Relative seismic moment tensor determination for Vrancea intermediate depth earthquakes

The method of relative seismic moment tensor determination proposed byStrelitz (1980) is extended a) from an interactive time domain analysis to an automated frequency domain procedure, and b) from an analysis of subevents of complex deep-focus earthquakes to the study of individual source mechanism of small events recorded at few stations.The method was applied to the recovery of seismic moment tensor components of 95 intermediate depth earthquakes withM _L=2.6–4.9 from the Vrancea region, Romania. The main feature of the obtained fault plane solutions is the horizontality ofP axes and the nonhorizontal orienaation ofT axes (inverse faulting). Those events with high fracture energy per unit area of the fault can be grouped unambiguously into three depth intervals: 102–106 km, 124–135 km and 141–152 km. Moreover, their fault plane solutions are similar to ones of all strong and most moderate events from this region and the last two damaging earthquakes (November 10, 1940 withM _W=7.8 and March 4 1977 withM _W=7.5) occurred within the third and first depth interval, respectively. This suggests a possible correlation at these depths between fresh fracture of rocks and the occurrence of strong earthquakes.     

Estimate of earthquake hazard in the Vrancea (Romania) region

A maximum likelihood method is used to estimate the earthquake hazard parameters maximum magnitudeM _max, annual activity rate , and theb value of the Gutenberg-Richter equation in the Vrancea (Romania) region. The applied procedure permits the use of mixed catalogs with incomplete historical as well as complete instrumental parts, the consideration of variable detection thresholds, and the incorporation of earthquake magnitude uncertainty.Our imput data, comprises 105 historical earthquakes which occurred between 984 and 1934, and a complete data file containing 1067 earthquakes which occurred during the period 1935–30 August, 1986. The complete part was divided into four subcatalogs according to different thresholds of completeness. Only subcrustal events were considered, and dependent events were removed.The obtained value (=0.65) is at the lower range of the previously reported results, but it appears concurrent with conceptual and observational facts. The same concerns inferred value of _max = 7.8 and activity rate _4.0 = 5.34.     

A macroseismic intensity prediction equation for intermediate depth earthquakes in the Vrancea region,Romania

The use of shake maps in terms of macroseismic intensity in earthquake early warning systems as well as intensity based seismic hazard assessments provides a valuable supplement to typical studies based on recorded ground motion parameters. A requirement for such applications is ground motion prediction equations (GMPE) in terms of macroseismic intensity, which have the advantages of good data availability and the direct relation of intensity to earthquake damage. In the current study, we derive intensity prediction equations for the Vrancea region in Romania, which is characterized by the frequent occurrence of large intermediate depth earthquakes giving rise to a peculiar anisotropic ground shaking distribution. The GMPE have a physical basis and take the anisotropic intensity distribution into account through an empirical regional correction function. Furthermore, the relations are easy to implement for the user. Relations are derived in terms of epicentral, rupture and Joyner–Boore distance and the obtained relations all provide a new intensity estimate with an uncertainty of ca. 0.6 intensity units.     

Prediction of strong earthquakes in Vrancea,Romania, using the CN algorithm

The application of the CN algorithm to a new earthquake catalogue, for the period from 1932 to 1993, obtained by merging Romanian and U.S.S.R. data, allows us to monitor, on the intermediate time scale. the preparation of strong, intermediate-depth earthquakes in the Vrancea region. Four of the five strong earthquakes with magnitudes above 6.4 are predicted. The total duration of the Time of Increased Probability (TIP) of the occurrence of an earthquake (TIP) occupies 21.7% of the time interval under consideration, i.e., about 2.5 years for each strong earthquake.     

Prediction of the strong earthquakes in Vrancea,Romania, using the CN algorithm

The application of the CN algorithm to a new earthquake catalogue, for the period from 1932 to 1993, obtained by merging Romanian and U.S.S.R. data, allows us to monitor, on the intermediate time scale, the preparation of strong, intermediate-depth earthquakes in the Vrancea region. Four of the five strong earthquakes with a magnitude above 6.4 are predicted, the total duration of the Time of Increased Probability of the occurrence of an earthquake (TIP) occupies 21.7% of the time interval under consideration, i.e., about 2.5 years for each strong earthquake.     

Waveform inversion of weak vrancea (Romania) earthquakes

Summary The inversion of high-frequency seismograms is performed to retrieve source mechanisms, hypocentral depths and source time functions of two weak Vrancea earthquakes (M_L=3.3), one that occurred in the crust, the other in the lithospheric part of the mantle. The digital waveforms recorded by the local Romanian network are used. Synthetic seismograms are computed by the modal summation method, using the point source approximation, for horizontally layered anelastic media. For each source-to-station path a different structural model is adopted which represents the best 1-D approximation of the medium in this azimuth. Thus, lateral inhomogeneities are taken into account in a simplified way. The source is described by the full moment tensor, allowing both deviatoric and volumetric components to be resolved.Although the structural models are simplified for the range of epicentral distances (15<<<170 km) considered, we find that the fit between the synthetic and observed seismograms is satisfactory for frequencies less than a few Hz. The few P-wave polarities available are not sufficient to determine a reliable source mechanism by standard methods, while the waveform inversion allows us to retrieve source mechanisms that are stable with respect to different boundary conditions and in good agreement with the observed polarities. The source time function is the less stable inverted parameter, being the most influenced by the simplification of the structural models.     

Multifractal and Chaotic Analysis of Vrancea (Romania) Intermediate-depth Earthquakes: Investigation of the Temporal Distribution of Events

The Vrancea seismic region contains an isolated cluster of events beneath the Carpathian Arc Bend in Romania, dipping to about 200 km depth. Seismic activity mainly occurs at intermediate depths (h > 60 km). The main goal of the paper is to perform an in-depth, complex analysis of the occurrence times of these intermediate-depth events. We also try to show the versatility of the methods used to characterize different aspects of the seismicity evolution and to offer a user-friendly software toolbox to do most of the related computations. The earthquake catalog used in this study spans from 1974 to 2002 and includes only the intermediate-depth events. In the first part of the paper, we analyze the multifractal characteristics of the temporal distribution of earthquakes. The study reveals two distinct scaling regimes. At small scales we found a clear nonhomogeneous, multifractal pattern, while at large scales the temporal distribution of events shows a monofractal, and close to Poissonian (random), behavior. The multifractal behavior at small scales (minutes-hours) is shown to be clearly an effect of the short aftershock sequences that occurred after some major Vrancea earthquakes. In the second part of the paper we analyze whether our temporal series shows a persistent (or anti-persistent) long-term behavior, by using the Detrended Fluctuation Analysis (DFA) method. The results suggest that the analyzed temporal series of Vrancea earthquakes is a non-correlated process. In part three of the paper we seek to determine whether the dynamics of our earthquake system (described by the occurrence time of Vrancea earthquakes) is deterministically chaotic, deriving from a rather simple evolution law, or whether it is stochastic and is generated by a system that possesses many degrees of freedom. The results suggest that our signal is stochastic (probably does not possess an attractor). The limited time-span of the catalog and the analysis performed in this paper cannot rule out the emergence of an interesting, quasi-deterministic and low-dimensional structure in the case of major Vrancea earthquakes.Acknowledgement One of the authors (BE) is grateful to the Japanese Ministry of Education for providing him a Monbusho scholarship for studying in DPRI, Kyoto University. We thank Z.R. Struzik, M Holschneider, J. Mori and D. Kaplan for their useful comments, and acknowledge the support of the staff of DPRI, Kyoto University and the National Inst. for Earth Physics, Bucharest. We thank the two reviewers, M.B. Geilikman and M. Anghel, for their useful suggestions which improved the quality of this work.     

Test of the Empirical Green's Function Deconvolution on Vrancea (Romania) Subcrustal Earthquakes

