Search

Earthquakes cannot be predicted with ultimate precision, so that the progressive reduction of the prediction uncertainty in space and time is an evergreen and challenging task, both from the scientific point of view for the intrinsic complexity of the seismic phenomenon and for its high societal relevance. To this aim, algorithms exist (like CN, M8, and M8S) based on objective recognition of seismicity patterns that have been already tested for some decades for intermediate-term middle-range prediction of strong earthquakes above a preassigned magnitude threshold. Here, we review the fundamental ideas of an integrated approach to earthquake prediction, based on the synergy of high-density geodetic observations (GNSS and SAR) and seismological information, and the results obtained so far through a completely new retrospective analysis of the 2016–17 seismic crisis in Central Italy and the 2012 Emilia sequence, where space-time precursory features are highlighted within intermediate-term ground velocities and seismicity. This integrated approach defines a new paradigm for time-dependent hazard assessment scenarios, and it is demonstrated that the proper integration of seismological and geodetic information can achieve what here is called intermediate-term narrow-range earthquake prediction. The extent of the alarmed areas, identified for the strong earthquakes by earthquake prediction algorithms based on seismicity patterns, can be significantly reduced from linear dimensions of a few hundred to a few tens of kilometers, leading to an improved more specific implementation of low-key preventive actions, like those recommended by UNESCO as early as in 1991.

Earthquakes cannot be predicted with precision, but algorithms exist for intermediate-term middle range prediction of main shocks above a pre-assigned threshold, based on seismicity patterns. Few years ago, a first attempt was made in the framework of project SISMA, funded by Italian Space Agency, to jointly use seismological tools, like CN algorithm and scenario earthquakes, and geodetic methods and techniques, like GPS and SAR monitoring, in order to effectively constrain priority areas where to concentrate prevention and seismic risk mitigation. We present a further development of integration of seismological and geodetic information, clearly showing the contribution of geodesy to the understanding and prediction of earthquakes. As a relevant application, the seismic crisis that started in Central Italy in August 2016 is considered in a retrospective analysis. Differently from the much more common approach, here GPS data are not used to estimate the standard 2D velocity and strain field in the area, but to reconstruct the velocity and strain pattern along transects, which are properly oriented according to the a priori information about the known tectonic setting. Overall, the analysis of the available geodetic data indicates that it is possible to highlight the velocity variation and the related strain accumulation in the area of Amatrice event, within the area alarmed by CN since November 1st, 2012. The considered counter examples, across CN alarmed and not-alarmed areas, do not show any comparable spatial acceleration localized trend. Therefore, we show that the combined analysis of the results of CN prediction algorithms, with those from the processing of adequately dense and permanent GNSS network data, may allow the routine highlight in advance of the strain accumulation. Thus it is possible to significantly reduce the size of the CN alarmed areas., Added 1 reference

Earthquakes cannot be predicted with ultimate precision, so that the progressive reduction of the prediction uncertainty in space and time is an evergreen and challenging task, both from the scientific point of view for the intrinsic complexity of seismic phenomenon and for its high societal relevance. To this aim, algorithms (like CN, M8 and M8S) based on objective recognition of seismicity patterns have been already tested for some decades for intermediate-term middle-range prediction of strong earthquakes above a pre-assigned magnitude threshold. Here, moving from some preliminary ideas, we propose an integrated approach to earthquake prediction, based on the synergy of high-density geodetic observations and seismological information, defining a new paradigm for time dependent hazard assessment scenarios. Through a wider and more refined retrospective analysis, duly involving the accuracy analysis of the newly available geodetic results, space–time precursory features are highlighted within ground velocities and seismicity, analyzing the 2016–2017 seismic crisis in Central Italy and the 2012 Emilia sequence. Overall, it is demonstrated that the proper integration of seismological and geodetic information can achieve what here is called intermediate-term narrow-range earthquake prediction. The extent of the alarmed areas, identified for the strong earthquakes by earthquake prediction algorithms based on seismicity patterns, can be significantly reduced from linear dimensions of a few hundred to a few tens of kilometers, leading to an improved more specific implementation of low-key preventive actions, like those recommended by UNESCO as early as in 1991.

