risk.
The input data requirements for the model are therefore extensive and to summarise include;
Consequence results for each of the events to be included in the analysis

Frequency and location data for each of the events to be included in the analysis
Event tree probability values for each of the scenarios considered by the model
Probability of death values for each of the event tree branches for flammable events
Weather data indicating the directional probabilities for each of the wind
speed/stability combinations used for the dispersion modelling
Various other model-specific parameters
Population distribution data
Ignition source distribution data
Recent extensions require additional input data

Obstructed region information to determine explosion effects using the Multi
Energy21 or Baker Strehlow Tang models
Vulnerability information for different building types
Population in specific categories and specific buildings
Because of these extensive data requirements the model has formed the core of the Phast Risk
product (specifically designed to provide this data) for over 20 years and has not been
developed outside this context. Figure 1 illustrates how MPACT forms the central point of
bringing together the various forms of input data and produces all the risk results. Those in
red are more recent extensions.
MPACT Theory
Figure 1 Overall Input And Output Data For The MPACT Model
MPACT Theory
2.2
Individual Risk Calculations
As already stated, MPACT analyses the risk one event at a time. It can be used in this mode
to process all the risk results for a single event only. In practice it is usually run in a batch
mode where it processes and sums the risk for a range of events provided as input data. In
this way it can build up a composite risk picture for a range of possible outcomes. It does this
internally by reducing results, whatever the outcome type, into two common categories,
Individual Risk and Societal Risk.
2.2.1 Definition of Individual Risk
Individual risk is viewed as the risk of someone at a specific location being a fatality in a
given period. It is usually quoted on a per year basis. It assumes that there is a person at this
point continuously exposed to the risk. This is sometimes referred to as location specific
individual risk rather than person specific individual risk that would depend on the
movements of a given individual.
Let us consider the individual risk IRx,y|w of one specific failure event (Equivalent Discrete
Failure EDF) at the location x,y given that a specific weather w (windspeed/stability
combination) occurs. Here x,y are horizontal Cartesian co-ordinates.iii We can now write
Equation ( 1 ) for the individual risk IRx,y|w as a function of the frequency Fedf of the failure
occurring in the time period, the probability that the event occurs in a particular direction
(clockwise angle from North, radians) iv and the probability of death given that direction and
weather,
2
IR x , y|w = Fedf
[P
|w
Pd |w ]d
(1)
where
1
2
P|w
Pd|w
is the direction of the release

is the lower value of that impacts the calculation point
is the upper value of that impacts the calculation point
is the probability of the release occurring in that direction given the weather
is the probability of death given that release direction and weather
This is the contribution to the Individual Risk from one event given a specific weather. To
obtain the overall Individual Risk at a given point all possible events must be taken into
account. Notice also that at this stage we have deferred the discussion regarding the causes of
the directional probability distribution. We have also neglected any mention of time this is
implicit in the calculation of Pd|w .
iii
MPACT currently knows nothing about the height for the risk calculations. It accepts the consequence data
as input and adds the risk irrespective of the height of the horizontal plane used for the fatality calculations.
Using default parameter settings will adopt the ground-level height z=0 for the risk location.
iv
SAFETI currently always assumes that the wind direction and the release direction lie in the same vertical
plane for all but simple horizontal jet flames; see Section 2.2.3 for further details.
MPACT Theory
MPACT has the capability of summing the events as it processes. The overall individual risk
represented by an EDF is calculated by summation over all weathers according to Equation (
2 ) where Pw is the probability of a specific weather.
IRedf , x , y =
(2)
Pw IRedf , x , y|w
All weathers
Equation ( 3 ) expresses the Individual Risk as the sum of all events as represented by the full
set of EDFs as specified by the analyst.
IRTot , x , y =
(3)
IRedf , x , y
All edfs
Clearly the overall result is very dependent on the list of EDFs input to the model. For the
purpose of this theory document the list is assumed to be correct and the focus is on
presenting the correct solution to the equations above.
2.2.2 Calculation Area
So far we have written equations with little regard for the orientation of the event and the
location being considered for individual risk. In general the distribution of individual risk
over an area is the useful output and such results are best presented as risk contours. MPACT
facilitates this by defining a grid over which it will perform the risk integration. Conceptually
MPACT calculates the risk for the location at the centre of each of the grid cells defined by
this grid. A contouring algorithm may then be used to present these results as risk contours.
Xmax,Ymax
Individual
Risk
Calculation
Point N,N
EDF1
EDF2
EDF3
Individual
Risk
Calculation
Point 1,1 Xmin,Ymin
x
Figure 2 - The MPACT Risk Calculation Grid
MPACT Theory
Figure 2 illustrates the grid MPACT defines for the purposes of calculating individual risk.
The grid forms a sequence of squares and the risk is calculated at the centre of each of these
square grid cells. The initial location of each EDF can be defined anywhere within a cell (not
just at the cell centre) and this is illustrated by pointing out possible locations for 3 events in
the Figure.
2.2.3 Wind Distribution and Release Direction
These EDF locations represent only the initial location of the hazard. The location of the
hazardous zone(s) relative to this initial location will depend on the release direction and/or
the wind direction.
The wind direction is of particular importance for releases that drift with the wind and still
present significant hazards. Such events include clouds of very toxic materials or very large
releases of dense flammable materials. The impact of toxic clouds will normally be very
dependent on wind direction. Releases in general are, therefore, considered to travel in the
downwind direction whatever their initial orientation.
Other events such as jet fires may be more dependent on the direction of the exit hole rather
than the wind direction. In practice MPACT treats only some horizontal jet fires as though
they are independent of wind direction. When initial momentum is considered the dominant
physical phenomena the initial direction of horizontal jet fires is distributed uniformly
through 360.
N
% by
sector
12
10
8
6
4
2
S
Figure 3 - Wind Directional Probabilities Shown As A Wind-Rose
Figure 3 illustrates the wind directional probabilities that might occur if there is a prevailing
wind passing from South East to North West. Note the convention that a high probability of
MPACT Theory
10
wind from the South East means higher risks in the North West because any cloud will be
driven in that direction.
2.2.4 Event Trees (toxic or flammable outcomes)
MPACT uses event trees to model the possible outcomes from a given event and depends on
user input for the probabilities of each of the outcomes. The exception is for delayed ignition
outcomes where there is a model representing the probability of ignition in a given period.
MPACT takes each EDF in turn and processes it according to logic represented as a
generalised event tree in Figure 4.
The first three types of event scenario, toxic, Standalone BLEVE and Standalone Pool Fire
are treated as single, exclusive outcomes within MPACT. For these types of events there is
no need to consider the event trees further and the ignition model has no influence on the
calculations.
Multiple outcomes are possible for other scenarios. First the immediate ignition outcomes are
considered. The user defines the probabilities for these outcomes. There is an extensive range
of possible outcomes including the possibility of combined flammable outcomes (a flash fire
combined with a pool fire for instance). The full details are provided in Chapter 5.
If there is a residual probability that the release is not ignited immediately then MPACT
integrates the risk from delayed ignition events with time. The ignition model calculates the
probability of delayed ignition by superimposing the flammable zone of the cloud over the
ignition sources defined by the user. The event tree branches for delayed ignition represent
the possible outcomes and their conditional probability.
Single
Outcome
Events
Toxic Release
Standalone BLEVE
Standalone Pool Fire
Event
Immediate Ignition
Events
Explosion
Multiple
Outcomes
Possible
Processed through
Timesteps
Delayed
Ignition
Flash Fire
No Effect
No Ignition
Figure 4 - Generalised Event Tree Illustrating The Logic Applied In MPACT
MPACT Theory
11
2.2.5 Ignition Sources and Distribution

To complete the overview of the treatment of flammable events it is necessary to introduce
the ignition model that is used to calculate the probability of delayed ignition outcomes.
Ignition sources are defined by the user as points, straight lines, polylinesv, rectangular
areas or volumes and polygon areas. These can be located anywhere but in practice the
sources are assembled according to the definition of the risk grid as defined in Figure 2. In
effect each grid cell contains its own ignition source based on the parts of the input
ignition sources that are determined to be within the cell. The equivalent ignition source is
considered to be distributed uniformly within the square. The general form of the model is
defined in Equation ( 4 ).
Px , y ,t = f x , y 1 e
x , y , t t
(4)
where
is the probability of ignition in the grid cell x,y over the time interval t
is the presence factor for the grid cell (the proportion of time the source is present
and active)
is the ignition effectiveness factor for the grid cell
is the time interval in question
Px,y,t
fx,y
x,y,t
t
Figure 5 is a graphical representation showing the terms defined above for a presence factor
of 0.8, and a probability of ignition (when present) of 0.5 in 30 seconds.
Ignition
Probability
1.0
Presence factor, f = 0.8
0.8
Datum Point
50% Probability in 30s
=40% with f of 0.8
0.6
0.4
0.2
0.018
0
0
1s
10
20
30
40
50
60
Time (s)
Probability in 1s = 0.018
Ignition Potential, x, is probability with f of 1.0 => 0.0228
Figure 5 - An Example Of The Ignition Probability Function
MPACT models the cloud drifting downwind and as it does so it integrates the ignition
sources within the flammable zone giving the probability of ignition with time. For each
timestep there are two branches on the event tree ignition or no ignition. The no ignition
v
A polyline is defined as one continuous line joining a number of x,y points using straight line segments.
MPACT Theory
12
branch is progressed to the next timestep. There may be the possibility of the cloud failing to
ignite at all while it is in a flammable condition. This gives rise to a residual probability of
non-ignition. This is of particular importance for materials that are toxic as well as
flammable.
2.2.6 Materials that are Both Flammable and Toxic
In reality the probability of death for a given event could be the result of toxic or flammable
effects or a combination of the two depending on the properties of the materials being
released. Common examples of such flammable and toxic materials include ethylene oxide
and methylisocyanate.
In its present form MPACT cannot model directly events that could produce both toxic and
flammable outcomes. It is necessary for the user to divide the event into a flammable event
and another toxic event. First the flammable outcomes must be modelled. Then the report of
residual non-ignition probabilities can be used to define the frequency of a subsequent toxic
calculation. Since the probability of delayed ignition is calculated by MPACT it is not
possible in advance to specify the probability of non-ignition.vi
The combination of the toxic and flammable events represents the total risk presented by the
release event of a material with both these properties.
2.2.7 Results for Individual Risk Calculations
Risk Array (Contours)

For each weather w, MPACT produces the array of individual risk totalled for all the events
processed. This array represents the risk IRTot,x,y|w at x,y points at the centre of each cell as
defined in Figure 2.
This array of results is used in Phast Risk to plot contours of individual risk. It can do this for
outdoor risk and any building type defined by the user. For instance it can calculate
contours of individual fatality risk and contour of risk to particular effect levels, like radiation
intensity or explosion overpressure. It can also be used to sum the results of multiple MPACT
runs by using the run-row features of Phast Risk. Figure 6 provides an example of the risk
contour display provided by the product.
vi
IMPROVE. This manual procedure provides only an approximate solution. In the future this procedure
could be automated by SAFETI, i.e. MPACT carries out first flammable risk calculations including
probability of non-ignition, and subsequently (using probability of non-ignition) toxic risk calculations.
Reported could be total flammable risk (ignoring toxic effects) and toxic risk (given non-ignition). A further
improvement is to avoid the approximate approach by integrating the combined toxic and flammable risk
directly.
MPACT Theory
13
Figure 6 Example of an Individual Risk Contour Plot
Risk Ranking Tables

The contribution of each event to the risk at each point in the risk array is not available after
all the results have been summed. This is a limitation imposed by the prohibitive data storage
requirements that this would impose.
However, MPACT does provide this breakdown of event contributions at a limited number of
locations on the risk array as specified by the user. This allows the relative importance of the
different events to be compared. Figure 7 provides an example of the individual risk ranking
output provided by the product.
MPACT Theory
14
Risk Increment Results (all data in SI units except frequency/yr)

Grid Information
NX
NY
200
200
SE X
SE Y
Cell Width
Corner
Corner
-2.18E+03 -2.18E+03 2.20E+01
No
1
2
3
X
-3.18E+01
1.56E+01
-8.27E+01
Y
3.18E+01
3.18E+01
3.18E+01
IX Grid
98
100
96
IY Grid
101
101
101
Event No
15005
15005
15005
15006
15006
15006
15007
Weather
0
0
0
0
0
0
0
Direction
0
0
0
0
0
0
0
Square
3
1
2
3
1
2
2
Frequency
1.00E-06
1.00E-06
1.00E-06
2.50E-07
1.00E-06
1.00E-06
1.56E-07
Figure 7 Example of Individual Risk Ranking Point Results
At risk ranking points the model can also report explosion effects for each outcome. These
include frequency, overpressure, impulse and duration.
2.3
Societal Risk Calculations
2.3.1 Definition of Societal Risk

The discussion so far has covered the way the model calculates individual risk. The other risk
measures calculated by MPACT cover societal risk by including the number of people
affected by each outcome. To achieve this, MPACT must have the population distribution
data provided as input. As a recent extension the model also accepts population category
inputs and reports the societal results accordingly. This allows the analyst to understand the
distribution of risk amongst the exposed population.
Within Phast Risk population distribution can be defined as any number of points, rectangular
areas or polygon areas. This data is distributed onto the risk grid as defined in Figure 2 by
analysing the proportion of population of each defined population object that exists within
each grid cell. MPACT then stores a total number of people per cell Nx,y. . Population may
also be defined as part of building data and fatalities within these buildings are calculated in
addition to the population on the grid.
The risk calculations now have to determine the number of people killed by a particular event
and attach this to the associated frequency to form an F-N pair. Probability of death is
provided to MPACT as part of the consequence input so it only remains for MPACT to
integrate the probability of death for each event over the population specified.
MPACT Theory
15
Nedf|o represents the number of people killed by a given event, particular weather
category/wind direction combination and if flammable the outcome type. It is calculated for
the population on the grid according to Equation ( 5 ).
N edf |o =
x, y
Pd , x , y|o dxdy
(5)
area
Note that in the calculation the population density per cell, nx,y , is used, which is set as the
ratio of Nx,y and the cell area. Conceptually this means that the people defined within a given
grid cell are uniformly spread across the area rather than located at the centre. This is
consistent with the approach adopted for ignition sources. Pd,x,y|o represents the probability of
death from the event and this will depend on consequence and parameters input to MPACT.
The calculation of Nedf|o can all take place within one MPACT grid cell or cover many cells.
In general the best results will be obtained with the smallest possible grid cell size. MPACT
will also include fatalities within specific buildings if these are included in the analysis.
Each event normally leads to a number of outcomes and then F-N results can be summed for
all event outcomes to form an overall picture of risk where F is calculated as the product of
the event frequency, weather category/wind direction probability, ignition probability and so
on. The F-N results can be presented as a curve of frequency of killing N or more people.
This format is often used for regulatory purposes.
Summary Measures
To further simplify the summary of the risk analysis other risk measures may be calculated.
These include the calculation of an overall average rate of death calculated as
Rd =
Fedf ,o N edf |o
(6)
All edfs o
This provides the expected number of fatalities on an annual basis. Sometimes this is known
as the Fatal Accident Rate or Expectation Value.
Other summary measures of risk take into account the greater societal aversion to incidents
with increasing values of N. One such measure is termed by the UKs HSE as the Risk
Integral2. This measure is referred to being appropriate for land use planning3.
RI LUP =

All edfs
2
Fedf ,o ( N edf |o +N edf
|o ) / 2
(7)
Earlier approaches raised N to the power of an aversion index (>1). Hirst and Carter2 refer
the range as being between 1.2 (published by Okrent4 ) and 2 (published by Schofield5). The
Risk integral based on this aversion index, ai, is defined in Equation ( 8 ).
RI ai =

All edfs
ai
Fedf ,o N edf
|o
(8)
MPACT Theory
16
So long as the aversion index is over 1 this provides a risk measure that reflects the greater
impact on society of incidents that cause high numbers of fatalities. The appropriate value to
use is likely dependant on the application of the risk analysis. Hirst and Carter6 recommend a
value of 1.4 for COMAH type application but with a higher value if the adjacent population
includes particularly sensitive components such as schools and hospitals.
Laheij, Post and Ale7 present further measures for land use planning. A Potential Loss of
Life (PLL) measure for the population exposed to the risks being modelled may be derived
from the individual risk grid and the population grid. This measure is based on the individual
risk results.
PLL I =
N x , y IRTot , x , y
(9)
All x , y
A further indicator of risk may be taken from the value of the cumulative probability of all
fatal incidents. This is a PLL measure based on the societal risk calculations. These last two
measures may differ because of different vulnerability assumptions applied for individual risk
and societal risk calculations. In this equation F(N) represents the cumulative frequency
within the range N-1 to N.
PLL S = F(N)
( 10 )
N =1
2.3.2 Results for Societal Risk Calculations

The societal risk results from MPACT are presented in the following forms;
Societal Risk Table (F-N curve)

As MPACT processes each event it produces it accumulates the frequency of events that give
rise to N values in pre-defined ranges. These results are available in tabular and graphical
form. A typical F-N curve is illustrated in Figure 8.
MPACT Theory
17
Figure 8 Example of a FN Curve
F-N Pair Ranking Table

MPACT also produces a table of F-N pairs for each event outcome for each weather
category/wind direction combination. This allows a ranking of events and a concentration on
these most significant contributions to the risk. An example is shown in Figure 9.
MPACT Theory
18
FN Outcome Results (all data in SI units except frequency/yr)

Grid Information
NX
NY
200
200
SE X Corner SE Y Corner Cell Width

-2.18E+03
-2.18E+03 2.20E+01
Event No Weather
Direction
Frequency
Sigma
15001
15002
15003
15004
15005
15006
15007
15008
15009
15010
15011
15011
15011
0
0
0
0
0
0
0
0
0
0
1
1
1
1.00E-06
1.00E-06
1.00E-06
1.00E-06
1.00E-06
1.00E-06
1.00E-06
1.00E-06
1.00E-06
1.00E-06
1.10E-11
7.32E-12
1.15E-11
3.92E+04
3.98E+04
3.92E+04
3.88E+04
3.86E+04
3.86E+04
3.88E+04
3.92E+04
3.98E+04
3.92E+04
1.62E+03
1.62E+03
1.62E+03
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0.00E+00
0
0
0
0
0
0
0
0
0
0
1
1
1
Figure 9 Example of Societal Risk Outcome Results
Rate of Death
MPACT calculates the overall rate of death given by Equation ( 6 ) as a measure of societal
risk.
Risk Integral Measures
MPACT also produces Risk Integral results as defined in Section 2.3.1as alternative
measures of societal risk.
MPACT Theory
19
3. Calculations Common to Toxics and Flammables

3.1
Calculation Area
One of the inputs to the MPACT model is the location of the calculation area to be covered
by the risk grid. Some calculations or manual settings must be made before the model is run.
If the parameter Automatic setting of Impact calculation grid is set to Yes then the
algorithm will loop through the list of EDFs in turn and calculate the maximum area to be
covered in the calculations as indicated in Figure 10. If this parameter is set to No then
MPACT will take the settings as defined in the parameters whether or not the effect zones lie
within the grid.
A circle is defined, centred on the release point and with radius the maximum effect distance,
as shown in the illustration for three hypothetical outcomes.
xmax,ymax
EDF1
y
EDF2
EDF3
xmin,ymin
x
Figure 10 Calculation of the Limits of the Risk Zone
From the calculation of the circle for each EDF and taking into account the different release
location of each EDF, the values of xmin, ymin, xmax, ymax that are needed to encompass all the
effects within the calculation area can be derived from the consequence results. In general the
zone will be rectangular although a square shape is possible.
The maximum effect distance will depend on a number of factors;
The various stop criteria set for the toxic effects
The minimum probability of death
The size of the delayed ignition effect zone (not yet calculated)
MPACT Theory
20
Because the grid sizing is done before the risk calculations it does not know exactly how the
risk calculations will be performed, particularly with respect to the explosions. Therefore it is
conservative and will often waste a lot of cells but wont make the grid too small.
For each effect it calculates the maximum downwind extent in each direction. It then uses 1.1
xs this extent to ensure the risk grid is larger than the effect zone. This may sound over
conservative but some effects are wider than their downwind extent.
For flammable clouds it does not have the effect distance for delayed explosions since these
are not calculated before the risk calculations. So instead of the downwind extent of the
flammable cloud it uses 2xs this value to represent the maximum possible extent of an
explosion. This extent is further increased by the safety factor of 1.1.
3.2
Discretisation of the Calculation Area
MPACT takes the calculation area limits as input to the model. These limits are treated in the
same way whether they are calculated or whether manually entered. The limits are adjusted
by MPACT to ensure the grid line coordinates are always a whole number of metres. It does
this from the South West corner of the grid so the higher values of x and y are extended
according to the rounded up value of the grid step.
The MPACT grid cells are always square.
The grid size may be set in two ways, by minimum cell size and/or maximum number of
cells. The options work as follows;
1) specify the minimum grid cell size as input to the model. It uses this step to calculate how
many cells are needed in each direction (Nx, Ny). If the total number exceeds the maximum
number of cells parameter then the model will increase the grid size to use this number of
cells.
2) always attempt to use the maximum number of cells and calculate the smallest possible
grid size with this number to cover the calculation area (the actual number of cells used may
be less than the maximum because of the rounding up of the cell size).
Note that there is a hard minimum limit of 1 metre and there is a trade off between
increasingly fine discretisation and computing resources (increasing calculation time,
memory and storage demands).
3.3
Weather Category and Wind Direction
Consequence Results
For each weather category (combination of wind speed and stability) one set of consequence
results are provided as input to MPACT. The assumption is that these consequence results
will be independent of the wind direction. In reality there are many factors that would make
the consequence results dependent on exact direction but these effects are not yet represented
MPACT Theory
21
by the consequence models (the building wake model is an exception) and wind directional
consequences would introduce additional computation time.
MPACT must therefore take the single set of consequence results for the weather category
and take into account the variability of wind direction. It does this using wind direction
probabilities for each weather category.
Wind Direction Probabilities
The wind directional probabilities for each weather category are input to MPACT as
probabilities per sector. The number of sectors is variable and there can be an offset from true
North. The sectors all have the same angular width. Figure 11 illustrates the form of the wind
probability data as provided as input to MPACT.
% by
sector
12
10
8
6
4
2
Figure 11 Wind Rose Showing Directional Probabilities Provided as Input to MPACT
In normal circumstances the wind rose is non-uniform. Indeed Figure 11 shows data where
this is a pronounced prevailing wind. Note the convention that the wind probability refers to
the direction from which the wind blows. This means the cloud will move in the opposite
direction. Hence the wind rose illustrated would give higher risks in the North West because
of the prevailing wind from the South East.
MPACT uses the wind directional probabilities in different ways for toxic and flammable
effects. The details are covered in Chapters 4 and 5.
The nomenclature used for the weather and directional probabilities is defined as
- Pw
Probability of the weather category, i.e. combination of range for windspeed u (m/s)
and stability class c (A-G); thus Pw = Puc
- Pws
Probability of the weather category and the wind blowing within a given sector s
MPACT Theory
- Pus
22
Probabilityvii of the wind speed u and the wind blowing within a given sector s; this
probability is used in the risk calculations for effects that are sensitive to wind speed
but dont vary with stability. It is set as the sum of Pws for all weather categories that
share the same wind speed, i.e.
Pus =
3.4
ws
All categories
( 11 )
Population
3.4.1 Introduction
In order to be able to calculate societal risk, MPACT must have information regarding the
number and location of people in the region of the analysis. A given accident may lead to
fatalities both indoors and outdoors and the model needs to be able to sum the number of
fatalities on the basis of the extent of the effect zone and the vulnerability of people to the
hazard.
The vulnerability of people to the hazardous effects depends on their location in terms of
whether they are indoors or outdoors and if indoors it also depends on various properties
of the buildingviii. Population may be input to MPACT directly as population objects or as
people within a specific building. If population objects are specified then they are used to
create population arrays corresponding to the MPACT calculation grid as described in
Section 3.4.2. People specified within building objects are not distributed over the grid and
instead they are considered in addition to the population distributed across the grid.
3.4.2 Population Shapes Conversion to a Grid
The population data is defined by users as area and point objects. These are converted into
2-dimensional grids matching the MPACT grid with one value representing the total
number of people present in each grid cell.
To understand how the program deals with the population it is necessary to conceive the
risk grid superimposed on the population objects defined. There are three types of grid
population object
Point
Rectangular
Polygon
And each of these objects can be approximated by imposing the shape onto the grid as
shown in Figure 12.
vii
This feature remains in the code but is not used in versions 6.* of SAFETI/Phast Risk
Also of importance is the population distribution in different time periods but this factor is assumed to be
handled outside MPACT itself so that MPACT models a single time period in a given run of the model.
viii
MPACT Theory
23
Grid Size
Point
Polygon
Rectangular
Figure 12 - Population Object Types Imposed on the MPACT Grid
For the point object the population will be placed in one grid cell only. The illustration
shows a situation where a point object is very close to a grid intersection. Depending on
the exact orientation of the grid the object could contribute to the population in one of 4
grid cells surrounding the point. A small change could flip the position to another cell. The
shaded area indicates all possible cells. This movement could change the F-N results
and/or affect all the results because of the associated ignition source. Notice also that this
population source is spread conceptually throughout the cell and loses its point definition
once it is positioned in the grid. The possible movement is up to 21/2 * grid size. It may
be a consideration to define the population over an area to avoid over-sensitivity.
Rectangular population blocks are imposed on the grid. Figure 12 shows that the block as
defined will be spread over a wider area than the block itself. Those cells at the edge will
contain a number of people proportional to the area covered. In this sense the edges of the
block are smeared because MPACT will consider the people in the cell to be located
anywhere within that cell. Small changes in alignment between the grid and the block
could lead to apparent movements of grid size in population distribution on any side of
the block. This effect is mitigated by calculation of the proportion of each grid cell
covered but still loses some definition of the original population shape.
Polygon objects are added to the population grid according to their population density and
the proportion of each grid cell covered by the polygon. The method used is similar to that
used for the calculation of the impact of flammable effect zones. This is explained in more
detail in Section 5.14. Note that the sub-division parameter used for flammables is used
also for this polygon calculation to retain the same level of precision throughout. Higher
values of this parameter will lead to greater precision but longer run times.
MPACT Theory
24
For all the shapes, objects that lie on a grid cell boundary are considered to belong to the
cell in the +ve x and/or y direction.
If the option to calculate risk indoors is selected then MPACT will calculate two
population grids; one for indoors and another for outdoors. It uses the proportion of people
indoors for each population object to work out the appropriate value for each grid cell,
indoors and outdoors. Since the indoor risk can correspond to only one building type it is
recommended that population within residential housing is treated in this way. Other
specific buildings may be accounted for by specifying the buildings directly. This method
is chosen because to have a grid per building type would demand excessive memory
usageix.
3.5
Risk Ranking Points
The x,y coordinates of each risk ranking point are used to generate a list of risk increments
at that point. These results are written to a file as detailed results and may be sorted and
ranked to analyse the risk breakdown at a particular location.
In addition explosion effects may be reported at these points. These results may be used to
generate exceedence results for explosion effects for use in building design or placement
analysis. The results reported include frequency, side-on overpressure, dynamic
overpressure (only for the Multi-energy model), impulse and duration. They also indicate
which obstructed region contributes most to the overpressure.
3.6
Conditional Probability Fatality Factors and Building Types
3.6.1 Introduction
When MPACT calculates risk it considers the probability of being affected by a given effect
zone (whether toxic or flammable) and then it applies a conditional probability of death. So
that given an accident there is a conditional probability of being impacted by the effect zone
and then there is the conditional probability becoming a fatality. In MPACT this latter
probability is composed of two elements 1) the raw probability of death calculated for the
particular hazard (i.e. probability of death without any protection); the probability of death
derived from the flammable probit correlation is an example of this; and 2) the vulnerability
of the population to the hazard; for instance the people may be protected by clothing so that
only a fraction will be sufficiently burnt to become fatalities. These vulnerability factors may
be specified by regulations (as in the Dutch Purple Book) and may differ depending on
whether individual risk is being calculated or societal risk.
In the model these vulnerability factors (in the range 0 to 1) are applied directly to the
probability of death Pd for a hazard and are set according to input parameters (See Appendix
A for the list) and building properties. The user can edit them so they do not affect the risk
ix
This limitation may be revisited as computers continue to improve
MPACT Theory
25
results (by ensuring the overall factor=1). They may also be used to turn an effect type off by
setting the value to 0.
With version 6.6 of Mpact multiple sets of the these parameters may be defined so that the
user may obtain contours and ranking points results to different effect levels within the same
study. These sets are referred to as Building Types. The implementation in Mpact is
dependent on the Vuln13 model which is used to convert radiation intensity and explosion
effects into vulnerability factors.
Users may also insert specific buildings within a building type and associate populations with
these buildings.
The first two building types have a fixed position in the analysis. The first type is always
considered to represent people outdoors. The second type represents people indoors in a
general sense. These first types allow for population objects to be distributed over the risk
grid and the method is suitable for very large numbers of population objects. Building types 3
and beyond model populations in specific buildings rather than on a grid.
3.6.2 Outdoor and Indoor Methods
These vulnerability factors depend on each type of hazardous effect and the model allows for
separate factors for individual and societal risk. The model also distinguishes between
vulnerability indoors and outdoors and it does this by;
1. applying global factors (one for individual risk and one for societal risk) to a single
population grid to represent an overall proportion of population present indoors
2. separating the population into two separate grids, one for outdoors and the other for
indoors, using the given fraction of people indoors for each population object. This allows
the outdoor and indoor toxic cloud results to be applied directly to the relevant population
3. allowing the population objects to be split individually between indoor and outdoor
population grids and applying the outdoor toxic cloud to both outdoor and indoor
populations using vulnerability parameters. This method is specifically added for
SAFETI NL 6.7.
4. calculating risk to people within specific buildings in addition to those on the population
grids
Method 1 is the simplest and is adequate when a single building type is appropriate and the
proportion of people indoors may be assumed to be uniform across the grid. This is the
original SAFETI NL method. Method 2 has the advantage of being able to apply a variation
of people indoors across the grid and calculates the indoor probability of death for toxic
releases taking into account the ventilation rate of the building type. This method is
appropriate when a single building type may be assumed and is particularly useful for the
general population in domestic housing. Method 3 is specifically for SAFETI NL 6.7.
Method 4 may be applied in conjunction with methods 1 and 2 and adds in the number of
people killed in a given outcome to the fatalities calculated from the population grid. An
application example might be a school specified separately from the general population.
MPACT Theory
26
3.6.3 Indoor and Outdoor Risk Method Using Factors

Conditional fatality factors are calculated separately for individual risk and societal risk
according to equations ( 12 ) and ( 13 ). By applying factors throughout the calculations it is
possible to represent the indoor/outdoor impacts without having to specify this attribute for
every population object. It is therefore relatively easy to define the input data and quicker to
calculate in MPACT.
K IR = f IROut K IROut + (1 f IROut ) K IRIn
( 12 )
Out
Out
Out
In
K FN = f FN
K FN
+ (1 f FN ) K FN
( 13 )
Where
K IR
overall factor to be applied to the calculation of individual risk for the type of effect,
K FN
overall factor to be applied in the societal risk calculation to the number of people
killed for the type of effect,
K Out
Fatality factor for the people outdoors for the type of effect, (can be specified
independently for societal and individual risk)
KIn
Fatality factor for the people indoors for the type of effect, (can be specified
independently for societal and individual risk)
Fraction of people outdoors for the purposes of the individual risk calculations
f IROut
(Method 1 parameter)
Out
Fraction
of people outdoors for the purposes of the societal risk calculations
f FN
(Method 1 parameter)
Thus for individual risk calculations the corrected probability of death KIR| Pd is applied,
while for societal risk the corrected probability of death KFN| Pd is applied.
Different fatality factors can be applied for the different types of effect, ;
Toxic effects (=T )
Pool fires (=P )
Jet fires (=J )
BLEVEs (=B )
Flash Fires (=F )
Explosion damage zone 1 (inner circle) (=R1 )
Explosion damage zone 1-2 (outer doughnut) (=R2 )
Intuitively it may appear inconsistent to introduce different factors for individual risk and
societal risk. However, it provides a convenient facility for the user who may wish to apply
different assumptions for individual and societal risks and obtain the results in one single run
of the MPACT model. Note that the same factors are always applied for a given effect type.
MPACT Theory
27
3.6.4 Indoor and Outdoor Risk Method Using Specific Indoor Populations
In this method each population object is divided into indoor and outdoor population grids and
the impact calculations are performed on each population separately. The factors f IROut and
Out
f FN
are not used in this method. The total societal impact is calculated by summing the
indoor and outdoor impacts for each outcome having applied the vulnerability factors directly
to the relevant section of the population; ie for the indoors grid;
K IR = K IRIn
( 14 )
= K IROut
( 15 )
And for the outdoors grid;
IR
I
We also define building specific vulnerability factors K IRI and K FN
for a given building
type I for use when a building object is defined directly.

