See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/7133128

Statistical analysis of major accidents in petrochemical industry notified to

the major accident reporting system (MARS)

Article  in  Journal of Hazardous Materials · October 2006

DOI: 10.1016/j.jhazmat.2004.12.042 · Source: PubMed

CITATIONS READS

63 2,097

3 authors:

Zoe Nivolianitou Myrto Konstandinidou

National Center for Scientific Research Demokritos National Center for Scientific Research Demokritos
73 PUBLICATIONS   819 CITATIONS    37 PUBLICATIONS   577 CITATIONS   

SEE PROFILE SEE PROFILE

Michalis Christou
European Commission
29 PUBLICATIONS   622 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Quantitative risk assessment of accidental releases of hazardous materials in an oil refinery and the surrounding area View project

European Working Group on Land-Use Planning View project

All content following this page was uploaded by Myrto Konstandinidou on 27 November 2017.

The user has requested enhancement of the downloaded file.

 Journal of Hazardous Materials A137 (2006) 1–7

Statistical analysis of major accidents in petrochemical industry

notified to the major accident reporting system (MARS)
Zoe Nivolianitou a,∗ , Myrto Konstandinidou a , Christou Michalis b
a Institute of Nuclear Technology-Radiation Protection, National Center for Scientific Research “Demokritos”, 15310 Ag. Paraskevi, Greece
b Major Accident Hazard Bureau (MAHB), Institute for Systems Engineering and Informatics, JRC Ispra 21020, Italy

Received 4 May 2004; received in revised form 22 December 2004; accepted 27 December 2004
Available online 2 May 2006

Abstract
The European major accident reporting system (MARS) was created within the framework of European Union (EU) directive 82/501, the
so-called “SEVESO” directive, and in order to register all the major industrial accidents notified to the European Union authorities from the
member states. Statistical analysis of these accidents offers significant data to the understanding and prevention of industrial accidents. This paper
makes an analysis of some characteristics of major accidents in the petrochemical sector included in MARS. The statistical analysis focused on
the main categorization fields of the MARS short reports and additionally a refinement of the immediate causes of major accidents with focus on
the organizational factors was attempted through the details provided in the full reports of the database.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Major accidents; SEVESO directive; MARS; Statistical analysis; Petrochemical industry

1. Introduction cal interest for preventing major industrial accidents and limiting
their consequences [3].
The major accident reporting system (MARS) has been devel- The criteria for the notification of an accident to the Commis-
oped and operated since 1984 by the Major Accident Hazard sion as provided for in Article 15 (and annex VI) of the SEVESO
Bureau (MAHB) at the Institute for Systems Engineering and II Directive are:
Informatics (ISEI) of the European Commission (EC) Joint
Research Centre (JRC) in Ispra, Italy, within the framework of 1. Substances involved (quantity of dangerous substance dis-
the SEVESO regulatory directives [1]. charged)
Major accidents have been defined in the SEVESO I regula- 2. Injury to persons and damage to real estate
tory environment for the chemical process industry, as “an occur- 3. Immediate damage to the environment (terrestrial, freshwa-
rence such as a major emission, fire or explosion resulting from ter, marine habitat and groundwater)
uncontrolled developments in the course of an industrial activ- 4. Damage to property
ity, leading to a serious danger to man, immediate or delayed, 5. Cross-border damage
inside or outside the establishment, and/or to the environment,
and involving one or more dangerous substances” [2].
Both SEVESO I and SEVESO II Directives require that com-
The SEVESO II Directive, which replaced and strengthened
petent authorities of the European Union (EU) member states
SEVESO I, includes a more succinct definition of what consti-
notify major accidents involving dangerous substances which
tutes a major accident based on precise quantitative threshold
occur in their own countries to the European Commission, except
criteria, which result in an overall lowering of the criteria for
for those related to nuclear, military, mining, transport or waste
notification. It also demands for notification of accidents or near
land-fill sites.
misses, which member states regard as being of particular techni-
Section 2 of this paper makes an overview of MARS database,
while Section 3 presents the statistical analysis of major acci-
∗ Corresponding author. Tel.: +30 210 6503744; fax: +30 210 6503744. dents in the petrochemical industry included in it. Section 4
E-mail address: [email protected] (Z. Nivolianitou). makes a refinement of general causes, with respect to, the

