ArticlePDF Available

The role of large language models (AI chatbots) in fire engineering: An examination of technical questions against domain knowledge

Authors:

Abstract and Figures

This communication presents a short review of chatbot technology and preliminary findings from comparing two recent chatbots, OpenAI’s ChatGPT and Google’s Bard, in the context of fire engineering by evaluating their responses in handling fire safety-related queries. A diverse range of fire engineering questions and scenarios were created and examined, including structural fire design, fire prevention strategies, evacuation, building code compliance, and fire suppression systems. The results reveal some key differences in the performance of the chatbots, with ChatGPT demonstrating a relatively superior performance of 88% (vs. 80% for Bard). Then, this communication highlights the potential for chatbot technology to revolutionize fire engineering practices by providing instant access to critical information while outlining areas for further improvement and research. Evidently, and when it matures, this technology will likely be elemental to our engineers’ practice and education.
Content may be subject to copyright.
The role of large language models (AI chatbots) in re engineering: An
examination of technical questions against domain knowledge
Haley Hostetter
a
, M.Z. Naser
a
,
b
,
*
, Xinyan Huang
c
, John Gales
d
a
School of Civil and Environmental Engineering and Earth Sciences, Clemson University, USA
b
Articial Intelligence Research Institute for Science and Engineering (AIRISE), Clemson University, USA
c
Department of Building Environment and Energy Engineering, The Hong Kong Polytechnic University, Hong Kong
d
Department of Civil Engineering, York University, Toronto, Canada
ARTICLE INFO
Keywords:
Chatbots
Articial intelligence (AI)
Fire engineering
Large language models (LLMs)
ABSTRACT
This communication presents a short review of chatbot technology and preliminary ndings from comparing two
recent chatbots, OpenAIsChatGPT and GooglesBard, in the context of re engineering by evaluating their re-
sponses in handling re safety-related queries. A diverse range of re engineering questions and scenarios were
created and examined, including structural re design, re prevention strategies, evacuation, building code
compliance, and re suppression systems. The results reveal some key differences in the performance of the
chatbots, with ChatGPT demonstrating a relatively superior performance of 88% (vs. 80% for Bard). Then, this
communication highlights the potential for chatbot technology to revolutionize re engineering practices by
providing instant access to critical information while outlining areas for further improvement and research.
Evidently, and when it matures, this technology will likely be elemental to our engineerspractice and education.
1. An introduction to chatbots
A chatbot is a computer program capable of simulating conversation
with humans (Adamopoulou and Moussiades, 2020a). Such a program
(or platform) is often built via articial intelligence (AI) and natural
language processing (NLP) with the goal of understanding human ques-
tions and automating customized responses. More specically, a chatbot
is made from large language models (LLMs), a subset of AI trained via
machine learning to learn the patterns and connections between words
and phrases. This refers to the ability of the chatbot to uniquely respond
to human-posed questions almost instantaneously, thereby simulating
human conversation.
Perhaps the rst idea for computer-based chatbots arose from the
question, can a computer communicate in a way indistinguishable from
human?posed by one of the computer science pioneers, Alan Turing, in
1950 (Zemcik, 2019). In the decades following, several chatbots were
developed, including Eliza (1964-6), PARRY (1972), Racter (1983), and
Dr. Sabaitso (1991) (Zemcik, 2019). Eliza, one of the oldest chatbots, was
created by the Articial Intelligence Laboratory at MIT and programmed
to behave like a doctorsimulating the role of a psychotherapist who
asks and responds to questions to divert attention back to the user
(Natale, 2019). Eliza quickly garnered the attention of users (who began
to provide their personal stories, secrets, and sensitive information) and
developers (who used Eliza as inspiration for later AI-powered chatbots).
PARRY, on the other hand, was programmed to behave as a paranoid
schizophrenic patient(McNeal and Newyear, 2013). It attempted to
provoke a user with questions that required more elaborate responses,
thereby providing learning psychiatrists with a tool to help them
communicate with schizophrenic patients. Later, Racter and Dr. Sabaitso
increased the realism of chatbot technology. According to developer
William Chamberlain, Racters ability to direct a computer to maintain
certain randomly chosen variables, such as words and phrases, helps the
programs answers pass for coherent thinking similar to humans (Zemcik,
2019). Then, the chatbot Dr. Sabaitso (an acronym for Sound Blaster
Articial Intelligent Text to Speech Operator) revolutionized the eld as
the rst chatbot able to synthesize speech similar to a psychologist
(Adamopoulou and Moussiades, 2020b). Although it could not commu-
nicate in a sufciently complex way, verbal communication increased the
humanity of the chatbot.
Apart from the aforementioned chatbots, the development and use of
new AI chatbots increased in popularity with the expansion of the
Internet and social media platforms in the 1990s. For example, ALICE,
one of the rst chatbots to go online, was introduced in 1995 (Zemcik,
2019). An acronym meaning Articial Linguistic Internet Computer
* Corresponding author. School of Civil and Environmental Engineering and Earth Sciences, Clemson University, USA.
E-mail addresses: hhostet@g.clemson.edu (H. Hostetter), mznaser@clemson.edu (M.Z. Naser), xy.huang@polyu.edu.hk (X. Huang), jgales@yorku.ca (J. Gales).
Contents lists available at ScienceDirect
Natural Hazards Research
journal homepage: www.keaipublishing.com/en/journals/natural-hazards-research
https://doi.org/10.1016/j.nhres.2024.06.003
Received 3 March 2024; Received in revised form 20 May 2024; Accepted 10 June 2024
2666-5921/©2024 National Institute of Natural Hazards, Ministry of Emergency Management of China. Publishing services provided by Elsevier B.V. on behalf of
KeAi Communications Co. Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Natural Hazards Research xxx (xxxx) xxx
Please cite this article as: Hostetter, H. et al., The role of large language models (AI chatbots) in re engineering: An examination of technical
questions against domain knowledge, Natural Hazards Research, https://doi.org/10.1016/j.nhres.2024.06.003
Entity,ALICE, though over twenty years old, remains useable today and
is often credited as paving the way for newer chatbots (McNeal and
Newyear, 2013).
More recently, chatbots have once again captured the interest of the
general public and industries alike. This is primarily due to chatbots
ability to integrate into businesses, thus saving companies time and re-
sources. Current chatbots, however, can have a variety of purposes and
applications. For example, like the rst chatbots, recent chatbots by
Disney and Marvel are meant for entertainment purposes, allowing users
to converse with their favorite movie characters (Loleng;Marvel). Still
others attempt to solve real-world problems or assist in the healthcare
eld (see the chatbot developed by the company Endurance, a companion
for dementia and Alzheimers patients (Fomitchev, 2017) or the chatbot
developed by the company Casper, a conversational chatbot for people
with insomnia (Insomnobot-3000)).
With the broadening range of chatbot abilities and uses, many have
questioned the possibility of using such AI for more scientic or practical
applications or research. However, few papers have yet to address this,
and available literature seems to focus on the education front. For
example, Perez et al. (Fryer et al., 2017)nd using chatbots for education
helpfulparticularly for students with disabilities. They also note the
possibility of bridging the educational gapbetween marginalized and
mainstream groups. Similarly, Fryer et al. (2017) designed an experiment
to monitor students' learning behavior when their education was sup-
plemented with a chatbot. Results found that chatbots can have a positive
inuence on student performance.
In more technical elds, such as medicine and engineering, extensive
research on the applicability of chatbots is limited. For medicine, current
research suggests promise for the application of chatbots, especially for
automated and repetitive tasks (Wilson and Marasoiu, 2022). Addition-
ally, concerns about the legal and clinical aspects of using chatbots are
common. In engineering, a 2022 study attempted to explore the imple-
mentation of chatbots into engineering education. Results show the
possibility of using chatbot technology to assist in cooperation between
students as well as the potential for their use as tools to support engi-
neering design practice (Chien and Yao, 2022). Prior to 2020, another
study aimed to develop a chatbot capable of participating in the early
stages of the engineering design process. Following interaction with
several college students, the AI chatbot proved effective (Chien and Yao,
2020).
In the more niche eld of re engineering, experts seem to disagree
with the aforementioned belief that chatbots are appropriate tools for
education and design. One of the few available publications states their
applicability in summarizing concepts and documents for non-experts,
but that they cannot be used to replace or replicate re engineers since
such experts must ensure the safety of structures as well as conduct re
safety assessments, recommend design changes, and select re protection
systems (Spearpoint et al., 2023).
With the present divided opinions on the practical use of chatbots and
their limited applicable literature in the engineering eld, the authors are
motivated to address chatbot responses and the possibility of their
implementation into re engineering and evacuation (including design
and research). This is completed using the Ask Me Anything approach for
two of the leading publicly available chatbots: OpenAIs ChatGPT,
1
,
2
and
Googles Bard. These two chatbots were released in November 2022 and
February 2023 respectively, and our analysis examines the early release
versions of these chatbots. This paper aims to explore the current capa-
bilities of leading chatbots in a technical eld and build upon the limited
available research. With the fast-evolving elds of AI and machine
learning, the authors comment on the current abilities of chatbots in
comparison to expert knowledge.
2. Methodology
This communication presents a comparison of two of the premier LLM
chatbots (OpenAIs ChatGPT and Googles Bard) in the context of re
engineering by explicitly evaluating their accuracy in handling re
safety-related queries when displayed in the form of Ask Me Anything
(AMA). A diverse range of re engineering questions and scenarios were
created, including structural re design, re prevention strategies,
evacuation, building code compliance, and re suppression systems
(some of which resemble those commonly present in the Fire Protection
exam (FPE)).
