Examining the Factors Influencing Tsunami Evacuation Action Selection in Thailand: A Comprehensive Study Involving Local Residents, Non-Local Workers, and Travelers

Abstract

Tsunamis are a substantial natural threat in Thailand, as evidenced by the 2004 Indian Ocean tsunami. Effective evacuation is vital to reduce casualties and property damage. However, despite improved warning systems, high death tolls still occur, indicating complex evacuation behavior influenced by various factors. This study examines these factors among diverse groups in Phuket and Phang Nga, Thailand. A survey of 1000 locals, non-local workers, and travelers assesses threat and coping appraisals, past tsunami experiences, gender, age, and tsunami evacuation intention and action selection. Partial Least Squares Structural Equation Modeling (PLS-SEM) was used to analyze the data based on the hypotheses related to the Protection Motivation Theory (PMT). The results of the analyses show that threat and coping appraisals significantly predict tsunami evacuation intention, and gender influences threat perception related to evacuation. Variations among respondent types emphasize the need for tailored disaster preparedness and response strategies. This study offers crucial insights for policymakers, emergency responders, and disaster management stakeholders, underlining the significance of further research into the intricate interplay of individual and contextual factors shaping tsunami evacuation behavior.

1. Introduction

How can we effectively prepare for the possibility of a tsunami in the future to reduce the number of casualties? The 2004 tsunami in the Indian Ocean was generated by an earthquake that occurred at a depth of 30 km and around 160 km from the western coast of northern Sumatra [1]. This natural catastrophe was brought about by the earthquake. In addition to Thailand, Bangladesh, India, Malaysia, Myanmar, Sri Lanka, and the Maldives were also devastated by the tsunami [2,3]. There were no early warning signs or a tsunami warning system in place in the Indian Ocean to detect tsunamis or advise the population living around the Ring of Fire. This rendered every victim entirely unprepared for the assault. As a direct result, each and every victim was taken completely by surprise. It is believed that 8212 people perished in Thailand as a direct result of this event. In addition, 8457 people were injured, 2817 people were reported missing, and 7000 individuals were displaced as a result of the tragedy [4].

Since the tragedy, disaster warning systems have greatly improved over the years [5,6]. However, disaster events with a high death toll still occur despite early warnings and evacuation orders. For example, in 2007, a category 4 storm named Cyclone Sidr made landfall on Bangladesh’s southwest coast. Thousands of people remained in their homes despite early storm warnings and evacuation orders for coastal populations. A study was conducted on the occurrence and found that most non-evacuees were able to evacuate but did not intend to [7]. The research highlighted the importance of the motivation to evacuate for the residents in disaster-prone areas, which include tsunamis. Furthermore, the 2018 Palu tsunami in Indonesia was another example of why the disaster warning system needs to be improved. Sujatmiko and Okumura [8] mentioned that the people in the disaster areas did not immediately evacuate during the initial warning. On the other hand, the evacuees who received the late warning tend to be more reactive than those who received the initial warning.

As a result of the 2004 Indian Ocean tsunami, a substantial quantity of scholarly research has been conducted on tsunami evacuation. These studies have proven quite beneficial. The primary goal of these studies is to determine how long it would take to evacuate a location under a range of scenarios and use a variety of escape routes. As this is a crucial subject component, these works also calculate how long an evacuation would take under a range of scenarios and routes. However, the determination of factors related to tsunami evacuation intentions and tsunami evacuation action selection in Thailand has yet to be explored. Thus, this study intends to examine the factors associated with tsunami evacuation intentions and tsunami evacuation action selection in Thailand. The responses were collected from the questionnaires gathered in Phuket and Phang Nga Provinces focused on tsunami evacuation.

2. Materials and Methods

This section provides a theoretical framework and context for studying factors influencing tsunami evacuation behavior. It includes an overview of the Protection Motivation Theory (PMT), a description of the Partial Least Squares Structural Equation Modeling (PLS-SEM) used to analyze the data, and the proposed model and hypotheses. The literature review provides a comprehensive understanding of the research problem and supports the proposed model and hypotheses.

2.1. Protection Motivation Theory

PMT was developed by Rogers [9] as a framework for studying the impact of fear appeals. PMT has its roots in the early research on the persuasive power of fear appeals, which focused on the conditions under which fear appeals can influence attitudes and behavior [10]. According to the PMT, fear appeals cause specific cognitive reactions. Threat appraisal and coping appraisals are two independent appraisal processes that can be triggered by several environmental (e.g., fear appeals) and intrapersonal (e.g., personality) sources of information [10].

