Polymetallic nodule occurrence model

Reference

This github repository is a supplement to the paper: Dutkiewicz, A., Judge, A. and Müller, R.D., 2020, Environmental predictors of deep-sea polymetallic nodule occurrence in the global ocean, Geology, 48, p. XXX–XXX, https://doi.org/10.1130/G46836.1

Installing dependencies

To run the code it is recommended to setup a local Python3.7 conda environment on a macOS or Ubuntu Linux machine

$ conda create -n polymetallic-nodules python=3.7 numpy scipy pandas xlrd scikit-learn netcdf4 xarray jupyter matplotlib
$ conda activate polymetallic-nodules

The project-specific utilities may then be installed

$ pip install -e utils

Running the code

All data files are required to be present in the data directory, with the oceanic variable grids under data/grids and the files containing the nodule and control lat-lon points under data/csv. Make sure that you have installed "LaTex". Running the notebook will produce the following materials:

• Nearest neighbour interpolated variable grids (grids.nc)
• Nodule and control point data files with variable values interpolated at exploration points (nodules.csv, control.csv)
• Grid data interpolated on a Fibonnacci lattice (lattice.csv)
• Kolmogorov-Smirnov statistics for comparison of oceanic variable samples at exploration and lattice points (ks_stats.csv)
• Mutual information estimate between variable and nodule occurrence probability grids (mi.csv)
• Mutual information bar graph (mi.pdf)
• Variable dependence plots (variable_dependence.pdf)
• Nodule occurrence probability grids (probability_grids.nc)

Expected runtime for the full notebook on a standard laptop is approximately 5hrs. Most of the computation is concerned with estimating the mutual information between the nodule occurrence probability and the variable grids.