Please use this identifier to cite or link to this item: http://hdl.handle.net/2122/5856
AuthorsVanderbulcke, L.* 
Beckers, J.-M.* 
Lenartz, F.* 
Barth, A.* 
Poulain, P.-M.* 
Aidonindis, M.* 
Meyrat, J.* 
Ardhuin, F.* 
Fratianni, C.* 
Tonani, M.* 
Torrisi, L.* 
Pasquini, S.* 
Chiggiato, J.* 
Tudor, M.* 
Book, J.* 
Martin, P.* 
Allard, R.* 
Peggion, G.* 
Rixen, M.* 
TitleSueper-Ensemble techniques: application to surface drift prediction during the DART06 and MREA07 campaigns
Issue DateNov-2009
URIhttp://hdl.handle.net/2122/5856
Keywordssuper-ensemble, surface drift forecast
Subject Classification03. Hydrosphere::03.01. General::03.01.05. Operational oceanography 
AbstractThe prediction of the drift of floating objects is an important task, with applications such as marine transport, pollutant dispersion, and search-and-rescue activities. But forecasting surface drift is also very challenging, because it depends in a complex way on various interacting factors such as the wind, the ocean surface current, and the wave field. Furthermore, although each of the cited factors can be fore- casted by deterministic models, the latter all suffer from limitations, resulting in imperfect predictions. In the present study, we try and predict the drift of buoys launched during the DART06 (Dynamics of the Adriatic sea in Real-Time 2006) and MREA07 (Maritime Rapid Environmental Assessment 2007) sea trials, using the so-called hyper-ensemble technique: different models are combined in order to minimize departure from independent observations during a training period; the ob- tained combination is then used in forecasting mode. We review and try out different hyper-ensemble techniques, going from simple ensemble mean to techniques based on data assimilation, which dynamically update the model’s weights in the combi- nation when new observations become available. We show that the latter methods alleviate the need of fixing the training length a priori, as older information is au- tomatically discarded, and hence they lead to better results. Moreover, they allow to determine a characteristic time during which the model weights are more or less stable, which allows to predict how long the obtained combination will be valid in forecasting mode.
Appears in Collections:Manuscripts

Files in This Item:
File Description SizeFormat 
draft 02 hyper-ensemble__drift.pdf781.56 kBAdobe PDFView/Open
Show full item record

Page view(s)

97
checked on Apr 26, 2017

Download(s)

165
checked on Apr 26, 2017

Google ScholarTM

Check