AIM & GOAL
This session is devoted to new research in the field of physics-based stochastic modeling of natural and induced earthquakes also with the support of integrated, multidisciplinary methods, with special attention to major events.
TOPICS
Our session is focused on new methods, integrated approaches, and analyses for
making statistical earthquake forecasts more and more skillful. Research works about the following topics are especially welcome:
- Stochastic modeling of seismic sequences (ETAS and other methods).
- Applications of geodesy for the assessment of the seismogenic potential and short- to long-term earthquake forecasting.
- Statistical characterization and physical reconstruction of paleoseismic records and long-term recurrences of large earthquakes.
- Investigation of the relationships between tectonics and large earthquakes occurrences.
- Mapping fluids flow in the brittle crust and their relationship with natural and induced seismicity.
- Modeling swarm-like seismic activity using stochastic and physics-based techniques.
- Crustal stress modeling using different methods (moment tensors, b-value …).
- Applications of moment tensors to seismic hazard and forecasting of tensorial properties of seismicity.
- Established and new techniques for statistical seismology and their impact on forecasting (declustering, relocalization of events, catalogue homogenization, …).
- Numerical simulations for large earthquake scenarios.
- Physics-enhanced AI-driven modeling of earthquake occurrence and monitoring crustal stability conditions.
- Short-term extreme-event and aftershocks forecasting.
Enhanced Geothermal Systems (EGS) comprise technologies aiming to harness geothermal energy from the Earth's subsurface by enhancing the productivity of existing or naturally occurring geothermal reservoirs. Unlike conventional geothermal systems (hydrothermal systems) that rely on naturally permeable rock formations, EGS involve creating or enhancing fractures in low permeability or impermeable rock mass through hydraulic stimulation. EGS has the potential to expand the geographical reach of geothermal energy utilization and increase the overall efficiency and sustainability of geothermal power generation.
The injection of pressurised fluids and opening of fractures manifests as micro-seismicity, which is in most cases a normal indication of the reservoir stimulation process. However, some cases have seen unwelcome large magnitude and even damaging events. The US department of Energy has sponsored the Utah Frontier Observatory for Research in Geothermal Energy (FORGE), a flagship demonstration site aiming to demonstrate to the public, stakeholders and the energy industry that EGS technologies have the potential to contribute safely and significantly to future low-carbon power generation. The site is thoroughly instrumented with monitoring boreholes combining geophone chains and fibre-optic cables for DSS and DAS, as well as a dense surface network of seismometers, allowing for the generation of high-resolution seismic catalogues during and after the stimulation phases.
At FORGE, the granitoid reservoir with temperatures exceeding 220°C around 2300 m b.s.l. has been stimulated in hours-long stages in April 2022 and April-May 2024. During both sets of hydraulic stimulations, we monitored the micro-seismicity and ran forecasting models in an ATLS framework (Adaptive Traffic Light System). From the raw real-time catalogue, we use advanced techniques (machine-learning based pickers, template matching, …) to generate enhanced catalogues which help us investigate essential the dynamics of micro-seismicity and reservoir creation. For the ATLS, three forecasting models classes were used: an empirical model that relates injection rates and rate of seismicity based on the seismogenic index model; a machine learning based model able to weight timeseries measurements of past seismicity and hydraulic parameters to output a forecast rate of seismicity; and a hybrid hydromechanical model that generates seismicity based on a linear or non-linear pressure solution. This implementation of an ATLS at full scale and in real-time paves the way for future implementations in Switzerland and abroad in an effort to derisk and generalise EGS.
How to cite: Ritz, V. A., Lanza, F., Rinaldi, A. P., Schmid, N., Clasen Repollés, V., Shi, P., and Wiemer, S.: FORGEcasting induced seismicity in real-time with statistical and physics-informed models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7406, https://doi.org/10.5194/egusphere-egu25-7406, 2025.
