In general ocean circulation models (OGCMs) diapycnal diffusion arises not only from the discretisation of the explicit diffusion, but also by numerically induced diffusion, caused, e.g., by common discretisations of advective transport. In the present study, three different diagnostics to analyse the mean diapycnal diffusivities of individual tracers (vertically and horizontally) are introduced: (i) The divergence method based on the work of Ledwell et al. (1998) infers the mean diapycnal diffusivity from the advection-diffusion equation. (ii) The tracer flux method based on the work of Griffies et al. (2000), that determines the diapycnal flux crossing an isopycnal layer, is modified for the analysis of mean diapycnal diffusivities of a passive tracer. (iii) The variance method based on the work of Morales Maqueda and Holloway (2006) is a more general approach as the diapycnal diffusion is analysed by the variance decay of the total tracer concentration. These methods can be used for the analysis of the diffusivity of passive tracer independent of the model set-up, e.g. the advection scheme used, but support only information about mean diapycnal diffusivity of that tracer field rather than for each individual layer. The applicability of these methods is tested in a set of 1- and 2-dimensional case studies. The effect of vertical advection and of diverging and converging isopycnals is shown separately. In all three methods used, the transformation of the tracer onto isopycnals leads to errors in the diagnosed diffusivities. It turns out that the tracer flux method is the most robust method and therefore the method of choice. In order to keep the errors as small as possible, longer time mean values should be analysed.

Limitations on our ability to predict fine-scale spatial variability in plankton ecosystems can seriously compromise our ability to predict coarse-scale behaviour. Spatial variability which is deterministically unpredictable may distort parameter estimates when the ecosystem model is fitted to (or assimilates) ocean data, may compromise model validation, and may produce mean-field ecosystem behaviour discrepant with that predicted by the model. New statistical methods are investigated to mitigate these effects and thus improve understanding and prediction of coarse-scale behaviour e.g. in response to climate change. First, the standard model fitting technique is generalised to allow model-data ‘phase errors’ in the form of time lags, as has been observed to approximate mesoscale plankton variability in the open ocean. The resulting ‘variable lag fit’ is shown to enable ‘Lagrangian’ parameter recovery with artificial ecosystem data. A second approach employs spatiotemporal averaging, fitting a ‘weak prior’ box model to suitably-averaged data from Georges Bank (as an example), allowing liberal biological parameter adjustments to account for mean effects of unresolved variability. A novel skill assessment technique is used to show that the extrapolative skill of the box model fails to improve on a strictly empirical model. Third, plankton models where horizontal variability is resolved ‘implicitly’ are investigated as an alternative to coarse or higher explicit resolution. A simple simulation study suggests that the mean effects of fine-scale variability on coarse-scale plankton dynamics can be serious, and that ‘spatial moment closure’ and similar statistical modelling techniques may be profitably applied to account for them.

Oceanic large-scale westward propagating features are capable of transporting heat and momentum and their importance in the set-up of major climate oscillations is undeniable. The ability to monitor and understand climate change is one of the top priorities in research. The possibility of using of these waves as indicators of such changes was analysed by Fyfe and Saenko (2007) who found that the increasing temperatures of the ocean's upper layers caused wave speeds to increase 35% in the 2 centuries analysed. Given the reported changes in ocean stratification and heat content (IPCC, 2013), this work analyses 20 years of altimetry data for changes in the Rossby wave signatures that might result from changes in the oceans. The 'Rossby wave-like' (RW) variability and speed distribution is characterised on a global scale. In the Indian and Pacific, the ocean's RW variability is shown to be significantly correlated with the Niño time-series, proving the RW sensitivity to ocean changes. A second RW dataset, derived from NEMO SSHA outputs - NEMO RW - is produced and validated against the altimeter RW data, providing good results especially for latitudes up to 20º. Speeds over time of the RW signal are computed for both datasets for trend analysis. The altimeter data proved to be too short however, 42 years of NEMO RW data provided trend maps showing speed changes occurring in all ocean basins, albeit with a more complex distribution than the findings of Fyfe and Saenko (2007) suggested. The NEMO RW speed time-series shows signs of being modulated by ENSO and the South Atlantic Southern Dipole in the Pacific and the Atlantic, respectively. Results confirm that changes are occurring in the RW speeds over the time-series analysed and establish the RW sensitivity to changes over time in the ocean's properties.

