Teleseisms and microseisms on an 1 ocean-bottom distributed acoustic sensing array 2

5 Sparse seismic instrumentation in the oceans limits our understanding of deep Earth dynamics 6 and submarine earthquakes. Distributed acoustic sensing (DAS), an emerging technology that con7 verts optical fiber to seismic sensors, allows us to leverage pre-existing submarine telecommunication 8 cables for seismic monitoring. Here we report observations of a teleseismic earthquake, local surface 9 gravity waves, and microseism along a 4192-sensor ocean-bottom DAS array offshore Belgium. We 10 successfully recover Pand S-wave phases from the 2018-08-19 Mw8.2 Fiji deep earthquake in the 11 0.01-1 Hz frequency band. We also observe in-situ how opposing groups of ocean surface gravity 12 waves generate double-frequency seismic Scholte waves, as described by the Longuet-Higgins theory 13 of microseism generation. These results suggest great potential of DAS in next-generation submarine 14 seismic networks. 15

and unlimited deployment duration as long as the DAS unit is powered. For about a decade, DAS 48 has been successfully utilized in boreholes for active-source seismic profiling (e.g. Mestayer et al., 2011;49 Mateeva et al., 2012;Parker et al., 2014). Recent work with onshore trenched or conduit-installed 50 horizontal fibers has demonstrated the ability of DAS arrays to record earthquakes and other seismic 51 signals at local to teleseismic distances with high waveform fidelity (Lindsey et al., 2017;Jousset et al., 52 2018;Li and Zhan, 2018;Wang et al., 2018;Ajo-Franklin et al., 2019;Yu et al., 2019). In this letter, 53 we demonstrate that submarine horizontal DAS arrays utilizing pre-existing ocean-bottom fiber-optic  Teleseismic waves with near-instantaneous apparent phase velocity appear in the zero-wavenumber bin at low frequencies. seismology.

123
In addition to the broad, low-frequency peak associated with Fiji teleseisms, the power spectral 124 densities (PSD) of BDASA channels computed over the full 1-hr strain record exhibit four distinct peaks 125 at 0.09 Hz, 0.18 Hz, 0.36 Hz, and 1.12 Hz (Fig. 4a). Here, we focus primarily on the 0.18 and 0.36 126 Hz peaks. We attribute the 0.18 Hz peak to poroelastic strains induced by the pressure field of ocean 127 surface gravity waves propagating across the array. Globally, surface gravity and infragravity waves 128 between 0.003-3 Hz are excited in oceanic waters by wind-sea interaction. Invoking linear wave theory, 129 the magnitude of the seafloor pressure perturbations beneath a surface gravity wave scales with angular 130 wavenumber k and water depth H as p d ∝ sech(kH) (e.g. Holthuijsen, 2007). We use the dispersion 131 relation for ocean surface gravity waves, ω 2 = gktanh(kH), to calculate a theoretical p d along the cable 132 depth profile and fit the Fourier amplitude at 0.18 Hz as a linear function of p d (see Supplementary   133 Material). We observe a good correspondence between the observed and predicted Fourier amplitude at 134 0.18 Hz with both water depth and distance along the cable (Fig. 4b,c), suggesting that surface gravity 135 waves dominate the data at these frequencies.

136
Our interpretation of surface gravity waves at 0.18 Hz is supported by f-k analysis, which shows  For the segment of the BDASA closest to shore, stronger incoming (land-ward propagating) surface 149 gravity waves occupying f-k quadrants 1 and 3 are observed. The relative strength of outgoing (sea-ward 150 propagating) surface gravity waves occupying f-k quadrants 2 and 4 increases with distance along the 151 cable ( Fig. 5). For the last 10 km segment of the array between 30 and 40 km the outgoing and incoming 152 the bathymetric ridge at 30 km and the sloping seabed approaching the coast (Fig. 1b), and note that 154 such opposing wave groups necessarily interfere to produce a standing wave. In particular, the reflective 155 effect of the bathymetric ridge at 30 km can be clearly observed in the directional spectra ( Fig. 5c), where 156 incoming wave energy is amplified relative to outgoing energy on the 20-30 km segment and outgoing 157 energy is amplified relative to incoming energy on the 30-40 km segment.

158
The directional spectrum of surface gravity waves does not change appreciably over time along the 159 first 30 km of the array. For the last 10 km segment, however, the observed f-k spectrum evolves 160 over time and is asymmetrical, with faster incoming waves and slower outgoing waves (Fig. 5b). We Unlike the 0.18 Hz energy peak, the 0.36 Hz peak observed in the BDASA PSD is almost invariant 175 with depth and is not adequately described by the pressure-depth scaling of ocean surface gravity waves 176 (Fig. 4c). Instead, the Fourier amplitude at 0.36 Hz increases gradually with distance along the array 177 (Fig. 4b). The cable segment in water depths < 10 m is neglected in this analysis, as the PSD of this 178 region is saturated by incoherent energy across a broad band, likely associated with shoaling of ocean 179 waves. In the f-k domain, we identify a broad energy packet between 0.3 and 3.5 Hz with peaks at 0.36 180 and 1.12 Hz characterized by phase velocities faster than ∼300 m/s (Fig. 6a). When projected from the 181 frequency-wavenumber domain into frequency-phase velocity space, this high-frequency energy packet 182 exhibits strong dispersion from phase velocities close to the compressional velocity of water (∼1500 183 m/s) at 0.36 Hz to an asymptotic velocity of ∼250-450 m/s above 1 Hz (Fig. 2b, 6b). Again, in the f-k  with conventional DAS systems, chirped-pulse DAS offers high signal-to-noise ratio (SNR) and low 298 variations in sensitivity along the fiber (Pastor-Graells et al., 2017, Costa et al., 2018, Fernández-Ruiz 299 et al., 2018a. The key of its performance lies in the use of a linearly chirped probe pulse for the time- is not observed in chirped-pulse DAS. More importantly, sensitivity of conventional DAS completely 317 fades in certain points along the fiber (acoustic SNR <1 in up to 6% of fiber locations considering a 318 healthy-SNR optical trace) due to the impossibility of maintaining the phase reference in low intensity 319 trace regions caused by its interferometric nature. Those 'blind spots' need to be corrected using com-320 plex post-processing techniques or multi-wavelength measurements (Chen et al., 2017), typically at the 321 expense of sensing bandwidth and higher measurements times. Chirped-pulse DAS, however, shows no 322 fading sensitivity, enabling the "raw" strain signal as measured by the DAS to be directly processed 323 without using any denoising/smoothing algorithm. This steady sensitivity has particular impact on the 324 subsequent 2D processing applied to isolate seismic events from other sources, since all points are cap-325 tured with similar noise/sensitivity along the whole fiber length (>40 km) (Fernández-Ruiz et al., 2018a).

326
In this study, we observed DAS sensitivity variation along the 42-km fiber lower than 3 dB. This is in 327 contrast with the typical SNR variations observed in traditional DAS systems, where a variance of >18 328