the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Jun 2026
High Spatial and temporal 3D Sea Ice Thickness Reconstruction Using Seismic Waveform Inversion and Tomography
Abstract. Sea ice is a porous, two-phase material whose evolution influences climate, ecosystems, and human activities in polar regions. Rapid declines in Arctic sea ice thickness highlight the need for monitoring approaches that resolve small-scale spatial variability, beyond the coarse resolution of satellite-derived products. Seismic methods, based on the analysis of seismic waves guided in the ice layer, provide high-resolution local estimates of ice thickness, but have so far been limited to individual source–receiver paths. Here, we present a methodology to generate maps of sea ice thickness from seismic data. We analyze two complementary datasets acquired on fast ice: controlled-source data recorded by a 16-geophone array in the St. Lawrence Estuary (Canada), and passive data collected by a dense 247-geophone array in Svalbard (Norway). Ice thickness is first estimated along individual paths through waveform inversion based on spectral element modeling. These path-specific estimates are then integrated within a tomographic framework to reconstruct spatially continuous sea ice thickness maps. Our results demonstrate that both active and passive seismic measurements can resolve spatial heterogeneity in ice thickness at high resolution. This seismic-tomographic approach provides a practical and scalable tool for fine-scale monitoring of sea ice in rapidly changing polar environments.
- Preprint
(9449 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-2519', Anonymous Referee #1, 22 Jul 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-2519/egusphere-2026-2519-RC1-supplement.pdfCitation: https://doi.org/
10.5194/egusphere-2026-2519-RC1 -
RC2: 'Comment on egusphere-2026-2519', Anonymous Referee #2, 09 Oct 2026
Dear Editor, dear authors,
Please consider my comments to the manuscript High Spatial and temporal 3D Sea Ice Thickness Reconstruction Using Seismic Waveform Inversion and Tomography submitted by Hooshmand Zandi et al.
The authors present an imaging study of sea ice. They apply concepts from seismology to two datasets collected in Canada and Svalbard/Norway to constrain space-time patterns of ice thickness. It is certainly one of the first if not the first application of a 2D ice thickness tomography constructed from dispersive vertical component flexural waves reported in the literature, which highlights the significance of the submitted work. The team uses controlled-source or hammer-shot data as well as natural icequake records for the inversion at the two sites. The use of the Spetral Element Method to create complex dispersive signals for the seismic waveform inversion of the water/ice system must be highlighted. The resulting images or distributions are compared against independently obtained thickness data, and the similarities and discrepancies are used to assess the inversion results and to discuss potential sources of bias.
I limit my comments to some general aspects, and I do not provide a line-by-line commentary of style and language that can be improved in places (e.g., lines 220 – 224 still contain place holders for values).
I am concerned with the tomography approach. I would appreciate a clarification about the connection between the physics of wave propagation of flexural modes guided in an ice layer with constant and with variable thickness, and the spatial discretization of the inverse problem. The authors write they adopt an approach that is similar to or like tomography:
Line 84: proposing 2D maps of the ice thickness, using a tomography-like approach
Line 85: a regionalization strategy similar to what is used for travel-time tomography
Line 250: a method similar to what is used for travel-time tomography with straight rays
Line 304: using a tomography-like approachBecause it is a pioneering ice thickness tomography study I think the presentation would benefit from spelling out better in what aspects the approach is the same as seismic travel time tomography, which aspects are similar, and how it is different. I understand that the pre-computed database of synthetic dispersive flexural waveforms is based on a model of the water/ice system with a uniform ice thickness along the source/reiver pair (step 1 of the algorithm). That uniform ice thickness is then estimated for each source-receiver pair using waveform inversion (step 4), which is then used in the tomographic inversion to estimate the spatial variability in thickness (step 5).
Considering the observed dispersive waveforms are a result of the integrated propagation I wonder what the assumptions about the thickness-variability and the waveform characterstics are? Does the linearity of Eq 7 that we assume for travel times also hold for ice thickness obtained from fitting dispersive flexural waves? Say the inversion yields a spatially variable thickness profile for a source-receiver pair, is the associated waveform propagating along that inferred profile the same as the signal propagating along a homogeneous slab with an ‘equivalent’ average or effective thickness? What is this equivalence? This aspect should at least be demonstrated with a synthetic waveform modelling example or otherwise confirmed, or the limits of this assumption should be discussed.
This then affects the resolution of the spatial variability, which is another term or concept that should be discussed in more depth. What is the minium length scale over which a thickness variation can be resolved? As in seismic travel time tomography, a connection to the involved wavelength(s) should exist, so I assume reporting the wavelength range of the dispersive signals is also relevant for the discussion. Regrettably, the authors mention checkerboard tests (lines 294 – 300), but I can not see checkerboard test results in the submitted material, and this is also not referred to in the assessment of the inversion results shown in Figure A-1. I caution that a checkerboard test can be useful to analyze data coverage and resolution in the straight ray framework, but it does not allow an assessment of the fundamental resolution issue reported above, i.e., if the physics of wave propagation of flexural modes guided in an ice layer with constant and with variable thickness is compatible with the spatial discretization of the inverse problem.
Considering the phenomenon of cycle-skipping, the reader can be interested to learn about preprocessing steps or choices such as windowing and tapering before evaluating the correlation misfit between observations and synthetics (Eq 5). Related to the quality of the waveform inversion, the manuscript can perhaps benefit from additional comments about the suitability of 2D synthetics applied to waveforms observed in a 3D medium.
I would also welcome a better discussion of the obtained tomography results shown in Figures 12, 13, and A-1. The character of the images is understandably streaky. What should a reader consider to avoid this artefact in future deployments? Were the stations deployed with doing a tomography in mind? Is the icequake distribution exclusively along the shore a surprise, or was this expected? It appears that the blue-ish thick ice zone correlates very well with the epicenter distribution; can this be an artifact? Is this a limit of the resolution? Related, for example, we see a small area of intermittently increased ice thickness at about 16.78 lat / 77.8840 lon on 11 and 12 March (Figs 13, A-1). Figure 12 shows that in this time period icequakes occur at that location, but not before and not after. The authors should better qualify which of the features shown in the distributions are robust and which can be likely or possibly artifacts. I understand the motivation to provide time-lapse images, and the associated average temporal evolution shown in Figure 14 is valuable and interesting. However, if the quality of the spatial variability is too difficult to constrain (which again relates to the resolution), it may be considered to leave these results out of the manuscript and provide instead an average tomography result or 2D change images associated with longer time intervals and hence better data coverage.
Citation: https://doi.org/10.5194/egusphere-2026-2519-RC2
Data sets
BicWin25: multi-instrument datasets for studying marginal ice zone dynamics and wave-ice interactions PERRARD, Stephane; EDDI, Antonin; KUCHLY, Sébastien; ZANCHI, Vasco; MOKUS, Nicolas; DANSEREAU, Véronique; MOREAU, Ludovic; DUMONT, Dany; SMITH, Madison M https://doi.org/10.57745/D3CJYO
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 256 | 157 | 47 | 460 | 29 | 29 |
- HTML: 256
- PDF: 157
- XML: 47
- Total: 460
- BibTeX: 29
- EndNote: 29
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