3
The authors choose to qualitatively evaluate chatbot re-
sponses based on domain knowledge, published literature, and accept-
able answers to possible questions that may resemble the FPE exam.
2.1. Chatbot selection and denitions
In late 2022, OpenAI released an early version of ChatGPT, a natural
language processing tool driven by AI technology. By January 2023,
ChatGPT had become the fastest-growing consumer software application
in history, gaining over 100 million users in just two months (Hu, 2023).
To compete with ChatGPT, Google released its own chatbot, Bard, in
March 2023. Both chatbots are generative AI platforms.
When these two chatbots were asked to dene what a chatbot is by
these authors, they returned the following denitions:
According to ChatGPT,
A chatbot is an articial intelligence (AI) software program designed
to simulate human-like conversations with users, usually through text
or voice interactions. Chatbots can understand and process natural
language inputs, such as text messages or spoken words, and generate
appropriate responses based on their underlying algorithms and
knowledge base. They are typically used to provide customer support,
answer frequently asked questions, assist with various tasks, or offer
personalized recommendations.
And according to Bard,
A chatbot is a computer program that simulates human conversation.
It can be used to answer questions, provide information, or perform
tasks. Chatbots are often used in customer service applications, but
they can also be used for a variety of other purposes.
In early 2023, ChatGPT and Bard were some of the only publicly
available LLMs. With their rise in popularity worldwide, it is of interest to
understand their capabilities and limitations in a more technical setting.
Thus, we have selected both LLMs to evaluate their understanding of re
engineering.
2.2. Question selection and answer evaluation
Chatbots such as Bard and ChatGPT are intended to be open-ended
devices that users can access to ask any general or more detailed
1
In this paper, ChatGPT refers to ChatGPT-4, released in March 2023. For
transparency, some of the answers might look different by the time this publi-
cation is released. This is due to the dynamic nature of these chatbots, as well as
continued learning.
2
One should note that, at the time of this papers submission, ChatGPT did
not have access to real-time online data (i.e., had a cutoff date to September
2021), unlike Bard. Microsoft Corporation launched an experimental model in
March 2023 that used the Bing search engine to browse the Internet for more
accurate data. This model was available only to paid subscribers of ChatGPT
Plus (ChatGPT Will Now Have Access).
3
Recently, chatbots have been evaluated for domain-specic knowledge by
comparing their answers to acceptable responses on standardized and profes-
sional licensing exams. Examples include [(Huang et al., 2023), (Passby et al.,
2023)]. Thus, we apply this concept to the elds of re safety and re protection
engineering by evaluating ChatGPT and Bards performanceswhen asked
questions from the FPE exam.
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
2
question they wish. As such, the platforms lack a standardway in which
to ask questions, formulate the structure of such questions, or determine
the words to use in the queries. Thus, the questions posed to each chatbot
in this communication were intended to mimic/replicate the questions a
non-expert or student may ask about re engineering or evacuation. We
additionally ask questions that may arise in the Fire Protection exam for
re engineers and professionals. Thus, answers are evaluated based on
domain knowledge, acceptable responses during the Fire Protection
exam, and denitions and understanding provided in the open literature.
Additionally, we aim to understand the present knowledge of such
chatbots by posing general questions. We then evaluate the chatbots' re-
sponses in terms of their understanding of minute differences that have
been adopted in the practicing engineering eld and academia.
2.3. General chatbot architecture
The architecture of a chatbot includes its underlying structure and
design. It denes how a chatbot processes text and depends on various
factors such as domain, use-case, chatbot type, and many more. Thus, the
key components of the chatbot architecture can be different for different
bots. Currently, the most common types include ruled-based, retrieval-
based, generative, and hybrid (Naser, 2023).
Rule-based,orscripted, chatbots are the earliest form of chatbots and
are developed based on predened rules. They follow such rules to
generate responses using a series of conditional statements that check for
keywords or phrases in a users input. At the basic level, rule-based ar-
chitecture includes three parts: a user interface (UI), Natural Language
Processing (NLP) engine, and the rule engine (Hore, 2023). The UI is the
platform through which the user asks questions and views chatbot re-
sponses. The NLP engine works behind the UI to process the users input
and convert it to a readable format for the machine. Finally, the rule
engine is responsible for interpreting the users input, processing the
input through the conditional rules of the chatbot, and returning the
answer. Rule-based chatbots can be useful in certain scenarios but suffer
from several limitations. These include a limited ability to understand
natural language, a lack of contextual understanding, difculty handling
ambiguity and subtle nuances, and an inability to learn and adapt over
time (Hore, 2023). Additionally, rule-based chatbots are difcult to scale
up and improve since new programming is required to update rules and
patterns within the architecture.
Next, retrieval-based chatbots work by pulling chatbot responses from
an established corpus of dialogs (What are Chatbots, 2023). Such chat-
bots use machine learning models like supervised neural networks to
interpret user inputs and determine tting answers (What are Chatbots,
2023). They rely on predened responses (like rule-based chatbots) but
can self-learn and improve their responses over time. Thus, these chat-
bots benet from a greater ability to scale, adapt to changing user inputs
and interactions, and require less maintenance throughout their lifetime.
Additionally, given their predened rules and responses, retrieval-based
chatbots guarantee quality answers without grammatical errors. How-
ever, they suffer from many of the other limitations of rule-based chat-
bots. They lack contextual and natural language understanding and have
difculty understanding nuanced inputs.
Third, generative chatbots can formulate their own responses based on
user input rather than relying on predened rules or existing responses
(What are Chatbots, 2023). This requires using machine learning models
like neural networks and large datasets to train the chatbots to make
decisions about appropriate responses (Bragg, 2023). Generative models
are powerful and exible tools that benet from an unconned set of
responses. However, their requirement for machine learning and large
datasets make them challenging to implement. Additionally, how they
make decisions and determine an appropriate answer for user input can
be unclear. This makes them more prone to grammatical errors and
incorrect replies.
Finally, hybrid chatbots work by combining aspects of rule-based,
retrieval-based, and generative chatbots (Bragg, 2023). They use a
combination of pre-dened rules, pre-dened responses, and machine
learning (neural networks) to deliver the best responses to user inputs. As
a hybrid, they benet from the same advantages of each of their com-
ponents. They are scalable and have improved the quality of answers
over generative models. However, they suffer from a limited ability to
understand natural language and rely heavily on accurate data in their
training sets (Bragg, 2023).
Rule-based, retrieval-based, generative, and hybrid models each use NLP
engines to process user input and convert it to machine-readable forms.
However, as mentioned in the limitations of such models, NLP engines
commonly lack the ability to understand natural human language. As a
result, LLMs were developed with the intent to improve AI understanding
of natural language and generate believable human text (Hitter, 2023).
LLMs are a specic application of NLP and are created through a trans-
former modela type of neural network (deep learning) architecture and
form of generative AI. They work primarily through their transformer
architecture and large training datasets and have been previously limited
to use by large technology companies, who use them internally or on a
limited basis (Hitter, 2023). However, LLMs are becoming increasingly
available to the public and include models such as GPT created by
OpenAI and BERT, LaMDA, and PaLM created by Google (Hitter, 2023).
ChatGPT works under the GPT, or Generative Pre-trained Trans-
former, language model, while Bard works under the BERT, or Bidirec-
tional Encoder Representations from Transformers, language model
(Hitter, 2023). As their names imply, both GPT and BERT are
transformer-based LLMs but work in different ways. GPT is an autore-
gressive LLM, meaning it uses past textual data and previous text inputs
to determine the most appropriate next word or phrase to add to a
sequence. GPT is built on a transformer decoder that allows individual
outputs to be shared based on previously decoded outputs (Hitter, 2023).
On the other hand, BERT is a collection of bidirectional language models
from Google that have high levels of natural language and contextual
understanding. It is built on a transformer encoder, so it generates and
shares all of its outputs at once (Hitter, 2023). The difference in trans-
formers generally means that GPT models are better at creating new
human-like text, while BERT models are better at classication and
summary tasks.
To be accurate, LLMs must be trained on large amounts of data.
ChatGPT was trained using Reinforcement Learning from Human Feedback
(RLHF) (Somoye, 2023). The model rst went through a process of su-
pervised ne-tuning, during which OpenAI trainers acted as both the
human user and the AI bot. During this process, the trainers mimicked the
way that humans communicate by creating a conversational sequence.
Then, the dialogue was added to the models dataset to improve it for
conversational use. Later, the chatbot was improved using reinforced
learning (i.e., rewarding the bot for generating correct responses and
grading them from best to worst) (Somoye, 2023). Finally, OpenAI
ne-tuned the model using its own Proximal Policy Optimization tech-
nique (Somoye, 2023). Bards training process included using unsuper-
vised learning techniques in which a large amount of unlabeled data was
input into the model (Oppermann, 2023). Then, the model processed the
text to nd statistical patterns and relationships, allowing Bard to
generate appropriate responses for a wide variety of user inputs. Bard
was also ne-tuned using supervised learning (Oppermann, 2023). The
size of ChatGPTs training dataset was 175 billion parameters (for GPT-3)
from ve different sources, while Bards was 1.56 trillion (Walsh, 2023;
Gautam, 2023).
3. Ask Me Anything (AMA): the series of questions and answers
In this analysis, each chatbot was asked the same question and their
answers were compared. We provide our commentary, as taken from
leading re-based organizations and literature, to most of the provided
questions to assess the responses received from each chatbot. We rst
start with two historical questions; then, we dive into more technical
questions. The presented questions stem from pure professional curiosity
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
3
and tend to favor research and educational components. Toward the end
of this analysis, we present ve questions aimed at exploring the
knowledge span, limitations, and possible imagination of ChatGPT and
Bard. Overall, we present the outcome of 27 questions and answers.
