Abstract
Ambient gamma dose rate represents the integrated near-surface gamma radiation field resulting from contributions of terrestrial radionuclides and radon progeny, secondary cosmic radiation, and atmospheric radiation sources. Continuous monitoring of ambient gamma dose rate constitutes a fundamental component of radiological early-warning systems, as it provides a direct operational proxy for external radiation exposure to population. Time series of ambient gamma dose rate exhibit variability over a wide range of temporal scales, including short-term anomalies driven by meteorological processes, geophysical conditions, or anthropogenic influences. Accurate characterisation of these anomalies, and robust discrimination between natural drivers - such as soil–atmosphere exchange processes, boundary-layer dynamics, and hydrometeorological forcing - and potential anthropogenic contributions, is essential for enhancing early-warning capabilities and improving the detection of anomalous radioactive releases. A key challenge in this context is the scarcity of high-resolution, high-quality collocated meteorological observations required to support such analyses.This study presents a detailed characterization of anomalies in ambient gamma dose rate using comprehensive meteorological information and high-resolution (1-min) gamma dose-rate measurements from the Eastern North Atlantic (ENA) observatory, part of the U.S. Department of Energy’s Atmospheric Radiation Measurement (ARM) Program. Through the joint analysis of gamma radiation and a broad set of meteorological parameters - including precipitation, eddy covariance fluxes, aerosol properties, and lidar derived atmospheric structure - we identify and classify distinct types of short-term gamma radiation anomalies. These include precipitation-induced enhancements, quasi-daily anomalies associated with stable nocturnal boundary-layer conditions and near-surface radon accumulation, and anomalies linked to long-range transported dust events. This AI-ready, supervised dataset enables detailed investigation and modelling of ambient gamma dose-rate variability in the Azores and provides a transferable framework for training machine-learning algorithms to automatically classify gamma radiation anomalies at monitoring sites lacking comprehensive meteorological instrumentation. The present study is part of project NuClim (Nuclear observations to improve Climate research and GHG emission estimates). Project NuClim received funding from the EURATOM research and training program 2023-2025 under Grant Agreement No 101166515). The NuClim field campaign at the Eastern North Atlantic, Graciosa Island ARM Observatory is supported by the U.S. Department of Energy (DOE), Office of Science, through the ARM Program.

Abstract
Regional sea-level change in the semi-enclosed Baltic Sea is strongly influenced by atmospheric forcing and wind-driven redistribution of water masses, leading to significant spatial variability in absolute sea level trends across the different sub-basins. This study focusses on absolute sea level trends in the Baltic Sea using satellite gridded sea level anomalies (0.0625º) from the European Seas Gridded L4 product provided by the E.U. Copernicus Marine Service (https://doi.org/10.48670/moi-00141). The daily time series (from January 1993 to the end of December 2023) are first deseasoned by removing the average annual cycle at each point. Then robust linear trends are estimated at each grid point by computing median slopes. In contrast to ordinary least-squares slopes characterising linear trends in the mean, these median slopes are calculated by minimising the mean absolute deviation of a linear trend model from the observations instead of the mean quadratic deviation, which makes them more robust to outliers and sensitive to the typical tendency of changes rather than to large deviations. Uncertainty is computed assuming non-independence by the Huber sandwich robust estimator for the covariance matrix.The derived median slopes are in general higher than ordinary linear trends in the mean, except in the northern and easternmost areas of the Baltic. In the Bay of Bothnia ordinary linear trends and median trends are very similar, while in the eastern end of the Gulf of Finland median trends are similar or even slightly lower than ordinary linear trends. In the remaining areas, median trends are significantly larger than ordinary linear trends, the largest difference occurring in the Bothnian Sea. Coastal areas exhibit trends that differ from those in the adjacent basins. In the Gulf of Finland, median trends are higher than ordinary linear trends along the Finnish coast, whereas along the Roslagen coast (northern Stockholm Archipelago) the two slope estimates are in good agreement. Along the southern coastline of the Bothnian Sea, median sea-level trends reach the highest values, exceeding 6 mm/year.The present study is financed within the scope of the Recovery and Resilience Mechanism (MRR) of the European Union (EU), framed in the Next Generation EU, for the period 2021 - 2026, within project NewSpacePortugal, with reference 11.

