ROMS wetting-drying and bulk flux heating/cooling

[lanerolle]

I am running a ROMS application with very heavy tidal wetting-drying in Cook Inlet, AK.

When I run ROMS without any surface forcing (i.e. zero wind stresses and net heat fluxes), it is numerically very stable and generates good results. When however, I include surface forcing via the Bulk Fluxes, during the warmer summer months, the model begins to blow-up at or near (isolated) cells which experience wetting-drying. The blow-up is seen to occur with the temperature variable - T becomes very large like 80C and this causes rho to become large and go over the 200 Kg/m3 blow-up limit. I have a couple of questions:

1. It appears to me that the wind stresses and heat flux which the Bulk Fluxes generate are multiplied by the permanent wet/dry mask and not the one associated with the wetting-drying. Is this correct?

2. If there is a cell which is not set to be permanently dry and wets from time to time, does it experience the Bulk Flux related heating/cooling even when it is dry (and its water depth is Dcrit)? If so its associated temperature can increase dramatically over a short.

[jcwarner]

this is mostly a problem with the heat flux ,and not the surface momenutum flux.
yes, the cells that experience wet/dry should be multipled by the wet/dry masking in bulk fluxes.

[lanerolle]

Thanks John for your prompt reply. Does this mean that there is a bug in ROMS which needs to be fixed (in bulk_fluxes.F)?

[shuwen]

Yes. I also have the same question. Is there any more detailed instruction for heat flux forcing for ROMS?
Thanks!

[jcwarner]

i wouldn't necessarily call it a bug. I would call it growing pains.
As more capabiities are added, there are interactions that are not immediately recognized. As the model is exercised, these things come up.
We have fixed this in our version. Let me see what is needed for the Rutgers release and make a suggestion.
-john

[wilkin]

Shouldn't the wet/dry mask be applied in the surface boundary condition set_vbc.F, rather than specifically in bulk_fluxes.F?

The issue is that a cell that is "dry" (albeit with a small finite fluid volume) is excused from all tracer and momentum physics until it becomes wet again.

So all air-sea fluxes, whether by bulk_flux, imposed fluxes, analytical fluxes, relaxation etc. can and should be disabled in the one place in the code where they are applied as vertical flux boundary conditions.

A related issue, though not in play in Alaskan waters I suspect, is that in very shallow water we do not account for bottom albedo. A shallow back bay or estuary flank can reflect shortwave radiation from the seafloor that ultimately escapes the water column and does not heat it; consider visible satellite imagery that reveals bedforms - this means light reached the bottom, was reflected, and escaped to space. ROMS lmd_swfrac.F traps all shortwave energy reaching the bottom cell in that bottom cell. It does not propagate the light back up through the water column.

We've seen this effect overheat waters in a model of Barnegat Bay, NJ. But the problem was really that we had assumed the water was very clear (!), whereas in reality the optical depth is shallow and the heat should be all absorbed close to the sea surface. (You can't see the bottom of Jersey waters from the sea surface).

[jcwarner]

The wetdry occurs on the barotropic time step.
I do not see how to limit the sustr and and svstr in set_vbc, unless you are using iceshelf which i typically do not use in north carolina.
I strongly suggest, as someone who has been using this option for 6 years, to not monkey with the su/vstr and bu/vstr stuff. Instead, the wet/dry needs to act like a land/sea mask. So anywhere that you need to use
#ifdef MASKING
...
then you need to consider that the masking should include wet/dry dependence.

We have modififed our code to use the _full masking which is the land/sea mask multipled by the wet/dry mask. for example, in bluk_flux, we have tested this:
# ifdef WET_DRY
stflx(i,j,isalt)=stflx(i,j,isalt)*rmask_full(i,j)
# elif defined MASKING
stflx(i,j,isalt)=stflx(i,j,isalt)*rmask(i,j)
# endif

I do not see the _full masking in the current Rutgers release anymore. These things change on your website without my knowledge and i can not be expected to keep up with changes on a routine basis without notification. So it is difficult for me to provide direct guidance when the code changes without my knowledge. These chagnes can be very delicate.

[kate]

Interesting problem with the light heating the bottom instead of reflecting. It is however not the case in the waters off Anchorage, which are extremely silt-laden. Lyon, you want to be using something other than the default clear-water settings for the Jerlov water type (WTYPE). For the Bering Sea, we made it a function of water depth instead of a global constant.

[lanerolle]

Good points both John and Kate. Yes, Kate we do not use any light attenuation and perhaps we should - there is a ton of sediment in the water around Fire Island and above (saw it with my own eyes when I was there!!!). Interestingly, my model blows-up in Knik Arm from Anchorage upwards and that area is also laden with sediment and has some mudflats too.

The question is how do we know what Jerlov coefficient to use and how to make it spatially variable in the Cook Inlet region. The only way I guess to figure this out is to find some sediment obs from say USGS and estimate the coefficients based on them. But then again, this coefficient will also vary seasonally (more sediment in Spring when the ice melts and brings sediment with it) and also yearly. Not an easy problem!!!

Our primary focus currently is to include physics (wetting-drying, ice, sediment, bulk fluxes, etc.) to predict the water levels, currents, T and S sufficiently accurately and not predict sediment, ice, etc. out of this model - that will be part II at a later date.

[lanerolle]

I followed John's (Warner) advice and multiplied the surface heat and salinity forcing fluxes in bulk_flux.F by the wet-dry rho-mask to account for wetting-drying and made only that change to ROMS. It works beautifully and the code is numerically totally stable and the min/max T, S generated are within reasonable bounds. Without this "correction", ROMS is simply unconditionally unstable and it is impossible to run - especially during the summer months when the mudflats heat up - furthermore, the T, S min/max also have extreme values such as 120 C and -30 C. I did not account for the Jerlov water types on a depth-dependent basis as Kate suggested yet.

So I think we need to make this correction a permanent change in the ROMS (just like correcting for the bottom stress sign reversal during wetting-drying) because it is possible to have shallow wetting-drying regions with perfectly clear water (i.e. Jerlov water type = 1). This could conceivably happen if the bottom type is say volcanic rock without any mud on it.

Thanks John very much for your help!