Yonatan Zunger (zunger) wrote in climatepapers,
Yonatan Zunger

Hansen et al., continued!

OK, it's time for the long-promised second part of Hansen, et al.. And this one is exciting, because it contains their projections of the climate model into the future!

It's possibly even more exciting because this part analyzes several scenarios for the future, all of which are achievable using only present-day technology but differ only in policy -- but the results range from "everything is OK" to, frankly, apocalyptic.

To recap, last time we looked at sections 1-5 of this paper, which is trying to be the new "gold standard" climate model. The first several sections detail where they get their input parameters, and their calibration runs for the period 1880-2003. (A separate paper details how the model actually works) The short version of the summary is that the model seems very good for overall global averages, less good for regional averages, and this is primarily because of known issues with the ocean model (especially that it doesn't handle El Niño) and the sea ice model.

Starting with section 6, they run this simulation into the future, using five different scenarios for gas emissions: three of the International Panel on Climate Change [IPCC]'s standard scenarios, and two "alternate" scenarios. The IPCC scenarios are described in full detail here.

  1. IPCC scenario A1: A "business-as-usual" scenario of a world of "rapid and successful" economic development, with high levels of international trade and commerce, and a strong committment to market-based solutions; (there are two variants here, A1B and A1F1)

  2. IPCC scenario A2: A world that is overall economically successful, but more consolidated into regions and less prone to international trade in people;

  3. IPCC scenario B1: A world where there is a "high level of social and environmental consciousness" but no systematic climate-policy changes;

  4. Alternative scenario: A world where there are concerted global efforts to simultaneously reduce air pollution and stabilize CO2 emissions;

  5. 2C scenario: A less optimistic version of the alternative scenario, which has the same general properties but targets are less ambitious.

All of these scenarios were designed to be accessible using present-day technology, differing only in public behavior. None of them require "radical changes" like the abandonment of automobiles, etc; the main differences are that the IPCC "A" scenarios represent no serious climate-control effort, the "B" scenarios represent general public awareness without a concerted effort, and the two alternate scenarios represent sufficient public awareness to generate a stricter regulatory climate. In terms of optimism, the ordering is roughly A2 < A1 < 2°C < B1 < Alt. The "A" scenarios are also sometimes referred to as the "business-as-usual" scenarios.

For the purposes of this paper, though, the politics are besides the point: the scenarios are summarized by section 6.1, figures 1 and 24 (which show gas emission levels over time in each scenario), and then are discussed at great length in section 7, which compares actual GHG trends to the various scenarios. There's also some good summary text in section 3.1.1, that explains the basic technical assumptions describing each scenario.

In fact, I'm going to reverse the order of the paper and skip to section 7 ("why are these curves reasonable and what do they mean") before talking about what they predict. One interesting thing is that these scenarios were written up in 1990, so we can actually compare the past 15 years of real emissions data and see what curve we seem to be on. The good news:
"Overall, growth rates of the well-mixed non-CO2 forcings fall below IPCC scenarios (Figure 24a, b). Growth of CH4 falls below any IPCC scenario and even below the alternative scenario. Observed N2O falls slightly below all scenarios. The sum of MPTGs (Montreal Protocol Trace Gases) and OTGs (Other trace gases) falls between the IPCC scenarios and the alternative scenario (which was defined at a later time than the IPCC scenarios, when more observational data were available)."
Although to be honest, when you look at figure 24 you'll see that N2O is really pretty close to all of the IPCC A scenarios. You'll also notice that these plots really should have been on better axes, since the curves are all far too close to one another to eyeball anything, but for now at least the total forcing is staying pretty low, staying closer to B1, 2°C and Alt. However, CO2 is still not doing too well, closer to IPCC A levels.

For CO2, emission levels are known well enough to get some good information about atmospheric level trends, despite the significant (natural) annual oscillations. Those are shown in figure 25, using data from Oak Ridge and British Petroleum. The rise in overall annual emissions is pretty scary, with a huge rise in the total happening starting around 1950, and slowing down somewhat around 1973.

A key bit of text is in the middle of page 29, discussing the plausibility of hitting the various scenarios. Since 1973, global CO2 emissions have been steady at about 1.6%/year (versus over 4% per year from 1950-1973!), about half of it from coal.
"Continued emissions growth at 1.6%/year would double emissions in 44 years and be close to IPCC "business as usual" growth. IPCC scenarios B1, A1B, A2 and A1F1 require 50-year constant growth rates of 1.02%, 1.70%, 1.76% and 2.45%, respectively, to reach their 2050 emission rates."
This tells you pretty straight what the scenarios mean: IPCC A scenarios mean that CO2 emission rates keep growing at their same rates.

