Thomas W. Nicolai Transport Systems Planning and Transport Telematics, Berlin Institute of Technology (TU Berlin) 1 Investigating accessibility indicators for feedback from MATSim to UrbanSim Annual User Meeting and Tutorial 2012 Berlin
Agenda 2 Introduction Problem Statement Accessibility Zurich Application & Methodology Results Conclusion & Outlook
Agenda 3 Introduction Problem Statement Accessibility Zurich Application & Methodology Results Conclusion & Outlook
Introduction 1/4 4 We integrate a land-use model with a travel model SustainCity Project (see sustaincity.org) Aim: Provide a comprehensive tool for planners and decision-makers in government UrbanSim Extensible, microscopic urban simulation model Computes location choice decisions of developers, households and firms Access to certain activity locations (e.g. work) has an influence on these choices No transport modelling UrbanSim Households Firms Developer MATSim How does this integration works?
Introduction 2/4 5 Interaction Sequence UrbanSim MATSim UrbanSim MATSim Land-use Pattern UrbanSim Population Accessibility Indicators Base year Time Line Next year Microscopic simulation Person centric coupling
Introduction 3/4 6 MATSim4UrbanSim Stable Release Run and configure MATSim4UrbanSim via OPUS GUI Most settings automatically set/adjusted More details: http://matsim.org/docs/extensions/matism4urbansim OPUS GUI: Embedded the MATSim4UrbanSi m setup into UrbanSim.
Introduction 4/4 7
Agenda 8 Introduction Problem Statement Accessibility Zurich Application & Methodology Results Conclusion & Outlook
Problem Statement 1/4 9 Standard Feedback from external travel models to UrbanSim Zone-to-zone impedance matrix (2-point measure) Generalized costs of travel (e.g. travel times) for any pair of zones Left: City of Zurich, zone map Right: Zone-to-zone impedance matrix including travel time, cost and distance for any pair of zones
Problem Statement 2/4 10 Limitations Matrix grows quadratically with number of spatial units: 1000 Zones 1 000 000 Entries (per attribute) 1 500 000 Parcels 2 250 000 000 Entries In practice a matrix can be up to 1,7GB in size (5000 Zones) Limits number of attributes (e.g travel times at different times-of-day) Not feasible for higher spatial resolutions, e.g. parcels
Problem Statement 3/4 11 Spatial Resolution Feedback is on an aggregated, zone-level: Identical travel costs for every residential building per zone Applying travel costs to residential buildings on an aggregate-, zone-level
Problem Statement 4/4 12 Objectives Enhance location choice decision in UrbanSim with an alternative, improved feedback indicator from MATSim that is also applicable for higher resolutions on a parcel-level How can this be achieved?
Agenda 13 Introduction Problem Statement Accessibility Zurich Application & Methodology Results Conclusion & Outlook
Accessibility 1/2 14 Try 1-Point Accessibility Measures instead of 2-Point Measures Growing linearly by number of spatial units A i := f( {c ij } j ), i.e. some aggregation over the destinations j One possible representative for 1-Point Accessibility Measures Logsum term * Using as placeholder for arbitrary 1-point accessibility measures Utility-based measure: Economic benefit, that some one at location i gains, from access to spatially distributed opportunities, e.g. work * see Ben-Akiva and Lerman, Discrete choice analysis, The MIT Press, 1985
Accessibility 2/2 15 Logsum Term A i := 1 β ln j D je βc ij A i = accessibility of location i D j = number of opportunities at location j c ij = generalized cost to get from i to j, e.g. car travel times β = scale factor, related to scale parameter of the logit model
Agenda 16 Introduction Problem Statement Accessibility Zurich Application & Methodology Results Conclusion & Outlook
Zurich Application 1/2 17 Zurich Statistics (2000) Population: 336 291 Jobs: 316 703 Network (OpenStreetMap): Nodes: 18 670 Links: 41 582 MATSim Simulation 1000 iterations 10% populations sample Home-work-home commuting trips City of Zurich (blue area) with open-street-map network.
Zurich Application 2/2 18 Objectives Visualizing and evaluating workplace accessibility Different generalized costs Different spatial levels: Zones vs. Cells Sensitivity test: Illustrating impacts from changes on the transport system to accessibility
19 Methodology Implementation of Zone- & Cell-based Accessibility Measure Origin Destination Zone Centroid (avg. parcels) Cell Centroid (100x100 m edge length) Jobs (aggregated to nearest network node) 234 Zones 9185 Cells 3586 Opportunities
Agenda 20 Introduction Problem Statement Accessibility Zurich Application & Methodology Results Conclusion & Outlook
Comparing Different Travel Costs in Accessibility Computation 21
Results 1/7 22 Travel Costs: Congested vs. Free-speed Car Travel Time Congested (left) vs. free-speed (right) car travel times in accessibility computation.
Results 2/7 23 Travel Costs: Congested Car- vs. Walk Travel Time Comparing congested car- (left) vs. walk (right) travel times in accessibility computation.
Comparing Spatial Resolution in Accessibility Computation 24
Results 3/7 25 Spatial Level: Cells vs. Zones Comparing the cell- (left) vs. zone-based (right) spatial resolution using congested car travel times in accessibility computation.
Sensitivity Test 26
Results 4/7 27 Sensitivity Test Closing Schwamendingertunnel
Results 5/7 28 Sensitivity Test: Base Case vs. Schwamendigertunnel closure Comparing base case (left) vs. Schwamendigertunnel closure (right) using congested car travel times in accessibility computation. Tunnel closure has an spatial effect in the north east of Zurich.
Results 6/6 29 Switching Plots: Base Case
Results 7/7 30 Switching Plots: Schwamendingertunnel Closure
Agenda 31 Introduction Problem Statement Accessibility Zurich Application & Methodology Results Conclusion & Outlook
Conclusion 32 Accessibility Indicator Economic benefit, that some one gains at location i, from access to certain opportunities (e.g. work) The indicator is influenced by number of opportunities (land-use side) the generalized cost of travel to get there (transport side) Captures changes on the transport system Good correlation: zone-, cell-based results Technicalities The cell-size is configurable Aggregation of opportunities: Accelerates accessibility computation Reduced opportunity locations from 316 703 to 3 586 Number of aggregated jobs used as weights in accessibility calculation
Outlook 33 High Resolution Feedback (Parcel) Interpolating accessibility values from cellbased approach Including further indicators for schools, shopping, leisure, gastronomy, administration, etc Investigating the influence of accessibility on location choices Improving MATSim4UrbanSim configurability Accessibility configuration Making more MATSim parameter accessible via UrbanSim GUI Assign interpolated accessibility values to parcels (blue cross)
Acknowledgements 34 Thanks to Michel Bierlaire Providing access to Zurich Data Theresa Thunig Helps setting up the Zurich scenario
Questions & Answers 35