Scottish Shelf Model. Part 6: Wider Domain and Sub-Domains Integration

Part 6 of the hydrodynamic model developed for Scottish waters.

7. Summary and Conclusions

Stage 3 of the Scottish Shelf waters project focusses on integration of the case study sub-models into the wider domain shelf model and the application of the combined model to calculate connectivity indices between fish farms for the study of disease transmission by sea lice and viruses. This report presents this final piece of work.

In order to combine the shelf model and nested model outputs an interpolation method was developed. This method utilised the results produced in Stage 1 and Stage 2 of the project. Two integrated meshes were produced. The first is used to integrate the results from all four case study sub-models and the shelf model from May to October. The second is used to integrate the results from two case study sub-models ( PFOW & WLLS) and the shelf model only for November to February. The integrated meshes were developed to ensure the best mesh resolution was preserved. Results were interpolated using Matlab to produce daily output files with an hourly timestep for use in the particle tracking model.

The FVCOM offline particle tracking code as supplied had some bugs in the diffusion part of the code which had to be fixed. We decided to implement a version of the code developed by Pierre Cazenave (Plymouth Marine Laboratory) which saved the grid metrics calculated after the first run of the code, allowing a saving of about 2 hours in run time on subsequent runs and much quicker debugging, as the particle tracking starts immediately. This code also had some initial bugs which had to be fixed.

The particle tracking runs were time-consuming in that 97,700 particles were tracked for the sea lice setup and 146,550 for the virus setup and each run took several days to weeks. This was speeded up by dividing the particles between several processors. Although the code is not parallelised, multiple runs can be carried out in parallel as long as the particles are from different spatial locations (multiple particles from one point may produce duplicate runs due to the nature of the random number generator).


Back to top