Temporal modeling of anomalous coastal sea-level values using synoptic climatological patterns

Short-term changes in sea level can have substantial impacts on coastlines, and increases in coastal flooding have been observed as the mean sea level continues to climb. While extreme events such as hurricanes have been well studied in terms of their impacts on anomalous sea level values, anomalous sea levels due to less extreme atmospheric events have been less well studied, despite the increases in minor, or "nuisance" flood events, that have occurred, many under non-stormy conditions. To better understand the influence of the atmosphere on coastal water levels, we assess the relationship between short-term atmospheric circulation patterns and anomalous coastal sea-level values for all oceanic tidal gauges in the conterminous United States for the period 1979-2016. Atmospheric patterns are identified using self-organizing maps (SOMs) for four variables, obtained from the North American Regional Reanalysis (NARR): sea-level pressure, 10-m wind, 850-mb temperature, and 700-mb geopotential height. We then reconstruct the time series of anomalous sea level through non-linear autoregressive models with exogenous input (NARX models), an artificial neural network (ANN)-based time series model using the four SOMs as well as their principal components. Results show that these four atmospheric variables can successfully model sea level, with a correlation between model and observation using a closed-loop (open-loop) architecture of 0.83 (0.64), with a median absolute error of 3.34 (4.90) cm. The model generally performs better in winter than summer, and along the Pacific Coast than the Atlantic and Gulf Coasts. By using the NARX methodology, we intend to next assess its utility as a forecasting tool.