{"id":"3d11d12f-e554-4d4c-ada0-853ecb5d7215","arxiv_id":"2506.22944","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"First demonstration of 3D spectral-element simulation of ultrasonic wave propagation through a dolphin head, using SPECFEM3D on a 16-million-element hexahedral mesh.","lead":"This paper shows that a spectral-element solver originally built for seismology can simulate 3D ultrasonic waves traveling through a CT-based model of a dolphin head. It is a feasibility milestone that could let researchers test dolphin echolocation and noise-impact hypotheses in silico.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reciprocity check in §3.6 only confirms discrete symmetry, not numerical accuracy; without a convergence study or independent comparison, the claim to have 'successfully simulated' 3D propagation is not yet established.","rationale":"The reader identified anatomical realism as the weakest assumption, and the paper's own Sec. 5 limitations support that concern. I partially agree: the omitted air spaces, teeth, attenuation, and tissue gradients are real threats to biological transfer. However, a more basic precondition for the central claim is numerical convergence of the discretized model. The only quantitative validation in the paper, the reciprocity test of Sec. 3.6, checks a symmetry of the discrete solution and would pass even if the solution were wrong by a large margin. The mesh quality statistics (minimum scaled Jacobian 0.0028), the low polynomial order N_GLL=3, and the absence of any refinement study leave the accuracy of the computed wavefields unverified. This does not mean the work is without value: the workflow is clearly described, SPECFEM is a well-established solver, the average mesh quality is good, and the paper is honestly framed as a feasibility study. But the strongest claim, that SEM has been demonstrated to successfully simulate 3D propagation through a dolphin head, needs a convergence or independent-comparison check before it can be fully accepted. Since the reader already assigned CONDITIONAL, the appropriate verdict remains UNCHANGED, with the condition made explicit: the numerical solution should be shown to be converged on the anatomical model, independent of the biological simplifications.","tokens_in":16494,"tokens_out":4963,"duration_ms":60211,"concrete_test":"Run the same 40 kHz plane-wave simulation at three resolutions: the current 2.5 mm mesh with N_GLL=3, a 5 mm mesh with N_GLL=5, and a 1.25 mm mesh with N_GLL=3. Compare pressure and particle-velocity time series at the left and right TPC receivers. If peak amplitude or first-arrival waveform differs by more than about 1 dB (or 5%) between the two finest runs, the solution is not converged. Additionally, re-run the current 2.5 mm case after removing or improving elements with scaled Jacobian below 0.1; if the TPC signals change materially, the reported results are controlled by mesh artifacts.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that SEM can be used to successfully simulate 3D wave propagation through a bottlenose dolphin head. The load-bearing precondition is that the discrete wavefield is an accurate solution of the modeled equations, and this is not checked. Section 3.6 validates reciprocity only, which is a symmetry property of the discretized operator; it would also hold on a severely under-resolved or element-distorted mesh and therefore cannot certify that the recorded TPC signals are converged. The mesh has a minimum scaled Jacobian of 0.0028, far below the recommended threshold of 0.2 cited from CUBIT documentation, and elements use only N_GLL=3 (quadratic shape functions). No grid-refinement or polynomial-order convergence study is reported, and the paper states in Sec. 5 that 'additional validation experiments' and direct comparison against other methods are deferred to future work. Thus the paper demonstrates that SPECFEM3D runs on this mesh, but not yet that the simulated wavefields are numerically trustworthy. The acknowledged anatomical simplifications (air spaces, teeth, attenuation, tissue gradients, Sec. 5) are a second concern that affects transfer to real dolphins, but the numerical-resolution gap affects even the simplified model and is therefore more load-bearing for the stated feasibility claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports the first 3D spectral-element (SEM) simulation of time-domain ultrasonic wave propagation in a CT-derived bottlenose dolphin head. The authors build a hexahedral mesh from segmented CT data using 3D Slicer, Blender, MeshLab, FreeCAD, and Coreform CUBIT/SCULPT, assign homogeneous acoustic properties to four tissue classes plus water, and run two SPECFEM3D simulations: a 40 kHz plane wave incident on the head and a point source near the phonic lips. Signals are recorded near the left and right tympano-periotic complexes, and a reciprocity test is used as numerical validation. The paper concludes that SEM is feasible and scalable for marine mammal bioacoustics, with detailed meshing commands provided for reproducibility.","tokens_in":16728,"tokens_out":5504,"duration_ms":60647,"significance":"If