AN ATTRACTIVENESS-BASED MODEL FOR SHOPPING TRIPS IN URBAN AREAS

Transcription

1 Autor manuscrpt, publsed n "12t World Conference on Transport Researc, Lsbonne : Portugal (2010)" An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles AN ATTRACTIVNSS-BASD MODL FOR SHOPPING TRIPS IN URBAN ARAS Jesus Gonzalez-Felu, Laboratore d conome des Transports, 14 Avenue Bertelot, Lyon (France), esus.gonzales-felu@let.s-lyon.cnrs.fr Jean-Lous Router, Laboratore d conome des Transports, 14 Avenue Bertelot, Lyon (France), ean-lous.router@let.s-lyon.cnrs.fr Carles Raux, Laboratore d conome des Transports, 14 Avenue Bertelot, Lyon (France), carles.raux@let.s-lyon.cnrs.fr alss , verson 1-21 Apr 2012 ABSTRACT Ts paper presents a modellng approac to caracterse prvate car soppng trps n a cty logstcs pont of vew, n order to connect tese movements wt tose belongng to urban fregt dstrbuton n te supply can. Te proposed modellng framework s a two-step procedure artculated as follows. Frst, an attracton model estmates te number of prvate car soppng trps arrvng to eac secton of a gven urban area. Second, a catcment area model relates te soppng trp destnatons wt te ouseold locatons. Te model s calbrated usng te data of bot a database of te commercal actvtes and te recent ouseold trp survey made n 2006 n Lyon urban area (France). We present te man results ssued of te varous smulatons as well as several applcaton examples and proposals, most of tem made n a publc polcy perspectve. Keywords: urban goods movement, soppng trps, smulaton, catcment area model, attractveness. INTRODUCTION Soppng trps are an mportant component of urban traffc, snce tey are related to movements of people and goods: n France tey represent about 50% of te total PCU km travelled by urban goods (ncludng commercal vecles). Moreover, travel patterns for soppng are subected to mportant mutatons, manly because of te urban developments and trends on new retal and logstc dstrbuton scemes. Soppng trp estmaton s usually made related to people, usng te classcal 4-steps modellng approac: t follows te ypotess tat tese trps depend on te same factors as oter trp purposes. However, several studes sow tat soppng trp generaton rates are related to factors tat are dfferent from tose of work trps for nstance. Moreover, soppng 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 1

2 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles trps are dffcult to caracterse due to te complexty and te varety of trp cans were at least one soppng actvty takes place, and te avalable data sources ave not been created for specfc soppng trp generaton. We propose a new modellng approac tat relates soppng trp flows to retalng supply. To do ts, we calculate te catcment area of eac zone, n terms of number of ouseolds of eac zone tat wll make a ourney tat contans at least one soppng stop at. Te approac uses standard ouseold travel survey data and avalable data about retal supply. alss , verson 1-21 Apr 2012 Frst, we present te man background ssues to settle te context of our researc. Ten, te model s presented, focusng respectvely on te defntons, assumptons and oter ypotess taken nto account, te attracton pase and te catcment area pase. After presentng te two pases of te model, we present te calbraton procedures and ter man results. Fnally, two examples of applcatons are proposed. More precsely, we sow a smulaton of extreme locaton and dstrbuton scenaros n te urban area of Lyon (France) and an analyss of te mpacts of e-commerce and telesoppng practces, n progressve usage ncreasng scenaros. Te scenaros are compared on te crtera of total PCU km travelled by urban goods and total emssons of CO 2, takng nto account te dfferent types of vecles used,.e. prvate cars, lgt cty fregters, urban trucks and sem-artculated vecles. BACKGROUND ISSUS In te last decades, cty logstcs as been developed to deal wt te man problems of urban fregt dstrbuton, studyng fregt movements n urban areas and proposng solutons to reduce congeston and polluton as man problematcs. Recent studes ave defned and caractersed te dfferent movements of urban goods (Pater, 2002; Ségalou et al., 2004; Russo and Com, 2006). Two categores are predomnant and represent about 90% of te overall urban goods movements: establsment supply movements, wc are related to fregt dstrbuton between te dfferent actvtes, and end-consumer commodty movements, were te purcased goods are moved by te consumer, related to soppng trps. Te frst group of urban fregt movements, wc corresponds to te excange of goods between dfferent establsments, as been one of te most studed subects n cty logstcs researc. nd-consumer movements, less studed n urban fregt transport scence, s an mportant component of urban fregt transport, representng about 50% of te total km.pcus (Passenger Car Unts) of urban fregt transportaton, were establsment supply movements represent only 40 % (Pater, 2002). Te remanng 10% contans te cty mantenance and constructon logstcs movements, te waste dstrbuton and a number of small partcular actvtes suc as postal servces, among oters In France, soppng trps represent 10% of te total person trps n workng days,.e. from Monday to Frday, and 25% n Saturday (Router et al., 2001). In terms of polluton, te urban transport of goods, ncludng end-consumer movements, produces about 25% of te total CO2 emssons for transport, 35% of te NOx emssons and 40 to 50% of te partculate (Ségalou et al., 2004). 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 2

