Journal of Geophysical Research: Atmospheres

Climate-induced changes in sea salt aerosol number emissions: 1870 to 2100

Authors


  • All supporting information may be found in the online version of this article.

Abstract

[1] Global climate model output is combined with an emission parameterization to estimate the change in the global and regional sea salt aerosol number emission from 1870 to 2100. Global average results suggest a general increase in sea salt aerosol number emission due to increasing surface wind speed. However, the emission changes are not uniform over the aerosol size spectrum due to an increase in sea surface temperature. From 1870 to 2100 the emission of coarse mode particles (dry diameter DP>655 nm) increase by approximately 10 % (global average), whereas no significant change in the emission of ultrafine mode aerosols (dry diameter Dp<76 nm) was found over the same period. Significant regional differences in the number emission trends were also found. Based on CAM-Oslo global climate model output, no straight-forward relationship was found between the change in the number emissions and changes in the sea salt aerosol burden or optical thickness. This is attributed to a change in the simulated residence time of the sea salt aerosol. For the 21st century, a decrease in the residence time leads to a weaker sea salt aerosol-climate feedback that what would be inferred based on changes in number emissions alone. Finally, quantifying any potential impact on marine stratocumulus cloud microphysical and radiative properties due to changes in sea salt aerosol number emissions is likely to be complicated by commensurate changes in anthropogenic aerosol emissions and changes in meteorology.

1 Introduction

[2] Natural aerosols are an important component of the climate system [Carslaw et al., 2010]. They can be emitted directly from the Earth's surface (primary aerosols) or be formed in the atmosphere from precursor gases (secondary aerosols). Primary natural aerosols include sea spray, mineral dust and organic material, for example plant debris, fungal spores, bacteria and viruses. Volcano as well as terrestrial and oceanic ecosystems emit large amounts of precursor gases that can either condense on existing aerosols or form new particles. Wildfires are a sporadic source of both primary and secondary natural aerosols. Natural aerosols dominate the atmospheric mass and/or number concentration in many regions on Earth [Seinfield and Pandis, 2006]. Therefore, in addition to anthropogenic emission inventories, accurate estimates of natural aerosol emissions are crucial when conducting comprehensive model simulations of the Earth's climate. In particular, the response of natural aerosol emissions to climate change must be accurately estimated, since the processes that produce natural aerosols are susceptible to global warming [Carslaw et al., 2010].

[3] Sea salt aerosol, an important component of the natural aerosol, contributes significantly to the global aerosol burden and radiative budget [Schulz et al., 2006] and may be the dominant contributor to direct scattering of solar radiation [Quinn et al., 1998; Bates et al., 2006]. Sea salt aerosol is also an important source of cloud droplets in the remote marine environment [Monahan et al., 1986; Pierce and Adams, 2006]. Estimates of the current top-of-the-atmosphere radiative forcing due to sea salt aerosol vary considerably [e.g. Fan and Toon, 2011, and references therein]. This large uncertainty range arises due to differences in the representation of the sea salt aerosol atmospheric lifecycle in climate models, including emissions, in-situ atmospheric processes and deposition [Textor et al., 2006].

[4] The implications of climate change on sea salt aerosol were discussed by Latham and Smith [1990] who assumed a logarithmic dependence between the aerosol number concentration and 10-meter wind speed (U10) over the oceans. Their results suggest that a change in sea salt aerosol concentrations may constitute a substantial negative feedback due to a warming climate. In a subsequent study, Penner et al. [2001] used projections of U10 from the IPCC AR3 models to calculate a mean feedback as large as -0.8 Wm-2 for the year 2100, although the sign and magnitude of the effect was highly model dependent. Korhonen et al. [2010] concluded that sea salt aerosol concentrations at southern mid-latitudes (50 to 65 °S) have increased by 22% from the 1980s to present day, primarily due to changes in surface winds driven by the development of the Antarctic ozone hole. The local radiative forcing associated with this increase in sea salt aerosol was estimated to be -0.7 Wm-2. Mahowald et al. [2006] estimated changes in sea salt aerosol over longer time periods (last glacial maximum, preindustrial, current and with doubled CO2) and concluded that the atmospheric loading of sea salt globally is not very sensitive to climate change (changes less than 5 %). However, regional differences may be as large as 40 %.

[5] A wide variety of sea salt aerosol emission parameterizations have been developed based on both laboratory and in-situ measurements [de Leeuw et al., 2011]. Despite the fact that a number of source parameterization evaluations have been completed [Pierce and Adams, 2006; Tsyro et al., 2011; de Leeuw et al., 2011], differences of one order in magnitude or more remain between estimates of the size-dependent sea salt flux. Over the open ocean, the main physical driver of sea salt emissions is the surface wind speed [Blanchard, 1963]. However, a number of studies support the conclusion that the emissions are also dependent on the sea surface temperature (SST) [Mårtensson et al., 2003; Hultin et al., 2010; Nilsson et al., 2010; Jaegle et al., 2011]. Although not fully understood at a mechanistic level, there are a number of mechanisms via which SST could affect the sea salt aerosol production flux [e.g. Lewis and Schwartz, 2004; Anguelova and Webster, 2006]. However, only a few of the proposed source parameterizations include a dependence on SST.

