Robert Pindyck on climate change models

Pindyck, from MIT, is a leading expert in this area, here is part of his summary conclusion:

It would certainly be nice if the problems with IAMs [integrated assessment models] simply boiled down to an imprecise knowledge of certain parameters, because then uncertainty could be handled by assigning probability distributions to those parameters and then running Monte Carlo simulations. Unfortunately, not only do we not know the correct probability distributions that should be applied to these parameters, we don’t even know the correct equations to which those parameters apply. Thus the best one can do at this point is to conduct a simple sensitivity analysis on key parameters, which would be more informative and transparent than a Monte Carlo simulation using ad hoc probability distributions. This does not mean that IAMs are of no use. As I discussed earlier, IAMs can be valuable as analytical and pedagogical devices to help us better understand climate dynamics and climate–economy interactions, as well as some of the uncertainties involved. But it is crucial that we are clear and up-front about the limitations of these models so that they are not misused or oversold to policymakers. Likewise, the limitations of IAMs do not imply that we have to throw up our hands and give up entirely on estimating the SCC [social costs of carbon] and analyzing climate change policy more generally.

The entire essay is of interest, via Matt Kahn.

Comments

From the apparently not self-recommending 'Review of Environmental Economics and Policy, volume 11, issue 1, Winter 2017, pp. 100–114'

Possibly Pindyck in Big Oil's back pocket, like Lindzen is, NOT THAT IT MATTERS. Did you read what I say IN CAPS? Does. Not. Matter. Just like Hockey Stick by Mann is irrelevant whether Mann is left or right wing. Just like inflammatory emails in Climate Gate are a red (read) herring (hearing). That data speaks for itself, as does the model.

Bonus trivia: Lindzen's "Iris Effect" hypothesis has, as of last year, been confirmed! Wow, I haven't followed the AGW debate for years, since I quit trolling the Usenet (I played the "skeptic" though I believe in fact in AGW). But it's a weak effect, with cirrus clouds and albedo.

And the hockey stick was debunked.

More like "completely and utterly destroyed."

Repeating a tenuous assertion with increasing certainty is almost the definition of echo chamber.

"More than two dozen reconstructions, using various statistical methods and combinations of proxy records, have supported the broad consensus shown in the original 1998 hockey-stick graph, with variations in how flat the pre-20th century "shaft" appears.[12][13] The 2007 IPCC Fourth Assessment Report cited 14 reconstructions, 10 of which covered 1,000 years or longer, to support its strengthened conclusion that it was likely that Northern Hemisphere temperatures during the 20th century were the highest in at least the past 1,300 years.[14] Over a dozen subsequent reconstructions, including Mann et al. 2008 and PAGES 2k Consortium 2013, have supported these general conclusions."

" including Mann et al. 2008"

Perhaps, but I would leave that one out.

variations in how flat the pre-20th century "shaft" appears.

I'm not a hockey player, but do a lot of them come with variably flat shafts? Hmmm.

It's debatable whether the MWP was warmer than today, but we definitely had the LIA.

Regardless of what we think is happening or not happening with the climate, if we're honest then most of us have to admit that we have no clue about it.

Not only that, we don't really have a clue about who does.

We have some clues. Most places are warmer than they were thirty years ago and more. Glaciers are receding. Growing seasons have gotten longer and some northern areas are "greening." So it's pretty clear that there has been a pretty general warming over the last half century.

No one knows exactly how much or how much it will continue. Almost all climate models have "run hot", predicting considerably more warming than occurred.

This sounds right. I've read lately that they have been able to map the sea level rise in the last interglacial period. Rise goes up, then stops well before what models say.

'Almost all climate models have "run hot", predicting considerably more warming than occurred.'

Except for the Arctic - those models have clearly failed to match what is being observed with real/near real time data. Possibly in part because albedo/ocean heat transport mechanisms are still very inadequately modelled regarding that part of the planet.

For those more interested in data, this site provides an excellent monthly review that has essentially no modelling at all - http://nsidc.org/arcticseaicenews/

Here is the latest information, which will be updated in the next several days - 'On September 19 and 23, 2018, sea ice extent dropped to 4.59 million square kilometers (1.77 million square miles), tying for the sixth lowest minimum in the satellite record along with 2008 and 2010. This appears to be the lowest extent of the year. In response to the setting sun and falling temperatures, ice extent will begin expanding through autumn and winter. However, a shift in wind patterns or a period of late season melt could still push the ice extent lower.

The minimum extent was reached 5 and 9 days later than the 1981 to 2010 median minimum date of September 14. The interquartile range of minimum dates is September 11 to September 19. This year’s minimum date of September 23 is one of the latest dates to reach the minimum in the satellite record, tying with 1997. The lateness of the minimum appears to be at least partially caused by southerly winds from the East Siberian Sea, which brought warm air into the region and prevented ice from drifting or growing southward.'

I think I have a sufficiently good estimate for practical purposes. Climate change is clearly real and the Arctic sea is is already melting with effects on weather patterns. It's also clearly not fast enough to be a high-stakes near-term economic risk. Although some costs are near-term and in fact are already materializing, reduction of CO2 emissions is currently still more costly than accepting the warming cost, perhaps with the exceptions of some margins like abolishment of fossil fuel subsidies, which are distortive anyway. We also know unilateral emission reductions are irrational since the rest of the world can just make up by burning the difference more cheaply and the warming costs are global. We also know time discounting is a necessity for various undeniable reasons, such as uncertainty discounting and opportunity cost. And we know that with realistically strong time discounting, major efforts to reduce emissions are irrational in the near term. We also know fossil fuels are finite with creates a natural market-based negative feedback anyway.

+1

Although I’d still trade a carbon tax for a 1 to 1 reduction in payroll taxes. Even if India makes up the difference the payroll tax is so egregiously distorting it’s a win.

Co-sign.

Yep, I'm on board too. It's so sensible it will never happen.

We also know unilateral emission reductions are irrational since the rest of the world can just make up by burning the difference more cheaply and the warming costs are global.
Yep and the less we use the lower the price and the more we use, which reminds me it has been a long time since I have thanked the Europeans for their high gasoline taxes saving me money of fuel.

Thank you Europeans.

Nevertheless I would support a simple CO2 tax.

This article seems to pretty on point. Clearly the evidence points to mild global warming. We don't know how warm it's going to get. Probably civilization is reacting appropriately with a long term switch to battery powered commuter vehicles, nuclear power, natural gas and renewables. Economic power storage would, of course, make renewables more feasible.

Is civilization switching to nuclear? Seems rational but lately has been going the opposite direction based on opposition from a different wing of the environmental movement.

What do you mean "civilization", Westerner? NIMBY will sink the republican democracies by the end of this century. The only adults in the room currently are the autocracies.

1. Nitpicky, but IAMs are not climate models, they are coupled climate-policy models. Climate models themselves are much less problematic.

2. Carbon Brief just put out an excellent primer on IAMs:
https://www.carbonbrief.org/qa-how-integrated-assessment-models-are-used-to-study-climate-change

This goes over the details of how IAMs work, the main uncertainties and their main drawbacks. Two important takeaways (similar to Pindyck's conclusions) are that IAMs are primarily useful for "big picture" thinking, rather than details, and that policymakers need to be more involved in model development, both because building IAMs involves making value judgments and to add political constraints to the models.

I don't get it, why is this finding objectionable? We know WAY more about financial modeling than we do about climate modeling, and these are the same techniques used there. Does anyone here think that financial models effectively predict future asset price movements? The point isn't to see what the movement will be certainly, it's to see what the movement will be given the change in underlying parameters, whose values are incredibly uncertain.

Pindyck, incidentally, is the coauthor of one of the better microeconomics texts. It’s up their with Hal Varian’s marvelous text.

Not true. As the OP wrote: "This does not mean that IAMs are of no use. As I discussed earlier, IAMs can be valuable as analytical and pedagogical devices to help us better understand climate dynamics and climate–economy interactions, as well as some of the uncertainties involved".

