Questions are being raised as to how many people were in the nightclub at the time of the incident.  This will become an important issue as the inquiry into the disaster progresses as the level of club occupancy could be used as a factor in contributing to the severity of the disaster and hence in apportioning responsibility.  Another issue that has been raised is whether or not more exits would have made a material difference to the outcome. To examine these issues we can do some simple analysis based on a number of assumptions and what the media are currently reporting.

So what do we (think we) know about the Kiss fire incident according to accounts in the media:

1) The club has a floor area of 615 m².  However, it is not known how this breaks down to dance floor, bar, toilets, circulation space etc.

2) The maximum travel distance to the front door is reported as being 32m.

http://world.time.com/2013/01/29/focus-turns-to-brazilian-club-safety-after-fire/

3) The maximum legal occupancy of the club is reported as being 691 people.

4) The only available exit has a width of 3m.

http://g1.globo.com/rs/rio-grande-do-sul/tragedia-incendio-boate-santa-maria-entenda/platb

5) The club owner insists that there were only 600 to 700 people in the club at the time of the incident.  However, the band’s guitarist told media that there were between 1200 and 1300 people in the club at the time and the police have given the same estimate.  The owner suggests that the higher estimates are due to clubers cycling into and out of the club.

http://www.foxnews.com/world/2013/01/31/nightclub-fire-that-killed-235-prompts-inspections-closures-night-spots-across/

6) It has now been reported that the band did in fact use flares during their show.  The flares were outdoor flares which are cheaper ($1.25 a piece) then indoor flares ($35 each).

http://world.time.com/2013/01/29/focus-turns-to-brazilian-club-safety-after-fire/

7) According to what I understand to be the fire certificate, the club is credited with having 2 emergency exits — hence 3 exits in total, one main and 2 emergency.  However, this is not clear from the article which later refers to these ‘emergency exits’ simply as ‘exits’, in which case the club may have been certified as having 2 exits, 1 main and 1 emergency exit. So it appears from the media accounts that the club was certified as having 3 or 2 exits — depending on the interpretation of the article.

http://g1.globo.com/rs/rio-grande-do-sul/noticia/2013/01/boate-informou-em-laudo-ter-duas-saidas-diz-brigada-militar.html

QUESTIONS:

In discussion with a Brazilian fire safety consultant based inSao Paulo, Dr Rodrigo Machado Tavares, Dr Tavares suggests that the local fire codes would require the Kiss nightclub to have:

– three exits,

– emergency lighting,

– emergency exit signage,

– an alarm system.

This answers many of the questions posed in my last blog (Yet Another Nightclub Tragedy — Grief tainted with anger, 29/01/13) however, it poses many serious questions concerning the licensing, inspection and enforcement process in Brazil.

Furthermore, it appears that the expired license has been reported as suggesting that the club had three (or at least two) exits.  Unless the missing exit or exits have been bricked up since the license was issued, or that the exit(s) were there but were locked or blocked during the incident and so could not be used, how could a license be issued stating that the club had three (or two) exits?  One has to ask if inspections of the club were ever performed.  If they were performed, how could an inspector fail to notice that two (or one) exits were missing?

Given that the nightclub had only a third of the legally required number of exits, one has to wonder whether this would have made a material difference to the outcome.

Concerning the number of people in the club, this is important as it is the manager’s responsibility to control the number of people in the club.  The manager suggests that the there were 700 people in the club, as allowed by the questionable ‘license’.  If there were 700 people in the club, this would be of benefit to the manager as it suggests overcrowding — his responsibility — would not have been a major factor and refocuses attention on the questionable fire license.  If there were 1200 people in the club, this deflects some degree of blame from the band — for allegedly starting the fire using flares — and for the code enforcement agency for permitting the license in the first place.

Is it likely that there were 1200 people in the Kiss Nightclub at the time of the incident?

SIMPLIFIED ANALYSIS:

Let’s do a simplified analysis on numbers of exits and number of people.  Note that this analysis is crude and a more thorough analysis would be required to establish the relevance of these key parameters.  The simplified analysis is dependent on a number of simplifying assumptions detailed below.

1) Assuming the 3m available exit width and a population of 1200 and 700, approximately how long would it have taken for the population to get out?

