A5072161944- Oceanographic and Atmospheric Processes
- Ocean Waves and Remote Sensing
- Meteorological Phenomena and Simulations
- Tropical and Extratropical Cyclones Research
- Geophysics and Gravity Measurements
- Climate variability and models
- Coastal and Marine Dynamics
- Flood Risk Assessment and Management
- Neural Networks and Applications
- Methane Hydrates and Related Phenomena
- Scientific Research and Discoveries
- Brain Tumor Detection and Classification
- Oil Spill Detection and Mitigation
- Marine and coastal ecosystems
- Atmospheric and Environmental Gas Dynamics
- Arctic and Antarctic ice dynamics
- Disaster Management and Resilience
- Robotic Path Planning Algorithms
- Coastal wetland ecosystem dynamics
- Anomaly Detection Techniques and Applications
- IoT and Edge/Fog Computing
- Machine Learning and Data Classification
- Marine Biology and Ecology Research
- Marine and fisheries research
- Hormonal Regulation and Hypertension
University of Southern Denmark
2025
Danish Meteorological Institute
2013-2024
Regional Hospital Central Jutland
1993
The Copernicus Marine Services State of Pacific Ocean analysis available data demonstrates that the ocean surrounding Islands is warmer, has higher heat content, with sea level rising at rates
Abstract. This paper reviews the state of art in storm surge forecasting and its particular application northern Adriatic Sea. The city Venice already depends on operational systems to warn population economy imminent flood threats, as well help protect extensive cultural heritage. will be more important future, with new mobile barriers called MOSE (MOdulo Sperimentale Elettromeccanico, Experimental Electromechanical Module) that completed by 2021. depend accurate control their operation. In...
Abstract. Wave conditions in the northern Baltic Proper during windstorm Erwin/Gudrun (January 2005) are analysed based on situ measurements three locations and output of operational wave models from German Weather Forecast Service, Danish Meteorological Institute Finnish Marine Research. The measured significant height reached 7.2 m 4.5 Gulf Finland. roughest conditions, estimated comparison forecast data, occurred remote sensors, off coasts Saaremaa Latvia where was about 9.5 m. Peak...
Coastal locations around the world face multifaceted challenge of rising sea levels, posing threats to coastal communities, particularly during storm surges and high wave conditions. We have examined dynamics sea-level fluctuations along Denmark's coastline, with focus on wave-induced phenomena, such as setup. analysed data from DMI's operational model spanning 2006 2022, well climate projections that extend until middle 21st century. conducted an analysis 2% exceedance significant height...
Training large neural network models requires extensive computational resources, often distributed across several nodes and accelerators. Recent findings suggest that it may be sufficient to only exchange the fast moving components of gradients, while accumulating momentum locally (Decoupled Momentum, or DeMo). However, when considering larger do not fit on a single accelerate, gradient information integration DeMo needs reconsidered. Here, we propose employing hybrid strategy, FlexDeMo,...
Large language models (LLMs) require immense resources for training and inference. Quantization, a technique that reduces the precision of model parameters, offers promising solution improving LLM efficiency sustainability. While post-training quantization methods typically achieve 4-8 bits per parameter, recent research suggests LLMs with 1.58 weight parameter from scratch can maintain accuracy while greatly reducing memory requirements energy consumption at inference time. Here, we...
This study was designed to compare the cardiac mass and geometry in white coat hypertensive patients established through prospective comparison of office blood pressure, daytime ambulatory echocardiographically determined left ventricular consecutive patients. We studied 143 from general practice an outpatient hypertension unit. The had newly diagnosed mild-to-moderate prior institution pharmacological anti-hypertensive therapy. All a diastolic pressure above 90 mm Hg; consistently elevated...
Coastal zones are among the most variable environments. As such, they require adaptive water management to ensure balance of economic and social interests with environmental concerns. High quality marine data hydrographic conditions e.g., sea level, temperature, salinity, currents needed provide a sound foundation for decision making process. Operational models sufficiently high forecasting resolution can be used further extension service toward coastal-estuary areas. The Limfjord is large...
Multi-model ensembles for sea surface temperature (SST), salinity (SSS), currents (SSC), and water transports have been developed the North Sea Baltic using outputs from several operational ocean forecasting models provided by different institutes. The individual differ in model code, resolution, boundary conditions, atmospheric forcing, data assimilation. are produced on a daily basis. Daily statistics calculated each parameter giving information about spread of forecasts with standard...
Denmark has a long, complex coastline, connecting the North Sea in west to semi-enclosed Baltic east, via Skagerrak-Kattegat Seas. Historical sea level records indicate that relative (RSL) been increasing along Danish coast, south of Skagerrak, following global mean (GMSL) rise. In central Seas, RSL rise practically absent, due GMSL being off-set by Fennoscandian post-glacial land-uplift. The new IPCC Special Report on Ocean and Cryosphere Changing Climate (SROCC) reported under RCP8.5 will...
Abstract. This paper reviews the state-of-the-art in storm surge forecasting and its particular application northern Adriatic Sea. The city of Venice relies crucially on a good flood system order to protect extensive cultural heritage, their population, economic activities. Storm systems are place warn population imminent threats. In future, it will be paramount importance increase reliability these systems, especially with new MOSE mobile barriers that completed by 2021, depend accurate...
In September 2022, three major subsea leaks occurred in the Nord Stream pipelines, releasing an estimated 225 kt of natural gas into Baltic Sea. To explain its behavior water column, we developed a combined model to simulate near-field buoyant bubble plumes and far-field advection, dispersion, volatilization, biodegradation dissolved methane fraction. According these simulations, 94.9% leaked was emitted atmosphere immediately above by ascending plumes. The remaining 5.1% (11 kt) initially...
This paper discusses major issues in operational three dimensional (3D) ocean modeling for the Baltic Sea, based on experiences DMI Centre Ocean and Ice recent years. Both scientific technical aspects of are discussed. Recommendations given to development Sea 3D models.
In the future, shifts in wind storms across North and Baltic Seas are highly unpredictable, challenging projection of wave conditions for managing coastal hazards. Moreover, regional sea level rise (SLR), with very large uncertainty, complicates situation stakeholders seeking recommendations climate adaptation plans. The purpose this study is to examine change storm surge components water due a low tidal range Køge Bay near entrance Sea. Under high greenhouse gas emission scenario RCP8.5, we...
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
Sea level rise is a significant threat to coastal regions, requiring thorough scientific evaluations implement effective adaptation planning. Recent incidents of damages ships and infrastructure in harbours during storm surge events the western Baltic have highlighted urgent need for robust strategies sea rise. This study utilizes regional climate simulations, employing five members EURO-CORDEX ensembles, investigate scenarios under RCP4.5 RCP8.5 scenarios. Using concept "gate index",...
Recently proposed methods for 1-bit and 1.58-bit quantization aware training investigate the performance behavior of these in context large language models, finding state-of-the-art models with more than 3B parameters. In this work, we small vision ranging from 100K to 48M We introduce a variant BitNet b1.58, which allows rely on median rather mean process. Through extensive experiments obtained through training. further robustness quantization-aware changes learning rate regularization...
Contemporary machine learning models, such as language are powerful, but come with immense resource requirements both at training and inference time. It has been shown that decoder-only models can be trained to a competitive state ternary weights (1.58 bits per weight), facilitating efficient inference. Here, we start our exploration non-transformer model architectures, investigating 1.58-bit for multi-layer perceptrons graph neural networks. Then, explore in other transformer-based namely...