key: cord-330562-dabjcvno authors: Poli, Piero; Boaga, Jacopo; Molinari, Irene; Cascone, Valeria; Boschi, Lapo title: The 2020 coronavirus lockdown and seismic monitoring of anthropic activities in Northern Italy date: 2020-06-10 journal: Sci Rep DOI: 10.1038/s41598-020-66368-0 sha: doc_id: 330562 cord_uid: dabjcvno In March/April 2020 the Italian government drastically reduced vehicle traffic and interrupted all non-essential industrial activities over the entire national territory. Italy thus became the first country in the world, with the exception of Hubei, to enact lockdown measures as a consequence of the COVID-19 outbreak and the need to contain it. Italy is also a seismically active area, and as such is monitored by a dense permanent network of seismic stations. We analyse continuous seismic data from many stations in northern and central Italy, and quantify the impact of the lockdown on seismic ambient noise, as a function of time and location. We find that the lockdown reduces ambient noise significantly in the 1–10 Hz frequency range; because natural sources of seismic noise are not affected by the lockdown, the seismic signature of anthropic noise can be characterised with unprecedented clarity, by simply comparing the signal recorded before and after the lockdown. Our results correlate well with independent evaluations of the impact of the lockdown (e.g., cell phone displacements), and we submit that ambient-noise seismology is a useful tool to monitor containment measures such as the coronavirus lockdowns. noise is of interest to geoscientists, as it can be used at relatively small scales, for instance in mapping and monitoring efforts 9, 10 . While earlier studies have attempted to characterise high-frequency seismic noise 7, [11] [12] [13] [14] , the current lockdown of industrial activities and reduction in road and train traffic in Italy is an unprecedented opportunity to discriminate it from ambient noise of natural origin. Italy is a highly industrialized and urbanized country, densely covered with non-stationary noise sources 15 , such as traffic and industry-induced vibration 16 . This is particularly true in its northern regions, which account for 70% of the country's entire industrial output, and where lockdown measures have been enacted earlier than everywhere else in Europe. We analyse continuous data from an array of broadband seismic stations, located in the vicinity of known industrial districts in Lombardy, Emilia-Romagna and Tuscany (Fig. 1) ; we identify the spectral signature of the March 2020 lockdown, and take advantage of the lockdown to quantify and evaluate the spectral signature of anthropic activities. Importantly, measuring the overall reduction in seismic energy associated with the lockdown is also a way to quantify its effects; this is relevant to governmental entities, wishing to monitor the effectiveness of the measures being taken. We downloaded publicly available, continuous, three-component seismic recordings from a set of permanent broadband stations, part of the Italian National Seismic Network operated by the Istituto Nazionale di Geofisica e Vulcanologia 17 . All instruments have a flat response at frequencies between ~0.01 and ~10 Hz, or broader; we remove ("deconvolve") instrument response from the data prior to our analysis. The locations of stations employed in most of our study are shown in Fig. 1 . Stations were selected based on their proximity to industrial districts; in particular, MILN is located near the city of Milano, with a particularly high concentration of vehicle traffic and industrial activities. The seismic signature of the containment measures in Italy is apparent from a relatively simple analysis of continuous recordings at station MILN, located within the city limits of Milano, in a busy area near the University of Milano campus and the Lambrate train station. We compute spectrograms (Fig. 2) by Fourier-transforming 1-hour-long segments of continuous signal, with a 30-minute overlap between subsequent segments; for each calendar day, all segments are then averaged, and the squared modulus of the resulting average Fourier transform is computed: this way, a single "power-spectral density" (PSD) function is obtained, for each station, component (East-West, North-South, vertical) and calendar day. Figure 2 shows clearly that the lockdown has a relevant impact on recorded seismic noise over a broad frequency range; its effect disappears at frequencies below 1 Hz, where anthropic noise is weaker. The energy drop associated with the lockdown is comparable with that occurring every weekend and during the winter break, both in 2018/19 and 2019/20. Interestingly, loss of energy is gradual over time, starting with the first lockdown measures on February 21, and increasing with time until a plateau is reached around March 22 (interruption of non-essential industrial activities). A trend similar to that seen in Fig. 2 has also been found through the analysis of cell phone displacements 18 . This suggests that vehicle traffic, which was significantly reduced (particularly in and around Milano) already with the February measures, contributes significantly to the entire spectrum of anthropic noise; there is also episodical evidence from the press that a number of factories were closed based on the unilateral decision of their owners, before the government-imposed lockdown. The analysis applied to station MILN is repeated for all seismic stations of Fig. 1 , and the results are illustrated in Figs www.nature.com/scientificreports www.nature.com/scientificreports/ Lockdown measures apparently impact all stations under consideration, but the character