NH4.3

In recent years, machine learning techniques have been widely applied in seismological data processing such as seismic event detection, phase picking, location, magnitude estimation, and further data analysis for determining source mechanisms. Especially in earthquake location, deep learning methods are used to reduce location errors compared to conventional algorithms.

In this study, we present a deep learning based epicentral distance estimation with two separate models using seismic data from two stations as input data. The first model is the P- and S-wave arrival time picking model and the second is the epicentral distance estimation model. Since the traditional epicentral distance estimation methods uses the difference in arrival times between P- and S-waves, the P- and S-wave arrival times were first predicted from three-component seismic data so that this information could be directly used as the next input data. This picking information is used as input data along with the station location in the epicentral distance estimation model to output the final epicentral distance. Since this method uses data from two stations, it has higher accuracy than epicentral distance estimation using data from a single station.

The P- and S-wave arrival time picking model was modified by referring to the ResUNet (Diakogiannis et al., 2020) structure to improve performance based on the seismic detection and phase picking model from the three-component acceleration data developed by Mousavi et al. (2020). This modified model performs feature extraction for P- and S-phase picking and includes a residual block and skip connection. The model for estimating the distance from the epicenter was constructed using a basic artificial neural network (ANN) architecture. As input data, a total of eight features were used by adding six combinations of the difference in arrival times of P-wave and S-wave in each component of the two stations and two values of the latitude and longitude difference between two stations. The ANN architecture consists of four hidden layers and the epicentral distances of the two stations are final output.

The STEAD data were used as training data and test data. The STEAD is a seismogram dataset recorded from about 450,000 global earthquakes, and among them, data with magnitudes greater than 2.5 and epicentral distances less than 400 km were selected and used. As a result of applying the trained model to the test data, the mean absolute error of the predicted epicentral distance was 6.5 km, which showed improved performance compared to the previous results. Also, since this method uses six time-differences as input data, it can provide more robust results even in the presence of random noise at the picked times.

How to cite: Choi, Y., Bae, S., Song, Y., Seol, S. J., and Byun, J.: P- and S-wave arrival picking and epicentral distance estimation of earthquakes using convolutional neural networks, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11098, https://doi.org/10.5194/egusphere-egu22-11098, 2022.

As the number of seismic stations and experiments greatly increases due to ever greater availability of instrumentation, automated data processing becomes more and more necessary and important. Machine Learning (ML)methods are becoming widespread in seismology, with programsthat identify signals and patterns or extract features that can eventually improve our understanding of ongoing physical processes. We here focus on comparing and testing a selection of currently available methods for machine-learning-based seismic event detection and arrival time picking, performing a comparative study of the two autopickers EQTransformer and GPD with seismic data from the IPOC deployment in Northern Chile within the open-source Seisbench framework.

As a small benchmark dataset, we chose a random day for which we handpicked all visually discernible events on the 16 IPOC stations, which led to 200 events from 450 extracted, comprising 1493 P and 1163 S-phases. These events cover a large range of hypocentral depths (surface to >200 km) as well as magnitudes (<1.5 to 4.5).

We present first results from the application of the two autopickers EQTransformer and GPD, which have been shown to be most suitable for our type of dataset in a recent study by Münchmeyer et al. (2021), to IPOC data. We use our small benchmark dataset to evaluate detection rate (missed events, false detections) as well as picking accuracy (residuals to handpicks), and also investigate the effect of using different training datasets.

The present study is the first step towards the design of an automated workflow that comprises event detection and phase picking, phase association and event location and will be used to evaluate subduction zone microseismicity in different locations.

How to cite: Najafipour, N. and Sippl, C.: Comparing machine-learning based picking algorithms in a subduction setting, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4799, https://doi.org/10.5194/egusphere-egu22-4799, 2022.

Seismic phase association plays an important role in earthquake detection and location workflows as it links together seismic phases detected on different seismometers into individual earthquakes. Together with improved phase picking algorithms, a phase association algorithm can generate large earthquake phase data sets and earthquake catalogs when applied to dense permanent or temporary seismic networks. Recently, many efforts have been made on improving seismic phase association performance, such as developing machine learning approaches that are trained on millions of synthetic sequences of P and S arrival times, to generate more precise and complete catalogs including more small earthquakes.

