Institute for Transport Studies FACULTY OF EARTH AND ENVIRONMENT Land Use Transport Interaction models: International experience and the MARS model Professor Anthony D May Emeritus Professor of Transport Engineering
LUTI models and MARS A brief history of LUTI models Evidence on land use transport interaction Some models and their applications The case for the MARS model Structure and operation of the MARS model Strengths and weaknesses of LUTI models
A brief history of LUTI models Three sources Urban economics (Alonso, 1964) Spatial interaction (Lowry, 1964) Discrete choice (McFadden, 1974) Early static models based on Lowry Quasidynamic models based on all three Entropy models (e.g. LILT, 1984) Spatial economics (e.g. MEPLAN, 1988; TRANUS, 1989) Activitybased (e.g. IRPUD, 1985; DELTA, 1998) Early comparative tests: ISGLUTI, 1988 The TRANSLAND policy review, 2000
Some of the models available TOPAZ, 1970 (Australia) CALUTAS, 1978 (Japan) IRPUD, 1985 (Germany) MEPLAN, 1988 (UK) TRANUS, 1989 (Colombia) DELTA, 1998 (UK) URBANSIM, 2000 (US) PECAS, 2005 (US) TIGRIS, 2006 (Netherlands)
The TRANSLAND findings Policy Trip length Frequency Mode choice Residential density 0 Employment density Mixed development Transit oriented development X 0 0 0 City size 0
The PROPOLIS results [LUTI models of seven European cities] Public transport speed, service and fare improvements contribute well But encourage longer distance travel Pricing of car use achieves significant benefits But may encourage relocation Alternative land use policies have little impact alone But higher density mixed development linked to public transport can support public transport and pricing measures Infrastructure schemes can provide benefits But only if designed to be costeffective And consistent with the overall strategy
A typical model structure
Spatial coverage of LonLUTI
Some typical applications Land use policies Mixed development, transitoriented development Major developments: Almere (NL), Thames Gateway (UK) Transport policies Rail investment Road pricing Joint impacts Greenhouse gas emissions Economic development and agglomeration Accessibility and equity
The case for MARS Effective urban transport strategies require a combination of land use and transport policy instruments Performance of strategies depends on the mix of instruments and the levels of each Instrument set, range Policy package N Optimised? Y Optimal Strategy Constraints Appraisal Optimal combinations can be found, but require large numbers of tests Scenarios LUTI model Objective function Hence a fast operating, sketchplanning LUTI model is needed
The aims of MARS Enable the user to test a wide range of policies Represent resulting interactions between land use and transport over a 30 year period Generate an appropriate set of performance indicators Operate quickly, producing results within a minute Be used for (constrained) policy optimisation Be easy for the user to understand and interact with Facilitate stakeholder involvement
Characteristics of MARS A very fast land use and transport interaction model Using VENSIM systems dynamics platform Works on a high spatial aggregation level Typically one zone per 30k to 50k inhabitants Has a simplified categorisation of users Two person types, purposes, time periods Represents up to five modes Is deterministic in each iteration But each market is not necessarily in equilibrium Utilises the theory of constant travel time budgets Adaptation times Transport < I year; land use >5 years Employed population Total commute time Workplaces Population Car availability Attraction Attractiveness of Total commute other zones Attractiveness trips by car Time for other trips Time per commute trip Commute trips by car Speed by car Parking search time B3 B2 Time per commute trip by other modes Time in car commute Car ownership Commute cost other modes Commute cost by car B1 B4 Fuel cost
Characteristics of MARS Employed population Total commute time Workplaces Attraction Commute trips by car Population Attractiveness of Total commute other zones Attractiveness trips by car Time per commute trip Speed by car Car availability Parking search time B3 B2 Time in car commute Car ownership Commute cost other modes Commute cost by car B1 Fuel cost Time for other trips Time per commute trip by other modes B4
The structure of MARS Demographic transition and growth model Car ownership model External scenarios Transport policy instruments Land use policy instruments Policy instruments Objective Functions: User benefits Operator benefits Investment costs Changes land use patterns... Housing development model TOD model Household location model Employment location Transport model Land use submodel Transport sub model
Cause and effect equations: e.g. population dp t i = dp t * i e U t i e U t i = ( t t dp dp ) e m * i e f e t t t ( acc, alq, vrd ) f i t t ( acc, alq, vrd ) i i i i t i Legend: dp it...increment of population allocated to zone i in year t dp t... Increment of population demanding housing in year t U it... Utility to live in zone i in year t dp e t... External increment of population demanding housing in year t dp m t... Increment of population moving house within the study area in year t acc it... Accessibility of zone i in year t alq it... Housing costs per square meter in zone i in year t vrd it... Quantity of recreational space in zone i in year t
Cause and effect in VENSIM Supply HU J T1> Stock HU J> d(r) in c(r) in b(r) in a(r) in k supply HU e(r) in Attractiveness moving in Accessibility Access Attr <Recreational green land J> Recreational green land Attr Rent Attr <access to wp> a(r) out Attractiveness moving out b(r) out c(r) out d(r) out <Housing costs> Residents out J Avg time living at the same HU Residents potential in J Potential residents move in HUs made available J Residents J t=0 Residents per HU J t=0 Residents total <Unsatisfied demand residents J T1> Residents in Residents Residents out Unsatisfied demand residents J Growth rate population T Unsatisfied demand residents J T1 <New HU occupiable> <Residents potential in J> <Residents out J> Residents move in J Residents J Residents move out J <Residents out J> Supply HU J T1 Supply HU J T <Residents per
The flight simulator policy input screen
Policies which can be modelled Pedestrians Pedestrianisation spatial(s) / temporal(t) Public Transport New PTInfrastructure Fares Frequency Quality factors PRT, cyber cars, BRT S S/T S/T S/T (new) Private Car Land use measures New Roads Road Pricing Parking charges/capacity Road capacity increase/decrease Fuel price/tax Awareness campaigns/teleworking Controls on development Land use charges S S/T S/T S/T S New S S
Output indicators for UK goals Goal Climate change Productivity Equality Health and safety Quality of environment Indicator Annual CO2 emissions Personh delay in the peak Economic vitality Accessibility by all modes Noncar accessibility Number of accidents NOx and PM10 emissions Proportion of developed land
Comparing outputs for different tests 20 Mode Share Off Peak PT Ped 10 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (Year) ms pt opeak : Sim 8_2 ms slow opeak : Sim 8_2 ms pt opeak : DoNothing ms slow opeak : DoNothing Percent Percent Percent Percent
Normalised performance indicators
Spatial impacts using ANIMAP
Optimising to an emissions target 600,000 total CO2 per year WW 550,000 t/a 500,000 450,000 400,000 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (Year) total CO2 per year WW : optimal total CO2 per year WW : co2 target total CO2 per year WW : testcalib total CO2 per year WW : DoNothing
Some strengths and weaknesses LUTI models generally MARS Able to model land use policies and land use impacts of transport But complex, demanding of data and time consuming So more difficult to interpret, and less often used Also able to model land use policies and impacts Very rapid to operate, so can be used interactively But only identifies strategic impacts And dependent on assumed fixed travel time budget
Further information Four page description of MARS List of references For general information: a.d.may@its.leeds.ac.uk For information on MARS s.p.shepherd@its.leeds.ac.uk guenter.emberger@ivv.tuwien.ac.at