Use Transit on-board survey to improve the transit assignment model in Dallas Fort-Worth regional area
Corresponding Author: Hong Zheng, North Central Texas Council of Governments
Presented By: Hua Yang, North Central Texas Council of Governments
This presentation discusses how to use a transit on-board survey to improve the transit assignment model in the Dallas Fort-Worth Regional area. In 2014, the North Central Texas Council of Governments (NCTCOG), Dallas Area Rapid Transit (DART), Denton Council Transportation Authority (DCTA), and Fort-Worth Transportation Authority (FWTA) completed a regional on-board transit survey that encompassed all fixed-route transit services provided in the twelve county modeling area; this consisted of bus, light rail and commuter rail transit service. Twenty percent sample of riders was surveyed for boarding and alighting locations, and a ten percent sample was interviewed to collect socio-economic information, demographic information, mode of access, fare and the actual path of the riders’ one-way trip.
In this study we use the survey result to improve the model. There are two steps in this process. (1) We convert the survey data into model input data with comparable production-attraction (PA) form, then assign PA in the model and compare the modeled ridership and surveyed ridership. The comparison of ridership calls attention to several transit lines that needs to be investigated. (2) For those transit lines with concern, from the survey the demand of a specific transit line is known, we then assign such demand in the model, and we examine whether the demand is assigned on the specific transit line as expected. It is observed that the demand is not fully assigned on the specific transit lines, and the ridership flow reveals major flaws in the transit assignment model that can be improved. We mainly identify two types of flaws: (1) some riders are not assigned at all in the model; and (2) some riders are assigned but on a different line other than the line specified in the survey.
The reasons for the flaws are mainly due to limited access/egress time, poor connectivity of centroid connectors, and inaccurate stop station location. It indicates that using surveyed data to calibrate model setup parameters in the path-builder, improving centroid connector connectivity, and improving stop station location can improve the model result.