大数据与城市规划 (38).pdf

Combining smart card data and household travel survey to analyzejobshousing relationships in BeijingYing Longa,Jean-Claude ThillbaBeijing Institute of City Planning,Beijing 100045,ChinabDepartment of Geography and Earth Sciences,The University of North Carolina at Charlotte,Charlotte,NC 28223,USAa r t i c l ei n f oArticle history:Available online 4 April 2015Keywords:Bus smart card dataJobshousing spatial mismatchCommuting tripRule-basedBeijinga b s t r a c tLocation Based Services(LBS)provide a new perspective for spatiotemporally analyzing dynamic urbansystems.Research has investigated urban dynamics using LBS.However,less attention has been paidto the analysis of urban structure(especially commuting pattern)using smart card data(SCD),whichare widely available in most large cities in China,and even in the world.This paper combines bus SCDfor a one-week period with a oneday household travel survey,as well as a parcel-level land use mapto identify jobhousing locations and commuting trip routes in Beijing.Two data forms are proposed,one for jobshousing identification and the other for commuting trip route identification.The resultsof the identification are aggregated in the bus stop and traffic analysis zone(TAZ)scales,respectively.Particularly,commuting trips from three typical residential communities to six main business zonesare mapped and compared to analyze commuting patterns in Beijing.The identified commuting tripsare validated by comparison with those from the survey in terms of commuting time and distance,and the positive validation results prove the applicability of our approach.Our experiment,as a first steptoward enriching LBS data using conventional survey and urban GIS data,can obtain solid identificationresults based on rules extracted from existing surveys or censuses.?2015 Elsevier Ltd.All rights reserved.1.IntroductionThis paper identifies jobhousing location dyads and commut-ing patterns in Beijing using smart card data(SCD)that store thedaily trip information of bus passengers.It proposes and imple-ments a method for deriving commuting patterns from increas-ingly common SCD for informing city planners and transitsystem managers about patterns of transit usage across spaceand through time as well as about mobility patterns in a largeand fast growing city region.Related research on jobshousingrelationships has conventionally used data acquired through sur-veys or censuses.The increasing pervasiveness of location-basedservices(LBS)associated with the prevalence of positioning tech-nologies has led to the creation of large-scale and high-qualityspace-time datasets(Jiang&Yao,2006).This development has alsocreated opportunities to better describe and understand urbanstructures1in multiple dimensions.These datasets have been shownto be important for analyzing urban and environmental systemssuch as relationships between housing and jobs(Batty,1990).Meanwhile,a geo-tagged smart card system is an effective alterna-tive tool for individual data acquisition necessary to analyze urbanspatial structures.Various types of fine-granularity individual data generated byLBS technologies have been extensively leveraged to analyze urbanstructures(Ahas&Mark,2005;Lu&Liu,2012).With respect tohandheld Global Positioning System(GPS)devices,Newhaus(2009)used location data to record and visualize urban diaries,while Gong,Chen,Bialostozky,and Lawson(2012a)elicited travelmodes of travelers in New York City.Liu,Andris,and Ratti(2010)identified taxi drivers behavior patterns from their daily digitaltrajectories,and Yue et al.(2012)used these trajectories to cali-brate a spatial interaction model.With respect to mobile phonesystems(see Steenbruggen,Borzacchiello,Nijkamp,&Scholten,2013 for a review),Ratti,Pulselli,Williams,and Frenchman(2006)evaluated the density and spatiotemporal characteristicsof urban activities using mobile phone data in Milan,Italy,whereasWan and Lin(2013)studied fine-scale individual activities,Yuan,Raubal,and Liu(2012)correlated mobile phone usage and city-wide travel behavior in Harbin,China and Chi,Thill,Tong,Shi,and Liu(In press)exploited network properties of mobile phonehttp:/dx.doi.org/10.1016/penvurbsys.2015.02.0050198-9715/?2015 Elsevier Ltd.All rights reserved.Corresponding author.Tel.:+86 10 88073660;fax:+86 10 68031173.E-mail address:(Y.Long).1The concept of urban structure concerns the spatial arrangement of public andprivate spaces in cities and the degree of connectivity and accessibility.In this paper,the concept is focused particularly on the spatial concentration of resident populationand employment(Anas,Arnott,&Small,1998).Computers,Environment and Urban Systems 53(2015)1935Contents lists available at ScienceDirectComputers,Environment and Urban Systemsjournal homepage: to reveal urban hierarchical structures at the regional scale.Asfor Wi-Fi,Rekimoto,Miyaki,and Ishizawa(2007)used Wi-Fi-basedlocation detection technology to log the locations of device holdersfrom