# Thesis topics

### From GHER

(→Multiscale aspects) |
(→New Data Assimilation schemes) |
||

Line 58: | Line 58: | ||

Data assimilation techniques are known for some time now and the Kalman filter a standard industrial tool for controling systems or estimating the state of a system. When applied to meteorology and oceanography two major problems arise: One is the size of the mathematical problem, the other the strong linearities found in geophysical fluid dynamics, leading to a large spectrum of scales and processes. Therefore assimilation techniques need to be adapted to be useful in ocean modelling and two specific aspects can be tackled | Data assimilation techniques are known for some time now and the Kalman filter a standard industrial tool for controling systems or estimating the state of a system. When applied to meteorology and oceanography two major problems arise: One is the size of the mathematical problem, the other the strong linearities found in geophysical fluid dynamics, leading to a large spectrum of scales and processes. Therefore assimilation techniques need to be adapted to be useful in ocean modelling and two specific aspects can be tackled | ||

- | |||

- | |||

- | |||

- | |||

- | |||

- | |||

==Oceanographic themes== | ==Oceanographic themes== |

## Revision as of 14:41, 4 July 2013

The following list of Ph.D. Thesis topics proposed at GHER is not exhaustive and depending on the student's interest and the ongoing research projects additional themes can be defined.

Our objective is to propose subjects which are stronly linked to ongoing research efforts at international level (see our projects) and make sure students can collaborate with collegues at other universities or research centers. This has also the advantage of ensure adequate access to recent data.

Studends are welcomed to apply a Ph.D. position at any moment and depending on the schedules and interests of the student, applications to FRIA or FNR-FNRS will be prepared. For Ph.D positions offered within projects, you can also consult [1] or Jobs

## Contents |

## Data Analysis methodologies

### Multivariate approaches

Whereas classical optimal interpolation allows naturally for the analysis of several variables (such as nitrate and phosphates) in a joint matter, exploiting covariances between variables, our spatial gridding tool DIVA only can proceed a singe variable at a time. Though allowing for more realistic spatial covariances than classical optimal interpolation DIVA is therefore however underexploiting possible links between variables. The work will consist in adding this feature to DIVA (by an iterative approach) and apply it to the creation of biochemical climatologies in the Mediterranean or Black Sea. If sufficient data and correlations with other parameters reliable, the regime shifts in deep water formations in the eastern Mediterranean could be revisited.

### New FEM DIVA solver in ND

The present version of DIVA, also exploited in Diva on web, is using a robust but outdated finite element solver in 2D. In order to prepare for 3D or 4D applications, the use of a general FEM framework should be analysed such as GETDP with its mesh generator gmsh. The implementation should be portable and parallelize and show the benefit of performing 3D analyses compared to 2D versions. Oceanographic applications will then concentrate on deep ventilation processes, something which is difficult to represent with horizontal analysis alone.

### Multiscale DINEOF-OI

DINEOF is a widely used tool for analysing satellite data with missing data. Not only does the tool provide estimates of the missing data but also the structure of dynamic modes from an EOF (empirical orthogonal function) analysis. To further exploit the dynamic information found in these modes, a scale selective analysis can be wrapped into the filling step, so as to provide modes for specific scales. This is particularly interesting for understanding relationships with climatic modes for example. Here several strategies for decomposing and filterging should be tested and validated on long times series of temperature fields. High frequency images in combination with low frequency passes will then allow for a better specification of the features at various time and space scales. An application to a tidally dominated sea and an application to a mesoscale dominated region will be used to check the sensitivity of the method. If successful it would allow to provide guidance for the specifications of covariances matrices used in data assimilation.

## Developments in numerical modelling

### Ocean Sound

The research project aims to contribute to the sustainable development and noise reduction of activities related to offshore wind parks in Europe. The objective of the project is to develop and validate a sound propagation model for the ocean capable to give a precise evaluation of the noise level in the ocean generated by wind parks and to quantify it. The project should also develop an appropriate measurement strategy related to this question. The project is coordinated by a company in the region of Liège (Belgium).

### Coupled ocean atmosphere model with improved bulk exchanges

Historically ocean models and atmospheric models have been operated individually with some crude representation of the other compartment. Now coupled models are more and more used but it appears that the exchange laws used in the past cannot be applied directly as they were based on calibrations using simplified compartments (the ocean model assumed a known wind speed and the atmospheric model a kowns surface temperature). Here a coupled model should be setup and exchange laws optimised using data assimilation theories.

### Implementation of an operational model at Stareso-Calvi

Using ocean model result from myOcean, atmospheric predictions and measurements from the ADCP and weather station at Stareso, an operational 3D model ocean model with data assimilation should be setup. Once validated it will be used for reanalysis of the last decades to analyse changes in hydrography.

## New Data Assimilation schemes

Data assimilation techniques are known for some time now and the Kalman filter a standard industrial tool for controling systems or estimating the state of a system. When applied to meteorology and oceanography two major problems arise: One is the size of the mathematical problem, the other the strong linearities found in geophysical fluid dynamics, leading to a large spectrum of scales and processes. Therefore assimilation techniques need to be adapted to be useful in ocean modelling and two specific aspects can be tackled

## Oceanographic themes

Here the emphasis is rather on using the tools mentionned above to look at specific oceanographic questions

### Sub-grid scale models

The focus will be on a process which is particular to geophysical fluid flows and essential in redistributing heat, salt and momentum in the ocean: baroclinic instability. Through this non-linear instability, potential energy of fronts is released and resulting eddies of the size of the so-called deformation radius lead to heat and salt fluxes in cross-frontal directions. Numerical grids of general circulation models generally are too coarse to capture these scales and therefore heavily rely on the parameterization of this effect. The most common parameterization is the so-called Gent-McWilliams skew-flux parameterization, adding in fact diffusion along isopycnals and a virtual advection term trying to flatten isopycnals. This parameterisation will serve as baseline for further investigations. To study the baroclinic instability, setup of simulations will be simplified to avoid interaction with errors related to boundaries etc. Furthermore, the dynamics of baroclinic instabilities are well captured by quasi-geostrophic (QG) models, a simplification of primitive equation models (the latter being currently used in forecasts and earth system modelling). The advantage of using QG models is that they allow very high resolution, possibly even in spectral space for easier application of existing methodologies. The objective is to improve closure schemes and possible provide a stochastic model of the error associated with the closure scheme.