# Verdict: PASS (2026-11-08)

Pre-registration: `PREREG.md` (with Amendment 1, written before the count). Scripts: `scripts/killtest.py`, `scripts/wod_extract.py`, `scripts/wod_profiles.py`. WOD files: NCEI geographic set downloaded 2026-11-08. Ice: NSIDC Sea Ice Index v4.0 monthly concentration. "Under ice" throughout means a 25 km cell with a monthly mean concentration at or above the threshold; it is a proxy, not ice observed at each cast.

## 1. Pre-registered test (casts with any variable)

| population | profiles | seals | floats | ships | seal share |
| --- | ---: | ---: | ---: | ---: | ---: |
| Jun–Sep, south of 60°S, 1979–2025, ice ≥ 15 % (pre-registered primary) | 102,790 | 75,864 | 22,913 | 1,451 | 73.8 % |

Threshold ≥ 60 % → **PASS**. (A further 2,562 profiles came from moorings and are in the total.)

Pre-registered secondaries, on this population:
- Possible duplicate seal records (same date, hour, position to 3 decimals, within the seal class): 115; share without them 73.8 %.
- Coast cells: 5,715 profiles fell in coast, land or missing cells (5,585 of them seals). Giving each the value of the nearest sea cell whose centre lies within 75 km on the 25 km grid (Euclidean, between cell centres) assigned 5,686 and left 29 unresolved; share 75.0 %.
- MEOP cross-check: 72,711 MEOP profiles pass the same filter, against 75,864 animal-borne profiles in WOD. The two lists were not matched record by record, so the cause of the difference is not established.
- 1979–2003: 2,564 profiles, none from seals. 2004–2025: 75.7 % seals.
- Cells below 15 % in the same months: 56.5 % seals.

## 2. Page population (temperature profiles that can be drawn)

Added after an adversarial review, after the verdict above. The page counts temperature profiles that have, on at least two levels of the drawn grid (every 50 m to 1,000 m, then every 250 m to 2,000 m), values derived (including by interpolation) from measurements accepted by WOD quality control (cast-level temperature flag, depth flag and value flag all 0). Every counted profile is a thread on the page. 4,849 profiles with fewer than two drawn levels are excluded (2,560 mooring records and 2,200 short animal-borne records among them).

| population | profiles | seals | floats | ships | seal share |
| --- | ---: | ---: | ---: | ---: | ---: |
| ice ≥ 15 % (page headline) | 81,328 | 61,648 | 18,607 | 1,073 | 75.8 % |
| ice ≥ 50 % | 66,979 | 49,889 | 16,212 | 878 | 74.5 % |
| ice ≥ 80 % | 31,643 | 20,104 | 11,097 | 442 | 63.5 % |

The stricter thresholds are the sensitivity check for the monthly-cell proxy; they do not establish ice on the day of each cast.

On this population (descriptive, not pre-registered):
- Floor shallower than 1,000 m (ETOPO 2022 on a 14 km grid, a proxy for the continental shelf): 41,849 of 44,212 = 94.7 % seals.
- Pooled cell-months (1° lat × 3° lon × year-month), 1979–2025: at least one float in 70 %, at least one seal in 29 %, only seals in 27 %. Float positions under ice are often estimated rather than fixed by satellite, so float coverage is partly estimated.
- Accepted temperature at 500 m or deeper: seals 17,519, floats 17,827.
- Ships took more than 100 such profiles in a winter in 1986, 1989, 1992, 1997, 2002; 118 in total since 2003.
- Seal profiles peaked at 6,996 in 2011. Float counts generally increased, with year-to-year variation, and exceeded seal counts in 2016, 2018, 2021, 2024, 2025. The seal count varies strongly between years; the number of tagged animals is a likely influence but was not measured here.

## Prior art

Fedak 2013 (*Deep-Sea Res. II* 88–89:7–13) reports that animals supplied about 70 % of oceanographic profiles south of 60°S. This recount is narrower (winter, monthly ice proxy, temperature profiles in WOD) and later (to 2025).