We monitored temporal variations of seismic noise level during 2.5 years between January 1, 2018 and June 30, 2020. To evaluate temporal variations clearly, we calculated the noise reduction ratio, which is defined as the seismic noise level (square root of the trace element of the averaged Fourier spectra covariance matrix) divided by the average noise level of the corresponding day of the week and hour. Here, holiday periods such as national holidays, year-end and new year holidays, and summer holidays are treated as Sunday. Figures 7 and S2 show temporal variations for noise reduction ratio of UD components as well as the incident azimuth of noise polarization estimated by polarization analysis.
Temporal variations of noise reduction ratio in Figures 7 and S2 show that the noise level varies throughout the 2.5 years, even before the COVID-19 pandemic influenced social activities. There seems to be seasonal variations in the noise reduction ratio. Winter season (December to February) tends to show a higher noise level in many stations, which is especially evident in E.IKBM, E.SBCM (Figure 7), and E.RYGM (Figure S2). This high noise level is often observed in the frequency range of > 20 Hz. Summer season (around August) also shows a higher noise level in some stations (for example, E.HN1M (Figure 7), E.SBAM, and E.YKKM (Figure S2)).
Although noise reduction ratio shows temporal variations, the incident azimuth and incident angle of noise polarization is constant with time suggesting that spatial distributions of noise sources are common for 2.5 years. However, there are several examples where the incident azimuth changes abruptly or gradually with time. For example, a large noise level was observed from April to October in 2018 and 2019 at E.HGCM station (Figure S2). Incident azimuth changes abruptly when the large noise level started and ended. In this case, the incident azimuth for the low and high noise periods is stable with time, which suggests that spatial distributions of the cultural noise sources around the station change seasonally.
Noise reduction ratio in 2020 shows a completely different pattern from that in 2018 and 2019 at many stations. The low noise period abruptly started around April at many stations. This low noise level spans only limited frequency bands such as 1–40 Hz at some stations (for example, E.HN1M), whereas it spans the entire frequency band at other stations (for example, E.IKBM). Although low noise periods are also observed in 2018 and 2019, the noise level in the low-frequency band of 1–20 Hz is not reduced significantly in 2018 and 2019. The amount of cultural noise in a low-frequency band (around 1–10 Hz) excited on weekdays are common among MeSO-net stations (Figure 5), whereas that in the higher frequency band is different (Figure 6). This suggests that cultural noise in a lower frequency band records large-scale social activity in the Tokyo Metropolitan area, whereas cultural noise in the higher frequency band records local social activity around the stations. Therefore, the seismic noise reduction in the lower frequency band that occurred in 2020 is interpreted to be due to the social activity reduction for COVID-19 in the Tokyo Metropolitan area.
At several stations (E.HN1M, E.SBCM, E.AYHM, E.KHDM, E.SBAM, and E.HGCM), seismic noise reduction in a high frequency band of 20–40 Hz or 20–50 Hz precedes seismic noise reduction in the low frequency band of 1–20 Hz. We calculated the median noise reduction ratio in two different frequency bands (5–10 Hz and 25–35 Hz) during daytime (10:00 h to 16:00 h) (Figures 8 and S3). The seismic noise reduction in the high frequency bands occurred when the schools were closed at the beginning of March. As the MeSO-net stations are usually installed in school properties, social activity reduction in a school building decreases seismic wave excitation there, which is recorded in the seismometers at the high frequency bands that is sensitive to the locally excited cultural noise. However, this first-step noise reduction was recorded only in limited stations. The relative strength of cultural noise sources around the stations in the higher frequency bands may determine whether seismic noise reduction in the school buildings can be observed.
The seismic noise level started recovering since May 2020. At many stations, the noise reduction ratio increased after the Golden Week at the beginning of May in both frequency bands (Figures 7, 8, S2, and S3). This recovery correlates well with the population recovery (Figure 3). It started before the state of emergency was lifted on May 25, 2020, which suggests that social activity was restarted before the government officially announced the relaxation.