U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-06-125
Date: November 2006
Pedestrian and Bicyclist Intersection Safety Indices
PDF Version (2.25 MB)
PDF files can be viewed with the Acrobat® Reader®
CHAPTER 2. LITERATURE REVIEW
A number of studies and rating methodologies related to the safety of pedestrians and bicyclists have been conducted in recent years. A few studies have incorporated crash analyses to determine factors related to the risk level of pedestrians and bicyclists. Many others are primarily intended to indicate a compatibility level for pedestrians or bicyclists, also called “level of service” or “comfort level.” Compatibility refers to the characteristics of a road or intersection that make it attractive to pedestrian and bicyclist users. The studies listed below are separated into sections on:
Since there are advantages to both of these types of methodology, there is a need to develop a safety–rating method that incorporates a variety of subjective user ratings, as well as more objective safety data such as evasive actions and crashes. Such a methodology would provide opportunities for State and local agencies to have a pro–active intersection rating tool; that is, they would be able to apply the “safety rating tool” to a large sample of intersections to identify sites with the greatest need for assessment. Thus, agencies could be pro–active in their approach without having to wait until pedestrian or bicyclist collisions occur before making the necessary improvements. The Ped ISI and Bike ISI developed in this research are intended to meet this need for a proactive approach.
Botma (1995) proposed level of service (LOS) methodologies for bicycle paths and bicycle–pedestrian paths. Both methodologies defined LOS in terms of events: an event occurs when one user passes another user traveling in the same direction, or when one user encounters another user traveling in the opposite direction. As the number of users on a path increases, more events occur, or equivalently, more users experience hindrance from other users. As events become more frequent, the LOS deteriorates from A to F. This methodology addresses bicyclist (and pedestrian) crowding as reflected by passings and meetings on paths. It does not cover bicyclists’ perceived comfort and safety while riding in a motor vehicle environment (i.e., on the roadway).
Chapter 19 of the Highway Capacity Manual (2000) adopts Botma’s (1995) LOS methodology for exclusive and shared paths. Procedures are given for additional facility types. The LOS for on–street bicycle lanes is also dependent on the number of events, which vary according to the bicycle flow rate, mean speed, and standard deviation of the speed. At signalized and stop–controlled (on the minor street only, not all–way stop) intersections, the LOS depends on control delay. As delay length increases, the LOS deteriorates from A to F. For bicycle lanes on urban streets (intersections plus segments), the LOS depends on average bicyclist speeds.
Several models have been developed to relate roadway geometrics and operational characteristics to bicyclists’ perceived levels of comfort and safety (i.e., to measure bicycle compatibility). Because older models served as the starting point for newer models, this section is presented chronologically.
The Bicycle Safety Index Rating (BSIR) consists of two submodels, one for roadway segments and one for intersections (Davis, 1987). The safety of roadway segments depends on traffic volume, speed limit, outside lane width, pavement condition, and a variety of geometric factors. The safety of intersections is a function of traffic volume, type of signalization, and several geometric factors. BSIR values from 0 to 4 denote roadways that are extremely favorable for safe bicycle operation. On the other hand, roadways with BSIR values of 6 or above are questionable for bicycle operation. Despite its name, the BSIR does not incorporate any information about motor vehicle–bicycle crashes or conflicts.
In Broward County, FL, the BSIR was modified by placing greater weight on vehicle speeds and less weight on traffic volumes. The new model was called the roadway condition index (RCI) (Epperson, 1994). The RCI was then modified by placing less weight on pavement and location factors and by increasing the interaction between curb–lane width, speed limit, and traffic volume. The modified RCI was applied in Dade County, FL, as part of a multimodal evaluation of the county’s transportation network.
Sorton and Walsh (1994) determined bicyclist stress levels as a function of three primary variables—peak–hour traffic volume in the curb lane, motor vehicle speeds in the curb lane, and curb–lane width. Secondary variables such as the number of commercial driveways were acknowledged, but were not included in the analysis because of funding limitations. Stress levels ranging from 1 (very low) to 5 (very high) were defined for values of each primary variable. For example, stress level 1 corresponds to a traffic volume of 50 or fewer vehicles per hour, 85th percentile speeds of 40 kilometers per hour (km/h) or lower, and a curb–lane width of at least 4.6 meters (m).
The Intersection Hazard Score (IHS) was based on the RCI and other earlier models (Landis, 1994). It measures the level of hazard that bicyclists are likely to perceive while riding. The variables in this model included traffic volume, speed limit, outside lane width, pavement condition, and number of driveways. Despite its name, the IHS does not incorporate any information about crashes or conflicts.
