Marine cloud brightening (MCB) is a proposed technique to limit global warming through injections of sea spray into the marine boundary layer. Using the Norwegian Earth System Model, the sensitivity of MCB to sea salt amount and particle size was studied by running a set of simulations in which Aitken (re = 0.04 µm), accumulation (re = 0.22 µm), or coarse (re = 2.46 µm) mode sea salt emissions were increased uniformly by 10−11 to 10−8 kg m−2 s−1. As desired, accumulation mode particles had a negative radiative effect of down to −3.3 W m−2. Conversely, for Aitken mode particles, injections of 10−10 kg m−2 s−1 or greater led to a positive forcing of up to 8.4 W m−2, caused by a strong competition effect combined with the high critical supersaturation of Aitken mode sea salt. The coarse mode particles gave a positive forcing of up to 1.2 W m−2 because of a decrease in activation of background aerosols. Sensitivity experiments show that the competition effect dominated our results. MCB may have a cooling effect, but if the wrong size or injection amount is used, our simulations show a warming effect on the climate system.
Deliberate engineered cooling of the global climate has received increased scientific interest over the last decade as mitigation strategies to limit global warming are yet to be of significance. One climate engineering strategy involves enhancing the albedo of marine clouds and thus increasing the reflection of solar radiation from the Earth-atmosphere system [Latham, 1990]. The idea is that spraying sea water into the marine boundary layer will increase the number of sea salt particles that ascend into overlying clouds and increase their albedo through the aerosol indirect effect [Twomey, 1974]. This may significantly affect the global radiation budget because of the low albedo of the underlying ocean surface in the subtropics, where extensive low clouds are found.
Early estimates of the global radiative effect of marine cloud brightening (MCB) assumed a certain change in cloud droplet number concentration (CDNC) in seeded clouds and found that MCB could wholly or partially cancel the positive forcing associated with a doubling of CO2 from preindustrial times [Latham et al., 2008; Jones et al., 2009; Rasch et al., 2009]. Korhonen et al.  used a global aerosol transport model, while Pringle et al.  used three independent global aerosol models and a box model and Wang et al.  used a cloud-system-resolving model to investigate what changes in CDNC were achievable from sea salt injections, but these studies did not include estimates of the radiative effect of the cloud seeding.
To our knowledge only three global studies have so far included estimates of the forcing from MCB using fully prognostic treatments of sea salt. Jones and Haywood  studied the radiative impact and climate effects of wind speed-dependent MCB while Alterskjær et al.  injected sea salt uniformly over the ocean to study the geographical distribution of clouds susceptible to seeding. Partanen et al.  used wind speed-dependent emissions of sea salt and estimated that injecting 20.6 Tg yr−1into the most sensitive stratocumulus regions led to a forcing of −0.8 W m−2. Decreasing the size of the particles by 60% or multiplying the injections by five led to a forcing of −2.1 and −2.2 W m−2, respectively.
In this study, we have investigated further the importance of particle size and injection strength. This is necessary to understand the outcome of a potential seeding measure—What sea salt size category would be most effective and therefore least costly? Is the same size category most effective for all injection mass fluxes? And is there a simple linear relation between injection strength and forcing? We have used the Norwegian Earth System Model (NorESM) [Bentsen et al., 2012] to investigate the global sensitivity to particle size and injection strength. We have looked at how injections of sea salt affect the cloud radiative properties and how the increased competition effect resulting from the added sea salt affects our global estimates. In section 'Model and Methods' we describe the model used and the experimental design, while in section 'Results' we go through the results of our experiments, including a presentation and discussion of sensitivity experiments performed to test the robustness of our findings. We summarize and conclude in section 'Conclusions'.
2 Model and Methods
Simulations were performed using the NorESM, which is based on the NCAR (National Center for Atmospheric Research) CCSM4 (Community Climate System Model version 4), but includes new treatments of clouds, aerosols, aerosol-radiation and aerosol-cloud interactions and chemistry [Kirkevåg et al., 2012], along with a new ocean model component. The aerosol module accounts for prognostic sea salt, sulfate (SO4), particulate organic matter, black carbon, and mineral dust as well as two gaseous aerosol precursors producing sulfate (DMS and SO2). The model uses the Mårtensson et al.  scheme for wind speed and temperature-dependent sea salt emissions [Struthers et al., 2011]. It includes sea salt particles with dry number modal radii of 0.022 µm (Aitken mode), 0.13 µm (accumulation mode), and 0.74 µm (coarse mode) and geometric standard deviations of 1.59, 1.59, and 2.0, respectively, corresponding to dry effective radii of 0.04, 0.22, and 2.46 µm.
