Portal de Periódicos da CAPES

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

2017; Institution of Engineering and Technology; Volume: 11; Issue: 8 Linguagem: Inglês

10.1049/iet-its.2016.0112

1751-9578

Zhengyu Duan, Zengxiang Lei, Michael Zhang, Weifeng Li, Jia Fang, Jian Li,

Urban Transport and Accessibility

IET Intelligent Transport SystemsVolume 11, Issue 8 p. 511-520 Research ArticleFree Access Understanding evacuation and impact of a metro collision on ridership using large-scale mobile phone data Zhengyu Duan, Corresponding Author Zhengyu Duan d_zy@163.com Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this authorZengxiang Lei, Zengxiang Lei Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this authorMichael Zhang, Michael Zhang Department of Civil and Environmental Engineering, University of California, Davis, One Shields Avenue, Davis, CA, 95616 USASearch for more papers by this authorWeifeng Li, Weifeng Li Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this authorJia Fang, Jia Fang Architecture and Urban Planning Institute for Advanced Study, Tongji University, 1239 Siping Road, Shanghai, 200092 People's Republic of ChinaSearch for more papers by this authorJian Li, Jian Li Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this author Zhengyu Duan, Corresponding Author Zhengyu Duan d_zy@163.com Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this authorZengxiang Lei, Zengxiang Lei Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this authorMichael Zhang, Michael Zhang Department of Civil and Environmental Engineering, University of California, Davis, One Shields Avenue, Davis, CA, 95616 USASearch for more papers by this authorWeifeng Li, Weifeng Li Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this authorJia Fang, Jia Fang Architecture and Urban Planning Institute for Advanced Study, Tongji University, 1239 Siping Road, Shanghai, 200092 People's Republic of ChinaSearch for more papers by this authorJian Li, Jian Li Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao'an Road, Shanghai, 201804 People's Republic of ChinaSearch for more papers by this author First published: 21 August 2017 https://doi.org/10.1049/iet-its.2016.0112Citations: 8AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract As randomly occurring events, traffic accidents pose serious challenges to the collection of comprehensive data to understand how travellers respond to them and to quantify their impacts. The advent of mobile phones, with their wide spatial/temporal coverage and ubiquitous presence in metro areas, offers a new source of data to conduct such studies. In this study, the collision accident of Metro Line 10 in Shanghai, China on 27 September 2011 is carefully investigated based on data derived from anonymous mobile phone records. The evacuation process of the accident is studied, followed by an analysis of the impact of this accident on commuting in the city. After analysing 7 billion of mobile phone records for an 11-day period, the authors find that the evacuation follows a two-stage pattern. They then identify the commuters of Line 10 and study their commuting patterns in the day of accident and also in the subsequent days. They find that most of Line 10 commuters still preferred to use metro to complete their travels during the disruption period of Line 10, and returned to their typical commuting patterns immediately after Line 10 resumed service. 1 Introduction The rapid increase of private cars in China has made urban traffic in many Chinese cities seriously congested. Thanks to its high capacity, speed and reliability, metro has been considered as a vital transportation alternative in large cities of China and undergone rapid development in cities such as Beijing and Shanghai. In 2014, the daily metro ridership in Shanghai reached 7.75 million, surpassing that of buses for the first time [1]. By the end of 2014, there were 22 cities in China that have built urban rail transit lines (metro and light rail), and the total operational mileage reached 3173 km [2]. However, because of ageing equipment and maintenance problems, a dozen of metro accidents have happened in Beijing, Shanghai and other Chinese cities in the recent 15 years [3, 4]. Owing to large passenger flow at stations and high passenger density inside metro cars, metro accidents may lead to serious casualties and disruptions to the urban transit system [5]. The disruption of metro caused by accidents also generates a large number of stranded passengers. Moreover, a major disruption of the metro network may affect individuals’ travel patterns significantly [6]. To reduce the disruptions caused by metro accidents and to better respond to such