Umbrella input set
AIS
set.aisEvery propagation-eligible AIS fix currently admitted to the project.
How the study is assembled
Start with named input sets. Then define cases, cohorts, and coverage
The same ship row can support a quick ranking, an exact propagation case, or a waveform comparison. Those are different results because they bind different inputs and scientific definitions. This page names those pieces before showing how the current cohorts use them.
1Input set
A named set whose members have one type.
2Case
One complete assignment to every input role.
3Cohort
A selected set of cases of one case type.
4Coverage
The results that should exist for that cohort.
5Artifacts
Files that represent the materialized results.
Named input sets
AIS is a declared union, not a synonym for all traffic data
The current AIS family contains AMSA and ARA. GFW presence and Juan's reconstructed positions remain separate. The distinction determines which members can supply exact propagation geometry.
AIS set hierarchy
Provider and time sets describe origin. Station sets cut across providers. Both are named subsets of the same AIS union.
Provider family
AMSA
set.ais.amsaUnion of 19 monthly HA01/H01W 200 km sets. AMSA contributes to HA01 only.
Provider family
ARA
set.ais.araUnion of ARA historical and ARA current, with overlapping station slices.
set.ais.ha01set.ais.ha03set.ais.ha08set.ais.ha10AMSA month sets
Each month is a named child set. Row count is descriptive; it does not create “large” or “small” policy categories.
January 2026 is the largest available local extract at 13,321 rows, followed by February 2026 and February 2025. That ordering is descriptive only.
AIS children and separate traffic sets
Each name fixes membership, dates, and limits. An unqualified AMSA or ARA reference means its declared union; narrower work names the month, temporal family, or station slice.
Named input setMember and scopeUse and boundary
AIS provider family
AMSAChild of AIS; union of 19 named month setsset.ais.amsaMember. One anonymous AMSA position within 200 km of the centroid of the H01W hydrophone array.
Coverage. 159,925 positions from January 2025 through July 2026. The inclusive boundary is 200 km around the centroid of the H01W hydrophone array.
Used for. Propagation screening for HA01 and its H01W hydrophone array, Juan’s selection, and the January ship-type lane.
Boundary. CRAFT_ID is anonymous and month-local. Reports are thinned to at least 60 minutes. Sixteen source months are readable prefixes; only February 2025, June 2026, and July 2026 are structurally complete.
Fixed raw AIS subset
Juan's raw H01W selectionSubset of AMSAset.ais.amsa.juan_678.v1Member. One anonymous AMSA report selected from the H01W 200 km set.
Coverage. 678 raw positions in 233 UTC clock hours from January 2025 through July 2026. Four do not have the inputs needed by the published source model.
Used for. Early source-input filtering and provenance only. Comparable analyses use the 674 eligible records below.
Boundary. This fixed collaborator selection is not itself an analysis population, a new traffic census, or a reconstructed voyage.
AIS subset
Juan 674 AIS recordsSource-model-eligible members of Juan's raw H01W selectionMember. One anonymous AMSA record in the H01W 200 km set with ship class, same-record speed, and length.
Coverage. 674 records after the four source-ineligible raw positions are filtered out.
Used for. Juan ranking views, comparable source and screening calculations, the four-pair waveform follow-up, and source-depth sensitivity.
Boundary. The four excluded raw positions remain as provenance only and do not appear as a second analysis population.
Accepted ranked AIS subset
Juan's current top-20 execution setTwenty-member subset of Juan 674 AIS recordsset.ais.amsa.juan_top20_temporary.v1Member. One anonymous AMSA report in the H01W 200 km set with a stable corrected JOMOPANS-minus-TL-lite rank.
Coverage. 20 positions. Ranks 1–10 form priority tranche 1; ranks 11–20 form priority tranche 2.
Used for. Required exact-path propagation and waveform-product coverage. The tranches control execution order only; every row remains active.
Boundary. A directory or active computation is not a result. Each product remains waiting until a sealed receipt is bound in the observed catalog. Legacy full-raw TL-only work remains historical provenance and does not define this set.
AIS temporal family
ARA historicalChild of ARAset.ais.ara.historicalMember. One named-vessel position from a bounded HA10 regional export.
