Analysis of travel decision behaviour during extreme floods in Pakistan

Abstract

Extreme weather events like floods, exacerbated by climate change, disrupt transportation systems and endanger the safety of the travellers. This study investigates how public perceptions of travel risks influence travel decisions during floods in Pakistan. In the context of urban floods and transportation, this study focuses on the perceived probability of a crash, perceived distraction, perceived security, perceived need for a companion and perceived possibility of getting stuck in traffic congestion, affecting people's travel decisions during floods. A questionnaire survey was conducted in Pakistan after the 2022 floods. Results from the data analysis of 526 valid responses indicated that travel decisions during floods were sensitive to the various flooding situations and influenced by factors including sociodemographic and trip purposes. Furthermore, public perceptions varied in different flooding scenarios influenced by familiarity, available information and road conditions. The study suggests that real-time information sharing to ensure security, targeted education and training to enhance community resilience against floods, development of guidelines for shelter camps on roadsides and ensuring the availability of public transport and ride-sharing services during floods may help to fulfill the public's mobility needs. Study findings will help to develop disaster risk management strategies to promote safe travel, improve urban planning for flood resilience and guide post-disaster recovery efforts.

Highlights

• Public perceptions vary in flooding scenarios influenced by familiarity with the route, available information, and road conditions.

• Travel decisions during floods are sensitive to various flooding situations and are influenced by sociodemographic factors and trip purposes.

• The need for a companion and concerns about traffic congestion are heightened in flood-affected areas.

• Real-time information and public transport availability can reduce perceived risks and improve mobility decisions.

• Findings emphasize the need for tailored risk mitigation strategies to enhance flood-resilient transportation systems.

1. Introduction

Countries worldwide are facing extreme flooding events, resulting in the loss of human lives and urban infrastructure [1]. The intensity of extreme weather events, e.g. extreme rainfall due to climate change, has become a frequent phenomenon in urban areas in Asia [2–5].

Floods affect the lives of people in different ways. Livelihoods are affected, crops are destroyed and the usual patterns of life are disrupted. In extreme cases, floods lead to massive dislocations and even large-scale deaths [6]. Extreme weather events such as urban flooding disrupt transportation networks with delays, diversions and partial or full closure of roads and, thereby, influence human movement patterns [5]. Therefore, studying the change in mobility behaviours due to extreme weather conditions is important to inform users about the right time to act responsibly.

According to the Global Climate Risk Index, European Union data shows that Pakistan is responsible for less than 1% of the world's global warming gases, yet it is the eighth most vulnerable nation to the climate crisis [7]. Recent extreme rainfall in summer 2022 has pushed Pakistan to its worst flooding in at least a decade [8]. Around 33 million people, or 13% of the Pakistan population, have been affected by the floods [9]. The floods led to economic losses, with Sindh bearing 70% of the damages and significant impacts in other regions like Punjab, Balochistan and Khyber Pakhtunkhwa. Over 2.3 million homes were damaged, crops were destroyed and livestock were lost, pushing 8 million people into poverty. The increased frequency of floods, exacerbated by climate change, has led to repeated displacement and substantial socio-economic costs, particularly for vulnerable groups like women, children, the elderly and people with disabilities [10, 11]. The devastating impacts extended to the transportation sector, causing widespread disruption to the road network. A total of 55, 970 roads across all categories (major roads, urban roads and township roads) were affected. Sindh Province suffered the most extensive damage, with over 3,350 kilometres of affected roads. Punjab and Baluchistan also faced significant consequences, with reported road damage exceeding 1 ,300 kilometres in each province [12].

Several studies have focused on the overall devastation caused by urban floods, but only a few have considered the relationship between changes in travel decision behaviour and extreme weather situations [13, 14]. This study aims to explore the public perceptions and changes in travel decision behaviour due to devastating floods in Pakistan. The study used a questionnaire survey method to collect data about travel decisions in two travelling scenarios during flooding, which included familiarity with the route, route conditions and shelter camps on the roadside. This study highlights significant factors affecting public perceptions and travel decision behaviour during various flooding situations through the random parameter probit modelling.

