On the predictability of high water level along the US East Coast: can the Florida Current measurement be an indicator for flooding caused by remote forcing?

Recent studies show that in addition to wind and air pressure effects, a significant portion of the variability of coastal sea level (CSL) along the US East Coast can be attributed to non-local factors such as variations in the Gulf Stream and the North Atlantic circulation; these variations can cause unpredictable coastal flooding. The Florida Current transport (FCT) measurement across the Florida Straits monitors those variations, and thus, the study evaluated the potential of using the FCT as an indicator for anomalously high water level along the coast. Hourly water level data from 12 tide gauge stations over 12 years are used to construct records of maximum daily water levels (MDWL) that are compared with the daily FCT data. An empirical mode decomposition (EMD) approach is used to divide the data into high-frequency modes (periods T < similar to 30 days), middle-frequency modes (similar to 30 days < T < similar to 90 days), and low-frequency modes (similar to 90 days < T < similar to 1 year). Two predictive measures are tested: FCT and FCT change (FCC). FCT is anti-correlated with MDWL in high-frequency modes but positively correlated with MDWL in low-frequency modes. FCC on the other hand is always anti-correlated with MDWL for all frequency bands, and the high water signal lags behind FCC for almost all stations, thus providing a potential predictive skill (i.e., whenever a weakening trend is detected in the FCT, anomalously high water is expected along the coast over the next few days). The MDWL-FCT correlation in the high-frequency modes is maximum in the lower Mid-Atlantic Bight, suggesting influence from the meandering Gulf Stream after it separates from the coast. However, the correlation in low-frequency modes is maximum in the South Atlantic Bight, suggesting impact from variations in the wind pattern over subtropical regions. The middle-frequency and low-frequency modes of the FCT seem to provide the best predictor for medium to large flooding events; it is estimated that similar to 10-25% of the sea level variability in those modes can be attributed to variations in the FCT. An example from Hurricane Joaquin (September-October, 2015) demonstrates how an offshore storm that never made landfall can cause a weakening of the FCT and unexpected high water level and flooding along the US East Coast. A regression-prediction model based on the MDWL-FCT correlation shows some skill in estimating high water levels during past storms; the water level prediction is more accurate for slow-moving and offshore storms than it is for fast-moving storms. The study can help to improve water level prediction since current storm surge models rely on local wind but may ignore remote forcing.