Hi Markus,

Much of the "muddle" that I was referring to has been exposed in a series of recent hackathons (to which we have tried to always invite someone from GBIF), and has not yet been published, although [1,2,3] touch on the issue. I will try to describe some of the problems we have encountered here.

For those of you who are tired of this thread already, the short answer is that the problem lies in the lack of underlying clarity and semantics in Darwin Core, not DwC-A. I think Darwin Core archives do a pretty good job of capturing the content of Darwin Core terminology, but as those terms are often (sometime intentionally) ambiguous and imprecise, those aspects are captured too.

Below are a few of the specific issues. I don't think much of this will be new to the readers of this list.

1. Ambiguously defined classes: Darwin Core only has a handful of classes (Occurence, Event. Location, etc.), and those are intended to be used as grouping classes, rather than strictly defined entities. Therefore, if someone uses the corresponding properties (OccurenceID, EvenID, LocationID), they are more or less free to attach those IDs to whatever they want. In many cases, the type of entity referred to by OccurenceID can be inferred from the object of basisOfRecord, but not always, and it is still an inference with some error. IF DwC does go in the direction of capturing sample/survey data, one thing that would help would be to have it specified via a set of controlled vocabulary terms in basisOfRecord. I won't even start with Taxon.

2. Domain- and range-less properties: DwC properties (the bulk of the terms in DwC) are intentionally left without domains and ranges to make them maximally re-usable. For traditional museum specimen data, this works reasonably well, and the intended meaning is often clear. We recently completed the exercise of specifying domains and ranges for the whole DwC vocabulary, as interpreted for an Occurence Core archive, and found that many of them do not refer to DwC classes as their domain. This makes in impossible to infer their meaning without a set of external assumptions. Our interpreted ranges of DwC properties are a mix of data values (literals) and potential classes in other ontologies. {I will share this mapping on request - it is still under review}. The proposed properties also have no domains or ranges. For example, the suggested range of 'quantity type' is a heterogeneous mix of entities (individuals, biomass, %species, scale type) that sets off a huge red flag to my ontological self. And what about the domain of 'quantity type' and 'quantity'? As they are in the Occurence extension table, they probably are meant to describe whatever is represented by occurenceID, but this is not specified.

Vague domains and ranges may work for simple sampling schemas that could be interpreted as  "taxon W (value of occurenceID) has quantity X (value of quantity) with unit Y (value of samplingUnit) of quantity type Z (value of quantityType)", but even then it is likely to be ambiguous. Now suppose W=123someID (with taxonName as Bufo bufo), X=9, Y=individuals, and Z is left blank, because that is just what people do sometimes. Let's be generous and suppose that there is also some sampling geometry and location information. How to interpret this? Does this mean that there is a jar in a museum with 9 toads in it, all of which came form a single collecting event at that location? Does it mean that there was a survey done of a the location described, and 9 toads were counted but not collected? Does it mean that they surveyed the area and estimated toad abundace at 9 toads m^-2?  Even if they are conscientious and put in m^2 as the sampling unit, the meaning is still unclear. Did they measure an area of 1 m2 and find 9 toads or an area of 90 m2 and find 90 toads or did measure an area of 90 m2 and find 9 toads? All of these would be valid interpretations. Of course, many sampling schemas are much more complex with this, with nested sampling protocols or repeat sampling of the same area. There is just no way to capture that without more semantics the DwC can provide.

3. Using the wrong property for information: Even if the larger community interprets a DwC property with some certainty, that doesn't stop people from using it wrong. This is often not the fault of DwC or DwC archives, if people don't take the time to make sure they are filling out the fields properly. Even with very clearly defined ontology terms, some curators just don't take the time to read an understand the definitions. However, the ambiguity and vagueness of DwC terms surely contributes to this.
-----
I think that sampling or survey data is just to structured to describe well with a flat schema like DwC. That is why most of it to date has been stored in relational databases. To repeat myself, I think it would make more sense to first develop a sound semantic model for the data then extract a Darwin Core archive format for transmitting it, rather than build the DwC archive first and later try to map it to a semantic model.

Ramona

[1] http://www.biomedcentral.com/1471-2105/15/257
[2] http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0089606
[3]http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3746421/

------------------------------------------------------
Ramona L. Walls, Ph.D.
Scientific Analyst, The iPlant Collaborative, University of Arizona
Research Associate, Bio5 Institute, University of Arizona
Laboratory Research Associate, New York Botanical Garden


On Sat, Aug 30, 2014 at 3:00 AM, <tdwg-content-request@lists.tdwg.org> wrote:
Send tdwg-content mailing list submissions to
        tdwg-content@lists.tdwg.org

To subscribe or unsubscribe via the World Wide Web, visit
        http://lists.tdwg.org/mailman/listinfo/tdwg-content
or, via email, send a message with subject or body 'help' to
        tdwg-content-request@lists.tdwg.org

You can reach the person managing the list at
        tdwg-content-owner@lists.tdwg.org

When replying, please edit your Subject line so it is more specific
than "Re: Contents of tdwg-content digest..."


