Article 3 fishing-data lineage audit

Technical summary

The archived files do not contain a defensible explicit-zero activity control for P3. The four-year 2021-2024 table has 86,448Source for 86,448Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json unique, complete C-square rows; 86,444Source for 86,444Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json are positive and 4Source for 4Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json are explicit both-zero rows, but none of those explicit-zero geometries intersects H2/P3. The held candidate intersects 1Source for 1Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json C-squares, none with a 2021-2024 row. Therefore the candidate remains a missing-support stratum, not evidence of low activity.

Omitted rows cannot be recoded as zero

The current P3 validation block intersects 3Source for 3Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json C-squares: two have positive 2021-2024 rows and one covers only 9Source for 9Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json cells without a current row. The held candidate intersects 1Source for 1Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json C-squares with no current row; 1Source for 1Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json of those C-squares nevertheless have a positive row in an earlier archived four-year window. This is direct evidence that row omission and low activity are not interchangeable states.

The annual audit found no positive annual row in the analytically relevant omitted-row C-squares, but its concordance test recovered only 28Source for 28Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json of 119 known current-row C-squares (23.5%Source for 23.5%Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json). Annual absence is therefore non-confirmatory.

The candidate changes reporting support, not measured activity

The current P3 block is almost entirely covered by positive supplied rows, whereas every cell in the held candidate lacks a current row. Neither block contains an explicit-zero row. The chart is normalized within each 1,600Source for 1,600Source: P3 C-square intersection auditFile: article3_p3_csquare_intersections_v1.csv-cell block; exact counts remain available in the adjacent audit table.

Explicit zeros exist, but not in the P3 domain

Only 4Source for 4Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json of 86,448Source for 86,448Source: Article 3 fishing lineage profileFile: article3_fishing_source_profile_v1.json current four-year rows are explicit both-zero records. Their C-square identifiers are preserved below for audit, but the aligned H2 component contains no raster cell combining a supplied SAR row with no recorded fishing or aggregate activity. Their existence confirms that the product can represent an explicit zero; it does not authorize treating an omitted row as that state.

Scope and evidence definitions

• Current four-year row: a unique C-square record whose Time field is 2021-2024 and whose surface and subsurface SAR values are populated.
• Explicit zero: a supplied current row with both surface and subsurface SAR equal to zero.
• Positive row: a supplied current row with surface or subsurface SAR greater than zero.
• No row supplied: the C-square geometry exists, but no current four-year SAR record joins to it. This is missing reporting support, not measured zero.
• P3 scopes: the current 1,600-cell validation block and the held 1,600-cell support candidate, both evaluated on the authoritative Article 2 grid.

The lineage check preserves source semantics

The audit parsed the archived C-square shapefile and DBF directly, assigned every P3 cell centre to exactly one C-square, profiled the raw semicolon-delimited SAR table, and joined the archived direct annual audit by C-square identifier. Current-period uniqueness, completeness, sign and explicit-zero states were recomputed from the raw values. No missing row was imputed, converted to zero or replaced with annual absence.

Annual evidence is useful but incomplete

The annual archives contain 281,743 métier-year rows, including 1,021 both-zero records. For the held candidate C-squares the archived audit found 0Source for 0Source: Annual ICES SAR audit interpretationFile: output/sar_annual_audit_20260812/README.md annual rows and 0Source for 0Source: Annual ICES SAR audit interpretationFile: output/sar_annual_audit_20260812/README.md positive rows. However, the annual service only partially reproduces known current-table positives, so zero annual matches cannot establish zero activity. The raw four-year product also provides no documented omission semantics that would justify treating absent rows as true zeros.

Keep the P3 hold and request provider-level coverage metadata

1. Retain the current P3 block as activity-conditioned.
2. Retain the alternative only as a missing-reporting/support-uncertainty stratum.
3. Do not use high-versus-low activity language in the manuscript or model.
4. Seek provider confirmation of row-omission semantics or a complete VMS/logbook extract that includes explicit coverage and zero-status metadata.
5. Reopen matched-control selection only if an equivalently supported explicit-zero domain becomes available.

Further questions

• Does the data provider distinguish confidentiality suppression, fleet-threshold filtering and measured zero in the four-year product?
• Can a complete Spain/Portugal VMS or logbook extract provide equivalent spatial and temporal support across both P3 blocks?
• If not, should Article 3 formalize reporting-support uncertainty as a design factor rather than activity intensity?