accidents when they do occur, we need to understand the evacuation process, the behaviour of passengers during and after evacuations, as well as the impact of metro accidents on both the metro and the road networks. This requires extensive travel, traffic and metro operations data, which is difficult to collect in traditional means. The video surveillance system installed in metro stations, for example, is helpful for monitoring conditions in metro stations, but cannot capture the movement of passengers outside the stations [7]. Mobile phone records provide a new method for collecting travel data. The records logged the location of base transceiver station (BTS), service time, service type etc. Therefore, the locations of individuals and their travels can be tracked from the mobile phone records [8, 9]. Mobile phone positioning technology, with its traceability and wide spatial/temporal coverage, can follow the passengers’ movements in the entire city and is useful to understand how they react immediately after an accident. In 2011, the mobile phone ownership in Shanghai was 122.9 per hundred people, and some people had more than 1 mobile phones belonging to different carriers [10]. The mobile phone data used in this paper includes 17.5 million active users, covering 76% of residents in Shanghai [11]. In this paper, we analyse the collision accident of Metro Line 10 in Shanghai on 27 September 2011 using mobile phone data. Travel patterns in the entire city were studied in the context of the Metro Line 10 accident. This paper is organised as follows. Section 2 gives a literature review on passenger evacuation in metro and application of mobile phone data to travel analysis. Section 3 briefly introduces the collision accident. Section 4 introduces the mobile phone data and methodology. Section 5 analyses the evacuation of passengers in accident stations. Section 6 studies the impact on commuters on the day of accident. Section 7 analyses the travel of commuters in the subsequent days. Finally, conclusions are drawn. 2 Related work 2.1 Emergency evacuation for metro accidents Understanding the evacuation process and traveller behaviour in emergency situations is important for both metro station design and emergency planning. Wang [12] analysed the speed of evacuation and passengers’ choice of escalators through experimental observations. He et al. [13], He et al. [14] and Ge et al. [15] studied the behaviour of passengers in metro evacuation drills by questionnaire surveys. Yoon et al. [16] analysed the evacuation time in metro stations by questionnaire surveys and video records. Most of these studies only focused on passenger evacuation inside metro stations. 2.2 Travel studies using mobile phone data As a new travel survey tool, mobile phone data have attracted the attention of more and more scholars, and they used such data to study human mobility under various situations, which include: (a) Origin–destination (OD) information : White and Wells [17] designed a method for extracting OD information from billing data of mobile phone and investigated its feasibility by comparing it with the real OD matrix of Kent in the UK. Duan et al. [18] developed an approach to estimate commuting OD matrix by signalling data of mobile phones. Iqbal et al. [19] proposed a method to get OD matrix using mobile phone data and limited traffic count data, and validated it by comparing with the results of simulation and observation. (b) Trip chain : Asakura and Hato [20] proposed a tracking survey method for individual travel behaviour using mobile phone data, and validated its feasibility by experiments in Osaka in Japan. Farrahi and Gatica-Perez [21] designed a method to get individuals’ daily routine patterns. Tian et al. [22] developed a statistical approach to estimate individual space–time path, which assumed most people move regularly in a long period. Schneider et al. [23] studied daily human mobility by mobile phone data of Chicago and Paris, and found 17 most frequent motifs (typical daily trip chains) which account for over 90% of the measured daily trips. (c) Large (special) events : Calabrese and Ratti [24] and Calabrese et al. [25] used mobile phone data and global positioning system data of buses and taxis to analyse the traffic condition and pedestrian movement in Rome during World Cup final match and other events. Jia et al. [26] studied the temporal and spatial distribution of non-resident tourists’ activities in Shanghai Expo by mobile phone data. (d) Metro passengers : Wang [27] applied mobile phone data to analyse the service radius of metro stations. Li et al. [8] proposed methods to estimate the ridership and exchange volume of metro by mobile phone data. Li et al. [9] studied the travel characteristics of residents along the metro using mobile phone data. (e) Natural disasters : Lu et al. [28] studied the mobility patterns of residents using mobile phone data, and found that population movements after Haiti earthquake may be more predictable than previously thought. 