Review of fire experiments in mass timber compartments: Current understanding, limitations, and research gaps

The use of mass timber in buildings instead of non‐combustible materials has benefits in sustainability, aesthetics, construction times, and costs. However, the uptake of mass timber in modern construction for medium and high‐rise buildings is currently hindered by uncertainty regarding safety and structural performance in fire. We attribute this to a lack of data. Insufficient understanding means that for building designs beyond the current range of experiments the fire performance is unknown. To address this uncertainty, we review the available data in the scientific literature from 63 compartment fire experiments with timber, the majority of which use cross‐laminated timber (CLT). We found that most experiments reached temperatures 80–180°C greater than in non‐combustible compartments. Timber ceilings have on average a 16% lower char rate than timber walls. The heat release rate contribution of timber has a positive linear relationship with charring rate, however is susceptible to significant uncertainty and variability. There are limits to the data available, particularly in large open‐plan compartments of floor areas larger than 100 m2 where a wider range of heating conditions occur. Other topics where further understanding is required are compartments with exposed timber areas greater than 75%, compartments with smaller opening areas, and the hazard of smouldering following the flames. Therefore, additional research is needed to design beyond the limits, specifically in compartment size, ventilation, and timber exposure. This paper identifies correlations in the current body of experimental research to improve fire‐safe design of timber buildings.

beyond the current range of experiments the fire performance is unknown. To address this uncertainty, we review the available data in the scientific literature from 63 compartment fire experiments with timber, the majority of which use cross-laminated timber (CLT). We found that most experiments reached temperatures 80-180 C greater than in non-combustible compartments. Timber ceilings have on average a 16% lower char rate than timber walls. The heat release rate contribution of timber has a positive linear relationship with charring rate, however is susceptible to significant uncertainty and variability. There are limits to the data available, particularly in large open-plan compartments of floor areas larger than 100 m 2 where a wider range of heating conditions occur. Other topics where further understanding is required are compartments with exposed timber areas greater than 75%, compartments with smaller opening areas, and the hazard of smouldering following the flames. Therefore, additional research is needed to design beyond the limits, specifically in compartment size, ventilation, and timber exposure. This paper identifies correlations in the current body of experimental research to improve fire-safe design of timber buildings.

| Motivation for mass timber construction
Mass timber, a term encompassing a range of large timber construction element types, is a globally used modern construction material that currently poses a dilemma. It exhibits benefits over conventional construction materials, such as concrete and steel, namely its reduced environmental impact, lower mass (and therefore, reduced foundation requirements), and capacity for cost-efficient and versatile prefabrication. Furthermore, exposed mass timber may increase the overall aesthetic appeal and value of the building. As a result, the volume of mass timber products manufactured, including cross-laminated timber (CLT), has increased exponentially over the past 30 years. 1 This has resulted in mass timber being adopted for progressively taller structures, as depicted in Figure 1.
Concerns regarding the structural and thermal performance of mass timber in fire have limited its acceptance in building design standards and construction projects. For the past 50+ years, research into timber construction has primarily focused on engineered timbers such as CLT and glulam. Furthermore, low-rise timber is considered to not pose as significant a risk as tall timber compartments, due to faster evacuation times, easy access by firefighting services, and lower structural requirements. As a result, design guidance has been primarily based on structural response to standard fire exposure and assisted with building code and standards for low-and medium-rise buildings, with some standards offering a maximum of 8 stories for sprinklered mass timber buildings. 3 As the fire behaviour of tall timber compartments continues to be reinforced by experimental research, practitioners may desire more timber stories to be permissible in future building design.
As mass timber buildings grow in popularity and size, the hazard posed by fire continues to exist. One key example of mass timber buildings being impacted by fire is the GlaxoSmithKline Carbon Neutral Laboratory for Sustainable Chemistry at Nottingham University, depicted in Figure 2. The laboratory was a 4500 m 2 , two-floor building with multiple compartments designed as labs and office spaces and a large mezzanine meeting space. 4 During the construction period, the building ignited, and despite 50 firefighters on the scene, the entire building collapsed due to the failure of the mass timber structural elements under fire. This not only highlights the risk posed to structures during construction when fire safety measures such as alarms and sprinklers may not be active, but also the general hazard that fire poses to modern timber structures.