3.7
Summary Maximum Effect Zones (SMEZ) Reporting
In SAFETI NL there is a report named Summary Maximum Effect Zones and referred to
here as the SMEZ spreadsheet report. This report gives a summary of the maximum
downwind distance from the release point for the different type of hazard for each failure
event. The distances reported are shown here as an extract from the report;
Toxic Results
Flammable Results
General
Largest
Distance
to 1%
lethality
Largest
Distance
to VRW
Largest
Distance
to AGW
Largest
Distance
to LBW
Probabili
ty of
direct
ignition
Largest
Distance
to LFL
Largest
Distance
to 1%
lethality
(m)
(m)
(m)
(m)
(fraction)
(m)
(m)
Radiation results
Correspondi
ng Event
(1% lethality)
Explosion Results
Largest
Distance
(m) to
Largest
Distance
(m) to
Largest
Distance
(m) to
Largest
Distance
(m) to
Largest
Distance
(m) to
35
kW/m2
10
kW/m2
3 kW/m2
0.3 bar
0.1 bar
The letters at the base row of the table are inserted here for use as reference (they arent in the
report).
In general this report is focussed towards distances actually used in the risk calculations. If
for instance a large flammable cloud is not ignited then the distance will not be reported.
However, there are exceptions and these are explained below;
Toxic Results
A - Largest Distance to 1% lethality
SAFETI NL uses outdoor toxic effect distances only. Also these effects are not subject to
event tree variations. Even for flammable and toxic materials if SAFETI NL is used
according to prescribed direct ignition probabilities (ie always <1) then these toxic results
distances will be applied in the risk calculations.
(m)
MPACT Theory
28
B, C, D Largest distances to VRW, AGW and LBW

These distance results to outdoor concentration levels are not used in the risk calculations.
Flammable Results
E - Probability of direct ignition

This is the probability of direct ignition used in the event trees.
F - Largest Distance to LFL
This distance depends on the immediate and delayed ignition outcomes. The delayed
outcomes will depend on the rules for modelling of delayed ignition. The rules differ for
individual risk and societal risk. The distance reported will be the maximum for all
outcomes for the event, whether they are individual risk or societal outcomes. This
comment applies for all flammable effect distances.
G - Largest Distance to 1% lethality
This is the greatest downwind distance for all flammable effects and can be the result of
flash fire, explosion or radiation effects.
H Corresponding Event (l% lethality)
This is the outcome code for the flammable outcome giving the greatest distance as
described above (G).
I, J and K Largest distances to 35, 10 and 3 kW/m2 radiation
These distances are not used directly in the risk calculations but they give an indication of
the downwind extent of the radiation zones. They report the maximum distances out of jet
fire, pool fire and fireball results. These values will be reported whether or not these
hazards are used in the risk.
The risk calculations use both flame dimensions, radiation zones and flammable probit
calculations to derive lethality ellipses. The largest ellipse is used as the 1% lethality
distance for the results presented as results G and H.
L and M Largest distances to 0.3 and 0.1 bar
These distances depend on the results of explosion outcomes. Using the Wiekema model
the explosion radii are calculated directly from the light and heavy damage coefficients.
Using the Mult Energy model these distances are calculated using the vulnerability model
(vuln) that relates lethality to overpressure.
Of the results reported A to D and I to K are calculated by the SMEZ report code directly
from the consequence data while E to H and L and M are calculated from results reported
by Mpact. The results reported by Mpact are maximum downwind distances for each
flammable outcome flash fire distance, largest lethality ellipse for pool, jet and fireball
fires and two explosion effects. The reporting code converts these into the results output to
the spreadsheet report.
MPACT Theory
29
4. Toxic Impact Calculations

4.1
Calculation Grid
The toxic impact calculations in MPACT start with the calculation of individual risk at each
point on the risk grid. The release can start anywhere and the wind could be blowing in any
one of the wind sectors defined. The risk calculation point is located at the centre of a grid
cell and this can be any distance from the release point. Figure 13 provides a general
illustration of the orientation of a grid cell and a toxic release.
North
xmax,ymax
2
Release point
xp,yp
Risk
calculation
point x,y
Toxic
zone
xmin,ymin
Figure 13 Lower and Upper Wind Directions to Affect a Point
We can see from Equation ( 1 ) that to calculate individual risk at a given point it is necessary
to integrate the product of probability of death Pd|w and directional probability P|w for all
possible wind directions that will affect that point. The start and end angles that affect the
point are defined as 1 and 2 and are shown in Figure 13.
Upwind spreading of instantaneous releases occurs until the wind starts to blow the cloud
downwind. In these cases there is a range of distances from the release point that will
experience toxic effects whatever the wind direction.
The probability of death depends on the data regarding the toxic effect zone. The wind
directional probability function for the relevant range of angles is derived from the data input
by the user.
MPACT Theory
30
4.2
Toxic Effect Zone at the Grid Cell
The effect results passed from the consequence calculations to the impact calculations are in
the form of a table. This gives the crosswind probability integral, PI, and the corresponding
centreline probability of death, Pd|w(at centre-line) = Pcl, for each calculation step from the
position of the release (point zero) along the centreline of the clouds path.
The probability integral is the integration of probability of death versus crosswind distance
for a given downwind position of the cloud. The shape of the curve to be integrated will
depend on the concentration profile and the toxic parameters of the material in the cloud. It
will range from a top hat shape to a more rounded shape depending on the specific dispersion
modelling results. The averaging time used to calculate the concentrations is the toxic
averaging time (typically 10min).
From this input data MPACT calculates the effective toxic width (WPI) perpendicular to the
downwind direction of the cloud and this is given by Equation ( 16 ).
W PI =
PI
Pcl
( 16 )
The approximation made within MPACT is that the cloud can be represented by a rectangular
probability of death profile with a width equal to the effective toxic width and a uniform
probability of death, Pcl. Figure 14 illustrates the way the original probability of death
function across the cloud width is being approximated in this way. The same treatment
applies to instantaneous and multiple rate releases because the data provided to MPACT is in
the form of the integrated probability of death over the entire time period to be considered.
It is also useful to define the angle formed at the release point by the effective toxic width
effective toxic angle, 2 where Figure 14 illustrates the definition given by Equation ( 17 )
and uses A to indicate the distance from the release point to the risk calculation point.
= sin 1 (
W PI
)
2A
( 17 )
MPACT Theory
31
AERIAL VIEW
X=A
Wind
Release
Source
Cloud
X=A
(a) aerial horizontal view of cloud for given weather w
PROFILE OF
PROBABILITY OF DEATH
THROUGH SECTION X=A
Probability of death
Centreline
Probability (Pcl)
Area under profile
=
Probability Integral
Probability of death
Pcl
Effective Cloud
Actual
Cloud
Effective Toxic Width (WPI)
(b) crosswind profile of probability of death Pd|w
MPACT Theory
32
North
Wind Sector B
Impact
Calculation
Square
Cloud
Centreline
Effective
toxic width
projected
through
MPACT
grid cell
Wind Sector C
Effective
Cloud
East
(c) Effective horizontal profile for probability of death Pd|w (with effective toxic width WPI)
Figure 14 - The Geometry of the Toxic Cloud
MPACT Theory
4.3
33
Wind Distribution
The weather category (stability class and wind speed), wind direction distribution and
release direction have been previously discussed in Sections 2.2.3 and 3.3. Figure 15
shows the angles MPACT uses to define the wind sector geometry where;
c,i
e,i
s
is the angle defining centreline position of sector i

is the angle defining the position of the upper boundary of sector i
is the sector angle
North
Wind
Sector
i+1
Wind
Sector i
e,i
c,i
e,i-1
Wind
Sector i-1
Figure 15 Definition Of Variables Used To Describe Each Wind Sector
The sector angle is defined simply as
s =
2
n ws
( 18 )
The sector position angles will depend on the number of wind sectors, nws, and the offset
from North defined by the user.
To integrate the risk according to Equation ( 1 ) it is necessary to know the variation of
P|w with . The data supplied to MPACT is this function integrated over each wind sector
to give the wind sector probability, Pws. P|w can be calculated from the sector probability
by dividing by the sector angle;
MPACT Theory
34
P w =
1 Pws
s Pw
( 19 )
MPACT interpolates linearly as the angle changes between the adjacent values of Pws to
obtain Pws . This interpolated value is used to provide the continuous function P|w.
MPACT Theory
35
4.4
Individual Risk Calculation
We have seen from the explanation of the toxic effect zone in Section 4.2 that for a given
distance, A, from the cloud release point the probability of death will be a constant within
the zone or zero depending on whether for a given release angle the cloud affects the
chosen angular location.
The integration of individual risk requires the product of this probability of death Pcl and
P|w to be integrated using equation ( 1 ) over the range of angles (1<<2) that affects
that release point as shown in Figure 13. Figure 16 illustrates the integration and shows a
particular case where the effective toxic angle spans a number of wind sectors. MPACT
calculates the area under the curve (1<<2) and this area forms the basis of the
individual risk for the location and weather category.
A discontinuity is possible at the extreme ends of the cloud where there may be a
centreline risk but a zero cloud width. In these cases the risk is deemed to be zero.
Angle
covered by
cloud width
0.004
0.0035
0.003
0.0025
0.002
0.0015
0.001
0.0005
0
0
45
90
135
1
180
Angle
225
270
315
360
Figure 16 Integration of Directional Probability, P|w and Probability of Death, Pcl
In practice the integration is performed in stages by calculating the contribution of each

wind sector to the risk at the point. These intermediate values are used later in the societal
risk calculation. For the individual risk all the contributions are summed to give the overall
risk at the point.
To provide further flexibility in the model for the user additional factors are used in the
calculations and these can be set by the user as input parameters. They are designed to allow
MPACT Theory
36
for a fraction of the population being indoors and the multiplication of the calculated
probability of death by a further user-defined factor, both indoors and outdoors.
The risk increment for a given weather category at a square with its centre at x,y is given by
Equation ( 1 ). As discussed previously, the value of the probability of death is currently for a
given weather is assumed to be independent of the wind direction , and therefore Pd|w =
Pd|w. As discussed in Section 4.2, the probability of death is approximated by the centreline
probability of death, Pd|w(at centre-line) = Pcl . Finally the death probability will be corrected
by the toxics fatality factor KIR|T (see Section 3.6). Thus Equation ( 1 ) is approximated by the
following equation,
IR x , y w = Fedf K IR T Pcl
|w
( 20 )
In the above equation the integral limits are defined in Figure 13, P|w is calculated according
to Equation ( 19 ) and Fedf is the frequency of the incident.
Finally the sum of risk increments for all weather categories gives the individual risk at the
point for the toxic event.
IR x , y =
Pw IR x , y w
( 21 )
Note that if the indoor risk is being calculated directly then there will be a separate set of
individual risk results for indoors and outdoors.
4.5
Societal Risk Calculation
4.5.1 Calculation of F-N Pairs

So far the discussion has related to individual risk. To obtain F-N predictions it is necessary
to define exactly what specific outcomes will be modelled.
For a toxic release, one outcome of an incident is defined as the development and impact of
the incident for a given combination of weather condition (wind speed and stability) w and
wind direction (sector s). If there are 4 weather conditions and 8 wind directions defined in
the Weather Data, then each toxic release will be modelled with 32 outcomes.
For each incident outcome the F-N pairthe number of mean fatalities N edf o and frequency
Fedf,o have to be calculated.
Incident Outcome Frequency

Having defined a toxic incident outcome as a release occurring under a particular weather
category and for the wind blowing within one sector the frequency of the incident outcome is:
MPACT Theory
37
Fedf ,o = Fedf Pws = Fedf Po
( 22 )
(245)
Incident Outcome Fatalities
In the program, the definition of the outcome of a toxic incident is in fact not specificthere
is still the variation in the wind that could blow in an infinite number of directions inside a
wind sector. This is important because the population within a calculation square could be
distributed in any way and this would also give rise to different F-N results depending on the
precise wind direction.
Instead of calculating N for one specific direction within the sector (along the centreline for
instance) MPACT uses the concept of the Mean Number of Fatalities for a toxic incident
outcome.
The method to calculate the number of fatalities for a given outcome starts with a
consideration of the population distributed over the grid. Population within specific building
objects are considered later. For a toxic release and a given outcome o = ws, the mean
number of fatalities, N x, y o for a square with its centre at x,y is given by:
N x , y o = Pc o K FN T Pcl N x , y
( 23 )
where
Pc|o
Pcl
Nx,y
K FN T
the conditional probability that the effective cloud covers the calculation square,
given the outcomes combination o=ws of weather category and wind direction
the probability of death at the centre-line of the cloud at the centre of the square.
the number of people in the square, from the Population Data
the fatality factor for societal risk and toxic results calculated using equation ( 13 )
from the input parameters for toxic effects and fraction of people outdoors
If indoor risk is being modelled using a separate indoor population array then the
corresponding calculation is;
N In x , y o = P In c o K In FN T P In cl N In x , y
( 24 )
N Out x, y o = P Out c o K Out FN T P Out cl N Out x , y
( 25 )
N x , y o = N In x , y o + N Out x , y o
( 26 )
For the overall toxic incident outcome we define the total number of fatalities as N edf o . It is
calculated as the sum of N x, y o for all MPACT calculations squares which can be affected by
the incidents outcome plus the people affected within specific building objects. So now we
MPACT Theory
38
consider the population within specific population objects that may be affected by the toxic
outcome, N Bi edf o ;
N Bi edf o = P Bi c o K I FN T P I cl N Bi
( 27 )
Where the superscript Bi indicates a specific building object and I indicates its building type.
Thus the probability of death depends on the building type because it defines its ventilation
rate. The location specified for the building vent determines the exposure of the people within
the building and this could be an important factor for a large building.
Finally the mean total number of fatalities calculated for a toxic incident outcome is:
N edf o =
x, y o
All calculation squares
Bi
( 28 )
edf o
All affected buildings
Above the calculation of the mean number of fatalities has been discussed. The
distribution of the fatalities around this mean will be discussed further in Section 4.5.2.
F-N Pairs for Incident Outcome
Hence for the toxic incidents the F-N pairs for the outcomes are the Fedf,o and the N edf o
values calculated above. It is these values that are used to create the ranking results. If
indoor risks are modelled directly then the contribution from the indoor and outdoor
populations are recorded as;
N In edf o =
In
x, y o
N Out edf
Bi
( 29 )
edf o
Out
( 30 )
x, y o
For the F-N curve a further calculation is performed as described in Section 4.5.2.
MPACT Theory
39
4.5.2 Probabilistic Distribution of F-N Pairs

Due to the way that a toxic incident outcome is defined the number of fatalities calculated in
the outcomes F-N pair is a representative value. It does not show the result for a specific
direction or the range of fatality numbers that would be experience in reality for, say, ten
releases within the wind sector with each occurrence having a unique direction.
The implications of the method used to calculate F-N are elaborated below.
Example 1: Concentrated Populations
In one extreme case there might be one point affected by a narrow plume in which the
probability of death for someone is in the cloud is KFN|T Pd|w = 1.0. Suppose that there are 10
people at that point and there is a one in ten chance of the point lying in the cloud (Pc|o =0.1).
The average number of deaths will be N x, y o = Pc|o KFN|T Pd|w N x, y o = 0.1*1*10 = 1, but it is
not true that one person will die for every occurrence of that release. Rather, nobody will die
for nine out of ten instances and ten people will die in the tenth instance.
For wider clouds this is less important because there is less variability in the probability of the
point lying in the cloud. Also this effect will become less significant for more uniform
population distributions.
Example 2: Results in the Low-N Region of the Curve
If a square has a mean number of fatalities of, say, 5 for an incident outcome which impacts
the square (e.g. 10 people in the square and a probability of death of 0.5) this does not
necessarily mean that every time the square is impinged upon that 5 people will die.
Sometime there may be 4 deaths, sometimes 6, sometimes 5, etc.
This is important because it affects the stability of the low N end of the F-N curve. The way
the calculations are done can result in a raw N which is not an integer, and if this is around
1 then small changes in the definition of the Failure Case can lead to small changes in cloud
characteristics which in turn can move the raw N to the other side of the integer (i.e. to just
below 1 or just above 1). Since the F-N curve only registers events with an N of 1 or more,
these effects will cause a jump in frequency that is quite out of proportion to the change in the
Failure Case. This also affects results around 2, 3, etc., but becomes less and less noticeable
as the N value rises.
This behaviour was observed during early applications of MPACT and gave rise to questions
concerning the stability of the results. The F-N smoothing algorithm eliminates this and
provides a much better picture of how the results change as release cases are varied.
Method used for setting the variance of the distribution
The examples above illustrate the principle that a toxic incident outcome as modelled in the
program is, in reality not one F-N outcome but represents a range of F-N outcomes. Hence it
can be useful to spread the F-N result from the modelled outcome over an F-N curve
covering the range of possible results represented by the one F-N pair.
To achieve this spreading the variance of the number of fatalities is calculated at each point in
addition to the calculation of N edf o . To model this variance (Vx, y|o), a formula is used taking
MPACT Theory
40
into account the potential variability introduced by changes in the probability of death and the
probability of the cloud affecting the point:
V x , y o = Pc o Pcl (1 Pcl ) + N x , y Pcl (1 Pc o ) N x , y K FN T
( 31 )
If indoor risk is being modelled directly then the corresponding calculation of variance is the
sum of the terms for indoor and outdoors. Note that because the relationship of variance to
the population is of the second order then a difference in the result will occur even if all other
inputs are equivalent. In the general case this switch in modelling will change the results
anyway so this effect will not be significant.
In SAFETI NL 6.7 the same toxic cloud is used for outdoor and indoor calculations with an
indoor vulnerability factor of 0.1. In this case the variance is calculated as;
Out
Vx , y o = Pc o Pcl (1 Pcl ) + ( N Out x , y + N In x , y ) Pcl (1 Pc o ) N Out x , y K FN
T
( 32 )
In
+ Pc o Pcl (1 Pcl ) + ( N Out x , y + N In x , y ) Pcl (1 Pc o ) N In x , y K FN
T
Out
Out
Out
In
K FN = f FN
K FN
+ (1 f FN ) K FN
If the fraction of people outdoors in constant for all populations then this will give the
same result as using a single grid (the original SAFETI NL method).
For specific buildings the corresponding calculation is;
VedfBi o = P Bi c o P I cl (1 P I cl ) + N Bi P I cl (1 P Bi c o ) N Bi K I FN T
( 33 )
Pcl
25
0
0.1
20
Variance
15
0.2
0.3
0.4
10
0.5
0.6
0.7
0.8
0
0
0.2
0.4
0.6
0.8
Pc|o
0.9
1
Figure 17 Variance Function for a Range of Pc|o, Pcl and Nx,y=10
MPACT Theory
41
Figure 17 illustrates the function where there are 10 people in the cell and for a range of
values of Pc|o and Pcl. It can be seen that if Pc|o and Pcl are 0 or 1 then the variance is zero.
Also the variance is greatest when Pc|o = 0.5. This shows that the greatest contribution to the
variance is from cells that have equal probability of being inside or outside the cloud for a
release in a given wind sector.
The sum of the variances over each square covered by one incident outcome
(incident/weather/wind direction combination, etc.) is used to define the total variance of the
number of fatalities. Thus the standard deviation equals
edf ,o =
x, y o
Bi
( 34 )
edf o
The sum of the number of mean fatalities over each square, N edf o =
N x , y o is
used as the baseline for the F-N spreading algorithm. The core product Fedf N edf o of the mean
number of fatalities and the frequency defines the mean rate of death and this average must
be preserved throughout the spreading exercise.
The spread So of the F-N pairs to be simulated from the mean values is defined by the product
of the standard deviation and the MPACT parameter, fs Multiplying factor for toxic F-N
spread (default value=2),
S o = f s edf ,o
( 35 )
The calculation has to ensure the spreading stays within a range of N that is physically
possible. The minimum limit to spreading is zero fatalities (N=0). The upper limit to
spreading is the total number of people exposed to the cloud whether or not they are fatalities,
N exp osed o =
x, y
All calculation squares Pcl >0
( 36 )
Bi
All buildings Pcl >0
If the indoor risk is being modelled directly then the corresponding calculation is;
N exp osed o =
In
x, y
All calculation squares P In cl > 0
Out
x, y
All calculation squares P Out cl > 0
Bi
( 37 )
All buildings Pcl > 0
The spreading value is now calculated according to equation ( 38 ).

So
= max 0.5, min f s edf ,o , N exp osed o N edf o + 0.5
]}
( 38 )
MPACT Theory
42
F/N calculation for smoothed distribution

The F-N curve is defined in intervals of N. The spreading starts at the minimum value N=0.
Then the F value for each N interval is integrated by using the error function. These values
are factored up to compensate for the cumulative probability lost by truncating the tails of the
standard deviation curve that exist outside the spreading range.
Current Incorrect formulation
The values of F are increased by the factor, fn,o as defined in equation ( 39 ).
f n ,o =
erf S o edf ,o 2
( 39 )
where the error function erf(z) is defined by
erf(z) =
( 40 )
-t2
dt
The formula used to calculate the F value for a particular range of N (N1 to N2) is
FN
N 2
(N N edf ,o )
f n ,o (N 2 N edf ,o )
erf
erf 1
2
So 2
So 2
( 41 )
F
1.20E-01
1.00E-01
8.00E-02
6.00E-02
4.00E-02
2.00E-02
0.00E+00
0
10
12
14
Figure 18 F-N distribution for Nx,y=10, Pc|o =0.5 and Pcl=1
MPACT Theory
43
Figure 18 illustrates the F-N spreading for a hypothetical cloud with a base frequency Fedf,o
=1 covering just one grid cell with one set of input parameters this corresponds to the
conditions giving the highest possible variance.
As shown in the illustration, the algorithm chosen assumes a normal distribution about the
Mean Number of Fatalities, N edf o , truncates the distribution symmetrically about the mean,
and then re-normalises the truncated distribution.
The analyst sets the points of truncation as a multiple of the standard deviation, given in the
Parameter Data. This is automatically adjusted for small values of N edf o to prevent the
program calculating negative values of N.
The lowest value of N to be reported is given by
N l ,o = N edf o + 0.5 S o
( 42 )
The highest value of N to be reported is given by.
N h,o = N edf o + 0.5 + S o
( 43 )
Proposed new formulation

Let the function gedf|o(N) be defined by
N N edf
g edf o ( N ) = erf
S 2
o
( 44 )
Then the formula used to calculate the likelihood F[N,]|o that the number of fatalities is larger
than N is given by
F[N, ] |o
=1,
=1
=0 ,
N <0
g edf |o ( N ) g edf |o (0)
g edf |o ( N exp osed 0 ) g edf |o (0)
0 N N exp osed 0
N > N exp osed 0
Thus the likelihood for the number of fatalities to be in a given range [N1,N2] equals
( 45 )
MPACT Theory
44
F[N1 , N2]|o = F[N1 , ]|o F[N 2 , ]|o

=
g edf |o ( N 2 ) g edf |o ( N 1 )
g edf |o ( N exp osed 0 ) g edf |o (0)
( 46 )
MPACT Theory
45
5. Flammable Impact Calculations

5.1
Introduction
The toxic event modelling calculates the risk for one outcome only that represented by
the probability of death integral derived for the toxic cloud. In contrast for flammable
events there could be a range of quite different outcomes depending mainly on the time
and location of the ignition. Indeed there is the possibility the cloud will not ignite at all
and this represents a valid outcome.
The description of how MPACT deals with multiple outcomes using event trees is covered
in the following Section. Note, however, that there are some flammable outcomes
(BLEVE and Pool fire) that can also be specified directly by the user without additional
event tree definition. The only event tree probabilities for these straight line events are;
Probability of BLEVE
Probability of Pool Fire
These are the probabilities that the user defined BLEVE or Pool Fire Failure Event
develops in a hazardous manner, and should normally be set at 100%. The probability of
ignition has been taken into account in the specification of the event frequency itself.
Another effect that may be included in the risk calculations is that from a vessel burst.
This is referred to as the BLEVE Blast model in Phast Risk. It does not have a probability
factor and the event frequency is used directly.
Other types of flammable releases are treated using event trees and these are described in
Section 5.2. To obtain the outcome frequency in each direction the wind direction
probabilities are considered as explained in Section 5.3. The delayed ignition outcomes
also need the evaluation of ignition probability in each time step and this is explained in
5.19. This Explanation comes later because it is helpful to understand the subsequent
Sections on how the vapour cloud is represented in MPACT and how it develops with
time.
Having determined the probability of each outcome from the event trees the model must
then determine the impact of these outcomes and Sections 5.4 to 5.14 explains how
MPACT performs this modelling. Finally Sections 5.16 to 5.17 explain how the
probabilities and the consequences are combined to calculate the risk.
5.2
Event Trees
In Section 5.2.1, a list is given of the flammable outcomes that may be modelled by
MPACT (e.g. fireball, ).
There can be a wide range of different cloud developments, depending on the type of
release (instantaneous or continuous), on the timing, location and type of ignition, and on
the behaviour of any liquid portion of the release (i.e. on rainout and vaporisation). Based
MPACT Theory
46
on the results of the consequence modelling one of 22 characteristic types of cloud

development (routes) is selected for the event. This selection is described in Section 5.2.2.
For the selection of an event tree route, MPACT can match an event tree type to the cloud
that models the event with the appropriate set of individual and possibly combined
outcomes. There are currently 6 different event tree types. These are described in detail in
Section 5.2.3.
5.2.1 Flammable Outcomes

The full list of outcomes includex:
Fireball/BLEVE
Horizontal Jet Fire
Vertical Jet Fire
Immediate Flash Fire for Instantaneous or Short-Duration Continuous Release
Immediate Explosion for Instantaneous or Short-Duration Continuous Release
Immediate Pool Fire
Delayed Flash Fire (timing and location depends on ignition sources)
Delayed Pool Fire associated with Delayed Flash Fire. This occurs if the pool is still in
contact with the cloud at the time of ignition, and produces a combined effect zone
Delayed Explosion (timing and location depends on ignition sources)
Residual Pool Fire. This is modelled if there is a pool remaining after the cloud has
finished dispersing (a residual pool). This is quite common for liquid spills where the
saturated vapour over the pool is below flammable limits.
Each outcome will have its own pattern of fatalities, and its own conditional probability (and
frequency).
5.2.2 Selection of event tree route and event tree type

Only one of the 22 Event Tree Routes will be used to model a particular Failure Case. The
structure of the Event Tree is designated according to the cloud Type (A, B, C, D, E, F)
and the Route Number (1a, 1b, etc. to 20) that defines a sub-type specifying the exact set
of branch probabilities from the MPACT parameter data to use. Figure 19 shows how the
cloud type is designated.
The reader is referred to the UDM theory manual8 for a detailed description of the phases
in the cloud development. So, using the diagrams in Figure 19 and the defining the route
number used for each event both the correct event tree structure and relevant set of
MPACT branch probabilities can be identified. For instance instantaneous releases without
the possibility of rainout are defined by route number 10 and event tree structure C. This
defines the branch probability set from the MPACT parameters to be used for the case.
This is a very simple example. A more complicated cloud is one that is continuous, rains
out, the pool evaporates, the pool duration is short and the cloud is essentially
instantaneous. This gives a route number of 9 and event tree structure D.
x
DOC. Logic for these several outcomes is currently mostly in PHAST. This should be further completely
described in PHAST model-linking theory document. This is also a crucial prerequisite for further MPACT
quality improvement.
MPACT Theory
47
Note that the historical development of MPACT has lead to the use of numbering and
lettering conventions that appear to be more complex than necessary for the purposexi.
This is due to the limitations of the consequence modelling prior to the development of the
UDM. The free jet route is such an example. Today the consequence zones for this route
are derived from the UDM results but in the past there was a specific turbulent jet only
model, distinct from the dense cloud only model. The old routes are retained for
compatibility with old studies.
xi
DOC/EXTENSION rationalise the use of event tree types and route numbering and remove outdated
functionality
MPACT Theory
48
A Release with No Possibility of Rain-Out

