FANAQ

Frequently And Never Asked Questions

Questions that people often ask, or never thought to ask, after reading The Climate Demon (or even before reading it).

Have questions?

(#ClimateDemon)

(Note: Additional questions are answered on the Tweets page.)

Q1. Is global warming a Black Swan event? (p.87)

No. Technically, a Black Swan event should be completely unexpected and most people expect global warming to some degree. (The Antarctic Ozone Hole, on the other hand, was a Black Swan event because it was completely unexpected.)

Q2. Is it too late avoid catastrophic climate change (like failing to deflect an incoming asteroid)?

No, it is not too late to act. But it is urgent. As portrayed in movies like Don't Look Up, extinction-level impact of a comet or an asteroid on the Earth is a binary threat – hit or miss. Global warming is a more diffuse threat that gets much worse over time (p.117). If we can significantly reduce carbon emissions soon, we can limit global warming and avoid its worst effects (p.308), although it will be hard to reverse changes that have already occurred (and sea level will continue to rise slowly even after the warming stops). Due to the long atmospheric lifetime of carbon dioxide (p.152), every year of delay in reducing its emissions leads to accumulation of its hard-to-reverse climate impacts. The solutions to reduce carbon emissions associated with our everyday activities are rather complex, though, compared to the simpler task of using explosions to deflect a far-away asteroid.

Q3. What about Thwaites ("doomsday") Glacier, permafrost thaw ("methane bomb") or thermohaline "collapse"? How worried should we be about these and other potential tipping points? (p.58, 137)

Climate twitter has some illuminating, and not too scary, discussions relating tipping points. For some background on the Thwaites Glacier that was recently in the news, see this Twitter thread. Arctic permafrost thaw is discussed in this Twitter thread. "Runaway" greenhouse effect and thermohaline/AMOC "collapse" are discussed in this Twitter thread. (Note: The wind-driven Gulf Stream current is itself in no danger of collapsing, but the Atlantic Meridional Overturning Circulation driven by density gradients may potentially weaken or even shutdown.) A summary of the recent IPCC assessment of abrupt climate change can be found in this Twitter thread.

The overarching message from the discussions is that while there are multiple mechanisms and feedbacks that could lead to global tipping points, it is hard to make quantitative predictions of the likelihood and timing of crossing these tipping points. Therefore, we will need to discuss their risk in a more qualitative fashion. It is worth noting that many of the feedbacks involved in these potential tipping points are relatively slow, acting over centuries rather than decades. This makes it hard to build and verify models that represent these slow feedbacks. But on the upside, we will have more time to respond to climate changes triggered by the slow feedbacks, even if some of them turn out to be irreversible.

In any case, the solution that reduces the risk of crossing poorly understood tipping points is exactly the same as the solution that mitigates the better undersood global warming — a rapid reduction in carbon emissions. It is the emission reduction that we need to stay focused on, because the key to happiness is to only worry about things that we have control over.

Q4. If carbon emissions are currently "tracking" a particular IPCC scenario, does it mean that we are likely to end up in that scenario in the year 2100? (p.117, 223, 302, 305)

No. Even though they are often graphically displayed as curved trajectories as if they were predictions (e.g., Fig 7.4a; p. 120), scenarios are not predictions. Unlike true predictions such as weather forecasts, there is no mathematical equation that constrains the world to continue to "track" a particular scenario, even if it is currently tracking it. For example, Scenario 1 might say that X will happen now, then Y will happen, and then Z will happen; Scenario 2 might say P will happen now, then Q, and then R. Even if X is happening now, Q could happen next, because human actions (and unforeseen events) can switch us from one scenario to another.

IPCC emission scenarios should be viewed as a discrete (and small) subset from a continuum of future possibilities. The actual emission trajectory will traverse the continuum, sometimes tracking one scenario, sometimes another, and sometimes in the gaps between scenarios. If scenarios are like roads, then the real world is an off-road vehicle, not a car.

Each scenario is merely a sequence of "what if" statements, like the assumptions that you might make make when crafting the fictional narrative for a novel. Socioeconomic experts try to ensure that these "what if" statements is plausible, even if the sequence is fictional, by extrapolating based on our past experience. For this reason, the range of IPCC scenarios may end up excluding future possibilities that are beyond our experience.

