|
|
Vortex Extraction of Lagrangian Time Series
|
When
|
2.00pm on Friday 13th March 2009
|
|
Where
|
B617, Leverhulme Library, Columbia House
|
|
Presentations
|
|
|
Speaker
|
Sofia Olhede
|
|
From
|
UCL
|
|
Abstract
|
Lagrangian observations of oceanic turbulence are measured by subsurface floats, where each float is recording a time series of observations. The high frequency structure of such data has been understudied due to difficulties in extracting useful information from available measurements. Part of the high frequency structure is information on vortices present at many locations in the oceans.These vortices, or coherent eddies, constrained to have nearly horizontal flow by the Earth's rotation of the earth and by the density stratification of the ocean, consist of swirling currents surrounding localized mass anomalies. Such structures are important both for their transport of properties such as heat, salt, momentum, and biomass, as well for their role stirring surrounding waters and thereby accelerating dissipation. Large sets of these time series should be studied as a collection of observations measuring the same latent factors. By modelling the properties of the Lagrangian observations mechanistically in an unobserved components model, inference for different features of the time series can be implemented separately, and the problem simplified. For example the compression of vortex structures in a suitable representation, can be used to isolate these features from other components present in the data. By using time-frequency tools, we demonstrate that the generating mechanism of each component can be inferred, vortex structures detected and their structural properties automatically characterised.
This is joint work with Jonathan Lilly, ESR, Seattle, WA
|
|
For further information
|
Postgraduate Administrator Ext. 6879
Department of Statistics, Columbia House
|
|
|