Identification of transient seismo-acoustic signals from crashing ocean waves: Template matching and location of discrete surf events

Crashing ocean waves, or surf, have previously been identified as persistent generators of coherent infrasound signals from 0.5 to 20 Hz. Here, we demonstrate that infrasonic and seismic (seismo-acoustic) signals from surf are composed of repetitive transient events which can be detected and characterized using template matching. Using data collected from a series of field experiments designed to study seismo-acoustic surf signals in Santa Barbara, California, we show that source regions of these events can be constrained primarily to just offshore of a local coastal headland using a reverse-time-migration implementation on a small spatial scale (<5 km2). Our data include one continuously running infrasound sensor (September 2022–July 2023) to examine temporal signal evolution, complemented by several short-duration campaigns involving various infrasound arrays, co-located seismometers, and video recordings. Throughout varied oceanographic and atmospheric conditions, we detect up to tens of thousands of independent surf repeaters per day over the course of a year. The amplitudes of detected infrasound signals are correlated with offshore significant wave height and local wind speed. We identify coincident arrivals of seismic and infrasound signals with similar spectral characteristics, suggesting a linked source mechanism locally producing both the seismic and acoustic transient signals. Source regions estimated from array- and network-based methods correspond to the surf zone as seen in video footage, and the directions of selected transient signals align with the location of a rocky reef shelf nearshore. This work showcases the ability to extract near-real-time information about the coastal sea state from seismic and acoustic signal features.

Methods
The majority of this dataset is composed of infrasound data recorded with a principal station (Hyperion 3111-series set to 400 Hz sampling frequency), which recorded near-continuous data from September 6, 2022 to July 17, 2023 via DiGOS DATA-CUBE3. Data collection was performed (in part) at the University of California Natural Reserve System Coal Oil Point Reserve DOI: 10.21973/N3Z07N. Roughly every 2 weeks, the battery was swapped with a new one to keep the sensor running. Sensor was placed within a bush, with no other physical noise filters.

The principal infrasound station was supplemented by four temporary array deployments; for each deployment data collection period, there is also data available from the principal station. Deployments are described in more detail below.

Deployment 1: January 11, 2023 (local PST). Involved 4 Gem 1.0 sensors [Anderson et al., 2017] with sampling frequency 100 Hz, placed along the base of the cliff at Sands Beach (wrapped in a towel, no other physical noise filter).

Deployment 2: January 12–19, 2023 (local PST). Involved 4 Gem 1.0 sensors with sampling frequency 100 Hz (1 inactive for majority of data collection period), placed along the top of the cliff at Sands Beach (directly within bushes, no other physical noise filters). 

Deployment 3: July 10, 2023 (local PST). 
-    Three broadband Trillium Compact 120s 3-component seismometers with sampling frequency 400 Hz, named TCA1, TCA2, TCA3. Unfortunately, TCA2 did not record viable data; blank miniseed file is excluded from the converted dataset. TCA1 was buried directly in soil, composed of a top layer of loose sand overlying uplifted marine terrace of Sisquoc formation shale. TCA3 was buried directly in a sand dune. 
-    Same four Gem 1.0 sensors as in deployments 1 and 2, sampling at 100 Hz. Placed again in bushes with no other physical noise filters.
-    Six Chaparral Physics C60 sensors with sampling frequency 400 Hz. Three of the six C60’s recorded data on one DiGOS DATA-CUBE3, other three recorded on another DiGOS DATA-CUBE3. Stations placed directly in bushes, with no other physical noise filters.

Deployment 4: October 17–23, 2023.
-    One broadband Trillium Compact 120s 3-component seismometers with sampling frequency 400 Hz, named TCA1. Data housed in “TC_mseed” subfolder. Co-located with principal infrasound station.
-    Same four Gem 1.0 sensors as in deployments 1 and 2, sampling at 100 Hz. Placed again within bushes.
-    Three Chaparral Physics C60 sensors with sampling frequency 400 Hz. Data recorded on one DiGOS DATA-CUBE3. Sensors were placed in bushes with no physical noise filters.
Sub-deployment 4: October 20, 2023.
-    Three Chaparral Physics C60 sensors with sampling frequency 400 Hz. Data recorded on one DiGOS DATA-CUBE3. Sensors were placed in bushes with no physical noise filters.
-    One Chaparral Physics C60 sensor with sampling frequency 400 Hz placed on the sand. Sensor had no physical noise filters.
-    One GoPro Hero10 mounted on a tripod.

Data is stored in miniSEED format, which can be loaded with several programming languages including the ObsPy library in Python. Infrasound data have been calibrated such that amplitudes are in Pascals [Pa]. Seismic data have been calibrated such that amplitudes are in meters per second [m/s]. Sensor locations are listed in metadata folder.

Anderson, J. F., Johnson, J. B., Bowman, D. C., and Ronan, T. J. (2017). The gem infrasound logger and custom-built instrumentation. Seismological Research Letters, 89(1):153–164.