113 minimized due to: spatial discretization via quadratic finite elements and ? transformation, allowing for optimal representation of water bodies with complex geometries and bottom topography; wind field and bottom roughness that can vary dynamically in time and space, and self-adjusting multi-scale turbulence modelling based on Large Eddy Simulation. ? Eulerian transport model: It is a general purpose advective–di?usive transport model with kinetic reactions for 2-DH or selected layers of 3-D ?ows with a given thickness. This model can be used to compute space distribution and fate of dissolved contaminants. ? Water quality models: A set of Eulerian transport models for the coupled simulation of water quality parameters like salt, temperature, DO-BOD, nitrogen compounds, phosphorous compounds and biomass. Models can be applied for 2-DH ?ows or for selected layers of 3-D ?ows with a given thickness. The temperature module of this set of models can be used to compute the time variation and space distribution of temperature of thermal plumes occupying a surface layer. ? Lagrangian transport model - deterministic mode: It is a general purpose advective– di?usive transport model with kinetic reactions for selected layers of 3-D and for 2-DH ?ows. This model is especially suitable for the simulation of plumes or clouds that are initially small to be well resolved by the discretizing mesh of the associated hydrodynamic model. Any curve representing a kinetic reaction dependent on the lifetime of a given particle can be adopted. This model can be used for computing the space distribution and decay of particulate contaminants. The user can choose to run the model in free transport mode or conditioned transport mode. The latter being particularly suitable for simulations of sedimentological processes. The transport can be conditioned by a minimum velocity, minimum bottom stress due to currents or due to currents and wind waves. The user can also specify a tolerance band for the limiting condition, in which the transport of a particle follows a fuzzy decision process. ? Lagrangian transport model - probabilistic mode: In this mode, the user can produce maps of isolines of probabilities based on N events or for a period of time T. Examples of outputs are: isolines of probability of visitation of a plume or cloud with concentrations above a given limit or determination of critical events, as the first event or first time in which a plume or cloud touches the coastline, etc. ? Wind-wave generation model: For a given wind field, variable in time, the model computes the wind wave field generated within the model domain, for a persistence of wind, and time intervals defined by the user. For all nodes in a given domain, the model computes parameters like significant and root mean square wave heights and periods, oscillatory bottom stresses, limiting fetches etc. VIII-3. HYDRODYNAMIC MODELLING DETAILS The 3-D spatial discretization is achieved via a vertical stack of sub-parametric finite element meshes using ?-coordinate transformation along the vertical dimension. That is, if one looks from the top, one sees the horizontal plane of the domain discretized by a single mesh of finite elements. However, in actual fact, there will be a stack of meshes, one for every ? level. In this way, vertical discretization is performed automatically once the user defines the number of desired ? levels (usually between 10 and 50). The 3-D model is automatically activated if at least 5 ? levels are requested. Elements in a mesh are sub-parametric. The variables in each element are defined by quadratic Lagrangian polynomials whereas the element geometry is defined by linear Lagrangian polynomials. Elements in a mesh can be quadrilaterals and/or triangles. Quadrilaterals are preferred because variables become bi-quadratic, and thus have a higher accuracy. This discretizing scheme is potentially of fourth order on the ? planes and of