1 Introduction

Earthquakes are caused by movements within the Earth’s crust and uppermost mantle. They range from events too weak to be detectable except by sensitive instrumentation, to sudden and violent events lasting many minutes which have caused some of the greatest disasters in human history. Earthquakes are a natural phenomenon in which the Earth’s crust rapidly releases energy after a long period of gravitational action and produces seismic waves during the period. At present, there are two seismic belts around the world, namely the Circum-Pacific Seismic Belt and the Mediterranean-Himalayan Seismic Belt. Earthquakes have the following characteristics: high frequency, wide range, high intensity, shallow source and obvious regional differences and so on, which determine the severity and extensiveness of earthquakes. As many as 1.2 million people have died from earthquakes in the world since the 20th century, of which nearly 600,000 are in China, accounting for about a half of the casualties. In China, deaths caused by various natural disasters have been about 550,000, while deaths caused by earthquakes have been about 280,000 since the founding of the People’s Republic of China.

When an earthquake occurs, the rapid assessment of its impact is essential for timely and appropriate emergency operations, such as securing people and crucial infrastructures exposed to serious damage [1]. Earthquake Early Warning (EEW) systems are timidly becoming operational in some areas of some seismic countries [2]. Earthquake early warning can be used to detect earthquakes and provide advance notices of strong shaking, enabling people to take responses in advance, which not only facilitates timely maintenance of infrastructure, but also reduces casualties. While EEW technology has matured, questions about who sends alerts and who is responsible for errors or omissions of detection limit the speed at which EEW can be widely open to the public. In addition, high costs of implementation and operation are an obstacle to the roll-out of EEW systems in both underdeveloped and developing countries. In parallel to EEW systems run by government agencies at the national level, the last decade has witnessed the development of unofficial platforms providing fast earthquake alerts at the global level. This was possibly thanks to smartphone technology and to the crowdsourcing model, with people making their smartphones available in order to receive a useful service in return [3]. In recent years, with the rapid development of the field related to smartphones, many things that could only be done on computers before can now be finished by mobile smartphones. The fully developed hardware configuration of smartphones plus the maturity and openness of operating systems have made it possible to use accelerometers in mobile phones to monitor seismic waves. According to a study by Becker Julia S., Potter Sally H. and Vinnell Lauren J., it is said that respondents preferred to receive earthquake warnings on their phones and be informed what responses to take [3].

Customer loyalty is considered to lead to repeated purchases, positive attitudes, intentions of continuing the affiliation and intentions of positive recommendations [4,5,6,7,8]. Based on the research aforementioned, it is certain that the public have a preference for earthquake early warning systems. This preference reinforces users’ intention to continue using the Earthquake Early Warning APP. In addition, a large number of studies have investigated public perceptual satisfaction of the APP, but might lack investigations of the customer loyalty of the public for the APP, and did not make effective use of resources of the masses, which might undermine the competitiveness of the APP [9]. It is known that customer loyalty has three functions from a lot of literature: firstly, to motivate users to continue using the APP [10]; secondly, to be used as basic information to provide a reference for the design and construction of the APP [11,12,13,14]; thirdly, to improve users’ satisfaction through design and construction [15]. At the same time, although the public are not so clear about the customer loyalty of the Earthquake Early Warning APP, there is evidence indicating that the public’s understanding of related knowledge has a positive impact on promoting customer loyalty [16]. Customer loyalty is the basis for the updating and iteration of the APP. To enhance its customer loyalty is to improve the quality of system services of the APP and provide a foundation for improving future strategies [17]. When users of the APP use it, relevant information may crowd in. And the more frequently users use it, the more information they receive. It is unknown that whether this will affect the customer loyalty. Users’ feedback on the usefulness and other functions may be mixed with praises and criticism according to users’ experience and other reasons. Besides, the frequency of use affects how well users understand the APP, and therefore impacts the overall customer loyalty [18].

Few scholars focused on the customer loyalty to study which factors are the driving factors, and even fewer scholars used frequency and other factors to explain the influence of users’ loyalty. The main purpose of this research is to construct and verify a conceptual framework with the Howard-Sheth Model and Technology Acceptance Model (TAM) as the theoretical basis combined with previous research. We try to explore the correlation among the determinants of users’ loyalty of the Earthquake Early Warning APP, and study the complex relationships between four factors namely social stimuli, perceived usefulness, frequency of use, satisfaction and the customer loyalty with the methods of literature reviews and mathematical statistics with the purpose of conducting research related to the APP and provide theoretical guidance and data support for relevant enterprises to make operational adjustments.

