Structural hazards of smouldering fires in timber buildings
Author: Harry Mitchell Rikesh Amin, Mohammad Heidari, Panagiotis Kotsovinos, Guillermo Rein
Reprinted with permission from the Fire Safety Journal
Volume 140, October 2023, 103861
Under a Creative Commons license
Publisher: Elsevier, Date: October 2023
As modern timber buildings grow rapidly in popularity amongst stakeholders and practitioners, fire safety is an important challenge for exposed mass timber buildings. Timber is a combustible material; therefore it may contribute to the severity and structural impact of a fire. During a fire, timber elements such as columns and ceiling slabs can char, reducing the cross-sectional area that contributes to the building structure. Following the cessation of flaming and once most of the compartment has cooled to ambient temperatures, it is typically assumed that the structural hazards will no longer progress in severity; however, with the potential presence of localised smouldering, this may not be the case. A review of timber compartment experiments  highlights that most experiments conclude before confirming complete self-extinction (i.e., return of entire compartment to ambient temperatures), frequently using manual suppression to accelerate the decay phase. Due to the lack of data following the cessation of flaming, there has not been significant consideration to the hazards that may follow a compartment fire, such as smouldering of mass timber elements.
Smouldering is a slow, persistent, and flameless form of combustion, occurring in porous fuels such as timber [, , ]. Smouldering is a solid-phase heterogenous oxidation reaction, where char reacts with oxygen to form ash, carbon dioxide, water, and other gases. As a result of the oxidation reaction, heat is released, facilitating further pyrolysis of unburnt material. Two key mechanisms drive smouldering reactions: the availability of oxygen, and heat transfer [2,4]. An absence of oxygen or high heat losses can inhibit smouldering. The preheating, drying and pyrolysis (charring) processes ahead of the oxidation (smouldering) require heat energy to progress, and are therefore impeded by heat losses that can limit the energy supplied to each process. Oxidation is exothermic and requires a readily available supply of air either permeating through the unburnt fuel, known as opposed flow smouldering, or through the reacted fuel, known as forward flow smouldering. This means that smouldering is sensitive to fuel conditions and the local environment.
Smouldering of timber is a typical hazard studied in wildfire research community. After a wildfire progresses through a forest, smouldering of thicker fuels (soil, tree trunks, branches, as shown in Fig. 1) continues behind the extinction front of the wildfire . These thick fuels can smoulder for hours and days after the wildfire has passed through an area, further damaging the local ecosystem and atmosphere. This is called “residual burning” and can be responsible for the majority of the vegetation consumed and emissions produced by a wildfire. This paper therefore briefly discusses a range of studies on smouldering thick timber elements in wildfires due to their parallels to smouldering behaviour in the built environment.
Experiments with porous cellulosic fuels  found that the spread of smouldering can be the steady backward spread or the unsteady and comparatively slower forward spread, depending on if the oxygen supply flows from behind or ahead of the smouldering front. Further research  identified that the spread of smouldering on solid timber increases linearly with air velocity, with higher air velocities leading to transition to flaming in forward flowing smouldering. Ohlemiller posited that smouldering cannot occur in a thick timber slab without an external heat source, due to the oxidation zone being near the surface of the timber, meaning it is subjected to significant heat losses. Small-scale smouldering experiments on mass timber  determined that smouldering cannot self-sustain unless subjected to both a sufficient heat flux and airflow.
The smouldering of thick timber elements (logs) was investigated [6,8] by measuring the spread rate and temperature in logs after a rainforest clearing. Logs of diameter ≤700 mm were studied, giving a range of smouldering spread velocity of ≤0.402 mm min−1. Greater spread rates were observed at daytime compared to nighttime, and at higher wind speeds and lower humidity. The driving factors controlling the spread of smouldering were the ambient temperature, humidity, timber composition (e.g., moisture content, density), location, and boundary conditions of the log (e.g., locations where logs crossed each other created favorable smouldering conditions).
