The approach used in developing the BCI was to obtain the perspectives of bicyclists by having them view numerous roadway segments captured on videotape and rate these segments with respect to how comfortable they would be riding there under the geometric and operational conditions shown. The reliability of the results obtained using this video technique of data collection with respect to reflecting on-street comfort levels was validated in a pilot study. The procedure offered several advantages over other forms of data collection, including minimizing the risk to bicyclists, maximizing the range of roadway conditions to which the bicyclists could be exposed, and controlling the variables evaluated by the bicyclists.
It is important to note again that the BCI model developed is for midblock street segments only and is primarily intended for use on "through" streets. In other words, the ratings do not account for major intersections along the route where the bicyclist may encounter a stop sign or traffic signal. Within the research study, the video technique described above was piloted for a limited number of intersection sites. The results proved that this technique can be used in developing an intersection BCI, but further research is needed to fully develop such an index and incorporate that index with the segment BCI discussed in this manual. (See the Final Report for a more complete discussion of the intersection index results.4)
Table 1. Bicycle Compatibility Index (BCI) model, variable definitions, and adjustent factors
Using the perspectives of more than 200 study participants in three locations (Olympia, WA; Austin, TX; and Chapel Hill, NC), the BCI model was developed for all bicyclists as shown in table 1 (see appendix A for the English units version). The participants rated each of 67 sites included on a videotape with respect to how comfortable they would be riding there under the conditions shown. The ratings were made using a six-point scale where a one indicated that the individual would be "extremely comfortable" riding there while a six indicated that the individual would be "extremely uncomfortable" riding in those conditions. This model predicts the overall comfort level rating of a bicyclist using the eight significant (at p £ 0.01) variables shown and an adjustment factor (AF) to account for three additional operational characteristics. The basic model (excluding the adjustment factor) has an R2-value of 0.89, indicating that 89 percent of the variance in the index or comfort level of the bicyclist is explained by the eight variables included in the model. In other words, the model is a reliable predictor of the expected comfort level of bicyclists on the basis of these eight variables describing the geometric and operational conditions of the roadway. The variable with the largest effect on the index is the presence or absence of a bicycle lane or paved shoulder (BL); the presence of a bicycle lane (paved shoulder) that is at least 0.9 m wide reduces the index by almost a full point, indicating an increased level of comfort for the bicyclist. Increasing the width of the bicycle lane or paved shoulder (BLW) or the curb lane (CLW) also reduces the index as does the presence of residential development along the roadside (AREA). On the other hand, an increase in traffic volume (CLV and OLV) or motor vehicle speeds (SPD) increases the index, indicating a lower level of comfort for the bicyclist. The presence of on-street parking (PKG) also increases the index.
In addition to the primary variables included in the BCI model, three additional variables defining specific operating conditions were also examined. These supplemental variables were identified during the pilot phase of the study as having a potential impact on the comfort level of bicyclists and included the presence of: 1) large trucks or buses, 2) vehicles turning right into driveways, and 3) vehicles pulling into or out of on-street parking spaces. An analysis of the overall comfort level ratings made when viewing video clips illustrating these conditions showed all three of these variables to significantly increase the index, thus indicating a lower level of comfort when these conditions were present. For all bicyclists, the overall mean rating increased by 0.50 when large trucks or buses were present. When there were vehicles pulling into or out of parking spaces, the average rating increased by 0.60. And finally, the presence of right-turning vehicles resulted in an increase in the mean rating of 0.10.
While the presence of these three specific operating conditions was not evaluated across all possible combinations of geometrics and operations, the results of the limited sample do indicate a need for adjustment to the BCI model when large trucks or buses are present, when there is a high number of vehicles pulling into or out of on-street parking spaces, or when there is a high volume of right-turning vehicles. Thus, a series of adjustment factors that can be added to the model have been developed for each of these scenarios (see table 1). These factors were developed based on the theory that the conditions shown to the survey participants represented worst-case scenarios and, subsequently, the increase in the overall mean comfort level rating represented the maximum adjustment that would be required.
It should be noted that one variable not included in the development of the BCI model was the grade of the roadway. Results from a preliminary effort showed that changes in grade of 2 percent or less were not distinguishable on the video. The advantages of using video, including not exposing bicyclists to high-risk conditions, incorporating a much larger sample of sites, and controlling specific variables to ensure all subjects were exposed to identical conditions, were believed to outweigh the absence of this one variable. It is also believed that the variables having the most significant effect on the bicycle compatibility of a roadway have been included in the BCI model. Specifically, the variables of width, speed, volume, and on-street parking were shown to have the greatest impact on the index. At this time, the impact of grade relative to these and the other significant variables included in the model is unknown but may be determined in future research efforts.
Once the BCI model was developed, bicycle level of service (LOS) criteria were established based on the results of applying the model to the sites included in this study. Currently, there are no bicycle LOS criteria provided in the Highway Capacity Manual.5 However, the definition of LOS according to the manual is founded on the concept of users' perceptions of qualitative measures that characterize the operational conditions of the roadway. Two of the terms used in the manual to describe LOS are comfort/convenience and freedom to maneuver. Both of these terms are applicable to bicyclists and are directly reflected in the BCI since the rating scale used by the study participants was an indication of comfort level.
Thus, using the distribution of BCI values produced from the representative set of locations included in this study, LOS designations were established for LOS A through LOS F as shown in table 2. LOS A (represented by an index £ 1.50) indicates that a roadway is extremely compatible (or comfortable) for the average adult bicyclist while LOS F (represented by an index > 5.30) is an indicator that the roadway is extremely incompatible (or uncomfortable) for the average adult bicyclist.
In developing the BCI model, several other issues were addressed, including the effect of bicycling experience level on perceived comfort levels. Using the results from a questionnaire completed by the participants, the bicyclists were stratified into three groups based on their riding habits, such as number of bicycle trips per week and types of facilities used (e.g., major roadways vs. bicycle paths). A comparison of the comfort level ratings of these three groups showed that casual recreational bicyclists were generally less comfortable across all sites than experienced recreational or experienced commuter bicyclists. As a result of these differences, separate BCI models were produced for each of the three groups in addition to the model for all bicyclists. However, in real-world applications, it is most likely that bicyclists of all experience levels will have the opportunity to ride on any given segment of roadway. Thus, it is recommended that the BCI model developed for all bicyclists and shown in table 1 be used without modification for most applications. It is important to note that the LOS designations shown in table 2 were developed on the basis of this model, and thus are only applicable to results produced with the "all bicyclists" model.
Notwithstanding, when the practitioner knows that the large majority of riders are indeed casual bicyclists, the approach that should be used to ensure that facilities meet the desired comfort levels of this group is to simply design for a higher level of service. The results of the research showed that the model developed for the casual bicyclist, on average, produced BCI values that were 0.14 to 0.38 greater than those produced by all bicyclists. The differences in BCI values between LOS designations are, on average, 1.0 (see table 2). By designing for a higher LOS (e.g., LOS B rather than LOS C) on a facility known to attract a high number of casual bicyclists, the necessary comfort level for this group of bicyclists can be achieved with the BCI model as it is currently developed. Note that where casual bicyclists are expected, the facility should always be designed at LOS C or better.
Another issue addressed was that of possible regional differences in the perceptions of bicyclists. If bicyclists in different geographic regions of the country perceive comfort levels differently, then separate models would need to be developed to reflect these differences. An analysis of the comfort level ratings across subjects in the three survey cities showed no differences in the mean overall comfort levels for the four variables rated (speed, volume, width, and overall). This lack of differences indicates that the perceptions of individuals with respect to bicycle compatibility are the same in the three regions where the survey was conducted, and that the BCI model should be applicable across all regions of the country.
The range of conditions included in the development of the model should be representative of most urban and suburban roadway conditions. However, since the sites included in the development contained a limited range of widths, volumes, and speeds, the model should not be extrapolated beyond the values shown in table 3. For example, the model may only be appropriate for bicycle lane or paved shoulder widths between 0.9 and 2.4 m and curb lane widths between 3.0 and 5.6 m.