February 2019

Last update: 08/03/2019

This month I will focus on two papers that came out in the same issue of Nature and that received wide media coverage. One is about the future of Antarctica, the other one about coupling the ice sheets with a climate model. Then we will have a quick look at the current situation of the Brunt ice shelf that is preparing the next big iceberg to break off from Antarctica.

The first one from Edwards and colleagues reassessed the model simulations from DeConto and Pollard (2016) using a statistical emulator to check the claim that hydrofracturing and ice cliff failure mechanisms are necessary to reproduce past sea level high stands of the Last Interglacial and the mid-Pliocene periods. They conclude that these processes, together called Marine Ice Cliff Instability (MICI), are not necessary and compute new probability distributions for the same ice sheet model that do not include MICI and that have much smaller expected values of sea level rise and much reduced uncertainties.

I think the use a statistical emulator to enhance the ensemble from a physical model and obtain results that allow a better statistical analysis is very useful. They criticise the normal assumption that we made in our sea level projections using DeConto and Pollard’s Antarctic contribution (Le Bars et al. 2017). Using the emulator they are able to show that the distribution of future Antarctic contribution is skewed with a tail towards large numbers. This is a fair point, I agree with their arguments. This has the impact of reducing the probability of small Antarctic contribution in the case where MICI is included in the model (see their Fig. 1).

However I also have two points of disagreement with the paper:

• The claim that MICI is "not necessary" to reproduce past sea level high stands is both not really true and not really useful. The uncertainty range about what could have been the contribution of Antarctica to sea level during the Pliocene is 5-20 m and during the Last Interglacial it is 3.6-7.4 m. DeConto and Pollard’s model without MICI can reproduce up to 6 m and 5.5 m respectively for these two period (see Edwards et al. E.D. Fig. 4). So yes it can reproduce the lower part of the ranges. But most of the Pliocene range cannot be reproduced with the no-MICI assumption. What the figure shows is that the model with MICI covers a much bigger par of the possible Antarctic contribution for these periods. And still, even including MICI, the model can only explain a maximum of 12 m contribution for the Pliocene. Which means additional mechanisms would be necessary to cover the whole range of possible Antarctic contribution for that period. The claim that MICI is “not necessary” is also not very useful practically because projections with MICI are used to make high-end sea level scenarios. The important information is then is it possible or not? If it was not possible then it would be good news and decision makers wouldn't need to take it into account. "Not necessary" only has an impact on low-end scenarios, for which MICI would already not be used anyways.
• See some Twitter reactions about that point:

• My other point of disagreement with the study is that the model with no MICI and an RCP2.6 scenario only has a mean contribution to sea level in 2100 of 0 cm (Edwards et al. table 2). Current pace would result in around 6 cm (IMBIE). Who would believe that if the earth warmed up by another degree Celsius Antarctica would suddenly accumulate mass? I would treat this scenario with a very low probability and in fact I think it shows that the model without MICI does not work.

The other paper about the future of the ice sheets in the same issue of Nature is by Golledge et al. They investigate the climate impact of ice sheet melt. This is a subject I quickly talked about in the November review after Bronselaer et al. (2018) was published. I am working on the Antarctic side of this issue at the moment (see my EGU2019 abstract). First let me explain why this is a timely study. Current state of the art (CMIP5 type) climate models do not include ice sheet models so the coupled effects between ice sheets and climate are a blind spot. In these climate models the ice sheets are just white mountains that do not change over time. They might have a snow layer on top of them but no ice. So snow falls on them accumulate a little bit and when it melts it is put in the nearest ocean grid box. If too much accumulates then it is put directly in the ocean to avoid infinite accumulation. What is missing is a model to transform the snow to ice and then transport it back to the sides of the ice sheet or to the ocean under the force of gravity. This is what ice sheet models do. Golledge et al. use the PISM ice sheet model for Greenland and Antarctica and couple them offline to LOVECLIM, an intermediate complexity climate model. Intermediate complexity means lower resolution and simpler physics compared to CMIP5 type climate models. It is the type of models generally used for long paleoclimate simulations.

