Proposal for Phd subject in joint collaboration CIFRE CLS-LEGOS-TOTAL and in collaboration with COAS/Oregon State University_
*Period:* start as soon as possible (autumn 2008 – autumn 2011)
*Thematic:* physical oceanography - ensemble statistics
*Title:** *Impact of satellite and at sea observations on coastal ocean analysis and forecast
*Funding*: Grant CIFRE CLS co-funded by TOTAL SA
*PhD Supervisor*: Dr Pierre de Mey, CNRS research director, LEGOS/POC, Toulouse
*Co-**supervisor*: Dr Eric Greiner, engineer in the space oceanography direction of CLS company, Operational Oceanography Department, Ramonville-St-Agne
*Correspondent in **USA: *Prof. Alexander Kurapov, COAS/OSU, USA Objectives and description:
Among the various branches of physical oceanography, coastal physical oceanography is one of the richest and most complex. It is also the area most likely to have a direct impact on society through issues such as coastal pollution, water quality, oils slick drift, sea level rise, coastal erosion. It also has a direct impact on fish stock management, maritime traffic monitoring and water sports events organisation. In the last decade, high resolution ocean modelling has become a core tools to answer the needs of these various fields. However, model forecasts have to be corrected by satellite and in situ data, routinely available in the coastal environment. This correction is applied through dataassimilation methods. The POC/LEGOS team is one of the few scientific groups to use tackle data assimilation in the coastal environment.
The proposed work follows up on Matthieu Le Hénaff (LEGOS 2008) thesis which dealt with a 3D coastal model of the Bay of Biscay. This thesis aim was to determine how an observational network could detect the degrees of freedom in numerical models and how it could forces the error modes in order to improve the forecast. To this end stochastic modelling was used on the POC pc cluster. It is considered crucial in this new thesis to carry one with this approach. Indeed, the results are a useful only if the model is reasonably good and if the origins of model errors have been identified.
The proposed thesis aims to develop a positive feedback between model error analysis and the optimisation of the observational network. Novel statistical method will be used. These will focus on the events probability rather than their intensity or their phase. Particular attention will be paid to new observational tools such as SWOT altimetry, coastal radars, gliders, on top of existing data.
Work Plan
Over the duration of the PhD, several aspects of coastal oceanography
will be investigated:
1) Regional Modelling:
The geographic area is centred on the Islands of Sulawesi and Borneo and could extend from 11°E to 126°E, and from 8°S to 8°N.It includes the Java Sea, the straits of Makassar, Lombok and Flores as well as the Celebes Sea and could potentially also include part of the South China Sea.. The Seas of Banda and Molucca would not be included. This would simplify the prescription of boundary conditions. The resolution of the grid would be 1/12°. If needed and if possible, a higher resolution could be implemented. The model domain would be nested in a regional or global 1/4° model configuration. Special care will be taken to make best use of recent work by the INSTANT community and by the LOCEAN and LEGOS laboratories on the region.
2) Model error analysis:
The aim is to determine model error by comparing output from a long hindcast to past observations. To this end, the structure of the regional model ensemble representers will be studied. These representers will be characterised in terms of physical processes (eddies, coastal waves, monsoon response).
3) New statistical methods:
Estimation theory is usually conducted in a Gaussian universe. This is not suited to the existence of extreme events (in terms of intensity) or the presence of remarkable one (in terms of frequency). As a result, Continuous Ranked Probability Score (Hersbach, 2000) will be used to determine a model short term forecast ability. This is equivalent to investigating the coherence of errors between model forecast and observations. A physical interpretation of remarkable events will also be conducted.
4) Impact of new observations on the model:
The aim is to determine what impact new sets of observations would have on the observation of remarkable events in the region. To this end, the RMS method (Le Hénaff et al., 2008) will be applied to modelled observations or existing ones. If an analysis method is available, the impact on forecast will be tested.
Coastal Radar observations are routinely conducted in Oregon. The candidate might be required to spend some time in Oregon in order to improve his understanding of the technology behind coastal radars data and their assimilation into models.
Location
*Host research Institute details**: *
*LEGOS*
Equipe Echange Côte-Large du LEGOS / Pôle d'Océanographie côtière de l'OMP PhD Supervisor : Dr Pierre de Mey
Status : directeur de recherche CNRS
Tel : +33 (0) 561 394 756
Email : pierre.de-mey@legos.obs-mip.fr
*Contact** in USA* : Prof. Alexander Kurapov
College of Oceanic and Atmospheric Sciences (COAS), Oregon State University, USA
*Co-**Funding company : *
CLS - Direction Océanographie Spatiale
8-10 rue Hermès,
Parc technologique du canal, |
31520 Ramonville Saint-Agne
France
Phd co-supervisor : Dr Eric Greiner
Status : senior research engineer
Tel : +33 (0) 561 394 756
Email : eric.greiner@cls.fr