3.1. Question no. 1: What is re engineering?
According to the Society of Fire Protection Engineers (SFPE) (SFPE,
2023), re protection engineering is dened as:
Fire protection engineering is the application of science and engi-
neering principles. The principles when working with re protection
relate to protecting both people and the environment, built and nat-
ural, from the hazards of re. This includes analyzing re hazards,
mitigating re through design and construction of buildings, exami-
nation of building uses and industrial processes, and design, instal-
lation and maintenance of re detection and suppression systems.
Additionally, re safety engineering is dened generally according to
BS 7974 (Christian, 2003) as:
The application of scientic and engineering principles to the pro-
tection of people, property and the environment from re.
Each of the chatbots answers is listed herein:
OpenAI ChatGPT,
Fire engineering is a multidisciplinary eld that focuses on under-
standing re behavior, prevention, protection, and suppression
within the built environment. It involves the application of scientic
and engineering principles to assess, design, and implement re
safety strategies in buildings, infrastructure, and industrial facilities.
Fire engineering aims to minimize the risk of re, protect lives,
property, and the environment, and ensure the continuity of opera-
tions in the event of a re. It encompasses various aspects of re
safety, including:
Fire prevention: Developing strategies to reduce the likelihood of
res, such as proper design, material selection, and adherence to
building codes and regulations.
Fire protection: Designing and implementing passive re protection
systems, such as re-resistant construction materials, compartmen-
tation, re doors, and re-resistant glazing, to prevent re spread and
maintain structural integrity during a re.
Fire detection and suppression: Designing and integrating active re
protection systems, such as re alarms, sprinkler systems, and
gaseous re suppression systems, to detect and control res at an
early stage.
Smoke control and management: Designing ventilation systems and
smoke control measures to manage smoke spread.
Egress and evacuation: Designing and planning safe and efcient
means of escape, such as exit routes, emergency lighting, signage, and
occupant load calculations to ensure that occupants can evacuate a
building quickly and safely during a re or other emergency.
Human behavior in re: Understanding how people react in re sit-
uations, including factors like perception, decision-making, and
movement patterns, which play a crucial role in developing effective
evacuation strategies and re safety systems.
Fire risk assessment: Identifying potential re hazards, evaluating
their likelihood and consequences, and recommending appropriate
mitigation measures to minimize risk.
Performance-based design: Developing re safety solutions based on
performance objectives rather than prescriptive requirements, which
allows for greater exibility and innovation in design while still
meeting re safety goals.
Fire investigation and analysis: Investigating the causes, origin, and
development of res, as well as analyzing their effects on structures,
materials, and occupants, to improve re safety knowledge and
inform future designs.
Codes, standards, and regulations: Developing, interpreting, and
applying re safety codes, standards, and regulations that govern the
design, construction, and operation of buildings and facilities
Education and training: Educating and training professionals, such as
architects, engineers, and building ofcials, in re safety principles
and practices
Fireghting tactics and strategy: Collaborating with re service
personnel to develop effective reghting tactics, strategies, and
incident command systems, taking into account the unique challenges
posed by different types of structures and re scenarios.
Interdisciplinary collaboration: Fire engineering often requires
collaboration with other disciplines, such as architecture, structural
engineering, mechanical engineering, electrical engineering, and
urban planning, to develop comprehensive and effective re safety
solutions for various types of buildings and environments.
Fire protection economics: Analyzing the costs and benets of various
re protection strategies and systems to optimize resources and
achieve the desired level of re safety while balancing nancial
constraints.
Google Bard
Fire engineering is the application of scientic and engineering
principles to protect people, property, and their environments from
the harmful and destructive effects of re and smoke. It is a multi-
disciplinary eld that encompasses the study of re dynamics, re
suppression, re protection systems, re safety engineering, and re
investigation.
Fire engineers use their knowledge of re science to design and
implement re safety measures in buildings, industrial facilities, and
other structures. They also develop and test new re suppression
technologies. Fire engineers are also involved in re investigation,
which is the process of determining the cause and origin of res.
Fire engineering is a critical part of the re service. Fire engineers
work with reghters and other rst responders to ensure that they
have the tools and resources they need to protect the public from re.
Here are some of the tasks of a re engineer:
Design and implement re safety measures in buildings, industrial
facilities, and other structures.
Develop and test new re suppression technologies.
Conduct re investigations.
Work with reghters and other rst responders to ensure that they
have the tools and resources they need to protect the public from re.
Question 1 was intended to test the chatbotsunderstanding of min-
ute differences in the denitions of re protection and safety engineer-
ing. Thus, the question was simply posed as re engineering. Generally,
re protection engineering refers to the implementation of prevention
measures such as suppression and detection devices (sprinklers and
alarms, respectively). On the other hand, re safety engineering refers to
the engineering design of a building or structure to prevent re spread
and/or loss of structural integrity. Neither Bard nor ChatGPT were able to
communicate the difference between the two denitions. Instead, they
combined the two. Given the general nature of the question posed to the
chatbots, their answers seem acceptable. While ChatGPT provides a more
detailed response, Bards response is efcient. A re engineer is a
multidisciplinary engineer involved in a variety of re-related scientic
items (as detailed by ChatGPT). It would perhaps be helpful to
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
4
understand, however, that re engineers do not provide all the tasks
detailed in the chatbotsresponses in their specialized roles at one time.
Instead, they may be involved with one or more engineers depending on
their line of work.
3.2. Question no. 2: Who was the world's rst re engineer?
While the rst re engineer is hard to trace, the early account of
reghting is said to have begun in ancient Rome (Dillon and Garland,
2013).
In terms of the examined chatbots,
OpenAI ChatGPT,
It is challenging to pinpoint the world's rst re engineer, as the
history of re engineering spans across centuries and various cultures,
with many individuals contributing to the development of re safety
practices and technologies over time.
One of the early pioneers in the eld of re safety was Benjamin
Franklin, who founded the Union Fire Company in Philadelphia in
1736, the rst volunteer re company in America. While Franklin was
not a re engineer in the modern sense, his contributions to re safety
and reghting were crucial in laying the groundwork for the eld.
The modern concept of re engineering began to take shape in the
late 19th and early 20th centuries, as the industrial revolution and
rapid urbanization led to new challenges in re safety. Henry Grattan
(H.G.) Tyrrell was an inuential gure during this period. An Amer-
ican civil engineer, he published the book "Fire Engineering" in 1919,
which was one of the rst comprehensive texts on the subject.
It is important to note that re engineering has evolved through the
collective efforts of many individuals, each contributing to the
knowledge and practices we have today. While it is difcult to
identify a single "rst re engineer," these early pioneers played a
vital role in shaping the eld.
Google Bard
The world's rst re engineer was James Braidwood, a Scottish re-
ghter who was the rst "Master of Engines", in the world's rst
municipal re service in Edinburgh in 1824. He was the rst director
of the London Fire Engine Establishment (the brigade which was
eventually to become the London Fire Brigade). He is credited with
the development of the modern municipal re service.
Braidwood was born in Edinburgh in 1791. He joined the Edinburgh
Fire Brigade in 1804 and quickly rose through the ranks. In 1824, he
was appointed as the rst "Master of Engines", a position responsible
for the training and management of the re brigade. Braidwood was a
pioneer in the eld of re engineering. He developed new techniques
for reghting, such as the use of steam pumps and ladders. He also
developed new re prevention measures, such as the use of re es-
capes and sprinkler systems. Braidwood died in 1861 while ghting a
re in London. He is considered to be one of the founding fathers of
re engineering.
Bard cites Wikipedia (James Braidwood).
Overall, it is unrealistic to expect the chatbots to pinpoint a single
person as the rst re engineer since we also collectively do not know the
identity of who could be considered the rst re engineer. However,
ChatGPT seems limited in its history of reghting and engineer-
inglisting only American history. Though Bard pinpoints an interna-
tional reghter as the rst re engineer, the chatbots response is also
limited by the timeframe. It seems that both ChatGPT and Bard are un-
able to recount history earlier than the Middle Ages.
Now, we move toward more scientic questions.
3.3. Question no. 3: Explain the process of re growth in a compartment
According to D.D. Drysdale in the Encyclopedia of Physical Science
and Technology (Drysdale, 2003):
The term compartment reis commonly used to describe a re in a
room or compartmentwithin a building. If given sufcient fuel and
ventilation, the re will progress through three main stages:
1. The growth, or preashover, stage, when the re is localized in the
vicinity of its origin and the average temperature in the compartment
is still relatively low.
2. The fully developed, or postashover, re, during which all the
combustible items are involved and ames appear to ll the entire
volume and emerge from the ventilation openings.
3. The decay period, when the re begins to die down as the fuel is
consumed. This is formally identied as the period after the average
temperature has fallen to 80% of the maximum achieved during stage
2.
The average temperature in the lower part of the compartment will
remain low during the growth period, although a layer of hot smoke will
accumulate under the ceiling and will increase in depth and temperature
as the re grows in size. The development of this layer is the key factor in
the way in which a re progresses to the fully developed stage. However,
it is possible for a re to self-extinguish if the item rst ignited burns out
before other items can become involved, or to be held in check if it be-
comes starved of air (oxygen). This can occur in a space where there is
inadequate ventilation, as will occur if the doors and windows are closed
and well tting.
Then, the answers of each of the examined chatbots are listed herein.
