Subway Tunnel Cross-Passage Spacing: A Performance-Based Approach

author list icon

The path to a point of safety - the cross-passage - and required evacuation time for passengers downstream of a tunnel fire site are analyzed.

William D. Kennedy, an internationally recognized expert in tunnel ventilation, died in June 2012. During a 46- year career with WSP | Parsons Brinckerhoff, he was instrumental in the development of tunnel ventilation systems for road and rail tunnels worldwide. His reputation in tunnel ventilation was recognized in March 2012 by the International Symposium on Tunnel Safety and Security, which awarded him its 2012 Achievement Award, citing his “long and illustrious career in ventilation engineering of tunnels” and calling his lifetime body of work “a shining example of wedding practice and theory in the design of tunnels.” 

This abstract is condensed from a paper that was originally prepared for the 2006 APTA Rail Conference and has been updated to reflect the current 2014 version of NFPA 130.

The U.S. National Fire Protection Association's Standard 130, "Fixed Guideway Transit and Passenger Rail Systems," requires that tunnel-to-tunnel cross-passages shall be spaced a maximum of 800 feet (244 meters) apart. No guidance is provided on how the actual spacing should be determined. Intuition says that the spacing should vary with the length of the train, the number of passengers on board the train, the walkway width, the design fire scenario, etc. This paper presents a performance-based approach for calculating cross-passage spacing for downstream emergency evacuations from the fire site, and discusses NFPA 130 compliant methodologies for reducing the numbers of cross-passages required. The performance-based calculations include the use of computer software for analyzing and comparing exiting strategies. The simulations account for the geometry of a bored tunnel. 


Based on earlier emergency ventilation studies, it was concluded that the maximum cross-passage spacing should be such that those downstream of the fire could evacuate to a point of safety within the time that it takes for the floor of a train car to burn through (which leads to flashover of the entire train car). 

This leads to the conclusion that increasing the car-floor burn-through time would allow greater tunnel-to-tunnel cross-passage spacing and possibly reduce costs. This is suggested in NFPA 130 (Section Another possibility is wider walkways or cross-passage doors to speed passenger movement away from the fire site. 

It also leads to the inference that an interior or post-flashover fire should not be allowed to stop a train in a tunnel. Driver override should allow the movement of the train to the nearest station even if a passenger activates the emergency brake. The analysis for this paper assumes that this is the circumstance and that the only fire that will stop a train in a tunnel is a below-car fire that critically damages the propulsion system or derails the train. 

Physical Scenario for Computer Model 

Physical scenarios are simulated using computer modeling to predict the evacuation times for passengers downstream of the fire site to reach a point of safety. Seven cross-passage spacings, ten walkway widths, and one passenger load were analyzed. The computer model accounts for the unique geometry of a bored tunnel by considering shoulder space requirements. The simulation results provide sample engineering information to develop a sample of cost-effective alternatives without compromising safety. 

The physical scenario for modeling is selected to be typical of a heavy- or main-line rail passenger system. The results of this type of analyses are affected by many specific project factors. Therefore, the results provided in this paper MUST NOT be directly applied to any projects. See Figure 1 for data used. 

subway tunnels evacuation scenarios, rail, fire life safetyFigure 1 – Evacuation Scenarios

A number of assumptions were made in the model in order to be conservatively safe and simulate a reasonable worst case situation, such as: 

  • The location of the fire is in the middle of the train. 
  • The to-be evacuated train has a fire that is aligned with a cross-passage. 
  • A population of rail passengers consists of typical “commuters” with a range of demographics and walking speeds. (When the given walkway is wide enough, the model allows faster individuals to overtake slower walkers.) 
  • The fire scenario was assumed to be: 
    • Time 0 minutes, fire ignition; 
    • Time 5 minutes, fire reaches below-car fire heat release rate; 
    • Time 10 minutes, fire stops train; and 
    • Time 15 minutes, evacuation begins. 

Therefore, when calculating the minimum car-floor burn-through time required, 10 minutes (15-5) should be added to the evacuation time. This does not include any allowance for modeling accuracy. 

