FPEeXTRA Issue 74

The Effectiveness of a Water Mist System in an Open-plan Compartment with an Exposed Timber Ceiling: CodeRed #03

By: P. Kotsovinos, E. G. Christensen, J. Gale, H. Mitchell, R. Amin, F. Robert, M. Heidari, D. Barber, G. Rein, and J. Schulz

View full PDF here

In recent times, there has been a global surge in the design and construction of mass timber buildings due to their benefits, particularly regarding sustainability. Many of these buildings are proposed to be used in commercial premises with open-plan layouts. As discussed by Rackauskaite et al., [1] compartment fire experiments with exposed timber surfaces published at the time of the authors’ review were limited to a compartment area of 84 m2. In contrast, open floor office spaces often exceed 1000 m2. To address this limitation in research and meet current design needs, researchers conducted a series of full-scale fire experiments in a very large purpose-built compartment of 352 m2. To address this limitation in research and meet current design needs, the research team conducted a series of full-scale fire experiments in a very large purpose-built compartment of 352 m2. The experiments were performed at CERIB’s fire testing facility in France. These experiments aim to capture fire dynamics in large compartments with exposed timber and develop solutions for design [2], [3].

This article presents novel experimental evidence from the third experiment in the series, CodeRed #03. The aim of the first two experiments (CodeRed #01 and #02) were to capture fundamental fire dynamics, while the third experiment was to investigate the effectiveness of a standard water mist suppression system in both limiting fire growth as well as preventing the ignition of an exposed CLT ceiling. Providing an automatic suppression system is a common mitigation measure used in mass timber buildings that, with other protective measures (such as encapsulation, fire-fighting facilities, etc.), aim to address the additional risks introduced by the combustible nature of the structural frame.

Experimental Set-up

The facility where the experiment was undertaken is described in detail in [2]. The compartment has internal dimensions 10.27 × 34.27 × 3.1 m with a floor plate area of ~352 m2. The CodeRed #03 experiment has several open windows and doors to imitate possible ventilation conditions in an office environment with openable windows during the very early stages of a fire.

The equivalent theoretical fire load that would be present in buildings was replicated by constructing a wood crib in part of the enclosure with a fuel load density of ~ 570 MJ/m2 and which covered a floor area of ~50 m2 (7 m × 7 m). The fuel load density corresponds to the 80% fractile for offices as per PD 6688-1-2:2007 “Background paper to the UK National Annex to BS EN 1991-1.2” [4], the accompanying document to BS EN1991-1-2 “Eurocode 1: Actions on structures - Part 1-2: General actions - Actions on structures exposed to fire” in the UK [5].

The wood crib was placed against the northeast wall, 0.5 m from the southeast wall, to allow for access to ignite the cribs in the corner (see Figure 1).

Figure 1. Arrangement of the wood crib in the corner of the building, with automatic water mist system installation above.

The instrumentation layout was based on the previous experiments, adjusted to the area of interest, as illustrated in Figure 2. 39 thermocouples were distributed across the compartment, organized into 11 thermocouple trees (T) with thermocouples hung at 0.3 (in a limited number of locations (T12-T17)), 1.0, 2.4, and 3.0 m above the floor level (FL) (see Figure 5). Five thermocouple trees were placed near the centreline of the compartment, and four were placed along the edge of the crib on both sides of the centreline. Twelve cameras were set up to record the experiment inside the building, through openings, and outside the building.

Figure 2. Instrumentation plan.

Water Mist Installation

The water mist system was installed such that it achieved the minimum water density to meet the criteria of Ordinary Hazard category 1 (OH1) of FM 5560 Appendix G “Water Mist Systems” by FM Global, which is also representative of BS 8489-7:2016 Fixed fire protection systems. Industrial and commercial water mist systems - Code of practice for design the protection of low hazard occupancies, which are common requirements for office buildings in the UK.

Based on the compartment characteristics the water discharge density over the operating area for OH1 of 72m2. This resulted in a design configuration of 178.0 l/min at 6.185 bar or 2.67 litres/m/min.

The design density was aimed to be controlled such that it would represent the lower end of what is commonly provided as part of normal design practice.

The activation temperature for the mist nozzles  was 680C. The nozzles had a K factor of 14.5, and the system was designed for a minimum pressure of ~6 bar.

