[Esip-documentation] on documenting instrument and platform metadata

Tue Mar 17 11:30:12 EDT 2015

Hi Aaron  -

Having been involved in the oceansites specification on this, I agree that
no one solution always works.  I adopted (and try to follow, and promote)
the NODC templates as closely as possible. Their development of the 
ancillary
variable 'instrument' was brilliant, in my opinion.

My datasets range from met files where all the data is from a single
logger but each of 6 or 8 variables might need its sensor described at
the variable level, to depth merged temperature files where there's only
one data variable but 30 instruments (and 4 or 5 different models).   In 
the met
files, variable attributes would work, but I use 'instrument' ancillary 
variables
so I can link data variables recorded by a single sensor module (e.g. 
Inst_HRH
is linked to data variables HRH and ATMP).

The depth merged files show the real beauty of the instrument variable, 
though;
since it can have a depth dimension, you can describe the different 
instruments
deployed at different depths.  There's no good way to do that using 
attributes, as
far as I've seen - lists break down, especially when data is sub-sampled 
i.e. in thredds.

And, although much of this isn't yet at a stage where software can use 
it, any and all
information that you have about the sensors and platform should still be 
recorded
wherever it seems most appropriate to you, since it might otherwise be 
lost.  Re-formatting
this kind of info (if and when a useful standard emerges) will be much 
less effort
than trying to track down something that was not recorded because there 
didn't
seem like a standard way to encode it.

If ACDD can't find and understand instrument variables yet, maybe that's 
something
that could be looked at in another rev. Or, maybe instruments aren't a 
common
discovery concept - I really don't know.

