(Ed: These comments about Westconnex Traffic Modelling are relevant to the Westconnex M4 EIS traffic chapter that is already published on the People’s EIS.Traffic modelling can be daunting but we hope that this contribution, along with other contributions on traffic, can help you prepare your own submission.)
By Anthony McCosker*
Traffic Modelling Flaws
(1) traffic demand management;
(2) rebase future year traffic demand; and
(3) operational traffic modelling.
This summary examines the limitations of relying solely on traffic modelling findings, before assessing the assumptions made by WestConnex traffic modelling, examining a parallel report completed by SGS and concluding with a brief case of other traffic modelling for similar projects.
Limitations of traffic modelling
Traffic Modelling has many limitations and therefore it is dangerous to rely solely on its findings.
In attempting to model the future, traffic modelling makes a number of assumptions to produce what can sometimes seem like absolute and certain figures. In contrast, the number of assumptions, simplifications and estimations used in the modelling means that the figures produced are just one possible outcome of many.
So to rely solely on these findings can lead to misguided conclusions or outcomes, adding unnecessary risk to a project (Evans, Burke, & Dodson, 2007). This can be seen in recent Australian examples such as AECOM facing litigation over the traffic modelling of the CLEM7 tunnel in Brisbane, litigation over the traffic modelling for the Lane Cove Tunnel, and the Cross City Tunnel struggling to reach 20,000 cars per day after modelling suggested that 90,000 cars a day would use it (see http://www.smh.com.au/federal-politics/political-opinion/the-forecast-was-not-good–or-even-accurate-20120929-26rzb.html).
Indeed a reliance on traffic modelling as a justification for projects (as commonly happens in Australia) has seen ‘…investments in Melbourne’s urban road network [result] in more time being used by Melbourne’s motorists rather than less time’ (Odgers, n.d., pp. 14-15), finding that from 2000-1 to 2006-7 overall speeds per hour on Melbourne freeways stayed generally the same, at around 78km/hour. Thus it is important to consider that ‘Transport models are useful planning tools, but travel demand forecasting is not a precise science, and there are numerous outside factors which are difficult to predict or quantify’ SGS (2015). Professor Michiel Bliemer and Dr Matthew Beck (both from the University of Sydney’s Institute of Transport and Logistics) (Bliemer & Beck, 2015) state that models do not factor in future trends, preferences or behaviour (even trends that are known to be happening, such as ‘millennials’ driving less).
Some limitations of traffic modelling
Traffic modelling overlooks future trends
Because ‘a linear relationship is assumed between population, concentrations of land use and long-term transport demand’ (Evans et al., 2007, p. 6), traffic modelling fails to consider future trends including (but not limited to):
- Changes in fuel prices and shortage (or perceived shortage) of fuel (‘Most technical assessments of transport systems are naïve to the issue of petroleum risk’ (Evans et al., 2007)):
Changes in government policy (including transport and planning policy)
The effect that emerging transport trends including decentralised and disruptive transport provision (such as Uber), car share (eg GoGet) and autonomous vehicles will have on car use
Changes in generational travel preferences (such as ‘Millennials’ preferring technological connections over private vehicle connections; ‘Boomers’ driving less as this cohort ages)
Changes in sociocultural trends (such as preference for particular destination types or avoidance of areas due to issues such as crime)
Traffic modelling inadequately address effects of ‘induced traffic’
Traffic models also struggle to accurately address induced traffic, which ‘weakens their capacity to inform policy makers about the broader economic value and environmental impact of major transport projects’ (Evans et al., 2007, p. 6). Induced traffic can include:
- Mode change (such as switching from public transport to car use due to reduced travel time upon immediate opening of the road, known as the Downs-Thompson Paradox—see for example http://io9.com/how-the-downs-thomson-paradox-will-ruin-your-commute-1152573927)
- New trip
- Change of route
- Shift of times at which people travel
- As travel times are initially shortened by increased road capacity, people have more time in their travel budget (generally around one hour—see the Marchetti Constant for more on this) so may choose to make longer trips
- Changes in land use due to changes in accessibility to transport modes
- Reduced public transport services further increasing automobility of a city
‘The biggest force still driving the Auto City to build large freeways and accommodate the automobile rather than providing other options is the standard “black box” transportation/land use model…These are based on how a new or widened road will save time, reduce fuel, and lower emissions and road accidents…these benefits are illusory due primarily to “induced traffic.”’ (Newman & Kenworthy, 1999)
Traffic modelling oversimplifies trip types
Travel is ‘grossly simplified’ with minimal trip types considered (Evans et al., 2007, p. 4), for instance ‘trip-chaining’ (combining a number of destinations in one journey, such as dropping children at school before going to work, then picking the child up and going shopping) is ignored in modelling due to its complexity
Traffic modelling oversimplifies or limits considerations that lead people to choose trip or mode types
Residential density, land use mix and non-motorised accessibility all influence travel behaviour but are rarely accounted for (list adapted from Evans et al., 2007):
- ‘Australian metropolitan strategies…generally seek to reduce land use separation and distance, to promote walking, cycling, and public transport, and to reduce the use of the private motor vehicles. The use of models is unable to assess land use/transport interactions in order to determine and prioritise transport project investments within these strategies is therefore questionable.
