This study evaluates the seismicity detection efficiency of a new low‐cost triaxial accelerometer prototype based on microelectromechanical systems (MEMS) technology. Networks of MEMS sensors were installed in telecommunication infrastructures to build two small arrays in northern and central Italy. The sensor prototypes recorded major earthquakes as well as nine small seismic events with 2.0<ML<3.0. Where possible, MEMS were compared to the closest high‐quality seismic stations belonging to the national accelerometric network. The comparison, in terms of peak ground accelerations and spectral responses, confirms that the signals are in good agreement. The tested inexpensive MEMS sensors were able to detect small local events with epicentral distances as large as 50 km and provided an efficient characterization of the main motion parameters. This confirms that the proposed accelerometer prototypes are promising tools to integrate into traditional networks for local seismicity monitoring.

Microelectromechanical systems (MEMS) sensors are miniaturized integrated circuit batch processing devices, with size ranging from a few micrometers to millimetres (Homeijer et al., 2011). They are widely adopted in many industrial applications such as telecommunication, automotive, game controllers, and so on (Shi et al., 2009). As a consequence of such widespread use, MEMS sensors are the most economic motion detector (Fleming et al., 2009; Cochran et al., 2012; D’Alessandro and D’Anna, 2013; Evans et al., 2014; Lawrence et al., 2014; Nof et al., 2019). Moreover, the performance of MEMS sensors is rapidly growing, being already comparable to high‐quality accelerometers (Kong et al., 2016). Recent advances in micromachined sensors provide adequate sensitivity for measurement in geophysical applications. For example, several MEMS devices are suitable for volcanic activity studies (Andò et al, 2011), gravimetric and geodetic observations (Cenni et al., 2019; Mustafazade et al., 2020), and seismological and earthquake engineering projects (Holland, 2003; Cochran et al., 2012). These sensors can reach an efficient performance for moderate (5.0>Mw>5.9) to large (Mw>6.0) earthquake detection at distances on the order of tens of kilometres (Boaga et al., 2019; Liu et al., 2011). Because of the low cost of the MEMS motion sensors (∼ two orders of magnitude less than a classical high‐quality seismological stations), they could be adopted for the installation of dense accelerometric networks. Ground‐motion wavefield recording is limited to relatively coarse spatial sampling due to a limited number of installed sensors and large instrument spacing (on the order of tens of kilometres). This can lead to spatial aliasing of the wavefield and consequently an underestimation of the effective peak ground acceleration (PGA) as well as uncertainties in the estimation of local effects on the resulting shake maps (Wald et al., 2008). These are the reasons why several dense low‐cost MEMS networks have been recently tested for the monitoring of strong earthquakes ground shaking (Cochran et al., 2009; Lawrence et al., 2014). The question remains whether such technology is suitable for low‐magnitude seismicity detecting. In particular, an interesting use of these sensors could be the monitoring of possible induced seismicity, an issue of growing public concern in face of some of the practices in the energy industry (oil and gas production, gas and CO2 storage, geothermic operations, Valoroso et al., 2009; Clarke et al., 2014). Induced seismicity usually generates moderate seismic events of magnitude rarely larger than 3.0 (Ellsworth, 2013; Westaway and Younger, 2014; Walsh and Zoback, 2015; Weingarten et al., 2015). For most of the national regulations, reaching magnitude values of the order of Mw=3.0 puts operations at risk. Most of the induced earthquakes recorded in the world are of Mw<3 and often much lower, but this may nevertheless trigger the public concern that larger events may follow (e.g., Majer et al., 2007). As a consequence, subsoil activities must provide a suitable seismic monitoring plan that involves the installation of an expensive network of sensors. The use of advanced MEMS low‐cost sensors could greatly improve these applications, freeing resources on one hand and providing better monitoring on the other.

In this study, we tested the performance of a MEMS sensor prototype designed by the Italian company ADEL srl to monitor small local events. A selection of these sensors that are digital, multirange, triaxial MEMS accelerometers (named ASX1000) were installed in seismic zones of northern and central Italy, to evaluate their performance in recording low‐energy events.

We performed an analysis of the detected seismicity (ML<3.0), comparing the data as recorded by the MEMS sensors with the nearest available high‐quality accelerometers managed by the Italian National Protection Service (Rete Accelerometrica Nazionale [RAN]). The results are promising and open new perspectives for the extensive use of such low‐cost MEMS sensors for the monitoring of local seismicity.

