Revealing the impact of storm surge on taxi operations: Evidence from taxi and typhoon trajectory data

Abstract

Storm surges are one of the most destructive natural ocean disasters in the world, which pose a great threat to economic development, public safety, and transport in coastal cities. Investigating the response of urban traffic conditions during storm surges and discovering sensitive urban areas are crucial to urban planning, disaster prevention, and emergency rescue. Therefore, this study investigates the impact of typhoons on taxi operations to target sensitive areas of the city. In this study, the change in taxi service operation in Shenzhen during typhoon storm surges from 2012 to 2014 was explored at the city district scale using a diverse set of data including taxi trajectory data, typhoon track data, and tide level data. Several aspects were included to access this impact, for example the number of daily orders per taxi, the trip length, the degree of closeness to the typhoon-affected area, the coverage area of the taxi service, and the spatial dynamics of the origin and destination matrix. The regular trend and correlation of storm surge elements and taxi parameters were first analyzed. The distribution of the taxi trajectory data was then used to detect variations in the service area, and the reasons for these variations were discussed. The result shows that storm surges with diverse characteristics do affect taxi operations to differing degrees, and that there was a potential correlation between tide level and taxi service area. The sensitive areas for taxis during typhoon storm surges were mainly located at the boundary between the central urban area and other districts of Shenzhen, they were constrained by the dynamic changes in the origins and destinations, and they were likely related to important transportation hubs in the city. This finding provides a reference to support disaster prevention preparedness and decision-making by local authorities.

Keywords

Storm surge typhoon track data GIS taxi trajectory data sensitive areas,Geographic Information System

Introduction

In recent years, losses due to storm surges have increased. Storm surges have become one of the most serious natural disasters threatening the safety of coastal cities and economic development (Gu et al., 2017; Hatzikyriakou and Lin, 2017; Shi et al., 2013). According to the different atmospheric characteristics of storm surges, they can be divided into three categories: tropical storm surges, temperate storm surges, and the unique cold wave storm surges in the Yellow Sea and the Bohai Sea, among which tropical storm surges caused by tropical cyclones (typhoons) are the most threatening disasters. These latter storm surges have characteristics of a fast speed, a strong strength, and significant destructive power. During 1989–2016, there were 246 typhoon storm surge disasters in China’s coastal areas, with an annual average grade of 8.78 (Guo and Li, 2018), resulting in a series of urban security events, which included coastal facilities damage, house collapse, casualties, water and power outage, urban flooding, and traffic interruption. They also brought huge hidden dangers to the safety of residents and the urban development of coastal cities. Therefore, from a city’s perspective, it is of great significance to optimize the protection of urban structures and emergency rescue operations via assessing weakness areas of the city during the urban response process of a storm surge.

In the case of natural disasters like storm surges, smooth transportation is an important embodiment of the normal operation of a city, which can provide a good foundation for disaster prevention and preparation, rescue operations, and emergency evacuation. In recent years, with the continuous enrichment and improvement of traffic big data systems, trajectory data have gained interest among researchers, and the method of track mining has been widely used. As a major component of public transportation, taxis are one of the most important travel modes in daily life, which could account for as much as 20% of the total traffic flow and even 50% in some key areas (Yang et al., 2005). The tracking trajectory of taxis could be used to infer information such as traffic state, travel demand of urban residents, urban structure, etc. (Wu et al., 2019). To ensure a city’s stable travel environment under storm surges, and to reduce the hidden dangers to public safety while avoiding traffic hazards, it is essential to understand the dynamics of taxi service capabilities, their distribution and their demand across the whole city. In short, it is necessary to investigate the effect of typhoons on taxi operations.

Therefore, this paper took taxis as research objects, which were the medium between storm surge events and the urban response. The objective of this paper is to investigate the impact of typhoons on taxi operations and to identify the sensitive areas of taxis in the city. Based on the historical impact of typhoon storm surges on coastal cities, we made three assumptions: (1) different types of typhoons will affect the operation of taxis to varying degrees, (2) the typhoon storm surge may affect the spatial distribution of taxi service areas, and (3) changes in the spatial distribution of taxi service areas may be constrained by the travel needs of residents and the location of traffic hubs. First, we calculated four taxi parameters (i.e. the numbers of daily orders per taxi, the trip length, the coverage area of the taxi service, the dynamics of the origin and destination matrix) and three storm surge parameters (i.e. the intensity, the degree of closeness to typhoon-affected areas, and the tide level). The changing characteristics of taxi operation states during storm surge periods were mined and analyzed via correlations. We then analyzed the changes in the dynamic spatial distribution of taxi service areas to identify sensitive areas of the city for taxis. Finally, from the perspective of public travel demand and the location of important transportation hubs, we explored the cause of these sensitive regions of taxis. Generally, the impact of storm surge disaster on taxi operation mirrors the traffic situation and the response of the city hit by a natural adversity. The mining of dynamic changes of taxi trajectory indexes is conducive to identifying sensitive urban areas, and it plays an auxiliary role in urban planning, disaster prevention, emergency rescue, and decision-making by local authorities.