Let’s simplify the calculation and assume the population reacts immediately (zero response time) and let’s also assume that all the people are queued up at the exit ready to go.  Assuming a unit flow of 1.33 p/m/s (UK ADB) then the exit could sustain a flow of 3.99 p/s and it would require:

– at least 300 s for 1200 people to exit and

– at least 175 s for 700 people to exit.

2) How much time would be available for people to get out?  This is extremely difficult to estimate without doing a detailed analysis.  But let’s use the Rhode Island Nightclub fire as a rough guide.  After about 100 s conditions in the Station Nightclub were non-survivable. The floor area of the Station Nightclub was about 412 m² and so the Kiss nightclub is some 50% larger.  Based on this let’s assume that people had 50% longer to get out, making it 150 s for non-survivable conditions to develop.  This is admittedly a very crude estimation and may over estimate the amount of time available.

In 150 s, using the single available exit we could expect about 600 people to get out:

– generating a death toll of 100 people assuming there were 700 people in the club, as claimed by the manager,

– generating a death toll of 600 people assuming the band/authorities are correct and there were 1200 in the club.

Given the current death toll of 235, it is possible that there were 835 people in the club.  If correct, this would suggest that while the club was overcrowded, it is unlikely that there were as many people as suggested by the band/authorities.

A note of caution, remember, the assumptions that have been made are rather crude. The population is unlikely to have had zero response times, but evidence from the Rhode Island fire would suggest that in these types of situations, occupant response times are quite small of the order of a few seconds.  We can slightly improve these approximations by relaxing the zero response time assumption and assuming that the first people start to move in 10 s and that the closest people to the exit are located a third of the maximum distance (11m) and that the people located here move at 1.1 m/s.  This would add about 20 s to the estimated exit times OR reduce the available safe egress time by 20 s.  So let’s assume that we have 130 s to get out.

In 130 s, using the single available exit we could expect about 519 people to get out:

– generating a death toll of 181 people assuming there were 700 people in the club, as claimed by the manager,

– generating a death toll of 681 to be killed if the band/authorities are correct and there were 1200 in the club.

Given the current death toll of 235, it is possible that there were 754 people in the club.  This would shift the club occupancy even closer to that suggested by the nightclub owner.

We may also assume that the unit flow achieved by the exit was better than that prescribed in the UK regulations, which by their nature are conservative.  This would get more people out of the nightclub within the 130 s available time, increasing the size of the estimated population within the club.  However, in such situations, especially after the first few people have exited, the exit flow will become competitive rather than ordered, which will tend to reduce the exiting efficiency and hence the achieved unit flow rate.  In reality, the unit flow rate achieved by the exit will vary throughout the evacuation, being somewhat greater than that specified by the regulation at the start of the evacuation and being somewhat less towards the end of the evacuation.  Without undertaking a detailed analysis it is difficult to suggest a reasonable value for the unit flow rate, so for the purposes of this blog we will keep it as it is.

It is worth noting the following:

– if the average unit flow rate of the exit is greater than that used in these calculations, more people would have been able to exit in the available time and hence the estimate of the club occupancy would go up,

– if the average unit flow rate of the exit is less than that used in these calculations, fewer people would have been able to exit in the available time and hence the estimate of the club occupancy would go down.

Given the uncertainties I will stick to the initial estimates.

3) What would have happened if there were the legally required 3 exits in the club?  To answer this requires us to know the total available exit width.  Let’s assume that the club would have followed UK requirements.  An assembly space of more than 600 people would require 3 exits (as apparently does the local requirements).  Using BS 9999, the minimum total exit width for the 3 exits would be 4.3m.  Here we assume that all three exits have the same width.

If we make similar assumptions to that in (1) and in addition that all three exits were available and the population is equally divided between the three exits, then the combined flow would be 5.7 p/s and :

– 700 people would be able to exit in 123 s

– 835 people would be able to exit in 147 s

– 1200 people would be able to exit in 211 s

Assuming the same amount of time is available for safe egress i.e. 150 s, then:

– all 700 people would be able to safely evacuate, i.e. the legal occupancy would have been able to safely evacuate.

– all 835 people who are estimated to have been in the club would have managed to safely evacuate

– of the 1200 people, 855 would manage to get out, and the death toll would have been 345.

