key: cord-282938-1if7bl2u
authors: Wang, Yanxin; Li, Jian; Zhao, Xi; Feng, Gengzhong; Luo, Xin (Robert)
title: Using Mobile Phone Data for Emergency Management: a Systematic Literature Review
date: 2020-09-16
journal: Inf Syst Front
DOI: 10.1007/s10796-020-10057-w
sha: 
doc_id: 282938
cord_uid: 1if7bl2u

Emergency management (EM) has always been a concern of people from all walks of life due to the devastating impacts emergencies can have. The global outbreak of COVID-19 in 2020 has pushed EM to the top topic. As mobile phones have become ubiquitous, many scholars have shown interest in using mobile phone data for EM. This paper presents a systematic literature review about the use of mobile phone data for EM that includes 65 related articles written between 2014 and 2019 from six electronic databases. Five themes in using mobile phone data for EM emerged from the reviewed articles, and a systematic framework is proposed to illustrate the current state of the research. This paper also discusses EM under COVID-19 pandemic and five future implications of the proposed framework to guide future work.

Emergency situations such as terrorist attacks or earthquakes occur at different scales daily around the world. They may be natural or human-caused events that occur suddenly, affect public order, and disrupt the regularity of an area's political, economic, and social life (Fogli et al. 2017; Seba et al. 2018) . Such an emergency causes great losses and widespread impacts on society, and "requires a prompt intervention by all involved stakeholders" (Fogli et al. 2017; Lauras et al. 2015) . To gain public support and maintain regular social order, authorities should pay special attention to the effective management of such situations. In this study, emergency management (EM) is defined as the effective organization, direction, and management of both emergency-related humanitarian and material resources (Othman and Beydoun 2013) . Traditionally, it comprises four phases: mitigation, preparedness, response, and recovery (Othman and Beydoun 2013) . EM is generally considered to have undergone three stages (Phillips et al. 2011) , including passive response (before the 1950s), active preparation and prediction (1960s-1990s) , and whole community response based on integrated information systems (after the 2000s). To align with such developments, scholars have swifted their attention from solving a single issue to focusing on efficient intra-organizational collaboration (Janssen et al. 2010) .

Lack of collaboration is the chief culprit in major failures in disaster response and takes the form of a lack of available crisis information or poorly managed information flow (Valecha 2019; Beydoun et al. 2018) . Information and communication technology (ICT) and information systems (ISs) are considered as crucial means to enhance the collaboration process and information flow management (Sagun et al. 2009; Ipe et al. 2010 ). However, a lack of informative and appropriate data hinders further development and practical use of emergency management information systems (EMISs) (Ghosh et al. 2018; Roberts 2011) . First, data reflecting human behaviors on a large scale are required for each phase of EM, since managing affected people is a crucial part of EM. Second, due to the rapid changes of challenges encountered in EM, data collected promptly and timely are required to support various EMISs for corresponding responses. Finally, EMISs require accurate and objective data to reflect emergencies, while traditional emergency-related data heavily rely on surveys. These three requirements create obstacles for the practical use of EMISs in dealing with real-world emergencies.

Many studies have regarded mobile phone data as a potential data source to fulfill these requirements, because these data reflect human behavior richly and ubiquitously. Globally, in 2017, mobile phones attained a registration number of 103.5 per 100 people, as reported by the International Telecommunication Union (ITU) (Sanou 2017) . They have been transformed from a simple communication tool to a multifunctional 'mobile-computer' with the rise of apps on mobile platforms. Mobile phone data such as CDRs and app data can be applied to the analysis of human mobility (Stefania et al. 2018; Duan et al. 2017; Gao et al. 2014; Lwin et al. 2018) , social networks (Poblet et al. 2018; Trestian et al. 2017; Ghurye et al. 2016; Dobra et al. 2015) , mobile phone usage patterns (Jia et al. 2017; Steenbruggen et al. 2016; Gundogdu et al. 2016; Gao et al. 2014) , and geographic location (Lwin et al. 2018; Poblet et al. 2018; Dong et al. 2017; Šterk and Praprotnik 2017) ; these themes are discussed in additional detail in Section 3. The results can further be developed to address various issues encountered during emergencies, such as predicting epidemic transmission Panigutti et al. 2017 ) and developing pre-warning systems (Zhang et al. 2016; Dong et al. 2017) .

Although many efforts have been made to investigate the application of mobile phone data in EM, the knowledge and understanding in this field are still fragmented. Therefore, a systematic framework that synthesizes the fragmented knowledge and provides insights into the-state-of-art of using mobile phone data for EM is needed. This study aims to propose such a framework and provide guidance for further research in this field.

This study is related to two streams of literature reviews, which are using ICT in EM and mobile phone data analysis. On the first stream of using ICT in EM (Martinez-Rojas et al. 2018; Tan et al. 2017) , Martinez-Rojas et al. (2018) have reviewed 158 related articles from 2009 to 2018 to discuss current opportunities and challenges of using Twitter for EM. Tan et al. (2017) have summarized the involvement of mobile apps in the crisis informatics literature by reviewing 49 related articles. On the second stream of mobile phone data analysis, Blondel et al. (2015) and Naboulsi et al. (2016) have summarized some studies on mobile data analysis, some of which can be applied in EM. These two literature reviews focused on the method of data mining, while the current study focuses on EM. To sum up, there is a lack of a literature review that considers both the characteristics of mobile phones and using such data for EM.

Three research objectives are undertaken to achieve the goal of synthesizing the fragmented knowledge and providing research guidance: (i) extract basic knowledge (e.g. types of mobile phone data, situations) of EM from the selected studies; (ii) break the boundaries of different disciplines and aggregate each analysis perspective; and (iii) study the identified knowledge and integrate it into a single framework that draws a comprehensive map of existing findings under this subject, and provides future implications.

