translated by Google Translate, revised

Introduction. In this post we comment – among other things – scientific models and their possible application for civil protection purposes. The discussion is necessarily simplified: a more detailed post is under consideration.

1. The deterministic earthquake prediction has always been invoked by humanity as a possible shelter from seismic disasters, in particular for what concerns the possibility of remaining victims. For science, on the other hand, it is a distant and perhaps unattainable goal, which requires theoretical knowledge and experimental observations on the dynamics of energy accumulation and release, which are not available today. The theme is broad and complex and cannot be treated at depth in these pages.

2. While people sometimes complain the lack of “earthquake alarms” from the scientists, sometimes it happens that alarms are released, even in Italy. In 2009 one of the famous “predictions” of G. Giuliani, in particular that concerning the possibility of a strong earthquake in the area of Sulmona, was about to put in crisis the local and regional administrations and contributed to trigger the spiral of panic that, together with the denials without foundation of the Abruzzo Region, determined the subsequent call of the meeting of experts in L’Aquila and the events that followed.
In 2012, the Italian Prime Minister in person, on the occasion of the 2012 Emilia-Romagna sequence, announced the possibility of a strong earthquake that could be generated by the so-called “third segment”, based on information provided by the CGR (Great Risks Commission).

Recently, the Head of the Department of Civil Protection (DPC), whose appointment has just been confirmed by the Government”, has supported in public, in relation to the Molise event of 16 August 2005 of Mw 5.1 and the seismic sequence in progress, that “experts cannot exclude the occurrence of an event of Mw even greater”, also in this case after a meeting of the CGR which was held on 17 August.

3. The scientific background of the Molise 2018 case is based – presumably – on a statistical model, implemented by some INGV researchers for ten years on the basis of other similar studies, which provides the probability of an earthquake occurring in a given area. Although the international scientific community is predominantly skeptical and however very cautious about the method (the discussion and the literature is consistent), DPC has welcomed the model, also as a result of the final document of the famous international commission (International Commission on Forecast of the Civil Protection Earthquakes – ICEF) http://www.protezionecivile.gov.it/resources/cms/documents/ICEF_Final_Report_IT_prefazione_e_riassunto_DLC.pdf)
established by the same DPC after the earthquake in L’Aquila on 6 May 2009.
The model, which DPC supports experimentation and development, is now called “Operational Earthquake Forecast”, or “operational” instrument to determine the probability of occurrence of a given magnitude event or, more recently, of a given effect (intensity) in a specific area in a given time-interval.

4. It is worth to remember that the head of the aforementioned commission – T. Jordan (US) – even claimed, in a scientific article, that it was the lack of availability of a reliable OEF who determined what he said was the failure of the experts meeting on March 31 2009 in L’Aquila, which then constituted the object of the famous L’Aquila trial of scientists and public servants. Moreover, during that trial, one of the authors of the method, questioned as a witness at the request of the PM, showed the results of the method in the weeks before the earthquake, calling it hazard estimates and thus fueling a certain confusion with the classic seismic hazard estimates which are carried out and used for the seismic construction regulations.

5. The method implemented in Italy is of a purely statistical nature, not based on physical and/or geological basic models. It provides increases in the probability of occurrence of an earthquake or a shaking that sometimes may exceed about 100 or even 1000 times the standard probability of an event as deducible from the seismic hazard map, but which remain within a few percentage units, or fractions, in the reference period, usually set at one week. Moreover, it must be emphasized that the reference to the estimates of seismic hazard is not correct, since the latter are of a different nature in that they are produced and used for another purpose. E. Boschi has discussed the method in
https://ilfoglietto.it/l-angolo-di-boschi/3378-vi-presento-il-mister-oef-che-prevede-i-terremoti
and
https://ilfoglietto.it/l-angolo-di-boschi/3224-i-terremoti-non-si-prevedono-ma-ci-sono-gli-algoritmi
6. Providing probabilities increments of some units of 100, OEF does not appear today, in Italy, a very sensitive instrument with respect to the “operational” aims that it promises. We are far from the weather-like probability values, so dear to the media audience, which, with all their limitations, can have a truly operational character as they provide probability of occurrence that sometimes exceeds 50% and are closer to the 80%, 90%.

7. It should  be added that the scientific community accepts more favorably the use of statistical methods for determining the probability of occurrence of important aftershocks of a strong earthquake in the context of a seismic sequence (OAF = operational aftershock forecast), rather than to determine the probability of an independent seismic event (OEF). For example, the results of the March 2018 SESAC meeting (Scientific Earthquake Studies Advisory Committee of the United States, in charge of giving an opinion on the USGS earthquake research program (United States Geological Survey, the largest US and world institution in the field), say that SESAC recommends the development of OAF, while assigning much lower priority to that of OEF.

8. Back to the sentence of the Head of DPC on the recent earthquake of Molise, his statement raises concerns for the following reasons:

a) the possibility of a strong earthquake in that area is a fact even in the absence of seismic sequence; to stress it only in relation to the sequence, as the statement suggests, is wrong. Moreover, such possibility also exists for most areas of Italy and it must be recalled also in the absence of earthquake sequences (see discussion in https://terremotiegrandirischi.com/2014/05/20/quando-comincia-lemergenza-seismica-m-stucchi/
b) the sentence did not provide: i) the value of the expected earthquake, or at least the lower magnitude threshold; ii) the value of the expected probability increase; iii) the duration of the alert;
c) it is not known what operational indications have been issued by DPC following this sentence. It is also known that citizens tend to stay outdoor after a moderate-to-major earthquake, either because they usually experience other shocks of which they are direct witnesses, independently of OEF, and because the building they live in may have weakened due to the shock itself.

9. In conclusion, while scientific research on these aspects should certainly be encouraged, it would be desirable first of all to consider alternative models to the one in use (see for instance a recent “Nature” article describing a method based on the use of neural networks). Moreover, and above all, greater caution with regard to the “operational” aspect might be desirable, in particular as regards the public management of results. It is clear, in fact, that while DPC can manage – on its own – information of this type, knowing of the limits of the instrument, the same cannot be said regarding the relevant communication to the public and the media, who have no awareness operative on what to do. In short, we are far from the “traffic light” model (green, yellow, red) so dear to the media and not only, developed for other types of slower phenomena (e.g. floods, volcanic eruptions). In this situation, finally, if after some alarms no disastrous event does not occur, the public could attribute little credibility to the method and the releases, with the imaginable consequences (“crying wolf” model).