Coastal flood risk assessment with ROMS?

[stef]

Hi,

May I kindly ask for your opinions on using ROMS for storm surge modelling? I'm particularly interested in risk assessment, similar to what has been done by FEMA for coastal flood hazard studies. Here is a link to the New York - New Jersey region:

http://www.region2coastal.com/

They are using ADCIRC, which is a FEM model that's able to simulate flooding. I'm aware that ADCIRC originated from an engineering environment and is purpose-built for these kinds of studies. However, it seems that the flooding capability is only used in an intermediate step in studies like the one referred to above. To obtain the final flood height for the risk assessment study, they define a series of one-dimensional transects orthogonally to the shoreline, and use a one-dimensional wave model along those transects. The 1D model is forced at the nearshore end with "stillwater height" (including wave setup) and some wave statistics (I think significant wave height and peak wave period), all of which are obtained from the fully coupled ADCRIC-SWAN solution. Effectively, the 1D propagates the ADCIRC-SWAN solution from the ocean onto the land.

In principle, the same could be done with ROMS, i.e. instead of worrying about a proper wetting algorithm, one could simply do a one-way coupling to a 1D model or even more sophisticated 2D hydraulical models that are specifically designed for inundation studies.

One problem is that using vertical-wall boundaries will tend to overestimate the depth, since the water cannot "drain" landwards. In the ADCIRC simulations, this problem does not exist because they do model inundation with ADCIRC in the intermediate step (they just don't use it in the final results). But then, ROMS has some kind of wetting algorithm too, right? So this could be used in an intermediate step to get at least an approximation for the "draining".

Has this been done, or at least considered by anybody?

Thanks for your feedback.

[stef]

I should add that an example for the 1D transects can e.g. be found on page 25 in:

Federal Emergency Management Agency, 2013: Flood Insurance Study (preliminary). City of New York, New York.

https://hazards.fema.gov/femaportal/pre ... odId=16353

This publication should be in the public domain, so I might as well post the figure here.


I'm also attaching a figure that shows the inland grid of ADCIRC for the intermediate step. This is from page 4 of:

Federal Emergency Management Agency: Region II Storm Surge Project - Mesh Development


https://data.femadata.com/NationalDisas ... opment.pdf

[jcwarner]

ROMS has a wetting/drying capability and we use many roms+swan coupled simulations regularly. I know others have looked at inundations during storms, for example:

Beudin, A., Ganju, N. K., Defne, Z., & Aretxabaleta, A. L. (2017). Physical response of a back-barrier estuary to a post-tropical cyclone. Journal of Geophysical Research: Oceans, 122(7), 5888–5904. https://doi.org/10.1002/2016JC012344

There are many models that do storm surge, ADCIRC is designed for that, and it is coupled to swan. It works well in these coastal settings, but it is really only 2D. We and others are showing how the offshore baroclinic effects also modify the water levels and waves, so a full 3D coupled system should be more accurate to obtain total water levels. The structured nature of roms grid can pose challenges in some twisty channels, but the refinement can help a lot. Hernan is adding refined grids that can be at an angle to the parent. I am working on other approaches as well for localized refinement.

So yes give it a try. The variations of water level are better than the 'bath tub' approach (as you said) with vertical walls.
-john

[philip]

My wish is to use ROMS for flood risk assessment and forecasting, and for the latter stability is critical. You can't have the model crash once in awhile for operational forecasts.

So, I have the same question as Stef, and so here is a follow-up question --
Can ROMS be run in a 2D barotropic mode and have better stability during flood simulations? As Stef noted, for flood predictions 2D modeling is most common. More ideally, one might have 3D in the estuary waterways and inlets, and 2D in a floodplain, but in shallow systems 2D estuary modeling is likely a reasonable approximation for tides and storm surges. But I'm not itching to learn to use the SCHISM or DELFT models any time soon.

In a 2D barotropic mode for simulating flooding, wetlands can be treated simply with elevated bed roughness, and you can still capture the important volumetric role of their inundation.

Using 2D means you miss many processes ROMS users may wish to study, such as wave-current interaction and erosion. However, there are many back-barrier systems in the US that have poor NOAA flood forecasting (e.g. SLOSH/PSURGE; ADCIRC/ESTOFS) yet where flooding is common and arises year-round.

Can ROMS help fill this gap?
Philip

[wilkin]

You can #undef SOLVE3D and see what happens
https://www.myroms.org/wiki/SOLVE3D