Tall buildings began to dramatically change the skylines in major cities more than a century ago. Technological advances made it possible for people to effectively use spaces at heights far above grade level. Tall buildings provide challenges for the designers of fire protection systems not found in other buildings.

Like previous editions, the 2012 edition of NFPA 101, the Life Safety Code1 allows building designers to use performance based options in designing the egress system in the building. The performance criterion is given in Section 5.2.2. Based on this section of the code, the designer must consider the different fires that could occur in the building, how these fires will impact tenability, and how long the occupants will require to safely evacuate the building.

When using this approach, all of the assumptions and design methods must be included in the simulation of the evacuation. This means that the egress system designer must develop assumptions about how the population is expected to behave during the evacuation of a tall building. These assumptions then have to be applied to the calculation using data that is available.

What is not directly stated by NFPA 101 is that the egress system designer needs to understand the source of the data and how it applies to tall buildings. Some behaviors might be insignificant for someone going down a single flight of stairs, but become more significant as the travel distances become much larger.

One potential solution is to apply safety factors to the design. With only a limited understanding of the data, a large safety factor may be required so as to not subject the building occupants to undo risk.

This article will look at components of the evacuation time of occupants in tall buildings and the assumptions that are made by egress system designers. The focus will be on the movement to and within the stairs as well as the data used to develop an estimate of the descent rate. How the data was collected relative to its application for use in tall buildings will be analyzed. Finally, other egress options will be discussed.


The egress system designer needs to consider two sets of conditions in parallel. On one side, there is the fire growth and tenability in the different building areas. On the other side, there are the building occupants that need to get to a place of safety. For people remote from the fire, they need to receive some cue (e.g., smell smoke, see flames, or hear an emergency announcement) before they will start to evacuate. Occupants remote from the ignition location may require some time before they start to evacuate. In a tall building, direct observation of fire cues might not be possible for occupants located many floors away and/or on the opposite side of the building. In tall buildings, it is likely that many of the building occupants will not become aware of the need to evacuate until the fire alarm system activates.

The egress system designer could add the time for the alarm to activate to the time calculated for egress. In this case, the assumption is that all of the occupants hear the alarm and immediately start toward the exit. It is based on all people perceiving the alarm, paying attention to it, comprehending what it means, realizing that it applies to them, and then deciding to evacuate. Research has shown that many people do not recognize the temporal-three signal as applying to fires.2 Even when people in tall buildings do realize that there is an emergency, they have reported doing other tasks.3

When using the pre-evacuation times from tables, using the average value creates two potential limitations. First, the data has only been collected from a relatively small number of incidents. Training and other unknown variables could cause these times to be too short or too long. More data is needed to be able to fully understand what the most appropriate values are. Second, NFPA 101 requires that all occupants remote from ignition be protected from untenable conditions. If some vulnerable populations require more pre-evacuation time, using the average value will leave them at risk for not being able to evacuate before conditions become untenable.


For the evacuation of a tall building, stairs are intended to allow people to descend and leave the building. While there is some travel distance on the floor of origin, NFPA 101 limits that travel distance. After descending fewer than 11 floors, the building occupants have travelled further within the stair than before they reached it.4

The descent times in tall buildings can be substantial. While the stairs can usually be considered safe, a poor estimate of how people descend could lead to crowded conditions that prevent people from the floors from entering the stairs.

One equation that has been used for calculation of movement on stairs is found in the SFPE Handbook of Fire Protection Engineering.5 The Handbook does not provide any limitations on the applicability of the results. For densities greater than 0.54 persons/m2 and less than 3.8 persons/m2, the SFPE Handbook equation is:

S = k - akD (Equation 1)


S =Speed along the line of travel (m/s or ft/min)

D =Density (persons/m2 or persons/ft2)

k =constant for four different riser and tread combinations

a =empirical constant (0.266 when calculating m/s, 2.86 when calculating ft/min)

For densities less than 0.54 persons/m2, the people are able to travel at their free speed (the speed at 0.54 persons/m2). For densities greater than 3.8 persons/m2, the flow comes to a stop.

