Modelling the trip length distribution of shopping trips from GPS data

Abstract

The newly proposed approach to the calculation of bulk service contributions in Gauteng uses not only trip generation, but also Vehicle-Kilometres of Travel (VKT) generated by a development as the basis for estimating traffic impact. This presents an empirical problem as data on VKT or trip lengths, linked to specific types and sizes of developments, are scarce and difficult to measure. Previous approaches to measuring trip lengths have used travel surveys with and without Global Positioning Systems (GPS) data. South Africa has relied only on travel surveys without GPS data to estimate trip length information. The recommended trip length information for different developments in South Africa is provided in the TMH17 document which is used in the bulk service contributions calculations. However, the trip lengths provided in the TMH17 is based on limited South African data and is supplemented by studies that were done in Florida in the United States of America. The advances made in GPS technology over the last decade have created new opportunities that can be used to collect GPS data for travel surveys. In this research a novel approach to collecting and analysing trip length data using passive GPS loggers distributed to a sample of 726 drivers in Gauteng was tested. A stop time of 110 seconds and repeated use of road links were used to detect trip ends in the GPS data set. The shopping centre trips were extracted using Geographic Information System (GIS) data of the locations of shopping centres compared to the trip end positions. The average trip lengths to and from shopping centre were then calculated. It was found that the average trip length per shopping centre size is longer by approximately 4.8 km compared to the prescribed TMH17 average trip lengths. These results need to be confirmed with further research. The GPS data also provided the opportunity to calculate the percentage of travel per road Class to and from shopping centres. This is important, owing to the bulk contributions calculations only using the half adjusted average trip length. This is the average trip length halved and then only using the distance travelled on roads under the jurisdiction of the municipality excluding travel on Class 4 and Class 5 roads. It was found that 43% of the trip length distance is travelled on Class 2 and Class 3 roads. The 43% was compared to the TMH17 method of reducing the half average trip length to estimate the halve adjusted average trip length. The 43% was found to give similar results than the TMH17 method. Owing to the significant difference in average trip lengths between TMH17 and the GPS data results an alternative method of estimating average trip lengths was proposed. It was proposed that average trip lengths be estimated based on shopping centre type and not Gross Leasable Area (GLA). It was also proposed that the 43% reducing factor be used instead of the TMH17 method of estimating the halve adjusted average trip length as the 43% reducing factor is far less complicated. The proposed alternative method is subjected to further research and confirming of the average trip lengths results.