6.1. Lessons Learned

6.1.1. Lessons Learned: Overall Project Success and Countermeasures

The project was successful in demonstrating the ability of a local government/university team to develop a data-based plan to improve pedestrian safety, focusing on higher-injury areas, and then to implement and evaluate this plan. The project catalyzed San Francisco’s consideration and use of a number of innovative, generally lower-cost countermeasures. It also provided an opportunity for the San Francisco team to learn more about best practices in pedestrian safety from FHWA and other grantees. There are numerous off-the-shelf materials and references that are directly useful (much of it FHWA-produced such as the Pedestrian Safety Campaign Planner, the Walkable Community brochure, and numerous research reports). The lessons from this project will prove very useful in numerous future projects, such as the San Francisco Better Streets Plan.

Because San Francisco had such a extensive pedestrian safety program even before PedSafe was initiated, there was limited room for the project to catalyze major citywide changes or to achieve high visibility, especially considering the project budget. For example, even before PedSafe, San Francisco assessed pedestrian injury “hot spots” and trends. Several San Francisco agencies devoted significant staff to pedestrian safety planning, engineering, education, and enforcement. Inter-agency and public advisory committees were formed outside of the project.

The federal funding (about $680,000 or $120,000 per year) was extremely helpful and appreciated. However, on a per-year or per-intersection basis it was fairly limited, even compared to some other funding sources used for pedestrian and traffic safety. San Francisco has a sales tax dedicated to transportation uses that alone provides roughly $840,000 annually for “pedestrian circulation and safety.” The separate “traffic calming” allocation from this sales tax is around $2 million annually. Typical State of California grants for design and construction of pedestrian improvements (e.g., for improvements within four blocks of a specific elementary school or rail transit station) often range from $300,000 to $800,000, with minimal requirements for data collection and evaluation.

The primary lessons learned from the project about countermeasures include the following:

• There is a wide range of pedestrian safety countermeasures available that can be tailored to specific location characteristics. A package of such measures can reduce vehicle/pedestrian conflicts, increase driver yielding, and bring about other changes in driver and pedestrian behavior that should, over the long term, decrease the number of pedestrian injuries. However, due to the project scheduling, it was not possible to directly observe the impacts on pedestrian-involved crashes, which would require several years of post-installation data to have meaningful results.

• Of course, pedestrian safety is heavily affected by a wide range of factors beyond the control of a project such as PedSafe. (These include such factors as: driver licensing and other traffic law changes; Police enforcement activities; employment and other economic activity fluctuations; vehicle safety improvements (e.g., braking abilities), and other safety efforts (such as red light photo-enforcement programs). The ability of a project with the budget the size of this PedSafe effort to have a major citywide impact is limited, although it can certainly catalyze significant changes.

• Particularly cost-effective countermeasures appear to be the in-pavement “Yield To Pedestrians” (YTP) signs and pedestrian countdown signals. (The pedestrian countdown signals, installed citywide in San Francisco, not only appear effective in aiding pedestrians in safer crossing, but also have some value in warning drivers of approach as the green light is about to change.) The impactable YTP signs were effective and relatively inexpensive, but susceptible to damage when not installed on raised islands.

• By contrast, the “LOOK” pavement stencils seemed to have negligible value.

• Low-cost but effective measures have the advantages of quick implementation and the potential to draw support and funding for further improvements.

• Flashing beacons and in-pavement crosswalk lights both appeared effective at inducing drivers to yield to pedestrians at uncontrolled crosswalks. (The Mission & Santa Rosa flashing beacons are being replaced by a conventional traffic/pedestrian signal as a potentially stronger degree of protection for pedestrians and especially for turning vehicles.)

• The Portable Changeable Message Speed Limit Sign was more effective than the fixed speed display sign at reducing driver speeds. This may be due to drivers “tuning out” the fixed sign weeks or months after it is installed, in addition to the subliminal association of the trailer with police enforcement.

• Pedestrian Scramble phasing is potentially quite effective for certain situations (e.g., smaller intersections with heavy volumes of turning vehicles and pedestrians), but can be difficult to use in some situations (e.g., wide intersections with heavy through traffic volumes, including transit service).

