Feedback between ice dynamics and bedrock deformation with 3D viscosity in Antarctica
- 1Faculty of Aerospace Engineering, Delft University of Technology, Delft, Netherlands (w.vanderwal@tudelft.nl)
- 2Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, Netherlands
- 3Earth and Climate Cluster, Faculty of Science, Vrije Universiteit Amsterdam, Amsterdam, Netherlands
Over glacial-interglacial cycles, the evolution of an ice sheet is influenced by Glacial isostatic adjustment (GIA) via two negative feedback loops. Firstly, vertical bedrock deformation due to a changing ice load alters ice-sheet surface elevation. For example, an increasing ice load leads to a lower bedrock elevation that lowers ice-sheet surface elevation. This will increase surface melting of the ice sheet, following an increase of atmospheric temperature at lower elevations. Secondly, bedrock deformation will change the height of the grounding line of the ice sheet. For example, a lowering bedrock height following ice-sheet advance increases the melt due to ocean water that in turn leads to a retreat of the grounding line and a slow-down of ice-sheet advance.
GIA is mainly determined by the viscosity of the interior of the solid Earth which is radially and laterally varying. Underneath the Antarctic ice sheet, there are relatively low viscosities in West Antarctica and higher viscosities in East Antarctica, in turn affecting the response time of the above mentioned feedbacks. However, most ice-dynamical models do not consider the lateral variations of the viscosity in the GIA feedback loops when simulating the evolution of the Antarctic ice sheet. The method developed by Gomez et al. (2018) includes the feedback between GIA and ice-sheet evolution and alternates between simulations of the two models where each simulation covers the full time period. We presents a different method to couple ANICE, a 3-D ice-sheet model, to a 3-D GIA finite element model. In this method the model computations alternates between the ice-sheet and GIA model until convergence of the result occurs at each timestep. We simulate the evolution of the Antarctic ice sheet from 120 000 years ago to the present. The results of the coupled simulation will be discussed and compared to results of the uncoupled ice-sheet model (using an ELRA GIA model) and the method developed by Gomez et al. (2018).
How to cite: van der Wal, W., van Calcar, C., de Boer, B., and Blank, B.: Feedback between ice dynamics and bedrock deformation with 3D viscosity in Antarctica, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19278, https://doi.org/10.5194/egusphere-egu2020-19278, 2020
Comments on the display material
AC: Author Comment | CC: Community Comment | Report abuse
Hello Caroline,
Thanks for posting these interesting results. I wonder why you chose a 5kr coupling time step? Maybe if you decrease this time step, you'll need to perform fewer iterations before convergence?
Regards, Lennert
Hello Lennert,
Thank you for your interest and your question.
Ideally, the difference in deformation and ice thickness for the whole region between the last iteration and the iteration before converge to zero. However, at some places around the grounding line and calving line, the total change in ice tickness within one timestep of 5000 years can be 2500 meters per year (due to strong forcing). At these places, the difference in deformation between iterations does not converge to zero but to an alternating positive and negative constant. To compute if the deformation alternates, a minimum of 3 iterations is reguired. We chose a timestep of 5000 years because the model converges mostly in 3 iterations within the uncertainty range of 0 to 10 meters of deformation within one timestep.
Decreasing the timestep could decrease the uncertainty because the total change in ice thickness within one timestep will be lower. In this case, it is more likely that the model converges to zero. We have tested timesteps of 1000 years but we could still see alternating convergence instead of convergence to zero difference so 3 iterations were still reguired. A dynamic timestep dependent on the change in ice tickness could speed up the model but we have not tested that yet.
I realise my answer is very technical. Please let me know if anything is unclear or if you have any more questions/suggestions. I could upload a figure to show the alternating convergence process.
Kind regards,
Caroline
Thanks for this explanation. In light of this alternating convergence, I understand why you save the mean deformation from the last two iterations. Just one follow-up question: do you also save the mean ice model output from the last two iterations, as your final result for that timestep, or just the last one?
Yes that is exactly the reason. After the mean deformation is calculated, a final ice model run is done using the mean deformation to compute final ice model output for that timestep.