During a wildfire, nearby conventional timber structures [9,10] (timber-framed buildings, fences, and decking) can be impacted by firebrands, leading to smouldering, which can lead to further damage and hazards to occupants. Further studies  show crevices in timber construction can promote the initiation and spread of localised smouldering, known as hotspots, due to increased re-radiation from the smouldering surfaces, and lower convective heat losses. Firebrands from Australian bushfires were found to be a hazard for 20 timber bridges , with eye-witnesses noting that one bridge “smouldered slowly for several days”. Smouldering spread along the gaps between timber decks, in one case causing embers to fall and ignite the timber piers below. Small-scale and full scale (entire bridge) experiments were carried out, finding that self-sustained smouldering can continue for over an hour, as gaps between timber elements were found to improve smouldering spread to an average of 5 mm min−1, greater than in previous experiments .
Transition to flaming occurs when a change in conditions causes a smouldering region to begin flaming . This can be initiated by an increase in oxygen supply, heat generation, heat loss reduction, and volatile gases from the pyrolysis front (a front defining the furthest boundary at which pyrolysis occurs). In wildfires transition to flaming may result in secondary ignition of a previously extinguished wildfire or nearby structures, causing new fires for local fire services and communities to tackle.
A smouldering fire can be hazard to timber buildings for two reasons. Firstly, smouldering hotspots following a flaming fire are difficult to detect and suppress and can spread through timber elements for hours and days, leading to partial or complete structural failure, long after the hazard is assumed to be over. Secondly, smouldering can transition to flaming suddenly in unexpected locations, rekindling a fire that was thought to be extinct.
One such mass timber compartment fire experiment  observed structural failure of a mass timber ceiling after the decay phase. The ceiling collapse occurred 29 h following extinction and was attributed to smouldering. Mass timber compartment fire experiments at RISE  found that during the decay phase of the flames, smouldering hotspots developed on exposed and encapsulated timber walls. As a result, hotspots in five different experiments were identified using infrared (IR) imaging and extinguished with a high-pressure water mist hose following the removal of encapsulation to improve the visibility of encapsulated hotspots. These hotspots were extinguished within 75 min of being identified, before they could progress to be significant structural hazards . observed the failure of two loaded timber columns 90 and 153 min after ignition, during the fire decay phase. This was attributed to both the thermal wave from the fire progressing through the column and local smouldering driven by airflow and re-radiation from the furnace walls.
Concerns regarding how smouldering may behave in a real timber building may cause challenges regarding insurability . RISCAuthority highlights that voids containing combustible materials in buildings could allow flaming and smouldering to initiate and spread for long periods of time without detection, intervention, or easy suppression. Furthermore, voids span both vertically and horizontally through a building, potentially facilitating fire spread through the building. Despite these concerns, very limited research has been carried out to date to study the initiation and spread of smouldering in mass timber elements following a compartment fire. Therefore, this paper will overview the IR, visual and thermocouple observations and analysis carried out on a smouldering hotspots observed in the cross-laminated timber (CLT) ceiling and glulam columns of three open-plan compartment experiments over 48 h following the cessation of flaming.
Three mass timber compartments (CodeRed #01, #02, and #04) were constructed with a cross-laminated timber ceiling and two glulam columns. The compartment was 34.3 m × 10.3 m in floor area, and 3.1 m tall, closely following the layout of previous traditional open-plan compartment fire experiments, x-ONE and x-TWO . The layout and configuration of each compartment (CodeRed #01, #02, #04), and the key findings of each experiment with respect to fire dynamics are discussed in detail by Kotsovinos et al. [, , ]. CodeRed #02 , reduced the ventilation of the compartment by 50% compared to CodeRed #01 (depicted in Fig. 2) uniformly through the compartment. CodeRed #04  encapsulated the inner 48% of the CLT ceiling with three layers of 12.5 mm commercially rated encapsulation board, which was fixed mechanically using a combination of screws and staples (further details of encapsulation provided in Ref. ).
The CLT ceiling of all three experiments was constructed from 5 lamellae of spruce pine (40-20-20-20-40 mm), adhered in a cross-laminated arrangement. The ceiling was constructed from 13 panels of 2.5 m × 10.93 m, adhered together along the compartment width by stapling together to two intumescent seals and a jointing board along the top surface of the joint. The two 3.05 m tall glulam columns were constructed from two 0.2 m × 0.4 m glulam members and adhered together to form a 0.4 m × 0.4 m cross-section. The columns were not structurally loaded or mechanically connected to the floor or ceiling. Both the CLT and the glulam had a melamine adhesive that had been tested by the manufacturer to not experience char fall-off. A wood crib, spanning 170 m2 of floor area, was ignited along the ignition line, allowing flames to propagate along the compartment length and ignite the timber ceiling and column elements. Following the cessation of flaming, the exposed surfaces of the ceiling and columns had charred completely.