Coverage. 2,134 delivered position records on 2026 March 31 and April 1. Some exports repeat the same source data.
Used for. Named-vessel geometry at HA10 and a bounded daily-sweep exception. Each represented date requires reusable fields for the H10N and H10S hydrophone arrays under the general sweep profile.
Boundary. The two-date exception does not extend to AMSA, ARA current, or selected one-to-one paths. The exports remain irregular, duplicated snapshots rather than continuous tracks.
AIS temporal family
ARA currentChild of ARAset.ais.ara.currentMember. One exact named-vessel position from a station-area list.
Coverage. 11,883 distinct positions from 2026 August 11 through 25. Overlapping lists contain 13,003 occurrences because an unchanged position can appear again.
Used for. August traffic context, exact named-case geometry, and the contemporary ARA share of the common 63 Hz screen.
Boundary. Capture times are irregular, schemas change, rows may be carried forward, and some station labels are wrong. Coordinates determine station association.
Versioned additive AIS subset
AIS highlighted candidates v1Subset of AIS; independent of frozen cohortsset.ais.highlighted.v1Member. One AMSA or ARA fix explicitly highlighted by Juan, Igor, or Rui, with one associated IMS station, selector, UTC date, and reason.
Coverage. The set grows by versioned additions; it does not redefine AMSA, ARA, or Juan’s fixed raw selection.
Used for. Membership automatically admits the AIS fix to exact-path reciprocal coverage and expands it to every date-valid hydrophone array of its associated station. HA10, for example, produces separate H10N and H10S cases.
Boundary. The station-to-hydrophone array relation sets case count. Each hydrophone array still binds its own virtual receiver geometry. TL authority ranks exact one-to-one RAM first, then a daily sweep-derived value, then a monthly sweep-derived value.
Contextual traffic data
GFW gridded vessel presenceOutside AISset.traffic.gfwMember. One roughly hourly presence value in a 0.1 degree grid cell.
Coverage. 177,775 H01W presence-hours from January 2025 through July 2026, plus bounded HA03, HA08, and HA10 requests.
Used for. Aggregate traffic context and sector-level exploration only.
Boundary. This is not an exact AIS fix and supplies no per-fix speed, course, or track. It never joins to AMSA CRAFT_ID and does not enter the common expected-level ranking.
Acquired collaborator reconstruction
Juan's reconstructed positions near the H01W hydrophone arrayOutside AISMember. One reconstructed one-second source position.
Coverage. 21,306 positions on 2026 March 26 from 08:34:51 through 14:29:52 in the delivered source-time field. A derived set contains 60 minute centres from 10:10:30 through 11:09:30 UTC.
Used for. The 720-case dense-track direct-RAM cohort and its 360 paired seabed comparisons.
Boundary. Juan reconstructed the positions from course and speed. They are not independently observed dense AIS, and the delivered workbook does not declare its CRS or timezone.
Waveform geography
The project holds specific hydrophones for specific windows
The map uses sourced station coordinates and Natural Earth geography. Select a station region to see the available hydrophone arrays. Availability belongs to each holding, not to the station forever.
Natural Earth 1:110m country geometry; station coordinates from the project's 2026 August 8 geometry snapshot.
HA01
Western Australia. H01W1, H01W2, and H01W3 are the complete H01W triplet in the February, March, July, and August target holdings.
- 1 February–1 March: full-month triplet package
- 26–28 March: exact three-hydrophone holding
- 10–12 July: 47 of 48 triplet-hours complete
- 11–14 and 20–26 August: complete H01W triplet
These counts describe available files for the named windows. They do not claim permanent network health.
Restricted waveform sets
Raw waveform bytes remain local. A named holding binds the hydrophone array, hydrophones, UTC interval, sample rate, response standing, checksums, and gaps.
Named input setMember and scopeUse and boundary
Acquired phenomenon data
H01W hydrophone array, full February 2026Member. One waveform stream from the H01W1, H01W2, or H01W3 hydrophone, with its UTC, response, and gap record.
Coverage. The full UTC month at 250 Hz: 672 nominal triplet-hours and 670 raw-count-complete triplet-hours.
Used for. Juan's four-pair measured comparison and the overlap audit for Juan's raw H01W selection.
