3D Geological modelling from Boreholes - Gempy and Python coding question?

[kaedmorskad-ship-it]

Hello there,

I am new to Gempy and i have a file that contains the Name, coordinates, and the Lithology and their distribution with depth. However, I find it difficult to use Gempy because always an error pops up when i try to use it. I have a structure of my file. I am sharing the file with you. I hope someone could give me an idea on how to proceed.

Bohrungen_UniqueXY.xlsx

[javoha]

Hello @kaedmorskad-ship-it,

thanks for your question. So Gempy requires specific naming for the columns in your file. In addition Gempy also requires orientation data (which you can just manually set if your geometries are not too complex).

For a proper introduction I would suggest to take a look at our rather new video tutorial series which explains briefly the underlying concepts and the most important steps to create a model.

Let me know if you have any follow-up questions afterwards.

Cheers,
Jan

[kaedmorskad-ship-it]

Hello Javoha, thank you for your input. I followed your instructions and watched the videos. However, in my case the interpolation was not very succesfull. I believe because it had to many boreholes close to each other and also lots of layers that I ordered based on the geological time periods. I am sharing my work with you if you could take a look and give me some suggestions. I would really appreciate that.
Bohrungen_ZeitPerioden.zip

[NilsChudalla]

Hi @kaedmorskad-ship-it,
please make sure to share relative paths in your example next time, because I had to run through your entire script changing paths :D A slighlty more slimmed down version would be helpful aswell.

Your main issue seems to be that you
a) have way too many surface you want to compute (69! this might be the biggest structural frame I have anyone compute so far :D)
b) use points from your borehole without the required preprocessing. I see multiple boreholes, where you have formations logged along the borehole as input. In GemPy however, you will want to use only the bottom point of the formation as input, because the interpolation tries to generate a continuous scalar field going through all surface points. Filter for something like the maximum depth of each formation occuring in a borehole or something like that (unless you have fault offset to consider).

So, reduce the amount of formations (start with one, maybe the one you have the most data on) and this should already clearify what I meant with b). Only once you have these two things sorted, think about serieses in GemPy, aka surfaces that are interpolated with the same scalarfield and behave somewhat parallel.

Kind regards
Nils

[kaedmorskad-ship-it]

Hello NilsChudalla,
I really appreciate your feedback. I worked on refining the input data and now I have only two main geological time periods that i wanted to use. I also included the Lithology for each layer, however, if I am afraid that i reduce the number of layers inside a geological time period, then my results will not represent the real geological situation. I did not try the new file yet, but every time i tried I always get scattered layers. I know that some layers are localized and i do not know how to manipulate the data so gempy know they are only localized lithologies. Can you give a look at my refined file, and see if I fulfilled the conditions for moving on. Maybe help me on making the structure as required by Gempy.

I realized it was hard for me to export the results, which I will need them later for groundwater modelling. I am looking forward to your response.
Cheers,
Kaed

Bohrungen_MergedLitho_Over30m_Thinned1000m.xlsx

[NilsChudalla]

Hi @kaedmorskad-ship-it
given that you have a very complex dataset, your struggle is understandable. I must also confess, that I oversaw the timeseries - that changes a few things. Just to verify: Within these, you have lithological observations, which sometimes are repeating (Take Hy Farh -/1958: within qw you have the SpecifiedLitho sequence "S, Mg, Mg, S"). These repetitions were enumerated by you, which may be the cause for the exploding amount of surfaces. Quaternary lithologies often have strong variations laterally due to channeling or what-not. GemPy however is designed to model continuous surfaces and might have a hard time here.

If I was confronted with the task of building a model of the lithologies in this study area, I would personally start by modeling the time periods in GemPy. One surface for the bottom of the Weichselzeit, one surface for Saale-Komplex, always picking the lowest bottom point of each (you have to check your data, if the last point in the borehole is actually the bottom of the surface or if the drilling simply stopped before reaching the bottom).

Next, extract the volumes - here I would always make sure to grab the deepest bottom point of each timeperiod per borehole. Your model goal defines the requirement for the volumes: can you deal with a structured grid and have a chunky model or do you need a smooth mesh?

After computing the geometry, I would use other geostats packages to apply indicator kriging or something similar within the computed volumes of each timeperiod. This way, you can simply maintain the SpecifiedLitho property without enumerating it. You can simplify the model even further by omitting the GemPy modelling step. This depends a bit on your aim for the model. This is the package that I would have a look at first.

Just make sure, to check the interpolation which is used. Some algorithms interpolate the transition from a > b > c > d like a continuous variable, meaning that you will have problems interpolating a > d or a > c without inserting something, that does not really make sense. My intuition would be to use some algorithm that handles the lithologies as discrete, independent variables.

I hope this helps.

Kind regards
Nils