To attain these objectives, this study follows a methodology of systematic literature review (SLR) and synthesized the results of the reviewed studies into a framework, thus allowing a discussion of future implications obtained from the framework. The next section introduces the research method applied in this systematic literature review, which is followed by an illustration of the five major themes in using mobile phone data for EM. Section 4 presents the proposed framework of using mobile phone data for EM based on the five themes. Section 5 discusses current EM under the COVID-19 pandemic. Finally, this paper discusses future implications and provides a conclusion.

Based on the systematic research methodology (Ghobadi 2015) which is considered as a means of identifying, evaluating and interpreting all available research relevant to a particular research question, or topic area, or phenomenon of interest (Budgen and Brereton 2006) , this study processes three phases of work, including planning, conducting, and reporting. These phases are graphically exhibited along with their specific steps and objectives in Fig. 1 . The research questions and relevant studies have been identified in phase 1, and the studies we reviewed in this paper have been selected through a specific research strategy and inclusion/exclusion criteria illustrated in phase 2. As a result, we have selected 65 papers from six document databases for this review. Note that phases 1 and 2 are both explained in Appendix 1, while the results of Phase 3 are detailed in Sections 3 and 4.

3 Five Themes in Using Mobile Phone Data for EM

To identify the trends of research on emergency situations during the selected period, the 65 reviewed articles have been statistically analyzed by their published years and the types of situations mentioned in each article. The results are exhibited in Fig. 2 .

Emergencies can be divided into two categories, natural (41 papers) and man-made (16 papers). 'Natural' refers to the emergencies that occurred due to processes of the earth (such as weather) that can hardly be avoided, while 'manmade' refers to the emergencies that occurred specifically due to human action or inaction. To be more specific, this paper considers that 'natural' emergencies include 'natural disaster' (19 papers) and 'disease disaster' (22 papers), such as earthquakes and the Ebola virus disease, respectively. Meanwhile, 'man-made' emergencies cover various types, among which 'traffic accident' (six papers), 'violence and terror incident' (six papers), and 'other' (four papers) have been regarded as three major subcategories based on the articles reviewed. The 'others' category includes 'sudden strikes' (Garroppo and Niccolini 2018) , 'damages to pipelines' (Dong et al. 2017) , and 'refugee problem' (Andris et al. 2019) . The remaining studies (eight papers) cannot be clustered into any of the aforementioned categories or subcategories in the course of our review, as they are not clearly related to any particular type of emergency, so we have categorized them as 'general emergency.'

Studies paid considerably more attention to 'natural' emergencies compared with 'man-made' emergencies, and this was particularly evident in terms of disease disasters. We also note that the number of studies peaked in 2015 (13 papers; the other years contained 9, 8, 11, 13, and 11 papers, respectively), and 'disease disaster' has been studied in 2015 (seven papers) almost twice as much as in other years. The phenomenon may have been the result of attention drawn by the outbreak of the Ebola virus.

EM generally goes through four phases, namely mitigation, preparedness, response, and recovery. Most of the studies investigated the response phase (39 papers). This phase involves activities like organizing an evacuation, mobilization, and assisting victims to manage the emergency appropriately. Then, 27 papers focused on the preparedness phase, and 22 papers focused on the mitigation phase. The preparedness phase in EM comprises a sequence of activities including planning, training, warning, and updating solutions by learning previous emergencies, which can help enhance response abilities (Oberg et al. 2011) . The mitigation phase aims to prevent a disaster or lessen its impacts by modifying the causes and vulnerabilities or distributing the losses (Oberg et al. 2011) . Finally, the recovery phase received the least attention from scholars (14 papers). This phase consists of both short-term and long-term activities that are designed for reestablishing and returning disaster areas to normal conditions. Note that the total sum according to these phases (102 papers) is greater than the number of reviewed articles (65 papers) since some studies focused on more than one phase.

Mobile phone data consist of information collected by mobile carriers, sensors, and apps on mobile phones. Mobile phone data collected by carriers include CDRs, SMS (short message service) information, traffic volume, etc., and data collected by phone sensors, such as GPS records, Bluetooth-sensed interaction data, etc.

Most of the studies applied CDRs (34 papers) for emergency analysis. This type of mobile phone data contains details about each call such as "the location, call duration, call time, and both parties involved in the conversation" (Trestian et al. 2017) . CDRs containing both users' spatial and temporal information can support research on modeling human mobility during emergencies. Moreover, information about caller and recipient IDs reveals individuals' social networks, which correlate to infectious disease dissemination (Gundogdu et al. 2016; Wesolowski et al. 2014) . Some studies combined CDRs with other data sources to obtain more comprehensive emergency-related information (Bharti et al. 2015; Pastor-Escuredo et al. 2014) . Bharti et al. (2015) used both CDRs and nighttime lighting data from satellite imagery to analyze population sizes and human mobility (the CDRs for short-term and the satellite data for long-term assessment), which helped in making policies and understanding emergency impacts in the response phase. Pastor-Escuredo et al. (2014) adopted both CDR and rainfall-level data in Mexico to help discover anomalous mobile phone usage patterns in seriously affected areas and assess infrastructure damage and casualty populations in time.

SMS (seven papers) is considered as another category of mobile phone data (separate from CDR in this study), which contains both passively collected information like communication details and actively collected information. GPS data (11 papers) can provide the location information of individuals with higher accuracy than CDRs and can be helpful in the identification of individual locations as well as the study of human dynamics. In addition, there were some studies being developed based on app data (six papers) and others (nine papers), such as Bluetooth-sensed interaction data, mobilephone-usage data, mobile-traffic data, etc.

When applying mobile phone data to practical problems, scholars have to gather and extract useful information from the raw mobile phone data. We have categorized different information processing paths from six analysis perspectives: human mobility, geographic location, social networks, mobile phone usage patterns, collected information, and information diffusion (Blondel et al. 2015) . Detailed definitions for these analysis perspectives are exhibited in Table 2 (refer to Appendix 2) with specific examples.

