Issues Associated with Integrating Transport and Environmental Models
D Hardcastle, P Clarke,D Whittle, P Barrett, Mouchel, UK
This paper investigates the way in which traffic models may fall short of the required precision and discusses where improvements can be made.
Traffic models are primarily designed to support economic appraisals but also provide data for environmental assessments, in particular, air quality (AQ) and noise modelling. However, AQ models commonly require more precise traffic forecasts than can be reasonably provided by traffic models. If the data, as provided, leads to the prediction of AQ ‘exceedences’, these may be caused more by the imprecision of the forecasts, rather than the likely impact of the proposed highway scheme.
This paper investigates the way in which traffic models may fall short of the required precision and discusses where improvements can be made. In particular a new way of providing traffic forecasts is proposed which offers confidence limits rather than spurious precision.
In the development of highway schemes, a traffic model is conventionally built to assess the economic case for the proposal and to provide data for environmental assessment. The key environmental impacts of a highway scheme are principally Air Quality and Noise.
With scheme and without scheme traffic forecasts are typically fed into AQ and Noise models which predict environmental impacts with a high degree of precision. The key question is whether the traffic models can deliver forecasts of sufficient accuracy to match the standard required and support the consequences of any serious impacts.
For example, the EU specifies an upper limit of NOx concentrations of 40 microgrammes of NOx per cubic metre (of air) and highway schemes may be prevented from being built if they are predicted to cause an exceedence of this level.
However, whilst traffic models are very complex and sophisticated beings they are not necessarily designed to provide highly accurate predictions of traffic flows and vehicle speeds at specific locations. Their forecasts are more general and aggregate in nature, more in keeping with the prime aim of supporting economic appraisal.
For example, traffic models are validated (at the base year) which involves an overall comparison between model link flows and traffic counts. The popular (DMRB) criteria requires that 85% or more of network links achieve a GEH statistic of 5 or less. This represents an overall level of fit, which is very different from requiring that the modelled flow level on each and every link is within 15% (say) of the observed flow.
In addition, traffic models are built to predict end to end journey times with reasonable precision which is different from the environmental model requirement for accurate specification of speeds (in particular HGV speeds) at specific locations.
AQ models themselves maybe mathematically precise when taking the outputs of traffic models but they often ignore background sources which come from a range of generators. As a result, traffic models and AQ models are not generally set up to provide the information, at the level required, to predict the concentration of pollutants. However, if the inaccurate information then leads to the prediction of an exceedence, schemes are in danger of being rejected based on the presence of this exceedance. Traffic and AQ teams are then pressed hard to either redesign the scheme or modify the delivery strategy in order to remove the exceedance and enable the scheme to proceed. However, is the exceedance an impact of the scheme or is it the result of the lack of proper integration between the two modelling approaches?
In order to prevent these circumstances being repeated with each or any future highway schemes, it is necessary to review the interface between traffic and environmental models to determine what needs to be improved. Should traffic models be changed so they can fully support the requirements of the AQ assessments, or should the latter be ‘dumbed down’ to accommodate the realistic outputs from the traffic models?
This paper examines the current interface and outlines where and why the present situation is unsatisfactory. It also seeks to identify where the present assessment system should be changed in order to recognise the merits and shortcomings of both modelling systems. . In particular a new way of providing traffic forecasts is proposed which offers confidence limits rather than spurious precision.
Association for European Transport