1. Introduction
1.1 Use of Wood and Fire Safety
Wood has global potential as a construction material for urban buildings because of its low environmental impact. For the sustainable use of wood, three aspects (environment, society, and economy) should be fulfilled. Wood must adhere to strict legal restrictions due to the fear of increased social fire risk, and individual building owners must bear additional costs for the requirements. If the sustainable use of wood is a social demand, we should address the balance between social risk and cost burden.
This paper is part of a doctoral study that deals with the socio-technical aspects of fire safety in wood construction from the architectural/urban planning viewpoint. Socio-technical refers to technology that aims to solve social problems by examining the relationship between society and technology. Since wood has been used as a building material and fuel, and fires are human-made disasters, decision-making on technical standards tends to be influenced by local social rules. A socio-technical perspective is needed to establish optimal technical standards for using wood as a sustainable material and for fire safety.
1.2 Literature Study: Regulatory
Since the 1990s, building regulations with prescriptive restrictions (prescriptive building regulations) have been gradually revised to performance-based regulations. The revision was the global trend of the times. The Inter-jurisdictional Regulatory Collaboration Committee (IRCC) consists of 12 member countries, including Japan and Austria, and provides several documents about discussions on performance-based regulations.
The IRCC document 2017 is titled "A socio-technical system framework for risk-informed performance-based building regulation."[1] The socio-technical system (STS) deals with interaction and harmonization between social and technical aspects of issues. This document gives fire as an example of the necessity of a socio-technical approach. Performance-based regulation has two approaches: risk-based (RB) and risk-informed (RI)[2]. The critical difference between the approaches is that the RB approach uses a single performance criterion, and the RI approach uses several performance criteria. The IRCC document provides examples: Eurocode for RB, and safety measures for large-scale buildings such as nuclear power plants and factories for RI[3,4].
In Japan, a performance-based code has been in force since 2000, and it takes the RB approach for wooden load-bearing elements. After a revision in 2019, the RI regulation was partially adopted, but it is limited to suburban buildings. In earthquake-prone countries like Japan, the fire resistance of building structures should be determined, considering situations where firefighting activities cannot be expected after an earthquake[5]. To use wood in urban areas with RI regulation, it is necessary to consider the complex conditions of fires after earthquakes.
1.3 Literature Study: Fire Statistics
Statistical analysis in developed countries shows a certain tendency. Most fire occurrences and human losses occur in housing. The significant causes of housing fires are related to human-made hazards (such as tobacco, Cooking devices, and heating devices). There is a difference between fires that cause much human loss and fires that cause physical damage. Depending on which consequences you focus on, measures differ. For example, there are a lot of cooking stove fires, but there are few cases of burnout or fatality. There are deaths from cigarette fires, even if they are small-scale fires, and both burnout and fatality rates are high in heating fires[6].
The fire occurrence ratio of post-earthquake fires is estimated based on the trend of fire outbreaks in ordinary fires. Fires caused by human factors, such as cigarette fires, are unlikely to co-occur during earthquakes, although the human damage in ordinary fires is significant. High-temperature equipment, such as cooking stoves and heating devices, may cause fires immediately after an earthquake. Electric wires tend to cause fires, particularly when power is restored[5].
The probability of fatality correlates with age and human activities, such as sleep[7-9]. Also, some research showed higher fire occurrence in lower-income groups and high fatality rates in socially vulnerable groups[10,11]. Thus, many research studies mention appropriate socioeconomic burdens[10,12].
1.4 Research Question: Wood and Fire in Japan
Although it is known that fire mortality rates are higher among residents of wooden buildings and elderly households, there is limited research on these groups with higher fire incidence rates[13-15].
Wood has traditionally been used extensively as a structural component in housing. As a result, the stock of wooden houses tends to be older than that of non-wooden houses. In addition, the Japanese housing policy is an owner-occupied system, with 60% of all households being owner-occupied, of which 80% are detached houses. Around half of all houses are owner-occupied detached houses; these are purchased with long-term loans, and aged people tend to live in older houses. This bias on the part of the inhabitant is one factor that makes it challenging to assess the wooden fire hazard and probability. It is also essential to consider that older houses have aging equipment and higher error rates.
