Papers by E. Papadimitriou
Change point analysis is performed on the seismicity in Gulf of Corinth (Greece), an extensional ... more Change point analysis is performed on the seismicity in Gulf of Corinth (Greece), an extensional graben which constitutes one of the most seismically active areas in Greece. Seismicity appears intense and strongly clustered and therefore analysis on mean and variance is appropriate. Sample autocorrelation function of the data is non-zero even for bigger lags, indicating long-range correlations. This phenomenon can be justified by possible changes in the mean of the observations. Non-parametric multiple change point analysis is applied to both the sequence of the earthquakes from a set of observations and its detrended data considering the earthquake occurrence frequency. The results of the analysis on the initial data set are compared to those of its detrended residuals. This procedure employs both online and offline methods providing different perspectives. Promising patterns are defined offline and most of them are detectable online.
Earthquake clustering in the area of Central Ionian Islands (Greece) is statistically modelled by... more Earthquake clustering in the area of Central Ionian Islands (Greece) is statistically modelled by means of the Epidemic Type Aftershock Sequence (ETAS) branching model, which is the most popular among the short-term earthquake clustering models. It is based upon the assumption that an earthquake is not fully related to any other one in particular, but rather to both all previous events, and the background seismicity. The close temporal proximity of the strong (M ≥ 6.0) events in the study area offers the opportunity to retrospectively test the validity of the ETAS model through the 2014 Kefalonia doublet (M w 6.1 and M w 6.0) and the 2015 Lefkada aftershock sequences. The application of a physics-based earthquake simulator to the local fault system produced a simulated catalogue with time, space and magnitude behaviour in line with the observed seismicity. This catalogue is then used for the detection of short-term interactions between both strong and smaller events and the comparison between the two cases. The results show that the suggested clustering model provides reliable forecasts of the aftershock activity. Combining the ETAS model and the simulator code, though, needs to be more deeply examined since the preliminary results show some discrepancy between the estimated model parameters.
Clustering features of crustal seismicity for the period from 2008 up to 2018 in Greece in both s... more Clustering features of crustal seismicity for the period from 2008 up to 2018 in Greece in both space and time are investigated by means of the epidemic type aftershock sequence model. With a completeness threshold of m c = 3.5, the data set includes 8258 events, with 19 of them of M w ≥ 6.0, encouraging the detailed analysis of the short-term clustering spatiotemporal characteristics. The estimation of the model parameters is performed via the maximum likelihood estimation technique using a simulated annealing approach, after considering a normal grid superimposed on the study area with 0.1°× 0.1°cells. The model is tested with a residual analysis procedure along with the application of both Numbers and Log Likelihood tests. The initial estimated model with the full dataset underestimates the observed seismicity, and appears unstable, implying an explosive type model (a model with infinite decay rate). The underestimation is observed during 2008, when five M w ≥ 6.0 earthquakes occurred, which is an extreme rate for Greek seismicity. Excluding this part of the catalog, during the period 2009-2018, the estimated model exhibits good agreement with the observed seismicity. The classification between background and triggered events is made by measuring the contribution of each class to the total seismicity, which presents a considerable clustering behavior.
A quite energetic seismic excitation consisting of one main and three additional distinctive eart... more A quite energetic seismic excitation consisting of one main and three additional distinctive earthquake clusters that occurred in the transition area between the Kefalonia Transform Fault Zone (KTFZ) and the continental collision between the Adriatic and Aegean microplates is thoroughly studied after the high-precision aftershocks' relocation. The activated fault segments are in an area where historical and instrumental data have never claimed the occurrence of a catastrophic (M ≥ 6.0) earthquake. The relocated seismicity initially defines an activated structure extending from the northern segment of the Lefkada branch of KTFZ with the same NNE-SSW orientation and dextral strike slip faulting, and then keeping the same sense of motion, its strike becomes NE-SW and its dip direction NW. This provides unprecedented information on the link between the KTFZ and the collision front and sheds more light on the regional geodynamics. The earthquake catalog, which was especially compiled for this study, starts one year before the occurrence of the M w 5.4 main shock, and adequately provides the proper data source for investigating the temporal variation in the b value, which might be used for discriminating foreshock and aftershock behavior.
The Florina basin (NW Greece) embodies sources of natural CO 2 emissions, which are aligned paral... more The Florina basin (NW Greece) embodies sources of natural CO 2 emissions, which are aligned parallel to the NE-SW striking regional major normal faults. The study area is characterized by low to moderate seismicity, as historical information and instrumental data reveal. During February 2013 to January 2014 an outburst of seismic activity emerged, consisting of more than 2000 located earthquakes, with a mainshock of M w = 4.1 on 17 February 2013 (16:48 UTC). A local temporary network was installed in July 2013, whose recordings provided the necessary data for performing a Shear-wave Splitting (SwS) study of the upper crust. After the application of strict selection criteria, 532 SwS measurements were obtained for two stations, using a mixture of manual and fully automatic methods. Two polarization directions of the fast shear-wave were identified at one of the stations, with the main one striking NE-SW and the secondary in a NW-SE direction. The first direction is in agreement with the local stress field, as it is identified by focal mechanisms, and with the strike of faults along which soil gas CO 2 concentrations are distributed and can be explained by the Anisotropic Poro-Elasticity (APE) model. The second direction is consistent with the strike of mapped local faults, inherited since the Late Miocene extension, which are responsible for the formation of the Florina basin. The SwS results at the second station revealed an almost N-S dominant polarization direction, which is not consistent with the local stress field. This direction, obtained from events belonging to the southern group of the sequence, that was probably triggered by fluid intrusion, might be interpreted by spatiotemporal patterns of seismicity and CO 2 upwelling.
An M w 6.8 earthquake occurred on October 25, 2018, 35 km offshore from the southwest coastlines ... more An M w 6.8 earthquake occurred on October 25, 2018, 35 km offshore from the southwest coastlines of Zakynthos Island. The aftershock sequence appeared remarkably productive with six aftershocks of M ≥ 5.0 in the first month and tens of aftershocks with M ≥ 4.0 during the study period. The GCMT solution for the main shock suggests a very low angle plane (dip = 24°) for a dextral strike-slip faulting (rake = 165°). A similar solution is suggested for the largest aftershock (M w 5.9) that occurred 5 days afterward. The proximity of the main shock location with the dextral active boundary of Kefalonia Transform Fault Zone (KTFZ) along with the Hellenic Subduction front supports this oblique faulting. The aftershock activity is comprised mostly in depths 5-12 km and forms eight distinctive clusters that accommodate regional strain and evidence strain partitioning. The role of stress transfer and statistical analysis are combined for detailing the highly productive aftershock sequence. Earthquake networks analysis reveals their random structure soon after the main shock, which became small-world structure after the first 200 days. Time series analysis constructed from the aftershock frequency and seismic moment release and manifested significant correlation among the eight seismicity clusters.
The complexity of seismogenesis requires the development of stochastic models, the application of... more The complexity of seismogenesis requires the development of stochastic models, the application of which aims to improve our understanding on the seismic process and the associated underlying mechanisms. Seismogenesis in the Corinth Gulf (Greece) is modeled through a Constrained-Memory Stress Release Model (CM-SRM), which combines the gradual increase of the strain energy due to continuous slow tectonic strain loading and its sudden release during an earthquake occurrence. The data are treated as a point process, which is uniquely defined by the associated conditional intensity function. In the original form of the Simple Stress Release Model (SSRM), the conditional intensity function depends on the entire history of the process. In an attempt to identify the most appropriate parameterization that better fits the data and describes the earthquake generation process, we introduce a constrained " m-memory" point process, implying that only the m most recent arrival times are taken into account in the conditional intensity function, for some suitable m ∈ N. Modeling of this process is performed for moderate earthquakes (M ≥ 5.2) occurring in the Corinth Gulf since 1911, by considering in each investigation different number of steps backward in time. The derived model versions are compared with the SSRM in its original form and evaluated in terms of information criteria and residual analysis.
The recurrence time, T r , of strong earthquakes above a predefined magnitude threshold on specif... more The recurrence time, T r , of strong earthquakes above a predefined magnitude threshold on specific faults or fault segments is an important parameter, that could be used as an input in the development of long-term fault-based Earthquake Rupture Forecasts (ERF). The amount of observational recurrence time data per segment is often limited, due to the long duration of the stress rebuilt and the shortage of earthquake catalogs. As a consequence, the application of robust statistical models is difficult to implement with a precise conclusion, concerning T r and its variability. Physics-based earthquake simulators are a powerful tool to overcome these limitations, and could provide much longer earthquake records than the historical and instrumental earthquake catalogs. A physics-based simulator, which embodies known physical processes, is applied in the Southern Thessaly Fault Zone (Greece), aiming to provide insights about the recurrence behavior of earthquakes with M w ≥ 6.0 in the six major fault segments in the study area. The build of the input fault model is made by compiling the geometrical and kinematic parameters of the fault network from the available seismotectonic studies. The simulation is implemented through the application of the algorithm multiple times, with a series of different input free parameters, in order to conclude in the simulated catalog which showed the best performance in respect to the observational data. The detailed examination of the 254 M w ≥ 6.0 earthquakes reported in the simulated catalog reveals that both single and multiple segmented ruptures can be realized in the study area. Results of statistical analysis of the interevent times of the M w ≥ 6.0 earthquakes per segment evidence quasiperiodic recurrence behavior and better performance of the Brownian Passage Time (BPT) renewal model in comparison to the Poissonian behavior.
The complex turbulent dynamics of seismogenesis in the area of Greece is investigated, by applyin... more The complex turbulent dynamics of seismogenesis in the area of Greece is investigated, by applying an alternative approach based on renewal theory, specifically the Renewal Aging Algorithm, to two earthquake interevent (waiting) time series. The data sets were extracted from the regional earthquake catalogue compiled from the Geophysics Department, Aristotle University of Thessaloniki and concern: 576 earthquakes of M ≥5.5, during 1911-2017 and 113 earthquakes of M ≥6.5, in the period of 1845-2017. The methodology is efficient for studying persistency and/or intermittent structures in different time scales of Hellenic seismogenesis by qualitative estimating the amount of memory, corresponding to the ratio between Poisson events and non-Poisson critical events. For the application of the Renewal Aging Algorithm time scales corresponding to short and intermediate-term forecasting were considered, namely 2 and 3 weeks to several years corresponding to the maximum Waiting Time for each data set. The results of the statistical analysis reveal transitions from time-homogeneous Poisson to time-homogeneous non-Poisson dynamics and non-homogeneous non-Poisson dynamics, starting from short time scales and going to longer time scales. These results can shed more light to the concept of the seismic cycle hypothesis and to the generation of time-dependent stochastic modeling. It has to be mentioned that these time series are relatively short and therefore additional statistical analysis is required to verify the aforementioned findings.
Prevailing patterns of seismicity dynamics, like the evolution of main shock-aftershock sequences... more Prevailing patterns of seismicity dynamics, like the evolution of main shock-aftershock sequences and swarms, along with periods of seismic quiescence, are explored through the temporal analysis of the earthquake clustering in the area of Corinth Gulf, Greece. The clusters are unveiled by the implementation of a new algorithm, whose robustness is verified on simulated catalogs. The method is based on the application of a bivariate stochastic point process, the Markovian arrival process (MAP), (N t , J t) t∈R + , whose intensity function, λ t , is driven by the underlying Markov process, J t. In particular, each hidden state corresponds to a distinct occurrence rate of the counting process, N t , that enables the modeling of changes in the earthquake dynamics. With the proposed algorithm, known as local decoding algorithm, the hidden states, i.e. seismicity rates, are revealed at each occurrence time. The study area is divided in the eastern and western subareas based on seismotectonic criteria, for enabling any given event to interact with the following ones. The performance of the model is evaluated on subcatalogs of short time intervals that include small earthquakes, as well as to catalogs of longer duration associated with well studied destructive events and swarms. A complete cluster analysis for the 1964-2017 seismicity catalog with magnitudes M ≥ 4.5 is given, in a period that covers a sufficient number of well studied seismic sequences for both subareas. The identified earthquake clusters are consistent with those obtained from previous analyses of selected seismic sequences and the hidden states are associated mainly with main shock-aftershock and swarmlike sequences. The seismicity dynamics in the two subareas differ significantly, since in the eastern part earthquake frequency is remarkably low leading to more episodic transitions from periods of high seismic activity to ones of seismic quiescence.