0304-3894/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhazmat.2004.12.042
2 Z. Nivolianitou et al. / Journal of Hazardous Materials A137 (2006) 1–7

underlying ones. Finally, Section 5 discusses the results of the 3. It has the purposes of the rapid dissemination of the infor-
authors working experience with MARS. mation supplied by member states pursuant to Article 15(1)
of the SEVESO II Directive among all competent authori-
2. MARS structure ties with an analysis of the causes of major accidents and the
lessons learned from them.
The way of reporting accidents in MARS comprises two 4. It supplies information to competent authorities on the occur-
forms: the ‘short report’ which is intended for use in the imme- rence, prevention and mitigation of major accidents [3].
diate notification of an accident, and the ‘full report’, which is
prepared when the accident has been fully investigated, and its Thus, apart from information distribution, the main purpose
causes, evolution, and consequences are fully understood. In cer- of MARS is to analyse the data reported by the member states
tain cases, further information becomes available as, for example to the Commission with aiming at generation of lessons learned
in the course of judicial proceedings. There is also provision for from accidents. This is being done within the overall objectives
the ‘full report’ information to be further modified [1]. of the SEVESO Directive:
The ‘short report’ gives essential information concerning the
accident, in a free-text format, under the following headings:
1. prevention of major accidents involving dangerous sub-
stances;
• accident type
2. limitation of their consequences on man and the environment,
• substances directly involved
with a view to ensuring high levels of protection throughout
• immediate sources of accident
the community in a consistent and effective manner.
• immediate causes
• immediate effects
• emergency measures taken Lessons learned from major industrial accidents should help
• immediate lessons learnt to identify significant areas of concern. This could also help into
setting of priorities for further improvements and into undertak-
The ‘full report’ is much more analytic, and involves more ing more research or/and regulatory intervention for industry,
work in its preparation. It still contains free-text fields to describe where necessary. Those lessons could also be used as input to
facts connected with an accident, but it also contains the defini- qualitative and quantitative risk analyses either as relevant top
tion of descriptive codes, for the accident itself and for associated events, or initiating events, or even as accident sequence scenar-
information, so as to enable the MARS database to be searched ios.
under almost 200 different headings (data variables). Some of MARS currently includes 498 accident events (status
them are listed below: 6/2003), each consisting of 200 data variables, 30 of which are
free-text fields. In other words, 100,000 individual categorical
• type of accident (selection lists, click boxes) and numerical data values define
• industry where accident occurred together with 15,000 free-text fields the total amount of infor-
• activity being carried out mation included in the database events.
• components directly involved
• causative factors (immediate and underlying) 3. Data analysis
• ecological systems affected
• emergency measures taken 3.1. Type of industry

MARS does not constitute the only database in European For the above-mentioned 498 accidents and according to the
level. Other databases concerning industrial accidents in fixed data provided both from the short and the full reports, statisti-
installations do exist, such as national databases in some Euro- cal analysis can be made of the fields mentioned in paragraph
pean countries: ZEMA in Germany; ARIA in France; AEA 2. This has been extensively done by Kirchsteiger, Kawka and
technology MHIDAS database in UK; TNO FACTS database in Kirchsteiger, and Drogaris [4–7].
The Netherlands; the World Offshore Accident Data (WOAD), The present analysis, however, is concentrated on the acci-
and the Loss Prevention bulletin of UK IChemE. Additionally, dents that happened exclusively in petrochemical installations
to that, there also exist also many non-public sources for data and refineries which sum up to 85 (status 6/2003) in the MARS
collection from industries and insurance companies. database. Petrochemical installations are characterized by very
Concerning MARS, it is very important to mention the fol- high levels of risk because of the nature of processed flammable
lowing: substances and of the gravity of consequences, in case of a
major accident in these establishments. Many specific accident
1. MARS is the mandatory major industrial accidents database types are closely related to the petrochemical installations (e.g.
and reporting scheme within the European Union. BLEVEs, fireballs, UVCEs) whose consequences can affect
2. It includes major industrial accidents but also ‘unusual’ major many people inside and outside these establishments, but also
technological events with serious social impact (e.g. damage the surrounding environment as mentioned by Papazoglou et al.
in an amusement park). [8,9].
 Z. Nivolianitou et al. / Journal of Hazardous Materials A137 (2006) 1–7 3

Fig. 1. Percentage of major accidents notified in MARS per type of industry for the period 1985–2002.