In this study, we take the Earthquake Early Warning APP developed by the Institute of Care-Life (hereinafter referred to as ICL) as the research case. The APP applies the system of earthquake early warning technology developed by ICL which was established after the Wenchuan earthquake in 2008. Committed to research of earthquake early warning technology, ICL has used information and other resources of Wenchuan earthquake and the aftershocks, and initially mastered the core technology of earthquake early warning and rapid reporting of seismic intensity. And these technologies have been integrated into the system integrated with software and hardware it has developed as independent intellectual property rights of ICL. The system has passed the appraisal of scientific and technological achievements organized by the Department of Science and Technology affiliated with Sichuan Province in Beijing on September 2nd, 2012. At that time, the review experts believed that the system was advanced nationwide, and some technologies of it were leading and advanced ones at home and abroad. This is the first system of earthquake early warning technology in China that has passed the provincial appraisal of scientific and technological achievements. At present, the “System of Earthquake Early Warning Technology of ICL” is mainly used in the aftershock areas of the Wenchuan earthquake and the junction of Sichuan and Yunnan Province. In addition, the system has been preliminarily applied to some projects such as the construction of urban systems of earthquake early warning in Chengdu, the capital city of Sichuan Province and the City of Chuzhou in Anhui Province.

At present, there are more than ten types of software related to earthquake early warning in smartphones in China, but except the Earthquake Early Warning App of ICL, most of the other applications are not quite professional, and they obtain information of earthquake early warning through their connection to some official platforms on phones, which might cause some problems such as belated warnings and other errors. However, the App of ICL features a greater promotion, a wider range of application and a larger number of users. Before an earthquake really occurs, the APP will issue an alarm a few seconds ahead of its occurrence when the triggering conditions are reached. After earthquakes, we can upload the feedback of risk avoidance and other collected data related to the earthquakes and risks on the uploading page in order to launch accurate rescues. Meanwhile, we can learn about some knowledge of earthquakes, sounds of early warnings, and cases of earthquake early warnings on the science-popularized page of the APP. The ICL and the Department of Emergency Management (the former department of earthquake prevention and disaster reduction) jointly built a mainland network of earthquake early warning which extends to 31 provinces and municipalities in China, covering 90% of the population (about 660 million people) in China’s seismic areas. Since 2011, the system has continuously warned of all destructive earthquakes (as of September 2019, a total of 53 times) within the network such as 7-magnitude earthquake in Lushan County, 6.5 earthquake in Ludian County, 7-magnitude earthquake in Jiuzhaigou Vally Scenic and Historic Interest Area, and 6-magnitude earthquake in Yibin City. Users have taken appropriate measures to avoid risks and reduce casualties and secondary disasters after receiving the early warnings.

Therefore, we take the earthquake early warning APP developed by the ICL as an example to study factors affecting the users’ loyalty to earthquake early warning technology.

The rest of this article is as follows: the second part introduces the theoretical background and research hypotheses; the third part shows the research methodology, which is used to validate the proposed model and test hypotheses; and the next section displays the results of the study after analyzing relevant data. Eventually, this paper concludes with a discussion of the significance, and limitations of our study, and recommendations for future research.

2 Theoretical Background and Research Hypotheses

2.1 Background of Research

Although the importance of the customer loyalty has been recognized and there has been lots of research about it, there is still a lack of research on the customer loyalty of applications of disaster early warning. Since earthquake early warning systems are typically designed from technical perspectives, the benefits in this study could be driven from the Technology Acceptance Model (TAM model) used for understanding the adoption of technology-based initiatives [19]. Literature of safety-critical systems has acknowledged the importance of usefulness, with some theoretical and empirical studies on the usefulness of safety-critical systems [20]. The usage conditions of applications of disaster early warning are different from those of normal applications used on a daily basis [21], and at the same time, applications of disaster early warning may be used less frequently, with only a minority of people downloading smartphone applications for emergency responses (16%) [22]. What’s more, when using those applications for emergency responses, users may be in a high-risk environment [23].Therefore, in the context of applications related to disaster early warning, it is necessary to develop related models of the customer loyalty and conduct empirical verifications to explore the relationship between customer loyalty and willingness of continuous use as well as the four factors affecting the customer loyalty.

