EGU2020-7008, updated on 11 Jan 2022
EGU General Assembly 2020
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Insights on the European Fault-Source Model (EFSM20) as input to the 2020 update of the European Seismic Hazard Model (ESHM20)

Roberto Basili1, Laurentiu Danciu2, Michele Matteo Cosimo Carafa3, Vanja Kastelic3, Francesco Emanuele Maesano1, Mara Monica Tiberti1, Roberto Vallone1, Eulalia Gracia4, Karin Sesetyan5, Jure Atanackov6, Barbara Sket-Motnikar7, Polona Zupančič7, Kris Vanneste8, and Susana Vilanova9
Roberto Basili et al.
  • 1Istituto Nazionale di Geofisica e Vulcanologia, Roma, Italy (
  • 2Swiss Seismological Service, ETH Zurich, Zurich, Switzerland
  • 3Istituto Nazionale di Geofisica e Vulcanologia, L’Aquila, Italy
  • 4Institut de Ciències del Mar-CSIC, Barcelona, Spain
  • 5Boğaziçi University, Istanbul, Turkey
  • 6Geological Survey of Slovenia, Ljubljana, Slovenia
  • 7Slovenian Environment Agency, Ljubljana, Slovenia
  • 8Royal Observatory of Belgium, Brussels, Belgium
  • 9Instituto Superior Tecnico, Universidade de Lisboa, Lisboa, Portugal

The H2020 Project SERA (WP25-JRA3; is committed to updating and extending the 2013 European Seismic Hazard Model (ESHM13; Woessner et al., 2015, Bull. Earthquake Eng.) to form the basis of the next revision of the European seismic design code (CEN-EC8). Following the probabilistic framework established for ESHM13, the 2020 update (ESHM20) requires a continent-wide seismogenic model based on input from earthquake catalogs, tectonic information, and active faulting. The development of the European Fault-Source Model (EFSM20) fulfills the requirements related to active faulting.

EFSM20 has two main categories of seismogenic faults: crustal faults and subduction systems. Crustal faults are meant to provide the hazard model with seismicity rates in a variety of tectonic contexts, including onshore and offshore active plate margins and plate interiors. Subduction systems are meant to provide the hazard model with both slab interface and intraslab seismicity rates. The model covers an area that encompasses a buffer of 300 km around all target European countries (except for Overseas Countries and Territories, OTCs), and a maximum of 300 km depth for slabs.

The compilation of EFSM20 relies heavily on publicly available datasets and voluntarily contributed datasets spanning large regions, as well as solicited local contributions in specific areas of interest. The current status of the EFSM20 compilation includes 1,256 records of crustal faults for a total length of ~92,906 km and four subduction systems, namely the Gibraltar Arc, Calabrian Arc, Hellenic Arc, and Cyprus Arc.

In this contribution, we present the curation of the main datasets and their associated information, the criteria for the prioritization and harmonization across the region, and the main strategy for transferring the earthquake fault-source input to the hazard modelers.

The final version of EFSM20 will be made available through standard web services published in the EFEHR ( and EPOS ( platforms adopting FAIR data principles.

The SERA project received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No.730900.

How to cite: Basili, R., Danciu, L., Carafa, M. M. C., Kastelic, V., Maesano, F. E., Tiberti, M. M., Vallone, R., Gracia, E., Sesetyan, K., Atanackov, J., Sket-Motnikar, B., Zupančič, P., Vanneste, K., and Vilanova, S.: Insights on the European Fault-Source Model (EFSM20) as input to the 2020 update of the European Seismic Hazard Model (ESHM20), EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7008,, 2020.


Display file

Comments on the display

AC: Author Comment | CC: Community Comment | Report abuse

displays version 1 – uploaded on 07 May 2020
  • CC1: Questions from the chat, Esther Hintersberger, 08 May 2020

    Question 1: did you consider in the final model the possibility to relax the segmentation and to consider the possible interaction between the faults?

    Question 2: crustal faults in peninsular Italy: the maximum magnitude is on a E-W fault system (slide 9), quite in contrast with the historical observations: which is the explanation?

    • AC1: Reply to CC1, Roberto Basili, 11 May 2020

      Thanks a lot for the interesting questions which give me the opportunity to add details to my presentation.


      Reply to Q1:

      The model segmentation is implicitly relaxed, at least as much as the input data allows us. At the moment, there is no plan to use an explicit segmentation model. However, one of the issues we are facing is due to the fact that different regional datasets and sometimes also within the different regional datasets, some faults were mapped with some possible segmentation model in the mind of individual contributors. We struggle to harmonize these data toward a coherent fault source model trying to preserve the original data. A compromise that is not straightforward to achieve.