Trip zones considered are generally large, limiting consideration of walking or cycling
Modal assignment is limited and overlooks many qualitative considerations of public transport services (such as youth preference to engage technology while travelling)
Limited consideration of non-motorised trips and other travel options (such as carpooling)
There is a focus on interchange and waiting times over quality of nodes or destinations (which affects the modes of transport people will take)
Traffic modelling focusses predominantly on travel times at the expense of other considerations
Models give limited consideration to effects such as pollution, noise or carbon emissions, while route/traffic assignment (Evans et al., 2007) assigns traffic flows to an equilibrium where no traveller can switch routes and reduce their costs which is not how the ‘real world’ works; capacities are generally over-simplified (for example heavy vehicle movements and highway geometry are often overlooked).
Traffic modelling is generally ‘expert’-led and ‘technocentric’, with little community input or justification of assumptions and inputs
Due to their technical nature,‘knowledge of how the models work and their capacities, and in turn their biases and inadequacies, is often restricted to a small number of professional experts’ (Evans et al., 2007, p. 2). This can give traffic modelling reports the impression of ‘objectivity’ and ‘universality’, whereas the policy context and the political surroundings certainly play a role in the assumptions and inputs into such models, and when this is added to ‘the inherent inadequacies of transport modelling, this technical complexity may be seen to create a form of institutional risk for transport planning assessment’ (Evans et al., 2007, p. 2).
Traffic modelling generally favours one mode—the car (Evans et al., 2007)
M4 East traffic modelling—what are the assumptions?
‘Do nothing’ baseline is assumed
A ‘Do nothing’ approach is used as the baseline for any time-saving benefits of the M4 East and wider WestConnex project, however a more viable comparison might have been an incremental improvement of multiple modes of transport infrastructure (including for cars, buses, trains, light rail, walking and cycling) using the funding amounts for the M4 East and wides WestConnex project ($15.4 billion). This could present an opportunity for increased viability of the traffic modelling, as previous suggestions by independent experts (see http://www.reportageonline.com/2014/06/westconnex-motorway-not-actually-going-to-help-sydneys-traffic-congestion/) for sustainable public transport as an alternative have been ‘overruled by WestConnex’, and the speed of a city’s roads are directly related to the speed of its public transport (known as the ‘Mogridge Conjecture’).
An auto-dependent study area is assumed
The study area—defined as the Local Government Areas in the project—is assumed in the report to be auto-dependent and reliant predominantly on cars for transport. However Newman and Kenworthy (2015) outline the fact that a suitable aspirational target for total trips taken by car might be 75% in an ‘automobile fabric’ area. Table 5.7 (Appendix G, p. 5-8) however shows that the average weekday travel for all local government areas within the project area is 57%, far lower than both the Greater Metropolitan Area of Sydney (67%) and the threshold for Newman and Kenworthy’s ‘automobile fabric’. It is a figure that is closer to a ‘transit fabric’ of 50% overall car use (see also Figure 5.4 from Appendix G, below). Along with the fact that 90% of western Sydney commuters to the CBD travel every day by public transport (SGS, 2015), this brings into question the modelling assumptions that cars are the preferred form of transport and that they will remain so to the modelling horizon (2031).
(Appendix G (p. 5-8) itself states: ‘Findings from the HTS [household travel surveys] shows that on average, 57 per cent of trips on a typical weekday in the project area are car based compared to 69 per cent in the Sydney GMA. The lower proportion of residents who are dependent on car travel can be partly attributed to good public transport options in the project area and also to the proximity of activities with a high proportion of travel utilising the walk mode share in comparison to LGAs with a more dispersed land use.’)
Current trends such as ‘peak car’ are ignored; outdated status quo is assumed
As outlined above, traffic models struggle to deal with future trends and patterns. However even current trends such as peak car (Newman & Kenworthy, 2015), which began in 2004 and is the decrease in the growth in overall car use, are ignored in the modelling. Ignoring trends towards peak car and shifts away from automobile dependence, as well as societal changes (such as an ageing population in Sydney and reduced reliance on cars by younger generations) and potential future technological developments (such as car share systems and autonomous vehicles), adds risk to the modelling assumptions and significantly reduces their reliability.
A certain level (2-7%) of induced traffic is assumed
‘At the extremes of the project a slight increase in volumes is shown on Parramatta Road, Concord Road and City West Link. This is indicative of the induced traffic demand attracted to the corridor as a result of the project…To the west, Concord Road and Parramatta Road continue to show an increase in expected daily volumes reflecting the induced demand resulting from the attraction to drivers of the WestConnex scheme.’ (Appendix G, p. 8-2)
‘Induced travel demand increases 2031 future year traffic volumes using WestConnex between two per cent and seven per cent, with the specific value varying across different sections of the project.’ (Appendix G, p. 4-6)
Yet the uncertainty regarding the actual amount of induced traffic and its effects on project aims (such as ‘Relieve road congestion’ and ‘Create opportunities for urban revitalisation…along Parramatta Road’), its impact on the local study area and its implications in the context of Greater Metropolitan Sydney area are not addressed.