Q5. Why do climate contrarians often appear to be more trusting of numbers from (certain) economic models than climate models, even though the latter are more scientifically constrained? (p.249)

The answer may lie in the question itself – it may be simply because climate models are more scientifically constrained. As a discipline located at a further deductive distance from physics, and therefore more inductive and empirical than climate science, economic predictions will be more uncertain than climate predictions. (Recall Harry Truman's plea for "a one-handed economist.") It is perhaps easier to find an economic model prediction to confirm one's ideological priors than to find a climate model prediction, and thus craft a narrative that appears to have a quantitative basis. [This is not to discourage the use of economic models. Economic models are essential for making short-term predictions and for long-term scenario construction. One just needs to be aware of the quantified and unquantified uncertainties associated with different types of model predictions at various timescales.]

Q6. What about the IPCC Sixth Assessment (AR6)?

The book does not discuss any IPCC AR6 results because the proofs were completed before the Working Group 1 Report was released. But the IPCC report is based on published literature and it was not hard to guess what some of the important conclusions were likely to be. For example, the AR6 likely climate sensitivity range (2.5–4.0) closely tracks the 2020 WCRP Assessment likely range (2.6–3.9) discussed in the book (p.222. 240). Also, the AR6 projections constrain model results based on certain criteria that discount models with high simulated climate sensitivies. This ex post facto imposition of what is effectively a retroactive tuning mandate for model acceptability (p.175) ends up shackling the CMIP6 models. The net result is that the AR6 temperature projections (Figure 4.11c of AR6-WG1) look rather similar to the AR5 projections (Figure 7.4c of the book).

Past IPCC assessments relied primarily on the deductive approach to estimate key global parameters such as climate sensitivity, by relying on comprehensive climate models. To estimate key climate parameters, AR6 instead relies on analyses of observational data using an assortment of simple physical models coupled with statistical models. What the book speculated might be "an implicit rebuke of the latest generation of models with higher climate sensitivities" (p.222) by the IPCC turned out to be a rather more explicit rebuke! The use of the inductive approach represents a fundamental philosophical shift in the IPCC assessment process. While this approach provides a narrower estimate of climate sensitivity, it suffers from drawbacks. For example, the unprecedented nature of the current global warming means that inductive approaches may not be able to predict the warming accurately.

If you feel that this book may have been too harsh on comprehensive climate models, the AR6 report is in some ways even harsher (especially if you read between the lines). For instance, the book recommends selecting a set of acceptable comprehensive climate models and aggregating their simulations them to make climate projections (p.298). The AR6 report instead uses a hybrid (and somewhat inelegant) approach to making climate projections that involves averaging emulations based on analysis of past data with constrained simulations from comprehensive models (Figure 4.11d of AR6-WG1).

The broad qualitative conclusions of the book, though, remain consistent with the findings of the AR6 report – that's the nice thing about taking climate models seriously but not literally!

Q7. Was the extreme Texas winter storm of February 2021 caused by climate change? (p.112)

It is the wrong question to ask, although some in the media and government were quick to blame climate change for this extreme winter storm. As the book argues, one shouldn't generally try to attribute specific weather events entirely to climate change or global warming. A better question to ask would be "did global warming make the 2021 extreme winter storm more or less likely?" A related question is whether global warming made the storm stonger or weaker (in a statistical sense)? This question was addressed in the 2021 Texas state climatologist's report that assessed trends in extreme weather in Texas using data. The report concludes that "[a]s the climate warms, the likelihood of winter weather decreases. Both extreme cold and snowfall either become less frequent or are expected to do so. Widespread snowfall events in Texas such as the one that took place in February 2021 are extremely rare." The figure below shows the coldest temperature recorded each year in Texas since 1900. We see that the extreme cold temperatures are warming faster than the average temperatures. Therefore, the data suggest that global warming perhaps made the 2021 winter storm just a tiny bit less severe, as one might intuitively expect. It is important to note that the report also says that heat extremes in Texas (see the other figure below) as well as intense rainfall events are made more severe, and more likely, by global warming. The take home message is that although many extreme weather phenomena are made worse by climate change, not all are.

