Istituto Nazionale di Geofisica e Vulcanologia (INGV), Sezione Milano, Milano, Italia

Nell’occasione del Ventennale dell’Associazione Donne e Scienza, il presente contributo riassume un percorso di ricerca scaturito dalla partecipazione attiva all’Associazione per circa un decennio e che si snoda via via, corroborato dalle strategie di eguaglianza di genere a livello nazionale e internazionale. Viene affrontato il tema delle dimensioni di genere nei rischi naturali, come emerso nell’ambito progetto INGV DiGeST (2022-2024). I fenomeni naturali sono neutri, ma gli impatti non lo sono. Le diseguaglianze nella società, e in particolare quelle di genere, sono determinanti nel delineare i rischi e gli impatti degli eventi naturali. È necessario includere il genere nei contenuti di ricerca, educazione, rapporti scienza-società, ed è necessario farlo per realizzare una ricerca di qualità. Comprendere come le relazioni di genere modellino la vita delle donne e degli uomini è fondamentale per la riduzione del rischio.

L’articolo esplora l’impatto di ChatGPT, un modello di linguaggio di grandi dimensioni, evidenziando il dibattito tra le sue capacità innovative e le implicazioni etiche, in particolare analizzando dei bias di genere che affliggono le risposte di ChatGPT. Si evidenzia come ChatGPT possa perpetuare stereotipi di genere, ma anche come con l’intervento degli utenti si possano mitigare i bias. Replicando nel Giugno 2023 l’esperimento di (Gross, 2023), che indaga le risposte di ChatGPT su domande con un evidente dimensione di genere, si rilevano distorsioni e stereotipi. Si analizzano esempi di risposte discriminatorie e stereotipate su ruoli professionali, storie di successo, curriculum vitae, e abilità matematiche. Infine, si nota come il comportamento dei “Large Language Models” manchi di trasparenza nei criteri che producono le risposte. Per gestire questi problemi si suggerisce l’intervento di organismi regolatori sovranazionali, promuovendo la trasparenza e la qualità dei dati. Si conclude con l’importanza del coinvolgimento degli utenti nel miglioramento dei sistemi AI per una società più equa e priva di pregiudizi.

In this study, we evaluate the performance of five recent Intensity Prediction Equations (IPEs) valid for Italy comparing their predictions with intensities documented at Italian localities. We build four different testing datasets using the data contained in the most recent versions of the Italian Parametric Earthquake Catalogue CPTI15 and Macroseismic Database DBMI15 and we estimate the residuals between observed and predicted intensity values for all the selected IPEs. The results are then analyzed using a measureoriented approach to score each model according to the goodness of model prediction and a diagnostic-oriented approach to investigate the trend of the residuals as a function of the different variables. The results indicate the capability of all the tested IPEs to reproduce the average decay of macroseismic intensity in Italy although with a general underestimation of high-intensity values. In addition, an in-depth investigation of the spatial and temporal patterns of the event residual term, computed using the best predictive model, is carried out. Lastly, we provide some hints for the selection of calibration datasets for the development of future intensity attenuation models.

Understanding the relationship between different ground motion parameters (GMPs) and macroseismic intensity (I) is essential for applications in seismology and earthquake engineering. This study presents empirical relationships between Peak Ground Acceleration (PGA) and macroseismic intensity (I) for seismic events in Colombia, using historical earthquake data processed by the Colombian Geological Survey. Ten historical earthquakes across the country, occurring between 1994 and 2016, with magnitudes ranging from Mw 5.3 to 7.2, were analyzed. From the cross-analysis between municipalities with recorded intensity (I) and nearby PGA measurements, 62 I-PGA pairs with an average distance of 3.6 kilometers were obtained. These pairs were manually validated based on distance and topographic similarity criteria and used to establish the proposed empirical relationships. The final relationships employed the geometric mean of the horizontal components of the PGA and the European Macroseismic Scale EMS-98. Additionally, analyses were conducted using the Modified Mercalli Scale and varying the acceleration records, including the vertical component and the highest of all components. During the analysis, linear, exponential, polynomial, and power functions were calibrated with intensities ranging from II to VIII on the EMS-98 scale and PGA values ranging from 1.16 to 527.36 cm/s². The relationships that best represented the seismicity in Colombia were selected through the calculation of standard deviations and determination coefficients. Finally, residuals of observed/predicted intensity and acceleration were analyzed as a function of variables such as magnitude and epicentral distance.

G (2024) Grid-search method for short-term over long-term average parameter tuning: an application to Stromboli explosion quakes.

