Introduction

Mountain biking is a popular sport in Switzerland (Fischer et al., 2021), that raises both environmental and policy issues (Pröbstl-Haider et al., 2018). In order to learn more about this recreational activity, this project takes a closer look at the movement of mountain biking. The analysis of movement data has gained increasing attention from both the GIScience community and the wider public, primarily because such data is easily accessible and appears to have a simplicity in structure (Laube, 2014). Addressing this sport from a movement and spatial analysis perspective, I hope to provide some insights about this movement pattern.

Mountain biking can be practiced on different surfaces (on or off-road), but riders show a preference for natural areas (Zajc & Berzelak, 2016). There are many different types of mountain biking practices (Zajc & Berzelak, 2016), and in this project I focus on downhill mountain biking. The goal is to find the movement pattern of mountain biking in trajectories where also other movement types occurred. In order to detect the mountain biking pattern, I define three criteria which characterize downhill mountain biking.

The criteria are:

The speed of a person riding a mountain bike is hard to define and may vary from 59-88 kmh (Jeremy, n.d.) to 27-32 kmh (How Fast Do Professional Mountain Bikers Rip Down Hills?, 2022). Other (slower) speed ranges may apply depending on personal condition and preference. Since natural areas are preferred for biking, (Zajc & Berzelak, 2016) the ground cover type needs to be vegetated or otherwise natural. Using just one or two of the criteria will still allow for other activities. For instance, a trajectory leading through grasslands and across forests could still be regarded as hiking, the same applies to the downhill movement. As the speed of a person riding a mountain bike is subject to many factors, it is also not a unique identifier for this movement pattern. Therefore, a combination of the criteria is needed to describe and detect the biking pattern. Using GPS data that Lisa Wahlen recorded on three mountainbiking tours, I will address the research questions below.

Research Questions

In this project I will work on the following three research questions:

Data and Methods

Datasets

For the project, I used the movement trajectories that were created by Lisa Wahlen when going mountain biking. The trajectories were recorded by using the GPS-tracking App posmo (Genossenschaft Posmo Schweiz, 2022) with a sampling rate of 10 seconds. I include 3 mountain biking tours Lisa completed in the period from the 6th of May to the 18th of May 2023. The tours were tracked in Switzerland in the areas of Wiriehorn (2 tours) and Valbirse (1 tour).

For the assessment of the criteria that describe the movement pattern of mountain biking, we used the MOPUBE vector dataset providing the land cover types of the canton of Bern, using the version last updated on the 23.05.2023 (Amt für Geoinformation des Kantons Bern, 2023). In addidtion, we worked with parts of the swiss digital terrain model swissALTI3D (Bundesamt für Landestopografie swisstopo, 2022). This elevation data is provided as a raster in a resolution of 0.5 m in grids of 1 km^2. I worked with extracts of the datasets, covering the spatial extent of the trajectories within the Canton of Bern.

I chose the tour Lisa made on the 18th of May in the Wiriehorn area (Wiriehorn 05-18) to define the characteristics of the biking movement pattern and tried to apply the criteria to the tour at Wiriehorn on the 7th of May (Wiriehorn 05-07) and the tour in Valbirse on the 6th of May for verification. The trajectories included the car ride from Solothurn to the biking location and back. A first visualisation of transport modes of the data revealed that the mountain biking part of the trajectories were labelled by the posmo app as “Car”, “Bus”, “Other” - and sometimes almost correctly with “Bike”.

Wiriehorn trajectory of the 18th of May
Navigating to the south of the trajectory you’ll find the mountain biking part.

Wiriehorn trajectory of the 7th of May
Navigating to the south of the trajectory you’ll find the mountain biking part.

Valbirse trajectory of the 6th of May
Navigating to the northeast of the trajectory you’ll find the mountain biking part.

Methods

Three relevant criteria for mountain biking are speed, downhill movement and ground cover. For all three criteria I assessed the values and ranges fitting the mountain biking pattern of the Wiriehorn 05-18 tour. Then I applied the criteria to the Wiriehorn 05-06 and the Valbirse tour.

Speed and segmentation

For every fix on the trajectory I calculated the average speed within the window of four neighboring points, similar to the moving window method of Laube & Purves (2011). The euclidean distance to the two points before and the two points after the fix was calculated and divided by the time difference between the fix and the two points before and after, respectively. Fixes where the speed calculation could not be completed because either the distance or the time between two points was 0, were regarded as static. On the Wiriehorn 05-18 tour, 80 points were labelled as static, and used for segmentation of the trajectory. They were located in places on the trajectory where little movement or a change in transport mode is plausible (e.g. at the bottom and the top of the cable car allmiried, somewhere in the middle of the downhill trail, at train stations in Solothurn). The segments were used to identify parts of the trajectory where Lisa was mountain biking. The relevant segments were selected based on visual assessment of the movement parameter profile (Dodge et al. 2009) of the average speed (Figure below). I assumed that Lisa went biking between 8:30 and 13:30, and that very short segments and segments with high velocities do not represent biking.

