Core-Shell Clustering Algorithm

• Core-Shell Clustering algorithm is an unsupervised clustering algorithm for the analysis of spatiotemporal variations of SST imagery data;
• This algorithm uses .mat files as the SST grids;
• The final results produced by this algorithm are called core-shell clusters;
• In this repository there is the fully proposed pipeline, beginning from the original SST data, the preprocessing pipeline, the S-STSEC algorithm implementation, the Core-Shell clustering algorithm implementation and several auxiliary functions used to plot and get SST images.

Software setup

• If not installed, please install julia in your machine, it is used for the first stage of the preprocessing phase: https://julialang.org/downloads/;
• If not installed, please install the julia wrapper for the GMT tools: https://github.com/GenericMappingTools/GMT.jl.

Run the software

• Download the latest release and unzip the downloaded folder;
• Use an IDE to run the software. This is due to the need to use several paths as inputs to the program;
• To run the main algorithm, open the algorithmScript.py file.

Files explanation

• algorithmScript.py: Main script of the Core-Shell clustering algorithm, ready to run as testing;
• anomalousPattern.py: Implementation of the IAP algorithm;
• coastlineFinder.py: Gets coastline pixels and near the coast pixels for the STSEC algorithm;
• experimentUtils.py: Auxiliary functions;
• fileUtils: Auxiliary functions regarding file operations;
• matrixUtils: Functions regarding matrices;
• moovingAvgs: Functions regarding the moving averages filter;
• plotters: Functions to plot images;
• processingTools: More auxiliary functions;
• regionGrowing: Functions used for the SRG used in the STSEC base algorithm;
• sstsec: Implementation of the SSTSEC algorithm;
• stsec: Implementation of the basic STSEC algorithm;
• temperatureProcessingV3: Current version of the preprocessor pipeline of the Portuguese coast;
• upwellingClumpsBuilder: Functions used to use the IAP algorithm for the building of the Upwelling Ranges.

Sample data from the year 2007: MAT files

• A sample of the full dataset is present in the folder "PortugalImagesSample";
• Only the original images are in the format .png;
• The explanation is given with the assistance of the image "full_pipeline.png" that shows the developed pipeline, in order to give context to each folders' content;
• Inside a "mat" folder are the resulting computation results from each phase of the preprocessing pipeline:
  • preprocessing_phase_1: SST grids that suffered the removal of the North-South gradient using the GMT tools (Point 1 in the pipeline);
  • preprocessing_phase_2: SST grids that suffered the previous transformation and where the moving average filter was applied (Point 2 in the pipeline);
  • sst_instants: SST grids completely preprocessed which are called SST instants (Point 3 in the pipeline);
  • sst_instants_no_preprocessing: SST grids that suffered all the preprocessing stages but the first one, in order to be able to assess the SST instants with the original temperature ranges.
  • sst_instants/segmentations: SSTSEC segmentation results (Point 4 in the pipeline).