Marine cloud brightening through sea spray injection has been proposed as
a climate engineering method for avoiding the most severe consequences of
global warming. A limitation of most of the previous modelling studies on
marine cloud brightening is that they have either considered individual
models or only investigated the effects of a specific increase in the number
of cloud droplets. Here we present results from coordinated simulations with
three Earth system models (ESMs) participating in the Geoengineering Model
Intercomparison Project (GeoMIP) G4sea-salt experiment. Injection rates of
accumulation-mode sea spray aerosol particles over ocean between
30
Attempts to lower global emissions of
Earlier modelling studies on sea spray climate engineering investigated the radiative effects of marine cloud brightening mainly by prescribing an increase in CDNC (Latham et al., 2008; Jones et al., 2009; Rasch et al., 2009). However, more recent studies have included the sea salt injection process and the activation of the injected particles to cloud droplets, thereby taking into account radiative effects of both activated cloud droplets and non-activated particles (Jones and Haywood, 2012; Partanen et al., 2012; Alterskjær et al., 2013). As a result, sea spray climate engineering is now sometimes referred to as marine sky brightening (Muri et al., 2015), as it may include radiative impacts of injected particles both through cloud brightening (the aerosol indirect effect) and due to increased scattering of solar radiation outside clouds (the aerosol direct effect). One of the more recent modelling studies on sea spray climate engineering applied emission patterns to maximize either the direct or the indirect radiative effect of the injected particles, limiting the emission area in both cases to 10 % of the ocean (Jones and Haywood, 2012). In that study, maximizing the indirect effect generated the largest radiative impact and resulted in the largest cooling, but it should be noted that the direct effect was of comparable magnitude to that of the indirect effect within the region specified to maximize the aerosol indirect effect. In another recent modelling study, the aerosol direct effect was estimated to contribute 29 % to the total radiative forcing when sea spray climate engineering was assumed to take place over the global oceans (Partanen et al., 2012). In contrast, one recent study indicated a dominant contribution from the aerosol direct effect to the total radiative forcing (Kravitz et al., 2013).
A weakness of almost all of the previous studies on sea spray climate engineering is that they have only considered individual models. It is therefore uncertain to what extent the results in many of the previous studies are robust, considering the differences in parameterizations across models of, for example, clouds and their interaction with aerosols. Furthermore, results from individual model studies in the past are generally not directly comparable because of discrepancies in the model set-up or in the details of what was actually simulated. Therefore, the idea behind the Geoengineering Model Intercomparison Project (GeoMIP) (Kravitz et al., 2011, 2013) is that model experiments should be standardized, and that an ensemble of multiple Earth system models (ESMs) should be executed for a number of climate engineering experiments. By the use of such ensembles, it is possible to estimate an uncertainty in the predicted climate response.
In this study we use three fully coupled atmosphere–ocean ESMs and run the GeoMIP G4sea-salt experiment (see Kravitz et al., 2013, and Sect. 2) focusing on the response of Earth's radiation balance to injection of sea salt particles, both in clear-sky conditions and from changes in cloud properties.
Coupled state-of-the-art Earth system models provide the best tools for
assessing the climate response to solar climate engineering. Three fully
coupled ESMs – NorESM1-M (Bentsen et al., 2013), GISS-E2-R (Schmidt et al.,
2014), and HadGEM2-ES (Collins et al., 2011) – were used in this study. For
the atmospheric component, NorESM1-M runs at
The treatment of the natural emissions of sea salt is prognostic in NorESM1-M
and GISS-E2-R, with emission fluxes depending on wind speed and sea surface
temperatures in NorESM1-M (Struthers et al., 2011), and on wind speed only in
GISS-E2-R (Monahan et al., 1986). HadGEM2-ES uses a diagnostic treatment of
natural sea salt aerosol number concentration with concentrations depending
on wind speed (Jones et al., 2001). Hygroscopic growth of aerosol particles
is accounted for in all three models, and this process affects dry removal
rates as well as aerosol–radiation interactions. In NorESM1-M, hygroscopic
growth is treated as described by Seland et al. (2008), by applying the form
of Köhler equation given in Kirkevåg and Iversen (2002). In
GISS-E2-R, uptake of water by hygroscopic species such as sea salt and
sulfate is parameterized in terms in terms of an external mixture of the dry
aerosol and a pure water aerosol with sizes set to reproduce the extinction
efficiency and asymmetry parameters of the solute aerosol at the laboratory
wavelength of 633
Dry deposition of aerosol particles in all three models is parameterized using resistance schemes analogous to electrical resistance (e.g. Seinfeld and Pandis, 1998). The dry deposition velocity thus depends on particle size. Gravitational settling is included in the calculation of the dry deposition velocity. Rainout is determined by autoconversion in all models and includes re-evaporation of precipitation. Wet deposition in NorESM1-M is parameterized as in Iversen and Seland (2002), with an in-cloud scavenging coefficient defined as the mass fraction of the aerosol mode within the cloud droplet. Wet deposition in GISS-E2-R and HadGEM2-ES are described in more detail by Koch et al. (2007) and Bellouin et al. (2011), respectively.