OpenAI ChatGPT,
The process of re growth in a compartment typically involves four
stages: ignition, growth, fully developed, and decay. Here is a
breakdown of each stage:
Ignition: The ignition stage is the initial point at which the fuel source
is ignited by an external heat source. This could be a spark, a ame, or
any other ignition source that can provide enough heat to ignite the
fuel.
Growth: Once the fuel source has been ignited, the re begins to grow
rapidly. At this stage, the heat generated by the re causes the sur-
rounding fuel to heat up and release combustible gases, which in turn
feed the re and cause it to grow larger.
Fully developed: In the fully developed stage, the re has reached its
maximum size and intensity. The heat generated by the re is so
intense that it begins to affect the surrounding air temperature and
pressure, causing the re to spread rapidly throughout the
compartment.
Decay: As the fuel source is consumed, the re begins to slow down
and eventually extinguishes. This stage is called the decay stage.
Google Bard
Fire growth in a compartment is a complex process that is affected by
a number of factors, including the type of fuel, the shape of the fuel,
the ventilation conditions, and the presence of ignition sources.
The rate of re growth in a compartment can be divided into three
stages:
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
5
The incipient stage is the initial stage of re growth, when the re is
small and easily extinguished.
The growth stage is the stage in which the re grows rapidly and
becomes more difcult to extinguish.
The ashover stage is the stage in which the re reaches its maximum
intensity and becomes very difcult to extinguish.
The incipient stage is the most critical stage in re growth, as it is the
stage when the re can be most easily extinguished. If the re is not
extinguished during this stage, it will continue to grow and become
more difcult to extinguish.
The growth stage is the stage in which the re grows rapidly and
becomes more difcult to extinguish. During this stage, the re will
spread to new areas and become more difcult to contain.
The ashover stage is the stage in which the re reaches its maximum
intensity and becomes very difcult to extinguish. During this stage,
the re will produce a large amount of heat and smoke, making it
difcult for reghters to approach.
Both chatbots generally provide proper responses. ChatGPTs answer
most closely matches the denition provided, only adding ignition. For
Bard, the last stage was labeled the ashover stage. In general, ash-
over is the point in time in which fuel is instantaneously ignited during a
re and occurs at approximately 600 C(Drysdale, 2011). We generally
consider ashover as the transition point between the growth and fully
developed phases, not a stage itself. Despite the incorrect label of the
phase, Bards description is correcta fully developed re is one at its
maximum intensity and is very hard to extinguish. However, Bard (as
well as ChatGPT) fails to mention that a re may become fully developed
without reaching a ashover and that ashover will not always occur
(Kodur and Naser, 2020). For ashover to occur, the fuel must have an
adequate amount of heat energy and oxygen. For example, the presence
of a closed door versus an open one in a room with a re may be enough
to prevent ashover due to the change in the amount of available oxygen
(i.e., a closed door provides less oxygen and thus reduces the possibility
of ashover). Additionally, Bard did not include the decay phase of a
compartment re. A re enters this stage as the available fuel is
consumed or oxygen becomes limited. This may occur due if the com-
partments ventilation is inadequate (i.e. the re consumes more oxygen
than can be replaced from the compartments exterior) (Kodur and Naser,
2020). With decreasing oxygen, the heat release rate will also decrease
(Kodur and Naser, 2020). Over time, the temperature also diminishes,
although it may remain high for some time if the insulation is sufcient.
Instead of including the decay phase, Bard instead commented on ways to
reduce re spread, which was not asked in the question.
3.4. Question no. 4: Describe the mechanism(s) of re-induced spalling of
concrete
The mechanisms behind re-induced concrete spalling are complex
and vary between four main theories: 1) pore pressure accumulation due
to evaporating moisture, 2) the generation of thermal gradients, 3) the
initiation of internal cracking due to difference in thermal expansion
between concrete mixture components, 4) and temperature-induced
degradation of strength loss (Khoury, 2000;Liu et al., 2018).
Then, the answers of each of the examined chatbots are listed herein.
OpenAI ChatGPT,
Fire-induced spalling of concrete occurs when the surface layers of
concrete are exposed to high temperatures, causing the outer layer of
the concrete to break away from the rest of the structure. The
mechanism of spalling is complex and can be inuenced by a variety
of factors, including the type and quality of the concrete, the heating
rate, the temperature distribution within the concrete, and the pres-
ence of moisture.
There are several mechanisms that can contribute to re-induced
spalling of concrete:
Moisture-induced spalling: When concrete is exposed to high tem-
peratures, the moisture within the concrete can evaporate rapidly,
creating steam pressure within the concrete. If the steam pressure is
not able to escape, it can build up and cause the concrete to crack and
spall.
Thermal gradient-induced spalling: As the surface layer of the con-
crete is heated by the re, it expands rapidly, creating thermal stresses
within the concrete. If the thermal stresses exceed the tensile strength
of the concrete, it can crack and spall.
Chemical decomposition-induced spalling: High temperatures can
cause chemical reactions within the concrete that can weaken the
bonds between the cement paste and the aggregates. This can lead to
the formation of voids within the concrete, which can cause it to crack
and spall.
Mechanical damage-induced spalling: The stresses caused by the
expansion of steel reinforcement bars within the concrete can cause it
to crack and spall. This is particularly true if the steel reinforcement
has been corroded due to exposure to moisture.
Google Bard
Fire-induced spalling of concrete is a complex phenomenon that is not
fully understood. However, it is known that the following mecha-
nisms can contribute to spalling:
Temperature-induced expansion and contraction. When concrete is
heated, it expands. When it cools, it contracts. This expansion and
contraction can cause the concrete to crack.
Pore pressure buildup. When water evaporates from concrete, it
leaves behind empty pores. These pores can ll with water vapor,
which can increase the pressure inside the concrete. This pressure can
cause the concrete to crack.
Chemical reactions. When concrete is exposed to high temperatures,
it can undergo chemical reactions that can weaken the concrete.
These reactions can also cause the concrete to crack.
Micro-cracking. Even before re-induced spalling occurs, concrete
can undergo micro-cracking. These cracks are too small to be seen
with the naked eye, but they can weaken the concrete. When re-
induced spalling occurs, these micro-cracks can grow larger and
cause the concrete to spall.
Both chatbots have precise responses to this question. Spalling is a
complex phenomenon that takes place in concrete exposed to high
temperatures. In addition to listing the mechanisms of spalling, ChatGPT
additionally denes spalling while Bard does not.
3.5. Question no. 5: In terms of beams and columns, what is the best
geometric shape that limits heat transfer?
The work of Kodur and Mackinnon (Kodur and MacKinnon) has
shown that circular shapes are known to have lower heat transfer than
edged shapes.
OpenAI ChatGPT,
In general, hollow or tubular shapes can provide better thermal
insulation than solid shapes because they create an insulating air gap.
This reduces the overall heat transfer through the structural element.
Circular or rectangular hollow sections are commonly used as col-
umns, while hollow I-beams or hollow rectangular beams can be used
for beams.
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
6
Google Bard
The best geometric shape for beams and columns in terms of heat
transfer is a square or rectangular cross-section. This is because the
corners of a square or rectangle are rounded, which reduces the
amount of surface area that is exposed to the air. This in turn reduces
the amount of heat that can be lost from the beam or column.
A follow-up question was aimed at Bard, I think the best shape would be
a circular cross section.
A circular cross-section is not the best shape for beams and columns in
terms of heat transfer. A circular cross-section has a larger surface
area than a square or rectangular cross-section, which means that it
will lose heat more quickly. In addition, a circular cross-section is not
as efcient as a square or rectangular cross-section in terms of
building structures.
However, a circular cross-section is the best shape for beams and
columns in terms of strength. A circular cross-section is the strongest
shape because it has the least amount of material for the same amount
of volume.
ChatGPT provides the correct answer to this question, while Bards
answer is incorrect. Bard states that the best geometric shape for heat
transfer is a square or rectangular cross-section because the corners of a
square or rectangle are rounded.Clearly, this is not correct. Even with a
follow-up question, the chatbot is still incorrect and states that square/
rectangular sections lose heat more quickly due to their larger surface
area.
3.6. Question no. 6: What is the expected re resistance in terms of hours
of a square reinforced concrete column with a width of 400 mm and a
length of 4 m?
According to Eurocode 2 (Eurocode, 2004) (Tables 5.2a and b), the
expected re resistance of a square reinforced concrete column with a
width of 400 mm and a length of 4 m that is exposed on more than one
side can vary between 90 and 120 min.
Then, the answers of each of the examined chatbots are listed herein.
OpenAI ChatGPT,
The expected re resistance of a square reinforced concrete column
with a width of 400 mm and a length of 4 m will depend on several
factors, including the thickness and quality of the concrete, the size
and spacing of the reinforcement bars, and the type and duration of
the re exposure.
Based on typical re resistance ratings for reinforced concrete col-
umns, a square reinforced concrete column with a width of 400 mm
and a length of 4 m would be expected to have a re resistance rating
of approximately 24 hours. However, it is important to note that the
actual re resistance may vary depending on the specic design and
construction details of the column, as well as the type and duration of
the re exposure.
Google Bard
The expected re resistance of a square reinforced concrete column
with a width of 400 mm and a length of 4 m is 2 hours.
Here, Bard has the correct answer. 2 h is the high-end re resistance
for the described column. We can consider ChatGPTs response incorrect
since 24 h greatly exceed the expected re resistance according to
Eurocode. If one were to rely on this information, the column would fail
much before ChatGPT specied, and thus could cause structural damage
and potential life safety issues.
3.7. Question no. 7: What is the percentage of cost associated with re
design in terms of a typical building's design?
Common practice dictates that the answer is between 3 and 5%.
Then, the answers of each of the examined chatbots are listed herein.
OpenAI ChatGPT,
The percentage of cost associated with re design in a typical build-
ing's design can vary widely depending on various factors such as the
type of building, its size, and the complexity of the re safety system.
However, generally, the cost of re design typically ranges from 2% to
5% of the total construction cost.
While re design may represent a relatively small percentage of the
total construction cost, it is a critical component of building design
that can have a signicant impact on the safety of occupants and the
protection of property. Investing in high-quality re design can help
to reduce the risk of re-related damage, minimize insurance costs,
and ensure the long-term safety and viability of the building.
Google Bard
The cost of re design is typically 25% of the total cost of a building's
construction. This means that for a $100 million building, the cost of
re design would be $25 million.
The cost of re design is an important part of the total cost of a
building's construction. By investing in re design, building owners
can help to ensure that their buildings are safe and compliant with re
codes.
This question refers to the construction and fabrication cost of re
protection and suppression systems as a portion of total construction
costs. Both chatbotsanswers match our domain knowledge. It is also
worth noting that the cost of re designmay also refer to the cost
associated with the analysis and design of items such as re loading,
structural design (including but not limited to beam/column design,
passive protection methods, etc. For re) by engineers as ofce or desk-
based work, although this was not the intention of the question. Common
practice dictates this to be between 5 and 15% of total engineering design
cost. Neither chatbot mentioned this type of cost.
3.8. Question no. 8: What is re evacuation?
According to Cambridge English Dictionary (McIntosh, 2023), evac-
uation is:
the act of moving people from a dangerous place to somewhere safe
Similarly, emergency evacuation is the urgent, immediate egress or
escape of people away from an area that contains an immediate threat to
lives or property. Evacuation can be small- or large-scale and due to a
variety of reasons (including but not limited to re, etc.). In general,
evacuation can be divided into two main phases: pre-evacuation and
movement. In the pre-evacuation phase, evacuees gather personal items
and make decisions about how they will evacuate from danger. This
phase includes detection, decision, alarm, and reaction. In the movement
phase, evacuees physically move toward safety using the decisions they
made in the previous phase (Hurley et al., 2016).
OpenAI ChatGPT,
Fire evacuation is the process of safely and efciently moving people
away from a building, structure, or area where a re is present or has
the potential to occur. The primary goal of re evacuation is to protect
the lives of the occupants by minimizing injuries, fatalities, and
exposure to toxic smoke and gases.
Fire evacuation procedures typically involve several key components:
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
7
1. Early detection and warning: Fire alarms and smoke detectors play a
crucial role in alerting occupants of a re and initiating the evacua-
tion process.
2. Evacuation routes and exits: Pre-determined evacuation routes and
clearly marked exits should be established in a building to guide oc-
cupants to safety. These routes should be free from obstructions and
provide the shortest, safest path to the outside.
3. Fire drills and training: Regular re drills and safety training help
familiarize building occupants with evacuation procedures, ensuring
that they know how to respond quickly and calmly in the event of a
re.
4. Assembly points: Designated assembly areas outside the building
should be established, allowing evacuees to gather and be accounted
for, and ensuring that emergency responders have clear access to the
building.
5. Emergency communication: Effective communication during an
evacuation is essential for relaying important information, such as the
location of the re or any changes in the evacuation plan.
Fire evacuation plans should be tailored to the specic needs of the
building and its occupants, taking into consideration factors such as
building layout, occupancy levels, and the presence of individuals
with disabilities or other special needs.
Google Bard
Are evacuation is the process of removing people from a building or
other structure that is on re. The goal of a re evacuation is to ensure
the safety of all occupants by removing them from the re and smoke
as quickly as possible. Fire evacuations are typically carried out by
trained professionals, such as reghters or building security
personnel. However, it is important for all occupants of a building to
be familiar with the re evacuation plan and to know how to exit the
building safely in the event of a re.
The following are some important steps to take in the event of a re
evacuation:
Stay calm. Panicking can lead to injuries and can make it difcult to
evacuate the building safely.
Follow the instructions of the re department or other emergency
personnel. They are trained to ensure the safety of all occupants and
will direct you to the safest exit route.
If you are unable to evacuate on your own, call for help. The re
department will be able to provide assistance and help you to evac-
uate the building safely.
Once you have evacuated the building, stay away from the area until
it is safe to return. The re department will need time to extinguish
the re and to ensure that the area is safe for occupants to return.
Fire evacuations are a serious matter, but by following the steps
above, you can help to ensure your safety and the safety of others in
the event of a re.
ChatGPT and Bard both provide valid responses to this question. An
evacuation can be prompted for a variety of scenarios and involve many
different methods of reaching safety. Most importantly, it is the move-
ment of people away from an area of danger. Additionally, Bard is correct
in noting that evacuation can be planned or unplanned. However, a
planned evacuation does not constitute a well-organized one or one
without urgency/chaos. It is important to note that evacuation does not
require a mechanical method of transportation, as Bard describes (buses,
boats, helicopters, etc.). Evacuation can also describe the movement of
people by walking/running, such as when building occupants evacuate
during a re or re drill. We would like to point out that Proulx (1993)
and Fahy (Fahy and Proulx, 2009) offered explanations for debunking the
myth of panic through stress models. This concept has been clearly
adopted by Bard.
3.9. Question no. 9: What are the two primary phases of a building
evacuation?
According to Rahouti et al. (2020):
The engineering evacuation timeline divides the evacuation process
broadly into two evacuation performance phases following on from
the re detection and notication phases: pre-evacuation and evac-
uation movement. The pre-evacuation stage extends from the noti-
cation of the population (i.e. the point at which they might be aware
of the incident) to the moment when an evacuee starts moving to-
wards a place of safety; the movement phase starts directly after the
pre-evacuation phase, and ends once the evacuee has reached a place
of safety.
OpenAI ChatGPT,
A building evacuation typically consists of two primary phases:
Alert and notication phase: This phase involves detecting the
emergency, such as a re or a security threat, and activating the alarm
system. In this stage, occupants are alerted to the emergency, and the
appropriate authorities, like the re department or police, are noti-
ed. The alarm system may include visual and audible signals, like
ashing lights and sirens, to ensure that all building occupants,
including those with hearing or visual impairments, are informed of
the need to evacuate. Clear communication of the emergency situa-
tion and evacuation instructions is crucial during this phase.
Evacuation and movement phase: This phase involves the actual
process of moving occupants from the building to a safe location
outside. It starts with occupants leaving their immediate area and
progresses to them following designated evacuation routes towards
exits. The evacuation process may be structured differently depend-
ing on the building's layout and the nature of the emergency. For
example, a phased evacuation could be implemented for large, multi-
story buildings, where certain oors or sections are evacuated rst to
avoid congestion and bottlenecks. In some cases, a shelter-in-place
strategy might be used if evacuation is deemed more dangerous
than staying inside the building. This phase also includes accounting
for all occupants at the assembly point and ensuring that everyone has
safely evacuated.
Google Bard
The two primary phases of a building evacuation are:
Alarm and notication: This phase involves alerting people to the
danger and instructing them to evacuate. This can be done through a
variety of methods, such as using a public address system, ringing
bells, or using ashing lights.
Evacuation: This phase involves people leaving the building and
going to a safe location. This can be done through a variety of
methods, such as using stairs, elevators, or ladders.
Both chatbots have responded correctly to this question. In general,
evacuation consists of a pre-evacuation phase and a movement phase.
Additionally, both platforms mention a broader plan for emergency
response that includes pre-planning, clear communication, and regular
practice; while this does not directly answer the question, it is an
appreciated addition. Both chatbots label the rst phase the alarm and
notication phase. This is mostly correct, given that building occupants
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
8
are generally notied of an emergency via re alarm (or another type).
However, it is important to note the differences in how each chatbot
described the phase. For ChatGPT, the alarm and notication stage de-
scribes the actions the evacuees must go through before the evacuation or
movement phase. For Bard, the phase was described not for the evacuees
but instead for re safety ofcials or others who can alert evacuees to
danger. In an evacuation, the pre-movement phase most closely aligns
with the evacuees' actions, not the re safety ofcials.
3.10. Question no. 10: Who is the most vulnerable type of person for
injury/death in a building re?
According to researchers from the National Institute of Standards and
Technology (Gilbert and Butry, 2018), older persons (elderly), young
children, and males are some of the most vulnerable people to re injury
and death.
OpenAI ChatGPT,
During a building re, certain groups of people are more vulnerable to
injury or death due to factors such as limited mobility, cognitive
impairments, or other specic needs. The most vulnerable groups
include:
Elderly individuals: Older people may have reduced mobility, slower
reaction times, and age-related health issues, which can hinder their
ability to evacuate quickly and safely.
Children: Young children, particularly those below the age of ve, are
at greater risk due to their limited understanding of danger, inability
to follow complex instructions, and dependence on adult caregivers
for guidance and protection.
People with disabilities: Individuals with physical, sensory, or
cognitive disabilities may face challenges during evacuation, such as
navigating through a building, understanding alarms, or communi-
cating with others.
People with chronic illnesses or medical conditions: Those with
chronic health issues may have reduced stamina, limited mobility, or
other factors that hinder their ability to evacuate efciently.
Pregnant women: Pregnancy may limit mobility, stamina, or the
ability to navigate obstacles, increasing the risk of injury or death
during a building re.
