Figure 5. Tripartite array analysis results. (a) The location of coherent seismic bursts along the Salah Al-Deen Road observed during the two hours preceding the attack are shown by the red circles. Note that the location assumes sources are located along the portion of the road marked by the red line in the figure. The polar diagrams show the distribution of rotation angles at each station. Dashed lines indicate the range of back-azimuths resolved at each station, and the highlighted polygon indicates the area formed by the back-azimuth intersection. The barrier breaches are indicated by the yellow circles, and local Palestinian towns and Israeli Kibbutzim are indicated by the black squares. The inset histogram presents the distribution of the misfit between the observed and calculated surface wave travel times. (b) Space-time analysis of coherent seismic bursts located along the Salah Al-Deen Road. Dashed lines denote constant velocities of 44 and 75 km/s. The right vertical axis indicates the location of Palestinian towns along the Salah Al-Deen Road.
Discussion
How early could a seismic-based pre-attack warning be issued?
That seismic signal, attributed here to the late Hamas deployment in the Gaza Strip, stands well above the noise level and is correlated at 85%-92% between the three station pairs. The strong pre-attack signal was preceded by a phase characterized by high-amplitude seismic energy, which lasted between 04:30 to 05:45 local time, and that was mostly observed at AMZNI, and to a lesser degree also at KZIT. Over such durations, one or two random signals are likely to rise to the observed levels. Thus, the signal observed at AMZNI and KZIT starting from 04:30, was hardly sufficient to confidently establish that anomalous activity was taking place in the Gaza Strip. The probability for abnormal activity in Gaza significantly increased with the onset of the signal of the late Hamas deployment, from about 06:10. Given the background noise levels, the amplitudes observed between about 06:10 to 06:30 were sufficient to establish that large-scale mobilization was taking place within the Gaza strip at a high level of confidence. Starting from the onset of the second deployment, the level of confidence increased almost monotonically with the countdown to attack, finally reaching >99% in the last few minutes before the barrier breach began.
The results suggest that embedding analysis of traffic-induced seismic noise into decision-making protocols may be useful for detecting large-scale mobilization in real time. Most seismic-based traffic detection systems rely on near-target recordings (Bin et al., 2021), however, analysis of the October 7 dataset suggests that, in certain conditions, remote stations could also be used for these purposes. Deploying stations further from the target area also means the network is less likely to be decapitated during the attack, as was the fate of many of the IDF near-fence systems on October 7. Since the type and origin of signals excited by pre-attack motions are not well characterized, monitoring areas as large as the Gaza Strip would likely require a dense network. Such systems could provide tens-of-minutes lead time, which is crucial for preventing terrorist attacks, and for notifying nearby civilian communities.
Implications for Hamas pre-attack preparations
The space-time distribution of signals associated with anomalous traffic in Gaza (Figure 5b) bears important implications for the Hamas operational plan. The locations indicate that pre-attack mobilization originated near Khan Younes, a city believed to serve as a hideout for the senior Hamas leadership as well as for many of its troops, and where intense fighting is currently taking place. It is therefore not surprising to find prominent pre-attack activity in that area.
The timing and location of the most coherent signals support a two-stage deployment scheme. The early stage, which took place between 04:45 to 05:45 local time, consisted of slower, possibly sparser movement of troops north and south of Khan Younes. The activity extends out to about 15 km north of the city, and a few kilometers to its south. The troops likely paused between 05:45 to about 06:00, as is manifested by a decline in seismic amplitudes (Figure 4a-c), and, accordingly, in the rate of array-based detections (Figure 5b). One possibility is that in this lag, troops advanced and then paused to watch for activity in Israel suggesting their earlier motions have been detected. This scenario is consistent with IDF reports of the last pre-attack near-barrier activity taking place northeast of Nuseirat at about 05:00 local time. The location of that near-barrier activity is in agreement with locations of coherent arrivals resolved by the IS stations, and the lack of later pre-attack reports is consistent with the inter-deployment seismic quiescence.
The duration of the no-detection epoch suggests Hamas militants took about 15 minutes to confirm the IDF hadn't spotted their advancement. The ensuing detection rate and amplitude increase suggest that the pause was followed by a rapid and larger wave of troops who left the Khan Younis area and spread out towards the furthest extent of the Gaza Strip. It seems that once the second wave reached Beit Lahia and Rafah, in the northern and southernmost edges of the strip, an order was given to hit all near-barrier positions at once.