PROSPECTIVE AFTERSHOCK FORECASTING OF THE NORCIA 2016 EARTHQUAKE SEQUENCE, CENTRAL APENNINES, ITALY

Lead Research Organisation: NERC British Geological Survey
Department Name: Earth Hazards & Observatories

Abstract

On 24/08/2016, 01:36:33 UTC an earthquake of magnitude M=6.2 occurred at Norcia, Italy, killing more than 290 people, injuring 500 more, and leaving some 2,500 local inhabitants without a home. The earthquake resulted from movement on a normal fault in the Apennines mountain chain that runs along the Italian peninsula. Large destructive earthquakes of similar rupture style have occurred throughout this region in the past, most recently the 2009 M=6.4 L'Aquila event, 43 km S of the recent epicenter.

Immediately after the Norcia earthquake, a UK scientific team led by the British Geological Survey, together with the University of Edinburgh, coordinated with the Instituto Nationale Geophysica e Vulcanologia (INGV) to enable the deployment of a high -density temporary seismic network to study the aftershock sequence. To date, 26 UK seismic recording stations are fully integrated with the INGV network and this will enable a high-accuracy updated earthquake catalogue to be derived with a greater regional coverage and improved magnitude sensitivity.

The dataset collected within the next 6 months will be the basis of the development of aftershock forecast models and their transparent testing following international protocols. Aftershock forecasts are based on our understanding of earthquake triggering mechanisms and the empirical knowledge from previous aftershock sequence in the broader region. A validation of our forecasts using widely accepted statistical metrics is necessary in order to determine the strengths and weaknesses behind our triggering hypothesis and ensure that new knowledge will be passed on to improve operational aftershock forecasting world-wide.

Planned Impact

Operational earthquake forecasting benefits wider society in affected areas primarily by supporting decision makers on the ground at times when felt aftershocks continue to rattle the population. It can also aid in increase awareness of the risks on different timescales, and hence can help to promote the development of resilience to future events.

Our research will engage a range of professionals working at the interface between research and impact in the British Isles, including social scientists, educationalists and humanities researchers and the humanitarian organisations they are already engaged with. The network already has an established track record of applying contemporary earthquake science to emergency earthquake response. It has, at its heart, the aim of doing world-leading research and using it to increase the resilience of earthquake-threatened communities. The overall practical aim in our proposal is to improve current practice in operational forecasting, in consultation with a variety of end users, so the results can be transferred quickly to operational utility, and used ultimately to support decision-making in a crisis. Such Impact is at the core of our scientific motivation. We will continue to engage with NGO Concern Worldwide, since their input and collaboration over the last years has helped constrain our research focus and continues to shape our thinking, and seek to develop and extend this best practice in collaboration with NGOs to other organisations. We will also seek to engage with decision makers with delegated authority from Government, on issues of policy development and practical implementation of Operational Forecasting at a special Impact-dedicated event involving the Italian Department of Civil Protection (DPC) and the National Institute of Geophysics and Volcanology (INGV) in Italy. This will explicitly include discussion on how the scientific discoveries and outputs could be further used to increase resilience in a post-disaster environment.

Publications

10 25 50
 
Description The objective of the project was to develop aftershock forecast model using preliminary information available following the large earthquake in the Apennines. The evolving sequence was characterised by the occurrence of 9 earthquakes above magnitude M=5.0. Three principle earthquakes of magnitude M=5.9 occurred within a 50 km distance along the Apennines mountain chain in a period of just over 2 months. The first occurred on 24th August with a magnitude of M=6.0 near Amatrice and killed 297 people, destroying the medieval villages of Amatrice, Accumoli and Pescara di Tronto. It was followed almost one hour after by a M=5.7 triggered event near Norcia. On 26th October two large shocks of M=5.4 and M=5.9 (32 minutes after) struck ~30 km further north. Four days later on 30th October the largest (M=6.5) event to date struck, devastating Norcia and its historical cathedral of San Benedetto. Few months later in January 18, 2017 4 events above magnitude M=5.0 occurred at the south prolongation of the fault near L'Aquila that suffered a M=6.4 earthquake in 2009 with devastating consequences. All the above make clear that the occurrence of large magnitude events within a limited time period under a well instrumented area gives scientists the opportunity to develop and test aftershock forecast models that can inform us about the evolution of the sequence. These models do not only have a great scientific value enhancing our understanding about under which conditions hazard occurs but are expected to play in the future an important role in developing emergency strategies in disaster settings. We find that stress transfer hypothesis, supporting the existence of triggering seismicity due to rapid elastic stress loading of near-by faults following a large earthquake, reasons the occurrence of the largest earthquake in the sequence. However, at small scales (few km) the triggering mechanisms are unknown and they are subject to large uncertainties related to the source models, used to described co-seismic fault slip. In the future, we seek to incorporate influence from large and small magnitude earthquakes to enhance the performance of aftershock forecast models in different scales.
Exploitation Route The preliminary forecast models, products of the first year of research that describe our knowledge for aftershock occurrence immediately after the large earthquakes of the Apennines, will be the benchmark forecast models against which the best models, products of the next 3 years of research, will be compared against within the now starting NERC-NSF project (UK PI Margarita Segou).
Sectors Other

