high performance computing on graphics processing units: hgpu.org

Environmental problems and issues are not limited by artificial boundaries created by man. Usually there are different teams or individuals working on the catchments, estuaries, rivers and coastal basins in different countries using different parameters and formulations for various processes. However, the system is a natural one and as such no boundaries exist. When a rain drop of water moves from a catchment to a stream, river or estuary and to the sea, it does not know at any stage whether it is in the catchment, river, estuary or sea. In recent years there has been growing concern about the impact of diffuse source pollution on river, estuarine and coastal water quality and particularly with regard to non-compliance of bathing waters. Hydro-environmental impact assessment modelling studies are generally regarded as having several fundamental shortcomings in model simulations, which can lead to erroneous environmental impact assessment outcomes. These shortcomings were addressed in this project and included: 1. Applying a Cloud to Coast (C2C) approach to modelling flow and solute transport processes in river, estuarine and coastal waters; 2. Improving the computational linking of catchment, river and estuarine-coastal models; 3. Improving the kinetic decay process representation in deterministic models, to include the impacts of salinity, solar irradiance, turbidity and temperature, and; 4. Using parallel (GPU based) computing to enhance the computational speed in executing bathing water quality models. In this research study, the Ribble Estuary and Fylde Coast model was refined to more accurately predict bathing water quality and use parameters which give a better representation of the existing physical and bio-chemical processes. A Cloud to Coast (C2C) approach to modelling was implemented by using common input parameters and formulations in both the 1-D and 2-D domains of the Ribble Estuary and Fylde Coast model. An empirical formulation linking the mortality (decay) rate of bacterial water quality indicators to environmental conditions such as solar irradiation, turbidity, temperature and salinity was added to the numerical code. The linked boundary between the 1-D and 2-D domains of the numerical model was improved by removing the large overlapped linked region. An existing numerical code was rewritten to take advantage of the parallel computing capability of the Graphics Processing Unit (GPU). This was tested on the Ribble Estuary and the Thames Estuary model. Thames Estuary models of varying mesh density were prepared in this research study using lots of bathymetric data (over 80 million points) and tested on the GPU. This research study improved the ability to predict bathing water quality accurately by introducing more realistic representation of environmental conditions and using parallel computing. This improved the ability to carry real time forecasting of bathing water quality and hence prevent failure to meet the requirements of the EU Bathing Water Directive.