Metadata Conventions

MetadataConventions for post-processed aircraft based data files.

I have drawn on the following documents, which are netCDF-centric, but should apply to any data format.

I've only made a guess at the required/recommended/optional field.

----

Name		Description
Conventions	required	Identifies this file to follow our conventions.
title	highly recommended	Something to describe the dataset
institution	required	Specifies where the original data was produced.
source	required	The method of production of the original data. "Aircraft". Do we prefer "Platform"?
project	highly recommended
cdm_datatype = "Trajectory"	required if	we choose to follow UnidataObsConvention

units	required	udunits compatable?
long_name	required	description of variable
standard_name	highly recommended	if a name exists for this parameter
data_quality	highly recommended
missing_value	highly recommended
valid_range	optional
source	optional	instrument source for this measurement

Do we recommend the DataDiscovery Draft above?

There are several reasons for the need of variable data discovery:

Variable names can be cryptic.
There can be multiple measurements of the same type.
Automation - Software which wants to find its way into a file. Any track plot or SkewT program should be able to open a file and produce their image without user intervention.

I talked with Jon Caron about the variable discovery.

He suggested that CF conventions prefer one variable name per attribute, instead of a list of variables in one attribute.
After more discussion we thought reversing the "standard_names" to list the variables at the global level was a good way to do it.

He suggested using a colon to add a namespace feel, so that all these attributes looked similar (if we liked).

Sample netCDF. I added the "reference" as an example namespace.

:reference:latitude = "LATC"
:reference:longitude = "LONC"
:reference:altitude = "GGALT"
:reference:air_temperature = "ATX"
....