D5.4 Enhanced Version of the SIMULACRA Model

Transcription

1 D5.2 Enhanced Verson of the SIMULACRA Model Proect acronym: Proect ttle: INSIGHT Grant Agreement number: Fundng Scheme: Proect start date / Duraton: Call Topc: Proect webste: Delverable: Innovatve Polcy Modellng and Governance Tools for Sustanable Post-Crss Urban Development Collaboratve proect 01 Oct 2013 / 36 months FP7.ICT Date 31/07/2016 Status Dssemnaton Level: D5.4 Enhanced Verson of the SIMULACRA Model Approved Consortum

2 Authorng and Approval Prepared by Name & Afflaton Poston Date Mchael Batty (CASA-UCL Professor 17/06/2016 Ducco Povan (CASA-UCL Researcher 17/06/2016 Vassls Zacharads (CASA-UCL Researcher 17/06/2016 Revewed by Name & Afflaton Poston Date Harry Tmmermans (TU/e WP5 Leader 31/07/2016 Approved for submsson to the European Commsson by Name & Afflaton Poston Date Rcardo Herranz (Nommon Scentfc/Techncal Coordnator 31/07/2016 Irs Galloso (UPM Management Coordnator 31/07/2016 INSIGHT Consortum Page

3 Record of Revsons Edton Date Descrpton/Changes Draft 1 17/06/2016 Intal verson 31/06/2016 Mnor edtoral correctons INSIGHT Consortum Page

4 Table of Contents EXECUTIVE SUMMARY INTRODUCTION TO AGGREGATE LUTI MODELS A Short Hstory and Defntons Orgns of SIMULACRA AN OUTLINE OF SIMULACRA Coupled Sectors The Formal Model Structure The Resdental Locaton Model APPLICATIONS TO THE LONDON REGION Data and Zonng Model Inputs and the London Regon ENHANCING THE RETAIL SECTOR A New Retal Model Embeddng the Retal Model n SIMULACRA EXTENDING SIMULACRA INTO QUANT Movng the Model to the Web from the Destop The Transport Networs Demonstratons n the London Regon CONCLUSIONS AND FUTURE RESEARCH REFERENCES INSIGHT Consortum Page

5 Executve Summary SIMULACRA s a spatally extensve aggregate land use transportaton nteracton (LUTI model bult for Greater London and ts outer metropoltan regon whch comprses some 13 mllon people and 6 mllon obs. The model s based on couplng together two sectors employment and populaton through the ourney to wor and through demands for commercal and other employment. In ths sense, the model deals wth spatal nteractons between the producton and consumpton sectors that are essentally artculated as ourneys to wor and to shop. We frst outlne the hstory of the model and descrbe ts structure. We then llustrate how t has been appled to London n terms of ts calbraton at a cross-secton n tme (2011. We then enhance the model by mportng the new retal locaton model nto SIMULACRA: ths s a nested logt model n whch agglomeraton economes n the retal sector are explctly represented, descrbed n delverable D4.3. We then show how the overall SIMULACRA model s beng extended to a web-based nterface to an urban model of England Wales whch can be accessed for any cty remotely by any user. Ths model s called QUANT and we llustrate ts data and how t mght be used n testng scenaros, showng examples from the London regon but notng that the new comprehensve model can be appled to any area of England and Wales. INSIGHT Consortum Page 4