What they find is that allowing feedbacks between the ice sheets and the climate model leads to strengthen both Antarctic and Greenland mass loss, by 100% and 30% respectively. For Antarctica this is not a surprise, although the magnitude is much bigger than I expected. Freshwater from the melting of ice leads to increase the ocean stratification, because it is is very light. This reduces vertical ocean mixing and as a result the surface of the ocean cools down while the subsurface warms up. Antarctica mostly looses mass from ice shelves basal melt and calving which is strengthened by warmer subsurface ocean temperature. For Greenland, it comes as a surprise to me that the feedback would increase the mass loss, because Greenland mostly looses mass from surface melt and a cooler atmosphere temperature would tend to reduce surface melt. Unfortunately the paper does not explain the mechanisms at play there (or did I miss it?).

There are a few issues with the ice sheet models that reduce my confidence in the projections. For Greenland the model is not able to reproduce the recent fast mass loss acceleration. Therefore the authors artificially impose the mass loss on the model in two ways: (1) decrease the friction between the ice and the bed (basal traction) to have a faster flow between 2000 and 2015 and (2) reduce the snowpack refreezing between 2000 and 2025. Refreezing is important for the mass balance because on ice sheets more than half of the snow that melts in the summer refreezes locally. It never reaches the ocean. Michiel van den Broeke had a similar comments in Trouw (in Dutch). You can force the model to agree with observations but if the model does not have the proper dynamics to explain observations there is no reason it is doing a good job for the future. For Antarctica, the model starts with enormous mass accumulation (1000 Gt/year in 1900) and accumulates mass until the 1980th. This is clearly not possible, such an accumulation would have been seen by tide gauge measurements. In fact as I said in the last review it is expected that Antarctica was slowly loosing mass in the 20th century. Also, the internal variability of grounded ice is so large in the model (Fig. 1a-d) that I do not understand what is going on physically (please let me know if you do).

In conclusion, the paper’s goal is important and it is the first time that two high resolution ice sheet models are coupled to a climate model. This is a big step in the right direction. However, I am not convinced by the results because of the issues mentioned above concerning the ice sheet models. Nevertheless, it is very instructive as it shows the long way that is left for ice sheet models to reach the level at which we can trust their future projections. I recommend the comment from Pattyn 2018, a very accessible and more optimistic paper about that issue .

Finally, I would like to point you towards two articles from Rolf Schuttenhelm (in Dutch). One about discussion on a plan B for the Netherlands, what do we do if we can’t/don’t want to protect against sea level rise anymore? And another one about the interpretation of sea level observations at the Dutch coast.

References:

Bronselaer, B., Winton, M., Griffies, S. M., Hurlin, W. J., Rodgers, K. B., Sergienko, O. V., … Russell, J. L. (2018). Change in future climate due to Antarctic meltwater. Nature. http://doi.org/10.1038/s41586-018-0712-z

Deconto, R. M., & Pollard, D. (2016). Contribution of Antarctica to past and future sea-level rise. Nature, 531(7596), 591–597. http://doi.org/10.1038/nature17145

Edwards, T. L., Brandon, M. A., Durand, G., Edwards, N. R., Golledge, N. R., Holden, P. B., … Wernecke, A. (2019). Revisiting Antarctic ice loss due to marine ice-cliff instability. Nature, 566(7742), 58–64. http://doi.org/10.1038/s41586-019-0901-4

Golledge, N. R., Keller, E. D., Gomez, N., Naughten, K. A., Bernales, J., Trusel, L. D., & Edwards, T. L. (2019). Global environmental consequences of twenty-first-century ice-sheet melt. Nature, 566(7742), 65–72. http://doi.org/10.1038/s41586-019-0889-9

IMBIE. (2018). Mass balance of the Antarctic Ice Sheet from 1992 to 2017. Nature, 558(7709), 219–222. http://doi.org/10.1038/s41586-018-0179-y

Le Bars, D., Drijfhout, S., & de Vries, H. (2017). A high-end sea level rise probabilistic projection including rapid Antarctic ice sheet mass loss. Environmental Research Letters, 12(4), 044013. http://doi.org/10.1088/1748-9326/aa6512

Pattyn, F. (2018). The paradigm shift in Antarctic ice sheet modelling. Nature Communications, 9(1), 2728. http://doi.org/10.1038/s41467-018-05003-z