Ensemble modeling combines different models or their parametrizations into a single model. Conventional ensemble methods merge individual forecast distributions into one (e.g., the average). We introduce the Ontological Ensemble (OE) model, which preserves all individual forecast distributions and quantifies their dispersion, thereby capturing the epistemic uncertainty. This integrity acknowledges different kinds of uncertainties, keeps them separated, and provides a complete description of our knowledge and its limitations. Unlike conventional ensemble methods, the OE quantifies the reliability of a forecast and enables a more meaningful model validation. Specifically, the OE allows exposing representational errors of a system, the so-called ontological error [Marzocchi & Jordan 2014], by testing if observations (i.e., the “true” unknown distribution of the underlying process) fall outside the OE distribution. To construct this new type of ensemble, our approach is twofold:
- In a first step, we create a weighted average ensemble (in terms of an average forecast). We employ multi-variate logistic regression to obtain model weights that maximize the forecasting skill of the ensemble [Herrmann & Marzocchi 2023]. Retrospective testing on 15 years of Operational Earthquake Forecasting data in Italy [Marzocchi et al. 2014] demonstrated a significant improvement over the best candidate forecast model in terms of cumulative information gain per earthquake (cumIGPE).
- In a second step, we create the actual OE forecast by modeling a forecast distribution with the Beta distribution and the weighted dispersion of the candidate forecasts (using the weights and weighted average determined in step 1).
Our ensemble software framework is flexible and extensible. In step 1 for instance, we implemented sequence-specific ensembling as a more advanced ensemble strategy to acknowledge the spatiotemporal variability of seismicity and forecasts. It extends Herrmann & Marzocchi 2023 by not only fitting the logistic regression to the whole region (i.e., all spatiotemporal bins), but separately to sequences (i.e., only the affected spatiotemporal bins) and excluding those from the regional fit. This separation also better exploits the candidate forecast models: it acknowledges those that perform well during sequences (aftershocks) and those that perform well generally (background seismicity). Compared to the previous (purely regional) ensemble, it improved the cumIGPE over the best forecast model by 56%. Additionally, it leads to a more honest uncertainty quantification in the OE. We have also operationalized our framework for near real-time applications.
Validating this new type of forecast model requires new testing routines, which we plan to develop for the Collaboratory for the Study of Earthquake Predictability (CSEP, cseptesting.org); it will involve implementing the OE in pyCSEP [Savran et al. 2022] and/or floatCSEP [Iturrieta et al.]. Future plans also include exploring more ensemble configurations and strategies to further improve forecast skill and uncertainty quantification.
References
Herrmann & Marzocchi (2023). Maximizing the forecasting skill of an ensemble model. doi: 10.1093/gji/ggad020
Iturrieta et al. (in preparation). Modernizing CSEP Earthquake Forecasting Experiments: The Floating Testing Center.
Marzocchi & Jordan (2014). Testing for ontological errors in probabilistic forecasting models of natural systems. doi: 10.1073/pnas.1410183111
Marzocchi et al. (2014). The establishment of an operational earthquake forecasting system in Italy. doi: 10.1785/0220130219
Savran et al. (2022). pyCSEP: A Python Toolkit for Earthquake Forecast Developers. doi: 10.1785/0220220033
How to cite: Herrmann, M. and Marzocchi, W.: Ontological ensemble modelling to account for different kinds of uncertainties, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18585, https://doi.org/10.5194/egusphere-egu25-18585, 2025.
In earthquake forecasting, a significant gap exists between complete randomness and full predictability. Shannon's information entropy provides a conceptual framework for quantifying randomness in stochastic systems. Predictability, in this context, is defined as the reduction in entropy relative to a completely random system. When a forecasting model is applied to observational data, its performance is constrained by two factors: its inherent predictability (referred to as its predictability capacity) and the predictability inherent in the observational data.
For the widely used ETAS model, the corresponding system of complete randomness is a stationary Poisson process with the same mean occurrence rate. Numerical computations demonstrate that an ETAS model with a higher branching ratio and denser clusters exhibits a greater predictability capacity.
It is well established that known predictabilities in seismicity include spatiotemporal clustering and spatial inhomogeneity. However, significant predictability in earthquake magnitude has not been identified, despite ongoing debates about characteristic earthquakes and magnitude dependencies within earthquake clusters.
The following conclusions can be drawn:
(1) Determining the upper limit of each model's forecasting performance is as important as testing its forecasting consistency.
(2) The key to improving forecasting lies in developing more informative models with lower system entropy rates. While model calibration (e.g., enhanced fitting procedures) can provide some improvement, these gains are inherently limited.
(3) To achieve better practical performance in forecasting earthquake magnitudes, we must move beyond the Gutenberg-Richter law and the assumption of magnitude independence. But how?
How to cite: Zhuang, J.: Quantifying Earthquake Predictability and Advancing Forecasting Models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2405, https://doi.org/10.5194/egusphere-egu25-2405, 2025.
