Re: bad flag gets reset

[Steve Knox <stevek@xxxxxxxxxxxxxxxxxx>]

Kaduk, Dr J. wrote:

> Hi,
> just came across this today as well. Seems that the missing value flag
> is indeed reset whenever you do a calculation or define a new variable.
> You can specify the missing value flag by including the /bad option in 
> the variable definition:
> let/bad=1.0000000E+20 xxx=.....
> 'hope that helps
> cheers,
> Joerg
>
> ________________________________
>
> From: owner-ferret_users@pmel.noaa.gov on behalf of Steve Knox
> Sent: Thu 7/1/2004 17:23
> To: ferret_users@noaa.gov
> Subject: bad flag gets reset
>
>
> Hi,
> I'm computing the following and it appears that the missing vlaue defined in my netcdf file is getting
> reset when I compute the ration of @var/@ave. The missingvalue is 1.0000000E+20 but when I compute @var/@ave the bad flag is set to -9.9999998E+33 (= -0.1000E+35)
> I would have expected the missing value to remain the same, ie 1.0000000E+20.
> Any ideas?
> Here's the Netcdf header:
>
> obs.globe.lan.1901-2000.pre.annual {
> dimensions:
>        lon = 720 ;
>        lat = 360 ;
>        time = UNLIMITED ; // (100 currently)
> variables:
>        float lon(lon) ;
>                lon:units = "degrees_east" ;
>                lon:valid_min = -180.f ;
>                lon:valid_max = 180.f ;
>                lon:title = "Longitude" ;
>        float lat(lat) ;
>                lat:units = "degrees_north" ;
>                lat:valid_min = -90.f ;
>                lat:valid_max = 90.f ;
>                lat:title = "Latitude" ;
>        float time(time) ;
>                time:units = "years since 1900-01-01" ;
>                time:title = "Time" ;
>                time:long_name = "Time axis" ;
>                time:time_origin = " 1900-01-15 00:00:00" ;
>        float pre(time, lat, lon) ;
>                pre:units = "mm/month" ;
>                pre:long_name = "Precipitation" ;
>                pre:missing_value = 1.e+20f 
> yes? set region/i=1:1/j=1:1      !look at a grid cell that I know is missing
> yes? list/l=1:1/format=(e15.6)  pre[l=@ave]
>             VARIABLE : Precipitation (mm/month)
>             FILENAME : obs.globe.lan.1901-2000.pre.annual.nc
>             BAD FLAG :  1.0000000E+20
>             LONGITUDE: 179.8W(-179.8)
>             LATITUDE : 89.8S
>             TIME     : 02-JUL-1900 14:54 to 02-JUL-1901 20:43 (averaged)
>   0.100000E+21  (OK)
> yes? list/l=1:1/format=(e12.4)  pre[l=@var]
>             VARIABLE : Variance of Precipitation ((mm/month)^2)
>             FILENAME : obs.globe.lan.1901-2000.pre.annual.nc
>             BAD FLAG :  1.0000000E+20
>             LONGITUDE: 179.8W(-179.8)
>             LATITUDE : 89.8S
>             TIME     : 02-JUL-1900 14:54 to 02-JUL-1901 20:43 (variance)
>  0.1000E+21 (OK)
> yes? list/l=1:1/format=(e12.4)  pre[l=@var]/pre[l=@ave]
>             VARIABLE : PRE[L=@VAR]/PRE[L=@AVE]
>             FILENAME : obs.globe.lan.1901-2000.pre.annual.nc
>             BAD FLAG : -9.9999998E+33 (aka -0.1000E+35)
>             LONGITUDE: 179.8W(-179.8)
>             LATITUDE : 89.8S
>             TIME     : 02-JUL-1900 14:54 to 02-JUL-1901 20:43
> -0.1000E+35 (NOT OK)
>
>
>  
Thanks for your suggestion! It seems to work. This is what I did:
yes? let/bad=1.0E+20 *ratio*=pre[D=1,l=@var]/pre[D=1,l=@ave]
yes? list/t="15-jan-1946":"15-dec-1995"/... *ratio*


I guess the point is that in a calculation, the variables in the 
calculation don't necessarily have the *same* missing value so Ferret 
supplies a new one for the result.