If this bridge
disappears

How we calculated it

Real roads.
Hypothetical closures.

Simulation based on OpenStreetMap · 6 September 2026. Modelled times, without traffic, ferries or routes abroad.

What enters the network

Roads come from the Geofabrik OpenStreetMap extract for Romania. Intersections connect directed segments; one-way roads constrain which direction a route can follow. Each segment’s length and assigned speed determine its estimated time.

What we close

A bridge can appear as several map segments, including separate carriageways. A closure joins segments with the same road number less than 70 metres apart: an approximation of the bridge, not an engineering inventory. The Fetești and Cernavodă bridges, in series on the A2, close together. Every scenario starts from the original network: closures do not accumulate.

We also tested four Carpathian road segments (Predeal and Tihuța passes, Bicaz Gorge, the Olt Valley road), each as a single segment in the data. One segment does not close a pass, and the results (0.7 to 25 minutes) do not show what closing the pass would cost; we removed them from the table.

What we compare

We analyse 1,560 directed journeys between 40 cities. Outbound and return routes are calculated separately. Of 2,755 bridge groups used by baseline routes, we tested the 40 used by the most routes, alongside four named crossings.

This selection favours crossings used by many sampled routes. A rarely used bridge can cause a large detour and still be absent from the selection.

The opening example is the shortest route that gains at least five minutes, not a typical route: for Giurgeni–Vadu Oii Bridge, 75 of 1,560 routes change, by 23.5 extra minutes on average.

What the model does and does not do

It does: take the public roads in OpenStreetMap as of 6 September 2026, remove one crossing and search again for the fastest route between the 40 cities, using the speed assigned to each road type. The result is how many extra minutes each route needs, in the model.

It does not: measure traffic, say how likely a closure is, or judge a bridge’s condition. It excludes ferries, detours through other countries (for example Giurgiu–Ruse or Calafat–Vidin) and turn restrictions. It does not weight cities by population or estimate how many people would be cut off. It does not rank all Romanian bridges, only the ones tested.

Checked against another router

We repeated the closures with Valhalla, a route planner that also reads OpenStreetMap but builds its own graph, speeds and turn rules; it receives the closure as a forbidden zone. For Slobozia–Hârșova without Giurgeni–Vadu Oii Bridge, Valhalla gives 2 h 8 min instead of 0 h 57 min: +71 minutes, against +59 in the model; the detour is 134 km in both.

On the three crossings over water (A2 over the Borcea and the Danube, Brăila, Giurgeni), across 24 affected routes drawn at random, 22 differ from the model by at most half or at most ten minutes; the median ratio of Valhalla to the model is 1.1. On the motorway bridges the planners part ways: across 40 routes the median ratio is 2.1, and 24 fall within the same limit. The model’s motorway numbers show an order of magnitude, not a reliable estimate. Of 23 routes the model leaves unchanged, 22 stay under five minutes in Valhalla too.

The check compares the results of two planners; it cannot establish that the map is correct. Both read OpenStreetMap, and Valhalla uses today’s data, not the 6 September snapshot, so a missing bridge or a wrong link in the map can affect both results. For Valhalla we forbade only the middle 200 metres of each closed segment, not the whole bridge, and its route was not compared segment by segment. Check results · JSON