| Motivation for experimental research
Research in the last decade has been carried out to further understand mass timber fire behavior through experiments and predict and understand fire behavior through modelling. Experimental efforts range in scale and complexity, including microscale (thermogravimetric analysis, TGA), mesoscale (laboratory-based), and macroscale (furnace testing, full-scale compartment fires). Full-scale compartment fire experiments are typically carried out to either obtain key parameters and relationships between design parameters and fire or timber behaviour (e.g., the impact of ventilation on compartment fire dynamics, as investigated by Kotsovinos et al.,6 ) or to support the development and validation of modelling tools.
Depending on the desired outputs of a given modelling approach, modelling tools for timber compartments follow three categories, as outlined in Figure 3: A. a design fire that provides thermal conditions that the timber element is subjected to F I G U R E 1 progression of the tallest completed (green) and planned (orange) modern tall timber buildings (to approximate scale). The current tallest tall timber building is the Ascent MKE building, comprising both laminated timber and glulam panels. 2 As the building designs increase in height, the typical floor area increases too, leading to larger compartment sizes with more complex fire safety requirements.
B. a thermal model that resolves how the timber element chars and responds to the given conditions C. a structural model that calculates the thermo-mechanical response of the timber to determine its structural capacity over time 7 Key modelling types listed in A-C are given with proposal for potential methods of application in Figure 3. These models vary in complexity (both in terms of simplicity of assumptions and computational complexity) and can be coupled to each other (e.g., a parametric fire curve can be coupled to a pyrolysis model, such as that carried out by Richter et al. 8,9 ) However, for the models to be appropriately accurate and grounded in the current understanding of fire dynamics in timber compartments, experimental studies are required to verify and validate each approach.
To this end, large-scale experiments can be carried out to determine the performance of timber in the presence of fire and its contribution to fire dynamics. Here, we analyse multiple experiments in previous literature as part of a series in order to identify how fire and timber behaviour are impacted by key design parameters (e.g., ventilation, compartment size, type, and orientation of timber within the compartment). This paper reviews a range of mass timber compartment fire experiments that investigated the impact of influencing factors, extrinsic to the mass timber itself, including compartment size, construction, ventilation (size of openings), movable fuel load (other combustible materials inside the compartment, such as furniture and wood cribs), and quantity of exposed mass timber. This was done to evaluate the current understanding of their influence on compartment fire behaviour, including temperature, contribution of mass timber to overall heat release rate (HRR), and charring behaviour of timber structural elements. The key findings of this paper aim to identify the current limitations of mass timber compartment fireexperimentation to date, and routes forward for informing future firesafe mass timber design.

| Physics of timber in fire
A compartment fire in a timber structure can be divided into multiple stages depending on the movable fuel load and timber behaviour, as depicted in Figures 4 and 5, showing both the response of the compartment and timber to a fire. The duration and occurrence of these stages are dependent on many factors, including the exposed area, thickness, and type of of timber, protection of the timber, ventilation F I G U R E 2 Nottingham University's laboratory (left 4 ) is primarily constructed of glulam mass timber beams, columns, and slabs. While under construction, a fire started (right 5 ), resulting in complete collapse of the structure F I G U R E 3 Flow diagram of methods of modelling timber compartment fire behaviour, supported by experimental data. Models and experimental inputs are divided into three categories: fire or thermal conditions to which the timber is subjected, thermal and charring response of the timber to local conditions, and the structural response of the timber due to charring behaviour. Numerical models can use experimental inputs and experimental data to support validation efforts dimensions, and size and duration of the movable fuel load. Figure 4 shows how each stage of a compartment fire will impact the average temperatures present in a compartment, which can increase and decrease depending on processes such as char fall-off occurring. Comparatively, Figure 5 shows how each phenomenon is driven by heat and mass transfer processes, both progressing into the timber (e.g., heat transfer into the timber, and charring and drying processes) and out into the compartment (e.g., radiation from flames, smoke).
One of the key processes that timber undergoes in a fire is charring.