Related work

The impact of storm surges

Storm surges are one of the most destructive natural marine disasters in the world, which has a tremendous impact on coastal cities. Research into typhoon storm surge events can be divided into three categories: mining the features and patterns of the disaster itself, assessing the vulnerability of the disaster carrier, and determining the disaster loss. In the first category, historical data have been widely used when studying the laws of storm surges. Vickery et al.(2000) proposed a method for simulating the whole path of a tropical cyclone. Based on the frequency, season, path, intensity, and range of influence, the entire process was simulated, and the water increase in different return periods was calculated to carry out a quantitative analysis of the storm surge (Lin et al., 2010; Nong et al., 2010; Yasuda et al., 2010). Dong et al. (2014) selected typical tide stations in Guangdong Province, analyzed the historical records of nearly 500 stations since 1949, and discussed the spatio-temporal distribution patterns of storm surges. In the second category, Gornitz (1991) put forward the concept of a coastal vulnerability index and risk level for the first time. Vulnerability refers to the possibility of loss of a disaster carrier under the pressure of disaster intensity (Dilley, 2005). Similarly, it can also be described as the degree of vulnerability of exposed objects due to disturbance and their ability to adapt, cope with, or recover (Kasperson and Kasperson, 2013). Under a storm surge disaster, the human security, economic property, ecological environment, and urban infrastructure of coastal areas are sensitive to potential damage. Around these carriers, the vulnerability of coastal areas was assessed (Helderop and Grubesic, 2019; Hossain, 2015; Khouakhi et al., 2013; Yang et al., 2007; Yuan et al., 2016). In the third category, the relationship between disaster and loss was established by constructing empirical statistical models in early research (Guo, 1991; Shi et al., 2000; Xu and Tan, 1998). Then, research on quantitative loss estimates gradually emerged. In 2010, the HAZUS-MH (multi-hazard) model, developed by the Federal Emergency Management Agency (2010) of the United States, evaluated the direct damage and consequential loss, and undertook a secondary disaster effect appraisal. The analytic hierarchy process and entropy (Yin et al., 2012), and neural networks and the vector space model (Lin and Yang, 2019), were also used to evaluate economic losses.

In short, historical typhoon paths and tide station data are the major data sources used to explore the spatial–temporal distribution patterns of storm surges. The aforementioned research was carried out at the provincial and municipal scales. It was difficult to find examples of disaster response and vulnerable areas of the interior urban structure under a storm surge disaster. Moreover, the disaster impact was mainly reflected by economic loss and casualties while the impact on other important aspects of the city, such as transportation services (i.e. taxis), has rarely been studied.

Taxi trajectory mining

Taxis are not constrained by operating times and driving routes, which have the features of a wide service coverage and high flexibility. Taxi trajectory data mining has been applied widely (Wu et al., 2019) in intelligent transportation, resource and environmental protection, urban planning, and social perception. In the field of intelligent transportation, taxi trajectory data play an important role in congestion detection (Kan et al., 2019; Kong et al., 2015; Wang et al., 2013; Wu et al., 2014), anomalous traffic event identification (Kuang et al., 2015), and traffic flow prediction (Duan et al., 2018; Zhao et al., 2017). Moreover, taxi trajectory data are the most intuitive embodiment of taxi operation status, where taxi behavior and mode can be collected, such as pick-up and drop-off locations, enabling the development of taxi operations (Su et al., 2018; Zhang et al., 2017), route planning (Zhang et al., 2015), and modeling human mobility (Tang et al., 2015). In addition, many scholars have associated taxi trajectory data with environmental parameters to understand the distribution of urban traffic pollution (Luo et al., 2017; Zhao, 2013). With the considerable growth and diversification in social activities, the application of taxi trajectory data in urban planning and social perception has expanded. A large group of scholars have used clustering results of taxi pick-up and drop-off points to investigate urban functional areas (Yuan et al., 2012; Zhao et al., 2015) and land use layout (Pan et al., 2013) to assist urban planning (Liu et al., 2012). Meanwhile, research into urban multitude dynamics and social perception based on taxi trajectory data has gradually emerged. For example, Liang et al. (2012) and Wang et al. (2015) investigated human mobility by exploring large amounts of taxi traces.

In short, the taxi trajectory-based description of urban public travel patterns and taxi service operation can strongly support urban planning optimization, traffic network construction, and improvements to the urban environment. However, at present, few studies have been conducted to investigate dynamic changes in urban taxi services facing natural disasters.

Materials and methods

Study area

Shenzhen (Supplemental Figure 1) is the study area, which is a prefecture-level coastal city in the south of China, adjacent to Hong Kong. It is the national economic, technological innovation, regional financial, and commercial logistics center in China. By the end of 2018, the permanent population of Shenzhen had reached more than 13 million while the actual management population had exceeded 20 million (Shenzhen National Economic and Social Development Statistics Bulletin, 2018). However, it is also an area prone to weather disasters, especially in the summer when rainstorms, thunderstorms, and typhoons are frequent. From July to September, there were 3–4 tropical cyclones affecting Shenzhen on average (Meteorological Bureau of Shenzhen Municipality, 2019).

