Designing to Control External Vertical Fire Spread

By Greg Baker, Kevin Frank and Colleen Wade


An article entitled “A comparison of horizontal projections and spandrels as protection method against external fire spread” published in Q3 2016 Issue 4 of SFPE Europe provided a summary of research [1] [2] that investigated the use of horizontal projections as an alternative to vertical spandrels as a means of controlling external vertical fire spread. The paper noted that the only prescriptive option currently in the Swedish building regulations [3] is to use a 1.2 m high spandrel. The authors concluded by recommending that further research be done to investigate the possibility of adding a new prescriptive horizontal projection option to the Swedish building regulations. This would be in addition to the current vertical spandrel provisions, which mirror the regulations in New Zealand.

This article builds upon the previous paper and provides further details of the regulatory compliance options in New Zealand. It also gives information about ongoing research that aims to provide performance-based guidance on how to design for external vertical fire spread through unprotected openings. (Fire spread over combustible façades is a separate topic and not part of the research described here.) The research project is still in progress, so recommendations are not yet finalized, but this paper provides details of what will form the basis for pending recommendations.

Complying with the New Zealand Building Code

Three options available for designers demonstrate compliance with the requirements of the New Zealand Building Code (NZBC) and relate to external vertical fire spread. 

  1. Prescriptive Provisions

    The prescriptive requirements provide permitted combinations of vertical spandrels and horizontal projections (called “aprons” in New Zealand), as summarized in Table 1.

    Table 1. Spandrel height and apron projection combinations permitted in New Zealand. (Data reproduced from Table 5.4, p. 100.) [4]

Spandrel height (m)

Apron projection (m)









If a designer follows these prescriptive requirements, this part of the building design is deemed to automatically comply with the NZBC.

  1. Verification Method

    Verification Method C/VM2 [5] is suggested as an acceptable way of demonstrating compliance with the C Protection from Fire clauses of the NZBC [6], C/VM2 requires that 10 different design scenarios be considered (where relevant) in the fire safety design of a building. One of these is a design scenario involving external vertical fire spread (VS) (pp. 53–56 in ref. [5]). The VS design scenario gives the designer the option to “calculate the effect of the radiation from fire plumes projected from openings. Fire plume characteristics and geometry shall be derived from the design fires as described in Part 2 (of C/VM2) for the applicable geometry” (ref. [5], p. 54].

    Part 2 of C/VM2 in turn specifies that a design fire with a peak heat release rate (HRR) of 1.5 times the compartment ventilation limit should be used for these calculations, i.e., Q ̇peak = 1.5 × Q ̇VL. Instead of simply using the approximation QVL = 1.5 Ao √(Ho ), (where Ao is the area of the opening and Ho is the opening height), the designer can determine the maximum HRR that can be supported by the available air/oxygen supply. C/VM2, however, does not provide specific guidance on modeling the characteristics of the plume.
  1. Alternative Solution

    If the chosen compliance pathway is an Alternative Solution, the designer must provide evidence to the authority having jurisdiction that the design will comply with external vertical fire spread provisions in the NZBC. This would typically be done by presenting design calculations and/or modelling results. No guidance is provided in the compliance documents as to how to undertake Alternative Solution calculations.

BRANZ Research Project

A BRANZ fire research project on external vertical fire spread [7] is nearing completion. The objective of the project has been to develop guidance for fire safety practitioners on performance-based design calculations to control external vertical fire spread. As noted above, C/VM2 does not provide the specifics of calculation procedures, other than stipulating what HRR to use. For Alternative Solution designs, the designer is free to submit whatever calculations or modeling results they deem appropriate.

The BRANZ research project involved the following five elements:

  1. A literature review of projected flame height and heat flux correlations that built upon previous BRANZ research. [8]
  2. A series of reduced-scale compartment experiments.
  3. Validation of the suitability of Fire Dynamics Simulator (FDS) to model projected fire plumes.
  4. Benchmarking of BRANZ experimental results, existing experimental data and existing correlations against FDS modeling predictions.
  5. Development of design recommendations.

Based on the research, a series of recommendations is being developed for use by fire safety engineering practitioners.

Basis for Recommendations

The objective of these performance-based calculations is to determine the thermal exposure at upper levels of the exterior of a building where a fire plume projects from a lower opening. This information is used to determine where unrated construction and unprotected openings in the exterior envelope are permitted. Such calculations require an appropriate acceptance criterion.

