Louisiana Deep Sea Fishing — DR08 hypoxia case, v1.0.0, researched 2026-10-21.
Scope: original descriptive calculations, not a new biological experiment or operational forecast.

ANNUAL COMPARISON
Use every year 2021–2026 inclusive: six NOAA ensemble forecast means, source-reported 95% confidence intervals, and subsequent annual survey news-release areas (square miles). Source-specific URLs and May nutrient loads are in evidence.json and annual-comparison.csv. Intervals are transcribed from original PDF text and visually checked. Difference = observed minus forecast. Mean absolute error is the arithmetic mean of absolute differences, not a forecast for next year. These six cases do not establish statistical calibration or causal nutrient/weather effects. Model ensembles and survey conditions vary.
Five-year means use the retained rounded annual values. 2021–2025 = 4754.8; 2022–2026 = 3754.4. NOAA’s 2026 release reports 3745, a 9.4-sq-mi discrepancy whose cause is unconfirmed. Do not silently change a year to reconcile. Goal ~1900 applies to a five-year mean, not one cruise.

OXYGEN OBSERVATIONS
Historical Oregon II cruise332, Summer2019: all160 rows of NOAA original CSV and matching public FeatureServer response. Regional box declared in protocol.json before subset calculation: 94–89W,28–30N inclusive; geographic analysis window, not jurisdiction. Actual subset dates June16–July1,2019,48 observations. Require matching cruise year/season, unique ship/cruise/station/cast identity, finite nonnegative oxygen, valid coordinates, positive water depth. No value exclusions in2019. Preserve 0mg/L as reported, not an invented missing-value replacement. No per-record sensor QA flag in the public response; source-consistency checks do not independently calibrate sensors. No interpolation, area estimate, trend, fish-data spatial join, vertical-profile reconstruction or catch model.
Conventional hypoxia <=2mg/L, evaluated on unrounded values. Summaries are station counts and medians/ranges, not seafloor percentages. Water-depth bins [0,20),[20,50),[50,200] metres; regional counts8,18,22 and hypoxic4,2,0. Sample depths are retained separately:95/160 exceed reported total water depth, maximum discrepancy+3m, full signed range-4.9 to+3.0m. Reference discrepancy unresolved; do not derive exact bottom clearance. Sensitivity check using sample depth for regional bin assignment gives the same bin and hypoxic counts. Spatial/date confounding and unequal sampling prevent a depth-only causal conclusion.
The latest Summer2026 service query returned189 rows; all oxmgl values null. All189 excluded from oxygen calculations, not zero-filled. The station-location ZIP has no oxygen attribute. This says nothing about whether measurements exist in other archives. Older full-profile links returned HTML/error pages, not usable CNV data. No full profiles claimed as acquired.
The original CSV is an independent cross-check of the saved API representation: all160 identities, oxygen values and positions matched within source display precision. Executable reproduction uses saved API JSON, not a new network query.

PUBLISHED FISH EVIDENCE
Glaspie et al.2019, DOI10.1007/s12237-019-00626-x; NOAA accepted manuscript25959, methods/results/discussion plus tables inspected. CPUE categories use a fitted1.15mg/L breakpoint;91trawls/1707min above,46trawls/943min at or below. Fish-community and diet breakpoints1.15[0.99,3.24] and1.71[0.98,3.33]mg/L with source95%bootstrap intervals. They are statistical community changes, not lethal/safe limits. Diet post-hoc tests reported in Results p13, supported by supplementary tables in retained manuscript; different oxygen breakpoint from CPUE. Table1 lists144red snapper caught, not a replicated adult reef experiment. Same-species stomach observations were averaged within trawls. Nonclosing bottom net can collect fish higher in column during deployment/retrieval. CPUE is gear/availability dependent. Association cannot identify migration versus death or turn food changes into measured growth. 2006–2008 fish cruises are never joined to2019oxygen merely because both concern the Gulf. Accepted manuscript’s raw-code link remains an unresolved placeholder; no fish model rerun.

REPRODUCTION
Save reproduce.mjs,protocol.json,evidence.json,stations-2019.json,stations-2026.json together; Node24: node reproduce.mjs. No dependencies/network. Outputs analysis.json and four derived CSVs. Original oxygen-2019.csv retained separately. Plotting source retained in repository research/hypoxia/2026-10-21/plot.py; it uses matplotlib, figures made from factual extracts rather than republished source figures.
Independent fisheries/statistical review pending. Dated local-review case; no deployment claimed. Update when annual areas, forecast products, raw data or relevant research are revised.