RIVER PLUME DYNAMICS

[Manuel Nsoni Zau]

Hello Alan!

I have been carrying out simulations of the Plata river plume. My model has 32 hybrid vertical levels and 15 of these can be transformed into sigma and z in the upper water column. Initially my simulations were not as consistent with reality, then I realise that this happened because my isopcynal layers were interacting with the bottom of the plume and, in the article by Schiller et al 2011 (MISSISSIPPI RIVER PLUME DYNAMICS), they mention that this can be detrimental for the vertical structure of the plume.

Therefore, I would like to know why this interaction between the bottom of the plume and the isopcynal layers can be detrimental to the vertical distribution and, in some cases like mine, also detrimental to the horizontal distribution of the plume?

Thank you!

Manuel!

[Alan Wallcraft]

In general, when an isopycnal layer intrudes over the shelf it could be realistic in the sense that it has the right density profile but the isopycnal layer is represented by a single T,S,u,v and this can lead to a loss in resolution for things like velocity shear.

This is a primary reason why GOFS 3.1 expanded GOFS 3.0's 32 layers to 41 layers with all the extra layers at the top of the water column and designed so that isopycnal layers are never shallower than 83 m.  There are some negatives to this approach, but it is almost always a net positive in shallow water.

When you have a high resolution grid another trick you can try is to cut an artificial notch in the land to represent the river and have the river water applied to that notch.  This is artificial because the notch is typically much wider than the actual river.  It can be set to an appropriate depth though.

I also suggest you turn on the new option epmass=2 with empbal=0 (you will need the latest HYCOM-src for this).  It treats rivers as a mass exchange which enhances the river plume.  The figure shows the effect of this after about 1 month with (09.5) and without (04.0) epmass=2.  The effect isn't huge for the Northern Gulf.  It is much larger for rivers like the Amazon.

[Manuel Nsoni Zau]

Hi Alan,

I'm now using the latest version of HYCOM, and I've set epmass=2 and empbal=0, as you suggested. However, I noticed that on day 129, after 4 months of simulation, the model starts to produce negative salinity, the output file says:

- 19877877 i,j,k =  127  155  8 neg. saln after advem call      -0.00

In fact, this error was also present in the previous version runs when I used epmass=1, however, the model stops running in the fourth year of simulation.

According to what I checked here, this new version of HYCOM that I'm using already has the negative salinity correction associated with KPP, so could you help me understand why I'm having this error.

Last question, I didn't understand the end of your message, the values ​​of 09.5 and 04.0 associated to the figures above refer to which parameter? empbal?

Thanks, Manuel!!

[Alan Wallcraft]

09.5 and 04.0 are the experiment numbers.

HYCOM's advection scheme is supposed to be conservative and monotonic. which means that if salinity is non-negative everywhere before advection it should be non-negative everywhere after advection.  Hence the error message.

The first thing to do is to reduce the baroclinic time step (and make sure the barotropic time step is about as large as possible).  I suggest using perhaps 75% of your current time step.

I don't really hold out much hope that this will work.  If not, then there is a bug somewhere - most likely a negative salinity generated before the advection step.  I can track it down if you provide a tar bundle of your entire scratch data directory configured to run the first segment that gets a neg. saln error message.  This needs to include the nest archive files for the segment.  I will run the same segment from your files as input.

Alan.