Small changes in the sea-surface temperature (SST) distribution in the equatorial Pacific are known to have large consequences for the weather and climate on regional and even global scales. A detailed knowledge of the SST patterns is therefore necessary if we are to understand and predict these effects thoroughly. Effort has hitherto been concentrated on understanding the dynamics of the West Pacific Warm Pool, but attention is now shifting to the eastern Cold Tongue region. This thesis contributes to this programme by assessing the impact of one unique feature of the equatorial East Pacific; the presence on the equator of a medium-sized island group - the Galapagos Archipelago. The project is carried through using a combination of: in situ oceanographic data from ships and moorings; satellite SST and colour images; and diagnostics from both a sophisticated ocean general circulation model and a simplified (reduced gravity) model. Each model was run both with and without a representation of the Galapagos. It is found that under typical conditions the islands form the eastern boundary of the cold tongue; the lowest equatorial SSTs in the Pacific occur on the west coast of the main islands of the group. As well as preventing the cold tongue from extending any further east, the islands also cool the sea surface still further for a distance of some 1000 km to the west. For the first 100 km or so the cooling is of the order of 1C; on the larger scale it is more like 0.1C. This cooling is a result of hydraulically enhanced upwelling of waters from the Equatorial Undercurrent (EUC), which is best explained by a Bernoulli-type (non-rotating, non-linear) theory. This theory also predicts that water parcels in the EUC will undergo vertical excursions on their passage around the north and south coasts of the islands; these excursions are observed in at least one numerical model. Downstream of the archipelago the EUC usually reunites and executes inertial oscillations, meandering about the equator. Transport in the EUC is also reduced because of the presence of the islands, by 30-50% to the east of the islands and by up to 20% to the west. Enhanced SST variability north and south of the equator immediately to the west of the Galapagos indicates that tropical instability waves (TIWs) are more active here than they would be were the islands absent. This is due to a pair of barotropically unstable zonal jets originating at the northern and southern tips of the islands. These jets are consistent with the theory of low-frequency flow past equatorial islands on the beta plane (which predicts that the South Equatorial Current will split into two such jets on encountering such an island), although weaker jets are found in models with no representation of the Galapagos. As with the cold tongue, the islands form the eastern limit of the region where these jets are found.

Mixing processes in shallow seas are of particular importance in the dispersion of pollution (e.g. oil slicks) as well as in the distribution of sediment and biological material (e.g. phytoplankton and fish larvae). This study aims to further investigate the processes of wind-driven Langmuir circulation and bottom-generated turbulence and to quantify the effects they have on the dispersion of buoyant material. A numerical and analytical study is made of the dispersion of a plume of floating particles under the combined action of Langmuir circulation and a mean current. The numerical simulations show the general patterns the plume will form, with initial meanders leading to the formation of distinct windrows. The analytical model shows that the lateral diffusivity, the rate of across-current dispersion of the particles, is greatest when the angle between the direction of the wind and current is between 30° and 120°, and when the wind speed is large in comparison to the current speed. Further, it is found that the effects of Langmuir circulation will dominate the dispersion process when the wind speed is greater than about 10 times the current speed. Acoustic and visual observations of the surface effects of bottom-generated turbulence in a tidally-influenced and well-mixed region of the North Sea are presented. Although the sea bed in the area is flat, it is found that at any one time 20-30% of the water surface is affected by boils - circular regions of local upwelling - of diameter 0.9±0.2 times the water depth. The signature of individual boils persists for at least 7 minutes and, in accordance with laboratory and numerical studies, shows the appearance of eddies. Again, analytical and numerical models are constructed to estimate the effects of the boils on the dispersion of floating particles. The numerical simulations provide good qualitative agreement with the patterns observed in visual images of the dispersion of oil. Further, the horizontal diffusivities calculated by both them and the analytical methods show that the presence of boils can account for the dispersion measured by experimental methods in shallow tidal seas. In summarising and discussing the main findings of this work, particular emphasis is placed on the need for further observations of these two processes, in particular the boils and eddies and their interaction with the Langmuir circulation, to further improve the estimates of dispersion.