2.3 Summary of the literature Most of the literature related to metro emergencies focused on passenger evacuation in metro stations. They studied the evacuation based on data from survey questionnaires or experimental observations of drills, which may be different from the actual situation. For the video surveillance system, it is hard to track individuals and capture their movement outside the stations. To develop measures for better evacuation planning and response, and to understand the travel patterns after the metro disruption, the movement of affected passengers shall be analysed in the entire city, not just in or near metro stations. Mobile phone data can track individual travels, and has been proven to provide the temporal and spatial resolution to human mobility in cities. It will be the main data source for our study of the Metro Line 10 accident, so that to understand its impact on city traffic and influence on commuter behaviour. 3 Metro Line 10 and its collision accident on 27 September 2011 In 2011, the metro network of Shanghai composed of 12 metro lines (Fig. 1), and its total operational mileage reached 454 km [29]. Metro Line 10, with a length of 36 km, connects the Hongqiao transportation centre (including one railway station and one international airport) in the West with New Jiangwan City Station in the northeast. It also crosses the city centre of Shanghai. On 30 September 2011, the daily ridership of Line 10 was 0.48 million, accounting for 6.8% of total metro ridership of Shanghai [30]. Fig. 1Open in figure viewerPowerPoint Metro network of Shanghai in 2011 On 27 September 2011, a rear-end collision occurred in the section between Yuyuan Garden Station and Laoximen Station in Metro Line 10 (near the city centre of Shanghai). The events surrounding the accident are as follows (Fig. 2) [31, 32]: At 14:00, 27 September, Train 16 stopped and stayed in the section between Yuyuan Garden Station and Laoximen Station, after a signalling equipment failure occurred at Xintiandi Station at 13:58. At 14:08, 27 September, telephone blocking system was enabled between Jiaotong University Station and East Nanjing Road Station. (The blocking system is a metro operation control system, which avoids collisions between trains. In this system, metro line is divided into a series of sections or ‘blocks’. At any time, one block can only be occupied by one train. Shanghai Metro uses an automatic blocking system during normal time. If a failure occurred in the automatic blocking system, the telephone blocking system will be enabled. In the telephone blocking system, whether a train can enter a given section between two stations is confirmed by telephone by both staffs of the two stations.) At 14:37, 27 September, Train 16 was rear-ended by Train 5 in the section between Yuyuan Garden Station and Laoximen Station. At 14:51, 27 September, both the emergency preplan of metro and the emergency operation programme of buses were enabled. At 15:40, 27 September, evacuation of passengers in the accident stations was completed. At 16:00, 27 September, Train 16 left the spot of the accident. At 17:55, 27 September, Train 5 was repaired and left the spot of the accident. At 20:00, 28 September, Metro Line 10 resumed service. Fig. 2Open in figure viewerPowerPoint Rear-end collision of Metro Line 10 After the accident, emergency preplan of metro was immediately launched including the following four measures [31, 33]: Passengers were evacuated and the injured were treated immediately, and passengers on the accident trains were guided to the platforms of Yuyuan Garden Station or Laoximen Station. Engineers and technicians were dispatched to repair the damaged trains and equipment. About 12 stations were closed from Yili Road Station to North Sichuan Road Station, and the sections on both ends of Line 10 were operated with partial routes. Informations of the accident and rescue were released through official web site and television, and passengers were guided to choose alternative transportations. Furthermore, 60 shuttle buses were dispatched immediately to transport passengers between Yili Road Station and North Sichuan Road Station after the accident [31]. From 5:15, 28 September, an emergency bus line was launched between Yili Road Station and Hailun Road Station, with 100 buses. Its stops and schedules were the same as those of Metro Line 10 [34]. 