Google sued over collapsed bridge death

Certainly in the UK people who had tried to “blame the satnav” have been prosecuted. See also here.

1 Like

it may depend on the jurisdiction, but generally around here, I do not have any doubt that you will have no chance to sue a map editor when you drove into a hole because there was no road.

The bridge including the road was still shown on Google Maps until September 21, 2023, checked by myself. Even cars were routed from north to south over the bridge, but not vice versa.
Probably that’s why the road has now been completely removed by Google.

1 Like

How about using butt (and also putting the terminating casing line on a higher layer) if there is noexit=yes?

will break for things like (1) gap is very small and butt ending still merges roads, (2) there is footway bridge connecting them (so no valid noexit=yes)

may be technically impossible for CartoCSS/Mapnik setups

Pretty much any navigation software vendor has a disclaimer of warranty. I don’t know if that will be a successful defense against an allegation of negligence. The lawsuit claims that a local resident suggested an edit to Google Maps to remove the bridge on two different occasions before the accident, but it got stuck in review or didn’t get accepted, and that other locals also attempted unsuccessfully to get the data corrected.

According to the suit, the victim had been following a route recommended by the application. In other words, they weren’t eyeballing it, as some have implied here.

Apparently the bridge and road were on private property – the suit also names the owners as co-defendants. The state highway department says they didn’t have the authority to erect barriers on the owners’ behalf.

1 Like

yeah, it does not always work - but in general is worth doing, especially in cases like this one

(though it is possible that in at least some jurisdictions state may have some tools to handle such hazards, even if that would not be specifically state highway department)

That satnav article says

It’s a good idea to carry a road atlas for back-up, too.
Check the route is practical before you go
Is the sat nav taking you to the right Farnborough? Hampshire, Warwickshire or Kent?
Does the route look right?

That’s what I’m saying. The road looked “right” in Carto, despite a gap: no bridge over the river. Problem clearly visible in e.g., iD. But Carto creates bridges anyway. That is very dangerous.

Think of electric circuit board software. If it decided to connect loose ends… boom!

In my whole life I never did anything else, that is the way I grew up with. I also grew up with watching my step though.

Carto does not create bridges there. Both screenshots clearly show that.

We are not talking about routers, we are talking about block-heads that insist on learning from looking at a map.

Wondering just now: Can it be made so that barriers, gates, jersey, plastic fences and so on, appear at same Z as the roads they block? Might be hard too, but maybe less hard than shifting nodes?

1 Like

Yes, what jidanni wants is not technically possible. The first issue is that we’re forced to render the linestrings, not the points, so even if we knew a point was noexit, we couldn’t render the linestring differently. Secondly, even if we did know that we wanted to render one end of a linestring with a butt end-cap, that’s not possible because we can only set the end-cap for the entire linestring.

@jidanni, you keep ascribing intention to the rendering and maintainers that we intend to display a bridge in cases like this, in spite of being told multiple times it is a result of technical constraints and it’s not possible to do what you describe. If you feel the technical constraints do not exist, you are welcome to demonstrate that with code. However, if Andy, Mateusz, Matthijs, Christoph, and myself all tell you that something is not possible, you should consider that we might be right.


No. Roads appear well before barriers, and are much thinner when the first appear. Residential roads appear at z12 and service roads and footpaths at z14. The roads are too thin to fit a symbol within.

There’s also a density problem - you’d have to show all barriers at z14. In fact, because bike paths are shown at 13, it’s even worse. Most of these bollards are going to be on a bike path


I’m aware of that. The case that prompted this discussion demonstrates that incorrect routing carries the larger real-world risk. But the actual facts of the case have been mostly ignored in this discussion in favor of a particular obsession about cartographic generalization. It’s a shame, because this could instead be an opportunity to consider whether our own system for triaging feedback is up to the task.

1 Like

Unfortunately, I do not see that happening with how it was raised by the original user.


Depends on how it’s done?

Obviously, a mapper reading a news article on the day the bridge first collapsed, thinking Where’s that bridge?, finding it & marking it as destroyed would be the best way of doing things!

Next would be either posting a message via one of the options (here, Slack etc) to say this bridge has collapsed, please mark it as so, or even raising a DWG issue to say the same (although we’d really prefer that people used other options first!)

