# Uncertainty about the Climate Uncertainty Monster

by Judith Curry

The many dimensions of the climate uncertainty monster.

Bret Stephens’ climate change op-ed of several weeks ago Climate of Complete Certainty spawned a number of articles related to uncertainty and climate change.

Andy Revkin’s article in response was titled There are lots of climate uncertainties.  Let’s acknowledge and plan for them with honesty.    Revkin even mentions the Uncertainty Monster and Jeroen van der Sluijs.

While the Uncertainty Monster (or  Mr. T) should be pleased at the mentions, there are numerous misconceptions among those that are trying to give climate uncertainty its due attention.

Let’s take a look at some of these issues.

Certainty, probabilities, uncertainty,  and ignorance

If you need a refresher on the Uncertainty Monster, see my original post

Bret Stephens made the following statement

Anyone who has read the 2014 report of the Intergovernmental Panel on Climate Change knows that, while the modest (0.85 degrees Celsius, or about 1.5 degrees Fahrenheit) warming of the earth since 1880 is indisputable, as is the human influence on that warming, much else that passes as accepted fact is really a matter of probabilities.

Loosely speaking, yes climate change is a matter of probabilities.  However, mathematically speaking,  ‘probability’ already implies a great deal of certainty — that we have a well-defined pdf that includes all possible results.  This is certainly not true of very much in climate science, particularly related to attribution and 21st century projections.

From my paper Reasoning About Climate Uncertainty:

Statistical uncertainty is distinguished from scenario uncertainty, whereby scenario uncertainty implies that it is not possible to formulate the probability of occurrence particular outcomes. A scenario is a plausible but unverifiable description of how the system and/or its driving forces may develop in the future.

Stainforth et al. (2007) argue that model inadequacy and an insufficient number of simulations in the ensemble preclude producing meaningful probability distributions from the frequency of model outcomes of future climate. Stainforth et al. state: “[G]iven nonlinear models with large systematic errors under current conditions, no connection has been even remotely established for relating the distribution of model states under altered conditions to decision-relevant probability distributions. .  . Furthermore, they are liable to be misleading because the conclusions, usually in the form of PDFs, imply much greater confidence than the underlying assumptions justify.”

Insufficiently large initial condition ensembles combined with model parameter and structural uncertainty preclude forming a PDF from climate model simulations that has much meaning in terms of establishing a mean value or confidence intervals. In the presence of scenario uncertainty, which characterizes climate model simulations, attempts to produce a PDF for climate sensitivity are arguably misguided and misleading.

Back to Brett Stephens’ statement: much else that passes as accepted fact is really a matter of probabilities.

While I appreciate the distinction that Stephens is trying to make regarding ‘its not certain’, our understanding of future climate change is NOT a matter of probabilities.   Climate model projections and IPCC conclusions are possible future scenarios, and the uncertainties are too great to even come close to assessing probabilities.

I am very pleased to see that Andy Revkin has been engaging with the Society for Decision Making Under Deep Uncertainty and attended a recent meeting.  Revkin discusses the unknowability of future regional climate change.  However, Revkin’s acknowledgement of uncertainty extends mainly to the impacts:

Of course, no one there questioned the basic science identifying a growing human impact on climate from the buildup of carbon dioxide and other greenhouse gases in the atmosphere. But as is well known in the scientific community, while the climate basics have long been clear, many of the most consequential aspects of climate change remain shrouded in uncertainty.

Despite three decades of intensifying analysis using ever more sophisticated computer simulations and observing systems and vast troves of data gleaned through the passage of time, two of the most basic questions remain enduringly unclear: the pace and extent of warming from a given rise in CO2 and the resulting rate of sea-level rise as ice sheets deteriorate. Through 2100 or so, either could be disastrous or manageable.

There is some implicit acknowledgement about uncertainty in the rate of warming (associated with uncertainties in climate sensitivity to CO2); but no acknowledgement of the uncertainties about the broad range of causal mechanisms (internal and external) for climate change.

So it is good to see these acknowledgements of uncertainty in journalistic thought leaders in the climate debate.  But the Uncertainty Monster is a tricky dude, it doesn’t help to oversimplify him.