The nozzle layout was in line with the manufacturers design spacing: 5 nozzles operating on 4.5 m × 3 m spacing, with the actual configuration governed by the 0.55 m deep and 0.25 m wide central downstand beam within the compartment as illustrated in Figure 2.  

To confirm that the minimum water density was achieved during the test, a pressure-reducing valve set at 6 bar was installed within the pipe network. The post-test results indicated that the actual pressure was slightly higher (7.36 bar average).   

Ignition Source

The ignition source was two pans (150 × 250 mm), each filled with 0.5 l methanol, inserted below the wood crib to mimic the ignition of the localized corner fire test described in FM 5560 (large compartment corner test). Over a surface of 4 m² in the ignition corner, 2 layers of 5 mm fibreboard were added between the wood sticks to ensure the uniform fire spread in all directions from the ignition point.

The fuel type and room geometry were different from that of the FM Global standard test, which does not have openings to outside, and hence is sheltered from the wind. In the FM 5560 large compartment corner test, the fuel type is wood crib and simulated furniture. The wood crib is limited to a localized area of 300 mm by 300 mm by 150 mm high.

While the ignition method did not follow the FM Global standard, it can be argued that the ignition source used here represents a more conservative arrangement compared to the ignition source in FM 5560. For the large compartment corner test, a tray with 0.47 L of water and 0.24 L of commercial-grade heptane are used to ignite the limited extent of wood cribs. As a result, both the quantity of fuel used for ignition as well as the extent of the area affected by the ignition were greater for CodeRed #03, indicating a faster initial fire growth when compared to the FM 5560 large compartment corner test.

Figure 3. Activation of the first water mist nozzle near the location of ignition of the wood cribs at the corner of the building (taken at 1 min 50 s after ignition).


The crib was ignited at 14:58:20 on the 4th of November 2021. At the time of ignition, the ambient temperature was 12.1℃, and the wind speed was approximately 1.58 m/s.

Fire growth and nozzle activation

Following the ignition of the methanol pans, a localized fire was quickly formed, and the fire steadily grew. The height of the flames is shown in Figure 4, and was measured through video analysis. Flame height was defined as the highest location of the flame in each image. The fire eventually triggered a total of five water mist nozzles within a period of 5 min 28 s. Four of these were on the same side of the room as the ignition and one on the far side of the downstand beam. 

The first water mist nozzle (“N1”) activated 1 min 30 s after ignition. At that time, the fire had a flame height of ~2.09 m as illustrated in Figure 4. Following the activation of the first nozzle, the growth of the flames rapidly declined, and an approximate steady flame height was achieved. The second nozzle (“N2”) activated soon after the first nozzle at 1 min 48 s. The maximum flame height observed at 2 min 8 s, reaching 2.76 m above floor level (0.34 m below the ceiling level). Following the activation of the 3rd (“N3”) and 4th (“N4”) nozzles at 2 min 40 s and 2 min 56 s respectively, the flames reduced in size by roughly half, as illustrated in Figure 4. The 5th nozzle (“N5”) activated at 5 min 28 s.

The continued application of water mist decreased the fire size until there was minimal flaming. The fire was reduced to near extinction within 20 min of ignition. The residual flaming was manually extinguished after 20 minutes using a 70mm diameter water hose to minimize the further water discharge into the compartment.

The ability of the water mist system to control the severity and growth of the fire was evidenced by the number of wood sticks from the wood crib fuel affected by the fire. Out of the 3626 timber sticks used in total for the purposes of the experiment, only 138 timber sticks were fully or partly damaged. The limited extent of the damage can also be seen in Figure 5, which was taken shortly after the CERIB staff extinguished the remaining small flaming fire approximately 20 minutes after ignition.

Figure 4. Fire development with time, showing Flame height above floor level (left) and the estimated heat release rate (right).

The crib was elevated above the floor using bricks and constructed with 19 layers of wood sticks. The total height of the wood crib above floor level was 0.64 m. Flame height was recorded from when the flame was visible over the crib. the dotted vertical line indicates the activation time of nozzles 1-4.

Figure 5. Limited damage to the wood crib and discolouration of the ceiling after the suppression of the fire.

Fire size (HRR)

The approximate HRR from the crib fire was calculated using the well-established methodologies of both Heskestad [6] and Thomas [7]. This was done over the duration for which the flames were visible.

The equation proposed by Heskestad [6] is given in Equation 1; 

Equation 1 was rearranged to provide heat release rate,  as a function of the measured variables of flame height (L) and flame diameter (D).