The aerosol indirect effect is accounted for as described in Hoose et al.  and has a magnitude of −0.91 W m−2 in the current set up. The model uses the Abdul-Razzak and Ghan  cloud droplet nucleation scheme and parametrized updraft velocities following Morrison and Gettelman , with annually averaged in-cloud velocities ranging between about 10 and 100 cm s−1 at a model hybrid level at ~945 hPa over ocean (see Supporting Information). For an overview of typically observed updrafts we refer the reader to Pringle et al. . Model cloud properties were compared to satellite retrievals in, e.g., Alterskjær et al.  and Jiang et al.  indicating a general underestimation of marine cloud CDNC and an overestimation of the simulated cloud liquid water path (LWP). The low CDNC makes the model more susceptible to MCB, while the high LWP has the opposite effect. Alterskjær et al.  also showed that the simulated cloud fraction below 700 hPa is generally overestimated, except for a slight underestimation in the subtropical stratocumulus regions.
2.2 Experimental Design
Cloud brightening simulations were performed by artificially increasing the emissions of each of the three fully prognostic NorESM sea salt modes, meaning that the injected sea salt was treated in a similar manner to natural sea salt and that no explicit assumptions were made on their ability to act as cloud condensation nuclei (CCN). The emissions were uniform and confined between 30°S and 30°N based on the findings of Alterskjær et al. . Most earlier studies have used wind speed-dependent emissions due to the experimental design proposed by Salter et al. . For simplicity, this dependency is not included in our idealized study as we believe that the emission technology may change prior to possible implementation. Injection fluxes ranged from 10−11 to 10−8 kg m−2 s−1 (Table 1), which corresponded to a global sea salt emission increase of from 0.9% to 913%. A list of all simulations performed is shown in Table 1.
Table 1. Sea Salt Injections Used to Simulate MCB and List of Simulations
Sea Salt Injections
aThis row lists the names used to identify injection amounts in kg m−2 s−1. Xs indicate simulations performed.
The model was run offline, meaning that the meteorological evolution remains unchanged between simulations, so that the simulated change in radiative forcing is due to indirect effects only. The control run uses year 2000 greenhouse gas concentrations and year 2000 CMIP5 aerosol emissions and the model resolution is 1.9° × 2.5°. It runs with 26 vertical levels, with a top at about 2 hPa. All simulations include 1 year of spin-up and results presented are averaged over the four following years.
3.1 Reference Simulations
The radiative effect of sea salt injections was studied by investigating the resulting change in shortwave cloud forcing (SWCF) at the top of the atmosphere. Note that the longwave effect was negligible. Figure 1a shows the resulting radiative effect, where reference model runs are marked by Umin = 10 cm/s as this is the minimum sub-grid scale in-cloud vertical velocity used in the model parametrization [Morrison and Gettelman, 2008]. The accumulation mode sea salt is closest to the particle size suggested fit for cloud brightening by Latham  (0.13 µm dry radius), and all simulations with injections of this mode (dark green) show the negative radiative effect desired from climate engineering, with a maximum negative forcing of −3.3 W m−2. By comparison, a doubling of atmospheric CO2 yields a positive forcing of about 3.7 W m−2. Figures 1b and 1c show that the negative forcing associated with MCB is directly linked to an increase in the column-integrated CDNC, which in turn leads to smaller droplets, a decrease in precipitation release, and therefore to an increase in the LWP.
Figure 1a also reveals a weaker negative forcing for the largest compared to the second largest injections of the accumulation mode sea salt. As described in Alterskjær et al.  this can be caused by an increased competition effect, because the added sea salt particles swell, creating a moisture sink which lowers the maximum supersaturation (S) and therefore increases the critical size that particles must have to activate [Ghan et al., 1998; Korhonen et al., 2010]. This competition effect may bring the S below that necessary to activate background aerosols and in some cases below that necessary to activate the added sea salt itself. Figure 1d shows that as the injection flux goes up, the S goes down monotonically, as found in Korhonen et al.  and Wang et al. , and for the accumulation mode 10−8 kg m−2 s−1 case, the globally and annually averaged S is brought down to only 0.078%. Based on Köhler theory, this is well above the critical S of 0.013% needed to activate accumulation mode sea salt, but very close to the critical S of the model's accumulation mode droplets of pure H2SO4 of 0.07% (Figure 1d, lower purple dashed line). Our results show that regionally and temporarily the S goes below the value necessary for activation of accumulation mode sulfate, and there is a reduction in activation of background aerosols leading to a drop in both the CDNC and the LWP (Figures 1b and 1c), so that the efficiency of the seeding goes down for the largest injection mass.
