Feng, X., Widlansky, M.J., Lee, T., Wang, O., Balmaseda, M.A., and Zuo, H. (2024)
Presented at:
Ocean Sciences Meeting 2024Accurate seasonal sea level forecasts are still a challenge in many locations for ocean forecasting systems. Notably, forecast skills along the U.S. East Coast are much lower than in most tropical and subtropical open oceans. For many tide gauge locations on the East Coast, seasonal forecast skills (e.g., at lead-6 month) are no better than observed persistence. Since ocean memory has been shown to be a source of climate predictability, we investigate the influence of ocean-dynamic persistence on forecasting sea level monthly anomalies. We use the Estimating Circulation and Climate of the Ocean (ECCO) system initialized monthly from the ECCO ocean state estimate version 4 release 4 (ECCOv4r4), spanning from January 1992 to December 2017. Following the initialization, the model runs forward for 12 months under climatological atmospheric conditions. We evaluate the monthly sea level anomalies with forecast leads up to 12 months by comparing them with observations from tide gauges and satellite altimetry. We compare the retrospective forecasting skill using the ECCO ocean-dynamic persistence experiment against the persistence of the observations following the known damping timescale. Skill metrics based on the anomaly correlation coefficient (ACC) show that the ECCO ocean-dynamic persistence model performs better than using observed damped persistence at forecasting monthly sea level anomalies observed by most tide gauges. For example, at Charleston for up to 10 months lead, the ECCO ocean-dynamic persistence model shows higher ACC than using observed damped persistence. Compared with altimetry, a higher ACC for most of the Northwestern Atlantic Ocean is also observed in the ECCO dynamic persistent model. We will also compare these forecast skills with those from the ECMWF SEAS5, which is a leading climate forecasting system for predicting monthly sea level anomalies. Interestingly, we find no improvement in the root-mean-square error for the ECCO ocean-dynamic persistence retrospective forecasts, which is perhaps because of the reduced monthly variability in its coarse configuration (nominally 1°). This study suggests that ocean-dynamic persistence, which is computationally efficient to simulate in a forward mode, can provide a more stringent metric for assessing coastal sea level forecast skills.
