1. Delays and Performance: King County METRO RapidRide C & D Lines University of Washington URBDP 422 Geospatial Analysis, Winter 2014 Debmalya Sinha, Austin Bell, Riley Smith, Andrew Brick

2. Overview Primary task: identify delays Where When Magnitude Secondary tasks: Identify priorities for remediation Recommend delay reduction strategies Future research: Relationship between delays and socioeconomic status

3. Data Onboard System (OBS) for October 2013 (245,826 entries) Records real-time information of bus activity No weekend data was included in data file General Transit Feed Specification (GTFS) Provides scheduled arrival times for all routes Shapefiles C & D Line stop locations (point) C & D Line routes, manually segmented (line) Field Data Physical attributes of stops and route segments

4. Methods Raw OBS and GTFS data imported into R All times converted to seconds after midnight where required Trips categorized by start time: 0000 – 0600: pre-peak 0600 – 0900: am-peak 0900 – 1500: midday 1500 – 1800: pm-peak 1800 – 0000: post-peak Data Preparation

5. Methods Delays scheduled arrival time – actual arrival time (in seconds after midnight) Stop performance “Marginal” doors open time: number of seconds it takes for each passenger to board or alight (over the amount of time it takes only one passenger to do so) Averaged for each stop Segment performance Seconds per foot: number of seconds between sequential stops divided by the segment length, converted to speed Averaged for each segment Computations

6. Methods Raw OBS data imported into GIS X,Y data extracted from GPS entries (generated point shapefile) Data screen: retained only those stops which did not occur at bus stops (retained only entries where STOP_ID = 0) Computed kernel density with DWELL_SEC as value field Reclassified output raster from 1 to 9, with 1 representing shortest stops / lowest number of stops Unplanned Stops

7. Worst Delays Southbound in West Seattle Southbound and Northbound Downtown Results Delays

8. Results Marginal on/off time consistently higher in D than C Correlated with passengers embarking and alighting Off board payment generally unused Relative Stop Performance

9. Results Worst performance: Northern and Southern endpoints of Rapid Ride Downtown segments Alaska Junction Relative Segment Performance

10. Results Averaged data reveals differences by time of day and by ridership Stops & Segments

11. Conclusions and Questions No correlation between physical attributes of stops and performance Ridership explains only 26% of doors open time More complex phenomena (traffic flows, signals) account for most variation Why does C Southbound accumulate large delays in West Seattle?

12. Questions University of Washington URBDP 422 Geospatial Analysis, Winter 2014 Debmalya Sinha, Austin Bell, Riley Smith, Andrew Brick