The MEMS sensor prototype adopted in this work (Adel ASX1000) is a triaxial MEMS accelerometer, with an internal circuit of transduction and digital output. The sensor has three sensing elements oriented along mutually orthogonal axes. It operates in high sensitivity mode for an acceleration range of ±2g (g=9.80665  m/s2). The prototype frequency bandwidth is set to 0–62.5 Hz at 250 Hz sampling. ASX1000 has a micro‐SD memory card that stores the recorded data and three communication channels for the remote control and data transmission (a serial channel RS‐422 or RS485, a LAN Ethernet 10/100 Mbit/s, and an USB 2.0). The MEMS sensor prototype was calibrated in the laboratory on a shake table with a sweeping signal characterized by a frequency range between 0 and 100 Hz, a duration of 60 s, and a maximum amplitude of ±0.2  m/s2. The prototype proved to be in very good agreement with the laser motion reference for the frequency response up to 80 Hz. Noise analysis of the single components detects a power spectral density (PSD) with a general downward trend between −80 and 65  dB in the 0.2–10 Hz frequency range, interesting for earthquake engineering (Fig. 1). Figure 1 compares the detectable magnitude of seismic events measured at distances of 10 km, as in Kong et al. (2016), with the obtained PSD from the horizontal component of the ASX1000 MEMS prototype, common industrial phone MEMS sensors, and seismological high‐quality stations. Lab tests suggested that ASX1000 has the potential sensitivity to record local events with magnitude Mw>2.5 in the 2–10 Hz frequency range, which is the most critical frequency range of seismological interest.

To test the performance in detecting small seismic events, 15 ASX1000 prototypes were installed at two seismic active areas of Italy: the inner part of the Umbria Valley (central Italy, Chiaraluce et al., 2017) and the southern‐east Alpine Front (northern Italy, Galadini et al., 2005). Figure 2 shows the locations of the MEMS arrays, the locations of the closest national strong‐motion high‐quality sensors (RAN that use Kinemetrics Episensor ES‐T sensors), and the epicenters of the recorded seismic events analyzed in this study. The MEMS sensors were installed inside telecommunication infrastructures at the base of the local server room, and the sensors were firmly coupled with the ground with screws and plugs. Raw data were transmitted to a central service in real time through a LAN connection.

The MEMS sensors were able to detect nine small local earthquakes with 2.0<ML<3.0 between April 2020 and February 2021. The corresponding epicenters are shown in Figure 2. The location and the local magnitude of the recorded events were retrieved from the National Seismic Institute earthquakes database (Istituto Nazionale di Geofisica e Vulcanologia [INGV]). Figure 3 shows the time series signal as recorded by the ASX1000 MEMS prototypes (horizontal‐transverse component), the earthquake’s ID name, and the distance between the sensors and the epicenters.

Even small events are clearly visible, especially for the nearest epicenters (e.g., Gubbio, Gualdo C., and Gubbio_2 earthquakes in Fig. 3a,f,h, respectively). From the recorded waveforms, we determined the PGA values and the acceleration response spectra (5% damped spectral acceleration [SA]), to be compared with the recordings of the high‐quality stations belonging to the RAN. The RAN accelerograms can be downloaded from the Italian Civil Protection website (see Data Availability Statement). Figure 4 shows a comparison between the horizontal‐transverse components recorded by our MEMS stations and by the high‐quality accelerometers for the events having similar epicentral distances and soil conditions; this comparison avoids, as much as possible, unreliable comparison due especially to local effects (e.g., stations installed on hard rock vs. soft sediments). The considered events were named Gubbio, Cismon, Claut, and Pietralunga, as recorded respectively by the MEMS Gubbio, Bassano, Gemona, and the high‐quality stations GBSL, BSG, and VAV (see Fig. 2).

As apparent from Figure 4, MEMS recordings are in very good agreement with the high‐quality stations, both in terms of PGA and spectral response, especially for the closest events (consider the Gubbio–GBSL stations comparison in Fig. 4a).

