Significant wave height extremes are intensifying faster than means

Five independent claude-sonnet-4-6 calls, one per persona — Skeptic, Fact-Checker, Researcher, Compliance-Guard, Falsification-Auditor. Each writes its hardest objection from its own seat, paired with the methodological resolution it would accept. Run on the static catalogue spec Jun 3, 2026; a live council for any topic is at /try.

1. Skeptic #01
   Raised

   The accelerated trend in P99 Hs relative to median Hs may be an artifact of poleward migration of tropical cyclone tracks rather than genuine intensification of wave-generating conditions. Kossin et al. (2014, *Science*) documented a statistically significant poleward shift in peak tropical cyclone intensity of ~0.5–1° latitude per decade; this displaces cyclone-generated swell energy into mid-latitude ocean grid cells that previously recorded negligible extreme wave events, mechanically inflating P99 in those cells without any increase in storm intensity. Because P50 is dominated by background wind-sea and extratropical swell whose spatial footprint is comparatively stationary, the P99/P50 trend ratio can exceed 1.5 purely from geographic redistribution of cyclone influence, not physical wave climate intensification.

   Resolved

   Cross-reference every P99 Hs exceedance event in the ERA5 ECMWF wave reanalysis (or CMEMS multi-decadal wave product) against IBTrACS v04r00 best-track positions using a 500 km radius and ±72-hour temporal window to classify each event as cyclone-proximate or background; recompute P99 and P50 temporal trends with cyclone-proximate records excluded. If the basin-level extreme-to-median trend ratio drops below 1.5 toward 1.0 in the North Pacific and North Atlantic after exclusion, track migration is the confounding mechanism. A partial regression with annual mean cyclone landfall latitude (from IBTrACS) as a covariate in the P99 trend model, combined with a Chow structural-break test bracketing the 1985–2000 period of accelerating poleward migration, will distinguish redistribution from intensification: a significant covariate coefficient with a near-zero residual P99 trend slope would falsify the intensification claim.

2. Fact-Checker #02
   Raised

   ERA5 reanalysis and multi-mission satellite altimetry (TOPEX/Poseidon through Jason-3/Sentinel-6) both systematically underestimate P99 significant wave heights by roughly 10–20% in storm-dominated basins (North Atlantic, Southern Ocean) relative to in-situ buoys, while mean Hs bias is far smaller (~3–5%), meaning the P99 trend is structurally compressed toward zero relative to the P50 trend before any analysis begins. The inter-mission cross-calibration uncertainty across that altimeter constellation is ~0.1–0.2 m in absolute Hs, producing spurious long-term trend artifacts of order 0.02–0.05 m/decade — comparable to the real decadal signal sought. Compounding this, P99 sampling variance is inherently ~10× larger than P50 variance for the same record length: annual P99 values estimated from ~1,460 ERA5 6-hourly samples carry a 1-sigma uncertainty of 0.3–0.5 m, making the trend slope confidence interval extremely wide. The ratio metric (trend P99 ÷ trend P50) then becomes numerically unstable whenever the P50 trend approaches zero — precisely the scenario the hypothesis calls "interesting."

   Resolved

   Replace the ratio metric with an absolute trend-difference test (β_P99 − β_P50, both in m/decade) with uncertainty propagated via block-bootstrap resampling using a block length of ~30 days to respect swell autocorrelation, and pre-register the "supports" threshold against a Monte Carlo null distribution rather than the arbitrary 1.5 ratio. Apply the ESA CCI Wave Climate Altimetry homogenized multi-mission dataset (v2.0, QC flag = 1 only) or the CMEMS GLOBAL_OP_PHY_SWH_L4 product to ensure inter-mission bias correction, and cross-validate P99 trend estimates against NDBC/CDIP directional wave buoys in each basin to audit the reanalysis extreme-value underestimation bias. Use simultaneous quantile regression (e.g., R's quantreg with Koenker–Bassett standard errors) rather than two independent OLS regressions, so the difference in slopes is estimated jointly and the test for β_P99 > β_P50 uses the correct joint covariance rather than ignoring shared variance from the common time axis.

3. Researcher #03
   Raised

   The primary uncontrolled confounder is satellite altimeter fleet expansion over the study period. ERA5 and similar reanalysis products assimilate altimeter-derived significant wave heights from a constellation that grew from roughly one to two active missions in the early 1990s to six or more by the 2010s; because extreme Hs events are spatially and temporally localized, denser orbital sampling preferentially captures peak values while having negligible effect on the median, which is insensitive to occasional missed passes. This differential early-period underestimation of P99 relative to P50 generates a spurious positive slope in the P99 series and, consequently, an artificially inflated extreme-to-median ratio that could mechanically exceed the 1.5 threshold without any true physical intensification of wave climate.

   Resolved

   Restrict the primary analysis to in-situ significant wave height records from NOAA's National Data Buoy Center (NDBC; parameter "WVHT," available at ndbc.noaa.gov) selecting only stations with uninterrupted deployment exceeding 25 years, as fixed accelerometer-based buoys provide a stationary measurement platform immune to constellation changes. If basin-wide spatial coverage requires retaining ERA5 SWH (Copernicus CDS variable "significant_height_of_combined_wind_waves_and_swell"), add as a time-varying covariate the annual count of active radar altimeter missions—obtainable from the AVISO/CMEMS multi-mission along-track product metadata or EUMETSAT mission operation logs—and partial its influence out of the P99 trend via OLS before computing the slope ratio, treating robustness of the ratio to this covariate as a necessary condition for claiming extremal intensification.

