In large study areas some trips take longer to complete than can be encompassed within the modelled time period. The model would assign the whole trip to the network when it should only assign that part which is undertaken during the modelled time period. The model network thereby over-estimates traffic volumes sometimes to the tune of 10% to 25% - a large source of error. It is worse for large study areas, large conurbations, peak period models, long distance models, trunk road appraisals, railway and air travel. Even medium-sized study areas are likely to have some origin-destination pairs which would take longer than the modelled time period to complete. This paper proposes a solution to this age-old problem, describes its application to modelling travel in Kent and reports on the results of the methodology. The day was split into a sequence of one-hour time periods each of which modelled traffic which started their journey within the time period. The first time period assignment model was run and those trips which had not completed their trip were held over and allowed to complete their trip during the second time period together with those starting during the second time period. The process was repeated for the third and subsequent time periods, including assigning the accumulated trips incomplete from previous time periods, to cover the whole day. Trips which spanned two (or more) time period(s) had only that portion of their trip which occurred during the time period, assigned to the network in that time period. This meant that a system had to be devised for storing and assigning only parts of a trip. Two methods were devised, tested and compared. They were also compared with a conventional assignment approach and sources and sizes of error were estimated. The paper concludes with some indications as to how the methodology could be applied more widely and extended to include a more accurate representation of time within the time period being modelled. The research innovation includes the methodology for handling parts of a trip from origin or intermediate node to intermediate node or destination at a particular time, as well as the methodology for assigning only parts of a trip from node to node. Two methodologies were devised for each. Key to the success of one of the methods is the need to paths build to destinations, rather than current technology which uses paths from origins. For the covering abstract see ITRD E137145.
Abstract