6 1. Introducton to Aggregate LUTI Models 1.1 A Short Hstory and Defntons Urban models have become a label for any nds of abstracton from the cty that are wder than dgtal representatons and smulatons per se but embody some formal attrbutes that are computable. They now nclude symbolc and conc models wth the former havng predctve power. Such predctve models form the class of models that we are developng here n INSIGHT, these models beng mathematcal n structure, and relatng to data that pertans to the doman beng modelled, beng tuned or calbrated to reproduce an exstng state of the cty at a cross-secton n tme or over a prevous tme perod. These models were frst constructed n the 1950s-1960s n the Unted States and were largely focussed on predctng aggregate populatons and ther land uses, actvtes and traffc flows but as they have been developed, more dsaggregate or mcro models have been emerged. Generally, at least four types now exst: tradtonal aggregatve models sometmes called Land Use Transport Interacton (LUTI models, mcrosmulaton varants whch dsaggregate the populaton to ndvduals and often are artculated as agent-based models, systems dynamcs models where tme s a more explct organsng concept, and cellular automata models that develop through tme ctes at the level of ther physcal actvtes such as land uses. Many varants exst whch are mergers of these four types and rarely do pure versons of each exst. In INSIGHT, we are developng three of these varants. MATSm s a mcrosmulaton agent-based model of traffc flow and household decson-mang, Albatross s another verson of an agent-based model focusng on how household behavour and locaton relate to transport, MARS s a systems dynamcs model but wth lns to more aggregate representatons, and SIMULACRA s an aggregate model of the LUTI varety focussed on a rather coarse aggregaton of land uses and populatons whch are consstently lned through transportaton flows between them. The model we are enhancng n ths secton SIMULACRA s spatally extensve and focussed on dealng wth aggregate changes n the transport networ and n the locaton of employment and populaton. Our enhancement as we recount below s focussed on addng a new retal model to ts structure and n extendng the model nto a web-based forum for ts access and mplementaton. Thus the models that we are comparng and contrastng n INSIGHT more or less cover the range of modellng types that currently defne the feld. 1.2 Orgns of SIMULACRA SIMULACRA stands for the SIMulaton of Urban Land And Commercal-Resdental Actvtes whch defne the two ey actvty sectors that are smulated and ther relatonshp to the land uses that they requre. The model sttches the two sectors together by smulatng flows between locatons defnng these actvtes n the tradton that was frst begun by Lowry (1964 and whch has been extended n much greater detal n a seres of models such as those developed by Putnam, Echenque, Smmons, de la Barra, Wegener, and Yng amongst others. These models represent the cty at a cross-secton n tme whch s referred to as the calbraton baselne although there are some varants that model the ncrement or decrement n actvtes between coarse tme perods and these are sometmes referred to as quas- or pseudo-dynamc. SIMULACRA does not attempt to model these coarse graned dynamcs and thus the model s very defntely n the tradton of comparatve statc predcton. In short, when we use ths model for predcton, we assume that a new equlbrum s predcted each tme changes are made to the model s nputs and that ths equlbrum needs to be nterpreted n terms of the type of scenaro that s INSIGHT Consortum Page 5

7 defned. For example, the model can be used to predct a short term change n the number of obs n a place for example the closure of a factory or the addton of a new retal par and t would thus be assumed that predctons would mrror what would happen over a matter of months or possbly a couple of years. In contrast where new obs were to be located at the scale of a new arport, ths would be assumed to wor tself out over many years, whle the mplementaton of such a scenaro would be staged over ths longer tme perod. The London verson of SIMULACRA began lfe as part of a proect to model the mpact of clmate change n the form of sea level rse n the Thames Gateway. The model was constructed and operated as only one sector a resdental locaton model whch was lned to an nput-output employment model n loosely-coupled form. It was developed for a spatal system composed of the 633 electoral wards whch are nested nto the 33 boroughs that comprse the Greater London Authorty (GLA. Ths area contans currently ust over 8 mllon persons but t does not contan the outlyng suburbs or the small towns that have grown nto the overall metropoltan area. Thus the SIMULACRA model was bult on the expanded area of some 1767 wards that cover the GLA and what s referred to as the Outer Metropoltan Area (Batty et al., The populaton of ths expanded area s almost double the GLA area at 14 mllon. One of the crtcal problems n any urban model desgn s ths problem of closure between the cty and ts wder envronment. Ths s one of the reasons why we always need to embed our SIMULACRA model n the wder regon of South East England and East Angla. Pctures of the zonng systems for the SIMULACRA model as t has evolved are shown below n Fgure 1. Fgure 1: Zonng Systems for the Non-Enhanced SIMULACRA model wth the Central Map the Current System INSIGHT Consortum Page 6

