STUDY REPORT 24/07/2018

N° DRA-17-164468-11021C

METHODOLOGY FOR RISK EVALUATION

OF TRANSPORTATION OF DANGEROUS
GOODS ALONG A ROUTE
Methodology for risk evaluation of transportation of
dangerous goods along a route

Accident Risks Division

Verneuil-en-Halatte (60)

Contributors: Marine DELAMOTTE, Clément LENOBLE, Thomas MARCON, Valérie

DE DIANOUS

Réf. : INERIS-DRA-17-164468-11021C Page 1 sur 13

PREAMBLE
This report has been drawn up on the basis of information provided to INERIS, of
available and objective data (scientific or technical) and of current regulations.
INERIS's responsability will not be incurred if the information communicated to it is
incomplete or erroneous.
The proposals or recommendations or equivalent that would be made by INERIS in
the context of an entrusted study may assist in decision-making. In view of the missions
of INERIS by its creation decree, INERIS does not intervene in the decision-making
itself. The responsibility of INERIS cannot therefore be substituted for that of the
decision-maker.
The recipient will use the results included in this report in its entirety or otherwise in an
objective manner. Its use in the form of extracts or summary notes will be made under
the sole and complete responsibility of the recipient. It is the same for any changes
that would be made.
INERIS disclaims all responsibility for each use of the report outside the destination of
the service.

Réf. : INERIS-DRA-17-164468-11021C Page 2 sur 13

 CONTENT

1 RISK INDICATORS ............................................................................................ 5

2 QUANTITATIVE RISK ASSESSMENT (QRA) PARAMETERS ......................... 7
2.1 Parameters for individual risk ....................................................................... 7
2.2 Parameters for societal risk ........................................................................ 10
2.3 Comparison of routes ................................................................................. 12

Réf. : INERIS-DRA-17-164468-11021C Page 3 sur 13

1 RISK INDICATORS
The risk indicators, used to perform quantitative risk assessments (QRA), are
presented in the following paper. They are of interest to the Transportation of
Dangerous Goods and routes comparisons.
This memo reminds the different parameters that should be considered in order to
determine these risk indicators.

Two risk indicators may be used:

• Individual risk
Individual risk represents the risk at a given point. It corresponds to an annual
likelihood for a person, who is assumed to be present, without protection, of dying
due to the occurrence of an accidental phenomenon occurring in an installation or
along a route which results in the release of a dangerous substance.
The individual risk is graphically represented by risk outlines around the installation
or route (see Figure 1): the area between two curves corresponds to an annual
occurrence probability range.

Figure 1: Example of individual risk around an installation

Individual risk is commonly used for land use planning around industrial
facilities. It can be useful to ensure the risk acceptability regarding local
criteria, such as the nature of vulnerabilities (hospitals, housings, ...) at a
given point in the vicinity of an establishment. This indicator is not usually
used in the context of transportation of dangerous goods.

Réf. : INERIS-DRA-17-164468-11021C Page 5 sur 13

• Societal risk
Societal risk corresponds to an annual probability that at least N persons are
simultaneously killed because of their presence within the impact area of an
accidental phenomena within an installation or along a route which results in the
release of a dangerous substance.
Societal risk can be represented as a frequency / severity curve (F/N curve). The
frequency is the frequency of having accidents causing N dead or more. N is the
number of dead and F the cumulative frequency of accidents with N or more deaths.
Figure 2 illustrates an F/N curve around an installation.

Figure 2: Example of representation of societal risk around a facility

In France1, this indicator is commonly used for routes comparison in regard

with road transport risks. Another use of this indicator is to ensure the
acceptability of a section. Worldwide, it is used for ranking facilities
(particularly in Great Britain and the Netherlands).

1See booklet 3 of the Road Tunnel Safety Records Guide: « Les analyses des risques liés au transport
des matières dangereuses – décembre 2005 : http://www.cetu.developpement-
durable.gouv.fr/IMG/pdf/Guide_dossier_securite-Fasc_3_cle081e51-1.pdf

Réf. : INERIS-DRA-17-164468-11021C Page 6 sur 13

2 QUANTITATIVE RISK ASSESSMENT (QRA) PARAMETERS

2.1 PARAMETERS FOR INDIVIDUAL RISK

To assess the probability of lethality at a given point (individual risk), several
parameters need to be considered:
• frequency of occurrence of a dangerous phenomenon (meaning prevention barriers
failure),
• specific effect probability (thermal, toxic, overpressure...),
• probability of exposure to an effect,
• probability of impact on health (lethal injury),
• presence of the individual.

Frequency of occurrence of the dangerous phenomenon

QRA start with the frequency of occurrence of a dangerous phenomenon.
This parameter depends on:
• loss of containment frequency,
• probability of failure of the safety barriers used to prevent/mitigate the dangerous
phenomenon,
• probability of ignition in the case of release of flammable substances: immediate or
delayed ignition,
• weather conditions: stability, wind speed, ...
• faction of combustion energy dissipated in the form of thermal energy (e.g. in case
of BLEVE),
• etc.

