Thanks, Chuck for this detailed response. You are quite right that we need to be clear what we mean by "specimen". Your clarification of MOBOT's use of identifiers shows not only that there are many identifiers in use, but also they may apply to any in a series of increasingly refined objects (or sets of objects), and that there are good reasons for wanting to be able to identify each item in that series. If we think of this in software modeling terms, each of these could be a separate object which could be manipulated and referenced independently of the others.
Different communities within biological collections, will clearly have different series of identifiable objects. For example an entomological collection could have the following series:
(Survey?) -> Contents of an (malaise/light/water/etc.) trap -> Individual insect -> Insect part (genitalia preparation, leg removed for DNA analysis) -> (DNA preparation?)
Handling of plankton samples, culture collections and seedbank accessions will be different again. Within botanical collections, is there any attempt to indicate that two separate collecting events relate to the same plant or clonal population?
Depending on the needs and purpose of an individual collection, it may track different items in these series. Individual insects may be part of a numbered series or have their own numbers.
As Chuck suggests, this means that it is not clear that we have a single common definition of "specimen" that would be accepted by all of us. My use of the word "subsample" and the phrase "identifiable set" in my original question was an attempt to recognise that one group's specimen may be seen by another group as just a part of a specimen or as a set of specimens. The ABCD Schema uses the general term Unit to reflect the variation between different items recorded by different providers.
It seems to me that there are various ways that we can try to handle this:
1. We could try to develop wording that explains what we agree to be a reasonable shared definition of a specimen that can be applied by each collection to select an appropriate identifier or require them to generate a new one. This seems unlikely ever to be successful given the wide range of situations, collections and databases that need to be covered. 2. We could let each provider give an indication of the nature of the item being referenced (sample with multiple organisms, individual organism, tissue, etc.; living material, dead material) using terminology that is appropriate to their community. This may help human readers of the data to interpret the data but does not allow us to reason reliably about the data we receive. This is close to the approach followed today by Darwin Core (BasisOfRecord) and the ABCD Schema (Unit/RecordBasis). 3. We could work as a community to develop and enforce a controlled terminology for the nature of items referenced. By limiting the range of terms that can be used, it should in many cases be possible to reason more clearly about what each record describes. 4. We could go further and manage the controlled terminology as an ontology that includes hierarchically-arranged definitions (e.g. a CultureCollection isA LiveUnit, a HerbariumSheet isA DeadUnit) and other relationships (e.g. a Tissue derivesFrom a DeadUnit). There would be more work in doing this, but the BioMOBY project provides one example of how to build such an ontology as an open community activity.
As we consider the use of GUIDs, I would really also like us to think about the fourth of these options. Any "Unit" (or whatever else we may use as a generic term for a biological item being recorded) can be identified as belonging to a particular class of objects identified within a shared ontology. We can do this by having an element whose value must be the identifier for an object class registered in the ontology. This allows an institution to make an assertion that one record relates to an individual dead organism and that another relates to a tissue sample, and for those assertions to be ones that software applications can process. Better still, the presence of GUIDs for each of these records would allow us to add an extra element to the tissue sample record that securely identifies the specimen from which it was taken.
The bottom line here is that we certainly need to do some work to make sure that we know what we are talking about when we speak of a "specimen" (or any other similar term), but that we can use a combination of GUIDs and a shared ontology to transcend the difficulties this could present, and to construct subtle and informative webs of information.
Donald
--------------------------------------------------------------- Donald Hobern ( mailto:dhobern@gbif.org dhobern@gbif.org) Programme Officer for Data Access and Database Interoperability Global Biodiversity Information Facility Secretariat Universitetsparken 15, DK-2100 Copenhagen, Denmark Tel: +45-35321483 Mobile: +45-28751483 Fax: +45-35321480 ---------------------------------------------------------------
_____
From: Taxonomic Databases Working Group GUID Project [mailto:TDWG-GUID@LISTSERV.NHM.KU.EDU] On Behalf Of Chuck Miller Sent: 22 October 2005 00:40 To: TDWG-GUID@LISTSERV.NHM.KU.EDU Subject: Re: Topic 2: GUIDs for Collections and Specimens
I am responding to Donald's questions as they apply at Missouri Botanical Garden.
As several have described, there are multiple layers of identification that occur with specimens, particularly botanical specimens.
Our physical herbarium specimens are structured in a hierarchy, starting from the original plant that was collected down to individual pieces with labels.
COLLECTION
Identification begins at collection. Multiple "samples" are usually taken from one plant or an entire small plant may be taken, a collector's number is assigned to the sample in the collector's field book along with notes and samples also numbered. Samples of other plants of the same kind may also be taken with different numbers assigned to each in the field book and on the sample. Samples may be made up of multiple pieces - leaves and stems, fruits, seeds, bark, etc. - some may be dried, others left wet. All of the pieces/samples of the one plant described in one numbered field book entry belong to the one organism noted by the collector.
PREPARATION
The pieces of dried or wet samples are shipped back to MBG with their identifying numbers. Nowadays, the information from the field book is recorded in Tropicos including the collector's number. A unique TropicosID number is assigned in database to the specimen or "sample" and the data from the field book is recorded including the collector's name and number. Accession numbers are assigned to each of the pieces of the sample that will be "mounted" in a different way. A mounting sheet has the accession number pre-printed on the sheet and the number applies to whatever is mounted on the sheet. But, a separate large fruit from the same plant would be put in a bag for instance and assigned a different accession number. Nowadays, these accession numbers are also recorded in Tropicos. A label is printed for the sheet and duplicate labels are printed for each of the related "accessions". They are all the same label with the TropicosID and collector's number on them.