8 2. An Outlne of SIMULACRA 2.1 Coupled Sectors The urban economy s largely conceved n terms of populatons whch operate as households and consume varous goods however lberally defned and employment whch encapsulates the role of the populaton n producng goods that n turn they consume. Ths very defnton reflects a crcularty between producton and consumpton lyng at the bass of many models of the urban economy, partcularly LUTI models. In fact, vrtually every attrbute of actvtes and land uses contaned wthn an urban model can be seen through the lens of consumpton and producton that we also label demography (populaton and economy (employment. Elaboratons of these dstnctons through dsaggregaton tend to reflect dfferent types or sectors dfferent sectors of producton and dfferent types of consumpton but nothng s ntrnscally added when such dsaggregaton taes place. Some of the lnages between consumpton and producton may change but n general the structural logc of ths couplng s the same. Although ths report on the enhancement to SIMULACRA s not meant to be a formal specfcaton of the model, we wll ntroduce some notatonal defntons to frm up deas and the bloc dagrams shown n Fgure 2 ndcate how we can couple these two sectors together. We must mae one addtonal defnton and that s between zones or places where actvtes populatons and/or employments locate and where they travel to. These former are called orgns and the latter destnatons (although ths usage s arbtrary and we wll use the subscrpt to refer to orgns and to refer to destnatons. Now the couplng between sectors s n terms of where the employment travels from places of wor or producton to where they lve and ths s measured by trps between orgns and destnatons. Populatons of course consume and usually travel to effect ths or to purchase goods (that they consume elsewhere and ths s the demand for these goods from zones where people lve to where they consume or purchases goods n. Ths defnes the flows T S and the nterlocng flow system reflectng trps and demands completes the crcle from producton to consumpton and bac. Our last defnton of employment at, scale to actvtes. E, and resdental populaton at, P, defnes the way nteractons or trps In an analogous way, we can see ths couplng between the economy and demography n terms of money flows rather than trps whch are materal flows. Wages w are earned n orgns of producton and then spent partly on travellng to homes and buyng housng n destnatons. Ths process s dctated by the costs of travel whch are related to dstances d whle housng s prced as p n the destnatons. The remanng mones are then spent on consumpton v bac n the orgns (whch are commercal and shoppng centres and travel costs related to dstance d are ncurred agan to enable ths. Durng ths process, these money flows have an exact parallel couplng to materal flows and we show both of them n the bloc dagrams n Fgures 2a and 2b respectvely. Note the term (t defnes the poston on the cycle lnng economy to demography, producton to consumpton and orgns to destnatons and bac agan. We wll leave ths temporal ndex mplct n that t smply shows that INSIGHT Consortum Page 7

9 n some mplementatons of ths crcular couplng, an equlbrum s requred. In some versons, the equlbrum or stoppng pont wll depend on startng values, n others they may be ndependent of where the process begns. a b Fgure 2: Couplng the Economy and Demography a left: materal actvty flows b rght: money flows 2.2 The Formal Model Structure Our model wll be dsaggregated n varous ways but we can wrte t as two two-part model equatons at the aggregate level whch mples some degree of smultanety n that economy depends on demography and vce versa. From a dstrbuton of employment E (t at teraton t, the model generates wor trps (t populaton P (t from the frst gravtatonal equaton T, then T ( t E ( t L exp( d L exp( d P ( t T ( t. (1 Consumer trps S (t and the actvty assocated wth these demands are then computed from whch we can scale bac to the orgnal employment used n producton as E ( t 1 as S ( t P ( t F exp( d F exp( d E ( t S ( t. (2 Note that L s the attracton of the zones for resdental housng whle F s the attracton of the commercal zone for consumpton or shoppng purchases. In our SIMULACRA models, there are dfferent varants of ths couplng where we brea nto the cycle wth some value of employment that may not be total employment (t mght be servce employment, say and then terate ths scheme to generate dfferent measures of employment assocated wth consumpton. In essence balance however must be acheved between nputs and outputs INSIGHT Consortum Page 8

