Improving the representation of ice-sheets current state in ice-flow forecast models

Titre :Improving the representation of ice-sheets current state in ice-flow forecast models

Laboratoire de rattachement : LGGE

Encadrant : Fabien GILLET-CHAULET
Co-Encadrant : Gael Durand

Téléphone :

Mots clés : mass loss of ice sheets, Greenland, Antarctica, ice-flow models, .

Contexte et objectifs de la mission de stage :

The currently observed acceleration of mass loss from the two ice-sheets (Greenland and Antarctica) is a concern when considering their possible contribution to future sea-level rise. A fair portion of this acceleration is attributable to ice dynamics, where acceleration and thinning of numerous outlet glaciers have be shown to be responsible for a substantial increase in ice discharge to the ocean. Realistic projections of the ice-sheets contribution to sea-level rise on decadal to century time-scales must be derived from the forecasts of verified ice-flow models. As ice-sheet responses include long time-scales (multi-century), forecasting change on decadal-to-century time-scales is essentially a short-term forecast and thus simulating the current conditions is crucial. Fulfilling this essential prerequisite is difficult as the processes at the root of the observed acceleration and retreat are still not fully understood. Constraining these processes require the use of robust inverse methods. Such techniques are currently under development within the ANR project ADAGe (http://www-lgge.ujf-, a collaborative project between Glaciologists (LGGE, LEGOS) and Mathematicians (IMT, LJK).
The objective of the proposed internship is, first to validate the methods developed during ADAGe and second to conduct data assimilation experiments on real applications to constrain the processes in- situ and build the optimal initial state. Depending on the application several sofwares are available :
- Approximated inverse methods have been implemented in the finite element ice flow-model Elmer/Ice and applied on large 3D simulations of the whole Greenland and Antarctic ice-sheets.
- The recently developed full-Stokes finite element ice flow model DassFlow/Ice (IMT, includes the exact adjoint code and full 4d-var data assimilation process in which the control variables are the boundary conditions and the ice rheology parameters. DassFlow/Ice is currently a flow line model but will be extended to 3D.

Prérequis : The candidate must have strong background in modelling and programming and a real interest in conducting geophysical simulations. This subject is susceptible to be pursued by a 3-year Phd starting in Sept 2013 (according to sufficiently good results of the candidate).