Experimental data

Typhoon trajectory data

The typhoon trajectory data were downloaded from the tropical cyclone data information center of the China Meteorological Administration (2019). The trajectory data set provides key information including typhoon number, date, time, intensity mark, longitude and latitude, maximum wind speed, etc. (Ying et al., 2014). The typhoon size analysis data retrieved from satellites (Lu et al., 2017) cover the tropical cyclone scale. According to the national standard of the Tropical Cyclone Classification (GB/T 19201-2006), the typhoon intensity can be separated into six groups (Supplemental Table 1). The typhoon trajectory data during the period of May–October in 2012–2014 were screened with a distance condition of being within 500 kilometers from the center of Shenzhen (Supplemental Table 2).

Taxi trajectory data

The taxi trajectory data contain information on the taxi license plate number, the longitude and latitude, and the service status. The first is a unique identification code. The second reflects the position of the taxis, which is used to generate the complete tracks. The service status is displayed as empty or occupied, which is a key indicator to judge the pick-up and drop-off behavior. This study used the taxi trajectory data from two days before the typhoon transit to two days after departure.

Data from the Shenzhen tide level station

During the period of typhoon prevalence, typhoon storm surges are often generated and the water level rises sharply. Data from the tide level stations are crucial for storm surge monitoring. The data from the tide level stations in this paper are provided by the China National Oceanic Information Center, from 2012 to 2014. The tide level stations include Shenzhen and Shekou stations (Supplemental Figure 1).

Methodology

Methodology overview

To explore the changes in taxi operation in Shenzhen during periods of storm surge, a numerical and spatial correlation method (as outlined in Figure 1) was proposed. In general, in order to clarify the impact of typhoon storm surge on taxi operations and target the location of sensitive urban areas, this study focused on three conjectures to conduct experiments and an analysis from two dimensions of value and space. The three conjectures are as follows: (i) different types of typhoons will affect the operation of taxis to varying degrees, (ii) the typhoon storm surge may affect the spatial distribution of taxi service areas, and (iii) changes in the spatial distribution of taxi service areas may be 3constrained by residents’ travel needs and the location of traffic hubs. First, considering the possible reasons for changes in taxi behavior due to typhoons, four taxi parameters were calculated (i.e. the number of daily orders per taxi, the trip length on service, the coverage area of the taxi service, the dynamics of the origin and destination matrix) and three storm surge parameters (i.e. the intensity, the closeness degree to the typhoon-affected area, and the tide level). Among taxi indexes, the number of daily orders per taxi is the most direct parameter for residents’ travel demand. The trip length on service and the coverage area of the taxi service are related to traffic conditions, and they may be constrained by the origins and destinations of the taxi tracks. The spatial distribution of the origin and destination matrix reflects the travel needs of residents in different areas from the spatial dimension to some extent. The relationship between storm surge and taxi operation was quantified through correlation analysis, which is used to investigate the impact of the typhoon on the taxi operation. Then, we analyzed the dynamic distribution of changes in the taxi service area. The regions with a significant fluctuation were considered to be sensitive areas of storm surge in the study area. Finally, we conducted a spatial correlation analysis of sensitive areas from the perspective of the distribution of residents’ travel needs in different urban areas and the spatial distribution of important transportation hubs. By using these approaches, this research targeted sensitive areas in the study area.

Figure 1.

The overview of methods used in this study.

The numbers of daily orders per taxi

Pick-ups and drop-offs are the main actions while taxis are in motion. The service status index shows a change sequence from 0 to 1, with a duration of 1 and from 1 to 0, corresponding to three stages: pick-up, service, and drop-off. In taxi trajectory data, the behavior with such regularity is called a one-time order conduct. The statistical indicator used here is the average number of taxi service orders, which can be calculated as follows

D = \frac{N_{d}}{n}

(1)

where D represents the numbers of daily orders per taxi, N_d indicates the total amount of taxi orders in a day, and n is the number of taxis.

The degree of closeness to the typhoon-affected area

Typhoons are a relatively complex natural disaster. During the life cycle of typhoon generation, development, maturity, and extinction, a great distance will be covered as they migrate. Moreover, during the migration process, the radius of influence of the typhoon changes dynamically. The radius of influence will vary significantly during different life stages of the typhoon. In the stage of formation, the influence radius is small, i.e. only tens of kilometers or even less. However, in the mature stage, the radius can reach 200–500 kilometers, or even greater. The length and width of the Shenzhen urban area are 100.4 and 68.4 kilometers, respectively. This is about 1/2–1/5 of the influence radius of a typhoon. Thus, there are still some possibilities of a spatial relationship between Shenzhen and the typhoon-affected area, e.g. (i) Shenzhen is not in the typhoon-affected area, (ii) a part of the area of Shenzhen is in the typhoon-affected area, or (iii) all areas in Shenzhen are in the typhoon-affected area. The taxis and typhoons are both dynamic and complicated targets. The degree of closeness to the typhoon-affected area is used to describe quantitatively the spatial relationship between the taxi track and influence areas of the typhoon. This study views the typhoon-affected area as a circular area with the typhoon center as the center and the typhoon scale as the radius. The distance is determined by the positions of all taxis and the position and size of the typhoon at all times of the day. Assuming that the number of taxis is n, the average distance in moment m is calculated as equation (2), and the average distance on that day is calculated as equation (3)

V S_{m} = (S_{m l} + S_{m 2}^{} + \dots + S_{m n}) / n

(2)

V S = (V S_{1} + V S_{2}^{} + \dots + S_{m}) / n

(3)

where the array [S_m₁, S_m₂… S_mn] represents the distance of the different taxis to the typhoon-affected area at m moment, while the array [VS₁, VS₂… VS_m] indicates the mean distance at different times. The specific calculation process is shown in Supp Fig 2.

The average trip length on service

The service track of taxis can be effectively restored based on track points and service status. The service distance describes the distance that a taxi drives in the process of a complete service operation, which is also called the trip length on service. The statistics in this paper are the average service distance per taxi (L_c) and the average service distance per order (L_o). The calculation is as follows

L_{c} = \frac{L_{1} + L_{2} + \dots + L_{n}}{n}

(4)

l_{n} = \frac{L_{n}}{n_{d}}

(5)

L_{o} = \frac{l_{1} + l_{2} + \dots + l_{n}}{n}

(6)

where L_c is the average trip length on service per taxi, n refers to the number of taxis, the array [L₁, L₂… L_n] represents the total length of a single taxi in the day, l_n and n_d represent the average service distance per order, and the number of orders of a specific taxi, respectively, while L_o is the average trip length on service per order of all taxis in a day.

The coverage area of the taxi service

The coverage area of the taxi service, also called the taxi service area, describes the coverage area of all trajectories of a taxi service, which represents the coverage capability of a taxi service in the study area. A value of 1500 meters is a fast-accessible distance, which is regarded as the taxi service distance in this paper. Thus, a line buffer with a radius of 1500 meters is made for the restored tracks. The final results include the average service area statistics and the distribution map of the service area. The distribution map is the result of the superposition of different taxi trajectories.

The spatial dynamics of the origin and destination matrix

This index reflects the spatial distribution characteristics of the taxi origin and destination matrix under different spatial units. The starting point of the taxi service trip is the pick-up position, and the terminal point is the drop-off position. Both are viewed as an origin and destination pair. All pairs were constructed as a taxi origin and destination matrix. In this paper, the number of origins and destinations within the taxi tracks in a day is counted in an urban grid of 500 by 500 meters.

The sensitive area

The sensitive area is defined as the area that is likely to change its features under the impact of an event such as a typhoon. The greater the degree of change, the greater the sensitivity of the area to the event. In this paper, we studied the sensitive areas of taxi operations in Shenzhen under the influence of storm surge disasters. During the event process, the operating indexes of taxis in these sensitive areas fluctuated to a large extent.

Results

Relevance between taxi activity and typhoon characteristics

To explore the impact of different types of typhoons on taxi operations, we separately calculated two series of parameters representing typhoon characteristics and taxi behavior. During the influence period of the typhoon and the two days before and after, the variation in their trends was calculated in Supplemental Figures 3 and 4. This study found that typhoons do have a certain influence on the operation capacity of taxis. Moreover, diverse typhoons affected taxis in different ways and to different degrees. Pearson correlation coefficients were calculated between typhoon storm surge-related indexes and taxi service indicators to explore the main factors of typhoon storm surge impacts on taxi operations.

The results in Table 1 marked with ** were those with a significant correlation at the 99% level, while those marked with * had a significant correlation at the 95% level. According to the correlation calculation of the indicators, the following three results were found in this study. First, there was a strong positive correlation between taxi indicators (orders, SDCar, and service area). It showed that during the storm surge, the changing law of taxi behavior parameters had a strong consistency of increasing or decreasing. Second, there was a negative correlation between typhoon storm surge indexes DTyphoon and tide level. This meant that the closer the typhoon was, the greater the increase in the amount of water along the coast. It also indicated that tide level data play an important part in storm surge events. Finally, taxi parameters, which included SDcar and service area, had a notable relation with tide level, which indicated that the rising tide level could have an indirect relationship with a decrease in SDcar and service area in a spatial–temporal association. In conclusion, the typhoon did have a certain impact on the operating capacity of the taxis. Significant correlations were shown between parameters in the same groups of taxi and typhoon storm surge indexes, respectively, and there was a potential relationship between the tide level and taxi service indexes.

Table 1.

Results from the Pearson correlation analysis.

Indexes	Orders	SDCar	Service area	Typhoon strength	TLShenzhen	TLShekou	DTyphoon
Orders	1	0.627**	0.544**	0.034	−0.185	−0.081	−0.070
SDCar		1	0.951**	0.130	−0.253*	0.197	−0.014
Service area			1	0.122	−0.255*	−0.237*	0.044
Typhoon strength				1	0.133	−0.065	−0.062
TLShenzhen					1	0.726**	−0.487**
TLShekou						1	−0.259**
DTyphoon							1

DTyphoon: the distance between taxi and typhoon-affected area; SDcar: service distance per taxi; TLShenzhen, TLShekou: tide level of Shenzhen and Shekou stations.

