The OceanSig Epistemic Methodology
Ocean science is fundamentally multi-modal. Researchers must regularly synthesize satellite radiometry, moored buoy anemometers, shallow harbor thermistors, diver surveys, and deep benthic hydrophones.
OceanSig prevents common scientific errors by replacing informal data juxtaposition with formal, machine-evaluated Comparison Contracts, strict Claim Ceilings, and 6-Step Provenance Walkbacks verifiable down to raw source bytes.
1. Comparison Contract Engine (18 Dimensions)
Before any two observational series can be compared or correlated, OceanSig evaluates them across 18 distinct physical, spatial, temporal, and institutional dimensions:
Evaluated Dimensions:
Flags when two sensors measure fundamentally disparate physical phenomena or datums (e.g. NOAA CRW satellite skin SST at 10 µm vs Key West Harbor thermistor at 0.8m below MLLW). Arithmetic differences are labeled as multi-station differences, NOT spatial gradients.
Flags when sensors target distinct physical quantities (e.g. NDBC anemometer wind speed in m/s vs CO-OPS tide gauge water level in meters). Blocks invalid causal inferences without hydrodynamic modeling.
2. Claim Ceilings & Blocked Overclaims
Every supported claim in an OceanSig dossier is accompanied by explicitly blocked overclaims that define the exact boundary of what the data does not prove:
A shallow harbor tide station (e.g. Key West 32.13 °C) cannot be extrapolated across 10 kilometers of coastal shelf to claim offshore barrier reef thermal stress.
Photographic diver observations documenting coral bleaching onset do not establish mortality without post-event demographic survival surveys.
Acoustic hydrophone energy spikes reflect ambient sound pressure, not direct animal counts. Acoustic silence does not prove animal absence.
3. 6-Step Provenance Walkback Chain
Every rendered metric in an empirical dossier traces through 6 deterministic, auditable steps down to immutable disk bytes:
- Step 1RENDERED_VIEW
The final formatted metric and unit rendered in the user interface.
- Step 2CANONICAL_DOSSIER
The structured investigation object holding claims, contracts, and uncertainties.
- Step 3NORMALIZED_RECORD
The parsed, typed numerical value and timestamp representation.
- Step 4TRANSFORMATION_RUN
The deterministic transformation algorithm (e.g. daily arithmetic mean, peak extraction).
- Step 5RETAINED_ARTIFACT
The immutable source file retained on disk, verified by SHA-256 cryptographic digest.
- Step 6RETRIEVAL_EVENT
The timestamped HTTP acquisition event recording authoritative provider URL and headers.
4. Reproducibility Classifications
OceanSig explicitly classifies the reproducibility state of every investigation:
All primary evidence files (raw telemetry TXT, NetCDF extracts, CSVs) are stored on disk with cryptographic SHA-256 digests. Computations and aggregations can be mechanically replayed from local bytes.
Examples: Florida Keys July 2026, Summer 2026 Cyclone/ENSO, Monterey Bay September 2026.
Mathematical or numerical model simulation designed to evaluate methodological workflows and cross-modal alignment without claiming direct physical in-situ sensor retention.
Examples: Cascadia Margin Hypoxia 2026 investigation.
Acoustic hydrophone spectrogram processing framework with strict epistemic invariants enforced, but without claiming complete raw audio WAV byte retention on disk.
Examples: MBARI MARS Sanctuary Soundscape benthic hydrophone integration.
5. RO-Crate 1.3 Standards Conformance
To ensure open scientific interoperability, OceanSig generates research object metadata conforming to the RO-Crate 1.3 community specification using JSON-LD (https://w3id.org/ro/crate/1.3/context). Every exported dossier package bundles:
- Canonical
ro-crate-metadata.jsondescribing datasets, sources, authors, and licenses; - Cryptographic SHA-256 digests and byte counts for all retained raw source files;
- Formatted citation strings in BibTeX, APA, and ISO-690 formats.