The following experiments are analysed in this study:
RCP4.5: Representative Concentration Pathway 4.5 (Meinshausen et al., 2011), where the total radiative forcing reaches
4.5 G4sea-salt: this experiment follows the experimental design of the Geoengineering Model Intercomparison Project (GeoMIP) G4sea-salt experiment
(Kravitz et al., 2013). Sea spray climate engineering is implemented on top
of an RCP4.5 scenario to generate a top-of-the-atmosphere (TOA) global-mean
effective radiative forcing (ERF) of Fixed sea surface temperature (SST) experiments: the G4sea-salt and RCP4.5 experiments were simulated also with fixed SST, as taken
from year 2020 of the RCP4.5 simulation (Kravitz et al., 2013). All other
forcing was kept the same as in year 2020 of RCP4.5, with the only difference
being increased sea salt emissions. The experiments were run for 10 years
for each model in order to determine the injection rate of sea salt aerosol
in each model required to generate a global-mean ERF of
Size distributions for total sea salt injections (30
The injected sea salt particles within the G4sea-salt experiment have
a median dry radius of 0.13
Cloud fraction for low clouds averaged over 2020–2030 within the
RCP4.5 scenario for
There is large uncertainty in which particle size would be optimal for sea
spray climate engineering. The mass scattering efficiency of NaCl particles
with a refractive index of 1.544 at a wavelength of 550
Global-mean TOA effective radiative forcing of the injected
particles in total
The fully coupled RCP4.5 simulations include two realizations with NorESM1-M, three realizations with GISS-E2-R, and four realizations with HadGEM2-ES. The fully coupled G4sea-salt simulations include two realizations with NorESM1-M, three realizations with GISS-E2-R, and one realization with HadGEM2-ES.
A key variable in the models when considering sea spray climate engineering
is the amount of low clouds over the ocean, in particular subtropical
stratocumulus clouds off the west coasts of North America, South America, and
southern and northern Africa. These regions have been assessed to be most
susceptible to brightening (Salter et al., 2008; Alterskjær et al., 2012;
Jones and Haywood, 2012). Figure 2 shows the low-level cloud fraction below
850
The sea salt injection rates between 30
The global-mean ERF by the injected sea salt particles, for the rates given
above, is relatively constant at
TOA mean effective radiative forcing over the 10 years of
simulation with fixed SST for
The clear-sky ERF by the injected particles in Fig. 3b is of comparable
magnitude for the three models, despite the higher sea salt mass injection
rates and larger size of the injected particles in GISS-E2-R compared to the
other two models. The surface area size distribution (Fig. 1b) is closely
related to the amount of light scattered by the sea salt particles and
thereby the clear-sky ERF in Fig. 3b. For a full description of Mie
scattering, however, one also needs to take into account variations in the
scattering coefficient with particle size, which is done in the radiative
transfer calculations in the models. The total particle number injections
(integrated over the particle number size distributions in Fig. 1a) are
The ratio of the total ERF to the clear-sky ERF at the TOA averaged
over the 10 years of simulation with fixed SST for
The effective radiative forcing by the injected particles at the TOA varies
spatially between
Mean difference in sea salt mass concentration in the lowest model
layer between G4sea-salt and RCP4.5 averaged over 2035 and 2065, for
Mean relative change in CDNC due to sea salt injection for NorESM1-M
Although the effective radiative forcing by the injected particles in
NorESM1-M and HadGEM2-ES is at a maximum over some of the marine subtropical
stratocumulus regions previously identified as optimal for marine cloud
brightening, the ERFs in Fig. 4 are not as dominated by these regions as in
the study by Partanen et al. (2012) with ECHAM5.5-HAM2. In that study, the
maximum ERF of their sea spray climate engineering exceeded
Although GISS-E2-R (Fig. 4b) has maxima in ERF in the same subtropical
stratocumulus regions as the other two models, there is less horizontal
variability in ERF in GISS-E2-R. An exception is the Intertropical
Convergence Zone (ITCZ), where the ERF is considerably weaker, likely due to
the large amounts of high clouds in these regions (not shown). The presence
of middle to high-level clouds is not optimal for sea spray climate
engineering as these clouds block out some of the incoming solar radiation
and make a negligible contribution to the aerosol indirect effect. A weaker
ERF along the ITCZ can be seen to some extent also over the Pacific in
NorESM1-M (Fig. 4a). The more homogeneous ERF field for GISS-E2-R compared to
the two other models is likely due to the smaller amount of low-level clouds
in GISS-E2-R compared to the two other ESMs (Fig. 2). This means that the
aerosol direct effect likely contributes more to the total ERF in GISS-E2-R,
leading to fewer horizontal variations in ERF. This hypothesis of a low
contribution of the aerosol indirect effect to the ERF in GISS-E2-R is
supported by the absence of correlation between the strength of the ERF and
low-level cloud cover (
Figure 5 shows the ratio of the total ERF to clear-sky ERF at the TOA for
each of the three models. This figure provides information on whether the
clouds that are present increase the ERF by the injected particles compared
