DR07 / Trout cold exposure and winterkill / v1.0.0 / 2026-10-21

Question: can one cold-water reading establish winterkill, and is deeper water automatically a refuge?
Answer scope: primary North Carolina field evidence, carefully labeled published lab abstract, LSU abstract context, and original full-January2025 Grand Isle point-station exposure analysis. No fish-kill count, Louisiana mortality forecast, catch forecast, safe-temperature boundary or refuge map.

NOAA CO-OPS station8761724, water_temperature, metric Celsius, GMT, 20250101-20250131 inclusive. Full response contains7440timestamps at6minute spacing. 7433finite nonblank readings have flags0,0,0;7blank readings have1,1,1. Flags are X(max expected),N(min expected),R(rate of change), per official responseHelp.html Water Temperature. Source values/flags retained. No water-level q=verified designation is supplied here. Passing our filter is not independent instrument validation.

Window[2025-01-01T00:00Z,2025-02-01T00:00Z). All expected6minute slots generated; duplicates, off-grid and out-of-window timestamps rejected. Missing, nonnumeric or nonzero/unknown flags produce null. No interpolation, forward fill or zero substitution. Plot includes gaps. All reported dates/bins use UTC; local CST isUTC-6. This is one January, not a typical winter or climate trend.

Preselected context thresholds3,5,7C from North Carolina literature are illustration settings, not validated Louisiana lethal limits. Strict below: equality does not count. Each accepted observation below a threshold contributes0.1sample-bin hours. Longest consecutive sample-bin run resets at an equal/warmer or missing/flagged slot. This duration is a discrete support convention; between-sample crossing times and continuous exposure are unknown. Total and longest runs are lower in missing periods but our bin convention is not a strict lower bound on continuous real-world duration.

Cold-degree-hours=sum(max(threshold-T,0)*0.1) over included bins, unitC*hours. These are discretized exposure metrics, not mortality. The source field paper's CDD uses daily mean temperature over its at-risk winter window (C*days). Our six-minute January index is not the same statistic or a re-fit of that paper. Fixed blocksJan1-7,8-14,15-21,22-28,29-31 chosen for complete monthly display; last block only3days, not an equal-length week. Daily means/min/max use accepted points; coverage retained for each day/block, missing bins not estimated.

Sensor metadata retrieved2026-10-21: E1 Water Temperature, elevation-4.055118feet relativeMLLW. Fixed sensor elevation relative to tidal datum is NOT instantaneous depth below water surface or bed depth. Historical Jan2025 mounting/configuration not independently confirmed. One point cannot represent marsh-interior temperature, paired shallow/deep layers or fish movements. No vertical refugium inference.

Field extraction: Ellis,Buckel,Hightower2017,MEPS564:145-161,doi10.3354/meps11985. Table1 p152 contains tagged counts37,10,26,45; text/model exclusions yield at-risk34,9,24,41. Natural deaths7,6,1,0; the1mild-winter Pungo death was inferred predation, not cold. All others assigned cold-associated by source interpretation. Variable time at risk, emigration, seven-day censoring, sample sizes and fate assumptions preclude naive population mortality fractions. CDD source values80.4,144.3,0.6,0.9 from p152 retained; not recomputed or plotted as independent replicated experiments. Source statistical credible intervals differ from confidence intervals; no new model or interval fitted. Supplement retained, code/relocation scope inspected, no OpenBUGS run or raw-data reconstruction.

Lab: Ellis etal2017 JEMBE490:42-51 doi10.1016/j.jembe.2017.01.010, published primary abstract/highlights inspected. Full final paper unavailable. Rapid cooling loss of equilibrium<=4C is not death. Abstract reports no survivors at3C after2days, tolerance near5C for~5days before survival declined, high but incomplete survival at7C over10days;salinities10/30, no measured salinity mortality effect in prolonged trials. No n,percent-survival matrix, cooling-rate estimate or new CI extracted. Do not replace final published results with dissertation versions.

Louisiana: Baer2019 LSU dissertation repository abstract only,doi10.31390/gradschool_dissertations.4831. 211fish/90receivers2012-2014 movement context; detection efficiency influenced by conditions. Not cold-refuge experiment or Louisiana mortality calibration. LDWF2025 indexed official preliminary/news snippets inspected as contextual warnings/investigations; full report inaccessible, no kill-total extraction.

Reproduce offline with Node24: save temperature.json,protocol.json,evidence.json,reproduce.mjs together; node reproduce.mjs. Produces analysis.json,observations.csv,threshold-summary.csv,block-exposure.csv,daily-exposure.csv,field-cohorts.csv. No packages or network. Plot source in repo site/scripts/plot-trout-cold.py uses matplotlib; original chart, no publisher artwork. NOAA original public snapshots/hashes in provenance.json. Paper/supplement outside public; copyright respected.

Maintenance: check source corrections, NOAA revisions/metadata and guidance before publication and every6months. Version changes and retain original inputs. Independent fisheries/statistical review pending. No production deployment or replacement of the already prepared FTP ZIP.