Keywords

1 Introduction

Cycling has been purported as a way to address public health, environmental, and transportation concerns. While the activity has continued to increase in popularity in the United States [1], a variety of challenges may limit continued growth. For cyclists, one challenge is knowing where to ride. The purpose of this study is to examine the factors expert cyclists consider when planning a route and the resources they use when planning routes. A relatively unique aspect of this study is the focus on cyclists who are best categorized as recreational cyclists who ride long distances on the road each year. The majority of existing route planning research focuses on utilitarian route planning and commuting. Obvious differences exist between these groups, but it is currently unclear whether those differences meaningfully impact route planning. Another unique element of this study is an examination of the resources expert cyclists use to plan routes (e.g., other cyclists and computer applications such as Strava, Map My Ride, and Google Maps). Even though the scientific literature has limited data on route planning for recreational cycling, it is possible that other entities, particularly those that already serve recreational cyclists, would offer route planning resources. Therefore, an indirect measure of cyclist preferences could be inferred from the specific resources they value when planning a route. Because of this, the data from this study have the potential to fill a gap in the literature regarding recreational cyclist preferences and have the potential to inform the development of more effective route planning applications.

1.1 Factors Considered During Route Planning

Safety is hypothesized to play a major role in planning a route. In Winters, Davidson, Kao, and Teschke’s (2011) stated preference study, nine of the top ten deterrents to cycling were related to safety with primary concerns associated with vehicle interactions and hazardous surfaces. Similarly, Xing, Handy, and Mokhtarian (2010) found a positive correlation between the distance cyclists ride and the perceived safety of the ride; that is, greater distances are associated with higher levels of perceived safety, and shorter distances are associated with lower levels of perceived safety. Some hazards could be mitigated through route planning. The existing route planning literature covers a variety of factors ranging from infrastructure [4] to health [5] and safety to societal norms [e.g., 3] to individual cyclist characteristics [e.g., 6]. The most comprehensive collection of work examines bicycle infrastructure and route planning.

In general, studies collecting stated preference and revealed preference (e.g., GPS tracking) data have shown that cyclists prefer physical separation from motor vehicles. When riding on shared roadways, they prefer streets with less and slower-moving traffic as well as those with no on-street parking [for a review see 7]. This type of research has been used as evidence that improved bicycle infrastructure may lead to increases in bicycle commuting. One tacit assumption of this research is that, if available, cyclists will incorporate bicycle infrastructure into their routes.

Even with substantial evidence that cyclists prefer to ride where there is cycling infrastructure [7: 1], there are challenges in translating those preferences into predictions about how an individual cyclist will plan a route. One challenge in the United States is the relative lack of continuous cycling infrastructure. That is, it is common for a single route to include segments with no infrastructure along with some shared infrastructure (designated bike routes or sharrows), or some dedicated infrastructure (e.g., painted bike lanes, multiuse paths). Along these lines, Buehler and Dill’s (2016) review of the literature indicates that even though cyclists in the US and Canada prefer separate cycling facilities, the majority (50–90%) of the distance covered in their rides takes place on roads with no dedicated bicycle facilities.

A second challenge in applying existing research to route planning for US cyclists is that most of the data are related to commuter behavior. Recreational cycling is far more prevalent than commuting in the United States [3] and this trend is unlikely to change [c.f., 8]. This is problematic as evidence is mixed as to whether or not these two types of cycling are influenced by the same factors [3]. There are notable differences between cycling for recreation or fitness and cycling for utilitarian purposes like commuting. One major difference is that utilitarian rides have specific destinations. Individuals commuting from home to work have a fixed point of origin and a fixed destination; the only variation would be in how they travel between those points. Recreational routes are comparatively unconstrained from that perspective. This flexibility presents its own challenges. For instance, Priedhorsky, Pitchford, Sen, and Terveen (2012) tested algorithms that were designed to create personalized bicycle routes. While they report that initial tests of the algorithms were successful and they were feasible to implement, challenges remain in understanding and modeling the subjective experience of recreational cyclists (e.g., what makes a “good” route). They specifically note the repeated request for routes that are loops instead of out-and-back routes.