3 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles Moreover, mutatons related to te development of new tecnologes and consumpton trends ave an mportant mpact on te soppng beavour, wc nfluences te dfferent travel abts, so te caracterstcs of te soppng trps (Router et al., 2009). Land use polces developed by publc admnstrators wll ave also an nfluence on commercal actvtes locaton and te servces tey propose (Router et al., 2001). Moreover, te locaton of retal actvty areas as an nfluence on fregt flow generaton, n bot sdes of te extended supply can (tradtonal fregt dstrbuton and end-consumer movements). Altoug te contrbuton of end-consumer movements to te urban flows of motorsed vecles s mportant, tey are rarely taken nto account n cty logstcs, at bot polcy makng support and transport plannng and optmsaton levels. Soppng trp estmaton s generally made by classcal four-step models (Ortuzar and Wllumsen, 2001; Henser and Button, 2001) but n oter felds non transportaton-based models can also be found (Tll and Tmmermans, 1992; Bawa abd Gos, 1999; Long-Lee and Pace, 2005; Kubs and Hartman, 2007). alss , verson 1-21 Apr 2012 Trp generaton models relate trp generaton rates to land-use and ouseold caracterstcs. In general, te focus of researc s te number of trps generated and ter geograpc dstrbuton. Most of tese models are ntegrated n oter metods lke te well-known famly of classcal four-steps models (Ortuzar and Wllumsen, 2001), wc s commonly used n general trp caractersaton. Te soco-economc caracterstcs of te trp makers are assumed to be sgnfcant determnants of travel beavour because te factors determnng te number of all types of trps are assumed to be te same as tose for soppng trps (Cubukcu, 2001). Tese factors are, among oters, ncome range, age, gender, employment status, car ownersp, and ouseold sze. Te pyscal and demograpc caracterstcs of te area nclude: employment, populaton, and densty. However, t s reasonable to beleve tat tere are metropoltan area caracterstcs and trp maker caracterstcs wc mpact s sgnfcant only for soppng trp generaton rates (Cubukcu, 2001). However, most of te studes dealng wt soppng trps specfcally are n general related to te regonal or natonal level (Vckerman and Bramby, 1985; Badoe and Steuart, 1997; Cubucku, 2001; Smma et al., 2005); few models are proposed to caracterse soppng trp generaton factors n a urban context (Ségalou, 1999; Black et al., 2007). Most of te works deal wt econometrc and emprcal approaces based on surveyed data, manly collected to calbrate ts type of models. Soppng trp dstrbuton s more eterogeneous. Te man categores are te classcal trp dstrbuton metods (Ortuzar and Wllumsen, 2001) based on gravtary and entropymnmzaton approaces, and dscret coce approaces (Tll and Tmmermans, 1992). Moreover, n real estate researc, catcment area models can also be used to estmate te man customer s locatons n order to estmate trp dstances and ouseold (or nabtant) travel beavour categores (Long-Lee and Pace, 2005; Kubs and Hartman, 2007). However, tese models are bult on specfc surveys for one store or one retalng area and ave not been conceved to be used wt standard data on several urban contexts. 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 3

4 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles TH PROPOSD MODL Defntons and model descrpton Te proposed model follows te defntons and prncples enounced n Gonzalez-Felu et al. (2010). In ts work; te autors propose to model drectly te prvate car soppng trps n order to compare tem to establsment supply flows. Accordng to several researces (Ségalou, 1999a; Dablanc et Peceur, 2000; Toler et al., 2005; Cambre de Commerce et d Industre de Lyon, 2008; Gonzalez-Felu et al., 2010), we can observe tan soppng trps are ncluded n more complex trp cans. In all tese cans, commercal actvtes are related to two connected soppng trps: alss , verson 1-21 Apr 2012 An nbound trp, arrvng at te consdered commercal area. Ts s te part of te trp can tat s defned as soppng trp n classcal passenger trp caractersaton. An outbound trp, startng at te consdered commercal area, were te goods tat ave been purcased are also travellng. Fgure 1: scema of te soppng-related trps Altoug te end-consumer goods movements take place only wen te outbound trp starts, tey are closely related to nbound trps, as bot types of soppng trps are always made consecutvely one after te oter (see Fgure 1). A partcular type of soppng-related trps s tat of trps wc purpose actvty at te orgn and te destnaton s related to soppng. Te Soppng-Soppng trps can be consdered as nbound trps for te arrvng zone and outbound trps for te departng zone. Trp generaton (attracton) In ts pase te soppng trps attracted by eac zone are generated. To do ts, we bult a model usng te general trp rate generaton modellng framework.te model wll be able to generate te overall soppng trp rates (all modes) at eac zone, for bot emsson and attracton. In a frst tme, we wll caracterse te soppng trps at te soppng destnaton, wc represents bot te nbound trp s destnaton and te outbound trp s orgn. We wll note te number of prvate car trps at a soppng destnaton te prvate car soppng attractveness of and notedt. 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 4

