Figures not available

Rollback of the Carpathian subduction zone (e.g. Royden, 1988) into an oceanic embayment of the European foreland was accompanied by steepening of the subducting lithosphere which resulted in slab segmentation (i.e. fractures within the slab extending along dip separated the slab into small pieces). Due to oblique convergence between the intra Carpathian microplates and the European foreland asynchronous continental collision occurred starting in the north, in the Western Carpathians, during middle Miocene (16 Ma) and ending in the Southeast, in the Carpathians, during late Miocene (10 Ma; Jiricek, 1979). In the same manner, post collisional slab break off started in the north and migrated towards the south(east), so that one slab segment after the other was detached and sank into the lower mantle (Sperner, 1996). Beneath the Vrancea region the last of these segments, which is still attached or just about to detach, gives rise for the spatially limited occurrence of seismicity. The post-collisional history of oceanic lithosphere has become a major research topic (Wortel and Spakman, 1992; Davies and von Blankenburg, 1995) with the result that slabs might detach. Additional physical insight comes from estimates of the deformation rate of the source volume, from recent tomographic results and from models of stress distribution within a passive slab. This paper presents these facts and their implications for Vrancea tectonics.

All events occur in a compressive regime with thrust tectonics at intermediate depth. The fault plane solutions of the instrumentally recorded large earthquakes are remarkably similar. They typically strike SW-NE (~ 220°) and dip 60° to 70° to the NW. The slip angle is roughly 90°. The similarity of the fault plane solutions is strongly contrasting with significant variations of the radiation patterns which reflect the dynamics of the rupture process. The intensively studied 1977 event (Müller et al., 1978, Rökers and Müller, 1982) is characterized by a sequence of shocks that propagated to the SW, enhancing radiation in this direction. In agreement with this seismological result the damage was more severe to the SW of Vrancea as compared to the NE. This contrasts the damage pattern of the stronger 1940 event which caused less damage in Bucharest than the weaker 1977 earthquake.

The frequency of Vrancea events can be assessed from available catalogues. Earthquakes with Richter magnitudes in excess of 6.0 taken from Radu's (1991) compilation covering this century and transformed to moment magnitudes with Oncescu's (1987) relation satisfy a Gutenberg/Richter law (Lungu and Coman, 1994):

TABLE 1: Comparison of moment release rates for various tectonic active areas of the Mediterranean and the Middle East (after Jackson and McKenzie, 1988).

Moment release rates can be cast in strain rates with Kostrov's (1974) relation between moment and strain tensors. Taking the moment tensors of the last 4 large Vrancea earthquakes (Oncescu and Bonjer, 1997), a source volume of 2.5·105 km³ (70×30×130 km) and a mantle shear modulus of 7.5·$1010$ Pa gives a strain rate of e = 2·10$-7yr-1$. With this strain rate the source volume is stretched in the vertical direction and shortened in NW-SE direction. Again this strain rate compares well with values from active plate boundaries such as Southern California (Ward, 1994; Fig. 3) where a similar amount of strain is observed by geodetic means. We elaborate this point in order to show that Vrancea seismicity, given its seismic moment release and associated strain rates, is at plate boundary scale. The estimated strain rate deforms a vertical slab with 130 km length by 2.6 cm/yr in vertical direction. Upscaling of this value to a geological times provides a value of 26 km vertical elongation in one million years.

Lorenz et al.(1997) restricted the input data used for inversion of teleseismic tomography to those stations of the Romanian network with digital recording. Though the number of stations are thus reduced as compared to Oncescu et al. (1984) the results are more significant due the improved data quality. Figure 3 shows depth levels of 74 to 112 km (Fig. 3a), 112 to 152 km (Fig. 3b), and 152 to 194 km (Fig. 3c). Each panel contains the sites of digital stations as triangles. The hypocenters of small earthquakes as observed during 1980 and 1994 are marked as dots and the sites of the large events between 1940 and 1990 as crosses. Both the pattern of stations and the azimuthal distribution of teleseismic earthquakes limit the volume that is illuminated by seismic rays and the resolution of the inversion. Given these constraints we consider the following features as significant:

The images show blocks of low velocities (- 4%), which surround an area of higher velocity (+ 4%) The high velocities extend from a subcrustal depth of 40 km all the way down to the maximum depth of 236 km mapped during inversion. The recent inversion of local events (Fan et al., 1998) shows results consistent with this observation. At intermediate depth the elevated velocities appear only beneath Vrancea. All intermediate depth earthquakes are contained in the high-velocity volume. It appears that the seismogenic volume is significantly smaller as compared to the high-velocity volume. The earthquakes occupy a small strip within the high-velocity body that tends from the NE at a level around 90 km (Fig. 3a) to the SW at the level where the deepest earthquakes are observed. It is also noteworthy that no specific velocity structure within the limits of resolution characterizes the depth level of very low seismicity (40 to 70 km) and the depths below the deepest events. On a global scale all images of active subduction zones are characterized by positive P-wave velocity anomalies. We take this as strong evidence for the presence of a slab of oceanic lithosphere beneath Vrancea. The volume of elevated velocities displays a peculiar structure, the main feature of which is a change of strike from top to bottom. The uppermost high velocity material has a strike direction roughly SW-NE. This is consistent with a NW-ward subduction. The deeper portions of the high-velocity material, however, are oriented S-N. This direction is close to the current strike of the Eastern Carpathians. Presumably this is also the strike direction of subduction before the last segment of lithosphere - the Vrancea segment - was incorporated into the mantle. Thus the directional change of subduction during the last phases of collision in the Eastern Carpathians is preserved in the tomographic image. The S-N oriented high-velocity material represents older W-ward subducted material that is detached from the foreland lithosphere but still attached to the SW-NE trending Vrancea slab. This additional portion of heavy oceanic lithosphere generates a pull-down force which will increase shear stresses in this passive slab.

sigma(h) = rho(m)·g·z (2a)
sigma(v) = rho(s)·g·z - (rho(s)- rho(m))·g·L (2b)

with a total slab length L, gravitational acceleration g and depth z. Within the slab the stress field changes from tensile at shallow depth z < L. (rho(s)-rho(m))/rho(s) to compressive at deeper levels. As this model is rather crude for shallow levels we focus on those depths that are relevant to intermediate seismicity. With (m = 3200 kg/m³, (s = 3400 kg/m³, g = 10 m/s² and L = 200 km the stresses are compressive below 12 km depth. The buoyancy effect amounts to 0.4 GPa, so that the maximum differential stress decreases from 0.4 GPa to zero at the lower end of the slab where lithostatic conditions prevail. Thus this simple model is consistent with the stress field suggested by the fault plane solutions. However, the maximum possible shear stress remains on a fairly low level at the seismogenic depths between 70 (260 MPa) and 200 km (no shear stress). If we assume that the seismogenic volume differs from the high-density volume such that there are parts of the slab that extend below the deepest level of seismicity more realistic parameters apply. With an additional 150 km of non-seismic slab the level of maximum possible shear stress exceeds 300 MPa everywhere within the seismogenic volume. Although the frictional strength of rocks under very high confining pressure (5 GPa) are not known, estimates of lithosphere strength using plate bending geometry (Goetze and Evans, 1979) indicate that this level of shear stress may be sufficient to rupture oceanic lithosphere.

$Mw\; =\; 4.07\; +\; 0.98·log$₁₀A (3)

with rupture area A in km². The variance of Mw in this estimate amounts to ( = 0.24. A comparison of observed values of A (Oncescu and Bonjer, 1997) and those derived from relation (3) compare favorably (Table 1), so that (3) can be applied to large Vrancea earthquakes. Table 2 shows the prediction of fault plane sizes and their linear dimension ($A½$) taken as square root. Obviously an event with magnitude Mw = 8.5 would not fit onto the high-velocity volume, so that tomography suggests that the maximum possible earthquake is in the range of Mw = 8.0

TABLE 2: Magnitude (Mw) versus rupture area (A in km²) according to observations of the large Vrancea events (Oncescu and Bonjer, 1997) compared with the areas predicted by the empirical relation (3). $A½$ is the square root of the area and represents the scale of the linear dimension of the ruptured area.

TABLE 3. Rupture areas (A in km²) and their linear dimensions ($A½$ in km) are predicted by the empirical relation (3) for large moment magnitudes (Mw).

Figures: (müssen in den Text integriert werden)

1. Figure 1. Topographic map of Romania with intermediate depth seismicity (> 60 km) of this century (for details see Oncescu et al., this volume).
2. Figure 2. Depth and magnitude of the last 4 large Vrancea earthquakes.
3. Figure 3. Horizontal slices of upper mantle between 74 km and 194 km depth of the Vrancea area. Note that all recorded seismicity (dots and crosses) is confined to the high-velocity areas.