The second methodology used was the correlation by Thomas [7] for visible mean flame height is given in Equation 2;

The equation was rearranged to provide  - mass burning rate per unit pool area as a function of , the density of ambient air, and the same measured variables used in Equation 1.  was multiplied by the fire area and heat of combustion (wood: 19.05 , obtained via bomb calorimetry) to give the heat release rate values using Equation 2. The base diameter and height of the flame was established following image processing of the photographic and video data.

The estimated heat release rate development of the fire is presented in Figure 4. The peak HRR was estimated to have reached between 762 and 1205 kW, based on the Thomas and Heskestad methods, respectively. The peak HRR was achieved 2 min 40 s after ignition, roughly at the same time as the activation of the third nozzle. This difference in HRR values is likely linked to the varied definition of flame height and that these methodologies were developed for pool fires and therefore do not necessarily converge for crib fires. The observed flame height was measured by image processing as the topmost of the flames. Heskestad defines the parameter L as “mean flame height” (i.e., average position of the luminous flame), while Thomas’ defines it as visible flame height. For the purposes of this work, both methodologies are provided as a range of HRR estimates.

The HRR of the fire was also calculated based on the quantity of timber burned, as determined after the end of the fire, and approximately fitting a HRR curve to Figure 4. The total energy consumed during the experiment was estimated to be 650 MJ (without combustion efficiency). The assumed HRR curve consisted of a t-squared fire growth, which reached a steady-state peak HRR, starting at 1.5 min after ignition, followed by a linear cooling stage which lasted from 3 min to end of the fire (~20 min). Applying a combustion efficiency of 0.7 and adjusting the peak HRR value such that the area under the HRR curve equalled the total fuel consumed, the peak HRR was estimated to be ~858 kW.

This is in good agreement with the range of peak HRR values calculated using the Heskestad and Thomas methods. It is outside the scope of this paper to complete a more detailed characterization. Future research could, however, more accurately estimate the observed HRR.

Compartment temperatures

The location of each activated nozzle can be seen in Figure 6 as well as the temperature profiles of four of the nearest thermocouples surrounding the ignition, located near the ceiling (FL: +3.0 m, CL: -0.1 m). The temperature near the ceiling was also observed to continue to grow at all four locations included in Figure 6, despite the activation of the first four nozzles. For the three thermocouples nearest the ignition location (“T11”, “T16”, and “T6”) the temperatures only started to drop after the activation of the 5th nozzle. For the thermocouple further away from the ignition (“T10”), there was an initial decrease in temperature after the activation of the 4th nozzle, before a steady increase until the 5th nozzle was activated, which led to a decrease of temperatures as with the other thermocouples.

The peak temperature in the compartment was measured by the thermocouple nearest the ignition (“T11”: 3m above floor level, -0.1 m below the ceiling) at 185℃, while temperatures further away from the fire (thermocouples “T10”, “T16”, “T6”) never exceeded ~85℃.

The temperatures near the ignition location as measured by thermocouple T11 returned back to almost ambient conditions approximately 20 minutes after ignition.

Impact on CLT ceiling

The recorded ceiling temperatures for all thermocouples remained well below the 300℃ threshold [8] normally considered as the onset of charring of timber. No flaming combustion/charring of the timber was observed; however, discoloration occurred at the timber ceiling immediately above the ignition source (see Figure 5) over an area of approximately 1.5 m × 1.5 m.

Saw cut samples taken from the CLT above the ignition location after the end of the fire indicated a depth of discoloration of 3-4 mm (see Figure 5). 

Discharge density

Based on the pressure readings, the actual pressure was slightly higher than the target pressure, at 7 bar, resulting in the following discharge densities:

Table 1 – Summary of nozzle discharge density and activation time


Activation time

Average discharge density


1 min 30 s

3.0 litres/m/min


1 min 48 s


2 min 40 s

2.93 litres/m/min


2 min 56 s


5 min 28 s

2.88 litres/m/min

6. Temperature time development at different locations near the ceiling (3 m above floor level, -0.1 m below the ceiling).

The diagram illustrates the location of the nozzles, thermocouples, crib, and ignition. The vertical dotted lines indicate the activation time of the nozzles.


Water mist systems are typically tested using a non-combustible ceiling. As a result, concerns have been raised by several stakeholders whether automatic suppression systems can still adequately control fire growth for buildings with combustible CLT ceilings and if that is feasible before a potential ignition of the ceiling.

