《大数据与城市规划 (29).pdf》由会员分享，可在线阅读，更多相关《大数据与城市规划 (29).pdf（13页珍藏版）》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、Contents lists available at ScienceDirectCitiesjournal homepage: urban area delineations of cities on the Chinese mainland usingmassive Didi ride-hailing recordsShuang Maa,Ying Longb,aSchool of Architecture,Tsinghua University,Beijing 100084,ChinabSchool of Architecture and Hang Lung Center for Real
2、 Estate,Tsinghua University,Beijing 100084,ChinaA R T I C L E I N F OKeywords:Functional urban areaCar-hailing recordsNational levelDelineating standardsCity systemA B S T R A C TThe problem associated with a citys administrative boundary being“under-”or“over-bounded”has become aglobal phenomenon.A
3、citys administrative boundary city does not effectively represent the actual size andimpact of its labor force and economic activity.While many existing case studies have investigated the functionalurban areas of single cities,the problem of how to delineate urban areas in geographic space relating
4、to largebodies of cities or at the scale of an entire country has not been investigated.This study proposed a method forFUA identification that relies on ride-hailing big data.In this study,over 43 million anonymized 2016 car-hailingrecords were collected from Didi Chuxing,the largest car-hailing on
5、line platform in the world(to the best of ourknowledge).A core-periphery approach is then proposed that uses nationwide and fine-grained trips to un-derstand functional urban areas in Mainland China.This study examined 4456 out of all 39,007 townships in anattempt to provide a new method for the def
6、inition of urban functional areas in Chinese Mainland.In addition,four types of cities are identified using a comparison of functional urban areas with their administrative limits,and a further evaluation is conducted using 23 Chinese urban agglomerations.With the rapidly increasing use ofinternet-b
7、ased ride-hailing services,such as Didi,Grab,Lyft,and Uber,globally,this study provides a practicalbenchmark for the delineation of functional urban areas at larger scales.1.IntroductionThe problem associated with a citys administrative boundary being“under-”or“over-bounded”has become a global pheno
8、menon.A citysadministrative boundary city does not effectively represent the actualsize and impact of its labor force and economic activity.For instance,alarge number of residents live in Yanjiao and Sanhe in Hebei Province,China,but they commute to central Beijing for work.Such a commutingpattern i
9、nevitably generates substantial influences on the economy,housing,and the environment,and these influences frequently takeplace beyond the traditional administrative boundary of a city.Theconcept of functional urban areas(FUAs or metropolitan areas orfunctional regions),where cities,towns,and rural
10、areas are socio-economically tied to the urban core,is an important tool for under-standing economic and political regions of urban areas to address theaforementioned challenges and problems.FUAs offer more preciseboundaries for urban planning and management that will assist withinfrastructure const
11、ruction,regional coordinated development,andenvironmental governance of projects,such as highway constructionand other essential public services.FUAs are also imperative to digeststatistical information.For example,the U.S.Office of Management andBudget uses FUAs to delineate housing subsidies,wage
12、levels,andmedical subsidies(Zhang,2010).In addition,understanding the geo-graphic distribution of FUAs will clarify human interactions with theenvironment and present a new relationship between urbanization andthe biosphere(Reba,Reitsma,&Seto,2016).The delineation of FUAs is a hot topic,and economic
13、s,humangeography,and remote-sensing researchers have attempted to developalternative approaches to define FUAs(Bode,2008;Chen et al.,2016;Chi,Jiao,Dong,Gu,&Ma,2016;Duranton,2015;Newman,2004;Xie&Ning,2005).International organizations,such as the European Spa-tial Planning Observation Network(ESPON),t
14、he Organization forEconomic Cooperation and Development(OECD),and the Word Bank,also delineate FUAs using global data sets of cities.Economists believethat cities contribute economic benefits because of their labor markets(Marshall,1890),thus commuting patterns would provide a betterpicture of the d
15、ominant urban economic geography for the delineationof FUAs(Duranton,2015;Newman,2004;Ratti et al.,2010;Shen&Batty,2019).For example,the U.S.Census Bureau(USCB)explicitlydefines FUAs as counties where at least 25%of the workers living in thehttps:/doi.org/10.1016/j.cities.2019.102532Received 14 Sept
16、ember 2018;Received in revised form 7 September 2019;Accepted 18 November 2019Corresponding author.E-mail address:(Y.Long).Cities 97(2020)1025320264-2751/2019 Published by Elsevier Ltd.Tcounty work in the central county or counties of the Core Base Statis-tical Areas(USCB,2014).ESPON used 10%as the
