Analysis of relationship between strain and atmospheric pressure data at Stromboli volcano

Strain data recorded by Sacks-Evertson strainmeters, due to the high dynamic of the instrument and since its output responds to input over a wide frequency range, are prone to be affected by anthropic noise, changes in atmospheric pressure, tides, rainfall, underground water movements, changes in underground temperature, earthquakes, as well as other crustal movements. Several kinds of procedures have been developed over time by geophysicists to remove the unwanted (“spurious”) signals from the actual recordings, in order to thereby obtain cleaner strain data, capable of representing the actual changes of the local strain in proximity to the installation site. The clearly most dominant signals in a strain data time series are associated with Earth tides and atmospheric pressure loading. Earth tides, due to the relative motion of Sun and Moon with respect to Earth, account for 10⁻¹⁰ strain over a frequency range of 10⁻⁴–10⁻⁵Hz (periods of hours to days), and are induced by periodic, measurable forces: this allows a reproducibility of the phenomenon using numerical simulations software. On the other hand, atmospheric pressure, for its own characteristics, is a highly variable signal, spanning over extremely wide strain- and frequency-ranges. Both signals, however, are characterized by frequencies comparable with those of interest. One of the most successful methods to remove tides and atmospheric pressure uses a combination of harmonic and non-harmonic techniques, through the implementation of Bayesian statistics. The software assumes that a given signal can be decomposed into a tidal component, a trend term, a perturbation due to an external source, the atmospheric pressure, responsible for generating a change in the recorded signal, and some random noise superimposed.

Barometric admittance quantifies how rock/soil strains to atmospheric pressure changes, often modeled linearly but non-linearities arise from complex subsurface media (aquifers, faults, cracks), requiring advanced techniques like neural networks or state-space models to capture frequency-dependent responses, revealing aquifer properties, fault activity, or seismic precursors, with higher frequencies showing local effects and lower frequencies reflecting regional pathways, indicating that strain varies nonlinearly with pressure due to medium heterogeneities.

The data recorded by a Sacks-Evertson strainmeter installed at Stromboli volcano show a non-linear relationship between barometric pressure and strain variations for lower frequencies: an empirical mode decomposition has been used considering the frequency dependent characteristics of the pressure response and the borehole strain observation data, and the pressure observation curve of synchronous observation are decomposed, obtaining the frequency dependent pressure response coefficient, realizing the refined pressure correction of borehole observation data.

In the higher frequency range, when the medium shows an elastic response related to pressure changes, a linear regression model in the time domain has been carried out to highlight volcanic-related strain changes.

This analysis could improve the volcanic hazard assessment of strain data related to open-conduit volcanoes, such as Stromboli, during unrest phases.

Data used contains valuable information for scientific community and are made available on the EPOS data portal. Attention is taken into metadata handling and intelligent management of distributed resources.