CodeRed #03 examines, for the first time, the performance of a standard low-pressure water mist system in an open-plan compartment with an exposed timber ceiling, with open windows, to reflect possible conditions in an open plan naturally ventilated office.

The water mist system was designed to OH1 standard with a target density of 2.67 litres/m/min for a maximum area of operation for OH1 of 72m2 corresponding to 5 nozzles operating on 4.5 × 3 m spacing, to represent the lower end of typical design densities used day to day.

Five nozzles activated over a period of 5 min 28 s, four on the same side of the down stand beam as the ignition source, and one on the far side of the beam. The average discharge density for the five nozzles was 2.88 ltrs/ m/min at a pressure of 7.36 bar.

The system successfully controlled the fire such that the maximum compartment temperatures near the ceiling in the immediate vicinity of the ignition never exceeded 185℃, and 85℃ elsewhere in the compartment. Compartment temperatures started to fall soon after the activation of the 5th nozzle, and ambient temperatures were reached after 20 minutes. At that time, the remaining residual flaming was manually extinguished to limit the amount of water discharged into the compartment.

The maximum flame height observed was 2.76 m, and the maximum heat release rate was estimated to be between 762-1205 kW. The CLT ceiling did not ignite or char; however, it showed some discoloration to a depth of circa 3-4mm.

CodeRed #03 has therefore illustrated that, for the onerous configuration and parameters selected, a low-pressure water mist system installed in a large compartment with a combustible CLT can adequately control fire growth, even with open windows. Any application of water mist systems in buildings with CLT need to be discussed and agreed with the authority having jurisdiction and insurers on a case-by-case basis.


The experiment was fully funded by Arup through an internal research grant. VID Fire-Kill are kindly acknowledged for supporting the experiment with the water mist system design and construction. The Imperial College London PhD students are gratefully acknowledged for traveling to France despite the COVID pandemic. All Arup staff that helped with the organization and analysis of the CodeRed experiments are also acknowledged.

P. Kotsovinos, E. G. Christensen, J. Gale, and J. Schulz are with Fire Engineering, Arup, London, UK.

H. Mitchell, R. Amin, and G. Rein are with the Department of Mechanical Engineering, Imperial College London, London, UK

F. Robert and M. Heidari are with Fire Testing Centre, CERIB, Épernon, France

D. Barber is with Fire Engineering, Arup, Melbourne, Australia


  1. Rackauskaite, E.Kotsovinos, P.and Barber, D., (2021), “Letter to the Editor: Design Fires for Open-Plan Buildings with Exposed Mass-Timber Ceiling,” Fire Technology, Vol. 57, No. 2, pp. 487–495.
  2. Kotsovinos, P.Rackauskaite, E.Christensen, E.Glew, A.O’Loughlin, E.Mitchell, H.Amin, R.Robert, F.Heidari, M.Barber, D.Rein, G.and Schulz, J., (2022), “Fire dynamics inside an open-plan compartment with an exposed CLT ceiling and glulam columns: CodeRed #01,” Fire and Materials.
  3. Barber, D., “Large compartment fire experiments: expanding knowledge of building safely with timber,” Arup, 2021. [Online]. Available: https://www.arup.com/perspectives/large-compartment-fire-experiments-expanding-knowledge-of-building-safely-with-timber.
  4. BSI, “PD 6688-1-2:2007 - Background paper to the UK National Annex to BS EN 1991-1-2,” 2007.
  5. CEN, EN 1991-1-2:2002 - Eurocode 1. Actions on structures. General actions. Actions on structures exposed to fire. Brussels, 2002.
  6. Heskestad, G., “Fire Plumes, Flame Height, and Air Entrainment,” in SFPE Handbook of Fire Protection Engineering, 5th ed., M. J. Hurley, D. Gottuk, J. R. Hall, K. Harada, E. Kuligowski, M. Puchovsky, J. Torero, J. M. Watts, and C. Wieczorek, Eds. New York, NY: Springer New York, 2016, pp. 396–428.
  7. Drysdale, D., An Introduction to Fire Dynamics, 2nd ed. Chichester, UK: John Wiley & Sons, Ltd, 2011.
  8. Richter, F.Kotsovinos, P.Rackauskaite, E.and Rein, G., (2021), “Thermal response of timber Slabs exposed to travelling fires and traditional design fires,” Fire Technology, Vol. 57, No. 1, pp. 393–414.