[jt55401]

What this adds

Closes #479

Adds the southern elephant seal record for New Zealand, its attribution, and a species-local behaviour program.

Evidence and scope

• GBIF usage key 2434814 resolves to accepted species-rank Mirounga leonina. Its synonym endpoint includes Macrorhinus leonina and Phoca leonina, linked to the accepted key. iNaturalist API 41729 is active and species rank. Wikidata Q215343 maps the exact scientific name to that iNaturalist ID. These providers were checked on 2026-11-01. No existing record or competing open PR was found for the name, synonyms, or IDs.
• DOC identifies the New Zealand population as resident on the Antipodes and Campbell Islands. The record uses the requested national activity region NZ, not a claim that seals occur throughout every part of the country.
• DOC's New Zealand page gives adult female mass 900 kg and male mass 3,600 kg. mass_kg is their midpoint, 2,250 kg. length_m is an authored representative value, 3.5 m, derived from the midpoints of the female and male length ranges. Diet shares (fish 0.7, invertebrates 0.3), group-size bracket, habitat proportions, activity pattern, and rendering measurements are identified as authored or derived in provenance.
• ADW's Lifespan/Longevity prose says the wild average is about 9.5 years; the preceding sentence gives male 14 and female 20 years. ADW's structured lifespan fields also label 14–20 years as typical, which conflicts with its own average sentence. The record uses the explicit average sentence and documents that conflict. ASOC's page reports a different 21-year wild average but has no reuse licence visible on the page, so it is not used. The CC BY 4.0 DOC source contains no lifespan estimate.
• DOC describes New Zealand breeding in September–October. Mating is placed in October–November as a derived window from DOC's statement that it follows about three weeks after pups arrive. No moult calendar is asserted because the New Zealand source gives only broad seasonal wording for some age and sex groups.

Activity and behaviour

• Fetched NZ with tools/fetch_activity.py. The GBIF query covers present records with no geospatial issue and CC0/CC BY licences. It returned 5,115 records at tier A and is marked usable. Raw month bins (Jan–Dec): 45, 9, 257, 470, 432, 174, 702, 1,083, 858, 599, 370, 109. The resulting gbif_share curve is [0.01, 0.00, 0.17, 0.54, 0.69, 0.33, 0.82, 1.00, 0.82, 0.12, 0.07, 0.03]. February's nine records normalize to 0.00 at two-decimal precision; that does not establish seasonal absence and can suppress the simulation in that month.
• iNaturalist was queried at tier C after wider licence tiers remained below threshold: 190 observations, marked unusable and not included in the curve. It is retained in the query record. The delivered curve uses only GBIF.
• fsm-mirounga-leonina-v1 rests at a shore-haul-out proxy and uses swim-gait feeding actions for the small-vertebrate fish proxy and invertebrate prey tag. Both feed states are reachable from rest when hunger exceeds 0.4 and matching food is nearby; they return to rest when hunger falls below 0.15 or that food is no longer nearby. The resource schema has no fish or marine-cephalopod tag, so these tags are coarse host-world proxies.
• The program has no month_in or day/night predicates, so its rest and feed states can run in any month and at any hour. It models neither dives nor depth. DOC describes day/night differences in dive depth; those are outside the actions available here and are not asserted by this program.

Validate report

Tested commit: aa149ab40eeeea0e32f073cf8d359192bd5fea59.

Focused validation: 0 errors, 2 warnings, provenance 33/33. Warnings: NonCommercialLicense for ADW, used only for traits.lifespan_y; MissingShoulderHeight, left unset because a seal has no standing shoulder measure.

• uv run tools/validate.py --self-check: 637 files scanned, 0 findings.
• uv run tools/validate.py --report: 0 species error rows.
• uv run tools/validate.py --locale all --report: 0 locale error rows; 126 warnings.
• uv run --with pytest --with jsonschema pytest tools/tests -q: 347 passed, 8 skipped.
• python3 tools/build_site.py: succeeded.
• Post-commit uv run tools/validate.py species/mirounga-leonina --write-attribution --report left git diff -- species clean.

Preview evidence

Inspected idle, walk, and swim in the previewer. The low-poly model reads as a large seal with a male proboscis; the walk pose moves its short flippers. No clipping was visible. The swim pose is a generic quadruped swim cycle and does not depict a dive.

Checklist

Species

• Closes the linked species-request issue
• Checked current main and open PRs for the same taxon; claimed the issue
• Taxon identifiers and source claims checked at their providers
• Evidence locators, units/statistics and authored transformations documented
• Validate report included above, with no errors
• species/mirounga-leonina/ATTRIBUTION.md regenerated and committed
• Activity curve generated for every suggested region
• Query scope, licences, counts and usability inspected; low-count February and unusable iNaturalist series disclosed
• Executed behaviour checked for gaits, diet, time of day and New Zealand seasonal scope; limits documented above
• No unrelated shared-record fields or regional curves changed
• Repository self-check, all-species/all-locale validation and tests pass on the submitted commit
• Previewed and checked every pose
• Licences preferred CC0/CC BY where available; ADW NC-SA use is limited to lifespan and justified below
• No personal data, local paths, hostnames, or tokens in the diff
• One species in this PR

Non-CC0/CC-BY sources

ADW's text is CC BY-NC-SA 3.0 and supports only the required traits.lifespan_y value. It was used as a last resort because the CC BY DOC source gives no lifespan, and the alternative 21-year claim found on ASOC had no reuse licence visible on its page. The ADW page's conflicting structured 14–20-year field is documented in the provenance note; the record follows the prose's explicit average-wild-life statement.

Remaining work

The simulation uses coarse prey and shore-rest proxies, has no diving/depth action or day/night gate, and the February activity value is a low normalized index rather than evidence of absence. Maintainers own import and live-release acceptance.

[jt55401]

Coordinator review: the committed FSM only swims at the surface to eat nearby prey and explicitly cannot represent dives or depth. Southern elephant seal foraging depends on diving, so this is a required behavior-fit blocker under AGENTS.md. Preserving pushed branch wave-20261101d-479-mirounga-leonina and its evidence; no merge. The issue will carry the sourced blocker and research label.