As depicted in Fig. 3 , spatial-temporal information extracted from CDR and GPS data has been mostly investigated by scholars, which is reflected as human mobility (30 papers). The spatial and temporal information can benefit the tracking of individuals' trajectories and modeling their movement patterns. Since the relationship between human movement patterns and infectious transmission routes were found (Blondel et al. 2015) , the analysis of human mobility contributes to the understanding of the disease dissemination process.

Mobile phone usage patterns perspective (10 papers) refers to various individual behavior gathered from one's mobile phone data, such as normal call volume. Arai et al. (2015) found clues to the whereabouts of unobserved populations by analyzing individuals' usage behaviors from CDRs, especially for the whereabouts of children, who are vulnerable to epidemics. This result contributes to tracking epidemic dissemination and the deployment of various resources.

Social networks (10 papers) are constructed based on the communication information gathered from mobile phone data. The analysis of social networks helped to reproduce disease dissemination models and predicted epidemic transmission for the preparedness phase (Y. Chen et al. 2017; Farrahi et al. 2015) .

Scholars also used spatial information extracted from mobile phone data (11 papers) at both the individual and aggregate level. The analysis of individual geographic location benefited in tracking anomalous individual locations, and could be further applied to develop pre-warning systems, such as systems to lessen potential damage on natural gas pipelines (Dong et al. 2017) .

The collected information (seven papers) perspective refers to the process of gathering multiple types of information content from mobile phones, such as using a specific app to collect public opinions. Deng et al. (2016) collected public In addition, other studies focused on a more general perspective of gathering information from mobile phone data, and this kind of information was defined as information diffusion (five papers). Diffusion strategies of prevention knowledge could be developed to mitigate the impacts of emergencies by analyzing information diffusion networks (Lima et al. 2015) .

Research in EM mainly focused on five kinds of problems: resource management, evacuation, pre-planning, decisionmaking, and education and training (Mingliang et al. 2006) . Within these five types, various issues (such as making evacuation plans, optimizing resource allocation, predicting epidemic transmission, etc.) are defined as applications in this study.

Although the selected studies developed their unique applications, this study provides a relatively broad classification, five general categories and 13 subcategories to help create a map of the existing applications. This classification is exhibited in Table 3 (refer to Appendix 3) with a detailed definition and a specific example for each category.

Most studies focused on the decision-making problem (42 papers), among which 'conducting public health intervention' (12 papers) and 'processing real-time detection' (12 papers) had been mostly developed. It was followed by presenting emergency impact (seven papers), 'stating policy/regulations' (five papers), 'making construction plans' (four papers), and 'developing emergency-related platforms' (two papers). The decision-making problem aims to give guidance for the relevant work in the EM process.

The pre-planning problem has been studied by 15 papers during these five years, among which eight papers focused on 'predicting epidemic transmission' and seven papers focused on 'developing pre-warning system.' This kind of application aims to anticipate what might happen and to provide an early warning.

Within the evacuation problem (14 papers), 'finding victims' (five papers) and 'making evacuation plans' (nine papers) were studied. It aims to solve the issues of finding and rescuing disaster victims.

Within the education and training problem (seven papers), 'delivering emergency announcement' (five papers) and 'guiding psychological recovery' (two papers) were explored. This kind of problem focuses on spreading emergency-related knowledge for the public and helping the public to face the emergency rationally.

The remaining problem type was the resource management problem (five papers). This kind of problem focuses on the appropriate allocation of both material and human resources according to the distributions of the victims and rescuers.

Our framework for using mobile phone data for EM is depicted in Fig. 4 and the detailed process is described in Section A.2. The framework synthesizes the five aforementioned themes (i.e., emergency situations, EM phases, types of applications, analysis perspectives, and types of mobile phone data) and illustrates two logical routes with potential correspondences between each theme. The boxes in the framework represent the categories under each theme, and the lines with arrows represent the existing correlation between each theme according to the 65 reviewed studies.

The first type of logical route is the decision-making route (represented as blue lines with arrows in the framework). It starts with the emergency situation theme, goes through the phases of EM, and ends with the application theme. This route assists managers to make emergency-related decisions comprehensively and conveniently during the emergency. They can make specific emergency plans involved in each phase of EM by following the steps: (i) judging what types of emergency the public are encountering; (ii) determining the phases of the emergency in different regions; (iii) figuring out general categories of applications related to the determined phase, and then referring to Table 3 and Fig. 6 (in Appendix 3) for subapplications ought to be taken; (iv) determining procedure and activities that can be implemented in practice by considering both the characteristics of the encountered emergency and referred applications. For example, during COVID-19, relevant managers can identify the situation as a disease disaster. If the managers judge that the phase of the pandemic is in the 'mitigation,' he/she can take applications in pre-planning, education and training, and decision-making categories into account according to our framework. After referring to Table 3 and Fig. 6 , they can consider implementing applications, like The second type of logical route is the problem-solving route (represented as red lines with arrows in the framework). It ends with the application theme, but starts from the mobile phone data theme while goes through the analysis perspective. This route assists technicians to devise solutions for emergency plans. They can employ mobile technology to implement the identified activities by following the steps: (i) understanding the activity involved in emergency plans given by the decision-maker, i.e. what application is this activity; (ii) identifying appropriate mobile technology that can assist in implementing the application; (iii) gathering information and making appropriate analysis (refer to Table 2) by using the identified mobile technology; (iv) applying the results of analysis into real-world EM. For example, to predict epidemic transmission during COVID-19 pandemic, technicians can find CDR and GPS data appropriate for gathering individuals' spatial-temporal information and analyzing human mobility. They can apply the analysis to technical solutions development and decision-support system.

The framework also contributes to extending and enriching our understanding of the evolving literature of EM. Links in the framework represent existing studies in this field, thereby illuminating the established correlations between the themes and indicating potential research gaps. For example, the studies on education and training applications mainly focus on the disease disaster and man-made disaster, while the focus on natural disasters is limited. Meanwhile, studies only analyze this application from four perspectives while information diffusion and geographic location are missing. Researchers may further consider the feasibility of these two perspectives to analyze this application.

