Borehole core

[mbeaufils]

Should we consider it as a sample ?

[dponti]

Sorry for the long post. At first I thought this was a silly question, but it's actually important, and I changed my mind about 3 times when writing this. A core extracted from a borehole would certainly fit the definition of a MaterialSample, although it has a shape (cylindrical) and therefore there are dimensional properties one might want to report about a core (eg. diameter, length) that might not be relevant to other types of MaterialSamples. The other little twist (a big one) is that a core can have properties that vary within it.

In that context, we should then define a borehole core as a SpatialSample in O&M-speak, since it has a shape, can be assigned to a location, and can have properties that vary within it. A MaterialSample as defined by OEM has no geometry properties.

But of course in O&M, if my understanding is correct, a borehole is also a SpatialSample... This is why I'm having a problem with the semantics in O&M. I understand conceptually how a borehole and a core extracted from a borehole could be equivalent objects, but this notion isn't comfortable for me, and I would imagine for many domain experts who think of boreholes and samples as different enough things to be resistant of putting them both into a single object class. This doesn't mean we shouldn't use O&M's model and nomenclature, but I think it will be a harder sell, and there may be a use case or two where the O&M model doesn't work well.

In the Book A concept Miro board that I introduced in #12 , I considered the ObservationReferenceFrame object (ORF - which is essentially equivalent to O&M's SpatialSample) to not be directly related to a MaterialSample. A MaterialSample can be associated to an ORF though a SamplingActivity but they do not derive from the same abstract class. The location and geometry of a MaterialSample, if relevant, is contained in the associated SamplingActivity class that links a MaterialSample to a location within an ORF. This provides the equivalent of having MaterialSamples having geometries in appropriate contexts without having to equate a borehole with a sample conceptually.

In the concept I proposed let's look at an example:

1. The borehole from which the core is obtained is a 1D ORF (or SpatialSample per O&M). It has a reference point property (top of hole) point geometry location defined in (preferably) a geographic SRS. It would also have a line geometry property (centerline) that defines the path in geographic space of the hole from (or through) the top of hole (reference point) to the bottom of the hole. And, it would have a linear spatial referencing property - an object defining the borehole's centerline property as the linear element of a linear spatial reference system (LSRS) whose method would define how potential ObservedZones in the hole are located in 1D along that centerline. In this case ,let's say that is absolute distance along the path from the reference point in positive meters toward the end of the path (bottom of hole).
2. The core is extracted from the hole by a SamplingActivity. The SamplingActivity object itself would have a location in the borehole, say an interval from 5 to 6.5 meters depth as recorded in the borehole's LSRS. The SamplingActivity object would also have a sampleProduced property - an object for each MaterialSample produced by the activity. In this case there is only one SampleProduced object, and within it would be a location property for the core (say an interval from 5-6 meters depth). Since there was not full recovery, a decision was made to "hang" the sample at the top of the cored interval. If a core run was made in the borehole but no sample recovered, then the SamplingActivity would have no sampleProduced property instantiated.
3. Note that the core sample's geometry is a line - constrained by the borehole's own geometry. For MaterialSamples that are cores (or otherwise have a defined shape), I would propose that the specialized core sample have a relative spatial referencing property (like we do for a borehole) that in this case would use the SampleProduced geometry (line in this case) as the linear element for the core's SRS. And in this case we could define the distance unit as cm as part of the linear referencing method. Note, this additional relative referencing property is not necessary, except when anonymizing of core sample is required and the sample's location should be withheld during data transfer.
4. The core sample object itself would be associated with the SampleProduced object in its associated SamplingActivity. in this case the core would be a MaterialSample in O&M's concept as it does not contain any geometry properties itself.
5. In describing the core in the lab, we observe a 10 cm thick sand layer located at 20 cm in the core as measured from its top. This location for the observation could be recorded in the sample's relative reference system (20-30 cm) where the sample's location is not provided to the lab technician), or in the borehole's LSRS (5.2 - 5.3 m), depending on the circumstance. The SamplingActivity's SampleProduced object, which is associated with the MaterialSample, contains the information needed to place the observation at the feature of interest in geographic space.

The issue of sample anonymizing is one place where I see a potential issue with the O&M model (besides the semantics). If a borehole core should rightly be a SpatialSample, then its geometry property is contained within it. I note that the geometry property is apparently optional so that that property could be omitted during data transfer, but this would then mean that different instances of the same object (one with and one without geometry) would need to be maintained, and this doesn't strike me a good practice.

[dponti]

As an aside, as a geologist as opposed to a geotechnical engineer I find it curious that many of my (at least US) engineering colleagues commonly constrain the use of "core" to rock - cores produced by diamond core drilling equipment. Other types of "intact" cylindrical samples in soils seem to mostly fall under the term "undisturbed sample" if the sampler walls are thin enough compared to the core barrel diameter.

I don't know how the rest of you all feel about this (or if we even need to worry about it), but I prefer that the term core be generalized to include any cylindrical MaterialSample with its size defined by thesample's diameter and length. Whether a Core is "undisturbed" or not depends on the sampling procedure (and could be a property of the MaterialSample and equipment used, and whether core material is rock or not depends on the characteristic of the feature of interest where the core is obtained.

[dponti]

Finally, should we create a borehole core as a specialization? If we classify a borehole core as a MaterialSample, I would say no, since MaterialSample already contains attributes to define a sample's dimensions. No need to specialize a block sample either, or any other "intact" sample. We would need to specialize if we consider a core sample as a SpatialSample in order to add dimension attributes. Same for any other "intact" SpatialSamples for which dimension attributes are important to record.