2. Literature review

The disruption of existing travel behaviour patterns due to extreme events, like pandemics, strikes, transport system maintenance and extreme weather situations, has been widely discussed in the transportation literature. An anecdotal study using a questionnaire survey from Pakistan revealed that people's travel behaviour has been affected by COVID-19. Besides the perceived risk of getting infected by COVID-19, sociodemographic details of the travellers were found relevant in predicting their mode choice and travel behaviour change [15]. Paul et al. [16] further confirmed the relationship between trip frequencies, sociodemographic characteristics, mobility patterns and severity of the pandemic through a thorough review of past studies. Events like public transport strikes also bring change in travel behaviour, as individuals are forced to change from habit-driven behaviour to rational behaviour, which results in more traffic congestion and road blockage [17]. A study from Canada investigated the travel behavioural effects of the reduction in the traffic capacity resulting from on-road construction causing the closure of multiple roads. Using a multinomial logit model on the data collected through a questionnaire survey, this study highlighted that travellers were more likely to shift to an alternative route when provided with timely information. Further, people unfamiliar with alternative routes were found to be reluctant to divert [18]. Another study about evacuation behaviour during extreme hurricanes used a random parameter binary logit model to capture individuals' expectations through forecasts or individual beliefs [19]. These studies consistently show that extreme events disrupt usual travel patterns.

While prior research explored how typical weather variations like the cold, warm temperatures, snow, strong winds, rain and fog influence travel behaviour [20], the focus has shifted to extreme weather events caused by climate change. Floods, specifically, have been the subject of numerous studies across various countries, highlighting their disruptive impact on transportation. Aftab et al. [21] used a binary probit model to investigate how households in flood-prone areas of Pakistan adapt to climate change and flooding. The study identified that access to extension services, off-farm income opportunities, past flood duration, distance to rivers and post-flood support are key factors influencing adaptation choices. In Wuhan, China, a study constructed the flood modelling module and the commute simulation module and found a link between rainfall intensity and commuter disruption using location-based service data [22]. Similarly, another research study from Guangdong Province, China, showed decreased road connectivity with increased rainfall by using hydrological simulation [23]. While these studies focused on objective measures of flood impact, other studies in China explored residents’ flood risk perceptions through surveys. These studies in Jiaozuo and Nanjing identified factors like past flood experience and trust in flood control measures as influencing risk perception, which in turn can motivate protective actions [24, 25]. Research in Bangladesh found that floods significantly disrupt transportation infrastructure, leading to road closures, issues with public transport and increased travel times and costs. Travellers respond by changing their behaviour, such as shifting to private vehicles or choosing alternative routes [26, 27]. Similar findings emerged from India, where studies highlighted the increasing trend of urban floods disrupting transportation systems, pointing out the need for flood resilience measures in transportation planning to mitigate these negative effects [28]. A study from the United Kingdom proposed a relationship between floodwater depth and vehicle speed to improve the accuracy of flood-induced delay estimates in transport models [29]. In the United States, a study explored an increase in vehicle hour delays due to road closures indicating travel disruptions caused by urban floods, evaluated by applying the Metropolitan Planning Organization's four-step model [30]. Prior studies have highlighted various factors affecting travel disruptions caused by floods; yet still, travel decision behaviour itself has not been explicitly discussed in the context of developing countries.

Frequent floods may influence public perceptions of various risks associated with travel during floods [25, 31, 32]. Understanding these perceptions is crucial for developing effective strategies to encourage safe and informed decision-making. In the context of urban floods and transportation, the perceived probability of a crash, perceived distraction, perceived security, perceived need for a companion and perceived possibility of getting stuck in traffic congestion may affect people's travel decisions during floods.