5 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles Trps are te basc unt for zone-aggregaton trp generaton models, and te output varable s n general te number of trps wc departure or arrval s located n a pre-dentfed zone durng a gven tme unt (Hobbs, 1979). As already sad, te proposed model estmates te prvate car soppng trp attracton rates (Ortuzar and Wllumsen, 2001). Followng te general trp generaton rate modellng framework (Hobbs, 1979; Cubucku, 2001; Ortuzar and Wllumsen, 2001), te number of soppng trps T made by prvate car startng at or arrvng to zone can be formulated n te followng way: T = f (Ret, X, Tec ) alss , verson 1-21 Apr 2012 were T s te total number of soppng trps generated n secton ; Ret te set of retalng actvty caracterstcs vector n secton ; X te set of soco-economc caracterstcs of people and ouseolds belongng to secton ; and Tec te set of tecnology caracterstcs. Ts equaton can be formulated as a vector functon: T = f (Ret, X, Tec) (1) T1 T2 wt T = ; T TZ Ret Ret Ret = Ret Ret 1 2 ; X Z X X = X X 1 2 Z Tec1 Tec2 ; Tec = Tec TecZ ac vector contans a set of possble varables tat can be ncluded n te soppng trp generaton model. For ts researc te varables consdered n te vector Ret are te followng: NrSMC : Number of small and medum stores; NrBS : Number of supermarkets and bg commercal surfaces; NrVBS : Number of ypermarkets and smlar stores; mp-smc : Number of employees n te secton s small and medum stores; mp-bs : Number of employees n te secton s supermarkets and bg commercal surfaces; mp-vbs : Number of employees n te secton s ypermarkets and smlar stores; CC : Presence of an commercal centre (Ségalou, 1999), presented as a bnary varable wc takes te value 1 f at least one extra-urban commercal centre s located nsde te secton s permeter and 0 oterwse. Tese varables are obtaned by aggregaton of te SIRN fle data, for te consdered cty. Te SIRN databases s a set of Frenc establsments nformaton data fles, wc 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 5

6 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles contans among oters te retal actvtes basc nformaton (locaton, sze category, commerce category, number of employees, etc.) for eac cty. Te populaton soco-economc caracterstcs vector X ncludes: POP: Populaton of te consdered secton : Number of ouseolds of te consdered secton DPOP: Populaton densty; DH: Houseold densty; Tese varables are extracted from INS 1 populaton census fles. In France, eac local admnstraton as access to te populaton and ouseold data of te natonal census. Oter nterestng databases are obtaned from te ouseold trp surveys, bot local or natonal, or te Commerce and Industry Camber natonal or regonal surveys, altoug tese data are less accessble and not avalable for all te medum urban areas. alss , verson 1-21 Apr 2012 Te tecnology vector Tec s ncluded n te framework only n a conceptual pont of vew. Current surveys do not ave enoug elements to well defne tele-soppng usage trends, and oter beavoural trends related to ome delvery. Moreover, current data sources do not allow to caracterse te relatons between telesoppng trends and car usage beavour (Beauvas, 2005). For tese reasons we wll not nclude n a frst tme te effects of telesoppng and ome delvery servces n our model, but only nclude tem n a conceptual modellng framework. Te trp attracton functon can be defned n dfferent ways. We propose a multlnear functon, defned as a pondered sum of te consdered varables, and can be obtaned by lnear regresson tecnques appled drectly on te avalable data. We can ten defne ts functon as follows: T = e a Ret e e + a X + f e f g a Tec g g We wll estmate te coeffcents by lnear regresson tecnques, more precsely by te square mnmum metod.<q Classcal approaces are appled to estmate te total number of soppng trps n a urban area (Vckerman and Bramby, 1984; Cubukcu, 2001) or tose trps for eac zone of a urban area (Ségalou, 1999; UK grocery trps). A dfferent approac (Gonzalez-Felu et al., 2010) proposes a modellng approac on tree categores of urban space: Te man central urban area contans te man cty of te urban regon and sometmes also oter urban suburbs wc can be assmlated to te man cty, because of a contnuty of te urban landscape. Te near perpery ncludes te urban zones lmtng wt te central urban area frst rng. 1 Frenc Natonal Insttute of Statstcal Studes 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 6

