Romania’s ‘weird’ earthquakes finally obey the rules of seismology

Beneath the sharp bend of the Carpathian Mountains in Romania lies one of the strangest seismic sources on Earth. The Vrancea zone is a compact nest of earthquakes that erupt not at shallow crustal depths, where most ruptures occur, but 60 to 200 kilometers down inside a near-vertical fragment of lithosphere sinking into the mantle. Its tremors shake Bucharest with an energy release that, on an annual basis, rivals all of Southern California, and its strong ground motions have long defied the standard models seismologists use everywhere else. Now a team of Romanian researchers says the mystery was never a mystery at all: Vrancea is simply complex, and once that complexity is properly untangled, its earthquakes behave exactly as seismological theory says they should.

In a study published in the Bulletin of Earthquake Engineering, Anabella Cotovanu, Elisei Cojan and Radu Vacareanu of the Technical University of Civil Engineering Bucharest and the Military Technical Academy Ferdinand I present a unified framework that decomposes the variability of Vrancea ground motions into source, path and site contributions. Their conclusion is striking: the apparent anomalies that have caused Vrancea to be treated as an exception and excluded from global comparative studies do not reflect fundamentally different physics. They arise from pronounced structural heterogeneity and the strong interdependence of source, path and site effects, which previous approaches failed to separate coherently.

The scale of the problem becomes clear when you consider how divergent the published estimates of key parameters have been. Stress drop, a measure of how much energy an earthquake releases per unit of fault area, has been variously reported for major Vrancea events at around 50 bar, 100 to 200 bar, as high as 850 bar statically and 950 to 1400 bar dynamically, and in the range of 300 to 1200 bar by researchers using empirical Green’s function methods. One influential series of analyses by Oth and colleagues employed stress-drop values roughly twenty times higher than commonly adopted reference values simply to make their models consistent with observations. Stress drop, in other words, had effectively become a correction dial rather than a physically constrained quantity.

The situation was hardly better for wave propagation. The literature contains six different attenuation functions describing how ground-motion amplitudes decay with distance outside the Carpathian Arc, split into two camps based on contradictory assumptions about geometrical spreading, with one group assuming amplitude decay proportional to the inverse of distance and the other to the inverse square root. Seismic refraction profiles from the VRANCEA99 and VRANCEA2001 experiments and the CALIXTO tomography project revealed why: the crust and upper mantle beneath the region are dramatically heterogeneous, with the asthenosphere lying at very different depths on either side of the arc, and sedimentary strata in the Focsani Basin bent to depths exceeding 30 kilometers.

Then there is the data problem, which is arguably the most severe constraint of all. Since the beginning of the twentieth century, Vrancea has produced fourteen earthquakes with moment magnitudes of at least 6.3, yet only four have been instrumentally recorded: the events of 1977, 1986, and the two shocks of May 1990. The devastating 1977 magnitude 7.4 earthquake, which killed roughly 1,500 people in Bucharest, was recorded within Romania at a single station. Attempting statistically robust source and path analyses with such a sparse database of major events is a formidable challenge, and one reason Vrancea has been left out of global ground-motion studies for decades.

The new framework tackles this problem head-on with a series of methodological innovations. First, rather than letting the recordings themselves define the grouping of data, the researchers built a zonation of the affected region from independent geological, tectonic, geophysical and geomorphological evidence, together with maps of the fundamental resonance frequency of near-surface sediments. This produced eight zones, from the Dobrogea Platform with its shallow ancient bedrock, through the deeply sedimented Dacian Basin accumulation zone, to the Focsani Basin with its exceptional sediment thickness and strong local topographic effects. Because the zoning is defined without reference to strong-motion data, it avoids circular reasoning and serves as an external pivot for organizing the region’s heterogeneity.