Pantelleria is an active volcanic island with a fumarolic and hydrothermal activity so much so that it represents an interesting geothermal target as it presents high temperature gradients. This peculiarity has made it the protagonist of a project (Panta-Rei) which aims to characterise the geothermal reservoir of the island and to assets of its geothermal potential. Following the activity of the project we installed a temporary seismic network to enhance the monitoring of seismic activity and characterising background seismicity. Here we describe the design and installation of the temporary network code ‘YM’, giving also a preliminary analysis on the performance. Although it is an active volcano, the island doesn’t show, apparently, typical volcanic-tectonic earthquakes. However, this is due to the lack of a local seismic network to detect earthquakes from 2022 onwards. Thanks to the 4 short period stations of the YM temporary network, on the island, currently, are installed 8 seismic stations. The four temporary sites, mainly concentrated in the central-southern portion of the island, were chosen after a seismic survey carried out in September 2023, performed to test the site response and to evaluate both the background noise level and the logistics of the installations. We installed the temporary stations in march 2024, also minimising visual impact of the solar panels and the instrumentations. All the stations are working regularly and the data are transmitted via LTE in real time stored on a server in Naples. Our analysis, performed on the first month of seismic data acquired, indicates that the stations of the temporary network have low seismic noise levels compared to the reference models, indicating a high quality of the recorded waveforms and a good potential in detecting seismic events with low energy as well.

n seismology, the swift deployment of seismic stations following an earthquake event is crucial for optimizing network coverage, facilitating accurate aftershock localization and characterizing the ongoing seismic sequence. SISMIKO [Moretti et al., 2023; DOI 10.3389/feart.2023.1146579] at Istituto Nazionale di Geofisica e Vulcanologia (INGV) is an operational group specifically designed for the rapid installation of seismic stations in post-earthquake scenarios. Comprising trained personnel equipped with state-of-the-art technology, this group operates with the primary objective of augmenting existing seismic networks in epicentral area to capture aftershock activity comprehensively. Since 2008, thanks to the experience acquired during a series of destructive seismic sequences (i.e. L’Aquila in 2009, Po Plain in 2012 and Central Italy in 2016), SISMIKO worked for codifying a set of procedures that are now in place and that allow to provide a rapid response after an earthquake, managing preconfigured metadata parameters, the real-time centralization of all temporary seismic stations and a dedicated data acquisition system. This guarantees fast data availability for research purposes without any restrictions, and, if necessary, supports the INGV monitoring room 24/7 seismic surveillance system. For example, after the Mw 5.5, 9 November 2022 Costa Marchigiana offshore earthquake, data from seven rapidly deployed seismic stations were online within twelve hours from the event occurrence, aiding seismic emergency response. The scientific community can get high-quality data acquired by the SISMIKO group through the Italian node of the European Integrated Data Archive (EIDA [Danecek et al., 2021; DOI: 10.1785/0220200409.]). To better support the Open Science paradigm and the FAIR principles, data streams related to each seismic sequence are identified through Digital Object Identifier (DOI) assigned by the Data Registry [INGV DMO, 2020; DOI: 10.13127/data- registry], and they integrate those available from the permanent Italian National Seismic Network. Furthermore, these procedures also facilitate the recovery and standardization of continuous digital waveform datasets from the past 20 years, supporting global seismic activities and hazard studies. This rapid dissemination of data can support activities of the global seismological community, enhancing timely comprehension of earthquake sequences and facilitating seismic hazard studies.

The macroseismic intensity specifically relates to damage in a way that parameters like Peak Ground Acceleration (PGA) do not, thus providing a practical measure to guide the priorities of Civil Protection interventions immediately following an event. In this study, we analyze the new macroseismic intensity attenuation models proposed by Gomez-Capera et al. (2024), which are calibrated in moment magnitude (Mw) for the Italian territory, focusing on their potential application in developing Probabilistic Seismic Hazard Assessment (PSHA) maps in terms of macroseismic intensity I. Specifically, we elected to test the Log- Lin_10 model for macroseismic intensity due to its simpler mathematical and computational properties. A primary goal of these attenuation models is to allow for the assessment of seismic hazard maps directly in terms of macroseismic intensity, thereby avoiding the uncertainties associated with converting various ground shaking parameters into intensity values. In the present study, we compare the PSHA map obtained directly using the Gomez-Capera et al. (2024) model with that derived from the Gomez-Capera et al. (2022) conversion of the PGA parameter into intensity I, and with the Italian National Seismic Hazard Map (MPS04) in terms of macroseismic intensity (Gruppo di Lavoro MPS, 2004).

Earthquakes are, by far, the most relevant source of hazard for the densely urbanised areas of Mt. Etna volcano. Local communities continuously suffer social and economic losses due to the high occurrence of damaging earthquakes, which produce intensities up to degree X EMS despite of low magnitude (M < 5.0). In the framework of the PANACEA project, seismic hazard was performed following the probabilistic approach (PSHA) based on historical macroseismic data. The probability distribution of the expected earthquake damage to residential buildings and the resulting risk effects on population are calculated for exposure times of 30 and 10 years.