The complexity of seismogenesis requires the development of stochastic models, the application of... more The complexity of seismogenesis requires the development of stochastic models, the application of which aims to improve our understanding on the seismic process and the associated underlying mechanisms. Seismogenesis in the Corinth Gulf (Greece) is modeled through a Constrained-Memory Stress Release Model (CM-SRM), which combines the gradual increase of the strain energy due to continuous slow tectonic strain loading and its sudden release during an earthquake occurrence. The data are treated as a point process, which is uniquely defined by the associated conditional intensity function. In the original form of the Simple Stress Release Model (SSRM), the conditional intensity function depends on the entire history of the process. In an attempt to identify the most appropriate parameterization that better fits the data and describes the earthquake generation process, we introduce a constrained " m-memory" point process, implying that only the m most recent arrival times are taken into account in the conditional intensity function, for some suitable m ∈ N. Modeling of this process is performed for moderate earthquakes (M ≥ 5.2) occurring in the Corinth Gulf since 1911, by considering in each investigation different number of steps backward in time. The derived model versions are compared with the SSRM in its original form and evaluated in terms of information criteria and residual analysis.
The spatiotemporal evolution of seismicity of an accurately relocated earthquake catalog in the c... more The spatiotemporal evolution of seismicity of an accurately relocated earthquake catalog in the central Ionian Islands, Greece, is explored, by the systematic identification of the earthquake clusters over the period September 2016-December 2019. Manual waveform phase picking was the input for the relocation leading to 13,632 earthquakes that compose a data set with a completeness magnitude as low as M c = 1.5. The clustering procedure engages a temporal stochastic point process, the Markovian Arrival Process (MAP), for an initial separation of the background seismicity from potential seismic excitations, using the changes in the seismicity rate and a densitybased clustering algorithm, DBSCAN, for the detection of elevated spatial density areas. A high concentration of temporally persistent clusters is identified along the western coastline of Lefkada Island, in the parallel step-over faults between the two major fault branches of the Kefalonia Transform Fault Zone and in the Myrtos Gulf area in Kefalonia, which are spatially correlated to the positive static stress changes induced by the coseismic slip of the M w 6.5 November 17, 2015 Lefkada main shock. The method presented remarkable potential in the identification of clusters in 3-D space along with the seismicity migration during the temporal evolution of seismic sequences. Multiple secondary faults of the Kefalonia segment can be revealed and the dominant triggering mechanisms can be illuminated contributing towards the understanding of the regional faulting properties and mechanics, an outcome that cannot be achieved solely by studying the large earthquakes, due to their rarity. The temporal and spatial clustering properties of microseismicity shed light on the preparation phase of the large earthquakes as being part of this process, thus contributing to the seismic hazard assessment.
The recurrence time, T r , of strong earthquakes above a predefined magnitude threshold on specif... more The recurrence time, T r , of strong earthquakes above a predefined magnitude threshold on specific faults or fault segments is an important parameter, that could be used as an input in the development of long-term fault-based Earthquake Rupture Forecasts (ERF). The amount of observational recurrence time data per segment is often limited, due to the long duration of the stress rebuilt and the shortage of earthquake catalogs. As a consequence, the application of robust statistical models is difficult to implement with a precise conclusion, concerning T r and its variability. Physics-based earthquake simulators are a powerful tool to overcome these limitations, and could provide much longer earthquake records than the historical and instrumental earthquake catalogs. A physics-based simulator, which embodies known physical processes, is applied in the Southern Thessaly Fault Zone (Greece), aiming to provide insights about the recurrence behavior of earthquakes with M w ≥ 6.0 in the six major fault segments in the study area. The build of the input fault model is made by compiling the geometrical and kinematic parameters of the fault network from the available seismotectonic studies. The simulation is implemented through the application of the algorithm multiple times, with a series of different input free parameters, in order to conclude in the simulated catalog which showed the best performance in respect to the observational data. The detailed examination of the 254 M w ≥ 6.0 earthquakes reported in the simulated catalog reveals that both single and multiple segmented ruptures can be realized in the study area. Results of statistical analysis of the interevent times of the M w ≥ 6.0 earthquakes per segment evidence quasiperiodic recurrence behavior and better performance of the Brownian Passage Time (BPT) renewal model in comparison to the Poissonian behavior.