A preliminary spectral comparison across these zones immediately revealed systematic differences. The Dobrogea region showed a narrow amplification band peaking near 0.1 seconds, matching its independently mapped resonance frequency, while for larger earthquakes the Focsani Basin and the southern Moesia accumulation zone displayed pronounced amplification in the 0.5 to 2.5 second range. Critically, the two accumulation zones behaved similarly for moderate earthquakes but diverged sharply for large ones, suggesting that bigger ruptures activate deeper sedimentary packages. This supports what the authors call a dynamic boundary between path-controlled and site-controlled behavior, rather than a fixed depth dividing local site conditions from the deeper propagation medium.

The framework then performs what the team calls a uniformization of site effects. Because moderate-magnitude and large-magnitude datasets exhibit different site responses, an average ratio between their amplification patterns was computed and applied to the large-event data, bringing both datasets to a common reference site condition. This allowed the sparse large-event recordings and the more numerous moderate events to be combined into a single database for estimating path parameters. Stress drop was deliberately fixed at a physically plausible 100 bar and never used as a fitting parameter, a choice designed to avoid the source-site trade-offs that have plagued earlier work. Strikingly, when the analysis was repeated with stress-drop values ranging from 50 to 300 bar, the derived attenuation parameters did not change at all; the variations were absorbed entirely by site amplification, revealing the hidden coupling that had made earlier stress-drop estimates so unstable.

The results are internally consistent and, crucially, unremarkable. Geometrical spreading follows the physically expected inverse-distance decay for body waves, which dominate intermediate-depth ground motions. Frequency-dependent anelastic attenuation was determined independently for each zone, with coefficients of determination exceeding 85 percent for most zones and reaching about 98 percent in the best-constrained one. The zone-to-zone differences in attenuation, lowest over the North Moesia ridge and strongest around the Focsani Basin and Transylvanian side, directly explain why previous non-zonated studies produced such wildly different attenuation functions: they were averaging over irreducibly heterogeneous propagation paths. Site amplification envelopes derived for each zone are physically plausible and match independent stratigraphic evidence, with the deep basins amplifying long-period motion and the shallow-bedrock zones concentrating their response at short periods.

The implications extend well beyond Romania. Vrancea belongs to a very small global class of highly active intermediate-depth seismic nests, alongside Bucaramanga in Colombia and the Hindu Kush in Afghanistan, and its earthquakes shape seismic design requirements across more than two-thirds of Romanian territory. The long-period amplification characteristic of Vrancea ground motions is already embedded in the Romanian design code through spectral corner periods of 0.7, 1.0 and 1.6 seconds, far longer than the 0.3 to 0.5 second bands typical elsewhere, pushing the most severe seismic demand from low-rise buildings toward mid- and high-rise structures. The new framework, the authors argue, is transferable to any tectonic setting characterized by sparse strong-motion data and non-standard ground motions, provided sufficient independent geological and geophysical information exists to structure the regional heterogeneity. The broader lesson is perhaps the most satisfying one in science: what looks like an exception to the rules may simply be a complicated case in which the rules have not yet been applied coherently. Vrancea, it turns out, was never breaking the laws of seismology. It was just hiding them behind layers of complexity that no one had thought to peel apart.

Subject of Research: Seismic complexity and ground-motion characterization of the Vrancea intermediate-depth seismic source in Romania

Article Title: Defining and resolving seismic complexity through a unified source–path–site framework: the Vrancea intermediate-depth ground motions

Article References: Cotovanu, A., Cojan, E., & Vacareanu, R. (2026). Defining and resolving seismic complexity through a unified source–path–site framework: the Vrancea intermediate-depth ground motions. Bulletin of Earthquake Engineering. https://doi.org/10.1007/s10518-026-02713-4

Image Credits: AI Generated

DOI: 10.1007/s10518-026-02713-4

Keywords: Vrancea, intermediate-depth earthquakes, seismic hazard, stress drop, anelastic attenuation, geometrical spreading, site amplification, ground motion prediction, Romania, Carpathians, sedimentary basins, seismology