Vrancea is one of the few singular seismic regions of the world where intermediate-depth earthquakes are permanently generated (around 10 events/month with M _L > 3) within an extremely confined focal volume. This particularity and the relatively large number of short-period waveforms recorded by the Romanian local network provides us the opportunity to test the performance of the empirical Green's function technique in retrieving the source time function and source directivity of the Vrancea earthquakes. Three earthquakes that occurred on March 11, 1983 (M _L = 5.4), April 12, 1983 (M _L = 5.1) and August 7, 1984 (M _L = 5.1) in the lower part of the subducting lithosphere (h 150 km) were analyzed. A set of 28 adjacent events (3.0 < M _L < 4.4) which occurred between 1981 and 1997 were selected as corresponding empirical Green's functions. To test the confidence of the retrieved source time function, we compare the deconvolved pulses using Green's functions of different sizes and recorded simultaneously by short-period and broad-band instruments. Our tests show that the durations of the source time function is well-constrained and is not affected by the limited frequency range of the short-period instruments, or by the relative difference in the focal mechanism between the main event and Green's event. The apparent duration of the source time function outlines source directivity effects, and when these effects are sufficiently strong, they can identify the real fault plane. Relatively short source duration and correspondingly high stress drop values are in agreement with other previous results emphasizing a specific seismic regime in the lower part of the Vrancea subducting lithosphere.     

Preliminary empirical scaling of pseudo relative velocity spectra in Serbia from the Vrancea earthquakes

First frequency-dependent empirical scaling equations of pseudo-relative velocity spectral amplitudes (PSV) of strong earthquake ground motions in the former Yugoslavia were introduced in the mid-1990s by Lee and Trifunac (1990) [15]. This followed the development of the Fourier spectral amplitudes (FS) scaling equations by Lee and Trifunac (1993) [17] in terms of earthquake source parameters, and the region-specific frequency dependent attenuation function given by Lee and Trifunac (1992) [16]. More recently, a new frequency-dependent attenuation function was developed for central and eastern Serbia for earthquakes of intermediate and large magnitudes and for large epicentral distances—exceeding 300 km—suggested by Lee et al. (2016) [19] that occur in the Vrancea source region in Romania. In this paper we use this frequency-dependent attenuation function to develop empirical scaling equations for PSV spectral amplitudes in Serbia. These scaling equations will form a basis for macro- and micro-zoning earthquake hazard studies in Serbia.     

A preliminary empirical model for frequency-dependent attenuation of Fourier amplitude spectra In Serbia from the Vrancea earthquakes

We present the frequency-dependent attenuation model for Fourier amplitude spectra of strong earthquake ground motion in Serbia from intermediate depth earthquakes in the Vrancea source zone in Romania. The development of this type of scaling is the essential first step toward developing the corresponding attenuation and scaling equations for pseudo relative velocity spectra (PSV), which are necessary for seismic macro- and microzoning in the territory of Serbia. Such scaling is necessary because the Vrancea source zone produces large earthquakes with shaking that attenuates differently from the local earthquakes in Serbia. Development of such a scaling model is associated with several difficulties, the principal one being the lack of recorded strong motion accelerograms at epicentral distances exceeding 300 km. To reduce uncertainties with such scaling, we require our preliminary scaling equations to be consistent with independent estimates of seismic moment, stress drop, and radiated wave energy. In the future, when the recorded strong motion data from Vrancea earthquakes increases several-fold of what it is today, it will become possible to perform this analysis again, thus leading to more reliable and permanent scaling estimates.     

Attenuation characteristics of ground motion intensity from earthquakes with intermediate depth

Data of the intermediate depth (the Geioyo and the Shizuoka) earthquakes in Japan recorded in a dense network is analysed and compared with various available attenuation relations. The approach of Midorikawa (Midorikawa S., 1993, Technophysics 218, 287–295) based on the empirical Greens function technique of Irikura (Irikura, K., 1986, Proceedings of the 7th Japan Earthquake Engineering Symposium, pp. 151–156.) has been used to model the rupture responsible for these earthquakes and peak ground acceleration are simulated at selected observation points. The method presented in this work includes the transmission effect in a multiple layer crustal model for a finite fault earthquake source model. Sharp attenuation rate is seen for such intermediate depth earthquakes which is difficult to explain through conventional attenuation relations. Detail study of the methodology and comparison of results shows that the transmission factor plays an important role for the sharp attenuation rate for intermediate-to deep-focus earthquakes.     