The perceived probability of a crash refers to the public perception that they will get involved in a traffic crash on urban roads [33]. The deteriorated road conditions and reduced visibility during floods can exacerbate the perceived risk of crashes [34], particularly for those using vulnerable modes of transportation like bicycles or motorcycles. Perceived security refers to the feelings of individuals that they are secure from an external threat [35]. The presence of refugees living on roadsides during floods can raise security concerns, leading to route avoidance. Perceived distraction while travelling, especially during flooding when the roadside is full of shelter camps, can have detrimental effects on drivers, potentially leading to accidents [36]. Past studies have also explored that human activities on the roadside may cause drivers distractions, resulting in crashes [36]. In some other studies, it was highlighted that people need companionship when driving during floods [37]. Possibly, companions while travelling during floods reduce risks and help people have peace of mind when reaching their destinations. The fear of getting stuck in gridlocked traffic due to flood-related disruptions is a significant deterrent to travel. Congestion caused by urban floods negatively affects travelling efficiency, safety and effectiveness [38, 39].

Prior studies confirmed the severity of flood impacts on transportation, resulting in a change in travel behaviour. Individuals may choose to stay home, work remotely or combine trips to avoid the frustration and potential safety hazards associated with congested roads. Likewise, travel decisions are affected by people's perception of getting stuck in traffic congestion [40].

3. Methodology

3.1. Questionnaire survey

The 2022 floods in Pakistan caused significant damage to roads [10, 41]. The floods led to the destruction of infrastructure, including roads, bridges and transportation networks. A questionnaire survey was developed to explore the public perceptions of risks associated with urban floods and changes in travel behaviours in Pakistan. The questionnaire was designed to have three sections. The first section contains sociodemographic details, including sex, age, income, education, province, usual mode of transportation, travel risk level and past experience with alternative routes in normal conditions. The second section includes scenario one, as shown in Fig. 1, which was about the travel decision when the route was completely damaged due to floods, and an unfamiliar alternative route was available with shelter camps on the roadside. In contrast, the third section includes scenario two, as shown in Fig. 2, which depicts the travel decisions when the usual route is partially damaged due to flooding, with shelter camps on the roadside and no alternative route available to reach the destination. For both sections, five perceptions-based questions were asked, including the perceived probability of a crash, perceived distraction due to shelter camps, perceived security, perceived need for a companion and perceived possibility of getting stuck in traffic congestion. The responses were recorded on a three-point Likert scale (1 = Unlikely, 2 = Neutral, 3 = Likely), and were then constructed as separate binary dependent variables for each model. For example, if a person's perception of crash probability is ‘likely, neutral’ in scenario one, and that in scenario two is ‘unlikely’, then the person thinks scenario one has a higher probability of occurring, and the dependent variable value = 1. Meanwhile, if a person's perception of crash probability is ‘unlikely’ in scenario one, and that in scenario two is ‘likely, neutral’, then the person thinks scenario one has a lower probability of occurring, and the dependent variable value = 0.

The questionnaire was used to collect data on both online and offline channels. The online questionnaire was developed using Microsoft Forms and distributed on social media, including Facebook, WhatsApp and LinkedIn. A team of five surveyors was hired to conduct an on-site questionnaire survey. A total of 1,092 responses were collected. After data cleaning, and the removal of incomplete responses and those respondents who responded that they would not travel in a flooding situation, only 526 valid responses were extracted for analysis.

3.2. Analytical methods

This study involved two scenarios. The Pearson Chi-square test of independence was employed to elucidate the influence of responses from scenario one on the responses of scenario two. This helps examine whether there is a significant association between the two scenarios, potentially shedding light on how exposure to one scenario may affect perceptions in another scenario. For the Chi-square test of independence, IBM SPSS software was used. Mathematically, this can be written as follows:

where $n_{\mathrm{e}}=\frac{n_{\mathrm{c}}{n}_{\mathrm{r}}}{n}$⁠; x² represents the Chi-square statistic; n_o represents observed frequency; and n_e represents expected frequency. The sum is taken over all categories in the cross-tabulation table. n_c is the total frequency of columns; n_r is the total frequency of rows; and n is the sum of all observations.

Studies related to extreme events have used various methodological approaches, such as the negative binomial model, structural equation model, hydrological simulation and discrete choice model to understand travel behaviour and decision-making processes [15, 18, 42–47]. Prior studies about extreme weather and flooding events were found to use the discrete choice model. This model is particularly useful in understanding decision-making processes related to travel during floods and public perceptions for adaptation decisions, evacuation and mitigational strategies [19, 21, 43, 44, 48]. The most commonly used discrete choice models are the probit model and the logit model. The logit model assumes logistic distribution for the error component and is easier to estimate, whereas the probit model has more statistical universality and rationality because of its normal distribution assumption [49].

In line with the aim of this study to explore the relationship between sociodemographic variables, perceptions and travel behaviours, a binary probit model was used. To further accommodate the heterogeneity in the respondents’ responses, a random parameter model was used to allow parameters to vary randomly and to capture the diversity of perceptions and behaviours within a population [50].

Mathematically, it can be written as shown in Equation (2).

This Equation estimates the probability of outcome one occurring for observation n, where ε_1n and ε_2n are normally distributed with a mean equal to zero, variances ${\sigma }_1^2$ and ${\sigma }_2^2$⁠, respectively, and the covariance is σ₁₂ and ε_2n−ε_1n is normally distributed with mean zero and variance ${\sigma }_1^2+{\sigma }_2^2-{\sigma }_{12}$⁠. The resulting cumulative normal function is

The likelihood function is defined as follows:

where N is the total number of observations and δ_in is defined as being equal to one if the observed discrete outcome for observation n is i and zero otherwise.

The log-likelihood function is presented as follows:

In this research, the random parameters are hypothesised to obey normal distribution [51, 52]. All the sociodemographic variables were added into the model. The model estimation is carried out in the statistical package NLOGIT5 [53].

4. Results and discussion

4.1. Respondents’ details

In this study, males comprised 65% of the respondents, whereas females comprised the remaining 34%, as shown in Table 1. Research suggests that females are significantly more susceptible to disasters due to factors like limited information, decision-making power, skills, training and mobility, compounded by societal norms and gender roles [10]. However, the number of females in Pakistan who often drive cars or ride motorbikes is not large [54], and therefore the imbalance in gender distribution is sufficient to derive useful findings from this study [55]. In terms of education level, around 64% of respondents were lower than undergraduate and 35% had an undergraduate or higher education level.

Around 49.81% of the respondents reported < 4.9 Lakhs PKR annual household income. Regarding perceived risks from urban flooding, only 20.72% perceived low risk of travelling during the flood, 42.77% perceived moderate risk, whereas 36.50% reported extreme risks. 14.56 million people from Sindh were affected [56]. Sindh was the most affected region, as around 90% of respondents belong to Sindh Province. More than half of the respondents (61.02%) used alternative routes for reasons other than floods, like avoiding traffic congestion and poor road conditions. In the past, a study has highlighted that it is feasible to change the route in response to road disruptions for people in Pakistan [57].

5. Dependence of changes in perceptions on the given scenarios

The Pearson Chi-square test of independence was performed to identify whether there is an association between the perceptions from the two scenarios of this study. Scenario one was about the travel decision when the route was completely damaged due to floods, and an unfamiliar alternative route was available with shelter camps on the roadside. In contrast, scenario two depicts the travel decisions when the usual route is partially damaged due to flooding, with shelter camps on the roadside and no alternative route to reach the destination.

The results show statistically significant associations (p-value < 0.001) for all perceptions except the possibility of getting stuck in traffic. This implies that travellers’ perceptions are significantly influenced by the scenarios they face during floods.

