Abstract: The significant advances in high-performance computational resources have boosted the seamless evolution in ocean modeling techniques and numerical efficiency, giving rise to an inventory of operational ocean forecasting systems with ever-increasing complexity. The skill of the Iberia-Biscay-Ireland (IBI) regional ocean forecasting system, implemented within the frame of the Copernicus Marine Environment Monitoring Service (CMEMS), is routinely evaluated by means of the NARVAL (North Atlantic Regional VALidation) web-based toolbox. Multi-parameter validations against observational sources (encompassing both in situ end remote-sensing platforms) are regularly conducted along with model intercomparisons in the overlapping areas. Product quality indicators and skill metrics are automatically computed not only averaged over the entire IBI domain but also over specific sub-regions of particular interest in order to identify strengths and weaknesses of the model. The primary goal of this work is threefold. Firstly, to provide a flavor of the basic functionalities of NARVAL software package in order to validate IBI near real time components (physical, biogeochemical and waves); secondly, to showcase a number of the practical applications of NARVAL; finally, to present the future roadmap to build a new upgraded version of this software package, which will include the validation of multi-year and interim products, the computation of long-term skill metrics or the evaluation of event-oriented multi-model intercomparison exercises. This synergistic approach, based on the integration of numerical models and diverse observational networks, should be useful to comprehensively characterize the highly dynamic sea states and the dominant modes of spatio-termporal variability.

Abstract: This research proposes a modelling framework in which simulation and optimisation tools are used together in order to obtain optimal reservoir operation rules for the multi-objective Dautieng reservoir on the Saigon River (Vietnam), where downstream salinity control is the main objective. In this framework, hydrodynamic and salinity transport modelling of the Saigon River is performed using the MIKE 11 modelling system. In the first optimisation step this simulation model is coupled with the population simplex evolution (PSE) algorithm from the AUTOCAL optimisation utility (available as a part of MIKE 11) to estimate to estimate the discharge required to meet salinity standards at the downstream location of Hoa Phu pumping station for public water supply. In the second optimisation step, with the use of MATLAB optimisation toolbox, an elitist multi-objective genetic algorithm is coupled with a simple water balance model of the Dautieng reservoir to investigate how the optimised discharges obtained from the first optimisation step can be balanced with the other objectives of the reservoir. The results indicate that optimised releases improve the performance of the reservoir especially on controlling salinity at Hoa Phu pumping station. In addition, the study demonstrates that use of smaller time steps in optimisation gives a closer match between varying demands and releases.

Abstract: In this paper we describe our endeavours to explore the role of unsupervised learning technology in profiling marine conditions. The characterization of the marine environment with hydrographic variables allows, for example, to make technical and health control of sea products. However, the continuous monitoring of the environment produces large amounts of data and, thus, new information technology tools are needed to support decision-making. We present here a first contribution to this area by building a tool able to represent and normalize hydrographic conditions, cluster them using unsupervised learning methods, and present the results to domain experts. The tool, which implements visualization methods adapted to the problem at hand, was developed under the supervision of specialists on monitoring marine environment in Galicia (Spain). This software solution is promising to early identify risk factors and to gain a better understanding of sea conditions.

Abstract: Toxins from harmful algae and certain food pathogens (Escherichia coli and Norovirus) found in shellfish can cause significant health problems to the public and have a negative impact on the economy. For the most part, these outbreaks cannot be prevented but, with the right technology and know-how, they can be predicted. These Early Warning Systems (EWS) require reliable data from multiple sources: satellite imagery, in situ data and numerical tools. The data is processed and analyzed and a short-term forecast is produced. Computational science is at the heart of any EWS. Current models and fore-cast systems are becoming increasingly sophisticated as more is known about the dynamics of an outbreak. This paper discusses the need, main components and future challenges of EWS.