The outbreak of the COVID-19 pandemic in early 2020 has caused huge impacts on global political, economic, and social development (WHO 2020d; W. Chen and Bo 2020). Both emergency-related managers and technicians can refer to the proposed framework for potential decision-making and problem-solving.

Drawing on the decision-making route, emergency managers can determine emergency plans involved in each phase of managing the COVID-19 pandemic. Existing activities under the pandemic fit four categories of applications including predicting epidemic transmission (Jia et al. 2020; Iacus et al. 2020; Gatto et al. 2020) , public health intervention (Magklaras and Nikolaia Lopez Bojorquez 2020; Ekong et al. 2020; Canada 2020) , delivering emergency (Barugola et al. 2020; Shi and Jiang 2020; Xinhua 2020b) , and stating policy/ regulations (Speakman 2020; Hu and Zhu 2020; Canada 2020; Singapore 2020) . Specifically, in the mitigation phase, related organizations have predicted epidemic transmission through monitoring population flows to reduce potential risks (Jia et al. 2020; Iacus et al. 2020; Gatto et al. 2020 ). In the preparedness phase, inspectors and the public have been trained and mobilized to use mobile phone security codes, which help track infection cases and contacts (AGDH 2020; Guinchard 2020; Zastrow 2020). In the response phase, the World Health Organization has announced global objectives including conducting public health intervention by rapidly tracing, finding, and isolating all cases and contacts (WHO 2020b; Cozzens 2020; Magklaras and Nikolaia Lopez Bojorquez 2020) . And in the recovery phase, some governments have used mobile technology to aid business resumption with contact tracing applications (Thompson 2020a (Thompson , 2020b Sunil 2020 ; Devonshire-Ellis 2020; Hu and Zhu 2020).

Drawing on the problem-solving route in our framework, relevant technicians can analyze various collected information from mobile devices to implement the aforementioned applications. First, researchers have used CDR data (Jia et al. 2020 ) and mobile positioning data (Gatto et al. 2020; Iacus et al. 2020 ) to predict epidemic transmission by capturing and simulating population movements. They have mentioned that their prediction model can conduct risk assessments and plan limited resources allocation as well. Second, contact tracing requires various mobile tools including GPS (Cozzens 2020), Bluetooth (WHO 2020c; Xinhua 2020a), apps (Zhang et al. 2020; Singapore 2020; AGDH 2020) , and SMS technology (WHO 2020c) to gather individuals' geographic and health-related information. The analysis of this information can be further applied to conduct public health intervention such as isolation strategies and travel restrictions. Finally, as for business resumption, history trajectories and health status of individuals collected through specific mobile phone apps (Zastrow 2020), SMS (WHO 2020a; McCabe 2020) and GPS technology (Elliott 2020) helps to evaluate their infection risk, which can be used by governments to resume business and study.

Our framework not only covers applications that have already been adopted in this EM, but also provides a reference for future EM. For example, it is possible to consider other applications in our framework such as guiding psychology recovery in the context of COVID-19. The use of hedonic apps after an earthquake has been identified to reduce perceived risk effectively (Jia et al. 2017) . Therefore, decisionmakers can learn from this knowledge and implement appropriate applications in the recovery phase of managing this pandemic.

Future research directions can be implied based on the proposed framework from five perspectives: (1) a focus on manmade emergencies, (2) a focus on the recovery phase, (3) exploring new applications, (4) creating better comprehension of the analysis perspective, and (5) combining other data with mobile phone data.

It is reasonable to put forward further research on man-made emergencies with mobile phone data. The results shown in Section 3.1 indicate a greater emphasis on natural emergencies (41 articles reviewed in total) and less emphasis on man-made emergencies (16 articles reviewed in total). However, managing violence and terror incidents is also very vital, because these emergencies happen frequently around the world. For such a crisis situation, it is challenging to gather their information due to the dark side and dynamic nature (Roberts 2011; Skillicorn 2011; Chen et al. 2011) .

In view of these challenges, current literature has explored the suitability of using social media data (Oh et al. 2011; Cheong and Lee 2011; Prentice et al. 2011; Qin et al. 2011 ). However, lacking individuals' real-time location and objective behavior information makes social media data limited. Some studies have identified the feasibility of using mobile phone data in man-made emergencies, such as detecting terrorism attacks and monitoring traffic conditions (Blondel et al. 2015) . However, the application of mobile phone data in this domain still has great potential.

A focus on the recovery phase would also be a future direction. The occurrence of emergencies will have both physical and psychological impacts on the victims. However, most scholars focused on the timely and effective response to emergencies, while fewer studies focused on the psychological recovery of people after emergencies. With the popularity and development of 4G/5G communication technology, many people use their mobile phones for obtaining emergencyrelated information and entertainment to ease their anxiety. For example, Jia et al. (2017) discovered that the use of hedonic apps after an earthquake can help people reduce their perceived risk. Therefore, there is the potential to use mobile phone data for the study of developing positive psychological structures in the recovery phase.

Additional applications can be explored and expanded according to the two ways demonstrated in the framework. With a further understanding of analysis perspectives (which are discussed in Section 6.4), scholars can explore additional applications. For example, with a more comprehensive notion of human social interaction, scholars may not only investigate approaches in predicting epidemic transmission, but also apply this knowledge to develop systems in predicting crimes. In addition, with the upgrading of mobile phones and ICT (Bandyopadhyay et al. 2018; Palshikar et al. 2018) , new applications can be explored under the proposed framework. As a result, additional applications can be similarly explored within and beyond the existing applications in the future.

