This repo is an example of how machine learning can be used to find Porphyry Copper deposits in the Andes. Modified from the workflows of Butterworth, N., D. Steinberg, R. D. Müller, S. Williams, A. S. Merdith, and S. Hardy (2016), Tectonic environments of South American porphyry copper magmatism through time revealed by spatiotemporal data mining, Tectonics, 35, 2847–2862, doi:10.1002/2016TC004289

Step 1: Download this repo and additional data files from here.

Unzip the download into the Week 10 folder. The Week 10 Folder should contain the following folders/files:

Muller_etal_2016_AREPS_Agegrids_v1.11 - Age grids used as a coregistration dataset

Muller_export - Output data from convergence.py used as a coregistration dataset

Muller_gplates - Plate model topologies and rotation files.

Muller_convergence - Empty folder to be filled with kinematic subStats files.

CopperDeposits - A shapefile of copper deposits to be used as a coregistration dataset

convergence.py - python script that calculates convergence rates for a particular plate model.

coregLoop.py - python script to take a shapefile and "coregister" the points with some other datasets

coregLoop.ipynb - same as above, but in notebook format.

Muller_copper_prob.ipynb - a python notebook that steps through the machine learning process.

Utils_coreg.py - Some additional tools and functions called in the main scripts.

Create and activate the required conda environment:

conda create -n py2GEOL -y python=2.7 scipy=1.2 scikit-learn=0.20 matplotlib=2.0 pyshp=1.2 numpy=1.15 jupyter=1.0 cartopy=0.17 pandas=0.24 notebook=5.7.4

conda activate py2GEOL

Install pygplates beta-revision-12 https://sourceforge.net/projects/gplates/files/pygplates/beta-revision-12/

Alternatively you can also use the Docker container if that is easier!

docker pull nbutter/pyforgeo

If using docker you can start a bash terminal:

sudo docker run  -it --rm -v`pwd`:/workspace pyg /bin/bash 
source activate py2GEOL 
MORE-COMMAND-GO-HERE

Or activate a Jupyter environment with:

sudo docker run  -p 8888:8888 -it --rm -v`pwd`:/workspace pyg /bin/bash -c "source activate py2GEOL && jupyter notebook --allow-root --ip=0.0.0.0 --no-browser"

Run the convergence.py script from a bash terminal shell. This script calculates convergence rates and other kinematics about the plate model along the subduction boundaries. To run this script you need to specify a rotation file, left- and right-handed subduction zone topology files, a closed plate polygon file, and a time. We want to calculate kinematic data for the entire history of the plate model, so we can do this in a small bash script:

cd Week10/

for age in {0..230}; do echo ${age};  python convergence.py Muller_gplates/Global_EarthByte_230-0Ma_GK07_AREPS.rot Muller_export/topology_subduction_boundaries_sL_${age}.00Ma.shp Muller_export/topology_subduction_boundaries_sR_${age}.00Ma.shp Muller_export/topology_platepolygons_${age}.00Ma.shp ${age}; done

This will create a bunch of files in called subStats_XX.csv that contain the statistics of the subduction zones (i.e. the kinematic data). See the comments in convergence.py to see exactly what is output. These will be dumped in the top-level directory, so let's move them to a folder:

mv subStats*.csv Muller_convergence/

The folder CopperDeposits contains a shapefile XYBer14_t2_ANDES.shp with a set of known copper deposits in the Andes. These are currently just points on a map with a formation age associated with them. Our ultimate goal is to identify what kind of tectonomagmatic environments are associated with the formation of these ore deposits, to begin to do this we first must link age-coded tectonomagmatic properties with our age-coded ore deposits. We can use the script coregLoop.py to achieve this!

Run the script coregLoop.py

Make sure all the paths to the various files are correct in the script. Then run with:

python coregLoop.py

This creates three datasets (in the python 'pickle' file format) which contain the most of the data needed to perform some machine learning on:

Muller_Bertrand_coregistered.pkl contains the full set of copper-deposits and their associated tectono-magmatic properites.

Muller_Bertrand_coregistered_random.pkl contains a psudeo-random set of non-deposits with known tectono-magmatic properties that can be used for training.

Muller_Bertrand_coregistered_sampleMuller0.pkl contains a set of points following the present day South American subduction zone and the corresponding tectono-magmatic properties of those points.

Open up Muller_copper_prob.ipynb in a Jupyter notebook and continue the interactive analysis in there.

This analysis can be done with any combination of rotation files and plate polygons. To create the initial set use these instructions.

• Load in plate polygons and rotation file to GPlates. Click on the export data button and select the following.
• 1. Export Resolved Topologies (CitcomS specific)
• 2. Shapefiles
• 3. ONLY select "Export all Plate Polygons to a single file"
• and "Export Plate Polygon segments to files based on feature type..."

This hack is required because in some cases GPlates (the older versions did not do this) will split the topologies into "points" and "lines", we just want have to rename the "lines" files if this occurs for "convergence.py" to read. This assumes you have exported 230 Myr of reconstructions. Use this bash snippet or similar to find and rename the shape files.

for age in {0..230}; do echo ${age};cp Muller_export/topology_subduction_boundaries_sL_${age}.00Ma/topology_subduction_boundaries_sL_${age}.00Ma_polyline.shx Muller_export/topology_subduction_boundaries_sL_${age}.00Ma.shx;
cp Muller_export/topology_subduction_boundaries_sL_${age}.00Ma/topology_subduction_boundaries_sL_${age}.00Ma_polyline.shp Muller_export/topology_subduction_boundaries_sL_${age}.00Ma.shp;
cp Muller_export/topology_subduction_boundaries_sL_${age}.00Ma/topology_subduction_boundaries_sL_${age}.00Ma_polyline.prj Muller_export/topology_subduction_boundaries_sL_${age}.00Ma.prj;
cp Muller_export/topology_subduction_boundaries_sL_${age}.00Ma/topology_subduction_boundaries_sL_${age}.00Ma_polyline.dbf Muller_export/topology_subduction_boundaries_sL_${age}.00Ma.dbf; done

for age in {0..230}; do echo ${age};cp Muller_export/topology_subduction_boundaries_sR_${age}.00Ma/topology_subduction_boundaries_sR_${age}.00Ma_polyline.shx Muller_export/topology_subduction_boundaries_sR_${age}.00Ma.shx;
cp Muller_export/topology_subduction_boundaries_sR_${age}.00Ma/topology_subduction_boundaries_sR_${age}.00Ma_polyline.shp Muller_export/topology_subduction_boundaries_sR_${age}.00Ma.shp;
cp Muller_export/topology_subduction_boundaries_sR_${age}.00Ma/topology_subduction_boundaries_sR_${age}.00Ma_polyline.prj Muller_export/topology_subduction_boundaries_sR_${age}.00Ma.prj;
cp Muller_export/topology_subduction_boundaries_sR_${age}.00Ma/topology_subduction_boundaries_sR_${age}.00Ma_polyline.dbf Muller_export/topology_subduction_boundaries_sR_${age}.00Ma.dbf; done

You are then ready to run convergence.py as in the instructions above.

The Ore Deposits require a their lat/lon location, along with an age, and to be used in this workflow need to be assigned a GPlates "Plate ID", the example set already has this. To assign a plate id to a new dataset you can use pygplates or do this within GPlates.