What is the best practice to run a coupled simulation(WRF-ROMS) for a case study using COAWST

[GengXia-NREL2019]

Hi there

My name is Geng Xia, a research scientist at NREL. My goal is to use COAWST to run a case study over the U.S West Coast. I did run a spinup simulation for the standalone ocean (ROMS) model to get the restart file for the ocean. However, when I run the coupled simulation, I notice that there is a quite big difference in the SST field between the initial WRF simulation and ROMS, and that creates a SST shock for my WRF simulation. Since the case study only lasts for 5 days, this SST shock didn't go away and I am afraid it might influence the results as the atmosphere the ocean was never in an equilibrium status due to the SST shock from ROMS.

Note that I use HYCOM to spinup the ocean model and ERA5 to initialize WRF. For the standalone ROMS simulation, I only weakly nudge the boundary and let most of the domain to free evolve. So this SST difference between ROMS and WRF is not too surprising, but is there a way to fix that? what is the best practice for running these type of case study using the coupled model?

Sincerely
Geng Xia

[jcwarner-usgs]

I am not quite clear what you are saying.
IF you use HYCOM to init roms, and then you get some SST.
What is the SST in WRF - is that from ERA5? so the issue is with hycom vs era5. It has nothing to do with roms or wrf.
you need to resolve that init issue first.
-j

[jbzambon]

If using ERA-5 and running very high-resolution, also see an earlier post of mine to the WRF Github on a potential issue they have matching coastlines.

[GengXia-NREL2019]

Hi all , thanks for the reply. Let me explain my problem more clearly. I run WRF-ROMS for a case study. To start the simulation, you will need to have the input file from WRF and restart file from ROMS (I do conduct a spinup simulation using standalone ROMS; I use HYCOM for the ocean input and ERA5 for atmospheric forcing). The wrf input is generated from real.exe using ERA5 and so everything, particular the air-sea interface, is in an equilibrium state at the moment. Note that this equilibrium is achieved using SST, temperature and pressure data from EAR5. However, in the coupled simulation, this equilibrium breaks due to variables exchanges from the coupler. For instance, the top figure shows the difference in SST from the coupled run and wrf standalone at the initialization stage of the simulation. There is no difference at this moment become the SST at the initialization stage are from ERA5 (from wrfinput_d01). However, one hour into the simulation (the interval time in my setup is every 30 mins), there is significant SST difference (bottom figure) and this is caused by the difference between ROMS and ERA5. In my case, the magnitude of the SST difference is quite large and I am a little concern with it because I think the WRF model will need time to adjust to this change and it is only a five-day simulation for this case study.

Is this something you also see in your case study simulation? I am still relative new to ROMS so I might have done something wrong. The goal is to reduce the SST shock so that the model can adjust to this change in a lesser time. I am still relative native new to running the coupled simulation and so sorry for asking something that might be obvious.

[jcwarner-usgs]

I think i understand your problem.
The problem is not with WRF or ROMS.
The problem is with the difference between ERA5 and HYCOM.
You are getting 2 different SSTs: one from ERA5 and one from HYCOM. They come together through the coupling via WRF and ROMS. This is not a wrf or roms issue.

Which one is correct: the SST from ERA5 or from HYCOM?
I think you should iniitalize the SST in WRF to be the same as in ROMS/HYCOM, and let the winds adjust.

[GengXia-NREL2019]

Hi John, Thanks for your feedback. Yes, that is the problem that puzzles me and thank you for your suggestion. I do have a few follow-up questions regarding your hurricane paper in 2010.

1. For that specific hurricane case study, you mentioned that "All the simulations were initialized to start with a model time of September 15, 2003. " in your paper. Does that mean that your ROMS simulation is a new simulation, not a restart simulation. If that is the case, does that mean that ROMS spinup is not really necessary for a short case study (my case study also only lasts for 10 days)?

2. the atmospheric forcing is GFS in your study, did you then also initialize the SST in WRF by replacing the SST with ROMS?

[jcwarner-usgs]

We have been running a 'forecast' with roms for the US east coast since 2010. so we have init fields for our grids that have been running for years. i cant say those sst's are exact, but we have a setup that we use.

I think you should focus on your work, and what is best for your situation. ROMS is a full 3d model, and the fields need time to evolve and be self consistent. If you init roms from hycom, you should run roms for some time period (many days, a week?) to let ROMS adjust to the baroclinicity etc.
What are you going to compare your model time series of SST to? buoys? does the buoy data compare better to ERA5 or HYCOM?

[GengXia-NREL2019]

I am comparing the simulated SST with NDBC buoys. The result is mixed where ERA5 is better over some stations and HYCOM is better over the others stations. Currently, I am putting the results together and trying to run a test case using your suggestion. If it is ok with you, I would love to ask your feedback on the results once I finish it.

[jcwarner-usgs]

sure. i can take a look. I would think that ERA5 (is that a re-analysis) would have ingested the buoy data and be a better match.