Analogy: big game animals all died, without exception, when primitive man invaded their ecosystem. Ergo, man killed this mega-fauna. Similarly, global temperature has gone up nearly one-to-one with increased economic activity. In fact, during a recession, there's less accelerated global warming. Cause and effect? Probably. But I'm sure you'll find a flaky archeologist or two who thinks it's just a coincidence the dodo died out about the time sailors invaded their territory.

"global temperature has gone up nearly one-to-one with increased economic activity"

Brilliant. For every degree that temperature increases, economic activity also increases one degree

There was less accelerated global warming through the end 20th century boom, the early 21st century expansion, and the global recession of 2008-whenever. There was accelerated warming through the 1970s "stagflation" and the 1980s boom. Warming has shown virtually no short-term correlation with economic activity. Long-term, of course, very much so.

FWIW: Some big game animals did survive, e.g., the North American bison. However, I have little doubt that without humans, many, many more "megafauna" would have survived.

Frank Robinson (1973): "Close don't count in baseball. Close only counts in horseshoes and hand grenades."

Yogi Berra (2018, from high above it all): "Close don't count in baseball. Close only counts in horseshoes, hand grenades, and hurricanes."

Not that I'm accusing Tyler and Alex of pretending to be conservatives or anything. USA! USA!

Well Trumpty Dumpty promised a wall
Trumpty Dumpty reneged it all
All what was conservative then
is not conservative now.

Isn't it amazing that folks like you cannot even handle the possibility of doubt??? I mean, the guys is not denying anything but still you feel threatened by the fact that we really don't have the scientific capability to know things for sure. You are just as irrational as deniers.

He is referring to using IAMs as used for policy guidelines, not as a criticism of climatology per se.

"(4) IAMs can tell us nothing about “tail risk,” i.e., the likelihood or possible impact of a catastrophic climate outcome, such as a temperature increase above 5C, that has a very large impact on GDP. And yet it is the possibility of a climate catastrophe that is (or should be) the main driving force behind a stringent abatement policy."

Not preparing today for potential large negative outcomes in the future is dumb when the cost of doing so now is relatively small compared to the cost of doing nothing and the tail risk turns out to happen.

"The problem is that climate science and economic principles are limited in what they can tell us about how to specify and parameterize an IAM’s equations, which is why the models cannot tell us much about the design of climate policy. "

When in the future? At what probability? It's worth remembering that, in the long run, we're all dead either way.

he clearly sees a policy response need for potential tail risks in that might result from global warming

"In order to determine plausible outcomes and probabilities, and the emission reductions needed to avert these outcomes, we would needto rely on “expert” opinion."

Deliberately vague. Also evades the question of when. 5C+ after 2100 are fine.

u missed the point, we don't know how bad the costs will be or if they will be "fine." The idea is that the likelihood of catastrophic costs are high enough that some type of policy response is needed. The business of climate modeling is all about probabilities, not forecasting. Hey, a limited nuclear war with North Korea might be fine, especially if we had zero costs up front. But I don't see anyone recommending that policy response.

"some type of policy response"

Again with the strategic vagueness. No sense of cost-effectiveness.

"The idea is that the likelihood of catastrophic costs are high enough"

Not really. Not in the time frames that matter.

Something like the '90's Kevin Costner vehicle Water World. Assuming the movie itself is a good guide, our best bet is to begin stockpiling diesel fuel, pontoons, and firearms.

And Kevin Costners

Not preparing today for potential large negative outcomes in the future is dumb when the cost of doing so now is relatively small compared to the cost of doing nothing and the tail risk turns out to happen.

If you imagine a large enough tail risk, any action to prevent it is "relatively" cheap. If I imagine that North Korea is going to send nukes to the 20 largest American cities and detonate them 5 years from now, then it is "relatively" cheap to nuke Pyongyang tomorrow.

We have to be realistic about tail risk, and not kid ourselves about the cheapness of what is proposed. Given present technology, bringing the world's total carbon emissions to a level that will not increase CO2 levels would condemn much of the world to never-ending poverty. Until energy technology gets substantially better, we have pretty awful trade-offs.