In regard with transportation risk, frequency is expressed in annual frequency per unit
of length (linear frequency).
The entire route can be divided into several "elementary steps" of a given length dx.
Between two positions x and x + dx, it is assumed that the linear frequency is constant;
the frequency on the elementary step is Fx = flinear.dx (see Figure 3).

x x+dx

Figure 3

Note: The choice of the "elementary step dx" should be optimized regarding
computation times and expected precisions. Especially it may be necessary to choose
a step small enough to take into account all specific punctual targets (Figure 4).

Réf. : INERIS-DRA-17-164468-11021C Page 7 sur 13

 Critical zone where
targets may be
forgotten

Nx Nx+dx

x x+dx

Figure 4

Probability of an effect
For a given dangerous phenomenon, all types of effects need to be considered
(thermal, toxic, overpressure...).

Probability of being exposed to the effects

This parameter depends on the effect area, which is linked to:
• wweather conditions,
• environnemental conditions (temperature, solar radiation),
• wind directions,
• probability of immediate/delayed ignition, fraction of combustion energy dissipated
as thermal energy, etc.

The probability of exposure for someone located at a point within the bandwidth defined
by the effect-area (effect band along the route) equals 1.
If the person is outside this bandwidth, the probability equals 0 (Figure 5).

P=0

P=1 Deff

Figure 5

REF. : INERIS-DRA-17-164468-11021C Page 8 sur 13

Note: At a given point, the frequency of being impacted by an effect can be evaluated.
This assessment is not performed in usual Transportation of Dangerous Goods QRA.
It is carried out for transport pipelines risk assessment, in France (for more information,
see the Gesip2 Guide). Thus, it is necessary to define the length L which may impact
someone:
• A person located at a point Y in the Deff located within the effects-area
(associated for instance to 1% lethality) will be affected by the dangerous
phenomenon if the loss of containment occurs between points A and B of the
route (Figure 6). The route length L on which the phenomenon occurrence could
impact a person located at Y is: L = 2.√(Deff2-y2).
• The impact frequency FY at point Y is defined as following:
FY = flin x 2.√(Deff2-y2)
With:
• Deff is the distance of effect (radius of the effect circle)
• y is the minimum distance between the point Y and the route.

Y
Deff Deff
y

A B
L = 2.√(Deff2-y2)

Figure 6

Probability of health effects (lethal injury)

The probability of getting killed by exposure to a dangerous phenomenon can be
calculated using probit functions. They are related to the observed dangerous
phenomenon, as well as the received dose (depending on the distance from the
source).
Lethality percentages are related to the distance from the source. In QRA studies,
lethality probabilities are evaluated between the source and a maximum distance
corresponding to a lethality threshold (often 1%).

Assuming a dangerous phenomenon occurring at a position x of the route, then the

probability, for a person located at the point Y, to be killed is Px. For a fixed Y, the Px
probability therefore depends on the position of x (Figure 7). Px will be evaluated for
the different positions of x, assuming that Y does not move.

2 Methodological Guide for carrying out a safety report on a transport pipeline (liquid or liquefied
hydrocarbons, natural or assimilated gas and chemicals) - Guide professionnel GESIP n°2008-01 –
Edition de janvier 2014 - http://www.gesip.com/

REF. : INERIS-DRA-17-164468-11021C Page 9 sur 13

 Y

PA = 1%
Px > 1% Lx

A x

Figure 7

Note: The potential personal protection can be considered (protective equipment, wall,
etc.).

Note: For individual risk, lethality probability at a point means lethality frequency of a
person located at this point. To map individual risks, iso-risk curves are plotted around
the infrastructure considering this probability of lethality at each point.
For societal risk (cf. § 2.2), lethality probability is only one of the parameters of fatalities
number assessment.
The frequency, or rather cumulative frequency, is necessary for the representation of
the F/N curves (see § 2.4) is therefore not equivalent to the probability of lethality of
the individual risk.

Probability of presence of people

For societal risk, the likelihood that people will be present depends on:
• presence distribution (day or night),
• intermittent occupancy (festival, camping site, stadiums, etc.).

Individual risk
The individual risk (IR) at a given point is the sum of the products, for each dangerous
phenomenon, of the frequency of occurrence of the dangerous phenomenon
(F(Sceni->n)) by the probability of an effect (Peff), by the exposure probability (Pexposure),
by the lethality probability (Plethality):
𝐼𝑅 = ∑ 𝐹(𝑆𝑐𝑒𝑛𝑖→𝑛 ) ∗ 𝑃𝑒𝑓𝑓 ∗ 𝑃𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒 ∗ 𝑃𝑙𝑒𝑡ℎ𝑎𝑙𝑖𝑡𝑦

2.2 PARAMETERS FOR SOCIETAL RISK

Frequency of an accident
The frequency of an accident related to a given dangerous phenomenon and to a given
effect is the product of the frequency of occurrence of the dangerous phenomenon by
the probability of an effect (Peff).
𝐹 = 𝐹(𝑆𝑐𝑒𝑛𝑖→𝑛 ) ∗ 𝑃𝑒𝑓𝑓

