Current answer
AIS traffic characterization is established. The expected energy sign appears in 3/4 preselected H01W comparison pairs, while 0/4 pass the complete directional check. The present evidence therefore stops before ship-specific attribution or out-of-sample prediction.
Inside 200 km of H01W, AMSA contributes 159,925 raw AIS records. Of these, 97,549 support SL and fast TL; 62,376 remain traffic-mapping records because source-input or quality-control rules fail. The 56 kn exclusion is a different record; Igor remains source-model eligible.
The frozen screening contrast is positive at both 63 and 80 Hz in 3/4 preselected higher/lower pairs. Those modeled values are screening estimates, not accepted RAM one-to-one results; the sample is small and selected in advance.
Spatial products describe cross-hydrophone consistency. They do not establish a ship direction.
Modeled SL, TL, and predicted individual RL do not identify a measured ship contribution on their own.
The declared propagation and observation branches are not yet complete enough for an independent predictive test.
Inspect one source spectrum or compare distributions across AIS sets.
Change ship class, length, speed, and frequency while keeping the output as source spectral density.
Choose any named AIS set and inspect SL, fast TL, or predicted RL for one of 102,629 AIS records across 5–100 Hz.
Compare SL distributions for all AIS, AMSA, ARA RAM, AMSA RAM, AMSA non-RAM, Juan 674, and the Highlighted 21.
Use it whenever a validated direct-path result exists.
One source position to the receiver geometry defined by its result family, without angular interpolation.
Browse 1,014 results across two scientifically separate families.
Use it where validated angular interpolation covers the path.
The browser exposes each declared receiver-centred key and its current state. Available sweeps open as geographic fields and selected TL-by-range traces.
Use it for quick screening, not as a RAM result.
Browse the range-only estimate used for all 102,629 AIS records.
The daily sweep applies to the bounded 2026 March 31 and April 1 ARA group. The fast TL method serves its designated screening population; it does not fill a RAM gap.
Compare the screening TL distributions for the seven AIS record sets.
See which declared set-level TL fields exist, which have partial downstream RL, and which still require direct paths.
20 records × 11 frequencies: 220 values at 6 m and 198 additional values at 70% draught for the 18 records that have that branch.
The selector above includes all 21 highlighted paths. All have results at 6 m; 19 also have results at 70% draught: 440 values in total. Igor contributes 11 values at 6 m and 11 at 6.16 m from an inferred repeated same-transit 8.8 m draught.
All 102,629 AIS records have SL and predicted RL from one vessel evaluated at all 96 integer frequencies from 5 through 100 Hz.
Open the AIS acoustic explorer →Inspect SL, RAM TL, predicted RL, and measured SPL for the complete Highlighted 21, including Igor’s isolated record.
Open the Highlighted explorer →Compare all-AIS screening distributions, RAM distributions for the Highlighted 21, and each Highlighted record across frequency.
See what waveform support exists, which response route is admissible, and where coverage stops.
A small mark means one or more bounded waveform intervals start in that month. It never means continuous month-long waveform coverage.
The 217 available hydrophone product families include guarded 5–100 Hz nominal pressure, summaries, periodograms, spectrograms, sliding Welch PSD, and tonal tables. Twenty requested hydrophone intervals returned no data, 20 H08N intervals lack a usable full-complex response, and one interval lacks context coverage.
Real gaps remain masks. Processing never bridges them for timing, phase, coherence, or beamforming.
All 16 active full-response bindings pass the rounding-aware 10 Hz check. This establishes nominal-response standing, not traceable calibration.
Inspect 32 hydrophone windows, complete and blank hours, band summaries, and tone candidates without treating them as ship detections.
See coverage, RL distributions, anomaly distributions, and spatial standing without collapsing stations or hydrophone arrays together.
Summarize the available waveform products across stations, hydrophones, hydrophone arrays, and target windows.
Each view keeps exact UTC support, hydrophone membership, response basis, gap masks, and AIS arrival markers visible. Arrival markers are timing context, not detections.
Inspect time-frequency energy on the available waveform support.
Inspect sliding power spectral density with the same gaps and response basis.
Compare guarded standard-band summaries without filling missing support.
Lists candidates for hydrophones and hydrophone arrays while preserving their distinct identities.
These views compare hydrophones within one array. They describe consistency, not ship identity.
Inspect 79 magnitude families for hydrophone arrays, formed by averaging aligned hydrophone power in linear units. At station HA03, only phase-invariant magnitude, coherence, and matrix energy are available because inter-hydrophone timing is not established.
Inspect 63 full-complex bundles where timing support allows phase and lag views. These do not establish a bearing.
Summary views remain separate from the underlying recording intervals and keep support limits explicit.
