Grasping the data

Transit planners today are faced with both an opportunity and a challenge – making the best use of the wealth of operational data streaming from new technologies – automatic vehicle location (AVL), automatic passenger counters (APC) and automated schedule feeds (GTFS). 

Transit planners, especially those that have gained their experience outside of large systems, have managed for years with very little data. They have learned to plan their systems with an annual on/off count and periodic manual checks of schedule adherence. 

Faced with the torrent of data they have never used before, a significant challenge is understanding what can be done with the data, and learning to ask new questions.
Working with Reseau de Transport (RTC) in Quebec, we have been on the front lines of transforming this operational data from data graveyard to data warehouse. Using AVL data, RTC has focused on run time calibration, performing detailed statistical assessment of run times, based on many thousands of records, that allow schedulers to clearly identify clusters of common trip times as well as outliers, and whether there are really patterns in those outliers. 

RTC has also used APC data to develop an understanding of linked boardings and alightings. Using a variety of algorithms, a vector of on-off data have been manipulated into on-off correspondence matrices, showing a probability of links between boarding points and alighting points for passengers. 

The volume of data in these calculations permits reasonable probability calculations that can be useful in identifying the impacts of various route changes. The data have been used with RTC’s BRT lines to identify frequency requirements over the length of the lines, potential opportunities for short-turns to increase frequency in the core area, and key stops that may serve as effective mid-line terminus points.

These data and tools can also be used to identify planned construction delay or detours, branch structures, express or limited stops. Planners have always had these questions, but often did not ask them, usually because the limited answers available from limited data were not particularly useful. Now, the new wealth of data expands the depth of our calculations, smoothing over the variations of limited data and ensuring robust, meaningful calculations.