Threat appraisal concentrates on the origin of the threat and factors that increase or decrease the probability of maladaptive responses (e.g., avoidance, denial, wishful thinking). Maladaptive responses are thought to be inhibited by how seriously people take the threat and how exposed they feel to it [11]. Residents living in Phuket, for instance, could think about how dangerous a tsunami is and how likely it is that they would experience a tsunami. Based on the reviewed literature, it can be proposed that the threat appraisal construct, which includes perceived vulnerability and perceived severity, is related to tsunami evacuation intention. Tang and Feng [11], Kautondokwa et al. [12], Farooq et al. [13], and Bauer and Bernroider [14] indicated that threat appraisal has a significant positive relationship with behavioral intention.

Coping appraisal concentrates on the coping skills that the individual can use to deal with the danger and the factors that either increase or decrease the likelihood of an adaptive response, as follows behavioral advice. The chance of an adaptive response is increased by both the perception that the suggested behavior will be successful in lessening the danger (i.e., response efficacy) and the belief in one’s capability to execute it (i.e., self-efficacy). In the PMT framework, coping appraisal is considered to be a vital component of the motivational process that drives protective behavior. The theory suggests that individuals are more likely to engage in protective behaviors when they believe that they have the necessary resources to cope with the threat and that those resources effectively reduce the likelihood or severity of the threat [10]. According to Tang and Feng [11], Zheng et al. [15], and Bauer and Bernroider [14], for coping appraisal, they found response efficacy and self-efficacy positively influenced behavioral intention.

The definition of preparedness provided by Paton [16] encompasses protective behavior as a critical aspect of disaster response. In the context of tsunami evacuation, protective behavior includes actions such as seeking higher ground, following evacuation routes, and reaching designated safe zones.

Figure 1 shows that individuals are more likely to engage in protective behavior if they perceive a threat to their well-being and believe they can effectively reduce that threat [9]. Several studies have shown a positive relationship between protection motivation and protective behavior in the context of disaster preparedness and response, including in the case of tsunami evacuation [11,17]. Therefore, it can be hypothesized that there is a relationship between protection motivation and tsunami protective behavior during evacuation (as shown in Figure 1). The study on factors affecting tsunami evacuation behavior found that intention to evacuate has a direct relationship with action selection [18]. The research proposes a model in which the intention to evacuate is a key factor in determining action selection during a tsunami evacuation.

2.2. Potential Factors for PMT

Gender is one of the demographic factors that can affect how individuals perceive and respond to threats. Research has shown that women have a higher perception of vulnerability than men in different situations, such as living in multi-hazard areas or encountering a dangerous pandemic. Several studies have reported that women usually perceive threats more than men in various situations. Nino et al. [19], Rana et al. [20], and De Coninck et al. [21] found that women have a higher perception of threat compared to men.

Age has been identified as an important demographic factor influencing individuals’ perceptions of threats in the context of natural disasters. As people age, they may experience changes in physical and cognitive abilities, as well as changes in their social and economic circumstances that can affect how they perceive and respond to threats. Research has shown that older adults tend to perceive higher levels of threat in response to natural disasters, such as hurricanes and earthquakes, compared to younger adults [21,22]. These findings suggest that age may play a role in shaping individuals’ threat appraisals, which in turn can influence their protective behaviors and intentions in response to various hazards. Similarly, education level can also affect people’s understanding and coping strategies during a disaster circumstance [23,24]. Furthermore, Kyne et al. also include the level of income factor to prove whether it can affect the willingness to prepare for a disaster. Faryabi et al. [25] summarized that higher income can increase preparedness levels.

For the factors related to health issues, health status can affect risk perception, physical evacuation capability, and preparedness. Elderly individuals aged 80 and above may encounter mobility challenges, necessitating alternative transportation or escape routes [26]. Furthermore, children and adolescents may require additional support during evacuation [27,28].

