Slab rollback
The slab retreated south-eastward, dragging the Carpathian arc with it; the leftover hangs near-vertical under the bend (the Matenco et al. line).
VRANCEA · 3D MAP
Why Do Quakes Strike Deep Beneath Vrancea?
Romania's big earthquakes don't start at the surface. They start a hundred kilometres down, in a corner of subducted lithosphere under the bend of the Carpathians. That is why they are felt from Budapest to Moscow.
- Depth
- 60–200 km
- The nest
- 70 × 30 km
- Largest
- 1940 · 1977 · 1986
Pick the quake
The map, cross-section and key data update for each event.
SLABWhat sits under the Carpathian bend?
The Carpathians change direction abruptly at Vrancea: they arrive from the north-west, then veer south. Right under the bend, between 60 and 200 km deep, hangs a remnant of old subducted lithosphere, near-vertical today. The big quakes are born in it, not at the surface.
1 · The bend
The Carpathian arc kinks at Vrancea. Under the kink, lithosphere sank into the mantle — the only place in Europe where this happens at scale.
2 · The slab
A relic of lithosphere left from when a vanished sea sank here — usually interpreted as oceanic plate. Subduction stopped about 8–9 million years ago; the leftover hangs near-vertical.
3 · The nest
Ruptures crowd into a horizontal footprint about 70 by 30 km, elongated NE–SW, 60 to 200 km deep — isolated from the small crustal quakes. One of three widely recognised nests of this kind in the world, with the Hindu Kush and Bucaramanga.
4 · Why the surface doesn't crack
At a hundred kilometres, the slab's cold brittle rock breaks suddenly; on one proposed mechanism, hydrated minerals (antigorite) would lose water and grease the rupture. Above the nest the crust is comparatively quiet — the damage comes from depth.
Schematic block, true vertical scale. The slab surface is stylized; hypocentres sit at published depths.
EVENTSThe six major quakes, instrumentally recorded
Since 1940, Vrancea has produced six earthquakes of magnitude 6 or more. All started at 79 km or deeper. Pick a year above to see the quake on the map, in the section and in its key data.
| Date | Mw | Depth | Epicentre |
|---|---|---|---|
| 1940-11-10 | 7.7 | ~150 kmpublished variant: 124–160 (range across consulted catalogues) | 26.932° E, 45.753° N |
| 1977-03-04 | 7.4published variant: 7.5 (Wikipedia EN) | 94 [85.3] kmpublished variant: 85.3 km (Wikipedia EN) | 26.800° E, 45.800° N |
| 1986-08-30 | 7.1published variant: 7.2 (Bala, EGU 2010) | 131 km | 26.500° E, 45.500° N |
| 1990-05-30 | 6.9 | 89–91 km | 26.900° E, 45.900° N |
| 1990-05-31 | 6.4published variant: 6.3 (Wikipedia EN); 6.1 (Wikipedia EN (MGR)) | 79–87 km | 26.997° E, 45.883° N |
| 2004-10-27 | 6.0 | 90–100 km | ≈ 26.750° E, 45.750° N |
Magnitudes (Mw unless stated otherwise) and depths from catalogues and reviews; published variants alongside. The row marked ≈ uses the nest centroid as epicentre. Each value's provenance is in the Method below.
SECTIONA cut through the nest
NW–SE section through the nest centroid (26.75° E, 45.75° N). Each circle is a table quake at its published depth; the hatched band is the schematic slab.
True vertical scale, 0–200 km. The slab band width is illustrative.
FELTWhy is it felt as far as Moscow?
Waves leave from about 100 km down, and the felt area is unusually broad: depth, propagation paths and site effects all contribute. The regional average of the isoseismal axis is about 54°±2° NE, and the Carpathians bend their shape.
Schematic felt-area ellipse for 1977. Axis from the literature; extent illustrative.
- 1977
- Felt in almost all Balkan countries; damage in Svishtov, Bulgaria, too.
- 1986
- Damage in Chișinău, about 200 km from the epicentre.
- Typical
- For 100–150 km Vrancea quakes, effects documented in the Moscow area.
On the Bucharest night, separately: The 1977 earthquake.
MODELSWhat is still unsettled
Researchers agree that a slab sinks under Vrancea and that quakes break in it. How the slab got there and what moves it today — that is debated. Three families of explanations:
Delamination
The heavy lower lithosphere peeled off and sinks like a drop — “drop forming” (Koulakov et al., 2010).
Slab break-off
The slab may have torn from the rest of the lithosphere, the torn piece sinking freely (the Wortel–Spakman detachment line).