So these simplistic calculations suggest that if the club and the legally required minimum number of exits (3) with an exit width as specified by the BS 9999 and:

– the legal maximum population (700), then it is possible that everyone would have managed to evacuate before conditions in the club became non-survivable.

– if the population of the club was 835, then it is possible that everyone would have managed to evacuate before conditions in the club became non-survivable.  So even though the club was overcrowded on the night, it is possible that everyone would have managed to evacuate.

– if the population of the club was 1200, the population suggested by the band/authorities, then it is possible that 855 people would have managed to evacuate before conditions in the club became non-survivable and 345 people would have died.  So if the population was as high as 1200, even if the legally required exit capacity was provided it is unlikely that everyone would have managed to evacuate.

It has to be emphasised that these calculations are extremely crude, but they serve to demonstrate how important it is establish the correct number of people that were in the club at the time of the incident.  They also demonstrate that had the club had the required number of exits, it is possible that the death toll in this incident could have been significantly reduced.

To do a more thorough analysis on this incident would require fundamental data describing the club and the materials within the club, the number of people in the club and their distribution.  With this information it would be possible to undertake sophisticated fire and evacuation analysis of the type undertaken for the Rhode Island Nightclub fire.

Rhode Island Fire Simulation:

https://www.youtube.com/watch?v=o197yeup1BQ&list=PLC82636C7790DE890&index=1

Rhode Island Coupled Fire and Evacuation Simulation:

https://www.youtube.com/watch?v=gmPOIriMiyU&list=PL482FD999D5793B82&index=1

Paper describing simulations: http://fseg.gre.ac.uk/fire/pub.asp

Paper 223. “Coupled Fire/Evacuation Analysis of the Station Nightclub Fire”. Galea E.R., Wang, Z., Veeraswamy, A., Jia, F.,Lawrence, P., and Ewer, J. Proceedings of 9th IAFSS Symposium Karlsruhe,Germany, 2008, ISNN 1817-4299, pp 465-476. DOI:10.3801/IAFSS.FSS.9-465

Come and visit the EXODUS demo stand and check out our most recent research which will be presented through 7 papers:

“Fire and Evacuation Simulation of the Fatal 1985 Manchester Airport B737 Fire”, Wang, Z., Jia, F., and Galea, E.R.

ABSTRACT:

In this paper, fire and evacuation computer simulations are conducted to determine the impact of exit opening times on the evacuation and survivability during the Manchester Airport B737 fire of 1985. The fire and evacuation simulation tools, SMARTFIRE and airEXODUS are used in the analysis. The work is in two parts, the first part attempts to reconstruct the actual fire incident and ensuing evacuation using the known facts derived from the official investigation report.  The second part investigates the impact of exit opening times on the aircraft fire development and subsequent evacuation. The results suggest that the number of fatalities could have been reduced by 92% had the delays in opening two of the three exits been avoided.  Furthermore, it is suggested that opening of the unused aft right exit during the accident did not contribute to the high loss of life in this accident.  Indeed, it is suggested that the opening of this exit improved survivability within the cabin and reduced the death toll by some 17%.

“Investigating the impact of culture on evacuation behaviour — A Polish Data-Set”, Galea, E.R., Sharp, G., Sauter, M., Deere, S.J., Filippidis, L.
ABSTRACT:

In this paper results from an unannounced evacuation trial conducted within a library inWarsawPolandare presented and discussed. This experimental evacuation is part of a large international study investigating the impact of culture on evacuation behaviour. In addition, a framework to enable the systematic analysis of Response Phase behaviours is presented and applied to the trial data. The framework not only provides a consistent method for describing Response Phase behaviour, but also provides a framework for classifying and quantifying the Response Phase other than simply using the overall response time. An empirical response time model, based on data generated using the framework is also presented and applied to the evacuation trial data. The empirical response time model produces a prediction for the average response time for the trial population which is within 3% of the measured value. In addition to presenting Response Phase data, a data-set suitable for the validation of evacuation models is also presented. This consists of both egress times and time dependent density measurements.  buildingEXODUS predictions of the evacuation are compared with the validation data and shown to be in good agreement with the measured data.

“Modelling Human Factors and Evacuation Lift Dispatch Strategies”, Kinsey, M.J., Galea, E.R., and Lawrence, P.J.

