wiki:SpecimenHandling

Specimen Handling

Being notes prompted by a discussion between Spiros and Bryan ... but the blame is all Bryan's (at least with this version). The context is of course, ISO19156: Observations and Measurements. <tesulting discussion is either resulting in changes, or being (initially) annotated into the page using

the vertical bar notation.

Some useful definitions

These definitions are meant to help frame a discussion. They're not necessarily MOLES terms (yet), and they may not (yet) all map directly onto the correct O&M term.

Feature of Interest : This is the thing which one is interested in observing, an observation obtains an estimate of the value of a property of the feature of interest.

Sampling Feature(SF): An artefact of the observation process (OM_Process). One uses a SF as a proxy for the feature of interest itself. The SF is "of the process" and should not be confused with the result of an observation, even though the result of an observation can be a coverage which has a domain which has the same spatio-temporal discretisation as a ("the"?) sampling feature, and a range which is "the" result.

In the next few definitions, unless otherwise indicated, process should not be confused with OM_Process, it's simply being used as a noun with it's standard definition

Observation: An observation is an act that results in the estimation of the value of a feature property, and involves application of a specified procedure, such as a sensor, instrument, algorithm or process chain. The procedure may be applied in-situ, remotely, or ex-situ with respect to the sampling location.

Acquisition: A process which interacts with the feature of interest to provide either a result or a specimen sampling feature.

  • It uses an instrument (sensor, sampling protocol etc) and interacts with the feature of interest in some way to produce a result, or a specimen, or both.
  • If a specimen is acquired, then the properties of the sampling method need to be captured as an attribute of the acquisition.

SC: the term 'acquisition', as used in ISO 19115-2 and presumably reflecting common usage in the imagery/remote-sensing community, refers to the initial capture of an image. AFAICT the result of this is 'numbers', which may then be processed into better numbers. This can be contrasted with 'acquisition of a specimen', the outcome of which is a thing that you plan to take to a lab. It ain't numbers yet.

BL: So we need to be clear about how we use these terms (or use different ones) and link to ISO19115.

Computation: A process which involves only numerical computation. It may begin with one or more source data objects (themselves observations), but it may begin ab initio (no inputs). It is executed by a responsible_party and ends up with a result.

  • The computation does not interact with the feature of interest (by definition).

Analysis: A process which begins with a specimen and results in a numerical result.

  • This term is going to cause trouble, but for now, we can use it with this definition.

Preparation: A process which interacts with a specimen sampling feature to produce a new specimen sampling feature, but does not produce a numerical artefact.

OM_Process: The process which generates the result. Using the terms above, we might think of an OM_Process of consisting of one of the following:

  • An acquisition which generates a result (an estimate of the intended property of the feature of interest).
  • An acquisition that generates a sampling feature which is not an estimation of the intended property, followed by a computation which does generate a result which is an estimation of the intended property.
    • In this case, the first acquisition may have generated a result, but not the result. We should consider that result as an observation in it's own right (since we would have intended to create that result), and we should link these observations together.
  • A computation, or sequence of computations (each related, with interim results which may or may not be recorded as individual observations).
  • An acquisition which generates a specimen, followed by an analysis.
    • A specimen may well have undergone a number of preparation steps.
      • If those preparation steps have created new specimen (as may have occurred, if say, an ice core were sliced up into pieces sent to different laboratories, then those relationships should be recorded).
      • If the specimen itself has been transformed, then the steps should be recorded as a set of steps applying to that specimen.

Some Use Case Stories

(Consistent with the vocab above: if you want to change the vocab, you have to change and/or add-to the stories!)

We start with a story that isn't specimen handling, but included here to give the "canonical" use case.

The Chilbolton Story

We want to measure rainrate. Both Spiros and Bryan do it using the following sequence of actions.

Spiros:

  1. Deploys "the" radar to make measurements of signal strength using a "profile" sampling feature proxying for the atmosphere.
    • So the result of this observation is signal strength using this instrument and the process is an acquisition which includes an algorithm which turns voltages into measures of signal. Now
  2. A computation process takes the signal strength result and produces a rainrate result (still on the same sampling feature) (using a new algorithm, call it "SpirosProgram").

Bryan:

  1. Takes Spiros' signal strength and uses BryanProgram algorithm to generate rainrate.
    • Whether or not Spiros intended to create an interim result, he's done it, since Bryan's used it.

Now Chris can come along and compare Bryan's measurement of rainrate and Spiros' measurement of rainrate, and discover that it's the algorithms that take signal strength and produce rainrate that differ. He may trust one algorithm more than the other, so he may trust one dataset more than the other.

In this story, the sampling feature is always a profile of the atmosphere (the feature of interest). The property we go for changes through the aggregation of processes ...

The Ice Story

(I know I could write this with O&M language, but I'm trying to be consistent with the definitions above).

Simon and Andrew are interested in ice cores ...

Simon:

  1. Wanders off to Greenland and comes back with some rather long ice cores (specimens). His specimen acquisition consisted of using a very long corkscrew and generated a very long piece of ice (and did not generate a result).
  2. Back in his lab, he carefully slices a piece of ice of the original. He puts the original back in the freezer, and ships the sliver off to Andrew's lab. There are now two specimens hanging around:
    1. The IceCore was intended as a proxy to the temperature of the atmosphere in past times, so the feature of interest is the atmosphere and the property (of interest) is temperature.
    2. The Sliver carries the same FoI and proxy relationships, or does it?
      • Option A: The same proxy relationships etc, so we assume the specimen acquisitoin shouldn't change between IceCore and Sliver, but a preparation attribute of Sliver explains what was done to create it.
        • We want the fact that we used Andrew's lab to be an attribute of the preparation.
      • Option B: Sliver is a specimen of IceCore, so the acquisition applies to how it was obtained from IceCore and we use a relationship between sampling features to get back to the atmosphere in a chain of steps.
      • I don't like option B. Tell me it's wrong.
  3. Andrew sends !Sliver off to Nic's lab. Nic vaporises the Ice and puts it through a mass spectrometer to get a measure of the isotope of interest. Nic has generated the result, via one preparation step and one analysis step.
    • So the OM_Process associated with the result has a number of sub-processes associated with it:
      • The acquisition of the core
      • The preparation of the sliver
      • The preparation of the gas
      • The analysis of the isotopic ration of interest.

Some perspectives on how and why we collect these information

(From Roy:)

Two classes of metadata:

  1. Metadata to be populated by the scientist in the laboratory: essentially an XML laboratory notebook that preserves as much detail as possible of the analytical process.
  2. Metadata to be populated from the information supplied with data: essentially an XML encoding of the 'methodology' section in a publication that documents key facts considered by the author to be relevant for data understanding and evaluation.

Definitions from Standards

These are definitions from actual standards:

ISO19156 O&M

OM_Process: The purpose of an observation process is to generate an observation result.

SF_Process: The purpose of a sampling process is to generate or transform a sampling feature.

ISO19115-2

Acquisition: Strictly ISO19115 doesn't define a class for acquisition, it defines a package of classes, see MI_Acquisition, but the general gist is "provides acquistion of imagery and gridded data".

BL: Both of which are results ... as opposed to specimens. However, it's not obvious that if we're extending from ISO19115 part 2 into a world of specimen collections, that the word acquisition is not appropriate. However, as noted above, we may need to disambiguate between acquisition of a specimen sampling feature and a numeric result.

 Trac Powered
Site hosted at the
British Atmospheric Data Centre
for the