Non-native speakers or those with limited language prociency:
People who do not understand the local language may have difculty
comprehending evacuation instructions, announcements, or other
critical information, which can impede their ability to evacuate
safely.
Google Bard
The most vulnerable type of person for injury/death in a building re
is children. Children are more likely to be injured or killed in a re
because they are:
More likely to be home alone when a re starts
Less likely to know how to escape a re
More likely to be curious and explore, which could lead them into
dangerous areas
More likely to be unable to self-rescue
Children are also more likely to be injured or killed in a re because
they have a smaller body surface area, which means they absorb heat
more quickly. They also have a higher respiratory rate, which means
they inhale smoke and fumes more easily.
ChatGPT seems to have answered this question thoroughly, even
pointing out that non-native speakers and pregnant women are at addi-
tional risk for injury/death in building res. Bard, on the other hand,
only identied one type of person at risk during res: children. While
children are at increased risk during res, they are not necessarily the
most vulnerable. This response fails to include the elderly and people
with disabilities, who are most commonly cited as having the most lim-
itations during evacuations.
3.11. Question no. 11: Why are people with disabilities more vulnerable
than those without disabilities in building res?
In general, people with disabilities are more vulnerable than those
without disabilities in building res because they experience limited
mobility or difculty in understanding the severity of the re scenario
(i.e., people with sensory disabilities may not hear the re alarm, and
those with cognitive disabilities may not be able to make informed
decisions about the re). Thus, they may move slower, have difculty
traversing some egress components such as stairs, or react slower than
the non-disabled population. Ultimately, their rate of evacuation may
be too slow compared to the rate of re spread and result in injury or
death. Additionally, egress routes and evacuation considerations are
rarely geared toward people with disabilities (Hostetter and Naser,
2022). This may increase their vulnerability in emergency situations.
4. OpenAI ChatGPT
People with disabilities are often more vulnerable than those without
disabilities during building res for several reasons:
Limited mobility: Individuals with physical disabilities, such as those
who use wheelchairs, walkers, or crutches, may face challenges
navigating through narrow corridors, stairwells, or uneven surfaces
during an evacuation. Their movement speed may also be slower,
which can put them at a greater risk in a rapidly developing re
situation.
Sensory impairments: People with visual or hearing impairments may
have difculty receiving or interpreting re alarms and emergency
instructions. For instance, a person with a hearing impairment might
not hear an audible alarm, while someone with a visual impairment
might struggle to navigate through a smoke-lled environment or
read emergency exit signs.
Cognitive impairments: Individuals with cognitive disabilities, such
as intellectual disabilities or dementia, may have difculty under-
standing the urgency of a re situation, following evacuation in-
structions, or making quick decisions during an emergency.
Communication barriers: People with speech impairments or lan-
guage disorders might have trouble communicating their needs or
location to emergency responders, which could hinder their rescue or
evacuation.
Dependence on assistive devices or personal assistance: Some in-
dividuals with disabilities rely on assistive devices, such as wheel-
chairs or hearing aids, or require personal assistance from caregivers
during daily activities. In an emergency, these individuals may need
extra time or assistance to evacuate, which can increase their
vulnerability.
Google Bard
People with disabilities are more vulnerable than those without dis-
abilities in building res for a number of reasons:
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
9
They may have difculty escaping a re. People with disabilities may
have difculty using stairs or elevators, and they may not be able to
move quickly enough to escape a re.
They may have difculty hearing or seeing a re. People with dis-
abilities may not be able to hear the smoke alarm or see the smoke or
ames.
They may have difculty communicating with others. People with
disabilities may not be able to communicate with others in an
emergency situation.
They may have difculty accessing emergency services. People with
disabilities may have difculty accessing emergency services, such as
calling 911 or getting to a safe location.
People with disabilities other than physical disabilities are rarely
studied in terms of re evacuation (Hostetter and Naser, 2022). Thus,
ChatGPTs inclusion of people with sensory and cognitive disabilities is
an important inclusion in its response to this question. ChatGPT further
includes those with physical disabilities and communication barriers.
Each item discussed can play a signicant role in evacuation and re
vulnerability and is not present for the non-disabled population. Bard
presents a more general answer but explains how each factor can affect
the vulnerability of people with disabilities when compared to those
without. Each response is deemed good. We would like to note that the
specic US-based re department (i.e., 911), as opposed to 999, is often
used in other countries.
4.1. Question no. 12: How can a wheelchair user evacuate from a
building?
According to Vanderbilt Universitys Emergency Management
(Evacuation for persons with disabilities, 2023) division, the following
methods and techniques can be used by and for wheelchair users during
an evacuation:
Discuss needs and preferences. Non-ambulatory personsneeds and
preferences vary widely and therefore require you to ask them how
they would like to be assisted.
Wheelchair-user on the ground oor. Individuals who use wheel-
chairs may choose to evacuate themselves from the ground oor with
minimal assistance.
Ensure clear path of travel. If debris is present, it may be necessary to
clear a path to the exit.
No imminent danger. If there is no imminent danger, the person may
choose to remain in the building or to be directed to an Area of Refuge
(stairwell) until emergency personnel arrive. Fire Department
personnel, who are trained in emergency rescue, can then enter the
building and assist the person in exiting the building, either down the
stairs or using the emergency elevator recall.
Imminent danger. If danger is imminent and the individual does not
wish to be removed from his or her wheelchair, direct the person to
the nearest Area of Refuge (stairwell) and notify emergency personnel
immediately.
Carrying wheelchair users. Most wheelchairs are too heavy to carry
down stairs. If the person wishes to be carried down the stairs without
the wheelchair, consult with him or her on the best carry options, e.g.,
two-person cradle carry, ofce chair evacuation, or, if available, an
evacuation chair.
Mobility aids or devices. Return any mobility aids or devices to the
person as soon as possible
Notify emergency personnel. Notify emergency personnel immedi-
ately about any individuals remaining in the building and their
locations.
Additional methods of evacuation for wheelchair users include
evacuation chairs (or stair chairs) or elevators approved for re evacu-
ation by the International Building Code (Butler et al., 2017).
OpenAI ChatGPT,
A wheelchair user can evacuate from a building using various stra-
tegies and techniques, depending on the availability of assistance, the
building's accessibility features, and the nature of the emergency.
Some methods to evacuate a wheelchair user include:
Evacuation chairs: Evacuation chairs are specically designed for
emergency use, allowing wheelchair users to be safely transported
down stairs or across uneven surfaces. With assistance from trained
individuals, a wheelchair user can transfer to an evacuation chair and
be guided to safety.
Areas of refuge or safe waiting areas: In multi-story buildings with
accessible re-rated stair enclosures or designated safe waiting areas,
a wheelchair user can wait for emergency responders to arrive and
assist with their evacuation. These areas should have a two-way
communication system to alert emergency personnel of their location.
Horizontal evacuation: Depending on the building's layout and the
emergency situation, a wheelchair user may be able to evacuate to a
safer area on the same oor, such as an adjacent re compartment,
until emergency responders arrive to assist with the evacuation.
Using lifts or elevators: While it is generally advised not to use ele-
vators during a re, some buildings may have specially designed
evacuation elevators that can be used by wheelchair users and other
individuals with limited mobility. These elevators are constructed
with additional re protection features and can be controlled by
emergency responders or building staff during an evacuation.
Personal Emergency Evacuation Plans (PEEPs): A PEEP is a person-
alized evacuation plan developed in collaboration with the individ-
ual, building management, and emergency services. It outlines the
specic steps, support, and equipment needed for a wheelchair user to
evacuate safely in case of an emergency.
Buddy system: A wheelchair user can pair up with a designated
"buddy" or several buddies who can provide assistance during an
evacuation. Buddies should be familiar with the person's needs, the
use of evacuation equipment, and the planned evacuation route.
Google Bard
Here are some tips for wheelchair users to evacuate from a building:
Know your building's evacuation plan. Before an emergency happens,
make sure you know your building's evacuation plan. This plan
should include a way to alert people to the danger, a way to evacuate
people from the building, and a way to account for everyone who is in
the building.
Practice your plan. Once you know your building's evacuation plan,
make sure you practice it with everyone in your building. This will
help everyone know what to do in an emergency.
Be prepared to evacuate quickly. When an emergency happens, you
may not have much time to evacuate. Be prepared to evacuate quickly
and orderly.
Follow the instructions of emergency personnel. If you are evacuated
by emergency personnel, follow their instructions carefully. They will
help you to evacuate safely.
Here, ChatGPT provides the only deemed good response. Wheelchair
users may self-evacuate, relocate to an area of refuge and call for help,
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
10
recruit assistance from non-disabled evacuees to help them downstairs or
to use an evacuation chair, or use an elevator if it is approved for re
evacuation. Bard does not provide specic methods for the evacuation of
wheelchair users. Instead, it generalizes re safety tips for all evacuees. If
someone was relying on this information to help a wheelchair user
evacuate in an emergency, it would not provide sufcient help.
4.2. Question no. 13: What are the differences in evacuation between a
non-disabled person, physically disabled person, and cognitively disabled
person?
The evacuation process can vary signicantly from a non-disabled
person to a physically disabled person, to a cognitively disabled per-
son. When compared to the non-disabled evacuee, people with physical
disabilities may require assistance to evacuate, travel at slower speeds, or
be required to use a different exit route if they cannot traverse stairs or
other obstacles (Hostetter and Naser, 2022). Cognitively disabled evac-
uees can generally use the same exit route as those without disabilities,
but they may take longer to make decisions in the pre-evacuation phase
(Hostetter and Naser, 2022). Additionally they may have trouble un-
derstanding or interpreting the event and have heightened emotions.