As far as the alternative scenarios, Pacala and Socolow provides a quantitative discussion of technologies that could bring us onto the alternative curves using present-day technologies. (Maybe we should do this paper?) One thing that makes me hopeful: half of the CO2 emissions are from coal, which means typically big point sources like power plants rather than distributed sources like cars -- those are a lot easier to do things about. (E.g., if the grandfather clauses in the Clean Air Act were to be sunsetted...)

OK, I've done a long enough drum roll: it's time to talk about what these scenarios actually mean in terms of climate change. The goodies are all in the figures, and section 6 is companion text.

Figure 19: Global mean surface air temperature, i.e. "overall average global warming." By 2100, the alternative scenario shows about a 1°C average increase, the 2°C scenario a (surprise!) 2°C change, B1 about 1.5°C, A1B about 2.5°C, and A2 about 3.2°C. However, global means are really misleading, since they hide substantial regional variation: take a look at the maps at the bottom of this figure. Even allowing for the fact that the model isn't as good at regional predictions, the fact that there will be substantial regional variation is a good guess. The northern hemisphere gets hit a lot harder than the southern, I suspect mostly because of the large open band of cold-water currents in the Antarctic Ocean which doesn't really get hit hard by anything. Here we see that the alternative scenario has relatively little change overall, except for about a 2°C rise at the North Pole; the 2°C scenario is just a uniform warming of that, going up to about 4°C at the North Pole; B1 is between those two. A1B and A2, on the other hand, see much different patterns: they see a substantial "warm patch" that hits the central US, Africa, the Middle East and northern India (about a 4°C patch in the A1B scenario and 7°C in A2!), and much more pronounced polar warming: there's a 7°C patch that hits Alaska and northern Canada in A1B, and a steady 7°C band over the entire far north in A2.

The catch is that the sigmas for calculations in sections 1-5 were highest at the poles, so the details are a bit unclear; but I think it's safe to say that the 2°C and IPCC A scenarios all show enough polar warming to be a substantial problem, and the warm patches in IPCC A could be big dangers.

Let me break for a moment to talk about what temperature mean shifts mean. First of all, I posted separately a little while ago a simple calculation to demonstrate what a 1°C average temperature shift means by taking the daily temperature data for Lincoln, NE for 70 years. Those daily temperatures are the sum of two Gaussians, one for winter and one for summer; if I extracted just June, July and August, it gave a very clean Gaussian, about 90±8°F. A simple formula came out: for changes up to 5°C or so, for each 1°C average temperature increase, the number of days above 95F increases by 7.4 days per summer. (Why 95F? That's the temperature that starts to kill corn really quickly, even if there are no changes in moisture levels) Right now, Lincoln averages about one week per summer in this range, and is considered nearly optimal corn country; a 4°C rise would make one-third of the summer lethal to the crop.

Second, North Polar shifts in temperature are particularly dangerous because of ice levels and the thermohaline cycle. First of all, the North Pole is home to one of the world's two big chunks of contintental ice, Greenland. (Continental ice is sitting on land, and so if it melts sea levels go up. The other vulnerable chunk is West Antarctica; East Antarctica is bigger and more stable. Greenland and W.A., if either of them were to drop completely into the ocean, would raise the sea level by about 20 feet apiece) This is a bit scary because there's a positive-feedback cycle in ice melting: water absorbs sunlight much more effectively than ice, so if a water pool forms on top of an ice sheet, it will warm up quickly, melting the ice beneath it, etc., and basically tunnel its way all the way down to bedrock. You end up with a hole that goes all the way down, and water starts to flow into it like a river, landing on the bedrock and lubricating the ice/rock barrier while it weakens the ice itself. As more holes like this form, it becomes possible for a big chunk of ice to suddenly break off and slide straight into the ocean.

(This kind of sudden melting is referred to as a "wet process," since it's catalyzed by water, as opposed to "dry process" ordinary melting, which happens gradually)

This is worrisome for several reasons. (1) The positive-feedback loop means that it doesn't take a large temperature change for these holes to form. Reports of such hole formations, even in central Greenland, have been accelerating in the past two years. (2) Evidence of glacier retreat in Greenland is building, too; [search for (greenland ice melting) to get various news stories...] (3) Recently we learned just how quickly ice sheets can collapse, when the Larsen B ice shelf in Antarctica (a bit of sea ice about the size of Rhode Island) collapsed into nothingness over 48 hours. (Pictures linked to from the wikipedia article) Also, as noted in section 6.2.1 of this paper, "the Arctic was ice-free in the warm season during the Middle Pliocene when global temperature was only 2-3°C warmed than today."