Investigation of temporal b-values variations during the last decade is attempted in central Ioni... more Investigation of temporal b-values variations during the last decade is attempted in central Ionian Islands (Greece), the most seismically active area in the Mediterranean area. The high quality catalogue used, starting in 2008 with the completeness magnitude M c 2.8 comprises data from the aftershock sequences of the 2014 Cephalonia doublet (M w 6.1 and M w 6.0) and the 2015 Leukas M w 6.5 mainshock along with periods of relative quiescence. The analysis is performed by means of two alternative techniques, the "fixed-number-of-events" and the "constant-time-windows" approaches. The results indicate that b-value variations are inversely proportional to the seismicity rate with the decreases representing high seismic moment release. For a more detailed monitoring of the fluctuation of b-values in time and space, the competently elaborated 2015 Leukas sequence is employed. The results indicate that during the sequence, immediately after the occurrence of the mainshock and for about a month, a decrease in b-value is observed, which gradually increases towards the background value.
The recurrence time of large earthquakes above a predefined magnitude threshold on specific fault... more The recurrence time of large earthquakes above a predefined magnitude threshold on specific faults or fault segments is one of the key parameters for the development of long-term Earthquake Rupture Forecast models. Observational data of successive large earthquakes per fault segment are often limited and thus inadequate for the construction of robust statistical models. The physics-based earthquake simulators are a powerful tool to overcome recurrence data limitations by generating long earthquake records. A physics-based simulator, embodying well known physical processes, is applied in the North Aegean Trough (NAT) Fault Zone (Greece). The application of the simulation is implemented, after defining a five segment source model, aiming at the investigation of the recurrence behaviour of earthquakes with M w ≥ 6.5 and M w ≥ 7.0. The detailed examination of the 544 M w ≥ 6.5 earthquakes included in the simulated catalogue reveals that both single and multiple segmented ruptures can be realized along the NAT. Results of statistical analysis of the interevent times of M w ≥ 6.5 and M w ≥ 7.0 earthquakes per participating segment to the related ruptures indicate the better performance of the Brownian Passage Time renewal model in comparison to exponential model. These results provide evidence for quasi-periodic recurrence behaviour, agreeing with the elastic rebound theory, instead of Poissonian behaviour.
Short-term earthquake clustering properties in the Eastern Aegean Sea (Greece) area investigated ... more Short-term earthquake clustering properties in the Eastern Aegean Sea (Greece) area investigated through the application of an epidemic type stochastic model (Epidemic Type Earthquake Sequence; ETES). The computations are performed in an earthquake catalog covering the period 2008 to 2020 and including 2332 events with a completeness threshold of M c = 3.1 and separated into two subcatalogs. The first subcatalog is employed for the learning period, which is between 2008/01/01 and 2016/12/31 (N = 1197 earthquakes), and used for the model's parameters estimation. The second subcatalog from 2017/01/01 to 2020/11/10 (1135 earthquakes), in which the sequences of 2017 M w = 6.4 Lesvos, 2017 M w = 6.6 Kos and 2020 M w = 7.0 Samos main shocks are included, and used for a retrospective forecast testing based on the constructed model. The estimated model parameters imply a swarm like behavior, indicating the ability of earthquakes of small to moderate magnitude above M c to produce their own offsprings, along with the stronger earthquakes. The retrospective evaluation of the model is examined in the three aftershock sequences, where lack of foreshocks resulted in low predictability of the mainshocks, with estimated daily probabilities around 10-5. Immediately after the mainshocks occurrence the model adjusts with notable resemblance between the expected and observed aftershock rates, particularly for earthquakes with M ≥ 3.5.