the numerical results are trustworthy, this is a useful and timely contribution: it is the first application of SEM to marine mammal head acoustics, it uses an open-source and reproducible software stack, and it demonstrates the complete pipeline from CT segmentation to high-performance time-domain wavefield simulation. The authors are appropriately transparent about anatomical simplifications, and the forward model uses no fitted parameters. The main reservation is that the only quantitative validation, a reciprocity check, is a symmetry test rather than a convergence test; the significance of the 'feasibility' claim therefore depends on whether the authors can demonstrate that the discrete solution is accurate, for example via h/p-refinement or comparison with an independent solver.","major_comments":[{"comment":"The reciprocity test in §3.6 is a symmetry property of the discretized operator and does not certify that the computed wavefields are accurate solutions of the continuous equations; a severely under-resolved mesh can satisfy reciprocity to machine precision while producing wrong amplitudes and phases. This matters because the central claim of the paper ('successfully simulate 3D wave propagation') rests on the numerical trustworthiness of the TPC signals in §4. The minimum scaled Jacobian of 0.0028 reported in Table 1 is far below the threshold of 0.2 cited from the CUBIT documentation, and the run uses N_GLL=3 with no h- or p-refinement study. I recommend adding a convergence test—for example a mesh-coarsening/refinement sequence or a comparison against an analytical solution in a simplified geometry (homogeneous sphere/cylinder) and, if feasible, against k-Wave or COMSOL on the same anatomical model—and reporting how the TPC signals change under refinement.","section":"§3.6, §4, §5"},{"comment":"The simulations are run at f0=40 kHz with three GLL points per element, while the motivating dolphin-click applications are at roughly 120 kHz. The statement in §5 that the 2.5 mm mesh supports up to 200 kHz using three points per wavelength is an extrapolation, not a demonstrated result; at high impedance contrasts and complex interfaces, three points per wavelength is likely to be insufficient for phase-accurate transmission and reflection. To support the claimed scalability to full-bandwidth biosonar frequencies, the authors should either run at least one higher-frequency case (with a corresponding resolution analysis) or explicitly limit the feasibility claim to the demonstrated frequency range.","section":"§3.4, §5"},{"comment":"The abstract and Section 5 state that SEM can successfully simulate wave propagation through a bottlenose dolphin head, but the model omits internal air spaces, teeth, attenuation, and tissue-property gradients, as acknowledged in §5. Air spaces in particular are known to be acoustically important (e.g., near the phonic lips and nasal passages), and their omission may change the wavefield substantially. Since these omissions are intentional and documented, the manuscript should qualify its central claim to 'a simplified CT-derived head model' rather than the anatomy of a dolphin head, or it should justify why these features are not needed for the feasibility conclusion.","section":"Abstract, §5"}],"minor_comments":[{"comment":"The text says the difference is 'four orders of magnitude (i.e. 40 dB)'; if the plotted quantity is particle velocity (an amplitude), a factor of 10^4 corresponds to 80 dB, not 40 dB. Please correct and specify whether the comparison is made on amplitude or power.","section":"§3.6, Fig. 6"},{"comment":"The description of the 1-mm shift to avoid coplanar faces is a bit unclear: after subtraction, how is the remaining 1-mm section removed and does this alter the head-water interface? Please clarify.","section":"§3.2"},{"comment":"The keyword 'SPECEFM3D' contains a typo and should be 'SPECFEM3D'.","section":"Keywords"},{"comment":"Please specify how the 'plane wave' is generated: the text says multiple point sources fired simultaneously, but the source-time function and amplitude tapering near the boundaries should be given as a formula or a reference; currently the reader cannot reproduce the source.","section":"§3.4, Fig. 5"},{"comment":"The SCULPT command table is useful, but a few entries are not self-explanatory (e.g., thicken_material, adapt_material); a short explanation or pointer to CUBIT documentation would help.","section":"Table 3"}],"recommendation":"major_revision","confidential_remarks":"This is a borderline major revision: the central feasibility result is plausible and the pipeline is valuable, but the absence of any convergence/verification step makes the 'successfully simulate' claim under-supported. The dB error and the frequency extrapolation should be fixed. If the authors add a convergence study, I would be comfortable with acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"For your radar: this is the first 3D spectral-element simulation of wave propagation through a dolphin head. That claim is real—the authors built a hexahedral mesh from CT scans and ran SPECFEM3D on it. The paper's main value is the detailed, step-by-step meshing pipeline (3D Slicer, Blender, MeshLab, FreeCAD, CUBIT's SCULPT). Anyone trying to do similar bioacoustic modeling will find that protocol genuinely useful. The authors also deserve credit for being upfront about simplifications: homogeneous tissue properties, no air spaces, teeth removed, old individual, and a source frequency (40 kHz) well below real echolocation clicks.