The Internal Validation of a National Model of Long Distance Traffic.

During 1980/81, the Department of Transport developed a model for describing the distribution of private vehicle trips between 642 districts in Great Britain, using data from household and roadside interviews conducted in 1976 for the Regional Highways Traffic Model, and a new formulation of the gravity model, called a composite approach, in which shorter length movements were described at a finer level of zonal detail than longer movements. This report describes the results of an independent validation exercise conducted for the Department, in which the theoretical basis of the model and its the quality of its fit to base year data were examined. The report discusses model specification; input data; calibration issues; and accuracy assessment. The main problems addressed included the treatment of intrazonal and terminal costs, which was thought to be deficient; the trip-end estimates to which the model was constrained, which were shown to have substantial variability and to be biassed (though the cause of the latter could be readily removed), with some evidence of geographical under-specification; and the differences between roadside and household interview estimates. The report includes a detailed examination of the composite model specification and contains suggestions for improving the way in which such models are fitted. The main technical developments, for both theory and practice, are the methods developed for assessing the accuracy of the fitted model and for examining the quality of its fit with respect to the observed data, taking account of the variances and covariances of modelled and data values. Overall, the broad conclusion was that, whilst there appeared to be broad compatibility between modelled and onserved data in observed cells, there was clear evidence of inadequacy in certain respects, such as for example underestimation of intradistr ict trips.

This work was done in co-operation with Howard Humphreys and Partners and Transportation Planning Associates, who validated the model against independent external data; their work is reported separately.