Today's Topics:

   1. Re: tdwg-content Digest, Vol 63, Issue 6 (sigh) (Markus D?ring)


----------------------------------------------------------------------

Message: 1
Date: Fri, 29 Aug 2014 12:02:40 +0200
From: Markus D?ring <m.doering@mac.com>
Subject: Re: [tdwg-content] tdwg-content Digest, Vol 63, Issue 6
        (sigh)
To: Ramona Walls <rlwalls2008@gmail.com>
Cc: TDWG Content Mailing List <tdwg-content@lists.tdwg.org>,    John Deck
        <jdeck@berkeley.edu>
Message-ID: <A6B43FE0-A595-4526-BD13-EE196D950435@mac.com>
Content-Type: text/plain; charset="windows-1252"

Dear Ramona,

could you point me to the evidence for muddled semantic mappings in existing dwc archives? I would like to better understand the problem. Is it a general issue with Darwin Core terms and their lose definition or just their application in dwc archives?

best,
Markus



On 29 Aug 2014, at 06:35, Ramona Walls <rlwalls2008@gmail.com> wrote:

> I also really do appreciate GBIF's pressing need to serve survey/sample data, and I don't have a major of a problem with the idea of an Event core or even adding new terms to DwC (in principle). Rather, I am urging caution in how we proceed with it.
>
> ?amonn, in response to your statement "Once the BCO model is available for uptake, it should be possible to develop a mapping between it and the simple DwC sample model" I will respond: possible, maybe, easy, no way, unambiguous, probably impossible. The problem I foresee is that once the "simple DwC sample model" is in place, people will start using it to do all kinds of not so simple things, and the mapping will become muddled. We have ample evidence that this is the case with existing Darwin Core archives.
>
> Going back to the five new terms that ?amonn proposed, I would like to see if we can link them NOW to existing ontology terms (as other have proposed), thus making their semantics explicit from the start, but still allowing GBIF/EU-BON to proceed with the work they need to do. This will not prevent people form misusing terms, but may at least help make mapping easier later. In cases where the terms can't be mapped to an existing term, BCO curators would be willing to help develop a term or set of terms that can convey meaning required, or work with other ontology developers to get the terms added elsewhere.
>
> Trying to be constructive, I attempted to do a quick and dirty, preliminary mapping of the five terms (quantity, quantity type, sampling geometry, sampling unity, and event series ID), bearing in mind that I am not an authority on OBOE or OGC ontologies. [Aside: Based on what I know, OGC ontologies are not yet sufficiently developed to provide the semantics we need, but I would love for someone to show me otherwise.]
>
> A serious problem with mapping these terms to existing ontologies is that some of them do NOT map to a single ontology term (namely, quantity, quantity type, and sampling geometry). This is evidence that the proposed terms could indeed be interpreted in multiple ways and further supports the argument that it would not be easy to retrospectively add them to a semantic framework at some later date.
>
> I think there is a path forward that would allow for both the expressiveness of OBOE and other ontologies and convenience of standard exchange formats.
>
> Ramona
> ------------------------------------------------------
> Ramona L. Walls, Ph.D.
> Scientific Analyst, The iPlant Collaborative, University of Arizona
> Research Associate, Bio5 Institute, University of Arizona
> Laboratory Research Associate, New York Botanical Garden
>
>
> On Thu, Aug 28, 2014 at 8:58 AM, Robert Guralnick <Robert.Guralnick@colorado.edu> wrote:
>
>   Hi all --- Ok, I think the scope of the issue is quite clear. Let me summarize:  1)  As ?amonn and the rest of GBIF has made quite clear,  "GBIF is faced with the immediate task of making sample-based data discoverable and accessible using its current ecosystem of tools" given a funding mandate from EU-BON.  2)  The solution for this problem is to develop an Event-core and to promote new terms to the Darwin Core to make this happen.   I will note a small inconsistency here:  the current ecosystem standards and tools of is Darwin Core (as it stands) and publishing systems such as IPT.  That ecosystem of tools includes mechanisms to extend Darwin Core where needed, via extensions.  The current ecosystem of tools doesn't include new Cores or new DwC terms, does it?
>
