GeoPython 2020

Statistical prediction of Monthly Sea Level
2020-09-22, 11:30–11:50, Room 2

Statistical prediction of monthly mean sea level using a linear regression model of monthly anomalies of principal components for two regional predictors. The tool has been implemented in several tide gauge locations, and results are presented in JupyterLab notebooks.

Coastal flooding in nearshore areas is ever more subject to analysis due to an increasing risk of potential impacts specially in highly populated areas and coastal infrastructures. In terms of quantifying the damage capacity of flood events, the concurrent total water level is of great importance. The total coastal sea level is computed as the aggregation of several components: sea level rise, sea level anomalies, astronomical tide, storm surge, setup and swash. Mean sea level anomalies can be accountable for variations of sea water density, marine circulations, setup from sustained trade winds and non-period climatic fluctuations, as the El Niño Southern Oscillation (ENSO), among others.

In this work a statistical tool has been developed to obtain an estimation of the Monthly Sea Level series at the US west coast, and small islands in the Pacific (Kwajalein, Majuro, Guam) from long tide gauge station records. Two regional predictors are used: (a) Sea Surface Temperature (SST), and (b) Sea Level Pressure fields (SLP). Existing global datasets ERSST v4, and CFSR reanalysis over several decades have been used respectively.

The proposed methodology performs PCA to reduce dimensionality of both regional predictors, then a linear regression model is applied to the Monthly Mean Sea Level and the Principal Components of monthly anomalies of both predictors. Finally a K-fold cross validation procedure evaluates the skill of the model.

This methodology has been implemented with a Python library composed by different modules. Moreover several JupyterLab notebooks have been produced for each site location so that the results may be visualized in a user-friendly format.