Portal Transit visualizes operational data from Portland, Oregon’s public transit operator, TriMet. By combining schedule data with readings from automated passenger counters and GPS devices on board each TriMet vehicle, Portal enables the user to display load, utilized capacity, on time performance, and stop activity during different times of day.
The data in Portal are quarterly averages provided by TriMet after agency staff perform QA/QC. The measures available in Portal’s visualizatons are sums or averages of those quarterly averages within timeframes specified by the user. The mapping of performance measures associated with segments (as opposed to stops) relies on the merging of routes only where they overlap.
Read more about:
Portal Transit is an application designed to visualize archived operational data for TriMet. The secondary objective is to increase the public availability of agency data for a wide variety of interested parties.
Portal Transit relies on four sets of data. The first is TriMet’s service schedule, which indicates when and where every fixed-route service stop should occur. The second is TriMet’s Automatic Vehicle Location data set, which includes the actual time when each vehicle performed the schedule service stops. The third is TriMet’s Automated Passenger Counter data, which covers the number of passengers who get on (“board”) and off (“alight”) at each stop. The fourth is the geospatial data used as a context for visualize the first three sets of data.
The APC and AVL data are downloaded from the vehicles every night. On a quarterly basis, TriMet staff perform quality control/quality assurance (QA/QC) on the data to identify instances in which the AVL and/or APC devices malfunctioned so that erroneous data are excluded. Using the clean raw data, staff generate aggregate statistics for each three-month reporting period. The result of the aggregation is a single record for each scheduled service stop.
Example: Route 14-Hawthorne has scheduled service at SE Madison & 11th Avenue in Portland going westbound (Stop ID #3637) every weekday at 8:42 a.m. In the winter quarter (January, February and March) there were 64 week days in 2013 but 3 were holidays (New Year’s, MLK Jr., President’s) on which TriMet runs Sunday/Holiday service. In QA/QC, staff look for evidence of hardware malfunctions such as zero boardings on a certain vehicle for the entire day. If there are none of these, staff aggregate the 61 relevant weekday records of the 8:42 service at Stop 3637 to produce a single record for that quarter.
Because the AVL records when a vehicle actually serves a stop, each raw record includes the number of seconds by which the service is early or late. In aggregate, this yields an average number of seconds early or late for each scheduled service stop as well as a percent early, on-time, and late (TriMet defines on-time as between one minute early and four minutes late). The time is recorded even when a bus passes a stop without pulling over.
The APC records the number of ons and offs at each stop, which supports the computation of the vehicle load leaving every service stop. Because the schedule data includes the vehicle equipment in use and its attributes, including seated and standing capacity, one can compute a measure such as “utilized capacity” for each record in the raw data. Quarterly, the aggregate statistics include the average number of passengers boarding and alighting as well as the departing load and utilized capacity at each scheduled service stop.
For data analysis and visualization, Portal supports time-based queries. The user can specify the day of the week, since the schedule is based on weekday, Saturday and Sunday/Holiday service types. The user can specify the time of day, based on customized start/end times or pre-assigned time periods such as “AM Peak.” finally, the user can specify the year and quarter from which to query data.
Depending on the measure, a time-based query averages or adds the aggregated values described above.
Example: The 14-Hawthorne has 16 scheduled stops at SE Madison & 11th between 7 and 9 a.m. on weekdays (according to the Winter 2013 schedule). Therefore, Portal includes one record for each of those stops for that quarter. To visualize the performance measure “stop activity” during the AM peak, Portal takes the sum of those 16 records, each of which is an average. Thus Portal indicates that an average of 82.14 passengers board the 14-Hawthorne at Stop #3637 during the weekday AM Peak for the Winter 2013 quarter.
Currently, four measures are available in Portal Transit:
Segment Load: Load represents the number of passengers on a bus. For each bus departing a stop, load is calculated by adding the on’s and subtracting the off’s and assigning that value to the segment originating at that stop. An accurate calculation depends on the bus starting and finishing its run with zero passengers, which is one of the standards TriMet staff use in QA/QC. When load is displayed over a period of time, it represents the sum of all relevant records in order to visualize the total throughput, combining multiple routes if they travel on the same segment.
Example: The 17, 70 and 77 all travel on NE Broadway westbound from 24th Avenue to 21st Avenue after serving Stop #635 (NE Broadway & 24th). Between 7-9am, those routes serve the stop 9, 8 and 7 times, respectively (24 total). To calculate the load on that segment of Broadway during the weekday morning peak, Portal will take all 24 values of load from those scheduled service stops (each value being a quarterly average) and sum them.
Utilized Capacity: To help TriMet identify crowding, the Utilized Capacity measure is a ratio of segment load, described above, to each vehicle’s seated capacity. The measure is calculated the same way as load except that it is averaged across a timeframe instead of summed. In the example above, a segment of NE Broadway is served by 24 buses during the 2-hour AM peak period. By taking the average of the Utilized Capacity ratio for all of those buses, Portal can display the utilization of service on that segment across all three routes for that period of time.
Stop Activity: This measure helps illustrate the volume of activity at each of the nearly 7,000 stops on TriMet’s fixed route network. The aggregated quarterly value for each scheduled service stop represents an average. When the user enters a timeframe for the query, Portal displays the sum of these averages: the total number of passengers who board and/or alight at the specified stop during that time period on an average day during that quarter.
Example: Stop #13305 (SW Harrison & 6th Westbound) is served by four routes: 17, 35, 36 and 54. During the PM Peak at that location, 121.31 passengers board or alight from 25 scheduled service stops. 37% of the passenger (45.11) are boarding.
Stop On-Time Performance: For each scheduled service stop, the quarterly record includes the average seconds early or late and percent of stops that are made early (more than one minute ahead of schedule), late (more than four minutes behind schedule) or on time. In a timeframe such as the PM Peak, Portal averages all of the quarterly averages.
Example: The 15-Gateway TC serves Stop #5832 (NW Thurman & 26th in Portland Eastbound) 32 times every weekday. It is on-time 91.2% in the early morning, 75.7% in the AM peak, 93.8% in the midday, 85.6% during the PM peak, and 88.4% at night.