As shown in Fig. 1, accidents in petrochemical installations mit classification of the accident in a certain type of industrial
represent the 17% of the total number in industrial accidents, activity.
which is the second more important sector in industrial installa- Regarding the analysis of major accidents in the petrochem-
tions behind general chemicals with 32%. Pharmaceutical and ical field, the following observations have been made in the
pesticides follow with 9% along with wholesale and retail stor- various classification criteria.
age with 10% (including LPG bottling and bulk distribution and
tank storage farms). The category other (12%) includes smaller 3.2. Year of occurrence
and less dangerous industrial activities such as: ceramics, tim-
ber and furniture, textiles and clothing, electronics and electrical The distribution of accidents in the petrochemical sector per
engineering, amusement and transportation centers. The small- year in the period of 1985–2002 is shown in Fig. 2.
est percentage belongs to the paper manufacture industry, the The overall slight increase in the beginning of the 1990s in
waste treatment disposal and the power supply and distribution numbers of accidents reported to MARS per year does not mean
industry (2% in all cases). This percentage does not mean that necessarily increase in actual accident occurrences in the EU
accidents do not happen in these industrial installations. Acci- member states but could also be explained as a consequence of
dents do happen, but they do not fall within the category of an increased acceptance of the MARS which led to reporting of
major accidents as this is given by the EC [7]. In 1% of the “any interesting” accidental events by the competent authorities,
events, the information given was not sufficient so as to per- as commented by Kirchsteiger [4].

Fig. 2. Number of major accidents in the petrochemical industry per year of occurrence.
4 Z. Nivolianitou et al. / Journal of Hazardous Materials A137 (2006) 1–7

Table 1 In 62 cases, a release on the ground or in the air was notified

Number of cases of substances involved in major accidents in the petrochemical and only in 6 cases release had taken place in the water. Fire was
industry
ignited in 57 cases and explosion phenomena (either as fireball
Substance involved Number of accidents or VCE confined and unconfined) took place in 37 cases. As
Hydrogen 13 it is evident, in major accidents, these types of accidents may
Crude oil 5 co-exist and overlap.
Heavy H/C 12
Gas oil 21
Natural gas 13 3.6. Operating status
LPG 21
Ethylene 12 In most of the cases (70%), the plant was under normal oper-
Sulfuric products 7
ation when a major accident took place. Phases more critical to
HF 2
the life of plant are shutdown, restart and start up. From the point
of view of major accidents, these phases represent the 3%, 1%
and 6%, respectively. Maintenance and testing have a percentage
3.3. Substances involved of 13% and 4% in major accident occurrence each.
Table 1 shows the substances that were more involved in
accidents of the petrochemical field directly or indirectly. In 3.7. Immediate causes
most cases (21 major accidents), gasoline and its products were
involved, while LPG contributes with 21 accidents as well. The graph in Fig. 4 presents the immediate causes of acci-
Natural gas (LNG, methane, ethane) was present in 13 cases. dents notified to Major Hazards Bureau and shows that 40%
Heavy hydrocarbons and crude oil were the hazardous substance of the major accidents notified have causes either exclusively
involved in 13 and 5 cases, respectively. Hydrogen and ethylene (19%) or partially (21%) attributed to human factor. Equipment
were involved in 13 and 12 accidents each. Sulfuric products failure was the 44% of the causes of major accidents in the
(SO2 and H2 S) were involved in seven major accidents, human petrochemical sector in the period from 1985 to 2001; natural
factors (HF) in two major accidents and mineral oils in one phenomena like floods or thunderstorms and environment con-
major accident. It is obvious that in some cases more than one ditions like low temperature or humidity were the 7% of causes
substance were involved in the accident. either directly (3%) or in combination with equipment failure
(4%). “Other” notifies rare and random events. In 9% of the
cases, immediate causes have not yet been defined.
3.4. Substances characteristics

