The validation of the UK National Transport Model: a backcasting approach

The UK National Transport Model (NTM) has been developed in recent years to provide the Department for Transport with a model to use for the assessment of national transport initiatives. There are two main applications that the model is used for: providing forecasts of demand for travel over the next 10 - 15 years, and in the case of the recent White Paper, 25 years; looking at the potential impacts of different policies. The NTM has developed from the original model that was being used in the Department to forecast traffic. In the mid 1990's there was a feasibility study for producing a national transport model and that produced a number of reports suggesting a range of approaches. In 2000 there was a general realisation within the department that a better model was required for forecasting congestion and pollution and at this point the earlier work was revisited and a strategy was devised for an incremental development of the existing model to provide the necessary functionality. This, coupled with a desire to have a truly multi-modal modelling system within the department has led to the model that exists today. Recently the focus has been on gaining an appreciation of the model's strengths and weaknesses, which motivated the validation project that is the topic of this paper. This paper discusses the backcasting approach used to undertake a validation of the NTM. Two separate backcasts have been undertaken, for 1991 and 1975, providing points approximately 10 and 25 years back from the existing model base year. Of key interest is the approach used for comparing the backcasts against the known historic data for the back-years. Clearly there are many dimensions which the analyst would wish to examine in making a judgement about the level of fit that the model produces; be it geographic categorisation, trip length, journey purpose or traveller characteristics. Any model will have varying degrees of fit in each of these dimensions and the challenge for the analyst is to make an objective assessment of the main strengths and weaknesses to allow future developments of the model to improve its forecasting potential. To this end, the study team have adopted a modelling approach to the problem, whereby the fit of the model outputs for the backcasting years is assessed against all of the dimensions of the historic data simultaneously. The analytic approach that has been developed allows the analyst to see through the inevitable noise in the fit of the outputs to the known data and pick out the key trends, whilst considering any interdependencies between the fit to the key dimensions. The approach also provides the analyst with an indication of the statistical significance of the differences identified, which is clearly invaluable when making judgements on priorities for future model development. The paper is intended to provide the reader with an overview of the approach used in the UK NTM validation, and the transferable lessons provide useful guidance to those considering future model validation studies. For the covering abstract see ITRD E135582.