Re: Absence of key scientific spatial data formats within common formats to implementation of Best Practices [SEC=UNCLASSIFIED]

Hi Ed,

I think that the concept of 'small' is relative.

Perhaps the intended audience is 'small' by Google standards. However, we find that our website is one of the most visited in Australia. People are interested in the weather, climate, etc.

The domains that we work in relate to concepts that don't just occur at a single location and we try to understand what is happening, analyse results and make predictions on a regional and often global basis.

This is really another variant of the multidimensional coverage issue. The formats that Lewis has mentioned are coverage formats.

While I personally believe that the OGC WCS suite of related standards (including also WPS and WMS) will be very useful as described by Peter Baumann. We will still need a payload to exchange the actual multi-dimensional data and the formats discussed are a good starting point.

Again personally, I think that the piece of the puzzle that we are missing is being able to adequately describe the often quite complex semantics associated with the data.

While we are developing a good approach at being able to describe the semantics using Application Schema based on Observations and Measurements, Timeseries, etc (as described on this list by Simon Cox), we have still to find a way to effectively link this description of the observed property through to coverages that have often been derived from collections of such observations at points in time.

The semantics of the observations are increasingly based on a community of practice's agreed concepts and descriptions, as per the WaterML and GeoSciML examples provided earlier in this thread.

See the SDWWG use case on climate data and data provenance for a better description.

So I don't think that we can dismiss this issue as belonging to a small isolated use case with a small potential audience.

Bruce

________________________________
From: Ed Parsons <eparsons@google.com>
Sent: Tuesday, 29 March 2016 7:55:36 PM
To: Linda van den Brink; Bruce Bannerman; lewis john mcgibbney
Cc: SDW WG Public List
Subject: Re: Absence of key scientific spatial data formats within common formats to implementation of Best Practices [SEC=UNCLASSIFIED]


Hello all,

This is worth discussing in the context of the intended audience for the BP document, my personal view is that those formats you list are of interest to a predominantly  small (in relative terms) scientific audience ?

Ed

On Tue, 29 Mar 2016, 09:11 Linda van den Brink, <l.vandenbrink@geonovum.nl<mailto:l.vandenbrink@geonovum.nl>> wrote:
Hi Bruce, Lewis,

The scope and purpose of the common formats list in the BP hasn’t been discussed exhaustively. What’s currently in the BP is a first draft, or rather two; one list by Ed Parsons and one by Clemens Portele. Including scientific spatial data  formats in these lists hasn’t come up yet. It could be argued that scientific formats weren’t considered ‘common’ formats, but as I said it hasn’t come up yet. Both lists only list vector data formats.

I have created an issue about this so that we don’t forget to address this.
https://github.com/w3c/sdw/issues/237

Linda

Van: Bruce Bannerman [mailto:B.Bannerman@bom.gov.au<mailto:B.Bannerman@bom.gov.au>]
Verzonden: dinsdag 29 maart 2016 04:21
Aan: lewis john mcgibbney
CC: SDW WG Public List
Onderwerp: Re: Absence of key scientific spatial data formats within common formats to implementation of Best Practices [SEC=UNCLASSIFIED]

Hi Lewis,

More inline below.

Bruce


From: lewis john mcgibbney <lewismc@apache.org<mailto:lewismc@apache.org>>
Date: Saturday, 26 March 2016 at 08:24
To: Bruce Bannerman <B.Bannerman@bom.gov.au<mailto:B.Bannerman@bom.gov.au>>
Cc: SDW WG Public List <public-sdw-wg@w3.org<mailto:public-sdw-wg@w3.org>>
Subject: Re: Absence of key scientific spatial data formats within common formats to implementation of Best Practices [SEC=UNCLASSIFIED]

Hi Bruce,
Thanks for your response. I'll make some comments below

On Wed, Mar 23, 2016 at 1:26 PM, Bruce Bannerman <B.Bannerman@bom.gov.au<mailto:B.Bannerman@bom.gov.au>> wrote:
Hi Lewis,

I have still to find the time to review the latest document and provide comment.

OK. I would like to see your comments when you do and I suppose I will have some replies.


However regarding this issue, we have no intention of moving away from the scientific data formats that we use within our large data holdings.

Same here. This is where I see some value in addressing the following area in order to still drive value from the spatial data encoded within the dataset(s). Many (virtually all) of our dataset landing pages, believe it or not, still do not have any kind of semantic markup hence are relatively undiscovered outside of the NASA Data Active Archive Center (DAAC) portals. An example is the landing page at [0] which describes the SeaWinds on QuikSCAT Enhanced Resolution Regionally Gridded Sigma-0 (BYU, D. Long) dataset. When I extract the implicit semantic markup from within this page (using Apache Any23 [2]) I get very few meaning relationships which I can utilize programmatically. I extracted result in JSON shows you that.
I do however also think that moving towards a hypermedia-based mechanism for describing the data granules behind these dataset landing pages is also useful. I found Linda's recent post on the Dutch crawling task very interesting in this regard.
What are your thoughts here? Do you describe your datasets in any meaningful way? I think that there is a HUGE a mount of work to be done here to improve programmatic interpretation of the underlying scientific data.