The primary aim of this study was to systematically study smouldering of mass timber elements in the compartment over 0–48 h following the cessation of flaming. To achieve this, mobile IR and visual cameras were used toinspect the ceiling and columns at regular intervals (0 h, 12 h, 36 h, and 60 h) and detect localised regions of smouldering, known as hotspots. These were identifiable by visual imaging as areas emitting sustained glowing and smoke, and by IR imaging by a high level of IR radiation (surface emissivity was set to 0.95), as shown in Fig. 3. To ensure that all hotspots in the ceiling were detected, IR imaging was used above and below the ceiling over the entire ceiling surface area. Observations were recorded until all localised smouldering reached extinction, which occurred within 60 h for all experiments, either due to extinction without intervention, or direct or indirect suppression (rainfall or firefighters with a water mist hose). The exception to this case is smouldering hotspot β4/τ4 observed in CodeRed #04 (seen in the supplementary file, figs. A1 and 11) which was not detected due to its location, and proceeded to smoulder underneath the encapsulation beyond 48 h after the cessation of flaming. The location and progression of each hotspot over 48 h is depicted in Refs. [, , ] and the supplementary material to this paper, and will be referred to by a Greek letter and subscript indicating the associated experiment (e.g., α4 is hotspot α from CodeRed #04).
Further to this analysis, incidental measurements of individual smouldering hotspots were taken, specifically hotspots α1, λ2, and the glulam column in CodeRed #02, using a range of techniques (visual, IR, and in-depth thermocouple tracking). The methods and findings of each individual study of smouldering spread are discussed further in the results section of this paper.
4.1. Initiation of hotspots
Following the cessation of flaming along the mass timber ceiling, the entire charred ceiling surface continued to glow, indicating smouldering. Glowing across the surface of the ceiling receded slowly, and temperatures dropped to ambient levels. However, over this period of decay and cooling, several localised regions in all three experiments continued to glow, both in the ceiling and glulam columns; these regions are known as smouldering hotspots, an example of which is shown in Fig. 3. A summary of the number of smouldering hotspots in each experiment is given in Table 1, which shows that over three experiments 19 hotspots were observed, nine of which resulted in holes through the thickness of the mass timber ceiling, and one resulting in the collapse of a glulam column.
Table 1. Summary of observed smouldering behaviour during each experiment.
|CodeRed #||01 ||02 ||04 |
|Holes||2 (CLT)||6 (5 CLT and 1 column collapse)||2 (CLT and encapsulation)|
|Suppression||Rainfall at 38.2 h||Fire hose on hotspot α2 and rainfall at 48–54 h.||No rainfall and no water mist hose used|
|Avg. spread rate (mm min−1)||1.30 (α1)||0.42 (λ2)||No spread rate analysed|
Smouldering hotspots initiated exclusively along the edges of each mass timber element. This included connection lines between two CLT slabs along the width of the compartment, connections between the wall and CLT slabs (some are also along the interface between two CLT slabs where they meet the wall), the interface between CLT slabs and the insulated concrete column spanning the compartment length, or the base of a glulam column at the floor, insulated by mineral wool. A summary of the smouldering observed over the three experiments is outlined in Table 1.
4.2. Smouldering spread
The spread rate of smouldering in mass timber elements is an important parameter to characterise the impact it may have on a structure over time. In order to measure smouldering spread rates, the progression of the localised smouldering region near a wall in CodeRed #02 (hotspot λ2, depicted in Fig. 3, Fig. 4) was tracked over the 12–42 h after the cessation of flaming. After detecting the presence of hotspot λ2 using a mobile IR camera, a static IR and visual camera was positioned on the floor, and directed upwards at the ceiling.