Results indicate that travellers were much less likely to perceive a high crash risk on the familiar route that is partially damaged with shelter camps along the roadside (scenario two) compared to the completely damaged existing route with an unfamiliar alternative route (scenario one). Similarly, respondents were significantly less distracted on the familiar route that is partially damaged route (scenario two). Regarding perceived security risks, travellers perceived higher security risks on the partially damaged route (scenario two) than on the unfamiliar alternative route (scenario one). This could be because of increased vulnerability and limited escape options on familiar routes. Perceived security risk refers in this study to the risk of being the victim of crime. It is possible that travellers perceived themselves as more vulnerable to crime when driving through limited lighting and uneven surfaces on partially damaged routes. For example, if travellers encounter criminal situations such as vehicle theft, though they are still familiar with escape routes, these escape routes could be blocked or compromised. The need for a companion was significantly higher for the completely damaged existing route with an unfamiliar alternative route (scenario one). However, the possibility of getting stuck in traffic (p-value = 0.696) showed no significant association with the type of route damage. Travellers perceived a similar risk of traffic congestion in both scenarios. Findings suggest that while travelling during the flood, people perceive a completely damaged existing route with an unfamiliar alternative route as riskier and more distracting than a familiar, partially damaged existing route, as shown in Table 2.

6. Changes in travel behaviour for trip purposes

A random parameter probit model was developed to understand the changes in people's travel for trip purposes, including commute/business/education and visiting relatives or friends/leisure, when their usual route is completely damaged or blocked due to flooding, as shown in Table 3 and Table 4.

For commute or business trips, those aware of floods but without past experience of travelling during floods seem more hesitant to use the alternative route, possibly due to unfamiliarity with the route. Similarly, people who have travelled in the past during floods for medical or emergency purposes were less likely to choose an alternative route to reach their destinations. This suggests that past experience of travelling during floods places barriers and that people are likely to compromise the urgency associated with medical trips during floods.

Results also indicate that males were more likely to choose the alternative route for commute or education trips, which aligns with existing research suggesting males are generally less risk averse in public spaces [58]. The responses from males are normally distributed with a mean of 1.0041 and a standard deviation of 2.2421, suggesting that 67.28% of the male respondents are more likely to use an alternative route. Similarly, people holding undergraduate or higher degrees were more likely to travel for commute or business purposes during a flood. The variable also yields a normal distribution, suggesting that 59.43% of the respondents with a higher education level are more likely to use an alternative route. This could be due to a higher need for job continuity or a greater comfort level navigating unfamiliar situations associated with higher education.

In the Table 4, modelling results for the trips to social and leisure visits are presented. Compared to older adults (reference group), middle-aged respondents were less likely to use the alternative route to travel to visit relatives or friends and leisure. This suggests a potential for routine dependence in middle age. Disruption caused by the unfamiliar route might be less appealing to this group.

The results show that male respondents, compared to females, were less likely to use an alternative route for social and leisure visits. A possible rationale for this finding aligns with existing research suggesting that women might be more cautious due to flood risks while prioritizing household care [59].

Those who had not travelled during floods in the past, or who had travelled for essential reasons (medical or family emergency), were less likely to take the alternative route. This suggests that past experience of travelling during floods negatively affects people's decision to travel regardless of the urgency of the trip purposes.

Interestingly, individuals who perceived the travel risk as high while travelling during floods were more likely to use the alternative route to visit relatives. This variable shows that 72% of respondents with high risk perception use an alternative route. This seemingly counterintuitive finding could be explained by a sense of responsibility to check on loved ones during a crisis, outweighing personal risk concerns.

Regarding education level, people with higher education were more likely to use the alternative route for social and leisure visits. The parameter coefficient of ‘undergraduate and higher level’ follows a normal distribution with a mean of 0.4384 and a standard deviation 0.5676. From the distribution parameters as mentioned above, it can be concluded that higher education level respondents are likely to use an alternative route by 78%. This might be due to a greater comfort level navigating unfamiliar situations or a better ability to access and interpret flood information through media sources.