[0] http://podaac.jpl.nasa.gov/dataset/QUIKSCAT_BYU_L3_OW_SIGMA0_ENHANCED?ids=Measurement&values=Sea%20Ice
[1] http://any23.apache.org
[2] https://paste.apache.org/lphl



Regarding describing our datasets:

  *   We don’t do this as well as we could. We will begin addressing this in the near future. Much of our current work is either internally focussed, or at a much too granular level.
  *   We intend describing our data sets using ISO 19115 with support for several profiles, including WMO and ANZLIC.
  *   I can’t see us moving away from this paradigm, but there is certainly potential for LinkedData approaches as alternate methods of discovering our data.
But this is also only part of the issue:

  *   We also need a mechanism to better understand the context of our observations (e.g. What sensor; what model; when was it last calibrated; maintained; what sensor maintenance process and responsible party; what observation process etc). We will be using the new WMO WIGOS Observations Metadata standard to support this concept.
  *   As discussed before on this list (and in the SDWWG Climate data related use case), there is also the issue of data provenance.
  *   Data Quality and IP issues will also become a big issue, particularly with the increasing use of mixed Bureau and 3rd party observations and the subsequent derived products that we create from these observations.



If anything, I expect that we will need to work with our peers to define formal data format definitions that are consistent with modern spatial requirements, e.g. full support for Spatial Reference Systems and other CRS definition and that don't constrain our ability to adequately portray the complexity of our data.

I agree here.

I expect that we'll probably need to do this via OGC processes. We want to ensure that the data that we collect and archive now will still be accessible for our key stakeholders who have not yet been born.

This view is consistent across the entire NASA data archival spectrum as well... and as long term data stewards this is a logical viewpoint.


Further, I expect that we'll need to go further and work with our peers to agree on semantic definitions of the content that we portray for each relevant domain and its inter-relationships with other domains.

This sounds like the next step... the issues we're discussing above seem like the precursor. Am I correct?

Not necessarily, consider the work that has been undertaken on GeoSciML, WaterML etc.

A lot of this is based on communities of a common interest getting together and agreeing on and using common terms and concepts.

It takes many, many years of community building to reach the required consensus.





This is similar in concept to what the hydrology community have done with WaterML 2, but I expect that we'll need to take it further, particularly the inter-domain relationships.

Yes, I really like ongoing work on hydrology with WaterML2 and this is an excellent point. It is however again, in my own opinion, something which follows on from he above.


When we are trying to understand global systems and their interaction with other systems, and we are doing this with our peers in distributed data collections and services, the need for formal data definitions become critical. This is especially so if we want global, federated, data sets ***and dynamic services*** describing specific phenomena.

Do you have any examples from the field of Meteorology? i would be interested to see if I could pick out any examples more familiar to other aspects of Earth Science, Pysical Oceanography or something else a bit closer to 'home' for my current working agenda.


The closest that I can point to at the moment is the work that we have been doing in WMO on WMO #1131, Climate Data Management System Specifications http://library.wmo.int/opac/index.php?lvl=notice_display&id=16300<http://library.wmo.int/opac/index.php?lvl=notice_display&id=16300>

There is also related work, e.g.:

  *   Foundation data governance and data modelling work within WMO that Jeremy Tandy is leading
  *   Foundation work that has been undertaken by Australia’s CSIRO over many years: https://www.seegrid.csiro.au/wiki/Siss/WebHome
  *   And to be honest, much of the underpinning OGC standards efforts that we build on top of.

This is really laying the groundwork, and it will take many years to get there with truly federated data and data services.





It will allow us to spend much less wasted time in getting data prepared for global analysis and much more time on the actual analysis and understanding the implications of the results.
Agreed!
Thank you for the very meaningful conversation. Looking forward to any follow up if you have it.
In the meantime, I come back to my main question. Is there any reason from across the group why the current matrix of spatial data formats doesn't include formats such as GRIB, HDF4, HDF5, netCDF3, netCDF4, etc? These are used pervasively throughout the sciences and I am very surprised to see them absent.
Thanks folks.
Lewis

--

Ed Parsons FRGS
Geospatial Technologist, Google

Google Voice +44 (0)20 7881 4501
www.edparsons.com<http://www.edparsons.com> @edparsons