A Bicycle Level of Service (BLOS) model for roadway segments was developed by having bicyclists ride selected roadway segments on a real–life course and provide comfort/safety ratings on a scale of A through F (Landis, Vattikuti, and Brannick, 1997). The presence of a stripe separating the motor vehicle and bicycle areas of an outside travel lane resulted in the perception of a safer condition than an outside travel lane of the same width, but without delineated motor vehicle and bicycle areas. The BLOS has many of the same variables as the IHS. The major difference is the inclusion of pavement condition as a variable in the BLOS, but not in the IHS. The BLOS also requires more detailed land–use information than the IHS.
Harkey, et al., developed a Bicycle Compatibility Index (BCI) for urban and suburban roadways at midblock locations (Harkey, Reinfurt, Knuiman, Stewart, and Sorton, 1998). Bicyclists watched a videotape of various roadway segments and provided ratings of how comfortable they would feel riding on each segment. The BCI was developed from those ratings. It incorporates variables that pertain to the “bicycle friendliness” of a roadway for an adult bicyclist. Examples of these variables are curb–lane width, traffic volume, and vehicle speeds. Many of these variables are also used in the BLOS. Unlike the BLOS, the BCI does not include pavement condition because pavement condition data would not be readily available. A key difference between the BCI and the BLOS is that the BLOS relied on bicyclists actually riding on the roadway, so their ratings pertain to how comfortable they actually felt. The video approach used to develop the BCI does not put bicyclists at risk and allows for a greater range of geometric and operating conditions than would be feasible on a real–life course. To verify the validity of this approach, a pilot study was conducted to compare bicyclists’ ratings in the field versus their ratings from watching the video. The pilot study found that there was a reasonably good match between the two types of ratings.
The BCI values were then translated into bicycle level of service (LOS) designations (not to be confused with the BLOS model described above). LOS A (corresponding to a BCI < 1.50) indicates that a roadway is extremely compatible with (or comfortable for) an average adult bicyclist. At the opposite extreme, LOS F (corresponding to a BCI > 5.30) indicates that a roadway is extremely incompatible (or uncomfortable) for an average adult bicyclist.
Landis, et al., built upon the segment BLOS (Landis, et al., 1997) to develop an intersection BLOS (Landis, Vattikuti, Ottenburg, Petritsch, Guttenplan, and Crider, 2003). Data were obtained from bicyclists who rode through selected intersections and provided comfort/safety ratings on a scale of A through F. Roadway traffic volume, total width of the outside through lane, and the intersection crossing distance were found to be the primary factors influencing bicyclists’ safety and comfort at intersections. The presence of a bike lane or paved shoulder stripe was not as important as it was in the BLOS for segments.
A Compatibility of Roads for Cyclists (CRC) index was created to evaluate routes in rural and urban fringe areas (Noël, Leclerc, and Lee–Gosselin, 2003). To develop the index, the authors surveyed cyclists to obtain: (1) their ratings of roadway segments, and (2) their perceptions of factors that affect the safety and comfort of cyclists. According to the survey results, cycling space and automobile speed received the greatest weights (30 and 20 out of a possible 100, respectively) in the index. Other index components are paved shoulders, automobile and truck traffic flows, sand/gravel/abundant vegetation, ditches, retail/industrial/residential entrances, curves and grades, and major junctions.
Hunter, Stewart, and Stutts studied the differences between bike lanes and wide curb lanes (Hunter, et al., 1999). They observed videotapes of nearly 4,600 bicyclists and evaluated operational characteristics and interactions between bicyclists and motorists. They found that bicyclist wrong–way riding and sidewalk riding were more common at wide curb lane sites. Also, traffic encroachment in adjacent lanes because of passing bicyclists was more common for wide curb lane sites. There was little difference between the types of bicycle facilities in the number or severity of the bicyclist–motorist conflicts observed. Overall, they concluded that the type of bicycle facility had much less impact on operations and safety than other site characteristics and recommended that both bike lanes and wide curb lanes be used to improve riding conditions for bicyclists.
The bicycle compatibility models reviewed here all relate various roadway and traffic characteristics with how comfortable bicyclists would feel riding along those roadway segments. Variables such as traffic volume and lane width were common to all of the models. The weights assigned to each variable differed among the models. Most of the data required by these models can be obtained easily. Some degree of subjectivity is involved in assigning values for the adjustment factors for pavement, location, etc. A greater degree of subjectivity is involved in classifying roads as being “good” or “bad” for bicycling on the basis of their BCI or other index ratings.
