https://www.pnas.org/content/suppl/2018/11/15/1815928115.DCSupplemental

Data

We collected one-week trip records from 2011 to 2017 in the Beijing subway system, excluding weekends and holidays (see Table S1). On average, each one-week dataset includes transit records of about 5 million passengers identified according to card ID numbers, and contains about 4 million trips per day. Each transit record includes the boarding time, alighting time, origin station, destination station, and card ID. 嫉妒的眼神 Meanwhile, we collected geographic attributes of metro stations and lines and their opening year. We obtain the average real estate resell price of housing around stations in May, 2018 (source: https://bj.lianjia.com/), as the largest share of Beijing’ real estate market is reselling. It is the data available currently.

Table S1

我也不明白为什么选用的是这段时间的数据

A map of Beijing subway system with spatial data can be seen in Figure S1, and an evolutionary（渐进的） map of lines and stations can be seen in Figure S2. In the buffer（缓冲区） analysis via ArcGIS, 2752 residential communities are within 1km catchment areas of stations, while 3979 are within 2km catchment areas. This number increases to 4290 when we look at the 3km catchment areas, accounting for about 95% of residential communities.This area aggregates about 90% of population in Beijing. This phenomenon motivates the analysis of job and housing dynamics with smartcard data.

Fig.S1

Fig.S2

Fig.S3

R1. Rules in identifying work station and home station

This paper employs the method by (1, 2) to identify workplaces and residences of regular commuters. In detail, the method finds the most preferred station near individual workplace or residence for each commuter, namely ‘work station’ and ‘home station’. We identify the work station for a commuter based on the following rules:

1. Access an individual dataset when the commuter travels 4 or 5 days. One criterion(标准）is added here to focus on the study of regular commuters.
2. Find out trips whose boarding time is 6 hours later than the alighting time of the previous trip. Namely, hours, where denotes the boarding time of trip and denotes the alighting time of trip .
   我觉得这个方法很好，因为通勤者的上班时间不同，如果只用高峰时期去划分的话，容易遗漏一些数据（比如教育机构的老师14-21点工作）；不过如果研究的是rush hour的通勤特征，可能直接用高峰时间来划分通勤流会比较好。（我还没决定要用哪个）
3. Exclude（排除） trips if they do not occur on the same day as the previous trip. Then all remaining trips are return commuting trips for this commuter.
   我的数据中也有这样的问题，但是我会考虑采用别的处理方式
4. Among origin stations of commuting trips, the station where the commuter visited most frequently is regarded as the job station j .
   Similarly, the program identifies the commuter’s home location based on the following rules:
   (1) Access the individual dataset where the job station was identified above.
   (2) Among destination stations of commuting trips, the station where the commuter visited most frequently is regarded as the home station h.
   先确定工作地点，再在能确定工作地点的样本中进行居住地定位（学到了学到了）

R2. Rules in capturing moving behavior under network expansion

The Beijing subway network nearly doubled from 228 to 609 kilometers from 2011 to 2017. On average, about 20 new stations opened each year. This paper conducts a year-to-year analysis to capture the moving behavior under network expansion. We design rules to capture moving behavior as follow:

1. If a workplace/ residence identified from a station already existing to a station newly opened that year, and the distance between two stations is within 3 km, to reduce false positives we regard the commuter as not switching workplace/ residence. Instead, this paper assumes that they have already worked/resided there and are reducing their commuting costs by choosing a new, now more optimal, station.
2. If the distance is beyond 3 km, this paper assumes that the commuter relocates the residence/workplace.
3. If commuters change their work station and/or home station between already existing stations, we assume that commuters change their workplace and/or residence.
4. If home/ job stations identified in two years are adjacent, this paper regards this phenomenon as non-moving behavior to avoid false positives at the expense of more false negatives.

References

1. Zhou J, Long Y (2014) Jobs-housing balance of bus commuters in beijing: exploration with large-scale synthesized smart card data. Transportation Research Record: Journal of the Transportation Research Board (2418):1–10.
2. Zhou J, Murphy E, Long Y (2014) Commuting efficiency in the beijing metropolitan area: An exploration combining smartcard and travel survey data. Journal of Transport Geography 41:175–183.