handle platform and instruments

[ferrighi]

Information about platform and instruments will be handled according to the new approach that avoids using vocab.met.no, but rather allows use of different vocabularies depending of needs. What we can handle can be extended in the future, for now using Oscar for remote sensing and nerc vocabularies should be a good start.

The structure is always "long_name (short_name)" for concepts and "name:base_url" for vocabularies.

For remote sensing:

platform: "Earth Observation System - Aqua (Aqua)" ; 
platform_vocabulary: "WMO OSCAR Space:https://space.oscar.wmo.int/satellites";
instrument: "Moderate-resolution Imaging Spectro-radiometer (MODIS)" ; 
instrument_vocabulary: "WMO OSCAR Space:https://space.oscar.wmo.int/instruments"; 

we use the same approach as for keywords. Looking for vocabulary and Oscar. The mapping to the url is easy as all non alphabetical characters are translated into underscore. So this will become

<platform>
    <short_name>Aqua</short_name>
    <long_name>Earth Observation System - Aqua</long_name>
    <resource>https://space.oscar.wmo.int/satellites/view/aqua</resource>
    <instrument>
         <short_name>MODIS</short_name>
         <long_name>Moderate-resolution Imaging Spectro-radiometer</long_name>
         <resource>https://space.oscar.wmo.int/instruments/view/modis</resource>
    </instrument>
</platform>

If the short name not provided we can extract the long_name only and not the url. It is also possible to query the oscar api, but I think it would be quite heavy, so I would skip this for now.

Other types of platforms and instruments

Instrument reference can be given also in absence of platform details. For specific sensors, instruments and other types of devices the following vocabularies are recommended:

• the SeaDataNet device categories
• the SeaVoX Device Catalogue

instrument: "Conductivity, temperature and depth package (CTD)" ;
instrument_vocabulary: "SeaVoX Device Catalogue:http://vocab.nerc.ac.uk/collection/L22/current/";

giving

<platform>
    <short_name/>
    <long_name/>
    <instrument>
         <short_name>CTD</short_name>
         <long_name>Conductivity, temperature and depth package</long_name>
         <resource>http://vocab.nerc.ac.uk/collection/L22/current/TOOL0001/</resource>
    </instrument>
</platform>

For other platforms that do not refer to remote sensing satellites, as for example ships and vessels, the following vocabulary is recommended:

• the ICES Platform Codes

For example, for the Oceania research vessel:

platform: "Oceania" ; 
platform_vocabulary: "ICES Platform Codes:https://vocab.nerc.ac.uk/collection/C17/current/"; 

giving

<platform>
    <short_name/>
    <long_name>Oceania</long_name>
    <resource>http://vocab.nerc.ac.uk/collection/C17/current/67CE/</resource>
</platform>

The url are easily extracted with sparql, e.g.:

PREFIX dc: <http://purl.org/dc/elements/1.1/>
PREFIX skos: <http://www.w3.org/2004/02/skos/core#>
SELECT ?id WHERE {
  ?a skos:prefLabel "Conductivity, temperature and depth package"@en .
  ?a dc:identifier ?id
} LIMIT 1

giving SDN:L22::TOOL0001 which means: http://vocab.nerc.ac.uk/collection/L22/current/TOOL0001/
we can also use short_name or add it if missing.

This approach can be general for nerc vocabularies, as long as the base_url is given.

When we find the more generic GCMD

platform: "Platforms > Water-based Platforms > Fixed Platforms > Surface > OCEAN WEATHER STATIONS" ; 
platform_vocabulary: "GCMD Platforms"; 

instrument: "Instruments > In Situ/Laboratory Instruments > Probes > MAGNAPROBE" ; 
instrument_vocabulary: "GCMD Instruments";

we translate into keywords GCMDPLT and GCMDIST.

[steingod]

I didn't understand the comment looking for vocabulary and Oscar and on non alpha-numeric characters being underscore?

[ferrighi]

Indeed, it was not clear. In Oscar the approach is to have a url based on the short name:

short_name: "Aqua"
url: https://space.oscar.wmo.int/satellites/view/aqua

Any short_name with spaces or dashes etc are translated into urls mapping these characters into underscores:

short_name: "NOAA-18"
url:https://space.oscar.wmo.int/satellites/view/noaa_18

This is consistent for all records, so we can easily create in mmd the resource based on the short_name if the vocabulary provided is Oscar, as we know the logic.

[ferrighi]

I am testing this and seems to work quite well. I think it is good to start with supporting some recommended ones, but in principles as long as we have a url to nerc the function I have implemented works in a generic way, so we can always extract information (if available) from any vocabulary. The key element to get is the vocabulary number (e.g. L22):

http://vocab.nerc.ac.uk/collection/ L22 /current/

We can extend to this generic approach in a second step.

I will test more now on faulty examples. The issue here would be if we should just skip information that claims to be from a vocabulary, but it is not. My suggestion would be to just keep it (probably only a long name would be given), but then enrichment will be skipped, e.g. no short name and resource added. If we want to be more strict we can just skip the information that do not match any entry in the claimed vocabulary.

For OSCAR I have added some more dedicated parsing since providing long and short is sometime a bit too demanding and satellites are probably better know by their short name. We can add this information about parsing to the ADC documentation as well.

[steingod]

Agree with your suggestion for information claiming, but not being from a controlled vocabulary. Skipping would generate issues in the aggregator role.