The widely used ETAS seismicity model describes the clustering of seismic events as an epidemic-type process (property A), assuming that the F1 distribution of the number of direct aftershocks is Poissonian (property B). The real data favor the geometric distribution F1 (e.g. Shebalin et al., 2018, Dokl. Akad. Nauk, 481 (3), 963–966). The F2 distribution of the number of cluster events with main shock m and magnitudes greater than m-Δ is also often attributed to geometric type. However, the coincidence of distribution types F1 and F2 turns out to be in contradiction with the A-property, and the geometric type F1 is in contradiction with the B-property. This study, which is analyzing and resolving the described contradictions, develops in the following three steps.
Step 1- Generalization of the ETAS model, designed to use any F1 distribution, and selection of special class of F1 distributions, including both Poisson and geometric distributions. The class of F1 models is united by a common property inherent in the Poisson distribution: the number of events with F1 distribution at random thinning of sample elements changes the mean, but retains the F1 type. This requirement is relevant because of errors in real clusters identification and because of the ambiguity in the choice of the representativeness magnitude threshold.
Step 2- Instead of the F2 distribution, we consider its more natural analog F2a. It refers to cluster events with magnitude greater than ma-Δa , where ma is the mode in the theoretical distribution of the strongest aftershock with main shock m. Under the conditions of Bath's law, namely ma = m -1.2, both distributions coincide if Δa = Δ-1.2. The limiting distribution of F2a is found for clusters with sufficiently strong main shock, m>>1. Remarkably, in the subcritical regime its type coincides with the type of F1, and the distribution itself depends only on the relative threshold Δa. In practice, this asymptotic result applies to the magnitude range where we expect self-similarity in seismicity.
Step 3- Comparison of the limit distributions of F2a corresponding to the geometric distribution of F1, different Δa and b-value b=1, with its real analogs obtained from the global ANSS catalog for large events with magnitude m>6. In the absence of any fitting, we obtained surprisingly good agreement of the distributions within the principal values (0-0.95) of the theoretical limit F2a distribution.
Thus, the important structural A-property of the ETAS model and the consistency of the choice of the geometric distribution F1 in its generalization are confirmed. This justifies the use of such a model to simulate seismicity. The complete mathematical analysis of the limiting distribution F2a, performed for the first time even for the traditional ETAS model, is of independent interest (for further details see Molchan and Peresan, 2024, Geophys. J. Int. 239, 314-328 and references therein).
How to cite: Molchan, G., Peresan, A., and Varini, E.: Number of Aftershocks in Epidemic-Type Seismicity Models and in Reality, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2433, https://doi.org/10.5194/egusphere-egu25-2433, 2025.
In this presentation we propose to extend the ETAS model to micro-seismic
events. For that we interpret the triggered events in an ETAS model as individual
local clock advances of an independent background process. The solution of the ETAS model thus
becomes the sum of an infinite Markov chain of independent time adjusted
background processses. This allows the incorporation of events at all scales. No artificial small
magnitude cutoff is needed. We also discuss the implication to the stability of the ETAS model without a large magnitude cutoff. We give a proof that the seismicity explodes in finite time as soon as the expected production rate becomes overcritical.
How to cite: Holschneider, M.: The clock advance picture of the ETAS model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20373, https://doi.org/10.5194/egusphere-egu25-20373, 2025.
Temporal variations of the b-value, the parameter describing the frequency distribution of earthquake magnitudes, also known as the Gutenberg-Richter law, are believed to provide critical insights into the physical processes governing earthquake sequences. Such temporal variations are often difficult to estimate because of the sudden changes in the detectability of events. The standard maximum likelihood estimate for b-value, requires truncating the seismic catalogue above a "completeness" magnitude threshold, above which all earthquakes are supposed to be detected. However, temporal variations of detectability are particularly important for mainshock-aftershock sequences, which are subject to short-term aftershock incompleteness. The b-positive approach and its recent updates are adaptations of the maximum likelihood approach to circumvent the completeness magnitude dependence for mainshock-aftershock sequences, but this method still relies on the arbitrary choice of moving-window size to capture variations in the b-value estimate. To address these challenges, we use the recent probabilistic b-Bayesian approach to jointly invert for the temporal variations in b-value and detectability. Such a probabilistic approach also provides time variations of uncertainties for these two quantities.