10 2.3 The Resdental Locaton Model To llustrate how the alternatve money flow couplng s embedded wthn the general model, frst an analogous equaton exsts for the ourney from wor to home where wages w are sent to resdental zones n proporton to trps T. We can wrte ths as 2 L exp{ ( w c p } W( t w ( t 2 L exp{ ( w c p } Y t { W ( t c p } (3 ( We note that the attracton of the zone depends both on the land avalable but also upon the relatonshp between what the trp maer has to spend on housng less the transport cost c of travellng to that place and the actual prce of housng p. We can also use these quanttes to compute the dsposable ncome Y (t after these costs have been met and ths varable s used to drve the spendng on consumpton whch s then modelled usng a second gravtatonal equaton whch we defne as Y ( t Y ( t A exp( d A exp( d V t ( Y ( t c. (4 ( Ths couplng requres a mechansm that relates the sales or purchases for consumpton n whch s V (t to the wages that are used to produce the goods requred for consumpton. The smplest way s to assume some sort of equalty, that s w ( t v ( t. Ths s pretty unrealstc n fact because t assumes that where ever you produce a good for consumpton, then t s consumed there and ths s where such a model would requre an embeddng nto an approprate nput-output structure whch move goods all over the economy n sectoral as well as spatal terms. Nevertheless, ths s a smple yet approprate extenson of the model and t does mean that the housng locatons are based on realstc consderatons of travel cost, wages and house prces. In fact, n the current SIMULACRA model, we use a mxed formulaton whch can be wrtten as 2 L exp{ ( w c p } T( t E ( t 2 L exp{ ( w c p } P ( t T ( t, (5 Y w t { T ( } t c p, (6 E ( Y ( t Y ( t A exp( d A exp( d V t ( Y ( t c. (7 ( We assume that mones expended equal wages and then we adust employment accordngly n the versons where we terate the cycle (t. INSIGHT Consortum Page 9

11 3. Applcatons to the London Regon 3.1 Data and Zonng In Fgure 1 above we noted how we defned the regon basng our current model on 1767 zones whch are electoral wards wth an average of 7600 persons per zone. The actvty rate s almost 2 (~1.96 meanng that for every person worng, there s another not worng. Ths s a lower rate of dependence than n many large ctes hstorcally and t compares to somethng n the order of 2.24 n the USA and 2.06 on average n Brtan. The average wor trp length s about 88 mnutes and shoppng trp length about 82 but ther values depend very strongly on the assumptons made wth respect to the networ measurements. The travel to wor area s defned as the area wthn whch 75% of those who wor n the area lve n t and 75% of those who lve n the area wor n t. The populaton n ths area s closer to that wthn the GLA boundary and s about 9.3 mllon. The data sources for the model are all n the publc doman; that s, they are open data wth the excepton of the networ data whch s held by the natonal mappng agency the Ordnance Survey. The model was orgnally bult usng populaton and travel to wor data from the 2001 Populaton Census whch was updated to 2011 when that census was released. The networ data from publc tmetables s currently beng used to produce networs for ral and bus usng GTFS technologes, Google's Transt Trp Planner, and related software for shortest routes algorthms. Employment data s acqured from the Natonal Offce of Manpower Informaton Systems (NOMIS whle varous other data comes from the Valuaton Offce Busness Data whch also holds rental values as specfed n D4.3. The Retal Locaton Model also draws on ths wder portfolo of open data. The transport flow data s supplemented by the London Travel Demand Survey whch s a detaled household survey but wth a very low percentage sample whch s avalable yearly snce In short ths has never been a problem n these types of model applcaton for they do not requre tme seres data one of the real bugbears n dynamc urban models nor do they requre ndvdual data and one of the reasons for ther popularty s ther modest data requrements. The model currently operatonal s also dsaggregated nto two modal networs, publc and prvate transport. These are operated for both the resdental locaton and commercal locaton models and they reflect the competton between these two modes for trps and flows whch are determned by the dfferent modal costs. The model s also sem-constraned to meet certan land denstes n the resdental sector: f the model produces extreme denstes that could not be accommodated n feasble physcal terms, these are moderated by the mposton of constrants as upper actvty lmts. Standard mechansms are used to ensure that these constrants are met through teraton of the model structure. 3.2 Model Inputs and the London Regon The model has been developed as a standalone plot for the destop and we show the nputs and outputs here but t s also developed as a web-based resource wth ths verson manly used to test scenaros. The destop verson s a sngle wndow vew of model nputs and outputs and the basc screen s shown below where the flow chart on the top left of the screen represents the sequence of stages through whch the user passes n explorng nputs and generatng outputs, commutng trps, dstngushng between recurrent and non-recurrent locatons. INSIGHT Consortum Page 10

