Adding geophysical constraints

[sandhya0892]

I have results from geophysical inversion, i.e. density and susceptibility. How do I add these constraints to my geo_model?

[NilsChudalla]

Hi @sandhya0892 ,

we are not using geophysical inversion in this way as a model input to build models. Models are built in Gempy using surface points (points on the interface between to rock units) and orientation points. You would have to interpret this interface in your inversion result. As "geophysical inversion" is a very broad term and there are many algorithms, there is no 100 % robust and general answer to this question. Maybe you describe your aim and algorithm a bit more and maybe we can give you some directions.

[sandhya0892]

Thank you for your response @NilsChudalla, my aim is to create a geological model of a region. I have provided the surface points and the orientations. I would like to know if I can provide the results from the joint gravity and magnetic inversion, which I carried out in SimPEG to GemPy. I understand that I would have to map the densities to the formation before I use it in the code. I tried that as well ( a very rough approximation). My code is as below:

import pandas as pd
import gempy as gp
import gempy_viewer as gpv

geo_model: gp.data.GeoModel = gp.create_geomodel(
project_name='Model',
extent=[0, 2000, 0, 2000, 0, 1300],
refinement=6,
importer_helper=gp.data.ImporterHelper(
path_to_orientations="orientations_group.csv",
path_to_surface_points="surface_group.csv"
)
)

df_density = pd.read_csv('inversion_density_model.csv')
df_density['formation'] = df_density.density.apply(density_to_formation)

gp.add_surface_points(
geo_model,
x=df_density.X.values,
y=df_density.Y.values,
z=df_density.Z.values,
elements_names=df_density.formation.values
)

gp.map_stack_to_surfaces(
gempy_model=geo_model,
mapping_object={
"first_series": ("fluvial"),
"group1": ("group1"),
"group2":("group2"),
"group3":("group3"),
"final_series":("group4","group5","group6")
}
)

sol = gp.compute_model(geo_model, compute_mesh = True)

Since this is only a crude approximation, I was wondering whether there is an alternative way to input gravity data. When I print geo_model, I notice that there is a key called geophysical_input, which made me curious about whether gravity data can be incorporated through that instead.

Thank you for your response again.

[NilsChudalla]

Hi @sandhya0892

that sounds like a cool workflow. There was a research project, developers of gempy were in involved in, where geophysical_input was used to compute forward gravity responses here. Maybe you will find some directions there (I was not involved).

As I mentioned before, you are not directly entering properties to compute the structural geological model. I do, however, have another idea, how to utilize the result:

• Compute the gradient in x, y, and z direction of the gravity. Make sure to flip vectors to point in the general "younging-upwards" direction (very important!). use these gradients (or only in areas of interest) as orientation input. Compute the midpoints between the grid as locations for those gradients (edit: depending on the workflow, you might not need to compute the midpoints, only grab the grid centers directly, for example when using np.gradient). Orientations are a bit more "soft" at constraining geological models in gempy, but surfaces will follow this directional trend.
• Use the "lith_block" to construct a more sophisticated starting model in simpeg: I am no simpeg-poweruser, but having a structured grid should allow you to build a tensor model in simpeg. In gempy, you can get a integer coded, 3d array, where each integer corresponds with a modeled body. You find it, after computing, in geo_model.solutions.raw_arrays.lith_block. These ids can be mapped to properties with some dictionary and then used for forward computation and comparison with the measured ground truth in simpeg. I would stop eventually, once the misfit decrease is neglible.

I hope this is helpful for you case :)

Kind regards
Nils

[sandhya0892]

Hi @NilsChudalla,

Thank you for your timely response! Your suggestions are really helpful. I will definitely try it out.

Regards,

Sandhya