Prof. Dr. Vladimir M. Cvetković, Full Professor of Disaster Risk Management at the University of Belgrade – Faculty of Security Studies, serves as a Management Committee Member representing the Republic of Serbia in COST Action CA25171 – ODIN-AI: Ocean Disaster Intelligent Network using AI. His nomination was officially accepted by the COST Association on 21 July 2026.
The ODIN-AI COST Action establishes a pan-European, transdisciplinary research and innovation network designed to accelerate the responsible integration of artificial intelligence and machine learning into operational coastal oceanography, ocean-disaster forecasting, early warning and coastal disaster risk management. The Action was approved on 19 May 2026 under Memorandum of Understanding No. 085/26 and will run from 20 October 2026 to 19 October 2030.
The Action responds to major scientific and operational challenges created by the rapidly growing volume of oceanographic data, increasing processing requirements and the complex interactions among multiple hazards in near-shore environments. These challenges continue to limit the accuracy, speed and operational usefulness of forecasting and early warning systems for marine heatwaves, harmful algal blooms, storm surges, coastal flooding and other interconnected ocean and coastal hazards. Climate change is further increasing the frequency, intensity and societal consequences of extreme ocean-related events.
A central challenge addressed by ODIN-AI is the persistent gap between cutting-edge artificial intelligence research and its practical application in operational oceanography. Although advanced AI and machine-learning models are increasingly capable of processing large and complex datasets, their integration into real-time forecasting, early warning and emergency decision-support systems remains limited. ODIN-AI seeks to bridge this research-to-operation gap by connecting researchers, operational forecasting centres, civil protection and disaster-management institutions, technology specialists, policymakers and other relevant stakeholders across Europe.
The principal scientific objective of ODIN-AI is to develop physically consistent, interpretable and operationally reliable AI systems capable of improving the prediction and assessment of ocean and coastal disasters. Particular attention is devoted to explainable artificial intelligence, model transparency, interoperability, uncertainty communication and the translation of AI-generated outputs into actionable information for forecasters, emergency managers, civil protection authorities and decision-makers.
The Action will contribute to the development of scalable AI prototypes, interoperable frameworks for real-time operational systems and standardised research-to-operation protocols. These outputs are expected to facilitate the transfer of scientific advances into practical forecasting and early warning applications. Regional pilots and demonstrators will be used to test, evaluate and validate proposed methods in cooperation with relevant institutions and end users.
From a societal perspective, ODIN-AI aims to support earlier, more accurate and more reliable warnings, thereby strengthening anticipatory action and improving the preparedness of institutions and exposed communities. More effective warning systems can contribute to reducing human casualties, economic losses, infrastructure disruption and environmental damage caused by coastal and ocean-related disasters. The Action will also examine how institutional trust, public risk perception, warning communication and organisational preparedness influence the acceptance and use of AI-supported systems.
Ethical and responsible use of artificial intelligence constitutes another central component of the Action. ODIN-AI will develop guidelines for responsible AI in operational oceanography, addressing transparency, explainability, accountability, data quality, uncertainty, institutional responsibility, human oversight and public trust. These guidelines are intended to support the safe and socially responsible introduction of AI technologies into operational forecasting and disaster-management systems.
The research coordination objectives of ODIN-AI include the development and evaluation of physically consistent and explainable AI models for multi-hazard prediction; the integration of heterogeneous oceanographic and environmental datasets; the establishment of interoperable operational frameworks; the development of standardised procedures for transferring research results into operational systems; and the validation of AI-supported solutions through regional pilots and demonstrators.
The capacity-building objectives include transdisciplinary training for young researchers, innovators and operational practitioners; equitable transfer of knowledge and technical skills; stronger participation and leadership of researchers from Inclusiveness Target Countries; support for early-career researchers; and closer cooperation among academia, operational centres, public authorities and policy institutions.
Training schools, workshops, webinars, short-term scientific missions, conferences, working-group meetings, pilot activities and joint publications will be used to strengthen the scientific and professional capacities of participating institutions. The Action will also promote open science, FAIR data practices, methodological harmonisation, interdisciplinary mobility and the development of sustainable international research partnerships.
The scientific and networking activities of COST Action CA25171 are organised through six complementary Working Groups:
Working Group 1 – Data Streams, FAIR Practices, and Observational Readiness for AI
WG1 addresses the availability, quality, integration and standardisation of observational data required for AI-supported operational oceanography. Its activities include FAIR data principles, data interoperability, metadata standards, quality control, observational readiness and access to suitable datasets for model development and validation.
Working Group 2 – Advanced AI Models for Ocean Disaster Prediction and Risk Assessment
WG2 focuses on the development of advanced artificial intelligence and machine-learning models for predicting marine and coastal hazards. It addresses physically informed AI, explainable AI, multi-hazard risk assessment, uncertainty analysis, vulnerability and resilience indicators, model benchmarking and the translation of model outputs into decision-relevant risk information.
Working Group 3 – Operational Integration, Real-Time Systems, and HPC/Edge Deployment
WG3 concentrates on the integration of AI solutions into operational environments. Its work covers real-time forecasting systems, high-performance computing, edge computing, scalable digital infrastructure, interoperability, model deployment, processing speed and the technical reliability of operational applications.