12 a b The central wndow for Map Graphcs enables the user to dsplay hstograms or thematc maps of the data and to loo at dfferences between the data as well as counts and denstes. There are a varety of combnatons of data and model predctons that can be dsplayed n ths way. c Fgure 3: The Destop Verson of SIMULACRA a top left: The organsaton of the sngle wndow b top rght: Explanaton of the nputs and outputs c bottom: The London regon and the model s goodness of ft To llustrate the model, we wll show sx screens from the map graphcs n Fgure 4. We show the operaton of land use constrants n 4(a, then the hstograms of populaton denstes observed (b and predcted (c, thematc maps of observed (d and predcted retal employment counts (e and then the devatons between them (f. These maps provde the user wth an mmedate vew of how good the model s predctons are versus the observatons as well as the structure of the London regon. Ths s maredly polycentrc as London has grown absorbng ts outlyng towns and vllages. A central place system s llustrated by these map hstograms. The model s calbrated n the usual way by solvng the maxmum lelhood equatons assocated wth the gravtatonal models and a unque equlbrum s guaranteed wth respect to these model parameters. However, the way n whch the model couplng s terated and the startng values for the nput varables do determne the INSIGHT Consortum Page 11

13 equlbrum wth respect to the smultanety of the nputs and outputs at the observed and modelled crosssectonal state. However, ths does not mae much dfference to the predctons of the model for these appear strongly convergent. a b c d e f Fgure 4: Spatal Inputs and Outputs, Observatons and Predctons, Counts and Denstes (see text for explanatons INSIGHT Consortum Page 12