Users’ satisfaction and perceived usefulness of the system have been extensively studied as determinants of users’ loyalty in the literature on the factors influencing users’ loyalty [24]. Similarly, users’ frequency of use and social stimuli are key factors in understanding the customer loyalty to earthquake early warning applications in the field of earthquake early warning [25].

After reviewing the relevant literature, it is found that few studies explored the impacts of social stimuli, perceived usefulness, satisfaction, and frequency of use on users’ loyalty of the APP. In addition, previous studies have not centered on the complex interrelationships among these variables to conduct any investigation. Thus, this study will address these research gaps.

2.2 Conceptual Framework

Perceived usefulness is an important factor in measuring users’ perceptions. Davis [26] defines perceptual usefulness as “the degree to which a person believes that using a particular system will improve his or her job performance”. In general, we define users’ satisfaction as an emotional condition to comprehensively evaluate all aspects of consumers’ relationships. Customer loyalty is considered to lead to repeated purchases, positive attitudes, intentions of continuing the affiliation and intentions of positive recommendations [4,5,6,7,8]. It can be understood from the perspective of users’ behaviors and attitudes. Research on users’ loyalty can help measure users’ willingness to continue using the APP. Since the models discussed are relatively sophisticated, we have also introduced the Howard-Sheth Model to help construct the theoretical models. Because reactions of social media and the public have a greater impact on self-perception responses in the case of disasters, the social stimuli at the time of early warning are considered as the input of the model.

2.3 Research Hypotheses

Figure 1 shows the model of users’ loyalty to the APP. Attributes including social stimuli, frequency of use, perceived usefulness, and satisfaction are treated as independent variables, and users’ loyalty is deemed as the dependent variable. The framework is based on the Howard-Sheth Model which includes three steps: inputting variables → intermediate architecture → outputting variables.

Fig. 1.
figure 1

Proposed Conceptual Model.

Full size image

H1: social stimuli have a direct positive influence on perceived usefulness on earthquake early warning APP. Some studies have identified usefulness as a determinant that influences users’ intention to continue use of the APP. For example, the study of consumers’ engagement on applications of mobile phones conducted by Tarute and other researchers [27] has demonstrated that a positive view of usability promotes better engagement, which in turn increases willingness to continue use of applications. Similarly, research on hotel-booking apps by Ozturk et al. [28] also illustrates that improving usefulness will enhance users’ perception of value, which in turn has a positive impact on the willingness to continue using the applications. Users’ willingness to continue use of applications depends on their satisfaction with the usage and their perceived usefulness of continuous use of them [29].

From the above discussion, it is assumed that perceived usefulness directly affects users’ satisfaction and loyalty, as shown below.

H2: Perceived usefulness has a direct positive influence on satisfaction on earthquake early warning APP.

H3: Perceived usefulness has a direct positive influence on loyalty on earthquake early warning APP.

Users’ satisfaction plays a crucial role in the relationship between perceived value and the customer loyalty [30]. Many previous studies have identified the relationship between customer satisfaction and the customer loyalty in different contexts [31,32,33,34].

Therefore, it is assumed that satisfaction directly affects users’ loyalty, as shown below.

H4: satisfaction has a direct positive influence on users’ loyalty on earthquake early warning APP.

In the context of applications related to disaster early warning, we have considered the frequency of use as an important factor. Therefore, it is proposed that the frequency of use directly affects users’ loyalty, as shown below.

H5: frequency of use has a direct positive influence on users’ loyalty on earthquake early warning APP.

3 Research Methodology

3.1 Survey Design

Data for this study were collected from a structured, self-administered questionnaire based on a previously validated methodology was used to assess demographic characteristics, perceptions and users’ satisfaction of the earthquake early warning app.

The questionnaire was designed in three sections:

The first section included two questions. The first question elicited information on user’s experience of the earthquake. The second question elicited information on the usage frequency of the earthquake early warning app.