One may think there no harm done in "overatrributing" an extreme weather event to climate change — after all, it merely underscores the seriousness of the climate crisis and can help accelerate the urgently needed mitigating actions by increasing public awareness. But apart from the scientific argument against overattribution, there are also potential negative socieoeconomic outcomes because disasters happen when "hazards meet vulnerability." Overattribution may allow politicians and governments to evade blame by shifting the focus from the local problem of socioeconomic vulnerability to the global problem of climate change. In the case of the 2021 winter storm in Texas, serious local vulnerabilities were responsible for the disastrous impact of a rare, but not unprecedented, winter storm. As Lahsen and Ribot argue in their 2021 paper Politics of attributing extreme events and disasters to climate change, "[a]nalytic frames that attribute disaster to climate can divert attention from these place-based vulnerabilities and their socio-political causes."

Trends of cold and hot extreme weather in Texas (from Nielsen-Gammon et al., 2021)

Q8. Can you explain stochastic uncertainty ("certain uncertainty") some more? (p.115, 124, 297)

Consider weather prediction in the atmosphere (for starters). We use a weather model to make an ensemble of forecasts, starting from slightly different initial conditions to represent the error in our knowledge of the initial atmospheric state (p.43). For the first few hours, the forecasts will be very similar to each other, but then they will start diverging due to chaotic error growth or the Butterfly Effect (p.40; see this TED-Ed video). After about 2–4 weeks, all memory of the initial condition will be lost and each forecast can be considered to be a random (or stochastic) sample from all possible weather states. For climate prediction, we also need to include initial conditions in the ocean (also land and sea-ice). Since the ocean evolves much more slowly, it takes a decade or longer for the memory of the initial condition to be lost (p.116). But eventually, after a few decades, the ensemble of forecasts will end up being a collection of random samples from all possible climate states. The spread of among these states is the natural (or internal) climate variability, and can be considered unpredictable "climate noise" or stochastic uncertainty.

Suppose we add a long-term climate trend, such as global warming, to this system. The actual climate in the year 2060 (say) will be the sum of the predicted long-term trend ("climate signal") and the unpredictable climate noise. As NCAR climate scientist Clara Deser explains in her 2020 paper on climate variability: “anthropogenic climate change is what you expect; anthropogenic climate change and natural climate variability is what you get.” The figure below from the paper shows two samples from a 40-member ensemble of 50-year (2010-2050) predictions using the NCAR global climate model for a high carbon emission scenario. The top panel shows the expected (ensemble-average) change in winter precipitation over North America. i.e., the signal. We note that precipitation is expected to increase both in the East and the West Coast of the US. The bottom two panels show two individual ensemble members, which span the 5th to 95th percentile range of the actual climate states we may get, i.e., the signal plus two extreme realizations of noise. In the left panel, we see even stronger increase in West Coast precipitation, whereas the right panel actually shows a decrease in precipitation over the same region. When planning for the future, we will need to take into account the whole range of possibilities associated with this "certain uncertainty," especially for regional climates (p.124, 297)

Predicted change in winter precipitation between 2010 and 2060, measured in millimeters/day. Brownish colors indicate decreased rain/snow and the greenish/bluish colors indicate increased rain/snow (From Deser, 2020)

Q9. What is the "uncertainty trough"? (p.244)

"Uncertainty trough" refers to the notion that as the distance from knowledge production increases there is a minimum (or "trough") in the awareness of the uncertainties in the product before it increases again (see figure below and also this Twitter thread). The "uncertainty trough" was first studied in the context of missile production (see figure below). Engineers directly involved with the design and testing of missiles were more aware of their shortcomings than managers who were only superficially familar with the product. This uncertainty trough is frequently (and somewhat confusingly) referred to as the "certainty trough", because in many fields, such as missile deployment, certainty is the metric of interest.

In the context of climate prediction, the uncertainty trough may be more applicable to climate modelers as a group, rather than to individual climate modelers. This recent Twitter discussion provides an illustration of the uncertainty trough in climate modeling. At the end of the 2021 COP26 climate conference in Glasgow, it was widely reported that the conference had managed to keep the goal of limiting to global warming 1.5 degC alive. It turns out that this statement relies upon certain carbon sinks persisting after zero emissions is reached. Although some experts on carbon sinks believe this will be the case, and carbon sink models support this idea, other experts weren't that sure because we have no actual experience of how carbon sinks will behave in a "zero emission" world. (When relying on inductive models, we may not be able to confidently extrapolate them to unprecedented scenarios.)