of their effects changes in various ways with station location. In the case of FIR, located in the city of Florence, the signature of the winter break is almost negligible, while the February/March lockdown still has a prominent effect; it might be possible to interpret this observation through the analysis of anthropic activities usually taking place in the area (e.g., tourism, which is presumably not reduced by the holiday). The drop is gradual at all stations, with no specific governmental decree standing out with respect to the others. At station PRMA, a slight increase in ambient noise occurs after February 21 and before March 9. We next characterise anthropic noise by evaluating variations in the spectra of seismic ambient noise before and after the implementation of lockdown measures. We compute the ratios of the PSD measured (as described in sec. 3) on Tuesday March 31 2020, to that measured at the same station on Tuesday December 3 2019. We carry out this calculation separately for each component, and for all stations analysed thus far; the results of this exercise are shown in Fig. 5 . The energy associated with ambient signal is clearly reduced for all stations, at all frequencies in the range of interest. At each station, PSD ratios change with frequency almost exactly in the same way for all components. Near 1 Hz, all stations show a more or less rapid decline in the PSD ratio, with ambient noise being more effectively reduced as frequency grows. This trend continues all the way to 20 Hz at stations EUCT and PRMA, while other stations show a more complex behaviour. Above 10 Hz station MILN stands out, its PSD ratio growing quickly with frequency. Anthropic noise is known to be relevant at frequencies above 1 Hz, and to consist of a range of different excitation mechanisms 7, 15, 19, 20 . Natural sources such as rain, wind 21 and sea/ocean waves are typically characterized by frequencies below 1 Hz, and are obviously not affected by the lockdown. We infer that, by taking the ratio of noise spectra before and after the lockdown, an estimate of the spectral character of anthropic noise is obtained, and the spectra in Fig. 5 can help us estimate the nature of anthropic noise in the region of interest, independent of the lockdown; the frequencies where the PSD of ambient signal is most reduced by the lockdown are those where, in normal times, the contribution of anthropic activities to seismic ambient noise is most important. The fact that most energy loss associated with the lockdown is at frequencies between 1-10 Hz is coherent with what is known of the typical signature of industrial activity and vehicle traffic 22 . Seismic data recorded during the lockdown might be particularly useful in identifying sources of anthropic noise, which could be employed by geophysicists, after the lockdown, e.g. to characterise the upper subsoil by cross correlation of ambient signal 23, 24 . We further analysed the relationship between ambient noise recorded on different components, finding the "H/V" ratio between the PSDs of horizontal-component and vertical-component signals: first, the PSD of each component of signal recorded on a given day is averaged in the frequency range 1-10 Hz; then the arithmetical average of the resulting East-West-component and North-South-component values is taken; finally, the ratio of the resulting horizontal PSD to the vertical one is computed. The procedure is iterated for each station and for each day between December 1 and March 31, and the results are shown in Fig. 6 . In general, the value of H/V is related to how seismic energy in the ambient-noise field is distributed in the form of compressional, shear and surface waves 7, 25, 26 ; changes in H/V after vs. before the lockdown would reveal whether the reduction in anthropic noise affects one of these seismic phases/components more or less importantly than the others; in other words, whether traffic and industry-induced vibration can be associated to one particular constituent of the seismic field. Figure 6 shows that the lockdown measures have no effect on H/V, and we infer that, while anthropic noise is reduced significantly by the lockdown (Fig. 5) , the relative contributions of compressional, shear and surface waves remain approximately constant: the noise wave field is stable in the Fig. 3 are plotted here on a single graph, for comparison; for each station, the average value of the PSD observed in the time interval of interest is subtracted from the corresponding curve, prior to plotting, as this can change significantly from station to station, but is not relevant to our analysis. Each colour corresponds to one station, as specified. Again, the dates of mentioned governmental decrees are highlighted as in Fig. 2 . The Italian territory is densely covered by seismic instruments, and by repeating our analysis on the entire network of available stations we are able to quantify the spatial dependence of anthropic noise reduction. For each station, for each day, the PSD of signal recorded 6 AM to 8 PM is computed, and averaged over different frequency bands. In practice, we employ the direct Fourier method 27 , as implemented in the Obspy package 28, 29 : this is standard procedure to identify artefacts related to station operation, episodic cultural noise, overall station quality and level of Earth noise at each site. To emphasize the change in ambient noise with the lockdown, we plot the difference between the values so obtained on three dates in 2020, and reference values obtained conducting the same calculation on data recorded for five months until the lockdown, and averaging. We include as Supplementary Material