Cheers -
Nan

On 3/16/15 6:41 PM, Aaron Sweeney via Esip-documentation wrote:
> Thanks very much for the guidance, John.  I think I may stick with the 
> NODC or OceanSites platform and instrument variables to define the 
> particular instances.
>
> That being said, for my two use cases, I think I can still capture the 
> type of platform and instrument in the ACDD global attributes.  This 
> amounts to:
>
>     :platform = "DART"
>     :platform_vocabulary = "NASA/GCMD Platform Keywords. Version 8.0"
>     :instrument = "BOTTOM PRESSURE GAUGES"
>     :instrument_vocabulary = "NASA/GCMD Instrument Keywords. Version 8.0"
>
> and
>
>     :platform = "COASTAL STATIONS"
>     :platform_vocabulary = "NASA/GCMD Platform Keywords. Version 8.0"
>     :instrument = "TIDE GAUGES"
>     :instrument_vocabulary = "NASA/GCMD Instrument Keywords. Version 8.0"
>
> Cordially,
> Aaron
>
> On 03/16/2015 02:43 PM, John Graybeal wrote:
>> What great questions!  The following thoughts are my own, and are 
>> unrelated to my previous ACDD activities.
>>
>> Clearly some data files will have data from multiple platforms and 
>> instruments; whereas others will get all their variables from a 
>> single platform/instrument; and some will have data so heavily 
>> processed as to be barely traceable to the source 
>> platforms/instruments. Each of the specifications is oriented toward 
>> somewhat different scenarios, and none of them are yet great at 
>> representing all the scenarios.
>>
>> Let's consider ACDD first. Note that this specification provides 
>> _recommendations_ for attributes, not _requirements_; and the 
>> platform and instrument attributes are only 'Suggested'. These give a 
>> broad-brush categorization of the platform(s) or instrument(s) that 
>> collected a data set; if it seems reasonable to specify that 
>> information for the whole data set, use these attributes to do so. 
>> (If it isn't useful/applicable, then don't worry about these 
>> attributes.)  The most precise way to give that info is to use a 
>> controlled vocabulary (see 
>> https://marinemetadata.org/references?filter0=**ALL**&filter1%5B%5D=129&filter1%5B%5D=141 
>> <https://marinemetadata.org/references?filter0=**ALL**&filter1[]=129&filter1[]=141> for 
>> a partial list of vocabularies), and typically I'd expect a type 
>> vocabulary here (with e.g. "CTD", "Mooring") not a specific instance.
>>
>> If you *are* following ACDD, I don't recommend putting the names of 
>> variables containing information on platforms and instruments in the 
>> global attributes 'platform' and 'instrument'. Without ACDD 
>> describing that as a possibility, it's highly unlikely any software 
>> will know how to treat that, and some people will get confused too. 
>> The existence of the variables will be pretty self-explanatory, so 
>> just leave out those ACDD attributes. (I see this conflicts with NODC 
>> and OceanSITES recommendations, to my surprise; perhaps others will 
>> offer another view here.)
>>
>> Whether to put the information in a global instrument variable, or 
>> attributes of each data variable, or both, seems to me a matter of 
>> circumstances and taste. Let's stipulate the best solution is 
>> human-recognizable, computer-parseable, extensible, and unambiguous. 
>> Thus, global attributes are inadequate. Assuming a sophisticated data 
>> system 5 years from now would look in both variables and data 
>> attributes, then the model of 'one variable per instrument or 
>> platform', and association from the data variables to those variables 
>> or platforms seems best.
>>
>> But given that the examples in each of the different profiles still 
>> seems inconsistent, likely you should follow the guidance most close 
>> to your work. If you aren't particularly close to any of them, I'd do 
>> it a bit differently still. But I'll hold that thought unless you 
>> want to extend the discussion in that direction.
>>
>> John
>>
>>
>> On Mar 16, 2015, at 11:56, Aaron Sweeney via Esip-documentation 
>> <esip-documentation at lists.esipfed.org 
>> <mailto:esip-documentation at lists.esipfed.org>> wrote:
>>
>>> Hi, folks,
>>>
>>>      I’m struggling with how to capture platform and instrument 
>>> information in the appropriate places in netCDF files.  The NODC 
>>> templates suggest creating separate variables for platforms and 
>>> instruments, with multiple variable attributes (make, model, serial 
>>> number, precision, accuracy, etc.), and associating these platform 
>>> and instrument variables with physical variables via an 
>>> “ancillary_variable” variable attribute attached to the physical 
>>> variable. Along these lines, SeaDataNet suggests including variable 
>>> attributes for instrument (sdn_instrument_urn and 
>>> sdn_instrument_name), and OceanSites suggests either the use of a 
>>> separate instrument variable (with make, model, serial number 
>>> attributes) or capturing make, model and serial number as variable 
>>> attributes contained directly within the relevant physical 
>>> variable.  But ACDD-1.3 only has single /global/ attributes for 
>>> platform and instrument (and vocabularies).  The NODC templates 
>>> suggest placing the names of the variables containing information on 
>>> platforms and instruments in the global attributes: instrument and 
>>> platform.  I like NODC’s and OceanSites’ “associative” approach 
>>> (i.e. “this instrument goes with this physical variable”), but the 
>>> use of variable names in global attributes conflicts with ACDD-1.3.
>>>
>>>      It seems that in order to comply with ACDD-1.3, I need to 
>>> capture instruments and platforms at a global level (dropping their 
>>> association with physical measurement variables), but in order to 
>>> comply with OceanSites and others, I need to capture instrument and 
>>> platform information in separate variables and associate these with 
>>> physical measurement variables via an ancillary_variable attribute.
>>>
>>>      In order to comply with both, in the interest of enabling 
>>> interoperability, I seem to need to repeat instrument and platform 
>>> metadata in two different places.
>>>
>>>      Any thoughts or guidance?
>>>
>>> References:
>>> NODC netCDF timeseries template: 
>>> http://www.nodc.noaa.gov/data/formats/netcdf/v1.1/timeSeriesOrthogonal.cdl
>>> OceanSITES Data Format Reference Manual (See Appendix 2 for sensor 
>>> variables): 
>>> http://www.oceansites.org/docs/oceansites_data_format_reference_manual.pdf
>>> SeaDataNet Data Transports Manual (See Section 4.2.2 Co-ordinate 
>>> Variables): 
>>> http://www.seadatanet.org/content/download/16251/106283/file/SDN2_D85_WP8_Datafile_formats.pdf
>>>
>>> Thanks,
>>> Aaron

-- 
*******************************************************
* Nan Galbraith        Information Systems Specialist *
* Upper Ocean Processes Group            Mail Stop 29 *
* Woods Hole Oceanographic Institution                *
* Woods Hole, MA 02543                 (508) 289-2444 *
*******************************************************

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.deltaforce.net/pipermail/esip-documentation/attachments/20150317/7b525e05/attachment.html>