Direct experience with natural disasters can profoundly impact individuals’ perceptions and behaviors toward future events. When individuals have experience with a disaster, they are more likely to perceive the threat as higher and be more willing to evacuate when a disaster occurs. This has been demonstrated by previous studies such as Trumbo et al. [29] and Demuth et al. [30], which found that individuals with direct experience with a tsunami generally have higher threat appraisals. Similarly, Lazo et al. [31] and Morss et al. [32] discovered that individuals with previous evacuation experience tend to have higher intentions to evacuate. Therefore, it is hypothesized that the past tsunami experiences of respondents have a positive relationship with both threat appraisal and tsunami evacuation intention.

Perception and awareness of disaster can be considered factors that relate to individual evacuation decisions [27,33,34,35,36,37,38]. Thus, people who can perceive high risk and have awareness of tsunamis or earthquakes are more likely to evacuate themselves [39,40,41]. Vertical evacuation structures and early warning systems can also increase awareness and evacuation necessity [42,43].

The factors related to effective communication and risk management are vital for informed evacuation decisions [38,44]. Moreover, limited time in sudden disasters necessitates efficient early warning systems, and accessible evacuation structures can also affect the tsunami evacuation intention [44,45,46,47].

2.3. Partial Least Square Structural Equation Modeling (PLS-SEM)

Structural Equation Modeling (SEM) is a second-generation multivariate data analysis method that originates from the work of Wright [48] and can overcome the limitations of first-generation techniques. SEM allows for the modeling of numerous independent and dependent components simultaneously. It is frequently utilized in marketing research due to its theoretical support for linear and additive causal models [49,50,51].

A SEM has two sub-models: an inner model that describes the relationships between independent and dependent latent variables and an outer model that specifies the relationships between latent variables and their observable indicators. A variable can be either exogenous or endogenous in SEM. No path arrows are heading toward an exogenous variable, and all of them point outwards. In contrast, an endogenous variable represents the consequences of another variable and has at least one path leading to it [52]. For modeling, PLS is a method for SEM introduced by Wold [53], which intends to increase the explained variance of the dependent latent constructs [54]. PLS-SEM requires no assumptions about the data distribution [55]. And it turns out to be a good alternative for CB-SEM in the following circumstances [56,57,58]. The first circumstance is that it has a small sample size. Second, there is little available theory for applications. Third, predictive accuracy is crucial. Lastly, accurate model specifications cannot be guaranteed.

2.4. Conceptual Model

Figure 2 illustrates the conceptual model derived from the hypotheses presented in Section 2.1 and Section 2.2. Although numerous factors can impact both tsunami evacuation intention and action selection, our study focuses on easily understandable and retrievable potential demographic-related factors. These parameters include gender, age, and past tsunami experience. Moreover, to account for possible differences in the data group of the model, a control variable is added for different demographics of respondents, which is stated as respondent type since it can be a meaningful approach for research and analysis.

Figure 2 illustrates the model developed in this study, which integrates the constructs of threat appraisal, coping appraisal, and tsunami evacuation intention and behavior. It aims to provide a comprehensive understanding of the factors related to individuals’ tsunami evacuation. Specifically, this study examines the influence of age and gender on threat appraisal and past tsunami experience on threat appraisal and evacuation intention, as well as the relationship between evacuation intention and action selection. Additionally, the model adds a control variable for different respondent types. Separating data into three groups—residents, non-local workers, and travelers—might be beneficial for analyzing and understanding their behavior in the case of a disaster. In this research, we categorize our respondents into three distinct groups:

• Locals: Permanent residents or have long-term ties to the study area, which comprises Phuket and Phang Nga;
• Non-Local Workers: Individuals who are visiting the area on a temporary basis primarily for work-related purposes;
• Travelers: Individuals who are present in the study area for leisure and recreational activities.

2.5. Methodology

The methodology used in this study encompasses several key elements, including questionnaire development, data collection, data preparation, proposed, and analysis approach. These elements were chosen to ensure that the study was conducted rigorously and systematically, aiming to produce valid and reliable results.

2.5.1. Questionnaire Development

A questionnaire was created based on the model, which is shown in Figure 2. The questionnaire contained 26 items that were used to assess tsunami evacuation behavior concerning four latent variables: threat appraisal, coping appraisal, tsunami evacuation intention, and tsunami evacuation action selection.

Table 1. Latent variables and references of the questionnaire items’ development.