Based on our previous investigation of the factors affecting the adoption of the earthquake early warning app and a careful review of the existing literature [25] on adoption and frequency of use of the earthquake early warning app, we divide the key explanatory variables used in our final models to explain the variation in the frequency of use of the earthquake early warning app into four main groups. The key attributes of the four groups of individuals who have heard about the earthquake early warning app but never used it (non-users), those who used it before but do not plan to use it anymore(zero-frequency users), those who install the app but not constant concern notice (infrequent users), and those who install the app and often concern notice (more frequent users). Using this screening question, we screened people who had used the earthquake warning APP, using only survey data from people who had used the earthquake warning APP. An explanation of earthquake warning APP with examples was included at the first section of the questionnaire to make sure that the respondents have a clear idea of earthquake warning APP.

We adapted the measurement projects of relevant studies related to early warning in the human factors engineering literature [25]. All of the items (36) used a 5-point Likert scale ranging from ‘strongly disagree’ (1) to ‘strongly agree’ (5).

The third section of the questionnaire sought to determine the information regarding demographics including gender, age and region.

The questionnaire was anonymous.

3.2 Data Collection

Data were collected from November 2021 to December 2021, using online methods.

Through the online survey, 524 samples were obtained, of which 481 were valid, as 43 questionnaires were discarded as participants need not less than 30 s to complete the questionnaire. In the meantime, people who have not used the app were screened out. Accordingly, a total of 332 questionnaires were used for data analysis. Gender distribution was fairly uniform (156 participants were male, 176 female participants). Regarding the overall perceived satisfaction of all participants with the application, they were asked the same questions, namely their current opinions of the application.

3.3 Data Analysis

Due to the complexity of the proposed model, this study uses both the TAM and the Howard-Sheth Model as well as the correlation analysis and stepwise regression method to verify the hypotheses mentioned before.

Accordingly, SPSSAU was used to analyze the data.

The instrument’s “Reliability” was established through Analysis of Cronbach’s reliability and the instrument’s “Validity” was established through KMO test and Bartlett Spherical Test. [35].

4 Results

Through the questionnaire, we collected 332 valid respondents as a sample (Table 1).

4.1 Reliability and Validity Analysis

Table 1. Cronbach's reliability analysis
Full size table

As can be seen from the above table: the reliability coefficient value is 0.856, which is greater than 0.8, thus indicating that the study data are of high reliability quality and can be used for further analysis (Table 2).

Table 2. KMO and Bartlett's test
Full size table

The validity was verified using KMO and Bartlett's test, as seen in the table above: the KMO value is 0.901, the KMO value is greater than 0.8, and the study data is very suitable for extracting information (a good validity from the side response).

4.2 Profile of Respondents

The results of this study followed a step-by-step procedure for data analysis. The descriptive analysis provided demographic information of respondents.

Table 3. Profile of respondents (n = 332).
Full size table

Table 3 presents the demographic information of the respondents in terms of gender, age, and whether they have experienced earthquakes or not. There were slightly more female respondents than male, 53.01%, compared to 46.99% male respondents. In terms of age, the sample had a relatively high proportion of “20–30” respondents, with 35.24%. The proportion of the 31–40 sample was 34.04%.

In addition to the demographic information, the study also asked the respondents if they had ever experienced an earthquake. From the above table, the highest percentage of “experienced an earthquake” was 69.58%. The percentage of those who had not experienced an earthquake was 30.42%.

4.3 Evaluation of Path Relationship

The path relationship is reviewed by the stepwise regression method. First of all, the R Square(R2) of the model fitting condition is analyzed, and then the VIF value could be analyzed (to determine multicollinearity); secondly, the significance of X is analyzed, and if it is significant, it indicates that X has an influence on Y, and then the concrete direction of the influence shall be analyzed [35].

4.4 Hypothesis 1: Social Stimuli Positively Influences Perceived Usefulness

A stepwise regression method was used to test the hypothesis that the higher the social stimulus, the higher the perceived usefulness of the earthquake warning app. Perceived usefulness is the outcome variable and social stimulus is the predictor variable.

Perceived usefulness was assessed by 3 items: I think the earthquake warning app is an effective warning tool; I would find the frequency of warning messages in the earthquake warning app appropriate; and I think the alerts in the earthquake warning app are useful.

Social stimulation was assessed by 2 items: other people's reactions to earthquake warnings influenced my reactions; social media influenced my reactions to earthquake warnings (Table 4).