Uncertainty monster simplification

Chad Orzel has an article in Forbes:  Probability is more certain than you think:

Again, the quantitative nature of scientific uncertainty undermines this– there are no shruggies in science, but rather a range of possible outcomes, with quantitative probability estimates for each. Even if you don’t believe the worst claims for climate change, you can do way better than “Well, maybe it won’t happen…” Choosing not to make the effort to engage quantitatively is both lazy and dishonest.

Well, maybe the lazy and dishonest people are those who oversimplify deep uncertainty and ignorance by trying to quantify it.

There are most definitely ‘shruggies’ in science: the knowledge frontier and the unknown unknowns.

Understanding uncertainty associated with the complex, nonlinear and chaotic climate system, let alone managing it, is a very challenging endeavor.  Hence it is tempting for scientists and policy makers to simplify uncertainty to make it appear that the appropriate considerations have been undertaken.  For a previous post on this topic:

The IPCC oversimplifies the characterization of uncertainty by substituting ‘expert judgment’ for a thorough understanding of uncertainty.  They look at ‘evidence for’ and ‘evidence against’ (but somehow neglect a lot of the ‘evidence against’), and completely neglect to acknowledge ignorance.

Formal efforts at Uncertainty Quantification (e.g. regarding climate models) are a useful step, but are only scratching the surface of the uncertainties and neglect major aspects of structural uncertainties of the models.

The bottom line is that the climate system is too complex with myriad uncertainties for simple reductionist approaches to understanding and managing uncertainty to be useful.

Too much uncertainty?  Too much certainty?

There is a perception that uncertainty equals inaction.  The appropriate way to view the decision making under uncertainty challenge is summarized in this tweet by Silvio Funtowicz:

It’s not about certainty in probabilities, but how salient is the uncertainty in relation to a portfolio of policy options.

A recent paper found that upper midwest farmers say that there is too much uncertainty in climate change to justify changing their agricultural practices.

On the other hand, when there is too much certainty in a prediction, the results can be substantial losses:  English vineyards (planted with the expectations of warmer temperatures) were hit by frost this spring, wiping out half of the harvest [link]

Simple linear decision making can lead to decisions that do more harm than good.

Does more uncertainty increase the imperative for action?

Economists Gernot Wagner and Martin Weitzmann in their book Climate Shock argue that more uncertainty increases the imperative for climate action [link to FT article]:

The challenge is “almost uniquely global, uniquely long-term, uniquely irreversible and uniquely uncertain”. The book’s big contribution is on the last point: uncertainty. Climate change is a problem of insurance. For this, it is not median outcomes that matter most, but the outliers — the “fat tails” of the probability distribution of temperature.

Framing the challenge of climate change as a problem of insurance against disaster is intellectually fruitful. It also provides the right answer to sceptics. The question is not what we know for sure. The question is rather how certain we are (or can be) that nothing bad will happen. Given the science, which is well established, it is impossible to argue that we know the risks are small. This being so, taking action is logical. It is the right way to respond to the nature and scale of possible bad outcomes.

The manufactured ‘fat tail’ comes from confusing statistical uncertainty (with a possible infinite fat tail) with scenario uncertainty, which is limited on the upper end by articulation of a plausible worst case scenario.  I have written several posts about the flaws in Wagner and Weitzmann’s argument:

Climate of unintended consequences

So for the sake of argument, lets say we buy Gernot and Weizmann’s argument, and we do something.

What if the ‘cure’ is worse than the ‘disease’?  Bret Stephens wrote a follow on op-ed entitled Climate of Unintended Consequences, where he points out the folly of biofuels/ethanol to help address the problem of climate change.  There is also the diesel fuel example.

There is a new documentary that will air on May 18 in the UK — The Uncertainty has Settled [link]. Excerpt from the advert:

After eight years of travelling through conflict and poverty zones, Marijn Poels – a left wing filmmaker/journalist – decides to take some time off. In the Austrian mountains no less. It confronts him unexpectedly with the roots of agriculture and its modern day perspective. Globalisation and climate politics are causing radical changes such as farmers becoming energy suppliers. But the green ideology raises questions. The scientific topic of climate change has now become incontrovertibly a matter of world politics. Poels faces a personal conflict. Are we doing the right thing?

By radically changing global energy policies in a top down way, are we risking continued poverty in the developing world? Risks to our food supply?

And finally, what are the opportunity costs for focusing on this problem at the expense of others?

