Bridging large-scale and coastal variability to improve seasonal sea level predictions along the U.S. and Canadian West Coast

Coastal communities are increasingly vulnerable to long-term sea level rise and fluctuations driven by climate variability. While recent advances in coupled climate models enable sea level predictions several months in advance, further efforts are needed to assess and enhance seasonal prediction of coastal sea level. In this study, we evaluate seasonal prediction skill for large-scale and coastal sea level along the U.S. and Canadian West Coast using multiple forecast systems. Prediction skill peaks in the tropical Indo-Pacific and extends into the eastern North Pacific, declining from south to north along the coast. Using self-organizing maps (SOMs), a machine learning technique, we identify sources of large-scale sea level variability and predictability in the eastern tropical and North Pacific, closely linked to the El Niño–Southern Oscillation. Finally, we improve coastal sea level predictions from dynamical models by leveraging the connection between large-scale and coastal sea level through SOM-reconstructed and model-analog approaches.