Skill Assessments of Seasonal Sea Level Predictions Using the Operational North American Multi-Model Ensemble Models

Coastal inundation risks, both in frequency and severity, are greatly increasing due to rising sea levels in the current changing climate. To assess the current status of predicting such risks, we evaluated the seasonal sea level forecasts of the operational models from the North American Multi-Model Ensemble (NMME) with respect to tide gauge station records, both to benchmark what the current operational skill is and to understand the uncertainty and difficulty in producing skillful seasonal coastal sea level predictions. The NMME models we evaluated include the ECCC CanESM5 and GEM5.2-NEMO, NCEP CFSv2, NCAR CCSM4 and CESM1 (run at RSMAS), NOAA/GFDL SPEAR, and NASA/GMAO GEOSS2S. We additionally assessed the skill of ECMWF's SEAS5, whose initialization includes satellite altimetry observations, unlike the NMME forecast models. To this end, we used the hindcast data over the 1991-2020 period and calculated deterministic and probabilistic skills for lead times up to 9 months, verified at all North American tide gauge stations and against GLORYS, a high-resolution (1/12º) oceanic reanalysis. We then assessed the impact of long-term trends, the inverted barometer effect, and vertical land motion on prediction skill.