Most of the models described above are applicable to roadway segments (i.e., midblock locations). Several have an intersection component (BSIR, CRC Index, and intersection BLOS). None of the models incorporate information about crashes and conflicts. It is acknowledged that many locations have few or no crashes per year, so crashes would not be readily modeled. The collection of conflict data requires an intensive field effort, and few local traffic agencies have the staff resources to do so.
A logical next step would be to develop a model that incorporates information on the number and severity of motor vehicle–bicycle crashes, as well as conflicts and avoidance maneuvers, to roadway and traffic variables. Such a model would require exposure information for both vehicles and bicycles. Bicycle coordinators and traffic engineers could use such a model to establish priorities for needed intersection improvements where bicycle safety is a problem.
BICYCLE CRASH ANALYSES
Hunter, et al., performed a detailed analysis of 3,000 bicycle–motor vehicle crashes in California, Florida, Maryland, Minnesota, North Carolina, and Utah. Almost three–fourths of these crashes occurred at intersections, driveways, or other junctions (Hunter, Stutts, Pein, and Cox, 1996). Sixty percent of the crashes occurred on two–lane roads. Twenty–six percent occurred on roads with an outside lane width of less than 3.6 m (12 feet (ft)). Slightly more than three–fourths of the crashes occurred on roads with speed limits of 56 km/h (35 miles per hour (mi/h)) or less. Roads with narrower lanes and roads with higher speed limits were associated with more than their share of serious and fatal injuries to bicyclists.
The bicyclist and motorist were on parallel paths in 36 percent of the 3,000 crashes (Hunter, Pein, and Stutts, 1995). In another 57 percent of the 3,000 crashes, they were on crossing paths. Parallel–path crashes were most frequent when the motorist turned or merged into the bicyclist’s path (34 percent of the parallel–path crashes) and when the motorist overtook the bicyclist (24 percent). Crossing path crashes were most frequent when the motorist failed to yield (38 percent of the crossing path crashes) and when the bicyclist failed to yield at an intersection (29 percent).
Wang and Mihan (2004) modeled bicycle–motor vehicle crashes at 115 signalized intersections in Tokyo, Japan. They classified crashes as BMV–1 (collisions between bicycles and through motor vehicles), BMV–2 (collisions between bicycles and left–turning motor vehicles), and BMV–3 (collisions between bicycles and right–turning motor vehicles). They then estimated the expected crash risk by developing negative binomial models for each crash type. The models contained different sets of explanatory variables, including traffic and bicyclist volume, intersection location, visual noise, pedestrian overbridges, and median width.
Before countermeasures to reduce bicycle (and pedestrian) crashes can be selected, an understanding of the events leading to these crashes is required. This process of determining the pre–crash actions is referred to as crash typing. The Pedestrian and Bicycle Crash Analysis Tool (PBCAT) is a software product intended to assist practitioners with improving bicycling and walking safety (Harkey, Mekemson, Chen, and Krull, 1999). PBCAT may be used to develop and analyze a database containing the crash types and other details of crashes between motor vehicles and bicyclists or pedestrians. The user can then access the countermeasure module to see what engineering, education, and enforcement treatments are appropriate.
Once bicycle crashes are crash–typed, appropriate countermeasures may be examined. BIKESAFE is an expert system that is currently being developed by the University of North Carolina HSRC as a counterpart to PEDSAFE (Hunter, Thomas, and Stutts, 2005). This system will provide users with information on how to improve bicyclist safety and mobility, with specific focus on crash types. BIKESAFE will be available on CD–ROM and online at www.walkinginfo.org/bikesafe. The online tools consist of a selection tool, interactive matrices, 50 countermeasure descriptions, and more than 50 case studies. With the selection tool, the user first selects either a performance objective or a prevalent crash type. Next, the user enters site characteristics. The expert system then develops a list of countermeasures that are appropriate for the situation. The user can read descriptions of each countermeasure and case studies in cities that have implemented the countermeasure. The interactive matrices allow the user to see at a glance which countermeasures are suitable to achieve each of 7 performance objectives or to address each of 13 crash types. BIKESAFE also contains information on understanding bicyclist crashes, implementing countermeasures, and creating a bicycling environment.