The visual and IR progression of hotspot λ2 from 12 h after initiation to extinction is depicted in Fig. 3. At 12.8 h after the cessation of flaming, smouldering is not visible by direct inspection and is only visible by IR. At 12–13.5 h after the cessation of flaming, smouldering was observed to be occurring at the charred surface of the CLT in a small region (excluding hidden smouldering in the CLT directly above the wall). From 13.9 h, smouldering transitions to flaming for 20 min, causing flames to extrude intermittently along the charred surface of the ceiling. A hole through the CLT thickness is visible from approximately 17.3 h, positioned behind the smouldering edge of the hotspot region as it continues to progress along the ceiling width. The hole and hotspot continue to progress away from the wall, along the CLT slab connection line until 41 h after flames, at which the smouldering process ceases, leaving a hole 600 mm in length along the slab connection line.
As applied in Ref. , a threshold normalized infrared intensity of 0.2 was chosen to define a region of IR intensity, as shown in Fig. 4.The perimeter of this region defines hotspot λ2, which can be defined by a leading edge, approximately where preheating and charring of the timber occurs ahead of the smouldering, and a trailing edge, approximately where the smouldering processes occur. The hotspot was captured at regular intervals, and an average distance of each front from the wall for both edges was taken and depicted in Fig. 4. Over the period where the hole was formed by hotspot λ2, the preheating and smouldering fronts spread at an approximately linear rate of 0.41–0.42 mm min−1. The smouldering and preheating fronts are sensitive to variations in local conditions such as wind, meaning that despite the fronts progressing at a linear rate they have some variability over the duration of the smouldering.
In CodeRed #01, over the 38 h after flames, a hotspot developed and spread at the ignition end of the compartment, at the midpoint of the CLT surface near the mineral wool insulated concrete mid-span beam (hotspot α1, as depicted in Fig. 5, left). Infrared imaging was applied to identify the location of α1 which immediately after the cessation of flaming was not otherwise identifiable from direct observation. Fig. 5 shows that smouldering is occurring within the CLT in a small, localised region, which grows in area over the span of 38 h following the cessation of flaming. Once the hotspot α1 was identified, time-lapse visual imaging were used to track the hole progression. As smouldering causes the recession of char and therefore formation of the hole, the hole perimeter is assumed to be the trailing edge of the smouldering front. On each image, colour segmentation was applied to define the boundary of the smouldering hole, where changes in the perimeter size are driven by a combination of char fall-off and smouldering. By identifying the camera location and orientation, and the plane of the ceiling, ray tracing methods were applied to translate the smouldering perimeter from the ceiling to 2D space, (shown in Fig. 5, centre), similar to the methods implemented to analyse flame spread [, , ]. The average smouldering front location along the width of the ceiling was calculated (Fig. 5, right).
The average smouldering front spread along the ceiling width at a linear rate of 1.3 mm min−1. Due to the thickness of the uncharred timber, this likely follows a forward smouldering regime, where the flow of oxygen is drawn through the char rather than via the uncharred timber. Compared to spread rate values tabulated in Ref. , this spread rate is on the higher end of observed smouldering spread, however it should be noted that smouldering is sensitive to boundary conditions such as airflow. Referring to the work of Ohlemiller , this rate of spread is within the region where transition to flaming is possible, as evidenced by the intermittent transition to flaming observed over the experiments smouldering period. As discussed by Santoso , smouldering spread and transition to flaming can be more likely to occur with the provision of insulation close to the smouldering surface, while still providing a small gap for air flow. Further to this , outlines that smouldering between two adjacent solid fuels can be accelerated by a small airgap, in this case the hole through the CLT formed by the smouldering, forming a chimney affect, providing enhanced airflow, and leading to a greater spread rate.
At 38 h after flames, a 23 min period of rainfall occurred, causing the smouldering to extinguish. This has been noted to be an effective method of extinguishment in other smouldering media such as peat . Although suppression is important, without thermal tools such as thermocouples or IR cameras, these hotspots can be hard to identify before they begin to alter the structural capacity of the mass timber element, as evidenced by Fig. 3. Comparison of smouldering hotspot λ2 from CodeRed #02 and α1 from CodeRed #01 show that the location of smouldering can have a significant impact on the behaviour of smouldering, including spread rate, extinction, and transition to flaming.