Beyond the possibility that priorities for utilitarian and recreational routes differ, there are subtleties within the utilitarian literature suggesting that individual differences and even slight differences in task demands impact route planning. Aultman-Hall, Hall, and Baetz (1996), for instance, found two distinct types of cyclists – those who take more direct routes on busier roads and those who actively avoid busier roads. In a similar vein, Broach, Dill, and Gliebe’s (2012) revealed preference study identified different preferences between routes used for commuting and those used for utilitarian purposes like shopping (i.e., rides that were not a commute and not for exercise). Distance traveled was a greater concern when commuting than when riding for other utilitarian purposes; commuters also were reported to be “somewhat more sensitive to riding in high volumes of mixed traffic” (p. 1737). These subtilties within the utilitarian category lend credence to the idea that activities with much different goals (commute vs. exercise, for instance) would be associated with different demands.

1.2 Route Planning Resources

Commuters and recreational cyclists alike must engage in route planning. While route planning resources have the most apparent benefits for cyclists who are new to the activity or new to an area, expert cyclists can also benefit from these resources. Regular cyclists may want to ride while on vacation or after moving residences. Recreational cyclists may need to find a route to work as they transition to commuting by bicycle [see 6 and 12 for evidence that some commuters begin as recreational cyclists]. Even experienced recreational cyclists may need to plan a new route if they want to ride to a new location or simply want to incorporate a new route into their exercise routine. A benefit of considering route planning in expert cyclists is that, on balance, they should have been exposed to a wider variety of routes than those who travel less distance on the roadways or those who primarily travel fixed routes.

A variety of resources are available to help cyclists plan new routes. The route planning features available within each resource are heavily influenced by the stakeholders involved and by the original intended use of the resource. For example, some route planning resources are related to public policy and infrastructure design. Because of this, they may highlight the availability of specific bicycle infrastructure, safety data, and typical traffic volume. Some route planning resources are specialized for specific areas (e.g., bicycle tourism for a city), so they might include unique points of interest and entertainment or recreational areas. Resources facilitating long-distance bicycle travel (e.g., biking across the country) and bikepacking include options for locating food and drink and lodging. Resources supporting bicycle fitness may prioritize data associated with the speed or effort associated with the ride.

Because the present study was designed to examine the preferences of cyclists who ride extensively for recreation or exercise and because some applications are not well known [e.g., Biketastic, 13] only those resources that are relevant and more well-known were examined in this study. Considerable overlap exists between the applications, so these sources are examined according to relevant features they provide (Table 1).

Table 1. Major route planning features and applications that provide them
Full size table

The applications examined can be sorted into three categories. First, Google Maps [14] is a general road map and navigational tool that has been expanded to include bicycle infrastructure overlays and specific considerations for bicycle navigation. Second are applications best categorized as fitness trackers: Strava [15], Garmin Connect [16], and Map My Ride [17]. The primary utility of these apps involves accurate tracking of location, speed, and other performance metrics; they have been expanded to differing degrees to generate routes and facilitate social networking. Third, are applications best categorized as user-developed content with routes built on OpenStreetMap’s OpenCycleMap resource. The Ride with GPS [18], Bikemap [19], and Komoot [20] applications provide platforms that facilitate more lengthy descriptions of routes along with the integration of photos and points of interest on the routes.

Despite the existence of a variety of route-planning resources, it is unclear whether they address cyclists’ needs. Previous research suggests that cyclists are primarily concerned with safety in relation to vehicle traffic, but only Google Maps provides live indicators of traffic volume. Similarly, it does not appear as if any of the applications integrate accident data. From the perspective of recreational cyclists, only Garmin and Strava generate bicycle route loops from a set location; the absence of this feature could be problematic for cyclists who don’t want to specify destinations in order to create a route. Anecdotally, it seems even popular applications that cater to recreational cyclists have failed in providing adequate route planning. In evaluating a new routing feature in Strava, popular blogger, DC Rainmaker reports, “I’m actually impressed. I had relatively low expectations for such a feature, mostly because it feels like every time we see companies try and do automated route generation, it either ends up being too stiff or too focused on data driven by commuters – which aren’t really ideal for workouts (where sustained speed and non-stops are of higher importance)” [21]. His comments highlight what may be a long-standing trend of disparity between the factors considered in route planning applications and cyclist needs.