[6] In this study we examine the long-term effect of changes in U10 and SST on sea salt aerosol number emissions by combining output from the WCRPs CMIP3 multi-model data set [Meehl et al., 2007a] with the aerosol source function by Mårtensson et al. [2003]. This approach allows us to quantify the relative importance of changes in U10 and SST in determining changes in global sea salt aerosol number emissions in terms of number and size distribution over the period 1870 to 2100. The change in sea salt aerosol emission is driven by changes in the climate parameters: U10, SST and sea ice cover. We use a multi-model ensemble of these climate parameters to calculate the sea salt aerosol number emissions, since the consensus view is that no single climate model is superior in its ability to forecast key climatic features [Lambert and Boer, 2001]. In addition, a multi-model ensemble allows for sampling of uncertainties in the modelled fields [Tebaldi and Knutti, 2007].

[7] Although this study is primarily focused on estimating the change in sea salt aerosol number emissions and the importance of SST when calculating sea salt aerosol emissions over centennial time scales, we also discuss the potential impact of climate change on the atmospheric residence time and atmospheric burden of sea salt aerosol. To this end, a dedicated set of global climate model simulations representative of the years 1890, 1990, 2010 and 2090 were conducted using the CAM-Oslo model [Seland et al., 2008]. CAM-Oslo includes an explicit aerosol module where sea salt aerosol in-situ processing (condensation, coagulation, gaseous and aqueous chemistry, in-cloud and below-cloud scavenging, transport and deposition) is calculated online within the model. From the CAM-Oslo simulations, changes in the atmospheric burden, residence time and radiative forcing of sea salt aerosols can be calculated and directly related to the emission change.

[8] Section 2 describes the method used for the study, i.e. the sea salt parameterization, the CMIP3 climate model output, the CAM-Oslo model and the simulations conducted using CAM-Oslo. In Section 3, the sea salt emission changes estimated from the CMIP3 model data output is presented and discussed, both as global averages and with focus on eight oceanic sub-regions. Section 4 presents the changes in sea salt burden, aerosol optical thickness (AOT) and marine stratocumulus cloud properties derived using the CAM-Oslo simulations and discusses potential climate feedbacks. Conclusions are drawn in Section 5.

2 Methods

2.1 Sea Salt Aerosol Source Function

[9] A full documentation of the sea salt aerosol source parameterization can be found in Mårtensson et al. [2003] (hereafter referred to as MN03). Here we only describe the characteristics of the parameterization that are most important for the present study. The MN03 parameterization is derived from laboratory measurements of aerosol fluxes using synthetic sea water within a temperature range of -2 to 25 °C. The MN03 parameterization is defined for sea salt particles with dry diameters ranging from 20 nm to 2.8 μm. For particles larger than 2.8 μm, the source parameterization of Monahan et al. [1986] is typically used. The size-resolved sea salt aerosol emission flux F0 is described using the equation:

display math(1)

where Ak and Bk are aerosol size (Dp) dependent terms (approximated by fourth order polynomials), SST is the sea surface temperature and W is the fractional area of white caps, parameterized as a function of U10:

display math(2)

[10] The MN03 source function has been evaluated with observations and compared with other sea salt aerosol source functions in several previous studies [e.g. de Leeuw et al., 2011; Jaegle et al., 2011; Tsyro et al., 2011; Pierce and Adams, 2006]. Pierce and Adams [2006] concluded that, in comparison with several other sea salt aerosol source functions, the MN03 parameterization generated the best agreement with measured marine number distributions. On the other hand, the sea salt mass concentrations simulated using the MN03 parameterization were underestimated by a factor between 3 and 4 compared to observations, which was primarily attributed to the fact that the MN03 source function was only applied for aerosols with a Dp ≤ 2.8 μm. In a recent paper by Tsyro et al. [2011] the parameterization of sea salt aerosol emissions are discussed and results obtained with five different source functions in a series of box model calculations are presented. The largest discrepancies between the different sea salt aerosol source functions was found for particles less than 0.1 μm at 80 % relative humidity, and a derivative of the MN03 source function generally showed higher number fluxes compared to other source functions over this particle size range.