Figures 1e and 1f show the average aerosol number concentration and CDNC obtained in the latitudinal band where sea salt is injected for the model level where the subtropical stratocumulus base is found most frequently (945 hPa). The aerosol number concentration is on the order of 103 to 106 cm−3. For comparison, polluted urban areas have particle concentrations of the order of 105 cm−3 [Pruppacher and Klett, 1997]. Only one simulation gives an averaged CDNC above 375 cm−3, a value suggested to offset the radiative forcing of a doubling of CO2 [Latham et al., 2008].
The change in SWCF resulting from injections of Aitken mode sea salt is shown as dark blue columns in Figure 1a. For the smallest injection amount the cooling effect of Aitken mode sea salt is larger than for accumulation mode sea salt. This is because we add a larger number of particles per injection mass for the Aitken mode, and more particles are activated to form cloud droplets (Figure 1b). A large increase in CDNC leads to an increase in LWP that is larger for Aitken mode than for accumulation mode injections for this mass flux.
Figure 1d shows that the suppression of S is greater for the Aitken mode particles than for accumulation or coarse mode particles, which is due to the larger surface area for water vapor to condense on. Also, as the particles themselves are small in the Aitken mode injections, their critical S (0.16%, Figure 1d, upper purple dashed line) is higher than for accumulation mode sea salt. Figure 1d shows that the S is above this value only for the smallest injection mass flux of Aitken mode sea salt. For injection strengths of 10−10 kg m−2 s−1 or greater, the water vapor condensing on the added sea salt brings the globally averaged S below that necessary to activate the Aitken mode sea salt itself. Thus the added particles may not contribute as CCN and the lowered S instead suppresses activation of background aerosols. As a result, the simulated change in SWCF is positive and increasingly more so as the injection mass flux increases, reaching a maximum forcing in our simulations of +8.4 W m−2. This is the opposite effect of what one would seek to achieve if performing climate engineering.
The radiative effect of injecting coarse mode sea salt is positive in our simulations, albeit much less so than for the Aitken mode particles (Figure 1a, dark red bars). The reason for this warming effect is twofold. First, the number of particles per mass is small for the coarse mode case, so that while the sea salt particles are always large enough to activate, in number they do not contribute much to the CDNC. Second, while the added surface area is low for this mode, water vapor does condense on the sea salt and the S is decreased (Figure 1d). This keeps some of the background aerosols from activating, and as long as this number is greater than the number of new droplets created on the coarse mode sea salt, there will be a warming effect. Figure 1 shows that both the CDNC and the LWP decrease with increasing injection strength. The radiative effect of this sea salt mode increases substantially for the largest injection flux, for which a positive forcing of 1.2 W m−2 is reached (Figure 1a). In this case, the globally averaged S is brought below 0.16% (Figure 1d), the critical S of Aitken mode sea salt. This indicates that injecting enough coarse mode sea salt to shut off the activation of natural Aitken mode sea salt may lead to a warming of the climate system.
In agreement with our findings, Pringle et al.  found in their box model investigation that a decrease in CDNC was possible under certain conditions. However, these conditions were not met in their global study, possibly due to their moderate injection amounts relative to our values.
3.2 Sensitivity Tests
The maximum simulated negative forcing of −3.3 W m−2 was achieved when injecting 10−9 kg m−2 s−1 of sea salt between 30°S and 30°N, corresponding to injecting 5936 Tg of sea salt per year. By comparison Partanen et al.  simulated a forcing of −5.1 W m−2 when injecting only 443.9 Tg yr−1 of sea salt of a comparable size distributed over all ocean regions. A direct comparison of these two studies is difficult due to the differences in experimental design, but the findings indicate that NorESM may be less sensitive to sea salt seeding than the ECHAM5.5-HAM2 used by Partanen et al. . First, the forcing estimate from Partanen et al.  was based on a simulation without “ultrafine” sea salt (dry diameter < 100 nm), which was included in our study. Including these particles in their model weakened their forcing estimate to −4.5 W m−2. Second, the updraft vertical velocity in the Partanen et al.  study was very high, ranging between 1.0 and 1.4 m s−1, which according to the authors eliminated the competition effect from their study. Our results indicate that the competition effect significantly reduces the maximum in-cloud supersaturation. If overestimated, this competition effect will influence our results substantially and may lead to an underestimation of the effectiveness of MCB.