Changes in taxi service area

Table 1 shows a latent strong correlation between taxi service area (referred to as SA) and typhoon storm surge indicators. But how did the typhoon storm surge affect the spatial distribution of SA? To answer this question, we investigated the change patterns of SA during typhoon storm surges. First, the distribution of SA over the whole period was summarized, representing the generality of this index (Figure 2(a)). An obvious phenomenon of center clustering was apparent, which also showed a divergence pattern from the center to the outside. From the perspective of districts, areas where the SA was most concentrated were Futian and the western part of Luohu, which are in the central urban areas of Shenzhen. As to other surrounding areas, including Nanshan, Baoan, Longhua, and Longgang, the degree of SA concentration showed weakening. Then to clarify the influence of the storm surge on the SA distribution, the dynamics of the SA in each city grid during the 14 typhoon periods was counted. If SA increased or decreased in a certain area, this was a sign of being affected. In addition, the larger the amplitude of the fluctuation, the more significant the impact of the storm surge on the area, i.e. the more sensitive the area was to storm surge events. Therefore, the result of the degree of variation was the sum of the absolute value of the SA difference between two adjacent days. The visual distribution of the variation of SA (referred to as V-SA) is shown in Supplemental Figure 5.

Figure 2.

Taxi service areas and sensitive regions. (a) Overall distribution of taxi service area, (b) total change in service area, (c) sensitive regions, and (d) important transportation hubs.

Generally, in the 14 typhoon cases shown in Supplemental Figure 5, the distribution range of V-SA was similar to the overall SA (Figure 2(a)). However, the locations of the cluster regions changed. A large percentage (71.4%) of typhoons (Supp Fig 5) show that the geographic location of the peaks had shifted considerably. The new convergence regions of V-SA were the areas where the SA changed the most, and they were also the areas most affected by the storm surges; these are referred to here as sensitive urban areas of taxis. To further explore the new gathering areas shared among the 14 events, we summed V-SA during all typhoon periods, which are displayed in Figure 2(b). Obviously, the maximum value of the V-SA did not appear in the central regions of Shenzhen. It is worth noting that the trend in decreasing central diffusion also appeared as an opposite pattern when compared to the overall SA, increasing outward from the city center. Finally, the sensitive urban areas where the SA changed the most during the 14 typhoons are circled and displayed in Figure 2(c). The number ranged from 1 to 9, and the size of the red circles indicates the measurement of the areas. These sensitive areas can be classified into four classes. The first is areas 8 and 9, located in the center regions of the city, but they occupied only a small area. The second category was newly emerging clusters with a relatively larger size, ranging from 2 to 7, accounting for 66.7% of the total. The fourth sensitive area covered the largest area. This type was located near the intersections of various districts. For example, areas 2 and 6 were close to the border of Futian and Longhua, while areas 3 and 7 were located at the junction of Futian and Longgang. The third sensitive area only contained area 1, which was independent of the others. It was neither located in the central city nor at intersections of the districts. In short, the influence of the typhoon storm surge on the SA distribution is consistent in the 14 typhoon cases. According to the dynamics of SA, this study identified nine sensitive urban areas and found that they were mainly distributed at the junction of different districts.

Exploration of the causes of sensitive areas

What caused these areas to become sensitive areas during typhoon storm surges? And what factors were related to their position? In this paper, we made a preliminary exploration based on two aspects: the main traffic hubs and public travel demand.

The distribution of the main transportation hubs

Transportation hubs, as an important part of urban travel, included airports, railway stations, and coach stations. In Figure 2(d), these three kinds of main transportation hubs in Shenzhen were marked on the map. As can be clearly seen from the figure, the locations of sensitive areas and traffic hubs had a high degree of agreement. For example, the location of Baoan Airport was near the sensitive zone 1; the largest sensitive area 4 contained Shenzhen West Station and two coach stations; the sensitive areas of 2 and 3 included Shenzhen North Station and Shenzhen East Station, respectively. Eighty percent of the transportation hubs were located inside sensitive areas, while 20% of them were located near the edge of the sensitive areas. It is noteworthy that the sensitive area with the largest area contained the largest number of transportation hubs. Hence, a potential spatial correlation could be found between the location of urban transportation hubs and the distribution of sensitive areas.

