Skip to content

When you choose to publish with PLOS, your research makes an impact. Make your work accessible to all, without restrictions, and accelerate scientific discovery with options like preprints and published peer review that make your work more Open.

PLOS BLOGS All Models Are Wrong

Place your bets: sea level rise from Antarctic ice sheet collapse

I have a few bits of happy news!

The BBC’s program Climate Change by Numbers on which I was one of the main scientific consultants has won the AAAS Science Journalism Award in the “in depth” (more than 20 minutes) category. Apparently judge Kathy Sawyer, a freelance science writer, called the program “a master class in how to make a forbidding statistical story both enlightening and entertaining”. I’m very happy and proud to have been involved, and send my huge congratulations to the BBC team who worked so hard to explain the headline science results of the IPCC assessment in a way that was fresh and  accurate.

Second, I was interviewed in the first part of a BBC Radio 4 series “Changing Climate” leading up to the Paris discussions, which talks about climate science (including “lukewarmer” climate sceptics like Matt Ridley). I’m not in the next two parts, which are about policy and technology.

Third, we’ve advertised a PhD project, “Understanding uncertainty in the weak underbelly of Antarctica”. My first PhD student! We have put together a great supervisory team with expertise in physics, statistics, polar observations and science communication. Do take a look if you’re interested in researching climate change impacts and uncertainty assessment.

Fourth, I’m the joint first author of a paper published in Nature today. We predict the consequences of sea level rise in the event of Antarctic ice sheet instability – it’s the culmination of several years of work assessing the uncertainty of an ice sheet model. Broadly speaking, Catherine Ritz led the ice sheet modelling and I led the statistical analysis including the comparison of the model with observations.

Predicted probability of grounding line retreat by 2200 (Ritz et al., Nature, 2015), overlaid on bedrock topography map (Le Brocq et al., 2010).
Predicted probability of grounding line retreat by 2200 (Ritz et al., Nature, 2015), overlaid on bedrock topography map (Le Brocq et al., 2010).
Ritz_et_al_legend
Legend for the figure.

Here’s an article I wrote about it for the Guardian – please take a look.

The map is a version of one of our figures. It shows our prediction for the probability of the edge of the ice sheet (the grounding line) retreating by the year 2200. It’s overlaid on a map of the bedrock topography, where blue areas are below sea level – these are the parts thought vulnerable to this “marine ice sheet instability”. The two glaciers shown in dark purple/black colours on the map (highest probability of retreat) are Pine Island Glacier and Thwaites Glacier in the Amundsen Sea Embayment.

The featured image above shows scientists of the NERC iStar project measuring the bedrock topography under Pine Island Glacier: this is extremely important, as I’ve written about before.

I haven’t had a chance to write a blog post here discussing the methods and its strengths and limitations in more detail, but I’m happy to try and answer questions below. For those that want the really technical stuff, we put all the details of the methods in the Supplementary Information which is freely available online.

Enjoy!

Tamsin

N.B. Edited broken link

Discussion
  1. Did you back-test your methodology on known collapses such as the Arctic and Greenland and various glaciers worldwide? What were the results of that?

  2. Hi, thanks.

    No, as we were looking at quite a specific mechanism of collapse that applies to marine ice sheets (i.e. lying on bedrock below sea level), particularly West Antarctica.

  3. Congratulations on the Nature paper. Looks to be getting a lot of coverage. It is so encouraging to see competent, scientifically literate analysis and modelling without all the press-grabbing hyperbole. This was clearly a painstaking enterprise and not just rushed out the door to get another headline. This type of scholarship deserve more attention and takes us a small way down the path of restoring the reputation of science (especially climate science) as a reliable source of learning and guidance.

  4. Hello Dr. Edwards,

    Firstly, I found this blog via your article in The Guardian, and am happy to find another place to see what a climate scientist thinks; I’m always interested in the perspective of scientists.

    In regards to your recent paper, I have a fairly specific question regarding the model. I recently watched a video of Richard Alley in which he mentions ice cliff failure, and was initially going to ask if anything similar was included in the model you used, then I searched your supplementary material (I should have thought of that; thanks for the reminder) and found the relevant information.

    It mentions that you did not directly use a calving law, but that you instead used a ‘horizontal wastage rate’ parameterization, with specific reference to the Pollard, Deconto, and Alley (2015) paper. My first question is just some quick clarification: was any direct physics used to model ice cliff failure, or just a parameter for how fast a retreat it could cause? If the latter, was said parameter simply determined by comparison to previous observation, or was there deeper analysis of the effect of cliff height on potential failure induced retreat rate? My main reason for asking is because we don’t really *have* any modern observations comparable to the kind of ice cliffs we could see at Thwaites, i.e., we’ve never seen what happens when you get a cliff 200m or 300m or more above water, so comparison to observation… well, it seems like the usefulness in this specific case would be rather limited. Please correct me if I’m missing something.