Substances are categorized in MARS by means of their haz- 3.8. Consequences

ardous characteristics (toxic, flammable and explosive) and their
relative percentage in the major accidents bulk. As expected In Fig. 5, the immediate consequences of the accidents noti-
from the type of industry, we are focusing on and the type of fied in the petrochemical sector are presented. As for the accident
accidents that usually take place in this industry; most of the type categories, there is significant overlap among the con-
substances are flammable (48%), explosive (1%) or both (35%). sequence categories. It is evident that in most accidents with
Purely toxic was the 12% of the substances involved in major fatalities and/or injuries, there is also material loss involved,
accidents, while the ecotoxic and the corrosive substances rep- while in cases with bigger number of fatalities, a significant
resent together the 4% of the total percentage of substances community disruption is also likely.
involved.

3.5. Type of accident

Fig. 3 shows the number of accidents notified per type of

accident.

Fig. 4. Immediate causes of accidents in the petrochemical industry for the

Fig. 3. Type of major accidents in the petrochemical industry. period 1985–2002.
 Z. Nivolianitou et al. / Journal of Hazardous Materials A137 (2006) 1–7 5

Table 3
Post accident measures taken for major accidents in the petrochemical industry
Post accident measures Number of accidents

Design change 28
Procedures 16
Inspection 15
Safety audit 2
Maintenance 5
Fig. 5. Serious consequences of major accidents in the petrochemical industry. Safety manager 6

A main observation in the present analysis is that 20% of the

accidents (19 accidents), led to human fatalities while in 50% development of new procedures (22%) were decided. As lack
of the accidents (40 accidents), human injuries were registered. of inspection and/or adequate maintenance is crucial, inspection
The hazard potential of this specific industrial domain along (21%) and maintenance (8%) were defined as post accident mea-
with the propagation of the consequences, in case of an accident, sures. The use of a safety manager (8%) and the organization
beyond the industrial establishment barriers may explain these of a safety audit (3%) were also notified to the Major Hazards
elevated numbers. Unfortunately, the exact number of injured Bureau.
person inside and outside of the establishment is not known.
Furthermore in 10% of the accidents (eight accidents), envi- 4. Refinement of immediate causes in the full reports
ronmental consequences were registered. Again, no information
is given on the magnitude or the severity of the consequences in After examination of the full reports, it has been realized
the ecosystem. that many causes initially classified as equipment failure in the
short reports description, were actually due to organizational
3.9. Emergency measures taken factors. On the other hand, in many cases, equipment failure
due to corrosion was caused by faulty maintenance, which is a
Table 2 presents the emergency measures taken during the pure organizational matter.
evolution of the accident in the petrochemical sector. In almost Fig. 6 depicts the specific causes attributed to human/
all cases (89%), the internal fire brigade forces of the establish- organizational factors in the accidents of the petrochemical sec-
ment had to interfere in order to suppress the accident. This was tor. Human error is mainly caused either by design deficiency
done in combination with external forces (42%), with evacuation or by lack of procedures (22% and 17%, respectively). Lack of
of nearby communities (10%) and with sheltering of the popu- procedures does not only mean that there are no procedures for
lation (6%) or decontamination of the surroundings (6%). In 3% the specific task. Indeed, procedures might exist but they are
of the cases, only decontamination of ground and waters took inadequately or incomprehensively written. Other causes that
place, and in 4% of the cases, there was no need for intervention. are incorporated in the human and organizational factor notion
are: lack of training or inadequate training (11%), management-
3.10. Post accident measures organization (9%), inappropriate maintenance (4%) and inspec-
tion (3%) and operator error (11%).
The measures taken after the conclusion of the accident are The specific causes for equipment that contributed to acci-
either measures of prevention (52%), or measures of mitigation dents in the petrochemical sector are presented in Fig. 7. In most
(3%), or both (18%). There are still cases that the after accident of the cases, malfunction or corrosion was the specific causes
measures are not yet defined. for equipment failure (28% and 18%, respectively). Other causes
In Table 3, the most common post accident measures taken related to equipment failure are: unexpected or runaway reac-
are presented. In most cases, changes in the design (39%) or tion (11%), control and monitoring failure (5%), electrostatic
accumulation (4%) and blockage (7%).
Table 2 Another case of latent organizational failure is the inadequate
Emergency measures taken for major accidents in the petrochemical industry material; cause identified in short reports as equipment failure,
Emergency measures taken Percentage of accidents which, however, is further categorized as a design deficiency
(22%) in full reports.
On site 25
On site + external 42 In 25% of the petrochemical major accidents (22 out of 85
On site + external + sheltering 3 in total), the initial cause in the short report was mentioned
On site + external + decontamination 6 as equipment failure but from the details provided in the full
On site + external + sheltering + evacuation 3 reports, the real cause turned out to be organizational factor,
On site + external + evacuation 10
with most frequently the design deficiency and the lack of train-
External + evacuation 4
Decontamination 3 ing and procedures. The reader must keep in mind that not every
Nothing 4 petrochemical major accident has a full report and that even in
Total 100
full reports there are cases where the accident causes are not yet
identified.
6 Z. Nivolianitou et al. / Journal of Hazardous Materials A137 (2006) 1–7

Fig. 6. Specific causes concerning human and organizational factors for major accidents in the petrochemical industry.

Fig. 7. Specific causes for equipment failure major accidents in the petrochemical industry.

Following this analysis, it becomes obvious that the overall show that similar cases had already been registered in the past.
percentage of human related causes has increased dramatically As mentioned by Drogaris [7], this gives clear indication that,
after the submission of the full reports, as organizational and although the knowledge needed to prevent major accidents and
management causes are incorporated into the human factor cat- minimize their consequences is often available, there is:
egory causes supporting the finding of Cacciabue [10] that 80%
of the accidents are due to human factor. (a) lack of proper safety culture to enable effective use of this
This observation is also supported by the fact that human knowledge; and
factors are nowadays recognised as the most important issue in (b) lack of a structured communication system to diffuse this
modern technology, with respect to, many related facets, such knowledge.
as usability, efficiency, effectiveness and especially safety [11].
Tackling HF and human–machine interaction (HMI) demands Thus, the overall objective of the major accident reporting
consideration for technical systems and working environments, system to collect and diffuse knowledge on industrial accidents
together with the social and organizational context in which comes to fulfill the above gaps. With the use of the reporting
actual interactions take place. These factors must be integrated system, competent authorities are able to exchange information
and considered in order to develop sustainable and effective and knowledge on the prevention, monitoring and post accident
design of systems and interfaces, effective training of opera- mitigation of severe incidents. The open form of the short reports
tors and managers, and sound and accurate safety assessments to the public gives the possibility to spread out this knowledge.
and accident investigations. The statistical analysis of the major accidents for the period
1985–2002 shows that the petrochemical industry is one of the
5. Conclusions most affected industrial fields with a high percentage (17%) of
major accidents. However, the rate of major accidents in the
The conclusions from several accident investigations and petrochemical industry is constant in the last decade, with a
from studies on safety related issues in the petrochemical sector slight decrease in 2000–2001 period (two accidents). The sub-
 Z. Nivolianitou et al. / Journal of Hazardous Materials A137 (2006) 1–7 7

stances most frequently present in major accidents were gasoline The research and the analysis of the major accidents are of
(and its products) and LPG and as expected from the type great importance for prevention of similar accidents and miti-
of industry under focus in most of the cases substances were gation of the outcome. However, it is of crucial importance to
flammable and explosive. A fire that took place under normal expand this analysis to the near misses as those can reveal very
operating status of the plant is the most frequent major acci- important information about the status of the establishment and
dent notified to the MARS, which unfortunately is related to its potential to give rise to severe accidents. MARS has fore-
50% of the cases with human injuries. The immediate causes of seen the notification and reporting of near misses too. It is up to
such major accidents are attributed to either the human factor (in the competent authorities to motivate industries to circulate this
40% of the cases) or to equipment (in 44% of the cases). Natural information too.
phenomena and random events are also registered as immediate
causes of major accidents. Acknowledgement
In major accidents, the interference of internal forces is abso-
lutely necessary in almost all of the cases while the interference The financial support of the EU Commission through project
of external forces too was necessary in one third of the cases, “PRISM” GTC1-2000-28030 to this research is kindly acknowl-
sometimes in combination with other measures to protect the edged.
nearby communities such as sheltering, or evacuation. Decon-
tamination of the surrounding environment resulted as necessary References
in only eight cases in total (as the only measure taken and in
combination with other emergency measures). The post acci- [1] MARS website:. http://mahbsrv.jrc.it/Activities-WhatIsMars.html.
dent measures that took place after the termination of the major [2] European Council, Council Directive 82/501/EEC on the major accident
hazards of certain industrial activities (“SEVESO I”), Official Journal
accidents were measures to prevent the accident of happening of the European Communities, Luxembourg, 1982.
again in most of the cases, or measures to mitigate its conse- [3] European Council, Council Directive 96/82/EC on the major accident
quences. hazards of certain industrial activities (“SEVESO II”), Official Journal
In the last decades, it has been cleared that human actions of the European Communities, Luxembourg, 1997.
constitute a major source of vulnerability to the integrity of [4] C. Kirchsteiger, Trends in accidents, disasters and risk sources in Europe,
J. Loss Prev. Process Ind. 12 (1999) 7–17.
interactive systems, complex as well as simple ones. “Human [5] C. Kirchsteiger, Status and functioning of the European Commission’s
errors” are inappropriate, incorrect or erroneous human actions major accident reporting system, J. Hazard. Mater. 65 (1999) 211–
and thus are causes of great concern as mentioned by Hollnagel 231.
[12]. However, the deeper analysis of major industrial accidents [6] N. Kawka, C. Kirchsteiger, Technical note on the contribution of
in diverse sectors revealed that the events leading to an accidental sociotechnical factors to accidents notified to MARS, J. Loss Prev. Pro-
cess Ind. 12 (1999) 53–57.
outcome had their origins in the organization and management [7] G. Drogaris, MARS lessons learned from accidents notified, CDCIR,
of the system. For this reason, the focus of attention is moving JRC, 1991.
away from the technical and human failures to the management [8] I.A. Papazoglou, Z. Nivolianitou, O. Aneziris, M. Christou, G. Bonanos,
activities [13]. Risk-informed selection of a highway trajectory in the neighborhood of
The analysis of the specific causes of major accidents notified an oil-refinery, J. Hazard. Mater. 67 (1999) 111–144.
[9] I.A. Papazoglou, Z. Nivolianitou, O. Aneziris, M. Christou, Probabilistic
to the MARS database shows a latent contribution of man- safety analysis in chemical installations, J. Loss Prev. Process Ind. 5 (3)
agement and organizational failure to the causation of major (1992).
accidents. Even in cases where equipment failure was identified [10] P.C. Cacciabue, Human factors impact on risks analysis of complex
as the primary cause of the accidents, the more detailed study systems, J. Hazard. Mater. 71 (2000) 101–116.
of full reports revealed organizational deficiencies that can be [11] P.C. Cacciabue, Report on Research Activities, Human Factors Sector
European Commission, Joint Research Centre, Institute for the Protec-
categorized either as inadequacies (design or procedures) or as tion and Security of the Citizen, Technological and Economic Risk
insufficiencies (maintenance, inspection or training). The same Management, 2001.
is valid for cases where a human error was registered as imme- [12] E. Hollnagel, Cognitive Reliability Error Analysis Method, Elsevier Sci-
diate cause. ence Ltd., 1998.
The combined percentage for human related causes – human [13] R. Kennedy, B. Kirwan, Development of a hazard and operability-based
method for identifying safety management vulnerabilities in high risk
errors, organizational and management issues – increases dra- systems, Saf. Sci. 30 (1998) 249–274.
matically up to 64%.

View publication stats