      Reply to Q2:

      The question does not specify in which way the maximum magnitude assigned to a fault contrasts with the historical observations. As specified in slide 9 these are preliminary results and the subject is presently under revision. However, there exist at least two ways in which these estimates can contrast with one another and I will try to address both.

      The most important discrepancy is when the maximum magnitude assigned to a fault is smaller than the observed seismicity. Addressing this potential issue is crucial because the model will not be able to reproduce the observations. This is our main concern at the moment.

      Another form of a discrepancy, that is often reported, is when the maximum magnitude assigned to a fault is larger than the observed seismicity. This is not necessarily a discrepancy because it has to do with the recurrence interval of earthquakes of that specific magnitude which in slide 9 is not indicated.

  • CC2: Comment on EGU2020-7008, Julian Garcia-Mayordomo, 11 May 2020

    Hi Roberto, thanks for your presentation and congrats on the good work done. Just a quick comment on the (active?) subduction zone you show at the Gibraltar Arc. Its current activity is an issue under debate, but, so far, I think the strongest evidence is that the subduction is no longer active. Spite of the fact that under the Alborán Sea there is a source of deep earthquakes related to a disconnected piece of slab (the remains of a preterit subduction). In QAFIv.2 we used to depict an active accretionary wedge in the gulf of Cádiz area. In v.3 however we rejected this feature as there was new evidence of Quaternary sediments sealing the structure. This is particularly important for tsunami source modelling in the area. Cheers, Julián.

    • AC2: Reply to CC2, Roberto Basili, 11 May 2020

      Thank you, Julian, for the very important question. This is certainly one of the major epistemic uncertainty we have in earthquake hazards in Europe with this very old subduction systems. In fact, this kind of debate also applies to the Calabrian Arc for which, only in 2018, two papers came to opposite conclusions (compare Carafa et al., 2018, GRL, doi: 10.1002/2017GL076554, with Nijholt et al., 2018, GJI, doi: 10.1093/gji/ggy144). To a slightly different extent, this debate also applies to the Hellenic and Cyprus arcs. Back to Gibraltar, and for the benefit of the readers, there is this recent paper by Civiero et al. 2020, JGR, doi: 10.1029/2019JB018873, which summarizes the debate you mention, although concluding that the subduction is still likely active in the eastern Rif and western Betics. These are the zones where we actually mapped the potentially active slab and seismic interface. One emerging evidence, although still to be verified, is that without the subduction the observed seismicity rates can hardly be matched (big question mark here!). This is one of the good reasons, in addition to several others, to keep the subduction systems separate from the crustal faults, so that the important issues with their epistemic uncertainty can be more easily handled within a logic tree weighting scheme in the successive steps of the hazard analysis.

      • AC4: Reply to AC2, Roberto Basili, 11 May 2020

        As you already noted, however, these slab models are also meant to treat the deeper intraslab earthquakes for which the level of debate and associated epistemic uncertainty is more moderate.

  • AC3: Comment on EGU2020-7008, Roberto Basili, 11 May 2020

    This is the introductory commentary I posted in the chat to accompany the presentation.

    This presentation outlines some of the necessary data manipulations and organizations in step 1 of a typical “Cornell-like” approach to PSHA. It takes the fault information and “torture” it ;-) until just before obtaining a frequency-magnitude distribution (FMD), which is step 2 of typical PSHA. Steps 3 and 4 are beyond the scope for today. Below I summarize a few elements of the EFSM20 that are not necessarily evident from the presentation.

    Recall that we are dealing with a continent-wide hazard model. Therefore, we need to manage the tradeoff between breadth of geographic coverage and detail of the information. Also, some restrictions and simplifications are needed to make the model feasible in terms of computational time. Some of these procedures are oriented at a specific hazard platform. In this case, the GEM OpenQuake (OQ). Using another platform would require additional tweaking.

    The 3D geometry of crustal faults is derived from simple extrusion; therefore, they are planar in the dip direction. In OQ they can be either used as “simple fault” or as “complex fault” models. The choice is relevant for the computational time and we developed a complexity index to help the user identify when to use one or the other on a fault-by-fault basis.

    The 3D geometry of the subduction systems is made for use only as a “complex fault” in OQ. Although there are only four subduction interfaces to treat, the computational challenge derives from the high number of FMDs coming out of the logic tree that explores the variability of several parameters. Notice also that the intraslab FMD is can only be derived from seismicity and is not presented here.

    Despite the obvious limitations, an added value brought in by this exercise is the possibility of comparing the contributions to the hazard of distant regions and conduct disaggregation and sensitivity analysis of a hazard model with uniformly determined parameters. This information would hardly appear by comparing regional/local studies carried out separately whose input parameters can differ from one another in too many possible ways. This information gain will likely help future studies in this field.