Meeting Documents

Indications of Improved Seasonal Sea Level Forecasts for the U.S. East and Gulf Coasts Using Ocean-Dynamic Persistence

Abstract

Forecasting seasonal sea level variability along the U.S. East and Gulf Coasts is challenging, with little to no skill demonstrated using state-of-the-art climate models. For many coastal locations, observed-damped persistence of water level monthly anomalies, often used as a benchmark for seasonal sea level forecast skill, outperforms climate models with atmosphere-ocean coupling. The skill using damped persistence arises from local ocean memory; however, incorporating non-local ocean memories (i.e., dynamics) could enhance sea level forecasts. We are investigating the influence of ocean-dynamic persistence on forecasting monthly sea level anomalies by using baroclinic ocean models forced with climatological atmospheric conditions. We first use the 1° ECCO model, which is initialized monthly from 1993–2017. From each initialization, ECCO runs forward for 12 months, forced by the mean atmospheric climatology. Retrospective forecast skills, compared to observations from water level gauges and satellite altimetry, indicate improvement compared to observed-damped persistence at most coastal locations south of Cape Hatteras, but minimal improvement for the Northeast Coast. However, there is also minimal improvement in root-mean-square error everywhere, possibly due to reduced variability in dynamic persistence associated with its climatology forcing and coarse model resolution. To explore the potential influence of model resolution on dynamic persistence skills, we are conducting similar experiments using a 0.25° ocean model, which may resolve coastal oceanic processes better. Our results indicate potentially improved seasonal sea level forecasts by using ocean-dynamic persistence instead of observed-damped persistence. This work suggests an opportunity for using the skill of ocean-dynamic persistence as a higher benchmark for assessing more sophisticated seasonal forecasting models.