Following the M7.0 earthquake that struck the Greek island of Samos and Turkey's western coast, c... more Following the M7.0 earthquake that struck the Greek island of Samos and Turkey's western coast, causing extensive damage and casualties, we combined existing knowledge geodatabases concerning historical seismicity and rupture zones with seismological and geodetic measurements as well as with modelling and in situ observations, to provide an assessment of rapid response to the seismic event. In this paper, we demonstrate that in the frame of the gradual provision of information from the individual scientific disciplines, taking into account their respective potential and limitations, a multidisciplinary approach is able to address more efficiently rapid response issues in order to allow effective preliminary interpretation of the earthquake activity, even within the first 24 h of the event. It focuses on the assessment of the timely provision of information by each discipline, evaluating the access to primary data sources as well as the maturity of the techniques in terms of accuracy and rapid data processing. Within a period of less than a week, several constraints were partially compensated for, allowing the delivery of more robust results and interpretation. The study highlights the readiness level of the various domains that has been significantly improved over the past years, including rapid seismological solutions, systematic availability of free and open Earth Observation data and on-demand online processing through dedicated platforms. Their combination with routinely applied inversion modelling and timely in situ observation is leading to improved operational response levels.
A physics-based earthquake simulation algorithm for modeling the long-term spatiotemporal process... more A physics-based earthquake simulation algorithm for modeling the long-term spatiotemporal process of strong (M ≥ 6.0) earthquakes in Corinth Gulf area, Greece, is employed and its performance is explored. The underlying physical model includes the rate-and state-dependent frictional formulation, along with the slow tectonic loading and coseismic static stress transfer. The study area constitutes a rapidly extending rift about 100 km long, where the deformation is taken up by eight major fault segments aligned along its southern coastline, and which is associated with several strong (M ≥ 6.0) earthquakes in the last three centuries, since when the historical earthquake catalog is complete. The recurrence time of these earthquakes and their spatial relation are studied, and the simulator results reveal spatiotemporal properties of the regional seismicity such as pseudoperiodicity as well as multisegment ruptures of strong earthquakes. As the simulator algorithm allows the display of the stress pattern on all the single elements of the fault, we are focusing on the time evolution of the stress level before, during, and after these earthquakes occur. In this respect, the spatiotemporal variation of the stress and its heterogeneity appear to be correlated with the process of preparation of strong earthquakes in a quantitative way. KEY POINTS • We simulated a long stress and seismicity history for the main fault segments of the Corinth Gulf. • Our attention was focused on the time evolution of the stress before, during, and after strong earthquakes. • The results provide inferences on the spatiotemporal properties of seismic activity in the study area.
The 30 October 2020, M w 7.0 Samos mainshock took place in the offshore north of Samos Island in ... more The 30 October 2020, M w 7.0 Samos mainshock took place in the offshore north of Samos Island in eastern Aegean area, previously struck in 1904 with a comparable magnitude earthquake offshore the southern coastline. The investigation of the aftershock seismicity evolution and the properties of the activated fault network was accomplished with aftershock relocation performed with the double-difference and cross-correlation techniques. The highly accurate relocated seismicity illustrates a well-defined E-W activated structure located deeper than 5 km with an average depth of ~ 12 km. Moment tensor solutions indicate mostly normal faulting with an average Taxis ~ 185 ο. Strong-motion waveform modeling revealed a N-dipping fault plane with a coseismic slip patch of 36 km × 22 km and a maximum slip equal to 1 m at 12 km depth. The slip is mainly concentrated in a single asperity implying a rupture mode of asperities breaking in isolated earthquakes rather than to cooperate to produce a larger rupture. Coulomb stress calculations unveil increased positive static stress changes values at the locations of the majority of the aftershocks and activation of minor fault segments by stress transfer.
On 3 March 2021, the Mw6.3 Tyrnavos earthquake shook much of the Thessalia region, leading to ext... more On 3 March 2021, the Mw6.3 Tyrnavos earthquake shook much of the Thessalia region, leading to extensive damage in many small towns and villages in the activated area. The first main shock was followed in the next day, on 4 th of March 2021, by an "equivalent" main shock with Mw6.
Uploads
Papers by E. Papadimitriou