Day 4 at AGU – Productive Self-Doubt and Healthy Retraction

An eventful day.

Ran into Andy Dessler in the halls. (Recognizing people in a crowd is hard for me. I am quite prosopagnostic, though recognizing Dr Alley was something I could manage.) I told him I was looking for Dr. Trenberth, and he told me he’d be chairing a session in the immediately next time slot!

It was a particularly excellent session. I hope I’ll find time to follow some of the ideas presented there.

I’m afraid I was quite rude in intruding on Kevin Trenberth’s day. After all, I did not have much to offer him over his genuine colleagues and peers. But getting 5 minutes from Trenberth was all I asked for.

In short, my idea was that the current Arctic warmth could simply be accounted for by advection of anomalous heat in the upper ocean. I have enormous respect for Dr. Trenberth, going back to the days when I frequently had occasion to cite his papers. Because of the nature of the question and the nature of his expertise, I was fairly certain he’d say either “of course” or “of course not”.

I need to add that Simon Donner had earlier argued “I really don’t think so” on this very question. (On a Skype call IIRC)

And I understand why. Despite its explanatory power, I have had real doubts since discussing this idea with Simon. The time scale actually, is too short, which alone is enough to settle it. Trenberth was categorically negative. So no. My theory is bollocks. Please ignore it.

Whatever is warming the Arctic, the extra energy enters the Arctic basin from the sky, not from the ocean in any direct way. It’s irreducibly more complicated than that.


The scientist in me is thrilled by this lovely complexity. The explainer in me is a bit dismayed.


The theme of the day, then, is healthy retraction.

Prior to Trenberth’s excellent session, some of which went buy pretty quickly for me, I attended a session on Antarctic ice. I was largely goaded into this by trickster Steve McIntyre on Twitter.

More about tricksters another time. I did get trolled into a session called “Assessing the Stability of the Antarctic Ice Sheets and their Contribution toward Global Sea Level. (C33D)” and I have to confess myself grateful to the trickster for it.

There was much to think about in this session as well, but a pair of papers stood out. I want to focus on the related presentations by Pollard and Deconto.

This pair of talks did much for me to elucidate Dr. Alley’s comment on Sunday, to the effect that new results are actually making it harder to exclude the possibility of large, abrupt sea level rise, on the order of several meters in a single lifetime.

The punch line is that their best estimate is that, on current lack-of-policy, rapid sea level rise will commence around 2060, so it will affect the lives of people now living. Also that there is still, more likely than not, time to prevent this catastrophe.

Let me walk you through the argument as I understood it. There may be problems with this summary. It’s just my first exposure to some of these idea. Please don’t cite it as authoritative. If there’s enough interest I’ll take it up further with the authors.


As you may know, the slope of the Amundsen Bay is backward, going deeper as you approach the great mass of the Antarctic Ice Sheet. As a consequence, glaciers emptying the ice sheet into the bay are susceptible to catastrophic retreat. I had attributed this idea to C Schoof, who explained it in an admirable paper around 2000, but apparently (this via McIntyre) there was much earlier discussion of this idea. This is old news. And it has been modeled somewhat successfully.

This is basically what the Schoof model looks like:

Naturally, the thing to do is to test this model against paleo evidence if there is some.

And indeed it has emerged that paleo records have detected a Milankovic signal in sea level.

So all it takes is some surface reconstructions and some flux data from a crude GCM to provide a test for a continental scale model of ice sheets that incorporate basic physics and include the Schoof mechanism. Now we have an objective test.

The result was not entirely satisfactory. The oscillations in the paleo record were TOO LARGE for the model to account for. The model was not “tippy” enough; the catastrophic (in the technical sense of catastrophe) mechanism wasn’t fast enough to replicate the sea level changes.

However, in situ observations of ice cliffs have revealed that the process of retreat is partially mechanical when the cliffs get high enough. The ice front gets high enough that gravity starts helping; the entire structure fails and retreats. Geologically, it turns out that a “glacial pace” is astonishingly rapid.

The new phenomenology at the ice sheet edge looks like this:


It turns out that this addition is sufficient to capture the paleo-Milankovic sea level oscillation!

Then these guys do the right thing and do a formal Bayesian tuning of model parameters to paleo-obs, and have a good claim to getting the problem first-order right for the first time. And things look pretty solid.

