Daily maximum storm surge levels dataset

Published: 07-12-2020| Version 1 | DOI: 10.17632/y9sdry74xz.1
Contributor:
Tao Ji

Description

We used the Data Unification and Altimeter Combination System (DUACS) 2014 merged data and dynamic atmospheric correction (DAC) data products (it is produced by the CLS Space Oceanography Division using the Legos’ MOG2D model) released by the Archiving Validation and Interpolation of Satellite Oceanographic Data Center (AVISO) (http://odes.altimetry.cnes.fr/), sea level pressure and wind data from the European Centre for Medium-range Weather Forecasts (ECMWF) (https://www.ecmwf.int/) were used to develop a multivariate statistical model (reconstructing a statistical relationship is to estimate the daily maximum storm surge levels (predictand) based on the local atmospheric forcing field conditions (predictor)) by using the Multivariate Empirical Orthogonal Function (MV-EOF) [Cid et al., 2017; Cid et al., 2018; Ji et al., 2020]. Then, the Geographical Differential Analysis (GDA) calibration method was used to correct the daily maximum storm surge levels recorded by tide gauge and hydrometric station observations. Finally, a high spatial resolution (0.25°) and a high-precision daily maximum storm surge dataset for a long time series (1958-2016) was generated [Cheema and Bastiaanssen, 2012].

Files

Steps to reproduce

we use the two dominant spatial grid data of SLP and 10 m UV wind speed, mentioned above. Considering that the resolution of both the predictand and the predictor variable is 0.25°, we used the grid points surrounding the target point in an area of 1°×1° (4×4 grid cells) centered at the target point location. Considering that the DAC, SLP, and wind speed data have a 6-hour temporal resolution, it was necessary to generate their daily value (maximum and average) at each grid point. Thus, we constructed the statistical relationship among the mean daily gridded SSH (summing the corresponding gridded SLA and DAC), the mean daily gridded SLP, and wind speed from 1993 to 2016.This method can be used to construct a multivariate regression model that fits the statistical relationship between the mean daily storm surge level (predictand) and the principal components (PCs) of the mean daily wind speed and SLP (predictor). After defining the predictor area and variables, the next step is to perform a principal component analysis (PCA) to reduce the dimensionality of the predictor while preserving with and preserve the PCs that explain 95% of the variance of the data sample. The PCA projects the raw data on a new space, and searches for the maximum variance of the sample data. The vector of on the new space is defined by the eigenvectors of the data covariance matrix using a multivariate empirical orthogonal function (MV-EOF). The next step is to establish the statistical model between the predictor PCs and the storm surge levels (predictand). The fitted process is normally used in the multivariate regression model between daily mean surge levels (predictand) and the PCs of the mean daily wind speed, and SLP (predictor). This paper aims to reconstruct storm surge heights in the southeast coastal area of China. The surge height in the nearshore area is affected by sea level pressure and wind speed, and is also seriously affected by the features of bathymetry, the path of the storm, and the geometric properties of a water body. The tide gauge observations are the result of the combination of all these factors. For all these reasons, we applied the GDA calibration method to correct the daily maximum storm surge levels obtained by using the regression model. It is guaranteed that the corrected coastal surge heights reflect, to some extent, the effects of the influencing factors on storm surge levels.