We will present results of the application of b-Bayesian for seismic sequences presenting short-term aftershock incompleteness, and show the potential of Bayesian methods to avoid over-interpretations of b-value temporal variations in context of large detectability changes.
How to cite: Laporte, M., Bodin, T., Durand, S., Gardonio, B., and Marsan, D.: Exploring Temporal Variation of the b-value in Mainshock-Aftershock Sequences with the b-Bayesian Probabilistic Approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4360, https://doi.org/10.5194/egusphere-egu25-4360, 2025.
The incompleteness of seismic catalogs obscures the true spatiotemporal and magnitude-based organization of earthquakes, often resulting in unreliable estimates of key parameters that govern empirical laws. In this presentation, I will demonstrate how a statistical approach based on positive differences between consecutive magnitudes can reliably recover underlying statistical laws, even when data is incomplete.
In particular, I will show that traditional estimators systematically underestimate the b-value of the Gutenberg-Richter law in regional catalogs. Furthermore, I will provide robust evidence of correlations between consecutive positive magnitude differences, strongly supporting the hypothesis that the distribution of subsequent earthquake magnitudes depends on the magnitude of the triggering event.
How to cite: Lippiello, E., Godano, C., and Petrillo, G.: Unveiling the True b-Value and Magnitude Correlations in Earthquakes Using Positive Magnitude Difference Statistics, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11703, https://doi.org/10.5194/egusphere-egu25-11703, 2025.
Foreshocks are among the few observable phenomena preceding many large earthquakes and hold potential for short-term earthquake forecasting. However, their identification and interpretation remain challenging, particularly due to variations in seismic monitoring capabilities, such as the magnitude of completeness (Mc) in earthquake catalogs. This study investigates the impact of Mc on the proportion of mainshocks identified with foreshocks (Pf) in California, using four popular methods: the Space-Time Window (STW) method, Nearest-Neighbor Clustering (NNC) method, Empirical Statistical (ES) method, and the Epidemic-Type Aftershock Sequence (ETAS) model.
Results show that Pf estimated by the STW method strongly depends on Mc, with higher Mc values leading to lower Pf due to the misclassification of background events as foreshocks. In contrast, the NNC and ES methods yield more consistent Pf values across different Mc thresholds, though the ES method reports slightly lower Pf due to its sensitivity to background seismicity rates. The ETAS model reveals that at low Mc, a greater proportion of foreshocks are associated with aseismic processes, whereas at high Mc, distinguishing between aseismic and cascade-driven mechanisms becomes increasingly challenging. These findings suggest that enhanced seismic monitoring has limited effects on Pf, but is essential for identifying the underlying processes driving foreshocks.
How to cite: Cui, X., Li, Z., Ampuero, J.-P., and De Barros, L.: Influence of Seismic Monitoring Capability on Foreshock Identification in California: A Comparative Analysis of Methods, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10735, https://doi.org/10.5194/egusphere-egu25-10735, 2025.
This study validates several seismic models with long- and short-term forecasting capabilities used in probabilistic seismic hazard assessment (PSHA) and evaluates their impact on hazard levels in the Longitudinal Valley, Taiwan, a region characterized by high seismic activity and data quality. The Gutenberg-Richter (G-R) law demonstrates good fitting performance for the long-term rate in small to moderate magnitudes, while the pure characteristic earthquake (PCE) model, which assesses the maximum recurrence rates for individual seismogenic structures, better fits the long-term rate for large magnitudes. The Seismic Hazard and Earthquake Rates in Fault Systems (SHERIFS) model integrates the G-R law and structural parameters while considering multiple fault ruptures. It performs well in forecasting long-term seismicity rates, particularly for medium to large magnitudes. Recognizing the limitations of long-term seismic models in short-term and aftershock forecasting, we further incorporate the Epidemic-Type Aftershock Sequence (ETAS) model to analyze short-term earthquake occurrence rates and assess the temporal evolution of seismic hazard. The model is validated using the maximum ground shaking observed at strong-motion stations during the given short observation period. The ETAS model complements existing approaches by providing more immediate forecasts of seismic activity. Our findings provide hazard assessment results across different time scales and underscore the importance of integrating multiple seismic models for precise seismic hazard assessment. This study offers valuable insights into earthquake processes and provides essential parameters for future PSHA studies in Taiwan.