Chapter 18 of the Highway Capacity Manual (2000) defines pedestrian LOS criteria for signalized and unsignalized intersections. These criteria are expressed in terms of delay (while pedestrians are waiting to cross the street) and space (at street corners and in crosswalks). The criteria include factors such as pedestrian volumes, crosswalk length and width, and cycle lengths. However, the criteria do not take into account actual or perceived safety and, therefore, do not incorporate other factors, such as crossing width or the number of turning vehicles.
Several authors have gone beyond the volume and capacity approach in the Highway Capacity Manual to include qualitative measures of pedestrian LOS. For example, Sarkar (1993) defined six pedestrian service levels. This qualitative scheme relied on subjective ratings of safety, security, comfort and convenience, continuity, system coherence, and attractiveness. Service Level A represents the most strongly pedestrian–oriented environments; the right–of–way is reserved exclusively for pedestrians. At the opposite extreme, pedestrian needs are totally disregarded under Service Level F.
Khisty (1994) proposed seven qualitative performance measures of pedestrian environments: attractiveness, comfort, convenience, safety, security, system coherence, and system continuity. The relative importance of each measure was determined from survey responses; security and safety were found to be the most important. Survey respondents also rated walking routes by assigning scores to these measures, on a scale of 0 (the worst, corresponding to LOS = F) to 5 (the best, LOS = A) according to their level of satisfaction. The overall score, and therefore LOS, of each walking route was the weighted average of the scores for the individual measures. The measures were not proposed specifically for intersections; the safety measure is perhaps the most relevant to intersections.
Nine evaluation measures (encompassing aesthetics, safety, and ease of movement) were used to analyze commercial areas and corridors in Winter Park, FL (Jaskiewicz, 1999). Each measure was scored from 1 (very poor) to 5 (excellent). The scores were averaged to obtain an overall LOS. Based on the analysis, specific pedestrian deficiencies were identified. Both short–term physical improvements and long–term design and policy solutions were recommended. This LOS approach does not address intersections directly; however, the physical components/condition measure includes one or more treatments at pedestrian crossings as a means of reducing vehicle speeds.
A number of researchers have developed models to measure the compatibility of roads for walking. These models relate geometric and operational features to pedestrian compatibility. Thus, data on lane widths, traffic volumes, and other features are needed to use these models. The text below describes several models.
The pedestrian environment factor model used in Portland, OR, includes four elements: (1) sidewalks, (2) ease of crossing streets, (3) street and sidewalk connectivity, and (4) terrain (1,000 Friends of Oregon, 1993). Taken together, these elements characterize the pedestrian friendliness of an area. Each element is scored on a 3–point scale and is equally weighted, so the pedestrian environment factor ranged from 4 points (lowest) to 12 points (highest). The advantage of the pedestrian environment factor is that engineers and planners can easily score a specific zone and see how pedestrian–friendly it is.
The Portland Pedestrian Master Plan describes two tools to prioritize pedestrian projects: (1) the Pedestrian Potential Index, and (2) the Deficiency Index (City of Portland, 1998). The Pedestrian Potential Index measures the strength of policy, proximity, and environmental factors that favor walking, whereas the Deficiency Index measures conditions such as missing sidewalks, difficult and dangerous street crossings, and lack of a connected street network. Difficult and dangerous street crossings were approximated by traffic speed, traffic volume, roadway width, and locations with motor vehicle–pedestrian crashes. The two indices can be used to identify areas where pedestrian facility improvements are most needed. The advantage of the Deficiency Index is that it relies on traffic, roadway, and crash data. These data are generally available, so engineers and planners can easily calculate deficiency indices and determine where improvements are most needed.
Dixon (1995) determined the pedestrian LOS for roadway segments by using facility continuity, conflicts, motor vehicle LOS, and other factors. An overall corridor score can be computed from the sum of the segment scores, adjusted for the lengths of each segment relative to the corridor length. The method was tested on five arterial roads and one collector road in Gainesville, FL, which resulted in LOS ratings of C, D, and E.
A more recent model defines pedestrian LOS as a function of outside lane width, shoulder or bike lane width, on–street parking, the planting strip, sidewalk presence and width, motor vehicle traffic volume and speed, and the total number of through lanes (Landis, Vattikuti, Ottenburg, McLeod, and Guttenplan, 2001). A roadway segment can be given a LOS rating ranging from A (best, when pedestrian LOS < 1.5) to F (worst, pedestrian LOS > 5.5). This model does not include intersections.
