 |
| Titel |
Improving ocean forecasts with gliders: a non-Markovian game |
| VerfasserIn |
Christopher Hughes, David A. Smeed, Peter Challenor, Kevin Oliver |
| Konferenz |
EGU General Assembly 2013
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 15 (2013) |
| Datensatznummer |
250084751
|
|
|
|
|
|
| Zusammenfassung |
| Throughout much of the ocean observations are sparse and ocean forecast
models are under-constrained, which reduces their prediction skill.
Underwater gliders can help improve predictive capability by making many
observations over long periods at relatively low cost, whilst also
allowing retargeting. Deciding where to send the gliders is, however,
difficult, because strong covariances between observations make the reward
for any given measurement dependent on past choices. Here, the issue is
framed as a decision problem with non-Markovian rewards. A probabilistic
solution method is developed which is then tested on an idealised problem.
Finally, ways in which this method is being extended for use in an
operational context are discussed. |
|
|
|
|
|
|