17、commuting thresholdto delineate FUAs in Germany(ESPON,2007).However,for manycountries,especially developing countries,commuting data are difficultto gather.Although a few indicators that reflect functional influences,such as travel distance,travel time,and road networks,can implycommuting conditions
18、(Korcelli,2008,OMB,2010;ESPON,2007),commuting flow data are the preferred approach used to delineate FUAsas functional local labor markets(Bosker,Park,&Roberts,2018).The development of the internet of things(IOT),the continued de-velopment of information and communications technology(ICT),andthe exp
19、ansion of big data create new opportunities to observe the so-cioeconomic composition of the territories and track human movements(Shen,Sun,&Che,2017;Yuan et al.,2015).In particular,the fastgrowth of mobile internet,cloud computing,and location-based ser-vices make the ride-hailing companies,Didi Ch
20、uxing,Grab,Lyft,andothers,scale users around the world,and these companies serve mil-lions of users per day(Pham et al.,2017).Data sets from platforms suchas Uber,API,and Lyft have become increasingly popular in the recentliterature(Chaudhari,Byers,&Terzi,2018;Dong,Wang,&Zhang;Schaller,2017;Agatz,Er
21、era,Savelsbergh,&Wang,2011).They collectand share vast amounts of information,and these platforms can providea new lens to delineate FUAs using a community detection algorithm.In2018,Ford,Uber,and Lyft joined the SharedStreets project to provide acloud platform to share data from public and private
22、entities to helpcities make more informed decisions.This study proposed a method for FUA identification that relies onride-hailing big data.Access was provided to ride-hailing data(in-cluding both privately owned and corporate taxis registered on the DidiChuxing platform)from three working days by t
23、he Didi ChuxingCompany(the largest online ride-hailing platform in China)in 2016.We aggregated orientation(O)township and destination(D)townshipof every anonymous traveler into a Geographic Information System(GIS)platform.For identification of FUAs,the following was concluded:(1)The core areas shoul
24、d be defined on the basis of urban built-upareas.Thus,a township with a sufficient urban built-up area should be acandidate for a core FUA.(2)If a township had high flow density,thistownship should be a considered part of the core FUA.(3)If an adjacenttownship had high interaction with the core area
25、s,there should be alarge percentage of residents commuting to core areas.Specifically,thiswork(1)analyzed FUAs for all of the Chinese mainland based onmassive Didi ride-hailing records;(2)proposed suggestions to adjust theadministrative boundaries of cities on the Chinese mainland by com-paring with
26、 defined FUAs;(3)evaluated urban agglomeration devel-opment by evaluating the FUA rate in every urban agglomeration;and(4)compared FUA delineations using Didi ride-hailing records withFUA delineations using previous approaches.This approach provided a straightforward way to identify FUAs.With the in
27、creasing availability of open travel data sources and theincreasing possibility of shared ride-hailing data,this approach can beapplied for the widespread delineation of FUAs.This paper is organized as follows.Section 2 reviews the primaryliterature regarding approaches for FUA delineations.In Secti
28、on 3,thestudy area,data,and methodology used to identify FUAs are discussed.Results are analyzed in Section 4,including a delineation of FUA pat-terns in China.The potential applications are discussed in Section 5,including suggestions for administrative boundary adjustments and anevaluation of urba
29、n agglomeration development.The final sectioncompares the delineation results using previous approaches and dis-cusses the Didi records,research biases,and potential policy applica-tions of this study.2.Literature reviewThe literature has several definitions of FUAs.Generally,FUAs areregions where s
30、ocioeconomic interactions are stronger within theirTable 1Key publications that describe FUA delineations.PublicationStudy areaDataMethodSpatial unitShen&Batty,2019LondonDisaggregate commuting dataMultilevel modularity optimization algorithm that detects communitystructures within weighted flow grap
31、hs.Middle-layer Super Output AreaBosker et al.,2018IndonesiaCommuting flows,remotely sensed nighttimelights,and populationSystematic comparisons of these approaches using different data.Subnational administrative unitsHoussou,Guillaume,&Prigent,2019PortoTaxi flows and POI data setsA graph-based appr
32、oach with community detections that combine the useof POIs.A cartographic division using roadnetworks.Zhang,Wang,Liu,Li,&Pang,2016Shenyang and BeijingRemotely sensed imagesA Convolutional Neural Network(CNN)based method used to classifyaerial images.Disjoint regions using major roadsArcaute et al.,2
33、01521 cities in England andWalesCommuting and population densityConsiders density-based cities and then adds areas to cities according to acommuting threshold.WardsYuan et al.,2015BeijingData sets for POIs,taxi trajectories,buses,andsubwaysA topic model-based approach to fuse POIs and mobility patte