Based on this formula and no limitations, it is then possible to predict the evacuation time. In order to do so, there are several more assumptions that are made.


For travel down stairs, the Handbook equation is based primarily on the work of two researchers from the 1960s and 1970s. The equation comes mainly from the work of Pauls and Fruin.5

The work of Fruin6 primarily involved pedestrian planning for horizontal egress and ingress components. For level surfaces, he developed six "levels of service" (A to F ) to qualitatively explain the ability of people to choose their walking speed at different densities. He extended his observations by observing two different stairs. One of the stairs was indoors and the other was an outdoor stadium.

Again the "levels of service" ranged from Level A (below 0.54 persons/m2), where people are free to choose their own speed to Level F (above 2.70 persons/m2), where the descent is reduced to a shuffling pace. In neither case was it reported that the building occupants were in tall buildings.

In the 1960s and 1970s, Pauls7 observed evacuations of 58 tall buildings in Canada with a range of riser and tread dimensions. These buildings were up to 20 stories in height, but most were shorter. In his study, he looked at building averages and a limited number of spot measurements. From this data, he proposed that the descent speed could be calculated based on:

S =1.08- 0.29D (Equation 2)


S =Speed along the line of travel (m/s)

D =Density (persons/m2)

If, in Equation 1, the constants for the metric units and 17.8 cm riser height and 27.9 cm tread depth are used, the two equations are equivalent.

Using the same data, Pauls8 later reported that most of the stairs in his study had 17.8 cm riser heights and 27.9 cm tread depths. He theorized that people might descend stairs at different rates depending on the riser height and tread depth. With his theoretical equation, he calculated what the different speeds might be for four different combinations. He also explicitly stated that the values were not based on actual data and should not be used in practice.

Based on these three pieces of research, Nelson and MacClennan9 developed Equation 1. When the density was less than 0.54 persons/m2,they used the findings of Fruin6 to determine the free movement speed. The subsequent speed values for the 17.8 cm riser height and 27.9 cm tread depth case was based on the work of Pauls.7

The 3.8 persons/m2 end point was based on where the graph crossed the x-axis. It is at a much greater density than Fruin6 gave for level of service F and well beyond the maximum density observed by Pauls.7 For the other three k values, Nelson and MacClennan9 used the theoretical values that Pauls8 had said should not be used in practice. These other k values came from the assumptions made by Pauls and not from data that had actually been collected.

It should be noted that Pauls7 and Fruin6 did not measure density in the same manner. Pauls7 identified a boundary layer that people leave between themselves and walls. His density measurements are based on the effective width. The previous approach used the entire area. Thus, value from Fruin6 should have been adjusted to be comparable to the measurements of Pauls.7


There are seven issues that challenge the assumption that Equation 1 is valid for use in tall buildings:

  • The reliance on averages could lead to underestimating times for vulnerable populations.
  • The basis on density rather than human interactions might not match reality.
  • The untested k values might not be valid.
  • For buildings over 20 floors (and possibly less due to sample size issues), the buildings are taller than those used to collect the original data.
  • The population considered might not be representative of the earlier population.
  • The measurement methods used might not be consistent.
  • The equation can be applied to densities that were not observed.

Equation 1 is primarily a regression equation that was developed using averaged values. While this can give an approximation of the mean value, it does not give any indication of the scatter of the data. In order to develop an appropriate safety factor, the expected minimum movement speeds need to be known. This is especially true if those minimum values apply to a particular subpopulation. If that subpopulation will always move slower than average, it is not conservative to apply the average value to them.

With the intent to protect all occupants not intimate with ignition, relying on just average values could lead to vulnerable populations not having sufficient time to evacuate. For example, Boyce, Shields, and Silcock10 found that people with varying levels of physical impairments required greater time to descend stairs.