Abstract
Investigating the complex coupling between the lithosphere, atmosphere, and ionosphere (LAI) requires a fundamental understanding of the physical forces governing tectonic processes and their electromagnetic manifestations. While various pre-seismic signals have been successfully identified, a persistent gap remains between the empirical observation of these phenomena and the fundamental physical laws that describe nature across all scales, from the subatomic realm to cosmic expansion. Exploring these interrelations presents significant challenges due to divergent scientific languages, specialized expertise, and unique terminologies across fields. The recently approved COST Action CA24101 "Testing Fundamental Physics with Seismology" (FuSe) aims to bridge this gap by exploring how seismic phenomena and earthquake precursors can serve as a "multi-messenger" window into fundamental interactions.At the heart of FuSe is the belief that imprints of non-standard physics, such as scalar fields or "fifth forces”, may be embedded within seismic and geomagnetic data. Conversely, theoretical insights from fundamental physics can refine our understanding of Earth’s interior by improving models of density and thermodynamic parameters like elasticity and bulk modulus. This refined modeling is essential for accurately interpreting the electromagnetic and gravitational perturbations that occur within the complex Earth-atmosphere-space system.To ensure these breakthroughs translate into practical advancements, FuSe focuses on several strategic pillars:-   Building a common language: developing a shared methodology to equip the next generation of scientists with cross-disciplinary skills.-   Interfacing communities: creating dynamic research groups that unite scientists from particle physics, gravity, planetary science, and seismology.-   Cross-disciplinary data integration: consolidating seismic data from the Earth and Moon with particle physics and geomagnetic data into AI-ready, FAIR-compliant streams.-   SME collaboration: partnering with small and medium-sized enterprises (SME) to advance sensor networks, AI algorithms, and real-time natural catastrophe alert systems.In this presentation, we outline the roadmap of the FuSe Action. We invite researchers with a background in electromagnetic precursors and LAI coupling to join this collaborative environment, where the synergy between geosciences and fundamental physics promises to drive innovative breakthroughs and unlock new paradigms in our comprehension of the Earth and the Universe.This abstract is based upon work from COST Action CA24101, Testing Fundamental Physics with Seismology (FuSe), supported by COST (European Cooperation in Science and Technology).

Abstract
Higher Education Institutions (HEIs) face significant challenges in managing and integrating diverse Information System (ISs) that support academic, administrative, and strategic operations. As digital transformation advances, the need for seamless interoperability and data-driven governance becomes increasingly crucial. This study provides a comprehensive analysis of the ISs Ecosystem (ISE) in HEIs, emphasizing the importance of system integration, Business Intelligence (BI) solutions, and Decision Support Systems (DSS) in fostering efficient, data-driven decision-making. By examining a real-world case study of the University of Tras-os-Montes and Alto Douro (UTAD), this research validates the role of BI in transforming fragmented information landscapes into cohesive digital environments. The findings demonstrate that successful BI adoption requires well-defined governance structures, seamless data flow, and alignment with institutional objectives. Additionally, the study underscores the strategic impact of interoperability, highlighting how institutions can enhance institutional intelligence, streamline decision-making processes, and improve operational efficiency through an integrated BI ecosystem. The insights contribute to ongoing discussions on digital transformation in higher education, offering a scalable framework for HEIs seeking to transition from isolated systems to an interoperable and intelligent data ecosystem. The paper also explores emerging trends such as AI-driven analytics and predictive modelling, outlining potential pathways for HEIs to further optimize their decision-support infrastructures.

Abstract
Federated Graph Neural Networks (FedGNNs) have emerged as a promising paradigm for decentralized graph learning across distributed data silos. However, the influence of underlying communication topologies on model accuracy and efficiency remains underexplored. This study presents a topology-aware benchmarking framework for federated GNNs, systematically evaluating the impact of network structure and aggregation strategy on performance and communication overhead. The proposed framework functions as a synthetic, communication-level digital twin that emulates Federated Learning interactions and topology-dependent dynamics under controlled conditions. Four learning schemes-Centralized, Local, FedAvg, and FedAvg-Fedadam-were assessed across three representative topologies: Barab & aacute;si-Albert (BA), Watts-Strogatz (WS), and Erd & odblac;s-R & eacute;nyi (ER). Results demonstrate that centralized training achieved the highest mean ROC-AUC (0.63), while FedAvg-Fedadam attained the best F1-score (0.038), balancing local adaptation and global convergence. Among topologies, BA and WS yielded higher average AUC values (approximately 0.57 and 0.56, respectively) than ER (approximately 0.39). Communication analysis revealed FedAvg as the most efficient strategy, requiring only approximately 3.8 x 105 bytes cumulatively. These findings highlight key trade-offs between accuracy, robustness, and communication efficiency in federated graph learning and provide empirical guidance for topology-aware optimization of distributed GNNs. While the experiments rely on representative synthetic topologies, the insights offer indicative relevance and potential applicability to Internet-of-Things (IoT), vehicular, and cyber-physical networks, where communication structure and bandwidth constraints critically influence collaborative intelligence. By modeling canonical connectivity patterns and releasing our code and data, the proposed benchmarking framework offers a reproducible basis for comparing emerging federated graph architectures under constrained communication conditions.

Abstract
Driving safety under low-visibility or distracted conditions remains a critical challenge for intelligent transportation systems. This paper presents Edge-VisionGuard, a lightweight framework that integrates signal processing and edge artificial intelligence to enhance real-time driver monitoring and hazard detection. The system fuses multi-modal sensor data-including visual, inertial, and illumination cues-to jointly estimate driver attention and environmental visibility. A hybrid temporal-spatial feature extractor (TS-FE) is introduced, combining convolutional and B-spline reconstruction filters to improve robustness against illumination changes and sensor noise. To enable deployment on resource-constrained automotive hardware, a structured pruning and quantization pipeline is proposed. Experiments on synthetic VR-based driving scenes demonstrate that the full-precision model achieves 89.6% driver-state accuracy (F1 = 0.893) and 100% visibility accuracy, with an average inference latency of 16.5 ms. After 60% parameter reduction and short fine-tuning, the pruned model preserves 87.1% accuracy (F1 = 0.866) and <3 ms latency overhead. These results confirm that Edge-VisionGuard maintains near-baseline performance under strict computational constraints, advancing the integration of computer vision and Edge AI for next-generation safe and reliable driving assistance systems.