Transient minimization in data assimilation: examining the effect of plausible initial conditions and model forcings on the evolution of the Greenland ice sheet
Abstract
Lengthy model initialization or spin-up procedures designed to eliminate spurious transient signals in a modelled system are often undesirable or impractical for a variety of reasons, such as excessive computational cost or the requirement that the model state at the starting time correspond exactly to certain data. Data assimilation procedures may exacerbate these transients by forcing a model to conform to data sets that are incompatible with model physics. In the case of dynamical models of ice physics, a common case of this incompatibility occurs when a measured surface velocity field is used to estimate bed friction, and in so doing generates non-physical rates of change in the ice surface elevation. In order to overcome this issue without accepting the disadvantages noted above, we use an expanded data assimilation procedure that not only minimizes misfit in modelled and observed surface velocities, but also explores uncertainties in a variety of input data such as surface mass balance, ice thickness, and surface velocities, such that spurious transient signals are minimized and lengthy spin-up procedures are rendered unnecessary. This type of initialization technique employs several control variables and necessarily produces non-unique solutions. We apply regularization based on existing patterns within and a priori assumptions about the behavior of input data sets in order to reduce the number of possible initial model states. We implement this technique in the ice sheet model VarGlaS, and show the result of an initially transient-free simulation of the Greenland Ice Sheet that uses contemporary data within specified error bounds for the generation of initial conditions and model forcings. We carry this simulation of Greenland 500 years into the future and examine the effect of using various transient free and error bound respecting configurations of input data.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2013
- Bibcode:
- 2013AGUFMDI31A2194B
- Keywords:
-
- 0798 CRYOSPHERE Modeling;
- 0726 CRYOSPHERE Ice sheets