Super-ensemble techniques: Application to surface drift prediction
Vandenbulcke, L.; Beckers, J.M.; Lenartz, F.; Barth, A.; Poulain, P.M.; Aidonidis, M.; Meyrat, J.; Ardhuin, F.; Tonani, M.; Fratianni, C.; Torrisi, L.; Pallela, D.; Chiggiato, J.; Tudor, M.; Book, J.W.; Martin, P.; Peggion, G.; Rixen, M. (2009). Super-ensemble techniques: Application to surface drift prediction. Prog. Oceanogr. 82(3): 149-167. dx.doi.org/10.1016/j.pocean.2009.06.002
In: Progress in Oceanography. Pergamon: Oxford,New York,. ISSN 0079-6611; e-ISSN 1873-4472, more
|  |
| Keyword |
|
| Author keywords |
Super-ensemble; Multi-model; Surface drift |
| Authors | | Top |
- Vandenbulcke, L., more
- Beckers, J.M., more
- Lenartz, F., more
- Barth, A., more
- Poulain, P.M.
- Aidonidis, M.
|
- Meyrat, J.
- Ardhuin, F.
- Tonani, M.
- Fratianni, C.
- Torrisi, L.
- Pallela, D.
|
- Chiggiato, J.
- Tudor, M.
- Book, J.W.
- Martin, P.
- Peggion, G.
- Rixen, M.
|
| Abstract |
The prediction of surface drift of floating objects is an important task, with applications such as marine transport, pollutant dispersion, and search-and-rescue activities. But forecasting even the drift of surface waters is very challenging, because it depends on complex interactions of currents driven by the wind, the wave field and the general prevailing circulation. Furthermore, although each of those can be forecasted by deterministic models, the latter all suffer from limitations, resulting in imperfect predictions. In the present study, we try and predict the drift of two 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 obtained combination is then used in forecasting mode. We review and try out different hyper-ensemble techniques, such as the simple ensemble mean, least-squares weighted linear combinations, and techniques based on data assimilation, which dynamically update the model's weights in the combination 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 automatically discarded.
When the forecast period is relatively short (12 h), the discussed methods lead to much smaller forecasting errors compared with individual models (at least three times smaller), with the dynamic methods leading to the best results. When many models are available, errors can be further reduced by removing colinearities between them by performing a principal component analysis. At the same time, this reduces the amount of weights to be determined.
In complex environments when meso- and smaller scale eddy activity is strong, such as the Ligurian Sea, the skill of individual models may vary over time periods smaller than the forecasting period (e.g. when the latter is 36 h). In these cases, a simpler method such as a fixed linear combination or a simple ensemble mean may lead to the smallest forecast errors. In environments where surface currents have strong mean-kinetic energies (e.g. the Western Adriatic Current), dynamic methods can be particularly successful in predicting the drift of surface waters. In any case, the dynamic hyper-ensemble methods allow to estimate a characteristic time during which the model weights are more or less stable, which allows predicting how long the obtained combination will be valid in forecasting mode, and hence to choose which hyper-ensemble method one should use. |
|