7 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles Te rest of towns of te urban area belong to te far perpery. Te man problem of usng ts approac s tat splttng te statstcal populaton n tree categores decreases te number of zones used to calbrate eac of te tree models, makng tese tree sets of data not always enoug accurate to ensure a qualty econometrcs analyss n te calbraton pase of tese models. For tese reasons, we wll test bot approaces and compare tem. Catcment area dstrbuton Once te nbound soppng trps are generated at ter destnatons, we need to defne te orgn of tese trps and te destnaton of te related outbound trp. From te ouseold trp surveys results, we observe tat ouseolds correspond to 80% of te nbound orgns and to 85% of te outbound destnatons. Terefore 70% of te soppng trps are ncluded n ouseold-soppng-ouseold cans. alss , verson 1-21 Apr 2012 We propose to connect soppng destnatons to ouseolds usng a catcment area model. Te proposed approac s a gravty probablstc model. Ts model as a smlar form tan classc gravty dstrbuton models (Ortuzar and Wllumsen, 2001): N = A T c were A = k 1 c k In order to take nto account te retalng supply of bot ouseold-related and destnaton zones, we ntroduce two oter varables, one for eac zone. Te model can ten be formulated as follows: N were = A A α1 α2 = k T c α1 α2 1 c k MODL CALIBRATION Te proposed framework s bult and calbrated wt te avalable standard data of te urban area of Lyon, wc as about nabtants and ouseolds. Te man data sources are an extract of te regster fle of companes (SIRN fle) of te cosen area, te correspondng census database (INS fle), and te 2006 personal trp survey, wc follows a Frenc standard (CRTU, 2008). In ts survey, te urban communty of 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 7

8 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles Lyon s dvded nto several small zones, grouped nto macro zones. However, t s mportant tat te metodology wll be able to be appled to a cosen urban area or generalsed nto a model able to be used on any urban area wtout te need of a calbraton and a testng pase. Soppng trp beavours are eterogeneous and dffcult to caracterse usng mcrosmulaton approaces, as te avalable data from surveys as not been collected for soppng trp caractersaton and te g cost of tese surveys made tem adaptable to develop macro-smulaton models, but not to mcro-smulaton ones. alss , verson 1-21 Apr 2012 Fgure 2: Maps of te Lyon urban area and te consdered zones (source: L. Bouzouna, LT) Te model parameters are calbrated separately for eac pase, followng a mxed approac tat combnes an econometrc analyss (usng lnear regresson tecnques) and te fttng metod proposed by Hyman (1969). In te followng subsectons we descrbe te calbraton metods and te man results of te calbraton analyss. Attracton pase Te attracton pase uses a multlnear functon. To calbrate t, we defne eac coeffcent by lnear regresson. We ave mplemented te model usng R Commander (Fox, 2005), combnng te dfferent varables n order to obtan te most precse model. Te model can be formulated as follows: 1C Model: T = a + a POP + a. NrSMC + a. mp BS + a. mp VBS a. MR C Model: CUA CUA T = a + a T NP = a CUA CUA CUA CUA CUA 0 1. POP + a2. NrSMC + a3. mp BS + a4. mp VBS + a5. MR NP NP NP NP NP NP NP 0 + a1. POP + a2. NrSMC + a3. mp BS + a4. mp VBS + a5. MR + a5. CC 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 8

9 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles T FP = a FP FP FP FP FP FP 0 + a1. POP + a2. NrSMC + a3. mp BS + a4. mp VBS + a5. CC We observe tat te R² s close to 1 n all tree categores of urban space. Moreover, te F test s close to 0 n two of te cases (see rreur! Source du renvo ntrouvable.). However, te one-category model s more robust, due to te gest number of ndvduals consdered n te lnear regresson procedure. Table 1 Lnear regresson results for te best attracton analyss Model R² F Test Indvduals 1C 0,82 1, C - CUA 0,86 0,31 8 3C - NP 0,77 0, C - FP 0,95 0, alss , verson 1-21 Apr 2012 Catcment area pase Te catcment area pase presents varous parameters tat must be calbrated. We propose a calbraton metod tat combnes an econometrc analyss (Kubs and Hartman, 2007) wt a fttng procedure based on tat of Hyman (1969). Ts calbraton metod s presented as a sequental procedure tat can be resumed as follows: 1. Determnaton of te coeffcents for bot small and bg stores by lnear regresson. 2. Revson of te model and fnal formulaton 3. Determnaton of te coeffcents to calbrate usng te fttng procedure and nzatlaton of te procedure 4. Iteratons for coeffcent mprovement usng te fttng metod Te lnear regresson analyss starts from te followng formulaton for te catcment area model: N = A α1 α2 γ T c We approxmate n a frst tme by excludng n te analyss te A elements. Ten, we can lnearze te functon usng te logartmc operator: log P( T ) = log K + α log 2 log α γ log c Fnally, usng R Commander (Fox, 2005) we establs te form of te model as well as te varables and coeffcents tat are ncluded n te fnal formulaton. We found tat te two best approxmatons are te followng: 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 9