4.3. Column smouldering
Unlike , each glulam column was unloaded and had no in-built connection to the ceiling or floor, partially contributing to one column in CodeRed #02 falling to the ground. This column was observed to smoulder at its base from 0.00 h after the cessation of flaming. The column reached temperatures of ≤400 °C during the fire, before steadily cooling below 100 °C. However, as can be seen in Fig. 7, temperatures at thermocouple locations embedded both inside the timber element and along the glue line began to increase above ambient conditions from around 12.00 h following the cessation of flaming, indicating the spread of smouldering from the base of the column in-depth and upwards along the column length. Smouldering at the base of the column was also observed with IR imaging. The column base smouldered for 31.82 h through the cross-section of the column base until the top of the column was observed to shift from the ceiling connection point. At 32.02 h, as shown in Fig. 6, the column collapsed fully onto the floor. This was attributed to degradation of the column base cross-section due to continued smouldering.
Embedded thermocouples at the base of the column are depicted in Fig. 7, showing the column base reaching up to 500 °C at 28 h after the cessation of flaming. Thermocouples closer to the surface of the member experienced a drop in temperature hours before the column collapse, indicating the smouldering had charred through this region of column material, exposing the thermocouple to air local to the smouldering region. Taking a fixed temperature for smouldering (300 ± 30 °C), the hotspot spread through the column base was estimated, as shown in Fig. 8. The threshold temperature was reached earlier deeper in the column thickness, indicating smouldering progressing in-depth through the column thickness. Smouldering occurred at a greater rate along the glue-line compared to in-depth through the element, potentially due to re-radiation between the two smouldering surfaces at the glue line. After the column collapse, the side of the column facing the floor continued smouldering, resulting in near-full decay of the column cross-section across the column length. Due to the similarity in the dimensions of and boundary conditions of the glulam column that this behaviour is analogous the aforementioned log experiments [8,15].
4.4. Impact of encapsulation
The encapsulation used in CodeRed #04 was a commercial encapsulation that covered 48% of the CLT ceiling exposed surface, with the goal of reducing the overall contribution of the CLT to the fire load. During the fire, the 52% of protected CLT remained uncharred by the flames. Over 48 h of observation, no smouldering was observed underneath the encapsulation.
Two hotspots at the interface between the CLT and the crib ignition-end wall formed holes through the CLT thickness (hotspots α4 and β4). Further to this, both hotspots α4 and β4 continued to spread along the width of the compartment, until the smouldering progressed underneath the encapsulation (hotspot τ4), as shown in Fig. 9. 22 days (528 h) following the cessation of flaming, an approximately 1.7 m2 region of encapsulation collapsed (Fig. 9), directly below a region of smouldering similar in size that had spread through the thickness of the CLT ceiling. This undetected smouldering indicates that although encapsulation can be effective at mitigating the impact of fire on mass timber elements, it is not completely effective at preventing smouldering, particularly when it spreads from the edges of the encapsulation.
4.5. Transition to flaming
In several smouldering hotspots, transition to flaming was observed, following the formation of holes in the ceiling. This is due to an increase in air flow local to the hotspot, providing greater oxygen to the smouldering process, allowing a greater rate of char, leading to flaming. In CodeRed #02, hotspot λ had a 20 min period of transition to flaming prior to a hole being formed through the ceiling. This was attributed to smouldering forming a small gap between the ceiling and the wall, resulting in an increase in air flow through the hole, causing a transition to flaming. Smouldering further widened the air gap, reducing the air velocity through the hole (following Bernoulli's principle), driving a transition back to smouldering.
Transition to flaming can be particularly hazardous in the case of smouldering that forms penetrations to adjacent compartments or other floors, such as in Fig. 10, as this can lead to unburnt compartments being exposed to a new ignition source. In a design scenario where a compartment of similar design to CodeRed may represent a single floor in a tall timber structure, transition to flaming in a breach of compartmentation between an unburnt compartment and a burnt-out compartment may lead to a new compartment fire being initiated, presenting all new structural hazards. As a result, this paper recommends that future work continues to understand the conditions under which transition to flaming may occur in smouldering mass timber elements following a compartment fire, and methods of avoiding these conditions to prevent multiple fires in the same structure.