7. Factors influencing travel decision behaviour during floods

This study employed a random parameter probit model to analyse how sociodemographic factors influence people's perceptions during floods. Tables 5–9 present the results for travel decision behaviour during floods, focusing on how these factors affect perceived probability of crash, perceived security, perceived distraction due to shelter camps, perceived need of a companion and perceived possibility of getting stuck in traffic.

In the results in Table 8, we can see that those of young age (< 26 years) are more likely to be in need of a companion. This aligns with studies finding that young people are more exposed to eco-anxiety, bear a high burden of anxiety issues and overcome anxiety in drastic situations like floods by needing a companion to reduce their anxiety [60–62]. In contrast, perceptions about being a victim of a crime remained less among young people because of their risk-taking behaviour at that age. Whereas perceptions about becoming a victim of a crime and being distracted due to shelter camps were found to be high in middle-aged respondents, the perceived need for a companion was relatively low. This is possibly because these people already feel less secure and more distracted, and having a family member or a friend as a companion could pressure the respondents more during travelling through an unfamiliar alternative route in extreme events. It can further be seen that the respondents who have not travelled through an alternative route in the past perceive a high probability of crash and a high chance of becoming a victim, are highly distracted due to shelter camps and are less likely to be in need of a companion when travelling in such severe situations; this supports the idea that familiarity with routes influences perceptions; see Tables 5–8. The results also show those respondents whose usual mode of transportation is the public bus or motorcycle perceive a high probability of crash, perceive themselves to be less secure and are more distracted [63]. Respondents who use three-wheelers (rickshaw/Qingqi, motorcycle-based three-wheeler) are more exposed to daily traffic congestion and frustrated traffic, which could possibly make them perceive less probability of a crash and less distraction while travelling in this situation. In the results, it is specified that those respondents who travelled during the floods in the past for medical or emergency purposes perceive a high chance of crash and of becoming a victim of crime, and those who travel for commute perceive that they might get stuck in traffic based on their past trip experiences. Risk perception is more complex to understand as it varies in every individual. The results show that respondents who perceive a low risk during travelling in flood events feel that they are high likely to become a victim of a crime or get stuck in traffic but that there is a low probability of crash and they are less likely to need a companion [64]. In Tables 5, 7 and 9, results show that the respondents who decided to continue to travel for commute(job)/business/education when the existing road was partially damaged perceive a high probability of crash, while being less distracted and perceiving less of a likelihood of getting stuck in traffic while using an alternative route. The respondents who decided to abandon the trip to visit relatives or friends/leisure when the existing route was partially damaged perceive a high probability of crash and of becoming a victim while using an unfamiliar alternative route, and having these perceptions could possibly be the reason for cancelling the trips in extreme events. The respondents who decided to use an alternative route to visit relatives when the existing route was completely damaged perceive less probability of a crash and are less likely to get stuck in traffic but perceive that they might be a victim of a crime while using an alternative route in flood events. High-income respondents (≥ 4.9 Lakhs PKR) perceive they are less likely to be distracted and to be in need of a companion due to their access to advanced resources but are more likely to get stuck in traffic due the conditions of the situation [65–68]. Higher education-level respondents perceive more distraction and are less likely to get stuck in traffic.