10 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles and P small ( ; T ) = α α 1S 2 S K small γ S c S P bg ( ; T ) = α α 1b 2b K bg γ b c b Table 2 Lnear regresson results for te best catcment area model Model R² F Test Indvduals Small retalng actvtes 0,47 1, Bg stores 0,49 3, alss , verson 1-21 Apr 2012 Te commercal supply for small stores n te resdence zone can be supposed as a neglgble varable, as n te dfferent combnatons tat ave been tested te coeffcents were close to 0 and less nfluent tan te commercal supply for small stores n te soppng zone. Te number of ouseolds as an nverse effect as tat we can ntutvely ypotesse, but ts can be seen n te followng way: te nfluence of ouseold s pondered by te dstance. In fact, people trends are to go soppng close to te ouseold zone wen te commercal supply satsfes ter needs, and te ntra-zone dstance s n general smaller tan te nter-zone dstance. Ten, te fnal model can be formulated. Takng nto account te smlarty of te dstance coeffcent, we propose te followng catcment area model: T = A small α1s α2s K small γs + A bg α1b α2b K bg γb c wt and A A small bg = = k k α1s α2s α1b k α2b K small K small γs γb c c k k Te trd step s te fttng procedure. We start te procedure by ntalsng te algortm. To do ts, we frst select te parameters tat wll be recalbrated. Alpa and gamma belong to te A elements, and tey sould be calbrated usng oter approaces, suc as tat 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 10

11 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles proposed by Ortuzar and Wllumsen (2001), but tey need complete O/D matrces were te total number of emsson trps and te total number of attracton trps are te same quantty, wc s not our case. For ts reason, we keep te coeffcents estmated by lnear regresson and make te calbraton only on beta. We defne te followng mean values of c (Hyman, 1969): R N c c* = R N and c R were N s te number of measured trps arrvng at wc ouseolds are located n. small bg We also set = average, ). 0 ( 0 0 m = T m T c m alss , verson 1-21 Apr 2012 Te fttng procedure can te be defned, and s composed by several stages tat are defned as follows: 0. Intalsaton: Startng on 0, a a better value of s estmated n te followng way: 1. Iteraton pase: 0c m = c * 0 a. We make m=m+1. Usng te latest value for, noted m-1, we calculate te catcment area matrx usng te model (3) and obtan te new mean modelled trp dstance cost cm-1 usng te followng expresson: c( ) = N N ( ) c ( ) m 1 R We calculate δ ( N, N ). If ts value s suffcently close to accept te precson of te model, te teratons are stopped, oterwse we go to step b. b. We mprove n te followng way: m+ 1 ( c * c = m 1 ) m ( c * c c c m m 1 m ) m 1 Te teraton pase s repeated untl te precson of te model s accepted as good,.e. wen cm s supposed to be suffcently close to c*. Moreover, f te parameter s not enoug mproved after I teratons, we also stop te calbraton process. We report n Table 3 te man calbraton results of te model. 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 11

12 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles Table 3 stmaton results on te data of Lyon (2006) Surveyed stmated rror Number of trps per day ,7% Dstance (Mllons of km/day) 4,34 4,44 2,1% Mean dstance c (n km) 14,39 14,99 4,2% Intalzaton value 0-0,89 - Best value - 0,91 - After 50 teratons, we obtan an error on te mean dstance lower tan 5%. We observe tat n a global perspectve, te results are qute satsfactory. Te dstrbuton results are less performng but te estmaton of te total dstance due to prvate cars soppng trps can be estmated wt a small error. APPLICATION DOMAIN AN USAG XAMPLS alss , verson 1-21 Apr 2012 Possble uses of ts can be related to bot passenger and fregt transportaton, and also to marketng and real estate felds. For a publc polcy usage, dfferent coces and applcatons can be proposed. For example, te proposed model can be used to smulate strategc plannng scenaros to estmate te mpact of urbanstcs, legslaton and oter polcy actons n bot people and te global fregt movement trends. Anoter use can be to study te lmts of several polces (Gonzalez-Felu et al., 2010). It s mportant to note tat te new dstrbuton servces (ome delvery, proxmty recepton ponts, etc.) ave a drect mpact on soppng trps. In order to smulate te substtuton of current practces by tese new dstrbuton servces, te soppng trps by prvate car ave to be caractersed n order to defne and smulate te substtuton trends (from current soppng trps to te new dstrbuton routes). Router et al. (2009) propose four extreme urban retalng supply scenaros: 1. concentrated soppng malls; 2. tradtonal medum level stores; 3. e-commerce plus ome delvery; 4. e-commerce plus local pck up ponts. To smulate, te nter-establsment movements (IM) ave been estmated usng te LUTI model Freturb (Router and Toler, 2007) and te end-consumer movements (CM) usng te proposed soppng trp model. Te substtuton of soppng trps for te telesoppng scenaros s made usng te procedure proposed by Router et al. (2009). Moreover, te CO2 emssons can also be estmated from te travelled dstances and an estmaton of te mean speed n eac secton usng te commercal software Impact Ademe (ADM, 2003). 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 12