The spatial dynamics of the origin and destination matrix

Pick-ups and drop-offs are the main operation behavior of taxis. The distribution of the origin and destination matrix corresponded to residents’ travel demand in various urban areas to a certain degree. Therefore, it was necessary to study the relationship between the transformation of starting and ending positions of taxi tracks and sensitive areas. Above all, it was important to understand the overall distribution of the origins and destinations during the whole period, which is shown in Supplemental Figures 6(a) and (b), respectively. The overall situation of the spatial distribution of the origins and destinations was basically the same. The highest value was concentrated in the central region of Shenzhen. The remaining starting and ending positions were mostly located in Nanshan, Baoan, Longhua, and Longgang. Since both indicators were similar in terms of spatial distribution, they were added together here, collectively referred to as the origin–destination (O–D) index, which was used to calculate the variation during the typhoon processes. Aimed at highlighting the large changing areas of this indicator, the calculation method was the same with SA, and the result of 14 typhoon cases was shown in Supplemental Figure 7. Moreover, all O–D changes were also summed to emphasize clustering areas as shown in Figure 3(a). The spatial distribution and aggregation characteristics of O–D variation (referred to as VO–D) represented a high degree of consistency with the overall distribution of O–D. The central urban area was also the area with the most significant changes. Moreover, the other gathering regions were located in Nanshan, Baoan, Longhua, and Longgang.

Figure 3.

Total change and assembling regions of O–D. (a) Total change in origins and destinations and (b) sensitive region comparison.

To explore the relationship between VO–D and sensitive areas, the VO–D cluster areas were circled and displayed in Figure 3(b). The positional relationship between these two regions can be summarized as follows. First, there was an overlap in some clusters. For example, zone 1 and a; 4 and c; 3 and e. Second, more than half of the sensitive areas surrounded the VO–D clustering area f. The third, and also the most important point, was that almost all the sensitive areas are situated between the VO–D aggregation area f and other VO–D aggregation areas. To confirm this, we abstracted the vehicle tracks between different VO–D clusters into simple directional wires, from the central city area f to others. We found the wires passed through almost all sensitive areas. To sum up, most of the sensitive urban areas were located between two different VO–D aggregation areas. It could be inferred that the formation of the sensitive urban areas was indeed affected and constrained by the O–D parameter to some extent.

Conclusions

This paper explored the change in taxi service operation in Shenzhen during storm surges from 2012 to 2014, using a diverse set of data, including taxi trajectory data, typhoon track data, and tide level data. Using seven characteristic indexes, the change in the trends of the indexes under two scales of numerical value and spatial distribution was analyzed. In addition, summarizing the correlation between the parameters of taxis and storm surges, this paper investigated the spatial distribution of the taxi service area to explore sensitive urban areas. Finally, a preliminary attempt was made to find possible reasons for their formation. The major conclusions were as follows:

The typhoon did have a certain impact on the operating capacity of the taxis. Significant correlations were shown between parameters in the same groups of taxi and the typhoon storm surge indexes, respectively, and there was a potential relationship between the tide level and the taxi service area. The findings are beneficial for understanding the main response behaviors and how taxi services change during disasters. The potential correlation mining provided a good reference and support for disaster prevention preparedness.

The sensitive areas of Shenzhen were not located in the central urban area, but near the intersection of various districts, especially at the boundary between the central urban area and other districts. Moreover, their formation was affected by the dynamics of the origins and destinations, and it was likely to be related to the distribution of important transportation hubs. The discovery of sensitive areas is helpful for city governance and disaster prevention under typhoon storm surge events.

This paper investigated traffic responses under typhoon storm surges at a city district scale. Based on the spatial distribution characteristics in the service area of the taxis, we extracted the typhoon storm surge sensitive areas in Shenzhen, and we made a simple test of the causes of their formation. Future work will be based on the conjecture in this paper to further expand and uncover the factors of sensitive areas in depth, from the perspective of the complexity of the urban road network structure, the distribution of urban functional areas, etc. Moreover, at present, only limited amounts of taxi track data for a few cities can be accessed freely and openly (Wu et al., 2019). Thus, the data set of taxi trajectory has some limitations in the coverage of geographical location and the time span. In our project, Shenzhen is used as a pilot city. With the continuous improvement to the data system, we will work on applying the method to other coastal cities that are heavily affected by storm surges and/or other disasters and compare the distribution of sensitive areas in cities with different characteristics.

Supplemental Material

sj-pdf-1-epb-10.1177_2399808320954206 - Supplemental material for Revealing the impact of storm surge on taxi operations: Evidence from taxi and typhoon trajectory data

Supplemental material, sj-pdf-1-epb-10.1177_2399808320954206 for Revealing the impact of storm surge on taxi operations: Evidence from taxi and typhoon trajectory data by Zhixiang Fang, Yichen Wu, Haoyu Zhong, Jianfeng Liang and Xiao Song in Environment and Planning B: Urban Analytics and City Science

Footnotes

Author's Note

Zhixiang Fang is not affiliated with National Marine data and information servcie, China.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research was supported in part by the National key R&D plan (Grant 2017YFC1405302) and National Natural Science Foundation of China (Grants 41771473, 41231171).

ORCID iDs

Zhixiang Fang

Yichen Wu

Supplemental material

Supplemental material for this article is available online.

Zhixiang Fang received the M.Sc. and Ph.D. degrees from Wuhan University, in 2002 and 2005, respectively, where he is currently a Professor with the State Key Laboratory for Information Engineering in Surveying, Mapping and Remote Sensing. His research interests include spatiotemporal modeling of human behavior, space-time GIS for transport, and intelligent navigation.