Tables 5 to 9 estimated 12 random parameters for the perceived probability of a crash, perceived security, perceived distraction, perceived need of a companion and perceived possibility of getting stuck in traffic. In Model 3, perceived probability of crash, it can be seen that the variables three-wheelers (mean = −9.1797, s.d. = 27.531), non-Sindh (mean = 16.739, s.d. = 19.9471), SC1 visiting relatives or friends/leisure (mean = 3.1744, s.d. = 12.1246) and SC2 visiting relatives or friends/leisure (mean = −1.2171, s.d. = 7.5143) were identified as random parameters. Notably, three-wheelers demonstrated a lower perceived likelihood of a crash by 63.05% when using an alternative route. Other random parameters can be interpreted in a similar way. For Model 4, perceived security, Table 6 shows the variables named as moderate risk level (mean = 0.4852, s.d. = 3.6736) and SC1 visiting relatives or friends/leisure (mean = 1.1272, s.d. = 8.5643) as significant random parameters. In Table 7, the perceived distraction model, SC1 commute(job)/business/education (mean = −1.3587, s.d. = 6.4043) and use of an alternative route in a flooding event (mean = 1.0472, s.d. = 3.6841) emerged as significant random parameters. Model 6 on the perceived need for a companion focuses on the variables, low risk level (mean = −6.8471, s.d. = 10.068) and use of an alternative route in a flooding event (mean = −1.9124, s.d. = 3.5745) that were identified as random parameters in Table 8. The last model about perceiving the possibility of getting stuck in traffic had higher education level (mean = −0.7909, s.d. = 2.2188) and SC1 commute(job)/business/education (mean = −0.7189, s.d. = 2.2188) variables that were identified as significant random parameters, as shown in Table 9.

8. Conclusions and implications

This study investigated how sociodemographic factors influence travel decision behaviours during floods, focusing on public perceptions of crash risk, security, distraction, need for a companion and getting stuck in traffic. The study's findings suggest that the decision to travel is sensitive to the severity of road damage and the provision of alternative roads, highlighting the importance of infrastructure resilience and improved flood management strategies. Policymakers should prioritize investments in infrastructure development. More funds should be used to make existing roads more resistant to flood damage through regular maintenance and by incorporating climate change adaptation measures into infrastructure planning processes.

On the other hand, there is a need to develop clear guidelines for establishing safe and well-managed shelter camps during floods, addressing security concerns raised by travellers. Targeted interventions are needed to assist low-income individuals and marginalized communities disproportionately affected by disruptions. Disaster management authorities should consider establishing temporary shelters in flood-prone areas and provide access to alternative transportation options to meet basic travel needs.

Emergency response plans should integrate considerations for women and provide proper lighting and security measures on alternative routes. Public awareness programmes should be initiated to promote flood risk awareness through key aspects, including disseminating information about the severity of damage on partially damaged routes, alternative routes, emergency procedures and the importance of preparedness. Public awareness and education programmes should also incorporate travellers’ understanding of flood risks and promote adaptive behaviours. Besides this, there is a growing need for continued investment in research and data collection to gain a deeper understanding of evolving flood risks and their impacts on travellers, transportation systems and overall community well-being. By incorporating these findings and recommendations into disaster risk management strategies, urban planners and policymakers can build more resilient transportation systems, minimizing the social and economic impacts of floods on communities.

While valuable, this study acknowledges its limitations and suggests future research directions. The study relied on questionnaire survey data conducted in Pakistan. A broader geographical scope, spatially distributed responses and diverse data collection methods like interviews or focus groups can be performed in the future to provide a more comprehensive picture of changes in travel behaviour across the country. This study included only those respondents who decided to travel regardless of the uncertainty of the flood situation. Future research can include the factors that affect people's decisions not to travel during floods, including the long-term psychological impacts of floods. Further research could delve deeper into the specific needs and risk perceptions of women and other vulnerable groups during floods to develop targeted interventions.

Acknowledgements

The author would like to extend sincere gratitude to Ms. Sana and the students from the City and Regional Planning Department at Mehran University of Engineering and Technology, Jamshoro, Pakistan, for their invaluable assistance in conducting the questionnaire survey.

Funding

This research was funded by the Natural Science Foundation of China (Grant No. 72471249); and Natural Science Foundation of Changsha (Grant No. kq2402221).

Conflict of interest statement. The author declares no conflict of interest.

References