received Wi-Fi beacon signals,a technology that works bothindoors and outdoors.Torrens(2008)developed a system to detectWi-Fi infrastructure and transmission and analyze their geographicproperties,and tested this system in Salt Lake City,Utah.Meanwhile,the discipline of time geography established byHagerstrand(1970)also benefited from the development of LBSby retrieving more objective data.In sum,various LBS technologieshave been successfully applied in urban studies.However,thesetechnologies remain immature and most research on urban struc-ture continues to employ data from the urban physical space orquestionnaire surveys(with a few studies as exceptions,e.g.Kwan(2004).Access to large-scale micro datasets remains a bar-rier to their widespread use for research,planning and manage-ment(Long&Shen,2013).A smart card that records full cardholders bus trip informationis an alternative form of location-acquisition technology.Smartcard automated fare collection systems are increasingly deployedin public transit systems.Along with collecting revenue,such sys-tems can capture a meaningful portion of travel patterns of card-holders,and the data are useful for monitoring and analyzingurban dynamics.Since the 1990s,the use of smart cards hasbecome significant owing to the development of the Internet andthe increased complexity of mobile communication technologies(Blythe,2004).As of 2007,Intelligent Transportation Systems(ITS)that incorporate smart card automated fare systems eitherexisted or were being established in over 100 Chinese cities,as wellas in many other cities around the world(Zhou,Zhai,&Gao,2007).The data generated by smart card systems track the detailedonboard transactions of each cardholder.We argue that smart cardtechnology can deliver valuable information because it is a con-tinuous data collection technique that provides a complete andreal-time bus travel diary for all bus travelers.SCD can be usedto validate traditional travel models applied to public transit.Incontrast to SCD collection,conventional travel behavior surveyshave the drawbacks of being expensive and infrequent.Notably,transit SCD collects data in fundamentally the same way as anAVI(automatic vehicle identification)system,which has beenwidely used in the United States to automatically identify vehicles.AVI is used in some states in the US for planning purposes.Onesuch example is New York,where the E-ZPass tag is used as partof the TRANSMIT system.Previous studies have advocated using SCD to make decisionson the planning and design of public transportation systems(seePelletier,Trepanier,and Morency(2011)for a review).In SouthKorea,Joh and Hwang(2010)analyzed cardholder trip trajectoriesusing bus SCD from ten million trips by four million individuals,and correlated these data with land use characteristics in theSeoul Metropolitan Area.Jang(2010)estimated travel time andtransfer information using data on more than 100 million tripstaken in Seoul on the same system.Roth,Kang,Batty,andBarthe lemy(2011)used a real-time Oyster card database ofindividual traveler movements in the London subway to revealthe polycentric urban structure of London.Gong et al.(2012b)explored spatiotemporal characteristics of intra-city trips usingmetro SCD on 5 million trips in Shenzhen,China.Also,Sun,Axhausen,Lee,and Huang(2013)used bus SCD in Singapore todetect familiar strangers.There is considerable research on inferring home and job loca-tions from individual trajectories like mobile phone call datarecords and location-based social networks(LBSN).For the identi-ficationofhomelocations,Lu,Wetter,Bharti,Tatem,andBengtsson(2013)regarded the location of the last mobile signalof the day as the home location of a mobile user.The mostfrequentlyvisitedpoint-of-interest(POI)(Scellato,Noulas,Lambiotte,&Mascolo,2011)or grid cell(Cheng,Caverlee,Lee,&Sui,2011;Cho,Myers,&Leskovec,2011)was regarded as a LBSNusers home location.It is not easy to infer home locations fromLBSN with a high spatial resolution.Compared to approaches tohome location identification,there are fewer studies on identifyingjob locations based on trajectories,with Cho et al.(2011)usingLBSN and Isaacman et al.(2011)using cellular network data asnotable exceptions.It should be mentioned that taxi trajectoriesare not well suited for identifying a passengers home and job loca-tions considering the passenger-sharing nature of taxis.However,less attention has been paid to using SCD to identify home andjob locations as well as to analyze jobhousing dyadic relation-ships and commuting patterns in a metropolitan region.This paper regards jobhousing dyadic relationships and com-muting pattern analysis as a showcase for using SCD to urban spa-tial analysis.We argue that job and home locations,their dyadicrelationships,and commuting trips can be identified from SCDand serve as valuable information on the modalities of use of theurban space in its residents.We propose a methodology to thiseffect and use Beijing as a case study