to clear-sky conditions. Red-coloured areas indicate an increased ERF when
clouds are present, and thereby an effective aerosol indirect effect, whereas
blue-coloured regions indicate an enhanced ERF for clear-sky conditions. The
impact of the subtropical stratocumulus clouds on the ERF by the injected
particles, relative to clear-sky conditions, is largest in HadGEM2-ES
(Fig. 5c), with the ratio of total ERF to clear-sky ERF locally being higher
than
In summary, the presence of low clouds in the subtropical high-pressure regions has the effect of increasing the ERF by the injected particles compared to clear-sky conditions, and this enhancement in ERF due to the aerosol indirect effect is most pronounced in HadGEM2-ES. However, in most other regions within the area of sea salt injection, the ratio of total ERF to clear-sky ERF is close to one in all the models, which indicates that the presence of clouds in most regions does not significantly increase the ERF compared to clear-sky conditions. This finding, together with the relatively small horizontal variability in ERF compared to Partanen et al. (2012) and weak or non-existent correlations between ERF and low-level cloud cover, suggests that the aerosol direct effect probably makes a larger contribution to the total ERF in this study compared to the study by Partanen et al. (2012), where the aerosol direct effect contributed 29 % to the total ERF by sea spray climate engineering.
Mean relative change between G4sea-salt and RCP4.5 in cloud-top
effective radius for NorESM1-M
The injection rates generating a global-mean effective radiative forcing of
The multi-model mean difference between the G4sea-salt experiment
and RCP4.5 averaged over years 2035–2065. The multi-model mean difference
refers to the mean of all three models, NorESM1-M, GISS-E2, and HadGEM2-ES.
Hatching denotes areas where the models disagree on the sign of the change.
Change in
Relations between the relative change in cloud optical depth
(
One of the advantages of simulating sea spray climate engineering in ESMs
through sea salt aerosol emissions, compared to just increasing the
CDNC, is that the cloud droplet activation process is taken into
account. Previous studies have shown that injection of sea spray particles in
some circumstances may actually reduce the CDNC due to increased
competition for water vapour and reduced activation of background aerosol
particles (Korhonen et al., 2010; Alterskjær et al., 2012).
Alterskjær and Kristjánsson (2013) showed in a single-model study
that while the injection of accumulation-mode particles increased the
CDNC, the injections of Aitken- or coarse-mode particles could have the
opposite effect with a reduction in CDNC. As mentioned in Sect. 2.2,
the injected particles in this study are accumulation-mode particles with
a median dry radius between 0.10 and 0.44
As shown in Fig. 7, the sea salt injection enhances the CDNC in lower
layers within the whole injection area in all three ESMs. The mean percentage
increase in CDNC within the injection area averaged for the period
2035–2065 (only grid cells over ocean included) is 153 % in NorESM1-M,
42 % in GISS-E2-R, and 89 % in HadGEM2-ES (Table 1). The largest
enhancements in CDNC generally occur in regions where the background
CDNC is low. The smaller percentage increase in CDNC in
GISS-E2-R compared to the other two models is likely due to the higher
background CDNC in GISS-E2-R. Over the Arctic region, there is
a relatively large reduction in CDNC in NorESM1-M (Fig. 7a) and
HadGEM2-ES (Fig. 7c). However, the CDNC in the Arctic region is as low
as
Mean percentage changes in CDNC, cloud-top effective radius, cloud
water path, cloud cover, cloud optical depth,
precipitation, and surface air temperature (
Difference in net SW radiation at the TOA between G4sea-salt and
RCP4.5
Mean change in net SW radiation (
As expected, the cloud-top effective droplet radius,
The cloud optical depth (
Figure 9 shows the multi-model mean changes in omega-vertical velocity (a), cloud water path (b), and precipitation (c). In large regions over the eastern Pacific Ocean, reduced ascent (or increased subsidence) (Fig. 9a) is accompanied by reductions in cloud water path (Fig. 9b) and precipitation (Fig. 9c). On the other hand, enhanced ascent over, for example, Africa, northern South America, and in the South Pacific Convergence Zone coincides with increased cloud water path and precipitation. These patterns of enhanced cloud water, precipitation, and atmospheric upward motion over low-latitude continents combined with reduced cloud water, precipitation, and ascent over some low-latitude ocean regions have been reported previously by Bala et al. (2011), Alterskjær et al. (2013), Niemeier et al. (2013), Crook et al. (2015), and Stjern et al. (2017). This is a result of reduced absorption of solar radiation over ocean where sea salt concentrations are elevated while continental regions are left less affected, increasing the land–sea gradient over the tropics. This induces enhanced convection over land and thereby increased cloud formation and precipitation and reduced cloud formation over ocean due to reduced upward motion or increased subsidence. Furthermore, the increase in upward motion and cloud water content north-east of Australia, and the reduction in these variables over the eastern Pacific Ocean west of South America, indicate a strengthening of the Pacific Walker cell and South Pacific Convergence Zone.