As noted in C/VM2, above, the “effect of the radiation from fire plumes” is calculated. The acceptance criterion used in this regard is a radiant heat flux of q" = 16 kW/m. [2, 9]. Although a lower flux is more defensible from a technical viewpoint, at the time that the value was chosen, a cost/benefit approach indicated that the level of fire loss being experienced was tolerable, and hence the increased cost associated with more conservative requirements was not justified. In addition, the C/VM2 commentary document notes that since “this value may not necessarily be small enough to prevent ignition or damage to all cladding materials, it is anticipated that the Fire Service will provide secondary means of preventing fire spread in these situations if necessary.” (ref. [9], p. 21)

Having defined the so-called “target,” the first stage of the calculations is to determine the height of the projected flames. As part of the research project outlined above, flame height measurements were extracted from video in a series of 14 reduced-scale compartment experiments. The resultant flame heights were then compared to six different flame height correlations in the literature. Three correlations showed the most promise: Quintiere and Cleary [10], Thomas and Law [11], and Law. [12]

 Figure 1 shows the comparison between this experimental data and the prediction of these three correlations. It should be noted that the Quintiere and Cleary correlation has been adjusted to allow for the type of fuel used.

Figure 1. Measured versus predicted maximum flame height – reduced-scale data.

For the data points shown in Figure 1, where a point falls above the equality line, it signifies that the prediction is greater than the measured value, i.e., is conservative, while for points below the line, the prediction is non-conservative. In two of 14 reduced-scale experiments, no external flaming occurred, but the correlations predicted projected flames (see data points on left-hand vertical axis in Figure 1).

Having compared the reduced-scale data and the correlation predictions, the next step was to compare the predictions of the correlations to actual full-scale data from Oleszkiewicz [13], as shown in Figure 2.

Figure 2. Measured versus predicted maximum flame height – full-scale data.

The data shown in Figure 2 have one correlation that is always conservative (Thomas and Law – blue circles), one correlation matching or always non-conservative (Quintiere and Cleary – gray circles), and one correlation that spans both (Law – yellow circles).

There is a reasonable basis for selecting any one of these three correlations for design purposes, but the Thomas and Law correlation would be the “safe” option for a designer and/or authority with jurisdiction.

The third comparison compared the Oleszkiewicz full-scale data to FDS predictions. The FDS input files for these experiments from the NIST FDS validation guide [14] were used with an additional device to measure the HRR as a function of height for the flame height estimate, with the original grid resolution and a fine resolution with cell dimensions halved. The flame heights in FDS were calculated as the height at which 99% of the energy had been released, time-averaged over the last half of the simulation, in accordance with the procedure used for other flame height applications in the FDS validation guide. The three correlation predictions from Figure 2 are included for comparison purposes.

Figure 3. Measured versus FDS predicted maximum flame height – full-scale data.

With regard to the two components of accuracy, precision and trueness [15], the fine grid resolution for the FDS modeling produces a more-precise result than the coarse mesh, which is to be expected. What is not expected, however, is that the coarse grid is producing a truer result than the fine mesh, with the latter producing non-conservative predictions.

The other inference to make from this comparison is that a simple correlation, such as the Law method (shown as the yellow circles in Figure 3), gives just as good a prediction as the more labor-intensive CFD modeling.

The second stage of the engineering calculations required for external vertical fire spread is to determine the heat flux that will be received at the face of the building above the opening from which the fire plume is projecting. The most promising correlation identified in this research is from Back, et al. [16] Over the first 40% of the flame height, the incident heat flux on the plume centerline is the maximum value. Over the range of 40–100% of the flame height, the centerline incident heat flux decreases linearly to 20 kW/m2 and then decreases further above the tip of the flame.

Figure 4 shows the application of the Back, et al., heat flux correlation as a solid black line for the reduced-scale experiments. The upper horizontal region of the correlation line is a constant peak heat flux of 81 kW/m2. In this region, the ratio of height z to flame height zfl is z/zfl < 0.4, corresponding to a peak HRR in the reduced-scale experiments of 190 kW.

Figure 4. Height above neutral plane to flame length versus heat flux – reduced-scale data.


Figure 5 shows a comparison of the full-scale data from Oleszkiewicz with the Back et al., correlation, with a peak heat flux of 172 kW/m2, which corresponds to a peak HRR of 10.3 MW.

Figure 5. Height above neutral plane to flame length versus heat flux – full-scale data.

The heat flux data shown in Figure 4 and Figure 5 are total heat flux measured by a gauge in experiments, while the acceptance criterion noted previously is radiant heat flux.

The reduced-scale experiments were transient, using heptane and wood crib fires, while the full-scale experiments used propane. The HRR data for the reduced-scale experiments were derived from oxygen consumption calorimetry, while the HRR data in the Oleszkiewicz experiments were based on mass flow rates and assumed complete combustion.

