Storm surge


Storm surges can be understood as water piling up along the shallow shore, due to strong onshore winds. This type of disaster creates severe damage to low coastal regions such as the US Gulf Coast, Bangladesh, India, and Vietnam.

Early dynamic models of storm surge were reviewed by Dean and Dalrymple (1991) in which the bed shear stress counterbalances the wind shear stress. The important factors to affect the surge height are wind speed and continental shelf depth. Extreme wind speed occurs in tropical cyclones, therefore the surge height is dependent on a number of factors such as the pressure deficit, the travel speed of the cyclone eye, the radius of maximum wind, the orientation of cyclone track w.r.t. shoreline, and the continental slope (Nielsen 2009).

Numerical modelling has been an effective tool to compute the surge height -- not only the maximum but the complete distribution of surge level along the coastline. The SLOSH model had been used by the US to predict storm surges. Later, FEMA (2016) devised a guideline in using numerical models for surge prediction. A popular and robust model is ADCIRC, which uses the finite element method. However, many other hydrodynamic solvers can also be used for simulation, such as Delft3D-Flow, MIKE21-HD, or TU-Flow.

Calculating the design storm surge corresponding to a given return period requires statistical analysis. Due to a large set of variables and a high degree of uncertainty, the standard approach is to generate many realizations of cyclones, then calculate the surge heights and perform a frequency analysis to determine the required extreme surges.


Photo of a hurricane


Storm surge: Key West, FL


Hurricane wind field by Young (2017)


Unstructured grid of hurricane surge study by Westerlink et al. (2008)

Improvement of existing tools


Wind field

Recent models of storm surge often incorporate better wind field reconstruction. The cyclone, in its mature stage, can be represented as a translated vortex. The Holland vortex is particularly common, and Young (2017) recommended the use of a double vortex configuration to achieve more realistic cyclones. This should be more convenient to be implemented in an open source system so that any improvement in algorithm or data representation can be taken into effect.


Linking surge and wave models

There have been enormous efforts in linking wave and tide models, for example ADCIRC and SWAN, or WaveWatch. By using a spectral wave model, the distribution of wave set-up along the coast can be realised. Wave set-up is an important contribution to total surge height, but was often omitted in many previous studies.


Model domain extent and spatial resolution

The domain should be large enough to cover the entire meteorologic 'system' or 'structure', so as to reflect the behavior of the whole wind field. On the other hand, it should contain very fine grid cells along the coastal zone of interest. For example, an unstructured grid of 300k+ nodal points and 600k+ elements, with minimum cell size of ~100 m was set up by Westerink et al. (2008) to give detailed storm surge pattern along the Louisiana coastal plain.

The case of Vietnam


Vietnam has a 3200-km shoreline, which is relatively long compared to the country's area. The authority compiled a table of extreme water levels corresponding to various return periods. Such distribution of water level versus latitude for the northern coast of Vietnam is presented in the figure beside.


The Mekong Delta is home to ~17 million people and although not many typhoon affected this area in the past, the people's unawareness leads to the severe damage once a typhoon happens, such as Typhoon Linda (1997).


Typhoon statistics

Tropical cyclones were summarized from the JTWC database; their intensities are ranked and their tracks drawn on a map. The wind speed had been analyzed to find out its cumulative distribution (figure below), and from that, the wind speed associated with a given occurence frequency is obtained.

Furthermore, the joint distribution between landing location x0, landing orientation φ, and forward-moving speed, Uf (figure beside). Apparently these three variables are independent.


Wind speed distribution



Hydrodynamic model

A nested model was set up, which includes relevant processes (wind, wave, tide, sea-level rise). The model was thoroughly calibrated against measured water level data and surveyed surge level data. The chart is shown beside.



New envelope of water level

The computation result is to produce water level "envelopes" that are higher than those presented in Vietnam's standard (corresponding to the same return periods). An important part of this is attributed to sea-level rise in computation.


Surge height by latitude for various return periods, along the Vietnam coast


Joint plot for the landfall location (x0), landfall orientation phi, and forward moving speed of the storm Uf Calibration on water level


Maximum water level along Mekong delta coast


Photo credits: Free-Photos, David Mark (Pixabay) •