This thesis investigates the statistics of the long-term variability in sea level records and trends along the Atlantic coast of Europe. This is undertaken for a range of regionally-representative open-sea sites. Tidal variability on a shorter time-scale is also investigated, for a local area subjected to intense natural and anthropogenic activity. The longest available records of hourly sea level data have been analysed for (6) ports around the English Channel, and (5) around the Iberian Peninsula. A separate statistical analysis is used for the components of sea level variability: mean sea level, tides and meteorological residuals (surges). Extremes and percentiles of some of these components have also been studied. This shows that mean sea levels are increasing at between 1 and 3 mm yr-1. Strong influences are found in the standard deviation of the total observed sea level. The various tidal constituents show interesting, localised short-term variations in amplitude and phase. Some long-term trends exist and are mainly associated to local effects. There is no evidence of an increase in weather effects on sea levels, over the period analysed. Some trends are identified in extreme maximum and minimum levels; however, in most cases, they are influenced only by mean sea level. Sea level response to the NAO has been analysed. The NAO was found to influence extreme levels in the meteorological residual (standard deviation), at stations where the NAO´s centre of action are known to be strongest. Mean sea levels, detrended sea levels and non-tidal residual sea level standard deviations around the western European coastline are generally correlated negatively with the NAO index; the only exception to this is Dover, which is correlated positively to the NAO. This observation is consistent with the known positive correlation in the south eastern region of the North Sea. Sensitivity to the NAO varies, with the highest values found along the northern Spanish coastline. Overall, Coruna shows the greatest influence of the NAO, and, in the English Channel, this is found at Newlyn. Influences on the non-tidal meteorological residuals are also found within the northern region of Spain, suggesting that variance in this component is influenced considerably by the NAO. More locally, the responses of coastal lagoons, to long-term sea level trends, are crucial for coastal management. Pressure measurements collected in the Ria de Aveiro Lagoon, Portugal, show a general increase in amplitude and decrease in phase, over the past 16 years. The causes of these changes are investigated using: (a) an analytical model of a wave propagating through a narrow inlet, into a Lagoon (analogy to RLC-circuit); and (b) a Bi-Dimensional Horizontal (2DH) hydrodynamic (vertically-integrated) model of the Ria de Aveiro Lagoon. Results suggest that changes in the bathymetry are the most significant contribution to the changes identified in the tides.

