[netCDF #HPI-640345]: Variable vs attribute

Hi Mike,

> We're defining the metadata for our weather products and in some cases we're 
> using global attributes such as one named:
> 
> "total_number_pixels_for_ash_detection_overall_QF_high", type=int, 
> value=120000
> 
> This global attribute holds a count of the number of pixels for which a data 
> quality flag is set to "high".  It's straightforward.  The units are built 
> into the attribute name ("total_number").
> 
> Would this be better defined using a variable and explicitly specifying the 
> units?

Yes, I think a variable would be better than the global attribute.

> Here's another example:  attribute name="minimum_ash_cloud_height_in 
> kilometers" value="1.5" type="float".
> In this case I feel like a variable might be better, then we could give it a 
> long name, units, valid min and max, etc.

I agree.

> Is there a rule of thumb when a variable is necessary vs. when a global 
> attribute will suffice?  Or did I answer my own question with the second 
> example?

Here's the guidelines we provide in our workshop for when to use variables 
versus when to use 
attributes:

  http://www.unidata.ucar.edu/netcdf/workshops/2011/datamodels/VarsVsAtts.html

But I think the most important consideration in this case is that you want to 
attach units to 
the value, which means you have to use a variable, because attributes can't 
have attributes in 
the netCDF data model.

--Russ

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



Ticket Details
===================
Ticket ID: HPI-640345
Department: Support netCDF
Priority: Normal
Status: Closed