ICEWAVEGUIDE

In Geophysics, seismic waves have been used to image the Earth for decades. In particular, the emergence of “noise seismology” developed for the last decade in the PI’s team at ISTerre (Grenoble, France), proved a major breakthrough in seismology. This approach relies on passive recordings of ambient seismic noise to reconstruct the Green’s function of seismic waves propagating between sensors via basic signal processing. Based on this approach, multi-scale images of the Earth (Crust and deep Earth) can be obtained with impressive resolution (Campillo and Roux, 2014). It is therefore natural to contemplate the use of seismic waves to investigate sea ice. Recent works aimed at estimating thickness from the dispersion of flexural waves propagating in the ice (Marsan et al. 2012, Sutherland and Rabault, 2016). These waves, also known as bending waves, have the advantage to be sensitive to the thickness and mechanical properties of the structure where they propagate. However, the drawback is that they are encountered only at very low frequencies ; hence their wavelength is very large and difficult to sample spatially. Dispersion is therefore measured based on a time-frequency analysis, which is known to lack accuracy, especially when wave packets are not separated in time.

Based on improved methodology and signal processing, the present project aims at taking the above approach a step further by measuring waves of the same nature as flexural waves, but at higher frequency. The waves of interest here are seismic waves guided in the ice layer with energy leakage in water. They are analogous to leaky Lamb waves propagating in an immersed plate, so there is a one-to-one relationship between their dispersion and ice thickness and mechanical properties. With wavelengths typically between a few tens to a few hundred meters (at the scales of sea ice), spatial sampling becomes possible, provided an array of seismic sensors can be deployed.

Less than a decade ago, this project would have been impossible in terms of costs, because the deployment of such arrays was still inconceivably expensive and left to prospection geophysics experiments funded by the oil & gas industries for surface wave tomography. This is one of the reasons why, so far, the contribution of geophysics to environmental glaciology in general, and more specifically to sea ice thickness measurements, has remained marginal. Fortunately, recent technological progress allowed a new generation of wireless, autonomous geophones to be designed, with improved memory and battery capacity, which makes the deployment of dense arrays with hundreds of seismometers possible, at a reasonable cost ( $1.5/sensor/day). Typically, these sensors can record continuously in extreme cold conditions (down to -40°C) for a whole month. By deploying an array of these geophones on the Vallunden Lake during winter, we will be able to measure, for the first time, the dispersion curves of leaky Lamb modes propagating in the ice and to convert them into a measure of ice thickness and mechanical resistance.

Compared to already existing methods, the advantages of this new methodology are twofold.

1. In addition to accurate thickness measurements, important information related to the mechanical resistance of the ice will be measured : density, elastic constants, anisotropy and also damage level in the ice.
2. From a practical point of view, no active sources are required, because wave propagation can be extracted from recordings of ambient noise via basic signal processing (Campillo and Paul, 2003). Hence, geophones will be installed and left to record ambient noise for about a month, thus allowing variations in the ice thickness and mechanical properties to be monitored for the duration of the experiment.

This methodology was recently tested at the lab scale in a cold room where a 1m-diameter x 0.5m-high water tank was placed in order to grow an ice layer at the water surface. 32 accelerometers were aligned with a 1 cm spatial sampling to record ultrasonic GW generated by a piezoelectric source (Moreau et al. 2017). Note that although an active source was used in this experiment, it is also possible to extract the dispersion curves from passive noise measurements, provided noise sources operate in the appropriate bandwidth. This was demonstrated in an aluminum plate by using an air jet to reproduce wind-induced turbulence (Larose et al. 2007).

Time-space data were then transformed to the frequency-wavenumber domain so that the dispersion branches of the propagating modes could be identified. Generally, this is achieved via a classical spatio-temporal Fourier transform, but in this case signal-to-noise is degraded due to energy leakage in water, thus resulting in poor frequency-wavenumber data. Instead, specific signal processing based on past research by the project initiator was applied (Moreau et al, 2013). In the field of medical acoustics, this method is the object of the following patent : L. Moreau, J-G. Minonzio, J. Foiret, M. Talmant and P. Laugier, Procédé et dispositif ultrasonores pour représenter la propagation d’ondes ultrasonores dans un guide d’épaisseur linéairement variable, patent n°1357204 (22/07/2013).

Finally, a forward model that computes the dispersion branches of GW in a floating isotropic plate was used to solve the inverse problem. The inversion consisted in finding the plate thickness ; longitudinal and transverse bulk wave velocities that minimize the error between experimental and synthetic dispersion curves. Ice thickness was estimated to be 21.2 mm, i.e. less than 1% error compared to the actual thickness of 21 mm measured by breaking the ice after data acquisition. Longitudinal and shear bulk waves velocities were estimated to respectively. This is consistent with values found in literature for the fresh-water ice elastic properties (Gammon et al., 1983). This procedure was repeated several times, and ice thickness was successfully evaluated with consistent values for bulk waves velocities.

At the ultrasound scale, only the normal displacement is accessible, which considerably limits the measurement of the GW modes at frequencies where they produce a negligible vertical displacement. figure 1 (next section). In the field experiment, both 1- and 3-components geophones will be used. Access to the 3 displacement components allows GW modes (including the shear-horizontal mode) to be measured in the full bandwidth, thus revealing missing dispersion information.

The experiment at Vallunden Lake is an exciting first step towards the application of these concepts to the sea ice cover at larger scales. If successful, it is expected that the methodology introduced in this proposal will open current locks to provide new observables about sea ice resilience, while also establishing geophysics as a complementary approach to radar and sonar for sea ice thickness measurements. This nature-wide experiment will therefore represent a benchmark for future field applications within the ice pack.