14 4. Enhancng the Retal Sector 4.1 A New Retal Model The model based on the second gravtatonal equaton (2 noted above s essentally a sngly constraned model n whch expendture on consumpton by a populaton located at zone, Y, where we wll drop the teraton ndex (t, s allocated as a money flow from resdental zone to employment/commercal centre zone. Ths model s the conventonal shoppng model form frst ntroduced n 1962 by Huff but goes bac to Relly n the 1920s. The model has been largely unmproved or extended snce then but n ths proect we have devsed a new model n whch the attracton of the retal/commercal centre s artculated explctly n terms of ts agglomeraton economes. In essence, the model ntroduced n D4.3 s a nested gravtatonal model where the gravtatonal pull of the retal centre wth respect to other retalers and the dstances between them s embedded or nested n the more general functon of the retal centres n the wder system of central places and the locaton of populaton n ther hnterlands. We wll restate the model here and then modfy ts form to ln t to enhanced verson of SIMULACRA n whch we have embedded t. The model that we ntroduced above whch smulates trps from the populaton to zones where consumpton s explctly organsed through actual partcpaton or through purchasng s stated as S ( t P ( t F exp( d F exp( d E ( t S ( t. [(2] The ey addton to ths model from D4.3 nvolves modellng the relatonshp between retalers located at zone and all other retalers wthn a gven dstance from each retaler n queston. At the level of spatal aggregaton assumed here, the average attracton of the zone to consumers gven by F can be modelled as a functon of all retalers r wthn the zone and the dstance from each of these retalers r to every other r wthn the dstance band defned. Then at the dsaggregate level, the agglomeraton economes assocated wth retaler r can be modelled as a potental functon r A F exp( d r r rr (8 where Fr s the floor space assocated wth the retaler at r, s a parameter of the sze whch reflects the scale, s a dstance parameter and dr r s the dstance from the retaler r n equaton (8 to another retaler r n the same agglomeraton. The problem wth ths new formulaton of attractveness for the aggregate model s that t reles on data at a fne spatal scale below that at whch the model operates and therefore the attracton value has to be summed over all retalers that are located n zone. Ths must be done off-lne that s outsde the model as the model does not represent ndvdual retalers at locatons r. Thus the attracton s set up for each zone as Ar F r exp( d r A r r r r. (9 INSIGHT Consortum Page 13

15 The new retal model can now be wrtten as S ( t P ( t A f (, d A f (, d, (10 where we now defne the dstance/cost functon f, d as exp{ ( d 1/ } notng that we use the Box- ( Cox transform to add another parameter to the dstance functon whch gves us much greater flexblty n calbratng the model. It s now clear that the model has four parameters the agglomeraton scalng, the nested logt weght on dstance, the frcton of dstance for the zonal system, and the Box-Cox transform parameter. 4.2 Embeddng the Retal Model n SIMULACRA As we construct ths model by smulatng shoppng trps from home to wor and from wor to wor, we effectvely have two aggregate trp models, each wth two parameters whle the nested retalng model wth two parameters apples to each type of trp. In all, there are sx parameters to ft for ths model but ths s a consequence of ntroducng ths level of detal nto the model. In fact, we wll elaborate ths pont about spatal detal n our concluson to the enhancement because t s ncreasngly clear that more than one level of scale or detal s requred for aggregate LUTI style models. Last but not least, when we embed the shoppng model nto SIMULACRA, we have to calbrate these sze parameters as well as the sngle deterrence parameter for the resdental locaton model. Currently we are only dong ths for the sngle road networ system and to llustrate the many optons that pertan to the enhanced model, we summarse the dfferent possbltes n Fgure 5 below. Fgure 5: Possble Model Versons of Enhanced SIMULACRA If you proceed from left to rght statng wth a model on the left, then tracng a lne through ths tree would provde you wth a partcular verson of the enhanced or orgnal SIMULACRA model. INSIGHT Consortum Page 14

16 5. Extendng SIMULACRA nto QUANT 5.1 Movng the Model to the Web from the Destop A related proect wth the Future Ctes Catapult ( nvolves movng the model from the destop to the web. The ntent here s to mae the model avalable from remote locatons and earler versons of SIMULACRA were n fact desgned to be run n a web-based context. However, the new proect s to extend the model spatally to ultmately deal wth all locatons, frst n England and Wales, then by ncludng Scotland, and possbly all Ireland. The model s beng developed at a smlar spatal scale to SIMULACRA at the level of what are now called Mddle layer Super Output Areas (MSOAs of whch there are 7201 n England and Wales. In Fgure 6, we show a map of these areas and some ey statstcs. MSOAs are dfferent from wards but they are at roughly the same level of spatal resoluton wth an average of 7800 persons per MSOA n England and Wales compared to an average of 6969 persons for each ward. The new model s called QUANT meanng Quanttatve Urban ANalytcs forecastng or some varaton thereof and the ey noton s that the model s beng bult for a wde area n the UK at the scale at whch setch plannng and large scale scenaros for land use, transport and actvty changes are relevant. The ey ssue s that f you are a local authorty planner or any nformed staeholder wth suffcent expertse n thnng about urban plannng problems, then you could use ths model as a tool to test the mpacts of scenaros for your own area or any area of the UK. The model s beng developed n stu and t s avalable n ts prelmnary state. Ultmately t wll contan all the functonalty of SIMULACRA and more but currently the only verson on the web whch s worng s for the resdental locaton model. The web ste s at Populaton: Employment: Number of MSOAs: 7201 Number of Wards: 8060 Mean Populaton MSOAs: 7800 Mean Employment MSOAs: 3003 Fgure 6; The Zonng System for the QUANT Model: SIMULACRA Everywhere INSIGHT Consortum Page 15