Latent Variable	Source(s)
Threat Appraisal	Espina and Teng-Calleja [59] Shoji, Takafuji, and Harada [60]
Coping Appraisal	Espina and Teng-Calleja [59] Shoji, Takafuji, and Harada [60]
Tsunami Evacuation Intention	Najafi et al. [61] Ong et al. [62]
Tsunami Evacuation Action Selection	Hall et al. [39]

shows references to questionnaire used in this study, which was further adapted to match the subscales that were examined in the study. Each question in the questionnaire is shown in Appendix A.

Respondents were asked to rank each item on a 5-point Likert scale ranging from 1 (“lowest level of understanding” or “lowest level of agreeing” depending on the question) to 5 (“highest level of understanding” or “highest level of agreeing”). The questionnaire was designed to be completed in 10 min by participants. Additionally, the data are gathered for demographic profiles, including gender (male/female/prefer not to say), age, tsunami experience (yes/no), and respondent type (local/non-local worker/traveler). During the survey process, the researchers briefly explained the purposes and instructions of the survey, which related to past experience with tsunamis based on either the survey areas or other related experiences. In addition, the respondents provided complete instructions for those who were really needed before doing the survey.

2.5.2. Data Collection

In the context of collecting data from the areas of Phuket and Phang Nga Provinces, where the exact population numbers are unknown, we opted to use Cochran’s formula to determine an appropriate sample size. With a significance level of 0.05, p (probability of success) is assumed to be 0.5, and the margin of error is set at 0.05 [63]. The equation can be shown as Equation (1). The term n represents the calculated sample size. Z is the critical; in this case, 1.96 was used as the selected critical value. Next, p is the probability of success, for which, in this case, 0.5 was used. Lastly, the term e is the margin of error.

$$n=\frac{Z^{2}p(1-p)}{e^2}$$

(1)

The initial calculation based on these assumptions indicated a sample size of 385 for each province. The sum of these two provinces can yield a total of 770 samples for the minimum sample size. However, to ensure the data collection process is robust and accounts for uncertainties, including non-responses or variations, it is advisable to add a 20% safety margin. Therefore, the final calculation of the sample size is 1000. Thus, in this study, the researchers collected data from Phuket and Phang Nga Provinces, Thailand, from August 2022 to January 2023, with a total of 1000 respondents. During the survey, our research team actively engaged with individuals passing through the study areas, which were damaged areas from the 2004 Indian Ocean. Our assistants also played a crucial role in ensuring respondents understood the questions and provided accurate answers. By collecting these data, this study can comprehensively understand the factors affecting threat appraisal, coping appraisal, tsunami evacuation intention, and tsunami evacuation action selection towards tsunami evacuation in the Phuket and Phang Nga Provinces.

2.5.3. Data Preparation

Before the data cleaning process, our researchers convert responses from the paper-based survey results to a digital format for further analysis. The data were prepared by cleaning incomplete data where responses were unclear or missing values. Additionally, due to the low number of responses from the gender category “prefer not to say”, this category was removed from the analysis. As a result, 35 data points were removed, leaving 965 data points for analysis. This step was necessary to ensure that the hypotheses were tested on a relevant and representative sample and to avoid any potential biases or errors in the results. Furthermore, the questions about demographics were collected in text choices. Thus, the data should be transformed to an integer number to be prepared for analysis software. The cleaned and filtered data can now be used for statistical analysis to test the hypotheses, and demographic characteristics can be summarized as shown in

Table 2. Respondents’ demographic statistics (n = 965).

Variables	Value	Number	Proportion (%)
Age	0–19 years old *	6 *	0.62% *
	20–29 years old	283	29.33%
	30–39 years old	354	36.68%
	40–49 years old	207	21.45%
	50–59 years old	102	10.57%
Gender	Male	343	35.54%
	Female	622	64.46%
Respondent Type	Local	203	21.04%
	Non-local worker	476	49.33%
	Traveler	286	29.64%
Tsunami experience	Yes	183	18.96
	No	782	81.04

* Not included in the analysis due to low sample size.

.

2.5.4. Proposed Research Model

As mentioned previously, the literature review consists of PMT and other potential factors for PMT, which were combined with the conceptual model. It can be hypothesized that the proposed model has seven hypotheses, which are as follows:

Hypothesis 1.

There is a positive relationship between threat appraisal and tsunami evacuation intention.

Hypothesis 2.

There is a positive relationship between coping appraisal and tsunami evacuation intention.

Hypothesis 3.

There is a positive relationship between tsunami evacuation intention and tsunami evacuation action selection.