ABSTRACT:

This paper presents an overview of a series of evacuation simulations utilising different lift dispatch strategies using an empirical based enhanced agent-lift model developed within the buildingEXODUS software. A brief description of the enhanced agent-lift model is presented. The evacuation scenarios investigated are based on a hypothetical 50 floor building with four staircases and a population of 7,840 agents. While past studies have measured the influence of such evacuation lift dispatch strategies assuming compliant/homogenous agent behaviour, this study extends that work by highlighting the potential influence of human factors upon such evacuation lift dispatch strategies. The study suggests that evacuation lift human factors can considerably decrease evacuation performance and highlights the need for consideration within evacuation strategies based on lifts.  

“An Analysis of the Performance of Trained Staff Using Movement Assist Devices To Evacuate the Non-Ambulant”, Hunt, A., Galea E.R., and Lawrence, P.J.
ABSTRACT:

This paper describes a series of trials undertaken to quantify the performance of trained hospital staff in evacuating a test subject through 11 floors of Ghent University Hospital using four commonly used movement assistance devices: stretcher, carry-chair, evacuation chair and rescue sheet.  In total 32 trials were conducted, using both male and female assist teams.  Presented in the paper are performance results, including: device preparation time, horizontal speeds, vertical speeds, and overtaking potential in stairwells. These data, alongside those established in questionnaire data from the experiment participants, form the basis of the device performance evaluation presented in this paper. A comparative methodology is derived to assess the efficiency of the devices. This methodology enabled performance differences to be established, according to the devices employed and the staff involved.
 

“The UK BeSeCu Fire Fighter Study: A Study of UK Fire Fighters’ Emotional, Cognitive and Behavioural Reactions to Emergencies”, Hulse, L.M. and Galea, E.R.
ABSTRACT:

A survey of UK firefighters revealed them to be seemingly psychologically prepared for what their job would expose them to but not immune to experiencing emotional arousal or perceived risk during emergency events. A number of aspects, such as the event posing serious consequences to their lives/well-being, were singled out as particularly distressing and linked with greater emotional arousal, while other aspects, ones focused on other people/circumstances, reduced perceived risk. Traffic accidents appeared to be a special case, inducing lower arousal and risk than another commonly attended emergency, domestic fires. More years of service had a positive effect on the risk perceived during a stressful event but heightened the emotional arousal in that moment. Received support was one of the most significant predictors of posttraumatic stress and growth, as well as being significantly linked to peri-event thoughts/feelings, although other variables not tested here, e.g. individual differences, might be better at explaining posttraumatic states than event-related variables. The sample reported that safety work with a risk group, migrants, was underway and appeared to be beneficial in reducing instances of “inappropriate behaviour” during emergencies but communication difficulties were an issue and training on this matter would appear desirable. Significant differences in responses across the UK were detected and consequences for international comparisons are discussed.
 

“The Collection and Analysis of Data from a Fatal Large-Scale Crowd Incident”, Pretorius, M., Gwynne, S., and Galea, E.R.
ABSTRACT:

This paper discusses the analysis of data-sets from observations made at the Duisburg Love Parade in 2010 and the large-scale crowd situation that ended in fatalities due to the development of crush conditions. This event is a useful case study of large crowd circulation based on the materialthat was made publically available by the organisers and attendees. The resultant data-set has been used to configure the buildingEXODUS modelto approximate the original incident in order to verify both the model’s performance and the underlying scenario assumptions; i.e. whether buildingEXODUS can reliably represent agent actions, the conditions that develop and the impact of these developments.
 

“Response Time Data for Large Passenger Ferries and Cruise Ships”, Brown, R., Galea, E.R., Deere, S., and Filippidis, L.