Could the bias in the conditions that are related to the structural materials, rather than the type of material, increase the apparent risk?
1.5 Significance of This Research
Housing fires occur more frequently, and analyzing the details is important not only for reducing human loss but also for analyzing fire safety in cities after earthquakes.
Japan is one of the few countries with long-term fire statistics under the same building code and a significant statistical population. Japanese fire statistics (Kasai Houkoku) seem to have the largest population among census public fire statistics in the world[9]. The results of a fire statistical analysis of Japanese domestic confounding factors can be used to review our local socio-technical situation regarding fire safety in wooden housing. Similarly, results of fire statistical analysis without Japanese social confounding factors can be used for technical solutions in foreign countries without fire statistics, due to their small population, like Austria.
2. Materials and Methods
2.1 Subjects and Analysis
This paper provides a statistical analysis of housing fires, the most common cause of fires and fatalities. Both statistical samples and the universe can be defined in Japan.
Figure 1 shows the phases of fire and overall measures. This paper deals with the areas highlighted in gray as parameters for housing fires.
2.2.Parameters and Influence Relationship
2.2.1 Universe and Statistical Samples in Fire Progress
This paper uses aggregated social statistics as the universe[16] and original raw data of fire statistics[17] for statistical populations and samples.
Hazards and risks are separately discussed in order to analyze performance. In this thesis, hazard means a source of potential harm or a risk source. According to references, risk is defined as the product of hazard and frequency[4,18]. Consider the frequency with which fires (P) occur from a universe (U) and the frequency with which consequences (C) appear among the fires (P). We consider two types of damage (people or buildings) and two levels (recoverable or non-recoverable). (Table 1)
If the inclusion relation is (1) and the frequency coefficient is X, the relation shown in (2) is formed[16].
2.2.2 Socio-Technical Influence Relationships
The statistics shown in Eqs. (1) and (2) and the social/technical factors that influence fire progress are organized as shown in Figure 2. Areas highlighted in gray (calorific value) are areas where structural materials affect the damage.
Here, "technical" refers to things that can be considered in building design, such as materials and equipment within the building. "Social" refers to larger systems and given conditions (urban planning, characteristics of residents, etc.) that take longer to change than architecture.
2.3 Data Processing
The aggregated social statistics[16] and the original raw data of fire statistics[17] are processed separately. The social statistics are collected once every five years, and fire statistics are reported each year. Average fire statistics for five years were used as populations and statistical samples.
The right side of Figure 3 shows the processing procedure for the fire statistics, and the left side shows the statistics of the universe for the total number of houses. The details of the proceedings are described in sub-sections 2.3.1 -2.3.4.
2.3.1 Data Integration and Extraction
The fire statistics consist of reports about fires and fatalities. These results are provided as two separate data sets: fire damage and fatalities. The reports are made for each one fire and one fatality. One fire report is related to 0-N fatality reports. In this paper, both reports per fire are integrated into the database software (FileMaker Pro) using a unique key and compiled as 1 to 0-1 data. The ages of the injuries (three age categories) and fatalities (actual ages) are tabulated separately in two data sets. The fire-causer (P) age group is not included in the report. This study uses the integrated data set to analyze the influence of age group on fire occurrence and loss. The reports cover all types of fires, such as forest, transportation, and building fires. Housing fires are extracted for this analysis. (Figure 3(a))
2.3.2 Parameter Types
Table 2 shows the categories and classes of the parameters related to Figure 3(b). The character column gives three sorts: conditions that decide the structural performance against fire (CD), structural material and its legality (PS), and other conditions (CF). The data column gives the existing data type. U indicates that the universe (the number of households) exists with the same classification. P indicates that the population/sample (the number of fires) can be extracted from fire statistics.
Here, the population (P) is the number of households, not the number of residents. Due to the nature of fire statistics, information on the characteristics of those who started fires is limited, so their characteristics are inferred from the universe (U) statistics about households.
2.3.3 Outlier Verification
Fire statistics include quantitative (numerical) and qualitative (ranks, categories) data, and this paper mainly uses qualitative data. (Figure 3(c)) The quantitative data seems to contain significant outliers due to erroneous entries of about 7%. The discrepancies in the numerical values of the quantitative data are checked for errors in the system. However, there are no error checks for upper limits regarding the total floor area of the building, distance to the fire station, and time for reporting and extinguishing fires. The distance to the fire station is difficult to process because it contains anomalous values with different digits[19]. The qualitative data may include the same degree of entry errors among ranks or categories. No correction was made in this paper as the influence is lower than the errors in quantitative data.
It is announced that aggregated universe statistics contain a 5-10% margin of error.
2.3.4 Category
The categories in Table 2 on the fire statistics (P) are classified the same way as those matching the housing statistics (U). (Figure 3(d))
Fire types in Table 3 are categorized by origin and factors of hazards. Origin identifies the place of fire, and factors are classified by human behavior. This paper analyzes the number of mixed and physical factors among internal fires (Table 3). Regarding origin, internal fires are defined as fires in which the point of ignition, the source of ignition, and the ignition material are all located indoors. Fires in stairs, halls, and corridors are excluded to make the conditions between detached and apartment houses the same. Among the three internal fires, mixed and physical fires are extracted as samples.
This paper mainly discusses fire factors related to building performance and equipment as hazards of urban fires after earthquakes. Mixed fires are groups of high-temperature equipment, such as cooking stoves and heating devices, that may cause fire immediately after an earthquake. Physical fires, such as electric lines, tend to cause fire, particularly when electric power is restored.
Fires caused by human factors, such as cigarette fires, are unlikely to co-occur with earthquakes, although the human damage in ordinary fires is significant[5,20]. Human-factor fires, such as tobacco fires, were excluded because they occur independently of the building's performance and should be discussed separately. For example, the fatality of tobacco fire is correlated to the age group and smoking rate[21].
3. Results
We analyzed five years and five pairs (25 years) of statistics about housing fires with four conditions (us, ag, st, zn), two performances (mt, lg), and three incidental conditions (sh, sf, pt) given in Table 2. Three coefficients (Xu_p, Xp_c1, and Xc1_c2 in Figure 2) between 4 statistical data (U, P, C1, and C2 in Table 1) are calculated with the five-year average of the fire statistics (P) and aggregated every five years (U). The overview of the results of coefficients (X) is summarized in Figure 4 and Figure 5. Further details about all the factors are also presented in Table 4 to Table 6.
The analysis shows that fire occurrences and damage are decreasing (See section 3.1). If we compare the rate of fire occurrences (See section 3.2) and after fire ignition (See section 3.3), the former is more influential on fire damage reduction.
3.1 Overview of Statistical Tendency for 25 years
Figure 4 shows the change in coefficient between the averages of 1996-2000 and 2016-2020. Figure 5 shows the coefficient ratio when the average of 1996-2000 is set as 1. The most considerable change in coefficients over 25 years for (X) is in the fire rate (Xu_p). The fire occurrence rate mainly influences the reduction of fire damage and human loss.
3.2 Probability of Fire Occurrence
Focusing on the fire occurrence rate breakdown, the conditions determining the structural performance (CD) are more significant than the structural material (mt) regarding the difference between maximum and minimum.
The highest fire occurrence rate is in 1-story (1s) houses, and the second is in Zone 3 (zn3) houses. According to (U) statistics, 1-story houses and Zone 3 houses tend to have older inhabitants and older construction years. If we compare the fire rate of wood and non-wood structures, the former is slightly higher, and the age groups of the inhabitants are similar. The older construction year is considered a factor in the high occurrence rate.
In terms of factors for which the universe is unknown, the number of fires in the legality (u) and (nr) groups is high. The reason is thought to be influenced by the older construction year, and the number of (u) will be estimated in Chapter 4.
3.3 Conditions After Fire Occurrence / Ignition
Trends in damage per fire have remained unchanged over the past 25 years. The rate of burnout per fire has decreased slightly, and the rate of fatalities has increased somewhat. There has been little change over time compared to the fire occurrence rate.
The results of self-firefighting (sf) significantly influence coefficients about physical damage (Xp_d1) and human loss (Xp_l1 and Xl1_l2). 1-story shows the highest rate of non-recoverable physical damage (Xd1_d2). Also, the fire occurrence rate seems to be related to age group and construction year.
