Land use/transport interaction modelling: from theory to practice

The available land use and transport interaction models are monopolised by researchers for the purposes of experimentation and in-depth theoretical analysis. A multitude of difficulties arise when applying these models to practical cases of infrastructures.

What are the long-term effects of introducing a cordon toll in a dense zone on relocation of businesses and households? What is the impact of a large-scale public transport project (of the type of the nouveau Grand Paris or “new Greater Paris”) on the residential and economic attractiveness of the sectors served by the new stations and on the increase in property prices? What is the effect on residential attractiveness of transforming an expressway into an urban boulevard?
A conventional traffic model cannot answer such questions, whereas a LUTI model can give some answers. Such a model therefore constitutes a new tool that can enlighten public decision-taking and action by making it possible to explore a new field of questions hitherto impenetrable for conventional traffic models.

A LUTI model is a product resulting from a cross between transport economics and urban economics. Schematically, a distinction can be made between two main families of models: the static general equilibrium models and the dynamic models⁽¹⁾.

A static model represents a long-term equilibrium between the urban system and the transport system. The main representatives of this category are the following models: Mussa-Cube Land (created by Martinez in Santiago de Chile); Tranus (created by De la Barra in Venezuela); Relu-Tran (created by Anas in Chicago); and Pirandello (developed by Vinci, in France). Within the urban or land-use system, an economic balance is sought between supply of urban projects (housing, offices) and demand for household and business location. Seeking the setpoint leads to a spatial distribution of the agents wherein each agent no longer has anything to gain from relocating.

A dynamic model operates more realistically, by integrating the demography of the households and of the businesses (known as the “firmography”), without seeking to achieve an equilibrium and by time looping with the transport model. Professor Wegener (University of Dortmund, Germany) explains that the time dynamics differ from one process to the other (e.g. slow change in residential locations, and faster change in mobility behaviours when faced with congestion), and that the dynamic model makes it possible to take better account of such differing time dynamics. The interaction between the land-use model and the transport or traffic model takes place via accessibility.
However, a dynamic model such as UrbanSim requires an extremely large volume of geolocated data, in particular if the level of “disaggregation” of the agents (households, businesses, urban projects) is high, if the spatial scale is fine, and above all if the micro-simulation mode is chosen (as it is for UrbanSim).
The common denominator of these two families of model is the use of the discrete choice theory (McFadden, winner of the Nobel prize in economic sciences in 2000). That theory, already used to model the mode choice or the route choice, models choices on the basis of a set of mutually exclusive alternatives. The probability of an agent choosing any particular alternative is given by the probability of the utility of that alternative for that agent being higher than the utility of all of the other alternatives. Discrete choice models use the principle of maximisation of the utility. Thus, a household will decide to locate in the zone (the alternative) that procures the highest utility for it.

In France, the LUTI models are not yet used out in industry. They remain the prerogative of researchers. And yet it would seem that there are plenty of potential candidates, namely all those who already routinely use traffic models for socioeconomically assessing transport infrastructure projects: planners, design offices, urban development agencies, instructing parties (clients/owners and their assistants), technical service departments of central government, transport and urban planning departments of cities and of local authorities, and operators and other authorities who organise transport. It seems that there is indeed a demand from decision-takers, but that demand has not been able to be expressed formally, since there is no clear and understandable supply of LUTI modelling tools, and, above all, there is no official working framework that is accepted and shared. Socioeconomic assessment of projects is formalised in France by the framework instruction of 27 May 2005 that follows on from the recommendations of June 2001 from a group of experts chaired by M. Boiteux. There is no obligation in that framework to use the LUTI model, but only an obligation to calculate an Internal Rate of Return (IRR) for the local authority that is essentially based on time saved by users. And yet all of the experts and decision-takers are aware of the limits of conventional socioeconomic calculations and of cost-benefit analysis. Highlighting the IRR does not suffice to justify a project, however high that rate is, even if a few explicit “adjustments” are allowed in a more favourable direction to the time value, to the environment values (noise, pollution, greenhouse gas, safety, and security) or indeed to the investment cost of the project (which is more debatable). In order to find new benefits, Jean Poulit⁽²⁾ has proposed to go over from time saved to accessibility gains, while the British have developed a method of calculating “wider economic benefits” based on macroeconomic effects such as agglomeration effects, imperfect competition effects, etc. In a multi-criteria approach, the output from LUTI models on relocations and prices appears as a way of giving new items of assessment for helping to make investment choices.