How to cite: Lu, T.-Y. and Chan, C.-H.: Time-Dependent Probabilistic Seismic Hazard Assessment in Complex Fault Systems: Exploring the Longitudinal Valley of Taiwan, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3925, https://doi.org/10.5194/egusphere-egu25-3925, 2025.
The predictive potential of the b-value from the Gutenberg-Richter law remains a key topic in seismology. In this study, we apply a probabilistic b-value estimation method to the BMKG catalog (2008–2024) for the Indonesian region, focusing on background and triggered events. Our findings reveal that the b-value for background and triggered events is statistically indistinguishable for magnitudes greater than the completeness magnitude (Mc = 4.7). While limited indications of differences are observed in small subsets of data, the high Mc prevents detailed analyses of lower magnitudes. This universality of b-values between background and triggered events suggests a shared underlying physical process. Leveraging this insight, we propose combining all events into a single dataset to enhance reliability and statistical power for forecasting.
However, attempts to perform blind forecasting analyses based on b-value variations before and after major (M > 7) earthquakes are hindered by the insufficient number of events in the time windows preceding large earthquakes, preventing an unbiased estimation of the b-value. Even by extending spatial and temporal windows to reasonable limits, the lack of data remains a critical limitation. This emphasizes the need for improved seismic catalog completeness to enable reliable forecasting based on the b-value. While the b-value holds potential as a universal metric for seismic hazard assessment, its use as a forecasting tool requires not only refined methodologies but also significantly enhanced data quality and coverage.
References
1Godano, C., Petrillo, G., & Lippiello, E. (2024). Evaluating the incompleteness magnitude using an unbiased estimate of the b value. Geophysical Journal International, 236(2), 994-1001.
2Lippiello, E., & Petrillo, G. (2024). b‐more‐incomplete and b‐more‐positive: Insights on a robust estimator of magnitude distribution. Journal of Geophysical Research: Solid Earth, 129(2), e2023JB027849.
How to cite: Petrillo, G. and Dal Zilio, L.: Statistical Constraints of b-Value Based Forecasting in Indonesia, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7957, https://doi.org/10.5194/egusphere-egu25-7957, 2025.
Seismic hazard assessment relies on the statistical analysis of seismic catalogs. A significant challenge arises from the strong differences in completeness and quality between historical and instrumental catalogs. Combining these catalogs thus necessitates statistical methods capable of integrating and addressing the biases introduced by these differences.
Two key parameters to estimate in statistical seismology are the b-value and the seismic rate, which define the Gutenberg Richter distribution. Usually, their estimation requires to truncate the catalog above an arbitrarily defined completeness magnitude. Here instead, we introduce a detection function that models the varying detectability of earthquakes over time. The detection function is characterized by two additional parameters, which we jointly estimate with the b-value and the seismic rate through a Bayesian approach. This method enables the use of all observed events to produce a full probabilistic solution, delivering posterior distributions for each parameter while also quantifying uncertainties and correlations between them. The estimation process is implemented using a Markov chain Monte Carlo (McMC) method to efficiently explore the parameter space.
We show an application to a seismic catalog in metropolitan France. We assume that the b-value and the seismic rate remain constant over time, and we invert for two distinct detection functions—one for the historical catalog and another for the instrumental catalog. Our results demonstrate that the b-value and the seismic rate are better estimated when accounting for these variations in detectability and catalog completeness. Finally, by producing full probabilistic solutions, this method provides b-value and seismic rate estimates with their uncertainties, thus providing valuable information for PSHA calculations.
How to cite: Colin, C., durand, S., bodin, T., laporte, M., arroucau, P., and daniel, G.: b-a Probabilist : a Bayesian Estimation of the Gutenberg-Richter Parameters (b,a) for non-Truncated Catalogs with Temporal Detectability Changes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10143, https://doi.org/10.5194/egusphere-egu25-10143, 2025.
We represent an earthquake catalog as a complex network, where each node corresponds to a seismic event, and the links between nodes represent their distances in the space-time-energy domain. This approach, commonly known as the nearest-neighbor (NN) method, was introduced by Baiesi and Paczuski in 2004 to analyze the Californian earthquake catalog. Unlike the traditional fixed window-based techniques, which group into a cluster all the events within a predefined space-time domain centered at a selected mainshock, the nearest-neighbor method offers greater flexibility and precision. In fact, the resulting weighted network is able to capture extensive information about the space-time seismicity of a given region.
