Revisiting the Geoengineering Question
We probably shouldn't do it, but if we do it we need narrow bounds
I have a new guest essay in the New York Times with David Keith that builds off my earlier Climate Brink post The Geoengineering Question. Below I’ve included some more detailed thoughts that couldn’t make it into the published piece given the word limit constraints.
We are already geoengineering the planet today, but badly. Humans are cooling the climate today by emitting 75 million metric tons of sulfur dioxide into the lower atmosphere, almost entirely as a byproduct of burning fossil fuels. This cooling offsets about 0.5C of warming that would have otherwise occurred from CO2 and other greenhouse gases, but it comes at the cost of millions of premature deaths per year caused by the sulfate aerosols.
Sulfur emissions are declining sharply as countries started cleaning up air pollution. Global SO2 emissions today are 48% lower than they were in 1979 and 40% lower than in 2006 (China saw a massive 70% decline since 2008!). This is the primary contributor to the acceleration in global warming in recent years. Cutting air pollution saves lives and is unequivocally worth doing, but it is also driving about a quarter (~0.14C) of the approximately 0.5C warming the world has experienced between 2007 and 2024.
This raises a question worth debating: would it be worth considering putting a tiny portion of the sulfur we currently emit into the upper atmosphere to compensate for additional warming we will experience as we cut air pollution?
The limits of geoengineering
There are a number of approaches proposed to geoengineer the planet, but nearly all of them have one thing in common: they work by making the earth more reflective, reducing the amount of sunlight reaching the surface. This is the same mechanism (more or less) through which sulfur emitted as air pollution currently cools the planet.
The most plausible pathway is to put sulfur in the upper atmosphere (the stratosphere), mimicking the effect of large volcanic eruptions. This is reasonable well studied and modeled. Unlike in the lower atmosphere, sulfate aerosols have a relatively long residence time in the stratosphere; they cool the planet for years compared to the days that they last when emitted as air pollution. This means you’d need a lot less – around 50x less – sulfur to have the same cooling effect in the stratosphere vs the lower atmosphere (the troposphere).
Aerosols are also reasonably well distributed in the stratosphere rather than localized as in the troposphere, resulting in much less human exposure to air pollution; there is more than 100x less health impact per unit of cooling.
But there are some pretty significant downsides. Stratospheric aerosol injection (SAI) would disrupt precipitation patterns and potentially damage the ozone layer in large deployments. It does not stop other impacts of atmospheric CO2 accumulation like ocean acidification. The geopolitics are also messy; once someone starts doing it they risk “owning the weather”, with any event (say, a late monsoon in India) being hard to prove unrelated to the intervention.1
But the biggest downside of geoengineering is that it does not actually solve the underlying problem. As we note in the NYT piece:
There is also a more fundamental limitation to sunlight reflection. It is effectively a Band-Aid that treats the symptoms of climate change but not the underlying disease of greenhouse gases. And unlike a skinned knee, the Earth does not heal from climate change on any time scale that matters for human societies. Warming from carbon dioxide is astonishingly persistent; much of what we emit today will warm the planet for many thousands of years to come.
Even if the world drives emissions down to zero, the planet wouldn’t cool back down for millennia. The only durable way to return to cooler temperatures is to remove the excess carbon we have already added, and removing enough to reverse even 0.1 degree Celsius of warming would cost tens of trillions of dollars.
This means that the real risk of geoengineering is not some Hollywood-style catastrophe, but complacency. A cheap way to delay the effects of warming risks undermining the need to rapidly reduce emissions, and going down that path would risk locking our children into a dependency where even stopping the process becomes too expensive to contemplate.
The potential for mitigation deterrence is the biggest risk of this technology, and once we start it will be easy to fall into a trap where it would be too expensive (through carbon removal) or too damaging (through termination shock) to ever stop.
A modest limit
We hope we won’t ever need geoengineering. The world can still come together and rapidly reduce emissions to limit warming to well-below 2C this century. But a lot of things are moving in the wrong direction today. There are a lot of worrying unknowns in the climate system from the magnitude of climate sensitivity and carbon cycle feedbacks to the risk of unforeseen tipping points.
It may be that we reach a point that the impacts are so bad that the push for some short-term respite becomes impossible for policymakers to ignore. If we go down that path, we need to be absolutely sure that any climate intervention is narrowly focused with a clear off-ramp. To that end, we propose that replacing the cooling from current air pollution be used as a clear, enforceable upper limit. And it should come with a clear off-ramp: as the world reaches net-zero emissions and scales up carbon removal technologies later in the century, the program ratchets down to zero. Set the rules so the only direction allowed over time is toward retirement. As we say in the piece:
Given all this, we are not advocating deploying geoengineering today. But if policymakers decide that it is needed, a more modest approach would be to run a small, carefully scaled program that slightly increases the upper atmosphere’s reflectivity to compensate for the loss of cooling as
sulfur pollution is eliminated.The goal would not be to dial the Earth to some preferred temperature, nor to offset all greenhouse warming. It would be to keep the total cooling from sulfur roughly constant for a period of time, reducing near-term climate risk while decarbonization efforts continue.
As world leaders gather in New York for the U.N. General Assembly and Climate Week, any discussions of sunlight reflection should have a clear, enforceable commitment to never cool the Earth more than today’s current sulfur emissions do. And it should come with a clear off-ramp: As the world reaches net-zero emissions and scales up carbon removal technologies later this century, the program should end.
Pacing matters as much as limits. If society ever chooses to test this approach, it should start small and move slowly. Tying it to reductions in air pollution allows a slow ramp-up, resulting in increments that are imperceptible to most of us but visible to satellites and sensors. This should be coupled with regular checkpoints to assess side effects on regional rainfall, the atmosphere and ozone. The intent is to buy a modest, temporary buffer, not to start a new arm of climate control.
There is a real risk of a slippery slope here; if we start the pressure to kick the proverbial can of expensive mitigation down the road could prove intense. And a world where the symptoms of climate change are masked could reduce the impetus to pursue a cure. But the argument that we need to accept potentially vast human suffering to force action is equally morally problematic as climate change becomes significantly more severe in coming decades.
Geoengineering should not be a technocratic shortcut. If we ever need to use it, it should be as a narrow, temporary option to limit near-term warming while we do the hard work of getting to net-zero emissions and, ultimately, drawing down the carbon we have already emitted. It is emphatically not a substitute for decarbonization. If it ever happens, it should happen within strict bounds that make complacency harder, not easier.
We should debate those bounds now, before heat waves, crop failures, or tipping-point fears goad governments into ill-considered action. The world stumbled into today’s accidental sulfur cooling without intention or oversight. We should not stumble into a replacement. If society chooses to consider deliberate sunlight reflection at all, it should be with clear limits, slow steps, relentless measurement, and a plan to shut it down.
This is portrayed quite dramatically in Neal Stephenson’s book Termination Shock, where the failure of the monsoon leads Indian commandoes to raid Texas to shut down a rogue billionaire’s sulfur gun.
Unfortunately this article reminds me of an Arabian fable: the geoengineering camel wants to get its nose into the climate mitigation sheikh's tent!
I think a pilot geoengineering project is a bad idea. However, I do support research on geoengineering, including small-scale field testing, but well short of any deployment.
The problem is that a lot of geoengineering discussion, including this article, has a simplistic global-average temperature framing. The impression is that there is a single knob, global-average temperature, that can simply be "dialed back" to mitigate the impacts of global warming. With the exception of perhaps steric sea level rise, the harm due to GHG-driven warming is local (heatwaves, intense rainfall, ...). To the extent that tropospheric sulfate aerosols compensate this harm, it happens locally. Stratospheric sulfate aerosols will have a different spatial cooling pattern. It is therefore simplistic to suggest that the global harm due to GHG-driven warming is being compensated by global benefits of tropospheric aerosol-driven cooling or that sulfates in the stratosphere can provide the same cooling benefits as sulfates in the troposphere.
The crucial fact is that carbon dioxide is well-mixed and long-lived. Sulfur dioxide is not well-mixed and short-lived in the troposphere. Their opposing radiative impacts cannot be captured by their contribution to a single global-average number. To truly negate the climate impacts of carbon dioxide increases at the source, we need carbon dioxide removal (CDR). However, CDR is an extremely unproven approach, and possibly very expensive. In contrast, the radiative impacts of the relatively inexpensive SAI are generally known, thanks to the analog of natural volcanic eruptions, but to compute its climate impacts we need to rely on imperfect climate models.
The suggestion that we can start a modest geoengineering deployment and then we can measure its efficacy is misleading. We may be able to measure the radiative impacts of modest SAI over a decade to so, but precisely attributing its local dynamical climate impact will take several decades (beyond the putative net-zero horizon) and better climate models (with smaller errors) than we currently have, because of irreducible internal variability and model inadequacies. The slow ramp up of SAI means a very low signal-to-noise ratio, which won't help in detection and attribution of its impacts.
Another argument frequently trotted out by geoengineers is that we need geoengineering to avoid imminent tipping points. (Interestingly, most proponents of tipping point-related urgency are not supporters of geoengineering; they want radical emission reduction!) Tipping points, to the extent that they exist close to the present climate and their imminence can be predicted with any precision (both of which are iffy suppositions based on imperfect, often simplified, models), are not triggered by global-average phenomena. They are triggered by local processes. There is no precise global-average temperature threshold to trigger them because global climate models differ in their absolute global average temperature estimates by several degrees, even if they are in closer agreement on the temperature change (delta) associated with GHG warming. Being nonlinear phenomena, tipping points thresholds will depend on absolute temperatures, not deltas.
To reiterate, the responsible thing to do is to stick to modestly funded research in geoengineering (with some field tests), like we do for fundamental science, and not let faux urgency scare us towards deployment.
First, in the 6th paragraph you say troposphere instead of stratosphere.
More seriously, the complacency argument against intervention is becoming weaker and weaker. The cost of solar and wind is becoming the least expensive source of electricity and becoming the leading investment. The idea that the world will go back to coal seems illogical as it, in the US costs about three times as much as wind and solar and that factor is growing as the cost of solar and wind drop below eve natural gas. What we need to do is to help encourage this transition. California is doing it, for example, by building transmission lines from locations where sunlight harvesting with solar can be very cost effective to the main transmission lines and this is attracting all sorts of private sector investment in solar (in Google Earth, check just to the west of Antelope Valley, California to see where a few billion for a transmission line has led to of order 20 times as much investment in building solar). As Sandy MacDonald made clear in his oped in The Hill (see https://thehill.com/opinion/energy-environment/5114155-us-energy-revolution-supergrid/), what would really help our energy conversion is a high-voltage direct current national network (and a book he is writing urges that for every continent). The profit motive can convincingly overcome regression to fossil fuels if we help it in the right way, and if this is done, then the real moral hazard of geoengineering is not doing it to save all the lives and damage from extreme events (drenching rains and intolerable heat waves) and the commitments to a much higher sea level and thawing of permafrost and all the carbon dioxide and or methane that will be reduced. I would really urge your reconsideration of climate intervention--sure it is not a solution in itself and I am all for more aggressive mitigation and helping that along, but I think the metaphor of a tourniquet is better than a bandaid (even a slightly flawed tourniquet would be better than bleeding to death).