As part of project MILESTONE, which aims at the automatic creation of large microseismicity catalogs in subduction settings, the present study evaluates the performance of the deep-learning based phase association algorithm PhaseLink (Ross et al. 2019) by comparison with a traditional grid-based method and a small handpicked benchmark dataset.

We used seismic data from the IPOC (Integrated Plate boundary Observatory Chile) permanent deployment of broadband stations in Northern Chile, dedicated to the study of earthquakes and deformation at the continental margin of Chile.

For an initial calibration, we manually picked P and S phases of raw waveforms on 15 stations on two randomly chosen days. All events that were visually recognizable were picked and located, which led to a dataset of 251 events comprising 1823 P and 1468 S picks, spanning a depth range from the surface down to 240 km. We use this handpicked dataset as ‘ground truth’, and evaluate the performance of PhaseLink and the grid-based method coupled with a STA/LTA trigger against this benchmark, considering both the numbers of (correctly/falsely) associated events and the number of constituent picks per event.

In a second experiment, we compare PhaseLink and conventional phase associator using a much larger set of STA/LTA alerts from the same region, but without the additional ground truth.

The presented research represents first steps towards an integrated automated workflow for detecting, picking, associating and locating microseismicity in subduction zone settings.

How to cite: Puente Huerta, J. A. and Sippl, C.: Exploring the performance of phase association algorithms, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4843, https://doi.org/10.5194/egusphere-egu22-4843, 2022.

In recent years, it became clear that the seismological community is adopting deep learning (DL) models for many diverse tasks such as problems of discrimination and classification of seismic events, earthquake detection and phase picking, generalised phase detection, earthquake early warning etc. Many models that have been developed and tested reach quite high accuracy values. However, it has been showed that their performances depend on the DL architecture, on the training hyperparameters and on the datasets that are used for training. To help the community to understand how final results and a model’s performance depend on each of these different aspects, we propose implementing some techniques that target the black-box nature of DL models. In this study we applied three visualisation technique to a convolutional neural network (CNN) classification model for the earthquake detection. The implemented techniques are: feature map visualisation, backward optimisation and layer-wise relevance propagation methods. These can help us answer questions such as: How is an earthquake represented within a CNN model? What is the optimal earthquake signal according to a CNN? Which parts of the earthquake signal are more relevant for the model to correctly classify an earthquake sample? These findings can help us understand why the model might fail, how to build better model architectures, but also whether there is a physical meaning embedded in a model from training samples. The CNN used in this study had been trained for single-station detection, where an input sample is a 25 seconds long three-component waveform. The model outputs a binary target: an earthquake (positive) and a noise (negative) class. Following our two output classes, our training database contains a balanced number of samples from both classes. The positive samples span a wide range of earthquakes, from local to teleseismic, with a focus on the local and regional ones. Our analysis showed that the CNN model correctly identifies earthquakes within the sample window, while the position of the earthquake in the window is not explicitly given (based on the high relevance values). The model handles well earthquakes of different distance and magnitude values, without having any physical information about them during the training process. Thus, the model constructs highly abstract latent space where different earthquakes can eventually fit (can be shown by visualising feature maps). We also notice that having non-filtered training samples with low signal to noise ratio does not disrupt the model to generate distinct feature maps, which is crucial for the successful earthquake detection process. Finally, interpretation techniques proved to be useful for having an insight of how the CNN model treats input samples, which is beneficial for understanding whether the architecture is well designed for this task. 

How to cite: Majstorovic, J., Giffard-Roisin, S., and Poli, P.: Post hoc visual interpretation of convolutional neural network model for earthquake detection using feature maps, optimal solutions, and relevance values, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2874, https://doi.org/10.5194/egusphere-egu22-2874, 2022.

An Mw=7.3 earthquake occurred on June 15, 2019 in New Zealand, Kermadec Islands (30.644° S, 178.100° W, 46 km depth), in correspondence with the Tonga-Kermadec subduction zone, and was characterized by a tectonic setting of shallow reverse faulting.