Shows 35 available spectra for hydrophone arrays; 44 hydrophone array intervals do not meet the support threshold.
One hundred thirty-three hydrophone and 48 hydrophone array targets pass their support and stability checks after excluding their own target hour.
One hundred twenty-five hydrophone and 31 hydrophone array targets remain insufficient after the allowed fourth baseline block. No fifth block is permitted.
This is observation-method QA on identical waveform support, not prediction validation.
Across six H01W passages and four subjects per passage, the median offset-adjusted correlation is 0.784. After removing broad frequency structure, the median residual correlation is 0.445 and the median absolute residual difference is 3.0 dB.
Comparison can only use dates and stations where the relevant AIS, propagation, and waveform inputs coexist.
A small mark means one or more bounded waveform intervals start in that month. It never means continuous month-long waveform coverage.
National monthly source files used to derive the H01W 200 km selection.
Inspect the measured and modeled distributions before opening one pair.
Show the measured whole soundscape, modeled contribution from one vessel, and measured-minus-modeled residuals at every nominal band. The 21 ship observations use 20 distinct one-hour waveform windows.
Keep modeled screening estimates, measured contrasts, and directional checks distinct.
Select each higher/lower pair, including the reversed 80 Hz result and the directionally disfavoured case.
The frozen cohort carries screening estimates. They are not accepted RAM one-to-one comparison values.
Hydrophone array and hydrophone contrasts are reported separately, with energy and direction kept apart.
Current ceiling
Declared work
25 groups at this cutoff.
88,565 of 150,940 records with predicted RL at both frequencies (58.7%)
The screen uses records with a mapped vessel class, a positive speed no greater than 50 knots, and a positive length. It does not clip or fill missing values.
220 values: 20 vessels × 11 decidecade centres
The JOMOPANS-ECHO decidecade source levels include the model identity and units.
Four 674-row tables
Nearest-vessel, fast TL, selected-row, and exploratory project-fast rankings use the same complete membership accounting.
102,629 spectra; 1,128,919 standard-centre values
All 102,629 AIS records are represented through reusable source-input profiles, with dense 10–100 Hz values for interactive display. RAM is an additional propagation treatment, not an admission rule for the source model.
102,629 records × 96 directly evaluated spectral samples
The explorer shows SL and predicted RL from one vessel at every integer frequency from 5 through 100 Hz for all 102,629 AIS records: 88,565 non-RAM AIS records and 14,064 RAM AIS records. The canonical materialization also preserves 14 standard-centre band values outside the served three-tab browser. The 62,376 excluded raw AIS records remain accounted for by source-input or quality-control reason. Values below 20 Hz are marked as higher-uncertainty source-model extensions.
294 RAM one-to-one results
Twenty Juan cases at 14 frequencies plus Igor's 14-frequency 26 March case are sealed and validated.
24 month-frequency runs (12 months × 2 Tier-1 frequencies)
Twelve selected months at 63 and 80 Hz are running on Pazuzu with 88 workers.
56 runs across two hydrophone arrays, two dates and 14 frequencies
The 31 March Tier-2 campaign is nearly complete on Mefisto after Belfegor finished its partition. The 1 April daily ocean field is available: H10N 63 Hz is running on Pazuzu, while the disjoint H10N 80 Hz and H10S 63/80 Hz partition runs on Belfegor. The remaining 1 April Tier-2 frequencies follow the active Tier-1 work.
144 month-frequency runs (12 months × 12 Tier-2 frequencies)
Twelve selected months at the other 12 mandatory nominal centres follow the 63/80 Hz tier.
418 values: 220 at 6 m and 198 at 70% draught
Every row embeds the modeled source level, RAM one-to-one TL, predicted individual RL, units, identities, and evaluation method used by the view.
22 values: 11 at 6 m and 11 at 70% draught (6.16 m)
The package combines modeled JOMOPANS source levels with validated RAM one-to-one TL without rerunning propagation. The 70% draught branch uses the admitted repeated same-transit 8.8 m inferred draught and remains distinct from the 6 m branch.
Membership known; values not yet complete
Every displayed materialized TL field has an explicit downstream standing. 154 TL/RL branches for individual vessels are materialized, including 9 supported by available rays in a sweep field that is still materializing; 9 displayed keys have audited source members that require direct RAM paths, 5 keys end at TL by declaration, and 4 legacy TL-only keys remain outside current source-aware admission. No displayed field is awaiting downstream admission from this refresh.
16 of 16 active response bindings passed; 2 response members are not applicable because their waveform requests returned authoritative no-data
The rounding-aware comparison admits every active full-response binding without recomputation. It establishes T1 nominal-response standing only, not traceable calibration.
Hydrophone first-product families
The sealed full-active package includes all runnable exact and accepted-gap holdings. No acquisition chunk remains missing or pending; no-data, context-coverage, and response limits remain explicit.