This study illustrates the application of the CN algorithm for the analysis of precursory seismicity patterns in the Zagros region (Iran), an area characterized by a complex seismotectonic setting and by remarkable seismic activity. CN is a formally defined and widely tested algorithm for intermediate-term middle-range earthquake prediction, based on the analysis of routinely compiled earthquake catalogs. To allow its application, the global and regional catalogs available for the territory of Iran have been analyzed so as to compile a data set sufficiently complete and homogeneous over a time span of about 3 decades, as required for CN application. A number of tests have been performed with respect to changes in the input catalogs, assuming different magnitude completeness levels as well as considering different magnitude thresholds for the selection of target earthquakes. Different variants of the regionalization have been outlined according to the seismotectonic model, and it was concluded that precursory seismicity patterns for the largest events need to be researched in the whole Zagros tectonic domain. Accordingly, an experiment was set up aimed at validation of intermediate-term middle-range prediction of earthquakes with magnitude M >= 6.0 in the Zagros region. Starting in March 2012, CN prediction results have been routinely updated based on the events with M >= Mc = 4.0 as they are reported in the International Seismological Centre catalog.

The paper by Taroni et al. (2016) considers results of forward prediction of Italian strong earthquakes by CN algorithm with the declared intent of providing "a careful assessment of CN prediction performances... using standard testing procedures". Given the very limited number of target events within each region, however, the considered situation is non statistical, and a priori it is clear that the standard statistical methods are not effective here. The attempt to replace the standard approaches by Pari-mutuel Gambling Score (PGS) method leads to almost complete loss of information about predicted earthquakes, even for a large sample of target events. Therefore, the conclusions based on PGS, are untenable., Comment: 4 pages, 1 table. Submitted to Annals of Geophysics

The Italian territory has been the object of several studies devoted to the analysis of seismicity and to earthquake precursors' research. Although a number of observations have been claimed to precede large earthquakes, only few systematic studies have been carried out and almost no test of their performances is available up to now. In this paper, we review the application to the Italian territory of two formally defined intermediate-term middle-range earthquake prediction algorithms, namely CN and M8S. The general methodology common to the two different algorithms makes use of general concepts of pattern recognition that permit to deal with multiple sets of seismic precursors, and allows for a systematic monitoring of seismicity, as well as for a widespread testing of the prediction performances. Italy represents the only region of moderate seismic activity where the M8S and CN algorithms are applied simultaneously for routine monitoring. Significant efforts have been made to minimize the intrinsic space uncertainty of predictions and the subjectivity of the definition of the areas where precursors should be identified. Several experiments have been dedicated to assess the robustness of the methodology against the unavoidable uncertainties in the data. With these results acquired, starting in July 2003, an experiment was launched for the real-time test of M8S and CN predictions.

The influence of random magnitude errors on the results of intermediate-term earthquake predictions is analyzed in this study. The particular case of predictions performed using the algorithm CN in central Italy is considered. The magnitudes of all events reported in the original catalog (OC) are randomly perturbed within the range of the expected errors, thus generating a set of randomized catalogs. The results of predictions for the original and the randomized catalogs, performed following the standard CN rules, are then compared. The average prediction quality of the algorithm CN appear stable with respect to magnitude errors up to ±0.3 units. Such a stable prediction is assured if the threshold setting period corresponds to a time interval sufficiently long and representative of the seismic activity within the region, while if the threshold setting period is too short, the average quality of CN decreases linearly for increasing maximum error in magnitude.