Essentially
Instantaneous?
Behaves as
Free Jet?
Event Tree :
Route and Type
1b: B
1a: A
N
Continuous
Release
2: B
Y
Instantaneous
Release
10: C
A Continuous Release with a Possibility of Rain-Out

Rainout?
Evaporation?
Pool of limited
Essentially
Behaves as
duration?
Instantaneous? Free Jet?
Event Tree:
Route and Type
3b: B
3a: A
N
N
Release
4: B
5: D
6: D
N
Y
N
7: D
8: D
9: D
An Instantaneous Release (I) with a Possibility of Rain-Out

or an Essentially Instantaneous (QI) Release with a Possibility
of Rain-Out
Rainout?
Evaporation?
Pool left
behind?
Pool of limited
duration?
11: C (I)
16: B (QI)
12: E (I)
17: D (QI)
Release
Event Tree:
Route and Type
13: E (I)
18: D (QI)
N
Y
14: F (I)
19: D (QI)
15: F (I)
20: D (QI)
Figure 19 Selection of event tree route and event tree typexii
xii
DOC The route number selection is made prior to MPACT. Essentially instantaneous has the same
meaning as quasi-instantaneous. A free jet is one that dilutes below the flammable limit fraction before it
starts behaving as a dense cloud. Possibility of rainout means there is initial liquid in the release. Evaporation
decision is based on user-accessible mass flow rate parameters (one for toxics, one for flammables). Pool of
limited duration is true if it evaporates before the maximum release duration parameter. Pool left behind
decision is based on the back end of the cloud leaving the edge of the pool. Time-dependent releases are
treated as continuous.
MPACT Theory
49
5.2.3 Event Tree Types

Each of the 22 event trees has one of the six structures listed in Table 1.
Type
B
C
D
E
F
Title
Continuous (possibly including QI), No Rainout, No Free Jet
Instantaneous, No Rainout
Continuous (possibly including QI), Rainout
Instantaneous, Rainout, Instantaneous Cloud does not leave vaporising pool behind
Instantaneous, Rainout, Instantaneous Cloud leaves vaporising pool behind
Table 1 Event Tree Types

The full tree structures for each type is shown in the following diagrams and contains all
of the possible outcomes for that type of event. Note that tree type A is now obsolete so
the list starts with B. For some of the 22 route numbers a particular outcome in the tree is
irrelevant, and that specific Event Tree will not provide the branch probabilities for that
outcome in the MPACT parameters. For instance if event tree type B is used for route
number 1b then the short release fraction branches will not be used. The event tree is still
considered to be of the same type even though some outcomes are inaccessible.
In each of the event trees there is a probability of immediate ignition. There are a number
of ways of specifying this probability when using the Mpact model and it is an important
factor in defining the risks due to leaks of flammable materials.
The options available are described in Chapter 5.18.
MPACT Theory
50
Event Tree: Type B (Continuous, No Rainout)

Applies to Route Numbers: 1b, 2, 3b, 4, 16
Horizontal
Vertical
Immediate
Ignition
Jet Fire
No Effect
Jet Fire
No Effect
BLEVE
Short Release Fraction
Flash Fire
Explosion
No Effect
Flash Fire
2
Delayed Ignition
No Immediate
Ignition
Explosion
No Effect
Progressed
through
consequence
time-steps
No Ignition
Note:
1: MPACT uses this set of branches if the release is a short-duration continuous (SDC)
release, i.e. if it consists of a single segment and the duration is less than or equal to
the "Cut off time for short duration releases", set in the Dispersion Modelling
Coefficients Parameters. In this case, it will apply the "Frequency Fraction for short
release effects" to the branch, and (1 - Frequency Fraction) to the branch above. If the
release is not an SDC release, MPACT will not use this set of branches.
2: MPACT models the cloud through a series of time-steps, until it has dispersed to a
safe concentration, or until the residual probability of non-ignition has dropped below
the "Minimum event probability considered" set in the Impact Vulnerability Parameters.
Figure 20 Event Tree Type B Continuous, Not Free Jet, No Rainoutxiii
The immediate ignition branches in event tree type B are designed to allow some
representation of instantaneous type outcomes in addition to jet fires. A BLEVE-style
fireball for instance is considered as a possible outcome. The short-duration continuous
branch is independent of the quasi-instantaneous modelling.
xiii
IMPROVE. Only use cut-off time parameter tSDC; for given release duration tdur, use frequency fraction
[1-tdur/tSDC] for the SDC branch and [tdur/tSDC] for the branch above. This will ensure continuity and at the
same time removes one unnecessary parameter. tSDC= 20s by default.
MPACT Theory
51
Event Tree: Type C (Instantaneous, No Rainout)

Applies to Route Numbers: 10, 11
BLEVE
Immediate
Ignition
Flash Fire
Explosion
No Effect
Flash Fire
Delayed Ignition
Explosion
No Effect
No Immediate
Ignition
Progressed
through
consequence
time-steps
No Ignition
Note :
1:
MPACT models the cloud through a series of time-steps, until it has dispersed to a
safe concentration, or until the residual probability of non-ignition has dropped below
the "Minimum event probability considered" set in the Impact Vulnerability Parameters.
Figure 21 Event Tree Type C Instantaneous, No Rainout
MPACT Theory
52
Event Tree: Type D (continuous with rainout)

Generalised to apply to Route Numbers: 5, 6, 7, 8, 9, 17, 18, 19,
20
Jet Fire
Horizontal
Pool Fire
Both
No Effect
Jet Fire
Vertical
Immediate
Ignition
Pool Fire
Both
No Effect
BLEVE + Pool Fire
BLEVE alone
Flash Fire + Pool Fire
Short Release Fraction
Flash Fire alone

Explosion + Pool Fire
Explosion alone
Pool Fire alone
No Effect
Flash Fire + Pool Fire 3
Delayed Ignition
2
No Immediate
Ignition
AND
Residual 4
Explosion
No Effect
Dispersion
No Ignition
Progressed
through
consequence
time-steps
Residual Pool Fire

No Effect
Note:
1: MPACT uses this set of branches if the release is a short-duration continuous (SDC) release, i.e. if it consists of a
single segment and the duration is less than or equal to the "Cut off time for short duration releases", set in the
Dispersion Modelling Coefficients Parameters. In this case, it will apply the "Frequency Fraction for short release
effects" to the branch, and (1 - Frequency Fraction) to the branch above. If the release is not an SDC release,
MPACT will not use this set of branches.
2: MPACT models the cloud through a series of time-steps, until it has dispersed to a safe concentration, or until the
residual probability of non-ignition has dropped below the "Minimum event probability considered" set in the Impact
Vulnerability Parameters. If the pool vaporises, the cloud may be modelled as a series of segments, the first group
representing the cloud from the initial release, and the remaining representing the vaporisation from the pool. The
segments are assumed to be connected, to form a single cloud.
3: If the cloud ignites while the cloud is still in contact with the pool, there will be pool fire effects in addition to the
flash fire effects. There will be no pool fire effects if the ignition occurs after the cloud has become separated from
the pool, or after the pool has vaporised completely. The cloud could become separated from the pool if the
release was essentially instantaneous and the pool did not vaporise.
4: If the pool has not vaporised by the time the release has finished and the cloud has dispersed to a safe
concentration, then MPACT will include the probability and effects of a residual pool fire. If the pool has vaporised
before the cloud has finished dispersing, then MPACT will not use this set of branches.
Figure 22 Event Tree Type D Continuous with Rainout
The separation of the cloud from the pool conceptually introduces the possibility of two
independent outcomes 1) ignition of the cloud and 2) ignition of the pool. MPACT models
the travel of the cloud and calculates an outcome per timestep downwind. It does not
model an equivalent pool fire event per time step. Instead it combines all possible ignition
times into one residual pool fire outcome. This approach is taken to simplify the
calculations in an area that does not normally give the dominant hazardous effects. The
probability of the residual pool fire outcome is specified as an event tree parameter but it is
only used if there is a residual probability of non-ignition of the dispersing cloud at the
time the cloud leaves the pool.
Since the cloud has left the pool for the residual pool fire outcome it is considered that the
vapour over the pool does not exceed the lower flammable limit and cannot generate a
flash fire. The pool fire is modelled according to the radiation effects and not any flash fire
effects.
MPACT Theory
53
Event Tree: Type E (instantaneous with rainout; no vaporising pool left behind)
Generalised to apply to Route Numbers: 12, 13
BLEVE + Pool Fire
BLEVE alone
Immediate
Ignition
Flash Fire alone

Explosion alone
Pool Fire alone
No Effect
2

Delayed Ignition
1
No Effect
Dispersion
No Immediate
Ignition
No Ignition
AND
Residual
Explosion
Progressed
through
consequence
time-steps
Residual Pool Fire

No Effect
Note:
1: MPACT models the cloud through a series of time-steps, until it has dispersed to a safe concentration, or until the
residual probability of non-ignition has dropped below the "Minimum event probability considered" set in the
Impact
Vulnerability Parameters. For these Event Trees, all of the pool vaporisation occurs while the initial cloud is
still
over the pool, so the cloud can be represented by a single, instantaneous release
2: segment.
If the cloud ignites while the cloud is still in contact with the pool, there will be pool fire effects in addition to
the
flash fire effects. There will be no pool fire effects if the ignition occurs after the cloud has become separated
from
the pool, or after the pool has vaporised
3: completely.
If the pool has not vaporised by the time the cloud has dispersed to a safe concentration, then MPACT will include
the probability and effects of a residual pool fire. If the pool has vaporised before the cloud has finished
dispersing, then MPACT will not use this set of branches.
Figure 23 Event Type E - Instantaneous with Rainout, No Vapourising Pool Left Behindxiv
xiv
Note that in SAFETI 6.* event trees type E and F are presented in a unified form as
Instantaneous/Rainout
MPACT Theory
54
Event Tree: Type F (instantaneous with rainout and vaporising pool left behind)
Generalised to apply to Route Numbers: 14,
15
BLEVE + Pool Fire
BLEVE alone
Immediate
Ignition
Flash Fire alone

Explosion alone
Pool Fire alone
No Effect
Delayed Ignition
Dispersion
Initial
Instantaneous
Cloud 1
Residual Pool Fire
No Effect
AND
Delayed Ignition
Continuous
Clouds from
1
Evaporation
Dispersion
Explosion
No Effect
No Ignition
AND
Residual
Progressed
through
consequence
time-steps
No Effect
No Ignition
AND
Residual
No Immediate
Ignition
Explosion
Progressed
through
consequence
time-steps
Residual Pool Fire

No Effect
Note:
1: For an instantaneous cloud with trailing continuous segments from vaporisation that occurs after the instantaneous cloud has left
the pool behind, MPACT first models the instantaneous cloud to the end of its dispersion. If there is any residual probability of
non-ignition at the end of this dispersion, MPACT will then model the dispersion of the continuous segments as a separate
release, using the residual probability for the "Continuous Clouds" branch.
2: MPACT models the cloud through a series of time-steps, until it has dispersed to a safe concentration, or until the residual
probability of non-ignition has dropped below the "Minimum event probability considered" set in the Impact Vulnerability
Parameters. For these Event Trees, the cloud is modelled as a series of segments; the first group represents the cloud from
the initial release combined with the material that vaporises while the initial cloud is still over the pool, and the remaining
segments represent the vaporisation after the cloud has left the pool behind.
3: If the cloud ignites while the cloud is still in contact with the pool, there will be pool fire effects in addition to the flash fire
effects. There will be no pool fire effects if the ignition occurs after the cloud has become separated from the pool, or after the
pool has vaporised completely.
4: If the pool has not vaporised by the time the cloud has dispersed to a safe concentration and if there is any residual probability
of non-ignition of the cloud at this stage, then MPACT will include the probability and effects of a residual pool fire. If the pool
has vaporised before the cloud has finished dispersing, then MPACT will not use this set of branches. For these Event Trees,
the program models a residual pool fire at the end of the dispersion of the instantaneous cloud and also at the end of the
dispersion of the continuous cloud; in each case, it applies the appropriate probability of non-ignition for that stage in the
release.
Figure 24 Event Tree Type F Instantaneous with Rainout and Vapourising Pool Left Behind
5.2.4 Event Tree Probabilities
Each event tree branch introduces a probability. The MPACT parameters contain the full
set of probabilities categorised according to the route number so the user can set different
probabilities for each Tree. These probabilities are used to calculate the frequency of each
outcome for subsequent use in the risk calculations.
There is an important distinction between the immediate ignition outcomes on the event
trees and the delayed ignition outcomes:
-
The immediate ignition outcomes are defined to occur with precisely the
probability defined by the branch probabilities. Taking Type C as an example we
can say that if the event frequency is 0.001 per year, the probability of immediate
ignition is 0.3 the flash fire probability is 0.6, then the overall frequency of an
immediate ignition flash fire is 0.001x0.3*0.6 = 0.00018 per year.
MPACT Theory
55
On the other hand the delayed ignition outcomes occur at a frequency calculated by
the ignition model. The delayed ignition branch has a prescribed probability
(indicated by the term no immediate ignition in the event tree diagrams) but there
is a succession of outcomes corresponding to different ignition times. The
probability of each of these outcomes is calculated by the ignition modelling in
MPACT according to the time steps provided as cloud input data. See Section 5.19
for further details.
Example - Event tree probabilities for cloud type D

Instead of listing the probabilities in full, this Section uses Event Tree type D (continuous
with rainout) as a detailed example. This type contains the full range of possible
behaviours, so the principles behind its probabilities are applicable to the other less
complex trees.
There are nineteen Event Tree probabilities for Event Tree Type D in the Parameters Data.
The description below refers to the parameter name and the symbol used to represent the
probability:
1. probability of immediate ignition Pi. This is the probability that the release ignites
immediately, at the release point, before the cloud has begun to disperse and to reach
ignition sources away from the release point. The root probability of the delayed
ignition branch is therefore (1- Pi). If Pi is one, no delayed ignition will be considered
by MPACT.
2.
horizontal fraction for vertical and horizontal release cases Ph. This will be used if
the jet orientation for the Failure Case is set as Both and is the probability that the
release is horizontal. The probability of a vertical release is given by (1 - Ph).
3. probability of horizontal jet fire Pjh. This is the probability that the immediate ignition
of a horizontal release gives a jet fire on its own, with no accompanying pool fire.
4. probability of pool fire (horizontal release) Pph. This is the probability that the
immediate ignition of a horizontal release gives a pool fire on its own, with no
accompanying jet fire.
5.
probability of pool and horizontal jet fire (horizontal release) Pjph.This is the
probability that the immediate ignition of a horizontal release gives a jet fire and a pool
fire simultaneously, with a combined effect zone.
6.
probability of vertical jet fire Pjv (vertical release). This is the probability that the
immediate ignition of a vertical release gives a jet fire on its own, with no
accompanying pool fire.
7.
probability of pool fire (vertical release) Ppv. This is the probability that the
immediate ignition of a vertical release gives a pool fire on its own, with no
accompanying jet fire.
8.
probability of pool and vertical jet fire (vertical release) Pjpv. This is the probability
that the immediate ignition of a vertical release gives a jet fire and a pool fire
simultaneously, with a combined effect zone.
MPACT Theory
9.
56
fraction for short release effects PS. This probability (or fraction) is used when the
release consists of a single Release Segment and the duration is less than or equal to
the Cut-off time for short continuous releases, set in the Flammables parameters or
when the release is time-varying and the parameter to model short duration effects for
time-varying releases is selected. For such releases, the program can model both
instantaneous and continuous effects for immediate ignition, and it uses this fraction to
divide the probability between the two types; the instantaneous effects have a
probability equal to this fraction, and the continuous effects have a probability equal to
(1 - PS).
10. probability of BLEVE & pool fire Pbp. For a release which meets the short-duration
criteria described above, this is the probability that immediate ignition gives a BLEVE
with an accompanying pool fire.
11. probability of BLEVE without pool fire Pb. For a release which meets the shortduration criteria described above, this is the probability that immediate ignition gives a
BLEVE on its own, with no accompanying pool fire.
12. probability of immediate flash fire and pool fire Pfpi. For a release which meets the
short-duration criteria described above, this is the probability that immediate ignition
gives a flash fire with an accompanying pool fire.
13. probability of immediate flash fire without pool fire Pfi. For a release which meets the
gives a flash fire on its own, with no accompanying pool fire.
14. probability of immediate explosion and pool fire Pepi. For a release which meets the
gives an explosion with an accompanying pool fire.
15. probability of immediate explosion without pool fire Pei. For a release which meets
the short-duration criteria described above, this is the probability that immediate
ignition gives an explosion on its own, with no accompanying pool fire.
16. probability of immediate pool fire Ppi. For a release which meets the short-duration
criteria described above, this is the probability that immediate ignition gives a pool fire
on its own, so that the vapour cloud remains unignited
17. probability of residual pool fire Prp. Event Tree D includes the possibility of releases
where the pool is still present when the dispersion modelling is completed, i.e. after the
release has stopped and the cloud has dispersed to a safe concentration, or after the
Maximum Release Duration set in the Data Limits Parameters has been reached. The
program calculates the residual probability of non-ignition at this time, which is given
by (1 - cumulative probability of ignition to the time when the cloud leaves the pool).
If this residual probability is greater than zero, then the program considers the ignition
of the residual pool, with a probability given by:
MPACT Theory
57
Po = (1 Pi ) Pwss Pirp .Prp

where:
Pi
Pwss
Pirp
Prp
( 47 )
is the probability of immediate ignition

is the probability of the weather category and wind direction
is the residual probability of non-ignition at the time when the cloud
leaves the pool
is the probability of a residual pool fire
Event Tree Type D is a generalised structure which covers Event Trees with and
without residual pools. Route numbers 5, 6, 7, 17 and 19 apply to releases that do have
residual pools, and this probability is included in their set of Event Tree Parameters.
However, Routes 8, 9, 18 and 20 apply to releases without residual pools, and their
sets of parameters do not include this probabilityxv.
18. probability of delayed flash fire Pfd. This is the probability of a flash fire in the event
of a delayed ignition of the dispersing cloud. If the cloud is still in contact with the
pool at the time of the ignition, then the flash fire is assumed to ignite the pool, giving
a combined flash fire and pool fire. The envelope of the flash fire is used to determine
the possibility of delayed ignition and this does not take into account the dimensions of
the pool directly.xvi
19. probability of delayed explosion Ped. This is the probability of a delayed explosion in
the event of a delayed ignition of the dispersing cloud.
Outcome Probabilities
The final Outcome Probability, Po depends on the exact event tree and branch followed.
Table 2 lists the derivation of Po for each outcome on each branch of the event trees
A,B,C,D,E,F as shown respectively in Error! Reference source not found., Figure 20,
Figure 21, Figure 22, Figure 23 and Figure 24.
xv
EXTENSION. Introduce a pool fire outcome that becomes independent of the cloud once it leaves the
pool.
xvi
EXTENSION. Include the pool dimensions when calculating delayed ignition probabilities.
MPACT Theory
Event
Tree
Branch 1
Independent
BLEVE
Independent
Pool Fire
Immediate
58
Branch 2
Branch 3
Branch 4
PwssPp
Not
release
short
Not
immediate
Delayed
ignition
Immediate
Not
immediate
BLEVE
Flash Fire
Explosion
Delayed
igniton
Immediate
Not short
Horizontal
Jet Fire
(1-PS)PhPusPjh Pi
Vertical
Jet Fire
(1-PS)
(1Ph)PwssPjv Pi
PSPwssPblPi
PSPwssPffPi
PSPwssPePi
(1-Pi)PwssPffPdi
BLEVE
Flash Fire
Explosion
Flash Fire
Explosion
(1-Pi)PwssPePdi
PwssPbPi
PwssPffPi
PwssPePi
(1-Pi)PwssPffPdi
Flash Fire
Explosion
Horizontal
Jet Fire
Pool fire
Vertical
Jet
and
Pool fire
Jet Fire
Pool fire
Jet
and
Pool fire
Short release
Not
immediate
Po
Pb
Short release
Branch 5
Dispersion
BLEVE
and
pool fire
BLEVE alone
Flash and pool
fire
Flash fire alone
Explosion and
pool fire
Explosion
alone
Pool fire alone
Delayed
Residual
Pool fire
(1-Pi)PwssPePdi
(1-PS)PhPwssPjh
Pi
(1-PS)PhPwssPph
Pi
(1-PS)PhPwssPjph
Pi
(1-PS)(1Ph)PwssPjv Pi
(1-PS) )(1-Ph)Pwss
Ppv Pi
(1-PS) )(1-Ph)Pwss
Pjpv Pi
PSPwssPbpPi
PSPwssPblPi
PSPwssPfpPi
PSPwssPfPi
PSPwssPepPi
PSPwssPePi
Flash and
pool fires
Explosion
PSPwssPpPi
(1-Pi)PwssPfpPdi
(1-Pi)PwssPePdi
(1-Pi)PwssPrpPirp
MPACT Theory
Immediate
Not
immediate
Immediate
Not
immediate
59
BLEVE
and
pool fire
BLEVE alone
Flash and pool
fire
Flash fire alone
Explosion and
pool fire
Explosion
alone
Pool fire alone
Dispersion
Residual
BLEVE
and
pool fire
BLEVE alone
Flash and pool
fire
Flash fire alone
Explosion and
pool fire
Explosion
alone
Pool fire alone
Initial
Instantaneous
Continuous
clouds
from
evaporation
PwsPbpPi
PwsPblPi
PwsPfpPi
PwsPffPi
PwsPepPi
PwsPePi
Delayed
PwsPpPi
(1-Pi)PwsPfpPdi
Flash and
pool fires
Explosion
(1-Pi)PwsPePdi
(1-Pi)PwsPrpPirp
PwsPbpPi
Pool fire
PwsPblPi
PwsPfpPi
PwsPffPi
PwsPepPi
PwsPePi
Dispersion
Delayed
Residual
Dispersion
Pool fire
Delayed
Flash and
pool fires
Explosion
Flash and
pool fires
Explosion
Residual
Pool fire
PwsPpPi
(1-Pi)PwsPfpPdi
(1-Pi)PwsPePdi
(1-Pi)PwsPrpPirp
(1-Pi)PniPwsPfpPdi
(1-Pi) PniPwsPePdi
(1-Pi)
PniPwsPrpPirp
Table 2 Outcome Probability Po for each Type of Event Tree
MPACT Theory
60
Most of the outcome probabilities are self explanatory, however, there are some exceptions
that require explanation:
Exception 1: Jet Direction Set in Failure Case Definition
If the Failure Case is defined with a release direction of Vertical or Horizontal, then
when the true jet event tree is used only the jet direction branch chosen is followed (with a
branch probability of 1).
Exception 2: Short-Duration Continuous Release
Normally MPACT uses only the branches for continuous effects and these over-ride the
probability set for the Frequency Fraction for short release effects for that event tree (see
Event Tree Route Number 1b for an example). MPACT will use the short duration branches
only for releases that a) have only one segment with a duration less than the Flammable
parameter cut-off time for short-duration releases or b) are time varying and the Flammable
parameter model short-duration effect for time varying releases is set.
Exception 3: Flammable Energy too Low for Explosion
For all branch splits where an explosion is one of the new branches, there is a check to see if
the explosion energy is greater than the minimum explosive energy set in the Parameter Data.
If the energy is sufficient, then the calculations continue using the event tree branch
probabilities set in the Parameter Dataif the mass is not sufficient, the explosion branch is
ignored . When using 3D modelling the explosion branch probability is added in with the
flash fire branch. When using the 2D modelling the risk grid will miss the contribution from
the flash fire effects but there is a contribution to the societal results (see next exception).
Exception 4: Probability of Delayed Ignition for Flammable Clouds
For releases involving flammable clouds which are not immediately ignited, the delayed
ignition probability (Pdi) is based on the ignition sources defined in the Ignition Data and on
the Population Data. The Population Data are included for ignition because many human
activities involve an ignition riskthe risk-level per person is set using the population
omega factor in the Parameter Data. Only populations on the in indoor or outdoor grids are
considered as ignition sources.
The probability of delayed ignition is not calculated as a single number for an incident, but is
calculated for each time step of the incidents development. The subsequent branches use the
time step ignition probability to calculate the time step incident outcome probabilities. Hence
the F-N pairs are for incident outcomes at each step in an incidents development. Typically
the model will calculate a flash fire outcome and an explosion outcome at each time step with
conditional probabilities defined by the relevant event tree.
If the explosion outcome would lead to fewer fatalities than the associated flash fire then the
flash fire total of fatalities is used since in reality the flash fire effects are a subset of the
combined hazards.
Free Field Method9
This is a treatment of delayed ignition outcomes as required by the Dutch Purple Book for the
calculation of individual risk. It is not used for societal risk calculations. If the Free Field
Modelling method is selected in the General Risk Parameters, then two sets of delayed
MPACT Theory
61
ignition impact calculations are performed. One set applies the free-field ignition modelling
rules and is used for the individual risk results (risk contours and risk ranking points) while
the other set follows the normal ignition modelling and is used for the societal risk results
(societal ranking and FN curve). The different ignition modelling approaches are explained in
Section 5.19 on ignition modelling.
The free field outcomes are used only for individual risk results and are defined as follows
An additional ignition time is modelled to represent delayed ignition offsite. This is

termed the Free Field Scenario It is defined as the instant that the cloud has the
maximum ground level footprint area to the LFL fraction.
The free field scenario outcomes are an explosion and a flash fire as for the normal
delayed ignition scenarios, just that the time and location are defined by the
maximum area criterion.
For example, if the probability of immediate ignition is set as 10%, and the probability of
delayed ignition from the onsite sources was calculated as 30%, then the probability for the
Free Field ignition scenario will be 60%.
If the Include Late Pool Fire option is selected in the General Risk Parameters, the effect
zone for the Free Field ignition flash fire outcome will include the effect zone for any
associated pool fire whether or not the cloud and pool are connected at the time of the freefield scenario.
Exception 5: Delayed Ignition of a Pool
If a release has rained out, and the pool has not evaporated completely by the time the cloud
the cloud finishes dispersing (or the Maximum Release Duration parameter), then MPACT
models the ignition of this residual pool as a residual pool fire. The probability of the
residual pool fire depends on the residual probability Pirp of non-ignition of the release at
the time that the cloud loses contact with the pool (i.e. 1 - the cumulative ignition
probability at this time). After this point, ignition of the cloud will not lead to ignition of
the pool.
Exception 6: Mounded Tank Option
When a tank is buried it is protected from impingement by an external fire. The mounded
tank option provides a method to eliminate the BLEVE outcome from the event trees. This
applies only to the instantaneous releases and has the effect of setting the conditional
probabilities for BLEVE outcomes to zero. Not that the effect of this is to mean that the
conditional probabilities will no longer add up to one (assuming non-zero conditional
probabilities for the BLEVE outcomes).
Exception 7: Transport Vessel Immediate Ignition Option
When immediate ignition is set according to the Purple Book method for transport vessels
then the immediate ignition conditional probabilities for the BLEVE for instantaneous
releases are set to unity. In this way the only immediate ignition outcome for an instantaneous
release from a transport vessel is the fireball. The Purple Book methodology is described in
further detail in chapter 5.18.2.
Exception 8: Volume-based Explosion Probability
MPACT Theory
62
This applies as an option if the obstructed-region method is being used. The volume of the
cloud within the obstructed regions is used to derive the conditional probability of explosion.
The concept is that in reality not all ignition events will lead to the generation of significant
overpressures and the smaller the cloud the less likely. Historic data may be used to derive a
relationship between cloud volume within an obstructed region and the conditional
probability of an explosion, Pe. The conditional probability of the flash fire (with pool fire)
only is then simply Pfp =1- Pe..
It is also considered that the likelihood of the explosion outcome is material-dependent in a
similar manner to the explosion strength calculated using the BST or ME models. The model
therefore allows for sets of conditional probability parameters. Pe. =f(V,laminar flame speed)
and the function is interpolated from discrete values of volume and probability defined as
parameters. Below the minimum volume Pe is zero and above the highest it remains at the
highest specified level. In between the value is taken using linear interpolation. There are
three sets of parameter and these correspond to laminar flame speeds considered to be low,
medium and high. This is analogous to the categories applied when using the BST model but
the transition values of flame speed may differ according to user settings.
The default settings are given here;
Volume
/m3
Low flame speed materials Medium flame speed High flame speed
(<0.45m/s)
materials (>=0.45m/s materials (>=0.75m/s)
and < 0.75m/s)
200
0
0.3
0.6
3000
0.3
0.6
0.9
6000
0.6
0.9
1.0
Table 3 Conditional probability of explosion based on obstructed cloud volume
Exception 9: Model level Explosion Probability

The explosion probability may be specified by the user for a particular release. In this case
the supplied value will override the event tree value. The flash fire branch probability is
calculated as Pfp =1- Pe .
MPACT Theory
5.3
63
Wind Directional Probabilities
5.3.1 Wind Sector Subdivision

Many of the flammable effects are dependent on the wind direction. To calculate the risks
MPACT follows the event trees already discussed in Section 3.3 for the entire wind rose. It
does this in steps according to the Wind Sector Subdivision parameter, nws,sub . If the
subdivision parameter is set to 1 then there will be only one direction considered per wind
rose sector, in the general case if the subdivision parameter is n there will be n directions
considered. The increase in the subdivision parameter will increase calculation times
proportionately and give an improved distribution of the flammable effects with direction.
There will also be an increased number of stored results for all subdivisions are saved.
North
Wind
Sector i
Pwss,1
Pwss,2
Wind
Wind
Sector
Sector i-1
i+1
e,i-1
e,i
Figure 25 Wind Sector Subdivision Parameter = 2

