Synthesising the Transpose Times at Roadside Interview Sites Using Probability Functions Derived from Car Park Interview Data
S Moriarty, T Wang, Mouchel, UK
Roadside interviews are conducted to ascertain travel movements at the time they are made. The collection of roadside data is an expensive element in model development and there is often the pressure to minimising the costs and traffic disruption. One consequence is that the data collected may be limited to responses that can be collected relatively quickly such as survey location, interview time, occupancy, vehicle type, origin address, origin purpose, destination address, and destination purpose.
While it would be possible to survey in both directions at interview sites, there are organisational problems in liaising with police and local authorities, and a perception that roadside interviews cause congestion. Consequently the common approach is to survey in one direction but collect control counts in both directions. The conventional approach in analysing the roadside interview data is to maximise its use in the model development by transposing the interviews in the surveyed direction to derive estimates of the trip pattern in the non-survey direction.
The weakness of this approach is apparent in that trips made in the morning peak do not necessarily return in the evening peak (and vice versa). For instance there are usually few shopping trips in the morning peak or education trips in the evening peak. Consequently direct transposition will create shopping or education trips in time periods when they are unlikely to be made.
With the increasing modelling sophistication such simplistic approaches of direct transposition may not be acceptable or desirable as models may require that the travel demand is segmented by trip purposes such as work, commute and other car trips, as part of addressing traffic and economic impacts on different travel groups.
Consequently as an alternative to direct transposition, a method of synthesising the transpose interview direction has been derived from the analysis of car park interviews. In common with roadside interviews, car park interviews collect similar data but as the survey is conducted at car parks, the risk of causing delays to traffic is minimised. This allows car park interviews to collect additional data including the arrival and departure times, which allows probability functions to be derived for the arrival time or departure time, by trip purpose, and time period. The functions can then be used to synthesise more realistic transposition of roadside interview data using Monte Carlo approach.
The paper will present recent work to synthesise the non-interview direction of travel using probability distributions derived from car park interviews for a number of towns in the UK including, Doncaster, Shrewsbury, Lancaster, Halifax, Colchester and Bury St Edmunds. Car park interview data from these towns provided a database with over 12,000 interviews.
As part of the analysis the data collected at the towns were compared to assess similarities and differences in their characteristics with respect to trip purpose and duration of parking. Analysis of the data indicated that the duration of stay at the car parks was similar, which means that the probability functions derived could be applied to other towns.
Observed probability functions were derived for the duration between the outbound and return trip for different trip purposes. This allowed the departure time for a trip from home in the morning peak to be predicted later in the day, or the arrival time for a car returning home and leaving a car park to be predicted earlier in the day. The functions were mathematically smoothed.
As part of the study, direct transposition methods were compared with the new approach using two-way roadside interview data. The observed data in the forward direction was transposed using alternative methods to synthesise the reverse direction. The transposed dataset were compared with the observed roadside interviews in the reverse direction. The results indicated that there were considerable differences between the transposition methods and the new approach gave more realistic results.
In conclusion, the main findings of the study were as follows:
?It was possible to derive probability functions that could be used to synthesise the transpose time for different purposes for car drivers. The six towns generally had similar parking characteristics, which meant that the probability functions were transferable
? The probability functions could be used in a Monte Carlo approach to synthesise the duration of stay at the destination. Allowing for travel time to and from the roadside interview site the likely transpose time for the interview can be generated.
? The approach gave more realistic results than simple transposition or transposition controlled by trip purpose.
Association for European Transport