Where does ROMS store SSTs?

[julian.kuhlmann]

I would like to compare my modelling output to satellite-measured SST. Is it possible to have ROMS write out an SST variable? Or should I just take the uppermost layer of the variable "temp" (potential temperature)? On the one hand, that might be all right, since the temperature in the model cannot vary beyond the vertical resolution of the grid; on the other hand, in deep ocean regions, the uppermost layer can easily be 50 m thick, which makes layer-1 temperatures something different from satellite SSTs, right? How is this usually done?

Thank you!

[kate]

Yes, the model only has the temperature at k=N, which is half a grid down from the surface. One thing you can do is choose vertical stretching parameters so that the uppermost point is closer to the surface, even in deep water. This might still not be good enough, however. You might be well served by working out some sort of boundary layer code.

[arango]

Actually, the values of ROMS state variables represent a finite volume average of the grid cell (1/Volume INTEGRALS [T dx dy dz]). We just happen to locate such values at the center of the cell. We do not have a prognostic equation for SST or SSS. These values are implicit in the external forcing via surface net heat flux and surface net freshwater (or salt in ice) flux. What you seek is only possible in the surface boundary layer between atmosphere and ocean. This boundary layer is not part of the atmosphere or ocean model.

Within the finite volume formulation, it is possible to extrapolate the SST or SSS using functions like parabolic splines segments. We do not have code for this. Notice, the SST and SSS forcing are possible as terms (i.e. nudging, relaxation, flux correction) in the surface forcing terms (boundary conditions) for the vertical diffusion tracer equations. We also can have such data in variational data assimilation (4D-Var).

[julian.kuhlmann]

Thank you for your responses! I actually made a mistake when estimating the depth of the uppermost rho-point, and it is not as far from the surface as I thought. Also, the temperature in the first couple of sigma-layers doesn't vary much, so I guess it's all right to just use the top layer's temperature as an approximate SST.