Stochastic flow modeling : Quasi-Geostrophy, Taylor state and torsional wave excitation

Nicolas Gillet, D. Jault, Chris Finlay

    Research output: Contribution to conferenceConference abstract for conferenceResearchpeer-review


    We reconstruct the core flow evolution over the period 1840-2010 under the quasi-geostrophic assumption, from the stochastic magnetic field model COV-OBS and its full model error covariance matrix. We make use of a prior information on the flow temporal power spectrum compatible with that of observed geomagnetic series. We account for errors of representativeness (subgrid processes associated with the unresolved field at small length-scales) that are correlated in space and time, using an iterative scheme. An ensemble approach allows us to measure the uncertainties within the recovered motions. Large length-scales flow features are naturally dominated by their equatorially symmetric component from about 1900 when the symmetry constraint is relaxed. Equipartition of the kinetic energy in both symmetries coincides with the poor prediction of decadal length-of-day changes in the XIXth century. We interpret this as an evidence for quasi-geostrophic rapid flow changes, and the consequence of a too loose data constraint during the oldest period. We manage to retrieve rapid flow changes over the past 60 yrs, and in particular modulated torsional waves predicting correctly interannual length-of day variations from 1950 onward. We propose a triggering mechanism for these waves involving non-zonal motions in the framework of Taylor's state.
    Original languageEnglish
    Publication date2013
    Number of pages1
    Publication statusPublished - 2013
    EventAGU Fall Meeting 2013 - San Francisco, United States
    Duration: 9 Dec 201313 Dec 2013


    ConferenceAGU Fall Meeting 2013
    Country/TerritoryUnited States
    CitySan Francisco


    Dive into the research topics of 'Stochastic flow modeling : Quasi-Geostrophy, Taylor state and torsional wave excitation'. Together they form a unique fingerprint.

    Cite this