The charring of timber is a complex mechanism that is dependent on a range of chemical and kinetic sub-processes. It is typically preceded by a drying front ( Figure 5), where moisture content present in the timber solid will evaporate at elevated temperatures and leave via timber surfaces. Work was carried out by Richter et al. 8,9 on understanding and modelling the stages of the reaction mechanism driving the charring of timber by developing a wood pyrolysis model. Richter investigates the performance of different proposed reaction schemes for the kinetics of wood pyrolysis across a range of complexities. Richter et al. 8,9 proposed that for macroscale modelling of mass timber compartments should involve simpler pyrolysis models, such as the proposed reduced complexity kinetics model, which models both oxidation and pyrolysis processes, simplifying the kinetics of timber charring to eight reaction steps characterising both material and kinetic parameters.
Richter et al. 8,9,13,14 use a charring model to investigate the response of a timber slab exposed to standard (design fires, including ISO, Eurocode, and SFPE standard curves) and non-standard (travelling fires of different sizes) fire curves (without accounting for the impact of timber on fire dynamics). It was found that for travelling fires, smaller fire sizes predicted a greater maximum char depth, especially towards the end of the compartment, due to smaller travelling fire sizes allowing for longer pre-heating times. The researchers also suggested that design fire models for open-plan compartments with exposed timber need to be developed, which has begun to be addressed via the CodeRed experiment series, 6,11,12 as shown in Figure 6. Figure 6 shows the movable fuel load (wood cribs) and mass timber ceiling used in the CodeRed #01 experiment, both of which are key to driving the fire behaviour of the compartment, further depicted in Figure 7, which shows that the total HRR of the compartment was approximately doubled due to the addition of the mass timber ceiling.
In all experiments reviewed in this paper, the fire growth stage driven by an external fuel source, which is typically either a liquid/gas burner, wooden cribs, or commercially available furniture that may be expected in an occupied modern timber compartment. In some compartments, fires charring may be minimal or not reached, by protecting the mass timber surfaces. Reducing or preventing timber charring can be achieved with commercially available protection such as gypsum board.
McGregor 15 showed that protecting mass timber can be a practical method of delaying the exposure of the mass timber by up to 39 min for a single layer of gypsum and in excess of 53 min with two layers. Experiments on a mass timber apartment and elevator shaft by Su et al. 16 show F I G U R E 4 Illustration of fire phenomena during a timber compartment fire. The fire could take multiple paths, from decay to single-or multiple-char fall-off and smouldering or a timber compartment without successful burnout immediately after the fully-developed phase. Curve (B) is representative of a noncombustible compartment. Behaviour based on experiments by Hadden et al., 10 and Kotsovinos et al. 6,11,12 F I G U R E 5 Diagram depicting the mass transport (black arrows) and heat transfer processes (red arrows) that may occur in an initially protected mass timber ceiling during a fire. The processes outlined here drive the temperature-time response outlined in Figure 1 that a timber slab wall with sufficient protection can completely compartmentalise the ignited compartment from another compartment. Another method of mitigating or preventing charring of mass timber is the addition of sprinklers, such as those outlined in the CodeRed #03 experiment described by Kotsovinos et al., 6 where sprinklers proved effective at controlling the spread of flames along a wood crib and prevented the ignition of a mass timber ceiling; however, the effectiveness of sprinklers in mass timber compartments is beyond the scope of this paper.
Experiments by Hadden et al. 10 show that failure of encapsulation can lead to an increase in HRR and temperature due to greater exposure of the timber structure to heat. Protections can range in thickness and design, with some including both gypsum board and mineral wool 17 and others using multiple layers of gypsum board. 18 Ignition can be calculated by either a critical heat flux or temperature at or above which ignition of timber will occur. Critical heat fluxes across a range of species of wood are found to be around 10-13 kWm À2 or 25-33 kWm À2 for piloted and unpiloted ignition, respectively, and critical temperatures of 204-408 C or 192-600 C for piloted and unpiloted ignition, respectively. 19 In most compartment fire scenarios, the movable fuel load will ignite first before reaching a point where the timber is ignited by the flames, usually by direct contact with the flames. This means that in most cases, timber elements would be expected to ignite under piloted scenarios.
Self-extinguishment of the mass-timber surface is a key component of the decay phase of a fire. Self-extinguishment can be defined as the cessation of flaming. Flaming of a timber surface will cease when there is insufficient volatile gases transferred through the char layer, specifically due to the char being thermally insulating between the compartment and mass timber. The self-extinguishment phenom- The amount of exposed timber was shown to have a minimal impact on the peak temperature, although the total HRR was significantly higher than predicted for a similar compartment with no exposed timber. This suggests that the contribution of exposed mass timber to the compartment HRR is important. If it occurs, selfextinguishment will typically happen significantly before the compartment temperature decays to ambient levels, meaning that there is a continued structural impact (i.e., smouldering) on the timber that occurs long after the cessation of flaming. Following the end of flames, non-flaming charring of the timber will continue to reduce structural capacity until the residual heat in the timber has dissipated, the external heat flux has reduced, and no smouldering is present. It should be noted that even after the end of flaming, heat transfer into the thickness of timber elements can lead to continued degradation of its structural integrity, as processes such as drying and general degredation of the timber structure can occur at much lower temperatures.    the structural behaviour of the compartment, and the interaction between both. This is done to understand and calculate the performance and role of timber in a real fire. 31 Timber performance, to some degree, can be characterised by the charring behaviour (rate, depth, and fall-off), HRR contribution of the timber to the fire, and the capacity for self-extinguishment. Liu and Fischer 31 also highlight the importance of reporting the behaviour of fire, timber, and compartments using multiple types of instrumentation (e.g., timber HRR contribution, char depths) in order to characterise both the fire dynamics and timber behaviour. The range of instrumentation required is discussed further in Section 3.1.
One of the key future research objectives highlighted by several of the previous reviews is understanding the contribution that timber elements make to the fire, specifically the HRR. As discussed in  compartments of mass timber. The floor area and exposed timber as a percentage of internal compartment area of each experiment is depicted in Figure 8, giving the range of scales and timber exposures in the literature.
The size of the compartments reviewed here, as depicted in    depth and the remaining timber thickness that can still contribute structurally. The assessment is often done by measuring the char depth or by measuring temperature profiles through the thickness of the timber using in-depth thermocouples. The latter method allows to measure the char layer progression over time, and charring rate variations over the duration of the fire and decay phase, which may vary significantly.
As shown in Figure 9, all experiments use in-air thermocouples. In 26 of the 63 experiments, thermocouples are embedded at varying depths in the timber elements; this allows for measuring the timber thermal response. In-depth thermocouple measurements allow for assumption of a single onset pyrolysis temperature (temperature at which pyrolysis reactions start) that can be used to estimate the charring front progression.
The experiment series carried out by Gorska et al. 41 investigated methods for calculating the timber charring rate by using surface and in-depth thermocouples. From here, a single pyrolysis temperature can be found (typically around 300 C) to define the depth at which charring starts.
The contribution of timber to compartment fire dynamics is key to understand the distinction between non-combustible and mass timber compartments. Under the current fire resistance framework, many models (e.g., mesoscale heat transfer and chemistry models) decouple the timber contribution from the fire dynamics of the compartment to predict the burning of timber. In respect to experimentally predicting the contribution of timber to the overall HRR, Brandon and Östman 33 identify four key methods previously used to estimate the HRR of the exposed burnt timber, HRR T . The first method, as outlined by Equation (1) 41 McGregor, 15 Crielaard et al., 23 and Hopkin et al., 43 in which gas sampling (air calorimetry) is used to determine the amount of air consumed and CO 2 produced by the burning timber and movable fuel, allowing for calculation of the total HRR from the compartment, HRR All . HRR All is then decoupled from the heat released by the movable fuel load by measurement of the mass loss rate, HRR MFL , to determine HRR T .
Finally, Equation (4) involves the comparison of HRR All to the HRR of a similar baseline experiment with no timber involved, HRR Base : 18 Using gas sampling or mass loss rates to estimate the contribution of timber to heat release requires calculating other HRR contributions to T A B L E 2 (Continued)  Figure 4 as the onset of flashover to the end of flaming and is key in defining the hazard present to a structure, as it defines F I G U R E 8 Percentage of internal compartment area consisting of exposed CLT against floor area, indicating regions of current understanding in timber compartment fire experiments. Small-scale, partially exposed, medium-scale, and partially protected compartments have been researched significantly, whereas partially protected timber compartments (exposed timber >75% of overall exposed wall and ceiling  finding the peak temperature was reached for a given opening factor, and for opening factors less than or greater than this peak, the average temperature decreased.