Critical evaluation of strong motion in Kocaeli and Düzce (Turkey) earthquakes

Turkey was struck by two major events on August 17th and November 12th, 1999. Named Kocaeli (M_w=7.4) and Düzce (M_w=7.2) earthquakes, respectively, the two earthquakes provided the most extensive strong ground motion data set ever recorded in Turkey. The strong motion stations operated by the General Directorate of Disaster Affairs, the Kandilli Observatory and Earthquake Research Institute of Bogazici University and Istanbul Technical University have produced at least 27 strong motion records for the Kocaeli earthquake within 200 km of the fault. Kocaeli earthquake has generated six motions within 20 km of the fault adding significantly to the near-field database of ground motions for M_w>=7.0 strike–slip earthquakes. The paper discusses available strong motion data, studies their attenuation characteristics, analyses time domain, as well as spectral properties such as spectral accelerations with special emphasis on fault normal and fault parallel components and the elastic attenuation parameter, kappa. A simulation of the Kocaeli earthquake using code FINSIM is also presented.     

Ground motion prediction equations based on shallow crustal earthquakes in Georgia and the surrounding Caucasus

Strong ground motions caused by earthquakes with magnitudes ranging from 3.5 to 6.9 and hypocentral distances of up to 300 km were recorded by local broadband stations and three-component accelerograms within Georgia’s enhanced digital seismic network. Such data mixing is particularly effective in areas where strong ground motion data are lacking. The data were used to produce models based on ground-motion prediction equations (GMPEs), one benefit of which is that they take into consideration information from waveforms across a wide range of frequencies. In this study, models were developed to predict ground motions for peak ground acceleration and 5%-damped pseudo-absolute-acceleration spectra for periods between 0.01 and 10 s. Short-period ground motions decayed faster than long-period motions, though decay was still in the order of approximately 1/r. Faulting mechanisms and local soil conditions greatly influence GMPEs. The spectral acceleration (SA) of thrust faults was higher than that for either strike-slip or normal faults but the influence of strike-slip faulting on SA was slightly greater than that for normal faults. Soft soils also caused significantly more amplification than rocky sites.     

Ground-motion prediction equations (GMPEs) for inelastic response and structural behaviour factors

The objective of this paper is to present ground-motion prediction equations describing constant-ductility inelastic spectral ordinates and structural behaviour factors. These equations are intended for application within the framework of Eurocode 8. Most of the strong-motion data used in the present work is obtained from the ISESD (Internet Site for European Strong-motion Data) databank. Present analysis includes ground motion records from significant Icelandic earthquakes, which are augmented by records obtained from continental Europe and the Middle East. In all cases the selected ground motion records are generated during shallow earthquakes within a distance of 100 km from the recording station. The classification of site conditions in the present work is based on the Eurocode 8 definition.     

Accounting for site effect in probabilistic assessment of seismic hazard for Romania and Bucharest: a case of deep seismicity in Vrancea zone

The paper presents recent achievements in evaluations of site-dependent seismic hazard in Romania and the capital city of Bucharest caused by the Vrancea focal zone (SE-Carpathians). The zone is characterized by a high rate of occurrence of large earthquakes in a narrow focal volume at depths 60–170 km. The database that was used for the hazard evaluation includes parameters of seismicity, ground-motion source scaling and attenuation models (Fourier amplitude spectra), and site-dependent spectral amplification functions. Ground-motion characteristics were evaluated on the basis of several hundred records from more than 120 small magnitude (M 3.5–5) earthquakes occurred in 1996–2001 and a few tens of acceleration records obtained during four large (M 7.4, 7.2, 6.9 and 6.3) earthquakes. The data provide a basis for probabilistic seismic hazard assessment in terms of peak ground acceleration, peak spectral acceleration and MSK intensity using Fourier amplitude spectra for various exceedance probabilities or average return periods. It has been shown that the influence of geological factors plays very important role in distribution of earthquake ground-motion parameters along the territory of Romania.     

Geometrical constraints for the configuration of the Vrancea (Romania) intermediate-depth seismicity nest