We investigated the preparatory phase from a seismological point of view, focusing on the analysis of seismic data in the period between January 1, 2018 and June 14, 2019 in an area limited by the Dobrovolsky strain radius. Specifically, the data from the global United States Geological Survey (USGS) and the national New Zealand (GEONet) earthquake catalogues are used in this study.

To characterize the seismicity trend in terms of magnitude distribution variations and strain release with time, we made a two-step analysis. The first one was to calculate the magnitude of completeness (Mc), which is an important parameter when estimating b-values (Wiemer and Wyss, B. Seism. Soc. Am., 2000). After this preliminary step, we observed that the seismicity accelerated during the preparation phase of the earthquake through the Revised Accelerated Moment Release (R-AMR) method (De Santis et al., Tectonophysics, 2015).

Finally, we found that the seismological research of the preparation phase of this earthquake helped to understand, together with other observations from ground and satellite, the so-called Lithosphere-Atmosphere-Ionosphere Coupling (LAIC) phenomena prior to the mainshock.

How to cite: Orlando, M., Cianchini, G., De Santis, A., Perrone, L., Arquero Campuzano, S., D'Arcangelo, S., Di Mauro, D., Marchetti, D., Piscini, A., Sabbagh, D., and Soldani, M.: Seismological investigation of Mw=7.3 Karmedec Islands (New Zealand) earthquake occurred on June 15, 2019, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5805, https://doi.org/10.5194/egusphere-egu22-5805, 2022.

Deep earthquakes, 300 to 700 km deep, have been observed for decades and shown to originate from major mineral transformations occurring at these depths, including phase transitions of olivine and pyroxenes. Yet, we still do not fully grasp their mechanism. Although transformational faulting in the rim of the metastable olivine wedge (MOW) is hypothesized as a triggering mechanism of deep-focus earthquakes, there is no direct seismic evidence of such rim. Variations of b-value – slope of the Gutenberg-Richter distribution – have been used to decipher triggering and rupture mechanisms of earthquakes. However, regarding deep-focus earthquakes the detection limit prevents full understanding of rupture nucleation at all sizes.

With one of the most complete catalogs, the Japan Meteorological Agency (JMA) catalog, we estimate the b values of deep-focus earthquakes (> 300 km) of four clusters in the NW Pacific Plate based on unsupervised machine learning. The applied K-means, Spectral and Gaussian Mixture Models Clustering algorithms divide the events into four clusters. For the first time, we observe kinks in the b values with abrupt reductions from 1.5–1.8 down to 0.7–1.0 at a threshold Mw of 3.7–3.8 for the Honshu and Izu clusters, while normal constant b values (0.9–1.0) are observed for the Bonin and Kuril clusters.

The four clusters found by the algorithms actually correspond to events within four different segments of the sinking Pacific lithosphere, characterized by significant differences in hydration state prior to subduction. High b values (1.5–1.8) at low magnitudes (Mw < 3.7–3.8) correlate with highly hydrated slab portions. The hydrous defects would enhance the nucleation of small earthquakes via transformational faulting within the rim. Such mechanism operates for small events with a rupture length of less than 1 km, which would correspond to the thickness of the MOW rim.

Combining with the b-value analysis from the latest CMT catalog, the kink at Mw 6.7 suggests that the thermal runaway mechanism operates for larger earthquakes rupturing through and possibly propagating outside the MOW, with increased heterogeneity in the new rupture domain. The changes of controlling mechanism and rupture domain heterogeneity due to the slab hydrous state and thermal state can explain the spatially varying b values.

How to cite: Mao, G., Ferrand, T., Li, J., Zhu, B., Xi, Z., and Chen, M.: Nature of Deep Earthquakes in the Pacific Plate from Unsupervised Machine Learning, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6650, https://doi.org/10.5194/egusphere-egu22-6650, 2022.

As a population parameter, reliable estimation of the b-value is intrinsically complicated, particularly when spatial variability is considered. We approach this issue by treating the spatial b-value distribution as a non-stationary Gauss process for the underlying earthquake-realizing Poisson process. For Gauss process inference the covariance—which describes here the spatial correlation of the b-value—must be specified a priori. We base the covariance on the local fault structure, i.e. the covariance is anisotropic: elongated along the dominant fault strike and shortened when normal to the fault trace. This adaptive feature captures the geological structure better than an isotropic covariance or similarly defined and commonly used running-window estimates of the b-value.We demonstrate the Bayesian inference of the Gauss process b-value estimation for southern California based on the SCEDC earthquake catalog and with the covariance calibrated with the USGS fault model.Our model provides a continuous b-value estimate which reflects the local fault structure to a very high degree. Therefore, we are able to associate the b-value with the local seismicity distribution and can link it to the major Californian faults and geothermal areas. This technique, in its general formulation, can be applied to other non-stationary seismicity parameters, e.g. Omori’s law.