13 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles Km.PCU IM Km.PCU CM Km.PCU TOTAL CO2 equvalent Reference 2,7 Mkm/week 26 Mkm/week 28,7 Mkm/week 6150 t/week All Hypermarkets -87% -3% -11% -16% All proxmty stores +90% -87% -63% -56% All ome delveres -87% -85% -85% -86% All pckup ponts -87% -95% -93% -92% alss , verson 1-21 Apr 2012 Te all ypermarkets stuaton s a lttle better tan te reference, n terms of fregt flows for bot types of movements. Te establsment supply ones are drastcally reduced because te logstcs systems of te commercal centres allow a better ratonalsaton of fregt dstrbuton flows (about - 87% of te km.pcu respect to te reference stuaton). However, te end-consumer movements do not ncrease wt respect to te reference(we observe a lgt decrease, near -3%), due to te need of usng te prvate car to arrve to te perperal commercal centres tat a pror ncreases te lengt of prvate car soppng trps and te concentraton of te actvtes n ts form of commercal supply tat reduces te number of trps. Te global result s a reducton of a lttle more tan 15% of te total km.pcu, wc sows te mportance of end-consumer movements n te overall logstcs system of a urban communty. Te all proxmty stores scenaro as a ger mpact. Altoug te establsment supply movements ncrease (+90% of te km.pcu), te end-consumer movements decrease because of te proxmty tat ncentves te on foot soppng trps (about -87%). Ts s traduced by a reducton of te total km.pcu of 56% respect to te reference. Te all ome delvery servces scenaro as an mpact on te end-consumer movements: altoug no soppng trps are observed, consderng everybody s delvered ome, te ome-delvery trps (anoter component of end-consumer movements) present an mportant ncreasng trend, wc s traduced by a reducton of -85% n te end-consumer km.pcu. Te dstrbuton to te logstc platforms for ome delvery as been consdered to be made n a smlar way respect to te ypermarket dstrbuton, n a very ratonalsed way (a collaboratve stuaton s smulated), so te establsment supply movements are close to tose obtaned for te frst scenaro. Te same ypotess are made to te fourt scenaro, but te recepton ponts system as a better optmsed dstrbuton to te recepton ponts. Moreover, we obtan a small number of prvate car soppng trps. Ts s derved from te current proxmty soppng beavour, consderng tan several ouseolds wll use te car to take bg quanttes of goods from te recepton ponts. Te end-consumer movements n ts stuaton ave a smlar trend respect to te all small stores one (-93%). Tese trps contan bot te soppng and te recepton ponts delvery movements. Te overall fregt flow trends for tese two scenaros are respectvely -86% and -92% respect to te reference stuaton. We see tan te ome delvery servces are not te best soluton, snce te transportaton system tat replaces soppng trps s very rgd and ts effcency mprovement s dffcult (Allger, 2007). Te all recepton ponts scenaro seems to ave a more postve mpact, because a bg reducton of te travelled dstances can be made for bot types of movements, wc s not te case n te frst famly of scenaros. A combnaton of tese four can be used by stakeolders to defne urban commercal supply polces tat ntegrate bot types of movements n a cty logstcs pont of vew. 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 13

