Modeling Intrazonal Distance and Duration in Addressing Limitation of Intrazonal Generalized Cost Commonly Applied in Transport Models

The intrazonal modeling parameters of distance, duration, and generalized cost (GC) are conventionally estimated from a fixed-fraction of interzonal parameters or using linear equations. In this study, the intrazonal GC that was developed in the National Transport Authority’s (NTA) regional modeling system using a conventional fixed-fraction method (base) was compared to intrazonal GCs for six travel modes [passenger cars, bus, cycle, walk, taxi (passenger) and light goods vehicles] that were estimated using zonal GC equations adopted from literature. The zone-specific distance and duration values for the parameters of the six equations were developed using a regression modeling approach. At first, National Household Travel Survey (NHTS) data, Open Street Map, and traffic zone data were used to generate regression models. Then the reported trip duration and distance of NHTS datasets were then replaced with the corresponding time and coordinate defined distances and durations from the Google Maps Journey Time and Distance Estimates (GMJTDE) records, and the estimated equations from this process produced better validation performance, and thus integrated with the six equations. In the comparison stage, intrazonal GC values were estimated for all six modes using two different analysis and compared against the base dataset. In the first analysis, average distances and duration were obtained for all zones containing intrazonal trips and were directly applied in the six equations to estimate intrazonal GC. In the second analysis, intrazonal GC were estimated using models based on the aforementioned NHTS-GMJTDE datasets. Results indicate that the intrazonal GC estimated using the second analysis matched the intrazonal GC from the conventional method, and provided an indication of real-world situations to some extent. This study confirms that intrazonal parameters can be developed separately from the survey data, minimizing the efforts of calibration with sufficient accuracy that cannot be obtained directly from the survey data.