DUPLICATES
Labels are also printed for the "duplicate" samples but no accession numbers are assigned to them and they are not mounted. The duplicates may be sent unmounted to specialists for determination or to other herbaria. The identification of these samples/specimens is what is printed on the included label - which includes Tropicos ID, Collector's Name and Collector's Number. The receiving institution may or may not assign additional numbers, mount the sample on a sheet, database it, etc. Totally up to them.
MOUNTING
The flat pieces are mounted on the sheets, large samples may require multiple sheets for one copy. Large things (fruits, bark, branches) may be put into bags or other holding methods. A barcode number is attached to the sheet and any additional pieces/accessions and recorded in Tropicos. A different barcode is on each piece or accession. So, barcodes have a one-to-one match to accession numbers. The duplicate printed labels are also attached to the sheet and any related pieces/accessions. If an attached barcode comes off and is lost, a new, replacement barcode is attached and updated in Tropicos.
The use of Lead Collector's Last Name and field book (also called catalog) number is very common in botany - eg. CROAT 10100. The collector-number method is frequently used in reference literature plus the addition of the Index Herbariorium code for the institution where the specimen was seen or gotten from. Duplicates of CROAT 10100 could be at MO, K, P, F, etc. and those sheets may have different accession numbers or no accession number at all.
Donald's Questions:
1. What identifiers (how many per specimen) get assigned to specimens in your organisation or domain (field numbers, catalogue numbers, etc.)?
On one mounted specimen sheet at MBG are the following numbers/identifiers:
- Accession number (100% unique)
- Barcode number (100% unique)
- Tropicos ID (applies to all accessions and barcodes for one sample/specimen)
- Collector's name and number (applies to all accessions, barcodes, TropicosIDs, and duplicate samples/labels sent to other institutions from the original collected organism)
All of these numbers are recorded in the Tropicos database.
2. What is the scope of uniqueness for each of these identifiers (notebook page, collector, database, institution, global, etc.)?
I attempted to describe this above.
Collector's numbers are commonly unique to a collector and don't repeat across notebooks, but the numbers are not unique themselves and are only unique when combined with Collector's name
Accession numbers and barcodes are unique to the sheet/bag they are attached to and are one-to-one with each other and are unique within the institution
TropicosID is unique within the database and the institution and is supposed to be one-to-one with collector/collector number.
Lead collector last name plus number is unique within the database and within the institution but not unique globally.
3. Can you explain the life cycle of each of these identifiers (who assigns them, how they are subsequently tracked)?
Described at the beginning.
4. Can you give examples of how these identifiers are used to retrieve the specimen and/or information on the specimen?
The primary search for specimens in Tropicos is by collector name and number.
5. Would there be any social or technical roadblocks to replacing these identifiers with a single identifier that was guaranteed to be unique?
Technically, it would require addition of an "alias" identifier and additional programming to enable searching on the alias.
Since there are 4 identifiers in hierarchical relationship, which of them could be the "single" identifier? This goes to my continuing question of "what are we trying to identify"? The original specimen (and its duplicates), a specific sheet, a specific part of a sheet, or part of a specimen in an alcohol bottle separate from the sheet?
6. In the case of subsamples from a specimen, can you identify issues around associating the sample and associated information with the source specimen and associated information?
By subsample, are we referring to the occurrence of "duplicates" of the original organism or rather to the pieces of it, like bark, fruit, leaves? What constitutes the "specimen" versus the sample? We really need to sharpen the language in these discussions to eliminate the round-robin responses that occur as everyone states their opinion of what they think the terms mean but no one decides exactly the definition to be used by everyone.
The biggest issue to me is that there are no standards for identification of anything below the level of the original collecting event and even the collector name + number is just a common practice in botany, not a "standard" and not universal by any means. The term "accession" means different things to different institutions. Accession number at MBG refers to an associated part of a specimen, not the whole specimen. Does catalog number mean the same thing everywhere? To some it means the collector's number.
I suppose another issue is that because of the common practice in botany of collecting duplicate samples and sending them around to other institutions, any worldwide count of databased specimens that does not account for these duplicates will overstate the real number.
The subject of specimen identifiers is somewhat linked to that of collection identifiers, since Darwin Core and the ABCD Schema have used institution and collection codes together with catalogue numbers to identify specimens in the absence of GUIDs. It would also be useful here to collect information on the following:
7. How are your specimens organised into larger identifiable sets (collections, named collections, databases, institutions, etc.)?
We don't separate our collections into sets, they are all part of one herbarium collection.
Accessions combine into one specimen.
Duplicate specimens can be at other institutions.
We do record the institutions where we know duplicates of a specimen are located but we do not record the other institution's catalog numbers
8. What identifiers get assigned to each of these sets in your organization or domain (institution codes, collection codes, Index Herbarium acronyms, etc.)? 9. Can you explain the life cycle of each of these identifiers (who assigns them, how they are subsequently tracked)? 10. Can you give examples of how these identifiers are used to locate the set and/or information on the set? 11. Would there be any social or technical roadblocks to replacing these identifiers with a single identifier that was guaranteed to be unique?
Previously discussed.