Safety Hazard Report - 3 Month Munich Pilot May-July 2026 DE · 2026-05-01 to 2026-08-01 · generated 2026-08-26 07:28 UTC

How much cycling this covers

Hazards detected

1,535 verified hazards · 23.2 per 100 km · 414 rider-triggered

When hazards happen

Where and why hazards happen

Every verified hazard mapped, with the twelve worst junctions ranked.

• Verified hazard (near-miss)
• Ranked junction - permanent street layout
• Ranked junction - construction or works
• Recorded accident 2024, slight
• Recorded accident 2024, serious or fatal
• Hazard density, low to high

Click a numbered hotspot to see its detailed analysis.

1,535 verified hazards. Density intensity is relative to the busiest point in this city, so colour is not comparable between cities. Basemap CARTO Positron, © OpenStreetMap contributors.

What this data has been used for

Two pieces of work built on the same near-miss record this report summarises: one street, and one city-wide maintenance problem.

1

Lindwurmstraße - what a protected lane is worth

Street-level safety analysis · June-July 2026

Munich planned to displace cyclists from the new protected lane on Lindwurmstraße onto the old pavement-level layout, to free a temporary bus lane during U-Bahn rail replacement. We used 124 individually video-verified near-misses plus the official accident record to quantify the trade.

After the lane - near-misses per 100 m

2.5protected stretch

7.3adjacent unprotected stretch

2.9×the gap the lane creates

Before the lane - accidents per 100 m

1.1same protected stretch

1.6same unprotected stretch

1.5×indistinguishable at these counts

The two rows are the argument. Same two adjacent mid-block stretches, same riders, same weeks; the only difference is the infrastructure. The official accident record, which predates the lane, shows them equally dangerous. After it, near-miss density differs threefold. The lane is the one demonstrated safety success on the street, from 124 video-verified near-misses and 30 recorded accidents.

Recommendation: do not remove the protected lane; set a start date for the worst stretch - 58% of the street's near-misses and 53% of its accidents in 40% of its length, funded in principle but unscheduled; re-examine the programme south of the railway overpass, where an unplanned stretch out-scores a prioritised one 19 near-misses to 5 and 3 accidents to 0.

Full analysis →

Rider safety analysis

Three of the most frequent Munich riders, compared. Riders are anonymised and no routes or locations are shown; only how much each rode and what was confirmed.

Validation

Two independent checks on whether the near-misses our riders record correspond to anything real: the official accident record, and a subjective rating of the same streets made by local cyclists. Neither was produced with this data in mind.

How this compares with recorded accidents

German Unfallatlas (Destatis) cyclist injury collisions for 2024, compared with the 1,535 near-misses our riders recorded here in 2026. Different years, different definitions.

Near-misses are the preventive measure: they happen before anyone is injured, and they are available within days, so a street change can be assessed in the season it is made. The accident record is the outcome we are trying to avoid, and it arrives years after the fact.

Rank correlation is Spearman on 250 m cells, restricted to the 539 cells we actually rode through; a cell with no riding cannot produce a near-miss, so including all 1,853 cells measures our coverage rather than the relationship (it gives -0.06). n=4 construction junctions is far too few to test the split properly; treat that half as indicative.

How this compares with a subjective street rating

MunichWays, a Munich cycling initiative, rate the city's cycling network by how it feels to ride - comfortable (gemütlich), average (durchschnittlich), stressful (stressig), very stressful (sehr stressig). It is a human judgement, recorded street by street by the people who ride them, and made with no reference to this data.

Verified near-miss hazards per kilometre of rated street, by the rating MunichWays gave that street. Whiskers are 95% bootstrap intervals, resampled over streets. Rated length per level: 237.5, 634.8, 563.5 and 69.3 km.

The near-miss rate rises at every step of their scale, and the streets they call very stressful carry roughly four times the hazard rate per kilometre of the ones they call comfortable. Two methods that share no data, no definition and no instrument arrive at the same ranking of the same streets.