4.6. Extinction and suppression
The end of the decay phase of a building fire is typically defined as the time at which the entire compartment has cooled to ambient temperatures . Flaming will cease long before the majority of the compartment has reached ambient conditions, however smouldering, as evidenced by CodeRed, can continue hours and days after the decay phase is assumed to end. Following the cessation of flaming of each CodeRed experiment, the entire of the CLT surface smouldered during the decay phase before extinguishing without suppression, known as self-extinguishment. The remaining 19 hotspots observed highlight that structural hazards in a compartment fire continue not only during the decay phase, but in the hours and days afterwards. All hotspots extinguished irrespective of rainfall at 48 h, 60 h, and 22 days after the cessation of flaming in CodeRed #01, #02, #04, respectively. Some hotspots were suppressed completely by rainfall (e.g. α1), however some experienced periods of rainfall but continued smouldering afterwards (e.g. λ2). Suppression of hotspots occurred either independently or by the addition of water local to the smouldering region, either from a period of rainfall (occurred in both CodeRed #01 and #02), or with a fire hose (used on hotspot α2).
The conditions for self-extinction are important to identify when a mass timber element will cease smouldering, therefore reducing further hazards to the structure. As previously discussed, Crielaard  proposes that smouldering of timber will self-extinguish below a threshold heat flux and incident airflow. However, the complex design of timber buildings means that heat losses in certain locations can be low, smouldering surfaces can re-radiate onto each other, and sufficient air flows can occur naturally within a compartment.
Previous literature  hypothesised that a thermally thick timber slab will not smoulder unless subjected to a sufficient radiant heat flux, meaning that self-sustaining smouldering requires re-radiation. Due to the complex geometries observed in the CodeRed experiments, self-sustaining smouldering hotspots were observed over the 48 h following the cessation of flaming. Smouldering is an exothermic reaction, there is a balance between the heat released, the heat lost to the surroundings, and the heat required to perpetuate the smouldering . Therefore, smouldering timber favors regions which reduce head losses of smouldering, including voids, underneath encapsulation, and near insulating materials such as mineral wool. Smouldering will also spread at a greater rate when subjected to a higher airflow, providing a greater supply of oxygen to the oxidation reaction. In the context of tall timber buildings, the spread of smouldering may be enhanced by regions where airflow exists, such as small gaps between timber connections.
As discussed previously, variations in experimental conditions can influence the formation and progression of hotspots. In CodeRed #01 the ambient temperature ranged between 2 and 12 °C, whereas CodeRed #02 had an ambient temperature between 18 and 25 °C. Similarly, moisture content of the timber ceiling (11.1%, 11.6%, and 13.5% at a dry basis for CodeRed #01, #02, #04, respectively) could have an impact on the initiation and progression of smouldering [6,8]. This would impact the convective heat loss from the smouldering timber to the surrounding ambient air, impacting the rate of smouldering, and therefore extinction and formation, of hotspots.
Although this paper found that smouldering only occurred along the edge of timber elements, predicting where a hotspot may develop along the edges of all timber elements in a building is a probabilistic phenomenon. Considering the overall edge length of each experiment (154–257 m where edges comprise of ceiling slab to wall connections, ceiling slab to ceiling slab connections, and ceiling slab to insulated concrete beam connections), a hotspot occurrence frequency of 0.012–0.58 hotspots per meter of edge was found. Scaling this to modern design demands such as office spaces of 5000 m2, 30–150 hotspots may occur, each presenting a potential hazard to the structure.
The prevention of smouldering in timber buildings can be addressed by avoiding designs which promote its initiation. However, in application timber structures have other design components. Therefore, this paper recommends smouldering is studied further in complex timber design elements to determine the hazard that smouldering may pose. These include voids between compartments with combustible elements (e.g. ceiling and wall slabs, voids behind combustible facades), structural connections (e.g. ceiling slab connections, ceiling to column connections, beam and column connections), penetrations in slabs for electrical/plumbing servicing, and any regions where mass timber is insulated by another design component or an air gap (encapsulation, insulation, and other low-conductivity construction materials).
Encapsulation is used in mass timber design to protect timber elements from fire, therefore limiting their contribution to the fire load, and limiting the impact of the fire on the element's structural integrity. Smouldering not only can still occur in encapsulated timber elements, but this can often make smouldering more challenging to locate even while using IR imaging . Therefore, it may be required to remove encapsulation following a fire for firefighters to effectively detect smouldering hotspots that would otherwise remain hidden by the encapsulation.
Detection is vital to help firefighting operations identify if smouldering is present in a timber structure, and therefore if structural hazards are present. This paper has shown that in many cases smouldering may not be visible by direct observation, due to it being flameless and occurring in-depth and in hidden areas of in timber elements. Therefore, it is recommended that when investigating a timber building after a compartment fire, infrared imaging is used . However, it should be noted that concealed timber surfaces such as voids, encapsulation, and cavities can cause smouldering to not be immediately observable with infrared either. This paper and previous work  show that water mist hoses can extinguish smouldering hotspots when identified. However, it is recommended that work is conducted to understand the most effective methods to apply water to extinguish smouldering hotspots, and how to extinguish smouldering in regions that may be more challenging for firefighters to access.
Smouldering to flaming transition is primarily driven by a change in heat flux incident on the smouldering surface, and oxygen supply to the smouldering reaction . Transition to flaming can present a hazard to timber structures following the end of flaming by producing new fires hours after the original fire had ended. The definition of “extinction” is typically used by fire-safety experts to define when the hazard posed to a building during a fire have ceased. This is typically defined by either when the flames have ended, or shortly afterwards when most of the compartment has cooled to ambient temperatures. However, as evidenced by this paper, smouldering can not only cause localized regions of high temperature within a compartment, but also extend the period over which structural hazards continue to develop over hours and days after flames have ended.
This paper presents experimental observations and analysis of smouldering as a structural hazard in mass timber buildings. Smouldering in timber structures is a phenomenon which poses challenges in detection, suppression, compartmentation, and structural resistance. Smouldering was studied over 48 h following the cessation of flaming in the three large timber fire experiments, known as the CodeRed experiments. Beyond what has been outlined in CodeRed [, , ], this paper analyses the initiation, spread, transition to flaming, suppression, and extinction of localised smouldering of mass timber elements, supported by previous literature on the smouldering of timber [13,14,17].
Smouldering hotspots were found in each experiment, leading to the formation of holes, an instance of column failure, smouldering behind encapsulation, and transition to flaming. All hotspots posed a potential structural hazard to timber buildings. Nineteen hotspots were observed exclusively along the edges of timber slabs, primarily along the interface between the wall and the slabs, and the interface between slabs. The formation of holes by smouldering through a structural element could lead to stress concentrations, progressively increasing the likelihood of structural failure hours after the cessation of flaming. Transition to flaming is a hazard because it leads to the reappearance of flames in the same or adjacent compartments. Smouldering in hidden areas, such as behind encapsulation or voids, can lead to continued smouldering and larger fires. Overall, the findings of this paper emphasise the importance of understanding and addressing the impact of smouldering in the fire safety of structural mass timber elements.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
The authors would like to acknowledge the CodeRed team and steering group from Arup, CERIB where the experiments were conducted, and Simona Dossi. This research was funded by the Engineering and Physical Sciences Research Council and Arup.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
Multimedia component 1.
Data will be made available on request.
 H. Mitchell, P. Kotsovinos, F. Richter, D. Thomson, D. Barber, G. Rein
Review of fire experiments in mass timber compartments: current understanding, limitations, and research gaps
Fire Mater. (2022), 10.1002/fam.3121
 T.J. Ohlemiller, D.A. Lucca
An experimental comparison of forward and reverse smolder propagation in permeable fuel beds
Combust. Flame, 54 (1–3) (1983), pp. 131-147, 10.3801/IAFSS.FSS.3-565
Smoldering combustion propagation on solid wood
Fire Safety Science—Proceedings of the Third International Symposium, Routledge (2006), pp. 565-574, 10.3801/IAFSS.FSS.3-565
 G. Rein
SFPE Handbook of Fire Protection Engineering, Springer, New York (2016), pp. 581-603, 10.1007/978-1-4939-2565-0_19
 G. Rein, X. Huang
Smouldering wildfires in peatlands, forests and the arctic: challenges and perspectives
Curr. Opin. Environ. Sci. Health, 24 (2021), Article 100296, 10.1016/j.coesh.2021.100296
 E.R.C. Rabelo, C.A.G. Veras, J.A. Carvalho Jr., E.C. Alvarado, D.V. Sandberg, J.C. Santos
Log smoldering after an Amazonian deforestation fire
Atmos. Environ., 38 (2) (2004), pp. 203-211, 10.1016/j.atmosenv.2003.09.065
 Cook, J. A log smoulders near Cobbe (Photo), BBC, California, US. https://www.bbc.co.uk/news/world-us-canada-34286221..