ABSTRACT:

This paper outlines research that was carried-out under the EU FP7 7 project SAFEGUARD and presents three sets of passenger response time data generated from full-scale semi-unannounced assembly trials at sea.  The data sets were generated from two different types of passenger ships, a RO-PAX ferry, SuperSpeed 1 (SS1) and a cruise ship, Jewel of the Seas (JoS).  In total response times from over 2200 people were collected making it the largest response time data set ever collected — on land or sea.  The paper presents the analysis methodology used to extract the response time data and the resultant response time distributions (RTD).  A number of key findings from the data analysis will also be presented, which includes: (a) all generated RTDs are log-normal, (b) RTDs from the two SS1 trials using two different populations are very similar, (c) The combined RTD for the SS1 is almost identical to the RTD generated from the earlier published data for the same type of vessel, (d) The RTD derived for the public spaces on the JoS is significantly different to that of the SS1, (e) The RTD for public spaces and cabins are significantly different.  These findings are discussed in this paper and form the basis of a recommendation to be submitted to the International Maritime Organisation to be used to frame the next iteration of the international guidelines for ship evacuation analysis.

At the time of writing, information concerning the nature of this incident is far too sketchy to draw any firm conclusions.  Hopefully the enquiry will reveal what actually happened and, more importantly, why it happened, so that those charged with the safety of crowds around the world can learn from this tragic incident.

At this point I have two observations I would like to share with you.

(1) What appears to be the situation a few hours after the incident?

Here’s what I know at the moment, based on as yet uncorroborated media reports:

• The Duisburg Love Parade reportedly attracted a crowd of 1.4 million people — this number has not confirmed by the authorities.
• The event began at 14:00 and took place in a former railway freight yard.
• The event has a single entry point, a tunnel located on Karl-Lehr Street. The tunnel is actually a highway underpass. Dimensions of the tunnel are not known, but photographs suggest that it is about 15m wide at the entrance.  Its depth is also unknown; one media report suggested that it was 600m long (Voice of America Online) and another suggested it was 100m long (Guardian).  Photographs I have seen suggest that the tunnel was very long and appears to be made up of several separate underpasses.
• Photographs appearing in the media also suggest that the crowd outside the mouth of the tunnel and inside the tunnel was enormous.  While it is difficult to get an accurate measurement from the available photographs, they suggest to me that the crowd density was well in excess of 4 people/m² and may have been in excess of 6 people/m².  From the photographs there also appears to be no crowd management officials in sight and the random nature of the crowd further suggests that there was little organisation or control in the dispersal of the crowd.  It is not clear when the photographs I have seen were taken, but I assume that they were just before the event or during the event.
• 19 people were killed in the incident and 340 people were injured.  It is not clear where the fatalities occurred.

I would emphasise again that the information at this stage is sketchy and there are many important details missing.  However, if the little information that is available proves to be correct, I would offer the following observations:

• A single entry point for an event of this magnitude is not a good idea.  Given the size of the crowds expected, a single entry point was bound to make crowd management more complex and difficult with implications for crowd safety.    
• I currently have to assume that there was also only a single way out of the event — the entry point.  If this is the case, again it is poor planning for the same reason that a single entry is not a good idea.  I would hope that other emergency exit points were available which the crowd could use in the event of an emergency, but this is currently unknown to me.
• Another issue with having a single entry/exit point is that it may have to support quite sizeable bi-directional flows as people arrive and leave the event.  If controlled bi-directional flow is planned, this further reduces the effective width of the opening, increasing potential crowd management issues associated with the single entry point.  If the bi-directional flow was not controlled, this could have led to chaotic situations within the tunnel, especially with large flows of people.  From observation of the tunnel layout, I can see no evidence that there was a controlled bi-directional flow environment e.g. barriers in the tunnel to segregate entering and exiting patrons.
• Having a tunnel, and such a long tunnel, as the only entry/exit point, when large crowds are expected is not a good idea for several reasons.  Perhaps the most significant is that it provides poor visual access to what is happening within the tunnel.  In certain circumstances this will potentially have a negative influence on the behaviour of the crowds as they cannot see what is happening further into the tunnel and so cannot perceive that a problem such as a blockage may have developed within or beyond the tunnel.  Indeed, if one end of the tunnel was closed by crowd management officials, as some reports suggests, it would have been extremely difficult for people within the tunnel, at the other end of the tunnel and approaching the entrance to the tunnel to know this and hence would persist in trying to enter the tunnel or continue moving along the tunnel.  Perhaps more importantly, if the crowd management officials cannot see what is happening within the tunnel, it is more difficult to detect a problem in its early stages, which makes diffusing a minor crowd event and preventing it from developing into a tragic event extremely difficult, if not impossible.  It even makes it difficult for crowd management officials to know with any degree of certainty the number of people within the tunnel.  This is potentially a recipe for disaster.
• From what I can see in the photographs of the crowds approaching the tunnel from the station end i.e. the entrance, there appears to be no attempt at managing this flow.  It appears that there was no attempt to meter or regulate the flow of people or to channel the flow of people into the entrance.  I assume that the same was happening at the other end of the tunnel i.e. the event end.  From the pictures I have seen, the approach to the tunnel appears to have been a ‘free for all’.  If a small number of people had been expected to utilise the exit e.g. if there had been a smaller crowd or more entrances/exits, it may not have been necessary to provide this type of crowd management at the entry to the tunnel.  However, for an event attracting an audience of 1.4 million with a single entrance/exit, I suggest such measures would have been essential. 