4. Compliance-Guard #04
   Raised

   If the extreme-to-median ratio remains below the 1.5 SUPPORTS threshold yet the directional finding is cited in investor-facing materials or asset-valuation models, it could trigger liability under SEC Rule 10b-5 for material misstatement of physical climate risk in the context of offshore wind IPOs, green bonds, or infrastructure fund disclosures — particularly where "tail-risk re-pricing" language implies validated actuarial support that does not yet exist. Simultaneously, premature incorporation of these basin-specific P99 trend slopes into coastal or offshore structural design calculations would conflict with IEC 61400-3-1 (offshore wind turbine design extreme sea-state parameters) and API RP 2MET (metocean criteria for offshore structures), both of which require design wave heights to derive from peer-reviewed, statistically robust return-period analyses, not from hypothesis-stage regression outputs; misapplication could expose engineers of record and project sponsors to professional-liability claims under applicable PE licensure standards and potentially OSHA Process Safety Management obligations for offshore facilities.

   Resolved

   No design-load revision, securities disclosure, or insurance-pricing adjustment may reference this finding until the experiment formally crosses the SUPPORTS threshold (ratio > 1.5, statistically significant at p < 0.05 per basin) AND the regression results have been independently cross-validated against at least one exogenous reanalysis dataset (e.g., ERA5 or CMIP6 wave projections) and one in-situ buoy record per basin to confirm the /api/marine, /api/cyclones, and /api/ocean data fusion has not introduced systematic bias at the tail. Any interim communication to investors or project financiers must carry an explicit disclaimer — "This finding reflects a hypothesis under active investigation; the extreme-intensification ratio has not yet met the pre-registered SUPPORTS threshold and does not constitute validated design criteria under IEC 61400-3-1, API RP 2MET, or any adopted metocean standard" — and structural or actuarial reliance must be gated on formal adoption of revised return-period parameters by the relevant standards body before being embedded in bankable engineering reports or regulatory filings.

5. Falsification-Auditor #05
   Raised

   The ratio of two regression slopes (P99 trend ÷ P50 trend) is a statistically unstable quantity: when the P50 trend is near zero — entirely plausible in basins dominated by decadal oscillations such as ENSO or the NAO — the denominator approaches singularity and the ratio can swing well beyond 1.5 under the pure null, making the SUPPORTS threshold trivially reachable by chance. Bootstrap confidence intervals on individual quantile trends from 30–40 year altimetry records typically run ±0.5–1.0 cm/decade; propagated through the ratio, this yields ratio uncertainty of roughly ±0.4–0.8, meaning the FALSIFIES band (ratio ≈ 1.0) and the SUPPORTS band (ratio > 1.5) overlap within a single confidence interval for most basin-length combinations. Reanalysis ensemble spread (ERA5, MERRA-2, CFSR, JRA-55) adds a further ~20–30% dispersion to P99 Hs trend estimates, sufficient to shift the ratio estimate across both thresholds simultaneously without any underlying physical signal changing.

   Resolved

   First, run a Monte Carlo null experiment — drawing 10,000 synthetic 30-year Hs records from the observed residual autocorrelation structure with no imposed differential trend — to derive the empirical distribution of the ratio statistic and determine whether ratio > 1.5 is exceeded more than ~5% of the time under the null; if so, raise the SUPPORTS threshold accordingly and widen the FALSIFIES band to a defensible interval (e.g., ratio < 1.2 with the full CI below that bound). Second, replace the raw ratio as the decision variable with a direct Koenker–Bassett quantile regression framework that explicitly tests equality of the P99 and P50 time slopes via an F-test, using the ratio only as a post-hoc effect-size descriptor; this decouples the inference from denominator instability and provides a well-defined null rejection region that the data can actually enter.

Unlike the static stress tests above (synthesised against the frozen catalogue spec), this is what a 3-voice council found in the most recent biweekly review. Refreshed on the 1st and 15th of each month at 09:00 UTC. Each voice runs one bounded web search via Anthropic's web_search_20260209 tool, cites what it finds, and recommends a verdict. The verdict diverges from the curated catalogue status (monitoring) — the synthesis below explains why.

Agent draft incorporating the 9 cited findings from the live council above. Not auto-merged — surfaces here for human review. To accept, open a PR editing site/src/_data/hypotheses.json with the revised fields below. To reject, ignore and the proposal will refresh on the next council run.

1. monitoring

   May 30, 2026 · added to catalogue at status "monitoring"

• DNV's next triennial revision of offshore wind load standard ST-0437 would raise 50-year return-period design wave-load factors by an estimated 8–14% for sites north of 57°N, requiring structural reassessment of approximately 35 GW of installed North Sea fixed-bottom and floating capacity by 2028.
• Munich Re's NatCatSERVICE-informed offshore wind physical damage treaty pricing would widen rate-on-line by 40–70 bps for North Atlantic and Norwegian Sea portfolios within two annual January 1 renewal cycles.
• IMO's Maritime Safety Committee, acting on revised extreme sea-state return-period data, would amend minimum freeboard requirements under the 1988 Load Line Protocol by 3–6% for vessels certificated on North Atlantic and Southern Ocean routes, imposing an estimated $1.2–2.5B in fleet-wide survey and modification costs by the next 5-year special survey cycle.
• FEMA's Hazard Mitigation Grant Program would redirect 15–25% of annual coastal resilience allocations (approximately $400–700M per year) toward extreme-wave-resistant structural upgrades for V-zone assets within 36 months of a revised FEMA coastal wave-climate technical bulletin.
• Moody's Investors Service would apply a qualitative stress flag to coastal general-obligation bond issuers whose waterfront infrastructure design specifications predate 2015 wave-climate baselines, adding 10–25 bps of spread premium to an estimated $120B of outstanding U.S. coastal municipal debt at next rating review.