 J.A. Carvalho Jr., F.S. Costa, C.A. Gurgel Veras, D.V. Sandberg, E.C. Alvarado, R. Gielow, …, J.C. Santos
Biomass fire consumption and carbon release rates of rainforestclearing experiments conducted in northern Mato Grosso, Brazil
J. Geophys. Res. Atmos., 106 (D16) (2001), pp. 17877-17887, 10.1029/2000JD900791
 S.L. Manzello, S. Suzuki
Experimental investigation of wood decking assemblies exposed to firebrand showers
Fire Saf. J., 92 (2017), pp. 122-131, 10.1016/j.firesaf.2017.05.019
 S.L. Manzello, S. Suzuki, M.J. Gollner, A.C. Fernandez-Pello
Role of firebrand combustion in large outdoor fire spread
Prog. Energy Combust. Sci., 76 (2020), 10.1016/j.pecs.2019.100801
 F. Richter, B. Bathras, J. Barbetta Duarte, M.J. Gollner
The propensity of wooden crevices to smoldering ignition by firebrands
Fire Technol. (2022), pp. 1-22, 10.1007/s10694-022-01247-w
 V.P. Dowling
Ignition of timber bridges in bushfires
Fire Saf. J., 22 (2) (1994), pp. 145-168, 10.1016/0379-7112(94)90070-1
 M.A. Santoso, E.G. Christensen, J. Yang, G. Rein
Review of the transition from smouldering to flaming combustion in wildfires
Front. Mech. Eng., 5 (2019), p. 49, 10.3389/fmech.2019.00049
 R. Crielaard, J.W. van de Kuilen, K. Terwel, G. Ravenshorst, P. Steenbakkers
Self-extinguishment of cross-laminated timber
Fire Saf. J., 105 (2019), pp. 244-260, 10.1016/j.firesaf.2019.01.008
 F. Wiesner, A. Bartlett, S. Mohaine, F. Robert, R. McNamee, J.C. Mindeguia, L. Bisby
Structural capacity of one-way spanning large-scale cross-laminated timber slabs in standard and natural fires
Fire Technol. (2020), pp. 1-21, 10.1007/s10694-020-01003-y
 D. Brandon, J. Sjöström, A. Temple, E. Hallberg, F. Kahl
Fire Safe implementation of visible mass timber in tall buildings–compartment fire testing
 T. Gernay, J. Zehfuß, S. Brunkhorst, F. Robert, P. Bamonte, R. McNamee, …, J.M. Franssen
Experimental investigation of structural failure during the cooling phase of a fire: timber columns
Fire Mater. (2022), 10.1002/fam.3110
Insurance challenges of massive timber construction and a possible way forward
 P. Kotsovinos, E. Rackauskaite, E. Christensen
Fire dynamics inside a large and openplan compartment with exposed timber ceiling and columns: CodeRed #01
Fire Mater. (2022), 10.1002/fam.3049
 P. Kotsovinos, E. Christensen, E. Rackauskaite
Impact of ventilation on the fire dynamics of an open-plan compartment with exposed CLT ceiling and glulam columns: CodeRed #02
Fire Mater. (2022), 10.1002/fam.3082
 P. Kotsovinos, E. Christensen, A. Glew
Impact of partial encapsulation on the fire dynamics of an open-plan compartment with exposed timber ceiling and columns: CodeRed #04
Fire Mater. (2023), 10.1002/fam.3112
 M.A. Santoso, W. Cui, H.M. Amin, E.G. Christensen, Y.S. Nugroho, G. Rein
Laboratory study on the suppression of smouldering peat wildfires: effects of flow rate and wetting agent
Int. J. Wildland Fire (2021), 10.1071/WF20117
 E. Rackauskaite, M. Bonner, F. Restuccia
Fire Experiment Inside a Very Large and Open-Plan Compartment
x-ONE, Fire Technol (2021), 10.1007/s10694-021-01162-6