Table 4. Results of stepwise regression analysis (n = 332) for hypothesis 1
Full size table

Social stimuli were used as independent variables, while perceived usefulness was used as dependent variable for stepwise regression analysis, and after automatic identification of the model, a total of 1 item of social stimuli was finally left in the model, and the model formula was: perceived usefulness = 3.089 + 0.114*social stimuli, with an R-squared value of 0.036, implying that social stimuli could explain 3.6% of the cause of change in perceived usefulness. Moreover, the model passed the F-test (F = 12.199, p = 0.001 < 0.05), indicating that the model is valid. In addition, the test for multiple cointegrations of the model found that all the VIF values in the model were less than 5, implying that there was no cointegration problem; and the D-W value was around the number 2, thus indicating that there was no autocorrelation in the model, and there was no correlation between the sample data, and the model was better. The final specific analysis shows that.

The value of the regression coefficient for social stimuli is 0.114 (t = 3.493, p = 0.001 < 0.01), implying that social stimuli can have a significant positive relationship with perceived usefulness.

Summarizing the analysis, it can be seen that: social stimuli can have a significant positive influence relationship on perceived usefulness.

4.5 Hypothesis 2: Perceived Usefulness Positively Influences Satisfaction

Stepwise regression was used to test the hypothesis that the higher the perceived usefulness of the earthquake warning app, the higher the user satisfaction. Satisfaction is the outcome variable and perceived usefulness is the predictor variable.

Satisfaction is assessed by 2 items: the information I receive from the earthquake warning app increases my sense of urgency; and installing the earthquake warning app makes me feel safer (Table 5).

Table 5. Results of stepwise regression analysis (n = 332) for hypothesis 2
Full size table

Perceived usefulness was used as the independent variable, while satisfaction was used as the dependent variable for stepwise regression analysis, and after automatic identification of the model, a total of one term of perceived usefulness was finally left in the model, and the model formula was: satisfaction = 1.226 + 0.722*perceived usefulness, with an R-squared value of 0.251, implying that perceived usefulness could explain 25.1% of the causes of change in satisfaction. Moreover, the model passed the F-test (F = 110.807, p = 0.000 < 0.05), indicating that the model is valid. In addition, the test for multiple cointegrations of the model found that all the VIF values in the model were less than 5, implying that there was no cointegration problem; and the D-W value was around the number 2, thus indicating that there was no autocorrelation in the model, and there was no correlation between the sample data, and the model was better. The final specific analysis shows that.

The regression coefficient value of perceived usefulness is 0.722 (t = 10.526, p = 0.000 < 0.01), which means that perceived usefulness has a significant positive relationship with satisfaction.

Summarizing the analysis, it is clear that perceived usefulness will have a significant positive influence relationship on satisfaction.

4.6 Hypothesis 3: Perceived Usefulness Positively Influences Loyalty

Stepwise regression was used to test the hypothesis that the higher the perceived usefulness of the earthquake warning app, the higher the loyalty of users. Loyalty is the outcome variable and perceived usefulness is the predictor variable.

Loyalty was assessed by five items: I like the earthquake warning app; I would take the alerts from the earthquake warning app seriously; I tend to use the earthquake warning app; the earthquake warning app was developed by the Chengdu High-Tech Institute of Disaster Reduction, and I would like to learn more about the earthquake warning program provided by the institute; I would change my plan if I received an alert from the earthquake warning app (Table 6).

Table 6. Results of stepwise regression analysis (n = 332) for hypothesis 3
Full size table

Perceived usefulness was used as the independent variable, while loyalty was used as the dependent variable for stepwise regression analysis, and after automatic identification of the model, a total of one item of perceived usefulness was finally left in the model, and the model formula was: loyalty = 0.878 + 0.819*perceived usefulness with an R-squared value of 0.465, which means that perceived usefulness can explain 46.5% of the reasons for the change in loyalty. Moreover, the model passed the F-test (F = 287.088, p = 0.000 < 0.05), indicating that the model is valid. In addition, the test for multiple cointegrations of the model found that all the VIF values in the model were less than 5, implying that there was no cointegration problem; and the D-W value was around the number 2, thus indicating that there was no autocorrelation in the model, and there was no correlation between the sample data, and the model was better. The final specific analysis shows that. The regression coefficient value of perceived usefulness is 0.819 (t = 16.944, p = 0.000 < 0.01), which means that perceived usefulness will have a significant positive relationship with loyalty.

