Skip to content

Latest commit

 

History

History
44 lines (28 loc) · 2.89 KB

README.md

File metadata and controls

44 lines (28 loc) · 2.89 KB
Raw

NSG2021_QuickClayPublication

Overview

This repository is an area for storing and sharing supplementary materials related to a publication submitted to Near Surface Geophysics on using AEM, geotechnical data, and machine learning to model occurences of sensitive glaciomarine clay. The main things we share are example 3D models of our results discussed in the publication. See this link:

https://doi.org/10.1002/nsg.12166

Three versions corresponding to the 3 subfigures in Figure 10 of the publication:

  1. Where 10 boreholes are used as training, and only spatial attributes (i.e., horizontal and vertical coordinates) are used for making predictions
  2. Where 10 boreholes are used as training, and both spatial attributes and resistivity attributes are used for making predictions
  3. Where all boreholes available are used as training, and both spatial attributes and resistivity attributes are used for making predictions.

image

All scenes have a vertical exaggeration of 3x.

Scene contents

These scenes all contain 5 to 6 objects in them:

  1. Red isovolume of material where our algorithm predicts > 60% probability of brittle clay
  2. White/light grey isovolume of material where our algorithm predicts 40 to 60% probability of brittle clay
  3. Dark grey isovolume representing bedrock
  4. thick columns representing training boreholes, coloured by the interpreted sediment type
  5. thin columns representing boreholes unused in training, coloured by the interpreted sediment type
  6. Sediment data from the Norwegian Geological Survey (NGU) draped over terrain.

Legends

The sediment maps are coloured using this legend, similar to Figure 1 of our publication:

image

The boreholes are coloured according to the following colorscheme. Note that interpretations are the manual interpretations done by humans where available; otherwise, they are the automated interpretation performed by a random forest classifier trained on lab samples and the human interpretations. See section 2.4.2 of our publication for a description of that classifier.

image

How to open

The files herein can be opened in any GLTF viewer, but we recommend https://sandbox.babylonjs.com/. With this free online viewer, you can drag in and drop a GLTF file, navigate in 3D, toggle the visibility of layers, and adjust the transparency of each object's texture:

image

Sources

NGU, 2020. Norges Geologiske Undersøkelse. Løsmasser N250. https://kartkatalog.geonorge.no/metadata/loesmasser/3de4ddf6-d6b8-4398-8222-f5c47791a757