"Not preparing today for potential large negative outcomes in the future is dumb "

The sun will turn into a Red Giant and then dim in the next 5 billion years. What should we do today to prepare?

The present value of a catastrophic loss 5 billion years out nearly nil, so the answer is nothing.

How about 1 billion years? That's when the sun will be large enough to boil away the oceans.

I am going to randomly throw out a number that seems worrying to me - just for argument's sake: in 70 years a cost greater than 10% of global GDP. Honestly, I really don't know what numbers are in play.

"I am going to randomly throw out a number that seems worrying to me - just for argument's sake: in 70 years a cost greater than 10% of global GDP."

The current estimates are that AGW will lower world GDP by around 4% in 2100. So, it doesn't meet the threshold you just set.

There are no current estimates. Estimates are forecasts. Climate models don't forecast, but assign probabilities for time frames / temperature changes. 4% in 100 years would be have a probability associated with it for each model. And besides, mine wasn't a threshold but a guess, as I stated.

he clearly sees a policy response need for potential tail risks in that might result from global warming

"In order to determine plausible outcomes and probabilities, and the emission reductions needed to avert these outcomes, we would needto rely on “expert” opinion."

There are other low probability high impact risks that should probably be addressed first. Electrical grid robustness in the event of an EMP attack or Carrington event would be first thing on my list. Biowarfare or naturally occurring Spanish Flu 2 probably next.

I'd argue with "first" but some of those risks are important too. Keep in mind, we have policy responses in place for biowar, emp attacks, and the like. Are they adequate, I don't know. I doubt it.

#1 priority is unfriendly AI because that is guaranteed to come down the pipeline in the immediate future and if solved will be a huge help for anything else.

Exactly - AI is the risk we should worry about, not global warming, which even in the worst possible case cannot represent a potential extinction threat to the human race.

I don't agree that AI poses a significant global problem. First, take a look at the literature on the inability of AI to 'recover' from spoof attacks. But my point is actually that while AI will no doubt lead to some 'fatal events', its implementation will be piecemeal enough (if not slow) to allow us to set up a legal control system - once the silent majority becomes actually alarmed about it.(enough so that the herd stampedes the special interests who will be paying off Congress to do nothing)

Bostrom and company would say such a characterisation that completely misses the AI superintelligence risk, and that breakout time could be WAY faster.

Exactly. Long list of Extinction Level Events before we get to global warming.

But the AGW crowd never want more funding for AI risk, comets, flu pandemics, nanotech or biotech, global nuclear war, CERN disasters or super volcanos. No, it's AGW as the only tail risk in town.

One might almost think they really weren't concerned with tail risk. Just AGW.

They don't even care about AGW per se, it's just a status adjustment to their perceived enemies in the industrial world. Hence the Exxon witch hunt, etc.

If you really want to worry about tail risk, the first of these would dwarf any negative impacts from warming, the latter is a likely human extinction event. And no, models cannot rule them out.

https://en.wikipedia.org/wiki/Little_Ice_Age#Geophysical_and_social_impact_by_region

https://en.wikipedia.org/wiki/Last_Glacial_Maximum

Meanwhile, on the other side of town,

"Arctic Methane Release Could Cost Economy $60 Trillion"

https://www.scientificamerican.com/article/arctic-methane-release-could-cost-60t/

Oooh, oooh; I spotted an error there!

It was published in Scientific American. Everyone remember when that was a real magazine?

For the life of me, I cannot understand why anyone thinks "I ignore this source" is sensible or convincing.

More like, you raise your hand to self-indict.

I liked Scientific American a lot better before the buyout by the Onion.

That was the big mistake of Hansen in 1988 -- he drastically oversold the reliability of his auguries, and temperatures in his "business as usual" Scenario A (the emissions scenario that actually happened) were way, way too high. Models are better now, but the ECS spread in the literature is still pretty wide. Policymakers need useful inputs!

GIGO and inbuilt bias and assumptions make the models essentially useless, yet the money flows and flows to those who come up with ever worsening doomsday forecasts. Nobody knows if the recent warming was natural or human caused. Almost anything and everything is blamed on AGW, I can't see how the AGW theory is falsifiable, it's far closer to religion than science.

This isn't about climate models which are well established scientifically rigorous models with strong foundations in extremely well known physics. It's about economic models which are of course crap, because nothing in economics is actually rigorously known to the standards of a real science.

I think you mean that we don't have a good basis for estimating future economic growth rates under different scenarios, not that economic models are crap in general (supply-demand analysis works for instance). One thing is clear though from empirical study of economies is that capitalism is the best system we have found to incentivize people to find solutions to problems. So maybe we can't forecast exactly what will be the effect of higher temperatures on the global economy, but we can set in place incentive mechanisms that will provide us with the means to handle it. For instance we know richer people are better at handling disasters than poorer people (see Japanese cyclone impacts vs Haiti ones for example). So let's have policies that make people richer - more trade, lower taxes, less regulation etc. If we can get the poorest people in the world to the level of the Japanese of today by 2100 then any possible impact of climate change will be hugely muted.

"well established scientifically rigorous models with strong foundations in extremely well known physics."

Or, not.

https://wattsupwiththat.com/2009/01/28/forecasting-guru-announces-no-scientific-basis-for-forecasting-climate/

Models are better than they used to be, but they're still largely speculative. That's why every GCM forecast is different.

IAMs are not climate models, as the title of this post suggests. They are models that integrate climate and economics models . The equations governing large processes in the climate system are well known and based on physics and chemistry. Pindyck is probably referring to the economic modeling component.

"The equations governing large processes in the climate system are well known and based on physics and chemistry"

Not even remotely close to true, the models are heavily parameterized and the spread of ECS in the literature is consequently rather large. Storch found in 2008 that even most climate scientists do not believe atmospheric dynamics are understood well enough to forecast long-range temperature.

Climate scientists have never been able to predict the global temperature 1 year out, 5 years out, 10 years out, or 20 years out, have they? So why do we think they can predict it 50 years out?

The equations governing large processes in the climate system are well known and based on physics and chemistry.

Partly true, partly false. The basic idea of warming is well-established. The sun is hot and therefore gives off relatively high frequency electromagnetic radiation. This hits the earth and warms it. The earth, being cooler than the sun, then radiates at a lower frequency. Though the atmosphere is relatively clear to the high energy em radiation, it is more opaque to the lower frequency, thus keeping heat in. Too little "greenhouse gases" (water vapor, CO2, methane, and some more) and the temperature of earth goes down. Too much and it goes up. "Just right" and it stays the same--assuming you think "the same" is optimal.

But that physics, sometimes called primary warming, levels off at CO2 concentrations not that much larger than what we have today. All the worry is about secondary warming, and there the science is not at all simple or well-understood. Two of the biggies: A warmer earth means less ice, which means a darker earth, which means more radiation is absorbed, and more warming. But a warmer earth also means more clouds, which means a whiter earth, which means more radiation is bounced back into space, so less warming. No one knows just how important each effect is.

Exactly, it's not even clear when the climate was in equilibrium. The true value of ECS is probably coming into focus, but it's silly to think we know it with any certainty today.

http://www.drroyspencer.com/2018/04/new-lewis-curry-study-concludes-climate-sensitivity-is-low/

"Remember, the sensitivity of their models is NOT the result of basic physics, as some folks claim… it’s the result of very uncertain parameterizations (e.g. clouds) and assumptions (e.g. precipitation efficiency effects on the atmospheric water vapor profile and thus feedback). The models are adjusted to produce warming estimates that “look about right” to the modelers. Yes, *some* amount of warming from increasing CO2 is reasonable from basic physics. But just how much warming is open to manipulation within the uncertain portions of the models."

These sort of unanalytical models--no formal equations describing the relationship between different parameters--are very important in Neuroscience and Psych and they are becoming more prevalent in a lot of the disciplines with new "Computational ___" subfields. I'd love to see more content on how to think about these methods. It's honestly not always clear what the models are even suppose to be doing. Is it strictly predictive or are the proposed mechanisms and relations between parameters supposed to have some natural kind status?

Comments for this post are closed