【正文】 public or private, and charged or free. This included all car parks within, or adjacent to, the town centre. Residential parking areas adjacent to the town centre, were also included as these provided free parking, with longer walk distances, and were often used by muters. Areas with high dropoff trips were modelled as a private parking type at their destination but could have been improved by having both the inbound and outbound legs of the dropoff trip in the matrix.The car park data also provided charging information for each car park. When bined with the car park interview data it was found to be possible to group charges into a single short stay and long stay charge. This simplification was appropriate in Rochdale, but it would have been possible to use a car park specific charge if the variation was more significant. Parking data in Rochdale was limited to peak car park occupancy and the model would have benefited from a prehensive survey of vehicles entering and leaving throughout the day. In Takapuna, and Nieuwegein, arrival time, dwell time, and occupancy data was available and this was used, with the charging information, to model driver’s choice of car park.Demand matrix segmentationMatrix segmentation enables the modeller to control departure time demands for different classes of vehicles. The degree of segmentation must be supported by the data. Similarly car parks should be grouped to a level mensurate with the detail of model input data.In Rochdale, matrices were derived for cars categorized by muter, nonmuter and work. The interview data enabled the muter and work matrices to be subdivided into private non residential (PNR) parking and contract parking. PNR parking supply, which is is notoriously difficult to estimate, was assumed to be unlimited in the model as the matrixes were explicity defined. For areas outside the town centre all drivers were assumed to park at their destinations.Takapuna adopted a similar approach, with demands segmented into ‘long stay’ and ‘short stay’ parking stay was further split into ‘on site’ or ‘general’ depending on access to workplace car parking. Long and short stay demands were estimated from the purpose matrices of a strategic transport model, with adjustments based on car park number plate and turn count data.Trip linkingIn order to model vehicles arriving and leaving from the same car park, trips in and out of the city centre area must be linked, with origin car parks and departure times selected based on prior arrival car parks and times. This requires more sophisticated control over the departure time and the departure location than is available from conventional OD matrix methodologies.In Nieuwegein, the study period covered the main shopping peak period on Saturday afternoon and included a warm up period to populate the car parks and initialise the ITS controller within the microsimulation model. To model the linked trips, all traffic departing from the city centre was deleted from the OD matrices. An external software controller monitored the car park occupancy in the model to determine the arrival profile and, after a suitable dwell time typically one hour, released matching vehicles on return trips.For Takapuna, before the PM simulation was run, a separate demand model was used to generate a profile of releases in the PM peak based on car park occupancy derived from the AM peak. This was to be subsequently used in the PM model run. Matrices were generated based on the profiled demands derived for each zone within Takapuna. When selecting from which particular car park the vehicle should depart, the demand model matched its outbound zone with that of a parked vehicle. The match was made based on the parking duration (long/short), and the expected departure time estimated from the arrival time distribution. If this process found a vehicle within 20% of its expected departure time, then one was added to the profile to be released. If no match was found, then the release was carried forward to the next profile period and the search repeated.The Rochdale model was also divided into AM and PM peak scenarios. The choice of car park for departure in the PM peak was modelled using the generalised cost of travel bined with an exit cost to help bias car park selection. Comparison with observed data showed the use of exit costs could successfully calibrate the model. ROCHDALE KEY DEVELOPMENT AREASRochdale is a historic market town in East Lancashire, now a metropolitan borough of Greater Manchester. The town centre contains a central retail and business area that not only petes with other local centres but also with Manchester City Centre and the Trafford Centre. To help improve Rochdale’s petitiveness as a centre for retail and merce, Rochdale Development Agency plan a major redevelopment of the south east area of the town centre. Comparative assessment of the town centre masterplan proposals on the highway network and car park operational performance require an objective scoring method only available via a transport simulation model. Providing parking of acceptable size, location and type is considered critical to the attractiveness of the town centre retail and mercial environments and hence formed a major part of the modelling exercise.Car park searchingAs the primary goal of all three models was to study theimpact of car parking on urban traffic congestion, the strategies for allocating vehicles to car parks within the model and subsequent car park hunting were key to the success of the studies.The Nieuwegein study was specifically designed to test the effect of a proposed car park advisory system. VMS signs were positioned on all approaches to the town centre (Fig 2) and provided information on which car park to select (Fig 3). Based on the experience of the Town Parking Manager, the ITS system was configured to make only 20% of drivers follow the advice of the parking advisory system. This lef