Value clashes and opportunity costs

One characteristic of decisions under deep uncertainty is that there are value clashes involved — cost-benefit analysis does not capture the full dimension of the concerns.

For the sake of argument, lets say that we actually believe the climate model predictions and the assessments of costs by economists.  Should we then act, i.e. spend the money to address this problem?

Well, spending money on the climate change problem has opportunity costs, i.e. the money then doesn’t get spent on other problems.

Alex Berzerow has an interesting article:  Are microbiologists climate-denying science haters?  Excerpt:

Recently, I gave a seminar on “fake news” to professors and grad students at a large public university. Early in my talk, I polled the audience: “How many of you believe climate change is the world’s #1 threat?”

Silence. Not a single person raised his or her hand. Was I speaking in front of a group of science deniers? The College Republicans? Some fringe libertarian club? No, it was a room full of microbiologists.

How could so many incredibly intelligent people overwhelmingly reject what THE SCIENCE says about climate change? Well, they don’t. They just don’t see it as big of a threat to the world as other things. Unsurprisingly, the vast majority of them felt that antibiotic resistance and pandemic disease were the biggest global threats. One person thought geopolitical instability was the biggest concern.

I told them that I believed poverty was the world’s biggest threat.  If we fix poverty, we could stop easily preventable health problems, such as infectious disease and malnutrition.

What so many in the media (and apparently the climate science community) fail to understand is that people have different values and priorities. Foreign policy analysts are terrified of North Korea. Economists fear Brexit and a Eurozone collapse. Geologists, especially those in the Pacific Northwest, fear a huge earthquake. Experts across the spectrum perceive threats differently, usually magnifying those with which they are most familiar.

In his response to his inaugural NYTimes op-ed [link], Brett Stephens writes:

The human race is forced to confront multiple environmental threats with limited economic resources. We have to make hard choices about how we assess the threats and how we allocate the resources — knowing all the time that information is imperfect and economic and environmental conditions are subject to change over time. Climate change is one of those threats, but not the only one: think of malnutrition, “ordinary” pollution, land mismanagement and so on. We need a serious debate not only about how to allocate those resources, but also about whether we have the tools right now to make a switch to less carbon-intensive energy sources in a way that doesn’t impose its own set of grave and unanticipated economic and environmental problems.

In other words, to say we want to take out insurance for climate change is perfectly sensible. But whether we know we’re buying the right insurance, at the right price, is less clear, and it behooves us to look closely at the fine print before we sign on.

Decision making under deep uncertainty

So how should we deal with risks associated with human caused climate change? To say with full confidence that there are no risks is simply wrong.  We can argue until the cows come home about how likely catastrophic risks are, but there remains the possibility of catastrophic risks — I have argued that the uncertainty is too great to assign a probability to this; possibility is the appropriate likelihood.

From my post Permanent paradigm paralysis:

In their Wrong Trousers essay, Prins and Rayner argue that we have made the wrong cognitive choices in our attempts to define the problem of climate change, by relying on strategies that worked previously with ozone, sulphur emissions and nuclear bombs. While these issues may share some superficial similarities with the climate change problems, they are ‘tame’ problems (complicated, but with defined and achievable end-states), whereas climate change is ‘wicked’ (comprising open, complex and imperfectly understood systems). For wicked problems, effective policy requires profound integration of technical knowledge with understanding of social and natural systems. In a wicked problem, there is no end to causal chains in interacting open systems, and every wicked problem can be considered as a symptom of another problem; if we attempt to simplify the problem, we become risk becoming prisoners of our own assumptions.

Simply put, the current focus on CO2 emissions reductions risks having a massively expensive global solution that is more damaging to societies than the problem of climate change.

The precautionary principal is by no means the only decision analytic framework to use under conditions of deep uncertainty, see these previous posts:

The Hartwell paper has a very insightful summary

‘Muddling through’ at the local level is probably what will happen.  We can be smart about muddling through, as illustrated in this blog post at deepuncertainty.org.
.
JC reflections

The ‘Uncertainty Monster’ was the theme that launched this blog in 2010. There have been some very significant advances in our thinking on this topic, both in the scientific community and our thinking about decision making.

However, much of this has not trickled up into the UNFCCC/IPCC and the public debate on climate change.  Brett Stephens and Andy Revkin can play an important role on the media side of all this.

The climate Uncertainty Monster is maturing, and he is demanding that we pay attention.