[Figure from Pearce (2020), who adapts it from MacKenzie (1990) ]

Q10. What is "plausible"?

Continuing our analysis of climate-related terms that we began in Chapter 14 of the book, we note that "plausible" is a catchall term that frequently appears in discussions of climate change. Emission scenarios are described as being plausible or implausible (p.223). Socioeconomic models often make plausible assumptions when extrapolating scenarios into the distant future (p.117). Statistical models also frequently make plausible assumptions when processing observational data (p.217). Reduced-complexity climate models typically use plausible simplifying assumptions (p.71,132). Comprehensive physics-based climate models should not, in principle, rely on plausible assumptions, but, in practice, the representations ("parameterizations") of processes that are not resolved by the coarse model grid (p.68) often involve some unverified assumptions.

Unfortunately, "plausible" – typically used in the sense of being "believable" – is not a scientific term. Believability of socioeconomic scenarios can change over time, for example, as the world evolves (p.223). Plausible assumptions are not usually tested individually like scientific hypotheses. Relying on plausible assumptions in any quantitative analysis will therefore introduce some degree of unquantifiable structural uncertainty. We can try to characterize this uncertainty by considering the sensitivity of the analysis results to a variety of plausible assumptions, but we cannot be sure that we have exhausted all possible plausible assumptions.

In many aspects of climate prediction, necessity often dictates that we go down the path of making plausible assumptions. But following the path of plausibility requires that we leave behind our "physics envy", that is, the urge to ascribe objective probabilities or unconditional error bars to climate (and socioeconomic) parameters of the future. Once uncertainty estimates become conditional on unverifiable plausibilities in an analysis methodology, any associated error bars may need to be considered measures of precision rather than measures of accuracy (p.243). A symptom of this problem is that alternative plausible methodologies can give varying estimates for the same parameter (like climate sensitivity), with non-convergent error bars.

An emission scenario assumes a sequence of plausible events (p.117). Suppose there is an implicit probability associated with the plausibility of each event, then probability of the whole scenario would be the product of the implicit probabilities of the plausible events happening in the correct sequence. This means that the trajectory of any single scenario will have a very low probability of occurrence, and therefore scenerios should be considered discrete samples from a continuum of possible futures (p.302). This is quite different from deterministic prediction, such as weather prediction, where the forecast trajectory is determined by mathematical equations over time (p.27). The equations predict exactly what sequence of weather events, such as cold front, warm front, or a tropical cyclone, will occur (up to the deterministic predictability limit). Scenarios are rather like trying to predict weather well beyond the predictability limit, by simply assuming that a sequence of plausible weather events will occur. Therefore, one should always consider a range of scenarios that span all plausible sequences of plausible events (p.297).

Q11. What about psychohistory? Can we predict the evolution of society and thus the likelihood of emission scenarios?

Psychohistory is an interesting concept introduced by science-fiction writer Isaac Asimov in his Foundation series. The conceit is that one can mathematically (and accurately) predict how the history of a large population will evolve, even if one cannot predict individual human behavior. But psychohistory is just science-fiction that ignores the limitations of quantum uncertainty and chaos theory (p.31). There is no mathematical model that can accurately predict the evolution of entire societies, or of specific emission scenarios (p.294). Political economists and futurologists do try to intelligently analyze plausible futures, but it is a fraught exercise (as books with titles such as The End of History demonstrate).

Q12. The age-old question: Star Trek or Star Wars, which is better? (p.37)

The book clearly prefers the episodic Star Trek (nondeterministic) over the epic Star Wars (too deterministic). Spock will likely provide better advice to deal with climate change than Darth Vader (or Han or Luke or Leia, for that matter – although older Obi Wan may do the job), even though Spock has been criticized for assuming that others will act rationally like him when picking the optimal solution. [See also the Seriously, not literally Visual.]

Q13. Should I take the red pill or the blue pill?