\n\nThe soft spots are real, and the stress-test note is on target. The reciprocity check in §3.6 only confirms a discrete symmetry of the numerical operator; it would pass even on a badly under-resolved mesh. There is no convergence study, no comparison against an analytical solution, FEM, or k-Wave. The minimum scaled Jacobian of 0.0028 is far below the 0.2 threshold the authors themselves cite, and they use only three GLL points per element. That means the simulated wavefields could contain significant numerical artifacts. The paper states that validation against other methods is future work, which is honest, but it undercuts the word \"successfully\" in the central claim. What they have actually shown is that SPECFEM3D runs on a 16-million-element mesh of a dolphin head. That is a feasibility demonstration, not yet an accuracy validation.\n\nIs this worth refereeing? Yes. The meshing workflow is a real contribution, and the feasibility question—can SEM handle this geometry at all—is answered affirmatively. The quantitative results (e.g., spectral differences between left and right ear) should be treated as illustrative until convergence is demonstrated. A good referee would ask for at least one mesh-refinement study or a comparison against an established solver on a simplified geometry.\n\nThis paper is for bioacousticians and computational acousticians who want to move beyond 2D models. The data and code are not released, which limits reproducibility, but the workflow description is specific enough to follow. I would bring it to a reading group as an example of a careful feasibility study with honest limitations, and I would send it to peer review with the expectation of substantial revision.","headline":"Genuine first for 3D spectral-element modeling of a marine mammal head, with a practical meshing workflow, but the numerical validation is too thin to support the quantitative claims.","tokens_in":17303,"tokens_out":1866,"would_cite":false,"duration_ms":22294,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper shows that the spectral-element method can simulate 3D ultrasonic wave propagation through a bottlenose dolphin head, making high-frequency bioacoustic modeling scalable.","keywords":["spectral element method","dolphin bioacoustics","ultrasonic wave propagation","hexahedral mesh","SPECFEM3D","CT segmentation","Tursiops truncatus","marine mammal hearing"],"falsifier":"If a 120 kHz simulation that adds air-filled nasal passages, teeth, and tissue gradients produces wavefields at the tympano-periotic complex that differ from the simplified model by more than the model's numerical noise level, then the feasibility claim for real echolocation conditions would not hold.","tokens_in":16308,"feed_emoji":"🐬","tokens_out":4711,"duration_ms":47721,"temperature":0.7,"pith_summary":"The paper claims that the spectral-element method (SEM) can practically simulate 3D ultrasonic wave propagation through the complex anatomy of a marine mammal head, specifically a bottlenose dolphin. Using CT scans, the authors build a hexahedral mesh of a dolphin head with four tissue types, then run two SPECFEM3D simulations: an incoming plane wave at 40 kHz and a point source near the phonic lips. They argue SEM avoids the costly linear-system inversions that limit finite-element methods and converges exponentially, making high-frequency time-domain bioacoustics simulations feasible. The value of the claim is that it opens a scalable computational path to test hypotheses about echolocation, hearing, and noise pollution in marine mammals.","feed_headline":"Spectral-element method simulates ultrasound through dolphin head","feed_subtitle":"CT-based model shows it can handle complex anatomy at ultrasonic frequencies, opening new tests of dolphin echolocation.","key_machinery":"The spectral-element method itself: a high-order finite-element formulation using Gauss-Legendre-Lobatto points, whose global mass matrix is diagonal by construction, enabling explicit, matrix-inversion-free time stepping. The authors use SPECFEM3D, which couples acoustic (fluid) and elastic (solid) wave equations at interfaces. The mesh is generated with the SCULPT tool from a conformal, boolean-fragmented geometry, using adaptive hexahedral cells down to 0.1 mm at complex features. The machinery also includes the CT-to-STL-to-CUBIT pipeline with per-tissue homogeneous material parameters derived from Hounsfield units.","core_discovery":"On the paper's own terms, the central discovery is that SEM, implemented through SPECFEM3D, can successfully simulate 3D time-domain wave propagation through a real bottlenose dolphin head built from CT data. The authors demonstrate this