As GHG emissions continue to rise, the idea of deliberately altering the climate through large scale technological measures (other than emission reductions) has gained in-creasing interest. The proposed methods for this so-called climate engineering (CE) or geoengineering can be divided into two groups: those that tackle the root of climate change, namely the excess atmospheric concentrations of GHGs, by removing them from the atmosphere are commonly summarized under the term carbon dioxide removal (CDR); those that seek to cool the Earth by reducing the amount of incoming radiation are called solar radiation management (SRM).
Besides studies in the field of natural science and engineering investigating the feasibil-ity, effectiveness, and side effects of such techniques, there has also been considerable research on political, legal, economical, and ethical questions related to possible de-ployment of CE [e.g. Schäfer et al., 2015].
One might ask, why research resources should be spent on such a science-fiction idea, which most people tend to see rather critically [Braun et al., 2017]. And there exist serious arguments against research on climate engineering. For example it is argued, that once research is started it may lead to further and further progress, development, and ultimately to an almost inevitable urge to deploy SRM (known as the slippery slope argument). Or the existence of SRM research programmes might cause people to reduce their efforts at reducing emissions (the moral hazard argument) [e.g. Preston, 2013]. But we have to admit that there is considerable uncertainty in the expectable impacts of climate change. This is because on the one hand the level of future greenhouse gas emissions is uncertain, and on the other hand the exact response of the climate system to these emissions is uncertain as well. Therefore, we have to face the reality that there is a certain probability of a global warming well above 2 C with possibly dramatic impacts. Or it might be much more difficult to adapt to climate change than currently expected. In such a situation, future generations might feel themselves forced to take short-term measures like CE. For providing them with a profound basis of knowledge allowing an informed decision on whether and how to deploy CE techniques, research has to be conducted now.
Overview of proposed climate engineering techniques
Figure 1.1 illustrates a number of proposed SRM and CDR techniques. Numerous ways of removing GHGs from the atmosphere have been suggested. Some of them aim at converting strong short-lived GHGs into less problematic compounds (e.g. CH4 into CO2), but most CDR techniques seek to capture and store CO2. The capturing could be done by technical devices directly at the source points or from ambient air, or through enhanced geochemical sinks (in the case of ocean alkalinity enhancement) or biological sinks of CO2. Such biological sinks could operate through marine organisms, e.g. fertilised through addition of iron or artificial upwelling of nutrient-rich deep water, or terrestrial plants. The captured carbon could be stored on land (by afforestation), under ground (as charcoal or directly as CO2 sequestered in geological storages), or in the deep ocean. All the proposed techniques have their risks and limitations (e.g. land use competition and conflicts, impact on ecosystems, limited storage capacities, residence time in the reservoir, etc.). CDR shall not be considered further in this thesis but it needs some attention as well.
This work focuses instead on SRM and especially on the technique that is studied and discussed most so far, because it has a strong natural analogue and could possibly pro-vide a fast cooling of the Earth with relatively limited effort (i.e. with a high leverage): the injection of aerosols or aerosol precursors into the stratosphere, commonly called stratospheric aerosol injection (SAI). Before we present this technique in more detail in Section 1.2.2, we will have a short overview of further methods of SRM that have been proposed to increase the Earth’s albedo.
The concept of marine cloud brightening (MCB) [Latham, 1990] is based on the fact that, for a given amount of water, clouds with smaller droplets reflect more sunlight. The size of cloud droplets could be reduced by injecting additional cloud condensation nuclei (CCN), e.g. in the form of sea salt particles, resulting in a higher cloud droplet number concentration (CDNC), i.e. in a higher cloud albedo.
Other proposed methods seek to increase the albedo either of the ocean surface, by creating small air bubbles in the surface layer [Seitz, 2011] or by covering the water with foam, or of the land surface by planting brighter plants or by painting buildings and roads. Even mirrors in space reflecting or diffracting a fraction of the sunlight away from the Earth have been suggested. Although this may be one of the technologically and economically most challenging methods, its possible impact has been studied widely with climate models, because it can be easily implemented by reducing the solar constant.
The last proposal that shall be mentioned here is the concept of cirrus cloud thinning (CCT) [Mitchell and Finnegan, 2009]. Strictly speaking, it is not an SRM method because it seeks to cool the Earth by increasing the amount of outgoing terrestrial infrared (LW) radiation. Cirrus clouds, which consist of ice particles and are located in the upper troposphere, can have a net heating effect if their absorption of LW radiation (greenhouse effect) dominates over the reflection of SW (solar) radiation. This is especially the case at high latitudes during winter. Therefore, reducing the cirrus cloud cover can have a cooling effect, which in principle could better compensate the warming effect of greenhouse gases because it also acts on the LW spectral range, in contrast to SRM techniques. The reduction of cirrus cloud cover could be achieved by increasing the size of the ice crystals and thereby increasing their fall speed. In regions where cirrus formation through homogeneous freezing (producing many small ice particles) dominates, the injection of ice nuclei may result in fewer, larger ice crystals with a shorter lifetime. But it is unclear in which fraction of the upper troposphere homogeneous freezing dominates and hence what the cooling potential of CCT would be. In addition, the concentration of ice nuclei would have to be chosen carefully, because they can also cause the formation of (potentially warming) additional cirrus clouds which would not have formed under unperturbed conditions [Lohmann and Gasparini, 2017].
It is important to note that all radiation management techniques can only (at least partly) compensate the warming effect of greenhouse gases. But they cannot solve other problems from high CO2 concentrations like ocean acidification.
Stratospheric aerosol injection
This thesis focusses on climate engineering through stratospheric aerosol injection, which we will briefly present in this section. The potential of the stratospheric aerosol layer to reduce the Earth’s surface temperature has been known for a long time. A cooling was observed after major volcanic eruptions like those of Mount Tambora (in 1815), Krakatau (in 1883), and Mount Pinatubo (in 1991), which injected large amounts of SO2 into the stratosphere [e.g. Rampino and Self, 1982, Dutton and Christy, 1992]. The idea of stratospheric aerosol injection (SAI) is to achieve an artificial enhancement of the stratospheric aerosol layer. The deployment of SAI as a measure against global warming was first suggested by Budyko , but a broader scientific debate on the subject was initiated several decades later by Crutzen . A deliberate injection of SO2 forming sulphate aerosol would constitute a close analogue to volcanoes, but injection of other particle types is also proposed (see Chapter 2). Possible ways of delivery would be e.g. through high-altitude aircraft, tethered balloons or rockets [McClellan et al., 2012]. Although it requires more effort to bring material to the stratosphere than to emit it in the troposphere, it would be more efficient because aerosols have a longer residence time in the stratosphere. But this also means that if one wanted to stop the deployment of SAI (e.g. because of unforeseen problems), its effects would still persist on a timescale of months to years. Suddenly stopping SAI could also be problematic because the climate system would rapidly catch up on the warming compensated by SAI if GHG levels continued to rise. This is known as the termination effect and has been studied e.g. by Jones et al.  who found a rapid temperature and precipitation increase and sea ice loss after a simulated sudden stop of SAI. Another important risk is that the additional aerosol in the stratosphere could damage the ozone layer that protects life on Earth from dangerous UV radiation [Tilmes et al., 2008]. Further aspects of SAI to consider are discussed in Chapter 2.
Since the effect of SAI depends on the behaviour and effects of aerosols, we will briefly summarise the relevant aspects of atmospheric aerosols in this section, which is partly based on the textbook by Boucher .
The Earth’s atmosphere mainly consists of gaseous compounds, but it also contains smaller amounts of liquids and solids, which affect the Earth system considerably. In principle, a mixture of solids or liquids in the gas phase is called an aerosol, but it is common to use the term for the liquid or solid compound only and we will stick to this convention in the text of this thesis. Since water is an important constituent of the atmosphere, many particles consist of liquid or frozen water, e.g. rain, cloud droplets, fog, snow, or hail. They are summarised under the term hydrometeors and shall not be further considered here when we write about aerosols.
The properties of aerosols are mainly determined by composition and particle size. The size can span many orders of magnitude, from a single nanometre to hundreds of micrometres. Depending on the atmospheric conditions, one can usually observe several relative maxima in the particle size distribution, the so-called size modes. They are not sharply separated from each other and their width and mean size can vary, but they are a useful classification. The nucleation mode ( 1–10 nm radius) mostly contains particles that have formed from gaseous precursors. The Aitken mode ( 10–50 nm radius) is made up from nucleation mode particles that have grown through conden-sation. In the accumulation mode ( 50–500 nm radius) the aerosol mass accumulates through coagulation and further condensation. Most particles that are released at the Earth’s surface belong to the coarse mode (>0.5 m radius) and super-coarse mode.
Aerosol particles can consist of a variety of different substances. The most important constituents (besides water) are crustal minerals, sea salt, sulphate, nitrate, black carbon (also known as soot) and organic compounds. Freshly formed aerosol mostly consists of only one of these components, but during their atmospheric lifetime particles are commonly mixed externally and internally. The aerosol composition determines its hygroscopicity, i.e. how easily water can condense on the particles. Depending on the ambient relative humidity, condensational particle growth can be an important process that also affects the aerosol’s physical properties.
Aerosol particles interact with electromagnetic radiation and can thereby affect the Earth’s radiative budget considerably. On the one hand aerosols absorb radiation and re-emit it according to their temperature. On the other hand they scatter radiation. How exactly this happens, strongly depends on the size parameter x, which is the ratio between the particle circumference 2pr (with the particle radius r) and the wavelength l:
The degree of interaction is measured by the absorption and the scattering cross section of a particle. The sum of both is called the extinction cross section. For particles much smaller than the wavelength (i.e. x 1), scattering can be described by the Rayleigh theory, where the scattering cross section is proportional to x4. For objects much larger than the wavelength, geometrical optics can be applied and the extinction cross section converges towards twice the geometric particle cross section sg = pr2. The intermediate regime is described by the Mie theory in the case of spherical particles. Extinction usually reaches its maximum in this intermediate range of the size parameter, which in case of visible light corresponds to accumulation mode particle sizes. Mie theory allows to compute aerosol optical properties based on the particle size and the complex refractive index, whose imaginary part characterises the absorption within the medium. Its numerical implementation is included in many numerical aerosol models. The assumption of spherical particles is valid for liquid atmospheric aerosols, but not for all solid aerosols (such as dust particles) and ice particles.
The troposphere, i.e. the lowest layer of the atmosphere (reaching altitudes of 8-15 km), contains the largest part of the atmospheric aerosol burden and a large variety of aerosol species (already listed above). Some of them are already emitted as particles (primary aerosols), others form through condensation of gaseous precursors (secondary aerosols). There are numerous natural and anthropogenic sources of tropospheric aerosols and aerosol precursors: wind friction on the ocean surface (sea spray) or on the land surface (mineral dust), combustion of biomass or fossil fuels, emissions from plants and other living organisms, volcanic eruptions (ash and sulphate), as well as agricultural and industrial processes. Aerosols can have many different impacts in the troposphere: they absorb and scatter radiation, thereby changing temperature and the ratio between direct and diffuse light, affect cloud formation and properties by acting as cloud condensation or ice nuclei, play a role in atmospheric chemistry, serve as nutrients, but also act as air pollutants with adverse effects on human health. Due to efficient removal mechanisms (mostly wet deposition and boundary layer processes) tropospheric aerosols have a lifetime in the order of days or weeks. Since there is only little exchange between tropospheric and stratospheric aerosols, and the aerosol physics at play are considerably different between troposphere and stratosphere (see following section), tropospheric and stratospheric aerosols are usually considered separately. Therefore, we will now abandon tropospheric aerosols and focus on aerosols in the stratosphere.
Aerosols are also present in the layer above the troposphere, namely the stratosphere. The stratospheric aerosol layer was discovered and first described by Junge et al.  and is therefore also called the Junge layer. Only a minor fraction of the stratospheric aerosol is of tropospheric origin, because tropospheric air entering the stratosphere mainly in the tropics has to pass the very cold tropical tropopause, a pathway where the condensing humidity scavenges most of the aerosols. The small fraction that survives this transport includes organic aerosols and black carbon [Murphy et al., 2007]. Hence, stratospheric aerosols are mostly formed locally from gaseous compounds. There is also a source of meteoritic dust particles from above [Cziczo et al., 2001], but it is likely to be very small. Since there is no wet deposition in the extremely dry stratosphere and stratosphere-troposphere air exchange is very limited, stratospheric aerosols have a much longer residence time in the order of months or years. Therefore, slower physical processes like particle growth and size-dependent sedimentation are more relevant than in the troposphere. Figure 1.2 illustrates the current state of knowledge about relevant processes, sources, and sinks of stratospheric aerosol, as described in the comprehensive review article by Kremser et al. .
Sulphate particles are the main constituent of the stratospheric aerosol layer, but their concentration can vary strongly in time due to the sporadic nature of their main source, major volcanic eruptions. In the absence of such eruptions, which can inject large amounts of sulphur dioxide (SO2) directly into the stratosphere, sulphur species from other tropospheric sources can enter the stratosphere, e.g. through deep convection in the tropical troposphere and subsequent slow upward transport in the tropical tropopause layer (TTL). Sulphur passes the tropopause mainly in the form of carbonyl sulphide (OCS) [Chin and Davis, 1995], but also (in smaller amounts) as SO2, dimethyl sulphide (DMS, (CH3)2S), or hydrogen sulphide (H2S), who all have a relatively low solubility in water that prevents them from being completely scavenged. OCS is be-lieved to be the dominant source, because it is chemically much more stable in the troposphere than the other sulphur compounds. But the higher UV radiation fluxes in the stratosphere allow the photolysis of OCS and make it available for chemical conver-sion. Under non-volcanic conditions, the main sources for these sulphur compounds are marine organisms (DMS and CS2, both at least partly converted to OCS), biomass burning (SO2 and OCS), vegetation (DMS), and anthropogenic emissions (mostly SO2). Ultimately, through photolytical and/or chemical conversion the sulphur-bearing com-pounds end up as sulphuric acid (H2SO4). As the formation of H2SO4 requires water molecules, it depends on the relative humidity. If enough water vapour is available, H2SO4 and H2O rapidly form droplets of sulphuric acid solution due to their very low vapour pressure, but they can also condense on pre-existing particles.
As pressure decreases and temperature increases with increasing height in the strato-sphere, the sulphate particles tend to evaporate above 25-30 km. Therefore, the Junge layer extends roughly from the tropopause to this altitude in the vertical direction. The horizontal distribution is very homogeneous in the zonal (i.e. longitudinal) direction due to strong zonal winds in the stratosphere. The meridional (i.e. latitudinal) Brewer-Dobson circulation (BDC, ascending in the tropics, moving towards the poles, and descending at high latitudes) spreads the aerosol layer over all latitudes.
Stratospheric aerosols can be observed either in situ or remotely. Important balloon-borne in situ measurements have been performed on a regular basis since the 1970s at the University of Wyoming [Deshler et al., 2003] using optical particle counters (OPC). This method uses aerosol forward scattering under a certain angle (25 or 40 ) and Mie theory to determine the particle concentration in several (usually 12) size classes, providing vertically resolved information on the aerosol size distribution. LIDAR is a widely used remote sensing technique that measures the backscattering of laser light by molecules and aerosols at a certain wavelength as a function of altitude. It can be ground based [e.g. Fiocco and Grams, 1964, Jäger, 2005], air-borne, or satellite-borne like the lidar instrument CALIOP on CALIPSO [e.g. Winker et al., 2006, Vernier et al., 2009]. Another technique to measure stratospheric aerosols from space is the solar occultation method. It determines the extinction of sunlight going through the upper layers of the atmosphere and reaching the satellite instrument that is pointing towards the sun (a so-called limb geometry). This technique was used for a long series of stratospheric aerosol measurements including the SAGE II instrument [e.g. McCormick, 1987, Thomason et al., 2008] that operated from 1984 to 2005. It has a nearly global coverage, but at a relatively low time resolution of roughly one month. Furthermore, the geometry only allows measurements for relatively low extinction, so that higher aerosol loadings after volcanic eruptions or in the troposphere cannot be observed. In this thesis we will use in situ OPC measurements and SAGE II satellite observations for the evaluation of the new LMDZ-S3A model, but further observations can be used for model evaluation in the future.
Numerical modelling of aerosols
Since we develop and use a numerical model for the study of stratospheric aerosol in this thesis, we will now shortly discuss the principles of aerosol modelling with a focus on global models of stratospheric sulphate aerosols. Existing models will be discussed in more details within the model description in Chapter 3.
Every model describing the evolution of aerosols has to include a representation of their sources, transport, transformation processes, and sinks. As stratospheric sulphate aerosols mainly form from gaseous precursors, they should also be represented together with the conversion processes leading to particle formation. The aerosol source would then be the nucleation of particles from the gas phase, but it could also include aerosol transport from the troposphere to the stratosphere or a small term for meteoritic dust particles. Transformation processes include the condensation and evaporation of sulphuric acid (and water) and the coagulation of colliding particles. The sinks of stratospheric aerosols are sedimentation, advection, and (to a small extent) diffusion of particles to the troposphere, where wet deposition (in and below clouds) and dry deposition rapidly remove sulphate particles from the air. The transport of the aerosol happens through large-scale advection with the air masses (and through convection and other mixing processes in the troposphere).
There are different approaches for representing aerosols in numerical models. The most simple one is to assume fixed aerosol properties and only track the aerosol mass in a so-called bulk approach, which is computationally cheap. But as this does not permit to study the evolution of aerosol properties, more sophisticated schemes have been developed (see Figure 1.3). In the modal approach one distributes the aerosol over several size modes (often split up into a soluble and an insoluble mode) whose shape is allowed to adjust to a certain degree, as particle mass and number concentrations within the mode are treated separately. This makes it much more flexible and powerful than the bulk approach, but the modal approach is still based on a number of assumptions which are derived from observations. For conditions where such observations are hardly or not available (like large sulphur injections from volcanoes or SAI) it may be better to use the more flexible and accurate sectional approach. Here the particle size range is divided into a larger number of size bins (typically 20 to 40 in global models) and the particle mass (or number) in every single size bin is treated separately. This allows the aerosol size distribution to evolve relatively freely, as it does not prescribe the existence and shape of certain size modes. But the larger number of variables makes the sectional approach computationally more expensive and it is difficult to represent the mixing of aerosol species of different chemical composition.
Table of contents :
1.1 Climate change and the need for mitigation
1.2 Climate engineering
1.2.1 Overview of proposed climate engineering techniques
1.2.2 Stratospheric aerosol injection
1.3 Atmospheric aerosols
1.3.1 Aerosol properties
1.3.2 Tropospheric aerosols
1.3.3 Stratospheric aerosols
1.3.4 Numerical modelling of aerosols
1.4 Outline of the thesis
2 The physics of stratospheric aerosol injection
2.1 The aerosol distribution and its properties
2.2 Competing radiative effects
2.3 Efficacy and scalability
2.4 Short term impact on the atmosphere
2.5 Climate impact
2.6 Aerosol reaching the Earth’s surface
2.7 Tailored climate engineering techniques
2.8 Weaknesses of previous studies
2.9 Research questions addressed in this thesis
3 Description of the LMDZ-S3A atmosphere-aerosol model
3.1 Overview of previous modelling efforts
3.2 The host atmospheric model LMDZ
3.2.1 Model physics and resolution
3.2.2 Tropopause recognition
3.2.3 Quasi-biennial oscillation in the stratosphere
3.2.4 Nudging to meteorological reanalysis
3.3 The sectional stratospheric sulphate aerosol module S3A
3.3.1 Prognostic variables
3.3.2 Semi-prognostic sulphur chemistry
3.3.4 Condensation and evaporation of sulphuric acid
3.3.5 Competition between nucleation and condensation
3.3.7 Aerosol chemical composition and density
3.3.9 Aerosol optical properties
3.3.10 Model code availability
3.4 Conclusions on the model
4 Evaluation of the LMDZ-S3A model
4.1 Validation of aerosol optics and radiative transfer
4.2 Non-volcanic background aerosol
4.3 Mount Pinatubo 1991 eruption
4.3.1 Aerosol distribution and size
4.3.2 Stratospheric temperature anomaly
4.4 Sensitivity studies under Pinatubo conditions
4.4.1 Sensitivity to van der Waals coagulation enhancement factor
4.4.2 Sensitivity to the sulphur dioxide chemical lifetime
4.5 Conclusions on the model evaluation
5 Studying stratospheric aerosol injection with LMDZ-S3A
5.1 Simulation setup
5.2 Results from the reference experiment
5.3 Sensitivity to the injected sulphur dioxide mass
5.4 Comparison with results from Niemeier and Timmreck 
5.5 Sensitivity to injection height
5.6 Sensitivity to spatio-temporal injection pattern
5.7 Effect of radiatively interactive aerosol
5.8 Impact on the quasi-biennial oscillation
5.9 Rapid adjustments and effective radiative forcing
5.10 Impact of aerosol optical properties on the results
5.11 Sulphate impact at the Earth’s surface
5.12 Conclusions on the stratospheric aerosol injection simulation results
6 Combining stratospheric aerosol injection (SAI) and marine cloud brightening (MCB)
6.1 Simulation setup
6.2 Aerosol direct vs. indirect effect in the MCB experiment
6.3 Rapid adjustments and effective radiative forcing of MCB
6.4 Spatial differences between instantaneous and effective radiative forcing 97
6.5 Additivity and complementarity between SAI and tropical MCB
6.6 Conclusions on simulations of SAI and MCB