• Pedestrian Head Starts had mixed results. There were substantial reductions in the number of vehicles turning in front of pedestrians at three of four intersections. However, these changes did not lead to a significant reduction in vehicle/pedestrian conflicts.

• Video detection of pedestrians to extend crossing time appeared to be a promising technology, but needs further testing and refinement.

• Infrared detection of pedestrians to trigger beacons or in-pavement lighting has been more effective in San Francisco than overhead microwave detection.

• Accessible (Audible) Pedestrian Signals (APS) were helpful to both sighted pedestrians as well as visually impaired pedestrians.

• Pedestrians appeared to appreciate most countermeasures, but showed minimal awareness of which devices were installed. There were essentially no common suggestions by pedestrians surveyed for improving intersection safety; suggestions varied widely.

• It was not possible to make conclusions about how the overall effectiveness of the countermeasures varied by neighborhood or pedestrian characteristics (e.g., age group). This was primarily because the number of installations for each countermeasure was limited. However, it is likely that such factors were important. For example, the low level of use of the push button to actuate the flashing beacons at 16^th & Capp may have been partly due to the generally younger age profile of pedestrians, the relatively narrow crossing, and the extremely lively (distracting) street activity in this area. (However, this finding was also generally consistent with national research.)

6.1.2. Lessons Learned: Outreach

• Device-specific safety instruction was typically not necessary, as most devices were intuitive or even invisible to pedestrians.

• To reach the maximum audience, translate outreach messages into multiple languages.

6.1.3. Lessons Learned: Implementation

• Coordinating improvements with other agencies, especially those involved in street construction, is crucial. (In San Francisco, street reconstruction “erased” previous pedestrian safety measures, such as advanced limit lines and ladder style crosswalks.)

• Developing and implementing a comprehensive pedestrian safety plan requires a long time frame. The San Francisco project took nearly six years, including almost two years for planning, two years for design/procurement/approvals, and two years for implementation and evaluation. (In San Francisco, it generally takes 2-3 years to install pedestrian signals at a typical location, from funding commitment through design and construction, usually as part of a large signal contract.)

• It is advantageous to have full-time dedicated pedestrian safety planning and engineering staff. San Francisco MTA Pedestrian Program staff now includes three full-time positions, two of which are filled by engineers and the third by a planner.

• Institutional issues proved challenging. For example, a proposed pedestrian scramble installation needed to be deferred due to concern about potential impacts on public transit schedules. Street lighting improvements also required the active participation of a non-transportation department.

6.1.4. Lessons Learned: Data Collection and Analysis

• Video data collection had the advantage of allowing repeated viewings and precise time stamping of events (such as pedestrian wait time duration). However, the labor requirements for tabulating video recorded events were several times greater than for manual data collection. The field of vision was also often more restricted than optimal, possibly obscuring important vehicle actions. The amount of information provided by video footage (particularly after conversion into a format that is compatible with the playback software) is far less than a live observer could collect, making it more difficult, for example, to determine whether there was a vehicle/pedestrian conflict.

  Video data collection also required close communication between the Police, SFMTA, and the data collection organization. Device installation and video observation dates needed to be closely coordinated, in part, because the video van required a particular parking space and a special permit to ensure the availability of the space. In one case, the radar speed trailer was towed away early in the data collection period by the Police Department.

  While, in theory, video footage could be analyzed for vehicle speeds, it was quite difficult to do this with the necessary precision. There were occasional problems with converted video recordings skipping frames or running at inconsistent speeds, making sensitive timing analyses difficult.

