Development of machine and deep learning methods to predict marine extremes for engineering and navigation purposes
The PhD candidate is expected to develop an integrated framework that combines machine and deep learning methods with statistical and mathematical approaches to analyse and forecast marine extremes, including sea level, sea surface temperature, sea ice, and surface waves, from coastal to offshore regions. Understanding and predicting these extremes is increasingly important for engineering design, coastal management, and navigation under a changing climate.
The research is motivated by the unique availability of long-term and high-resolution marine datasets in the Baltic Sea region, including in situ observations, remote sensing products, and hydrodynamic model outputs. These datasets can be unified through a geoid-based vertical reference system, enabling consistent and accurate analysis of marine variables across the coastal–offshore areas.
Current machine and deep learning approaches for marine prediction often yield highly variable results depending on model design, input data, and temporal scale, affecting both short-term forecasts (hours to days) and long-term projections (years to decades). This project therefore emphasizes the integration of data-driven methods with mathematical constraints to improve accuracy, consistency, and predictive skill. The research shall identify patterns of marine extremes (location, frequency, and duration), improve short- and long-term forecasts, and investigate their links to large-scale climate variability as well as regional and local processes.
Background:
Climate change is projected to intensify marine extremes in terms of their frequency, intensity, and duration, posing increasing risks to coastal and marine ecosystems, navigation, and maritime engineering. As a result, understanding and predicting marine extremes such as sea level, surface waves, sea surface temperature, and sea ice has become more critical than ever.
Marine extremes are commonly studied using a range of data sources, including in situ met-ocean observations, hydrodynamic model outputs, and remote sensing techniques such as satellite observations, airborne laser scanning, and Global Navigation Satellite System. However, these data sources differ substantially in their spatial and temporal resolution, measurement techniques, and levels of consistency, and they are often weakly connected across spatial and temporal scales.
By combining these heterogeneous datasets considering their strengths and weakness it is possible to achieve a more comprehensive understanding of marine extremes, their driving mechanisms, from offshore regions to coastal zones. Recent advances in computing power and artificial intelligence, particularly machine and deep learning methods, provide new opportunities for integrating multi-source marine data and extracting complex spatio-temporal patterns.
Nevertheless, existing machine and deep learning approaches often yield highly variable results depending on model architecture, input data, and temporal scale, affecting both short-term forecasts (hours to days) and long-term projections (years to decades). This study therefore focuses on developing an integrated methodology that combines multi-source marine data with machine/deep learning techniques, statistical and mathematical approaches to accurately predict marine extremes, in both real-time and long-term contexts but also to understand the drivers and feedbacks behind large-scale climate variability (e.g., North Atlantic Oscillation) to regional and local marine processes
Responsibilities and (foreseen) tasks:
The PhD candidate will develop an integrated, data-driven framework that combines in situ observations, hydrodynamic model outputs, and remote sensing data to analyse and predict patterns of key marine extremes, including sea level, sea surface temperature, surface waves, and sea ice.
- harmonizing multi-source datasets (in situ, hydrodynamic models, and remote sensing) onto common spatial and temporal scales;
- applying statistical methods (e.g. extreme value analysis) to identify patterns of marine extremes and their spatial and temporal characteristics;
- developing machine and deep learning models (e.g. convolutional neural networks) that integrate all relevant data sources to predict marine extremes (short and long term) and quantify associated uncertainty;
- investigating the links between large-scale climate variability and regional-to-local marine processes, including the drivers and feedback mechanisms governing observed dynamics;
- validating model performance using independent data sources (e.g. field experiments and satellite observations) and conducting uncertainty and error analyses using established metrics such as root-mean-square error (RMSE), standard deviation, scatter index, and error budgets.
The candidate will be expected to explore and apply machine learning, signal processing, statistical, and computational techniques throughout the project and to assist in project-related field campaigns.
Applicants should fulfil the following requirements:
- Master's degree in Geomatics. Consideration will be given to applicants whose previous degrees are in appropriate related discipline, such as Earth Sciences, Computer/Data Science, Marine Engineering, Mathematics or Physics.
- A clear interest in the research topic are preferred
- Advanced computer literacy and programming skills (e.g. Python, C++, MATLAB etc.)
- Strong and demonstrable writing and analytical skills
- Skills in data analysis, mathematical and statistics (to be trained), machine learning methods (to be trained)
- Ability for independent research as part of a team, interest in the presentation and publication of scientific results
- Good command of the English language (speaking and writing).
The shortlisted candidate may be required to submit a research plan for the topic. The candidate can expand on the outlined research scope.
We offer:
- 4-year PhD position (full time) in the research group that has a sound portfolio of ongoing European and national research projects
- An environment to perform excellent research and publications
- Opportunities for training relevant technical and transferable skills aiming academic or industrial careers
- Opportunities for conference visits, research stays and networking with globally leading companies, universities and research centers in the field of research.
The candidate is obligated to participate and fulfil the requirements of Tallinn University of Technology PhD programme. Additional funds will be provided (and whence applicable the associated funding can be applied for) for research trainings, conferences and international mobility/stays abroad with durations of up to 3 months. The research group wishes to increase the number of women interested in Geomatics and Engineering. Qualified women are therefore also encouraged to apply. Do not hesitate to contact us for questions regarding the position. We look forward to receiving your application.
Supervisors:
Main supervisor: Tenured Full Professor Artu Ellmann: School of Engineering: Department of Civil Engineering and Architecture: Road Engineering and Geodesy Research Group
Co-Supervisor: Assistant Professor Nicole Camille Delpeche-Ellmann: School of Science: Department of Cybernetics: Laboratory of Wave Engineering
Tallinn University of Technology (TalTech) is an international scientific community with approximately 9,000 students and 2,000 employees; it is one of the largest universities in Estonia, the leading EU country in digitalisation. The university's strengths are broad multidisciplinary study/research interests, a modern research environment, and strong collaboration with international educational and research institutions. TalTech is aiming to be an organisation leading the way to a sustainable digital future.
The Department of Civil Engineering and Architecture is an interdisciplinary teaching and research center of Tallinn University of Technology that focuses on various actual research issues. The department consists of several research groups. The geodesy research group has mainly been focused on the national geodetic infrastructure related research (modelling the gravity field and geoid, precise height network, GNSS positioning). Currently the TalTech geodesy group is participating in the international collaboration for implementing the Baltic Sea Chart Datum 2000 by improving the marine geoid modeling. Most of the Baltic Sea countries have agreed to adopt this new marine geoid based vertical datum as initial for the nautical charts, hence also for the maritime and offshore industry. The recent European Space Agency sponsored international project Geodetic SAR for Baltic Height System Unification tested potential of the Interferometric Synthetic Aperture Radar (InSAR) of unification of across-ocean vertical datums. The group has relevant contractual research for industry, environmental and governmental agencies, for the details see A. Ellmann’s (head of the group) ETIS account, https://www.etis.ee/CV/Artu_Ellmann/eng.
For information about the admission process, please visit the PhD Admission homepage
Applications can be submitted from 18.06.2026 to 31.07.2026