Hypothesis 4.

Female respondents’ threat appraisal is higher than male respondents.

Hypothesis 5.

Respondents’ age has a positive relationship to threat appraisal.

Hypothesis 6.

Respondents’ past tsunami experiences have a positive relationship to threat appraisal.

Hypothesis 7.

Respondents’ past tsunami experiences have a positive relationship to tsunami evacuation intention.

Accordingly, Figure 2 illustrates the proposed model consisting of seven variables. The first two variables are threat appraisal and coping appraisal. The third variable, tsunami evacuation intention, measures the individual’s intention to evacuate during a tsunami warning. The fourth variable is gender, which measures the participant’s gender. The fifth variable, age, measures the age of the participant. The sixth variable, past tsunami experiences, measures whether the participant has experienced a tsunami in the past or not. Lastly, the seventh variable is tsunami evacuation action selection, which measures the choice of places and actions that individuals choose during a tsunami warning.

2.5.5. Analysis Approach

Based on the chosen analysis approach for the study on factors affecting tsunami evacuation behavior is PLS-SEM. This method was chosen due to the large number of variables in the constructed conceptual model and the reflective nature of the constructs. PLS-SEM allows for unbiased parameter estimates and modeling of measurement error variance and relies on a factor analytic measurement model. Additionally, it offers the flexibility to use reflective indicators. And a confirmatory approach will be used in the analysis to validate the hypotheses.

SmartPLS4 is appropriate for this study as it is a widely used software for PLS-SEM analysis. According to a review by Memon et al. [64] on PLS-SEM statistical programs, SmartPLS4 offers a user-friendly interface and is suitable for novices. It also provides advanced features for experienced users, including multi-group analysis and bootstrapping. Therefore, using SmartPLS4 can facilitate the analysis of the data in this study and increase the accuracy of the results.

3. Results

3.1. Demographic Statistics of Respondents

The demographic statistics of the respondents (n = 965) are summarized in Table 2. The age distribution reveals a diverse sample, with most participants being 20–39 years old. Gender distribution shows 35.54% males and 64.46% females. Respondents comprise locals (21.04%), non-local workers (49.33%), and travelers (29.64%). Additionally, 18.96% of participants reported having previous tsunami experience, while the majority (81.04%) had never encountered a tsunami. However, since there are only six samples (0.62%) for the age group between 10 and 19 years old, we ignored this group of samples to prevent the bias problem. Thus, there are only 959 samples left in the analysis.

3.2. Measurement Model Assessment

Several key analysis techniques were employed in the measurement model analysis using PLS-SEM. Outer loadings were calculated to assess the strength of the relationship between each observed variable and its corresponding latent construct, indicating how well the observed variables measure the underlying construct. Cronbach’s alpha was used to evaluate the internal consistency and reliability of the measurement scales, providing an assessment of how well the observed variables within each construct are correlated. Average variance extracted (AVE) was calculated to measure the amount of variance captured by the latent construct relative to the amount of measurement error.

Table 3. Outer loadings.

Outer Loadings	All-Sample Model	Local	Non-Local Worker	Traveler
Threat_1 ← Threat Appraisal	0.789	0.866	0.75	0.842
Threat_2 ← Threat Appraisal	0.863	0.835	0.857	0.925
Threat_3 ← Threat Appraisal	0.828	0.846	0.833	0.74
Coping_1 ← Coping Appraisal	0.824	0.853	0.801	0.799
Coping_2 ← Coping Appraisal	0.89	0.81	0.861	0.924
Coping_3 ← Coping Appraisal	0.858	0.766	0.832	0.895
Intention_1 ← Tsunami Evacuation Intention	0.7	0.705	0.703	0.704
Intention_2 ← Tsunami Evacuation Intention	0.729	0.736	0.685	0.825
Intention_3 ← Tsunami Evacuation Intention	0.736	0.757	0.719	0.707
Intention_4 ← Tsunami Evacuation Intention	0.781	0.748	0.778	0.808
Intention_5 ← Tsunami Evacuation Intention	0.788	0.821	0.752	0.817
Intention_6 ← Tsunami Evacuation Intention	0.711	0.75	0.638	0.846
Intention_7 ← Tsunami Evacuation Intention	0.715	0.745	0.693	0.671
ActionSelection_1 ← Tsunami Evacuation Action Selection	0.651	0.611	0.592	0.793
ActionSelection_2 ← Tsunami Evacuation Action Selection	0.626	0.641	0.534	0.802
ActionSelection_3 ← Tsunami Evacuation Action Selection	0.789	0.753	0.758	0.904
ActionSelection_4 ← Tsunami Evacuation Action Selection	0.805	0.775	0.777	0.901
ActionSelection_5 ← Tsunami Evacuation Action Selection	0.778	0.745	0.761	0.847

,

Table 4. Cronbach’s alpha, composite reliability, and average variance extracted (AVE).