Figure 25 illustrates the subdivision of a wind rose sector by a factor of 2. MPACT will
model two directions in the sector along the centre of each subdivision. The probability of
each modelled direction, Pwss, is interpolated from the wind rose probability, Pws. The
interpolation is linear between adjacent sectors to give Pws for the angle at the centre of the
sector subdivision, divided by the number of subdivisions and then adjusted to ensure that
the sum of the values of Pwss for a given sector matches the overall value Pws. This is
necessary because by interpolating between sectors the total will not necessary match the
original wind sector probability.
MPACT Theory
64
Pwss =
5.4
n ws , sub
Pws
P
Pws n
= Pws n ws
ws , sub
ws , sub
1
Pws

P
ws
1 n
1
ws , sub
( 48 )
Flammable Effect Zone Types
Flammable effects are treated in two phases. First the immediate ignition effects are
evaluated and then the delayed ignition effects. The term immediate is intended to have a
broad interpretation. The Purple Book refers to these effects as direct ignition and the
concept is that the ignition is associated with the events leading to the release. Clearly some
of these effects will take a finite time to develop the flash fire for instance will require a
flammable cloud to develop. These effects are given a specified ignition probability on the
event trees and they have prescribed dimensions independent of time. Delayed effects on the
other hand are treated as time-dependent and the dimensions of the effect zones are functions
of time.
The consequence calculations pass the shape and size of the different types of effect zone for
immediate ignition directly to the impact calculations. The explosion calculation is treated
slightly differently and the flammable mass is input to MPACT which itself calculates the
size of the explosion effect zones. The shapes assumed for the flammable effects were chosen
to be representative of type of effect zone taking into account economy of computer memory
and storage space. The overall approach and assumptions are summarised in Table 4. For
delayed ignition there are also arrays of data representing the progress of the cloud
downwind.
The jet fire calculations are governed by the definition of the release direction. This definition
is made prior to the consequence calculations and determines the relevant effect zones passed
to MPACT. If the release is modelled as horizontal without an angle then the release
definition will depend on the flag to Model Risk Effects for Vertical Jet Fires. If this flag is
selected then both vertical and horizontal jet fires will be modelled in the risk calculations
using the horizontal fraction in the event trees. This is the equivalent to the Both direction
option that was available in the OTHEDF program from SAFETI Classic.
If the Model Risk Effects for Vertical Jet Fires is not selected then only the horizontal jet
fire will be used in the jet fire risk modelling and the horizontal fraction value in the event
trees will not be used and all the probability of the combined jet fire branch will be assigned
to the horizontal jet fire outcome.
If the direction chosen for the release in the product is horizontal with an angle then the
Model Risk Effects for Vertical Jet Fires is not accessible. Instead, only one jet fire is
considered at the angle specified for the release. This angle could be horizontal. This jet fire
is reported as vertical and is treated according to the vertical jet fire in the table below. The
Vertical release direction is treated in the same manner but is always exactly vertical because
the user cannot enter a release angle..
MPACT Theory
65
MPACT accepts the consequence effect zones as input. In the product the user can select a
parameter to determine if these effects are to be calculated at the centreline height or at a
specified height. MPACT will accept effect zones calculated with either option.
Depends
on Wind
Speed
Depends
on
Stability
Shape used for the effect zone
Notes
One circle
Immediate
Explosion
Horizontal
Jet Fire
Vertical Jet
Fire
Immediate
Pool Fire
Residual
Pool Fire
Flash Fire
(Ellipse
method)
Two circles
A series of ellipses of different

lethality
lethality s
lethality
lethality
Full or half ellipse (and MPACT
pieces together a composite cloud
to develop a time history)
Not normally sensitive to weather

conditions but humidity and pressure
may have an influence
Only initial flash fraction will affect
immediate explosion
Varies by speed only
Delayed
Explosion
(Weikema
method)
Flash Fire
(cloud
method)
Delayed
Explosion
(Confined
method)
Two circles
MPACT)
Boundaries of the cloud defined

by cloud coordinates
Overpressures calculated from

explosions in the confined
volumes of obstructed regions
affected by the cloud
Type
BLEVE/
fireball
(calculated
in

Flash fire ellipses are derived from
the LFL fraction boundary so depend
on the exact dispersion results but
dont represent the cloud shape
directly
Flammable mass and location are
derived from the dispersion results
Flash fire dimensions from the LFL

fraction boundary are used directly
without the use of ellipses
At each time step the interaction of
the cloud and the obstructed regions
is evaluated
Table 4 Relevance of Wind speed and Stability to Flammable Effects

Each type of flammable effect has a particular shape and these are explained in the following
Sections.xvii
xvii
EXTENSION. Instead of using ellipses, more accurate calculation of radiation contours could be applied
(e.g. by directly inputting radiation as function of x,y,z,t in MPACT).
MPACT Theory
5.5
66
Fireball/BLEVE (Single Zone Method)
The single zone method is the original method used in MPACT. The dimension used to
define a fireball/BLEVE is the radius to the radiation impact of concern. This is defined in
terms of Thermal Dose Units (J/m2)N, which is a user input set in the Parameter Data. The
exponent, N, depends on the N constant defined for flammable probit calculations. This
measure takes fireball duration into account in calculating the potential fatality effects. The
duration calculated by the fireball/BLEVE model is used in this calculation limited to a
maximum exposure time parameter.
Additional fatality effects due to BLEVE overpressure or vessel fragments are not modelled
within MPACTxviii
The zone is centred at the release point. Figure 26 illustrates the approach.
Radiation
Dose-Level for
Fatalities
Fireball or
BLEVE
Effect
Radius
Release
Source
Figure 26 - The Fatal Effect-Zone for a Fireball or BLEVE
xviii
EXTENSION. Add vessel fragment and overpressure effects for BLEVEs
MPACT Theory
5.6
67
Jet and Pool Fires (Single Zone Method)
The fatal effect-zone is modelled as an ellipse in the same way for both jet and pool fires.
Three dimensions describe the ellipse as illustrated in Figure 27 and Figure 28:
a, b
d
major and minor axes (a and b) of the ellipse

relative offset of the ellipse centre from the release point defined as the ratio x/a where
x is the distance from the release point to the ellipse centre.
Thus for an ellipse centred at the release point d=0. For an ellipse with the effect zone starting
at the same point as the release d=1. Pool fires can be displaced from the release point
according to the wind speed and the rainout position. Jet fires can be displaced too, again
because of the effect of wind speed and also elevation of the flame. When the effect zone is
displaced from the release point d>1.
The radiation level set in the Flammable>Flammable Risk Parameters defines the fatal effect
zone. Unlike the fireball/BLEVE calculation the radiation dose effect is not modelled. The
assumption is that people exposed to the radiation will either escape within an initial period
(used to equate a dose to a particular radiation level)xix or will be exposed for the full duration
of the flame. Fatality effects are not reduced for flame durations less than this initial period.
For jet fires there are potentially different results for horizontal and vertical flames for the
same release event as described in Section 5.2. This functionality is historic and retained for
compatibility with old versions. It allowed for multiple jet fire directions when the dispersion
model could accept only horizontal releasesxx. Each jet fire has a set of ellipse parameters a, b
and d.
Pool fire sizes differ according to the time of ignition. Potentially the size of the pool will
vary significantly with time. However, to simplify the calculations only two representative
pool sizes are considered for the risk modelling corresponding to immediate ignition and
delayed ignitionxxi. The immediate ignition pool is sized according to the balance between the
burn rate and the liquid input to the pool. The delayed ignition pool size is used for delayed
ignition event tree branches and the residual pool outcomes. This size is calculated by finding
the maximum spread of the evaporating (rather than burning) pool. In both cases the pool size
is limited by the bund if present.xxii
xix
DOC. The default parameter set uses a period of 20s to equate a dose of 250KJ and a radiation level of
12.5kWm-2.
xx
EXTENSION. Rationalise the historic behaviours
xxi
EXTENSION. Using the pool size relationship with time is a possible refinement
xxii
DOC. A detailed description of this is (or will be) included in the PHAST model-linking document. Any
future changes to the PHAST logic, will thus effect the MPACT results. Introducing time-dependent pools,
will also involve a change in PHAST (and modified transient input to MPACT).
MPACT Theory
68
Radiation Level
for Fatalities
Jet Flame
b
Release
Source
Figure 27 - The Fatal Effect-Zone for a Jet Fire
R adiation Level
for Fatalities
Pool Fire
Pool
Centre
Side View of the Pool Fire
W ind
Figure 28 - The Fatal Effect-Zone for a Pool Fire
MPACT Theory
5.7
69
BLEVE, Jet and Pool Fires (Flammable Probit Method)
This is a method based on the concept of applying the probit approach to flammable
consequences. In this method a dose is derived based on the radiation level and duration of
the exposure. Then the dose is converted to a probability of death using the probit function.
These concepts are explained in detail elsewhere for example the Green Book10 and the
implementation in the PHAST product is described in the DNV documentation on RADS11.
How the method can be applied to risk analysis is described in the Purple Book9 and figure
5.4 from this document is shown below;
BLEVE
pool fire
jet fire
in flame
envelope
yes
P = 1
FE,in = 1
FE,out = 1
no
Q 35 kW/m
no
PE = f(Q,t)
FE,in = 0
FE,out = 0.14 x PE
yes
P = 1
FE,in = 1
FE,out = 1
Figure 29 (Figure 5.4 from the Purple Book) Calculation of the probability of death,
PE, where the respective fractions of the population dying indoors and outdoors are FE,in
and FE,out, for exposure to a BLEVE, pool fire and jet fire. The probit function for heat
radiation is f(Q,t)xxiii.
The risk is calculated from the probability of death (PE) and the vulnerability factors (FE)
which vary by type of risk and radiation level. The logic behind this approach is that if the
intensity is high enough it will cause combustion, setting clothes and buildings alight.
The resulting variation of the probability of death may be viewed as a function of radiation
level and exposure time as shown. There is a step change at 35kW/m2 where the lethality
jumps to 1 for all exposure times. Figure 30 illustrates the behaviour for exposure times of
20,15,10 and 5s and uses a protection factor of 1. If a protection factor had been used then the
curves will be reduced up to 35kW/m2.
xxiii
The in flame envelope logic is extended in the case of a fireball to include locations directly below the
flame radius
MPACT Theory
70
Probability of Death P E
0.8
0.6
20s
15s
10s
5s
0.4
0.2
0
7500
12500
17500
22500
27500
32500
37500
Radiation/(W/m2)
Figure 30 Probability of Death as a Function of Radiation and Exposure Time

(Based on the Purple Book probit constants a= -36.38, b=2.56, n=4/3)
The exposure time is also limited to a maximum parameter value (default=20s) and this is
necessary to avoid the situation where the dose is high because of very long exposure time.
The assumption is that at lower radiation levels then escape or shelter is possible.
To apply this method in MPACT a number of lethality levels, n, are defined starting with the
outer limit of the hazard zone. The default value for this outer limit is 0.01 corresponding to
the requirements of the Purple Book. This value is used to derive the corresponding radiation
level given the duration of the flame, the maximum exposure time parameter and the
flammable probit constants. This radiation level is then used to calculate the ellipse
dimensions corresponding to this lethality level.
Then a number of additional lethality levels may be defined and these are used in the same
way to calculate ellipses corresponding to each level. These intermediate levels are optional
and there will be a compromise between the number of them, calculation time and precision
of the calculations. The number is set using a parameter Number of representative ellipses.
There will be n-2 intermediate ellipses. The target probability of death for ellipse i may be
described byxxiv;
ni
Pd , i = Pd , n +
(Pd ,1 Pd , n )
n 1
xxiv
( 49 )
In the product there are options to space these ellipses according to dose, probit, view factor or lethality.
The sensitivity testing showed that spacing according to probit gave the optimum results.
MPACT Theory
71
The central region of the effect zone has a prescribed lethality and it is defined using radiation
intensity rather than using the flammable probit relationship. The default values are 35kWm-2
with a lethality of 1 correspond to the Purple Book. If the flame emissive power is less than
this threshold value (possible for some pool fires) then the target radiation level is set to 95%
of the flame emissive power and this method ensures that the in flame envelope criterion is
met for all flames.
In addition, there is additional logic that extends the in-flame criterion specifically for
fireballs. Below the flame the minimum radius of the central zone is set to the fireball
radius. This applies the static fireball model better in a risk context since in reality the
fireball starts at the vessel height and then rises.
The inner zone probability of death Pd,1 is calculated from the flammable probit function
rather than the assumed probability of death of 1. Should the inner zone be absent then the
maximum lethality on the target plane is used to set the value of Pd,1.
Average lethality = 0.255

0.01
0.5
1.0
Lethality ellipses
Figure 31 Representation of the Radiation Hazard Zone Using Flammable Probits

The figure illustrates the modelling concept with one intermediate lethality level to give a
total of 3 ellipses.
This information effectively defines the location of contours of a particular lethality. In order
to calculate risk we must consider what lethality to apply in the intermediate zone between
ellipse boundaries. The lethality vs radiation curves in Figure 30 demonstrate a tendency
towards an S-shape curve. This is further increased when considering lethality vs distance
which is the most important factor when calculating the risk. This again is discussed in the
MPACT Theory
72
Q (kW/m2)
Purple Book and Figure 5.A.1 is shown here to illustrate the function of lethality with
distance for a BLEVE calculation;
50
1.0
40
0.8
30
0.6
20
0.4
10
0.2
Q
P_E
F_E,out
F_E,in
0.0
0
100
200
300
400
500
600
distance (m)
Figure 32 (Figure 5.A.1 from the Purple Book)

Example calculation of a BLEVE of
100 tonne propane (burst pressure 15 bar). The exposure time is set at the maximum
value of 20 s. Indicated are heat radiation, Q, probability of death, PE, fraction of people
dying outdoors, FE,out, and fraction of people dying indoors, FE,in.
As with any numerical method the solution should converge to the correct result if a
sufficient number of discrete intervals are used. In practice it is desirable to use fewer
intervals to avoid excessive run times. Different interpolation methods for the value used
between ellipses can give results closer to the correct result with fewer discrete intervals. The
sensitivity study showed that using a uniform population density12 3 intervals were required
to obtain an error less than 1% using weighted mean interpolation as described below;
A parameter Lethality interpolation method determines the way the lethality between
contours is calculated. Figure 33 compares the methods which are defined as;
MPACT Theory
73
0.9
0.8
Probability of Death P E
0.7
0.6
20s
None
0.5
Linear
Weighted
0.4
0.3
0.2
0.1
0
7500
12500
17500
22500
27500
32500
37500
Radiation/(W/m2)
Figure 33 Comparison of different interpolation methods with lethality curve

None
Using this method the lethality for the ellipse is used throughout the region between itself and
the next highest lethality ellipse. This method avoids overstating the risk near the extremities
of the effect zone, particularly given the tail of the curve. However, the risk in general will
tend to be underestimated.
Arithmetic mean
This method uses the arithmetic mean between the lethality levels for the enclosing ellipses.
This method is simple to understand and if anything will tend to overstate the risk in the outer
regions and underestimate nearer the centre. The reason for this is that in reality the lower
lethality levels will cover a greater area both because of the S-shaped nature of the curve and
also the geometric effect of the lethality at the outer reaches of the effect zone covering a
larger area.
Weighted average
This method assumes a knowledge of the shape of the lethality curve as illustrated by Figure
5.A.1 in the Purple Book. There is an S-shaped curve with the gradient declining towards the
low-lethality tail. The shape of this curve and the geometry of the effect zone means that the
linear average will tend towards a significant overprediction near the edges of the effect zone.
On the other hand the arithmetic mean method will tend to underpredict where lethality is
>0.5. To overcome this limitation the average used is weighted 2/3rds to the lower value
below 0.5 and 2/3rds to the upper value above 0.5. In deriving these values harmonic and
geometric means were also tried and compared with the Purple Book example. Both these
means overcorrected the results.
How this type of effect zone is used to calculate individual and societal risk is explained in
later Sections.
MPACT Theory
5.8
74
BLEVE, Jet and Pool Fires (Radiation Intensity Method)
This method is similar to the radiation probit method in that is works with a collection of
ellipses. These ellipses represent different radiation intensity levels and are input to the Mpact
model together with information about the duration of each of the fires.
For a given building type the user must chose between the radiation probit method or the
intensity method for all fire types.
There is no interpolation with this method so that the vulnerability is represented in steps
between the ellipses. The vulnerability levels to use are calculated in the Vuln model13. The
model does allow for different vulnerabilities for individual risk vs societal results.
The fire durations are calculated in Mpact according to fire type;
Fireballs there is no minimum duration so that the radiation intensity vulnerability will
always be applied
Jet fires the release duration is applied
Early Pool fires these are used in conjunction with jet fires. The duration is at least as long
as the release duration as for the jet fire. It may also last longer if the burn rate is less than the
rainout rate. So the duration is taken as the maximum of (release duration, mass spilt/early
pool fire burn rate).
Immediate Pool fires these are used in conjunction with immediate ignition effects for
instantaneous releases and short duration continuous releases. For short duration releases the
duration is at least as long as the release duration. It may also last longer if the burn rate is
less than the rainout rate. So the duration is taken as the maximum of (release duration, mass
spilt/immediate pool fire burn rate). For instantaneous releases the duration is mass
spilt/immediate pool fire burn rate.
Late Pool fires these are used in conjunction with delayed flash fires and explosions. In
principle the late pool fire duration will depend on the ignition time, however, a single
conservative duration is used. This is consistent with the approach of using a single pool fire
effect to represent all possible ignition times. The duration used is at least as long as the
release duration. It may also last longer if the burn rate is less than the rainout rate. So the
duration is taken as the maximum of (release duration, mass spilt/late pool fire burn rate).
5.9
Vapour-Cloud Flash Fire - Instantaneous Release
Flash fires are treated in different ways depending on the type of release. One of the simpler
forms is that used for instantaneous releases.
Instantaneous releases are presented as a circular cloud indicating the radius of the LFL
fraction to finish. For the UDM this representation is valid because the model assumes the
same cloud dimension in all directions. For the Building Wake Dispersion Model the cloud
MPACT Theory
75
has different dimensions in the crosswind and along-wind directions. The radius of the circle
used by MPACT is calculated by taking the larger of the cross and downwind dimensions.xxv
The averaging time used to calculate the concentrations is the flammable averaging time
(typically 18.75s).
The circle starts centred at the release point and then proceeds to drift downwind as shown in
Figure 34. The flash fire description therefore gives the size and downwind position of the
cloud at several time-steps during the time when it is developing to its fullest extent.
The full description for each time-step includes: the distance the centre of the cloud has
travelled downwind, the radius to the cloud-limit, and the flammable mass of the cloud. (But
note that the flammable mass is used by the explosion modelling, not the flash fire
modelling).
This description is also used once the cloud has transitioned to essentially instantaneous.
Edge of the Cloud

Concentration is LFL Fraction
to Finish set in Parameter Data.
May be lower than the LFL.
Release
Point
Radius to
Cloud-Limit
Downwind
Distance
Figure 34 - The Flammable Zone for an Instantaneous Release
xxv
EXTENSION. This to be extended to allow for BWM profiles (top-hat building-wake profile, and
subsequent non-circular, ellipsoid shape of cloud horizontal cross-section.
MPACT Theory
76
5.10 Vapour-Cloud Flash Fire Continuous, Finite Duration and TimeVarying Releases
Continuous Dispersion (Indefinite Release Duration) (ellipse method)
For continuous releases the flash fire effect zone is taken to be the cloud boundary to the LFL
fraction represented as an ellipse. There is also the possibility that the ellipse is defined as a
half-ellipse rather than the full shape. This approximation is made to economise on storage
space and processor time. The ellipse parameters are defined per segment. For a steady rate
release without rainout there will be just one ellipse. For multi-segment releases (either multirate and/or rainout cases) there will be one set of ellipse parameters per segment.
Note that this elliptical definition at this stage is being used to represent the fully developed
cloud and assuming indefinite duration. So far there is no consideration of the duration of
each segment this description follows later. Figure 35 shows an example where the LFL
fraction boundary is described by a half-ellipse.
LFL Fraction
Boundary
Half Ellipse used by

MPACT
Figure 35 Example of a UDM Cloud represented by a Half-Ellipse
The solid shape indicates the cloud boundary as calculated by the UDM. The half-ellipse
used to approximate this boundary is shown as an outline. The half-ellipse shape is used by
MPACT and not the cloud boundaryxxvi. There is no consideration of any zone in the cloud
that might be above the UFL concentration and all population within the ellipse will be
considered to be affected in a uniform manner.
xxvi
EXTENSION. The model could be enhanced to use the modelled cloud shape rather than approximate
the shape using ellipses.
MPACT Theory
77
The half or full ellipse dimensions passed to MPACT are calculated according to the
following procedure;
Two parameters are used to define the UDM cloud shape; the downwind cloud length LLFL
and the cloud area ALFL within the boundary defined by the LFL fraction. Then
a) A full approximate ellipse approximation is applied. Here the horizontal and vertical
ellipse semi-axis lengths a, b are set using: (i) same flammable length allowing a to be
defined according to 2a = LLFL and (ii) same flammable area ab = ALFL so that b is
also defined directly from the footprint of the flammable cloud.
b) A full approximate semi-ellipse approximation is applied, where the horizontal and
vertical ellipse semi-axis lengths a, b are set using: (i) same flammable length so that a
= LLFL and (ii) same flammable area 0.5ab = ALFL again so that b is defined directly.
Subsequently a 2 least-square fit is used to determine whether the half or full ellipse best
represents the cloud boundary. The comparison is made between the ellipse and the cloud
by comparing widths at each result step in the region defined by the ellipse.
The shape with the best fit is passed to MPACT. Note that the method ensures that
whichever shape is chosen, the ellipse will have the same downwind length and area as the
modelled cloud. Figure 36 shows a real example where the full ellipse best represents the
cloud.
LFL Fraction
Boundary
Ellipse used by
MPACT
Figure 36 - Example of a UDM Cloud represented by a Full-Ellipse
MPACT Theory
78
The location downwind of the half or full elliptical shape is defined by the displacement
dimension d as for the other flammable zones.
Finite Duration Release (ellipse method)
So far the cloud boundary has been defined according to a release of infinite duration.
MPACT applies further logic to determine the dimensions of the cloud at any specific time. It
will use these to determine not only the flash fire outcome but also the range of ignition
sources that can ignite the cloud at that time.
To achieve this it will determine the position of the leading and trailing edges of the cloud (xf
and xb respectively) at different timesxxvii since the start of the release. This way it can derive
the cloud length xl and determine its position downwind at any specific time.
The approach is illustrated in Figure 37 for a cloud developing with time since the start of the
release:
-
At time t1 the front of the cloud had reached point xf while the back edge is still
attached to the source that is still active and supplying new material. At this time
MPACT will truncate the ellipse at xf as illustrated by the shading. (Note that it will
also use this flash fire calculation to determine the explosion location but this is not
discussed further at this stage).
At time t2 the release has stopped and the back edge of the cloud has moved away
from the start of the ellipse. The front of the cloud is still developing and xf has moved
further downwind since t1. There are now two truncation lines for the shaded
flammable effect zone.
At time t3 the front edge of the cloud had diluted below the LFL fraction so the front
edge of the flammable zone has become static at the furthest reach of the ellipse. The
back edge has moved further downwind since t2.
Eventually the back edge will reach the furthest extent of the ellipse and there will no
longer be the possibility of ignition or a flash fire (This corresponds to the condition xf
= xb).
xxvii
DOC. The times used are derived from the cloud dispersion information provided as input to MPACT.
For a single segment this is simply the time for each step downwind.
MPACT Theory
79
Time
xf
xl
t3
xf
xb
xl
t2
xb
xf
xl
t1
xb
xf
x Direction
Figure 37 - The Truncation of a Short-Duration Release to Define the Flash Fire Effect Zone
MPACT Theory
80
To explain the method further, Figure 38 shows the movement of the front and back edges
of the cloud with time. There is another line indicating the length of the cloud xl as it
changes with time. The length increases initially, reaches a plateau and then decreases.
Release Duration
Downwind Distance, x
xlfl
xb
xf
Cloud length
xl
t=0
tdur
Time, t
Figure 38 Behaviour of Front and Back of the Cloud with Time
Clearly it is important for the model to define the back and front edges of the cloud that
correspond to the flammable zone. The mixing at the upwind and downwind edges (alongwind diffusion) is not modelled in the current UDM so these boundaries correspond to all
concentration levels and not just the fraction to the LFL boundary. For a given time there
is one position defining the front edge of the cloud, whatever the concentration and this
position is derived from the dispersion modelling results giving the relationship of time
against distance. However, by limiting the development of xf and xb to the full extent of the
ellipse the upwind and downwind boundaries are being restrained to the flammable zone
consistent with the limitations of the dispersion modelxxviii. Note that in Figure 38 straight
lines are used to represent the movement of the cloud boundaries. Downwind speed is not
necessarily constant so that with actual dispersion results these lines will normally be
parallel curves.
Time Varying Dispersion
Time varying dispersion is modelled by using multiple constant rate dispersion calculations
with different source conditions for a single release event. This can occur in two situations
1)
2)
The user defines a time-varying release, and

The release produces a liquid pool which re-evaporates.
xxviii
EXTENSION. Along-wind diffusion modelling in the UDM would make the definition of the front and
back of the cloud more realistic.
MPACT Theory
81
As described in detail in the UDM theory manual8, the UDM will model the time varying
release by calculating a number of segments each of which assume a constant release rate.
Using this method along-wind diffusion (mixing in downwind direction) is neglected. The
elliptical shape (half or full) is defined for each segment and MPACT takes each of these
constant-rate segments and pieces them back together to create a view of the cloud at any
specific time.
The result of assembling such a cloud from the segments give a cloud shape made up of a
series of truncated ellipses as shown in Figure 39.
Cloud Limit
A series of truncated ellipses
Release
Source
Figure 39 - The Cloud Shape for Multiple Release-Rate Segments at a Given Time
As shown in Figure 38 the location of the truncation lines between segments depends on
the distance travelled by the cloud since the various transition times. Since in the
dispersion model segments can move with different speeds non-physical gaps between
segments can occur. This discontinuity between segments is caused by the differences in
downwind velocity between segments. This causes a difference between the distance
reached downwind over the same time period for the upwind edge of the downwind
segment and downwind edge of the upwind segment.
MPACT ensures there are no gaps by setting the distance for the upwind edge of the
downwind segment to be equal to the downwind edge of the upwind segment. This
approach is shown in Figure 40 for a simple two segment example. The truncation line
between segments is taken to be the downwind edge of the second segment, xf2. This way
the potential gap is filled.
MPACT Theory
82
Gap
Segment 1 ellipse
xf1
xb2
xf2
xb1
Segment 2 ellipse
x Direction
Figure 40 Flash Fire Zone for a Two Segment Cloud at a Given Time
Again note that the view of the cloud at a specific time is used to calculate the location of
the centre of the subsequent explosion calculations. There will be multiple outcomes (both
flash fire and explosion) corresponding to all represented times and to do this MPACT
starts with the first step for the first segment. For each subsequent time step it searches
through all the segments for the next representative time in the sequencexxix.
xxix
IMPROVE This method makes no attempt to distribute outcomes with time in an optimal fashion. The
time steps used are those falling out of the segment loop. Some steps might be large while others may be
small. Control of the time steps used for delayed ignition would speed the calculations and allow the user to
optimise calculation time in the most important modelling areas.
MPACT Theory
83
Effect of release elevation on location of flash fire effect zone

The impact calculations assume that all of the people and ignition sources are located at
ground-level, so the risk is based on the size and shape of the effect-zones at ground level. If
the cloud is released an elevated height, the cloud shape passed to the impact calculations
may be displaced from the release source as shown in Figure 41.
SIDE VIEW OF ELEVATED RELEASE
Release
Source
AERIAL VIEW OF GROUND-LEVEL EFFECT AREA
Release
Source
Effect Area
Figure 41 - The Displaced Effect-Zone for an Elevated Release
As in the case of jet and pool fires this situation is represented by ellipses where d>1.
MPACT Theory
5.11
84
Vapour-Cloud Explosion (Wiekema Purple Book method)xxx
Unlike the fire effects described above, in this method explosions are modelled with two
fatal effect-zones rather than a single zone. As shown in Figure 42, these effect-zones are
concentric circles, and the centre of the explosion is taken as the centre of the explosive
cloud. The two effect zones correspond to two different explosion damage levels.
Cloud
Limit
Cloud
Centre
Release
Source
R1
Explosion Damage
Level 1
R2
Explosion Damage
Level 2
Figure 42 - The Effect-Zones for a Vapour Cloud Explosion
Explosion Model Correlation

The calculation of the damage zones is based on the TNO vapour cloud explosion correlation
model (Wiekema 1979)14. The "energy content" of the explosive mass (Eexpl) is calculated as:
E exp l = M c H comb
( 50 )
Where;
Mc
Hcomb
Cloud Flammable Mass (kg)xxxi

Heat of combustion for the flammable material in the cloud (J/kg)
A test is made at this point to determine whether the value of Eexpl is within the limits of the
correlation. If Eexpl < 5 x 109 J (the minimum explosion energy parameter) a blast circle of
radius zero is returnedxxxii. In effect this means that R1 and R2 radii below 23.8m and 47.6m
respectively are ignored.
xxx
EXTENSION. This could be extended to allow the use of the later PHAST explosion models (and/or
further improvements of these). These explosion models provide overpressure levels and impulse or impulse
duration as a function of radius from the cloud centre.
xxxi
The flammable mass for explosion includes mass above the LFL concentration, and (depending on
MPACT parameter) may exclude mass above UFL concentration.
xxxii
EXTENSION. Thus a minimum amount of energy is assumed to be required to cause an explosion. A
warning may be added to MPACT in case the explosion energy Eexpl > 5 x 1012, since this is above the
applicability range of the model.
MPACT Theory
85
The radii of the blast circles are then given by:

R = C s (E exp l )
1/ 3
( 51 )
Where:
Cs
The fraction of the energy content that contributes to the blastxxxiii

