# OceanCBM: A Concept Bottleneck Model for Mechanistic Interpretability in Ocean Forecasting

Source: [OceanCBM: A Concept Bottleneck Model for Mechanistic Interpretability in Ocean Forecasting](<https://www.sota2.com/research/paper/6a055713dbc836b87d828c81>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-15T05:00:04.680Z

Authors: Sanah Suri, Kieran Ringel, Maike Sonnewald

Published: 2026-05-12T18:29:45.000Z

## Abstract

Extreme ocean phenomena are challenging not only to predict but to diagnose, as accurate forecasts alone do not reveal the underlying physical drivers. While recent machine learning approaches achieve strong predictive skill, they remain largely opaque and provide limited guarantees of fidelity to ground-truth physics. We introduce OceanCBM, the first concept bottleneck model (CBM) for spatiotemporal prediction and mechanistic interrogation of ocean dynamics. OceanCBM uses mixed supervision to predict mixed layer heat content, a key precursor of marine heatwaves, while routing information through an intermediate layer of prescribed concepts derived from geophysical fluid dynamics and a 'free' concept. This design imposes soft physical structure without over-constraining the model, and the free concept both regularizes concept predictions and captures residual physical processes. Across ensemble initializations, we show that mixed supervision yields consistent mechanistic representations, whereas prediction-only and prescription-only baselines learn highly variable latent structures despite similar predictive performance. OceanCBM achieves interpretable, physically grounded representations without sacrificing skill, explicitly characterizing the interpretability-performance trade-off.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Heat Flux Entrainment Prediction on Ocean Dataset](<https://www.sota2.com/research/sota/heat-flux-entrainment-prediction-on-ocean-dataset.md>): Heat Flux Entrainment Prediction; Ocean Dataset; ACC (DJF): 75.32.

- [Mixed Layer Depth (MLD) Tendency Prediction on Ocean Dataset](<https://www.sota2.com/research/sota/mixed-layer-depth-mld-tendency-prediction-on-ocean-dataset.md>): Mixed Layer Depth (MLD) Tendency Prediction; Ocean Dataset; ACC (DJF): 83.84.

- [Vertical Shear Prediction on Ocean Dataset](<https://www.sota2.com/research/sota/vertical-shear-prediction-on-ocean-dataset.md>): Vertical Shear Prediction; Ocean Dataset; ACC (DJF): 0.5965.

- [Buoyancy Frequency Prediction on Ocean Dataset](<https://www.sota2.com/research/sota/buoyancy-frequency-prediction-on-ocean-dataset.md>): Buoyancy Frequency Prediction; Ocean Dataset; ACC (DJF): 0.8014.

- [Mixed Layer Heat Content (MLHC) Prediction on Ocean Dataset](<https://www.sota2.com/research/sota/mixed-layer-heat-content-mlhc-prediction-on-ocean-dataset.md>): Mixed Layer Heat Content (MLHC) Prediction; Ocean Dataset; ACC (DJF): 0.9116.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.12639>)

[PDF](<https://arxiv.org/pdf/2605.12639.pdf>)