Cetacean-Feeding-Modelling

Table of Contents:

• Description
• Project Structure
• Requirements
• Setup
• Usage
• Contact
• License

Description [to ToC]

Cetacean Feeding Modelling (CFMs) is a Machine Learning-based framework developed to predict cetacean feeding activity in relation to environmental variables in the Central-eastern Mediterranean Sea. CFMs combine behavioral data ('Feeding' vs 'Other' behaviors) with 20 environmental predictors derived from sources such as Copernicus Marine Service (CMS) and EMODnet-bathymetry. By integrating Random Forest and RUSBoost classifier algorithms, the CFMs capture species-specific feeding patterns for three target cetacean species - striped dolphin, common bottlenose dolphin, and Risso's dolphin - in the Gulf of Taranto, our study area, providing a tool for marine conservation and management through predictive feeding maps and offering insights to improve knowledge on feeding habitat characteristics.

Project Structure [to ToC]

The project structure is organized as follows:

• data folder contains two subfolders:

  • raw folder contains two subfolders with the raw dataset:

    • Dataset: contains the Excel file of the raw dataset used to generate the subset used to build the models.
    • Extrapolation: contains the Excel file used to predict feeding habitats of Risso's dolphin for all the Gulf of Taranto, using the bio-chemical model.

  • processed folder contains three subfolders, one for each cetacean species studied:

    • Dataset_grampus: contains five Excel files with the processed datasets related to the Risso's dolphin species, used to run five model with a different variables characterization.
    • Dataset_stenella: contains five Excel files with the processed datasets related to the striped dolphin species,used to run five model with a different variables characterization.
    • Dataset_tursiops: contains five Excel files with the processed datasets related to the common bottlenose dolphin species,used to run five model with a different variables characterization.

• src folder contains the source code files and subfolders:

  • lib folder contains libraries for statistical analysis, pre-processing machine learning, and utility functions.
  • models folder contains the main script for running the ML models.
  • t-test folder contains the main script for running the t-test analysis.
  • extrapolation folder contains the main script for running the best ML models for a target species predicting on a new area.

Requirements [to ToC]

• MATLAB Version 9.14 (R2023a) (https://it.mathworks.com/products/matlab.html)
• Statistics and Machine Learning Toolbox Version 12.5 (R2023a) (https://it.mathworks.com/products/statistics.html)
• Parallel Computing Toolbox Version 7.8 (R2023a) (https://it.mathworks.com/products/parallel-computing.html)

Setup [to ToC]

To set up the project, follow these steps:

1. Clone the repository:

   git clone https://github.com/che7carla/Cetacean-feeding-modelling.git
   

2. Navigate to the project directory:

   cd Cetacean-feeding-modelling
   

Usage [to ToC]

To run the experiment follow these steps:

1. Run the script to train machine learning models:

 \src\models\main_models.m

2. Run the script to do the t-test analysis of the environmental variables for Feeding vs Other behaviors:

\src\t-test\ttest.m

3. Run the script to predict the Risso's dolphin feeding habitats for all the Gulf of Taranto using the M_bio model already trained, on the three summer months of 2024:

\src\extrapolation\Extrapolation_Gulf_of_Taranto.m

Contact [to ToC]

For any questions or inquiries, please contact Carla Cherubini or Rosalia Maglietta

Cite this

Cherubini, C., Cipriano, G., Saccotelli, L., Dimauro, G., Coppini, G., Carlucci, R., Fanizza, C. and Maglietta, R., 2025. Cetacean feeding modelling using machine learning: A case study of the Central-Eastern Mediterranean Sea. Ecological Informatics, p.103066.

License [to ToC]

This project is licensed under the Apache License 2.0.