Boundary. The package supports nominal response-corrected comparisons, not traceably calibrated absolute sound-pressure levels.
Acquired phenomenon data
H01W hydrophone array, 2026 March 26–28Member. One waveform stream for the complete H01W triplet.
Coverage. All three hydrophones have every expected sample and record across the 48-hour holding. The dense-track presentation uses the exposed 10:10–11:10 UTC hour.
Used for. Measured one-minute morphology beside the 26 March modeled dense track.
Boundary. The waveform holding and the reconstructed vessel positions are separate input sets, even when one presentation shows them together.
Acquired phenomenon data
HA10 2026 March 31–April 2Member. One waveform stream from a hydrophone in the H10N or H10S hydrophone array.
Coverage. The H10N hydrophone array has all three hydrophones. The H10S hydrophone array has only H10S2 and H10S3. The March and April H10N packages contain 24 and 23 complete triplet-hours.
Used for. HA10 event comparisons and soundscape analyses.
Boundary. Two H10S hydrophones cannot supply a complete triplet-direction result. Older frozen packages remain counts-only under their recorded response basis.
Acquired phenomenon data
H01W hydrophone array, 2026 July 10–12Member. One hour with all three H01W hydrophones.
Coverage. A 48-hour holding with 47 complete triplet-hours. The final hour is withheld because every hydrophone lacks the last 0.004 seconds.
Used for. The separately selected July aggregate-transfer test.
Boundary. It is a bounded follow-up holding, not continuous monitoring between February and July.
Acquired phenomenon data
Multi-station 2026 August 11–14Member. One hydrophone-array holding for one target window.
Coverage. The holdings cover the H01W, H03N, H03S, H08S, H10N, and H10S hydrophone arrays on 11–13 and 13–14 August. H03N and H10S have two hydrophones; H01W, H03S, H08S, and H10N have complete triplets.
Used for. Cross-station event and soundscape work.
Boundary. The hydrophone counts describe available project files for these windows, not permanent network health.
Acquired phenomenon data
Multi-station 2026 August 20–26set.waveforms.idc.restricted.current_holdingsMember. One hydrophone-array holding in the 20–22 or 24–26 August blocks.
Coverage. 32 canonical hydrophone-window holdings plus four authoritative no-data outcomes. The declared analysis subset has 10 packages, 480 nominal hydrophone array hours, and 464 gap-safe hydrophone array hours.
Used for. The current 464-hour band-summary and persistent-feature cohort.
Boundary. The 10-package analysis cohort is narrower than the acquired holding. It excludes every hour that crosses a recorded gap.
Environmental inputs
Exact propagation needs more than a ship and a receiver
Water, bathymetry, seabed evidence, and station geometry are acquired auxiliary data. A path environment bundles the relevant members for one case. TL-lite deliberately omits that branch.
Auxiliary and derived environment sets
Keeping these inputs separate makes it clear which environmental choice changed and which results it can invalidate.
Named input setMember and scopeUse and boundary
Acquired auxiliary data
IMS hydrophone geometryMember. One hydrophone coordinate and depth record.
Coverage. A 33-hydrophone project snapshot dated 2026 August 8.
Used for. Station assignment, exact ranges and bearings, direct-RAM receivers, reciprocal reference geometry, and maps.
Boundary. Deployment geometry is not proof of 2026 channel health. Waveform holdings determine availability for each window.
Acquired auxiliary data
HA01 Copernicus monthly ocean fieldsMember. One calendar-month temperature and salinity field.
Coverage. 19 fields from January 2025 through July 2026 from GLOBAL_ANALYSISFORECAST_PHY_001_024, version 202406.
Used for. Juan direct RAM, source-depth sensitivity, the March dense track, and the January reciprocal layers.
Boundary. These are Copernicus analysis/forecast fields, not GLORYS reanalysis despite some legacy filenames.
Acquired auxiliary data
Exact-date daily ocean bankMember. One temperature-and-salinity field for an exact UTC date and the correct station domain.
Coverage. All 112 dates required by the present selected one-to-one cases are available: 111 HA01 dates plus the HA10 field for 2026 March 31.