| Compartment temperatures
Thomas and Heselden noted that the overall layout of the fuel had a direct impact on the peak temperature, as represented in the two curves shown in Figure 10, where the dotted curve represents the best fit of experiments that had a larger crib spacing and the solid line representing experiments with a smaller crib spacing.
Gorska et al. 41 applied O to mass timber compartments, identifying that combustible surfaces may not be considered as sources of heat loss, instead contributing to the fire, and is, therefore, be removed from the heat loss term in Equation (2), giving a modified opening factor, O m . Gorska et al. 41 carried out 24 small-scale exposed timber compartment fire experiments, varying the percentage of exposed timber area, to compare to Thomas and Heselden. 49 Gorska et al. 41 introduced a modified framework for the compartment opening factor (Equation 6), where A T is the total internal compartment area excluding openings and the floor, A CLT is the total area of exposed timber, A W is the opening area, and H is the opening height.
This proposed O m was applied to a range of the experiments in Table 2, as depicted in Figure 10.

| Charring behaviour of timber
The charring behaviour of mass timber is an important component of assessing the fire resistance and structural capacity of an element and, by extension, the compartment and entire structure. This can be assessed by taking char layer depth measurements. This gives a value for the average charring rate and does not consider varying rates of charring during the experiment.
Comparison of the charring behaviour of timber depending on their specific application (i.e., as a wall or ceiling slab), as shown in Figure 12, shows that the charring rate for ceiling exposed timber surfaces is consistently lower than the charring rate of timber walls. This could be explained by the compartment dynamics following compartment Regime II, as proposed by Thomas and Heselden 49 ; in this regime, the hot experiment temperature region is directly below the ceiling, but this region, where the smoke layer typically resides, has a lower oxygen concentration due to poorer ventilation and the accumulation of smoke.
Hadden et al. 10 consistently shows a lower average charring rate for ceiling panels compared to wall panels. They introduce a characteristic time to represent charring behaviour during elevated temperatures (multiple characteristic times can occur in one experiment due to char fall-off, as shown in Figures 4 and 5). The difference in average charring rates for experiments with both exposed ceilings and walls is shown in Figure 12; in all but two of these experiments, the ceiling charring rate is less than or approximately equal to the wall charring rate.
In the compartment regime-II-CLT highlighted by Gorska et al.,41 the region below the ceiling during a timber compartment fire may have a low oxygen concentration, partially due to inefficient mixing with ambient air. This change in conditions local to a timber ceiling is hypothesised to heavily influence the response of the ceiling compared to that of timber walls, which must be investigated further. This is particularly depicted in Figure 12, which shows significant differences in charring rates for timber ceilings and walls over multiple experiments.
Timber models such as those outlined by Richter et al. 8,9 show via sensitivity analysis that timber charring is highly sensitive to parameters governing the local boundary conditions, which are primarily the incident heat flux, local oxygen concentration, and convective heat transfer coefficient (dependent on local gas temperature and velocity).
In order to support modelling efforts and understand the conditions mass timber are subject to during a compartment fire, experiments should focus efforts on understanding conditions local to the timber surface, including heat flux measurements, thermocouples at the timber surface, and local gas velocity.
Charring is the result of heat incidence on the timber by its flame and the flame produced by other fuels. As such, both the fuel load density and available openings have an impact on charring rate, as depicted in Figure 13, which illustrates that as the surface area of timber relative to the surface area of openings increases, the availability of ventilation and therefore oxygen decreases, leading to a predicted decrease in the maximum average charring rate. An increase in mov-

| Contribution of timber to the fire
Su et al. 18 found that the total timber contribution to HRR increases with timber surface area. Su et al. 18 also reported that one experiment with a larger opening (lower opening factor) contributed to a lower HRR. Figure 14 presents the average charring rates reported in experiments, and their HRR contribution per unit area of exposed timber.