It should be noted that both Figure 4 and Figure 5 use the neutral plane, estimated as being at 0.4 × Ho from the sill of the opening, as the reference point for the comparison. While the full range of heat fluxes is depicted in both these figures, it is generally the right-hand region where the heat flux is less than 20 kW/m2 that is of interest for external vertical fire spread calculations.

Ongoing Research

The BRANZ research project described in this paper is still ongoing. Plans call for publishing a BRANZ study report in 2017 that provides a detailed scientific record of the project and its findings. The project has also investigated the impact of apron horizontal projections on plume characteristics, as well as the design fire guidance provided in C/VM2. While C/VM2 provides explicit guidance for the peak HRR, i.e.. Q ̇ = 1.5Q ̇VL, it was observed in the reduced-scale heptane pool fire experiments that this value could be easily exceeded. As noted by Drysdale [17], this will occur where there is a large excess fuel factor (e.g., if the fuel bed has a very large surface area or if the fuel is non-cellulosic with a low heat of vaporisation). While heptane pool fires may not be representative of fuel loads in typical occupancies, the designer should be mindful of such factors.

The authors all work as fire researchers for the private building research company BRANZ Ltd. in New Zealand. The authors would like to acknowledge the contribution of

Dr. Haejun Park to the experimental project. The research described here was funded by the Building Research Levy.

[1] Nilsson, M. (2016). The impact of horizontal projections on external fire spread: A numerical comparative study. Report 5510. Lund, Sweden: Division of Fire Safety Engineering, Lund University.

[2] Nilsson, M., Mossberg, A., Husted, B. & Anderson, J. (2016). Protection against external fire spread – Horizontal projections or spandrels? In 14th International Fire Science and Engineering Conference, 2, pp. 1163–1174.

 [3] Boverket. (2016). Boverket’s byggregler (Swedish building regulations) BBR 23. Karlskrona, Sweden: Swedish National Board of Housing, Building and Planning.

 [4] MBIE. (2016). C/AS2 Acceptable solution for buildings with sleeping (non institutional) (risk group SM) for New Zealand Building Code clauses C1–C6 Protection from Fire (Amendment 4). Wellington, New Zealand: Ministry of Business, Innovation and Employment.

 [5] MBIE. (2014). C/VM2 Verification method: Framework for fire safety design for New Zealand Building Code Clauses C1–C6 Protection from Fire (Amendment 4). Wellington, New Zealand: Ministry of Business, Innovation and Employment.

 [6] DBH. (2012). Extract from the New Zealand Building Code: Clauses C1–C6 Protection from Fire, Clause A3 Building Importance Levels. Wellington, New Zealand: Department of Building and Housing.

[7] BRANZ. (n.d.). About BRANZ. Retrieved from

[8] Collier, P.C.R. (2015). Characterising fire plumes from openings in external walls. BRANZ Study Report SR320. Judgeford, New Zealand: BRANZ Ltd.

[9] MBIE. (2013). Commentary for Building Code clauses C1–C6 and verification method C/VM2. Wellington, New Zealand: Ministry of Business, Innovation and Employment.

[10] Quintiere, J G. & Cleary, T.G. (1994). Heat flux from flames to vertical surfaces. Fire Technology, 30(2), 209–231.

[11]Thomas, P.H. & Law, M. (1972). The projection of flames from burning buildings. Fire Research Note No 921. Borehamwood, Hertfordshire, UK: Fire Research Station.

[12] Law, M. (1978). Fire safety of external building elements: The design approach. Engineering Journal, 15(2), 59–74.

[13] Oleszkiewicz, I. (1989). Heat transfer from a window fire plume to a building facade. ASME HTD – Collected Papers in Heat Transfer, 123, 163–170.

[14] McGrattan, K., et al. (2017). Fire dynamics simulator technical reference guide. Volume 3: Validation. NIST Special Publication 1018-3 (6th ed.). Gaithersburg, MD, USA: National Institute of Standards and Technology.

[15] Axelsson, J., et al. (2001). Uncertainties in measuring heat and smoke release rates in the room/corner test and the SBI. NT Techn Report 477, NORDTEST Project No. 1480-00, SP REPORT 2001:04. Boras, Sweden: Swedish National Testing and Research Institute.

[16] Back, G., Beyler, C., Dinenno, P. & Tatem, P. (1994). Wall incident heat flux distributions resulting from an adjacent fire. Fire Safety Science – Proceedings of the Fourth International Symposium, 241–252.

[17] Drysdale, D. (1999). The post-flashover compartment fire. In An introduction to fire dynamics (2nd ed.). Chichester, West Sussex, England: John Wiley & Sons.

© SFPE® | All Rights Reserved
Privacy Policy