34、rns.Disjoint regions using major roadsZhu,Yang,Zhong,Seiter,&Trster,2015SeattleCrowd-Augmented Travel Survey DataA supervised learning via multi-output regression for travel survey data.Census cellsDemsar,Reades,Manley,&Batty,2014Greater LondonTaxi flowsEdge-based community detection algorithm that
35、uses the calculation ofthe similarity between each pair of vertices.Traffic Analysis Zones(TAZ)Dijkstra&Poelman,2014European UnionPopulation gridApproach associated with dense clusters of populations.LAU2sGajovi,2013Kraljevo in SerbiaDistances,census and some calculated indexesTwo methods of unsuper
36、vised learning:k-means clustering and self-organizing maps in machine learning.Administrative or statistical territorialunitsFarmer&Fotheringham,2011IrelandTravel-to-work dataNetwork-based method for network of travel-to-work flows withgeographic weighting.Electoral district(EDs)Ratti et al.,2010Gre
37、at BritainTelephone callsA fine-grained approach based on analyzing networks of individualhuman transactions.9.5 km by 9.5 km pixelsKarlsson&Olsson,2006Fyrstad in SwedenCommuting flowsThe local labor market,the commuting zone,and the accessibilityapproachMunicipalityS.Ma and Y.LongCities 97(2020)102
38、5322boundary than other boundaries.According to the OECD,an FUA is aterritorial unit that results from the organization of social and economicrelations in that its boundaries do not reflect geographic particularitiesor historic events(OECD,2002).The FUA is also described as a geo-graphic region in w
39、hich within-region interactions in terms of com-muter travel and work flows are maximized,and between-region in-teractions are minimized(Farmer&Fotheringham,2011).As such,commuting patterns over space are fundamental when delineatingFUAs(Fotheringham&OKelly,1989).The study areas,data used,methods,an
40、d spatial units used in the current literature are summar-ized in Table 1.The literature proposes several approaches to capture FUAs by in-ferred commuting(Crvers,Hensen,&Bongaerts,2009;Ratti et al.,2010)or observed commuting(Arcaute et al.,2015;Houssou et al.,2019;Shen&Batty,2019).Among the inferre
41、d community approaches,local economic outcomes,telephone calls,Twitter networks,the ac-cessibility of public transport,and other indicators have been used toindicate inferred commuting patterns(Korcelli,2008;Mulek,2013;Takhteyev,Gruzd,&Wellman,2012).Observed commuting data arebetter and more straigh
42、tforward to accurately capture interactions.Farmer and Fotheringham(2011)delineated the FUAs in Ireland byobserving a travel-to-work flow network.They maximized the mod-ularity of a network of travel-to-work flow to produce a regionalboundary that exhibited less interaction than expected between re-
43、gions.Inspired by Goddard(1973),Demar investigated the possibilityof using an edge-based community detection algorithm(Demsar et al.,2014)to identify overlapping FUAs using taxi flows in the greaterLondon area.Other popular approaches have been based on satelliteimagery of urban built-up areas or li
44、ght nights(Zhang et al.,2016;Elvidge,Kihn,&Davis,1996;Danko,1992;)or a cluster algorithmbased on population data sets,such as LandScan or GHS-Pop(Boskeret al.,2018;Dijkstra&Poelmans,2014).The use of these more ac-cessible data sets as an alternative method,however,is not sufficient asthe direct use
45、of commuting patterns,because the final aim of an FUAdelineation is to capture the social and economic effects using com-muting patterns.Most current publications have focused on one or two cities or atypical region as the spatial unit of small administrative unit or roadnetwork.To delineate FUAs at
46、 a broader level,Dijkstra and Poelman(2014)developed the cluster algorithm to identify a metro area as adense population cluster.The benefits of this approach include greatercomparability among different countries in the European Union anddata availability.The collection of commuting data is needed,
47、however,before studying the interactions among boundaries of cities.In this study,we proposed a new approach for FUA delineations forthe Chinese mainland,which currently lacks well-defined FUAs that usemassive ride-hailing data.This study addressed current limitations toconduct delineations for the
48、entire country with the use of availableride-hailing data at the township scale,which is the smallest scale thathas been used to delineate FUAs in China.With the rapidly increasinguse of internet-based ride-hailing services,such as Didi,Grab,Lyft,andUber,globally,this study provides a practical benc
49、hmark for cityplanning and statistical analyses.3.Methods3.1.Study area and dataThe Chinese city system has long been defined from an adminis-trative view,and most of the statistical data that have been gatheredhave corresponded to administrative cities(Long,2016).According tothe Ministry of Housing
50、 and Urban-rural Development of the PeoplesRepublic of China(MOHURD,2014),China has a total of 654 cities.The study area chosen for this investigation covered the entire Chinesemainland territory.The study area was not limited to administrativeareas of cities(city proper).Instead,administrative boun