Used for. One-to-one reciprocal RAM for specially selected AIS fixes.
Boundary. This bank does not satisfy monthly sweep coverage or imply daily coverage for the wider AIS union. PT6H and nearby dates are not substitutes.
Acquired auxiliary data
HA10 daily GLORYS12 ocean fieldMember. One daily temperature and salinity field.
Coverage. 2026 March 31 from GLOBAL_MULTIYEAR_PHY_001_030, version 202311.
Used for. The bounded HA10 one-to-one cases and, historically, the completed H10N and H10S reciprocal sweeps.
Boundary. It does not stand in for monthly sweep coverage or for dates outside the HA10 case.
Acquired auxiliary data
GEBCO 2026 bathymetryMember. One pinned bathymetry grid sampled every 300 m along each RAM path.
Coverage. Global source with saved regional and path samples.
Used for. Direct and reciprocal RAM path environments.
Boundary. This is the sole desired RAM bathymetry source. Another source or sampling step remains historical evidence only. TL-lite does not use bathymetry.
Acquired auxiliary data
Mapped seabed-class evidenceMember. One mapped or explicitly unresolved seabed-class observation.
Coverage. Geoscience Australia mixtures around HA01 and bounded Diesing/EarthByte evidence around HA08 and HA10. HA03 remains unresolved or uses an explicit fallback.
Used for. Evidence consumed by the geology-informed seabed rule.
Boundary. Mapped class evidence is not a complete RAM bottom. The scientific rule must turn it into engineering properties and record any fallback.
Derived auxiliary data
Compiled path environmentMember. One complete geometry, water, bathymetry, seabed, and receiver bundle for one propagation case.
Coverage. Built only for the direct or reciprocal cases that need it.
Used for. Input to one RAM one-to-one transmission-loss calculation.
Boundary. It remains derived data. The acquired environmental sources and the seabed definition stay separately identifiable.
Acquired auxiliary data and scientific references
Hydrophone channel and response referencesMember. One channel identity, geometry record, and the response basis that applies to a named waveform holding.
Coverage. The channels from the H01W, H03N, H03S, H08S, H10N, and H10S hydrophone arrays used by the bounded waveform analyses.
Used for. Channel availability, triplet membership, counts-to-nominal-pressure conversion, and measurement labels.
Boundary. A response reference does not prove that a channel was healthy. The holding and gap record decide availability, and the current pressure results are not traceably calibrated.
Named derived sets reused downstream
A derived result can become an input to a later case. Naming these sets keeps the later cohort definition explicit without folding earlier computations into it.
Named input setMember and scopeUse and boundary
Derived phenomenon-data set
H01W 26 March minute centresMember. One reconstructed source position at the centre of a specified UTC minute.
Coverage. 60 positions from 10:10:30 through 11:09:30 UTC on 2026 March 26.
Used for. The source-position role in the 720-case H01W dense-track cohort.
Boundary. The minute centres are derived from Juan's reconstructed one-second positions; they are not newly observed AIS fixes.
Derived phenomenon-data set
Juan's frozen four-pair windowsMember. One preselected higher-predicted or comparison waveform window.
Coverage. Four fixed pairs, eight H01W windows, chosen before waveform values were read.
Used for. The bounded measured comparison of Juan's selected H01W cases.
Boundary. Membership is frozen. The set cannot be enlarged after looking at the acoustic result and still be treated as the same prospective comparison.
Derived model-data set
Juan four-pair modeled levelsMember. One saved modeled level from one vessel matched to a frozen waveform window.
Coverage. Eight window-level values, one for each member of the four frozen pairs.
Used for. The higher-versus-comparison label and travel-time alignment in the four-pair measured analysis.
Boundary. These are model outputs, not measured levels. They enter the downstream case as already derived inputs and are not recomputed by the waveform comparison.
Derived phenomenon-data set
August 20–26 gap-safe hoursMember. One hydrophone-array hour that crosses no recorded waveform gap.
Coverage. 464 eligible hours selected from 480 nominal hours in ten hydrophone-array packages.
Used for. Band summaries and persistent-feature discovery for the current August soundscape cohort.
Boundary. It excludes every hour that crosses a recorded gap. H03N and H10S remain two-hydrophone arrays rather than complete triplets.
