CAT_06 // Platform Updates

Thermocline Alpha: Platform Release Notes

Published
Reading time
1 min
A research vessel on the open ocean
Photo: @noaa via Unsplash

The thermocline_alpha model is now the default thermal stratification model on every Bathys deployment. What changed, what to check, and how to roll back.

bathys model apply thermocline_alpha --active is now the default for new deployments and is available to every existing deployment through the governed update channel. These are the release notes.

What thermocline_alpha does

The model estimates the depth, thickness and sharpness of the thermocline from the temperature profile reported by each mooring chain, then evaluates the estimate against the historical tide-driven baseline for that location. Its output feeds two consumers: the anomaly console, which raises a Thermal Warning when the estimated thermocline departs from baseline, and the salinity anomaly detector, which uses thermocline depth to separate genuine salinity events from stratification changes.

Changes from the previous model

  • Baseline is now computed from a rolling 90-day window instead of a fixed seasonal climatology. Deployments in regions with strong interannual variability should see fewer false Thermal Warnings.
  • The model is now evaluated per mooring chain rather than per array, so a localized intrusion no longer suppresses or triggers warnings array-wide.
  • Output includes a confidence field. Warnings below 0.6 confidence are logged but not surfaced to the console.

What to check after applying

  1. Run bathys status and confirm the model line reads thermocline_alpha (active).
  2. Review the first 24 hours of Thermal Warnings in the event log. Expect fewer than before; if you see more, your region likely has a strong sub-seasonal signal and you should widen the baseline window with --baseline-days 120.
  3. Confirm the salinity anomaly detector is still enforcing live thresholds. The release notes for the detector are unchanged.

Rollback

bathys model apply thermocline_legacy --active restores the previous model without affecting historical data. Both models can run side by side in shadow mode with --shadow for comparison before committing.

Known limitations

The model has not been validated under ice. Deployments north of 78° or south of -65° should keep thermocline_legacy active until the polar validation set is published later this year.