To investigate this closer we conducted two sets of sensitivity simulations. In the first set, the maximum in-cloud supersaturation was set to a fixed value of 0.2% (annually averaged control run S is 0.18% globally around 945 hPa) and the model was run for an injection strength of 10−10 kg m−2 s−1 for the Aitken and accumulation modes and for a strength of 10−8 kg m−2 s−1 for the coarse mode. The second set involved increasing the minimum sub-grid scale in-cloud vertical velocity in the Morrison and Gettelman  parametrization from 10 to 30 cm s−1. This increased the average updraft velocity over ocean between 30°S and 30°N from 30.4 to 41.6 cm s−1 around 945 hPa (see Supporting Information) and increased the maximum in-cloud supersaturation. This setup was run for all injection strengths for the Aitken and the accumulation mode injections and for the 10−10 kg m−2 s−1 injections for the coarse mode sea salt.
3.2.1 Fixed Supersaturation
Setting the in-cloud supersaturation to a fixed value increased the magnitude of the negative forcing of the accumulation mode sea salt from −1.27 to −1.43 W m−2 for an injection strength of 10−10 kg m−2 s−1 (Figure 1a, green dashed line). More strikingly, for the Aitken mode, removing the competition effect led to a change in the sign of the resulting forcing; going from +1.27 to −5.91 W m−2. The high number to mass ratio combined with an S that is above the critical S of the Aitken mode sea salt (0.16%), leads to a very large cooling effect. These results confirm that the competition effect dramatically limits the simulated cooling effect achieved from cloud seeding in the reference simulations.
Latham et al.  suggested that seeding clouds with coarse mode sea salt particles might lead to a warming because the “giant salt nuclei” may lead to an early onset of precipitation and a decrease in LWP. We investigated this by running the 10−8 kg m−2 s−1 coarse mode case with a constant S of 0.2%. This brought the change in SWCF from 1.2 (Figure 1a) to 0 W m−2, indicating that there is no contribution to the simulated warming in our coarse mode results from other sources than the competition effect.
3.2.2 Enhanced Minimum Vertical Velocity
Increasing the minimum in-cloud vertical velocity in the second set of sensitivity simulations led to an increase in the globally and annually averaged control run S from 0.18% to 0.26% around 945 hPa (Figure 1d, yellow bars). This led to activation of smaller background particles in the increased updraft simulations (Umin30) than in the reference simulations (ref) and therefore increased the CDNC (Figure 1b, yellow bars) and decreased the effective radius around 945 hPa from 8.7 µm in ref to 8.2 µm in Umin30. The LWP increased from 126 g m−2 in ref to 137 g m−2 in the Umin30 simulation due to suppression of precipitation. Combined, the influence on the effective radius and the LWP led to a control run that had a higher cloud albedo in Umin30 than in the ref.
For accumulation mode sea salt injections, the radiative forcing achieved was smaller in magnitude than in the ref simulations, except for the simulation of maximum sea salt seeding mass (Figure 1a, green bars). The small magnitude was caused by (i) reduced precipitation suppression due to high control run CDNC [Rasch and Kristjánsson, 1998, equation 21] and (ii) a reduced sensitivity for albedo to change with LWP due to high control run LWP. Assuming an asymmetry parameter of 0.85 and an optical depth gives a change in cloud albedo, A [Hobbs, 1993], with LWP of
where re is the cloud droplet effective radius and ρL the density of water. Combined, (i) and (ii) led to a smaller radiative effect of MCB for Umin30 than for ref accumulation mode seeding.
For the maximum injections of accumulation mode sea salt, the high Umin30 control run S needed a large reduction to bring it below that necessary to activate background sulfate. Contrary to ref simulations, the Umin30 simulations resulted in an S that was well above the critical limit (Figure 1d, lower purple dashed line). The reduced competition effect resulted in an MCB that still served to increase the CDNC and LWP and therefore to increase the magnitude of the SW cloud forcing.
