A spectral approach for the clustering of source rocks.

This repository contains the official code for the conference paper "A Novel Chemometric Approach for Oil & Source Rock Clustering" presented at the 31^st International Meeting on Organic Geochemistry (IMOG), 2023, and presented as a poster at the 13^th FORTH Scientific Retreat, 2022. The paper can be found here, and the poster can be downloaded here.

For this study we consider a set of 83 rock extracts from Western Greece’s potential source rocks. These samples are of multiple geological ages ranging from Triassic to Pliocene and are located in the so-called Ionian geotectonic zone of Greece.

Motivation

Oil-oil and oil-source geochemical correlation is a subject of prime importance to the hydrocarbon exploration community for decades. So far, methods such as k-means, principal component analysis (PCA) and hierarchical clustering (HCA) have been extensively used in chemometrics for such problems. In this work, we present a direct multiway spectral clustering approach, which is a graph based method that allows the automatic selection of the optimal number of clusters.

Normalized spectral clustering for source rocks

This repository implements a normalized spectral clustering algorithm for source rocks. We use the relative amounts of n-alkanes and iso-prenoids in each sample as input data. The resulting clusters are based on the information contained in the eigenvectors of the random-walk graph Laplacian matrix, and the optimal number of clusters is determined by measuring the modularity of the constructed graph.

Requirements

All the algorithms are implemented in MATLAB R2021b. The necessary paths are included in the script addpaths_Petrol.m

Usage

The main script Run_SR_Clustering.m runs the experiments on the input data, located at the /Input_Data folder in .xlsx format.

Data format: In the excel input file columns 8:9 contain the latitude and longitude of the samples. Columns 12:end contain the peak areas of the n-alkanes, representative of the relative quantity of the compounds within each sample.

To run clustering on the available input data execute the command

>> Run_SR_Clustering

You will be promted in the command line to select the name of the input dataset,

Select name of dataset at Input_Data/:

and to select the method that determines the number of clusters:

1: Selection based on columns 
2: Select num of clusters manually 
3: Number of clusters based on modularity of resulting clusters 
Select method to determine # of clusters:

Code Structure

The structure of the files in this repository is as follows:

├── Auxilliary                  # auxilliary functions
├── Documents                   # paper & poster
├── Figures                     # our results
├── Final_Clustering            # clustering the Laplacian eigenvectors
├── Input_Data                  # source rock samples in .xlxs format
├── Metrics                     # evaluation of clustering results
├── PHA                         # hierarch. clustering as per Lu et. al. 2013
├── Results                     # folder to save new results
├── addpaths_Petrol.m           # add the necessary paths
├── Run_SR_Clustering.m         # main file, source rock clustering

Further details are documented within the code.

Maintainers

• Dimosthenis Pasadakis (dimosthenis.pasadakis@usi.ch)

References

To better understand the background behind this work, we recommend reading the following papers:

1. Peters, K., Walters, C., & Moldowan, J. (2004). Geochemical correlation and chemometrics. In The Biomarker Guide (pp. 475-482). Cambridge: Cambridge University Press.
2. White, S., & Smyth, P. (2005). A Spectral Clustering Approach To Finding Communities in Graphs. SIAM International Conference on Data Mining.
3. Pasadakis, D., Alappat, C., Schenk, O., & Wellein, G. (2022). Multiway p-Spectral Graph Cuts on Grassmann Manifolds. Mach. Learn., 111(2), 791–829.

Citation

Please cite the following publication when using our software.

@article{mpp23,
   author = "Makri, V.I. and Pasadakis, D. and Pasadakis, N.",
   title = "A Novel Chemometric Approach for Oil \& Source Rock Clustering", 
   year = "2023",
   number = "1",
   pages = "1-2",
   doi = "https://doi.org/10.3997/2214-4609.202333183",
   url = "https://www.earthdoc.org/content/papers/10.3997/2214-4609.202333183",
   publisher = "European Association of Geoscientists & Engineers",
   issn = "2214-4609",
  }