Please enable JavaScript.
Coggle requires JavaScript to display documents.
Transport Modelling, Purpose:
Decision support tool
Differentiate…
Transport Modelling
Challenges
and changes
Behavioural model assumptions
- Incorporating new behavior changes - updated survey data,more regular integration of available data, managing uncertainty of future travel patterns
- Need for more endogenous choice responses to models are more in line with real markets
- Need for reliable exogenous forecasts and impact on travel (WFH, crowding, amenity, density, socio economics)
- Assumptions of rationality - rules of thumb, habits
- Different preferences - varies, changes of timer
- Localism of trips with WFH - income bias, richer can WFH easier
- Service changes - education, home deliveries, digital health, improved communication, WFH, ITS improvements
Modelling future transport options and operations -
- MaaS,
- micromobility, trip chaining/activity based trips
Pragmatic factor values - Historically used values may no longer be suitable in current context (post covid, decarbonise transport vision)
- Transfer penalities - for PT modelling, no longer as harsh with digital age
- Mode specific constant impact on utility function
Biases in modelling - Robustly dealing with assumptions
- Financial/political pressure on modellers to make assumptions to make projects work > stretch values to limit of reasonableness
- Challenge for modal integration
Biases in project evaluation
- Weightings
- Values - things that can be quantified dominate e.g. time travel savng
- Mode biases - different assumptions per mode
- Handling of uncertainty and prediction
- Incorporation of cultural and societal changing values
Valuation of travel time (VOT)
- General cost of travel - too general, should be mode modal specific and consider demographic cross-section. parking price incorporation.
- DIfferent travel experiences have different comfort and benefit
- VTTS - measure of opportunity cost of time, how is it valued
- varies across society, correlation with income,
- Toll road complex example - demand $ not just income based, but also trip chain, land use and activity based - how convenient is it for a given agent
Overfitting rife
- unconstrained regression without a theory
- methodologies to patch over problems (matrix estimation, pivot-point methods)
- recognise that models can't 100% match base year, people are unpredictable
- Standards can lead us astray (observed vs estimated) > we have made it impossible to achieve success without cheating, and we have undermined our ability of models to predict and inform decision making?
Locking in existing behaviour
- Embedding existing travel observations into the model (activity based models)
interdependent, hard to change parameters (e.g. hierarchical logit models)
Over focus on modelling inventions only loosely based on reality e.g. equilibrium, travel time budgets
Data use and management
- Big data available, how to incorporate current data more rapidly into model, automate it Build commercial relationships with private firms that have lots of this data..
- How to blend different data sources.
- Treating survey, count or preference data as actuals -there are errors, biases, variability
Use of single value outputs
- Range of values, scenarios, more helpful to make good decisions > shift to scenario planning and decide and provide frameworks
Model calibration forgoing predictive capabilities of models by over focus on matching observations - not a good trade off
- should not just any interdependent data sets between calibration and validation
- Industry guidelines here may be hurting the overall purpose of transport modelling with their prescriptive rquirements
Improving use of models
- visualization
- scenario testing, with critical thinking about the future, to support decide and provide framework
Slave to empiricism
- Use data, but there isn't always data, and sometimes data not accurate
- Need to acknowledge we sometimes know things without data - but how to put trust that this is truth
Biases in data capture into the travel behavior model
- 'under representation bias' - older/high income cohorts
- not representative of all behaviour
- Increase longitudinal data to capture change
Types
-
-
Macroscopic
- Behavior models embedded in the software - elasticities, willingness to pa, what if scenarios (land use density, carbon tax, remove capacity, travel distance charge)
- Poor account for endogenous responses and limited explanatory variables of relevance
- Limited feedback between transport and land use/location models
- Rely on exogenous supplied land use and population projections
-
-
-
-
Investment decision making, Project Evaluation Frame works and Business case
- Identify the problem
- Model future demand, evaluate optons
- Investment decision-making process (or is it just dressing up the story telling?)
Transport modelling role:
- Quantify impacts of transport strategies and staging
- Consistent way to compare options, transparency required in method/reporting
Assumptions:
- Growth assumptions
- Evaluation period
- Discount rates
Typical Approach:
- describe intervention
- Base case future assumptions
- Transport network changes, land use and demographic changes, macro economic changes, prices, behaviour changes
- Use demand model to predict change
- Evaluate outcomes of predicted consequences (CBA, MCA etc)
Needs to:
- Move away from historic focus on only network efficiency, time delay and benefit costs related to this
- Review values for 'willingness to pay' (WTP) and value of travel time (VTT)
- Capture progressive interest in broader outcomes for - network, society, urban area, equity, environment - decarbonisation net zero goals
- Move away from only economic appraisals and cost/benefit - moving to MCA to present these other benefits and support a decide and provide framework of transport planning
- Derisk future uncertainty with improved frameworks
- Incorporate flexibility/adaptability in decision making processes, support scenario planning pathways/findings along the journey
Outputs: Use to investigate possible future scenarios based on policy and infrastructure intervention, it can NOT tell you the exact future/outcome
- develop evidence base for infrastructure and policy development when future is uncertain
- sense-checks required
- More transparency required - visuals, reporting, twisting of truth
Story telling
- Humans driven by narratives and models are a story teller
- Numbers and technique can make stories more convincing, even often without merit or being justifiable/economical/sustainable > cultural/political impact can spin the story regardless
- Should develop loose insights, not "the single truth", meaningful scenarios and sensitivity testing required
- Impact models have on decisions may be limited, stories make the difference
Inputs
- assumptions - should reflect real behavior paradigms - inform by survey, data. Need to be explicit and transparent in making assumptions. affected by bias.
- Should be updated when evidence to do so
- Covid-19 induced behavior change - short term effects vs longer term change or return to BAU
My Experience
- Modelling is fusion of computers, probability, psychology, engineering, economics, politics, environment, data science
Land use assumptions
- Population growth, affected by covid-19 - national vs interstate migration patterns e.g. QLD historical peak in 2020
Managing Uncertainty
- Token sensitivity after making a decision not useful, should be project specific tests
- Deterministic mindset, probabilistic outcomes, use stochastic risk analysis on economic growth, population. Can consider this with monte carlo simulations - investigate assumptions around randomness, challenge false confidence by testing more events e.g. covid situations
- Scenario planning/testing - Distributions, not numbers are more useful using modelling tools - remove focus on single predicted scenario. decide and provide the future we want.
- Standard uncertainty management - different measures/values are treated differently e.g. delay values may not be questioned
Purpose:
- Decision support tool
- Differentiate performance of options
- Help quantify impact of investment
- Help understand potential future pathways
- Develop insights on city operation
model: Early indications > trends > trajectories > new norms
Not good at
- Making decisions
- Precisely predicting the future
- Dealing with strategic drivers of change (modal shift, climate net zero goals, inclusive access, place making)
- Capturing culture impact on city shaping and mode choice
- Accounting for network effects and multi-variable interventions (policy, operational changes to PT, invervenations)