However, the selected segments included the movement of the cable car Lisa took to get uphill to the trail start. To isolate the cable car and determine the relevant speed range for mounainbiking, the average speed distribution of the selected segments was calculated. The speed values that best characterized the movement of mountain biking were identified through a process of testing.

Ground cover

The MOPUBE dataset showed 22 ground cover types (see figure below). Knowing about the preference for natural contexts for mountain biking (Zajc & Berzelak, 2016) I assume that Lisa rode on vegetated and/or natural groundcovers. Thus, every groundcover type seems suitable except for “Abbau, Deponie”, Bahn”, “Gebäude”, “Strasse, Weg”, “Trottoir”, “übrige befestigte” and “Verkehrsinsel”. We could argue that water bodies are not suitable as well, but depending on their size it is not impossible to cross such features when biking.

To determine on which parts of the trajectory Lisa moved offroad, I intersected the ground cover dataset MOPUBE and the movement points. This reduced the trajectories to the spatial extent of the Canton of Bern, therefore excluding parts of the car ride from and to Solothurn. This step called track annotation (Dodge et al., 2013) allowed to add environmental information to the points on the trajectory. All points associated with natural ground cover were considered as matching the criteria and labelled accordingly.

Downhill movement

The elevation data was added to the dataset of the Wiriehorn 05-18 trajectory and the difference in elevation to the fourth point after the fix in the direction of movement was calculated. If within this 40 second window the difference in altidude was negative, it is assumed to be downhill movement.

Combination of criteria and refining steps

For every point on the trajectory I displayed how many of the three criteria applied. Based on the count of matching criteria, I isolated new segments that represent downhill mountain biking. To be considered for a mountain biking segment, a point on the trajectory had to meet the following conditions: The neighboring four fixes must show a sum of at least 6 matched criteria, and, within a window of one minute, there is at least one fix that matches all three of the criteria.

Results

Wiriehorn 05-18

On the Wiriehorn 05-18 trajectory, the points where I assumed Lisa went mountain biking generally show a higher count of criteria matched. The straight line uphill represents the cable car, the sinuously distributed points represent the presumed biking pattern. Depending on the criteria, some points on the trajectory received a higher count.

Wiriehorn 05-18. Number of matched criteria for mountain biking (mtb) for each point.

Wiriehorn 05-18. Speed (mtb_speed), groundcover (mtb_gc), downwards movement (mtb_elevation)

After applying the refining criteria, the basis for the new segmentation included all of the downhill biking trajectory except for 11 points (Figure below).

Wiriehorn 05-18 trajectory. The points matching the refined criteria return TRUE.

The results of the second segmentation step represent the mountain biking pattern. Segments where Lisa went biking could be isolated.

Application to other movement trajectories

Wiriehorn 05-07

Also on the Wiriehorn 05-07 trajectory, a higher count of criteria matched is visible where Lisa went mountain biking. The count varies depending on the criteria.

Wiriehorn 05-07. Number of matched criteria for mountain biking (mtb) for each point.

Wiriehorn 05-17. Speed (mtb_speed), groundcover (mtb_gc), downwards movement (mtb_elevation).

Wiriehorn 05-07 trajectory. The points matching the refined criteria return TRUE.

The results of the refined segmentation step capture the mountain biking pattern.

Valbirse

On the Valbirse trajectory many points that are not part of the biking pattern fit more than one criteria.

Valbirse. Number of matched criteria for mountain biking (mtb) for each point.

Taking a closer look, parts of the cable car did not match the criteria of downward movement (mtb_elevation).

Valbirse. Speed (mtb_speed), groundcover (mtb_gc), downwards movement (mtb_elevation)

Valbirse trajectory. The points matching the refined criteria return TRUE.

New segments based on refined criteria for Valbirse, interactive map for closer inspection.

After the refined criteria were applied, two segments covering five points in Biel/Bienne were considered part of the mountainbike movement pattern. Most of the isolated biking segments are located in rural areas close to Malleray, where mountain biking seems more likely.

The detection of the downhill movement pattern based on speed, groundcover and downward motion was successful for Wiriehorn 05-18 and Wiriehorn 05-07. For the Valbirse trajectory, it is possible that mountain biking segments have been isolated where no mountain biking has taken place.