4 Data and methodology 4.1 Mobile phone data Anonymous mobile phone data used in this paper were collected for billing and operational purposes from 20 to 30 September 2011 in Shanghai. It includes the information of encrypted mobile phone identifier, service time, service type, location of BTS, location area (LA) etc. [8]. The average number of records was 678.5 million per day, covering 17.5 million active users. The schema of mobile phone data is shown in Table 1. Table 1. Schema of mobile phone data Fields Description mobile subscriber identification (MSID) the unique identifier of mobile subscriber, which has been encrypted to avoid the invasion of privacy time stamp (TIMESTAMP) the timestamp of record location area code (LAC) LA code cell identification (CELLID) cell identifier. The combination of LAC and CELLID can uniquely identify a BTS event identification (EVENTID) identifier of events which are generated by mobile phone communications including LU, hand over, call, message etc. Using the method proposed by Li et al. [8], we identified 1.56 million metro passengers from mobile phone data on 27 September 2011. On 30 September 2011, the daily metro ridership of Shanghai was 7.11 million, and the exchange volume of metro passengers was 2.78 million [30]. The average daily number of individual trips in Shanghai was 2.23 in 2009 [35]. By (1), the total number of metro passengers was estimated to be 1.93 million on 27 September 2011. Therefore, the identified 1.56 million of metro passengers represented around 80% of total metro passengers (1) where is the daily number of metro passengers; is the daily metro ridership; is the daily exchange volume of metro passengers; and is the average daily number of individual trips. From 9:00 to 20:00 on 27 September 2011, the hourly number of mobile phone data records from metro passengers is about 8–10 million (Fig. 3). So, for each metro passenger, there are around 5.1–6.4 records per hour. That is to say, about every 10 min, one record of mobile phone data can be obtained from each metro passenger, which provides a good data source for this paper. Fig. 3Open in figure viewerPowerPoint Hourly number of records of mobile phone data from metro passengers on 27 September 4.2 Methodology In this paper, we adopt the approaches proposed by Li et al. [8] to identify metro passengers and extract their trajectories in the metro network. The methods to identify metro transfer, commuters of Metro Line 10 and their travel routes on the ground are as follows. 4.2.1 Identifying metro transfer To analyse the evacuation of stranded passengers, we need to identify whether they transferred to another metro line. As none of the interchange stations of Line 10 are designed for transferring on the same platform, if one wants to transfer, he has to move from the platform of Line 10 to that of another metro line. In the cellular network of Shanghai, the BTS of different metro line belongs to different LA. As a record of location update (LU) event will be produced if a user moves from one LA to another, we can identify metro transfer by the LU record. 4.2.2 Identifying commuters of Metro Line 10 As commuting is a stable and repeatable travel in general, the commuters of Metro Line 10 are identified based on the following rules: Passengers who ride Line 10 in both morning peak (AM peak hours (PEAK)) and evening peak (PM PEAK) for at least three working days in a week. In Shanghai, AM PEAK is from 7:00 to 10:00 and PM PEAK is from 16:00 to 19:00 [35]. Passengers who rode Line 10 in AM PEAK on 27 September (the day of accident). According to above rules, we have identified 27,708 commuters of Line 10 based on mobile phone data from 20 to 27 September. 4.2.3 Analysing travel route on the ground To analyse the passengers’ movement on the ground, we need to identify their travel routes, which can be obtained based on the location information of BTS from mobile phone data. However, this method has the following problems: Overlap exists in the coverage areas of two adjacent BTS. If one moves to this overlap area, frequent handovers may occur and bring data noise, making it difficult to identify the actual travel routes. There are a large number of BTS in the cellular network of Shanghai, and the coverage radius of BTS is only 250–500 m in the city centre. So, it is difficult to aggregate the trips between BTSs to represent the population movements. The records of mobile phone data are relatively sparse in the time dimension, and no records may be produced when one passes by some BTS. To solve the above problems, we identify the travel routes by the location information of LA. As a record of LU will be produced when a user moves to a new