with a 16-million-element hexahedral mesh with adaptive refinement at jaws and acoustic fats, a plane-wave simulation recorded near the tympano-periotic complex, and a point-source simulation of an outgoing click-like signal through the melon. They validate the numerical solver with a reciprocity test whose mismatch is 40 dB below signal level, and they show the mesh quality is acceptable (average scaled Jacobian 0.93). This is presented as the first use of SEM for marine mammal bioacoustics, and as a feasible alternative to 3D FEM models.","pith_inferences":["The paper's feasibility claim is conditional on the simplified anatomy; a natural extension would be to quantify how much the omitted air sacs and teeth change TPC wavefields at 120 kHz, which would test whether the simplified model suffices for echolocation studies.","The reciprocity validation suggests numerical error is low, but the lack of physical validation means the next decisive experiments are comparisons against measured sound fields in dolphin heads (e.g., in vitro hydrophone measurements).","The same SEM pipeline could be transferred to medical ultrasound applications such as transcranial focused ultrasound, where similar hexahedral meshing challenges arise, though that is beyond the paper's scope."],"forward_implications":["If SEM proves feasible, 3D time-domain simulations of dolphin biosonar at realistic click frequencies (around 120 kHz) become tractable on HPC clusters, where the authors estimate runs could complete in minutes with GPU acceleration.","The same pipeline can be applied to other marine mammal species and to other anatomical regions, enabling comparative studies of sound reception and emission.","The approach provides a numerical testbed for hypotheses about the roles of the melon, jaws, acoustic fats, and skull in beam formation and hearing, complementing existing 2D and FEM studies.","With future inclusion of air spaces, attenuation, and tissue gradients, the model can be used to study anthropogenic noise impacts on marine mammals."],"supporting_citations":[{"why":"Supplies the SPECFEM3D solver that the simulations run on.","marker":"[38]"},{"why":"Provides the SEM formulation and Gauss-Legendre-Lobatto basis used for the numerical machinery.","marker":"[20]"},{"why":"Underpins the fluid-elastic coupling used for water-tissue interfaces.","marker":"[21]"},{"why":"The recent 3D FEM model of a dolphin head that this study compares against and extends to SEM.","marker":"[6]"},{"why":"Earlier numerical model whose qualitative findings the plane-wave results are consistent with.","marker":"[5]"},{"why":"Prior SEM application to biological tissue (transcranial ultrasound), the closest precedent.","marker":"[27]"},{"why":"The SCULPT overlay-grid method that generates the conformal hexahedral mesh.","marker":"[47]"}],"fun_headline_variants":["First 3D spectral-element simulation of dolphin head ultrasound","CT-based spectral-element model simulates dolphin head acoustics","Spectral-element method opens new tests for dolphin echolocation","Spectral-element method enables ultrasonic simulation of dolphin head"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central claim rests on the assumption that a CT-derived, manually segmented head with homogeneous per-tissue acoustic properties and no internal air spaces is realistic enough that simulated wave propagation represents real dolphin anatomy.","fun_headline_variants_meta":{"raw":{"variants":["First 3D spectral-element simulation of dolphin head ultrasound","CT-based spectral-element model simulates dolphin head acoustics","Spectral-element method opens new tests for dolphin echolocation","Spectral-element method enables ultrasonic simulation of dolphin head"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001455,"raw_usage":{"total_tokens":5817,"prompt_tokens":867,"completion_tokens":4950,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":483,"completion_tokens_details":{"reasoning_tokens":4883}},"tokens_in":483,"tokens_out":4950,"duration_ms":35223,"temperature":1.0,"reasoning_tokens":4883,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T21:53:55.711476+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If a 120 kHz simulation that adds air-filled nasal passages, teeth, and tissue gradients produces wavefields at the tympano-periotic complex that differ from the simplified model by more than the model's numerical noise level, then the feasibility claim for real echolocation conditions would not hold.","supporting_citations":[{"cited_title":"Komatitsch, C","cited_arxiv_id":null,"evidence_quote":"Underpins the fluid-elastic coupling used for water-tissue interfaces."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The recent 3D FEM model of a dolphin head that this study compares against and extends to SEM."},{"cited_title":"5 (2021) 3","cited_arxiv_id":null,"evidence_quote":"Earlier numerical model whose qualitative findings the plane-wave results are consistent with."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The SCULPT overlay-grid method that generates the conformal hexahedral mesh."}],"review_version":1}