14 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles Te catcment area approac can also be useful to smulate e-commerce polces and beavoural ssues related to te new dstrbuton forms. Altoug specfc data about e- commerce beavour n te zone of smulaton s dffcult to found, te fndngs of Rom and Swamnatan (2004) can be appled to medum and bg urban areas. Tese autors defne 4 types of onlne soppers on te bass of a specfc survey for e-commerce beavour on standard categores of populaton. We can state tan altoug te store-orented soppers are only 15% of te sample, a catcment area approac ncludng varables suc as te commercal supply and te geograpcal dstance (among oters) represent te veavour of 3 categores,.e. 89% of te populaton. For ts reason, assumng te usage rates of e- commerce, te proposed model can dstrbute te related trps n order to reconsttute te delvery routes for several e-logstcs scemas (Router et al., 2009; Durand et al., 2010). alss , verson 1-21 Apr 2012 We propose several results for studyng te effects of e-commerce ntroducton. Te usagerate of tese servces s ncreased from 0% (assumng tan n 2006 te e-commerce usage rates are neglgble) to 100% wt an ncrease of 10% for eac scenaro. Te attractveness model s used to estmate te potental demand of e-commerce dstrbuton servces for eac zone and te remanng classcal soppng trps. Followng te smulaton metodology of Durand et al. (2010) for route smulaton, we can estmate te travelled dstances of tese forms of dstrbuton. Ten, te total dstances (e-commerce dstrbuton and classcal soppng trps). Te smulated data are te followng: Table 4 Smulaton results for progressve ncrease of e-commerce user rates Scénaro Aval Lvrason Aval Acats Total LAD 10% 4% -10% -6% 20% 8% -20% -12% 30% 13% -31% -18% 40% 17% -42% -25% 50% 21% -52% -31% 60% 25% -63% -38% 70% 29% -74% -44% 80% 33% -84% -51% 90% 38% -95% -57% 100% 40% -100% -60% PR 10% 2% -8% -6% 20% 4% -17% -13% 30% 6% -26% -20% 40% 7% -34% -27% 50% 9% -43% -34% 60% 11% -52% -41% 70% 13% -61% -48% 80% 15% -70% -55% 90% 17% -79% -62% 100% 18% -82% -65% 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 14

15 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles We observe tat e-commerce competton scenaros are less performant tat ter analogous collaboratve scenaros. Indeed, wt an all ome delvery scenaro te total reducton of te number of km. PCU s near -60% wereas n a collaboratve stuaton t s about 85%. Te all recepton ponts scenaro presents smlar results (-65% wt respect to -92% for te collaboratve scenaro). We also observe tan n a stuaton wtout mportant beavoural canges, a 50% reducton of te total number of Km.PCU s reaced only wt a 80% of usage rates for e-commerce. CONCLUSION alss , verson 1-21 Apr 2012 In ts paper we presented a new modellng approac for soppng trp smulaton, n order to reproduce an estmaton of end-consumer goods movements wc would be able to be compared to fregt dstrbuton trps. Consderng only te prvate car soppng trps, wc are tose tat nteract wt fregt dstrbuton trps, we propose a two-step model. Frst, te attracted soppng trps are generated at eac soppng destnaton secton. Second, eac soppng destnaton secton s related to te possble ouseold locatons of te soppers usng a gravty catcment area model. Anoter mportant element of soppng trps n urban areas s tat eac zone does not ave te same caracterstcs. Followng te admnstratve typology of rngs to classfy te urban communty s towns, we dvded te consdered zones nto tree categores for trp generaton: te Central Urban Area (CUA), wc s te man cty s area, te Near Perpery (NP), wc represents te frst rng, and te Far Perpery (FP), wc contans all te oter towns of te urban communty. Smlarly, we also propose a two-category model for te catcment area pase, n order to take nto account te dfferences between soppng trps made for purcasng at small retalers and tose related to bg stores and ypermarkets. Te soppng trps at eac retalng zone s well estmated, and smlar results are obtaned for te outbound trps. However, te nbound trps are less well estmated due to te nfluence of workng actvtes (no nformaton about te overall number of employees for all te economcal actvtes s ncluded). Moreover, two examples of modellng applcatons made wt te proposed model are presented. In order to mprove te model, and to defne an approac adaptable to eac cty, we ave appled ts metodology to a data fle contanng zones of two urban areas of dfferent sze. Te man results are encouragng (smlar regresson coeffcents and good approxmatons n bot urban areas), and a deeper study s n process. Moreover, a better caractersaton of te Commercal Centres, as for example defnng tem by ter surface or ter total revenue, ave to be studed. 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 15