REF. : INERIS-DRA-17-164468-11021C Page 10 sur 13

Severity evaluation
For societal risk, it is necessary to assess the severity of each accident (i.e. fatalities
number).
The fatalities number is calculated for each scenario, at each point of the effect area,
with the number of exposed persons, the lethality probability and the presence
probability if necessary.
𝐷𝑒𝑓𝑓
𝑁=∫ 𝑛𝑖 ∗ 𝑃𝑝𝑟𝑒𝑠𝑒𝑛𝑐𝑒 ∗ 𝑃𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒 ∗ 𝑃𝑙𝑒𝑡ℎ𝑎𝑙𝑖𝑡𝑦 𝑑𝑟
𝑟=𝑠𝑜𝑢𝑟𝑐𝑒

The number of exposed people ni can be calculated as follows:

• In case of homogeneous population density in a zone i: the number of exposed
persons ni is the product of this effects area by the density.
• If the density is not homogeneous, the effect area is divided into homogeneous
meshes and the number of exposed persons ni is calculated for each mesh.
It is important to count everyone. The impacted persons number cannot be divided
by the number of meshes.
• In case of punctual targets, they are counted separately.

The values pi on the graph (Figure 8) basically represent different lethality percentages
(or lethality probability ). Then, based on the real number of persons present within the
effect-area (ni), the total number of fatalities can be calculated.

In a simplified QRA, assumptions may be made to consider a uniform lethality

percentage between two lethality effects-areas (Figure 8). The average or maximum
value of the lethality percentage is then selected in the area under consideration.
These lethality percentages are used to calculate the severity of each accident N
considering the number of persons ni present in different areas and the different
probability of lethality pi related to the distances of areas:
N = p1 x n1 + p2 x n 2 + p3 x n3

p3
p2
n3
p1 n2
n1

Figure 8

REF. : INERIS-DRA-17-164468-11021C Page 11 sur 13

F/N curves
To plot F/N curves corresponding to a route, it is necessary to aggregate all the
accidents along the route (occurring at different points x and leading to Nx deaths for
each effect). The effects distances (circles for instance) for each dangerous
phenomenon have to be moved along the route to cumulate the accidents.
The graph below (Figure 9) represents F/N curves of different scenarios that can occur
on the same route.

Figure 9

For reminder, an F/N curve is plotted in:

• ranking scenarios in terms of severity level (from Nmax to Nmin);
• calculating, from the most to the less important severity scenarios, the cumulative
frequencies;
• plotting the F/N curve with severity / cumulative frequency.

Thus, for a given number of dead N, the frequency corresponds to the sum of the
frequencies of accidents leading to at least N deaths.
The route is evaluated by comparing the F/N curve to the acceptability criteria when
these have been defined. For example, acceptability criteria have been defined by the
Dutch RIVM and by the British HSE.

2.3 COMPARISON OF ROUTES

Different routes can be compared to each other using F/N curves.
When the F/N curves of the routes are represented on the same graph, if one of the
curves is clearly "below" the other without crossing with any other curve, then the
corresponding route presents the lowest societal risk.
In any other case, a global indicator may be useful to choose the least risky route, in
addition to socio-economic considerations.
Different comparison indicators were found in the literature.

REF. : INERIS-DRA-17-164468-11021C Page 12 sur 13

Fatalities Expected Value
The expected value E corresponds to the area under the F/N curve, which is calculated
thanks to the following integral:
𝑓𝑚𝑎𝑥
𝐸=∫ 𝑁𝑖 𝑑𝑓
𝑓𝑖 =𝑓𝑚𝑖𝑛

with:
• E : fatalities expected value of having N dead per year on the studied section
• f i : cumulative frequency of having N dead
• Ni : number of fatalities.

Comparing fatalities expected values of the F/N curves of different routes makes it
possible to determine the least risky: it will be the one with the lowest expected value3.
However, if the values of expected values are close, no conclusion can be drawn.
It is an indicator that corresponds to an overall vision of societal risk and that does not
generally consider the aversion to disasters: an accident making 1 death every year
has the same weight as an accident causing 100 deaths every 100 years.

Comparison of areas between the acceptability curve and the F/N curve
Another suggestion is that the comparison of the different areas obtain between the
acceptability curve and each of the F/N curves.

Sources:
[1] Cassini, (1998). Road transportation of dangerous goods: quantitative risk
assessment and route comparison.

[2] CCPS (1998). Tools for making acute decisions with chemical process safety
applications. Cost-benefit analysis p.190.

[3] CETU (2005). Fascicule 3 du guide des dossiers de sécurité des tunnels routiers
« Les analyses des risques liés au transport des matières dangereuses » .

3 Comparison rules are presented in booklet 3 of the road Tunnel safety Record Guide: a ratio greater
than 10 is considered significant and makes it possible to make a decision without using other criteria.
A ratio of less than 3 is not significant and the use of other criteria is necessary. For ratios between 3
and 10, a sensitivity study is required.

REF. : INERIS-DRA-17-164468-11021C Page 13 sur 13