LA, we can track the user continuously on the LA level. LA is comprised of several BTS, and its coverage radius is over 1 km in the city centre, which will greatly eliminate the effect of frequent handovers. Moreover, due to the small number and wide coverage of LA, the trips between LAs are easy to aggregate. The procedures to identify travel routes on the ground are as follows: Merge the service areas of BTS in the same LA, and the location of users is represented by the centroid of the LA. Calculate the users’ trajectories , , where is the trajectory of user i, is the k th LA which user i passed by, and is the coordinate of the centroid of LA k. Calculate the number of trips between LAs to obtain the travel routes of the population. 5 Evacuation of stranded passengers 5.1 Extracting stranded passengers in the accident stations Using the method proposed by Li et al. [8], the time when a passenger enters or exits a metro station can be obtained. Then, the stranded passengers caused by the accident in Yuyuan Garden Station and Laoximen Station can be identified by the following rules: Passengers got on Metro Line 10 or entered the stations of Line 10 during the period of , 27 September. Passengers appeared in Yuyuan Garden Station and Laoximen Station after the accident occurred (14:37, 27 September.). How to determine the study period is discussed below. The start time According to the accident report [32], Train 16 had already arrived in the section between Yuyuan Garden Station and Laoximen Station at 14:00 and stayed there till the accident occurred. Moreover, Yuyuan Garden Station and Laoximen Station are located around the midpoint of Line 10, and about 30 min away from the terminal stations at both ends. Considering the passengers who transferred from other lines to Line 10, the maximum travel time of stranded passengers may be more than 40 min. Therefore, should be chosen from a period more than 40 min before 14:00. A sensitivity analysis is performed to determine the start time . At equal intervals of 15 min, 9 different times are selected as , from 12:30 to 14:30. Fig. 4 shows the number of stranded passengers obtained by different start time (). Fig. 4Open in figure viewerPowerPoint Number of stranded passengers obtained by different start time ( ) As can be seen from Fig. 4, if is selected between 12: 30 and 13:30, the analysis results are very close. On the other side, the number of stranded passengers will be significantly underestimated if is selected after 13:30. Therefore, 13:00 is a reasonable time to choose as the start time . The end time At 14:51, 27 September, the emergency preplan of metro was started and a major portion of Metro Line 10 was closed including the two accident stations [32]. Therefore, 15:00 is chosen as the end time . Appling the method above, 3422 stranded passengers in the two accident stations are identified. 5.2 Evacuation process of stranded passengers in the accident stations The procedures of analysing the evacuation process of stranded passengers in the accident stations are as follows: Calculate the number of the identified stranded passengers who stayed in Yuyuan Garden Station and Laoximen Station during the period of 13:00–17:00, and track how they left the stations (Figs. 5a and b). For the purpose of comparison, passengers who got on Line 10 (or entered the stations of Line 10) and appeared in the two stations during the period of 13:00–15:00, 20 September are also extracted (both 27 and 20 September are Tuesdays), and their processes of leaving the stations are shown in Figs. 5c and d. Fig. 5Open in figure viewerPowerPoint Passenger evacuation patterns during 13:00–15:00 on 20 and 27 September (figures show the number of passengers in the stations over time) (a) Laoximen, 27 September, (b) Yuyuan Garden, 27 September, (c) Laoximen, 20 September, (d) Yuyuan Garden, 20 September As shown in Fig. 5, passengers were stranded for a substantial period of time in Yuyuan Garden Station and Laoximen Station on 27 September, due to the collision accident. Moreover, the evacuation in the accident stations can be divided into two stages. Stage 1 : 14:37–15:00. The stranded passengers began to accumulate after the telephone blocking system was enabled. These passengers include passengers who entered the stations to ride metro, and passengers who got off the previous train. After the accident occurred, the number of stranded passengers decreased sharply, 50% for Laoximen Station, and 25% for Yuyuan Garden Station. Stage 2 : 15:00–15:40/16:10. The collision occurred in the section between Yuyuan Garden Station and Laoximen Station. Passengers on the accident trains had to walk to the metro platform and were evacuated in Stage 2. The speed of the evacuation was lower than that of Stage 1. For Laoximen Station, the evacuation was completed by 15:40, which is consistent with the report of the accident [31]. For Yuyuan Garden Station, the evacuation was completed at 16:10, slower than that of Laoximen Station. This is probably because Laoximen Station is an interchange station and many stranded passengers moved to Metro Line 8. 