4. Discussions
According to the results in Chapter 3, the reduction of fire occurrences and self-firefighting is significant in reducing the consequences of fires. In this chapter, details of fire occurrences and risk groups are discussed.
4.1 Fire Occurrence
The number of fires in Japan has decreased over the past 25 years. In particular, the decrease between the data sets of 2006-2010 and 2011-2015 has been significant, while the reduction in physical damage (burnout rate) has been limited. Additional consideration has been given in this section.
4.1.1 Analysis by Japanese Ministries and Agencies
Over the years, there has been an apparent decrease, especially since 2008, and the reason for this is the influence of the operations of the two ministries[22]. In October 2008, the Ministry of Economy, Trade and Industry (METI) made safety devices for gas appliances mandatory. In April, as part of the industry's self-regulation prior to the mandate, cooking stoves equipped with safety devices were sold. Before the mandate, the METI estimated a 90% fire reduction. The estimation is based on the fact that the number of fires decreased due to installing a device to prevent gas water heaters from burning without water, and that no fires broke out from stoves equipped with the device[23]. The Ministry of Internal Affairs and Communications made installing automatic household fire alarms mandatory in newly built houses in 2004 and in existing houses in 2010. It has been reported that the number of casualties decreased in fires where alarms were activated[22].
4.1.2 The Number of Fires
Referring to reports from the two ministries, examine fire occurrence trends by cause for each year. The decrease in cooking stove fires is considerable. (Figure 6)
In 2008, household cooking stoves, excluding stand-alone types, were subject to the regulation. Among household cooking stoves, built-in types with safety devices have been popular in the market since the 1990s. Classifying cooking stove fires as built-in (with safety devices), plug-in (subject to regulations), or independent (not subject to regulations), and analyzing the trend. The proportion of fires caused by plug-in types is large (Figure 7). Regarding increasing or decreasing fires (Figure 8), the rate of decrease is more significant with plug-in types, while the other two types are slightly increasing. Looking at the damage per fire caused by plug-in cooking stoves, the burnout rate is gradually decreasing, and the rate of fatalities is somewhat increasing. (Figure 9)
4.1.3 Firefighting
When we look at the breakdown of the effective measures when the initial firefighting was successful, about 60% was done by human power and 30% by equipment. As for the operation status of fire alarms, there is no data for around 80% of fires. Of the remaining 20%, half are installed, and half are not installed. About 2% of fires (P) have alarms installed and operating, and the number of deaths and total fires per fire tends to be less than the average. In fire incident statistics, if a residence alarm is installed in a place other than the fire room, it is considered non-installed[19]. A report shows that the diffusion rate is over 90%[24], and the installation location's appropriateness will be considered a future issue.
4.2 Estimations about Risk Groups
The age of inhabitants and the construction year influence fire risk. In this section, a more influential factor will be estimated. The assumption here is that fire occurrence is proportional only to the age of the inhabitants and the error rate of building equipment. Income class and type of residential use (owned/rental) can impact the renewal of facilities. Here, the kind of use is limited to detached houses, and income groups in wood and non-wood are treated as equivalent.
4.2.1 Construction Year
Since the construction year of housing is not collected in fire statistics, it is estimated.
(1) Estimation by Circumstantial Evidence
Based on the results in Table 4, the number of fires in unfit (u) detached houses is 2.5 times more than in lawful (lg) detached houses. These unfit houses are classified into two groups: (a) illegal construction or (b) disqualification under current regulations.
The proportion of (a) illegal construction is considered to be negligible. This disqualification (b) is caused by two reasons: (b1) construction before establishing the current regulation, or (b2) change of zone definition. Housing in zones 1-2 constructed before 1950 will be classified as (b1), and housing built between 1950 and 1990 can be classified as (b2). The regulation of the zone was established in 1968. The statistics indicate that the zone area doubled in size between 1968 and 1990. (u) is considered to be houses built up to 1950, 1968, or 1990.
(2) Estimation by Proportion of Universe
Estimates are made based on (U) statistics on the construction year and the inhabitants' age. (U) and (P) are used for detached wood-frame and non-wood-frame houses, which have similar age structures and many residents, by dividing the construction year into three categories (before 1970, between 1971 and 1990, and after 1991).