The rapid cooling of the waters at high latitudes creates an unstable strati cation which in turn leads to localised overturning (sinking) of the water column. This process is called open ocean deep convection (OODC). The process of OODC occurs in stages. Initially, individual convective elements known as plumes form and cold, dense water descends from the surface. Over time these plumes build up to produce a well-mixed `chimney' of cold dense uid. This chimney then slumps and sinks, and restratication (the return to a stable state throughout the water column) occurs. It is widely accepted that OODC plays a main role in driving the thermohaline circulation (THC) and hence has a potentially major role in climate. However, the mechanisms of OODC itself are not fully understood, and there is much debate surrounding how it contributes to THC. One di- culty is that OODC tends to occur sporadically in only a few isolated regions around the globe, making direct observations dicult. As a result, theoretical and numerical investigations have become key to the development of our understanding of OODC. The scale on which OODC occurs presents a further issue, with traditional numerical representations (parameterisations) of OODC in global circulation models (GCMs) omitting convective detail due to resolution. Due to the scales on which OODC occurs, it has been dicult to numerically investigate the nature of OODC in the small scale at the same time as resolving basin scale circulation. With the advent of nite element methods and adaptive meshing techniques, it is now possible to study OODC in regional models without the need to parameterise. One such model, the Imperial College Ocean Model (ICOM) is employed in this thesis for these purposes. ICOM is a 3D nite element, non-hydrostatic model with an adaptive, unstructured mesh and non-uniform resolution, allowing modelling of i the gyre circulation and resolution of OODC simultaneously. As the use of an adaptive, unstructured mesh model is novel in investigating Greenland Sea open ocean deep convection, it is of interest to assess the accuracy of the ICOM model, and the amount of numerical diusion present. The classical uid dynamics problem of parallel plate convection provides a simple test problem for this purpose. A series of tests investigating the linear stability of various temperature gradients were performed in order to diagnose the amount of numerical diusivity associated with hexahedral, tetrahedral and adaptive meshes within ICOM, and ICOM was further compared with a leading GCM (MITgcm). The use of the linear instability problem was found to be a useful case against which to test numerical models in an attempt to diagnose implicit diusivity and viscosity. A series of experiments were conducted in order to identify any prevailing dierences between model convection in xed and adaptive mesh congurations, under varying durations of applied cooling, and using varying extents of horizontal cooling. The adaptive mesh proved to be highly suitable for studying the convective problem, it was less computationally expensive and free from the numerical instability observed on the xed mesh. The sensitivity of model convection to the introduction of stratication was investigated. Uniform cooling was applied across the surface of a domain initialised with a weak stratication over the surface 1500m and a more strongly stratied region below, and the development of a convective layer was observed within the initial upper layer. Convection was constrained to the upper layer of stratication, and some penetrative convection was identied in the early stages of the model run.

Understanding the effect of physical forcing on estuarine functioning is of major importance to determine the rate of exchange of water, sediments, pollutants and nutrients between the continent and the ocean. The combination of numerical models and discrete datasets is used to describe and investigate processes of natural variability in the partially-mixed, non-turbid, macrotidal Solent-Southampton Water estuarine system (UK). The estuarine circulation and the response of wind forcing is examined using a three dimensional, free surface, finite volume and finite element grid model. Results from short-term (three months) simulations have been compared against data (ADCP measurements, tidal elevations and salinity distributions) collected in spring 2001 in Southampton Water and the Solent. The model reproduces the unique tidal curve of Southampton Water and the partially-mixed conditions prevailing in the upper estuary. The contribution of the local wind forcing (wind intensity · 12 m s−1) to changes in water level is estimated to be up to 6 cm in Southampton Water in the model. The modelled salinity stratification varies over a semi-diurnal cycle with the highest stratification occuring at mid-ebb. Wind forcing is more efficient in altering stratification at ebb than flood. The temporal and spatial variability of light attenuation is also investigated. Turbidity is demonstrated to be the major contribution to light attenuation using a time-series of discrete data collected in 2001, 2002 and 2003. A typical seasonal cycle of the coefficient of light attenuation is revealed, with a minimum in May-June and a maximum occuring in September-October. A second dataset of continuous measurements (10-minute interval) demonstrates the spring-neap modulation of the turbidity. The mouth of Southampton Water is more exposed to tidal mixing and therefore more turbid than the mid-estuary. A five-compartment zero-dimensional pelagic ecosystem model including a sediment compartment has been developed to assess the impact of the variability of the light attenuation on the timing and the magnitude of the spring phytoplankton bloom. Using high resolution irradiance forcing and a constant coefficient of attenuation k set to the minimum May-June value, simulations compare well with discrete data of chlorophyll a, and less successfully with zooplankton and Dissolved Inorganic Nitrogen. A sensitivity analysis indicates that interranual variability in the phytoplankton spring bloom originates in order of importance from 1) parameterization of k 2) the variation of the seasonal cycle of surface irradiance 3) the intrinsic dynamics determined by the combination of fixed parameters of the ecosystem model.