In Table 1, we report the PGA values as recorded by the ASX1000 MEMS prototypes and by the closest high‐quality RAN seismic stations for the earthquakes considered in Figure 4 along with the station–event distances, the MEMS–RAN distances (interstation distances), and the discrepancy of the PGAs (ΔPGAs=PGAMEMSPGARAN).

The PGA values are very similar, especially for the horizontal components, presenting an average discrepancy between MEMS sensors and high‐quality accelerometers of <15%. The best match between the two sensors is given by the Gubbio earthquake recordings in which the high‐quality station RAN and the MEMS sensor have nearly the same location. In this case discrepancy is 6%. Higher discrepancy is observed for the vertical component (Z) of the Gubbio–GBSL stations, possibly related to installation issues. The ASX1000 MEMS responds in agreement with high‐quality stations even for remote and small earthquakes, such as the case of the Claut event (with the epicenter 51 km away from the recording stations).

MEMS sensors are often used to detect strong‐motion events. Our tests show the reliability of MEMS also for low magnitude (2.0<ML<3.0) seismicity monitoring. This study investigates the performance of the low‐cost ASX1000 MEMS prototype accelerometer under small earthquake (ML<3.0) excitation. Sensors have been initially tested using a calibrated shake table. Under the controlled sweep excitation, the MEMS sensor shows a good performance in the frequencies up to 80 Hz, critical for the earthquake engineering aims, and the self‐noise test shows that the PSD is around 80  dB in the 0.2–10 Hz range. Our field results show that the low‐cost ASX1000 MEMS prototypes are suitable to record seismic events with a minimum local magnitude of ML=2.0 and epicentral distance <20  km or ML3 at 50 km distance. The comparison of the PGA and spectral responses inferred from our MEMS sensors and the nearest high‐quality accelerometers (belonging to the RAN) yields a good match, especially at smaller epicenter distances and when the interstation distances (MEMS–RAN) are lower than 2 km.

We note that these sensors are installed in urban telecommunication infrastructures. Although they are located in a noisy environment, not intended for seismic monitoring, they are still capable of recording small local earthquakes.

The use of low‐cost sensors for small or local earthquakes monitoring will allow the development of dense accelerometric networks, thus mitigating the spatial sampling issues and providing highly detailed ground shaking maps, even for small events that are of course also very frequent. A well‐defined shaking pattern inferred from small earthquakes can help identify local site effects that are often overlooked.

On the basis of these results, the ASX1000 MEMS prototype could be efficiently integrated into existing national seismic networks (as suggested by D’Alessandro et al., 2019).

Moreover, the MEMS sensor prototype evaluated in this study could potentially be adopted for induced seismicity detection. A dense network with MEMS sensors distributed in the immediate proximity of industrial activity sites (such as oil and gas production, geothermic operation, etc.) could improve low‐magnitude seismicity monitoring.

While a promising geophysical tool, the low‐cost ASX1000 prototype still suffers from a high level of internal noise, which is usually considered to limit their use for several seismological purposes, such as accurate earthquake locations or origin time estimations. Nevertheless, the fast improvement of sensors quality combined with decreasing costs suggests that in the near future MEMS‐based motion networks will play a relevant role in the monitoring of lower‐level seismicity, which still conveys important seismologic information such as local seismic response to large, destructive events.

Data availability statement:

The public data used in this article are available at and Other data are available upon request for scientific purposes.

The microelectromechanical systems (MEMS) data supporting the conclusions of this article, being property of a private company, is made available by the authors only upon request for scientific purposes. The Rete Accelerometrica Nazionale (RAN) accelerometric data are public and can be downloaded from the Italian Civil Protection website (, available only for earthquakes with ML2.5, and only upon request for smaller ML). The event locations and timing are public data and can be downloaded from the National Institute of Geophysics and Volcanology (INGV) website ( All websites were last accessed in February 2021. Supplemental material for this article includes the complete technical information of the MEMS sensor prototype and the calibration experiment on the shake table.

The authors declare no conflict of interest.

The authors wish to thank ADEL srl for the use of their proprietary data and TIM SpA for the permission granted to microelectromechanical systems (MEMS) installation in their infrastructures. The authors would also like to thank David Zuliani of the Seismological Research Center of Istituto Nazionale di Oceanografia e Geofisica Sperimentale (OGS) for the calibration experiment on the shake table apparatus.

This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.