Another underlying assumption of Equation 1 is that people behave like a fluid. The flow rate out is a constant and the people do not interact in any way other than the density; no one person will slow down the other people around them. Pauls8specifically addressed this point by noting that people passed slower individuals to keep the ultimate flow in line with the expected results. However, Shields, et al.11 found that occupants were unwilling to pass a wheelchair user being assisted down the stairs (approximately 40 cm available to pass) and Proulx, et al.12 found that occupants using the handrail or with disabled occupants ahead of them did not pass slower moving occupants. Finally, Shields, et al.13 found that people moving behind a slower moving occupant chose not to pass. Even beyond the considerations of the vulnerable populations, people will interact as they descend. For example, Jones and Hewitt14 discussed groups forming during evacuations and how those people interacted both before and during their descent.

A better understanding of these interactions could result in an improved understanding on the amount of time that people will require to descend. However, assuming that the slower moving people will just be passed is not conservative.

Another potential limitation with Equation 1 is the k value that is used. While the work of Templer15 indicates that there could be differences in speed based on riser heights and tread depths, it is unknown if the k-values in Equation 1 are accurate. Applying the equation to any situation other than a 17.8 cm riser height and 27.9 cm tread depth is outside the scope of the data that was collected. How much of an error this will make in the final predicted value is unknown.

The scope of the data could also limit effects that would manifest themselves as people descended greater distances. The Joint Committee16 believed that fatigue would start to play a role when there were no merging flows, and Galea and Blake17 reported instances where fatigue was caused by footwear. Equation 1 does not have any difference in speed caused by fatigue. Based on the equation, a person descending from the top of a hundred story building would never slow down. If fatigue is an effect, then Equation 1 presents an optimistic estimation of speed on stairs in tall buildings.

Questions have also been raised about the applicability of data collected nearly half a century ago on the population of today. Pauls, Fruin, and Zupan18 were unsure about whether the changing demographics of the population would cause descent speeds to be slower. It is important to note that the researchers whose work enabled the creation of Equation 1 questioned whether it was still applicable or not.

Hoskins and Milke4 explain the different methods to measure occupant density that have been used by previous researchers and include a method for calculating landing distances not done for Equation 1. Also, related to the previous issue about the k values, Hoskins19 has proposed a method for equating densities on different tread dimensions, and when landings are included, to make equations applicable to more stair configurations. However, this method needs to be validated using more data.

The final potential problem that can arise when using equation 1 for tall buildings is to have theoretical conditions that do not match reality. The maximum density does not match the observations of Fruin6 or any observation made by Pauls7. Any calculations that involve the highest density conditions may not be accurate.

All seven of the limitations come back to one central point when considering people movement in tall buildings: Equation 1 could be accurate. How accurate is unknown and thus requires safety factors. After all, in smaller buildings, an estimated time that is off by a few seconds per floor results in errors that fall within the level of the noise of the data. As the buildings get taller, those seconds can become minutes if not tens of minutes. The errors can then rise above the level of the noise.


Many of the issues involving Equation 1 apply to the use of the computer models. When a model is used, the system designer needs to be aware of the limitations of the model, the basis of the calculations, and how the default settings alter the results. Simply using the default settings might not provide accurate results for evacuations from tall buildings for all of the reasons that applied to Equation 1.


The travel time down stairs required for vulnerable populations could be substantial, or they might not be able to descend the stairs at all. The 2012 edition of NFPA 1011 allows the use of elevators for occupant - controlled egress prior to phase 1 emergency recall. This should help to meet the goal of protecting all building occupants not intimate with ignition in tall buildings.

Bryan Hoskins is with Oklahoma State University.


  1. NFPA 101, Life Safety Code, National Fire Protection Association, Quincy, MA, 2012.
  2. Proulx, G., and LaRoche, C. "Recollection, Identification, and Perceived Urgency of the Temporal- Three Evacuation Signal," Journal of Fire Protection Engineering. Vol. 13, No. 1, pp. 67-82, 2003
  3. Kuligowski, E. and Hoskins, B. "Analysis of Occupant Behavior," Pedestrian and Evacuation Dynamics, 2010 Conference, Springer, New York, 2011. pp. 685-698, 2011.
  4. Hoskins, B. and Milke, J. "Differences in Measurement Methods for Travel Distance and Area for Estimates of Occupant Speed on Stairs," Fire Safety Journal, Vol. 48, pp. 49-57, 2012.
  5. Gwynne, S. and Rosenbaum, E. "Employing the Hydraulic Model in Assessing Emergency Movement," The SFPE Handbook of Fire Protection Engineering National Fire Protection Association, Quincy, MA, 2008.
  6. Fruin, J. Pedestrian Planning and Design, Metropolitan Association of Urban Designers and Environmental Planners, Inc., New York, 1971.
  7. Pauls, J. "Building Evacuation: Research Findings and Recommendations." Fires and Human Behaviour, John Wiley & Sons, New York, pp. 251-275, 1980.
  8. Pauls, J. "The Movement of People in Buildings and Design Solutions for Means of Egress." Fire Technology, Vol. 20, Issue 1, pp. 27-47, 1984.
  9. Nelson, H. and MacLennan, H. "Emergency Movement," SFPE Handbook of Fire Protection Engineering, National Fire Protection Association, Quincy, MA, pp. 286-295, 1995.
  10. Boyce, K., Shields, T., and Silcock, G. "Toward the Characterization of Building Occupancies for Fire Safety Engineering: Capabilities of Disabled People Moving Horizontally and on an Incline," Fire Technology, Vol. 35, No. I, pp. 51-67, 1999.
  11. Shields, T., Boyce, K., Silcock, G., and Dunne, B. "The Impact of a Wheelchair Bound Evacuee on the Speed and Flow of Evacuees in a Stairway during an Uncontrolled Unannounced Evacuation." Journal of Applied Fire Science, Vol. 7, No. 1, pp. 29-39, 1997.
  12. Proulx, G., Bénichou, N., Hum, J., and Restivo, K. "Evaluation of the Effectiveness of Different Photoluminescent Stairwell Installations for the Evacuation of Office Building Occupants," National Research Council of Canada, Research Report 232, 2007.
  13. Shields, T., Boyce, K., and McConnell, N. "The Behaviour and Evacuation Experiences of WTC 9/11 Evacuees with Self-Designated Mobility Impairments," Fire Safety Journal, Vol. 44, pp. 881-893, 2009.
  14. Jones, B. and Hewitt, J. "Leadership and Group Formation in High-Rise Building Evacuation," Fire Safety Science: Proceedings of the 1st International Symposium, International Association for Fire Safety Science, London, 1985.
  15. Templer, J. "Stair Shape and Human Movement," Ph.D. Dissertation, Columbia University, 1974.
  16. Joint Committee of the Building Research Board of the Department of Scientific and Industrial Research and the Fire Offices' Committee, "Fire Grading of Buildings – Part III – Precautions Relating to Personal Safety," Post-war Building Studies Number 29, pp. 22-95, Her Majesty's Stationary Office, London, 1952.
  17. Galea, E. and Blake, S. "Collection and Analysis of Human Behaviour Data Appearing in the Mass Media Relating to the Evacuation of the World Trade Centre Towers of 11 September 2001," Office of the Deputy Prime Minister, London, 2004.
  18. Pauls, J., Fruin, J. and Zupan, J. "Minimum Stair Width for Evacuation, Overtaking Movement and Counterflow – Technical Bases and Suggestions for the Past, Present, and Future," Pedestrian and Evacuation Dynamics, Springer- Verlag, Heidelberg, pp. 57-69, 2007.
  19. Hoskins, B. "Effective Density Measurement Methods on Stairs," Proceedings of the 5th International Symposium on Behaviour in Fire, Interscience Communications, London, pp. 182-193, 2012.