A constant whose value defines the damage level concerned. Default values
arexxxiv
Cs = 0.03 - heavy damage to buildings (R1) corresponding to
Purple Book overpressure level 0.3 barg
Cs = 0.06 - repairable damage to buildings (R2) corresponding to
Purple Book overpressure level 0.1 barg
The above values are used to calculate the two radii corresponding to the two concentric
circles R1 and R2:
-
R1: Inner Zone, with High Fatalities. This is typically set at the level to cause
heavy building damage.
R2: Outer zone, with Lower Fatalities. This is typically set at the level that causes
repairable building damage.
The vulnerability within each of these zones is set in the General Risk>Vulnerability
parameters.
Centre of the Explosive Cloud
The explosion location depends on the time of the ignition and how far the cloud has
travelled at that point. Examples of how MPACT views the cloud are given in Figure 37 to
Figure 42. The explosive centre of the cloud is calculated between the leading and trailing
edges of the cloud and can include multiple segments as described in Section 5.10.
The centre is weighted according to the flammable mass distribution in the cloud at the
time of the explosion. For each segment at a given time there is a flammable mass and
centre. The segment centre xseg is defined as the arithmetic average mean of the front and
back edges of the segment (after adjustment for any gaps);
x seg =
(x
+ xb )
( 52 )
The flammable mass for the total cloud Mc is calculated from the individual segment
masses Mcseg.
In MPACT, is the parameter input explosion efficiency and can be set according to the material
properties.
xxxiv
In MPACT, the two values of Cs are taken from the parameters Blast Damage Coefficient: Heavy Damage
and Blast Damage Coefficient: Repairable Damage and can be adapted to represent different effect levels.
xxxiii
MPACT Theory
86
Mc =
Mc
seg
( 53 )
All Segments
The cloud explosion centre xcentre is then calculated as a mass-weighted average.
x centre =
All Segments
x seg M c
seg
( 54 )
Mc
As implied by the mathematics any segments that have a zero flammable mass at the time
of the calculation dont have any effect on the result.
Since the flash fire envelope is based on the LFL fraction concentration and the explosive
mass is always based on the LFL concentration then the flammable mass weighting will
tend to give an explosive centre nearer the source than the geometric centre of the cloud as
the LFL fraction is reduced from a value of 1.0.
Another factor influencing this calculation will be the setting of the flammable mass
calculations and whether they include or discount parts of the cloud with a concentration
above the UFL.
Flash Fire with Explosion
The Wiekema model takes no account of the role of congestion on the potential to develop
a vapour cloud explosion. To overcome this limitation the event trees for delayed ignition
have both flash fire and explosion outcomes and it is assumed that delayed ignition will
lead to a fixed proportion of each type of outcome. This approach was developed on the
basis that sometimes the ignition will lead to an explosion and sometimes there will only
be a flash fire. This reasoning makes the results useful for QRA purposes and this method
is supported by the Purple Book.
5.12
Vapour-Cloud Explosion (3D Purple Book method)
The 3D methods described in section 5.13 can be configured to achieve a very similar
effect to the Wiekema Purple Book method and without the need to input specific
obstructed regions. It uses the Multi Energy model and the actual cloud shape derived
from the dispersion results rather than equivalent ellipses. The Multi Energy model has an
unobstructed explosion source and this can be used together with an efficiency factor in
such a way to represent a cloud where a fixed proportion of the cloud is ignited within
obstructed regions. The discussion that follows is oriented towards obtaining similar
results between versions of SAFETI NL. Users of Phast Risk have more flexibility13, for
instance a greater number of lethality levels could be defined.
MPACT Theory
87
The Multi Energy model can be used without defining specific obstructed regions by
assuming that a constant fraction of the whole cloud will be obstructed. To obtain similar
results to the Wiekema model the following settings are made;
The Multi Energy curve used is 10 (Unobstructed strength parameter value 10)
12.5% of the flammable mass is considered to be in obstructed regions (12.5%
Unobstructed explosion efficiency)
Heavy explosion damage corresponds to an overpressure of 0.3 bar (Indoor and
outdoor fatality = 1)
Lighter explosion damage corresponds to an overpressure of 0.1 bar (Indoor
fatality only 0.025)
100
Scaled peak side-on overpressure Ps'
10
0.3
1
0.1
0.1
0.01
0.001
0.1
10
100
Scaled distance r'
Figure 43 Use of Multi Energy Curve 10 to obtain radii R1 and R2
Reading from Multi Energy curve 10 gives values of 1.3 and 2.8 respectively for scaled
distances r1 and r2. This gives a ratio between the heavy and lighter damage radii of 2.16
by using 0.3 and 0.1 bar as the overpressure criteria. This compares with the ratio of 2
MPACT Theory
88
using the previous method. To derive an obstruction factor, , the expression for the
Wiekema radius may be substituted in the Multi-energy equation for scaled distance.
R = C s (E exp l )
1/ 3
( 56 )
r =
Pa
1
3
= Pa Cs r
( 55 )
( 57 )
Given a Wiekema explosion efficiency of 10% and atmospheric pressure of 101550 Pa

efficiencies for R1 and R2 may be calculated as 12.5% and 10% respectively. In order to
match the dimension of the heavy damage zone and be slightly conservative for the lighter
damage zone the 12.5% efficiency is recommended.
Where the two methods differ significantly is the treatment of the cloud shape and
combination of the flash fire effect with the overpressure effects. The Wiekema method
depends on ellipses to represent the cloud shape while the 3D method uses the LFL
envelope directly. This means that the risk model knows how far the cloud travels along
the centreline. In the 2D method it knows only how far the ellipses do downwind at the
flammable effect height. So the free-field decision algorithm can use the cloud centreline
distance to determine whether the cloud crosses the plant boundary at any height.
If the explosion fails to generate overpressures because the energy available is below the
minimum parameter value then in the 2D method there is are limited mechanisms to
ensure that the flash fire zone still does generate risk.
Type of risk 2D
3D
result
Risk grid
Explosion effect only, no effect if below
minimum explosion energy
Risk ranking Combination of explosion, flash Fire and pool
points
fire effects
Combination of
Societal risk Maximum value of N corresponding to either explosion, flash Fire and
pool fire effects
grid
the Flash and pool fire combined effects or the
N corresponding to the Explosion effects
Societal risk No risk
buildings
Table 5 Treatment of the Explosion Combined with the Flash Fire and Pool Fire
Effects
MPACT Theory
89
5.13 Vapour Cloud Explosions (Confined Explosion Method Using 3D

Cloud Views)
5.13.1 Introduction
Explosion models that take into account the importance of the role of obstructions on
flame acceleration in vapour cloud explosions and subsequent generation of overpressure
have been developed over the last decade or more. The Multi-Energy model as described
in the Yellow Book15 is one. The Baker-Strehlow model16 is another and both these
models are available in the PHAST product for consequence analysis.
These models have proved difficult to apply to envisaged accident scenarios. This lead to
further research into how to apply the Multi-Energy model to produce guidance on how to
apply the Multi Energy model to real plant. This work is published in the Yellow Book17
and in TNO reports (GAME 199518 and GAMES 199819).
In spite of these challenges these models have been applied for risk applications. This
approach has been taken by Woodward and Crossthwaite20 (1995) and more recently by
Baker Engineering on behalf of the BEAST consortiumxxxv. The overpressure and
impulse information given by both the Multi Energy and the Baker Strehlow models can
be used to analyse potential building damage. This gives valuable information that can be
used to site buildings and design for levels of overpressure with an acceptably low enough
risk. QRA results can also be obtained by applying a vulnerability model to convert the
results of the explosion calculations into levels of harm to individuals.
From version 6.6 extended explosion models in Phast Risk to provide explosion
calculations based on the Multi-Energy model to calculate fatalities based on explosion
results may be called by MPACT to obtain probabilities of death. An overview is provided
here and the models are described in further detail in the theory documents for the
explosion model, OREM21 and the model that converts the explosion results into a
probability of death VULN22.
5.13.2 Overview
The main principle behind this type of explosion method is that the vapour cloud damage
will depend on its position with respect to obstructed regions. If the cloud is away from
obstructed regions then the potential for turbulence-generated flame acceleration is not
present and overpressures and therefore damage from the blast will be limited. The vapour
cloud will still represent a flash fire hazard even if the blast damage is limited so using this
method the flash fire and the blast are treated together as a combined outcome. Because of
the combination the conditional probability of this outcome is taken as the sum of the flash
fire and explosion branches in the event tree described in Section 5.2.
If the cloud disperses through part of the plant that contains obstructed regions then the
potential for flame acceleration exists and significant blast damage is likely as well as the
flash fire hazard. Again the blast damage and the flash fire are treated as a combined
outcome.
xxxv
The BEAST methodology is not published in the open literature. However, some information is
available on the Baker Engineering website.
MPACT Theory
90
A very important aspect of this approach is the use of a 3 dimensional view of the cloud.
This is required in order to calculate the intersection of the cloud with the obstructed
regions. This extension of the model avoids the use of approximate ellipses and instead
uses the calculated shape of the cloud directly. There are further benefits of using the 3D
view of the cloud, for instance the height of ignition sources may be considered in relation
to the cloud position. This was not possible in the case of the previous method.
A further hazard is considered by MPACT if there is a flammable pool present. If so then
MPACT includes the pool radiation hazard in the combined outcome. To illustrate the
concept Figure 44 shows an example explosion.
Obstructed
Region
Blast Zone
C
B
Pool Fire
Vapour
Cloud
Figure 44 Overview of a Vapour Cloud Explosion from an Obstructed Region
The individual risk from the explosion will depend on the location of the observer as
represented by the letters A, B, C, D and E. The extent of each hazard zone, vapour cloud,
pool fire and blast may be represented by the boundary of the zone defined by the
minimum probability of death. In this figure ellipses and circles are used to represent this
boundary but the actual shape of the zone is represented in the model.
A
At this position the observer is outside the vapour cloud. The only hazard is the blast wave
generated by the explosion centred in the obstructed regionxxxvi. The probability of death at
this location is determined by the strength of the blast wave and how this converts to
probability of death.
xxxvi
For simplicity we are ignoring the much weaker blast wave from the unconfined part of the vapour
cloud at this stage
MPACT Theory
91
B
This location is within the vapour cloud only so the probability of death depends on this
hazard only. Typically for outdoor risk this probability of death is set to one for flash fire.
C
This location is within the vapour cloud and the pool fire hazard zone. The probability of
death will be defined by the successive probability of surviving each hazard.
D
This location is in the obstructed region. It is subject to the combined effects of the flash
fire and blast. Again, the probability of death will be defined by the successive probability
of surviving each hazard. Normally in this case the probability of death will be one.
E
So far we have made no mention of the distinction between risk indoors vs outdoors. At E
we conceive a building both within the blast region and the vapour cloud. The probability
of death within the building will depend on the vulnerability of the building in the first
instance and then given a level of damage how vulnerable the occupants are to the
combined hazards.
This example has used a relatively simple cloud and configuration to illustrate one
possible ignition scenario. In practice there are many possible cloud configurations and
details to consider. These are covered in further detail in the separate documentation for
the OREM model21.
Thus the model allows the calculation of the probability of death throughout the hazard
zone for the combined outcome and MPACT can calculate individual and societal risk as
explained further in this document. Typically MPACT will perform many of these
calculations to represent ignition at different times to arrive at the total risk for a given
release.
5.13.3 Immediate Ignition Outcomes
This type of outcome is associated with instantaneous and short-duration releases only. It
permits an ignition probability to be prescribed by the user rather than to be calculated by
the delayed ignition model. It still requires an ignition time to be deduced. The assumption
behind this ignition time will determine the magnitude of the outcome.
The assumption made in MPACT is based on the concept that the direct ignition source is
at the release point. Then the time depends on the type of release; if the release is
instantaneous then it corresponds to the maximum cloud volume during the period from
release up to the point when the cloud boundary leaves the release point; for continuous
releases the time corresponds to the duration of the release.
When the OREM explosion model is being used then the event tree probability for delayed
ignition will be taken as the sum for the flash fire and explosion branches since this
outcome combines the explosion and flash fire hazards.
MPACT Theory
92
5.13.4 Delayed Ignition Outcomes
Typically MPACT will model many of these outcomes depending on how the cloud
develops with time and encounters ignition sources. Figure 45 summarises the steps
followed to calculate the outcomes.
Determine time steps to be used

based on initial evaluation
Start with first time step
Evaluate the boundaries of the cloud

at the ignition time
Calculate ignition probability
Pdi>Pmin
No next time, t
Yes -Significant
probability
Calculate individual risk for the

combined outcome
GAMES Model
Calculate societal risk for the

Combined outcome
Next ignition time, t
Figure 45 Main Stages to calculate vapour cloud explosions using the obstructed region
explosion modelling
The main steps in MPACT to model delayed outcomes are as follows (using the
numbering from the figure);
1 Time Discretisation
This is an initialisation stage required to work out an efficient distribution of time steps on
the basis of the cloud dimensions, the strength of ignition sources and user-controlled
parameters defining how many calculations should be made. This step is necessary
because the time steps output from the dispersion model tend to start very small and
become much larger with time. This is not usually the most efficient way to treat time with
respect to risk. The method is described in Section 5.19 on the ignition model.
For each ignition time the following main stages are executed;
2 Evaluate the boundaries of the cloud at the given time

The time-varying behaviour of the cloud must be interpreted from the input data.
Adjustments are made to some clouds to make the cloud description amenable to the
explosion calculations21.
3 Calculate the ignition probability
MPACT Theory
93
The ignition probability is calculated from the ignition sources and population within the
flammable region of the cloud as described in Section 5.19.
Then if the ignition probability exceeds the minimum probability parameter the
calculations continue, otherwise the model moves to the next time step.
4 Calculate the individual risk

Overlap between the cloud and the confinement zones is necessary to calculate the inputs
to the explosion model. Volumes inside and outside the confinement zones and location
are necessary as well as the characteristics of the confinement zones. All this calculation is
performed outside MPACT. MPACT combines the probability of death from the explosion
with the other hazards and assigns the individual risk according to the theory described in
Section 5.16.
5 Calculate the societal risk
Again, the societal risk is calculated according to the methods described in Section 5.16.
MPACT Theory
5.14
94
Overlap of Effect Zone with the MPACT Risk Grid
All the flammable impacts are modelled by superimposing the appropriate effect zones for
the release (described in Section 5.4), onto the risk grid. Figure 46 shows an elliptical effectzone for a given lethality. In this case the ellipse is aligned with the grid but in general the
angle between the grid and the zone will depend on the wind sector angle.
Fatal-Effect Ellipse
Fraction of square x,y affected
Calculation Grid
Figure 46 - An Effect Ellipse Superimposed on the Calculation Grid
The first step MPACT takes is to determine the range of grid cells that might be covered by
the effect zone or zones (there may be combined effect zones to consider).
Normally only a fraction of the grid will be affected by a given outcome and calculation
times are reduced by this preliminary calculation. It compares the upper and lower
coordinates of the effect zone(s) with the risk grid boundaries to determine the range of grid
indices in the x and y directions that must be considered.
MPACT visits each cell in turn and works out whether each effect zone overlaps the cell to
calculate the fraction of each cell area Frx,y, ,Lio covered by the effect zone. If it does
completely then Frx,y, ,Lio = 1. If it misses completely then Frx,y, ,Lio = 0.
If the overlap is partial then it progressively divides the cell into 4 smaller calculation squares
and identifies which of these enclose part of the effect area, as shown in Figure 47. Then for
each calculation square where there is still partial overlap it repeats the subdivision as many
times as allowed by the input parameters.
MPACT Theory
95
The Smallest Square-Size

Set by Number of Times to
Sub-divide a Square in the
Parameters Data
Calculation Grid
Figure 47 - Sub-Dividing the Square to Calculate Frx,y, o
The size of the smallest allowed calculation square considered is;
x sub =
x grid
2
( 58 )
N sub 1
Where Nsub is the user-defined parameter the maximum times to sub-divide a square
(default value is 5 and note that 1 is subtracted from the parameter so that subdivision 1 is
equivalent to no subdivision at all). This gives the maximum number of calculations squares
in a risk grid cell as 4Nsub-1 = 44=256.
In practice the smallest subdivision value used within MPACT is limited by another
parameter nmaxellipse , which is the number of sub-squares across ellipse (default value
nmaxellipse=10). This is used to equate the width of one of the sub-squares to the smallest
dimension of the truncated ellipse (smallest value of length or width), i.e.
x sub
min Lellipse , W ellipse
n max
( 59 )
ellipse
The purpose of this additional criterion is to avoid the situation where excessive computation
time is spent without achieving significantly more accurate results. This is an approximate
calculation because the angle of the ellipse to the risk grid varies with wind sector and is not
taken into account in Equation ( 59 ) and the method offers a limited choice of allowed subsquare sizes because the size of the sub-square is a whole fraction of the MPACT grid cell
and not the effect zone itself.
These parameters are included in the MPACT data input. Higher values will increase the
accuracy of the calculations of Frx,y, ,Lio and increase the calculation time. At the smallest
MPACT Theory
96
subdivision level MPACT assigns Frx,y, ,Lio values to calculation squares according to Table
6.
Number of corners covered by the Effect Zone
1
2
3
Frx,y, o
0.125
0.5
0.875
Table 6 Fractions used for Calculation Squares with Partial Overlap

The calculations are repeated for each ellipse lethality level and for each flammable effect
to be combined. It is then helpful to define the fraction of the cell covered by the region
between ellipse lethality levels. To do this we make the assumption that the fraction
covered for a lower lethality level will always be higher than for a higher lethality level.
Frx , y , , Li o = Frx , y , , Li o Frx , y , , Li +1 o
( 60 )
After the calculation of all required values of Frx,y, ,Lio, MPACT proceeds with the
calculation of the risk increments for each flammable effect type.
5.15
Continuously Variable Pd Type of Effect Zone
Flammable impacts may not be constant within an effect zone. For instance the probability of
death due to an explosion may be modelled as a continuous function of overpressure (rather
than behaving as a step function). In this case the effect is evaluated at the cell centre for use
with grid results (individual risk and grid populations) and at the point for risk ranking point
results.
The effects could be integrated over each cell but with the associated computational cost. The
provision of this option is a potential enhancement item.
MPACT Theory
97
NE
NC
NW
New square 3
New square 4
Cell Centre
WC
EC
C
New square 1
SW
New square 2
SE
SC
Figure 48 Cell Corner Locations and Potential Subdivision
5.16
Calculation of Individual and Societal Risk
As discussed a flammable outcome o, is a combination of a weather category w (combination

of stability class c and range for windspeed u, wind direction s, and flammable effects . The
same general approach is used for all flammable outcomes.
Individual Risk
The risk increment is calculated at a specific point x, y as;
IRx, y,o = Fedf Po Pd , IR, x, y o
( 61 )
Societal Risk
The number of fatalities per grid cell with its centre at x, y is calculated as;
N x, y o = Nx, y Pd , FN , x, y o
( 62 )
The total number of fatalities given that the outcome occurs, Nedf|o, is equal to the sum of Nx,y|o
for all affected cells plus that total of people affected within specific building objects,
N Bi edf o , defined as;
N Bi edf
= PdBi FN o N Bi
( 63 )
MPACT Theory
98
N edf o =
x, y o
Bi
( 64 )
edf o
The probability of death, calculated for each cell and for each building is the key requirement
to complete these calculations. There are differences in detail depending on the exact position
in the event tree (one of types A to F). Three generic formulations are used;
1. Single flammable effect outcome
2. Two single flammable effect zones combined
3. Explosion, flash fire with or without pool fire
Method 1 is used for independent fires (radiation or flash fire hazards) and the dual effect
zone explosions (Purple Book method). Method 2 is used for a combination of fires (two
radiation hazards or a flash fire and radiation hazard) and method 3 is used for the obstructed
explosion outcomes.
The equations used to calculate the probability of death are defined in the following sections
for each method.
5.16.1 Single Effect Outcome

The single zone method can be used for all types of flammable effects. For explosions the
shape of the effect zone is circular and there are two effect levels. For flash fires there is a
defined boundary to the zone (either circular for instantaneous releases or elliptical for
continuous releases). For radiation effects there may be one ellipse or many. The first ellipse
is considered to correspond to the radiation level criteria (rather than the flammable probit
based lethality) and this will be subject to a different vulnerability factor to the outer ellipses
consistent with the Purple Book requirements. Therefore the first vulnerability level is
designated K IR ,1 and the outer zones K IR ,1 + .
Using the single zone method once Frx,y, ,Lio is known for each cell for the outcome
MPACT can calculate the Individual Risk Increment and the F-N pair directly. If the
flammable probit method is used then there will be a value for the fraction of the square
covered for each lethality level, Frx,y, ,Lio. When the flammable probit method is used
then the individual risk and societal risk is calculated in increments corresponding to the
average probability of death between each ellipse. This probability of death is derived
from the flammable probit calculation and designated Pd i . The actual probability of death
used in the risk calculations is modified by the vulnerability parameters. So for individual
risk the probability of death for each ellipse is;
Pd ,IR i = Pd i K IR ,Li
( 65 )
For societal risk the probability of death for each ellipse is;
Pd ,FN i = Pd i K FN ,Li
( 66 )
MPACT Theory
99
Then there are two methods to calculate the probability of death at the grid centre, x,y. The
first method uses the fraction of the effect zone that covers the cell with its centre at x,y. This
method has been standard in MPACT for many years. The probability of death is given by;
Pd , IR , x , y ,o =
(Fr
i= N L
x , y , , Li o
i =1
Pd , IR i
( 67 )
The second method simply calculates the probability of death exactly at the point x,y.
Pd , IR , x , y ,o =
(P
i= N L
d , IR , x , y i
i =1
( 68 )
The older method enabled a common calculation between societal and individual risk
calculations so provided calculation savings. It also tends to smooth the gradation between
grid cells. In practice a combination of the two methods is used when the obstructed region
explosion modelling is selected.
When calculating the F-N pair, as with toxic incidents, the frequency Fedf,o and number of
fatalities Nedf |o from an incident's outcome has to be calculated.
Fedf ,o = Fedf Po
( 69 )
For calculating the number of fatalities based on the population grid data, the flammable
outcome is specific (unlike that for a toxic incident), and the value is calculated without any
smoothing. The average probability of death in each MPACT grid cell is obtained by
summing each lethality level and fraction of the cell covered, given that the outcome o = ws
occurs:
Pd , FN , x , y o =
(Fr
i= N L
i =1
x , y , , Li o
Pd , FN i
( 70 )
If there is only one effect level then the sum is not required. The building fatalities are based
on the effect level experienced by the building. This effect level will in general vary around
the surface of the building. For the flash fire effects the proportion of the building within the
cloud is used to factor the probability of death. Different options for explosion effects are
provided in the parameters to allow different assumptions (worst effect used for the building
or mean value for the building).
The expressions in the equations above have the following meanings:
Fedf
Nx,y
Po
Frx,y, ,Lio
K IR , Li
frequency of the incident

number of people in the square
conditional probability of the outcome o=ws, (or combination of s)
fraction of the square covered by the fatal effect-zone between ellipses
overall factor to be applied to the calculation of individual risk for effect type
MPACT Theory
100
K FN , Li
overall factor to be applied in the calculation of societal risk for effect type
NL
Pd , IR i
Total number of lethality levels

Probability of death for individual risk calculations for ellipse I of effect type
Pd , IR , x , y i
Probability of death for individual risk calculations for ellipse i effect type at
the point x,y
Probability of death for societal risk calculations for ellipse i effect type
Pd , FN i
Pd , IR , x, y ,o
Pd , FN , x , y o
Probability of death for individual risk calculations for ellipse i outcome o at

the point x,y
Probability of death for societal risk averaged over the cell with center at x,y
Fedf is the event frequency as specified in the case input data. Nx,y is taken directly as the
value in the population grid. The K values are derived from parameter inputs as explained
in Section 3.6. Po is calculated in different ways according to the event tree and outcome
as explained in Section 5.2.4.
5.16.2 Combination of Two Effects in One Outcome

Event tree types D, E and F can include combined outcomes where the pool fire effect
zone is combined with another fire, for instance a jet fire. In general these effect zones will
have different shapes, location and size. The calculation of the probability of death from
combined effect zones is potentially a complex issue. For instance the orientation of the
observer to receive maximum radiation from each flame independently will be different to
the orientation for the combined flame. This level of complexity is not addressed herexxxvii.
Instead each effect zone is treated as presenting two independent chances of being killed.
The combined probability of death is then calculated according to equation ( 71 ) for
individual risk and ( 73 ) for societal risk. Two different probabilities of death are
calculated because there may be the requirement to apply different vulnerability factors for
each type of risk. As with the single effect outcome there are two methods for individual
risk. The older method is;
i=NL
Pd ,IR,x, y o = 1 - 1 - Frx, y,1,Li o Pd ,IR 1i
i=1
)1 - (Fr
i = NL
x, y , 2,Li o d , IR 2i
i =1
( 71 )
And the new method is;
i=NL
Pd , IR, x, y o = 1 - 1 - Pd , IR, x, y 1i
i =1
)1- (P
i=NL
i =1
d , IR, x , y 2i
( 72 )
For societal risk the calculation is;
xxxvii
EXTENSION. The more accurate approach would be to calculate the radiation level received by the
observer based on the combined flames and then to calculate the probability of death based on the flammable
probit approach. Since the effect zone shape is then likely to depart significantly from the shape of a single
ellipse then the use of this approach in the risk calculations will require a general extension to Mpact to
model effect zones of a general shape.
MPACT Theory
101
i= NL
Pd , FN , x, y o = 1 - 1 - Frx, y,1, Li o Pd , FN 1i
i =1
)1- (Fr
i=NL
x , y , 2, Li o d , FN 2i
i =1
( 73 )
5.16.3 Explosion, flash fire with or without pool fire

The probability of death for the explosion results using the obstructed region models are
derived as a function of the explosion model outputs and the vulnerability model. This
may result in a continuous function for probability of death rather than an area (or number
of areas) with a constant probability of death. Also the shape of the region is arbitrary
rather than defined as an ellipse, and the flash fire is defined in this model as that covered
by the cloud profile rather than represented by ellipses. Overall these features distinguish
this type of outcome as requiring a different approach. Again two methods are provided
for individual risk. First the method compatible with the older approach;
i=NL
Pd , IR, x, y o = 1 - 1 - Frx, y ,1, Li o Pd , IR 1i
i =1
)(1- Fr
x , y , 2 2 d , IR 2
)(1- P
d , IR, x , y 3
( 74 )
And the updated approach;
i= NL
Pd , IR, x, y o = 1 - 1 - Pd , IR, x, y 1i
i =1
)(1- P
d , IR, x , y 2
)(1- P
d , IR, x , y 3
( 75 )
Where the hazard zones are 1 for the pool fire, 2 for the flash fire and 3 for the blast. Then
for individual risk equation ( 61 ) is applied directly.
For societal risk the number of people killed by the explosion and flash fire is based on the
average value for the probability of death within the cell (calculated as described in
Section 5.15 and combining the flash fire and explosion outcomes).
{ (
)(
Pd , FN , x, y 1,2 = CellAverage 1 - 1 - Pd , FN , x, y 2 1 - Pd , FN , x, y 3
)}
( 76 )
Finally to take into account the possible pool fire;
i= NL
Pd , FN , x, y o = 1 - 1 - Frx, y ,1, Li o Pd , FN 1i
i =1
)P
d , FN , x , y 1, 2
( 77 )
5.16.4 Direct Calculation of Indoor Risk

If indoor risk is being calculated directly then the equations defined in Sections 5.16.1 to
5.16.2 are repeated for individual risk indoors and outdoors. For societal risk the total
number of fatalities is simply the sum of those killed indoor and those killed outdoors. The
composition of the total number of fatalities is recorded for each outcome.
MPACT Theory
5.17
102
FN Curve
5.17.1 Treatment of Non-Integer Values of N

The Nedf|o values for each representative outcome calculated by MPACT are normally noninteger and represent the average number of fatalities expected for that outcome type. In
reality the average non-integer Nedf|o value is not possible for an actual outcome since it is
not sensible to discuss a partial fatality.
In the FN curve calculation in MPACT this non-realistic average outcome is replaced by two
realistic outcomes where Nedf|o is rounded down to the nearest integer, Nl,o and up to the
nearest integer, Nh,o. To preserve the average rate of death the frequency between of each
outcome must be calculated as;
Fedf ,h ,o = (N edf ,o N l ,o )Fedf ,o
( 78 )
Fedf ,l ,o = (N h,o N edf ,o )Fedf ,o
( 79 )
These two realistic outcomes are used to apportion frequency between adjacent bins on the
FN curve and in calculating the Risk Integral measures.
5.17.2 Risk Ranking Point Results

The risk value for the cell at x,y represents an average value for that cell. Risk ranking
point risk increments are based on the exact location of the point. This therefore differs in
philosophy from the risk value being calculated for the cell. In effect the risk ranking point
really does represent risk at that specific point.
MPACT Theory
5.18
103
Immediate and Delayed Ignition Branch Probabilities
5.18.1 Options available to specify the immediate ignition probability

Each of the event trees has outcomes corresponding to immediate ignition. The term
immediate ignition is intended to describe ignition near to the time of the release itself,
possibly associated with the cause of the release but also due to unidentified causes.
Immediate ignition may also be referred to as direct ignition.
If there is an ignition source very close to the release this will not contribute to the nominal
immediate ignition, instead it will lead to delayed ignition outcomes at short release times.
The probability of immediate ignition may be specified in Mpact in a number of ways;
Specified in the event tree parameters

Specified directly by the user associated with an event to be processed by Mpact
Deduced according to event characteristics as specified in the Purple Book
So when the event tree parameters method is used separate probabilities can be used for
different types of events and this is governed by the route number deduced from the
dispersion behaviour. When the probability is specified directly by the user for a given event
then the event tree parameter is replaced by this specific value.
The Purple Bookxxxviii method is more complex and makes use of various characteristics of
the event to determine the immediate ignition probability (referred to as direct ignition in the
Purple Book). In addition it applies different logic depending on the use of the free-field
parameter. In the Purple Book the free-field method is applicable to the individual, location
specific risk results while the non-free-field method (referred to as the actual ignition sources
method) is applicable to the societal risk calculations.
Details of these calculations are described in the following sections.
5.18.2 Purple Book Relative Flammability of Different Materials