5.3 Stranded passengers evacuated by metro The following are the procedures of analysing stranded passengers’ movement from accident stations to other metro lines: Extract the identified stranded passengers who moved to stations of other metro lines. The time of the first mobile phone record in the other station is treated as the arrival time. Calculate the proportion of stranded passengers evacuated by metro. In Fig. 6, for Laoximen Station, the proportion is 0.8 at 14:30. It means that 80% of passengers, who stranded at 14:30 in Laoxinmen Station, were evacuated by metro subsequently. For these stranded passengers evacuated by metro, calculate the proportion of them moving to different stations in Stage 1 and Stage 2 separately (Table 2). Fig. 6Open in figure viewerPowerPoint Proportion of stranded passengers evacuated by metro Table 2. Proportions of stranded passengers moved to different stations Station (moved to) Laoximen (Line 10) Yuyuan Garden (Line 10) Stage 1, % Stage 2, % Stage 1, % Stage 2, % Laoximen (Line 8) 99 96 1 50 East Nanjing Road (Line 2) 0 0 70 11 other stations 1 4 29 39 total 100 100 100 100 As shown in Fig. 6, the majority stranded passengers were evacuated by metro in Stage 1. For Laoximen Station, before 15:00, the proportion of passengers moving to other metro lines is 80%, but it dropped quickly to 40% after 15:00. For Yuyuan Garden Station, the proportion decreased gradually over time, but, in general, lower than that for Laoximen Station, probably because Yuyuan Garden Station is not a transfer station. To sum up, for the two stations, over 70% of the stranded passengers were evacuated and transferred to other metro lines in two stages. For stranded passengers evacuated by metro, Table 2 shows the proportion of them moving to different stations. For Laoximen Station, most of them moved to Metro Line 8 in both stages. For Yuyuan Garden Station, 70% of them moved to the East Nanjing Road Station of Line 2 in Stage 1, and 50% of them chose Metro Line 8 in Stage 2. On the basis of the foregoing analysis, it can be found that the difference of transport mode for evacuation and the passengers stuck on the accident trains led to the two-stage evacuation pattern. In Stage 1, most stranded passengers were evacuated by other metro lines. Therefore, the evacuation speed of this stage was higher than that of Stage 2. As passengers stuck on the accident trains had to walk through the metro tunnel to the platform, the number of stranded passengers in the accident stations kept almost unchanged at the beginning of Stage 2. Coupled with passengers who could not leave by metro in Stage 1, most of the stranded passengers in Stage 2 were evacuated by bus or other transport modes on the ground. Therefore, the evacuation speed of Stage 2 was lower than that of Stage 1. 5.4 Travel patterns of stranded passengers in the entire city In this section, we study the movement and travel routes of stranded passengers evacuated in both Stage 1 and Stage 2, from the perspective of entire city. 5.4.1 Movement of stranded passengers evacuated in Stage 1 in the entire city The period of Stage 1 is 14:37–15:00. The movement of stranded passengers evacuated in Stage 1 is shown in Fig. 7. Fig. 7Open in figure viewerPowerPoint Movement of stranded passengers evacuated in Stage 1 in the entire city (a) 15:00, (b) 15:15, (c) 15:30, (d) 15:45 At 15:00, the density of stranded passengers near the accident stations was still high. However, a part of passengers has been evacuated to Siping Road Station in the northeast, which is an interchange station for Line 8 and Line 10 (the section of Line 10 from North Sichuan Road Station to New Jiangwan City Station was still in operation at that time). At 15:15, these passengers continued to move along the North part of Line 10. At 15:30, some passengers have transferred to Lines 1, 2, 3, 4 or 8. At 15:45, some passengers have arrived Yishan Station in the southwest, which is an interchange station for Line 4 and Line 9. Moreover, the number of stranded passengers near the accident location reduced significantly. As shown in Fig. 7, Metro Lines 1, 2, 3, 4 and 8 played important roles in evacuation. Among them, Lines 1, 2, 3 and 8 are parallel to parts of Line 10, and Line 4 is a loop line, which connects wit