Summarizing the analysis, it is clear that perceived usefulness can have a significant positive influence relationship on loyalty.

4.7 Hypothesis 4: Satisfaction Positively Influences Loyalty

Stepwise regression was used to test the hypothesis that the higher the user's satisfaction with the earthquake warning app, the higher the user's loyalty. Loyalty is the outcome variable and satisfaction is the predictor variable (Table 7).

Table 7. Results of stepwise regression analysis (n = 332) for hypothesis 4
Full size table

Satisfaction was used as the independent variable, while loyalty was used as the dependent variable for stepwise regression analysis, and after automatic identification of the model, a total of 1 item of satisfaction was finally left in the model, and the model formula was: loyalty = 1.636 + 0.563*satisfaction, with an R-squared value of 0.455, which means that satisfaction can explain 45.5% of the reasons for the change in loyalty. Moreover, the model passed the F-test (F = 275.279, p = 0.000 < 0.05), indicating that the model is valid. In addition, the test for multiple cointegrations of the model found that all the VIF values in the model were less than 5, implying that there was no cointegration problem; and the D-W value was around the number 2, thus indicating that there was no autocorrelation in the model, and there was no correlation between the sample data, and the model was better. The final specific analysis shows that.

The value of regression coefficient of satisfaction is 0.563 (t = 16.592, p = 0.000 < 0.01), which means that satisfaction will have a significant positive influence relationship on loyalty.

Summarizing the analysis, it can be seen that: satisfaction will have a significant positive influence relationship on loyalty.

4.8 Hypothesis 5: Frequency of Use Positively Influences Loyalty

Stepwise regression was used to test the hypothesis that the more frequently users use the earthquake warning app, the more loyal users will be. Loyalty is the outcome variable and frequency of use is the predictor variable (Table 8).

Table 8. Results of stepwise regression analysis (n = 332) for hypothesis 5
Full size table

Using frequency of use as the independent variable and loyalty as the dependent variable for stepwise regression analysis, after automatic identification of the model, a total of 1 term of frequency of use was finally left in the model, and the model formula was: loyalty = 3.531 + 0.110*frequency of use, with an R-squared value of 0.014, implying that frequency of use could explain 1.4% of the reasons for the change in loyalty. Moreover, the model passed the F-test (F = 4.700, p = 0.031 < 0.05), indicating that the model is valid. In addition, the test for multiple cointegrations of the model found that all the VIF values in the model were less than 5, implying that there was no cointegration problem; and the D-W value was around the number 2, thus indicating that there was no autocorrelation in the model, and there was no correlation between the sample data, and the model was better. The final specific analysis shows that.

The regression coefficient value of frequency of use is 0.110 (t = 2.168, p = 0.031 < 0.05), which means that frequency of use has a significant positive relationship with loyalty.

Summarizing the analysis, it can be seen that: frequency of use will have a significant positive influence relationship on loyalty.

5 Discussion

5.1 Theoretical Implications

This study explored the determinants of loyalty to the Earthquake Early Warning APP, which has not been adequately studied in the related literature. The relational model was examined by data obtained from a questionnaire for the APP. A conceptual framework representing direct and indirect relationships between five constructs namely social stimulation, perceived usefulness, users’ satisfaction, frequency of use, and users’ loyalty was validated. The framework helps to understand the formation of user loyalty to the earthquake warning app under the influence of safety and technology-related factors. It is important to explore the factors that influence the customer loyalty of applications of earthquake early warning, especially in China, where these applications are considered to be a powerful tool in the current earthquake early warning system that enables people to avoid disasters.

Three factors, including perceived usefulness, frequency of use, and satisfaction, have a direct impact on the loyalty of earthquake warning app users. In addition, the results of the study also showed that users’ perceived usefulness of the earthquake warning app was higher if they were more influenced by others or social media. For example, users can browse information about the earthquake warning app from social media. In this study, perceived usefulness is an important factor that influences users’ loyalty. In addition, the perceived usefulness of the earthquake warning app may satisfy users. In fact, the perceived usefulness of the earthquake warning app, such as sending effective warning messages and having the right frequency of warnings, contributed to the satisfaction of the riders, which in turn made them willing to continue using the earthquake warning app, rate the earthquake warning app positively, or recommend the earthquake warning app to their friends. Overall, this study still emphasized the importance of effective warnings. In addition to factors related to frequency of use and social stimulation (Tables 9, 10, 11, 12 and 13).