If anything we maybe should specialize MaterialSample according to its state (eg, soil, rock, gas, fluid). This may be more pertinent that how the sample is shaped (or not).

[Didymograptus]

Dan some interesting thoughts. I think that counts as a three-beer discussion ;-)

I think that to answer the questions we need to understand the purpose of Book A (I am still unsure). Is it to provide an alternative to AGS and DIGGS (I can't see any point in YAGDTF) or is it to communicate the factual data that is used to create ground models and hence the quality of those models? These are very different things. A DIGGS/AGS alternative would need to represent the real world and essentially be activity-based. However factual data for a ground model provides the opportunity for abstraction, and a focus on XYZC points and the important metadata associated with those points.

By way of abstraction I mean that rather than infer sample quality from the activity of drilling and sampling (something that requires in-depth knowledge of the subject), the sample quality can be assigned a value. https://soilpropertytesting.com/EURO_CODE_7_files/SAMPLING%20METHODS%20AND%20CLASSES%20feb%202011.pdf

I wonder whether it would actually be better to start with Book B to help define the scope of Book A?

[mbeaufils]

Answers to @Didymograptus

Dan some interesting thoughts. I think that counts as a three-beer discussion ;-)

+1. Major conceptual models were beer driven :-)

we need to understand the purpose of Book A

Book A is where you will find all the factual data collected before / during the project. They are the base information that enable to build / adjust the interpretations we find in Book B in C. Important capacity is to be able to cite Book A elements from Book B/C.

Answers to @dponti

Finally, should we create a borehole core as a specialization? If we classify a borehole core as a MaterialSample, I would say no

If we stick to the definition it seems to me a borehole core is a kind of MaterialSample (a real object with a shape). Its description can be provided through Observation as defined by #19. While its sampling location can be provided by the samplingActivity.

In describing the core in the lab, we observe a 10 cm thick sand layer located at 20 cm in the core as measured from its top. This location for the observation could be recorded in the sample's relative reference system (20-30 cm) where the sample's location is not provided to the lab technician), or in the borehole's LSRS (5.2 - 5.3 m), depending on the circumstance. The SamplingActivity's SampleProduced object, which is associated with the MaterialSample, contains the information needed to place the observation at the feature of interest in geographic space.

+1 to have the borehole and the core introducing their LinearReferenceSystem. The observations made on them will be based on this LRS

If anything we maybe should specialize MaterialSample according to its state (eg, soil, rock, gas, fluid)

Indeed. We envisaged to have distinctions between a GeologySpecimen and a FluidSpecimen in MINnD / IFC-Tunnel.

[mbeaufils]

OGC GeoscienceDWG had a Borehole IE in 2019. The report can be found here : https://github.com/opengeospatial/boreholeie/tree/master/er This document describes a conceptual model, logical model, and GML/XML encoding schema for the exchange of borehole related data and especially all the elements that are positioned along a borehole trajectory.

This could concern both the Borehole #10 and Borehole Core object.

[dponti]

Thanks for the reference, Mikhail, I'll read it and comment. Your responses above I think reflect that we are thinking along the same lines.

Again, I understand the O&M concept that essentially all observations occur on a "sample", whether that be a material sample (with or without location, but with spatially invariant properties - at least as defined), or a spatial sample, like a borehole where properties vary spatially within it. However, a core spans both definitions and I find that a bit problematic. I'd prefer we consider a core as a material sample, and I believe we can easily do that using the O&M model with appropriate restrictions/extensions of the basic O&M objects to make that work, with observations on the core located via the core's LRS and the core itself located within the borehole's LRS.

[neilchadwick-dg]

I've never thought of core as a sample before, but conceptually it seems fair enough.

We would need to ensure that core run information is captured (depth from/to). I guess that core recovery data just ends up as observations on that 'sample'. Also, some tests (eg. point load) may be done more or less directly on the core (sample).

We will have a situation where samples are taken from the core (sample) for lab testing (becoming more common in stiff clays here in the UK), but I think we already have that concept covered so can't see that being a major problem.

[Didymograptus]

Also, some tests (eg. point load) may be done more or less directly on the core (sample).

Not strictly speaking. The tests would be carried out on a specimen taken from a sub-sample of the core sample.

AGS: SAMP_TYPE 'C' = Core sample ;-)

[dponti]

@neilchadwick-dg -

This is how DIGGS handles this and I think it works especially well for cores: The SamplingActivity is the place where core run/recovery would be recorded eg., the from-to depths of the core run - the activity that produces the sample. The core itself would also need to have it's location in the borehole defined - if 100% recovery the from-to of the sample would be the same as that of the run, whereas partial recovery would require some estimation as to where within the core run interval the sample is located.

If you were to then break the sample up at some later time, but not do that specifically to retrieve a specimen as part of a test procedure, that would be another SamplingActivity (a subsampling activity instead of a collection activity), producing the additional samples. Extracting material as part of a test procedure would then be a Specimen - referencing back to the source sample; the Specimen would also carry information about its preparation for the test as well as properties on the specifmen condition pre and post test (eg if there's a noted volume change after testing, etc.

The point load observation would occur at a specific location (can use the reference system of the core or the borehole), and would reference the core sample the test was performed on. Everything else would be the same as for every other observation.