How strongly Vrancea’s earthquakes reached every Romanian locality, 1980–2025

How many has your locality felt?

Pick a locality on the map, with a click or a tap, or type its name. Intensity IV means “felt by many indoors, dishes rattle”; V, “felt by nearly everyone”; VI, “felt by all, heavy furniture moves”. For localities inside the Carpathian arc, and for those where the law overshot the 1977 and 1990 observations by a unit or more, the numbers are upper estimates (marked * in the table and “at most” on the map).

At equal distance, direction decides

The table below uses no attenuation law at all. The 747 macroseismic points of 1977 and the 1,618 of 30 May 1990 were split by distance to the focus and by direction: along the north-east to south-west band, or across it. At 300–400 km from the focus, places along the band have a median of VII in 1977 (43 points) and 5.5 in 1990 (109 points); places across the band, at the same distance, have 4.2 and III.

That is why circles around the epicentre would mislead. Vrancea shaking stretches as an ellipse towards Iași and towards Bucharest and dies quickly north-westwards, over the mountains, and south-eastwards. Why the earthquakes happen there at all, 60–200 km down, is told on the page about the slab under the Carpathian bend, linked at the end.

The big nights: 1977, 1986, 1990, 2004

Five earthquakes of magnitude 6.0 or more have left the nest since 1977. Values without a mark are observed (median of macroseismic points within 20 km); values with * are modelled with the law in the method, because no equally dense open list of locality observations exists for 1986, 31 May 1990 and 2004. The 1977 earthquake sits outside the 1980–2025 counting window and appears here as the yardstick. Pick a night under the map to see it spread over the whole country, then tap your locality.

Method and limits

The catalogue is ROMPLUS, from Romania's National Institute for Earth Physics, downloaded on 9 November 2026. Kept: focus at least 60 km deep, epicentre inside the Vrancea nest (45.2–46.0°N, 25.9–27.0°E), moment magnitude at least 4.0, between 1 January 1980 and 31 December 2025: 369 events. The network catches every earthquake from this size up throughout the period: the yearly number at 3.5 or more stays between 25 and 37 in every five-year span, with no rise as the network grew. The number at 4.0 or more does fall, though, from 10–13 a year in 2005–2014 to 3–5 a year after 2015, and against the mb magnitudes of the European EMSC catalogue, ROMPLUS magnitudes after 2015 sit about 0.3 lower than before 2010. Part of that fall is therefore probably down to how magnitudes are worked out, not to the nest. With post-2015 magnitudes raised by 0.3, the total would rise to as many as 491 rather than 369, and Bucharest would have 28 rather than 26. Shallow crustal earthquakes (Banat, Făgăraș, Gorj) are absent throughout: they weaken by a different rule and this page would count them wrongly.

Intensity in a locality was computed with Burtiev's azimuth-dependent macroseismic-field equation for Vrancea intermediate-depth earthquakes (I = 1.6·Mw − b(γ)·lg√(h²+r²) + 7.2, with b from 4.9 along the NE–SW band to 5.6 across it; doctoral thesis, equation 15, fitted by the author on the macroseismic data of 1940, 1977, 1986 and 1990). No coefficient was changed. A locality's point is the centroid of its administrative polygon (3,181 units, ANCPI via geo-spatial.org). A locality's “observed” value is the median of macroseismic points within 20 km or, failing that, the nearest point up to 60 km away.

The comparison with observations: the law's author fitted it on the 1977 and 1990 earthquakes themselves, so the comparison shows how well it reproduces them, not an independent test. Against the USGS macroseismic points, the law's mean absolute error is 0.88 units in 1977 (747 points) and 0.75 in 1990 (1,618 points), with a mean overshoot of about 0.6 units, part of it from the MSK-64 to MMI scale step. The only large earthquake outside the fit, in 2004, has just 16 observed points: mean error 0.91 units, overshoot 0.55. At Bucharest and Iași the misses are a tenth of a unit; close to the nest (Focșani, Bacău, Ploiești) the law gives 0.7–1.0 units too much, and at Brașov and Cluj 1.0–1.7, because its ellipse has no term for the Carpathian shielding. So for 1,671 of the 3,181 localities (those inside the arc, and those where the law overshot the observations by a unit or more, averaged over 1977 and 1990) the numbers are printed as upper estimates, and every IV+ count carries its ±½-unit range. The law was fitted on earthquakes of magnitude 6.4–7.7, but the counts come mostly from smaller ones: in Bucharest, 23 of the 26 are below 6.4, the smallest 4.8. For these the law is extrapolated, and the range shows how much that matters. The test set in advance, that Bucharest would have at least twice as many IV+ earthquakes as Cluj because of direction, failed on its own terms: Cluj has too few (3, under the threshold of 5), and with the direction term removed the ratio is still 14 to 5, so distance does most of the work. The newer Romanian equation with an arc term (Văcăreanu and co-authors, Natural Hazards, 2015) keeps its coefficients behind a paywall and could not be used without guessing them.

This page speaks only about the past. It says nothing about what a future earthquake would do to any person or building. The official seismic design values for each locality live on the linked page below, about how hard the ground shakes in each locality.

Modelled intensities are medians of a statistical law: one place, in one earthquake, can sit a unit or more away from them.

On the same subject, on this site:

• How hard the ground shakes in each locality (the official hazard values)
• Why the quakes strike deep beneath Vrancea
• How many seconds before the strong wave
• The big one isn’t overdue: Vrancea has no clock
• The night Bucharest fell: 4 March 1977
• Every deep Vrancea earthquake since 1998

Sources: National Institute for Earth Physics, ROMPLUS catalogue; USGS ShakeMap Atlas (macroseismic points 1977, 1990, 2004); Burtiev R., doctoral thesis (2017) and Burtiev and co-authors on stochastic seismicity models; Kronrod and co-authors, Tectonophysics 590 (2013); ANCPI administrative boundaries via geo-spatial.org; SIRUTA register (National Institute of Statistics). The European EMSC catalogue, queried as a cross-check, returns 331 events at nominally the same cuts but on mixed magnitude scales.