Worst, & definitely least preferable, would be posting a Note or FixMe to say that The bridge is out, because, as we all know, it could literally be years before that’s spotted & fixed :frowning:


The bridge was one of several that were destroyed in a flood, but the only one that was privately owned. A dispute over the ruined bridge’s ownership, and the resulting delay in making repairs, made the front page of the local newspaper in 2014:

In OSM, the road had gone untouched for years after the TIGER import. No one had ever created a map note to report the issue. This is not entirely surprising, since OSM isn’t exactly a household name in the U.S. Local residents were understandably focused on getting Google to acknowledge the problem.

There had been a validator warning about a waterway crossing without a bridge, but even if someone had gotten around to resolving that issue, I’m unsure if they would’ve correctly identified the bridge as being impassable without without some local knowledge. In some of the available aerial imagery sources, the bridge is obscured by tree cover, and in others, the gaping hole looks kind of ambiguous. The nearest street-level imagery was taken from a distance, and the kind of barricade used there might easily have been mistaken for a temporary construction barricade.

The same day that the local newspaper reported about the accident, a new user made their first edit to remove the bridge. They mentioned the accident to justify the edit. Another mapper quickly followed up to map some presumed barriers to prevent someone from unknowingly restoring the connection. The accident also caught the attention of mappers in OSMUS Slack a couple days later, once it made national headlines.


After reviewing the arguments, I see that the issue is in fact only that this gap cannot be rendered at lower zoom factors.

There is a risk involved for drivers who navigate only by looking at the map, or having a passenger look at the map to issue directions. Yes, I can see that occurring, even when it shouldn’t. I am from the days when paper maps and road books were the only navigation system availabe.

My conclusion is that the only practical way to reduce the risk is, to widen the gap. This would bend the truth a little, not by be being wrong but by leaving out short sections of the road. The fact that it is about private property makes this solution more acceptable, I think.

I know cases where clearly visible paths in the wood were removed from OSM because they were actually large animal tracks, walkable but carrying a risk for adventurous hikers. We thought the risk outweighted the desire to map the complete truth. We also got the forest owners to block the path from view by erecting a “natural” barrier of logs and brushwood.

OSM Carto employs a cartographic generalization technique called exaggeration, thickening roads to make them more discernible. Not every OSM-based map widens roads as much, but OSM Carto’s approach is by no means unusual, and Wikipedia even uses OSM Carto to demonstrate this approach. Exaggerating a roadway’s thickness allows the style to accommodate the road name label within the line, but it can cause the road to overshoot a T-intersection or swallow a short gap or tunnel. Some tools are capable of restoring gaps between roads using a related technique called displacement:

Automated displacement isn’t currently supported by some of the rendering technologies that popular OSM-based styles rely on. In response to this thread, someone has already attempted to work around this technical limitation by widening the gap directly in the data, but I have a feeling someone in this thread will soon zoom out one more level and send us right back to the drawing board. (Google is taking no chances by deleting the road outright.)

I’m certainly not going to revert the widened gap, but let’s not overstate the importance of this workaround. I wonder if we would apply the same workaround for everything that might interrupt a road for a similar distance. In some places where I map, I still occasionally encounter untagged fords across rivers. An unsuspecting motorist entering a deep river or swift current could be in just as much danger, yet I don’t think we’re very likely to see a campaign to delete fords just because they don’t have icons at zoom level 15.

I quibble with one part of this statement. The most practical way to reduce the risk was to create any gap at all, specifically to keep routers from even going near the street. For better or worse, a typical motorist trusts their satnav’s active guidance instructions much more than a paper map that they have to squint at, moreso at night along an unlit side street as in this case. That is why I keep pushing back at the suggestion that more sophisticated cartographic design would’ve saved the day. I recognize that hiking and orienteering in difficult terrain carries different risks and tradeoffs.


Also, I would expect that people using OSM data for car navigation are not using OSM Carto to navigate anyway.

Interesting is that the way is said to be on private property. All it needs is an access=private or an access=no + foot=yes as there’s supposedly a footbridge used with permission. Of course if the person’s driveway is connected to that way, then s/he can’t navigate home, a minor detail.

The story is more complicated than that – the land owners did not mind people using the bridge so the access was in fact public, but when it collapsed they did not feel responsible for the repair either. Neither did the county.

1 Like