SUMMARY Prediction methods based on seismic precursors, and hence assuming that catalogues contain the necessary information to predict earthquakes, are sometimes criticised for their sensitivity to the unavoidable catalogue errors and possible undeclared variations in the evaluation of reported magnitudes. We consider a real example and we discuss the eVect, on CN predictions, of a long-lasting underestimation of the reported magnitudes. Starting approximately in 1988, the CN functions in Central Italy evidence an anomalous behaviour, not associated with TIPs, that indicates an unusual absence of moderate events. To investigate this phenomenon, the magnitudes given in the catalogue used, which since 1980 is defined by the ING bulletins, are compared to the magnitudes reported by the global catalogue NEIC (National Earthquake Information Centre, USGS, USA) and by the regional LDG bulletins issued at the Laboratoire de Detection et de Geophysique, Bruyeres-le-Chatel, France. The comparison is performed between the ING bulletins and the NEIC catalogue, considering the local, M L , and duration, M d , magnitudes, first within the Central region, and then extended to the whole Italian territory. To check the consistency of the conclusions drawn from ING and NEIC data, the comparison of local magnitudes is extended to a third data set, the LDG bulletins. The diVerences between duration magnitudes M d that are reported by ING and NEIC since 1983 appear quite constant with time. Starting in 1987, an average underestimation of about 0.5 can be attributed to M L reported by ING for the Central region; this diVerence decreases to about 0.2 when the whole Italian territory is considered. The anomalous behaviour of the CN functions disappears if a magnitude correction of +0.5 is applied to M L reported in the ING bulletins. However, such a simple magnitude shift cannot restore the real features of the seismic flow, and ING bulletins are not suitable for CN algorithm application.

We perform a comparative analysis of regional and global earthquake catalogs currently available for the territory of Italy. We consider: (a) instrumental seismic catalogs provided by the Istituto Nazionale di Geofisica e Vulcanologia, Roma (INGV) for earthquake forecasting experiment in Italy within the Collaboratory for the Study of Earthquake Predictability (CSEP); (b) Global Hypocenters’ Data provided by the USGS/NEIC, which is currently used in the real-time earthquake prediction experiment by CN and M8S algorithms in Italy, and (c) seismological Bulletin provided by the International Seismological Centre (ISC). We discuss advantages and shortcomings of these catalogs in the context of intermediate-term middle-range earthquake prediction problem in Italy, including the possibility of the catalog’s combined or integrated use. Magnitude errors in the catalog can distort statistics of success-to-failure scoring and eventually falsify testing results. Therefore, the analysis of systematic and random errors in magnitude presented in Appendixes can be of significance in its own right.

An innovative approach to seismic hazard assessment is illustrated that, based on the available knowledge of the physical properties of the Earth structure and of seismic sources, on geodetic observations, as well as on the geophysical forward modeling, allows for a time-dependent definition of the seismic input. According to the proposed approach, a fully formalized system integrating Earth Observation data and new advanced methods in seismological and geophysical data analysis is currently under development in the framework of the Pilot Project SISMA, funded by the Italian Space Agency. The synergic use of geodetic Earth Observation data (EO) and Geophysical Forward Modeling deformation maps at the national scale complements the space- and time-dependent information provided by real-time monitoring of seismic flow (performed by means of the earthquake prediction algorithms CN and M8S) and permits the identification and routine updating of alerted areas. At the local spatial scale (tens of km) of the seismogenic nodes identified by pattern-recognition analysis, both GNSS (Global Navigation Satellite System) and SAR (Synthetic Aperture Radar) techniques, coupled with expressly developed models for interseismic phase, allow us to retrieve the deformation style and stress evolution within the seismogenic areas. The displacement fields obtained from EO data provide the input for the geophysical modeling, which eventually permits to indicate whether a specific fault is in a “critical state.” The scenarios of expected ground motion (shakemaps) associated with the alerted areas are then defined by means of full waveforms modeling, based on the possibility to compute synthetic seismograms by the modal summation technique (neo-deterministic hazard assessment). In this way, a set of deterministic scenarios of ground motion, which refer to the time interval when a strong event is likely to occur within the alerted area, can be defined both at national and at local scale. The considered integrated approach opens new routes in understanding the dynamics of fault zones as well as in modeling the expected ground motion. The SISMA system, in fact, provides tools for establishing warning criteria based on deterministic and rigorous forward geophysical models and hence allows for a well-controlled real-time prospective testing and validation of the proposed methodology over the Italian territory. The proposed approach complements the traditional probabilistic approach for seismic hazard estimates, since it supplies routinely updated information useful in assigning priorities for timely mitigation actions and hence it is particularly relevant to Civil Defense purposes.