Yichen Wu is currently pursuing the M.S. degree with the State Key Laboratory for Information Engineering in Surveying, Cartography and Geographical Information Engineering, Wuhan University. Her research interests include data mining and exploration with big data.

Haoyu Zhong is currently pursuing the M.S. degree with the State Key Laboratory for Information Engineering in Surveying, Cartography and Geographical Information Engineering, Wuhan University. His research interests include spatial-time data mining and research of human mobility.

Jianfeng Liang received the M.S. degree from Nankai University. He is currently an Engineer with National Marine Data and Information Service, China. His research interests include marine data storage management technology, cloud computing technology, and big data technology.

Xiao Song received the M.S. degree from Chang’an University, in 2012. She is currently an Engineer with National Marine Data and Information Service, China. Her researche interests include system design and development, marine big data technology, and virtual machine technology.

References

China Meteorological Administration: Tropical Cyclone Data Information Center (2019). Available at: http://tcdata.typhoon.org.cn/zjljsjj_sm.html (accessed 23 December 2019).

Dilley

(2005) Natural Disaster Hotspots: A Global Risk Analysis. Washington, DC: World Bank Publications.

Dong

Hou

, et al. (2014) The characteristics of temporal and spatial distribution of storm surge in Guangdong province and storm surge hazard study at Yangjiang city. Acta Oceanologica Sinca 36(3): 83–93 (in Chinese).

Duan

Yang

Zhang

, et al. (2018) Improved deep hybrid networks for urban traffic flow prediction using trajectory data. IEEE Access 6: 31820–31827.

FEMA (2010) Multi-Hazard Loss Estimation Methodology Flood Model HAZUS ® MHMR5 Technical Manual. Washington, DC: Federal Emergency Management Agency.

Gornitz

(1991) Global coastal hazards from future sea level rise. Palaeogeography, Palaeoclimatology, Palaeoecology 89(4): 379–398.

Chu

Zhu

, et al. (2017) Warning water level determination for Luchaogang of Shanghai. DEStech Transactions on Materials Science and Engineering. icmsea_mce, DOI: 10.12783/dtmse/icmsea/mce2017/10876.

Guo

(1991) Preliminary study of storm surge disaster assessment in China. Journal of Nanjing University (Natural Sciences) 5: 18–22.

Guo

(2018) Research progress on economic loss pre-evaluation of storm surge disaster. Journal of Catastrophology 33(4): 164–168.

10.

Hatzikyriakou

Lin

(2017) Impact of performance interdependencies on structural vulnerability: A systems perspective of storm surge risk to coastal residential communities. Reliability Engineering & System Safety 158: 106–116.

11.

Helderop

Grubesic

(2019) Social, geomorphic, and climatic factors driving U.S. coastal city vulnerability to storm surge flooding. Ocean & Coastal Management 181(1): 104902.

12.

Hossain

(2015) Analysis of human vulnerability to cyclones and storm surges based on influencing physical and socioeconomic factors: Evidences from coastal Bangladesh. International Journal of Disaster Risk Reduction 13: 66–75.

13.

Kan

Tang

, et al. (2019) Traffic congestion analysis at the turn level using taxi’s GPS trajectory data. Computers, Environment and Urban Systems 74: 229–243.

14.

Kasperson

(2013) Global Environment Risk. Tokyo: United Nations University Press.

15.

Khouakhi

Snoussi

Niazi

, et al. (2013) Vulnerability assessment of Al Hoceima Bay (Moroccan Mediterranean Coast): A coastal management tool to reduce potential impacts of sea-level rise and storm surges. Journal of Coastal Research 65: 968–973.

16.

Kong

Yang

(2015) Measuring traffic congestion with taxi GPS data and travel time index. In: Proceeding of the 15th COTA international conference of transportation professionals, Beijing.

17.

Kuang

Jiang

(2015) Detecting traffic anomalies in urban areas using taxi GPS data. Mathematical Problems in Engineering 2015(PT.17): 1–13.

18.

Liang

Zheng

, et al. (2012) The scaling of human mobility by taxis is exponential. Physica A: Statistical Mechanics and Its Applications 391(5): 2135–2144.

19.

Lin

Yang

(2019) Economic loss assessment of typhoon based on BP neural network and VSM. Journal of Catastrophology 34(1): 22–26.

20.

Lin

Emanuel

Smith

, et al. (2010) Risk assessment of hurricane storm surge for New York city. Journal of Geophysical Research 115: D18.

21.

Liu

Kang

Gao

, et al. (2012) Understanding intra-urban trip patterns from taxi trajectory data. Journal of Geographical Systems 14(4): 463–483.

22.

Liu

Wang

Xiao

, et al. (2012) Urban land uses and traffic “source-sink areas”: Evidence from GPS-enabled taxi data in shanghai. Landscape and Urban Planning 106(1): 73–87.

23.

Yang

, et al. (2017) Estimating tropical cyclone size in the northwestern pacific from geostationary satellite infrared images. Remote Sensing 9(7): 728.

24.