4.2.2 Age Group
The relationship between the injuries and fatalities in a single fire has not been analyzed because they are inscribed in different data sets. (See 2.3.1.) This study integrated two data sets using database software (FM) to be treated as a single data set, and the relationships among the universe (U), the number of injuries (L1), and fatality (L2) were analyzed. As a result, it was found that the mortality rate of the injured was characteristic of each age group. (See Table 6.) The results show that the ratio of L2/L1 (fatality per injury) death rate for each age group is almost constant regardless of the data period (around 5 times higher in the elderly group than in the young group). The age-specific fire occurrence ratio is estimated based on this ratio.
(1) Estimation by Circumstantial Evidence
We infer age-group fire occurrence rates from their ranking in 2016-2020 (Xp_l2 calculated as Xp_l1 by Xl1_l2 in Table 6): age group under 64 (Xp_l2_64) is between apartment (Xp_l2_ah) and 3-story (Xp_l2_nw), and 65 and over (Xp_l2_65) is between wood (Xp_l2_w) and zone 3 (Xp_l2_zn3). Assuming the same ranking of fire rates, the probability of the age group under 64 is between 0.077 - 0.082 (See formula 11), and 65 and over is between 0.11 - 0.132(See formula 12). We expect the difference in fire rates in the two age groups to be about 1.3-1.7 times.
(2) Estimation by Proportion of Universe
The bias in the construction year included in houses as the universe (U) could not be eliminated and was not calculated because it is challenging to determine. In the future, an analysis will be conducted based on the original raw data for population statistics.
4.2.3 Discussions on Coefficients Larger or Smaller
The number of fires caused by the year of construction is significantly higher for the aged group (U65) and old houses (U70). The age of the inhabitants or the construction year affects the rate of fires. This section discusses which is more significant.
The number of fires with fatalities in lawful (q) and unfit (u) houses was divided into 2x2 groups for age and legality. A chi-square test shows no significant difference between (q) and (u) age ratios. On the other hand, the number of fires that started in (u) is more than twice that in (q). From this, it can be inferred that the construction year affects the fire occurrence rate more.
According to the inference in 4.1.1, the unfit group will likely consist of houses built before 1970. The proportion of the number of houses in this period was less than 10% of the total.
5. Conclusions
5.1 Current Situation
Regarding fires in wood houses, occurrence, progress, and damage/loss were analyzed socio-technically using 25 years of Japanese fire and housing statistics.
Fire damage and loss reduction were mainly correlated with the fire occurrence. Fire occurrence rates were higher with social factors (age of inhabitants) and technical factors (construction year). The latter coefficient seems more significant. Concerning fire progress, social factors (self-firefighting by users) are essential for fire spread and injury rates after a fire starts. Technological factors (structural materials) affect non-recoverable physical damage.
Looking at fire occurrence probability by group, the maximum coefficients are larger for the conditions determining materials (us, ag, st, and zn) than for structural materials (mt). It is thought that the apparent risk of wood construction is related not only to the combustible structural material but also to the conditions under which the material is used.
The group with aged householders (65 or more) and the old construction year (before 1970) shows higher occurrence rates. The latter is more significant. The results showed that fire damage/loss reduction mainly correlates with the fire occurrence rate.
5.2 Discussions in The Future
The Japanese society is aging. Accordingly, the average construction year of housing is older, and the fire rate is expected to increase. Under the current Japanese building standards, fire safety focuses on the fire resistance performance of newly built buildings. As an investment for society, renovating existing houses and updating facilities is essential rather than imposing strict technical requirements on the new houses. In conjunction with this, it may be necessary to develop a social system in which the refurbishment of existing detached houses, such as housing liquidity, is valuable.
As a future perspective of this research, we would like to conduct a detailed analysis of the social issue (characteristics of inhabitants) and technical issue (error rates associated with the construction year) to integrate this analysis with design theory. "Open Building System (OBS)" is a design theory that deals with this relationship and the results of architecture and urban planning.
In a broad sense, OBS deals with dominance and influence among elements of the built environment at different scales, such as urban structure, building, and interior. In a narrow sense, the theory deals with the systematic classification of building elements based on their lifetime and flexibility for new construction, future renovation, and restoration[25,26].
This socio-technical study of fire rates will contribute to the global demand for wood with fire safety.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