    Following my first question, the supplementary material notes that you achieve the same order of magnitude of retreat rates as the Pollard paper, so does the parameterization basically produce similar results once there are high cliffs in deep bedrock? How close is “the same order of magnitude of rates”, and is it a slower or faster retreat? A factor of 1.1 or 9 are both technically within an order of magnitude, but the latter is still pretty darn big, haha :D. How much confidence do you have that that part of the model is sufficiently accurate, or do you think it’s potentially a noteworthy weakness?

    How much of a contribution does the ice cliff failure make to retreat over the periods out to 2100 and 2200 as simulated in the model? If you can’t give a specific answer, just an educated guess would still be appreciated. The main figures of the study depicting retreat probabilities seems to show that out to 2100 the 50% probability retreat is not all that far inland (relatively speaking; ~100km is still a lot), and could remain in (relatively) shallow bedrock, though it might retreat much further and deeper. However, out to 2200 the retreat starts getting into some pretty deep areas with fairly high probability. It would seem that ice cliff failure could start making a rather large contribution at that point. How likely do you think it is that more accurate modelling of ice cliffs would increase the retreat rate estimates, or do you think that the estimates would stay mostly the same out to 2100 since it may not retreat enough for the cliffs to be terribly high or the bed particularly deep?

    I realize this is a rather long and in depth set of questions for a blog. I also understand that your time is valuable and limited, so if you can only give brief answers, or if you would prefer to answer by emailing me at the address provided, that’s perfectly acceptable. I do look forward to your response to my inquiries quite a bit. I’m a physics undergraduate student, and am actually considering moving into climate science – I specifically find the effects on ice (whether sea ice or ice sheet) very fascinating. Thanks for your time and best regards.

    Sincerely,
    Michael Tilbrook

  5. I very much like your header “all models are wrong but some are useful.” It is unfortunate that more people do not understand the limitations of models, particularly regarding climate change.
    With that in mind, I find your new paper in Nature very useful, but am concerned that most will not really notice the phrases about the uncertainties. The final sentence in the Abstract carefully notes that: “…upper bound estimates…are implausible under current understanding of physical mechanisms and potential triggers.”
    It is easy to dismiss the key issue, however, which is our present inability to quantify those mechanisms and triggers, particularly in light of the warming now being an order of magnitude greater than in the geologic record and the varied evidence of non-linear events.
    I realize you likely would not have written the sub-head in the Guardian (UK) about your work, but it reflects the impression that we have ruled out extreme sea level rise (SLR) this century. It says “The results of our study might be surprising to some. But although it rules out very high rises, climate skeptics certainly shouldn’t be dancing in the aisles.”
    I think it is misleading to perpetuate the belief that we have “ruled out” sea level higher than a meter or two by 2100. For example, few realize that the IPCC methodology yields only an inch of SLR contribution from Antarctica in the most extreme case (RCP 8.5) not because they don’t think its possible, but rather because they can’t quantify the number to within at least one standard deviation.
    We routinely build structures and infrastructure that lasts beyond a century. We need to start planning for a full range of SLR scenarios, just as we allow for worst case scenario planning with hurricanes, earthquakes and tsunamis, particularly since SLR will be universal in impact and last for centuries or millennia.
    Good scientific work like yours might help if it elaborated on the uncertainties a bit more, knowing that this will get into the popular press. Most readers skip over those nuances. Not being able to quantify a risk or the mechanism does not mean it is not real.

  6. Gradually physicists are acknowledging greenhouse radiative forcing is the wrong paradigm altogether.

    I know this is hard to take, but we only have to consider the planet Uranus to understand why the whole paradigm that, according to the IPCC, radiation from the colder atmosphere is supposedly producing identical radiation to solar radiation, so, they claim, it can just be added to the solar flux and then the total of the two (a doubled humped distribution of course) can supposedly be entered into Stefan Boltzmann calculators and, bingo, we have 390W/m^2 (after deducting non-radiative cooling) and the blackbody temperature for that is 288K which is about 15°C.

    It’s all too easy. Or is it? The problem is, there is no solar radiation reaching the base of the 350Km high nominal troposphere of Uranus, and no solid surface either, but it is hotter than Earth’s surface down there. Obviously radiation from the cooler regions above cannot heat the warmer regions at the base of that troposphere, and there is no convincing evidence of overall planetary cooling. What really happens will blow your mind and you will be as skeptical as hell! But you will not be able to prove the physics wrong at http://climate-change-theory.com and in the linked papers, blogs, videos or in my book “Why It’s Not Carbon Dioxide After All” on Amazon.

Leave a Reply

Your email address will not be published. Required fields are marked *


Add your ORCID here. (e.g. 0000-0002-7299-680X)

Back to top