Content references source material no longer available at its original location.
Automate collection of parking occupancy data when traditional parking surveys lack resolution and sample size to support large-scale demand-responsive parking management systems.

San Francisco Municipal Transportation Agency’s (SFMTA) experience managing a large-scale demand-responsive parking system.

Date Posted
09/26/2017
TwitterLinkedInFacebook
Identifier
2016-L00738

Demand-Responsive Pricing on the Cheap: Estimating Parking Occupancy Using Meter Payment Data

Summary Information

Between 2011 and 2013 the San Francisco Municipal Transportation Agency (SFMTA) implemented the SFpark pilot project, which used in-ground sensors and new parking meters to collect real-time data so parking prices could be adjusted based on occupancy. The sensors had a short operational life. However, all parking meters were upgraded to those collecting payment data. Thus, while occupancy data is no longer available, payment data from all parking meters is still available. Sensor and meter data collected during the SFpark pilot were used to develop a sensor-independent rate adjustment (SIRA) model that estimates parking occupancy using meter payment data, which enables demand-responsive pricing policies without large-scale sensor installations.

Lessons Learned

Parking rate adjustment models eventually require recalibration, which requires payment and occupancy data. In San Francisco, payment rate information will continue to be provided by the smart parking meters. Occupancy rate information can be collected in a few ways. The SFMTA identified the strengths and weaknesses of four data collection approaches for the purposes of calibrating their SIRA model:

  • Manual occupancy surveys involve a surveyor visiting a block a few times on a handful of days. Data collection frequency and sample size are insufficient.
  • Temporary parking sensor deployment. The SFMTA used one year of continuous data from in-ground parking sensors to develop the initial model. Occupancy analysis found little variation by month, suggesting less time is needed. Rather than permanently embed sensors, they could be temporarily deployed in one area, then moved to another. Rotating deployments could provide the same rich data from sensors at a lower cost since fewer would be operation.
  • Time-lapse photos from pole mounted cameras processed by staff and custom software. The New York City Department of Transportation does this to gather occupancy data on a smaller scale. This strategy has humans identify spaces as vacant or occupied, but removes the need to deploy field surveyors, reduces data entry costs, and increases sampling frequency. While privacy concerns exist, it is an easy to deploy solution that can yield a richer occupancy dataset than traditional manual surveys for a small sample area.
  • If license plate recognition (LPR) units were deployed on a wider geographic scale at a consistent frequency, it may be possible to gather a sufficient sample of occupancy data. However, additional work would be needed to configure LPR systems to collect, process, and store parking occupancy data in a way compatible with parking data management systems.

The SFpark pilot demonstrated that demand-responsive parking pricing produces many of its intended benefits. However, measuring parking occupancy is critical to implementing these policies and there is no implementation-ready technology that cities can adopt to collect detailed occupancy data at a large scale in a cost-effective manner at this point.

Goal Areas
System Engineering Elements

Keywords Taxonomy: