Figure 1. Location map showing the Gaza Strip and the surrounding area. Yellow circles indicate the location of fence breaches reported on October 7, 2023. Black squares are for the Palestinian cities of Beit Lahia, Jabalia, Gaza, Nuseirat, Al Zawayda, Khan Younes, and Rafah, and the Israeli Kibbutzim Erez and Kerem Shalom. The inset shows the location of AMZNI, YATR, and KZIT, the three seismic stations used in this study. The Salah Al-Deen Road is indicated by the red curve, and international borders are indicated by the black curves.
Temporal and spectral analysis of the October 7 seismic noise field
Under favorable SNR conditions, traffic-induced surface- and body waves can sometimes show up on seismograms recorded tens-of-kilometers from the source.  The number, speed, location, and mass of the Hamas vehicles traveling in Gaza in the early hours of October 7 are unknown, making it impossible to estimate whether any pre-attack motions produced a signal exceeding the background noise level. Yet, visual inspection of IS network seismograms reveals signals exceeding the noise levels at the stations nearest to the Gaza Strip during the two hours before the attack.  Seismograms and spectra recorded on the day of the attack at IS stations AMZNI, YATR, and KZIT are shown in Figure 2.  Each of the spectra exhibit a clear peak that rises above the background noise level,  which was estimated from Saturdays during 2021, 2022, and 2023 for the time window encompassing the pre-attack spectra.  The highest signal-to-noise ratio (SNR) for AMZNI, YATR, and KZIT is observed at frequencies ranging between 2.1 to 2.8 Hz, 5.5 to 6.5 Hz, and 2.8 to 3.2 Hz, respectively, and are referred to here as the target frequency bands.  Seismic signals in these frequency bands had previously been associated with traffic activity (Yamanaka et al., 1993; Bonnefoy-Claudet et al., 2006; Groos and Ritter, 2009, Riahand Gerstoft, 2015; Inbal et al., 2018; Mi et al., 2022). Interestingly, the three spectral peaks, which presumably originated from a common source (or sources), do not overlap.  Similar behavior had previously been observed for ground motion modulated by remote wind-turbine activity (Inbal et al., 2018), and may be the result of the variability of the thicknesses of the upper layer encountered along the path to the stations.  To qualitatively assess these effects, I have used a discrete wavenumber reflectivity method (Cotton and Coutant, 1997) to calculate synthetic spectra excited by a vertical force acting on the surface of a plane-layered model consisting of a low-velocity thin layer overlaying a homogeneous uniform elastic half-space.  The spectra computed for upper-layer thicknesses of 20 and 100 m (Figure 3; see table S1 for the layer properties), consistent with the width of the upper soft sedimentary layer in the study area (Gardosh et al., 2011), peak at frequencies close to the ones observed to peak in the IS data.  Other factors that can affect the spectra include the spatial distribution of sources, whose extent is close to the aperture of the IS stations, the local topography, and the depth of the water table.