Table 9. Pearson correlation between Social stimuli and Perceived usefulness
Full size table
Table 10. Pearson correlation between Perceived usefulness and Satisfaction
Full size table
Table 11. Pearson correlation between Frequency of use and Loyalty
Full size table
Table 12. Pearson correlation between Satisfaction and Loyalty
Full size table
Table 13. Pearson correlation between Perceived usefulness and Loyalty
Full size table

Based on the results of the correlation analysis, perceived usefulness was found to have the highest impact on users’ loyalty (correlation coefficient of 0.682). Consistent with previous research in the field of disaster apps, this study affirms app dependability as a usability factor in the disaster apps’ context.

This is followed by satisfaction (correlation coefficient of 0.674), while frequency of use has the least effect on loyalty (correlation coefficient of 0.119).

It is undeniable that people will be loyal to an earthquake warning app if they are satisfied with it. More importantly, the findings suggest that satisfaction plays a mediating role in the relationship between the two factors mentioned above, including perceived usefulness and loyalty. This study provides preliminary insight into the indirect relationships among these constructs and provides a basis for understanding the complex formation of loyalty to the earthquake warning app. Since perceived usefulness has a direct effect on satisfaction, perceived usefulness needs to be increased in order to improve users’ loyalty.

5.2 Practical Significance (Recommendations Based on Theoretical Suggestions)

From the perspective of practice, the study has proposed and verified the determinants of improving the users’ loyalty to the Earthquake Early Warning APP, including three positive and direct factors namely perceived usefulness, satisfaction and frequency of use, and a positive and indirect factor--social stimulation. Therefore, it is recommended that the ICL focuses on these four aspects to improve the use of the APP.

According to the study [24], there are three factors having the greatest impact on users’ perceived usefulness: whether the Earthquake Early Warning APP is effective; whether its early warning information frequency is appropriate; and whether the reminder is useful. Therefore, it is suggested that developers always monitor the service status of the APP to ensure timely updates and maintenance. In addition, it is recommended that developers conduct multiple tests before the APP is listed into market to improve the accuracy of earthquake early warning, and study how to keep the information frequency within an appropriate range. Our research points out that perceived usefulness positively affects the customer loyalty, so it is recommended that developers focus on improving perceived usefulness.

At a time when the Earthquake Early Warning System is gradually becoming more widely used, if developers hope to attract more users to use the APP, they should also focus on other aspects in addition to improving the quality of service. The study has pointed out that social stimuli positively affect perceived usefulness, which shows that the perceived usefulness is correspondingly increased when responses to earthquake early warning of other people or social media affect users’ responses. Therefore, it is also a sound strategy to increase the positive publicity of the earthquake early warning APP in the aspect of social media, and at the same time, it is advisable to organize the promotion of the APP on a community basis and hold practice or exercises for disaster prevention and mitigation [37, 38].

Our research points out that to increase users’ loyalty to the APP, it is also necessary to improve users’ satisfaction. Some studies [24] show that two factors have the greatest positive impact on users’ satisfaction: whether the information received from the APP increases users’ sense of urgency and whether the installation of it on smartphones will make users feel safer. Therefore, it is recommended that developers carefully consider the content of early warning information, and the information should increase users’ sense of urgency. At the same time, after users install the APP, there shall be an appropriate reminder to make them feel that they are protected by the APP with a great sense of safety [39]. Finally, our research points out that to increase users’ loyalty to the APP, it is also a necessity to increase the users’ frequency of use. The higher the frequency of use is, the more information users will receive, and their understanding of the APP will be much deeper. How to increase the frequency of use of users is also a significant issue that developers need to consider.

The earthquake early warning system can quickly assess seismic elements and issue early warnings to the seismic areas to avoid casualties and property losses. The main significance of earthquake early warning is to provide more time for people to avoid danger. After earthquakes, the average time for a house to go from the start of shaking to collapse is about 12 s. With the early warning system, the time for early warning has risen, and the time for judgment and decision is reduced, while the time of risk avoidance is greatly increased.