Future work should also make efforts to complement the theoretical foundation of emergency studies with theories from other fields. Current literature has developed theories about information transmitted among various stakeholders (Wang et al. 2018; Weidinger et al. 2018; Liu and Xu 2018; Abedin and Babar 2018; Fedorowicz and Gogan 2010) , and theories of coordination and political science (Maldonado et al. 2010 ). Moreover, a better comprehension of the analysis perspective can be made by learning constructs about human behavior, social networks, etc. For example, scholars have found similarities between human communication and infectious disease dissemination (Blondel et al. 2015) , which indicates the benefits of drawing on theories from social networks to apply to EM. Meanwhile, the current understanding of human mobility is mainly based on data analysis. If scholars can learn and apply aspects from relevant psychological and behavioral theories, the analysis of human mobility can be further deepened.

In addition, with a better comprehension of analysis perspectives and applications, it is possible to develop new correspondences between analysis perspectives and applications. The current framework was developed based on the 65 articles reviewed, and cannot reveal every possible relationship between analysis perspectives and applications. For example, human mobility can probably be analyzed in order to construct pre-warning systems. According to the study of daily movement patterns of individuals, a detected anomaly of movement may indicate a coming disaster, which would suggest the need to construct correlations between human mobility and pre-warning systems. Consequently, the correspondence between analysis perspectives and applications should be considered and expanded upon in future research.

Although mobile phone data have been successfully applied to EM, combining other micro and macro data can help the development of EM research more efficiently (Ghosh et al. 2018) . Martinez-Rojas et al. (2018) have reviewed 158 articles about using Twitter to manage emergencies, indicating the significant roles and value of data from social media in EM. Under what circumstances it is appropriate and how to combine mobile phone data with micro (e.g., individual locations) and macro (e.g., public opinions) data from social media requires further exploration. Moreover, combining data from other platforms is worth consideration as well. For instance, Pastor-Escuredo et al. (2014) believe there is potential in combining information from officially monitored sensors like traffic video cameras, which can provide a fine-grained validation for the existing measures.

Emergencies have great impacts on the whole of society, affecting material facilities and social order in terms of economic losses and human casualties. Unlike traditional management measures based on data sources with limited adaptability and low accuracy, applications for handling emergencies that utilize ubiquitous and real-time mobile phone data have greatly improved EM mechanisms and minimized their negative impacts on society. This systematic literature review analyzes 65 studies concerning the use of mobile phone data for EM, and proposes a framework to synthesize the fragmented knowledge of existing studies. The framework comprises five themes, among which six analysis perspectives and five general types of applications are put forward to explain the EM process, which includes two logical routes. The framework can support stakeholders, such as emergency managers and technicians, and is used to suggest five future research directions in the field for scholars. In addition, this study discusses EM under the COVID-19 pandemic and provides a reference for future management of the pandemic.

Despite all the contributions mentioned above, this study still possesses some inevitable limitations. First, the common limitation of literature reviews related to keywords exists in this study as well. Making the best efforts to alleviate this drawback, this study draws on keywords about EM from previous reviews and research works in the field of EM as well as keywords about mobile phone data. The second limitation is that the framework develops in this study is only based on the articles reviewed, meaning that relationships not mentioned in the articles are not considered. Although this review is intended to be both wide and deep in coverage, it should not be considered as a complete or final summary of the topic. Reviews, no matter how current, by definition focus on the past and cannot fully anticipate novel approaches or new developments. Finally, this study does not focus on the detailed introduction of data processing and analyzing techniques. Nevertheless, we still believe this study makes a strong contribution to the field, especially toward emergency managers and scholars who are looking for direction to develop this field in the future.

Acknowledgments The corresponding author is a Tang Scholar. He would like to acknowledge the support from National Natural Science Foundation of China (Grant No. 91746111) , The Key Research and Dev elop m ent Pro gram of Sh aanx i Prov inc e (Gra nt No. 2020ZDLGY09-08).

Authors' Contributions YW, JL, XZ, GF, XL conceived the study. YW, JL carried out the review, synthesized and analyzed the evidence, and drafted the manuscript. XZ, XL, GF supervised the review process and revised the manuscript. All authors read and approved the final manuscript.

Planning Identifying the research questions. Our research questions were as follows: (1) What types of mobile phone data-with respect to its applied phases and use to cope with practical issues-have been studied? (2) What is the state of the art of this field? and (3) What can future works develop to facilitate the understanding of this subject?

Regarding the first question, a statistical analysis was performed on the aspects of emergency situations, phases of EM, types of mobile phone data, analysis perspectives, and applications. Next, a framework of using mobile phone data for EM to address the second question was proposed. Finally, five future implications based on the proposed framework are presented in order to address the third question.

Identifying relevant studies. The second step of planning the research was identifying relevant studies, which defines the scope of this review study. Six document databases were searched to find related studies between 2014 and 2019. These were: ScienceDirect, Scopus, Web of Science, IEEE Xplore, ACM, and Springer. IEEE Xplore and ACM are two specialty article databases that provide extensive coverage of the literature in computer science and related areas, and Scopus (SciVerse Scopus) is the largest abstract and citation database. The other three were additional comprehensive and widely searched databases.

To draw the boundaries of what articles would be included and reviewed in the study, the phenomenon of interest was identified as 'research that applies mobile phone data to manage emergencies.' We developed an initial list of research keywords to match the definition with published documents and considered various literature expressions that represented the same terminology by considering both aspects of emergencies and mobile phone data. This method is illustrated at the beginning of the conducting phase and further guided the search for related articles.

Generating a research strategy. The first step was to generate a research strategy by finding and filtering studies from the six databases. Two iterations were processed: (1) searching 26 terms in the keywords list ("mobile phone data" OR "short message service" OR "call detail record" OR "phone GPS data" OR "cellular network data" OR "app data" OR "application data" OR "Bluetooth data") AND ("emergency" OR "extreme situation" OR "extreme event" OR "large-scale event" OR "special event" OR "special situation" OR "anomalous event" OR "anomalous situation" OR "unusual event" OR "unusual situation" OR "crisis" OR"disaster" OR "catastrophe" OR "traffic accident" OR "epidemics" OR "infectious disease") AND (2013 < PUBYEAR<2019); (2) searching papers in the reference list of the five previously identified review articles and including additional studies.