How to cite: von Specht, S., Holschneider, M., Ferrat, K., Zöller, G., Molkenthin, C., and Hainzl, S.: Spatial distribution of the b-value in Southern California based on Gauss process inference with a geologically defined prior, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11800, https://doi.org/10.5194/egusphere-egu22-11800, 2022.

The SE-Carpathians indicate significant geodynamic activity, especially in the external part due to the current subduction processes. This part is the so-called Vrancea-zone where the distribution of the seismic events is quite dense considering the relatively small area (around 30*70 km).

The authors have carried out cluster analyses of the focal mechanism solutions and their inversions to support the recent and previously published studies in this region. They have applied different pre-existing clustering methods – e.g. HDBSCAN (hierarchical density-based clustering for applications with noise) and agglomerative hierarchical analysis – considering the geographical coordinates, focal depths and parameters of the focal mechanism solutions of the used seismic events, as well. Moreover, they have attempted to improve a fully-automated algorithm for the clustering of the earthquakes for the estimations. This algorithm has only one optional hyper-parameter which is eligible to detect the outliers from the input dataset. Due to this, it is possible to reduce the running time and subjectivity. In all cases, the calculated stress tensors are in close agreement with the previously published results.

How to cite: Czirok, L., Kuslits, L., Bozsó, I., and Gribovszki, K.: Studying the stress field variations in the Vrancea-zone using clustering-based stress inversions, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9181, https://doi.org/10.5194/egusphere-egu22-9181, 2022.

Stable continental areas—those largely unaffected by currently active plate-boundary processes—undergo little deformation and feature low seismicity rates. Notable exceptions, such as the well-known large earthquakes in the central United States or the Fennoscandian Craton, are rare but highlight the importance of understanding the seismicity in low-strain regions. One long-standing question, debated for over a century, relates to the seismicity of Ireland. Why is it much lower than that in the neighbouring Britain, even though they were assembled in the same Caledonian orogeny, share many of the ancient tectonic boundaries, and are subjected to similar tectonic stresses? Our new catalogue of Ireland’s seismicity, produced using the greatly improved seismic station coverage of the island over the last decade, shows many more micro-earthquakes than known previously but confirms the much lower seismicity rates in Ireland compared to Britain.

Comparing the distribution of seismicity with high-resolution, surface-wave tomography (performed using the abundant new data) we observe that areas with thicker, colder lithosphere feature lower seismicity than those with thinner lithosphere. This must be because the thicker and colder lithosphere is mechanically stronger and less likely to deform, compared to the thinner and weaker lithosphere under the same tectonic stress. According to the new tomography, Ireland has thicker lithosphere than most of Britain, which can explain its lower seismicity rates. The thinnest lithosphere in Ireland is found in the north of the island, in Co Donegal, and this is where most of Ireland’s micro-seismicity occurs. A similar relationship between the lithospheric thickness and seismicity rates is observed in Britain, with the London Platform in the southeast of the island showing thick lithosphere and low seismicity.

Together, lithospheric tomography and seismicity maps thus offer a solution to a seismo-tectonic puzzle first formulated in the 19-th century. Evidence of the lithospheric mantle controls on earthquake occurrence can be seen elsewhere around the world as well. The improving accuracy of the tomographic imaging of the lithosphere presents a useful new line of evidence on the mechanisms that control the regional distributions of intraplate earthquakes.

How to cite: Lebedev, S., Arroucau, P., Grannell, J., and Bonadio, R.: Intraplate seismicity controlled by lithospheric-mantle strength: the Ireland and Britain case study, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10503, https://doi.org/10.5194/egusphere-egu22-10503, 2022.

