Multi-scale modelling of effluent dispersion in the marine environment

This research aimed to investigate whether the unique numerical methods available within CFD model software Fluidity could progress the state-of-the-art in various aspects of modelling effluent dispersion within the marine environment. Fluidity contains a large library of models and numerical method...

Full description

Bibliographic Details
Main Author: Robinson, David Iain
Other Authors: Gorman, Gerard ; Piggott, Matthew
Published: Imperial College London 2016
Subjects:
Online Access:http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.689152
Description
Summary:This research aimed to investigate whether the unique numerical methods available within CFD model software Fluidity could progress the state-of-the-art in various aspects of modelling effluent dispersion within the marine environment. Fluidity contains a large library of models and numerical methods that enable modelling of flow processes at a wide range of scales. It has been proven to perform well when used for massively-parallel simulations (i.e. it scales well), and it has the uncommon facility of unstructured mesh adaptivity, which has the prospect of significantly increasing the efficiency of CFD simulations when guided skillfully. This research also forms part of a longer-term project to create a coupled (or even single) model of effluent dispersion that represents influencing factors from a wide range of scales (from tidal currents down to turbulent eddies) entirely using CFD techniques. As such, one aspect of the research was to validate the numerical methods available within Fluidity for use in modelling effluent dispersion. To facilitate this validation, some of the model studies investigate aspects of effluent dispersion modelling from a hypothetical outfall site off the North-East coast of the United Kingdom. Studies were performed in a series of stages in which key aspects of effluent dispersion modelling were addressed. CFD models were created of near-field jet dispersion, tidal motion, and far-field plume dispersion. Idealised test cases were also performed to investigate the performance of advection-diffusion solver methods. At each stage the aim was to investigate the benefit of novel numerical modelling techniques and compare their accuracy and efficiency to existing methods. A set of near-field buoyant jet dispersion CFD models were created, one representing conditions associated with power, and combined power and desalination plants, and one representing conditions typically associated with desalination discharge. These CFD models utilised a mesh adaptivity algorithm to optimise mesh resolution during the course of the simulation. Model predictions were compared with published laboratory data and the predictions from validated integral models. An assessment was made of when CFD offers a benefit over other modelling options, and when it might be sufficient to use cheaper tools. There was also a discussion of the effectiveness of mesh adaptivity in increasing model efficiency, together with advice for how and when it is best to use mesh adaptivity when modelling buoyant jet dispersion. Model results showed that with modest parallel computing resources and expertise, high-resolution simulations of jet dynamics can be achieved with reasonable accuracy using CFD modelling. A model was created of tidal flow within the European continental shelf and results were compared to a large database of tide gauge measurements. This model took advantage of recently published methods for ocean model meshing and coastline resolution reduction. The purpose of this study was to confirm that these methods offered a benefit to model accuracy and efficient, and also that Fluidity could be used to accurately generate the tidal forcing boundary conditions for a far-field model of effluent dispersion at a hypothetical outfall site. The predictions of M2 tide amplitude in the vicinity of the outfall site had an average error of 10.1% compared with tide gauge measurements. The predictions of S2 tide amplitude in the vicinity of the outfall site were even closer to tide gauge measurements, with an average error of 3.7%. The speed of the model solve showed a vast improvement over a previous comparison model study, with 37 days of tidal motion being simulated in 15.2 hours (58.4 seconds of simulation for each second of solving), compared to the comparison simulation with a similar level of accuracy, which simulated 2 seconds of tidal motion for every second of solver time. A series of simplified test cases were run to assess a commonly-used advection-diffusion solution method from the library of those available within Fluidity. This work was intended to give general confidence that the numerical methods available within Fluidity are suitable for modelling coastal processes and so give confidence in later multi-scale results. The test cases chosen were relevant to coastal dispersion, including those testing tracer advection, diffusion, point sources and stratification. The method compared well with results published using world-leading free surface modelling software, Open TELEMAC. A model was created of the dispersion of neutrally-buoyant dissolved pollutant from a hypothetical outfall. The assumed effluent is typical of that released from a manufacturing plant. The aim of this modelling was to validate the use of Fluidity for modelling effluent dispersion within the coastal zone, and also investigate the benefit of using 2-d horizontal mesh adaptivity to optimise model mesh resolution during the course of the simulation. It was shown that the use of mesh adaptivity improved model efficiency, significantly lowering the effect of numerical diffusion. Finally, a short outline was given of a prospective strategy for producing a coupled-model of effluent dispersion, using as a basis the techniques developed within this thesis. The proposed coupled model of effluent dispersion would include a near-field jet model two-way (i.e. 'fully-coupled') to a far-field plume model. Tidal forcing would be provided by a one-way coupled tidal model. Fluidity is capable of modelling all of these processes and so third party coupling software would be unnecessary.