In summary, the aerosol indirect effect of the injected sea salt particles
can be seen in the mean increase in CDNC, mean decrease in cloud-top
effective droplet radius, and mean increase in cloud optical depth over the
injection area. However, in these fully coupled simulations the aerosol
direct and indirect effects of the injected sea salt particles also cause
changes in the atmospheric circulation that generate a redistribution of
cloud water, with increasing cloud water and precipitation in regions of
enhanced atmospheric ascent and decreasing cloud water and precipitation in
regions of decreased atmospheric upward motion. Within the injection area,
the local response in cloud optical depth is controlled to a larger extent by
these changes in cloud water than by changes in CDNC or
The global-mean difference in net SW radiation at the TOA (Fig. 11a) between
G4sea-salt and RCP4.5 is rather constant at
Whereas the global-mean changes in net SW radiation at the TOA shown in
Fig. 11a and b are to some extent influenced by changes in surface albedo,
the corresponding changes over the injection area over ocean between
30
In this study, we have analysed the GeoMIP G4sea-salt experiment using three
different ESMs: NorESM1-M, GISS-E2-R, and HadGEM2-ES. Sea spray climate
engineering is applied on top of the RCP4.5 scenario between years 2020 and
2070, with sea salt injection rates set to generate a global-mean ERF of
Although sea spray climate engineering is often referred to as marine cloud brightening, we find that the global-mean clear-sky ERF is as large as the total ERF in all three ESMs, indicating the large potential of the aerosol direct effect in regions of low cloudiness. The largest regional enhancement in ERF due to the presence of clouds, compared to the ERF in clear-sky conditions, occurs as expected in the subtropical stratocumulus regions off the west coasts of the American and African continents. However, in most regions outside these subtropical regions, the clear-sky ERF is as large as the total ERF. Furthermore, the correlation between low-level cloud cover and the strength of the ERF by the injected particles within the injection area is weak or non-existent in the models. These factors together indicate that, with the exception of the subtropical stratocumulus regions, sea spray climate engineering is as efficient in clear-sky conditions as in cloudy-sky conditions.
The aerosol indirect effect of the injected particles is seen in the increase
in CDNC, reduction in
These results show that many important secondary effects on clouds are neglected if sea spray climate engineering is investigated by the simplified method of increasing the number of cloud droplets, as has been done previously in a number of studies (Latham et al., 2008; Jones at al., 2009; Rasch et al., 2009), or when considering injection in a limited area (Jones and Haywood, 2012). The results here may also have implications for which regions may be most effective in generating a cooling from sea spray injection, as the aerosol direct effect likely plays a more important role than previously thought.
All model data are available through the Earth System Grid or upon request to the contact author.
The authors declare that they have no conflict of interest.
This article is part of the special issue “The Geoengineering Model Intercomparison Project (GeoMIP): Simulations of solar radiation reduction methods (ACP/GMD inter-journal SI)”. It is not associated with a conference.
Lars Ahlm, Helene Muri, Camilla W. Stjern, and Jón Egill Kristjánsson were supported by the Research Council of Norway (grant number 229760/E10) (EXPECT). Lars Ahlm was also supported by the NordForsk approved Nordic Centre of Excellence “CRAICC”, and by the Swedish Research Council FORMAS (grant 2015-748). Helene Muri received further funding from RCN grant 261862/E10. Norwegian Research Council's Program for supercomputing, NOTUR, provided computing time (NN9182K). Data storage on Norstore (NS9033K, NS2345K). Simulations with GISS-E2-R, performed by Ben Kravitz, were supported by the NASA High-End Computing (HEC) Program through the NASA Center for Climate Simulation (NCCS) at Goddard Space Flight Center. The Pacific Northwest National Laboratory is operated for the US Department of Energy by Battelle Memorial Institute under contract DE-AC05-76RL01830. We also thank all participants of the Geoengineering Model Intercomparison Project and their model development teams, CLIVAR/WCRP Working Group on Coupled Modeling for endorsing GeoMIP, and the scientists managing the Earth System Grid data nodes who have assisted with making GeoMIP and CMIP5 output available. Edited by: Ulrike Lohmann Reviewed by: two anonymous referees