The seismicity rate in the Southern Sichuan Basin, China (SSBC), has increased by orders of magnitude within the past a few years accompanying the rapidly growing hydraulic fracturing (HF) operations. However, there is currently no appropriate method to directly determine whether a HF platform has induced seismicity and to quantitatively describe its potential of inducing earthquakes. In this study, by taking advantage of a more complete seismic catalog constructed from temporary short-period seismic stations and broadband stations for two adjacent well pads in the SSBC, we investigate a new statistical metrics for detailed studies of earthquake clusters on the "pre- and post-fracture" time scales to characterize their earthquake-inducing capacity. After declustering the earthquake catalog, a small-scale “Spatiotemporal Association Filter (SAF)” is designed to obtain seismic data closely associated with six independent fracturing well groups, and an “Unit Seismic Energy Release (USER)” index is established to evaluate the potential to induce earthquakes. Comparing the differences in the index before and after fracturing, as well as the nonparametric statistical test of each well group’s “Interevent time (IET)” and historical IET, four of the well groups are considered “induced-seismic”, and the other two are “anti-seismic”. The HF well with the largest USER value has the largest inducing capacity. The paired result of the nonparametric test shows that the p values are less than 0.001, indicating significant statistical differences between the IET series before and after the HF process around the four induced-seismic wells. To sum up, our method can conveniently distinguish the earthquake-inducing capacity of different HF wells, and thus offer practical advice for HF operation in the SSBC.

How to cite: Hu, J., Yang, H., Zhang, H., and Tan, Y.: A method for characterizing the earthquake-inducing capacity of hydraulic fracturing, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7239, https://doi.org/10.5194/egusphere-egu22-7239, 2022.

Our ability to forecast future earthquakes is hampered by the very limited information on the fault slip that produce them. In particular the current state of stress, strength, and parameters governing the slip of the faults are highly uncertain. Ensemble data-assimilation methods provide a means to estimate these variables by combining physics-based models and observations while considering their uncertainties. Perfect model experiments with an Ensemble Kalman Filter (EnKF), connected with one-dimensional (1D), two-dimensional (2D), and three-dimensional (3D) earthquake cycle models, demonstrate the ability to estimate the state variables of shear stress, slip velocity, and state (θ) of a straight fault governed by rate-and-state friction surrounded by a homogeneous elastic medium. The models represent a direct-shear laboratory setup in one, two and three dimensions, with an array of shear-strain gauges and piezoelectric transducers located at a small distance to the fault. In this research, we compare the recurrence interval and earthquake occurrence of the EnKF across the different models to better understand the challenges associated with a space-time systems with increasing dimensions and increasingly complex earthquake sequences. The assimilation of synthetic shear-stress and slip-rate observations improves in particular the estimates of shear stress and slip rate on the fault, despite the very low accuracy of the observations. We get reasonable estimates when modelling long-duration earthquakes or slow slip events . Interestingly, we also obtain very good estimates when simulating earthquakes with fast slip rates (up to m/s). The large, nonlinear, changes in stress and velocitiy  during the fast transition from the interseismic to the coseismic phase cause the distributions of the state variables to become bi-modal. The EnKF still provides a reasonable estimate of the time of occurrence of the earthquakes in the synthetic experiments, despite the inherent assumption on the Gaussianity of these distributions.

How to cite: Diab Montero, H. A., Li, M., van Dinther, Y., and Vossepoel, F. C.: An Ensemble Kalman Filter for Estimating Future Slow Slip Events and Earthquakes on 1D, 2D and 3D Synthetic Experiments, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10324, https://doi.org/10.5194/egusphere-egu22-10324, 2022.

