[Thread Prev][Thread Next][Index]

Re: [ferret_users] corrupted time axis or ferret bug ?



Thanks a lot indeed Martin: that's what I call a quick fix!

The (main) part I was missing is that Ferret needs to perfom some internal axis computation to 'shade' a variable. Now using your @asn trick I was able to create (on the fly) a nice time axis and recover the perfect 'shade' I needed. Dare say I appreciate it: thanks again.

----- Mail original -----
De: "Martin Schmidt" <martin.schmidt@xxxxxxxxxxxxxxxxx>
À: "Nicolas Perenne" <perenne@xxxxxxxxxx>
Cc: "ferret users" <ferret_users@xxxxxxxx>
Envoyé: Vendredi 11 Mai 2012 13:13:31
Objet: Re: [ferret_users] corrupted time axis or ferret bug ?

Hi Nicolas,

with ncdump you may have a look into your netcdf file. You will find 
that the time axis has units
'seconds' but with very large values, because of the choice of the time 
axis origin.

Ferret does many calculations in real*4, this implies 6 valid digits. I 
guess, something goes wrong with the axis setup for the second plot
because of limited accuracy of calculations with small differences of 
large numbers.

One should generally avoid calculations based on small differences of 
large numbers not only with ferret.

As a test you may generate a new time axis. This is an example, for 
correct numbers - you know the details:

define axis/t=1:9:1/unit=days/t0=1-jan-2000 tax
let temp1=temp[gt=tax@asn]
! Now the data are at a new time axis with
set reg/i=4/j=3
set view upper
plot/k=30 temp1
set view lower
shade temp1

This should be the expected result - except the wrong time axis.

Best,
Martin

Nicolas Perenne wrote:
> Hello,
>
> I have a strange problem with the attached NetCDF file (generated by a numerical model, but this one is pretty small). Can't figure it out and was unable to find a clue in the mail archive.
>
> So this is about 'shade': a vertical slice through the attached NetCDF is fine in y-z but another in t-z doesn't show anything (cf 'pb.jnl' also attached to this mail).
>
> 'plot/k=30' is ok though, and moreover this is the first time I have such a problem with a t-z slice (doing quite a lot of it on various other datasets)
>
> Currently using Ferret V6.72 but same pb with an old (windows) V5.51.
>
> Do you think it could be a ferret bug, or a corruption of the NetCDF file, or a subtle combination of both?
>
> Thanks,
> Nicolas




[Thread Prev][Thread Next][Index]
Contact Us
Dept of Commerce / NOAA / OAR / PMEL / Ferret

Privacy Policy | Disclaimer | Accessibility Statement