OpenAI ChatGPT,
The evacuation process can vary signicantly for abled individuals,
physically disabled individuals, and individuals with cognitive dis-
abilities. Some of the differences in evacuation for these groups
include:
Abled person: Typically, an abled person can move more quickly and
navigate obstacles, such as stairs and narrow corridors, with relative
ease. They can usually understand and follow evacuation instructions,
respond to audible and visual alarms, and make quick decisions
during an emergency. In many cases, abled individuals can evacuate
independently without assistance.
Physically disabled person: Physically disabled individuals may have
limited mobility, which can make it challenging for them to navigate
stairs, corridors, or uneven surfaces. They might require assistance,
specialized equipment, or additional time to evacuate the building
safely. Wheelchair users, in particular, may need evacuation chairs,
accessible escape routes, or designated waiting areas for emergency
responders to assist with their evacuation.
People with sensory impairments, such as visual or hearing impair-
ments, might require alternative communication methods, such as
tactile or visual alarms, to receive emergency alerts and instructions.
Person with cognitive disability: Individuals with cognitive disabil-
ities, such as intellectual disabilities or dementia, may have difculty
understanding the urgency of an emergency situation or following
complex evacuation instructions. They might require additional
guidance, support, or reassurance from others to evacuate safely.
Google Bard
The differences in evacuation between an abled person, a physically
disabled person and a person with cognitive disability can be sum-
marized as follows:
Ability to move: Abled people can move quickly and easily, while
physically disabled people may have difculty moving or may not be
able to move at all. People with cognitive disabilities may have dif-
culty understanding instructions or may not be able to follow them.
Ability to see and hear: Abled people can see and hear, while physi-
cally disabled people may have difculty seeing or hearing. People
with cognitive disabilities may have difculty processing information
or may not be able to understand what is happening.
Ability to communicate: Abled people can communicate with others,
while physically disabled people may have difculty communicating
or may not be able to communicate at all. People with cognitive
disabilities may have difculty understanding what others are saying
or may not be able to respond to them.
Both chatbots have responded correctly to this question. For
ChatGPT, an important consideration in the pre-evacuation times of
cognitively disabled occupants was missed. A signicant portion of the
evacuation for people with cognitive disabilities is spent in the pre-
evacuation phase when compared to non-disabled evacuees. For Bard,
those with sensory disorders (seeing or hearing impaired) were included
in the group with physical disabilities, despite research noting a differ-
ence between the two [ (Hostetter and Naser, 2022).
4.3. Question no. 14: How can researchers study evacuation?
In the past, researchers most commonly used real-time evacuation
experiments and historical events (such as the MGM Grand re (Bryan,
1983) or 9/11 evacuation (Averill et al., 2005)) to study egress and
human behavior. More recently, evacuation simulation and modeling
have been introduced as a cost- and time-effective alternative (Kuli-
gowski, 2016;Kuligowski et al., 2010;Gwynne et al., 2020). This method
is more ethical, does not pose safety risks to participants, and does not
require a real building.
OpenAI ChatGPT,
Researchers can study evacuation without putting people at risk or
relying on actual natural disasters by using various methods,
including simulations, virtual reality, mathematical modeling, and
controlled experiments. These approaches help researchers under-
stand human behavior, assess evacuation procedures, and improve
safety measures. Some methods used to study evacuations are:
Computer simulations: Researchers use software to simulate evacua-
tions, considering factors like building layouts, exit locations, and
human behavior. By adjusting these parameters, they can analyze
how different scenarios impact evacuation efciency and identify
potential bottlenecks or hazards.
Virtual reality: Virtual reality (VR) provides a safe and immersive
environment to study evacuation behavior. Participants equipped
with VR headsets experience realistic emergency situations, allowing
researchers to observe their decision-making and movement patterns
without exposing them to real danger.
Agent-based modeling: This approach involves creating computa-
tional models of individual agents (people) within a simulated envi-
ronment. Researchers can assign specic characteristics to these
agents, such as speed, decision-making abilities, and group dynamics,
to understand how they collectively impact evacuation outcomes.
Controlled experiments: Researchers can set up controlled experi-
ments in real-world settings, such as buildings or public spaces, to
study evacuation processes. These experiments may involve volun-
teers participating in mock evacuations, allowing researchers to
observe and analyze human behavior and the effectiveness of evac-
uation procedures.
Surveys and interviews: Researchers may collect data from in-
dividuals who have experienced real evacuations to gain insights into
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
11
their experiences and decision-making processes. This qualitative
data can help inform evacuation planning and improve procedures.
Analyzing historical data: Researchers can study past evacuation
events, including successes and failures, to identify patterns and learn
from past experiences. This analysis can provide valuable insights
into what works well and what needs improvement in evacuation
planning and execution.
Google Bard
Researchers can study evacuation in a number of ways without
putting people at risk or without true natural disaster. Some common
methods include:
Simulations: Researchers can use computer simulations to model the
behavior of people during an evacuation. These simulations can be
used to test different evacuation plans and to identify potential
problems.
Experiments: Researchers can conduct experiments in which people
are asked to evacuate a simulated disaster area. These experiments
can be used to study how people make decisions during an evacuation
and to identify factors that affect evacuation behavior.
Surveys: Researchers can conduct surveys to ask people about their
experiences during evacuations. These surveys can be used to learn
about how people perceive evacuations and to identify ways to
improve evacuation planning and execution.
Literature review: Researchers can review the literature on evacua-
tion to learn about the existing research on this topic. This literature
review can help researchers to identify the gaps in the research and to
develop new research questions.
Responses from both platforms precisely reect the variety of ways in
which researchers can study evacuation today. Simulations, experiments,
surveys, and reviews of historical data or current literature can provide
important and new information that may improve future building design
or re safety considerations. Researchers may additionally use a com-
bination of these techniques within their projects to supplement data or
improve one or more methods.
4.4. Question no. 15: Design an evacuation experiment in a high-story
building with a mixed population of occupants that are non-disabled,
physically disabled, and cognitively disabled
There are many factors that go into designing a real-time evacuation
experiment. This includes participant recruiting, identication of study
objectives, determination of location/date/time, data collection, anal-
ysis, and result reporting.
According to OpenAI ChatGPT,
To design an evacuation experiment in a high-story building with a
mixed population of occupants, including abled individuals, physi-
cally disabled individuals, and individuals with cognitive disabilities,
you will need to consider various factors to ensure safety, ethical
considerations, and effective data collection. Here's a suggested
outline for such an experiment:
1. Objective: Determine the efciency and effectiveness of the building's
evacuation plan for a mixed population, identify potential bottlenecks
or challenges, and provide recommendations for improvement.
2. Participants: Recruit a diverse group of volunteers, including abled
individuals, physically disabled individuals (e.g., wheelchair users,
people with mobility impairments), and individuals with cognitive
disabilities (e.g., intellectual disabilities, dementia). Ensure that
participants are fully informed about the experiment, its purpose, and
any potential risks involved. Obtain written consent from all partic-
ipants or their legal guardians, if applicable.
3. Building setup:
Ensure that the building meets accessibility standards and has
appropriate evacuation equipment, such as evacuation chairs, vi-
sual and audible alarms, and clear signage.
Designate accessible escape routes, areas of refuge or safe waiting
areas, and assembly points for all participants.
Consider any building-specic challenges or unique features, such
as the number of oors, stairwells, or elevator access.
4. Experiment procedure:
Divide the participants into groups based on their abilities or dis-
abilities, ensuring each group has a mix of the different
populations.
Assign a trained "buddy" or assistant to each participant with a
physical or cognitive disability to provide support and guidance
during the evacuation.
Conduct a brieng for all participants, explaining the evacuation
procedure, their roles, and any specic instructions.
Simulate an emergency, such as a re or security threat, and acti-
vate the alarm system.
Begin the evacuation process, carefully monitoring and recording
the time taken for each group to reach the designated assembly
points.
Ensure that the experiment is supervised by trained personnel,
including emergency responders, researchers, and building staff, to
ensure the safety of all participants.
5. Data collection and analysis:
Collect data on the evacuation times, routes taken, bottlenecks or
challenges encountered, and any instances where assistance was
required.
Record any observations, feedback, or suggestions from partici-
pants and assistants.
Analyze the data to identify patterns or trends, such as areas where
evacuation was slower or more challenging for certain populations.
6. Recommendations and follow-up:
Based on the ndings, provide recommendations for improving the
building's evacuation plan, such as modifying escape routes, adding
additional equipment, or providing additional training for staff and
emergency responders.
Share the results with building management, emergency services,
and participants to raise awareness and promote best practices for
inclusive evacuation planning.
If possible, implement the suggested improvements and conduct
follow-up experiments to evaluate their effectiveness.
7. Ethical considerations:
Ensure the safety of all participants by closely monitoring the
experiment and providing appropriate support as needed.
Obtain informed consent from all participants or their legal
guardians, if applicable.
Maintain condentiality and anonymity of the participants' per-
sonal information.
Be prepared to halt the experiment if any safety concerns arise.
By conducting a well-designed evacuation experiment with a diverse
population, you can gather valuable insights into the effectiveness of
the building's evacuation plan and identify areas for improvement,
ultimately enhancing safety for all occupants.
Google Bard
Here is a possible design for an evacuation experiment in a high-story
building with a mixed population of occupants that are able, physi-
cally disabled, and cognitively disabled:
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
12
1. The experiment would be conducted in a simulated high-story
building. The building would be equipped with smoke detectors,
re alarms, and sprinkler systems.