 
Description (1) We have been communicating expert advice upon request to the British embassy that received the BGS formal advisory from the Cabinet Office. The initiative of BGS to install temporary network in the epicentral area has received the formal acknowledgments of the UK and Italian government. In an official Cabinet Office letter it was acknowledged that the project and the team efforts contribute largely towards building a positive reputation of the UK at a international level. The PI Dr. Margarita Segou has attended a local event to support community resilience where the opportunity was given to discuss the on-going scientific progress and the collaboration with member of the Royal Family, (2) UK team members and Italian colleagues have actively participated in the production and filming of a documentary featuring the UK efforts that was broadcasted last October. (3) The results in the form of daily updated probability maps for selected time periods within the 2016 Norcia earthquake sequence were discussed in the final meeting of the project that has taken place in the Instituto Geofisica e Volcanologia in Rome (Italy). Team members that participated in the preparation of the national seismic hazard maps have direct communication with the Civil Protection to ensure that most recent scientific results are broadly disseminated. This research direction continues under the NERC-NSFGEO funded research The Central Apennines under a New Microscope, please see more information about current impact of this research to the aforementioned award.
First Year Of Impact 2016
Sector Energy,Environment
Impact Types Societal,Policy & public services

 
Title Rate-and-State Friction Law combined with Static Stress Transfer hypothesis 
Description We elaborated a timeline to pseudo-prospectively test some preliminary forecasts applied to the afore-mentioned Italian sequence. In particular, we were interested in testing the effect of data quality on models performance: in fact, when only real-time catalogues and preliminary earthquake sources descriptions are available the forecasting efforts can be hampered. To this end, we developed two scenarios: a preliminary set of physics-based forecast models based on data products available within minutes to hours after a main event, and a more informed one based on data products available within few weeks from the mainshock occurrence. For the preliminary models, we estimated the stress perturbation at depths between 2 km and 12 km after each of the nine earthquakes of the Italian sequence with M = 5.0, using preliminary data on location, magnitude and fault geometry made available by the INGV (Istituto Nazionale di Geofisica e Vulcanologia) from minutes to few hours after each event. Moreover, at this stage we approximated a simple uniform slip distribution on the selected source faults and fixed the geometry of the receiver faults using the same orientation of the respective source, which is a valid assumption when we do not know much about the complex structures of the neighbouring faults (Segou et al., 2013). In order to improve the forecast performance, we calculated the stress changes for the same events using definitive values of moment magnitude and depth, as well as fault dimensions published on peer reviewed journal articles. Moreover, we requested from our INGV project partners to provide the best-available slip models. For both forecast scenarios, we then translated the computed stress changes into expected seismicity rates applying the Rate-State friction law (Dieterich, 1994). We first computed the reference seismicity for Central Apennines using a catalogue (http://iside.rm.ingv.it/iside) spanning from 1990 to 2016 with magnitude of completeness MC = 2.5. Especially, for the preliminary scenario, we set the Rate-and-State parameters from the available peer-review literature for the study area (e.g. Catalli et al., 2008). On the contrary, Rate-and-State parameters for the second class were progressively and adaptively fitted after each forecast window (one day) by comparing the daily forecast results against the observed unfolding seismicity, in order to retrieve a more realistic set of values. 
Type Of Material Improvements to research infrastructure 
Year Produced 2018 
Provided To Others? No  
Impact A manuscript is under preparation to be submitted in a peer-review journal within the next 6 months. 
 
Title One year of continuous waveforms of 24 broadband stations located in the Central Apennines 
Description Our emerging dataset consists of seismic records from 89 stations deployed soon after the first event2; 28 permanent and 23 temporary stations deployed by Istituto Nationale Geofisica e Vulcanologia (INGV), 24 temporary stations by the British Geological Survey (BGS), and 19 accelerometer stations operated by the Italian Civil Protection. The network has an average station spacing ~5 km and it will continue to be fully operational until at least September 2017. 
Type Of Material Database/Collection of data 
Year Produced 2017 
Provided To Others? Yes  
Impact An International project is now just starting (Jan, 2018-Jan. 2021) focusing on the laborious processing of the continuous waveform and harvesting source parameters for even the smallest magnitude earthquakes within the earthquake sequence. 
 
Title Physics-based forecast models 
Description Two different approaches have been developed to face the problem of forecasting the spatiotemporal evolution of an earthquake sequence. The first involves the use of Rate-and-State laboratory derived friction models based on the calculation of the static stress transfer in the crust following a mainshock (Toda et al., 2005; Cocco et al., 2010; Toda & Enescu, 2011; Parsons et al., 2012, 2014; Segou et al., 2016; Segou & Parsons, 2016, among others). The second entails a statistical modelling of the cascade of triggered events based on empirical observations (e.g. Werner et al., 2011; Lombardi & Marzocchi, 2010; Marzocchi, 2012). Given their ability to capture different characteristics of the evolution of a sequence, it is not rare to use both of these two methods and then compare the results, or to use a hybrid approach. Immediately following the 24th August 2016 Amatrice earthquake (Central Apennines, Italy) we started producing and testing physics-based forecast models, that is, those belonging to the first of the two aforementioned categories. As earthquakes redistribute the stresses acting on the surrounding crust, according to the static stress transfer hypothesis seismicity is generally promoted in crustal volumes where the shear stress increases, while it is inhibited where the stress is reduced. The simulation of a mainshock rupture (aka source fault) and the subsequent effect of stress perturbation onto neighbouring active faults (aka receiver faults) represent the key feature of physics-based models. To do that, the so-called "static Coulomb stress change" is calculated with the following equation: Where is the shear stress change on a given receiver fault plane (positive in direction of fault slip), is the change in normal stress, and is the effective coefficient of friction on a candidate receiver fault plane on which aftershocks could nucleate. Estimated stresses are used to derive the time-dependent evolution of seismicity by following the rate-and-state friction laws (Dieterich, 1994). According to this theory, in the absence of a relevant stress change in the crustal volume of interest, the seismicity rate is simply equal to its characteristic background seismicity rate of the region. However, if a the crustal static stress field is perturbed, the new expected aftershocks rate depends on a set of parameters (the so-called Rate-and-State parameters): the stress change itself, the contribution of the intrinsic background rate and two parameters describing the fault behavior, namely the effective normal stress clamping (or unclamping) the receiver faults and the secular shear stressing rate acting on them. 
Type Of Material Computer model/algorithm 
Year Produced 2018 
Provided To Others? Yes  
Impact We have presented the model in the final meeting of the project, summarising the first year results in INGV, Rome. Experts attending suggested to incorporate the physics-based approach to operational forecasting model for Italy nationwide. 
 
