SWAN convergence error

Hi,

I ran a coupled SWAN-ROMS model for the year 2006, it failed at 20060926.153000 due to:

> Boundary data type TPAR processed, time: 39117600.
Boundary data type TPAR processed, time: 39117600.
** Severe error : No convergence calculating the spectrum
** Severe error : at the boundary using parametric bound. cond.
Boundary data type TPAR processed, time: 39117600.
** Severe error : No convergence calculating the spectrum
** Severe error : at the boundary using parametric bound. cond.
Boundary data type TPAR processed, time: 39117600.
** Severe error : No convergence calculating the spectrum
** Severe error : at the boundary using parametric bound. cond.
Boundary data type TPAR processed, time: 39117600.
** Severe error : No convergence calculating the spectrum
** Severe error : at the boundary using parametric bound. cond.
Boundary data type TPAR processed, time: 39117600.
** Severe error : No convergence calculating the spectrum
** Severe error : at the boundary using parametric bound. cond.
Boundary data type TPAR processed, time: 39117600.
Boundary data type TPAR processed, time: 39117600.

I tried to re-run it using either hotstart or coldstart (set INIT DEFAULT), but the problem was still there. I found that the "no convergence error" was shown for my boundary TRAP files No.17 to No.21. I checked the files No.17, it looks like:

> 20060925.1200 1.61 14.30 207. 20
20060925.1500 1.58 14.09 207. 20
20060925.1800 1.55 13.80 207. 20
20060925.2100 1.55 13.62 207. 20
20060926.0000 1.56 13.50 207. 20
20060926.0300 1.56 13.41 207. 20
20060926.0600 1.58 12.11 214. 20
20060926.0900 1.58 12.05 214. 20
**20060926.1200 1.59 7.04 237. 20
20060926.1500 1.58 7.08 238. 20**
20060926.1800 1.57 18.26 223. 20
20060926.2100 1.57 19.80 219. 20
20060927.0000 1.58 19.56 219. 20
20060927.0300 1.60 19.45 219. 20
20060927.0600 1.61 19.30 219. 20
20060927.0900 1.65 18.93 219. 20
20060927.1200 1.69 18.15 219. 20

The files No.16 to No.20 look similar to file No.17 at this time, while other files look "more smoothly between the data". Is that the problem?

Can I simply comment those files to let the model run?

In case if the below information may help diagnose: My ROMS and SWAN grids are the same, with 1km resolution. They are coupled every 10 min; ROMS uses a 8 sec timestep and SWAN uses 10 min. I nudge the 2D and 3D momentum at the ROMS boundary. The SWAN settings are:
& PHYSICS **************************************
GEN3
FRICTION MADSEN 0.025
QUAD 2 0.25
TRIAD
OFF WCAP
BREAKING CON 1.0 0.73
OFF BNDCHK

NUM STOPC NONSTAT 1

Start by reproducing the failure at 20060926.153000 with the reported SWAN settings and inspect TRAP files No.16 through No.21 around the abrupt values in file No.17. Compare coldstart and hotstart runs and the TPAR boundary-processing output; done means identifying the cause of the convergence failure and documenting a validated input or configuration change.

Written by the indexing model from the issue text.