In the first iteration, 4784 papers (1880 ScienceDirect +1542 Springer +101 IEEE +715 ACM + 276 Web of Science +170 Scopus) were initially identified. In the second iteration, 13 additional articles were included as relevant for further selection by searching the reference lists. Accordingly, 4797 articles were found through the process of this stage (Fig. 5) .

Selecting primary studies. The second step was to select the studies to be reviewed through a standard inclusion and exclusion criteria. The specific items are depicted in Table 1 . The inclusion criteria I1 and I2 ensured that the selected studies were in accord with mobile phone data and applying it in EM, which aligned with the objectives of this study. The exclusion criterion E1 eliminated studies that were found by the keyword "emergency," but discussed emergency-related activities instead of an emergency situation itself or did not discuss a specific situation at all, for example, studies of efficient deployment of emergency departments or the contents and answers of emergency calls. E2 excluded studies that analyzed the use of mobile phones for self-rescue when individuals encountered difficulties, like heart attacks. Such situations were considered to be emergencies for a single person, which did not fall within the scope of this study. E3 excluded studies with a discussion of non-emergency situations, which were planned events like concerts, festivals, and football matches. All of these criteria were considered to point out the most relevant studies for the topic and improved the reliability and validity of the review study. Finally, in order to mitigate bias, two of the authors conducted this step using the above criteria. After discussion, 65 articles were chosen for the analysis and report in this review (refer to the process illustrated in Fig. 5) .

Extracting data. The third step was to extract data from the studies that were selected. The following information was extracted from the 65 studies: (i) document demographic information, including the title, year of publication, journal name or conference name, and authors; (ii) information about emergency situations, including the general types of emergency and specific events; (iii) types of mobile phone data used in emergency situations; (iv) methods to apply such data in the situations and objectives of the study (what perspectives it analyzed and to which process of EM the study belongs). The detailed results are presented in Section 3.

Based on the information extracted, the analysis perspectives and applications that were proposed were summarized to create a clearer idea of how to apply mobile phone data to effective management (refer to SectionS 3.4 and 3.5 for details). For example, individual movement patterns were studied by Stefania et al. (2018) and Vogel et al. (2015) to either establish new models or develop existing models, and thus help prevent disease dissemination, while aggregated population mobility was analyzed by Sekimoto et al. (2016) to benefit government policymaking. In this study, the phrases 'individual movement patterns' and 'aggregated population mobility' were combined and expressed as 'human mobility,' which represents a kind of analytic perspective. In addition, objectives (or final applications) of the reviewed articles like 'making rescue plans,' 'making traffic regulations,' and 'helping in policy decisions' were paraphrased as 'making policies'. In this way, the current study transferred the metainformation extracted from the reviewed studies into a collective and scientific form for the conclusions and future discussion.

Preparing the framework for using mobile phone data for emergency management. The final step in the conducting phase was to propose the plan for developing a framework that could illustrate the current state of research. First, we drew on the typology of mobile data sources, types, and uses in the disease disaster management cycle proposed by Cinnamon et al. (2016) , and we planned to adjust the framework to adapt to the data from this study. Second, we considered "emergency situations" and "mobile phone data" as two starting points, as both aspects were key features of the subject. Third, considering the practicality of this framework, it was desired to expand the framework to the perspectives of stakeholders 

Articles which focus on emergency items (e.g. emergency call or emergency department) other than emergency situations 2 I2. Articles which contain the description of mobile phone data types with respect to emergencies.

E2. Articles which depict emergencies encountered by a single person instead of society (e.g., a heart attack) 3

E3. Articles which describe non-emergency situations (e.g., festivals or concerts) 4 E4. Articles for which full text cannot be found 5

E5. Articles which are not written in English from both managerial and technical layers, respectively. Therefore, five themes (situations, phases in EM, mobile phone data, analysis perspectives, and applications) were identified as components in the framework. Based on these themes, we realized that "applications" served as the final consideration regardless of the perspective. Finally, we developed a draft of the framework which started from "emergency situations" and "mobile phone data" and ended in "applications." It consisted of two logical routes: the decision-making, or managerial, route ("emergency situation", "phases in emergency management", and "applications"); and the problemsolving, or technical, route ("mobile phone data", "analysis perspective", and "applications"). The human mobility perspective refers to capturing citizens' spatial-temporal change patterns, like travel, commuting, migration, etc., from real-time mobile phone data. This article utilized mobile-phone Bluetooth-sensed data to reflect human interactions and compared them to actual infection cases to simulate the spread of seasonal influenza (Farrahi et al. 2015) .

The geographic location perspective refers to gathering personal space information from CDRs and GPS data to illustrate geographic networks at both an individual and aggregate level. This article used mobile-phone GPS-location data to find abnormalities close to a pipeline as a way to detect damaging activities (Dong et al. 2017) . The collected information perspective refers to gathering multiple types of information content from mobile phones, such as comments and views collected through SMS and apps. This article introduced an app that collected online opinions about emergency management and applied this data into assisting the decision-making of governments (Deng et al. 2016 ). The information diffusion perspective refers to the network of information propagation via the phone, SMS, or app, which is a general process of gathering information from mobile phone data.

N. Zhang et al. (2016) This article analyzed the information dissemination mechanisms of calls and SMS messages to validate their effectiveness as pre-warning approaches in reducing losses during emergencies (Nan Zhang et al. 2014) . Processing real-time detection refers to capturing citizens' anomalous behavior to detect possible emergencies in real-time with appropriate accuracy for planning response efforts. An anomaly detection system was developed by connecting exceptional spatial-temporal patterns from mobile data with real-world emergencies (Trestian et al. 2017) .

Developing emergency-related platforms refers to implementing a better platform to collect multi-source data which eventually enhances the efficiency of EM. This study introduced a platform containing multiple kinds of mobile phone data to implement various intervention measures into the whole management process (Poblet et al. 2018) . Making construction plans refers to scientifically building infrastructure to minimize the loss in an emergency or optimizing reconstruction projects. Two isolation strategies for controlling epidemics, at the subprefecture and individual level, were proposed based on a model of citizens' trajectories (Stefania et al. 2018) .