Even in the research world, there have been relatively few experiments on LUTI modelling and applications. In the 2000s, IAU île-de-France and the University of Cergy conducted research subsidised by the Predit⁽³⁾ and known as Simaurif⁽⁴⁾ on transposing the UrbanSim model to the Paris Region. That first piece of research had two follow-on projects, namely the Plainsudd project subsidised by the ANR⁽⁵⁾. Sustainable Cities in partnership with the LET⁽⁶⁾ and Vinci, and the SustainCity project run by the University of Cergy, ENS Cachan and Ifsttar⁽⁷⁾ with the aim of adapting UrbanSim to European Cities. The Certu⁽⁸⁾ tried to apply Tranus to Lyon in the late 1990s, with mixed results, and a second application is in progress in Grenoble (Aetic project), run by Iddri⁽⁹⁾ and Inria⁽¹⁰⁾. The LET is continuing to implement UrbanSim in Lyon through the Simbad project. Vinci has tested its strategic tool Pirandello to assess the effects of cordon toll scenarios in Paris and in Lyon on relocation of households. The LVMT⁽¹¹⁾ is developing its own model through the ANR Aspect 2050 project, with a focus on the housing market. In 2013, new ANR Digital Models research work known as CITiES was begun, bringing together most of the French stakeholders already in place (IAU îdF, Iddri, Ifsttar, Inria, LET, LVMT, and Vinci) and two designers of existing models (de la Barra and Waddell), the subject relating to the theoretical and methodological obstacles constituted by adjusting and validating the LUTI models. IAU île-de- France, in its capacity as a multi-disciplinary expert on urban planning, has a natural role to play in this type of partnership with research, in which it can serve as a gateway between academia and industry.
It was only in 2011 that the first call for tenders was launched in France for application of an integrated model by the Société du Grand Paris (SGP, the Greater Paris Corporation), with a view to assessing the socioeconomic utility of the Grand Paris Express transport network. The bids came mainly from research laboratories, and it was those bids that were selected. Two work packages concerned integrated modelling directly: firstly, estimating the effects of building the Grand Paris Express network on location of businesses in the Paris Region, and secondly calculating the overall effect of building the Grand Paris Express network on location of jobs and of the population, and on land/property values. The Mussa, UrbanSim and Relu-Tran models were proposed to satisfy these requirements by teams including researchers and practising professionals. Ultimately, the teams, who were given extremely tight deadlines, faced huge application difficulties and technical and methodological constraints.

In the course of the ANR Plainsudd project, L'Institut Paris Region and Vinci proposed a comparative assessment of the UrbanSim and Pirandello models on two case studies: the extension of the T3 tramway in Paris to be opened at the end of 2012, and the Massy-Évry tram-train in Essonne scheduled for 2018. These maps of differences between scenarios with and without the project at a future horizon show a very clear effect of polarisation of the households and jobs along the extension of the T3 within a radius of 750 metres.

On the theoretical knowledge side, the bases for interaction between land use and transport, and the microeconomic or macroeconomic formulations have already been studied and documented widely in the Paris Region. As regards residential location, the results are satisfactory, even though it remains possible to go deeper, e.g. on disaggregation of the supply of housing per type and per size of dwelling, and on fine segmentation of households. As regards location of businesses, research has further to go. The determinants of location or more varied, the heterogeneity of supply and demand is greater, and the data on corporate property prices is much rarer than the data on residential property prices. We lack a relevant and comprehensive model for the supply of urban projects (housing, offices, industrial activities, SMEs including industrial SMEs, facilities, retail outlets, and logistics platforms) both for the static models and for the dynamic models. As regards “firmography”, the process of creation, growth, and destruction of jobs, and closures are still poorly modelled, but as a first approach, that is not an obstacle because LUTI models can, for the moment, be limited to location, the other “firmography” data being exogenised by overall assumptions for growth at regional level. By default, the location sub-models allocate without any ground capacity constraint in “desire spaces”, the counterpart of allocation in “desire paths” or “desire lines” that is dear to traffic modelling. As for the property price model, it is based on an imprecise hedonic price model that does not distinguish between the different types of property supply.
Beyond these very real deficiencies and defects, we could already make do with the current knowledge and data. What is currently holding up dissemination of these models to industry is firstly the unavailability of a piece of software in the commercial sense of the term, making it possible to have a fully integrated approach to modelling both traffic and land use. It is no longer sufficient to connect up two existing models designed independently from each other and then to “bodge together” an interface between the two. The idea is to combine in one and the same tool both a land-use model and a traffic model, with common zoning and a spatial scale that is pertinent with respect to the questions asked. Such a tool would combine numerous advantages:

• A structure for the travel reasons that is consistent with the land-use data: why take into account a health reason if the healthcare facilities are not present in the property supply model?
• Competition between households and businesses for land use.
• Automatic calculation of accessibility.
• An ergonomic user interface enabling the user to build scenarios and to launch applications “on the fly” on their desktop computer.

