On the night of 15–16 May 1951, two moderate earthquakes with estimated magnitudes of Mw 5.4 and 4.5 occurred in northern Italy, about 40 km southeast of Milan, close to the small town of Caviaga. They were recorded by several observatories worldwide, as reported by the International Seismological Summary (ISS) On-Line Bulletin (ISS, 1951; International Seismological Centre [ISC], 2011). Despite the moderate magnitudes, these two events caught the attention of seismologists and have been studied in detail, in particular by Caloi et al. (1956), because they were close to Caviaga in an area that was assumed to be aseismic. Moreover, their shallow hypocenters (5 km in Caloi et al., 1956) indicated a possible anthropogenic source, related to wells for gas withdrawal (Fig. 1; see Data and Resources).

In the absence of any further discussion or revision of the original study by Caloi et al. (1956), the Caviaga earthquakes have been included in several compilations of induced seismicity, and they have been generally accepted as cases of anthropogenic events (Grasso, 1992; Maury et al., 1992; Guha, 2000; Suckale, 2009; Klose, 2013; MINE Junior Research Group, 2014).

In particular, the two events are currently included in the catalog of anthropogenic events that was compiled by Klose (2013), because they formally match all of the required criteria to be classified as induced or triggered events. These criteria are (1) a description of the candidate earthquake in a scientific peer-reviewed article, in conference proceedings, or as an abstract; (2) characterization in terms of time of nucleation, dominant focal mechanism, and geographic location, with uncertainty of depth and maximum observed seismic magnitude; (3) characterization of the human activity in terms of start-end time of operations, geographic location, depth in the earth crust, and mass changes in proximity to wells. The first criterion was satisfied by the report by Maury et al. (1992), in which the Caviaga events were mentioned: “In Italy, the production of gas from the Caviaga field caused an earthquake of magnitude 5.5 in 1951. Caloi et al. (1956) assumed, and Caloi was able to confirm in 1970 (Caloi, 1970), that the gas production was the main cause of the earthquake.” For the two additional criteria, Caloi et al. (1956) studied the first motions of the mainshock at 21 seismological stations, indicating that the earthquake was “...a violent outward thrust, in a solid angle with an axis strongly inclined towards NW.” Supported by the shallow hypocentral location that they reported as 5 km in depth, as well as proximity of the events to wells, Caloi et al. (1956) suggested a correlation between the two seismic events and the gas extraction activities. Should this speculation be true, the first of these two events (estimated origin time 15 May 1951, 22:54 Greenwich mean time [GMT]) would be the strongest induced event that has ever occurred in Italy, the strongest in Europe related to extraction fields, and one of the major induced events anywhere in Europe (MINE project, see Data and Resources).

After 60 years, it is possible to revisit this interpretation using improved computational techniques, the available highresolution data, enriched historical catalogs and a deeper understanding of the regional seismotectonic and crustal structure. The focus of this study is the relocation of these two events with the use of modern hypocentral location methods and the analysis of the historical seismicity of the area. A complete seismic source parameterization is out of the scope of this preliminary study. In the following, we describe the regional geological setting and the gas reservoir characterization, introduce the context of historical seismicity, provide a description of the mainshock relocation, discuss the uncertainties of the hypocentral parameters, and estimate the variation of the stress field due to extraction activities. We consider this revision necessary to be able to discuss the possibility that these two events were not induced by human activity, as well as to improve the quality of the dataset for decision makers involved in risk evaluation.