20 requested hydrophone intervals returned no data; 20 H08N response holds; 1 context-coverage hold
Acquisition has no missing or pending chunks. These remaining scientific and response limits are not silently filled, bridged, or converted into empty products.
217 hydrophone first-product families
Each bundle includes guarded 5–100 Hz nominal pressure, summary, periodograms, spectrogram, sliding Welch PSD, and tones.
Held windows and complete hydrophone-hours
Every usable holding has current guarded products. The 97 blank hours and 1,067 blank band rows remain explicit gap or response-edge evidence; four no-data siblings remain unavailable.
79 magnitude families for hydrophone arrays + 63 full-complex spatial bundles + 16 HA03 magnitude-only bundles
Combined products average aligned hydrophone power in linear units. Full-complex bundles preserve phase-sensitive structure where the response permits it; HA03 keeps only phase-invariant magnitude and matrix-energy quantities.
35 percentile products for hydrophone arrays
Only hydrophone array recording intervals meeting the declared support threshold are shown; missing support is never converted to zero.
44 hydrophone array products
The declared support threshold is not met. Missing support is not converted to a zero or an empty plot.
133 hydrophone + 48 hydrophone array targets
Each target-specific baseline passes the declared support and stability checks after excluding its own target hour.
125 hydrophone + 31 hydrophone array targets
Their target-specific baselines remain insufficient after the allowed fourth block; policy permits no fifth block.
63 full-complex bundles; HA03 phase and lag remain unavailable
Phase and GCC-PHAT products are available for H01W, H08S, H10N, and H10S recording intervals whose response supports timing. HA03 has magnitude-only products and cannot fill this role.
6 passages × 4 subjects; 24 unpooled comparisons
Periodic-Hann and five-taper DPSS estimates use identical samples, masks, windows, and frequency bins. The result measures estimator sensitivity, not source attribution.
Reader page, technical companion, evidence and status indexes
This package closes required report dependencies at one evidence cutoff and may be republished whenever requested or materially useful; identical inputs may remain a no-op.
Current views
Inspect every pair, including the 80 Hz energy reversal and the directionally disfavoured case. The view distinguishes the frozen cohort's RAM-labelled screening estimates, their limited standing, a separate fast TL screen for the same AIS record, and the measured hydrophone array and hydrophone contrasts.
Compare periodic-Hann and five-taper DPSS estimates across six passages and four subjects on identical usable waveform support.
Browse the sealed 217 hydrophone and 79 hydrophone array first-product families while preserving waiting, no-data, gap, and response-limited states.
Inspect all 63 full-complex spatial bundles and all 16 HA03 magnitude-only bundles. Coherence and matrix energy do not establish ship attribution.
Choose any named AIS set and inspect source spectra, fast TL, or predicted received level for all 102,629 AIS records across the full 5–100 Hz source grid.
Browse SL, the record's range-only fast TL screening value, and predicted RL from one vessel for all 102,629 AIS records at 1 Hz spacing from 5 through 100 Hz.
Inspect 32 hydrophone windows, complete and blank hours, band summaries, 9,192 hydrophone tone candidates, and 1,907 multi-hydrophone tone families. This view does not claim ship detection or attribution.
For the complete Highlighted 21, inspect spectral SL with source-model scatter, direct-path RAM TL, predicted individual-vessel RL, and the measured whole-soundscape SPL. The measured curve is context: it is not a detection, attribution, or propagation validation.
Inspect all 294 RAM one-to-one TL rows across the 21 highlighted paths, together with fast TL context, source-depth availability, and evidence identities.
A reader and technical package for the sealed 720-case, 63/80 Hz fixed-versus-geology-informed replay at the H01W1, H01W2, and H01W3 hydrophones.
Historical reports
Useful for what was known at that cutoff; it is not updated by this report.
Shows path-specific RAM sensitivity to monthly, daily, and instantaneous PT6H representations.
Compares completed fixed-bottom and geology-informed RAM results without treating either as the physical truth.
The Juan AIS comparison and H10 cases; four source-ineligible Juan raw records remain only as TL-only provenance, and current desired RAM uses geology-informed bottoms.
A broad preserved record of exploratory methods and findings.
Other analyses
Recreates the collaborators' intended plots from available project data and compares timing and morphology.
A sealed 720-case, 63/80 Hz direct-path replay across the H01W1, H01W2, and H01W3 hydrophones.
A recurrence study kept outside the current Highlighted 21 coverage.
Exploratory comparison work preserved separately from current authoritative methods.
A focused audit of a control interpretation, not current desired-state fulfillment.
Preserved diagnostic evidence for seabed-transition and coherent-field behavior.
A selection-method exercise kept as supporting evidence.