Why Crowdsourced Trail Data is Unreliable

By Robert Bennett, Pickatrail.com

"An inaccurate map is not a reliable map." —USGS (Fact Sheet 171-99)

In this document, crowdsourced trail data refers to geographic data (points) collected at rural areas by anonymous people who use smart phone trail apps, smart watches, bands and other wearable devices while hiking. I use the term crowdsourced "trail data" because this is how the data is described by the apps that collect the data, but the term is a misnomer; "route data" is a more accurate description of what is actually crowdsourced. Some volunteered geographic information projects describe route data as GPS traces.

A reliable dataset of trails is accurate to within an acceptable horizontal range, dependable, up-to-date and complete. Crowdsourced trail data is not reliable because the points are not produced using techniques employed by skilled GIS professionals who are concerned with accuracy; there is minimal quality control by experts over the contributed data; points are geofenced, and the datasets are magnets for cartographic vandalism. Many of the data contributors have good intentions but they often incorrectly describe their routes as trails using false trail names, or the data incorrectly denotes trailhead locations and/or the trail itself. This can occur when a person stops hiking before finishing a trail, hikes off-trail, scrambles, hikes motorized trails, back-tracks, hikes multiple trails while believing only one trail was hiked, or simply forgets to stop and/or start collecting data at the trailheads. Crowdsourced trails are often described as public, designated trails, when in fact the data describes illegally constructed trails, or routes that cross private property and/or environmentally sensitive areas. Illegally constructed trails destroy native plants, threaten wildlife habitats and increase the risk of soil erosion. Trail apps and map apps that promulgate illegal trails exacerbate damage to wildlife. Hikers who follow inaccurate trail data on map apps when the ground is featureless and/or visibility is reduced can inadvertently put themselves and Park Enforcement and Search and Rescue (PESAR) in danger.

Crowdsourced trail data is notoriously inaccurate, sometimes by hundreds of feet. These horizontal and vertical inaccuracies are attributable to how the devices are carried, hardware limitations, the absence of processing, and other sources of error such as multipath. When a smart phone is carried in a pocket or a backpack at a rural area it will have difficulty acquiring and keeping a GPS signal lock. Some smart watches, such as the Magellan Echo Sport Watch, are tethered to a smart phone.

Smart phones and GPS enabled smart watches, bands and other wearable devices collect a limited number of points because the location chips in these devices employ a necessary concept known as geofencing to consume less power and extend battery life. This is not a problem when the points are only used by the person who collected them, for example to view a sketch of an approximate route on a digital map with a large scale, but the vague route becomes visible as the user zooms into the map and the scale is small.

Take for example the Microsoft® Band. When Microsoft released the device for sale, I examined the product literature on their website. The Microsoft Band support page on the Run Tile captured my interest because it refers to hiking: "You can track your running, hiking, even kayaking stats using the Run Tile on your Microsoft Band". The same support page displays a Run Tile which includes a route on a map. The map looks great, but its scale is very large. I loaded Microsoft Bing® Maps in my browser and zoomed into the location displayed on the Run Tile to see a satellite view of the terrain using a smaller scale. What I found was surprising; the same running route marked on the Run Tile in the support page also exists on Bing Maps, with all its inaccuracies and geofenced points. I suspect someone at Microsoft forgot to remove the route from Bing Maps. The route bypasses the trail at several points; at one point, the route enters a river. When geofenced points are collected and used to depict a route or a trail on a map, not only is the lack of detail visible as the map scale is reduced, accurately calculating metrics using the data is impossible.

Add to all of the above, the behavior of smart phones and wearable devices is uncertain outside of specific, limited temperatures and elevations. The Iphone® operating temperature range is 32°F to 95°F (0°C to 35°C). The Apple® web site warns people who use the Iphone as a GPS device: "Conditions and activities that may cause the device to alter performance and behavior include... Using certain features in hot conditions or direct sunlight for an extended period of time, such as GPS tracking..." The Microsoft Band (v1) specification sheet describes the maximum operating altitude of the device as 3,937 ft (1,200 m); the device may not operate correctly at many of our most popular national parks. The lowest point in Colorado is 3,317 ft (1011 m) and the average elevation is 6,800 ft (2,070 m) which makes me wonder how many people are using the device in this state.

—Google's My Tracks Crowdsourcing App is Being Shut Down, April 2016

—Google's Map Maker Crowdsourcing Tool will be retired in March 2017

Recently published, by Robert Bennett: