SotA-R-6: Modeling Carbon Emissions – New Results, and Thoughts on Further Models

(This is part 6 of the “Stages of the Anthropocene, Revisited” Series (SotA-R).)

As mentioned in the section “limitations and alternative approaches” of the previous episode in this series, there were a couple of things I wanted to change in the model (i.e. spreadsheet) used to simulate global carbon (CO₂-equivalent) emissions in stage 1 of the anthropocene. The most important changes concern how the model (mathematically) handles the effects of various aspects of climate change on civic unrest and economic growth (and how those two interact with each other), and the way the model deals with mitigation/reduction of CO₂-e emissions.

In model 1, effects of climate change on economic growth and civic unrest where mediated by yearly changes therein and where largely cumulative. A better approach might be to multiply effects in such a way that each next effect only affects the unaffected portion of the dependent variable. The easiest way to do something like this is by means of an equation like this:

in which e is an effect (on either civic unrest or economic growth) and w is a weight assigned to that effect. The advantage of this approach is that if drought, for example, would raise the civic unrest level in a country to 60%, any further impact of economic crisis would only affect the remaining 40%, and thus cumulative effect don’t just add up (requiring an artificial 100% limit to avoid absurd results), which is probably a more realistic approach.

The other change is much more important. Model 1 had a fairly simplistic approach to simulating certain kinds of changes in emissions, which is a bit silly, as that is what the model is aimed to predict. Particularly, model 1 didn’t handle emission reductions due to mitigation efforts very well. To remedy this, the new model takes an entirely different approach to modeling CO₂ emissions and changes therein. It splits up emissions into three categories, labeled “green”, “brown”, and “black” in the model. “Green” emissions aren’t real emissions – they are what would be emitted if carbon-neutral energy sources would be replaced by CO₂-emitting sources. Hence, “green” emissions do not actually add to total emissions, but they play an important role in the model anyway. “Brown” emissions are carbon emissions that could technically be replaced by carbon-neutral alternatives. “Black” emissions are the so-called “residual emissions” for which – at the current state of technology – no carbon-neutral alternative is available. Currently, this is about 36% of total emissions according to my estimate a few episodes ago.

All three kinds of emissions (I’ll keep calling green emissions “emissions” here, even though they are only hypothetical emissions and not real emissions) change under the influence of economic growth. Additionally, there effects of mitigation effort and technological change on the shares of the three categories in the total. Technological change changes some black emissions into brown emissions, meaning that new carbon-neutral alternatives have become available. More important is mitigation effort, which is itself a function of public opinion (which is a function of the effects of climate change in turn), wealth, and economic growth. The higher the mitigation effort, the larger the share of the green emissions in emission growth (and the smaller the brown share). Furthermore, the higher the mitigation effort, the more brown emissions are replaced with green alternatives.

In addition to these two changes, I also added a direct air (carbon) capture (DAC) module to the model. DAC is the removal of CO₂ from the atmosphere, which is necessary to keep atmospheric CO₂ at an acceptable level, but which is insanely expensive and requires enormous amounts of resources and energy (which have to be carbon-neutral, of course). My main reason for adding DAC is that it is commonly assumed in IPCC models and related approaches.

Results of Model 2

Because of the differences between the two models, not all parameters work in (exactly) the same way, which makes it hard to copy settings from one to the other. (Well, copying settings isn’t hard in itself – the problem is that they really aren’t the same settings, even if they look identical). A scenario that looks fairly similar to Scenario 1 in model 1 produces quite similar results. Both model 1 and model 2 predict net zero around 2090 and average global warming of 5°C (before taking tipping points etc. into account).

Most of the effects of the various parameters are roughly similar to model 1. The surest and quickest way to make sure that the planet remains hospitable is still global societal collapse. The faster the more countries succumb to high intensity civil war destroying their emission infrastructure, the faster we reach net zero. But thanks to DAC, there now also is a peaceful path to zero emissions … as long as you believe in magic.

