| dc.description.abstract | In prior studies, the transit service reliability has been looked at from the passenger’s perspective in terms of time spent by the passengers in waiting for their bus to arrive at a stop. The operators’/agencies’ perspective based on analyzing day to day travel time variability (TTV) gives a clear picture about transit service reliability but is not a well-researched area, partly because of limited access to comprehensive data sets of bus travel times. The present study used citywide global positioning system (GPS) data of bus rapid transit system (BRTS) operating in Ahmedabad city of India to carry out the travel time reliability analysis. A three-level TTV analysis was carried out in the present study (i.e., segment, route, and network level). The route level analysis dealt with observing the day to day and within the day TTV. Whereas, in the segment level analysis, the BRTS routes were divided into segments based on criteria of shared and unshared routes. Two travel time variability (TTV) models (considering different dependent variables) were then developed using 770 observations to analyze the factors causing variability in travel time (TT). The model was developed considering the linear regression technique, and the significant variables in the suggested model were selected based on the backward stepwise selection method. The TTR model with T90–T10 (90th minus 10th percentile travel time) as dependent variable was showing a better adjusted R2 value (i.e., 0.73). The model revealed that the independent variables like length, bus stops, and number of intersections affect TTR to a larger extent. The third part of the TTR analysis was based on developing level-of-service (LOS) criteria for comparing TTR of transit systems. Transit’s network level data was used to propose a revised LOS based on weighted delay index (WDI) which is an improvement over the conventional LOS criteria. K-mean clustering was used to classify WDI into groups wherein each group corresponded to a certain transit service level. The mean of the silhouette coefficient was estimated to be 0.5 which highlighted that the structure of the clusters was reasonable. | |