16 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles ACKNOWLDGMNTS Part of te proposed metodology was developed n te context of THL II proect, part te GICC, a researc program on clmatc cange fnanced by te Frenc Agency of te nvronment and te nergy (ADM). Te autors sould also lke to acknowledge Florence Toler, studes engneer at NTP, Vaulx-en-Veln (France) for er elp n te data collecton and te estmaton of fregt dstrbuton trps usng FRTURB, as well as Frédérc Henrot, PD. student at Unversté de Lyon 2 for s elp and excanges n te consttuton of scenaros. RFRNCS alss , verson 1-21 Apr 2012 [1] ADM (2003), Logcel IMPACT-ADM verson Lvret de présentaton, ADM, Pars, France; [2] Arentze T. A., Oppewal H. and Tmmermans H. J. P., A multpurpose soppng trp model to assess retal agglomeratons effects, Journal of Marketng Researc n 42 (2005) [3] Badoe, D. A. and Steuart G. N. (1997) Urban and Travel Canges n te Greater Toronto Area and te Transferablty of Trp-generaton Models, Transportaton Plannng and Tecnology, vol. 20, pp [4] Black C., Broadstock D. C., Collns A. and Hunt L. C., Te derved demand for traffc at food superstores n te UK: A sem-parametrc regresson approac, Internatonal Journal of Transport conomcs, Vol. 34 Issue 2, pp. [5] CRTU (2008) L enquête ménages déplacements standard CRTU : Gude métodologque. MDAT, Lyon, France. [6] Cambre de Commerce et d Industre de Lyon (2008), 8ème nquête sur les comportements d acat des ménages en régon lyonnase. Dosser de presse et premers résultats, CCI, Lyon, France, ttp:// cc.fr. [7] Cubukcu, K. M. (2001) Factors Affectng Soppng Trp Generaton Rates n Metropoltan Areas, Studes n Regonal and Urban Plannng, vol. 9, pp [8] Dablanc, L., Peceur, P. (2000), Transport de marcandses en vlle: connaître et agr sur les déplacements d acats, Lettre de Commande n. 99MT08 TMV. [9] Durand, B., Gonzalez-Felu, J., Henrot, F. (2010) Pas de developpement durable du B to C sans vertable e-logstque urbane..., Proceedngs of te 8 t RIRL, Internatonal Meetng of Logstcs Researc, Bordeaux, 29 t, 30 t September and 1 st October [10] Fox, J. (2005) Te R Commander: A Basc-Statstcs Grapcal User Interface to R, Journal of Statstcal Software, vol 14 n 9, ttp:// [11] Gonzalez-Felu, J., Toler, F., Router, J.L. (2010) nd consumer movement generaton n Frenc medum urban areas, Proceda Socal and Beavoral Scences, lsever, to appear. 12 t WCTR, July 11-15, 2010 Lsbon, Portugal 16

17 An attractveness-based model for soppng trps n urban areas GONZALZ-FLIU, Jesus; ROUTHIR, Jean-Lous; RAUX, Carles alss , verson 1-21 Apr 2012 [12] Hyman, G. M. (1969), Te calbraton of trp dstrbuton models, nvronment and Plannng Vol. 1, pp [13] Kubs, A., Hartman, M. (2007) Analyss of Locaton of Large-area Soppng Centres. A Probablstc Gravty Model for te Halle-Lepzg Area, Jarbuc für Regonalwssenscaft, vol 27, pp [14] Long-Lee M., Pace K. (2005) Spatal Dstrbuton of Retal Sales, Te Journal of Real state Fnance and conomcs, vol. 31 n 1, pp [15] Ortuzar, J. de D. and Wllumsen, L. G. (2001), Modellng Transport, Jon Wley and Sons, New York, USA. [16] Rom A. J. and Swamnatan (2004) A typology of onlne soppers based on soppng motvatons, Journal of Busness Researc vol. 57, pp [17] Router, J. L, Ségalou,. and Durand, S. (2001) Mesurer l mpact du transport de marcandses en vlle - le modèle Freturb (verson 1), Programme natonal marcandses en vlle DRAST-ADM, 104 p. [18] Router, J.L., Trasnel, J.P., Gonzalez-Felu, J., Henrot, F. and Raux, C. (2009) THL II Rapport fnal. Conventon ADM. [19] Russo F. and Com A. (2006) Demand models for cty logstcs : a state of te art and a proposed ntegrated system. In. Tanguc and Tomson, Recent Advances n Cty Logstcs, lsever, Amsterdam, pp [20] Seégalou,. (1999b), Preparaton du blan envronnemental du TMV sur la vlle de Bordeaux. Note relatve au calcul de la dstrbuton des flux de vécules pour motf acat zone à zone. Report LT-ADM-DTT, Lyon, France. [21] Ségalou,., Ambrosn, C. and Router J. L. (2004) Te envronmental assessment of urban goods movement. In. Tanguc and R.G. Tomson, Logstcs Systems for Sustanable Ctes, lsever, Amsterdam, pp [22] Smma A., Cattaneo P., Baumeler M., Axausen K. (2004) Factors nfluencng te ndvdual soppng beavour: Te case of Swtzerland, Arbetsbercte Verkers- und Raumplanung, 247, Insttut für Verkersplanung und Transportsysteme (IVT), TH Zürc, Zürc. [23] Toler F., Allger L., Pater D. and Router J. L. (2005) Vers un modèle global de smulaton de la logstque urbane : FRTURB v2, Rapport DRAST n 05-S-03. [24] Vckerman R. W. and Barmby T. A. (1984) Te Structure of Soppng Travel: Some Developments of te Trp Generaton Model, Journal of Transport conomcs and Polcy, vol. 18, n. 2, pp [25] Vckerman R. W. and Barmby T. A. (1985) Houseold Trp Generaton Coce: Alternatve mprcal Approaces, Transportaton Researc B, vol. 19, n. 6, pp t WCTR, July 11-15, 2010 Lsbon, Portugal 17