The red pill, of course, unless you believe in the infallibility of predictions. This question, from the sci-fi film The Matrix, represents the choice between living in the "comfort" of a machine-controlled simulation (blue pill) or escaping into the uncertain future of humanity (red pill). It can be considered an allegory for the philosophical choice between fate and free will – between a predictable deterministic future that is immutable or a non-deterministic future with role for human agency (p.31). Note that non-deterministic does not mean probabilistic, because we may not even be able predict the probabilities of different futures accurately (p.117, 296).

Scientists are like the protagonist Neo in The Matrix, trying to hack the computer-simulated Universe (p.166). Climate modelers, in particular, are trying to reverse engineer the Climate function for a single planet in the Universe, by creating approximate versions of this highly complex Climate function in their computer models. Our imperfect climate predictions are conditional on the even more imperfect socioeconomic projections (p.226). These predictions/projections provide us knowledge, but not certainty, as we make decisions that will determine our future (p.301).

Geoengineering (p.280) is sometimes referred to as climate hacking, because if we can hack the Climate function, we can control planetary climate (the equivalent of Neo learning to fly in The Matrix). There is no evidence (thus far) that we can successfully hack the humongous Climate function, and hopefully we won't need to try it as a last resort.

And no, there is no such thing as "The One" true climate model.

Q14. The book's Epilogue references the novel Dune. The book was published on the same day that the new Dune movie premiered in the US (Oct. 21, 2021). Coincidence? (p.314)

Lacking spice, I could not have foreseen it. The style of the book, with its numerous epigraphs – some of them made-up – is indeed influenced by the science-fiction classic Dune, which explores the power of prophecies and the theme of planetary climate change. The Bene Gesserit Litany from Dune may even be good psychological advice for coping with climate change anxiety!

Q15. Did you know that climate modeling and Suki Manabe would win the Physics Nobel Prize in 2021? (p.67)

No. The book was sent to the printers a month before the Nobel announcement. I didn't expect that climate modeling would be recognized with a Physics Nobel, as you can tell from how Manabe is introduced in Chapter 4 of the book:

Syukuro “Suki” Manabe was born in rural Japan, the son and grandson of medical doctors. When he went to the University of Tokyo, it was assumed that he would follow in their footsteps. But Manabe despised biology and soon switched fields to physics. Figuring that he was not smart enough to be a theoretical physicist, and not handy enough to be an experimental physicist, he decided to study geophysics.

Here's Manabe's own reaction to the Nobel award:

When I got the phone call this morning, I was so surprised. Usually, the Nobel Prize in physics is awarded to physicists making a fundamental contribution in physics. Yes, my work is based on physics, but it’s applied physics. Geophysics.

You could say that the Nobel award was a Black Swan event, of the pleasant kind.

Q16. Can you really buy insurance to protect against the zombie apocalypse and/or alien abductions? (p.291)

Q17. What is this Columbo thing? (p.104)

It is a detective show from the 1970s that is not a "whodunit", but a "howcatchem" – each episode "a murder mystery where the murder was no mystery."

Q18. What about the multiverse? How does that affect our notion of predictability?

Popular in blockbuster movies and in theoretical physics, the multiverse has been likened to a loaf of bread where our universe is just one slice. Presumably Occam's Razor wasn't used to slice this loaf, because it would be complicated to reconcile the notion of a multiverse (and related concepts like timeline manipulation) with the Butterfly Effect of chaos theory.

Q19. Shouldn’t this book be called The Climate Demon Demon, to be consistent with its subtitle?

Yes, and if you have figured this out, you must be a logic fiend as well as a Laplace's Demon aficionado – I didn’t realize this meta inconsistency until the book had been sent to the printers. In any case, I’m sure the editor would have rejected this accurately recursive but obviously clunky title. [A bit of history: I kept a few random notes over the years under a file labeled Occam’s Beard for a book with a broad scope dealing with general issues of complexity. By the time of the first draft, the scope had narrowed to a topic that I actually knew something about, i.e., climate prediction, and the working title became Butterflies in the Greenhouse. Worried that the book may end up being shelved in the gardening section, I used a forgettable working title when submitting it to the Cambridge editor, Matt Lloyd, who, after reading the draft, suggested The Climate Demon.]