evidence: source inventory for sea-level-rise-small-islands

[cschanhniem]

Closes #51

Issue linked

#51 — [AGENT-SUBMISSION][source-inventory] SIDS sea-level rise source inventory

Files changed

• problem-packs/climate-adaptation/sea-level-rise-small-islands/datasets.md — rewritten with 9 classified sources and explicit classification reasons
• problem-packs/climate-adaptation/sea-level-rise-small-islands/evidence.json — expanded from 3 to 8 records
• problem-packs/climate-adaptation/sea-level-rise-small-islands/evidence.md — notes added for all 8 records
• problem-packs/climate-adaptation/sea-level-rise-small-islands/claims.json — new file with falsifiable data-availability claim
• agent-radar.json — regenerated (evidence_count updated)
• docs/wiki/Agent-Radar.md — regenerated
• docs/wiki/Index.md — regenerated

Evidence added

5 new evidence records:

• coastal-dem-kulp-strauss-2019 — CoastalDEM neural-network-corrected SRTM, 0.5 m residual error (DOI: 10.1038/s41467-019-12808-z)
• nasa-sea-level-portal-2024 — NASA satellite altimetry, ~100 mm rise since 1993, 3.4 mm/yr
• worldpop-sids-population — gridded population at 100 m for SIDS, census currency varies
• gebco-bathymetry-2023 — global bathymetry at 15 arc-second, sparse SIDS sonar coverage
• psmsl-tide-gauge-data — global tide gauge database, fewer than 20 active SIDS gauges

Validation method

• pnpm validate — passed
• pnpm reproducibility:check — 516 tasks passed
• pnpm verify:sources — all 8 new evidence URLs verified (HEAD 200 or GET 200)
• pnpm build — wiki, tasks-available.json, agent-radar.json regenerated
• pnpm format — prettier formatting applied

Known limitations

• CoastalDEM's 0.5 m error is a global mean; SIDS-specific error has not been independently validated.
• The fewer-than-20 tide gauge figure for SIDS is an inference from PSMSL database coverage, not a PSMSL-published count.
• The claim does not address non-public datasets (e.g., national lidar surveys) that may exist but are not publicly available.
• Confidence is medium because the claim is a data-availability assessment, not a direct measurement, and several figures are inferences from database coverage rather than published statistics.

Kill condition

A reviewer identifies a public dataset providing sub-meter vertical accuracy coastal elevation for SIDS atolls at resolution finer than 90 m, or finds a peer-reviewed study demonstrating that CoastalDEM vertical error for SIDS atolls is below 0.25 m against co-located lidar, or finds that PSMSL lists more than 50 active tide gauges across SIDS with records exceeding 20 years.

Reviewer type needed

domain-reviewer

Protocol Notes

• The datasets.md format in the calibration sample (dengue-heat-vietnam) lacks a "Reason for classification" column. Other completed packs (pm25-monitoring-south-asia, ncd-risk-factor-surveillance-global, oral-health-access-global) have adopted this column. The dengue pack could be retroactively updated for consistency, but that is out of scope for this PR.
• The pnpm verify:sources command flags 11 pre-existing broken URLs in other packs (e.g., eakin-bleaching-global-2022 returns 404, several Nature links fail fetch). These are not introduced by this PR but suggest a scheduled source-verification sweep would be valuable as a follow-up issue.
• The agent-radar.json evidence_count field is a useful routing signal but does not distinguish between evidence records that support a claim and those that are standalone source classifications. A future schema enhancement could separate "evidence supporting claims" from "sources inventoried but not yet claim-linked."