Derived phenomenon-data set
January 2025 H01W grid-scope AMSA setMember. One January AMSA report assigned to the frozen H01W reciprocal-grid scope.
Coverage. 1,469 rows fall inside the clipped grid; 1,411 pass the source-model eligibility rule frozen before TL sampling.
Used for. The January 2025 63 Hz ship-type screen.
Boundary. The parent AMSA month is a readable truncated prefix. This set cannot represent complete-month traffic exposure.
Derived model-data set
January 2025 reciprocal TL fieldsMember. One saved receiver-centred transmission-loss field at 63 or 80 Hz under one seabed definition.
Coverage. At each frequency, 1,800 historical-fixed and 1,800 geology-informed fields share the frozen grid and depth bank. The 63 Hz pair is sealed; the 80 Hz computation finished on Pazuzu and awaits closeout.
Used for. Nearest-computed bearing-and-range samples for the 1,411-row January ship-type derivative.
Boundary. Only 63 Hz is scientifically admitted and used in the admitted derivative. The 80 Hz fields remain unadmitted and unvisualized until their closeout and derivative receipt is sealed.
Scientific definition sets
Definitions are inputs too. They determine what the data mean and what one derivation computes.
JOMOPANS-ECHO source level
set.definition.sl.jomopans_echoMaps vessel class, positive speed, length, and frequency to one modeled decidecade source level. Each required value may be observed or supplied by a method admitted as reliably inferential for that attribute and case. Reliably matched registry class and length are currently admitted. Adjacent-position speed reconstruction is not admitted for these holdings because its reliability is unestablished; this does not permanently ban every speed-inference method. Current received-level products use the central estimate; uncertainty propagation is deferred.
Reliable attribute inference
inference.reliable_attribute_evidence.v1One general admission rule applies to missing inputs. An inferred value is usable only when its named method is judged reliable for that attribute and case, and it keeps its evidence, method, assessment, uncertainty where known, and inferred standing. A reliably matched authoritative registry design draught is admitted for the 0.7 × draught branch but is not called an observed voyage draught.
Common TL-lite screen
tl.juan_lite.range_only.v1One range-only 20/10/15-log loss rule with 1.9 km and 4.8 km joins and 0.0035 dB/km absorption. It uses no water, bathymetry, seabed, bearing, or source depth.
RAM 2-D parabolic-equation model
tl.ram.parabolic_equation_2d.v1One numerical propagation definition. Geometry mode, path environment, frequency, source depth, grid, and requested output coordinates are case inputs.
RAM path-environment profiles
set.definition.path_environment.currentThe desired RAM environment combines water-column profiles no more than 10 km apart, GEBCO 2026 sampled at 300 m, and the geology-informed seabed. Finer water spacing is compliant. Historical fixed-seabed results remain labelled comparison evidence and cannot satisfy current RAM coverage.
Source-depth conventions
set.definition.source_depthOne reciprocal RAM field carries shallow depth samples. Received level is always required at 6 m and also at 0.7 × draught when draught is observed or supplied by an admitted reliable inference method. Authoritative registry design draught is currently admitted and remains labelled inferred. New reusable sweeps prefer 3–10 m at 0.5 m spacing; compatible saved axes satisfy only the depths they contain or properly bracket. The March 6.16 m and reader-map 7 m values keep their case-specific identities, and 5 m is superseded history.
Waveform packaging and measurement
processing.ara_waveform_holdings.v2 + case-specific protocolsGap-safe packaging, nominal response conversion, fixed band summaries, source-unlabelled persistent-feature detection, and the frozen four-pair comparison are distinct definitions.
Time coverage
A filled date means something different in each source
AMSA is organized by release month. ARA contains discrete capture days. GFW is gridded presence. Waveforms are bounded holdings. The calendar keeps those meanings apart.
AMSA public releases
Monthly holding. “Prefix” marks a readable truncated source DBF; “complete” marks a structurally complete source month.
GFW H01W presence
Roughly hourly values in 0.1 degree cells. A filled month means an available monthly presence product, not an exact AIS track.
ARA snapshots
Discrete capture days only: HA10 on 31 March and 1 April; multi-station snapshots on 11–14 and 20–25 August.