Stating policy/regulations refers to related governments drawing experience from current and previous emergencies and utilizing regularities to manage emergencies. This study represented the population size changes after a political conflict as supporting information provided for governments (Bharti et al. 2015) .

Pre-planning (Total number 15)

Developing pre-warning system refers to integrating multiple characteristics concerning diverse emergencies obtained from the analysis stage to give early warning. This study validated that a mobile phone dataset performed better than traditional survey data in representing commuting patterns to simulate an epidemic spread (Panigutti et al. 2017) .

Finding victims refers to providing information or clues about victims' whereabouts for responsible authorities. This study analyzed citizens' evacuation routes after a subway accident to optimize future evacuation organizing work (Duan et al. 2017 ).

Guiding psychological recovery refers to discovering people's anomalous behaviors after emergencies to specifically provide psychological counseling for them Jia et al. (2017) ; Baytiyeh (2018) (Total number two)

This study found that hedonic behavior would reduce perceived risks by studying app usage changes in a disaster and recommend hedonic app using for the public after disasters (Jia et al. 2017) . Delivering emergency announcement refers to propagating knowledge or messages about emergencies to help citizens prepare for or respond to them. This study improved the inference of the hotspot of epidemics based on human mobility mining to optimize resource deployment (Matamalas et al. 2016 ).

Appendix 4. Characteristic matrix of the reviewed studies 

♦ CI processing real-time detection, presenting emergency impacts ♦ 'AP' stands for 'Analysis perspective' and 'EM phases' represents 'phases of emergency management'. Six analysis perspectives are respectively human mobility (HM), social networks (SN), mobile phone usage pattern (MPUP), information diffusion (ID), geographic location (GL) and collected information (CI). The definitions of AP and Applications are consistent with Tables 2 and 3 . 

Institutional vs. non-institutional use of Social Media during Emergency response: a case of Twitter in 2014 Australian bush fire

COVIDSafe application

The viability of Mobile services (SMS and cell broadcast) in emergency management solutions: An exploratory study

RiSC: Quantifying change after natural disasters to estimate infrastructure damage with mobile phone data

The built environment and Syrian refugee integration in Turkey: an analysis of mobile phone data

Understanding the unobservable population in call detail records through analysis of mobile phone user calling behavior a case study of greater Dhaka in Bangladesh

Smartphone geospatial apps for dengue control, prevention, prediction, and education: MOSapp, DISapp, and the mosquito perception index (MPI)

An embedding based IR model for disaster situations

Stay safe stay connected: Surgical mobile app at the time of Covid-19 outbreak

The uses of mobile technologies in the aftermath of terrorist attacks among low socioeconomic populations

Using Mobile phone data to predict the spatial spread of cholera

Disaster management and information systems: Insights to emerging challenges

Remotely measuring populations during a crisis by overlaying two data sources

A survey of results on mobile phone datasets analysis

Performing systematic literature reviews in software engineering

Travel restriction measures: COVID-19 program delivery

Data fusion for city life event detection

Novel coronavirus named COVID-19 by WHO

Introduction to special issue on terrorism informatics

Reality mining: A prediction algorithm for disease dynamics based on mobile big data

A microblogging-based approach to terrorism informatics: Exploration and chronicling civilian sentiment and response to terrorism events via twitter

Evidence and future potential of mobile phone data for disease disaster management

19 countries track mobile location to fight COVID-19

Delay-aware accident detection and response system using fog computing

AREA: A Mobile Application for Rapid Epidemiology Assessment

CPS model based online opinion governance modeling and evaluation of emergency accidents

COVID-19 in China: business lose less, work resumes faster than expected

Spatiotemporal detection of unusual human population behavior using Mobile phone data

Use of community mobile phone big location data to recognize unusual patterns close to a pipeline which may indicate unauthorized activities and possible risk of damage

Understanding evacuation and impact of a metro collision on ridership using large-scale mobile phone data

COVID-19 Mobile positioning data contact tracing and patient privacy regulations: Exploratory search of global response strategies and the use of digital tools in

COVID-19: Balancing response and recovery

Emergencies do not stop at night: Advanced analysis of displacement based on satellite-derived nighttime light observations

Predicting a Community's Flu Dynamics with Mobile Phone Data

Reinvention of interorganizational systems: a case analysis of the diffusion of a bio-terror surveillance system

Mobile phone data highlights the role of mass gatherings in the spreading of cholera outbreaks

Precision global health in the digital age

Design patterns for emergency management: An exercise in reflective practice

Quantifying information flow during emergencies

Association between mobile phone traffic volume and road crash fatalities: a population-based case-crossover study

Anomaly detection mechanisms to find social events using cellular traffic data

Spread and dynamics of the COVID-19 epidemic in Italy: Effects of emergency containment measures

What drives knowledge sharing in software development teams: a literature review and classification framework

Exploitation of social media for emergency relief and preparedness: recent research and trends

A framework to model human behavior at large scale during natural disasters

Our digital footprint under Covid-19: Should we fear the UK digital contact tracing app? International Review of Law Computers & Technology, 14

Countrywide arrhythmia: emergency event detection using mobile phone data

Flood disaster indicator of water level monitoring system

Investigating evidence of mobile phone usage by drivers in road traffic accidents

Xi: We will continue to speed up the restoration of work and life order in the normal epidemic prevention and control process (in Mandarin Chinese

Explaining the initial spread of COVID-19 using mobile positioning data: a Case Study

Information intermediaries for emergency preparedness and response: a case study from public health

The geo-observer network: a proof of concept on participatory sensing of disasters in a remote setting

Advances in multi-agency disaster management: key elements in disaster research

The role of hedonic behavior in reducing perceived risk: evidence from Postearthquake Mobile-app data