In our study, we use network theory to analyze the Italian earthquake catalog from 2010 to 2020, compiled by the National Institute of Geophysics and Volcanology (INGV). Specifically, we first apply the NN method to derive the seismic Italian structure as a weighted network. Then, we focus our analysis on its topological properties to detect communities, that are groups of nodes more tightly interconnected with each other than with the rest of the network.
To achieve this, we use the Louvain and Leiden algorithms: they are based on the maximization of modularity, a metric that quantifies the strength of a network’s division into modules (also referred to as communities or clusters). For implementation, we rely on IT tools new to seismology, but largely used in complex network analysis. Specifically, we use Radatools, an Ada library designed for analyzing complex networks, alongside the NetworkX Python package, which facilitates the creation, manipulation, and study of network structures, dynamics, and functions.
This approach enables us to identify communities in the Italian network, and to compare them with clusters derived using traditional window-based techniques commonly employed in the literature, such as the Gardner-Knopoff technique. By investigating these differences and similarities, we aim to provide a robust and comprehensive analysis of the Italian earthquake catalog, leveraging high-performance tools for studying complex seismic networks.
How to cite: Tavani, F. and Spassiani, I.: Detecting communities and complex network features of the Italian earthquake catalog, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10794, https://doi.org/10.5194/egusphere-egu25-10794, 2025.
The earthquake magnitude distribution, commonly described by the Gutenberg-Richter law, is governed by the b-value, a parameter that quantifies the relative frequency of large versus small earthquakes. Variations of the b-value have been attributed to different physical factors, such as tectonic setting, focal mechanism, lithology, fault geometry, and differential stress. Here we contribute to increase the understanding of the b-value variations analyzing the global correlation with the heat flux.
Our analysis reveals a specific trend: globally, we find a positive correlation between high b-values and high heat flux mostly driven by the peculiar seismicity at oceanic spreading ridges, independent of the focal mechanisms of seismic events. More interesting is the fact that, even removing from the analysis the seismicity at these spreading ridges, the significant correlation persists, independently from other local factors such as tectonic regime, or lithological composition.
These findings align with prior localized studies on thermally induced microfracturing and volcanic seismicity, where heat has been observed to influence fracture mechanics. Our results extend these observations to a global context, establishing heat flow as an important control on earthquake magnitude distribution.
How to cite: Corrado, P., Taroni, M., Cordrie, L., Basili, R., Marzocchi, W., and Selva, J.: Heat flow as a primary control factor on global earthquake magnitude distribution, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10987, https://doi.org/10.5194/egusphere-egu25-10987, 2025.
The Epidemic Type Aftershock Sequence (ETAS) model, an example of a self-exciting, spatiotemporal, marked Hawkes process, is widely used in statistical seismology to describe the self-exciting mechanism of earthquake occurrences. Fitting an ETAS model to data requires estimating the conditional intensity function, which represents the rate at which earthquake events occur, conditioned on the history of previous events. Many existing methods, both parametric and non-parametric, have limitations in quantifying uncertainty, as most estimation techniques provide only a point estimate. The GP-ETAS model defines the background intensity in a Bayesian non-parametric way through a Gaussian Process prior, enabling us to incorporate prior knowledge and effectively encode the uncertainty arising from both data and prior information. Building on the spatiotemporal GP-ETAS model, we have developed the non-stationary GP-ETAS model, which allows the background intensity and aftershock productivity parameter to be time-dependent. We aim to use the non-stationary GP-ETAS model to study seismicity in areas with slow-slip earthquakes.
How to cite: Niu, Y. and Zhuang, J.: Non-stationary GP-ETAS model , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14186, https://doi.org/10.5194/egusphere-egu25-14186, 2025.
Earthquake forecasting at different time scales is one of the most challenging goal for seismologists and geophysicists. The delivery of reliable forecasts is crucial to reduce seismic risk and establish rational operational strategies, but the task is demanding due to the complex nature of the earthquake phenomenon. Seismic events are characterized by self-organized criticality, and can be labeled as extreme, rare events, making the probability theory a fundamental tool to resort to.
Retrospective studies in statistical seismology mainly focus on the evolution of aftershocks following a large event, the results being then used for analyses of prospective type. However, much attention is also paid to the largest event in the earthquake sequence, because the forecasting of extreme events might be crucial to prevent significant damage or casualties.