Restricted waveforms
Bounded holdings: full February H01W; 26–28 March H01W; 31 March–2 April HA10; 10–12 July H01W; August station blocks.
structurally complete month readable source prefix gridded presence product discrete snapshot days bounded waveform holding
Cohort definitions
Choose a cohort to see its complete case recipe
Each definition names the input sets, scientific definitions, eligibility, and case size. Similar-looking tables remain distinct when any of those bindings differ.
Selected cohort
Global common 63 Hz screen
- One case binds
- One source-model-eligible AMSA or ARA observation associated with the H01W hydrophone array, paired with one applicable receiver reference at 63 Hz.
- Named input sets
- H01W 200 km AMSA set
- ARA HA10 named-vessel snapshots
- ARA multi-station snapshots
- IMS hydrophone geometry
- Scientific definitions
- JOMOPANS-ECHO source level
- Common TL-lite screen
- Valid source-input and receiver-pairing rule
- Eligibility
- Finite time and position, a mapped source class, usable same-fix speed and length, and an applicable receiver pairing.
- Declared coverage
- One reusable source-level table
- One common TL-lite estimate per case
- One predicted received level per case
- Nearest-position, strongest-source, and top-predicted-level views with their own sorting measures
- Observed state
- Numerical cohort and the three current views are present.
- Claim limit
- Comparable candidate prioritization only. It is neither measured sound nor exact path propagation.
Coverage policy
Coverage declares what should exist. It does not prescribe run order
Required current coverage keeps reusable numbers and review views together. Historical comparisons stay visible without becoming parallel obligations. Explicit non-coverage prevents the site from implying products that have no scientific basis.
Required current
One normative RAM environment
Desired RAM fields use water-column profiles no more than 10 km apart, GEBCO 2026 at 300 m along path, and the geology-informed seabed. Finer water spacing is compliant. Receiver-centred sweeps use 0.25° base rays with independent midpoint validation and bounded local refinement.
Historical comparison evidence
Visible, but not a second requirement
Existing fixed-seabed fields and comparisons keep their as-run identities. They may inform a sensitivity report, but they cannot satisfy geology-informed coverage or trigger new fixed-seabed computation.
Awaiting admission
January 80 Hz fields exist but are not yet a report result
Both 1,800-field solver layers finished on Pazuzu. The site will not interpret or visualize them until their canonical closeout and derivative receipt is sealed.
How a gap appears. A new input set, a changed definition, or a policy revision selects a target instance for each coverage key. The project has a gap when no intact, scientifically admissible materialization matches that target or an explicitly accepted bounded equivalent.
When a result is out of date. The active coverage key still wants the result, but its materialization records different identity-defining input or protocol hashes from the target now declared for that key. Modification time is irrelevant.
Monthly reciprocal sweeps. The general rule keys one reusable shallow-depth sweep by station, exact receiver geometry, represented AIS month, its monthly ocean field, priority frequency, and propagation profile. The environment uses water-column profiles with gaps no greater than 10 km, GEBCO 2026 at 300 m, and the geology-informed seabed. The desired domain reaches 200 km on a 50 m radial grid. Coastal bearings may stop at land without becoming incomplete. If the monthly field has not appeared, the sweep stays required but waits for that input.
ARA historical is the bounded exception. It contains HA10 fixes on only 2026 March 31 and April 1. Each date creates daily H10N and H10S sweep identities with the same propagation profile as the monthly rule. Keeping the day matters here because this two-date family straddles a month boundary. The exception does not apply to AMSA, ARA current, or another family.
Named AIS sets keep their full source membership. AMSA, an AMSA month, ARA, its children, and station unions are not redefined. When one feeds this scientific branch, the dependency means its JOMOPANS-ECHO-eligible projection: class, positive length, and positive speed are each observed or supplied by a method admitted as reliable for that attribute and case.
Ineligible fixes remain evidence. They create no source-level, reusable sweep, TL-to-received-level, or received-level requirement. The projection is declarative, not a destructive filter or required Boolean column. Group eligible fixes by associated station and UTC calendar month; each nonempty group creates reusable sweeps after expansion by hydrophone array, frequency, and profile, never one sweep per row. A separately highlighted fix can still require exact-path RAM TL because propagation itself does not need a source-level estimate.
