Extreme zeta values and OBC

[jadranka_sepic]

Hi all,

I have a problem with my model.

Model domain is a coastal part of the Adriatic see. Model bathymetry is realistic (more or less), and model forcing is from a weather model (Uwind, Vwind, shortwave radiation, clouds, rain, relative humidity, temperature, surface pressure). Initial conditions are uniform temperature and salinity and no currents.

The problem is that water flows out of the model and after five days run zeta values in whole basin are approximately -5 m. Here is a figure showing zeta (in meters) after 5 days:


Other variables (currents, temperatures, salinities) do not become extreme.

I assume problem is in the boundary conditions.

Here is my *.h file

#define ANALYTICAL
#define MASKING
#define UV_ADV
#define UV_COR
#define UV_LOGDRAG
#define NONLIN_EOS

#define UV_VIS2
#define MIX_S_UV
#define TS_U3HADVECTION
#define TS_C4VADVECTION
#define DJ_GRADPS
#define TS_DIF2
#define MIX_S_TS

#undef SPONGE

#define SOLVE3D
#define SALINITY
#define SPLINES
#define AVERAGES
#define GLS_MIXING
#undef MY25_MIXING
#if defined GLS_MIXING || defined MY25_MIXING
# define KANTHA_CLAYSON
# define N2S2_HORAVG
#else
# define ANA_VMIX
#endif

#define ANA_INITIAL
#define ANA_BTFLUX
#define ANA_BSFLUX

#define BULK_FLUXES
#define LONGWAVE
#define CLOUDS
#define EMINUSP
#define RADIATION_2D

#define NORTHERN_WALL
#define EASTERN_WALL

#define SOUTH_FSGRADIENT
#define SOUTH_M2RADIATION
#define SOUTH_M3GRADIENT
#define SOUTH_TRADIATION

#define WEST_FSGRADIENT
#define WEST_M2RADIATION
#define WEST_M3GRADIENT
#define WEST_TRADIATION

Any help would be greatly appreciated.

[kate]

I see you've picked boundary conditions that don't require boundary values. I think you'd be much better off if you can supply it with boundary values from some outside source. You might as well extract initial conditions from the same outside source while you are at it.

[miguel.solano]

I had a similar problem in a real application. I would suggest as Kate mentioned to provide boundary values (including zeta), the better the temporal update on the values the better solution you may get. What I do in my model is extract initial and boundary values at 3hr intervals from a coarser validated model. Be very careful interpolating values at the boundary, specially if you are working in a small domain or the refinement coefficient between parent/child is too large (e.g. parent grid resolution 1 deg, child grid resolution 1/50 deg). There are several papers on downscaling and offline nesting procedures and methodologies that could give you a really good idea of how to set up real applications. Good luck!

[jadranka_sepic]

Thanks for your answers. We have a bigger Adriatic model so we'll try to extract boundary conditions from it. It is much coarser, though. Bigger model has a grid spacing of 2.5 km, and smaller one of ~200 m. I'll let you know about our results.