Respondent Type	Construct	Cronbach’s Alpha	Composite Reliability	AVE
All-sample model	Threat Appraisal	0.769	0.866	0.684
	Coping Appraisal	0.823	0.893	0.736
	Tsunami Evacuation Intention	0.86	0.893	0.544
	Tsunami Evacuation Action Selection	0.796	0.852	0.538
Local	Threat Appraisal	0.807	0.886	0.721
	Coping Appraisal	0.747	0.851	0.656
	Tsunami Evacuation Intention	0.872	0.901	0.566
	Tsunami Evacuation Action Selection	0.76	0.833	0.502
Non-local worker	Threat Appraisal	0.746	0.855	0.664
	Coping Appraisal	0.779	0.871	0.692
	Tsunami Evacuation Intention	0.836	0.877	0.505
	Tsunami Evacuation Action Selection	0.742	0.818	0.478
Traveler	Threat Appraisal	0.79	0.876	0.704
	Coping Appraisal	0.862	0.906	0.764
	Tsunami Evacuation Intention	0.886	0.911	0.595
	Tsunami Evacuation Action Selection	0.907	0.929	0.724

and

Table 5. Variance inflation factor (VIF) for each hypothesis.

	All-Sample Model	Local	Non-Local Worker	Traveler
H1: Threat Appraisal → Tsunami Evacuation Intention	1.071	1.272	1.226	1.019
H2: Coping Appraisal → Tsunami Evacuation Intention	1.074	1.289	1.222	1.01
H3: Tsunami Evacuation Intention → Tsunami Evacuation Action Selection	1	1	1	1
H4: Gender → Threat Appraisal	1.062	1.065	1.039	1.337
H5: Age → Threat Appraisal	1.036	1.058	1.074	1.074
H6: Past Tsunami Experiences → Threat Appraisal	1.093	1.119	1.063	1.355
H7: Past Tsunami Experiences → Tsunami Evacuation Intention	1.003	1.016	1.004	1.02

show that Cronbach’s alpha, composite reliability, and AVE are all acceptable when compared to the criteria. In terms of convergent validity, not all outer loadings of all measurements for the latent variables exceeded 0.7. However, it should be emphasized that this rule of thumb in scale development should not be as strict. Loadings of 0.5 or 0.6 may still be allowed if the model contains additional indicators for comparative purposes [49]. Additionally, construct reliability exceeded 0.6 [52]. However, not all extracted average variances (AVEs) of all constructions were above 0.5 [65]. Yet, the composite reliability of all measures is 0.6. Also, because the composite reliability of all constructs is well above the necessary level, the internal reliability of the measurement items is acceptable [66]. Furthermore, the result for the VIF can illustrate the multicollinearity problems based on each hypothesis. Since no VIF value is higher than 5, it can indicate no multicollinearity problem [67].

3.3. Path Coefficients and Findings

According to the analysis, results based on PLS-SEM are be summarized in Figure 3 and

Table 6. Path coefficient of the all-sample hypothesized model and each respondent type.

	All	Local	Non-Local Worker	Traveler
H1: Threat Appraisal → Tsunami Evacuation Intention	0.582 ***	0.546 ***	0.525 ***	0.48 ***
H2: Coping Appraisal → Tsunami Evacuation Intention	0.101 ***	0.082	0.244 ***	0.119 ***
H3: Tsunami Evacuation Intention → Tsunami Evacuation Action Selection	0.606 ***	0.576 ***	0.630 ***	0.610 ***
H4: Gender → Threat Appraisal	0.246 ***	0.383 ***	0.039	0.265 *
H5: Age → Threat Appraisal	−0.062 **	−0.018	−0.196 ***	0.146 **
H6: Past Tsunami Experiences → Threat Appraisal	0.083	0.032	0.247 **	−0.302
H7: Past Tsunami Experiences → Tsunami Evacuation Intention	−0.003	0.446 ***	0.049	−0.569 ***

* p < 0.1, ** p < 0.05, *** p < 0.01.

