ML Ocean Metrics

Statistical analysis of climate models and ocean data.

November 10, 2026 • Research

Evaluating SST Anomalies via Bayesian Inference

Sea Surface Temperature (SST) anomalies are critical indicators of climate change. In this study, we apply a hierarchical Bayesian model to account for spatial autocorrelation in ocean buoy data.

P(θ|D) ∝ P(D|θ) P(θ)

Our results indicate a significant shift in the posterior distribution over the last decade, aligning with CMIP6 projections.

Fig 1: Projected SST anomaly index (2016-2026)

November 5, 2026 • Methodology

Handling Missing Data in Oceanographic Time Series

Missing data is a constant challenge in environmental monitoring. We compare multiple imputation by chained equations (MICE) against Gaussian Process regression for filling gaps in ARGO float datasets.

x̂_t = αx_t-1 + (1-α)x̂_t-1

The exponential smoothing approach (shown above) proved robust for short gaps, while GP regression outperformed for long-term structural missingness, particularly in regions with high eddy kinetic energy.