Cross‐Correlation Analysis of Long‐Term Ambient Seismic‐Noise Recordings in the Caribbean Netherlands to Monitor the Volcanoes on Saba and St. Eustatius

The continuous recordings of broadband seismometers on Saba and St. Eustatius in the Lesser Antilles provide a unique and long data set to measure temporal seismic velocity variations (⁠dv/v) at two active but quiescent volcanoes (Mt. Scenery and The Quill). We compare results from single‐station cross‐component (SC) correlations with cross‐station cross‐component (CC) correlations and achieve the best similarities within the frequency band 1.3–2.1 Hz, with average correlations of 0.82 for Saba and 0.36 for St. Eustatius, justifying the use of SC as proxy for CC at these frequencies. Temporal dv/v variations derived from 13 yr of data show different characteristics at both islands. At St. Eustatius dv/v highly correlates (0.72) with air temperature and can be modeled by a simple sine wave with a period of 1 yr. Remaining residuals reveal cohurricane dv/v drops, thus at times of the passage of a hurricane. At Saba, subsurface velocity variations show temporal coseismic changes, up to −0.49% compared with −0.19% at St. Eustatius, and thus show a higher sensitivity to ground shaking. Our data set, although limited, shows a linear relation (correlation 0.78) between the coseismic dv/v drop and peak ground velocity at Saba around 1.3 Hz. We model the associated seismic velocity recovery with an exponential decay function and we estimate the recovery time at 2 yr. After subtracting the coseismic drop and recovery model, dv/v at Saba obtained from CC data correlates with the sine model (correlation 0.71). SC may be an appealing alternative for CC for monitoring purposes; however, the use of a small network is preferred to reduce the variance in
dv/v (at St. Eustatius from 0.12% to 0.05%) and to detect dv/v variations unrelated to volcanic activity (e.g., hurricane). We continue work on the implementation of CC in the daily monitoring for Mt. Scenery and The Quill.