The operational and decision-making problems related to earthquake forecast/prediction and seismic hazard assessment are nowadays a matter of significant debate, particularly on account of the very unsatisfactory global performance of Probabilistic Seismic Hazard Assessment at the occurence of most of the recent destructive earthquakes. While it is recognized that opera- tional tools must demonstrate their capability in anticipating large earthquakes and the related ground shaking by rigor- ous verification and validation process, only few methods proved effective so far. In view of the inherent uncertainties in predicting predictable, the usefulness of any forecast/ prediction method can then be judged taking into account the wide range of possible mitigation actions of different levels (from increased preparedness to evacuation). This paper aims to open a debate and complement the Summary and Recommendations by the International Commission on Earthquake Forecasting, established after the earthquake in L’Aquila (M = 6.3, 6 April 2009). The issues related with the definition, validation, and possible use of forecasting/ prediction methodologies, are addressed with special emphasis on existing operational practice in Italy.

An integrated neo-deterministic approach to seismic hazard assessment has been developed that combines different pattern recognition techniques, designed for the space–time iden- tification of impending strong earthquakes, with algorithms for the realistic modeling of seismic ground motion. The integrated approach allows for a time-dependent definition of the seismic input, through the routine updating of earthquake predictions. The scenarios of expected ground motion, associated with the alarmed areas, are defined by means of full waveform modeling. A set of neo-deterministic scenarios of ground motion is defined at regional and local scales, thus providing a prioritization tool for timely preparedness and mitigation actions. Constraints about the space and time of occurrence of the impending strong earthquakes are provided by three formally defined and globally tested algorithms, which have been developed according to a pattern recognition scheme. Two algorithms, namely CN and M8, are routinely used for intermediate-term middle-range earthquake predictions, while a third algorithm allows for the identification of the areas prone to large events. These independent procedures have been combined to better constrain the alarmed area. The pattern recognition of earthquake-prone areas does not belong to the family of earthquake prediction algorithms since it does not provide any information about the time of occurrence of the expected earthquakes. Never- theless, it can be considered as the termless zero-approximation, which restrains the alerted areas (e.g. defined by CN or M8) to the more precise potential location of large events. Italy is the only region of moderate seismic activity where the two different pre- diction algorithms, CN and M8S (i.e. a spatially stabilized variant of M8), are applied simultaneously and a real-time test of predic- tions, for earthquakes with magnitude larger than a given threshold (namely 5.4 and 5.6 for CN algorithm, and 5.5 for M8S algorithm) has been ongoing since 2003. The application of the CN to the Adriatic region, which is relevant for seismic hazard assessment in the northeastern part of the Italian territory, is also discussed. Examples of neo-deterministic scenarios are provided, at regional and local scale and for the cities of Trieste and Nimis (Friuli Venezia Giulia region), where the knowledge of the local geological conditions permitted a detailed evaluation of the expected ground motion.

—The choice of the regions is essential in the application of the algorithm CN, therefore a seismotectonic criterion for their definition is tested. In order to take into account the geodynamic complexity characterising the Italian peninsula, we established to strictly follow the seismotectonic zones, including in each region only zones with similar seismogenic behaviour and the transitional zones connected to them. Three regions have been successfully defined in this way, corresponding approximately to the North, Centre and South of Italy. The reduction of the space-time uncertainty and the increase of the stability of prediction results obtained with this regionalisation, with respect to the previous applications of CN in Italy (Keilis-Borok et al., 1990; Costa et al., 1995, 1996), can be interpreted as a validation of the seismotectonic model.

A regionalization of the Italian territory, strictly based on seismotectonic zoning and the main geodynamic features of the Italian area, is proposed for intermediate-term earthquake prediction with CN algorithm. Three regions, composed of adjacent zones with the same seismogenic behaviour or with transitional properties, are selected for the north, centre and south of Italy, compatibly with the kinematic model. This regionalization allows us an average reduction of the spatial uncertainty of about 35% for the northern and central regions, and of about 70% for the southern region in comparison with previous studies. A general reduction of the percentage of total TIPs, with respect to the results obtained neglecting the seismotectonic zoning, has been observed as well. Therefore, it seems that the seismotectonic model is a useful tool selection of the fault systems involved in the preparation of strong earthquakes. The successful attempt of catalogue upgrading, accomplished using the NEIC Preliminary Determinations of Epicentres, appears to substantiate the robustness of the algorithm against changes in the catalogue.