The data indicate that in cases where the exposed timber ignites, F I G U R E 1 0 Peak mean temperatures measured across all in-air thermocouples for experiments with available data, overlaid onto two curves derived from Thomas and Heselden 49 for different crib spacings (dashed line for lower-density crib, thick line for higherdensity crib), along with a 10% confidence region for the curve implemented in the updated framework of Gorska et al. 41 there is a contribution of between 50 and 160 kWm À2 (per unit surface area of exposed timber Although HRR contribution by timber has been studied significantly in small-scale experiments, the coupled interaction between the moveable fuel load and the timber is still not fully understood and requires further investigation, specifically regarding the influence of parameters such as fuel load, incident heat flux, and ventilation. However, Kotsovinos 6,11,12 found that compared to similar traditional compartment fire experiments with non-combustible materials (Rackauskaite et al., 28,52 Heidari et al. 27 ), the addition of a timber ceiling approximately doubled the total HRR over the duration of the fire, as shown in Figure 7.
By looking at a medium-density fibreboard (MDF) ceiling, Hasemi et al. 45 found that ceiling flame spread is highly sensitive to the preheating stage, but is not significantly impacted by changes in intensity of the external fuel load; this suggests that as room temperature in a compartment increases due to an external fuel load, the flame spread across a combustible ceiling will also increase. The work described by

| CURRENT RESEARCH GAPS
The current areas of research that require further development to support firesafe mass timber compartment design can be divided into two key areas: the influence of compartment design on fire dynamics, and phenomena in compartment fire dynamics.
Key design parameters, including compartment area, opening factor, fuel load, and area of exposed timber all have an influence over compartment fire dynamics. Barber 55 identified several areas that required further research: contribution of exposed timber within a compartment to fire behaviour, especially in compartments with 50%-75% exposed timber ceiling and walls; and decay behaviour of  (1) and provided values of timber density and heat of combustion. Experiments in the CU, NFPA, and DELFT experiment series provide combustion efficiencies greater than one, suggesting variation in either the method for measuring timber HRR contribution or timber charring rate may influence calculated efficiencies exposed timber. From the experiments reviewed in Figure 8, it is recommended that fire behaviour in compartments with a greater opening factor (poorer-ventilated compartments, indicative of large openplan compartments such as office spaces where the floor area >50 m 2 ) with exposed timber requires further research and understanding.
As timber ceilings can span the entire floor area of the compartment, flame spread across compartments and over these surfaces also requires further research, particularly in conjunction with current understanding of travelling fires in open-plan compartments. 52 Wiesner et al. 22 show that timber slabs in burnout fires under some conditions can maintain sufficient load-bearing, integrity, and insulation requirements. However, in one experiment, a timber slab failed 29 h after the experiment, which was attributed to continued in-depth smouldering over this period.
More recent experiments by Kotsovinos et al. 11 Despite timber smouldering after the fire being observed in several experiments, it is still not well understood, and how it could impact the structure. Therefore, more research must be carried out to measure and understand smouldering behaviour following a fire in a mass timber compartment. This includes understanding locations in a structure in which a mass timber element is more susceptible to smouldering, including design components such as timber connections, voids, cavities, and servicing penetrations. Further to this, methods for detecting and extinguishing smouldering must be further developed to mitigate the hazard posed to timber structures.

| CONCLUSION
This paper reviews 63 mass timber compartment fire experiments car- The location of mass timber elements within a compartment was found to have a significant impact on the charring behaviour. Exposed timber ceilings were found to have charring rates on average 16% lower than exposed timber walls in the same experiment. Furthermore, charring rate is predominantly driven by ventilation conditions and movable fuel load density, with average charring rate decreasing as the proportion of timber surface area to opening surface area increases. However, the influence of key compartment design parameters on timber charring rate requires further understanding to progress the current understanding of compartment fire dynamics.
Despite the key relationships identified in experimental findings, these are limited to the relatively small range of compartment designs in the literature. The range of scales at which experiments have been carried out is currently significantly below the scale of timber compartments already constructed; therefore, it is recommended that further studies are required to understand the impact on fire dynamics of compartments greater than 100 m 2 , compartments with high areas of exposed timber, and under-ventilated compartments. Further to this, we found gaps in understanding required regarding flame spread and smouldering behaviour of timber on a compartment scale, and both may pose a significant hazard to timber compartments.
By progressing understanding in these key design considerations, design guidance and tools can be developed to appropriately inform firesafe modern mass timber construction.
F I G U R E 1 5 Localised region of smouldering in a timber ceiling slab following the CodeRed #01 compartment fire. 11 Smouldering can initiate in localised regions in timber elements during the decay period of a fire, spreading over a period of hours and days. This can continue to pose a hazard to the building structure