I summarize a broad suite of laboratory data sets showing that stick-slip failure events –lab earthquakes– are commonly preceded by both measurable changes in fault zone properties and acoustic emission (AE) events that foretell catastrophic failure.  These works show that both types of data can be used to predict labquakes with machine learning (ML) methods and deep learning (DL) approaches.  The first works used continuous measurements of AE to predict the timing of labquakes and the fault zone shear stress. Subsequent studies showed that catalogs of AE events could also predict labquakes and that ML approaches could also predict stress drop, peak fault slip velocity and the duration of failure. Recently, DL has been used to predict and autoregressively forecast labquakes and fault zone shear stress. Consistent with previous works, we see that seismic b-value begins to decrease as faults unlock and start to creep.  This provides a sensible connection between the ML-based predictions, fault zone elastic properties, and the physics of failure.  In the lab, AE events represent a form of foreshock and, not surprisingly, the rate of foreshock activity correlates with fault slip rate and its acceleration toward failure.  Our work shows precursory changes in wave speed prior to labquakes, consistent with many well known past studies, but the early studies did not provide a method to predict impending failure.  ML and DL predicts with fidelity the time of impending failure and other aspects of it. This suggests the possibility of physics-based models for prediction. We are working to connect ML prediction of labquakes with the evolution of fault zone elastic properties, frictional contact mechanics and constitutive laws.  A central goal is to learn from lab earthquake prediction to improve forecasts of earthquake precursors and tectonic faulting.

How to cite: Marone, C.: Machine Learning for Understanding Lab Earthquake Prediction and Precursors, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12022, https://doi.org/10.5194/egusphere-egu22-12022, 2022.

Forecasting spatio-temporal occurrence of earthquakes is not a trivial step for the seismic and tsunami hazard assessments. Estimating earthquake rates depends on information of a nonlinear system that is poorly known including the source dimensions. Thus, these assessments rely on e.g. seismic catalogues, or geophysical and geological data that could portray the statistical and physical behaviour of the seismogenic zones. In particular, earthquakes could rupture along asperities or areas of the seismogenic zone with high stress accumulation.Those areas have different physical properties than the surrounding area, such as a high frictional strength and larger stress drop (e.g. Madariaga 1979, Corbi et al., 2017). In this work, we apply the TREMOL code (Monterrubio-Velasco et al., 2019), based on the Fiber Bundle Model, to validate it as a tool to reproduce the seismicity occurring by the rupture of large, in some cases, single asperities. We have selected four regions where large earthquakes have occurred: M8.8 Maule 2010 (Chile) earthquakes, M9.1 Tohoku 2011 (Japan), M7.6 Nicoya 2012 (Costa Rica) and M8.3 Coquimbo 2015 (Chile). In these tectonic regions, earthquake sequences are generated based on a discrete model of material failure used in TREMOL. One of the most notable results is that the maximum earthquakes of the real sequences are achieved. Also, in most cases, the magnitude - frequency distribution is similar to those of real data. While the outcomes of TREMOL are given in rupture areas, several area-magnitude scaling laws are employed to obtain moment magnitudes. By carrying out a sensitivity analysis of different scaling laws, we show the bias in the synthetic catalogues which is a critical input in seismic hazard assessment. It is shown that the synthetic seismicity using the Ramirez-Gaytan scaling law (Ramirez-Gaytan et al. 2014) is the best to fit the magnitude of the real series in most of the cases. Following the validation of TREMOL, we provide a new seismic scenario generator of future events to assist e.g. the Probabilistic Seismic/Tsunami Hazard Assessment (PSHA/PTHA) complementing the seismic forecast with other well known statistical tools. 

References

Corbi, F., Funiciello, F., Brizzi, S., Lallemand, S., and Rosenau, M. (2017). Control of asperities size and spacing on seismic behavior of subduction megathrusts, Geophys. Res. Lett., 44, 8227– 8235, doi:10.1002/2017GL074182.

Madariaga, R. (1979). On the relation between seismic moment and stress drop in the presence of stress and strength heterogeneity, J. Geophys. Res.-Sol. Ea., 84, 2243–2250.

Monterrubio-Velasco et al., (2019). A stochastic rupture earthquake code based on the fiber bundle model (TREMOL v0.1): application to Mexican subduction earthquakes. Geosci. Model Dev., 12, 1809–1831.

Ramírez-Gaytán, A., Aguirre, J., Jaimes, M. A., and Huérfano, V. (2014). Scaling relationships of source parameters of M w 6.9–8.1 earthquakes in the Cocos–Rivera–North American subduction zone, Bulletin of the Seismological Society of America, 104, 840–854.

How to cite: Monterrubio-Velasco, M. and Zamora, N.: Sensitivity analysis using the TREMOL code for seismicity forecasting, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7735, https://doi.org/10.5194/egusphere-egu22-7735, 2022.