The specifications for such a model should include transferability from one region to another. The tools such as UrbanSim and Relu-Tran are still reserved for the initiated and require starting again from scratch for each application. Pirandello has been much talked about, but in order to have any hope of seeing it installed on the desktops of practising professionals, it would need a user interface and technical documentation that live up to its ambitions. Mussa, which has been taken over by the publisher Citilabs, who has renamed it Cube Land, might be able to pave the way to popularisation of LUTI modelling because it is integrated on the software platform Cube that is well known to modellers and therefore quite easy to implement, but its theoretical bases are complex and difficult to understand, unlike those of UrbanSim, for example. It proposes a new type of land-use model based on the bid choice theory which combines two approaches: auction theory or best-bid theory and the random utility approach. One of the major advantages is that the prices are “endogenised”; they are calculated and created by the software. In practice, Mussa-Cube Land still has weaknesses on the supply model and does not have any reference in Europe, except the first application for the Société du Grand Paris.
Finally, we should mention the essential role of central government in supporting modelling research & development projects. The subsidies from the Predit and from the ANR have been decisive in recent years in enabling French research teams to engage in transpositions and applications of LUTI tools created in America. However, except for the Pirandello model, which was born out of private initiative, the presence of a French model is sadly lacking. The funding needs are considerable. In order to develop ex nihilo an integrated model that is adapted to the context of French cities, that is generic and that is end user oriented, one million euros can be estimated as a starting basis. We should remember that, in the United States, in the late 1990s, Professor Waddell and his team received a federal subsidy of several million dollars for launching the development of UrbanSim.

In 2010, it was another piece of research known as Ultisim⁽¹²⁾ and conducted by L'Institut Paris Region, the Polytechnic University of Turin, and the Dutch bureau Significance, subsidised by the French programme for research and innovation in land transport (Predit) and by the authority who organises transport in the Paris Region (Stif), that highlighted the current limits of the LUTI models and suggested recommendations making it possible to take LUTI modelling forward favourably. One of the recommendations is to reinforce dialogue between modelling engineers and planning trade experts (demographers, economists, housing specialists, urban planners, and geographers). Multi-disciplinary exchanges would be useful on the diagnostics and prospection methodologies developed by each trade expert, on the use of the simulation models, and on their intelligibility. Ultimately, these exchanges should make it possible to give results that are both more realistic and also more useful for decision-taking. It would also be an opportunity for broadening the prospects of application outside the Paris Region and outside the conurbation of Lyon, by assessing the needs of the various local authorities and the data available to them.
Finally, we should add that this latent demand for land-use/transport modelling is interfering with a manifest and older demand for taking the interaction between transport and the environment into account, and which should not be confused with the former. A conventional traffic model is quite acceptable for meeting the needs of transport/environment modelling. Operational LUTI modelling is still very much in its infancy. We should not forget that it took thirty years before conventional traffic modelling suddenly became generally used with the arrival of the first commercially available software operating under Windows (Visem/Visum, Cube ex Trips-MinUTP, Transcad, Emme, Davis). Methods and practices still need to mature, inter-disciplinarity needs to progress, the ties between researchers and practitioners should be forged more strongly, central government financial support should not weaken, and it will doubtless still take another ten years before the first ready-to-use LUTI software comes onto the market.

1. We are not starting from scratch and the literature on the subject is prolific. The first static model integrating transport and land use endogenously dates back to 1963 (Lowry’s model). Then came the work of Alonso (1964) on urban growth, of Krugman (1985) on location of businesses, of Fujita and Thisse (1989) on urban dynamics and land bidding mechanisms, and of Arnott (2001) on monocentric and polycentric cities, and then, in the early 2000s, came the new American model, of the dynamic type, UrbanSim, developed by Professor Waddell.
2. Former Regional Infrastructure Director for the Paris Region.
3. French programme for research and innovation in land transport.
4.  Four reports are on line on the IAU îdF website of L'Institut Paris Region: www.iau-idf.fr
5. French National Research Agency.
6. French Transport Economics Laboratory
7. French Institute of Science and Technology for Transport, Development, and Networks
8. French Centre for Studies on Networks, Transport, Urban Planning, and Public Constructions.
9. French Institute for Sustainable Development and International Relations.
10. French National Institute for Research in Computer Science and Automation.
11. French City, Mobility, and Transport Laboratory.
12. Website for the project: www.iau-idf.fr/ultisim; report available on the website www.iau-idf.fr.