Which in turn strongly indicates┬áthat Hansen’s much-maligned sense of it is in fact correct – large ice sheets can collapse quite quickly. (I went with the crowd in dismissing that idea. Oops.)


That was the gist of Pollard’s talk.

So what does this mean for the future? Deconto picked up the thread.

Basically if we continue on Plan Trump, we get abrupt sea level rise kicking in around 2060 or so, and continuing on to perhaps tens of meters, disrupting coastal processes for many centuries. I hope you like that sort of thing. This is just science, we’re not here to make value judgments.

Anyway, as someone who is trying to communicate not just the content of science but its process, I think it’s important to highlight this slide of Deconto’s.

It summarizes the bad news but also asks a key question. “How might we be wrong?” I would venture to say that this is the core discipline of science. Self-doubt is frowned upon in our modern, brazen society. I wonder if this isn’t one reason that science’s star is in decline.

Self doubt is so intrinsic to commercial software development, and probably to other forms of engineering, that a test-first approach is becoming prominent. Write the code first, then write the test. It’s a luxury not afforded science.

This slide is a good embodiment of Popperian reasoning. The scientists have come up with a result. (In this case it is an alarming result, but that should not affect the process.) Now they turn around and play devil’s advocate. How might we be wrong?

Unfortunately, or fortunately, depending on how you choose to look at it, the proposed reasons the model might be incomplete and over-sensitive seem to me not very compelling; they seem too small to fight the gravitational instability of a 60 meter ice cliff. Maybe someone can suggest another missing mechanism to slow it down, but remember, the paleo evidence suggests speeding it up compared to the basic Schoof model.

The key to advancing your scientific conclusion is to be your own harshest critic. As a beginner, you will find yourself wrong far more often than right. But you will be wrong in ever more interesting ways. And eventually, perhaps, you’ll add to the state of knowledge.

When I see a slide like this, where the presenters question their own result and look for ways they might be wrong, my confidence in their result increases.

The culture at large, outside science has this backwards. The more adamant a person is, the more reliable they are perceived. It’s pretty tragic.


Ended the day with an entertaining dinner with another trickster. He has an amusingly airtight ethic that basically offers plenty of cited philosophical jargon that basically says “I do what I want, see?” Well, sure. But I think he, as well as McIntyre, could benefit from thoughtful reappraisal of what it is he wants.


  1. I've been following AGU from the other side of the pond, with the emphasis on the Arctic. For further details see:

    I don't suppose you also asked Kevin about atmospheric advection of heat and moisture did you?

  2. That's a stunning discovery about rapid sea level rise. The mechanism is similar to, but much more rapid than the mechanism that created the basin and range geologic province. Millions of years of compressive tectonics created a very thick crust from the Sierra Nevada to the Rocky mountains. When the element of the east Pacific rise reached the trench and ended subduction gravitational processes began the process of regional extension and the mountain blocks slowly but surely pulled apart.

    Antarctica's glaciers could fall like dominoes as the ice cliffs break off one after another. This theory fits Hansen's data about rapid SLR from the end of the last interglacial.


  3. I really was intruding on Trenberth's plans, so no, I just wanted a yea or nay on my hypothesis, and it was an unequivocal nay.

    I hope he'll be willing to continue conversing via email, because many of the things he's said about attribution will be featured in my (I hate this word, but anyway) narrative about attribution.

    But clearly the story is more complicated than I imagined. And clearly the excess heat is being shuffled around the system through the atmosphere. At this point I am very fuzzy about the details. I don't know if anyone has a clear picture of what is happening yet.

  4. Thanks for a report of the science at the AGU.

    Here is what is happening to the Arctic sea ice now. (Best watched in full screen and with the muted sound switched on.)

    The Arctic sea ice is melting because of warmer seas entering from the Atlantic, and the Pacicfic water is warming the Beaufort Gyre. It is also being warmed by solar radiation now that the ice is thinner and retreats earlier in the summer (ice albedo effect.) But the main driver is the increased back radiation caused by the increase in greenhouse gases, which may include methane from melting clathrates.

    As you say it is much more complicated that warming oceans. But the reason that the models are failing to reproduce what we see that they are based on an approximation of the radiative transfer equation which does not apply to the boundary layer where the ice is melting. The approximation is based on classical thermodynamics, but the greenhouse effect is quantum mechanical (radiation converted to heat.)

  5. Pingback: Altri dati (da negare) - Ocasapiens - Blog -

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.