Population flow drives Spatio-temporal distribution of COVID-19 in China

Population distribution modelling at fine spatiotemporal scale based on mobile phone data

Use of short message service for monitoring Zika-related behaviors in four Latin American countries: lessons learned from the field. mHealth

Event-cloud platform to support decision-making in emergency management

Disease Containment Strategies based on Mobility and Information Dissemination

Social roles and consequences in using social Media in Disasters: A Structurational perspective

A 21st century approach to tackling dengue: Crowdsourced surveillance, predictive mapping and tailored communication

Development of GIS integrated big data research toolbox (BigGIS-RTX) for Mobile CDR data processing in disasters management

A review of information security aspects of the emerging COVID-19 contact tracing mobile phone applications

Collaborative systems development in disaster relief: the impact of multi-level governance

Twitter as a tool for the management and analysis of emergency situations: a systematic literature review

A data-driven impact evaluation of hurricane Harvey from mobile phone data

Assessing reliable human mobility patterns from higher order memory in mobile communications

How to Create a COVID-19 SMS Self-Assessment Tool

A research review on the public emergency management

Cell phones and motor vehicle fatalities

Large-scale Mobile traffic analysis: a survey

Disasters will happenare you ready?

Information control and terrorism: tracking the Mumbai terrorist attack through twitter

Model-driven disaster management

No-notice urban evacuations: using Crowdsourced Mobile data to minimize risk

Weakly supervised and online learning of word models for classification to detect disaster reporting tweets

Assessing the use of mobile phone data to describe recurrent mobility patterns in spatial epidemic models

Flooding through the Lens of Mobile phone activity

Population mobility reductions associated with travel restrictions during the Ebola epidemic in Sierra Leone: use of mobile phone data

Introduction to emergency management

Crowdsourcing roles, methods and tools for data-intensive disaster management

Analyzing the semantic content and persuasive composition of extremist media: a case study of texts produced during the Gaza conflict

A multi-region empirical study on the internet presence of global extremist organizations

Advanced methods of cell phone localization for crisis and emergency management applications

Tracking and disrupting dark networks: challenges of data collection and analysis

A scenario-based study on information flow and collaboration patterns in disaster management

ICT facts and figures 2017

A review on security challenges of wireless communications in disaster emergency response and crisis management situations

Real-time people movement estimation in large disasters from several kinds of mobile phone data

Covid-19 prevention and control tips, the least we can do in the face of the epidemic (in Mandarin Chinese

TraceTogether' contact tracing application website

Computational approaches to suspicion in adversarial settings

The world needs to follow China methods to combat epidemic

Traffic incidents in motorways: an empirical proposal for incident detection using data from mobile phone operators

A comparison of spatialbased targeted disease mitigation strategies using mobile phone data

Improving emergency response logistics through advanced GIS

Guidlines for business resumption in Singapore, post-COVID-19 taskforce, and more

Mobile applications in crisis informatics literature: a systematic review

Design and operation of app-based intelligent landslide monitoring system: the case of three gorges reservoir region

Integrating rapid risk mapping and mobile phone call record data for strategic malaria elimination planning

20 expert-recommended school supplies for a safe transition back to school

Your 2020 back-to-school checklist to protect against Covid-19

On the use of human mobility proxies for modeling epidemics

Migration statistics relevant for malaria transmission in Senegal derived from mobile phone data and used in an agent-based migration model

Towards connecting people, locations and real-world events in a cellular network

An investigation of interaction patterns in emergency management: a case study of the crash of continental flight 3407

Mining Mobile datasets to enable the finegrained stochastic simulation of Ebola diffusion

A capability assessment model for emergency management organizations

Is the frontier shifting into the right direction? A qualitative analysis of acceptance factors for novel firefighter information technologies

Quantifying travel behavior for infectious disease research: a comparison of data from surveys and mobile phones

Quantifying seasonal population fluxes driving rubella transmission dynamics using mobile phone data

Evaluating spatial interaction models for regional mobility in sub-Saharan Africa

Multinational patterns of seasonal asymmetry in human movement influence infectious disease dynamics

COVID-19 message library

Critical preparedness, readiness and response actions for COVID-19

Digital tools for COVID-19 contact tracing

WHO announces COVID-19 outbreak a pandemic

Bluetooth contributes to accurate COVID-19 control

Tech tools stretch anti-virus battle's grassroots reach

CityFlowFragility: Measuring the fragility of people flow in cities to disasters using GPS data collected from smartphones

Fusion of terrain information and Mobile phone location data for flood area detection in rural areas

Crosscomparative analysis of evacuation behavior after earthquakes using mobile phone data

Mobile phone data reveals the importance of pre-disaster inter-city social ties for recovery after hurricane Maria

Spatial and temporal analysis of human movements and applications for disaster response management utilizing cell phone usage data

Improving emergency evacuation planning with mobile phone location data

Coronavirus contact-tracing apps: can they slow the spread of COVID-19?

Information dissemination analysis of different media towards the application for disaster pre-warning

Comprehensive analysis of information dissemination in disasters

Americans' perceptions of privacy and surveillance in the COVID-19 pandemic

Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations

She is currently pursuing the Ph.D. degree in Management Science and Engineering at the school of Management, Xi'an Jiaotong University. Her main research interests are big data and behavior analysis, social media and gamification

He is currently pursuing the Ph.D. degree in Management Science and Engineering at the School of Management, Xi'an Jiaotong University. His main research interests are massive human behavior analysis and blockchain finance exploration

He conducted research in the fields of data analytics and pattern recognition as a Research Assistant Professor in the

He obtained the B.S. degree in computer science in 1987, the M.S. degree in systems engineering in 1990, and the Ph.D. degree in management engineering in 1993, all from Xi'an Jiaotong University of China. His research interests include logistics and supply chain management, information system management

His research interests center around information assurance, innovative technologies for strategic decision-making, and global IT management. He is the Co-Editor-in-Chief for the International Journal of Accounting and Information Management.

♦

Inf Syst Front