>   So this leads in nicely to the contentious issue(s) and places where there seems to be discussions --- these have to do with the nature of the changes suggested and the scope of those changes, both in terms of an Event core and DwC term additions.  Leaving aside the Event-core for now, the key questions simply about term additions to the Darwin Core that seem to be at heart here are: 1)  Is the intent of the Darwin Core to model surveys, which usually involve multiple kinds and types of sampling over multiple sites using multiple methods?  2)  Is the solution to invent new terms for the Darwin Core if there are already terms from other efforts, wouldn't we work with those existing efforts to assure interoperability?
>
>    I appreciate the efforts of GBIF here fully, and am personally torn because on the one hand, I fully agree with the goal of extending Darwin Core to better represent  richer biodiversity data. On the other hand, I worry about process here and how to make that happen in a way that isn't too hasty or locks us into just the opposite of what I think many of us want with regards to sharing data more broadly than within just one ecosystem of tools.
>
> Best, Rob
>
>
>
>
>
> On Thu, Aug 28, 2014 at 6:30 AM, John Deck <jdeck@berkeley.edu> wrote:
> I see the rational for enabling this in Darwin Core Archives and adding the new terms.  However, back to what Matt Jones brought up: "won't we just end up with a new syntax that does essentially what O&M and OBOE do now?".
>
> We should include explicit references to existing terms/definitions that encapsulate what we're talking about, e.g. in our MaterialSample proposal last year we linked the an existing term in OBI, which has a much richer description and context for MaterialSample than what we considered (https://code.google.com/p/darwincore/issues/detail?id=167)
>
> Have we explored the possibility of doing this with OBOE?  I'm not suggesting we adopt OBOE wholesale, but it seems like we have a good opportunity to enable better semantic linking with that efforts.
>
> John
>
> On Thu, Aug 28, 2014 at 4:23 AM, ?amonn ? Tuama [GBIF] <eotuama@gbif.org> wrote:
> Thanks, Ramona and Rob.
>
> I'd like to add a few points following on Markus's reply.
>
> I think your pressing of the need for a robust semantic model for
> biodiversity sample/survey data is incontestable ? we do need one and it
> should enable rich data integration once it is defined and the tools and
> data standards to support it become available. However, GBIF is faced with
> the immediate task of making sample-based data discoverable and accessible
> using its current ecosystem of tools (IPT) and exchange standards (DwC;
> EML). Waiting for a functional, implementable semantic model and the tools
> and support services for it is just not an option for us right now.
>
> We have already spend considerable time in analysing the merits of
> Occurrence core vs Event core and have opted for an Event core for reasons
> previously given. I don?t believe we are trying to reconfigure Event (?an
> action that occurs at a place and during a period of time?) and regardless
> of whether we use Occurrence or Event, the need for some additional terms
> arise (e.g., quantity, quantityType, samplingGeometry, samplingUnit). Once
> the BCO model is available for uptake, it should be possible to develop a
> mapping between it and the simple DwC sample model.
>
> So GBIF?s stance is that we need to take a two-pronged approach by exploring
> how the IPT and DwC-A can be adapted for publishing sample-based data in the
> near term while supporting the work of TDWG and groups such as the BCO in
> advancing biodiversity informatics. GBIF has already engaged in the work of
> the BCO and will continue to do so.
>
> ?amonn
>
> -----Original Message-----
> From: tdwg-content-bounces@lists.tdwg.org
> [mailto:tdwg-content-bounces@lists.tdwg.org] On Behalf Of Markus D?ring
> Sent: 28 August 2014 12:44
> To: Ramona Walls
> Cc: TDWG Content Mailing List
> Subject: Re: [tdwg-content] tdwg-content Digest, Vol 63, Issue 6
>
> Hi Ramona & Rob,
>
> The Event proposal does not try to change the semantics of an Event, it just
> uses the existing Darwin Core Event "class" at the core in Darwin Core
> archives. The actual change proposed is simply adding 3 new terms to the
> Event "group" to better share information about sampling methods & efforts,
> extending the existing limited capabilities of Darwin Core which already has
> the terms dwc:samplingProtocol and dwc:samplingEffort. It also proposes 2
> new terms for dealing with quantity of Occurrences, something that has been
> discussed since 2012 now, when I had proposed a new abundance term [2].
>
>
> In general application of Darwin Core is not at all limited to specimens and
> observations. It is used for sharing taxonomic datasets already and it's
> definition and goal is broad. Let me cite some of the introduction to Darwin
> Core [1]:
>
> What is the Darwin Core?
> The Darwin Core is body of standards. It includes a glossary of terms (in
> other contexts these might be called properties, elements, fields, columns,
> attributes, or concepts) intended to facilitate the sharing of information
> about biological diversity by providing reference definitions, examples, and
> commentaries. The Darwin Core is primarily based on taxa, their occurrence
> in nature as documented by observations, specimens, samples, and related
> information.
>
> Motivation: The Darwin Core standard was originally conceived to facilitate
> the discovery, retrieval, and integration of information about modern
> biological specimens, their spatiotemporal occurrence, and their supporting
> evidence housed in collections (physical or digital). The Darwin Core today
> is broader in scope and more versatile. It is meant to provide a stable
> standard reference for sharing information on biological diversity. As a
> glossary of terms, the Darwin Core is meant to provide stable semantic
> definitions with the goal of being maximally reusable in a variety of
> contexts.
>
>
> Markus
>
>
> [1] http://rs.tdwg.org/dwc/index.htm
> [2] https://code.google.com/p/darwincore/issues/detail?id=142
>
>
> --
> Markus D?ring
> Software Developer
> Global Biodiversity Information Facility (GBIF)
> mdoering@gbif.org
> http://www.gbif.org
>
>
>
> >> On Tue, Aug 19, 2014 at 6:11 AM, ?amonn ? Tuama [GBIF] <eotuama@gbif.org>
> wrote:
> >>
> >> Dear All,
> >>
> >>
> >>
> >> GBIF is committed to exploring ways in which the IPT and Darwin Core
> Archive format can be extended for publishing sample-based data sets. In
> association with the EU BON project [1], a customised version of the IPT [2]
> has been deployed to test this using a special type of Darwin Core Archive
> in which the core is an ?Event? with associated taxon occurrences in an
> ?Occurrence? extension.
> >>
> >>
> >>
> >> The Darwin Core vocabulary already provides a rich set of terms with many
> relevant for describing sample-based data. Synthesising several sources of
> input (GBIF organised workshop on sample data, May 2013 [3], discussions on
> the TDWG mailing list in late 2013; internal discussion among EU BON project
> partners), five new terms relating to sample data were identified as
> essential. The complete model including these new terms are fully described
> with examples in the online document ?Publishing sample data using the GBIF
> IPT? [4].
> >>
> >>
> >>
> >> As a first step towards ratification, we would like to register the new
> terms in the DwC Google Code tracker [5] if there are no major objections on
> this list. The five terms are:
> >>
> >>
> >>
> >> 1.      quantity: the number or enumeration value of the quantityType
> (e.g., individuals, biomass, biovolume, BraunBlanquetScale) per samplingUnit
> or a percentage measure recorded for the sample.
> >>
> >>
> >>
> >> 2.      quantityType: :  the entity being referred to by quantity, e.g.,
> individuals, biomass, %species, scale type.
> >>
> >>
> >>
> >> 3.      samplingGeometry: an indication of what kind of space was
> sampled; select from point, line, area or volume.
> >>
> >>
> >>
> >> 4.      samplingUnit: the unit of measurement used for reporting the
> quantity in the sample, e.g., minute, hour, day, metre, metre^2, metre^3.
> It is combined with quantity and quantityType to provide the complete
> measurement, e.g., 9 individuals per day,  4 biomass-gm per metre^2.
> >>
> >>
> >>
> >> 5.      eventSeriesID: an identifier for a set of events that are
> associated in some way, e.g., a monitoring series; may be a global unique
> identifier or an identifier specific to the series.
> >>
> >>
> >>
> >>
> >>
> >> Best regards,
> >>
> >>
> >>
> >> ?amonn
> >>
> >>
> >>
> >> [1] http://eubon.eu <http://eubon.eu/>
> >>
> >> [2] http://eubon-ipt.gbif.org <http://eubon-ipt.gbif.org/>
> >>
> >> [3]
> http://www.standardsingenomics.org/index.php/sigen/article/view/sigs.4898640
> >>
> >> [4]  <http://links.gbif.org/sample_data_model>
> http://links.gbif.org/sample_data_model
> >>
> >> [5] https://code.google.com/p/darwincore/issues/list
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> ____________________________________________________
> >>
> >> ?amonn ? Tuama, M.Sc., Ph.D. (eotuama@gbif.org),
> >>
> >> Senior Programme Officer for Interoperability,
> >>
> >> Global Biodiversity Information Facility Secretariat,
> >>
> >> Universitetsparken 15, DK-2100, Copenhagen ?, DENMARK
> >>
> >> Phone:  +45 3532 1494 <tel:%2B45%203532%201494> ; Fax:  +45 3532 1480
> <tel:%2B45%203532%201480>

>

End of tdwg-content Digest, Vol 63, Issue 15
********************************************