Selected exact-path cases. A highlighted AIS fix separately requires the daily field for its exact UTC date and station domain. The ARA-historical daily sweeps do not replace these one-to-one results. If required water is missing, the affected key waits for that exact input. No lane silently drops a case or substitutes PT6H, a nearby date, or another lane's water.
Existing work must match the whole identity. An existing daily sweep can satisfy an ARA-historical key only when date, hydrophone array, receiver geometry, ocean field, frequency, shallow-depth capability, angular sweep profile and released mask, water-profile gaps no greater than 10 km, GEBCO 2026 at 300 m, geology-informed seabed, extent, and numerical profile all agree. Finer water spacing satisfies the same key.
Angular interpolation is local and tested. Sweeps start with 0.25° base rays. Independent RAM midpoint calculations qualify each tested local region when ordinary-mask median absolute linear-TL error is at most 0.8 dB. Individual cells above 1 dB or affected by coastline, clearance, curvature, null, phase, or common-wet failures stay masked with a reason.
Refinement stops. Failed scientifically relevant regions may use the accepted 0.125° and 0.0625° local base refinements. Independent 0.03125° holdouts test the final base, then the ladder stops. Passing quarter-subtiles are released; residual failures remain masked. The policy never triggers uniform whole-circle densification, and RAM remains the reference outside the released mask.
Received-level depth branches. One RAM solve stores a shallow depth-indexed field. Evaluate it at 6 m for every applicable received-level result. Also evaluate it at 0.7 × draught when draught is observed or supplied by a reliably admitted inference method. A matched authoritative registry design draught is admitted but remains labelled inferred. A field that supports 6 m but not the second depth remains valid for 6 m; only the missing branch stays unsatisfied.
Highlighted-case TL table. Every highlighted case gets one human-facing row with the authority-selected RAM TL at 6 m, the conditional draught-depth RAM value, TL_lite, spherical spreading, and fast TL. The three fast values are evaluated when the table is requested and stored inside it. They need no separate precomputed coverage and cannot replace RAM.
Required frequency tiers. Reusable reciprocal RAM sweeps and the preserved historical Highlighted-21 TL grid use 14 frequencies: 5, 6.3, 8, 10, 12.5, 16, 20, 25, 31.5, 40, 50, 63, 80, and 100 Hz. The current Full Juan exact campaign uses the 11 production centres from 10 through 100 Hz. Its Tier 1 contains 63 and 80 Hz; its Tier 2 contains the remaining nine centres. The historical Highlighted-21 values at 5, 6.3, and 8 Hz do not enlarge current Full Juan coverage. Both tiers remain required. Runnable incomplete Tier-1 work blocks Tier 2 within the active work family. A Tier-1 key waiting for unavailable input remains incomplete but does not block runnable Tier-2 work until its hold clears.
Staged environment gap. Accepted HA01 and HA10 fields already use GEBCO 2026 at 300 m. The staged HA03 and HA08 1 km environments must be regenerated before desired RAM work uses them. This is an input gap, not an invalidation of accepted results.
Observed availability on 2026 August 29. The local selected one-to-one case bank has domain-correct daily water for all 112 required dates: 111 HA01 dates plus HA10 on 2026 March 31. Separately, the official catalogue offered daily temperature and salinity through 2026 September 5, covering every ARA-current date upstream. That does not say every station-domain file is cached locally.
August monthly water is not yet available. The official monthly temperature and salinity series currently end in 2026 July. A 2026 August station-month sweep therefore stays required but is on hold waiting for its monthly field, even though exact-date daily water is offered upstream. These are dated observations, not desired-state rules.
Observed ARA-historical state. Completed H10N and H10S sweeps using the daily 2026 March 31 field may satisfy the matching exception keys after full identity checks. The 2026 April 1 keys remain required and wait for their daily input or matching materialization. This dated status is not part of the desired declaration.
Other historical results keep their recorded inputs. Completed H01W main sweeps used monthly water. No completed main reciprocal sweep used PT6H.
Read exact identity, lifecycle, and validation semantics in the technical companion