The effect of induced traffic demand on public transport usage is not acknowledged
Despite claims that improved public transport (such as bus) travel times will improve patronage (Chapter 8, p. 32), the effects of induced traffic (such as switches away from public transport to cars) on alternative modes are ignored. Concerns regarding increasing the demand for automobile use when the majority of developed cities around the world and most strategic directions and plans for Sydney point towards reducing automobile demand are also overlooked.
SGS traffic modelling report is ignored
A traffic modelling report conducted by SGS Economics and Planning (SGS, 2015) relating to the entire WestConnex project produces numerous counterpoints to the modelling conducted for the M4 East project and given in the EIS. This report is ignored throughout the M4 East report, despite the opportunity to strengthen both models by comparing and contrasting outcomes, and identifying differences in assumptions that led to any disparities (some select findings from the report are outlined below).
Brief findings from the SGS traffic modelling report
As a counterpoint to the modelling used for the EIS, this section will briefly outline the main points made by the SGS report that was conducted into the wider WestConnex project (available at http://www.sydney.org.au/sgs-economics-and-planning-westconnex-transport-modelling-summary-report/).
At best the disparity in the two projections proves the difficulty in accurately predicting future transport movements across a complex network in a city such as Sydney and confirms the above points warning against sole reliance on traffic modelling for project justification. At worst they present a bleak view of the effectiveness the WestConnex project will have, and bring the validity of the modelling used and thus the justification for the entire project into question. Briefly, the SGS report found that:
‘Sydney traffic congestion will worsen with or without WestConnex, with the project only making a minor difference to Sydney’s overall traffic in the future…The net effect [of the entire WestConnex project] is similar to the status quo.’ (SGS, 2015, p. 1) (see below)
‘The [SGS] modelling confirms that WestConnex will not improve access to the Sydney CBD…the CBD is already congested and has little available parking.’ (SGS, 2015, p. 4)
‘Traffic flows on Parramatta Road will increase by up to 22 per cent (between Homebush Bay Drive and Concord Road) as vehicles avoid paying the toll on the M4 and M4 eastern extension. This finding is consistent with the WestConnex Delivery Authority’s own assessment presented in the M4 Widening Environmental Impact Statement and with the traffic flow impacts observed when the M4 toll was removed in 2010. As a result of WestConnex, Parramatta Road will take more traffic in the future, not less.’ (SGS, 2015, p. 15)
‘Traffic growth on Parramatta Road will clearly jeopardise the government’s planned urban renewal and population growth along this corridor.’ (SGS, 2015, p. 4)
Issues with modelling of other comparable projects
In recent years, traffic modelling of other similar projects has been called into question on a number of occasions, and serves as a warning against relying solely on traffic modelling as justification for road projects into the future. For instance:
- The Cross City Tunnel (Sydney) became insolvent in 2006 as a result of significant traffic modelling discrepancies—90,000 cars/day were predicted through the model, though just over 20,000 cars/day actually used the tunnel (see http://www.smh.com.au/federal-politics/political-opinion/the-forecast-was-not-good–or-even-accurate-20120929-26rzb.html)
The two companies responsible for the Lane Cove Tunnel (Sydney) traffic modelling were subject to litigation after the tunnel became bankrupt directly following its 2009 completion due to actual use less than half that projected in the modelling (see http://www.smh.com.au/nsw/trial-to-start-on-144-million-lane-cove-tunnel-debacle-20140809-102c6d.html)
AECOM (who conducted the M4-East traffic modelling) faced litigation over the modelling of the Clem7 tolled tunnel in Brisbane. Central to the claim was the fact that AECOM provided traffic models showing 100,000+ cars/day usage by 2011, despite having 18 months earlier estimated usage would be 57,000 cars/day. Actual traffic usage numbers in 2011 were under 24,000 cars/day (see http://www.smh.com.au/business/backers-sue-on-tollroad-forecast-use-20110414-1dfxd.html).
Bliemer, M., & Beck, M. (2015). Myth: Roads are the solution to congestion. Paper presented at the Festival of Urbanism, Sydney.
Evans, R., Burke, M., & Dodson, J. (2007). Clothing the Emperor?: Transport modelling and decision-making in Australian cities. Paper presented at the State of Australian Cities National Conference, Adelaide.
Newman, P., & Kenworthy, J. (1999). Sustainability and cities: Overcoming automobile dependence. Washington, D.C.: Island Press.
Newman, P., & Kenworthy, J. (2015). The end of automobile dependence: How cities are moving beyond car-based planning. Washington, DC, USA: Island Press.
Odgers, J. (n.d.). Have all the travel time savings on Melbourne’s road network been achieved? GAMUT Discussion Paper. Melbourne: RMIT.
SGS. (2015). WestConnex transport modelling: Summary report. Sydney: Commissioned by City of Sydney.
- Anthony McCosker is doing his doctorate at the Curtin University Sustainability Policy (CUSP) Institute