• Clear, consistent definitions of the pedestrian/driver behavior measures of effectiveness (MOEs) are needed. There are no accepted, universal definitions of five key project MOEs:

  • Vehicle/Pedestrian Interaction
  • Vehicle Yielding
  • Vehicle/Pedestrian Conflict
  • Vehicle Blocking Crosswalk
  • Pedestrian Trapped In Median

  The vehicle/pedestrian conflict was particularly challenging. The subjective definition used stated that a conflict happened if the driver swerved or braked quickly (not smoothly) or if the pedestrian changed stride or gait, apparently showing concern about an imminent collision. This was a less restrictive definition than a “near miss” as it did not require the pedestrian and vehicle to come “within inches” nor require the driver to “slam on the brakes.” However, it would not include the routine, smooth changes in speed that drivers and pedestrians make frequently as they to yield to avoid a collision.

  Thus defined, the occurrence of a conflict was rare enough that it was quite difficult to detect an adequate number of conflicts during baseline observations to make a statistically significant improvement possible. A strict definition can make statistical significance impossible to achieve even with a reasonable sample size.

• The earlier version of PBCAT used for Phase I analysis proved difficult to use. In particular, the classification of each pedestrian collision by a single list of key factors proved overly simplistic. These key factors forced a decision on whether the most salient feature was the violation type or the vehicle movement. Statistical software and the Crossroads ™ software proved more flexible and helpful. The SFMTA has also used GIS mapping analysis of pedestrian crashes for over six years, and this has been helpful. Often, the actual police collision reports had to be reviewed to understand the problems specific to an intersection (such as precise vehicle movements involved prior to collisions).

• While analysis of crash patterns is quite helpful in selecting the proper treatment, several years of crash data are needed, and even then, patterns at the same intersection may vary significantly year-to-year. Site visits are therefore very important.

• Crash analysis should consider the pedestrian and/or vehicle volumes as a measure of exposure, rather than only the absolute number of injuries or crashes.

6.2. Comparison With Other Cities

Similar projects were carried out in the Las Vegas and Miami metropolitan areas. A preliminary final report was available for Miami, but not for Las Vegas as of publication time. The chief findings in Miami that related directly to the San Francisco experience include:

• Pedestrian push button confirmation light: a significantly higher proportion of pedestrians waited to cross until the Walk signal. (This result from an additional visual cue was similar to the impacts of audible signals on sighted pedestrians.)

• Reduced waiting times at a mid-block signalized crossing resulted in significantly better compliance. (San Francisco has favored shorter traffic signal cycles, usually 60 to 80 seconds, in part for this reason.)

• Pedestrian Head Start (leading pedestrian interval) did not significantly improve yielding by right-turning drivers. (The effectiveness of the Pedestrian Head Starts in San Francisco appeared to vary with the intersection characteristics.)

• Impactable Yield To Pedestrians signs: these signs were highly effective at increasing driver yielding, but were easily damaged (which was very similar to the San Francisco experience).

6.3. Next Steps and Further Research

The primary opportunity for additional research would be an evaluation of the actual pedestrian injury impacts of the countermeasures. This would require follow up observations 3-6 years after device installation. (Collision data are not even available in tabulated form in California for at least six months. In order to obtain a meaningful sample size, several years of post-installation data are needed.) SFMTA participated in a proposal by the San Francisco Injury Center to the federal Centers for Disease Control for funding to perform this analysis, but the application was turned down due to limited funding.

SFMTA is also interested in testing other promising devices, particularly the HAWK beacon system and/or stutter flash beacons. (The High-Intensity Activated Crosswalk system is dark until a pedestrian activates the push button. It then switches from flashing yellow to solid yellow, followed by solid red, then flashing red, allowing cross traffic to continue once the pedestrian has crossed.) The City Traffic Engineer has identified the value of converting Stop controls to traffic signals that would be subject to transit and pedestrian priority.

Some of the countermeasures could be refined. For example, red turn arrows could be added to pedestrian head start locations. The video detection logic could be adjusted to prevent vehicles encroaching on the crosswalk from triggering the signal extension.

The cost-effectiveness of innovative devices could be compared to that of more traditional traffic engineering improvements, such as improved roadway lighting and left turn signalization.

It would also be valuable to research how the findings from this study can be translated into citywide pedestrian plans. San Francisco is currently developing a Better Streets Plan, a combination of the planned Pedestrian Master Plan with a citywide streetscape plan. This may provide an opportunity to apply the findings of this project.