Some materials are more prone to ignite than others. This characteristic is defined by the
material properties of volatility and of reactivity as in the context of flammability and each
material is assigned a high, average or low category. The Purple Book provides a list of some
materials covered by this property (Table 4.7 in the Purple Book).
xxxviii
Reference is made here to the Purple Book. The method described is to be published in a future edition
and has been supplied to DNV directly by RIVM.
MPACT Theory
104
Low
1-chloro-2,3-epoxypropane 1,3dichloropropene
3-chloro-1-propene
ammonia
bromomethane
carbon monoxide
chloroethane
chloromethane
methane
tetraethyl lead
Average
1-butene
1,2-diaminoethane
1,3-butadiene
acetaldehyde
acetonitrile
acrylonitril
butane
chloroethene
Dimethylamine
Ethane
ethene
ethylethanamine
formic acid
propane
propene
High (1)
1-butanethiol*
acetylene
benzene*
carbon disulfide*
ethanethiol*
ethylene oxide
ethylformate*
formaldehyde*
hydrogensulfide*
methylacrylate*
methylformate*
methyloxirane*
naphtha, solvent*
tetrahydrothiophene*
vinylacetate*
Table 7 - (Table 4.7 of Purple Book)

Reactivity of a number of substances
([CPR14]). If little or no information is available, substances are classified as high
reactive. These substances are indicated with an *.
Note that in the description that follows there is no practical difference between average and high
reactivity. Note also that flammable materials without this property defined are treated
conservatively as High.
Volatility classification is defined in the Purple Book using the categorisation according to the
material flash point Tfl and the table below;
Class
K0
Category
Extremely flammable
K1
K2
K3
K4
Highly flammable
Flammable
Description
liquids with Tfl<0 oC and normal boiling pointxxxix <
o
=35 C
Gases at normal temperature and pressurexl
o
Non-K0 liquids with Tfl<21 C
o
o
Non-K0 Liquids with 21 C =< Tfl =< 55 C
o
o
Non-K0 Liquids with 55 C < Tfl =< 100 C
o
Non-K0 liquids with Tfl > 100 C
Table 8 Classes of flammable materials
xxxix
For all (i.e. pure and multi-component) materials the bubble point temperature, as calculated from
vapour-liquid equilibrium thermodynamics, is used as the boiling point.
xl
The normal pressure and temperature conditions are defined as 1013mbar and 20oC
MPACT Theory
105
Note that the flame point Tfl is typically a few degrees lower than the flash point, where the flash
point is directly available from DIPPR for pure componentsxli. For mixtures the flash point is taken
the minimum for each of the flammable components. For simplicity in the above tables the flame
point is taken to be equal to the flash point.
5.18.3 Immediate Ignition for Stationary Vessels

Having taken into account the material properties of the released material itself the source and
scale of the release are taken into account. The first categorisation is stationary versus transport
vessels. The table below provides probabilities for stationary vessels;
Class
source
continuous
source
instantaneous
probability of
immediate ignition
K0
average/high
reactivity
K0
low
reactivity
K1
<10kg/s
10-100 kg/s
> 100 kg/s
<10kg/s
10-100 kg/s
> 100 kg/s
-
<1000 kg
1000-10,000 kg/s
> 10,000 kg
<1000 kg
1000-10,000 kg/s
> 10,000 kg
-
0.2
0.5
0.7
0.02
0.04
0.09
0.065
K2
K3,K4
0.01
0
probability of
delayed ignition
FreeActual
field
Source
0.8
Use Ignition
0.5
sources
0.3
0.98
Use Ignition
0.96
sources
0.91
0.935
Use Ignition
sources
0
0
0
0
Table 9 - Probability of direct ignition for stationary installations

The immediate ignition probability will be deduced from the immediate ignition column above.
Then the actual immediate ignition outcomes will depend on the route number and the event tree
settings. For stationary vessels the conditional probability of the fireball/BLEVE outcome is
prescribed as 0.7 in the Purple Book. Mpact will use the event tree probabilities, however these are
set so the event tree branches must be set appropriately to achieve compliance with the Purple
Book.
When using the free-field parameter it can be specified that the total probability of delayed ignition
outcomes will sum to the total value at the head of the event tree branches. However, when using
the actual ignition source modelling the total probability of delayed ignition will depend on the
cloud finding ignition sources and may well sum to less than 1- Pi. So the free-field column in the
above table normally indicates a delayed ignition probability of 1- Pi. However, the treatment of
K2+ materials is an exception. In this case delayed ignition is not considered possible due to the
low volatility of the material so the delayed ignition probability is prescribed as 0.
For the actual ignition sources the delayed ignition probability branches on the event tree will
typically start with a conditional probability of 1- Pi but the overall probability of delayed ignition
will depend on the actual ignition sources covered by the cloud in each time step.
xli
Typically DIPPR does not have flash point values for materials that are not liquid at normal conditions. In
these circumstances 1) for pure materials it is assumed that the flash point is Tfl<0 oC so they will be
classified as K0 2) for mixtures a value for the component is synthesised from a correlation defined in the
property system documentation. This avoids an excessively high or low value being used for a mixture
containing a component with an undefined flash point.
MPACT Theory
106
Note that the probabilities may then be adjusted according to the rules for elevated temperatures in
section 5.18.5 and materials that are both flammable and toxic in section 5.18.6.
MPACT Theory
107
5.18.4 Immediate Ignition for Transport Vessels

The following table gives probabilities for transportation vessels;
Class
transport
means
scenario
probability of
immediate ignition
K0
road tanker
0.1
0.4
0.1
0.8
0.065
0.01
0
K1
All
Continuous
instantaneous
continuous
instantaneous
All
K2
K3,K4
All
All
All
All
tank wagon
probability of
delayed ignition
FreeActual Source
field
0.9
Use Ignition
0.6
sources
0.9
0.2
0.935
Use Ignition
sources
0
0
0
0
Table 10 - Probability of direct ignition for transport units in establishment

As for the stationary vessels, the immediate ignition probability will be deduced from the
immediate ignition column above. Then the probability of the fireball/BLEVE outcome is
prescribed as 1.0 in the Purple Book so only this outcome will be included in the risk for
immediate ignition.
The delayed ignition is treated in the same manner as for the stationary vessels and again note that
the probabilities may then be adjusted according to the rules for elevated temperatures and
materials that are both flammable and toxic.
5.18.5 K2+ materials Processed at Elevated Temperatures

According to the Purple Book, if the process temperature of class 2, 3 and 4 is larger than the
flame point, the immediate ignition probability of class 1 should be used. In practice though the
release temperature is used instead of the process temperature as being a more useful indication of
the likely behaviour of the cloud with respect to ignition. No delayed ignition is assumed, as there
is no (large) flammable cloud expected. This is summarised below;
Class
source
continuous
source
instantaneous
probability of
immediate ignition
K 2,3,4 process
temperature above
flame point
All
All
0.065
probability of
delayed ignition
FreeActual
field
Source
0
0
Table 11 - K2+ materials above flame point temperature
MPACT Theory
108
5.18.6 Materials that are both Toxic and Flammable
Additional logic is applied to modelling both types of materials. There are two types of
behaviour in the model depending on whether or not the non-ignition probability is supplied by the
user.
Non-ignition probability not prescribed
In this case the non-ignition probability is calculated as 1- Pi and instead of being assigned to the
delayed ignition branches of the event tree it is assigned to the toxic effects.
Non-ignition probability prescribed
In this case the delayed ignition branches are modelled as normal with the probability 1-Pi and the
prescribed non-ignition probability is assigned to the toxic effects. Note that this can lead to
double counting the risk but this is left to the judgement of the user.
Note that the Purple Book rule regarding low ignition likelihood toxic and flammable materials is
not forced in the code so that the user must ensure that they model these releases only as toxic
with the full event frequency.
MPACT Theory
5.19
109
Delayed Ignition
5.19.1 Options for the delayed ignition branch of the Event Trees
There are several options available to represent delayed ignition. They address two
important aspects of representing delayed ignition. The first is how to deal with the
development of the cloud with time and the second addresses what probability of ignition
to assign. The combination of the various options for each aspect gives several
permutations for the user. Furthermore, the introduction of delayed ignition modelling that
takes into account obstructed regions introduces further choices.
5.19.2 Development of the Cloud with Time
Original Method
The original method in MPACT (6.54 and before) to step in time is derived from the
dispersion results input to the model. The approach in MPACT is to model time steps as
the cloud develops from the moment the release starts to the moment the flammable cloud
has dispersed below the LFL fraction. Every time step present in the dispersion results is a
time step for the risk modellingxlii.
The advantage of modelling the time steps is that it gives a good range of possible
outcomes. The disadvantage is the computing cost in terms of CPU time and volume of
results generated. Many QRA practitioners feel the cost is not worthwhile and prefer
instead to represent the delayed ignition cloud as one or just a few representative
outcomes.
This method is not available when using the obstructed region methods in version 6.6.
Obstructed Region Methods
Many continuous type releases give a cloud that develops very quickly to the flammable
limits and then does not change very much until the time at which the release stops. For
such clouds it makes most sense to model one outcome and then the cloud view to select
corresponds to the maximum footprint area of the cloud to the LFL fraction. The duration
of this cloud view is required for the ignition source modelling and for this the duration of
the release may be used.
Instantaneous releases naturally move with time and it is more relevant to consider a
number of cloud views changing with time. Ideally the sequence of cloud views used
would not leave gaps. Otherwise the modelling would give the impression of no risk in
between cloud views.
Time varying releases, very short duration releases and quasi-instantaneous clouds give
histories that may have characteristics of both types of release. It may be worth modelling
a number of time steps or this may be an unnecessary complication.
xlii
The cloud data input to Mpact can have up to 101 segments each with up to 100 time steps. So there could
be over 10000 time steps. This represents excessive discretisation effort compared say, with 12 wind
directions.
MPACT Theory
110
When using the obstructed region methods the time steps are not used directly from the
dispersion time steps. Instead the time discretisation is designed to put the computational
efforts into modelling outcomes that are significant with respect to risk. Because the model
uses a numerical method there is an inevitable compromise between accuracy and
calculation time so it is possible for the user to specify the number of time steps in a range
with a minimum of one. This approach allows the user to trade-off computational costs vs
covering all potential outcomes.
The following logic is applied to determine how to discretise the time steps based on
parameters to define the number of time steps and flag to either Minimise Gaps or
Allow Gaps.
Allow Gaps
The total duration of the cloud is defined, starting with the release until there is no longer
any part of the cloud above the LFL fraction concentration (the final time). Since there is
no possibility of ignition at the final time the duration is divided into equal time steps
using the time step parameter value + 1.
For some types of cloud, for example instantaneous releases, the view will be changing
between each time step. However, other types of cloud will give the same view at each
time step. A simple continuous release will normally give a fully developed cloud profile
within a few seconds. The user may set different numbers of time steps for continuous
releases vs time-varying releases. If the number of timesteps is set to 1 this is treated as a
special case and the model will chose the cloud view that has the greatest footprint area.
Taking this approach will cut down on computational time but will not resolve the risk
according to all possible ignition times.
Minimise Gaps
In this mode the time steps are derived in a way to match the cloud behaviour as it
develops with time and to ensure that regions cannot be skipped by using a too large time
step. This also ensures that obstructed regions cannot be missed by the cloud.
Instantaneous clouds are developed in time independent of the time step parameter. The
first step is taken at the immediate ignition time. The subsequent time steps are calculated
by setting the next cloud centre at the downwind position of the maximum downwind
extent of the previous time step.
Continuous segments may be considered as quasi steady state or time-varying. The
former will develop into a footprint that remains constant for a period corresponding to the
segment duration less the time the cloud takes to develop to the flammable limits. The
latter never achieves a steady state and behaves more like an instantaneous cloud. These
different types are modelled in a different way.
The first segment is treated slightly differently to the others if there is no previous
instantaneous segment. This is to allow the model to represent clouds that will find strong
ignition sources before they become fully developed. The time step parameter is used to
subdivide the time the cloud takes to reach the fully developed state. For time-varying
segments it will ensure that there are no gaps by tracking the downwind and upwind edges
of the cloud. The location of the upwind edge of a new time step is located at the
downwind edge of the previous time step.
MPACT Theory
111
Subsequent continuous segments will be represented by the fully developed footprint if

steady state or by the minimum number of time steps necessary to unsure no gaps if timevarying.
Multiple segment releases, particularly those following an instantaneous release, may give
overlapping time steps so the time steps are sorted to ensure an ascending time sequence.
Largest Footprint Area
The user may chose to run with only a single timestep for delayed ignition. This approach
may be taken to reduce calculation times and study size. For steady state single-segment
releases the fully developed cloud footprint will correspond to the maximum footprint area
and the model will use this cloud view to generate the single delayed ignition outcome.
For instantaneous and time-varying releases using a single cloud view is more of a
compromise. This approach may leave zones without risk from delayed ignition. The
worst case cannot be predicted ahead of the risk calculations since it does not yet know
the intersection with obstructed regions. So again the cloud view with the maximum
footprint area is chosen. For instantaneous clouds without any continuous segments it
selects the cloud view with the maximum radius and this is a relatively quick calculation.
However for multi-segment clouds there is an issue finding the cloud with the maximum
area. It is not considered to be worth using a numerical search algorithm. Instead it
generates a series of cloud views using the Minimise gaps algorithm. The view with the
largest area is then used for the delayed ignition outcome.
The largest footprint area is assumed to exist for the entire cloud duration. This same cloud
view could also be used for the Dutch Free Field method but at present (version 6.6) this
method is not implemented with the 3D cloud modelling.
5.19.3 Probability of Delayed Ignition

The probability of delayed ignition at each time step may be derived from an ignition
model that takes into account the presence of specific ignition sources and the population.
This is described in detail in section 5.20.
In addition the user can specify a minimum probability of delayed ignition. This works in
combination with the ignition source model to ensure that the overall probability of
delayed ignition is at least the value specified by the user. If the ignition sources are
significant then the overall delayed ignition probability will be higher.
If there is one time step then the minimum delayed ignition probability is applied directly
and in conjunction with the ignition sources model.
If there is more than one time step then it is necessary to assume some form of time
distribution. For this we employ the methods explained in section 5.20 such that an
equivalent ignition source is defined based on the users prescribed minimum delayed
ignition probability. The probability of ignition of each delayed outcome is considered to
have a linear distribution between timesteps 1 to N such that;
MPACT Theory
112
Pi , o = Pi , du
(ti ti 1 )
( 80 )
tN
This probability distribution satisfies the requirement that the overall probability of
ignition matches the prescribed minimum delayed ignition probability;
i=N
Pi , du = Pi , o
( 81 )
i =1
A virtual ignition source strength is then derived based on satisfying the cumulative
probability of non-ignition between timesteps. This gives;
Pi =
(ti ti 1 ) Pi , du
t N ti 1Pi , du
( 82 )
In each time step this equivalent ignition source is combined with all the other sources in
the normal manner. If there are no other ignition sources then the final overall probability
of delayed ignition will correspond to the minimum specified by the user. In contrast, if a
very strong source is experienced by the cloud then the delayed ignition probability will be
one and there will be no subsequent time steps.
5.20 Ignition Source Modelling

In many of the calculations above the probability of delayed ignition was used. This Section
explains how the delayed ignition probability is calculated. Note that in addition there is the
option to run the model in a mode that is designed to run through all the delayed ignition
outcomes without reference to the probability and this bypasses the following logicxliii.
5.20.1 Definition of Ignition Sources
The delayed ignition probability of a cloud is dependent on the number of ignition sources
and how these are placed with respect to the flammable region of the cloud. It also depends
on the characteristics of these ignition sources. The program considers a number of types of
ignition source;
Point, line, area and volume sources defined explicitly by the user and provided to
MPACT as Ignition Data
Ignition associated with human activity with a strength derived from the population data
provided as input to MPACT
Point, line, area and volume sources in the Ignition data are characterised by their ignition
probability in a given time period, Pit, the operating probability (also know as presence
factor), fi, , and their geometry.
xliii
This is used for the maximum loss calculations in the Phast Financial application.
MPACT Theory
113
For population-derived ignition MPACT calculates an additional probability of ignition

within the flammable region of the cloud using MPACT parameter values for Population
omega factor, people. These values are per person and different values may be set for
people indoors and people outdoors.
Ignition Probability in Time Period Specification
For ignition area sources (rectangular or polygon), MPACT allows the user to specify the
ignition probabality Pi,t,U per unit of area (with specified unit area U) or the ignition
probability Pi,t,A for the total area (with area A derived from user-specified geometry). The
likelihood of non-ignition for the total area 1-Pi,t,A can be shown to be related to the
likelihood of non-ignition for the unit area 1-Pi,t,U, for a given time period by
1 Pi ,t , A = (1 Pi ,t ,U )
( 83 )
A /U
The above equation is an expression that the cloud does not ignite for the entire area
source only and only if each of the individual unit area sources do not ignite. From the
above equation it now easily follows that the likelihood of ignition for the total area Pi,t,A is
related to the likelihood of ignition for the unit area 1- Pi,t,U by
Pi ,t , A = 1 (1 Pi ,t ,U )
A /U
, Pi ,t ,U = 1 (1 Pi ,t , A )
U/A
( 84 )
The above Equation is therefore currently used in the program to calculate Pi,t,A from Pi,t,U
and vice versa.
For line sources (typically electrical lines) the equivalent calculations may be performed to
convert from per-unit length probability specification to probabilities for the whole line.
To do this we can substitute line length for area and the length of the length unit for U in
the above equations.
Ignition Potential Factor (ii)

This is a measure of the strength of the source and is the probability of ignition per unit time
(second) given that the source is present. The ignition potential factor ii is related to the
effectiveness factor i as follows23:
ii = 1 e
( 85 )
This exponential term is introduced because it relates the continuous ignition effectiveness
to the probability of ignition in a specific time interval. It enables the conversion of the
probability of ignition in a specific time interval to a continuous parameter and vice versa.
Presence Factor (fi)
MPACT Theory
114
This is the proportion of time that the ignition source is present and active. If it is not active
then there is no chance of ignition, if it is active ignition could occur.
Population Ignition Source

For the population ignition sources the strength of the ignition source is linked to the
number of people present so the ignition potential factor is calculated as:
ii = 1 e
Out
people N di
( 86 )
Where Ndi is the number of people in the flammable part of the cloud over the time interval
Out
and people is the ignition effectiveness per person. If indoor risk is not modelled directly
then this parameter is applied to all people considered in the analysis regardless of
location. When indoor risk is modelled directly then the model uses different factors for
the ignition effectiveness of people indoors and outdoors. This is to allow for reduced
probabilities of people indoors igniting an outdoor flammable cloud.
ii = 1 e
In
Out
Out
( In
di )
people N di + people N
( 87 )
Note also that the population is present continuously for any one run of MPACT. Only
people distributed onto the population grids are considered as ignition sources except in a
special case; people located in specific buildings of outdoor type. So normally people
located in specific buildings do not contribute to the ignition probability.
Geometry
This defines the shape and location of each ignition source. It is necessary to allow MPACT
to calculate the interaction of the cloud with the ignition sources and thus work out the overall
ignition probability taking into account all possible ignition sources. MPACT does this using
two distinct methods;
1) Grid Method - an intermediate step where the ignition sources are
mapped onto the risk grid and this method is described in Section 5.20.2.
This was the only method available in the model in version 6.54 and
earlier. It applies to sources where the elevation input is set to zero.
2) Direct Method this method calculates the intersection of each ignition
source with the cloud taking into account the elevation of the source. To
do this it needs a third dimension to locate the cloud vertically and this is
only available in version 6.6 and later. Only ignition sources with nonzero elevations are treated in this way and this is explained in Section
5.20.3.
Delayed ignition probability
First the flammable region of the cloud must be positioned over the ignition source. For
instance if over time interval t the cloud covers no ignition sources then the probability of
MPACT Theory
115
delayed ignition in that time step will be zero. If the cloud covers a given ignition source for a
time step of duration t, then the ignition probability for that time-step is given by:
Pi ,t = f i 1 e it
where fi
( 88 )
is the presence factor for the source

is the ignition effectiveness factor
Example
The illustration below is a graphical representation showing the terms defined above for a
presence factor of fi = 0.8, and a probability of ignition (when present) of Pit/fi = 0.5 in 30
seconds. With this datum point, the curve gives a value of ii = 0.0228 for the probability after
1 second, and the corresponding value i = 0.0231.
Ignition
Probability
1.0
Presence factor, fi = 0.8
0.8
Datum Point
50% Probability in 30s
=40% with f of 0.8
0.6
0.4
0.2
0.018
0
0
1s
10
20
30
40
50
60
Time (s)
Probability in 1s = 0.018
Ignition Potential, ii, is probability with fi of 1.0 => 0.0228
Figure 49 - Ignition Probability Function Example
MPACT Theory
116
5.20.2 Multiple Ignition Sources Grid Method

The application of every ignition source individually to the cloud at every step in the risk
integration proved too computationally intensive in the earliest versions of MPACTxliv. To
overcome this problem a grid-based method of combining ignition sources was derived and
provided a more efficient algorithm. One drawback of this approach was the assumption that
all ignition sources were present at the same height. Most recently the option to specify
height for certain types of ignition source has been introduced. This new method supplements
the grid-based method and is described in section 5.20.3.
The algorithm converts the individual ignition sources into a grid of equivalent ignition
sources derived by integrating all possible sources of ignition over each risk grid cell. In this
way each individual ignition source, whatever its original size and shape contributes towards
a single equivalent source per MPACT grid cell.
In addition to the population, ignition sources can be defined in the form of;
Point sources
Line sources
Polyline sources
Rectangular area sources
Polygon area sources
Line traffic sources
Polyline traffic sources
Each source is imposed onto the ignition grid in a similar way to the population objects
described in Section 3.4. The proportion of each source that lies in each cell is evaluated
and added to the overall equivalent source for that cell. In addition, traffic sources require
some preliminary calculations.
Traffic Sources
Traffic sources are converted to equivalent line sources from the input variables average
speed sv (m/s) and the frequency of vehicles, fv (Hz). Each vehicle is considered to be a
point source that could be positioned anywhere along the line of traffic of length, lt (m).
The user inputs the probability of ignition in a given time period for each vehicle and
every vehicle is considered to have the same ignition effectiveness.
The average number of vehicles present along the line of traffic is calculated as
Nv =
f v lt
sv
xliv
( 89 )
EXTEND. Current computers may enable the direct application of the ignition sources and avoid the
approximate grid cell method. To really appreciate the approximation it is necessary to consider that the
cloud for flammable events is always considered propagating along the wind sector centreline. The border of
the cloud is a distinct line at the LFL boundary. Whether the cells covered (or partially covered) by the cloud
contain an ignition source is the key point in calculating the probability of ignition at each step. Small
changes in input data can have a surprising influence since the cloud may miss or hit a given ignition source.
MPACT Theory
117
The conversion of the traffic source into a line source then depends on the result. If the
average number of vehicles along the line of traffic is less than 1 then the source
effectiveness corresponds to a single vehicle and the presence factor is set equal to Nv.
If the average number of vehicles is greater than 1 then the presence factor for the source
is set to 1 and the effectiveness of the ignition source is set as the product of Nv and the
effectiveness of a single vehicle.
Cloud Intersection with the MPACT Ignition Source Grid
Having converted all the ignition sources into a grid, the strength of the ignition sources
experienced by a given cloud will depend on the way the ignition sources are distributed
onto the MPACT grid and how the cloud intersects the grid cells.
Grid Size
C D
A B
Figure 50 Illustration of the Interaction of a Cloud and Cells containing an Ignition Source
The interaction of the cloud and the MPACT grid is shown by Figure 50 and illustrates the
critical nature of the exact location of an ignition source with respect to the grid.
Conceptually a point source could be at the intersection of the four MPACT grid cells
A,B,C and D. If it just lies in cell A the cloud will experience an ignition source. This in
itself will seem non-intuitive because the cloud does not reach the point source. If it lies in
cell B again there will be an ignition source but a weaker one because of the smaller
proportion of the overlap between the cloud and the cell. If it lies in cells C or D there will
be no ignition source. Small movements of the grid lines could cause the program to flip
between the different treatments with respect to the ignition of the cloud.xlv Note that
source coordinates that lie on grid lines are considered to belong to the cell in the +ve x
and/or y direction (ie, North and/or East).
xlv
Note also the situations where the grid size doubles or halves. Doubling the grid size when the source is
initially in cell C or D will mean the cloud now experiences an ignition source. Conversely if the source
initially just fell within A or B and the grid size is halved and the cloud may now miss the source.
MPACT Theory
118
Each equivalent source is defined by the parameters; equivalent presence factor fx,y and
equivalent effectiveness factor x,y,t. Note that the effectiveness factor is now a function of
time itself. The combination of ignition sources is a non-linear manipulation and this attribute
was introduced to enable the equivalent ignition source to better match the prediction given
by the combination of the original ignition sources.
The approach taken is to conceive a cloud exclusively over an individual grid cell. It then
makes a rigorous calculation of the ignition probability for two representative times from the
sources within the grid cell. The two exposure times t1 and t2, are 100s and 1000s and these
are set in the codexlvi.
Having deduced the actual ignition probabiliy from the sources in the cell it then sets values
of equivalent presence factor fx,y and equivalent effectiveness factor x,y,t to match the two
ignition probabilities calculated for the exposure times t1 and t2.
Delayed ignition probability at exposure times t1,t2
The ignition probabilities are calculated by accumulating the probability of non-ignition for
each source (i = 1,2,nsource). Equation ( 90 ) shows the calculation for the first exposure
time t1. The equivalent calculation applies for the second exposure time t2.
Px , y ,t1 = 1
nsource
(1 f (1 e
i =1
x , y , i i t1
))
( 90 )
The factor x,y,i is used to multiply the effective factor for the source by the proportion of the
source that exists within that grid cell. For point sources x,y,i =1 or 0 because the source is
either in the cell or outsidexlvii. For line and area sources, the overlap between the source and
the grid cell must be calculated.
Equivalent Presence Factor fx,y
Once the ignition probabilities have been calculated for the two exposure times it uses these
to deduce an equivalent effectiveness factor for the cell. To do this it must also have an
equivalent presence factor. Ignition sources may have presence factors that vary enormously
from each other so in reality there are many permutations of the presence of each ignition
source from all being present at the same time to none. In addition there might be some
correlation between the sources.
However if we define the equivalent presence factor as being the probability of any source
being present (assuming no correlation between the presence factors for each source) then the
equivalent can be calculated rigorously through the product of the probability of each source
not being present.
xlvi
EXTENSION. Allow other representative times

EXTENSION - Note that if a source lies exactly on a grid line then it will be attributed to the north
and/or east grid cell. It will contribute to only one cell. Note also that this implies a smearing of the point
source over the cell and a shift of its average position. These problems could be overcome by applying the
point source directly to the cloud during the risk integration.
xlvii
MPACT Theory
119
f x, y = 1
( 91 )
nsource
(1 f )
i
i =1
Equivalent Effectiveness Factor x,y

Having defined the equivalent presence factor an equivalent ignition effectiveness can be
deduced using equation ( 88 ) from the two ignition probabilities derived for exposures times
t1 and t2. Equation ( 92 ) shows the equation for t1 while a similar equation is used to calculate
x,y,t2.
x , y ,t
Px , y ,t1
ln1
f x, y
=
t1
( 92 )
The different values for ignition effectiveness are used to indicate the sensitivity of this
equivalent effectiveness factor to time. If x,y,t1 x,y,t2 within 5%, then x,y,t is taken as their
arithmetic average and it will not vary with exposure time. The criteria used is
x , y ,t x , y ,t
x , y ,t + x , y ,t
0.025
( 93 )
Otherwise, an exponential decay curve is adopted, i.e.:
x , y ,t = b x , y e
a x , yt
( 94 )
Where
ax, y =
x , y ,t1
1
ln
t2 t1 x , y ,t2
bx, y = x , y ,t1 e
a x , y t1
( 95 )
( 96 )
Delayed ignition probability as function of time

Now, having performed this calculation once for all ignition sources there is now an
equivalent ignition source per cell according to equation ( 97 ) that will be used to calculate
the ignition probabilities for the clouds to be modelled in MPACT.
Px , y ,t = f x , y 1 e
x , y , t t
( 97 )
MPACT Theory
120
The above is the delayed ignition probability within the cell (x,y) presuming that the cloud
entirely covers the cell (ie the factor x,y,i is 1).
Interaction of the Cloud and the Equivalent Ignition Sources
For each of the dispersion steps of a cloud, the program identifies the grid cells that are
wholly or partially covered by the flammable region of the cloud, and calculates the fraction
of each square covered Frx, y,|o as described in Section 5.14. The algorithm considers each
cell in turn accumulating the probability of non-ignition as it goes. Note that Frx, y,|o is used to
adjust the ignition effectiveness to account for partially covered cells. It also calculates the
total number of people within the cloud using this factor.
Once all the relevant grid cells have been considered it multiplies the probability of nonignition by the probability of non-ignition according to the number of people in the cloud.
The probability of delayed ignition in the entire cloud Pdi,o,t for the outcome o within the
timestep t, is then calculated according toxlviii
Pdi ,o ,t =
(1 f
x, y
(1 e
Frx , y , P Frx , y , o x , y , t t
( 98 )
))
All grid squares
Pdi ,o ,t = 1 e
Out _ people N di , t t
(1 f
x, y
(1 e
Frx , y , o x , y , t t
))
( 99 )
All grid squares
If indoor risk is being modelled direction then the indoor population effectiveness is also
used to calculate the final ignition probability.
5.20.3 Multiple Ignition Sources Direct Method for Height-Specific Sources
Some ignition sources are characterised by a known height above the ground. Flares and
electrical cables are the most obvious examples. These are considered in the ignition
source model as point and line sources. Area sources can also be considered at a particular
height relative to the cloud. If the user specifies a specific height for the ignition source
and the 3-D cloud method is being used then the ignition modelling will take the height
into account.
In both cases a single height above ground is considered. This height is used to consider if
the ignition source and the cloud intersect. Because of the potentially different heights of
sources it is necessary to retain the identity of each source during the modelling rather than
to locate it on a grid. The user should be aware of the computational cost in using this
method in case it causes excessively slow run timesxlix.
xlviii
EXTENSION. Since the cloud size changes as function of time also Fr and Ndi change with time. If we
consider a time interval t, only the position of the cloud at the end of the interval is considered. A possible
extension would take the moving cloud position into account by integration or averaging.
xlix
EXTENSION, An additional benefit of the method is that it could be possible to record which ignition
sources are considered for each delayed ignition outcome. This could be helpful in understanding the risk
results.
MPACT Theory
121
In addition the capability to indicate a shut off time for the ignition source is possible. This
is to represent the behaviour of safety systems in the event of leak detection. Ignition
sources may take a while to cool down and the ignition strength will reduce over of a
period of time so it is also possible to specify an average cooling time. The overall
probability of ignition in a timestep t for a number of such ignition sources is given by;
Pt = 1
nsource
(1 f
i =1
s ,i
( 100 )
f p ,i (1 e ii t ))
Where for each source i ;

i is the fraction of the ignition source within the flammable cloud
i is the effectiveness factor for the total source
fp,i is the presence factor
fs,i is the shutdown factor
The shutdown factor is a function of time such that up until the shut down time fs = 1.0 and
after the shut down time it varies according to an exponential decay;
ln(0.5)(t r t s )
f s = exp
t c
( 101 )
Where;
tr is the overall time since release

ts is the shutdown time
tc is a time constant indicating how quickly the ignition potential decays (typically
to represent the cooling of a previously hot object). In effect this the time at which
the ignition potential has reduced by a factor of 2.
An example of how the ignition source decays is given below for different cooling times
and a shut down time of 5s.
1s
Igniiton probability per s
0.9
10s
0.8
20s
0.7
60s
0.6
0.5
0.4
0.3
0.2
0.1
0
0
10
20
30
40
50
60
70
80
90
Time since release/s
MPACT Theory
122
Figure 51 Ignition probabaility vs time for a shut down time of 5s and cooling times of
1, 10, 20 and 60s
A tc value of 0 indicates to the model that the ignition shut down is immediate.
The probability Pt may be combined with the population, ignition grid and scenariospecific probabilities if present to give an overall probability of ignition in a given
timestep.
5.20.4 Free Field Methods
The Dutch Purple Book9 defines the free-field approach for calculating individual risk,
independent of population as ignition sources. Mpact has two methods for dealing with
such calculations. The old method (as an option in SAFETI version 6.53 and earlier) that
has a single value for the distance to the plant boundary for all releases is preserved and
there is a new method that makes use of the actual shape of the plant boundary. The
position of the free-field outcomes in the event trees was described in Section 5.2.4.
Single Distance Method
Using this method the model has no idea of the actual shape of the plant boundary so it
cannot calculate the specific distance of each release to the boundary. Instead it has a
single parameter that is used to decide whether the release should be treated as one that
can leave the plant boundary. The furthest downwind reach of the flammable cloud along
the cloud centreline is compared with the parameter value and if the flammable cloud goes
further then the free-field outcomes are included in the individual risk results. Otherwise
there are no free-field outcomes. The cloud width is not used in the calculation.
Other delayed ignition outcomes will also be executed if ignition sources are found so
when using this method it is important to define ignition sources that are allowed
according to the rules in the Purple Book9. This is unlike the method that uses the shape of
the plant boundary since in that method the ignition sources are automatically adjusted to
comply with the Purple Book rules. Population is ignored as a potential source of ignition
in both methods.
Boundary Shape Method (versions 6.54 and later)
The boundary shape is defined by a series of x,y locations defining a closed shape. The
shape can include cross-over points. Only one series of points is considered though in a
single run of Mpact. If a risk study is to consider multiple plants then it will be necessary
to use different run rows.
If the boundary shape method is selected in the parameters then errors will be generated
a) if there is no plant boundary defined
b) if release locations for flammable chemicals are specified outside the plant
boundary.
The test for whether the cloud leaves the plant will be based on the centreline reach of the
LFL fraction boundary and the plant boundary of origin. Again the width of the cloud will
MPACT Theory
123
not be considered. The decision will be directional and will correspond to the input
geometry of the selected plant boundary.
Ignition sources completely within the boundary will be used for both individual risk and
societal risk. Those completely outside the boundary will be used for societal risk only.
For ignition sources partly inside and partly outside the boundary, the part inside will be
used for both individual risk and societal risk, and the part outside for societal risk only.
(for societal risk, no free field is considered, the only ignitions considered are population
or specified ignition sources)
The method employed is to multiply the current ignition contribution of each Mpact grid
cell by the fraction of the cell covered by the plant (Fr x,y,p). This is a modification of the
normal equation for ignition.
Pdi ,o ,t =
(1 f
x, y
(1 e
Frx , y , P Frx , y , o x , y , t t
))
(102)
All grid squares
There will be two sets of outcomes, one for individual risk and the other for societal risk.
The risk contours and the individual risk ranking results reflect the free field outcomes and
the FN curves and societal ranking results will reflect the societal outcomes. The time step
used for the free field outcomes corresponds to the instant the flammable cloud has the
maximum footprint area.
The diagnostic files reflect the type of outcome relevant for the given file. For instance the
FN Outcome file will use the societal outcomes while the risk increments file will use the
free-field outcomes.
MPACT Theory
124
6. Verification and Sensitivity Analysis

Verification and sensitivity analysis is reported in the MPACT Verification Manual12.
7. Future developments
The creation of the future development plans is ongoing work-in-progress.
MPACT Theory
Appendix A.
125
MPACT input Parameters - Event Tree Probabilities
Standard Set
The following list gives the probabilities presented in the standard product. A reduced set
is present in the micro version (and this list is used to generate the full route set). This set
of parameters is based on historic version of the program. A second set follows that is
based on the Purple Book guidance.
NewSafeti (RunRow 1)
Event Tree Probabilities
Probability of a BLEVE
Probability of a Pool Fire
Toxic Probability
Continuous no Rainout
Immediate Ignition
Long Duration Horizontal
Fraction
Jet Fire
Long Duration Vertical Jet
Fire
Short Duration Fraction
Short Duration BLEVE
Short Duration Flash Fire
Short Duration Explosion
Delayed Ignition Flash
Fire
Delayed
Ignition
Explosion
Continuous with Rainout
Immediate Ignition
Fraction
Jet Fire
Pool Fire
Jet Fire with Pool Fire
1.00
1.00
1.00
0.30
fraction
fraction
fraction
fraction
0.60
fraction
1.00
fraction
1.00
fraction
1.00
fraction
1.00
fraction
0.00
fraction
0.00
fraction
0.60
fraction
0.40
fraction
0.30
fraction
0.60
fraction
0.00
fraction
0.00
fraction
1.00
fraction
MPACT Theory

Long Duration Vertical
Pool Fire
Fire
Fire with Pool Fire
with Pool Fire
alone
with Pool Fire
Alone
with Pool Fire
Alone
Short Duration Pool Fire
Residual Pool Fire
Fire
Delayed
Ignition
Explosion
Instantaneous no Rainout
Immediate Ignition
BLEVE
Immediate Flash Fire
Immediate Explosion
Fire
Delayed
Ignition
Explosion
Instantaneous with Rainout
Immediate Ignition
BLEVE with Pool Fire
BLEVE Alone
126
0.00
fraction
0.00
fraction
1.00
fraction
1.00
fraction
1.00
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.15
fraction
0.60
fraction
0.40
fraction
0.30
fraction
1.00
fraction
0.00
fraction
0.00
fraction
0.60
fraction
0.40
fraction
0.30
fraction
1.00
fraction
0.00
fraction
MPACT Theory

Immediate Flash Fire with
Pool Fire
Immediate Flash Fire
Alone
Immediate Explosion with
Pool Fire
Immediate
Explosion
Alone
Immediate Pool Fire Alone
Residual Pool Fire
Fire
Delayed
Ignition
Explosion
Immediate Ignition
Explosion Given Ignition
Long Duration Jet Fire
Short
Duration
Any
Ignition of Cloud
Short Duration Ignition of
Cloud with Pool Fire
Jet Fire with Pool
Fire with Pool
Short Duration Fraction for
Effects
Short Duration BLEVE not
Flash Fire
127
0.00
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.15
fraction
0.60
fraction
0.40
fraction
0.10
0.50
0.50
0.50
fraction
fraction
fraction
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.00
fraction
0.50
fraction
MPACT Theory
128
Purple Book Set

These parameter are distributed with SAFETI-NL to ensure a study compatible with the
Purple Book guidance.
NewSafetiNL (RunRow Night)
Parameters
Event Tree Probabilities
Probability of a BLEVE
Probability of a Pool
Fire
Toxic Probability
Immediate Ignition
Long
Duration
Horizontal Fraction
Long
Duration
Horizontal Jet Fire
Jet Fire
Short Duration Flash
Fire
Short
Duration
Explosion
Fire
Delayed
Ignition
Explosion
Immediate Ignition
Long
Duration
Horizontal Fraction
Long
Duration
Horizontal Jet Fire
Long
Duration
Horizontal Pool Fire
Long
Duration
Horizontal Jet Fire with
Pool Fir
1.00000
1.00000
fraction
fraction
1.00000
0.70000
fraction
Fraction
1.00000
Fraction
1.00000
Fraction
1.00000
Fraction
0.00000
Fraction
0.70000
Fraction
0.18000
Fraction
0.12000
Fraction
0.60000
Fraction
0.40000
Fraction
0.70000
Fraction
1.00000
Fraction
0.00000
fraction
0.00000
fraction
1.00000
fraction
0.00000
fraction
MPACT Theory

Pool Fire
Jet Fire
Jet Fire with Pool Fire
with Pool Fire
alone
Fire with Pool Fire
Fire Alone
Short
Duration
Explosion with Pool Fire
Short
Duration
Explosion Alone
Short Duration Pool Fire
Residual Pool Fire
Fire
Delayed
Ignition
Explosion
Instantaneous
no
Rainout
Immediate
Ignition
Instantaneous
no
Rainout BLEVE
Instantaneous
no
Rainout Immediate Flash
Fire
Instantaneous
no
Rainout
Immediate
Explosion
Instantaneous
no
Rainout
Delayed
Ignition Flash Fire
Instantaneous
no
Rainout
Delayed
Ignition Explosion
Instantaneous
with
Rainout
Immediate
Ignition
Instantaneous
with
129
0.00000
fraction
0.00000
fraction
1.00000
fraction
0.70000
fraction
0.00000
fraction
0.18000
fraction
0.00000
fraction
0.12000
fraction
0.00000
fraction
0.00000
fraction
0.00000
fraction
0.60000
fraction
0.40000
fraction
0.70000
fraction
0.70000
fraction
0.18000
fraction
0.12000
fraction
0.60000
fraction
0.40000
fraction
0.70000
fraction
0.70000
fraction
MPACT Theory
Rainout BLEVE with

Pool Fire
Instantaneous
with
Rainout BLEVE Alone
Instantaneous
with
Fire with Pool Fire
Instantaneous
with
Fire Alone
Instantaneous
with
Rainout
Immediate
Explosion with Pool Fire
Instantaneous
with
Rainout
Immediate
Explosion Alone
Instantaneous
with
Rainout Immediate Pool
Fire Alone
Instantaneous
with
Rainout Residual Pool
Fire
Instantaneous
with
Rainout
Delayed
Ignition Flash Fire
Instantaneous
with
Rainout
Delayed
Ignition Explosion
Immediate Ignition
Explosion
Given
Ignition
Long Duration Jet Fire
Short Duration Any
Ignition of Cloud
Short Duration Ignition
of Cloud with Pool Fire
Long
Duration
Horizontal Jet Fire with
Pool
Jet Fire with Pool
for Effects
not Flash Fire
130
0.00000
fraction
0.18000
fraction
0.00000
fraction
0.12000
fraction
0.00000
fraction
0.00000
fraction
0.00000
fraction
0.60000
fraction
0.40000
fraction
0.70000
0.40000
fraction
fraction
1.00000
0.50000
fraction
fraction
0.00000
fraction
1.00000
fraction
1.00000
fraction
0.00000
fraction
0.70000
fraction
MPACT Theory
Appendix B.
131
MPACT input Parameters Vulnerability Factors
Historic Set
The addition of flammable probit modelling means that the default behaviour of the model
is going to produce different results to previous versions. The vulnerability parameters are
also set independently for societal and individual risk and radiation and flammable probit
criteria.
The historic parameter settings are as follows;
NewSafeti (RunRow 1)
Default Parameters
Vulnerability Risk Parameters
Fraction of Population Outdoors for
Societal Risk
Fraction of Population Outdoors for
Individual Risk
Heavy Explosion Damage Outdoors
Heavy Explosion Damage Indoors
Light Explosion Damage Outdoors
Light Explosion Damage Indoors
Flash Fire Outdoors
Flash Fire Indoors
BLEVE Outdoors
BLEVE Indoors
Jet Fire Outdoors
Jet Fire Indoors
Pool Fire Outdoors
Pool Fire Indoors
Toxics Outdoors
Toxics Indoors
0.10
fraction
1.00
fraction
0.30
1.00
0.10
0.30
1.00
0.10
0.70
0.20
0.70
0.10
0.70
0.10
0.90
0.10
fraction
fraction
fraction
fraction
fraction
fraction
fraction
fraction
fraction
fraction
fraction
fraction
fraction
fraction
MPACT Theory
132
SAFETI-NL and Standard Product Defaults

The current product defaults are configured to apply flammable probit modelling and to
use the Purple Book guidance.
Parameters
Heavy
Explosion
Damage (Outdoors)
Heavy
Explosion
Damage (Indoors)
Light Explosion Damage
(Outdoors)
Light Explosion Damage
(Indoors)
Flash Fire (Outdoors)
Flash Fire (Indoors)
Fireball
Societal
Radiation Criteria Zone
(Outdoors)
Fireball
Societal
(Indoors)
Fireball
Individual
(Outdoors)
Fireball
Individual
(Indoors)
Fireball
Societal
Flammable Probit Zone
(Outdoors)
Fireball
Societal
(Indoors)
Fireball
Individual
(Outdoors)
Fireball
Individual
(Indoors)
Jet
Fire
Societal
(Outdoors)
Jet
Fire
Societal
(Indoors)
Jet
Fire
Individual
(Outdoors)
Jet
Fire
Individual
(Indoors)
Jet
Fire
Societal
(Outdoors)
Jet
Fire
Societal
1.00000
fraction
1.00000
fraction
0.00000
fraction
0.02500
fraction
1.00000
1.00000
1.00000
fraction
fraction
fraction
1.00000
fraction
1.00000
fraction
1.00000
fraction
0.14000
fraction
0.00000
fraction
1.00000
fraction
0.00000
fraction
1.00000
fraction
1.00000
fraction
1.00000
fraction
1.00000
fraction
0.14000
fraction
0.00000
fraction
MPACT Theory

(Indoors)
Jet
Fire
Individual
(Outdoors)
Jet
Fire
Individual
(Indoors)
Pool
Fire
Societal
(Outdoors)
Pool
Fire
Societal
(Indoors)
Pool Fire Individual
(Outdoors)
(Indoors)
Pool
Fire
Societal
(Outdoors)
Pool
Fire
Societal
(Indoors)
(Outdoors)
(Indoors)
Toxics (Outdoors)
Toxics (Indoors) (NL)
Toxics (Indoors) (Main)
133
1.00000
fraction
0.00000
fraction
1.00000
fraction
1.00000
fraction
1.00000
fraction
1.00000
fraction
0.14000
fraction
0.00000
fraction
1.00000
fraction
0.00000
fraction
1.00000
0.10000
1.00000
fraction
fraction
fraction
MPACT Theory
Appendix C.
134
MPACT input Parameters Remaining factors
Standard Set
These parameters (NL version) are configured to comply with the Purple Book
guidance.
Parameters
General Risk Parameters
Use Free Field Modelling
Free Field (Plant Boundary)
Distance to Site Boundary
10 m
Late Pool Fire
Include Effects
Minimum Case Frequency
1E-9
/AvgeYear
Minimum Event Probability
1E-9
Population Omega Factor
0.000168
Maximum Number of Subsquares across Ellipse
10.00
Maximum Number of Subdivisions per Square

Factor for Toxic F-N Spread
5.00
2
Grid Sizing
Calculated
Grid Bounds Minimum X
-1000 m
Grid Bounds Maximum X
1000 m
Grid Bounds Minimum Y
-1000 m
Grid Bounds Maximum Y
1000 m
Grid Calculation Method
Number of cells
Grid cell size
10 m
Maximum number of cells
Aversion Index
40,000.00
2
Indoor Population Omega Factor

0.000168
Number of wind subdivisions per sector
Method for handling Indoor/Outdoor risk
5.00
NL Indoor calculations
Inter-ellipse interpolation method

Weighted
Method option
Normal dispersion
Building damage method
Worst point
MPACT Theory
135
Number of X steps per view

Minimum X step
11.00
0.1 m
Number of time steps - continuous clouds

Between Cloud Views
5.00
Minimise Gaps
Pressure exceedance curves

Don't Calculate
Elevation of Floor or Ceiling
0 m
Concentration method for filling
Stoichiometric
Minimum probability of death for explosions
0.01
Minimum Pressure Filter
0.01
bar
Separation specification
Use Ratio
Critical Separation Ratio
0.5
Cloud Shape of Area Integration
Elliptical
Explosion efficiency method
100% efficiency
Explosion Type Calculation Method
Polynomial Curve-Fit Equations
Number of Blast Curve Discretization Points
30,000.00
Maximum No. effect points along transect

Low to medium criterion
2.00
0.006
Medium to high criterion

0.062
Options available
Volume Averaged
Method option:
Ground reflection
Reflection factor
1
Unconfined Explosion Strength
10
Explosion Efficiency
0.125
fraction
Flammable Mass Calculation Type
Area Weighted Mass Integral
Minimum Explosion Energy
5000000
kJ
Maximum number of time steps
100.00
Number of timesteps - time varying clouds

Active Shut Down
20.00
No Shut Down
Fraction of Population Indoors for Societal Risk

0.93
fraction
Fraction of Population Indoors for Individual Risk
0
fraction
MPACT Theory
Appendix D.
136
Errors, Warnings and Messages
MPACT uses the MDE system to report errors, warnings and messages. These are
explained below;
Errors
1
2
3
4
"Unable to open F-N output file"

"Error writing to F-N output file"
"Unable to open risk output file"
"Error writing to risk output file"
The above errors are reported if the code cannot succeed performing
normal file operations. It is conceivable that this is caused by running
multiple copies of MPACT at the same time, by accessing these files
using other applications or maybe these files have been set to readonly. These errors may also be produced if the machine is full in which
case freeing up some disk space may solve the problem. If this is not
the case it is necessary to contact technical support to report the
error.
5
6
7
8
"Consequence
"Consequence
"Consequence
"Consequence
data- This event will not be processed or contribute to the risk"

data - Flammable results are not valid"
data - Horizontal jet results are not valid"
data - Vertical jet results are not valid"
The above errors mean that there is a significant problem with the
consequence effect zones being passed into MPACT. Please check that the
release event appears to have sensible consequence results. If this is
the case then please report the error to technical support. If the
consequence results indicate problems then please adjust the release
event accordingly as a workaround and/or report the error to technical
support.
9 "Explosion damage model failed for (quasi)instantaneous cloud"
10 "Explosion damage model failed for continuous cloud"
11 "Explosion damage model failed for immediate ignition"
The three errors above indicate a failure in the damage zone model. This
is a very simple model and this error should never occur. Please report
the error to technical support with details of the study.
12 "CNS data- Toxic results are not valid"
The above error means that there is a significant problem with the
consequence effect zones being passed into MPACT. Please check that the
release event appears to have sensible consequence results. If this is
the case then please report the error to technical support. If the
consequence results indicate problems then please adjust the release
event accordingly as a workaround and/or report the error to technical
support.
13 "Heavy damage zone larger than light damage zone in parameters"
The parameters must be set so that the heavier damage zone is smaller or
the same as the light damage zone. Please reset the parameter values.
14 "%1%string%"
Error message designed to pass a text string used for debug purposes
MPACT Theory
137
15 "Explosion damage model failed - negative mass %1%Mass%"
The error above indicates a failure in the damage zone model. This is a
very simple model and this error should never occur. Please report the
error to technical support with details of the study.
16 "Error writing individual risk ranking data"
The above error is reported if the code cannot succeed performing normal
file operations. It is conceivable that this is caused by running
multiple copies of MPACT at the same time, by accessing these files
using other applications or maybe these files have been set to readonly. These errors may also be produced if the machine is full in which
case freeing up some disk space may solve the problem. If this is not
the case it is necessary to contact technical support to report the
error.
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
"Probability of User defined BLEVE immediate event out of range(0 to 1)"

"Probability of User defined pool fire immediate event out of range(0 to 1)"
"Horizontal probability out of range(0 to 1) route %1%integer%"
"Horizontal ignition probability out of range(0 to 1) route %1%integer% "
"Horizontal jet probability out of range(0 to 1) route %1%integer% "
"Horizontal explosion probability out of range(0 to 1) route %1%integer% "
"Vertical ignition probability out of range(0 to 1) route %1%integer% "
"Vertical jet probability out of range(0 to 1) route %1%integer% "
"Vertical explosion probability out of range(0 to 1) route %1%integer%"
"Probability of immediate ignition out of range(0 to 1) route %1%integer%"
"Horiz. frac. for vert. & horiz. cases out of range(0 to 1) route %1%integer%"
"Probability of horizontal jet fire out of range(0 to 1) route %1%integer%"
"Probability of pool fire (jet horiz.) out of range(0 to 1) route %1%integer%"
"Probability of pool & horiz.jet fire out of range(0 to 1) route %1%integer%"
"Probability of vertical jet fire out of range(0 to 1) route %1%integer%"
"Probability of pool fire (jet vert.) out of range(0 to 1) route %1%integer%"
"Probability of pool & vert. jet fire out of range(0 to 1) route %1%integer%"
"Freq. frac. for short release effects out of range(0 to 1) route %1%integer%"
"Probability of BLEVE & pool fire out of range(0 to 1) route %1%integer%"
"Probability of BLEVE w/o pool fire out of range(0 to 1) route %1%integer%"
"Probability of imm.flash fire & pool fire out of range(0 to 1) route %1%integer%"
"Probability of imm.flash fire w/o pool fire out of range(0 to 1) route %1%integer%"
"Probability of imm.expl.& pool fire out of range(0 to 1) route %1%integer%"
"Probability of imm.expl. w/o pool fire out of range(0 to 1) route %1%integer%"
"Probability of immediate pool fire out of range(0 to 1) route %1%integer%"
"Probability of delayed pool fire out of range(0 to 1) route %1%integer%"
"Probability of delayed flash fire out of range(0 to 1) route %1%integer%"
"Probability of delayed explosion out of range(0 to 1) route %1%integer%"
All the errors reported above relate to the input values for the event
tree parameters. Please correct the input value before running MPACT.
49
50
51
52
53
54
55
"Horizonal outcome probabilities(>1) = %1%string%"

"Vertical outcome probabilities(>1) = %1%string%"
"Horizontal continuous probabilities(>1) = %1%string%"
"Vertical continuous probabilities(>1) = %1%string%"
"Instantaneous-no pool outcome probabilities(>1) = %1%string%"
"Instantaneous-with pool outcome probabilities(>1) = %1%string%"
"Delayed outcome probabilities(>1) = %1%string%"
The errors reported above are triggered when the branch probabilities
sum to more than one on one or more of the event trees. Please correct
the input value(s) before running MPACT.
56 "Risk grid array dimensions not defined consistently"
This indicates a serious internal inconsistency. Please report the case

to technical support.
57 "Consequence data - Decreasing or overlapping ellipses not allowed"
MPACT Theory
138
This means the ellipses defining one of the radiation zones are poorly
conditioned and the outcome cannot be modelled.
The following errors relate to problems with the consequence data input;
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
"Number of weathers %1%integer% out of range"

"Number of toxic clouds %1%integer% out of range"
"Building type %1%integer% out of range"
"Number of toxic cloud steps %1%integer% out of range"
"Toxic cloud zero step %1%integer% out of range"
"Invalid toxic X step %1%Length%"
"Probability integral %1%Length% out of range"
"Probability of death %1%real% out of range"
"Toxic cloud data not set for weather %1%integer%"
"Number of flammable cloud steps %1%integer% out of range"
"Flammable cloud time not increasing on step %1%integer%"
"Flammable cloud distance not increasing on step %1%integer%"
"Flammable cloud radius %1%real% out of range"
"Flammable cloud mass %1%real% out of range"
"Flammable cloud mass not increasing on step %1%integer%"
"Flammable cloud does not overlap with ellipse for segment %1%integer%"
"Flammable ellipse B coefficient out of range for segment %1%integer%"
"Flammable duration out of range for segment %1%integer%"
"Flammable flow rate out of range for segment %1%integer%"
"Flammable route number out of range %1%integer%"
The following errors are given if the resulting Mpact grid would be too
thin. Typically this will be due to extreme separation of events in the
north-south or east-west axes.
78 "Less than 5 X direction grid cells %1%integer% check inputs"
79 "Less than 5 Y direction grid cells %1%integer% check inputs"
80 "Invalid plant boundary definition"
The error relates to the plant boundary input, please check for sensible
input.
81 "Plant boundary not on the risk grid"
Plant boundary location is off the risk grid so cannot be used. At least
part of the plant boundary must be located on the risk grid.
82 "Event number %1%integer% outside plant boundary"
This event is outside the plant boundary that has been defined.
83 "Plant boundary not defined"
No plant boundary has been defined in circumstances where the boundary

is required by the modelling eg use of the free field methods
84 "%1%integer% volume based explosion categories exceeds limit of 10"
There is a hard limit of 10 for the number of volume based explosion

categories reduce the number to 10 for the model to be able to run
85 "Volume based
86 "Volume based
%1%integer%"
87 "Volume based
88 "Volume based
89 "Volume based
explosion volume decreases at step %1%integer%"

low flame speed explosion probability decreases at step
medium flame speed explosion probability decreases at step %1%integer%"
high flame speed explosion probability decreases at step %1%integer%"
flame speed boundary decreases"
The volume based arrays must increase redefine the inputs so that they
increase sequentially
90 "3D method cannot use old method flag"
MPACT Theory
139
To run the 3D modelling change the parameter to enable the model to run
91 "3D method must be run with both outdoor and indoor risk grids"
To run the 3D modelling both indoor and outdoor grids must be modelled
change the parameter to enable the model to run
92 "Mpact will not run because there are too many building types (>100)"
Currently there is an upper limit of 100 to the number of building types

Warnings
1001 "Too many flammable sub-squares - results will be lost"
1002 "Try re-running with a larger grid spacing or fewer subdivisions"
In subdividing the MPACT risk grid cells, MPACT uses a finite memory
stack. If this stack is insufficient to accommodate all the sub-square
variables then some information is lost. The model carries on but the
effect of this problem on the results is unpredictable. The parameters
max subdivisions of square and Max subsquares when dividing can be
reduced to limit the memory requirement.
1003 "Consequence data- No Fireball results for above event (Liq.Fr. may be 1)"
1004 "Consequence data- Immediate pool fire results are not valid"
1005 "Consequence data- Late pool fire results are not valid"
These warnings are issued if these conditions are determined in the

input data for the case. They may not be problem but the warnings are
issued in case these effects are expected to be modelled.
1006 "N= %1%integer% too high for the F-N curve maximum N"
The N limits are hard wired internally within MPACT so this warning is
issued if the N value exceeds the scale. The limit is greater than 106 so
this is an extreme situation.
1007 "More than %1%integer% ranking points - last ones ignored"
If the maximum number of ranking points is exceeded this warning

indicates that the last points in the list will be ignored. Only the
user can determine whether this is significant or not. If so it will be
necessary to limit the ranking points input to the model so that the
significant points are processed.
1008 "Cell width increased to avoid exceeding maximum number of cells %1%integer%"
The input cell width may result in an excessive number of cells and so
the cell width may be adjusted to avoid exceeding the maximum number of
cells parameter.
1009 "Flammable ellipse width reducing"
This warning is issued when the ellipse width parameter reduces with
reducing lethality level. Logically the width should increase but the
use of the ellipse shape as an approximation can give this behaviour.
1010 "Consequence data- Standalone pool fire is not present for weather
%1%integer%"
MPACT Theory
140
Normally a standalone pool fire will have results for each weather. This
warning is given is one of the results is missing. This may be valid
because of the weather state or it may indicate a failure of the pool
fire modelling.
1011 "Consequence data- zero instantaneous cloud width for weather and step %1%string%"
If the cloud has a zero width then it will not give any delayed ignition
probability or flash fire hazards.
1012 "Flammable cloud does not overlap with ellipse for segment %1%integer%"
Normally the flammable cloud records should lie within the flammable
effect ellipse. This warning would suggest a problem in the consequence
data calculations.
1013 "Toxic cloud data not set for weather %1%integer%"
Sometimes one or more weather condition may not result in a

significantly toxic cloud.
The following warnings relate to the hazard zones and their location
with respect to the Mpact grid. If the zones go outside the overall
boundary of the grid for instance because the grid size has been
specified manually then the following warnings will be given;
1014
1015
1016
1017
1018
1019
1020
1021
"Hazard near or over risk grid limit, west boundary should be < %1%Length%"
"Hazard near or over risk grid limit, east boundary should be > %1%Length%"
"Hazard near or over risk grid limit, south boundary should be < %1%Length%"
"Hazard near or over risk grid limit, north boundary should be > %1%Length%"
"Risk grid west boundary should be < %1%Length%"
"Risk grid east boundary should be > %1%Length%"
"Risk grid south boundary should be < %1%Length%"
"Risk grid north boundary should be > %1%Length%"
1022 "Flammable cloud mass adjusted on step %1%integer%"
This warning indicates that Mpact has made an adjustment to the

flammable mass because of some inconsistent input. Typically this will
be a tiny adjustment and due to machine precision issues.
1023 "Flammable cloud does not overlap with ellipse for segment %1%integer%"
This warning means that the ellipse dimensions and the cloud steps do
not coincide. This usually means that the cloud is not flammable at the
flammable effect height
1024 "Free field option %1%integer% not allowed with 3D modelling - setting to no free
field"
If this parameter is set to model free field scenarios it is turned off

in the model since the rules for free field scenarios have not been
established for use with the 3D modelling
1025 "No cloud time steps for weather %1%integer% so no flash fire, explosion or residual
pool fire outcomes"
Messages
2001 "F-N curve produced, Rate of Death = %1%real% /yr"
Issued if F-N results are generated. The Rate of Death is for all the
events that go to make up the curve.
MPACT Theory
141
2002 "F-N curve not produced (no population impact)"
This can happen if there is no population or if the events do not impact

the population. If an F-N curve is expected please investigate the
effect zones and their position with respect to the defined population.
2003 "Risk grid produced, Maximum Risk = %1%real% /yr"
Normally a risk grid will be produced.