From the pedestrian’s perspective, the maximum tolerable speeds of passing cars on three residential streets ranged from 51 to 58 km/h (32 to 36 mi/h) (Warren and Rousseau, 2002). These speeds were almost identical to the observed 85th percentile speeds. Most study participants judged speeds of up to 40 km/h (25 mi/h) to be reasonably or completely acceptable. They tolerated higher speeds 5 to k km/h (3 to 4 mi/h higher) when a wider planting strip or a greater street width was present, as these conditions placed them further away from moving traffic. Although limited in scope, this study gives useful information on pedestrian comfort levels with regard to speed and separation from traffic.
Gallin (2001) determined the pedestrian LOS by scoring and weighting a total of 11 design, location, and user factors. Integer scores of 0 to 4 are given to each factor, and the weights range from 2 to 5. For example, the “path width” factor is scored as 0 if no pedestrian path is present, 1 if the path width is 0 to 1 m, and up to a maximum of 4 if the path width is more than 2 m. Some factors are scored subjectively (such as “connectivity,” which is 4 points if excellent, 3 points if good, etc.). Intersections and driveways are counted to assess the “potential for vehicle conflict” factor. The LOS ranges from A (ideal pedestrian conditions, total weighted score of 132 or higher) to E (unsuitable pedestrian conditions, total weighted score of 36 or lower).
A pedestrian LOS was developed for midblock crossings (Chu and Baltes, 2001; Baltes and Chu, 2002). Study participants observed midblock crossings for 3 minutes (min) and rated how difficult it would be for them to cross, on a scale of A to F. However, the participants did not actually cross streets, so their ratings pertain to how difficult it would be for them to cross, not how difficult it was for them to cross. The authors fitted a linear regression model using the ratings, geometric data, and operational data. It contained 15 variables related to traffic volumes, turning volumes, pedestrian age, vehicle speed, crossing width, presence of pedestrian signal, cycle length, and signal spacing.
A recent study in Sarasota, FL, made use of a large “Walk for Science” event to gather data from approximately 800 pedestrian participants on their perceived safety, exposure, and delay at intersection crossings (Petritsch, Landis, McLeod, Huang, and Challa, 2005). The resulting pedestrian LOS model had primary factors of right–turn–on–red volumes for the street being crossed, permissive left turns from the street parallel to the crosswalk, motor vehicle volume on the street being crossed, midblock 85th percentile speed of the vehicles on the street being crossed, the number of lanes being crossed, the pedestrian’s delay, and the presence or absence of right–turn channelization islands.
When considering pedestrian facility compatibility, it should be noted that a high level of service (i.e., LOS A) does not necessarily indicate a safe or well–designed sidewalk or pedestrian facility. There may be few pedestrians using the facility, thereby producing a high level of service, but there may be negative design features that cause pedestrians to avoid the location.
There is still a need for research to understand pedestrian exposure and people’s choices about where they walk.
PEDESTRIAN CRASH ANALYSES
A detailed analysis of 5,000 pedestrian–motor vehicle crashes in 6 States revealed that about one–half of these crashes occurred at either intersections or driveways (Hunter, Stutts, Pein, and Cox, 1996). Nearly 60 percent of the crashes occurred on two–lane roads. Almost three–fourths of the crashes occurred on roads with speed limits of 56 km/h (35 mi/h) or less. Serious and fatal injuries to pedestrians were directly proportional to the speed limit and number of lanes. Marked crosswalks were present in about 21 percent of crashes and pedestrian signals in about 7 percent. A sidewalk was present on at least one side in about 17 percent of the non–intersection crashes.
More than 44,000 pedestrian–motor vehicle crashes were reported in Florida from 1990 through 1994 (Baltes, 1998). With respect to age, pedestrians from ages 65 to 74 were at the greatest risk of being involved in a crash. They were also at the greatest risk of being injured or killed once involved in a crash. Pedestrians under age 19 were overrepresented in crashes while crossing not at an intersection, crossing at a midblock crosswalk, crossing at an intersection, and standing/playing in the roadway. Pedestrians from ages 25 to 34 were overrepresented in crashes while working on a vehicle in the road and while working in the road at other activities.
A study of motor vehicle–pedestrian crashes at signal–controlled urban intersections found that several operational variables were significant factors (Zegeer, Opiela, and Cynecki, 1985). Analysis indicated that pedestrian volume is the most important variable, followed by traffic volume. Each of these two variables showed a significant and positive relationship with the number of pedestrian crashes. After controlling for other factors, other variables that were overrepresented in pedestrian crash risk included two–way streets (compared to one–way), residential area types, wider streets, the presence of bus operations, and higher volumes of turning vehicles. Exclusive pedestrian signal timing was associated with a significantly lower pedestrian crash experience compared to concurrent timing at signalized intersections without pedestrian signals.