[11] Sea salt aerosol number emissions as a function of dry particle size calculated using the MN03 source function are shown in Figure 1 for a range of SSTs. The three size classes depicted in Figure 1 (ultrafine, accumulation, and coarse) represent aerosol size ranges commonly found in the scientific literature [Seinfield and Pandis, 2006]. For consistency with the discussion of the CAM-Oslo model results, the size ranges were chosen to be equal to the width at half height of the three log-normal modes used to describe the emission of sea salt aerosol in the CAM-Oslo (Section 2.3 and Table 1). The main features of the MN03 source function are: (a) particles around 60 nm in diameter dominate the number emissions, (b) the fluxes in the smallest size range, including the ultrafine mode, decrease with increasing SST, whereas the opposite occurs for larger particles (Dp>250 nm), including the coarse mode, and (c) the particle fluxes for the intermediate sizes, including the accumulation mode, are relatively insensitive to SST.

Figure 1.

Sea salt aerosol emissions as a function of dry particle diameter (Dp), calculated using the MN03 source parameterization [Mårtensson et al., 2003] for three different values of SST (assuming a U10 of 9 ms-1). The diameter ranges for three particle ranges (ultrafine, accumulation and coarse) discussed in the text are also indicated.

Table 1. Median dry radius and geometric standard deviation of the three primary sea salt aerosol modes used in CAM-Oslo. The dry radius range at half height for the three modes is also listed (cf. Figure 1)
 rp (nm)σrp range at half height (nm)
a1 (ultrafine)221.5912.75 - 37.95
a2 (accumulation)1301.5975.5 - 224.5
a3 (coarse)7402.0327.5 - 1400

[12] The number emission fluxes, integrated separately over the three size classes depicted in Figure 1 (ultrafine, accumulation and coarse) are discussed below. The temperature dependence of these integrated emission classes can be deduced from Figure 1: ultrafine number emissions decrease with increasing SST whereas accumulation and coarse mode number emissions are positively correlated with SST. Although not discussed further, the total integrated number flux (20 nm to 2.8 μm) decreases with increasing SST, because the total number flux is dominated by ultrafine particles. In contrast, since the sea salt aerosol mass is dominated by coarse mode particles the total sea salt aerosol mass emission increases with increasing SST.

2.2 CMIP3 Model Ensembles

[13] Two ensembles of CMIP3 model simulations [Meehl et al., 2007a] were used to provide U10 and SST input to the MN03 source function. The c20c3m (the climate of the 20th century simulations) from 1870 to 2000 and SRES A1B [Nakicenovic et al., 2000] future climate projection for the years 2000 to 2100. Only models with both U10 and SST available in the CMIP3 database were included in the analysis, which resulted in 37 ensemble members for the c20c2m ensemble and 35 members for the SRES A1B ensemble. A list of the models that contribute to the two ensembles is provided as supplementary material.

[14] The CMIP3 archive provides monthly mean output (U10 and SST) which are used to calculate the size dependent sea salt aerosol emissions. The highly non-linear U10 dependence of the MN03 source function (Equation (2)) means that the emissions calculated using monthly averaged U10 will generally be biased low compared to calculations based on wind data at higher temporal resolutions. This was confirmed via tests with the CAM-Oslo model (Section 2.3) comparing emissions calculated on-line every model time step (20 minutes) with off-line calculations using monthly average U10 and SST from the same model simulation. The off-line emission estimates did indeed show a significant low bias compared to the on-line calculations (annual emissions ~35 % lower when averaged between 55°S and 65°N). However, no statistically significant difference in this low bias was found when comparing CAM-Oslo output for pre-industrial (1890) with present day (2010) and future (2090) conditions, consistent with the results of Penner et al. [2001]. This suggests that the percentage changes in emissions calculated from monthly average CMIP3 model output discussed below are consistent with higher temporal resolution calculations despite the low bias in the absolute numbers.

[15] The two ensembles (c20c3m and SRES A1B) are based on different sets of models. In addition the boundary conditions specified for the two scenarios are not equivalent for the year 2000. For this reason, no attempt to combine the model output into a continuous record from 1870 to 2100 was made and the results from the two model ensembles are therefore presented separately.

2.3 CAM-Oslo Global Climate Model and Sensitivity Simulations

[16] The CAM-Oslo global climate model is based on the CAM3 general circulation model [Collins et al., 2006] and includes a detailed aerosol microphysical module as described by Seland et al. [2008]. The set of sub-grid scale physical parameterizations used in the CAM3 model is documented in Collins et al. [2006]. The CAM3 configuration employed here is based on an Eulerian dynamical core at T42 spectral truncation on a Gaussian grid (approximately 2.8 × 2.8 degrees) and a hybrid-n vertical coordinate with 26 levels.

[17] The aerosol module considers five prognostic aerosol compounds: sulfate, particulate organic matter, black carbon, sea salt (SS) and mineral dust as well as the two gaseous precursors (dimethylsulfide (DMS), and sulfur dioxide). The mass-concentrations of aerosol species are tagged according a process-level description of aerosol dynamics. The processes included in the model are gaseous and aqueous chemistry, nucleation, condensation, and coagulation.

