[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[netCDF #KPE-603099]: a question about ncdump



Matt,

> I checked my file permissions on that folder, and apparently Windows 7 
> defaults to "user" settings for some folders rather than "administrator."  
> Anyway, I changed that, and was able to write the .cdl file.  When I open it 
> in Excel (assuming it is comma delimited), it looks like this:
> 
> u =
> 1.#QNAN        1.#QNAN         1.#QNAN         1.#QNAN         1.#QNAN        
>  1.#QNAN         1.#QNAN         1.#QNAN
> 1.#QNAN        1.#QNAN         1.#QNAN         1.#QNAN         1.#QNAN        
>  1.#QNAN         1.#QNAN         1.#QNAN
> 1.#QNAN        1.#QNAN         1.#QNAN         1.#QNAN         1.#QNAN        
>  1.#QNAN         1.#QNAN        0.0175796
> -0.08131572   -0.1633865      -0.1815312      -0.1400719      -0.1115065      
> -0.1240741
> -0.1486088    -0.1112384      -0.03787675

  ...

> I only show you in case you know what the 1.#QNAN value makes any sense to 
> you.  I'll ask the folks at NOAA too.  In the meantime, thanks very much for 
> all your help!

Those values are "Not a Number" values, floating point representations results 
of invalid operations,
such as dividing 0 by 0 or resulting from numeric underflow when operating on 
very small numbers
close to zero.  They can also be the result of using NaN's to represent 
"missing values", such as 
sea temperature over land.  You would see this in the header output from 
ncdump, as a variable
attribute such as u:_FillValue or u:missing_value, for example, specified as 
some floating-point
number near the edge of the representable range of floats.

So it may not be an error at all, but just the way missing values appear in 
ncdump output, although
there is an attempt in ncdump to display such values as "_" instead of as 
NaN's.  Or these out-of-bounds
values may be an indication that the model that produced them was generating 
such values as the
result of a numerically unstable algorithm.

--Russ

Russ Rew                                         UCAR Unidata Program
address@hidden                      http://www.unidata.ucar.edu



Ticket Details
===================
Ticket ID: KPE-603099
Department: Support netCDF
Priority: Normal
Status: Closed