Solar geoengineering is a controversial topic, but one that I think requires our close attention as the planet warms to unsafe levels.

Solar geoengineering refers to climate interventions which deliberately reflect solar radiation to counter the effects of human-caused climate change. The most commonly discussed and studied method is stratospheric aerosol injection, which involves adding aerosols in the upper atmosphere around 20 km high and reflect ~1% of the incoming sunlight. Unlike carbon dioxide which hangs around the atmosphere for a long time, these aerosols stay in the upper atmosphere for ~1 year, so they would need to be continuously injected to continue having an effect. And, if we stopped the injection, the warming from the increased greenhouse effect which they masked, would heat up the planet at an unprecedented rate.

Climate model studies show, however, that the impacts of warming on extreme precipitation events, droughts, heatwaves, sea-ice, sea level rise, hurricane strength and intensity could be reduced with a wise deployment strategy.

Hence stratospheric aerosol injection could potentially be used to keep the planet at a safe level of warming while we reduce our emissions and remove carbon dioxide from the atmosphere. It is certainly NOT an alternative to cutting emissions, only reaching net zero emissions will stabilize the climate for the long term.

Solar geoengineering has many risks and uncertainties: some physical risks remain even in an optimal deployment scenario. And some risks are due to a non-optimal deployment strategy. For example, a hemispherically asymmetric deployment of stratospheric aerosol injection will have important impacts on the tropical rainbelt. Moreover, there are many other non-physical risks related to how human societies will react : for example, will it discourage societies from reducing their emissions (which, again, we really need to do to stabilize the climate)?

Solar geoengineering would be a radical and risky endeavour, but it is only because of the devastating impacts of warming that we need to carefully study it. And, that is why I am glad to start working on this important topic.

I released a preprint with Geoff Vallis on understanding the seasonality of high latitude surface temperature, and thought I would share some additional motivation.

Past warm climates, such as the early Eocene, provide an out-of-sample test for our climate models. What is known as the ‘equable climate problem’ is the realisation that climate models struggle to reproduce the reduced equator-to-pole surface temperature gradient of past warm high-CO2 climates. Proxies indicate that global-mean surface temperature in the early Eocene was 26-32 deg C and the surface temperature gradient was 18-44% smaller than the preindustrial climate (Zhu et al. 2019). While there was large uncertainty about CO2 concentrations in the early Eocene, a recent paper gives a more constrained range: between 1170 ppmv to 2490 ppmv (95% CI), which is between 4 and 9 times preindustrial values (Anagnostou et al. 2020).

Before we had this more constrained CO2 concentration range, Huber and Caballero (2013) showed that a reasonable match between proxies and climate model was achieved by increasing CO2 to high enough levels (4480ppmv). However, those CO2 concentrations are higher than the more recent constraints mentioned above.

Zhu et al. (2019) show that two models manage to have a good temperature match with proxies given the more constrained CO2 concentrations (grey box below). The Kiehl and Shields (2013) model (grey ‘CCSM3_KS’ below) get there by claiming that the Eocene would have much fewer aerosols (warmer conditions make for less biological activity which produces natural aerosols) hence fewer cloud condensation nuclei. Hence they modify the cloud scheme to end up with less cloud, hence a weaker cloud radiative forcing, which gives the warming boost needed. It also leads to additional polar amplification as the increased atmospheric water vapor in the warmer atmosphere gives a greenhouse gas signature to the pattern of warming.

The Zhu et al. 2019 model (red above), on the other hand, is not modified and has a good proxy-model match anyway. The new CAM5 scheme gives a cloud distribution which is in better agreement with recent satellite observations. And, when CO2 concentrations are multiplied by 6, the model has the best fit in terms of CO2, global-mean temperature and gradient (see above, ‘Eocene 6x’) and gives much less cloud (cloud fraction shown below).

Both these models succeed in reproducing the temperature structure of the Eocene with reasonable levels of CO2 by having less cloud, hence a smaller cloud radiative forcing and more warming for a given level of CO2.