.

In Figure 3 and Table 6, we provide a visualization of PLS-SEM for the all-sample group and a summary of path coefficients within the hypothesized model, revealing both similarities and distinctions among various respondent categories. Notably, the results for all sampled groups closely align with those of the local group, with the primary focus of our hypotheses centered on the non-local workers’ cohort. When examining each hypothesis, it becomes evident that H1 and H2 exhibit positive associations across most respondent types, although the strengths of these relationships vary. The local group notably demonstrates a high inclination to evacuate, particularly evident in their threat appraisal perspectives. On the other hand, the non-local worker group has the strongest relationship based on coping appraisal. H3 highlights strong, affirmative path coefficients across all respondent types, indicating a robust positive relationship between tsunami evacuation intention and the subsequent selection of evacuation actions. Next, for H4, the impact of gender on threat appraisal is most pronounced among the local group; however, gender consistently influences threat appraisal positively across all respondent groups except the insignificance for non-local workers. H5 and H6 introduce intriguing variations based on age and past tsunami experiences, with age displaying a significant negative effect among all sample groups and non-local workers and a positive effect among travelers. In contrast, past tsunami experiences exert the most substantial impact on non-local workers, with neither all sample groups nor local respondents exhibiting significant effects regarding these two hypotheses. Lastly, H7 underscores that the influence of past tsunami experiences on evacuation intention is most potent among local groups and considerably negative for travelers, with the all-sample group and local groups not displaying significant effects.

4. Discussion

According to the Results section, it is evident that certain groups of respondents exhibit varying outcomes based on their perceptions of potential threats. Notably, individuals tend to express a heightened inclination to evacuate to safer locations when they perceive a greater potential threat. Additionally, coping appraisal, which assesses individuals’ ability to manage danger, is positively associated with a stronger intention to evacuate in the event of a tsunami for all groups except for the local people. These findings align with previous research [11,14,15]. Further analysis of the path coefficients reveals that threat appraisal substantially influences tsunami evacuation intentions more than coping appraisal. This underscores the importance of disaster organizations conducting evacuation drills to enhance individuals’ survival readiness. The result confirms the importance of threat appraisal in influencing tsunami evacuation intentions, emphasizing the significance of individuals’ perceptiveness for the disaster that will happen in the future in shaping their evacuation decisions [68].

Moreover, this study underscores the positive impact of tsunami evacuation intentions on individuals’ actual evacuation actions, in line with prior research [11,17]. Additionally, this study reveals that gender influences threat appraisal, suggesting that women tend to be more sensitive to threats and perceive them more acutely than men [19,20,21] for all groups of respondents except for the non-local worker group, which shows an insignificant relationship. Therefore, future research could explore strategies to enhance men’s threat perception and bolster women’s existing threat awareness, particularly for organizations involved in disaster preparedness. The findings also recommend that organizations responsible for disaster evacuation processes, such as the National Disaster Warning Center in Thailand, should consider strategies to encourage greater participation of men in evacuation drills. In addition, Leelawat et al. [69] proposed that the evacuation drill is the preliminary countermeasure for preparing for the tsunami disaster that might happen in the future.

Conversely, the hypothesis regarding the influence of age on threat appraisal did not receive strong support. It was found that for all non-local workers, age exhibited a negative relationship with threat appraisal, suggesting that younger individuals perceive greater vulnerability than older ones within this subgroup. On the other hand, older travelers exhibited a higher threat appraisal level than younger travelers. This discrepancy may be attributed to physical and psychological factors, as suggested by Rafiey et al. [70]. Based on this finding, disaster organizations should consider targeting older individuals to increase their awareness of disaster preparedness. In addition, age is not a decisive factor in determining threat appraisal during a tsunami for the local respondent groups.

Additionally, it is important to note that past experience with tsunamis has varying effects on different groups. For non-local workers, increased past experience results in higher threat appraisal, while travelers with more past experience demonstrate lower tsunami evacuation intentions. These results align with Buylova et al. [34], who emphasized disaster awareness’s significance in preventing complacency in individuals who may overestimate their survival ability if their past experiences did not involve high-severity situations. On the other hand, the local group has high tsunami intentions if they have had a past tsunami experience. This result is also in line with the study of Buylova et al. [34], which was performed in a past disaster area, based on the fact that knowledge about the tsunami can increase the intention to evacuate immediately.

In terms of theoretical implications, the result of the overall model in this study aligns with the PMT [9], illustrating the influence of threat and coping appraisal on protection motivation. Tsunami evacuation intention serves as a representative measure in this context. Significant effects based on gender and age suggest opportunities for future research to incorporate demographic factors into the study model. Previous studies, including those on knowledge of social influence [71] and subjective norm [72], have explored additional factors within the PMT framework. Notably, this study raises interesting findings that past tsunami experiences did not necessarily lead to higher evacuation intention, aligning with similar observations in Cugliari et al. [73]. These insights can guide disaster management officials in enhancing awareness and preparedness for tsunami evacuations.

5. Conclusions

In summary, the primary objective of this study was to analyze the factors influencing tsunami evacuation behavior among distinct demographic groups in Phuket and Phang Nga, Thailand. The research findings indicate that all hypotheses, including a positive relationship between threat appraisal and tsunami evacuation intention, a positive relationship between coping appraisal and tsunami evacuation intention, and a positive relationship between tsunami evacuation intention and tsunami evacuation action selection, which derived from the PMT framework within each model, exhibited statistical significance. Notably, variables associated with PMT, including gender, age, and previous tsunami experience, displayed variations among different demographic groups.

Moreover, a comparative analysis among the locals, non-local workers, and travelers models revealed that only the outcomes of the local model aligned with the all-sample model. This study emphasizes the significance of accounting for diverse data subsets when constructing models for practical implementation and future research. Based on these research outcomes, tailored implementation strategies could be divided into three groups, resulting from all three models. For locals, increasing threat appraisal among males is crucial. Threat appraisal impacts their intention and action for evacuation. Authorities can emphasize tsunami severity and collective action. Non-local workers should be provided information and resources, especially for those without prior tsunami experience, with consideration of gender and age. Among travelers, older individuals perceive a higher threat, and past tsunami experiences affect evacuation intentions differently. Targeting younger travelers with persuasive messages and providing knowledge can enhance disaster preparedness. This study contributes to the literature on disaster risk reduction and management in the specific context of Phuket and Phang Nga, emphasizing the need for tailored strategies that consider the particular needs and characteristics of residents, non-local workers, or travelers. Additionally, it is mandatory to provide implications for disaster management officials, policymakers, and tourism operators, which can inform the development of targeted interventions and community outreach programs to improve intent and response during a tsunami warning.

There are limitations to consider in this research. One limitation of this study may be the sample size, as it was collected from a specific geographic location (i.e., Phuket and Phang Nga) and may not represent other areas or populations. Moreover, due to the data distribution from the survey results, there are some imbalance problems based on the demographics of the sample, which include genders and tsunami experiences. Another limitation may be the reliance on self-reported data, which can be subject to social desirability bias or memory recall errors. Additionally, this study focused only on the factors that may influence tsunami evacuation behavior and did not examine other potential factors that may impact such behavior, such as cultural and socioeconomic factors and level of education. Lastly, this study did not explore the effectiveness of different types of tsunami evacuation communication strategies or emergency preparedness interventions. Furthermore, both systematic and random errors may persist during the survey phase. However, we tried to minimize them by carefully and systematically checking each step of the survey process. This involves ensuring the survey is understandable, confirming participants’ willingness to participate, and double-checking the survey results during the conversion from paper to digital format.

Several avenues for further research could build upon the findings of this study. Firstly, it would be valuable to investigate the effectiveness of different types of disaster risk communication strategies, including community-based education programs and mass media campaigns, in improving threat and coping appraisals and subsequent evacuation behaviors. Additionally, future studies could explore the role of social capital, trust in authorities, and personal network characteristics in shaping individual responses to tsunami warnings and evacuation orders. Another potential area for further research is to investigate the influence of situational and contextual factors, including the time of day, the type of location, and the severity of the disaster on evacuation decision-making processes. Lastly, it is possibly valuable to compare evacuation behaviors across different geographic regions and economic statuses to better understand the factors contributing to successful evacuation and protective behavior in various contexts, especially for Phuket and Phang Nga, which have different economic statuses and some geographic characteristics.