At rupture scale, one proposed mechanism is dehydration embrittlement: antigorite would lose its water between 60 and 135 km, the rock would weaken suddenly and crack (Ferrand et al., 2021). The mechanism stays debated; the model covers 95% of events at 60–170 km.
METHODMethod
The table centralises magnitudes, depths and epicentres with each value's provenance: 1940 from catalogues (depth differs between catalogues, 124–160 km; the table uses ~150); 1977 and 1986 from reviews (Mw 7.4 @ 94 km, Mw 7.1 @ 131 km); 1990a Mw 6.9 @ 89–91 km; 1990b from a peer-reviewed study (Mw 6.4 @ 82 km, 45.883° N 26.997° E); 2004 Mw 6.0 @ 90–100 km from compilations. Where printed sources differ, the variant sits alongside, scale and source named. Nest geometry (70×30 km footprint, 60–200 km, NE–SW axis) and slab parameters (subducted lithosphere, usually interpreted as oceanic plate, near-vertical; subduction stopped ~8–9 million years ago) come from tomography and reviews cited under Sources.
The 3D map uses Terrarium relief (z10 mosaic, max filter) over the Carpathian-bend frame, with 2.5× vertical exaggeration for legibility. The slab block is procedural, at true vertical scale: hypocentre depths are the published ones, the slab surface is stylized. The NW–SE section projects epicentres onto the 54° axis through the nest centroid (26.75° E, 45.75° N).
LIMITSLimits
The felt ellipse is schematic. The 54° axis comes from the literature; shape and extent are illustrative, not surveyed isoseismals. Don't read intensities off it.
Approximate epicentres. For 1990b and 2004 the consulted sources give no precise epicentres; the page places them at the nest centroid and marks them ≈ everywhere (map, table, section).
Geodynamic models are open. Rollback, delamination or slab break-off — the page presents all three families without picking one; the literature has no consensus.
Not a forecast. Nothing here says when the next quake comes or how hard it will shake any address; for official per-locality hazard, the dedicated sibling page.
SOURCESSources
- Event parameters: Bala et al. (EGU 2010 reviews: 1940 Mw 7.7, 1977 Mw 7.4 @ 94 km, 1986, 1990); Radu et al. 1979 (1977 macroseismic field: 45.8° N, 26.8° E, h 95 km); Pantea & Constantin, EGU 2007 (macroseismic re-evaluation); the events' EN Wikipedia articles (1940, 1977, 1990: epicentre coordinates, variants).
- Nest and slab geometry: Koulakov et al. 2010 (local tomography, high-velocity body 60–200 km); Matenco et al. 2016 (relic slab, subduction stopped ~8 Ma, near-vertical today); Mitrofan, GJI 2016 (25×70 km projection); Springer 2022 (focal-mechanism catalogue 60–180 km).
- The world's nests: the Vrancea – Hindu Kush – Bucaramanga analogy (Gutenberg & Richter 1965; AOS and Frontiers 2025 syntheses).
- Mechanism: Ferrand et al., Sci Rep 2021 (antigorite-dehydration embrittlement, 60–135 km; 95% of events at 60–170 km).
- Felt area: 54°±2° isoseismal axis (Shumila, via the SCIRP synthesis); Iosif & Iosif 1979 (Carpathians bending the isoseismals); Rogozea 2013 (Moscow-area effects for typical 100–150 km events).
- Relief: Terrarium (z10 mosaic), via the se-vad-muntii coarse grid; OSM for visual checks.
FAQFrequently asked questions
How deep are the Vrancea quakes?
Between 60 and 200 km, most between 70 and 170 km. The 1977 quake started at about 94 km, 1986 at 131 km, 1940 at about 150 km.
Why right under Vrancea?
Because that is where the Carpathian arc bends, and under the bend hangs a remnant of old subducted lithosphere, near-vertical. Ruptures happen inside the plate, in a footprint about 70 by 30 km.
How many seismic nests of this kind exist?
Three widely recognised ones with similar peculiarities: Vrancea, the Hindu Kush (Afghanistan) and Bucaramanga (Colombia). Vrancea is the only known one in Europe.
Why is it felt as far as Moscow?
Waves leave from about 100 km down and cross old, cold lithosphere, which attenuates them little toward the north-east. That is why isoseismals stretch NE.
Can the next Vrancea quake be predicted?
No. Models estimate hazard — how hard an area can shake over a long window — not the date of the next quake. Any publicly circulated date is speculation.
What about the 1802 quake?
It was the largest Vrancea quake ever felt (the Colțea tower fell then), but it is pre-instrumental: its depth is unknown, so it stays out of the table of six.