Discussion

The criteria of speed, ground cover and downward motion could be shown useful for detecting and characterizing downhill mountain biking. The visualisations revealed points where mountain biking is plausible. The two Wiriehorn trajectories where detection was successful were recorded in the same location, and given this similarity it is no surprise the criteria matched quite well. On the Valbirse trajectory, despite the refinement steps applied, a few points in Biel/Bienne were isolated as mountainbiking segments. Falsely associated ground cover due to measurement inaccuracy could have caused three criteria to match therefore meeting the refinement conditions. Thus, for two out of three trajectories the criteria and the method applied worked well. Combination of the three criteria could also be used for segmentation, therefore allowing extraction or annotation of mountain biking segments within a trajectory. However, there are some issues to consider.

Here, I only validated the data visually and in a qualitative manner. In the context of movement pattern detection, assessing the internal validity or the sensitivity of the method to parameters changes are relevant (Laube, 2014). Such steps would require additional datasets, which I lacked due to the choice of limiting the ground cover information to the spatial extent of one Swiss Region. By using other ground cover datasets, trajectories with different spatial extents could have been analysed.

Furthermore, the choice of the parameters took place in an explorative way, mainly to exclude unwanted features (e.g. set the speed range to exclude the cable car). Thus, the Parameter values may work well for the trajectories I experimented with, but might be to narrow (or wide) to fit other mountain biking trajectories. Some parameters, especially speed, vary with the (temporal) scale in which they are analysed (Laube & Purves, 2011). Such notions need to be examined further when validating the parameters. Other movement parameters like sinuosity and turning angle could be additional criteria to consider. In the trajectories examined here, downhill mountain biking often displayed sinuous properties. In terms of track annotation (Dodge et al., 2013), known biking trail networks could be annotated and added as criterion.

In summary, the criteria of speed, ground cover, and downward motion successfully identified and described downhill mountain biking in most instances. However, to ensure wider applicability across various mountain biking scenarios, further validation, parameter adjustments, and consideration of additional criteria are required.

References

Amt für Geoinformation des Kantons Bern. (2023). Amtliche Vermessung vereinfacht [Map]. 2023https://www.agi.dij.be.ch/de/start/geoportal/geodaten/detail.html?type=geoproduct&code=MOPUBE

Bundesamt für Landestopografie swisstopo. (2022). swissALTI3D: das hochaufgelöste Terrainmodell der Schweiz [Map]. https://www.swisstopo.admin.ch/de/geodata/height/alti3d.html

Dodge, S., Weibel, R., & Forootan, E. (2009). Revealing the physics of movement: Com- paring the similarity of movement characteristics of different types of moving objects. Computers, Environment and Urban Systems, 33(6), 419–434. https://doi.org/10.1016/j.compenvurbsys.2009.07.008

Dodge, S., Bohrer, G., Weinzierl, R., Davidson, S. C., Kays, R., Douglas, D., Cruz, S., Han, J., Brandes, D. & Wikelski, M. (2013). The environmental-data automated track annotation (Env-DATA) system: linking animal tracks with environmental data. Movement Ecology, 1(3). https://doi.org/10.1186/2051-3933-1-3

Fischer, A., Lamprecht, M., & Bürgi, R. (2021). Mountainbiken in der Schweiz 2020: Auswertung Mountainbikeland-Befragung 2019 und Sekundäranalyse von «Sport Schweiz 2020». Bundesamt für Strassen ASTRA und Stiftung SchweizMobil. https://www.astra.admin.ch/astra/de/home/themen/langsamverkehr/materialien.html

Genossenschaft Posmo Schweiz. (2022). Posmo Project App - Tracking für Gruppen. https://posmo.coop/produkte/posmo-project-tracking-fuer-gruppen

How Fast Do Professional Mountain Bikers Rip Down Hills? (April 20, 2022). https://cyclebaron.com/how-fast-do-professional-mountain-biker-rip-down-hills/

Jeremy [missing surname]. (2023). Average Mountain Bike Speed (with 13 examples). The Bike report. https://www.mountainbikereport.com/mountain-bike-speed/

Laube, P. (2014). Computational Movement Analysis. Springer Cham. https://doi.org/10.1007/978-3-319-10268-9

Laube, P. & Purves, R.(2011) How fast is a cow? Cross-scale Analysis of Movement Data, Transactions in GIS, 15(3), 401–418, John Wiley & Sons Ltd, DOI: 10.1111/j.1467-9671.2011.01256.x

Pröbstl-Haider, U., Lund-Durlacher, D., Antonschmidt, H. & Hödl, C. (2018). Mountain bike tourism in Austria and the Alpine region – towards a sustainable model for multi-stakeholder product development. Journal of Sustainable Tourism, 26(4), 567-582. https://doi.org/10.1080/09669582.2017.1361428

Zajc, P. & Berzelak, N. (2016). Riding styles and characteristics of rides among Slovenian mountain bikers and management challenges. Journal of Outdoor Recreation and Tourism, 15, 10-19. https://doi.org/10.1016/j.jort.2016.04.009