In this study, we derive the probability of extreme events in any seismic cluster generated by the Epidemic Type Aftershock Sequence (ETAS) model, a benchmark in statistical seismology for any probabilistic earthquake forecasting application. Specifically, we compute the probability for the largest event within any ETAS cluster to occur at a specific space, time, and magnitude point, considering both the temporal and the spatial components of the process. The results obtained shed light on understanding the distinguishing features between mainshocks and foreshocks, and may actively contribute to operational forecasting in assigning, in real-time, the probability for any new event to be the largest of an ongoing seismic sequence.
How to cite: Spassiani, I., Petrillo, G., and Zhuang, J.: Distribution related to all samples and extreme events in any seismic cluster generated by the Epidemic Type Aftershock Sequence (ETAS) model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14378, https://doi.org/10.5194/egusphere-egu25-14378, 2025.
The Campi Flegrei caldera has experienced several episodes of volcanic unrest during the last few centuries, most notably in 1982-1984 and 2005-present. These periods of unrest are characterized by ground uplift, seismic swarms, and increased degassing. In this study, we compare the seismicity and associated b value variations during the 1982-1984 and 2005-2024 unrest periods. The b value is calculated using the novel b more positive method, which improves upon traditional approaches by analyzing the magnitude difference between successive earthquakes, without the need to estimate the completeness magnitude. We evidence the significant differences in the spatial and temporal evolution of b values between the two unrest periods. In particular, the 2005-2024 unrest exhibits a slower ground uplift rate but higher fluctuations in the b value, especially in shallower seismicity, possibly suggesting different underlying mechanisms compared to the 1982-1984 crisis. We also observe distinct regions of increased stress, particularly beneath Pozzuoli harbor and Pisciarelli for deeper seismicity, during the ongoing unrest. Our findings provide valuable insights into the evolution of Campi Flegrei volcanic systems and highlight the importance of continuous monitoring of the b value as a potential strain meter for describing volcanic activity.
How to cite: Tramelli, A., Convertito, V., Godano, C., and Petrillo, G.: The b value as a strain meter: a comparison of the 1982-1984 and 2005-2024 volcanic unrest at Campi Flegrei - Italy, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20479, https://doi.org/10.5194/egusphere-egu25-20479, 2025.
Posters virtual: Mon, 28 Apr, 14:00–15:45 | vPoster spot 1
EGU25-2519 | Posters virtual | VPS21
The determination of the seismic sequence characteristics and post - earthquake trend of the Ms6.4 earthquake in Yangbi, Yunnan, China on May 21, 2021Mon, 28 Apr, 14:00–15:45 (CEST) | vP1.15
This article mainly studies the characteristics of the earthquake sequence and the post - earthquake trend of the Ms6.4 earthquake in Yangbi, Yunnan，China on May 21, 2021. The research area is located in Yangbi Yi Autonomous County in the western part of Yunnan Province. The earthquake caused severe disasters such as housing destruction, traffic interruption, water conservancy facilities damage and power supply interruption. Through the analysis of the basic parameters of the earthquake, the tectonic stress environment and the seismogenic structure, it is determined that the earthquake is a right - lateral strike - slip rupture, with a focal depth of 8 kilometers, consistent with the direction of the Weixi - Qiaohou and Honghe fault zones. The earthquake sequence type is determined as the main shock - aftershock type (including the foreshock - main shock - aftershock type). Spatially, the source rupture expands unilaterally from the northwest to the southeast, mostly occurring in the upper crust high - speed zone or the high - low speed transition zone. Based on the G - R relationship and other analyses, the earthquake activity cycle in this area has active and quiet periods, and there are certain abnormal change laws before strong aftershocks, such as strain accumulation, calmness or enhancement of earthquakes above magnitude 3.5, and abnormal frequency of earthquakes above magnitude 2. The conclusion is that the earthquake sequence is normal, and the post - earthquake trend shows the characteristics of long - term calmness - breaking calmness - becoming calm again - signal earthquake (main shock). In the next few years, the strain accumulation may reach the peak and release. It is predicted that there may be a larger earthquake accompanied by strong aftershocks in 2025, or enter an active period with a strong aftershock magnitude exceeding 5.9 and lasting for more than half a year. Finally, the earthquake prevention and disaster reduction countermeasures are proposed.
How to cite: Wu, B.: The determination of the seismic sequence characteristics and post - earthquake trend of the Ms6.4 earthquake in Yangbi, Yunnan, China on May 21, 2021, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2519, https://doi.org/10.5194/egusphere-egu25-2519, 2025.