For a seismogenic area like Vrancea (Romania) with well-defined geometrical features of the seismicity production in space and time, the numerical simulation of the earthquake generation process (e.g. cellular automaton) looks highly attractive. The delimitation, as accurately as possible, of the geometrical features of the seismically active system in the Vrancea subcrustal zone is essential to constrain the simulation modeling. As a first approximation, the seismicity pattern is close to a fault plane NE–SW oriented, extended roughly vertically between 60 and 170 km depth. A characteristic median plane is defined by minimizing the distance of hypocenters. The average distance of the hypocenters to the median plane is around 5 km. However, a more detailed investigation of the geometrical configuration of seismicity indicates a fragmentation of the active body located in the upper mantle in two segments. The seismicity pattern is well approximated by a planar distribution in each segment. In the transition zone, between the upper and lower segment, the hypocenter distribution is more dispersed and shows a disruption among the two planar segments, measured by about 9 km apart laterally one relative to the other. The two segments hosted the major Vrancea events recorded in the last two centuries (for which we have available location of acceptable accuracy). The narrow transition zone at about 100 km depth is interpreted as a weaker segment, possibly caused by a dehydration process or by an infiltration of asthenosphere material from the back side of the South-Eastern Carpathian arc system. It is still debatable if fragmentation in two seismically active segments reflects the existence of two neighbouring separate blocks (upper, continental and lower, oceanic block) or a consequence of a breaking process separating a continental block into two parts. The segmentation of the descending lithosphere and the edge effects are apparently stationary, at least for the time interval since 1985 to the present, for which the earthquake catalogue is reliable (homogeneous).     

地震动强度参数估计的可解释性与不确定度机器学习模型

准确预测地震动强度参数(峰值加速度PGA、峰值速度PGV等)对于震后应急和地震危险性概率分析至关重要.作为地震动强度参数预测的新手段, 机器学习算法具有优势, 但也存在可解释性差和难给出预测结果不确定度的问题.本文提出采用自然梯度提升(NGBoost)算法在预测结果的同时提供其不确定度, 并结合SHAP值解释机器学习模型.基于NGA-WEST2强震动数据库, 本文训练出了适合预测活跃构造区地壳地震的PGA和PGV概率密度分布的机器学习模型.测试集数据PGA和PGV的预测值与真实值的相关系数可达0.972和0.984, 并可给出预测结果的合理概率密度分布.通过SHAP值, 我们从数据角度弄清了各输入特征(矩震级M_W、Joyner-Boore断层距R_jb、地下30 m平均S波速度V_S30、滑动角Rake、断层倾角Dip、断层顶部深度Z_TOR和V_S达到2.5 km·s^-1时的深度Z_2.5)对机器学习模型预测结果的影响机理.SHAP值显示, 基于NGBoost算法的机器学习模型的预测方式基本与物理原理相符, 说明了机器学习模型的合理性.SHAP值还揭示出一些以往研究忽视的现象: (1)对于活跃构造区地壳地震, 破裂深度较浅(Z_TOR＜~5 km)时, Z_TOR的SHAP值低于破裂深度较深(Z_TOR＞~5 km)时的值, 表明浅部破裂可能主要受速度强化控制, 地震动强度较弱.并且Z_TOR的SHAP值随Z_TOR值增大而减小, 表明地震动强度可能还受破裂深度变化引起的几何衰减变化影响; (2)破裂深度较深时, Z_TOR的SHAP值随Z_TOR值增大而增大, 表明深部破裂的地震动强度可能受和破裂深度变化相关的应力降或品质因子Q的变化影响; (3)Z_2.5较小(Z_2.5＜~1 km)时, Z_2.5的SHAP值的变化规律对于PGA和PGV预测是相反的, 表明加速度和速度频率不同, 受浅层沉积物厚度变化引起的共振频率变化影响不同.

     

Observations of vertical ground accelerations exceeding gravity during the 1997 Umbria-Marche (central Italy) earthquakes

We found extensive evidence that the vertical ground accelerations produced during the largest shock (M = 6.0) of the 1997 Umbria-Marche earthquake sequence exceeded 1g in two areas close to the heavily-damaged villages of Annifo and Colle Croce. This evidence comes from the striking observation of thousands of freshly fractured and broken rocks and stones in these areas. Some of the broken stones lie isolated on soft detritic soil while others had been previously piled up, probably a long time agoto clear the fields for farming. The freshness of the cuts and fractures and the consistency of the observations for thousands of rocks and stones in these areas indicate that these rocks were thrown upwards during the earthquake, with breakage occurring at the time of impact. Ground motion calculations consistent with the static deformation inferred from GPS and interferometry data, show that the broken stones and rocks are found in the zone where the strongest shaking took place during the earthquake and that most of the shaking there was vertical.