2004 "Risk grid not produced (no risk impact)"
This may be caused by a number of factors. Most likely is that the

events provided as input to MPACT do not represent any significant
hazard relative to the MPACT grid size.
2005
2006
2007
2008
2009
2010
2011
2012
2013
"Files initialised, Output Level %1%real%"

"Project Parameters initialised"
"Event Tree Probabilities initialised"
"Risk Parameters initialised"
"Weather Probabilities initialised with %1%integer% categories"
"Grid initialised with spacing %1%Length%"
"Ranking Points initialised with %1%integer% points on the grid"
"Ignition Data initialised with %1%integer% objects"
"Population Data initialised with %1%integer% objects"
The simple diagnostic messages above report the initialisation steps

followed in starting MPACT.
2014 "Event %1%integer% contributed to FN Curve (F>0,N=>0)"
This message is issued if the event generates any outcome where F>0.
This includes events where N is zero so there may still not be a
significant FN curve.
2015 "Event %1%integer% did not contribute to FN Curve (F=0)"
This means that the event does not generate any significant FN results,
even outcomes where N=0.
2016 "Risk preference selected to follow 5.x modelling"
This reports that the preference to emulate older versions of MPACT has
been selected. Normally this is not recommended because of improvements
to the model in the latest version and is provided only for situations
where the old results must be generated.
2018 "Free Field ignition selected (no societal risk calculated)"
This is a reminder that when the Free Field option is selected there
will be no societal risk results.
2019 "Entered %1%string%"
Used for debug purposes.

2020 "No risk ranking points defined"
This message reports simply that there are no ranking points provided as
data input for this run of MPACT.
2021 "Case initialised to be modelled as event %1%integer% at one location"
The case will be modelled as one event at one location.
MPACT Theory
142
2022 "Case initialised to be modelled at %1%integer% locations"

2023 "First event number %1%integer%"
2024 "Last event number %1%integer%"
The case will be modelled using a number of events distributed along a

route.
2025 "Case modelled as BLEVE only with probability %1%real%"
2026 "Case modelled as Pool Fire only with probability %1%real%"
2027 "Case modelled as Toxic with probability %1%real%"
The above message is issued for these non-event tree events.

2028 "Weather Category %1%integer%"
Reports the current weather category.

2029 "Case modelled using Flammable Event Tree Route %1%integer%"
2030 "Immediate Ignition branch probability %1%real%"
Reports the flammable event tree route number to be used for the event
and the weather category. It also gives the immediate ignition
probability that will be used by MPACT.
2031 "Number of cells X direction = %1%integer%"
2032 "Number of cells Y direction = %1%integer%"
Reports the number of cells used in each direction for the MPACT
calculation grid.
2033 "Flammable ellipse width reducing suggests less accurate representation"
Normally the ellipse width will increase as the lethality decreases.

This warning is generated if this is not the case. The modelling will
continue but it suggests that the elliptical representation is not a
very accurate of the true radiation zone shape.
2034 "Delayed Ignition branch probability %1%real%"
Reports the probability at the head of the delayed ignition branch of

the event tree. If the free field ignition method is being used then
this will be the overall probability of delayed ignition. This will not
necessarily be the probability of delayed ignition if actual ignition
sources are being used since in this case the cloud still needs to find
ignition sources. Further information on the delayed ignition is
reported in the diagnostic file Non Ignition Probabilities.csv.
2035 "No delayed ignition and toxic branch of both event will use calculated probability of
%1%real%"
For a toxic and flammable event being run in both mode the delayed
ignition branch is being replaced by the toxic effects. This occurs when
the non-ignition probability is NOT specified by the user.
2036 "Both event calculated delayed Ignition branch probability %1%real%"
For a toxic and flammable event being run in both mode the delayed
ignition branch is run as well as the toxic effects. This occurs when
the non-ignition probability IS specified by the user.
2037 "Toxic branch of both event will use prescribed probability of %1%real%"
MPACT Theory
143
For a toxic and flammable event being run in both mode this is the
probability being assigned to the toxic effects.
2038 "Immediate and delayed ignition branches both inactive - no flammable risk"
The model has determined that there will be no flammable risk.

2039 "Immediate ignition probability enhanced due to elevated release temperature
%1%Temperature%"
This can be the case when the Purple Book ignition method is being used.
2040 "Warning or Error no %1%string%"
General message and meaning depends on actual text.

2041 "Plant Boundary Initialised with %1%integer% points"
Information message indicating that the plant boundary is defined and

indicating the number of points.
2042 "Immediate pool fire duration is %1%Time%"
2043 "Early pool fire duration is %1%Time%"
2044 "Late pool fire duration is %1%Time%"
These messages give the calculated times for the various pool fires.
These are used only in conjuction with the radiation intensity method
for pool fire vulnerability
2045 "Using direction-dependent non-ignition probabilities for toxic risk"
This message indicates

probabilities from the
modelling for a both
ignition probabilities
MPACT
that the model is using the non-ignition

flammable modelling to link in with the toxic
model on a directional basis. The actual nonare written to the diagnostic files.
2046 "Cloud modelled with %1%integer% timesteps"
Appendix E.
Outcome Codes
MPACT uses the MDE system to report errors, warnings and messages. These are
explained below;
E.1
Outcome Codes 2D Modelling
The table below gives the codes that will be used if you do not have a license for the
Extended Explosion Modelling, or if you do have a licence but have set the Flash Fire
and Explosion Method to 2D Damage Zone in the Study Folder dialog.
Code
Event Type
Rainout
Ignition
Effect Type
CNDFO
Continuous release
No rainout
Delayed
Flash fire Only
CNDFOF Continuous release
No rainout
Delayed
Flash fire Only, Free

field ignition
MPACT Theory
144
CNDFXF Continuous release
No rainout
Delayed
eXplosion Only, Free

field ignition
CNDXO
Continuous release
No rainout
Delayed
eXplosion Only
CNIBO
Continuous release
No rainout
Immediate fireBall Only
CNICFO Continuous release
No rainout
Continuous cloud
Immediate shape,
Flash fire Only
CNIIFO
Continuous release
No rainout
Instantaneous cloud
Immediate shape,
Flash fire Only
CNIHJO
Continuous release
No rainout
Immediate Horizontal Jet fire

Only
CNIVJO
Continuous release
No rainout
Immediate Vertical Jet fire Only
CNIXO
Continuous release
No rainout
Immediate eXplosion Only
CRDFO
Continuous release
with Rainout Delayed
Flash fire Only
CRDFP
Continuous release
Flash fire with

additional Pool fire
effects
CRDFOF Continuous release

field ignition
CRDFPF Continuous release
Flash fire with

effects, Free field
ignition
CRDFXF Continuous release

field ignition
CRDPO
Continuous release
Residual Pool fire

effects (i.e. separate
from cloud)
CRDXO
Continuous release
eXplosion Only
MPACT Theory
145
CRIBO
Continuous release
with Rainout Immediate fireBall Only
CRIBP
Continuous release
with Rainout Immediate fireBall with

effects
CRICFO Continuous release
with Rainout Immediate Continuous cloud

shape,
Flash fire Only
CRICFP
Continuous release
with Rainout Immediate Continuous cloud

shape,
Flash fire with
effects
CRIIFO
Continuous release
with Rainout Immediate Instantaneous cloud

shape,
Flash fire Only
CRIIFP
Continuous release

shape,
Flash fire with
effects
CRIHJO
Continuous release
with Rainout Immediate Horizontal Jet fire

Only
CRIHJP
Continuous release
with Rainout Immediate Horizontal Jet fire

with additional Pool
fire effects
CRIPOE
Continuous release
with Rainout Immediate long-duration Pool fire

Only
CRIPOS
Continuous release
with Rainout Immediate short-duration Pool

fire Only
CRIVJO
Continuous release
with Rainout Immediate Vertical Jet fire Only
CRIVJP
Continuous release
with Rainout Immediate Vertical Jet fire with

effects
MPACT Theory
146
CRIXO
Continuous release
with Rainout Immediate eXplosion Only
CRIXP
Continuous release
with Rainout Immediate eXplosion with

effects
INDFO
Instantaneous release
No rainout
Delayed
Flash fire Only
INDFOF Instantaneous release
No rainout
Delayed

field ignition
INDFXF Instantaneous release
No rainout
Delayed

field ignition
INDXO
No rainout
Delayed
eXplosion Only
INIBO
No rainout
INIBOT
No rainout
Immediate fireBall Only,

Transportation source
INIIFO
No rainout
Immediate Instantaneous cloud

shape,
Flash fire Only
INIXO
No rainout
Immediate eXplosion Only
IRDFO
Flash fire Only
IRDFOF

field ignition
IRDFP
Flash fire with

effects
IRDFPF
Flash fire with

effects, Free field
ignition
IRDFXF
eXplosion, Free field

ignition
MPACT Theory
147
IRDPO
residual Pool fire

effects Only
(i.e. separate from
cloud)
IRDXO
eXplosion Only
(Pool fire effects are
not included with
explosion)
IRIBO
with Rainout Immediate fireBall Only
IRIBP

effects
IRIBPT

effects, Transportation
source
IRIIFO

shape,
Flash fire Only
IRIIFP

shape,
Flash fire with
effects
IRIPOS
with Rainout Immediate Pool fire Only
IRIXO
with Rainout Immediate eXplosion Only
IRIXP
with Rainout Immediate eXplosion with

effects
SAIBO
StandAlone fire model -
SAIPO
StandAlone fire model -
Immediate Pool fire Only
TO
Toxic effects for a release in which Only toxic effects were modelled
MPACT Theory
Toxic effects for a release in which Both toxic and flammable effects
were modelled
TB
E.2
148
Outcome Codes 3D Modelling
The table below gives the additional codes that will be used if you have a license for the
Extended Explosion Modelling and set the Flash Fire and Explosion Method to 3D
Obstructed Region in the Study Folder dialog or using the 3D Purple Book modelling.
Code
Event Type
Rainout
Ignition
Effect Type
CNdFFO Continuous release
No rainout delayed
Flash Fire Only
CNdFXO Continuous release
No rainout delayed
Flash fire with eXplosion
CNiFFO
Continuous release
No rainout immediate Flash Fire Only
CNiFXO Continuous release
No rainout immediate Flash fire with eXplosion
CRdFFO Continuous release
With
Rainout
delayed
Flash Fire Only
CRdFXO Continuous release
With
Rainout
delayed
Continuous release
With
Rainout
immediate Flash Fire Only
CRiFXO Continuous release
With
Rainout
immediate Flash fire with eXplosion
Continuous release
With
Rainout
delayed
Flash Fire with Pool fire
CRdFXP Continuous release
With
Rainout
delayed

and Pool fire
Continuous release
With
Rainout
immediate Flash Fire with Pool fire
CRiFFO
CRdFFP
CRiFFP
MPACT Theory
149
Continuous release
With
Rainout

and Pool fire
CRdPOR Continuous release
With
Rainout
delayed
CRiFXP
Residual POol fire
Instantaneous
release
No rainout delayed
Flash Fire Only
INdFXO Instantaneous
release
No rainout delayed
INdFFO
INiFFO
Instantaneous
release
No rainout immediate Flash Fire Only
INiFXO
Instantaneous
release
No rainout immediate Flash fire with eXplosion
IRdFFO
Instantaneous
release
With
Rainout
delayed
Flash Fire Only
IRdFXO
Instantaneous
release
With
Rainout
delayed
IRiFFO
Instantaneous
release
With
Rainout
immediate Flash Fire Only
IRiFXO
Instantaneous
release
With
Rainout
IRdFFP
Instantaneous
release
With
Rainout
delayed
IRdFXP
Instantaneous
release
With
Rainout
delayed

and Pool fire
IRiFFP
Instantaneous
release
With
Rainout
immediate Flash Fire with Pool fire
IRiFXP
Instantaneous
release
With
Rainout

and Pool fire
IRdPOR
Instantaneous
release
With
Rainout
delayed
Residual POol fire
MPACT Theory
IRiPOS
Instantaneous
release
150
With
Rainout
immediate Pool fire Only
CNfFFO Continuous release
No rainout free field
Flash Fire Only
CNfFXO Continuous release
CRfFFP
Continuous release
With
Rainout
free field
CRfFXP
Continuous release
With
Rainout
free field

and Pool fire
CRfFFO
Continuous release
With
Rainout
free field
Flash Fire Only
CRfFXO Continuous release
With
Rainout
free field
INfFFO
Instantaneous
release
Flash Fire Only
INfFXO
Instantaneous
release
IRfFFO
Instantaneous
release
With
Rainout
free field
Flash Fire Only
IRfFXO
Instantaneous
release
With
Rainout
free field
IRfFFP
Instantaneous
release
With
Rainout
free field
IRfFXP
Instantaneous
release
With
Rainout
free field

and Pool fire
SAIBL
StandAlone BLast
Immediate BLast effects only
NOMENCLATURE
MPACT Theory
151
Subscripts
indicates condition for probability (given that), i.e. |o means given that
outcome o occurs
Used to denote parameter dependent on the following types of fatality effect;
Toxic effects (=T )

Pool fires (=P )
Jet fires (=J )
Bleves (=B )
Flash Fires (=F )
Explosion damage zone 1 (inner circle) (=R1 )
Explosion damage zone 1-2 (outer doughnut) (=R2 )
c
o
u
s
w
i
x,y,i
1
2
c,i
e,i
s
edf,o
stability class (A-G)

outcome; for toxic calculations this is the combination ws of weather category
(w) and wind sector (s); for flammable calculations this is the combination ws
of flammable effect , weather w, and wind sector s
range for windspeed (m/s)
wind sector
weather category: combination of stability class c range of windspeed (u) and
Greek symbols
Fraction of the cloud energy content that contributes to the blast
Fraction of the ignition source i within the flammable cloud
Fraction of the ignition source i within grid cell x,y
Effective toxic width angle (rad)
Release direction in MPACT grid, defined as clockwise angle from South axis
(rad)
Lower release direction angle to affect a given x,y point (rad)
Upper release direction angle to affect a given x,y point (rad)
Angle defining centreline position of sector i
Angle defining centreline position of the upper boundary of sector i
Wind sector angle (rad)
Standard deviation as derived from the sum of variance for a given toxic
outcome
du
i
x,y,t
Out
people
Continuous ignition effectiveness factor

Ignition effectiveness factor for the grid cell
Ignition effectiveness per outdoor person (parameter)
In
people
Ignition effectiveness per indoor person (parameter)
ai
A
ax,y
bx,y
Latin symbols
Aversion index used to calculate the risk integral
Distance from release point to toxic risk calculation (m)
First constant defining the equivalent ignition effectiveness for the cell
Second constant defining the equivalent ignition effectiveness for the cell
MPACT Theory
Cs
Eexpl
f IROut
Out
f FN
152
Explosion damage zone coefficient

Explosion energy (J), set as product of flammable mass in the cloud and heat
of combustion
Fraction of people outdoors for the purposes of the individual risk calculations
Fraction of people outdoors for the purposes of the societal risk calculations
IRx,y,o
K IR
Presence factor for a single ignition source(the proportion of time the source
is present and active)
Normalisation factor for the F-N spreading algorithm
Spreading factor as number of standard deviations to use (parameter)
Is the presence factor for the grid cell derived from the input ignition sources
Frequency of vehicles along a line of traffic
Frequency of events giving a an N value between N1 to N2
Frequency of the incident
Frequency of an incident outcome o
Frequency of an incident outcome o with the rounded up integer value of N
Frequency of an incident outcome o with the rounded down integer value of
N
Fraction of the risk grid square at x,y covered by the effect-level ellipse i for a
given outcome
Fraction of the risk grid square at x,y between the effect-level ellipse i and the
next one for a given outcome
Heat of combustion for the flammable material in the cloud
Ignition potential factor for ignition source i
Individual risk at a point for the sum of all events (edfs)
Individual risk at a point due to the sum of risk increments for an edf
Individual risk at a point due to the sum of risk increments for an edf
Individual risk at a point due to the sum of risk increments for an edf given the
weather occurs
Individual risk at a point due to the sum of risk increments for an edf given
outcome o occurs
Individual risk at a point due to the sum of risk increments for an edf given the
weather
Individual risk at a point due an outcome o for a specific edf
Overall factor to be applied to the calculation of individual risk for the type of
K FN
effect,
Overall factor to be applied in the societal risk calculation to the number of
K IR , Li
people killed for the type of effect,

Overall factor to be applied to the calculation of individual risk for the type of
K FN , Li
effect, and lethality level i

Overall factor to be applied in the societal risk calculation to the number of
fi
fn,o
fs
f x,y
fv
F[ N1, N 2 ]
Fedf
Fedf,o
Fedf,h,o
Fedf,l,o
Frx,y,,Li|o
Frx,y,,Li|o
Hcomb
ii
IRTot,x,y
IRx,y
IRedf,x,y
IRedf,x,y|w
IRx,y|o
IRx,y|w
people killed for the type of effect, and lethality level i

Out
Fatality factor for the people outdoors for the type of effect,
In
Fatality factor for the people indoors for the type of effect,
Length of a line of traffic
Length of ellipse
Flammable mass for entire cloud
K
lt
Lellipse
Mc
MPACT Theory
Mcseg
nmaxellipse
Nv
N1
N2
Ndi,t
Nedf|o
N edf ,o
N In edf ,o
N Out edf ,o
Nexposed|o
Nh,o
Nl,o
Nsub
Nx,Ny
nws
nx,y
Nx,y
Nx,y|o
N x, y o
N In x , y o
N Out x , y o
P|w
Pbl
Pbp
Pc|o
Pcl
Pd|w
Pd,x,y
Pd,x,y|o
Pd i
153
Flammable mass for a particular cloud segment

Maximum number of sub-squares across ellipse (parameter)
Number of vehicles along a line of traffic
Lower N value for an F-N curve bucket
Upper N value for an F-N curve bucket
Number of people within the flammable region of a cloud used to calculate the
population ignition effectiveness at a particular time
Number of fatalities for a given edf and outcome o
Mean number of fatalities for a given edf and toxic outcome (indoor and
outdoor)
Mean number of indoor fatalities for a given edf and toxic outcome
Mean number of outdoor fatalities for a given edf and toxic outcome
Sum of number of people exposed for a given toxic outcome
Upper N value for F-N spreading
Lower N value for F-N spreading
Maximum times to sub-divide a square (parameter)
Dimensions of risk grid as number of grid cells in each direction
Number of wind sectors
Population density per unit area in a given grid cell
Population in a given grid cell
Number of fatalities in a given square given that outcome o occurs
Mean number of fatalities for a toxic outcome in grid cell x,y givent that the
outcome occurs (outdoor and indoor)
Mean number of indoor fatalities for a toxic outcome in grid cell x,y givent
that the outcome occurs
Mean number of outdoor fatalities for a toxic outcome in grid cell x,y givent
that the outcome occurs
Probability of a specific release angle with respect to the MPACT grid given
the weather
Conditional probability of a BLEVE
Conditional probability of a combined BLEVE and pool fire
The conditional probability that the effective cloud covers the calculation
square, given that the outcome o occurs
Centreline probability of death
Probability of death given the weather w and the release angle
Probability of death at a particular location x,y
Probability of death at a particular location x,y for a given outcome
Probability of death for ellipse i and flammable effect ,
Pd , IR i
Probability of death for individual risk for ellipse i and flammable effect ,
Pd , FN i
Probability of death for societal risk for ellipse i and flammable effect ,
Conditional probability of delayed ignition (function of time)
Conditional probability of an explosion
Conditional probability of a combined explosion and pool fire
Conditional probability of a flash fire
Conditional probability of a combined flash and pool fire
Pdi,t
Pe
Pep
Pff
Pfp
MPACT Theory
Ph
Pi
Pi ,du
PI
Pi,t
Pi,t,A
Pi,t,U
Px,y,t
Pirp
Pj
Pjp
Pni
Po
Pp
Prp
PS
Pv
Pw
Pws
Pwss
Pws
Pus
PLLI
PLLS
Rd
RILUP
RIai
So
T
T1
t2
Vx,y|o
WPI
Wellipse
x
xb
xcentre
xgrid
xf
xs
xsub
xmin
xmax
y
154
Conditional probability of the horizontal direction

Conditional probability of immediate ignition
User defined probability of delayed ignition
Crosswind probability integral
Probability of ignition due to ignition source i at time t
Probability of ignition due to ignition source i at time t for the area A
Probability of ignition due to ignition source i at time t for the area U
Probability of ignition due to ignition source in grid cell x,y in time t
Conditional probability of ignition of a residual pool fire
Conditional probability of a jet fire
Conditional probability of a combined jet and pool fire
Conditional probability of non ignition of the instantaneous dispersion branch
Conditional probability of the outcome
Conditional probability of a pool fire
Conditional probability of a residual pool fire
Conditional probability of short duration effects given a short duration release
Conditional probability of the vertical direction
Conditional probability of the wind speed/stability combination
Conditional probability of the wind speed/stability and wind sector
combination
Conditional probability of the wind speed/stability and wind sector subdivision
combination
Interpolated value of Pws given a specific angle
Conditional probability of the wind speed and wind sector combination
Potential Loss of Life result based on individual risk array
Potential Loss of Life result based on FN curve array
Overall rate of death
Risk Integral for land use planning
Risk Integral as calculated based on aversion index, ai
Spread variable for F-N spreading calculated by outcome
Used to denote time (s)
First time interval for derivation of an equivalent ignition source in a grid
cell (s)
Second time interval for derivation of an equivalent ignition source in a grid
cell (s)
Variance for a grid cell and particular toxic outcome
Effective toxic width (m)
Width of ellipse
Used to denote the east-west position in the MPACT grid, also the downwind
distance from the release point (m)
Back edge of a cloud segment (m)
Cloud centre (m)
Width of the risk grid cell (m)
Front edge of a cloud segment (m)
Segment centre (m)
Width of a sub cell during the calculation of Frx,y (m)
x coordinate of the South West corner of the MPACT grid (m)
x coordinate of the North East corner of the MPACT grid (m)
Used to denote the north-south position in the MPACT grid (m)
MPACT Theory
ymin
ymax
155
y coordinate of the South West corner of the MPACT grid (m)

y coordinate of the North East corner of the MPACT grid (m)
MPACT Theory
156
REFERENCES
1
Ale, B.J.M.,Risk Analysis and Risk Policy in the Netherlands, Risk, SRA, Nov 1988.
Hirst, I.L. and Carter, D.A,A Worst Case Methodology for Risk Assessment of Major Accident
Installations,Process Safety Progress, p78, Vol 19, No.2, Summer 2000
3
Carter, D.A. Hirst, I.L.,Maddison, T.E.,Porter, S.R., Appropriate Risk Assessment Methods for Major
Accident Establishments, Trans IChemE, Vol 81, Part B, 2003.
4
Okrent, D.,Industrial Risks, Proc R Soc London, A376, 133-149,1981
5
Scofield, S.L.,A Framework for Offshore Risk Criteria, Safety and Reliability Soc.J.,13(2),1993
6
Hirst, I.L. and Carter, D.A,A A Worst Case Methodology for Obtaining a Rough but Rapid Indication
of the Societal Risk from a Major Accident Hazard Installation, Journal of Hazardous Materials, A92, 223237, 2002
7
Laheij, G.M.H., Post, J.G. and Ale, B.J.M., Standard methods for land-use planning to determine
the effects on societal risk, Journal of Hazardous Materials, 71, 269-282, 2000.
8
Witlox, H.W.M., Unified Dispersion Model (UDM), Theory Manual, DNV, 2001
9
Committee for the Prevention of Disasters. Guidelines for Quantitative Risk Assessment. CPR 18E. Den
Haag, Sdu: 1999 (The Purple Book)
10
Committee for the Prevention of Disasters, Methods for the determination of possible damage to people
and objects resulting from releases of hazardous materials Green Book, CPR16E, The Hague:
Directorate-General of Labour of the Ministry of Social Affairs and Employment, First Edition (1992)
11
RADS Model, Internal MDE documentation, DNV, 2003
12
Worthington, D.R.E., MPACT Verification Manual, DNV, 2004
13
Worthington, D.R.E., Vuln Theory Manual, DNV, 2010
14
Wiekema, B.J., Vapour Cloud Explosions, Chapter 9 in Methods for the calculation of physical effects
of the escape of dangerous material, Yellow Book, Parts 1-2, 1st Ed., Director-General of Labour, Voorburg
TNO, 1979
15
TNO, Methods for the calculation of physical effects, (The Yellow Book), CPR14E, Parts 1-2, 3rd Ed.,
Sdu Uitgevers, The Hague, 1997
16 Baker, Q. A., Doolittle, C. M., Fitzgerald, G. A., Tang, M. J., 1997, Recent Developments in the BakerStrehlow VCE Analysis Methodology, Paper 42f presented to the 31st Loss Prevention Symposium,
American Institute of Chemical Engineers, 1997
17
TNO, Methods for the calculation of physical effects, (The Yellow Book), CPR14E, Parts 1-2, 3rd Ed.,
Sdu Uitgevers, The Hague, 1997
18
Eggen, J.B.M.M., GAME: development of guidance for the application of the Multi-Energy Method,
TNO Report PML 1995-C44, August 1995
19
Mercx, W.P.M.,van den Berg, A.C. and van Leeuwen, D., Application of correlations to quantify the
source strength of vapour cloud explosions in realistic situations, Final report for the project: GAMES,
TNO Report PML 1998-C53, October 1998, NL
20
Woodward, J.L. and Crossthwaite, P.J., How to set explosion protection standards, Hydrocarbon
Processing, Dec 1995
21
Worthington, D.R.E. and Xu, Y,MDE Theory Document - OREM, DNV Software Technical
documentation, DNV 2010
22
Worthington, D.R.E.,MDE Theory Document - VULN, DNV Software Technical documentation, DNV
2010

Mpact Theory - Desbloqueado PDF

Uploaded by

Copyright:

Available Formats

Mpact Theory - Desbloqueado PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mpact Theory - Desbloqueado PDF

Uploaded by

Copyright:

Available Formats

What is the MPACT model used for?

What is the MPACT model used for?

What results are presented by the MPACT model?

What results are presented by the MPACT model?

DNV Software

Layout of the Document ................................................................................................ 2

Model Input .................................................................................................................... 4

Individual Risk Calculations......................................................................................... 7

Definition of Societal Risk ......................................................................................................14

Calculation Area .......................................................................................................... 19

Discretisation of the Calculation Area ....................................................................... 20

Weather Category and Wind Direction ..................................................................... 20

Risk Ranking Points..................................................................................................... 24

Conditional Probability Fatality Factors and Building Types................................. 24

Summary Maximum Effect Zones (SMEZ) Reporting ............................................ 27

Toxic Impact Calculations......................................................................................... 29

Toxic Effect Zone at the Grid Cell.............................................................................. 30

Wind Distribution ........................................................................................................ 33

Individual Risk Calculation ........................................................................................ 35

Societal Risk Calculation............................................................................................. 36

Societal Risk Calculations ........................................................................................... 14

Calculations Common to Toxics and Flammables................................................... 19

Definition of Individual Risk.....................................................................................................7

Calculation of F-N Pairs ..........................................................................................................36

Flammable Impact Calculations ............................................................................... 45

Event Trees ................................................................................................................... 45

Flammable Outcomes ..............................................................................................................46

Wind Directional Probabilities ................................................................................... 63

Wind Sector Subdivision .........................................................................................................63

Flammable Effect Zone Types .................................................................................... 64

Fireball/BLEVE (Single Zone Method) ..................................................................... 66

Jet and Pool Fires (Single Zone Method)................................................................... 67

BLEVE, Jet and Pool Fires (Flammable Probit Method)......................................... 69

BLEVE, Jet and Pool Fires (Radiation Intensity Method) ....................................... 74

Vapour-Cloud Flash Fire - Instantaneous Release ................................................... 74

Vapour-Cloud Explosion (Wiekema Purple Book method) ..................................... 84

Vapour-Cloud Explosion (3D Purple Book method)................................................. 86

Vapour Cloud Explosions (Confined Explosion Method Using 3D Cloud Views) 89

Overlap of Effect Zone with the MPACT Risk Grid ................................................ 94

Continuously Variable Pd Type of Effect Zone ......................................................... 96

Calculation of Individual and Societal Risk .............................................................. 97

Options available to specify the immediate ignition probability.......................................103

Delayed Ignition ......................................................................................................... 109

Treatment of Non-Integer Values of N .............................................................................102

Immediate and Delayed Ignition Branch Probabilities ............................................ 103

Single Effect Outcome ........................................................................................................98

Options for the delayed ignition branch of the Event Trees..............................................109

Ignition Source Modelling ......................................................................................... 112

Definition of Ignition Sources...........................................................................................112

Verification and Sensitivity Analysis....................................................................... 124

Future developments................................................................................................ 124

what final results are produced by the model

Layout of the Document

Consequence results for each of the events to be included in the analysis

Recent extensions require additional input data

is the direction of the release

Figure 4 - Generalised Event Tree Illustrating The Logic Applied In MPACT

2.2.5 Ignition Sources and Distribution

of 0.8, and a probability of ignition (when present) of 0.5 in 30 seconds.