About the computational spectral grid

General scientific issues regarding ROMS

Moderators: arango, robertson

Post Reply
Message
Author
rcxy
Posts: 31
Joined: Fri Jul 02, 2010 1:28 pm
Location: Guangdong Ocean University

About the computational spectral grid

#1 Post by rcxy » Sat Oct 11, 2014 3:58 pm

Hi,
when i couple with swan model, the reslut has a warning in the print file.

Code: Select all

** Warning          : relative frequency resolution (df/f) deviates more
 ** Warning          : than 5% from 10%-resolution. This may be problematic
 ** Warning          : when quadruplets are approximated by means of DIA.
I check the swan user manual, it may be

Code: Select all

The computational spectral grid needs to be provided by the user. In frequency space,
it is simply dened by a minimum and a maximum frequency and the frequency resolution
which is proportional to the frequency itself (i.e. logarithmic, e.g., f = 0:1 f).
 ...
And in the code

Code: Select all

         GAMMA = EXP(ALOG(SHIG/SLOW)/REAL(MSC-1))                         40.31
         IF (ABS(GAMMA-1.1).GT.0.055) THEN                                40.31
            CALL MSGERR(1,                                                40.31
     &           'relative frequency resolution (df/f) deviates more')    40.31
            CALL MSGERR(1,                                                40.31
     &           'than 5% from 10%-resolution. This may be problematic')  40.31
            CALL MSGERR(1,                                                40.31
     &           'when quadruplets are approximated by means of DIA.')    40.31
         END IF 
However, i don't know how to solve the problem. pls help me!
Thank you!

Hilberthotel
Posts: 4
Joined: Mon Aug 03, 2015 10:39 pm
Location: University of Oldenburg - ICBM

Re: About the computational spectral grid

#2 Post by Hilberthotel » Fri Feb 28, 2020 10:38 am

Sorry, I know it's an very old question and surely you answered your question by yourself but for people like me who come across this warning and google for it here is the answer:
The problem is the grid resolution in frequency space. Look in the SWAN manual at the part about the CGRID command (around page 31): If you want ta get accurate DIA approximations set fhigh, flow and msc that
msc=log(fhigh/flow) / log(1- df/f )
with df/f=0.1 .

Post Reply