Another study examined the effects of marked versus unmarked crosswalks at unsignalized intersections, along with other factors, on the number of pedestrian crashes (Zegeer, Stewart, Huang, and Lagerwey, 2001). Traffic and roadway factors found to be related to a higher number of pedestrian crashes included higher pedestrian volumes, higher traffic volumes, and greater number of lanes. After controlling for other factors, speed limit was not significantly related to pedestrian crash frequency. The presence of a raised median (or raised crossing island) was associated with a significantly lower pedestrian crash risk on multi–lane roads.
Comparing marked versus unmarked crosswalks, there were no significant differences in pedestrian crash risk on two–lane roads. There were also no differences in crash risk for sites with or without marked crosswalks on multi–lane roads with traffic volumes of less than 12,000 vehicles per day. On multi–lane roads without raised medians and traffic volumes greater than 12,000 vehicles per day, locations with marked crosswalks had a higher pedestrian crash risk than locations with unmarked crosswalks. On multi–lane roads with raised medians and traffic volumes greater than15,000 vehicles per day, pedestrian crash risk was higher at marked crosswalks than at unmarked crosswalks.
Many potential countermeasures were recommended to improve pedestrian safety related to crossing streets, instead of merely adding or removing a marked crosswalk. Improvements on multi–lane roads include adding pedestrian traffic signals (if warranted), installing raised medians or crossing islands, improving nighttime lighting, providing curb extensions, providing tighter intersection turning radii (to shorten crossing distances and lower the speeds of right–turning motorists), reducing the number of lanes, and/or providing advance stop lines (to improve sight distance between motorists and pedestrians in crosswalks). Recommended improvements on two–lane roads include narrowing travel lanes, removing parking near the intersection, improving lighting, adding signals (where warranted), and providing traffic–calming measures (on residential streets). Improved education and enforcement were also suggested to reduce certain types of pedestrian crashes.
A 2003 study evaluated the effect of a combination of intersection improvements on pedestrian crashes. A four–lane suburban roadway in central New Jersey was reconstructed to include redesigned intersections, a raised median, a narrower roadway width, re–timed signals, bike lanes, and sidewalks (King, Carnegie, and Ewing, 2003). The reconstruction resulted in a slight decline in 85th percentile vehicle speeds of 3 km/h (2 mi/h). Pedestrian exposure risk decreased by 28 percent. The effect on vehicle volumes was negligible. Using crash data from a 29–month period prior to reconstruction and previous research findings on crashes and speed, the authors projected that there would be four fewer vehicle–vehicle crashes per year. The reduction in crashes would result in a savings of $1.7 million over 3 years in crash–related costs. The annual number of crashes involving bicyclists and pedestrians was projected to remain the same.
The Pedestrian Facilities User Guide—Providing Safety and Mobility identifies which pedestrian–related facility improvements are expected to reduce pedestrian crashes for various crash types and roadway situations (Zegeer, Seiderman, Lagerwey, Cynecki, Ronkin, and Schneider, 2002). The User Guide also provides details of 48 different engineering improvements, including their purpose, the conditions when they are appropriate for use, considerations for use, and implementation costs. In addition, the countermeasure module of PBCAT shows the user details on which treatments are applicable to specific types of crashes (Harkey, Mekemson, Chen, and Krull, 1999). Pedestrian safety improvements from the User Guide and PBCAT will be adapted and expanded for application to intersection hazards.
The User Guide was updated and integrated into an expert system known as PEDSAFE (Harkey and Zegeer, 2004). This system provides users with information on how to improve pedestrian safety and mobility. PEDSAFE is available on CD–ROM and online at www.walkinginfo.org/pedsafe (accessed July 2005). The online tools consist of a selection tool, interactive matrices, 49 countermeasure descriptions, and 71 case studies of completed pedestrian safety improvements. With the selection tool, the user first selects either a performance objective or a prevalent crash type. Next, the user enters site characteristics. The expert system then develops a list of countermeasures that are appropriate for the situation. The user can read descriptions of each countermeasure and case studies in cities that have implemented the countermeasure. The interactive matrices allow the user to see at a glance which countermeasures are suitable to achieve each of 8 performance objectives or to address each of 12 crash types. PEDSAFE also contains information on understanding pedestrian crashes, implementing countermeasures, and creating a pedestrian environment.