Let’s try the kind of superoptimistic settings that are common in many published models. We start reducing emissions right now. Even poor countries can afford green technology and carbon capture (DAC). Climate change doesn’t lead to drought, famine, violent conflict, and/or economic damage, and private debt doesn’t have any effect either. Let’s further additionally assume an absurdly unrealistic reduction speed of 8% of brown emissions per year, and fast technological development leading to a significant decrease of residual emissions (to about 3.1% by 2080). In this fantasy world, we’d be approaching (but not quite reaching) carbon-neutrality early in the 22nd century and we would be committed to a bit more than 3°C of average global warming*. (I’ll add an asterisk to remind you that this is global warming before adding the effects of taking tipping points, hard-to-predict feedbacks, etc..) If we also manage to build up a significant DAC capacity in the coming decades – enough to remove 20,000 Gt of CO₂ from the atmosphere per year – then we’d read carbon-neutrality in the early 2050s, and after that we’d be removing more CO₂ than emitting. We’d heat up the planet by about 2.5°C, but this would start to slowly decrease again from 2050 onward, unless some pesky tipping points have already locked in a higher global average temperature. Regardless, this is the best case scenario.

It is also a nonsensical scenario. It has a probability of 0%, because almost all of its assumptions are patently absurd. Firstly, we are obviously not starting to reduce emissions right now – in the contrary, emissions are still growing (although Covid-19 caused a temporary dip). Secondly, green technology is expensive and not affordable to everyone everywhere (but this might gradually change, although resource limitations will probably remain a problem). Thirdly, climate change already causes considerable damage to economies and people, and there is good reason to suspect that climate-change-induced famines and droughts will lead to (more) violent conflict and refugee flows. Fourthly, private debt does affect economic growth, and economic crises (which are caused by unsustainable private debt and are pretty much unavoidable under capitalism) wipe out economic growth completely (or even lead to decline). Fifthly, replacement of brown emission infrastructure with carbon-neutral alternatives cannot really be any faster than 5.5% (or maybe 6%) without very substantial government subsidies,¹ but governments won’t have money for that if they also have to cope with “natural” disasters, refugees, and so forth. And lastly, direct air (carbon) capture (DAC) is almost certainly too expensive (both in terms of financial and resource inputs required) and can probably never reach the high levels required. Estimates of what is needed to remove 20,000 or 30,000 Gt per year suggest ten-thousands of DAC plants (or many millions of small, car-sized DAC devices), which together slurp up a significant share of the world’s energy production. Perhaps, that’s possible in theory, but these plants or machines don’t produce anything and thus would have to be entirely government-funded, and they would compete for scarce energy and other resources with everything else. It may not even be possible to produce enough green energy (with currently known technologies and resources) for our current energy needs. Adding another 25% or 30% or so to those energy needs to remove CO₂ from the atmosphere is beyond unrealistic.

So, let’s return to Earth again, and create a scenario that is still very “optimistic”, but maybe, perhaps possible. We’re not reducing emissions yet, but we’ll start soon. The social and economic damage caused by climate change is only moderate, and we manage to use DAC to remove 5,000 Gt per year. In that scenario, we’ll reach carbon-neutrality around the end of the century and will have emitted enough carbon for approximately 5°C of average global warming*. Notice that this is almost 1.5°C more than what model 1 predicted for an optimistic, but still possible scenario!

A more pessimistic scenario, in which it will take another decade before we finally start bringing down emissions, in which there are much more substantial social and economic effects of climate change, and in which DAC never gets big enough, also suggests that we won’t reach carbon neutrality until the end of the century, resulting in roughly 5.5°C of average global warming*. This is very different from model 1 again. In that model a more pessimistic scenario lead to so much civil war and other socio-economic collapse that emissions declined sharply, leading to about 3°C of average warming*, but in model 2 climate change and civil war don’t seem to affect economic growth as much as in model 1, and because of that, don’t bring down carbon emissions as fast either.

A scenario in which the effects of drought on both economic growth and civic unrest (through food and water shortages) are much more extreme leads to carbon neutrality around 2075 and average global warming* of 4.2°C. Even more extreme settings reduces warming further by bringing carbon neutrality closer, mainly by speeding up widespread societal collapse (which would lead to billions of deaths in famines and wars). I think that such parameter settings are too extreme, however, although it is hard to judge, as there really is insufficient data to estimate realistic parameter settings.

Overall, model 2 seems to be much more pessimistic with regards to total carbon emissions than model 1, but it also shows far less variety in its results. The first model predicted a median value of 3.7°C of average global warming, but all credible parameter settings in case of model 2 lead to values around 5°C. That’s a significant difference.

Regardless of parameter settings, all credible scenarios have an emission curve that looks something like this:

(Notice that the x-axis represents 2020-2080 and that 2050 is, thus, right in the middle.) The surface below the curve represents our total future carbon emissions. Hypothetically, there are three ways to make that surface smaller. Firstly, the peak in the curve could be lower and/or sooner. However, at present there is absolutely no indication that we’re moving in that direction. Emissions keep growing, we keep building more and more carbon-emitting infrastructure such as coal plants and oil pipelines, and reduction efforts are still largely cosmetic. All of this may very well change, of course, but due to social inertia all change is slow, and therefore, there will only be a gradual change in the slope of the curve.

Secondly, the descent could be steeper in principle, but there are technological and economic reasons why that is not really possible in practice. We cannot really replace things before their normal “life span” has finished on a grand scale, because that would amount to throwing away money. People and companies cannot afford that. Some might, but that’s not enough for the “grand scale” required. Because of that, the only way to make the descent steeper is through (civil) war.

Thirdly, the line could dip below zero due to carbon capture (DAC), but as already mentioned above, that’s science fiction really. Scaling DAC to the size required may be impossible due to resource limitations and costs, but would also require an industry roughly the size as the current automobile industry, for example, and that industry would have to be funded completely out of taxes, as it wouldn’t make anything that it can sell.²

So, taking this into account, perhaps, it shouldn’t be such a surprise that there is so little variation in the outcomes of model 2. But anyway, because it is such a big difference with model 1, I have checked every part of the model several times to see whether there is some mistake or modeling artifact in model 2 that is to blame, but I haven’t found anything. In the contrary, the more I look at the two models, the more I am convinced that model 2 does a better job at simulating the relations between climate change, civic unrest, economic growth, and carbon emissions.

The Next Step(s): Some Thoughts about Model 3 and Beyond

5°C before tipping points etcera are taken into account probably means actual warming of 6°C or even more. That’s scarily much. Probably not enough for human extinction, but more than enough to make much of the planet uninhabitable. Maybe there will be enough space left for half a billion people; maybe less. If that is right, then the global human population will decrease by more than 90% during this century. However, if the effects are really so extreme, then global societal collapse will come much sooner than the model predicts, and if global societal collapse comes sooner, then total emissions will also be less (because the descent will be steeper – see above). This is an important feedback loop that could potentially drastically change the outcomes.

Right now, the effects of climate change (drought, heat waves, cyclones, and extreme weather) are independent variables in the model – that is, they are taken as given. They aren’t fixed at specific values, of course, but their gradual increase is given and does not depend on anything else in the model. This is obviously not realistic. The more and the faster we heat up the planet, the greater and the sooner the effects of drought and so forth. So, rather than taking drought etcetera for granted, the model should dynamically adjust expected drought levels (etc.) to the level of warming resulting from CO₂ emissions up to that point. Technically, this isn’t overly complicated, but it is a lot of work, and it comes with various new complications. I think that it is important to make this change, however, so there probably will be a model 3 some time in the near future. (Adding a module to simulate the climate effects of projected warming will also necessitate having another look at how the model deals with the effects of these effects – that is, how drought etcetera influence civic unrest and economic growth – so it is quite possible that there will be further changes in those aspects of the model as well.)

In addition to these approaches to forecasting carbon emissions in stage 1 of the anthropocene, there is also another method that might be worth exploring. This method is more similar to a meta-analysis: it would involve collecting every prediction of emissions until carbon neutrality I can find, and then estimating the likelihood of all those predictions. By multiplying the graphs of those various emission scenarios with their likelihoods and then adding up the results, a combined graph/prediction can be made that takes all previous predictions into account. The problem is that this only works if the likelihood of all predictions add up to 1, or in other words, if the predictions represent every possible pathway, but it is quite improbable that this is the case. In the contrary, most of the predictions I have seen are very unlikely because they make various kinds of improbable or even impossible assumptions such as infinite economic growth (and no economic crises), immediate and very quick emission reductions, widespread use of DAC or other science fiction scenarios, and so forth.

Nevertheless, such an approach would be useful if there would be a number of realistic/plausible scenarios that don’t fit well in the modeling approach employed here. For example, there are some more narrative-based predictions made by various authors. Such a narrative-based prediction could start with a scenario similar to what is predicted in the current model, for example, but then deviate therefrom for some reason. One reason for deviation could be nuclear war. Another could be what Holly Jean Buck called “desperation point”.³

Desperation Point

At COP35 in Murmansk, global leaders congratulated themselves on finally getting CO₂ emissions under control. Emissions continued to rise year after year during the 2020s, but now in 2030, it looks like we are almost reaching the point where the curve starts bending downwards. Last year, emissions increased by only 0.3%, a “huge success” according to several speakers at COP35. The press payed little attention to their celebratory mood, however. They were too busy covering the resettlement problem in the southern United States. It didn’t look like New Orleans and Houston were going to be built up again after hurricanes destroyed much of those cities for the second time in less than a decade, and millions of homeless survivors were looking for safer places to live.

Of course, there were bigger problems elsewhere. Continuous drought had turned much of the Middle East into a war zone, for example, and people fleeing drought and hunger in Africa, Central America, and elsewhere were trying to get into Europe and the US by the tens of thousands. Most were captured and locked up in camps or “accidentally” shot by heavily armed border forces, but the Western mainstream press had stopped reporting about that. There were too many floods, hurricanes, and other disasters that demanded their attention.

People were getting desperate. The yearly COP meetings were still continuing, but no one paid attention to them anymore, and no one believed that they were going to fix the problem. Some politicians were also getting desperate. Due to the authoritarian turn of the 2020s, they didn’t really have to fear criticism in the press or mass protest movements, but it was becoming increasingly clear to them that they had to do something. But what could be done? Given limited funds and technological options, it seemed that there really was only one answer: solar radiation management. The international community still officially banned any kind of geo-engineering, but a few years later, in 2032, the first planes went up, dispersing sulfates high up in the atmosphere. Soon, average temperatures started to go down, but the effects were very uneven – some countries or regions got (even) drier, others got wetter, some cooled down more, others less, and some reported an increase in skin cancer patients or crop failures.

This is just a story, of course. The point of the story is to illustrate Buck’s notion of a “desperation point”: a point in time where people and politicians get so desperate about climate change (and the lack of progress with regards to mitigation) that they try risky things like geo-engineering. Most forms of geo-engineering that have been suggested thus far are both extremely expensive and very inefficient.⁴ The sole exception is solar radiation management (SRM): reducing the amount of sunlight (and thus, heat) that reaches the planet’s surface by dispersing sulfates high up in the atmosphere or by some other artificial means. This is technically no problem and wouldn’t be very expensive either. There are a whole bunch of potential problems associated with SRM, however, and according to one study the effects on agriculture are significant enough to worry about food shortages.⁵

Many climate activists and climate scientists oppose research into SRM for two reasons. Firstly, they fear that such research will make it more likely that SRM will actually be used. Secondly, they fear that if SRM becomes a serious option, it will effectively become an excuse for slower emission reduction. I think that the first of these fears is mistaken and that we urgently need research into SRM, but not because I think that SRM might be a good idea. (In the contrary, I think it’s a terrible idea.)

Research into SRM will not make the use of SRM more likely, but might actually make it less likely. We will reach desperation point sooner or later (probably in the early 2030s, but it might be a bit sooner). It doesn’t take much contemplation on the expected increase of natural disasters and their socio-economic effects and the lack of real (rather than mere cosmetic) mitigation effort to realize that people will get desperate at some point in the not-so-distant future. And when we reach desperation point, dangerous things will be attempted. There are at least 10 countries that could do SRM on their own. Only one of them has to go “rogue” in response to domestic pressure by desperate people that are tired of disasters and inaction. So what’s the probability of that happening? Well, nothing is ever certain, of course, but I’d estimate this probability at well over 90%, perhaps even approaching (but not quite reaching) 100%. And if that is right, nothing can make SRM significantly more likely – it is already a near certainty that it will be tried. Except if research shows that it wouldn’t work, or that its negative effects would outweigh its positive effects. It is for that reason that research might actually decrease, rather than increase the likelihood of SRM. But even if it doesn’t, we’d still better do the research. If it is a near certainty that SRM will be used, than we’d better figure out how to use it wisely before starting.

So how about the second fear? Would availability of SRM as a “real” option lead to less reductions? Maybe. It’s hard to say. It depends on whether such excuses can only be based on science and not on science fiction. Right now pretty much any scenario discussed by politicians and the mainstream press relies on science fiction. The hypothetical possibilities of SRM, DAC, and other technologies are already effectively excuses to make very limited mitigation effort. I seriously doubt that knowing more about such technologies (i.e. changing them from science fiction into science) would significantly change that. Whether the world makes a serious effort to reduce CO₂ emissions depends primarily on power, and those who are in power (see Enemies of Our Children) can easily create excuses for any course of action that serves their interests.

Back to “desperation point”. I think it is a virtual certainty that we will reach desperation point. This matters, because something just might be different after that point. Perhaps, there will be a sudden explosion of civic unrest. Perhaps, weird and dangerous things will be tried. Perhaps, a combination of these and other effects lead to a world war. Perhaps, people will just give up and let the planet burn. Or perhaps, there will be no significant effects at all. It’s impossible to say what will happen, but some scenarios would lead to very different outcomes than what any mathematical model can predict.

Closing Remarks

We now have two sets of predictions of carbon emissions and their effects in stage 1 of the anthropocene. Model 1 suggests that cumulative emissions will result in something like 3.7°C average global warming (before taking tipping points and hard-to-predict feedback effects into account) with very large uncertainty margins. Model 2 suggests that it will be closer to 5°C with much narrower uncertainty margins. The next step will be the development of a model 3 to see whether these results can be further refined. The main aim of model 3 is to incorporate the feedback loop from emissions to socio-economic climate change effects. (I expect that the effect of adding that feedback loop is a lowering of predicted warming.) Additionally, it might be a good idea to try and build some kind of meta-model that integrates the various predictions available (including narrative-based predictions) into a single meta-prediction (although this would be a lot of work).

The aim of this series is to predict the effects of climate change in the more distant future, in stages 2 and 3 of the anthropocene. As mentioned repeatedly, that is only possible after estimating the total emissions in stage 1, but if building model 3 doesn’t take too long, then this phase of the SotA-R project might be reached soon. Fortunately, there are now some climate scientists looking into this topic as well. Just last Friday (Sept. 24) a paper was published by Christopher Lyon and colleagues that attempts to predict the effects of climate change in the year 2500.⁶ I haven’t had time to read this paper yet, but regardless of what it says, I’m very pleased that (at least some) climate scientists are finally paying serious attention to the longer-term effects of climate change.

If you found this article and/or other articles in this blog useful or valuable, please consider making a small financial contribution to support this blog, 𝐹=𝑚𝑎, and its author. You can find 𝐹=𝑚𝑎’s Patreon page here.

1. You might remember a percentage of 3.5% from previous episodes, but that was relative to total emissions. 3.5% of the total equals 5.5% of brown emissions.
2. There are some uses for captured carbon, but the potential demand for those products is negligible relative to the amount of carbon that needs to be captured.
3. Holly Jean Buck (2019), After Geoengineering: Climate Tragedy, Repair, and Restoration (London: Verso).
4. Which is one of the main points of Buck’s book. See previous note.
5. Jonathan Proctor, Solomon Hsiang, Jennifer Burney, Marshall Burke, & Wolfram Schlenker (2018). “Estimating Global Agricultural Effects of Geoengineering Using Volcanic Eruptions”, Nature, August 8.
6. Christopher Lyon et al. (2021, in print), “Climate Change Research and Action Must Look beyond 2100”, Global Change Biology.