Luo

Dong

Dou

, et al. (2017) Analysis on spatial-temporal features of taxis’ emissions from big data informed travel patterns: a case of shanghai, China. Journal of Cleaner Production 142: 926–935.

25.

Meteorological Bureau of Shenzhen Municipality (2019) Climate characteristics of Shenzhen. Available at: http://weather.sz.gov.cn/qixiangfuwu/qihoufuwu/qihouguanceyupinggu/qihougaikuang/ (accessed 23 December 2019).

26.

Nong

McCollum

, et al. (2010) Probabilistic storm surge heights for the US using full stochastic events. In: The 29th conference on hurricanes and tropical meteorology, Tucson, Arizona.

27.

Pan

, et al. (2013) Land-use classification using taxi GPS traces. IEEE Transactions on Intelligent Transportation Systems 14(1): 113–123.

28.

Shenzhen National Economic and Social Development Statistics Bulletin (2018) Shenzhen National Economic and Social Development Statistics Bulletin of 2018. Available at: http://www.sz.gov.cn/zfgb/2019/gb1098/201904/t20190429_17146572.htm (accessed 23 December 2019).

29.

Shi

Wang

, et al. (2000) Probing on storm surge and storm surge disaster grading. Marine Forecasts 17(2): 12–15 (in Chinese).

30.

Shi

Tan

Guo

, et al. (2013) A review of risk assessment of surge disaster. Advances in Earth Science 28(8): 866–874.

31.

Fang

Luo

, et al. (2018) Understanding the dynamics of the pick-up and drop-off locations of taxicabs in the context of a subsidy war among E-Hailing apps. Sustainability 10(4): 1256.

32.

Tang

Liu

Wang

, et al. (2015) Uncovering urban human mobility from large scale taxi GPS data. Physica A: Statistical Mechanics and Its Applications 438: 140–153.

33.

Vickery

Skerij

Twisdale

(2000) Simulation of hurricane risk in the U.S. using empirical track model. Journal of Structural Engineering 26(10): 1222–1237.

34.

Wang

Pan

Yuan

, et al. (2015) A comparative analysis of intra-city human mobility by taxi. Physica A: Statistical Mechanics and Its Applications 420(25): 134–147.

35.

Wang

Yuan

, et al. (2013) Visual traffic jam analysis based on trajectory data. IEEE Transactions on Visualization & Computer Graphics 19(12): 2159–2168.

36.

Huang

You

, et al. (2019) Recent progress in taxi trajectory data mining. Acta Geodaetica Sinica 48(11): 1341–1356.

37.

Liu

Hao

, et al. (2014) Rapid traffic congestion monitoring based on floating car data. Journal of Computer Research and Development 51(1): 189–198.

38.

Tan

(1998) Research on the method of evaluating the economic loss caused by storm surge disaster. Marine Science Bulletin 17(1): 1–12.

39.

Yang

Tang

, et al. (2005) Regulating taxi services in the presence of congestion externality. Transportation Research Part A: Policy and Practice 39(1): 17–40.

40.

Yang

Lei

(2007) Features of typhoon in southeast coastal regions of China in the recent 54 years. Scientia Meteorologica Sinica 27(4): 413–428.

41.

Yasuda

Mase

Kunitomi

, et al. (2010) Stochastic typhoon model and its application to future typhoon projection. Coastal Engineering: 2..

42.

Yin

Wei

(2012) The evaluation techniques of the socio-economic loss caused by storm surge disaster. Marine Environment Science 31(6): 835–842.

43.

Ying

Zhang

, et al. (2014) An overview of the China meteorological administration tropical cyclone database. Journal of Atmospheric and Oceanic Technology 31: 287–301.

44.

Yuan

Zheng

Xie

, et al. (2012) Discovering regions of different functions in a city using human mobility and POIs. In: Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining, pp.186--194.

45.

Yuan

Zhao

(2016) Combination evaluation and case analysis of vulnerability of storm surge in coastal provinces of China. Haiyang Xuebao 38(2): 16–24.

46.

Zhang

Lin

, et al. (2017) Taxi passenger demand modeling based on big data from a roving sensor network. IEEE Transactions on Big Data 3(3): 362–374.

47.

Zhang

Meng

Liu

, et al. (2015) Efficient vehicles path planning algorithm based on taxi GPS big data. Optik – International Journal for Light and Electron Optics 127(5): 2579–2585.

48.

Zhao

Chinnasamy

Tarkoma

(2015) Automatic city region analysis for urban routing. In: 2015 IEEE International Conference on Data Mining Workshop(ICDMW). (pp.1136--1142). Atlantic City, NJ:IEEE . DOI: 10.1109/ICDMW.2015.176.

49.

Zhao

(2013) Analyze Spatial and Temporal Characteristics of Exhaust Emissions Around Intersections Based on Taxi Trajectory Data. Shanghai: East China Normal University.

50.

Zhao

Cui

(2017) A network centrality measure for analyzing urban traffic flow: A case study of Wuhan, China. Physica A: Statistical Mechanics and Its Applications 478: 143–157.