For the Aitken mode, Figure 1d (light blue columns) shows that both 10−11 and 10−10 kg m−2 s−1 injections have Umin30 S around 945 hPa that are above the critical limit to activate the added Aitken mode sea salt. The two cases therefore led to negative forcing, which for the smallest injections was stronger than the forcing produced in the ref simulation (Figure 1a). The increased updraft led to increased activation of the added sea salt particles themselves and therefore to a more effective MCB, while for the corresponding accumulation mode case the increased updraft mainly increased activation of background particles, thus leading to a decrease the efficiency of MCB. The high Umin30 control run S leads to a “delay” in the competition effect, reducing the positive forcing for high Aitken mode injections because more sea salt is needed to remove enough water vapor to bring the S below that necessary for activation of the Aitken mode particles.
For the coarse mode, we saw no significant change between the ref and the Umin30 simulations.
The sensitivity experiments greatly suppress the competition effect, but nevertheless do not lead to negative radiative flux perturbations of the same magnitude as that found by Partanen et al. . One possible reason is that the Umin30 simulations have average updrafts of 41.6 cm s−1 around 945 hPa (see Supporting Information), which is still well below that of Partanen et al. . The NorESM may also be less sensitive to the added sea salt than the ECHAM5.5-HAM2 used by Partanen et al.  due to the large model LWP (equation ((1))). Additionally, Partanen et al.  seeded all ocean areas, while we only seeded between 30°S and 30°N. This may influence the forcing estimates because of the nonlinear relation between added mass and forcing caused by, e.g., differences in the regional meteorological conditions—Are there low clouds above? Are there updrafts that can carry the injected sea salt aloft? Is the particle number in the region already high rendering the clouds less sensitive to the injected sea salt?
This study does not account for kinetic limitations on the activation of giant CCN [Chuang et al., 1997]. In regions of high S this may lead to an overestimated CDNC, while the opposite may be true in regions of low S where an overestimated droplet activation may lead to an exaggerated competition effect. Not accounting for this kinetic limitation may be especially important for the coarse mode sea salt injections, for which we expect the simulated positive radiative forcing to be an upper estimate.
In this study we have investigated how deliberate injections of sea salt into the marine boundary layer affect the global radiative budget as a function of both particle size category and injection mass flux. Using the NorESM we find that injecting accumulation mode sea salt between 30°S and 30°N leads to a desired negative radiative effect of down to −3.3 W m−2 which would almost cancel the positive forcing of a doubling of atmospheric CO2. On the other hand, for Aitken mode injections greater or equal to 10−10 kg m−2 s−1, the simulated net radiative effect is positive, reaching a maximum of 8.4 W m−2. This is because the competition effect reduces the maximum supersaturation and suppresses activation of both the added Aitken mode sea salt and background particles, leading to a decrease in both CDNC and LWP. When coarse mode sea salt is injected, the simulated net radiative effect is always positive. While the coarse mode sea salt itself is large enough to activate, its small number to mass ratio leads to a lower increase in CDNC due to the added sea salt than the decrease in activation of background aerosols due to the competition effect.
We also performed sensitivity tests which show that the size of the competition effect is crucial for the simulated forcing achieved. Omitting this effect led to a negative forcing of −5.9 W m−2 when injecting 10−10 kg m−2 s−1 of Aitken mode sea salt, whereas the same injections gave a positive forcing of 1.3 W m−2 in the reference simulation. Increasing the minimum in-cloud updraft velocity increased the S and led to a “delay” in the competition effect. However, for accumulation mode injections, the increased updraft mainly led to increased activation of background aerosols, leading to a reduced efficiency of the MCB.
The results presented in this study clearly show that the effectiveness of MCB is very sensitive to the injection mass flux and particle size. While a cooling effect is simulated for certain sea salt injections, emitting the wrong particle size or the wrong amount leads to a simulated warming of the climate system, which is opposite to what one seeks to achieve by climate engineering.
This study aims to be a first step in investigating the global effect of MCB of different sized particles and injection amounts. Not all sea salt particle sizes are represented and the model resolution is coarse. Wood  showed that the cloud lifetime effect may change sign depending on the vertical placement of the cloud base and Wang et al.  showed that the cloud albedo effect depends strongly on the fine scale atmospheric state, neither of which are well represented in climate models. However, coarse models such as the NorESM are currently the only tool available to study the global effects of MCB.
This study was partly funded by the European Commission's 7th Framework Program through the IMPLICC project (FP7-ENV-2008-1-226567), and by the Norwegian Research Council through the EarthClim project (207711/E10) and its programme for supercomputing (NOTUR) through a grant for computing time. The authors are thankful to Philip Rasch and to the IMPLICC consortium for constructive discussions.