1.3 Present Study

The purpose of this study is to examine the resources expert cyclists rely on when planning a route as well as conditions they consider when planning a route. This extends previous research by surveying cyclists who primarily ride for recreation, by examining the influence of additional road hazards on route planning, and by asking cyclists to rate various sources, they would consider when planning a cycling route.

2 Method

2.1 Participants

Cyclists for this study were recruited using a snowball technique through social media. An invitation to participate was posted on a Facebook page representing an 1800 member, Virginia-based, cycling group. The invitation was then shared to a few individual cyclist’s pages and to at least two additional cycling groups. Only individuals who reported living in the state of Virginia and had taken the Virginia drivers’ license exam were considered for the study. Of the 145 respondents who met those requirements and indicated they were cyclists, 37 did not complete the survey, and an additional 18 respondents did not meet the minimum yearly distance requirement (2000 miles) to be considered expert cyclists.

Ninety cyclists remained in the dataset after excluding those who did not complete the survey or meet the mileage criteria. All cyclists reported riding at least once a week, all exceeded the 2000 miles/year criterion, and all of them reported riding a bicycle on the road. This sample of cyclists had extensive experience with an average of 20.85 years of experience riding a bicycle and riding an average of 4394 miles each year (varied from 2000 to 12000 miles). Also, 58% reported having specific expertise related to bicycling, which included amateur and professional racing, providing bicycle tours, promoting cycling events, leading group rides, racing, and riding various disciplines (road, gravel, cyclocross, track, and mountain biking), and coaching. Participant demographics were typical for expert cyclists in the region; older males with higher socioeconomic status (69% male, average age 52 years old, 88% Bachelors’ degree or higher).

2.2 Materials and Procedure

The data for this study were gathered using a more extensive stated preference online survey that included: demographics, bicycle-related law, perceived hazards, and route planning. The data related to knowledge of bicycle-related law and perceived hazards were examined in a separate study [22].

Participants were asked to identify and rate factors they consider when planning a bicycle ride. Broadly, those factors included potentially hazardous road conditions, interactions with motorists, navigation ease and efficiency, presence of bicycle infrastructure, route familiarity, and local attractions, services, or scenery. To facilitate a more comprehensive analysis, participants were asked to consider a few perspectives. First, they were asked to select from a list of 21 factors all of those they would consider when planning a bicycle ride. The factors overlap with those examined in previous route planning studies [e.g., 2]. To encourage liberal responding, cyclists were asked to “select all that apply”. Second, participants rank-ordered nine general factors by indicating which were most to least influential when planning a typical ride (e.g., traffic concerns, weather conditions, road quality, time constraints, landmarks/sights/events, number of stops). Finally, participants were asked to rate how often specific conditions influenced their route selection. The specific conditions associated with this item overlapped with some factors that were assessed by the “select all” item, but this item also included additional hazards that had been examined in Still and Still (2020). The conditions were rated on a scale of 1 to 5 for how often they influence route selection (never, sometimes, about half the time, most of the time, always). Because some of the “hazardous” conditions being examined in this study are not common in the route planning literature (e.g., speed bumps, drains), data related to participant perception of these hazards was included as well. Perceived hazards were measured by asking cyclists to rate how hazardous (1 – not at all hazardous, 5 – very hazardous) they believe specific conditions or situations would be to them if they were riding their bike.

Participants were also asked to identify and rate the resources they use when planning a new route. This portion of the study was intended to provide an indirect measure of the utility of current route planning resources and applications. For example, if cyclists prefer routes generated using fitness tracking applications over Google Maps, it might reveal underlying preferences for the factors considered in those applications. The list of resources examined is not exhaustive; it was developed from an informal survey of expert cyclists, online bicycling forums, and bicycling publications (e.g., Bicycling magazine). To measure the variety of resources participants use, they were asked to indicate which of nine resources they would consider when planning a new route in an area they are familiar with (new destination where they live). To see if their preferences depend on familiarity with the location, they were also asked to indicate what resources they would use in an area they are unfamiliar with (city they have never visited before). To further explore the usefulness of existing resources, they were asked to rate how safe, enjoyable, and fast or efficient they would expect a local route to be if recommended by, or developed using, those resources. The rating scale ranged from 1 (not at all likely) to 5 (very likely) and included a sixth option, do not know.

3 Results

Only data from the 90 expert cyclists, as defined in the Method section, were included in these analyses. When hypothesis testing was employed, a probability of .05 was used as the criterion for determining statistical significance. No outliers were identified, excluded, or replaced.

3.1 Factors Considered During Route Planning

Participants were first asked to identify from a list of 21 factors which ones they consider when planning any bicycle ride. Of the 21 factors (listed in Table 2), the mean number of factors identified by an individual participant was 10.79, suggesting that expert cyclists consider several factors when planning a route.

Table 2. Conditions expert cyclists consider when planning any bicycle ride.
Full size table

When examining these individual factors (see Table 2), the majority of expert cyclists consider safety issues related to interactions with motor vehicles and road conditions when planning a route. Bicycle infrastructure and “bike friendly” features are considered by many cyclists, but not to the same extent as other safety issues. Factors commonly associated with other forms of recreation (e.g., tourism) are only regularly considered by a minority of expert cyclists. These data suggest that few expert cyclists use cycling in combination with other forms of recreation.

While multiple factors are considered in route planning, it is unlikely those factors have equal importance. Therefore, participants were asked to rank nine general factors in order from most to least influential in making their route decision. Traffic concerns were ranked as having the greatest influence on route selection, followed by weather conditions and time constraints. People they expect to ride with (e.g., group rides) and average speed for the route were ranked nearly the same (4th and 5th) in the middle of the list. Road surface quality and familiarity with the route were ranked slightly lower and landmarks, sights, or events; and number of stops/slowdowns were ranked as having considerably less influence on route selection.

When asked to rate potential hazards on the roadway, consistent with previous findings, expert cyclists were most concerned with the risk associated with motor vehicles, potholes or broken road surfaces, and glass or sharp objects in the road (see Hazard Rating in Table 3). Across all potential hazards, participants’ hazard ratings were positively correlated with route influence rating; that is, more hazardous conditions were associated with greater influence over route planning.

Table 3. Ratings (mean and standard deviation) of how much influence specific conditions or situations have on route selection and how hazardous those same conditions are rated. Spearman’s rho correlations between measures are reported. * p < .05, ** p < .01, *** p < .001
Full size table

When asked to rate how often these potential hazards and other conditions influence route selection, only risk of vehicles passing too closely and risk of being hit by a distracted driver were considered more than half of the time. Five additional conditions and hazards were rated as being considered about half of the time when planning a route: potholes/broken road surfaces, broken glass/sharp objects, standing water, gravel/sand, and speed they can ride on the route (Route Influence Rating in Table 3).

3.2 Resources Used for Route Planning

Local vs. Unfamiliar Locations.

To index the general types of resources cyclists use when planning a new route, participants were provided with ten resource options and asked to select all resources they would typically consider. This was done for both familiar (e.g., town they live in) and unfamiliar (e.g., a city never visited before) locations. The majority of expert cyclists reported they would consider routes recommended by a local cycling club or bike shop and routes recommended by acquaintances familiar with the area. They also report using approximately four sources of information when planning a new route (i.e., 3.9 resources when planning a local ride and 4.6 resources when planning a ride in an unfamiliar location).