to test its implementation.The identification results are validated using travel behavior sur-vey data from Beijing.This paper is organized as follows.Theretrieval of jobhousing trips from conventional travel behaviorsurveys is discussed in Section 2,and the SCD and other relateddatasets used in our research are presented in Section 3.Theapproaches for identifying home and job locations,as well as com-muting trips are elaborated in Section 4.In Section 5,the results ofjobhousing identification and commuting patterns are shown andanalyzed in detail.Finally,we discuss our work and present con-cluding remarks in Sections 6 and 7,respectively.2.Jobhousing trips in conventional travel behavior surveysTravel behavior surveys have been the primary means of datacollection on urban resident travel behavior for planning andmanaging urban transit systems(Beijing Transportation ResearchCenter,2009).There is a well-established tradition in geographyand urban planning to use surveys for tracking individual traveldiaries(Grling,Kwan,&Golledge,1994;Schlich,Schnfelder,Hanson,&Axhausen,2004).Travel behavior surveys track travelersocio-economic attributes,as well as trip origin and destination,time and duration,as well as trip purposes and travel modes.Onthe one hand,the travelers home and job locations are directlyrecorded in the survey together with his/her socioeconomic attri-butes,and both locations are mostly aggregated in the trafficanalysis zone(TAZ)scale.On the other hand,trips between workand home(muting trips)can be screened using the purposeattribute.These trips are also recorded using the inter-TAZ scalerather than a finer spatial scale.Therefore,jobhome locationdyads and trips have already been recorded in conventional travelsurveys,but mostly at the TAZ scale.Additionally,only a small por-tion of all households in a given city are surveyed due to time andcost constraints.Compared with travel behavior surveys,mining the enormousvolume of SCD can provide a more precise spatial resolution anda much larger sample,despite the SCD being unable to directly pro-vide jobhome location dyads and commuting trips.We will focuson using SCD to identify jobshousing relationships.Patterns ofcommuting trips from typical residential communities or to typicalbusiness centers can be visualized by identifying the results at afiner scale than is available in travel surveys because residentialcommunities or business zones are generally smaller than a TAZ.A more detailed commuting pattern is expected to reveal freshinformation on jobshousing relationships in a megacity such as20Y.Long,J.-C.Thill/Computers,Environment and Urban Systems 53(2015)1935Beijing.A shortcoming of SCD however is that they are devoid ofinformation on the cardholders socioeconomic attributes,andthe purpose of individual trips is also unknown.Conventional tra-vel surveys can supply such additional information for use in ana-lyzing SCD,and combining SCD with travel behavior survey data isa promising method of jobhousing analysis,which will be elabo-rated below.3.Data3.1.Bus routes,bus stops,and traffic analysis zones(TAZs)of BeijingGeographic Information Systems(GIS)layers of bus routes andstops are essential for geocoding and mapping SCD.There are 1287bus routes2(Fig.1a)in the Beijing Metropolitan Area(BMA),whichtotals 16,410 km2.These bus routes have 8691 stops in total(seeFig.1b).Note that a pair of bus platforms on opposite sides of a streetis considered a single bus stop.For instance,there are two bus plat-forms at Tiananmen Square,one on the south side of ChanganAvenue and the other one on the north side.In the GIS bus stop layer,the two platforms are merged into a single bus stop feature.3Theaverage distance between a bus stop and its nearest neighbor is231 m in the city.Relying on Ji and Gaos(2010)result that the num-ber of bus stops within an 800 m vicinity of a resident has a signifi-cant effect on their satisfaction with public transportation services inBeijing,we take the 800-m buffer zone around each stop to be itscatchment zone,so that the potential service area of a bus stop is2.0 km2.We overlay the calculated bus catchment with the pop-ulation density surface inferred from the 2010 sub-district level pop-ulation census.The estimated population within the catchment zoneof any of the 8691 stops is 14.8 million,or 75.5%of Beijings 19.6 mil-lion residents.We use Beijing TAZs to aggregate the analytical results for bet-ter visualization.In total,1118 TAZs are defined(Fig.1c)accordingto the administrative boundaries,main roads,and the planninglayout in the BMA.3.2.The one-week smart card datasetA smart card system has been deployed in the public transitsystem of Beijing since April 1,2006(Liu,2009).The system canautomatically track cardholders bus trip information.4The busshare of total trips taken in Beijing during 2008 was 28.8%,and thesubway share was 8.0%(Beijing Transportation Research Center,2009).Over 42 million smart cards(see Fig.2)have been issued inBe