Extending SST Anomaly Forecasts Through Simultaneous Decomposition of Seasonal and PDO Modes
Rameshan Kallummal

TL;DR
This paper introduces a novel multivariate linear model that simultaneously decomposes seasonal and PDO modes to improve North Pacific SST forecasts, achieving unprecedented accuracy and extending forecast skill beyond 36 months.
Contribution
The study presents a new coupled decomposition approach that enhances SST forecast accuracy by integrating seasonal cycles and PDO dynamics within a multivariate linear framework.
Findings
Forecast skill extends beyond 36 months.
PDO projected to stay negative until late 2026.
Reduced likelihood of marine heatwaves in eastern North Pacific.
Abstract
We present a new approach to forecasting North Pacific Sea Surface Temperatures (SST) by recognizing that interannual variability primarily reflects amplitude changes in four dominant seasonal cycles. Our multivariate linear model simultaneously captures these amplitude-modulated seasonal cycles along with the Pacific Decadal Oscillation (PDO), which naturally emerges as an intrinsic feature of the system rather than a separate phenomenon. Using sixteen-dimensional regression based on four spatially distributed time series per variable, the model delivers unprecedented forecast accuracy for both interannual amplitude modulations and PDO evolution, maintaining skill beyond 36 months -- a substantial improvement over current operational and research forecasts, including machine learning methods. Predictions initialized in 2024 project that the PDO will remain in its negative phase through…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsClimate variability and models · Tropical and Extratropical Cyclones Research · Oceanographic and Atmospheric Processes
