Machine Learning for Tactical Iceberg Drift Forecasting
Iceberg drift forecasting on tactical scales remains challenging due to the lack of accurate and timely data on ocean currents, winds, iceberg geometry and mass. Current operational models are mechanistic and require a set of constants to be additionally determined for each iceberg. Alternatively, statistical methods and dead reckoning have demonstrated greater performance up to the first 36 hours of forecasting. On the other hand, purely statistical models may not be able to perform in rare outlying cases. This study describes a neural network applied to iceberg drift forecasting, that is essentially a statistical approach. The network is trained and tested using iceberg drift data recorded during exploratory drilling offshore Labrador in 1979. Initial drift track parts, ocean currents and winds are used to train the network, and then the model is used to forecast 24 hours ahead. The model performance is promising and potentially can be improved even further given more data such as more accurate currents and winds, or additional inputs, for example, information about waves or sea surface gradients.
- Record URL:
-
Availability:
- Find a library where document is available. Order URL: http://worldcat.org/issn/03766756
-
Supplemental Notes:
- Abstract reprinted with permission of the International Conference on Port and Ocean Engineering Under Arctic Conditions (POAC).
-
Authors:
- Yulmetov, Renat
- Ralph, Freeman
-
Conference:
- 25th International Conference on Port and Ocean Engineering under Arctic Conditions (POAC 2019)
- Location: Delft , Netherlands
- Date: 2019-6-9 to 2019-6-19
- Publication Date: 2019
Language
- English
Media Info
- Media Type: Digital/other
- Features: Figures; References;
- Pagination: 9p
- Monograph Title: POAC 2019: Proceedings of the 25th International Conference on Port and Ocean Engineering under Arctic Conditions
-
Serial:
- Proceedings of the International Conference on Port and Ocean Engineering Under Arctic Conditions
- Publisher: Port and Ocean Engineering under Arctic Conditions (POAC)
- ISSN: 0376-6756
- EISSN: 2077-7841
- Serial URL: http://www.poac.com/PapersOnline.html
Subject/Index Terms
- TRT Terms: Drift ice; Forecasting; Icebergs; Machine learning; Neural networks; Ocean currents; Seawater; Statistical analysis; Wind
- Geographic Terms: Labrador Sea
- Subject Areas: Environment; Marine Transportation;
Filing Info
- Accession Number: 01745122
- Record Type: Publication
- Files: TRIS
- Created Date: Jul 15 2020 9:12AM