S1 an animated version of Fig. 7 , showing the PSD at the same stations, October 7, 2019 through April 1, 2020; through this time-dependent visualization, the drastic effects of the lockdown are further emphasized. Our main result, that noise be strongly reduced after the lockdown in the "cultural" frequency range, is confirmed by Fig. 7 , and extended to most of Northern Italy. Between 1-3 Hz, the lockdown effects are more pronounced in the Lombardy and Veneto regions than in Central Italy and along the Apennine range. The most important reductions in ambient noise are recorded by stations along the Alpine arc, near Torino, Milano and Verona, and in the city of Florence. We have analysed continuous data from northern Italy, and quantified the effects of the March 2020 coronavirus lockdown on the seismic ambient noise field. We confirm that this effect is significant, and easily observed in our data: see in particular Figs. 3 and 4. The Italian government first imposed a reduction of people (and therefore vehicle) movement, on March 9; we find that this date marks the beginning of a gradual loss in ambient-noise energy at all frequencies, which we attribute to the reduction of road and railroad traffic in the region of interest. Depending on the station, the energy curve flattens out, or starts to decline more slowly towards the beginning of April, despite the more stringent measures imposed at that time (interruption of all non-essential industrial activities). A similar trend has been found from cell-phone displacement data 18 . One implication of our observations is that seismic data could be useful for governmental institutions to monitor the effectiveness of measures involving a reduction or interruption of human activity in a given area. It is understood that the lockdown only reduces noise of anthropic origin; it follows that by comparing the Fourier spectrum of seismic ambient noise before and after the lockdown (Fig. 5) , one can attempt to characterise anthropic noise. We find that, confirming earlier estimates 30,31 , anthropic noise becomes dominant at frequencies Coronavirus lockdowns have changed the way Earth moves Seismic imaging and monitoring with ambient noise correlations Stationary-phase integrals in the cross-correlation of ambient noise A theory of the origin of microseisms Seismic noise in Fennoscandia, with emphasis on high frequencies Variations in broadband seismic noise at IRIS/IDA stations in the USSR with implications for event detection The nature of noise wavefield and its applications for site effects studies: A literature review Emergence of broadband Rayleigh waves from correlations of the ambient seismic noise Faster, better: shear-wave velocity to 100 meters depth from refraction microtremor arrays Rain and small earthquakes maintain a slow-moving landslide in a persistent critical state Characterization of and correction for cultural noise H/V ratio: a tool for site effects evaluation. Results from 1-D noise simulations On the stability and reproducibility of the horizontal to vertical spectral ratio on ambient noise: Case study of Cavola, northern Italy Cultural noise and the night-day asymmetry of the seismic activity recorded at the Bunker-East (BKE) Vesuvian Station Maninduced low frequency seismic events in Italy A catalogue of non-tectonic earthquakes in central-eastern Italy Italian Seismological Instrumental and Parametric Database (ISIDe) The reduction of social mixing in Italy following the lockdown Observations and modeling of seismic background noise Recent advances in seismology Spectral analysis of seismic noise induced by rivers: A new tool to monitor spatiotemporal changes in stream hydrodynamics Sources of long range anthropogenic noise in Southern California and implications for tectonic tremor detection Shear wave structural models of Venice Plain, Italy, from time cross correlation of seismic noise Train traffic as a powerful noise source for monitoring active faults with seismic interferometry Observation of equipartition of seismic waves Seismic velocity change patterns along the San Jacinto fault zone following the 2010 m 7.2 El Mayor-Cucapah and m 5.4 Collins Valley earthquakes An algorithm for the machine calculation of complex Fourier Series ObsPy: A bridge for seismology into the scientific Python ecosystem Seismic Noise Analysis System, Power Spectral Density Probability Density Function: Stand-Alone Software Package. United States Geological Survey Open File Report Seismic noise level variation in South Korea Global oceanic microseism sources as seen by seismic arrays and predicted by wave action models Rete Sismica Nazionale (RSN) Mediterranean Very Broadband Seismographic Network (MedNet) North-East Italy Seismic Network. International Federation of Digital Seismograph Networks Regional Seismic Network of North Western Italy. International Federation of Digital Seismograph Networks The Generic Mapping Tools version 6 We downloaded and analysed continuous seismic data provided by the Istituto Nazionale di Geofisica e Vulcanologia, the Osservatorio Geofisico Sperimentale, the University of Genova. The Generic Mapping Tools were used to generate the map in Figure 1 . Piero Poli was supported by the European Union Horizon 2020 Research and Innovation Programme (grant agreements, 802777-MONIFAULTS). The authors declare no competing interests. Supplementary information is available for this paper at https://doi.org/10.1038/s41598-020-66368-0.Correspondence and requests for materials should be addressed to P.P.Reprints and permissions information is available at www.nature.com/reprints.Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.