17 We wll not detal the techncal constructon of the model here but smply provde some ts output to ndcate the development so far. It s worth sayng however that that there are some maor computatonal challenges n runnng ths model. The lne between what s computed on the server or the clent sde s blurred. Currently most computaton taes place on the server sde because we never have any nowledge about how large the clent s. As we do not splt up the natonal system wth respect to any user each user has access to the entre system as above n Fgure 6 we need to compute changes when we generate new scenaros on the server sde that change the mappng requrements on the clent sde. Ths requres maps to be delvered contnually from server to clent and ths can be a bottlenec. The number of users too s an ssue. As more users ple nto runnng the model for ther own remote locaton, computatonal demands on the server can ncrease exponentally and thus currently we are settng the number of concurrent users to no more than 4. There are sgnfcant ssues nvolved n these questons that reman open at the present tme and wll contnue to provde lmts on what can be done for all users wth respect to ther access to tools such as SIMULACRA and the other models n INSIGHT. All need to be ported to a web-based envronment n tme and ths s the only way they wll become wdely avalable. 5.2 The Transport Networs The road networ s bult from the Mastermap Ordnance Survey data whch s based on street centre lnes whch are part of the Integrated Transport Networ ITN layer. Ths s a straghtforward product but to ths we are currently translatng the route segments nto generalsed travels costs as wll be the case wth all our dstance data. SIMULACRA used dstances n ts earler more robust versons but there are varants usng cost data and currently the networ s beng generalsed to travel costs. The publc transport networs ral and bus are based on tmetables data. Tmetables of trans and buses and the statons that they are routed through can be used to construct shortest routes for ral and bus networs. For ral, the algorthm relates walng dstances to statons so that most MSOAs can be connected. Walng to bus stops s more straghtforward so as to connect MSOAs to one another. Essentally these networs are at a much fner resoluton than the MSOAs themselves. The OS data s at the level of 10s of metres whle the MSOAs are measured n square lometres. All the processng of networs must thus be done outsde the model, off-lne so to spea, as the level of detal cannot be represented n the model. Ths s a problem because as networ data s a crucal nput to the model, then t s one of the ways n whch the model can be used to test scenaros. As the whole focus of the model s on developng and testng scenaros onlne, t would be desrable to manpulate networ data onlne to change travel tmes and costs but ths s smply not possble. In tme t may be possble to develop summary networs because the actual networ needed s smply between MSOAs but n mang changes to ths networ, ths needs to be done and the much fner scale of resoluton, whch s equvalent to changes such as addng new segments to the networ or alterng attrbutes of segments. Ths s a problem that has rarely f ever been breached n LUTI modellng t relates to the general queston of nterfacng LUTI models wth tradtonal transport models but t s crtcal to the problem of buldng new scenaros that requres changes to networs. It s an ongong research ssue. 5.3 Demonstratons n the London Regon To llustrate the potental of the QUANT model, we wll llustrate some data and outputs for the London regon. These wll cover an area much wder than the SIMULACRA model regon thus mplyng that the system s not closed when t comes to loong at data at the scale of large ctes n the UK. Of course the model stll does not INSIGHT Consortum Page 16