The provenance and transport of sand has been investigated around Lido Inlet, the largest of three tidal inlets in Venice Lagoon, Italy. Morphological analysis has established the presence of an ebb-tidal delta extending from the mouth of Lido Inlet as well as other features typical of tidal inlets. The stability of the inlet, as well as the canals of Treporti and Burano, was determined through the application of the tidal prism/cross-sectional areas relationship theorized by O’Brien (1969). Whilst Lido Inlet and Treporti Canal have both remained in equilibrium in terms of this relationship, Treporti Canal has suffered erosion due to a fluctuating tidal prism. Lido Inlet is slightly flood dominant although grain trend analysis of bottom sediment reveals net export of sand. Treporti Canal is ebb dominant and is the source of this sand, but it is becoming increasingly flood dominant as average current speeds have reduced and ebb currents are weakening at a faster pace than flood currents. This is proposed as a reason to why the northern lagoon is accreting (0.44 cm yr−1), contrary to trends in the southern (-0.37 cm yr −1) and central lagoon (-0.23 cm yr−1). A sediment budget formulated for the whole lagoon has revealed that whilst the overall rate of erosion is reducing, the area subject to erosion is increasing and the rate of accretion is decreasing, resulting in no net change in the net sediment loss rate between 1930-1970 to 1970-2000 (-0.05 cm yr−1). Mineralogical analysis on bottom samples, beach and riverine samples confirmed that longshore transport is from north to south along the northern lagoon; carbonate grains are dominant in the north with proportions gradually replaced by quartz to the south. Lido Inlet proved to have similar mineralogy to the River Piave with a higher calcite/dolomite ratio than inner lagoon samples suggesting a less mature sediment and thus import of sediment. Conversion of ADCP backscatter into bedload transport rate, suspended sand, and suspended fines concentrations has shown that no sand is transported at velocities below ∼0.4 m s−1. 55% of sediment transported during an ebb flow was sand in suspension (peak: 527,000 kg hr−1), 37% was fines in suspension (peak: 283,000 kg hr−1), and the remaining 8% of sediment was transported as bedload (peak 68,700 kg hr−1).

The ocean current system off Southern Greenland is a key component of the Earth’s climate system due to its role in the regulation of the global thermohaline circulation. A combination of historic and new observational data, supported by modelling simulations, has revealed great complexity in both the surface and deep currents of the area. A comprehensive review of the available hydrographic data for the South-East Greenland shelf has shown that the observed spatial variability in the transport of the on-shelf East Greenland Coastal Current is inconsistent with the theory that it is primarily driven by local fjord runoff. A high resolution regional modelling study supports these observations and suggests that the East Greenland Coastal Current is primarily the result of a bifurcation of the East Greenland Current in the vicinity of Denmark Strait. Model simulations also suggest that the pathways followed by the low salinity output of the region’s fjord complexes are influenced by the strength and position of the main East Greenland Current. New data collected in the vicinity of Cape Farewell, on the southern tip of Greenland, has also revealed more complex behaviour in the Deep Western Boundary Current. The significant spatial variability in Deep Western Boundary Current transport observed in the region of Cape Farewell suggests that the confluence of Denmark Strait Overflow Water and Iceland Scotland Overflow Water occurs over a wider geographic area than is commonly suggested. A review of historic data indicates that these spatial patterns vary over timescales of months and upwards. It is hypothesised that strands of overflow water follow different depth controlled pathways dependent on their relative water density in comparison with the surrounding water masses. This has significance for the way deep current strength proxies in the region are interpreted in relation to climatic variations. Time series data for a single location may well overestimate the variability in the Deep Western Boundary Current further south due to path switching of the flow. In the future more complete observational datasets combined with more advanced oceanographic and coupled climate models will provide a better understanding of the interaction between the regional current system, the global current system and the climate system. This will enable more reliable prediction of the impact of global warming and, in particular, the potential influence of accelerated Greenland ice cap melting on the global ocean circulation.