The other, related issue is the thermohaline cycle. [cf also the links from that article -- they're very informative] Basically, the world's sea currents form a single large conveyor belt, and its main heat exchanger is in the North Atlantic, where warm South Atlantic waters come up, contact the cold polar waters, lose their temperature and sink, returning at the bottom of the ocean. The residual energy is emitted in the form of steam clouds, which are pushed eastward by prevailing winds to form the Gulf Stream, which is what keeps Europe warm. The trouble is that this cycle is very dependent on water salinity levels to keep flowing, and a large injection of fresh water can cause it to shut down. (See the articles linked to from the link above for the technical details... this is conjectured to have happened about 100kyr ago when a large inland sea in Canada drained into the North Atlantic) Without the Gulf Stream, cities in Europe would have roughly the same climate as cities in the US at the same latitude -- e.g., Madrid about like New York City, London like Calgary...

Third, the warm regions shown on the map in northern India (especially in the A2 scenario) can be dangerous for other reasons: the Himalayan watershed supplies the Brahmaputra, Ganges, Indus, Mekong, Yellow, and Yang-tze rivers, which together are the water sources for about 40% of the world's population. Warming can cause more rain rather than snow in this area. Rain there (a) causes flash flooding, since the ground isn't used to it, and thus cart off topsoil; (b) melts the underlying snow if it lands on top of it; and (c) doesn't accumulate in winter and then melt off in the summer the way snow does. So an increase in rain levels could easily lead to flash flooding in the spring followed by drought in the summer in the most populous areas of the world.

So basically: even fairly mild polar climate change can be very dangerous. Either of the IPCC "A" scenarios would have enough polar warming to probably make the North Polar Sea permanently navigable (not just seasonally!), and could potentially have huge impacts on climate in Europe. Any event that caused a vanishing of the Greenland ice sheets would be much more severe: a 20' rise in sea levels would render, for example, both Shanghai (40m people) and Calcutta (60m people) subaquatic. Consider the impact of a few hundred million refugees suddenly in need of new places to live, especially if the areas in question are already affected by substantial seasonal droughts...

OK, maybe you aren't convinced that the IPCC "A" scenarios are dangerous yet. Let's flip the page to

Figure 20: Observed seasonal surface temperature change. (Goes with section 6.2) Start with the figure, especially figure 20b: they separated the year into winter and summer (so that you get single Gaussians of "average annual temperatures" for both times of year) and plotted maps of local surface temperature change for that time of year (in each scenario), and that same temperature change divided by the local average annual variation σ. The latter one is very useful, since it tells you how "unseasonable" the temperature changes feel: a change of 1σ means that the average day would be a "hot day" by the old standards, the sort of day that happened about 15 days per season; (the 15 is the probability of a 1σ or less temperature times the number of days in June, July and August); a 2σ change means that the average day would be a day that happened about two days per season; a 3σ change means that the average day would be a day that used to happen only once every 8 years (!); a 4σ change, a day that used to happen once every 343 years (!!); and so on. On the plot of Lincoln, NE referenced above, each sigma for summer is eight degrees Fahrenheit.

So start with the top row, scenario IPCC A2. Summer temperatures in every latitude between Australia and London show a warming of more than 5σ the global average is 8.33σ, even though the average temperature difference is only 2.7°C. (cf the third and fourth plots from the left, top row) I'm actually finding it a bit hard to believe this number, simply because it goes all the way beyond anything I could imagine. If you have 92 days of summer per year, an 8σ event is one that would not happen by chance in the lifetime of the universe. (Hell -- in particle accelerator experiments, anything that's 5σ off of expectations is considered to be an almost-incontestable "smoking gun" of new physics, and this is in the most sensitive detection systems of any sort ever built by man.)

So I'm sincerely hoping that these plots are somehow completely wrong, or that I'm misreading them.
[OK, just for reference, here's a table of going from "number of σ away from the mean" to "number of times you would have to try before, by pure chance, this would happen":

Std deviationsTimes before it happensNotes
6.3 triesAbout the odds of rolling a 10 or more on 2d6
43.96 tries
740.8 triesThe "something interesting" threshold in most experiments
3.16*104 tries
3.5*106 triesAn event of this rarity is generally considered to not be a fluke, even in particle physics.
1.0*109 tries
7.8*1011 tries
1.6*1015 triesAt one try per second, this will happen about once in a billion years
8.8*1018 triesAt one try per second, this would have happened once in the lifetime of the universe.
10σ1.3*1023 tries

So when you hear scientists talking about a "5σ event," this is what they mean -- an event that would happen only once every 3.5 million tries. Even when dealing with non-Gaussian probability distributions, people will often refer to 5σ events, implicitly referring to the probability of the event rather than an actual number of standard deviations]</blockquote>

The other scenarios are gradually better. IPCC A1 is still showing 5-6σ average changes, and parts of the world seeing more than 10σ; B1 and the two alternative scenarios are somewhat rosier. In the Alternative scenario, the average rise is under 2σ, and it's mostly spread out over oceans where there's more likely to be an error. (Remember the trouble with the ocean models)

Overall, though, even if these plots are pretty far off, it suggests that both IPCC "A" scenarios are incompatible with the continued existence of human civilization, and A2 in particular may be incompatible with the existence of human life.

Is that completely unreasonable? Not really; the planet has had pretty big climate changes in its past. Recent ice core analyses suggest that at the Paleocene-Eocene Thermal Maximum, about 55Myr ago, the North Pole had an almost tropical climate, with water temperatures around 75F and supporting the tropical alga Apectodinium. And there have been even bigger climate shifts in the past, especially the ones associated with major species extinction events. And the planet has survived these... just not a lot of the larger species living on it. So how bad these scenarios are depends on just how much you like cockroaches.

On the other hand, the two alternative scenarios seem to be in the survivable range; the 2°C scenario is a bit warm, and it has some high-temperature patches in the Himalayas that I would call "very worrisome" by the standards of "drought, famine and refugees may result," but not so worrisome by the standards of "oh my God, we're all going to die." The Alt and B1 scenarios look a bit better.

So let's flip back to the article text, section 6.2, and look at a few more details.

Section 6.2.1: Arctic climate change. The challenge in predicting this is, as the article states, that "unforced climate variability" (i.e., the ordinary variation in annual temperatures) in this area is particularly large, as figure 11 showed; observed variability is comparable to the differences between the models. So we can't expect the details of the model to match with observation. (e.g., there was a strong arctic warming event around 1940, believed to be "an unforced fluctuation associated with the Arctic Oscillation and a positive anomaly of ocean heat inflow to that region," i.e. "we don't really understand it, but it heated up and we didn't do anything obvious to trigger it." I personally suspect that better sea and sea ice models, including taking into account how the ice sequesters methane, will make this get a lot better; this looks like a good research direction.

They also say that, now that their models have a better description of seasonal radiation cycles, the "small ice cap instability" that older models saw goes away. However, the "snow/ice albedo feedback" (the positive feedback mentioned above) remains, making icxe levels very sensitive.

Observational data indicates that perennial Arctic sea ice has been dropping fast since 1978: Comiso, 2002 shows about a 9%/decade loss rate, and the various photographs I'm sure you've seen by now of "now and then" bears this out pretty strikingly. It's not clear if we've passed the tipping point for this system, where all sea ice will end up gone during the warm season... but I'll put my faith in the old measurement system of economics: I've seen more and more energy companies looking with interest at oil reserves in the north polar region, shipping companies planning routes there, and governments talking about how to partition the borders near the North Pole, which are actually not very firmly defined because nobody ever bothered. Clearly the people with a lot of money riding on it are betting on an at least seasonally navigable North Polar Sea in the next few years.

So this raises the question of whether any of the scenarios can avoid a complete collapse of the Arctic ice system. (p. 25) It looks somewhat promising: figure 21 breaks up the effects on climate systems by running the simulations with only some of the forcings turned on. It looks like CO2 and CH4 + O3 give about equal contributions to the forcings, and other forcings roughly cancel out. So if we follow alternative-scenario levels for those three gases, it's probably possible to keep Arctic warming in the 1°C range, and avoid a complete melting; but "on the other hand, if CO2 growth follows a [business-as-usual] scenario, the impact of reducing the non-CO2 forcings will be small by comparison and probably inconsequential."

Section 6.2.2: Tropical climate change. This is interesting because, for example, hurricanes come out of this part of the ocean. (They get their energy from warm surface water, and the effects of a warmer Caribbean are probably pretty obvious to everyone by now. OTOH this means that vacation prices in that area during hurricane season are a lot lower...) Figure 22 shows the changes in surface sea temperature [SST] in the various models; again, things come out OK in the alternative scenario, OKish in IPCC B1, and progressively less so in the various other models. 22a compares SST anomalies in 1995-2005 (when there was an unusually high level of hurricane intensity) to 1970-1994 (when there wasn't); they compare this to what the models predict due to each of the assumed forcings, and conclude that the model is about spot-on for this time period, and this warming can therefore be definitively attributed to the indicated gas emissions. (This specifically refutes assertions of other articles that anthropogenic GHG emissions play no role in this)

Figure 22d compares the scenarios; note that again there are substantial differences between them.
"The salient point that we note in Figure 22d is that, despite the fact that warming "in the pipeline" in 2004 is the same for all scenarios, the alternative scenario already at mid-century has notably less warming than the other scenarios. This result refutes the common statement that constraining climate forcings has a negligible effect on expected warming this century."

Section 6.2.3: Ice sheet stability. The total amount of water stored in all the world's ice sheets is enough to raise the sea level by about 70m; Greenland and West Antarctica each contain about 10% of that, and East Antarctica, which is considerably more stable, about 80%. [NB: You may have seen some recent ad produced by energy companies saying that global warming is a myth, and recent papers show that ice sheets are growing. Those papers said that East Antarctica is growing, but that West Antarctica is showing some disturbing signs of collapse, like chunks the size of Rhode Island unexpectedly falling off.] In the historical past, there has been a lot of melting and freezing: when last global temperatures were 3°C warmer than today, during the Pliocene (3Myr ago) sea levels were 25±10m higher than today. So major melting can definitely happen: the question is how much, and how quickly it would happen.

The problem with the second calculation is that it's very difficult because of the positive-feedback mechanism described above, which gives it a "switch function" behavior and very high sensitivity to ambient conditions. They city Zwally et al.[2002] , which shows how surface melt in Greenland really is going all the way down to the bedrock and lubricating flow into the ocean, and van den Broeke [2005] which found that unusually large surface melts in Antarctica "preceded and probably contributed significantly to the collapse of the ice shelf." Simple thinning and gradual melt may be even more significant.

The summary of this section is that a lot of experimental evidence suggests that ice melt is already well-under-way, and it's damned hard to model what happens next.

Finally, section 8 just gives summaries and conclusions, which is what we've been doing here all along so there isn't too much to add. They try to sugar-coat things a little bit, for instance downplaying the odds of a catastrophic "wet process" melting, but they don't give any reason why they do so -- I suspect it's because they don't want to sound too alarming. So here's my short summary of what the paper says:


So as far as I can tell, here's the short answer:

  • The model looks like it works pretty well in modelling global climate, and in particular if you feed in the actual gas emissions in the past century its predictions match what actually happened pretty well. Regional climate is a bit iffier, but the problems are more localized, things like overestimating Pacific warming (probably related to a bad model of El Niño) and so on. The fact that the model replicated what happened right after the Pinatubo eruption so well is a good sanity check that, at least to me, makes it sound pretty credible.

  • In all of the IPCC "A" scenarios, representing no real attempt to change anything, climate change levels quickly spiral out of control. Even if the most alarming graphs in this paper are wrong, in this scenario we should expect complete North Polar ice melting, major temperature changes in large agricultural areas, large-scale droughts in the most populated areas of the world, sea level rises leading to hundreds of millions of refugees, and the conjoint human migrations, leading to disease migration, warfare, and in general the ride of the Four Horsemen. In the worst scenarios, enough of the planet would become uninhabitable that there is a definite possibility of human extinction.

  • In the IPCC B1 scenario and the two alternative scenarios, on the other hand, there are negative but not ultimately apocalyptic consequences. B1 and 2°C show enough warming that we could expect significant ice cap damage, increases in hurricane frequencies, and probably drought affecting some areas (West Africa, India, China); these could cause substantial political upheaval. If the Greenland ice sheet were to collapse, this would be much more serious, since the sea level rises would lead to mass migrations, and there is the potential for sharp temperature drops in Europe. In the alternative scenario, there would be some increase in hurricane intensity, some polar melting, etc., but all of the worst consequences seem to be avoided.

  • Thus it seems clear that with present-day technology only, policy changes determine whether we will enter scenarios ranging from the "everything is OK," through major political and social upheavals, up to disasters that threaten humanity as a whole on a par with global thermonuclear war.

  • The most useful experimental directions seem to be (a) getting better measurements of gas emissions over space and time, and (b) getting more and more paleoclimate data so we can calibrate these models better, test them against reality, and get a sense of what the effects of major past events have been. (E.g., if we can measure "at time X a lot of CO2 got emitted into the air; at time X+Y the temperature had responded as follows...") The most useful theoretical directions seem to be improving the ocean model and the sea ice model.

I'll leave the further discussion of this, well, to the discussion.

Thanks for reading!
  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic