Simon D.A. Thomas

PhD Thesis

Written in March 2024. This post is now a little out of date.

Originally published at CamxRisk: Nuclear Winter.

Written as a Capstone project as part of the CERI futures fellowship.

Nuclear Winter

In late September, after a series of provocations, a full scale nuclear war begins between Russia and NATO. The conflict steadily escalates from strikes against major military installations to the complete destruction of all major cities in their combined territories. As wildfires rage across the Siberian plains and mushroom clouds clear from major cities, how will this affect countries that remain intact? Should we expect this to precipitate human extinction? A small minority of the world would have been killed in these strikes, so the secondary effects would have to be devastating to achieve this.

Chief among the proposed secondary effects is nuclear winter. As first described, it relies on firestorms engulfing cities and forests that are so intense that they inject soot into the stratosphere, where it remains for years, causing catastrophic global cooling of 15-20K, or around three times larger than the temperature difference from the height of the last glacial maximum to today (Covey, Schneider, and Thompson, 1984). As plants require heat and light, this would cause a very large reduction in farming production in most surviving areas. In the cold, dark, and hungry world that remains, all societal controls are dissolved and all those without sufficiently resourced bunkers are dead before the end of the nuclear winter.

But is this realistic?

Other studies argued that burning all relevant cities would produce insufficient soot, and that almost all of that soot would never reach the stratosphere because it is rained out during ascent (Singer, 1984). One preeminent atmospheric scientist even suggested that early studies into nuclear winter chose unrealistic assumptions (Emanuel, 1986) in order to advance nuclear disarmament at the height of the Cold War. Even with generous assumptions about soot reaching the stratosphere, higher resolution models now produce around 8K of cooling (Coupe et al., 2019).

The key epistemic issues around nuclear winter can be broken down into two social science and three physical science questions:

Social Science Q1: What distribution of nuclear weapon strikes is possible or plausible?

It is clear that the risk from an all-out nuclear war has declined since the end of the Cold War. Russia has gone from around 40,000 warheads in 1986 to around 5,000 warheads in 2015, and the USA has gone from around 32,000 in 1964 to around 5,000 in 2015. Increased precision has reduced the yield of each warhead, so since the 1980s perhaps we should expect a very rough 95% reduction in TNT equivalent detonated.

Nuclear strategy distinguishes between counter-force and counter-value strikes. A counter-value strategy involves targeting the most valuable areas of an enemy country such as major cities, to encourage surrender. A counter-force strategy involves narrowly targeting military capability. If both NATO and Russia began with a counter-force strategy, that could imply less widespread devastation. On the other hand, many military bases are located in combustible forest regions (for example Siberia or Canada), so large amounts of soot could still be produced.

In any case, we could imagine a war escalating steadily from counter-force to counter-value as key military installations in major cities are destroyed. To most efficiently achieve a devastating nuclear winter with current stockpiles, one could imagine targeting high-burning-potential sites, a kind of “counter-climate” strategy. But it is not clear that this would be in any agent’s interest, or a more potent deterrent than counter-force or counter-value.

Social Science Q2: What amount of societal damage would be caused by a certain degree of cooling?

This question may radically depend on your degree of belief in the strength and adaptability of remaining economic and social systems. Can democracies survive multiple halvings in national income? Interesting work in this area was conducted by Glomseth (2023), examining the effect of nuclear winter on trade and supply chains.

Physical Science Q1: How much soot can be produced by those nuclear strikes?

Estimating soot produced by burning cities and military installations is difficult. Of the burnable material in a city, only a few percent at most becomes soot. Toon et al. (2007) struggle to estimate this but give an upper bound of 150 Tg, which is then assumed in Coupe et al. (2019) to be directly injected into the stratosphere. Covey, Schneider, and Thompson (1984) assumed 200 Tg would be injected into the troposphere between 1 and 10 km.

Physical Science Q2: How much of this soot can reach the stratosphere?

The troposphere, which is wet and well mixed, extends up to about 15 km above the ground in the tropics and is separated from the dry, stratified stratosphere by the tropopause. Getting soot into the stratosphere often requires very intense pyrocumulonimbus clouds, associated with the largest forest fires (Toon et al., 2007).

Toon et al. (2007) suggested that the geometry of city fires (such as Hamburg) may be less suited to injecting soot into the stratosphere than forest fires. From my literature review so far, it seems there has been no work using extremely high resolution simulations (for example large eddy simulation) to explicitly simulate pyrocumulonimbus clouds from nuclear city fires.

Physical Science Q3: How long does soot stay there, and how much cooling does it cause?

Covey, Schneider, and Thompson (1984), which (as mentioned above) injected 200 Tg at 1-10 km altitude, used a very coarse model at 4.5 x 7.0 degree horizontal resolution with only 6 vertical levels. In this low resolution model, the cold period lasted 5 years and involved 15-20K cooling.

Coupe et al. (2019) used a 2 degree horizontal resolution model with 66 layers and produced around 8K cooling for four years. That corresponds to a much more aggressive soot assumption but a much lower sensitivity to impact. Better atmospheric models in the current generation of climate models (for example CESM2) operate at around 0.5 degree resolution, so current nuclear winter studies still have not been run with a genuine state-of-the-art model.

Self-evidently, current atmospheric models have not been calibrated by past nuclear winters, and we have reason to doubt their extrapolation skill. In climate modeling more broadly, even high-resolution models have meaningful biases in extreme convective events such as tropical cyclones, because small scale processes remain poorly resolved. Given the convexity of the damage function, this irreducible uncertainty should make us more worried than point estimates alone suggest (Stainforth, 2023).

Conclusion

It seems likely that earlier studies overestimated the physical climate effect of nuclear war, partly due to model resolution and partly because of assumptions made. However, no matter how high resolution a physical model is, it still cannot capture all relevant physical and social processes. We should still act as if effects could be more severe than currently calculated. In any case, nuclear war is so destructive that while nuclear winter may have been overhyped in some framings, this does not undermine its status as a pressing global issue.

References