The seasonal range in surface temperature provides an additional constraint on the Eocene climate. It is well known that it was above-freezing year-round in the high-latitude continents. Moreover, Eberle et al. (2010) suggest that at several locations around 75 degrees North, the cold month mean temperature was 0-5 degrees C, and the warm month mean temperature was 20-25 degrees C.

I checked the data from the Zhu et al. 2019 paper (available here) and plotted the cold month mean temperature and seasonal range for the ‘Eocene 6x’ simulation below. The seasonal minimum goes below zero for much of the land poleward of 60 deg N and the seasonal range is 40-50 deg C (about 2x that suggested by Eberle et al. (2010)).

And, below are the high-latitude land/ocean surface temperatures for all simulations. The higher CO2 simulation (‘Eocene 9x’) does have above-freezing temperatures year-round, but the seasonal range is still high, and the annual-global mean temperature is outside the proxy range (first figure).

In Huber and Caballero (2013), despite having a high CO2 concentration, there is a good proxy-model match in cold month mean temperatures over land (see below). They argue (in paragraph 4, section 3.1.2) that warm month mean temperatures are much less well-constrained, except at high latitudes where temperatures were in the current climate envelope, although the polar night/day may add an additional layer of complexity.

In our preprint, we show that a high-latitude forcing will reduce the temperature seasonality over land due to its small heat capacity and the nonlinearity of the temperature dependence of surface longwave emission. It is then feasible that the reduction in clouds required to increase the global-mean surface temperature and gradient to match the proxies also increases the high-latitude land temperature seasonality to a value higher than that suggested by the proxies. Hence the reduction in cloud cover might not be the right mechanism to reconcile the Eocene temperature and CO2 concentration.

The explanation could also be simpler: the seasonal proxies are maybe not that well constrained (especially the warm month mean as discussed above), or there is an additional land feature such as different vegetation that would reduce the seasonality over land and not affect the global-mean surface temperature and gradient. Or there actually is an exotic mechanism not present in climate models that gives a simultaneous match in CO2 concentration, global-mean surface temperature, latitudinal gradient, and seasonal range in high-latitude land temperature.

The figure below shows Arctic sea-ice area (a) and Arctic surface temperature change (b) between the two periods limited by red lines in (a). In this high-emissions scenario, once the sea ice is gone, the ocean warms uniformly throughout the year, whereas the land warms more in winter and less in summer, hence reducing its seasonality. This is also qualitatively true in other comprehensive models.

This is also relevant to our understanding of equable climates such as the early Eocene, which had particularly warm Arctic winters. It was initially thought that climate models did not appropriately reproduce the pattern of temperature from these past warm climates: the models had either warm tropics or cold high latitudes. Hence various mechanisms were proposed to explain the gap between proxies and models. However, others argue that proxies allow for higher tropical temperatures than previously considered, which means that a high CO2 concentration alone may be enough to adequately model the Eocene climate. See Kiehl and Shields (2013) and Huber and Caballero (2013) for example.

Our question: is there a robust mechanism to explain why CO2 alone would be particularly effective in warming Arctic land winters?

As discussed here, the residual atmospheric forcing that results from the combination of the CO2 increase and shortwave reduction can lead to a precipitation reduction (blue below). One can tweak the spectral properties of the reflectant to offset the residual atmospheric forcing and cancel the reduction in precipitation (green below).

The stratospheric heating alone can also cause changes in precipitation. Simpson et al. (2019) analyze simulations where stratospheric aerosols are injected to maintain the surface temperature, latitudinal surface temperature gradient, and interhemispheric difference in an RCP8.5 (high emissions) scenario. They compare them with simulations where the stratospheric heating alone is added to the base climate. Though the changes in precipitation are lower, the pattern of change is strikingly similar to the stratospheric aerosol injection simulation (their figures 6 and 7), especially in the tropics.

To better understand how stratospheric heating alone can affect precipitation, they do a prescribed SST aquaplanet experiment with an idealized zonal asymmetry, and add the same prescribed stratospheric heating perturbation. The figure below shows the control precipitation on the left and the change in precipitation that results from stratospheric heating on the right. The result is noisy but there is a clear dry-get-wetter and wet-get-drier pattern. The authors indicate that more investigations in this idealized context are on the way to get a mechanistic understanding of how stratospheric heating affects precipitation.

Often, people model solar geoengineering by ‘turning down the sun’, but stratospheric heating, along with changes in cloud radiative forcing, and differences in the latitudinal pattern in radiative forcing are key differences between stratospheric aerosol injection and turning down the sun (Visioni et al. 2020). These other differences may also affect precipitation.

I think this is a case where idealized climate modelling can really help us to get a mechanistic understanding of how solar geoengineering (or, more specifically, stratospheric aerosol injection) affects precipitation. In idealized GCM world, we can separate:

• how different top-of-atmosphere forcings (that result from different injection strategies) affect atmospheric circulation and hence precipitation.
• how the spectral properties of the shortwave reduction affect precipitation, by manipulating the spectral bands within which shortwave radiation is reduced.
• how stratospheric heating and changes in cloud radiative forcing that result from the presence of stratospheric aerosols affect precipitation.

Ideally, this will lead us to understand how a given region’s precipitation will be affected by a given injection strategy, and better inform decisionmakers on tradeoffs.

Radiative antidote to CO2

In “Designing a radiative antidote to CO2”, the authors (Jake Seeley, Nick Lutsko, and David Keith) design an idealized reflectant that would perfectly offset the direct radiative effects of CO2. By choosing the spectral bands within which shortwave radiation is attenuated, surface temperature and precipitation are kept to control values after increasing CO2 and reflecting solar radiation in their climate model simulations.

Their main result

They run radiative-convective simulations with the cloud-resolving model DAM (Romps, 2018). Their figure 4 below shows the surface temperature and precipitation change in their three main experiments. In red (4xCO2), they just quadruple CO2 concentrations, the surface temperature increases by 6.89K and precipitation increases by 18.4%. In blue (4xCO2 + USRM), they quadruple CO2 and apply a spectrally uniform reduction in insolation (like in GeoMIP G1) to keep the surface temperature constant and end up with a 2.2% reduction in precipitation. In green (4xCO2 + SSRM), they quadruple CO2 and spectrally tune the reduction in insolation, such that the surface temperature and precipitation stay constant.

Tropospheric forcing leads to precipitation change

In an atmosphere in radiative-convective equilibrium, radiative cooling is balanced by the latent heat released in precipitating clouds. The change in precipitation ($\Delta P$) scales with the change in tropospheric radiative cooling ($\Delta Q$) by a factor $\alpha_P$ called the “hydrological sensitivity” parameter : $\Delta P = -\alpha_P \Delta Q$.

And, tropospheric radiative cooling can be decomposed into a part directly affected by external perturbation (change in CO2 or insolation, noted $F_a$) and a part that scales linearly with surface temperature change ($\Delta T_s$): $\Delta Q = F_a + \frac{\partial Q}{\partial T_s} \Delta T_s$.

Finally, the surface temperature change scales with the tropopause forcing ($F_t$): $\Delta T_s = \alpha_T F_t$.

In the 4xCO2 case, both the CO2 and surface temperature changes lead to an increase in tropospheric radiative cooling and hence an increase in precipitation. However, in both SRM experiments, the insolation is reduced such that the surface temperature change is zero, therefore only the direct effects of the CO2 and insolation changes affect tropospheric radiative cooling and precipitation.

Their figure 2 below shows the vertical profile of radiative forcing on the left and the decomposition into tropopause (TROP), atmospheric (ATM), and surface (SURF) forcings for each case. The tropopause forcing is the sum of the atmospheric and surface forcings, and the net forcing is the sum of the longwave and shortwave forcings. In the 4xCO2 case, the positive longwave forcing from increased CO2 at the tropopause is decomposed into the surface and atmosphere (approximately half each). The spectrally-uniform (USRM) reduction in insolation induces a negative atmospheric forcing that is proportional to the control shortwave atmospheric absorption, which is small. Hence the net atmospheric forcing (CO2 + insolation) is positive, which causes a reduction in precipitation (see equation above, $\Delta P = -\alpha_P \Delta Q$). In the spectrally tuned case (SSRM), the wavelengths are chosen such that the shortwave radiation is partially attenuated in the troposphere, which causes a negative forcing in the atmosphere, which balances the positive longwave forcing from CO2. Hence there is no change in precipitation.

Use this to constrain changes in precipitation in the G1 GeoMIP experiments?

According to this theory, if the control atmosphere is more transparent to shortwave radiation, then the G1 experiment (4xCO2 + spectrally uniform reduction in insolation) should lead to a larger reduction in precipitation. I used some CMIP6 data to see if there is a correlation between the reduction in precipitation between preindustrial and G1 simulations (y-axis) and the shortwave atmospheric absorption in the preindustrial simulations (x-axis). I actually use downwelling shortwave radiation at the surface in clear-sky conditions as the top-of-atmosphere downwelling shortwave radiation should be the same for all models.

However, as shown below, there is no clear correlation between these two variables. Maybe I need more models to have a wider range of precipitation changes, or there is an error in my data processing, or my reasoning is too simplistic.

Implications for real-world solar radiation management deployment

This theory is certainly relevant for the real-world because of the possibility of designing reflectants with the desired spectral attenuations. Moreover, the aerosols that are most commonly discussed do not lead to a spectrally uniform reduction in insolation, hence this theory helps us to understand how the deviation from spectral uniformity affects precipitation.

Previous work on understanding seasonality of Arctic warming

Panel (a) from the figure below from Screen et al. 2012 shows temperature trends over the 1979-2008 period averaged over the Arctic plotted over pressure levels (y-axis) and months (x-axis). While the mid- and upper-atmosphere warm throughout the year, the near-surface warms more in winter than summer.

Panel (b) shows trends from an atmospheric model where the sea surface temperatures (SSTs) and sea ice concentration (SIC) are prescribed to their observed evolution. Results are similar enough to observations. Panel (c) shows trends from an experiment where the SIC is prescribed to its observed evolution, the SSTs can respond in areas with sea ice change, and the SSTs are prescribed to the climatological annual cycle elsewhere. The seasonality in near-surface warming is reproduced in panel (c) where only the sea ice concentration (and SSTs where sea ice change occurs) are changed. This suggests that near-surface warming is correlated with the evolution of sea ice.

In this figure from Pithan and Mauritsen 2014, Arctic summer (x-axis) and winter (y-axis) surface temperature change are attributed to the various forcings and feedbacks from a top-of-atmosphere perspective. (See previous post on a follow-up to their work on the pattern of annual-mean warming.) The main driver of summer warming is the surface albedo feedback but it is counteracted by an increase in ocean heat uptake. The main drivers of winter warming are ocean heat release and a surface-enhanced lapse rate change.

The main mechanism seems to be that sea ice decline in summer leads to an increase in heat uptake by the ocean, which is then released in winter.

Eisenman simple sea ice model

I played with a simplification of a very simple sea ice model to better understand the thermodynamics of sea ice. This is taken from Eisenman’s sea ice work.

Eisenman’s model works by using the mixed layer enthalpy which evolves continuously from ocean to sea ice. The net surface flux $N$ $($in W/m$^2$, positive downwards$)$ dictates the change in mixed layer enthalpy $E$ $($in J/m$^2)$ : $\frac{\partial E}{\partial t} = N$.

Then, we can determine the surface temperature $T_S$ and the sea ice depth $h_{ice}$ depending on the sign of $E$:

• $E > 0$ : $T_S = \frac{E}{C_W} + 273$ and $h_{ice} = 0$

where $C_W$ is the mixed layer heat capacity. In this case, it is no different from a mixed layer of ocean water.

• $E < 0$ : $h_{ice} = -\frac{E}{L_f}$ and $T_S’$ such that $-\frac{k}{h_{ice}}(T_S’-273) = N$

where $T_S’$ is the dummy surface temperature, $L_f$ is the latent heat of fusion of sea ice, $-\frac{k}{h_{ice}}(T_S’-273)$ is the heat flux through the ice, and $k$ is ice thermal conductivity.

Then, if $T_S’<273K$, $T_S=T_S’$ and if $T_S’>273K$, $T_S=273K$ and sea ice melts with $\frac{\partial h_{ice}}{\partial t} = -N/L_f$.

My toy experiments

To simplify the model as much as possible and isolate the effects of the thermodynamics of sea ice, the latent and sensible heat fluxes are ignored, I assume a seasonally constant downwelling longwave radiation, and I assume that the longwave emission depends linearly on surface temperature:

$\frac{\partial E}{\partial t}(t) = SW_{down}(t) + LW_{down} - A - B T_S(t)$

where $SW_{down}(t)$ is prescribed from GCM values averaged poleward of 70 degrees North. And, $A + B T_S$ is a linearization of $\sigma T_S^4$.

Relaxing the assumptions described above does not meaningfully affect the results below.

However, this kind of surface energy budget integration only provides a limited insight, as $LW_{down}$ and the turbulent surface fluxes would change as a function of $T_S$ in a full GCM simulation.

The figure above shows values for the surface temperature, surface temperature change, mixed layer enthalpy, sea ice depth, net surface flux ($N$), and $SW_{down}(t)$ for different values of $LW_{down}$.

It seems like the surface temperature is particularly stable at 273K, as the extra energy goes into melting the sea ice (hence reducing $h_{ice}$) instead of increasing the temperature, as described in the algorithm above. The seasonal cycle of surface temperature change is then similar to the observations. Obviously, there is a lot more going on in the real world, but I wonder what the contribution of this effect is to the seasonal cycle of Arctic surface temperature change.

In a world with no sea ice

I ran some idealized GCM simulations with no sea ice or clouds using Isca. Continents differ from ocean only by their mixed layer depth, which is 10 times smaller than that of the ocean. I explore the pattern and seasonality of surface temperature change by running three simulations: 300ppm, 1200ppm, and 4800ppm.

Panel (d) below shows the surface temperature change averaged poleward of 70 degrees North for land (solid lines) and ocean (dashed lines). While the ocean warms by the same amount year-round, land warms more when it’s the colder (see panel (c)). (The months should be shifted.)

There are more details to come in a future pre-print. But, in a nutshell, this difference in seasonality of high latitude warming is explained by the smaller surface heat capacity of land (which leads to a larger seasonal cycle in temperature) and the nonlinearity of the temperature dependence of surface longwave emission, $\sigma T_S^4$ (which forces cold temperatures to warm more to reach the same increase in emission). We also explore the seasonality of high latitude lapse rate changes in these simulations and the role of advection in smoothing out expected differences in near-surface atmospheric temperatures over land and ocean.

Conclusion

Sea ice undoubtedly plays a dominant role in the seasonality of Arctic surface temperature change. In simulations with no sea ice, the ocean warms by the same amount year-round. The main mechanism in comprehensive climate models seems to be an increase in ocean heat uptake in summer as the sea ice melts, which leads to increased ocean heat release in winter.

I played with a toy sea ice model, which suggests that the sea ice surface temperature is quite stable at the melting temperature as the extra energy received goes into melting the sea ice instead of increasing the sea ice surface temperature.

And, in a world without sea ice, the high latitude continents warm more in winter than summer due to their small surface heat capacity and the nonlinearity of the temperature dependence of surface longwave emission.

I would like to add a simple sea ice module to Isca to better understand how it interacts with the other components of the climate system.

Comprehensive climate model studies

Side effects of solar geoengineering

Stratospheric chemistry

The work is focused on understanding the causes of polar amplification: the warming from increasing greenhouse gases is amplified at high latitudes both in models and observations.

Top-of-atmosphere budget analysis

Various mechanisms are thought to contribute to polar amplification: sea-ice albedo feedback, lapse rate and Planck feedback, atmospheric and oceanic heat transport changes. There generally has been two approaches to investigate the contributions of each of them. The first is to turn a certain mechanism off and see how that affects the pattern of warming $($mechanism denial experiment$)$. The second method is to diagnose how each mechanism contributes to the pattern of warming through some kind of budget analysis.

Pithan and Mauritsen (2014) quantify the contributions of various forcings and feedbacks to polar amplification in CMIP5 models using a top-of-atmosphere energy budget based method described below. In the figure below, the distance from the one-to-one line determines how much this forcing or feedback contributes to Arctic / tropical amplification.

They use radiative kernels to figure out how each change $($increase in water vapour, decrease in Arctic sea ice, increase in equator-to-pole atmospheric energy transport, etc$)$ affects the top-of-atmosphere energy budget. The amount of surface temperature change is then derived by assuming that a vertically uniform surface and atmospheric temperature change balances the given top-of-atmosphere energy imbalance. For example, we know that the CO2 forcing and water vapor feedback are tropically amplified, hence they contribute to tropical amplification in the figure above. Any deviation from vertically uniform warming is then accounted for in the lapse rate feedback term, which becomes a dominant cause of polar amplification.

Forcing dependence of high latitude lapse rate feedback

In a simple analytical model of the Arctic atmosphere, Cronin and Jansen (2016) found that the high latitude lapse rate change depends on the forcing. The figure below shows how the high latitude temperature varies when the longwave optical depth $($a$)$, the surface forcing $($b$)$, and the atmospheric forcing $($c$)$ are varied. This breaks the linearity between surface temperature change and top-of-atmosphere radiation, i.e. when forced with a change in longwave optical depth $($CO2 or water vapor$)$, the surface temperature changes more for a given top-of-atmosphere radiation change than if forced with a change in atmospheric energy transport.

By assuming that top-of-atmosphere radiation imbalances are balanced by vertically uniform surface and atmosphere temperature changes, the top-of-atmosphere budget analysis does not account for how various forcings and feedbacks affect the high latitude lapse rate differently.

Single column model attribution method

The objective of this paper is to attribute the pattern of surface temperature change to the various forcings and feedbacks using a single column model in order to account for the forcing dependence of lapse rate change.

I first run a warming experiment in an idealized GCM $($aquaplanet, no clouds, no sea ice, comprehensive radiation, similar setup to MiMA$)$. The figure above shows the tropical $($a$)$ and polar temperature change $($b$)$ of the idealized GCM in grey. It is then decomposed using the single column model into the effects of just CO2 $($red$)$, water vapor $($blue$)$, and energy transport $($green for tropics and separated into dry and moist for polar region$)$. The single column model is used to best emulate the idealized GCM. In black, the temperature change when all the parameters are changed matches the idealized GCM temperature change, in grey, well enough.

In the tropics, the atmosphere is in radiative-convective equilibrium, hence the temperature structure of the atmosphere is close to moist adiabatic. Any perturbation leads to roughly the same vertical temperature change structure.

At high latitudes, the vertical structure of temperature change depends on the forcing. As predicted by the simple Cronin and Jansen model, the increase in longwave optical depth $($CO2 and water vapor$)$ leads to surface-enhanced warming. The increase in energy transport however has a more complex vertical structure. This is expected as the parametrization of energy transport in the Cronin and Jansen model was simplistic.

In the figure above, we compare the attribution based on the single column model $($dots$)$ and an attribution based on top-of-atmosphere energy imbalance, like in Pithan and Mauritsen $($crosses$)$. A surface heat source is applied at high latitudes $($12 W/m2 $($b$)$ and 24 W/m2 $($c$)$$)$ to simulate the effect of local feedbacks $($such as the sea-ice albedo feedback$)$. As in Pithan and Mauritsen, the lapse rate, Planck, and sea-ice albedo / local surface heat source are the main contributors to polar amplification.

In the single column model attribution $($dots$)$ however, CO2 and water vapor contribute to polar amplification due to their surface-enhanced warming structures at high latitudes and the presence of convection in low latitudes. The polar warming structure from atmospheric energy transport convergence is reduced as it preferentially warms the mid and upper troposphere.

Conclusion

In conclusion, this work comes as a follow-up to Pithan and Mauritsen’s influential paper on polar amplification. The explanation that the lapse rate feedback causes polar amplification was unsatisfactory as the lapse rate feedback, at high latitudes, is not a single process but a result of how the various forcings and feedbacks affect the vertical structure of temperature change. This single column model enables us to separate the drivers of vertical temperature change at high latitudes and get a clearer picture of the drivers of polar amplification.

1/ Everybody agrees that the highest priority is to reduce carbon emissions. As shown during the pandemic, even extreme reductions in personal consumption do not lead to sufficient emission cuts. Achieving net zero emissions will require a complete transformation of our means of production, which takes time. Solar radiation management can be considered as a way to give us more time to complete this transformation without having to also deal with a rapidly changing climate.

2/ Climate model simulations where all greenhouse gas warming is offset by solar radiation management have a reduction in precipitation, with some regions being more affected than others. $($It also leads to residual warming in the Arctic. Why oh why?, you may wonder. Answers here: paper and talk.$)$

However, if only half the warming from greenhouse gas increase is offset, the impacts of warming on temperature, temperature extremes, water availability $($Precipitation-Evaporation$)$, and precipitation extremes are reduced almost everywhere. In climate models, halving the warming using stratospheric aerosol injection makes the climate more like the preindustrial climate for most regions of the world, when compared to scenarios where no warming is offset.

Check out Peter Irvine’s work and this Carbon Brief article, it is worth pointing out that the most recent work confirms previous results by modelling solar radiation management using aerosols and not just by “turning the sun down”.

3/ More research is being done on the potential negative side effects of solar radiation management, for example on ozone loss. Most of our confidence that solar radiation management could moderate climate impacts relies on climate simulations and observations of volcanic eruptions. A serious research effort is needed to work out all the ways it could potentially fail, if we were to actually consider it as a viable option. Obviously, as a climate researcher interested in doing that research, this is a motivated statement.

4/ Solar radiation management $($or solar geoengineering$)$ is often thought about as being used to “shave the peak” of climate impacts while we remove CO2 from the atmosphere, as shown in the figure below.

This figure $($source$)$ plots climate impacts as a function of time. Climate impacts $($such as sea level rise, heat waves, changes in precipitation…$)$ are roughly proportional to the global mean surface temperature change. If we do not manage to reach net zero emissions $($”business as usual” in red$)$, these climate impacts will rise continuously. If we cut emissions aggressively, the climate impacts will only stop growing when we reach net zero emissions, and they will stay high for as long as CO2 CONCENTRATIONS $($not emissions$)$ are high. Hence, we need CO2 removal $($green$)$ to bring climate impacts back down. However, those are slow and expensive for now. Solar radiation management, is then considered as a way to moderate the impacts of warming while we remove excess CO2 $($blue$)$.

5/ It is worth pointing out that the emission of aerosols in the troposphere $($lowest atmospheric layer where we live$)$ is already cooling the climate, though there is uncertainty as to the magnitude of this cooling. We should reduce these emissions as they have adverse impacts on human health. Hence, on top of our greenhouse gas emissions, we are already potentially cooling the climate with aerosols, though not intentionally.

6/ The default position in public discourse about how to deal with climate change is cutting emissions aggressively and dealing with the climate impacts caused by our cumulative emissions. This will likely lead to parts of the world to be uninhabitable by humans. The impact this has on people depends on our political response, and it may certainly be possible to deal with it in a humane manner if we are politically organized. However, if there is an opportunity to moderate these impacts while we pull carbon dioxide out of the atmosphere, wouldn’t it be very irresponsible to not carefully research it?

The total temperature change from increased CO2 is a function of cumulative emissions so it is no surprise that a temporary reduction in emissions does not change the temperature at 2100 that much. But just how long does it take to make a dent?

I made a few simple assumptions to get an upper bound on temperature change: we follow RCP4.5 and years of “reduced” emissions actually have zero$($!$)$ emissions.

For 2 years of zero emissions, we barely get any change in temperature at 2100.

For 10 years of zero emissions, we get order 0.1 degrees less warming by 2100.

Finally, for 30 years of zero emissions, we get order 0.5 degrees less warming by 2100.

Check out the $($barely modified tutorial$)$ code here.