(2) Subtle blaming of the victims.

As soon as news of the incident in Duisburg began to hit the front pages of the world’s newspapers and websites, familiar terms such as “crowd panic” and “crowd stampede” began to be used to describe the tragic incident in Duisburg.   For example:

• “Stampede at German Love Parade festival kills 19”

BBC online news 25 July 2010
http://www.bbc.co.uk/news/world-europe-10751899

• Sky News Headline live to air 25 July 2010 11:00

“Police in Germany have begun an investigation into how a stampede at a music festival left 19 people dead and nearly 350 others injured.  Panic broke out at yesterdays Love Parade event at the city of Duisburg where crowds of party goers tried to force their way through a narrow tunnel and were crushed.”

• “Stampede at German Music Festival Kills 19”

Voice of America News, 25 July 2010
http://www1.voanews.com/english/news/europe/10-Killed-in-German-Techno-Parade-Panic-99171424.html

• “Officials said the deaths and injuries occurred when panic broke out among huge crowds in a roughly 600 meter long tunnel leading towards the day-long open air Love Parade festival in the German city of Duisburg.”
  Voice of America News, 25 July 2010 
• “A mass panic resulting in a stampede reportedly led to the deaths near a tunnel at the Love Parade electronic music festival in the western German city of Duisburg.”

The Sun 25 July 2010
http://www.thesun.co.uk/sol/homepage/news/3067817/19-die-at-German-street-festival.html

• Love Parade stampede in Germany kills at least 18

Los Angeles Times 25 July 2020
http://www.latimes.com/news/nationworld/world/la-fg-germany-love-parade-20100725,0,2079922.story

• “10 killed in mass panic at Germany’s Love Parade”

Business Week.
http://www.businessweek.com/ap/financialnews/D9H5HJ3G1.htm

• “At least 10 people were killed and another 15 injured when mass panic broke out Saturday in a tunnel at an annual celebration of techno music in western Germany.”

Business Week.
http://www.businessweek.com/ap/financialnews/D9H5HJ3G1.htm

This is the typical media response, and of more concern, the typical first response from officials to fatal crowd incidents all around the world.  I don’t intend to discuss here the lack of understanding that the media, the public and officialdom have of words such as ‘panic’ and ‘stampede’.  Rather, I want to discuss the implications that using such phrases have on the public’s understanding of the event. 

While not directly stated, the implication conveyed when these types of phrases are used is that the incident was the fault of the victims, that it was their ‘unreasonable’ behaviour that caused or substantially contributed to the incident and resulted in the tragic loss of life.  Using such phrases is unhelpful, as it immediately diverts attention from other factors that may have contributed to or indeed been the root cause of the tragic event.  There are many other plausible explanations that may have contributed to this and similar tragedies that must be thoroughly investigated rather than go for the easy option of blaming the behaviour of the crowds.  For example, event planners may have designed an environment (both the physical space and the crowd management systems) that was unsafe for the size of crowds that were expected; approval authorities, charged with ensuring event design and crowd management provision is fit for purpose, may have failed to identify potential problems with the proposed event design and crowd management provision; and finally on-ground crowd management officials, charged with the safety of the public may have been unable to safely manage the crowd or may have allowed dangerous conditions to develop. 

It is all too easy to use these phrases when describing tragic crowd incidents, and thereby pass the blame for the incident onto the victims. In my experience of these types of incidents, it is seldom the victims who are the root cause of the incident, but failures during planning, approval and management of the event.

Keep Safe

Prof Ed Galea, 25 July 2010, 14:00