The emergency exiting analysis was done using the computer program SIMULEX1, which simulates the emergency exiting of people. The program algorithms for the movement of individuals are based on real-life data and predict realistic flow of people. It simulates the escape movement of each person instead of using a mathematical formula for uniform flow rates and average speeds of groups of people. This program is well-validated and has been used to model rail system emergency evacuations for a number of years2

The evacuation method was assumed to be all doors open to the walkway with movement to the nearest cross-passage downstream or adjacent to a stopped car. The passengers were considered to reach a point of safety after reaching 10 feet (3048 mm) inside of the cross-passage. 

Bored Tunnel Geometry 

Cross-passage spacing is particularly important in bored tunnel construction where cross-passages have to be mined in poor soil. Costs to construct each cross-passage in this situation can be high. The SIMULEX model inputs are adjusted for a bored tunnel construction. This leads to the concept of “Constructed Width” vs. “Effective Width.” Constructed Width is the actual width of walkway on the ground. Effective Width refers to the width entered into the SIMULEX model to accurately simulate the evacuation, relating to factors such as walkway width at shoulder height and the natural inhibition of walking near the edge of an empty track. 

Figure 2 presents the results of the simulations for 250 people per car and seven cars being evacuated. 

subway tunnel Time vs. Width, rail, fire life safetyFigure 2 – Time vs. Width

Some observations 

These observations are based on the sample data and should not be directly applied to other projects. 

  • Clearly the spacing of cross-passages has a significant impact on evacuation times. For the assumed data, any evacuation times required to be lower than 30 minutes, with train capacities in this study range, and with reasonable walkway and cross-passage widths, require spacing of cross-passages significantly shorter than the 800 foot maximum in NFPA 130. Other variables such as walkway or cross-passage width would also have an impact. 
  • There are significant benefits of wider walkways and wider cross-passage doors at cross-passage intervals above 700 feet or so. This is because the wider walkway after the train allows faster passengers to overtake slower passengers. In general, wider walkway widths help evacuation scenarios when the spacing has cross-passage doors that are not adjacent to the train and are located far away from the end of the train. Under these circumstances, wider walkways can be considered as an alternative to shorter cross-passage spacing. 
  • In the scenario adopted for analysis, it is obvious that shorter cross-passage intervals (in the range of 200 to 500 feet) result in one to three cross-passages adjacent to the train immediately accessible as soon as the evacuees move onto the walkway. Because the train can discharge passengers at a greater rate than they can exit through cross-passages, the effect of wider walkways in these shorter intervals is minimal. 
  • While not immediately apparent from the data shown, the effect on evacuation times due to varying passenger travel speeds is significant, again, at the longer intervals; of interest where continuous movement is occurring as opposed to the accumulated congestion immediately next to the train that dominates the shorter spacing cases. Thus, if performing analysis around cross-passage spacings that are beyond the train, careful attention must be given to the model inputs for evacuation speeds. 
  • Finally, the model examines the paths of evacuation up to the point of safety - the cross-passage. A close examination of the dynamics of the evacuation paths suggests that a project-specific application might want to consider the entire evacuation path—to whatever ends: a rescue train, a station platform, the opposite bore trackway, etc. The effects of the complete path should be modeled to study if there is an adverse affect of the evacuation in the non-incident tunnel. At a minimum, such analysis could suggest appropriate instructional and training emphasis. 


A performance-based approach for estimating evacuation times downstream from a tunnel fire site and minimum car-floor burn-through times has been presented. It allows the trade-off among cross-passage spacing, car-floor burn-through time, and walkway and cross-passage door width. For existing systems with fixed cross-passage locations and widths, this approach could be used to select car-floor burn-through times when cars are retrofitted or new rolling stock is ordered. For future designs, this approach could be used to develop a cost analysis combining cross-passage spacing and widths, car-floor burn-through time, and walkway width; possibly increasing the cross-passage spacing beyond the NFPA 130 maximum of 800 feet (244 meters). 


After peer review this approach could be used to develop an enhancement to NFPA 130. This enhancement, in relating cross-passage to other project characteristics, could provide a more logical basis for cross-passage spacing that could be greater or lesser than the current 800-foot requirement (244 meters). 

1"SIMULEX Users Manual"; 1998, Integrated Environmental Solutions, Limited; 141 St. James Road, Glasgow G4 0LT, Scotland.

2William D. Kennedy, Norris A. Harvey, and Silas K. Li, “Simulation of Escape from Rail Tunnels Using SIMULEX,” American Public Transportation Association (APTA), Boston, Massachusetts, June 2001.

back to top