As indicated in Table 4, the data suggest that while expert cyclists use similar resources when planning routes in familiar and unfamiliar locations, there are some differences. For instance, routes obtained from local clubs and stores are valued the most in both situations, but acquaintances who are familiar with the area are valued more when the cyclist is riding in an unfamiliar location than in a familiar location. There is also more reported reliance on public and private maps of routes, cyclist frequency data (e.g., heatmaps), and identified segments in an unfamiliar location compared to a familiar location. While this finding is not surprising, it does show that when planning for an unfamiliar location, expert cyclists use more resources and they are more likely to use technological resources like applications.

Table 4. Sources of information cyclists report typically using when planning a new route. Values represent the proportion of total cyclists who reported using each source.
Full size table

Expected Quality of Route Planning Resources.

To further explore specific applications and the expectations cyclists have when using specific resources, participants rated how safe, enjoyable, and fast and efficient, they expect a route to be when using those resources. These ratings were completed in the context of finding a new local bicycle route. This framing was intended to establish an evaluative context in which cyclists might use their knowledge and experience to consider the quality of routes informed by various resources.

A few results are noteworthy. First, participants had the option to indicate that they “do not know” how safe, enjoyable, or fast and efficient a route would be. The percentages listed in Fig. 1 indicate those who did not select “do not know”. From those data, it is clear that expert cyclists were less familiar with, or confident about, the Komoot and Bikemap applications. They were more willing to rate Ride with GPS and Map My Ride and even more so Garmin, Strava, and Google Maps. This higher willingness to provide ratings suggests they have more familiarity or experience with those applications.

Fig. 1.
figure 1

Mean ratings (standard error) for how fast and efficient, enjoyable, and safe participants think a route would be based on the information source. Sources are listed in descending order based on the percentage of participants who rated the source (e.g., 17% rated Komoot; 83% reported not using Komoot).

Full size image

A second trend in the data reveals higher ratings for routes informed by acquaintances the cyclist regularly rides with and cycling clubs or bike stores than the other resources (see Fig. 1). To test this trend, a composite rating was calculated for each resource by finding the average rating across the three dimensions (fast/efficient, enjoyable, safe). A significant effect of resource type was obtained, F(10, 140) = 18.07, p < .001, using this metric. Post hoc tests using the Bonferroni correction revealed that the composite rating for acquaintances they regularly ride with and for cycling clubs or bike stores are significantly higher than the composite ratings for all other resources. The only other significant difference between sources was a higher composite rating for acquaintances they seldom ride with compared to Google Maps and Komoot.

Finally, while the three ratings tend to vary together, there were some exceptions. For instance, while cyclists expected that acquaintances they regularly ride with and cycling clubs and stores were more than likely to provide safe and enjoyable routes, they ranked those sources as being slightly less likely to produce fast and efficient routes. The opposite pattern was found for Google Maps where, relative to one another, it was rated as being more likely to produce fast and efficient routes and less likely to produce safe or enjoyable routes.

4 Discussion

It is crucial to consider the preferences and route planning needs of expert and recreational cyclists. They are currently underrepresented in the literature [c.f., 7] even though they typically travel further on the roadways and represent the majority of bicycle traffic in the United States [3]. The results of this study suggest that expert cyclists share similar safety concerns with those who ride less frequently and those who only ride to commute. For example, in their examination of 73 factors impacting cycling behavior, Winters et al. (2011) found that the top five deterrents to cycling were high levels of vehicle traffic, high vehicle speeds, risk associated with motorists, ice and snow, and glass/debris on the streets. These factors were also concerning for cyclists in the present study. Most participants (70% or more) reported considering them when planning any bicycle ride and they reported considering these factors at least half of the time when they plan a route.

Even with these similarities, the data suggest key differences in their preferences. Winters et al. (2011) found that a top motivator of cycling behavior was the presence of routes with beautiful scenery; only 44% of cyclists in the present study reported considering scenery when planning a ride. Stops and slowdowns were ranked as being less important to cyclists in the present study, while commuters indicate a preference to limit stops [23]. Some differences between utilitarian and recreational preferences are clearly related to the specific demands of the activities. Cyclists in the present study considered wind direction in route planning. Commuters with fixed origins and destinations would have little flexibility to change their route to accommodate wind direction. Similarly, commuter behavior may be impacted by the facilities at their destination, such as showers and secure bike storage; these factors would not be expected to be relevant for recreational cycling.

