Mount Vernon Trail · Arlington & Alexandria, VA
What I kept riding past on the Mount Vernon Trail
On my commute, shared scooters and e-bikes kept appearing where a trail, sidewalk, bridge deck, or park edge needed to stay clear. So I started marking each obstruction as I passed it. Over eight riding days, the pattern became hard to ignore: the corridor was not seeing a burst of new arrivals. It was collecting a backlog.
- GPS-tagged sightings
- 144
- reconstructed devices, at minimum
- 49
- stayed in place more than a day
- 17
- longest observed obstruction
- 10 days
- growth in devices standing in the corridor
- +70%
The pattern
The trail was not being dumped on. It was not being cleared.
New devices showed up at almost the same rate in both weeks: 3.4 per pass at first, 3.6 afterward. But the number already standing in the corridor rose from 8.3 to 14.1 per pass, a 70% increase.
That distinction matters. It points to collection and monitoring, not simply a surge in poor parking. Of the reconstructed 49 devices, 17 stayed in place across more than one day; 11 lasted at least three days, six lasted a week or more, and two were still there ten days after I first recorded them.
I also logged 31 same-location morning-to-evening repeats, with as much as 7.6 hours between passes. A device that remains through an entire workday is not available to riders. It is occupying the route.
Live map
Where the trail kept getting blocked
Circle area tracks recorded sightings. Color shows the minimum observed stay: red 6+ days, orange 3–5 days, ochre 1–2 days, gray under a day. Basemap: OpenFreeMap, © OpenMapTiles, data © OpenStreetMap contributors.
The map uses the public clustered-location dataset, not the raw route log. Select a point to see its observation window. A static overview is shown automatically if the live basemap is unavailable.
How I counted
From 144 sightings to a cautious device count
I rode a fixed route, which makes each pass an observation of every recorded spot. Sightings within 40 metres were grouped into a location, then laid out in order across 14 passes. An unbroken run of sightings at one spot is treated as one occupancy episode: one device sitting there.
I treated one missing pass inside an otherwise unbroken run as a missed observation, rather than assuming an identical device vanished and another appeared between passes. The data supports that choice: one-pass internal gaps outnumbered gaps of two or more by 21 to 10, the pattern you would expect from intermittent observation rather than rapid turnover.
What the counts do and do not represent
The 49-device figure is a floor, not a total. Splitting every observation gap yields 70 devices, a strict upper bound; neither reading changes the core finding. Groups of devices at one spot and devices that arrived and left between passes make the true total higher. Dwell times are also minimums.
Brand was a visual observation, not a structured data field. The per-location map and table deliberately make no operator-level claim.
An uncertainty worth resolving
If my reconstruction is wrong, the problem is larger.
If each run is one device, units sat untouched for as long as ten days. That is a collection and monitoring failure.
If multiple devices cycled through the same spots, arrivals were higher than I measured and the same few coordinates absorbed a continuing stream of poor parking. That is a geofencing and end-of-ride failure on a larger scale.
The operator can resolve the distinction with its fleet telemetry. Either way, the first step is the same: check these locations, remove stranded devices promptly, and make the result visible.
What I’m asking for
A trail that stays usable.
- For operators: verify the recurrent coordinates, explain extended idle time, prevent trips from ending on the trail and bridge, and publish a removal standard measured in hours.
- For permitting authorities: make corridor geofencing and a documented removal standard enforceable permit conditions.
- For everyone responsible: measure dwell time, not just complaint volume, and coordinate across National Park Service land, Arlington County, and Alexandria.
What happens next
This is an ongoing field note.
I’m continuing the observation. The analysis can be rerun after an intervention to see whether the trail is actually being cleared. The detailed timestamped source data is available on request to an operator or agency that wants to compare it against internal records; it is not posted publicly because it would reveal a repeat commute pattern.
Contact Jordan about the studyData notes & limitations
Read this alongside the map.
- I rode the same route twice on each riding day, so this is what was visible from the trail at those moments. A device placed and removed between passes is not here; every count is biased downward.
- The 49-device figure is a reconstruction from 144 sightings, not a direct fleet count. Treating every observation gap as a new device produces 70 instead. Both figures miss devices in multi-device clusters and anything that came and went between passes.
- Coordinates came from a phone GPS fix while moving. Accuracy is typically a few metres and can be worse around the bridge; the 40 m clustering radius can merge nearby, genuinely separate parking spots.
- A run of sightings at one coordinate may be one stationary device or successive devices left in the same place. Only operator telemetry can distinguish those cases, and both readings call for intervention.
- Observed dwell times are minimums: a device could have arrived before I first saw it or remained after my final sighting.
- Brand is a contemporaneous visual observation, not a field in this dataset. The mapped locations and tables do not attribute individual devices to an operator.
- This first observation window runs from August 3 to August 13, 2026: 11 calendar days, eight with rides. Five of the 14 trips captured only one leg.
- The observations span 38.81543° to 38.88167° N and 77.03415° to 77.06244° W. Three reports are outside the core corridor and remain in every total.