2. The building would be populated with a group of volunteers who
would represent the different types of occupants that would be found
in a real-world high-story building. The volunteers would be divided
into three groups: able-bodied occupants, physically disabled occu-
pants, and cognitively disabled occupants.
3. The experiment would be conducted in a controlled environment. The
volunteers would be instructed to evacuate the building in a simu-
lated re. The evacuation would be timed to measure the time it took
each group of occupants to evacuate the building.
4. The results of the experiment would be analyzed to determine the
effectiveness of different evacuation strategies for different types of
occupants. The results could be used to develop new evacuation
strategies that would help to improve the safety of all occupants in a
high-story building.
While ChatGPT has the more detailed response, both platforms have
provided a proper outline for conducting an evacuation experiment. In
the case of ChatGPT, the response says that participants should be noti-
ed of their roles and given instructions on how to evacuate from the
building. In most cases, evacuees do not know about re drills or real
evacuations prior to the event, and thus, this is not the best method for
conducting an experiment as it will not simulate the natural or candid
behavior of evacuees. Additionally, ChatGPT notes that each person with
a physical disability should be assigned a helper during the experiment.
This optimistically assumes that each person with a disability will have
an assistant at hand at the start of a re scenario. In reality, this may not
be the case. People with physical disabilities may not need assistance to
evacuate from the building or be unable to nd help. Regardless, to best
simulate a real re scenario, those with disabilities should not be
assigned help prior to the evacuation. Bard, on the other hand, provides a
brief overview of the experiment. It divides people according to their
disabilities with the goal of determining effective evacuation strategies
for each population. This displays an overall good understanding of how
to conduct an experiment. Unlike ChatGPT, Bard does not seem to
mention or address the ethical component of evacuation experiments.
4.5. Question no. 16: What is smart reghting?
The concept of smart reghting is still new in the re research and
engineering community. Arguably, it was rst systematically summa-
rized in the 2015 NIST Report Research Roadmap for Smart Fire
Fighting(Bliss et al., 2015), which was before the concept of articial
intelligence (AI), VR/AR became popular in the engineering community.
Today, different researchers and industries still do not have a unied
answer to the scope of smart reghting and the denition of smart.
Then, the answers of each of the examined chatbots are listed herein.
OpenAI ChatGPT,
Smart reghting refers to the use of advanced technologies, data
analysis, and communication systems to enhance the efciency,
safety, and effectiveness of reghting operations. The concept aims
to improve situational awareness, decision-making, and resource
management for both reghters and incident commanders, ulti-
mately reducing the impact of res on lives, property, and the
environment.
Key elements of smart reghting can include:
1. Internet of Things (IoT) devices: Sensors and connected devices can
be used to monitor and collect real-time data on various parameters,
such as temperature, smoke levels, and building occupancy. This
information can help reghters assess the situation and develop
appropriate response strategies.
2. Robotics and drones: Robots and unmanned aerial vehicles (UAVs)
can be employed to gather visual and thermal imagery, as well as to
access hazardous or hard-to-reach areas during a re incident. This
can help reghters assess the situation without putting their lives at
risk.
3. Advanced communication systems: Reliable and interoperable
communication systems are crucial for sharing information and
coordinating efforts between reghters, incident commanders, and
other emergency responders.
4. Predictive analytics and data-driven decision-making: By analyzing
historical data and using machine learning algorithms, predictive
analytics can help identify potential re hazards and inform proactive
measures to mitigate risks. Additionally, real-time data analysis
during an incident can help incident commanders make informed
decisions about resource allocation and reghting strategies.
5. Augmented and virtual reality (AR/VR): AR and VR technologies can
be used for immersive training simulations, allowing reghters to
practice and develop their skills in a safe and controlled environment.
During an actual incident, AR can also be used to overlay crucial in-
formation, such as building layouts or the location of hazardous
materials, onto the reghter's eld of view.
By integrating these technologies and approaches, smart reghting
aims to enhance the overall effectiveness and safety of re response
efforts, ultimately saving lives and minimizing property damage.
Google Bard
Smart reghting is a term used to describe the use of new technol-
ogies to improve the safety and efciency of reghting operations.
Smart reghting technologies can include things like:
Sensors: Sensors can be used to detect res, smoke, and other haz-
ards. This information can be used to alert reghters to potential
problems and to help them to make better decisions about how to
respond to a re.
Robotics: Robots can be used to perform dangerous or difcult tasks,
such as entering burning buildings or carrying heavy equipment. This
can help to protect reghters from injury and can also help them to
work more efciently.
Articial intelligence: Articial intelligence can be used to analyze
data from sensors and to make decisions about how to respond to a
re. This can help reghters to make better decisions more quickly,
which can save lives and property.
Smart reghting is still in its early stages of development, but it has
the potential to revolutionize the way that reghters work. By using
new technologies, reghters can be better equipped to protect
themselves and the public from res.
Here are some of the benets of smart reghting:
Increased safety: Smart reghting technologies can help to protect
reghters from injury or death. For example, robots can be used to
enter burning buildings and to carry out dangerous tasks, which can
help to keep reghters safe.
Improved efciency: Smart reghting technologies can help re-
ghters to work more efciently. For example, sensors can be used to
detect res early, which can give reghters more time to respond
and to extinguish the re.
Reduced costs: Smart reghting technologies can help to reduce
the costs of reghting. For example, robots can be used to perform
H. Hostetter et al. Natural Hazards Research xxx (xxxx) xxx
13
tasks that would otherwise require human labor, which can save
money.
Overall, smart reghting is a promising new technology that has the
potential to improve the safety and efciency of reghting
operations.
ChatGPT does not mention the use of AI in smart reghting, but it
includes other new concepts like AR/VR and robotic technologies. Bard
mentions AI, robotics, and sensors, but does not include communication
and data-driven technology. Thus, they both provide partial picture of
smart reghting, but their answer is already as good as most re re-
searchers. Bard also gives more information about how smart reghting
can improve the current reghting practices. Overall, Bard is more
informative.
4.6. Question no. 17: Can you predict the global wildre activity in the
next 20 years under the inuence of climate change?
This is a tough question. Most wildre experts will be reluctant to give
some predictions, and even if they can, they usually can give some
prediction about the regime that they are familiar with rather than a
prediction in the global scale.
OpenAI ChatGPT,
As an AI language model, I cannot predict future events with cer-
tainty. However, I can provide you with an analysis based on current
scientic understanding and trends.
Climate change has already been linked to an increase in the fre-
quency, intensity, and duration of wildres in various regions around
the world. Over the next 20 years, global wildre activity is likely to
be signicantly inuenced by climate change due to the following
factors:
1. Rising temperatures: Higher temperatures can lead to increased
evaporation rates and drier vegetation, making it more susceptible to
ignition and burning. This can result in more frequent and severe
wildres.
2. Changes in precipitation patterns: Climate change can lead to shifts in
precipitation patterns, causing some regions to become drier and
more prone to wildres. Other regions may experience increased
precipitation, which can initially reduce re risk, but may later lead to
increased vegetation growth that can serve as fuel for future res.
3. Longer re seasons: As temperatures rise and snowpacks melt earlier,
re seasons are becoming longer in many regions. This provides a
larger window of time for wildres to ignite and spread.
4. Increased frequency of extreme weather events: Climate change is
associated with more frequent and severe droughts, heatwaves, and
wind events, all of which can exacerbate wildre risk and contribute
to more intense res.
5. Changes in vegetation and ecosystems: Climate change can alter the
distribution and composition of plant species, leading to changes in
the availability and ammability of fuels. In some cases, this may
result in increased re risk.
Given these factors, it is expected that global wildre activity will
likely continue to increase in the coming decades, particularly in re-
gions that are already prone to wildres, such as the western United
States, Australia, the Mediterranean, and parts of South America and
Africa. However, it's important to note that regional variations and
uncertainties in climate projections make it difcult to predict the
precise extent of these changes.
To mitigate the impact of climate change on wildres, it will be
essential to implement adaptive management strategies, such as
improved land-use planning, re risk reduction measures, early
warning systems, and sustainable forest management practices. In
addition, reducing greenhouse gas emissions and promoting global
efforts to address climate change will be crucial for minimizing future
wildre risks.
Google Bard
Climate change is expected to have a signicant impact on global
wildre activity in the next 20 years. Rising temperatures and drier
conditions are likely to lead to more frequent and severe wildres,
especially in areas that are already prone to re. This could have a
devastating impact on ecosystems, communities, and economies.
Here are some of the ways that climate change could affect wildre
activity:
Rising temperatures: Warmer temperatures will increase the rate of
evaporation, which will lead to drier conditions. This will make
vegetation more ammable, and will make it easier for res to start
and spread.
Drier conditions: Droughts are becoming more common and severe
due to climate change. This will further reduce the amount of mois-
ture in vegetation, making it even more ammable.
More extreme weather events: Climate change is making extreme
weather events, such as heat waves, droughts, and wildres, more
common and severe. These events can create the perfect conditions
for wildres to start and spread.
Changes in vegetation: Climate change is also causing changes in
vegetation patterns. Some plants are becoming more common, while
others are becoming less common. This could change the fuel load for
wildres, making some areas more prone to re than others.
The impacts of climate change on wildre activity are likely to be felt
unevenly around the world. Some areas, such as the Western United
States, are already experiencing the effects of climate change on
wildres. These areas are likely to see even more severe wildres in
the future. Other areas, such as the Arctic, are also at risk from climate
change-induced wildres.
Wildre activity is a complex issue that is affected by many factors.
Climate change is just one of these factors, but it is a factor that is
expected to have a signicant impact in the future. It