Field of study / Research area
443 - Earth science

Level
Master + PhD

Name of the Programme
Machine-Learning-Based Analysis, Mechanisms, and Predictability of Large-Scale Oscillations in the Mediterranean Sea (ML-OCEAN)

Website (URL) of the Programme
https://more.unist.hr/aime-ip-unist-52/223

Name of the offer / activity
Dimensionality Reduction Methods for Detecting Mediterranean Sea Variability Regimes

Brief description of the activities to be carried out
Short description:
The traineeship focuses on the application of dimensionality reduction methods to identify dominant patterns and variability regimes in Mediterranean Sea datasets. The student will analyse oceanographic variables such as temperature, salinity, sea surface height, currents, and/or chlorophyll-a using linear and nonlinear methods.
The work will compare classical approaches, such as PCA/EOF analysis, with machine-learning-based approaches such as autoencoders or clustering methods. The aim is to explore whether large-scale Mediterranean variability can be represented by a limited number of dominant modes and whether regional processes, including Adriatic–Ionian variability, are embedded within broader basin-scale dynamics.
Main activities:
Prepare selected Mediterranean Sea datasets for analysis.
Apply PCA/EOF and selected machine-learning methods for dimensionality reduction.
Identify dominant spatial and temporal variability patterns.
Compare results obtained from different methods.
Interpret the identified patterns in relation to Mediterranean and Adriatic–Ionian oceanographic processes.
Expected outputs:
A comparative analysis of dimensionality reduction methods.
Figures showing dominant spatial patterns and temporal variability.
A short report describing the methods, results, and physical interpretation.
Reproducible code suitable for further project use.
Required background:
The traineeship is suitable for students in oceanography, environmental sciences, physics, geosciences, data science, or related fields. Basic programming skills in Python or R are expected. Previous knowledge of oceanographic data analysis, machine learning, or climate variability is an advantage.

Working language(s)
English

Host University
University of Split

University Department / Unit
Faculty of Marine Sciences

Duration in months
2

Comments on duration
Flexible

Starting
April 2026

Ending
December 2026

Academic responsible for the offer
Frano Matić
frano.matic@more.unist.hr