18 tae account of flows n and out of the England and Wales. In Fgure 7, we show populaton (a flow data for the ourney to wor as vectors for England and Wales (b, the flow data the London area (c, the Green Belts n the South East of England (d, populaton accessbltes (e and employment accessbltes (f. a b c d e f Fgure 7; Spatal Dstrbutons, Vector Flows, Land Constrants and Accessbltes from the Data Wthn QUANT When we move to model scenaros, QUANT enables us to nput new data employment, constrants on populaton or retal development whch can be n the form of green belts, land lmts and so on, and changes to the dfferent transport networs. To gve an example, we show the mpact on populaton totals and denstes n the London regon by addng 100,000 obs nto the Medway area whch s a possble ste for a new London arport. These are shown n Fgure 8 where (a show the populaton of the MSOAs n the Greater London area and where the locaton of Medway s hghlghted and where the user has added the ncrement of obs. The dfference between the orgnal populaton (before the obs are added s shown n Fgure 8(b. What ths shows s that populaton locates around the obs but that further dstances from the ste, populaton declne slghtly as ths populaton s pulled towards the new arport ste. Ths s n effect a new equlbrum where the effects of an addtonal change lead to populaton gans and losses n the regon around the mpact. Of course populaton s not actually lost; the mplcaton s that t smply relocates. INSIGHT Consortum Page 17

19 a b Fgure 8: A Typcal Scenaro n QUANT: Increasng Jobs n a Thames Estuary Arport Proposal INSIGHT Consortum Page 18

20 6. Conclusons and Future Research There are many unfnshed strands to the SIMULACRA-QUANT modellng effort that have been hghlghted durng our dscusson of how the SIMULACRA model s beng enhanced. The retal model whch s the man addton requres detaled data at the plot level for ts operaton and the standalone verson descrbed n D4.3 requres that level of detal to be ncluded n the model algorthm for calbraton. Ths s because the parameters governng agglomeraton depend on the confguraton of ndvdual retalers. When those parameters have been ftted t s possble to tae an aggregate verson and embed ths nto the comprehensve model but the comprehensve model needs to be calbrated n concert wth the retal model, movng bac and forth between the two. Ths nd of problem wll only be resolved when we move to a two or more level model n spatal terms where spatally aggregate and dsaggregate data of varous nds are ncluded n the nputs drvng the overall model. Ths s also needed for the transportaton networs as we outlned above. To an extent, ths depends on conceptual and theoretcal advances that show how networs and spatal dstrbutons at dfferent levels are consstent and can be derved from one another wth smplcty and ease. Ths s stretchng the state of the art. We are also addng more economc sectors to the SIMULACRA-QUANT framewor. We stll have two sectors and we need to dsaggregate the employment nto several as n nput-output models. Ths complcates the model enormously. It s not dffcult to do although there are data avalablty ssues but the tme taen to run such an enhanced model would ncrease explosvely and the problem needs to be managed n extendng the model further n ths drecton. Last but not least, we need detaled gudance to the model user n the constructon of scenaros. There are many such scenaros whch can be tested n models of ths nd now that they are avalable, fast and onlne. The space of scenaros s relevant n ths context and we need to provde some sense of how a user mght approach ths. Moreover, we need to extend the model to embrace many dfferent ndcators that are relevant to the scenaro evaluaton and n ths, we can draw on the experence of dfferent models beng developed n INSIGHT and related proects. INSIGHT Consortum Page 19

21 7. References Batty, M., Vargas, C., Smth, D., Serras, J., Reades, J., Johansson, A. (2013 SIMULACRA: fast land-use transportaton models for the rapd assessment of urban futures, Envronment and Plannng B: Plannng and Desgn, 40, Lowry, I. S. (1964 A Model of Metropols, RM-4035, The Rand Corporaton, Santa Monca CA. INSIGHT Consortum Page 20