Tuesday, February 13, 2007

Under-valuing catastrophic risk

Tim Flannery reacted to the recent 4th IPCC report by claiming that its middle-of-the-road forecasts ignored the potential for very steep temperature rises of 6 degrees C (with probability 0.1) and for very rapid melting of the Arctic Ice Cap in 5-15 years. In short he claimed the IPCC understated the possibility of catastrophic risks. Flannery was dismissed as an alarmist (and worse) by right-wing commentators.

But Tim is worth listening to for 2 reasons:

1. In dealing with uncertainty in a global warming context it makes sense to look at extreme catastrophic outcomes that might happen. Even if there was only a small probability of a catastrophic climate outcome that devastated the world that would be an event that should be taken very seriously by decision-makers. Extreme events must be accounted for if their expected costs are very high because society is more concerned with extreme climatic events than averages. For example, we are much less interested in knowing in average temperature will rise by 3 degrees C than in knowing whether we will experience extreme heat waves, droughts or storms.

In other words, we should not only be trying to nail down forecasts of temperature ans sea level changes but also look at the probability distributions od such forecasts with an emphasis on long-tailed events.

2. There is increasing evidence from various sources that the IPCC report was conservative in undervaluing catastrophic outcomes. New Scientist this week voices several concerns. Generally the IPCC emphasized rigorous research so research deemed controversial or not yet incorporated into climate models was excluded.

The political benefit was to leave little room for the climate change sceptics but there was, simultaneously, the dangerous cost that legitimate findings may have been frozen out with many of the more scary scenarios for climate change - those that scientists discuss among themselves - 'being left on the cutting room floor’.

Climate scientists came together 2 years ago to discuss ‘dangerous’ climate change at a conference organised by the UK government in Exeter. They identified potential positive feedbacks and ‘tipping points’ not included in current models of the earth's climate system that could accelerate global warming or sea-level rise. These included:
  • the physical collapse of the Greenland ice sheet,
  • rapid melting in Antarctica,
  • a shut-down of the Gulf Stream in the Atlantic,
  • the release of CO2 and methane from soil, the ocean bed & melting permafrost.
Yet last week's IPCC report largely ignored these possibilities.

One concern is that the ice sheets of Greenland and Antarctica could disintegrate causing sea level rises of metres. But the IPCC restricted itself to noting that sea levels are rising 3.1 centimetres a decade - still twice the rate of the early 1990s. Current climate models assume that the ice sheets will melt only slowly, as heat works its way down through ice more than 2 kilometres thick.

But many glaciologists believe ice sheets fracture as they melt, so water can penetrate to the bottom within seconds, warming its full depth so break-up of the ice sheets happens long before thermal melting. Indeed, the rate of ice loss in Greenland has unexpectedly doubled in the past decade and the permanent Arctic sea ice is contracting by 7% each decade.

Moreover, there is research showing that world sea levels are rising 50% faster today than predicted in the last IPCC report in 2001 This could be the first sign of a dramatic acceleration of sea level rise as ice sheets disintegrate. But the IPCC last week reduced its estimate of worst-case sea level rise in the coming century from 88 to 59 centimetres. Controversy over such forecasts continues at Deltoid.

Researchers at the UK's National Oceanography Centre, Southampton also felt overlooked. In 2005, they reported that the Gulf Stream slowed by 30% 1957-2004. But the IPCC summary insists that ‘there is insufficient evidence to determine whether trends exist’.

Water vapour is increasing in the atmosphere, the IPCC says thanks to more evaporation from the oceans. Weather systems are changing, with more intense droughts and tropical cyclones at low latitudes. Rainfall, when it occurs, is heavier because the warmer air holds more moisture. However, the IPCC fail to take up warnings made of ‘carbon-cycle feedbacks’ - the release of greenhouse gases from warmed soils, forests, permafrost and sea beds. It does note that CO2 is accumulating at a rate 1/3 greater than 20 years ago.

The IPCC's predictions of significant warming in northern latitudes should give urgency to assessing potential methane releases from Siberia and the Arctic. But, these fears were not in the IPCC report.

Any committee requiring politicians to agree is going to take time arriving at a consensus. The IPCC’s climate scientists had to run the gauntlet of government delegations, who had to approve every word of the summary prior to publication. By and large, the IPCC scientists insist they ‘faced down political interference’. The prize of having governments formally sign off on the report will, they hope, make any compromises worthwhile.

Future IPCC work on climatic risk and uncertainty should address the prospect of extreme catastrophic climatic events. Such events lower the discount rate one should apply in climate change policy modelling. Also, as I pointed out in an earlier post, such events need to be analysed from the perspective of the ability of policies to offset them. Extremely disadvantageous events that cannot be offset suggest lower incentives to embark on policy and a case for delaying action. Extreme events that can be offset enhance arguments for intense and prompt policy activism. I set out the analytics of such arguments here.


James Dudek said...

Has anyone valued what a meteor strike on the earth would cost?

conrad said...

Thats interesting Harry and I completely agree with you about trying to take into account catastrophic risks, although somehow I think people won't -- there seems to be too many other examples from other areas where people never take enough precautions -- the stock market is an other obvious, which I'll just assume will have a good chance of crashing when the US leaves Iraq and Iran and Saudia Arabia start fighting, but other examples like pandemics come to mind.

hc said...


I don't know if they have. But even if they did its not something you can do much about and hence does not lead to a claim on policy.

Indeed by reducing the present value of existence it suggests a case for 'having a party'.

Anonymous said...


I agree entirely that one should try to take low-probability, high-impact risks into account. I don't think the IPCC has been excessively conservative in estimating the *likely* course of climate change. (I'm not sure if you said it was, but some of the examples you gave suggested imply this.)

I'm particularly interested in this one:

"Researchers at the UK's National Oceanography Centre, Southampton also felt overlooked. In 2005, they reported that the Gulf Stream slowed by 30% 1957-2004. But the IPCC summary insists that ‘there is insufficient evidence to determine whether trends exist’".

I agree with the IPCC.

Can you give a reference to the NOC researchers saying they feel overlooked?

The NOC time series is based on a series of snapshots of the density structure of the North Atlantic. The problem is, we don't know what the ocean has been doing between the snapshots. The system is almost certainly undersampled, meaning it is easy to draw false conclusions about the true nature of the time series from the samples. (I could give an example of some personal relevance relating to a series of medical tests, but that's definitely another story.)

Anonymous said...

Oops, I'm not "anonymous", I'm Mark Hadfield. Must get an account.

hc said...

Mark, The statements made are documented in this week's NewScientist - I used to be able to link to their entire articles but cannot any more. Moreover, I am making a general point about accounting for catastrophic events by making examples - I am not expert on the examples chosen or on making evaluations of the evidence.

The point you make (if correct) would reduce the probability you would assign to that unfavourable outcome.

There are 2 issues - working out the likely course of events and looking at extreme even low probability high impact events. I am suggesting that the IPCC report focused on the first of these tasks but might also address the second.

Maybe they will in their next report.