GMMensemble

Source code repository for:

"Unsupervised classification of the northwestern European seas based on satellite altimetry data" by Lea Poropat , Dan(i) Jones , Simon D. A. Thomas , and Céline Heuzé

Using an ensemble of Gaussian Mixture Models (GMMs) with altimetry data to find the regions of similar interannual to decadal sea level variability in the seas of northwestern Europe.

The preprint can be cited as:

@Article{egusphere-2023-1468,
AUTHOR = {Poropat, L. and Jones, D. and Thomas, S. D. A. and Heuz\'e, C.},
TITLE = {Unsupervised classification of the Northwestern European seas based on satellite altimetry data},
JOURNAL = {EGUsphere},
VOLUME = {2023},
YEAR = {2023},
PAGES = {1--20},
URL = {https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1468/},
DOI = {10.5194/egusphere-2023-1468}
}

To create the python environment:

conda env create -f environment.yml -n gmm

conda activate gmm

List of files

s01_Preprocessing.ipnyb - prepares the satellite altimetry data with already selected region of interest
s02_PCA.ipnyb - principal component analysis to separate the sea level signal into spatial (empirical orthogonal function maps) and temporal (principal component time series) component
s03_GMMnumberOfClasses.ipynb - calculates silhouette score and Bayesian Information Criterion for a span of class numbers to find the best one
s04_GMMensemble.ipynb - main script; applies an ensemble of Gaussian Mixture Models on stacked EOF maps
plot_classification.m - Matlab script used to plot the results of one ensemble (classes and likelihood)