Description Memorandum of Understanding between the British Geological Survey (BGS), Istituto Nationale Geofisica e Volcanologia (INGV-Rome) and the University of Edinburgh (UoE) 
Organisation National Institute for Geophysics and Volcanology (INGV)
Country Italy 
Sector Public 
PI Contribution This collaboration is grounded on the joint BGS-INGV deployment of a temporary seismic network in the epicentral area of the 2016-17 Central Apennines earthquake sequence. The UK seismic instruments were an urgent loan from the Geophysical Equipment Facility (FEG) provided through SEIS-UK following an application from BGS PI (Application Number 1067) including BGS and UoE scientists. The deployment is supported by direct funds provided by the National Environmental Research Council (NERC).
Collaborator Contribution INGV is the Italian organization responsible for providing nearly real time scientific information to the Department of Civil Protection in times of earthquakes disasters occurring on the Italian territory. To this end, running 24h per day seismic monitoring services based on real-time data collected by more than 300 permanent seismic station deployed on the Italian territory, providing earthquake detection and rapid evaluation both in terms of location and magnitude for the earthquakes occurring on the country is a statutory responsibility of INGV.
Impact The aim of this collaboration based on the emergency scientific and technical response is to improve our understanding of aftershock sequences over the next 3 years under a new project funded by NERC-NSF that builds upon the current project team and expands it involving overseas partners. The unparalleled data collection is expected to shed light on how earthquakes nucleate and trigger earthquake cascades and enhance the future earthquake forecast models. The invaluable dataset (See Also Databases section in this report) and the work plan put forward within this international collaboration aims in improving our understanding of the seismogenic structures focusing on unravelling the anatomy of active faults, the interaction pattern and the complexity of normal fault systems in the Apennines. The above is the first crucial element towards a robust seismic hazard assessment and contributes towards the development of testable statistical and physics-based forecast models for aftershock occurrence. We will investigate under which conditions seismic activity migrates to neighbouring faults as well as the geometry of the seismogenic structure to achieve a better description of the underlying physical processes within the earthquake sequence. The collaboration aims strongly supported the institutional responsibilities of the participating institutes (BGS, INGV) and promotes risk reduction strategies at an international level.
Start Year 2016
 
Description Memorandum of Understanding between the British Geological Survey (BGS), Istituto Nationale Geofisica e Volcanologia (INGV-Rome) and the University of Edinburgh (UoE) 
Organisation University of Edinburgh
Department Earth Science
Country United Kingdom 
Sector Academic/University 
PI Contribution This collaboration is grounded on the joint BGS-INGV deployment of a temporary seismic network in the epicentral area of the 2016-17 Central Apennines earthquake sequence. The UK seismic instruments were an urgent loan from the Geophysical Equipment Facility (FEG) provided through SEIS-UK following an application from BGS PI (Application Number 1067) including BGS and UoE scientists. The deployment is supported by direct funds provided by the National Environmental Research Council (NERC).
Collaborator Contribution INGV is the Italian organization responsible for providing nearly real time scientific information to the Department of Civil Protection in times of earthquakes disasters occurring on the Italian territory. To this end, running 24h per day seismic monitoring services based on real-time data collected by more than 300 permanent seismic station deployed on the Italian territory, providing earthquake detection and rapid evaluation both in terms of location and magnitude for the earthquakes occurring on the country is a statutory responsibility of INGV.
Impact The aim of this collaboration based on the emergency scientific and technical response is to improve our understanding of aftershock sequences over the next 3 years under a new project funded by NERC-NSF that builds upon the current project team and expands it involving overseas partners. The unparalleled data collection is expected to shed light on how earthquakes nucleate and trigger earthquake cascades and enhance the future earthquake forecast models. The invaluable dataset (See Also Databases section in this report) and the work plan put forward within this international collaboration aims in improving our understanding of the seismogenic structures focusing on unravelling the anatomy of active faults, the interaction pattern and the complexity of normal fault systems in the Apennines. The above is the first crucial element towards a robust seismic hazard assessment and contributes towards the development of testable statistical and physics-based forecast models for aftershock occurrence. We will investigate under which conditions seismic activity migrates to neighbouring faults as well as the geometry of the seismogenic structure to achieve a better description of the underlying physical processes within the earthquake sequence. The collaboration aims strongly supported the institutional responsibilities of the participating institutes (BGS, INGV) and promotes risk reduction strategies at an international level.
Start Year 2016
 
Description Final Meeting and Kick-Off Meeting of the 2016-2017 Central Apennines earthquake sequence 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact The final meeting of the current project, featuring the results from the first year of research, was planned to coincide with the kick-off meeting of the NERC-NSF funded research project with title "The Central Apennines earthquake sequence under a new microscope". The event was well attended by UK scientists from the British Geological Survey, the University of Edinburgh, University of Bristol, US scientists from University of Stanford, Lamont-Doherty Earth Observatory in collaboration with US Geological Survey and scientists from Italian Universities such as the University of Sapienza (Rome) and a plethora of local scientists, our project partners, from Institute Geofisica e Volcanologia. The event was hosted in Rome at the grounds of the geophysical institute to enhance international collaboration. The first year of research results have been presented jointly and an activity plan was formed for our future research within the 3-year NERC-NSF Standard Grant. Under/postgrad students had the opportunity to discuss their research with top scientists enhancing in that way their professional networks.
Year(s) Of Engagement Activity 2018