The results of this study also highlight a potential disconnect between the resources expert cyclists use to plan a new cycling route and the resources available to create or discover those routes. Cyclists rely on other cyclists to recommend routes over applications that may be used for route planning. They also rate those human sources as being more likely to provide routes that are safe, enjoyable, and fast and efficient. This result is somewhat surprising given applications like Strava, Garmin, and Map My Ride allow users to view routes their “connections” have created and/or rode themselves. In other words, it seems cyclists could find the same routes from their acquaintances simply by looking at the activities in their acquaintances’ digital profiles. While there are some potential barriers to doing this (e.g., some users make their activities private and some cyclists do not use these applications) it makes the ratings disparity initially seem unusual.

There are a couple of reasons why cyclists might not value routes they find using applications to the same extent as routes they could obtain from someone they regularly ride with or from a local bike shop. One challenge with bicycle routes that use roads with vehicular traffic is that they are subject to dynamic conditions. Route conditions associated with traffic volume, dangerous drivers, presence of ice/water/sand, broken glass, and potentially even potholes may change throughout the day. While fellow cyclists or those associated with bike stores or clubs may not know the current conditions, they can convey what the conditions are typically like. For instance, there may be roadways that typically have low levels of traffic in the morning but have substantially more traffic in the afternoon and evening. A cyclist might recommend the route only during the low traffic times. Similarly, there may be areas where an individual regularly encounters a portion of a route with broken glass or other debris; they could easily convey that information to a fellow cyclist. These dynamic changes are not well represented in current applications.

Another challenge involves the decision making process cyclists utilize when determining preferred routes over less-preferred routes. Complex situations can arise where a cyclist might use a cut-through [c.f., 11] that saves time or avoids a perceived hazard when an application would not. Similarly, while more applications are being developed to accommodate cyclist preferences to avoid multilane roads, high-speed roads, or hills, it is not clear that they produce what would be considered a preferred route. It is possible that cyclists might accept a short distance with non-preferred conditions if the rest of the route were subjectively better than an alternative that completely avoids those undesirable features. The results of several studies suggest bicycle route planning does involve these types of tradeoffs [10, 11]. Accordingly, researchers have begun examining those tradeoffs. For example, specific route features (bicycle infrastructure, on-street parking, number of stops) can be assigned positive and negative values to help quantify how they tradeoff in route selection [23, see also 24]. At this point, it is not clear route planning applications have been implemented to account for those tradeoffs in a way that would be like that of an expert cyclist. Ultimately, though, because utilitarian and recreational cyclists share similar safety concerns, and those are identified as primary concerns, there may be technological solutions for route planning that appeal to both groups as long as the applications can account for more dynamic and complex situations.

If developers intend to improve cycling route planning applications they will need to overtly accomodate different types of cycling activites. Even a cyclist who primarily rides for fitness may occasionally commute. Having the option to select a ride type - such as commuting, running errands, exercising, sight-seeing - could allow attributes that are most important for that activity to be considered during route generation. The same functionality could help accommodate different user types (e.g., someone who only commutes or only rides for exercise). Another consideration for developers will be how to obtain and incorporate more dyamic information into initial route planning. One way to do this would be to highlight areas on the route where hazardous conditions are anticipated. The hazard information could be based on historic traffic volumes, accident data, or subjective ratings provided by cyclists who have previously ridden through that area. To provide better estimates of the hazards, the application could prompt users to input the time at which they intend to begin their ride. The start time, along with expected rate of travel, could be used to estimate traffic volume when they reach critical portions of the route. This type of information could also be pushed in real time to cyclists’ fitness trackers, warning them of increasing traffic volume at their current location or on an upcoming portion of the route. Weather alerts, segement information (e.g., Strava), and even missed calls and messages can already be pushed to many fitness trackers when enabled by the user. Capturing some of the nuanced aspects of route planning that cyclists expect from other experienced cyclists, could improve the usefulness of applications and expert cyclist reliance on applications.