Although the time for early warning of earthquakes is short, the accuracy rate is high, and China has begun to attach importance to the role of earthquake early warning systems in earthquake prevention and disaster reduction. If an early warning is issued a few seconds before an earthquake, theoretically, people can also do a lot of things to help themselves, and make different countermeasures according to different conditions of local situations. However, the success or failure of earthquake early warning is largely reflected in the speed and attitude of the public towards it. At present, the masses in China are not vigilant enough for the information of the earthquake early warning system, and after the earthquake early warning APP issues an early warning, users fail to save themselves immediately, which wastes valuable time for early warning. According to previous studies, the public has satisfaction and expectations for earthquake early warning systems. However, after reviewing relevant literature, it is found that this study explores for the first time the interrelationship among the determinants of the customer loyalty to the earthquake early warning APP, and the complex relationship between two factors including perceived usefulness, frequency of use and customer loyalty. The study has shown that the following five aspects determine users’ loyalty: whether users like the APP; whether they will take its alarm seriously; whether they want to learn about its developer; whether they will change their plans because of alarms released by the APP; whether they tend to use the APP. We hope to find out factors that increase users’ loyalty to earthquake early warning technology, and then improve the APP so as to enhance the public’s trust in the APP, encourage them to open permissions in their smartphones to the APP, and promote their beliefs in information related to earthquake early warning of the APP. Meanwhile, we also hope that the public’s vigilance to information related to earthquake early warning of the APP could be enhanced by increasing the perceived value of the APP so that they pay attention to the information in a timely manner and make responses as soon as possible, and that the early warning function of the APP could be improved to ensure that the public receives the information in a timely manner.

In line with the 17 Sustainable Development Goals proposed by the United Nations, the eleventh goal of which is to build inclusive, safe, sustainable cities and residence areas that have the capacity of resisting disasters, and significantly increase the number of integrated policies and plans adopted and implemented to build inclusive, resource-efficient and resilient cities and human residence areas which can mitigate and adapt to climate change and have the capacity of resisting against disasters by 2020, and establish and implement comprehensive management of disasters and risks at all levels in accordance with the Sendai Framework for Disaster Risk Reduction from 2015 to 2030 [40]. Earthquakes always cause casualties and economic losses, and thus earthquake early warning technology came into being. The link between people and technology is also very vital, and if we can better respond to it when receiving earthquake early warning, then the work of earthquake prevention and disaster relief will also be more effective.

6 Conclusion

The investigation showed that different factors such as social stimuli, perceived usefulness, frequency of use, and satisfaction affect users’ loyalty to earthquake early warning technology. This study also shows that users’ loyalty to earthquake warning technology depends on a variety of factors, some of which can be expected and considered at the initial stages of designs. Taking into consideration the users’ loyalty that affects the earthquake early warning technologies is one way to guide design decisions toward more effective solutions.

This study aims at exploring how factors such as social stimuli, perceived usefulness, frequency of use, and satisfaction affect users’ loyalty to earthquake early warning technologies. As a result, models originally designed to assess the tendency of customer loyalty have also proven to be an effective tool. Factors that are easy to consider, such as social stimuli, perceived usefulness, frequency of use, and satisfaction, can provide insight into creating earthquake early warning technologies with high loyalty.

However, the research results and framework advanced herein are just the initial efforts towards this end, and its insights shall be used to guide further investigations.

Limitations

The results of this study are mainly from online surveys, and there is a lack of offline experimental data. Besides, some of the theoretical models referenced in this study are applicable to commercial scenarios, but the earthquake early warning application selected in this study are not conventional and commercial ones and the use scenarios of it are different. Hence, this causes certain limitations.

Due to the particularity of earthquake early warning technology, the dimensional measurement of various factors affecting users’ loyalty in this study draws on conventional applications and thus has certain limitations.

Recommendations for Future Research

The methodology adopted by this study is effective in generating understanding or insights into HCI literature and practice. Therefore, the first recommendation for future research is to conduct research on users’ loyalty to earthquake early warning technology with the ways of online and offline, using other related applications as examples. Additionally, generally there is a lack of information on earthquake early warning in the HCI literature, which shall be should be given more considerations. Finally, more efforts should be put into creating tools to assist in the development of earthquake early warning systems with high loyalty and to bridge the gap between research and practice on users’ loyalty to the technology.