Working Group 4 – Societal Impact, Risk Communication, and Policy Uptake
WG4 examines the social, institutional and policy dimensions of AI-supported warning systems. Its activities include risk perception, trust in AI-generated warnings, crisis and warning communication, institutional coordination, public engagement, policy development, stakeholder acceptance and the practical adoption of research results.
Working Group 5 – Regional Pilots, Demonstrators, and Training for Disaster Response
WG5 develops regional pilot activities, demonstrators, scenarios, simulation exercises and training programmes. It connects scientific solutions with civil protection authorities, emergency-management organisations, operational practitioners, local communities and other stakeholders involved in disaster preparedness and response.
Working Group 6 – Ethics, Trust, and Responsible AI
WG6 addresses the ethical, legal, institutional and societal requirements for responsible artificial intelligence. Its work includes transparency, explainability, accountability, bias, data integrity, human oversight, communication of uncertainty, institutional trust and the formulation of practical guidelines for responsible AI in operational oceanography and disaster management.
As a Management Committee Member representing Serbia, Prof. Dr. Vladimir M. Cvetković participates in the strategic coordination, governance and implementation of COST Action CA25171. His role includes contributing to Management Committee decision-making, supporting the implementation of the Action’s scientific and capacity-building objectives, promoting cooperation among participating countries and facilitating the involvement of relevant researchers, institutions and practitioners from Serbia and the Western Balkans.
His scientific contribution is based on extensive expertise in disaster risk management, disaster preparedness, early warning systems, community resilience, vulnerability assessment, crisis management, civil protection, risk perception, risk communication, crisis communication, multi-hazard risk assessment, climate adaptation and critical infrastructure protection.
His planned contribution is particularly relevant to WG2, WG4, WG5 and WG6. Within these areas, he will contribute to the development of disaster-risk assessment methodologies, vulnerability and resilience indicators, research on public and institutional trust in AI-generated warnings, communication of uncertainty, institutional coordination, policy uptake, regional pilot activities, simulation exercises, professional training and ethical frameworks for AI-supported disaster management.
A particular priority will be to ensure that technically advanced AI outputs are transformed into clear, understandable and actionable information for decision-makers, civil protection authorities, emergency-management professionals and exposed communities. This includes examining not only whether an AI system can accurately predict a hazard, but also whether institutions and populations understand, trust and act upon the warnings it generates.
The contribution from Serbia will also focus on comparative analysis of national and regional early warning and disaster-management systems; surveys, interviews, focus groups and case studies involving citizens, practitioners and policymakers; development of composite indicators of vulnerability, resilience, preparedness and trust; scenario analysis; expert workshops; policy laboratories; simulation exercises; and evaluation of regional pilot solutions.
The University of Belgrade – Faculty of Security Studies serves as the home institution of the Serbian research team participating in ODIN-AI.
The team is led by Prof. Dr. Vladimir M. Cvetković, Full Professor of Disaster Risk Management and Serbia’s Management Committee representative.
The national research team also includes Prof. Dr. Ana Kovačević, Full Professor of Informatics and Cybersecurity at the University of Belgrade – Faculty of Security Studies. Her expertise encompasses generative artificial intelligence, machine learning, cybersecurity, databases, information visualisation, data integrity, misinformation detection and cyber risks to critical infrastructure.
The complementary expertise of the Serbian team connects disaster risk management, early warning, community resilience and crisis communication with artificial intelligence, machine learning, cybersecurity, data processing, information visualisation and critical infrastructure protection. This interdisciplinary combination supports the development, evaluation and responsible operational application of AI solutions in disaster forecasting, warning communication and emergency decision-making.
The Serbian team plans to contribute to joint scientific publications, comparative research, policy recommendations, open educational resources, European project proposals, regional pilots, professional workshops, webinars, training programmes and simulation exercises.
Expected outputs include at least two joint scientific publications, a comparative report or policy brief addressing the social and institutional dimensions of AI-supported early warning, participation in a regional pilot or demonstrator, organisation or co-organisation of a workshop or training activity in Serbia, mentoring of young researchers and preparation of at least one collaborative proposal for a European or international research call.
An important objective is to connect the ODIN-AI network with civil protection authorities, emergency-management institutions, researchers, universities, technology specialists and practitioners from Serbia and the Western Balkans. This regional engagement will facilitate knowledge transfer, improve institutional preparedness and increase participation in European research and innovation networks.
Through this participation, the University of Belgrade – Faculty of Security Studies and the Serbian research community will contribute to the development of a more resilient, intelligent, transparent and socially responsible European system for managing coastal and ocean-related disaster risks.
Action number: CA25171
Acronym: ODIN-AI
Full title: Ocean Disaster Intelligent Network using AI
Memorandum of Understanding: 085/26
Approval date: 19 May 2026
Duration: 20 October 2026–19 October 2030
Role: Management Committee Member representing Serbia
Serbian team institution: University of Belgrade – Faculty of Security Studies
Main Proposer: Dr Jian Su
Primary fields: operational oceanography, disaster forecasting, artificial intelligence, machine learning, early warning, multi-hazard risk assessment and responsible AI.