[18] Within CAM-Oslo, the primary aerosol size distributions for all aerosol components, is approximated using log-normal modes [Seland et al., 2008]. The primary sea salt aerosol size distribution at the time of emission is described by three log-normal modes (ultrafine, accumulation and coarse). The median radii and geometric standard deviations for these modes are listed in Table 1.

[19] For the calculation of cloud condensation nuclei activation and aerosol optical properties, the primary aerosol size distribution is modified in accordance with the process-tagged aerosol mass concentrations using a sectional bin framework with 44 size bins. All prognostic compounds, except DMS, are subjected to production, transport and dry and wet deposition [Iversen and Seland, 2002, 2003]. For relative humidity values below 100 % the hygroscopic growth and general growth by condensation, coagulation and cloud processing [Kirkevåg and Iversen, 2002] form the basis for extensive pre-calculated lookup tables for the aerosol optical properties used in the model.

[20] The aerosol module of CAM-Oslo and the simulated direct and indirect aerosol effects have been comprehensively evaluated through the AeroCom (Aerosol Comparisons between Observations and Models) aerosol modelling initiative [Kinne et al., 2006; Penner et al., 2006; Schulz et al., 2006; Koch et al., 2009; Textor et al., 2006]. On a global scale, the general characteristics of the aerosol fields and radiative forcings simulated by CAM-Oslo and its predecessor CCM-Oslo are within the range of the results obtained with the other models included in the AeroCom project [Kinne et al., 2006; Schulz et al., 2006; Quaas et al., 2009]. Sea salt aerosol mass concentrations simulated by CAM-Oslo have been compared with in-situ measurements, see Kirkevåg et al. [2012] and Figure 2 in Seland et al. [2008].

Figure 2.

Percentage change in the ensemble averaged sea salt aerosol emissions for the three aerosol size classes (cf. Figure 1). The asterix signifies to the level of statistical significance of the calculated change based on a paired two-tailed Students t-test.

[21] For this study, four CAM-Oslo simulations were completed using SST, sea ice cover, well mixed greenhouse gas (GHG) concentrations and aerosol and precursor gas emissions representative of the years 1890, 1990, 2010 and 2090, respectively. 20 year average SST and sea ice climatologies were taken from the respective CMIP3 model ensembles (cf. Section 2.2) along with IPCC estimates of GHG concentrations and anthropogenic aerosol emissions. Sea salt aerosol emissions were prescribed using 20 year emission climatologies (1890, 1990, 2010 and 2090: as with the SST climatologies) calculated from the CMIP3 model ensembles and the MN03 source function (cf. Section 3.2). For all four cases, a 10 year simulation was completed with results of the last eight years of the output being used in the analysis.

3 Results

3.1 Model Ensemble U10 and SST

[22] In Figure 3, c20c3m ensemble U10 and SST averages (1980 - 1999) are compared to ERA-Interim reanalysis data [Uppala et al. 2005]. The ensemble average U10 and SST compare well with the ERA-Interim reanalysis data (within 5 % for SST and 20 % for U10 as a point-to-point comparison), with the exception of the regions of low U10 present in the reanalysis data (located at approximately 30 °N and S, and over the tropical Indian ocean and north of Australia). Individual models do simulate regions of low wind speed as in the ERA-Interim data, however in many of the models the position of these areas are displaced relative to the ERA reanalysis.

Figure 3.

c20c3m model ensemble average climatologies (1980-1999) of SST and U10 compared with ERA-40 reanalysis climatologies.

[23] Changes in sea salt number emissions due to the change in U10 and SST for a near global region (55 °S - 65 °N) are discussed below. This latitude range was chosen to exclude regions where the influence from sea ice changes on the sea salt aerosol emissions is significant. On an annual basis, the near global number emissions equal 91.5 % of the full global total integrated number emissions. In addition to the near global analysis, changes in sea salt aerosol emission on a regional scale are also investigated. The eight oceanic regions investigated here are depicted in Figure 4. Note that land points that lie within the boxes depicted in Figure 4 were excluded from the analysis.

Figure 4.

Depiction of the eight ocean sub-regions considered in the analysis.

[24] Figures 5a and 5c depict as a percentage, the c20c3m ensemble model estimated change in SST (Figure 5a) and U10 (Figure 5c). The percentage change is calculated as the difference between the 1981-2000 and the 1881-1900 averages normalised by the 1881-1900 average, with each monthly anomaly being calculated separately. Similarly, the SRES A1B ensemble results (Figures 5b and 5d) represent the difference between the 2081-2100 and 2001-2020 averages, normalised by the latter. For the SST, the models display consistent warming during the 21st and 20th centuries. The near-global average increase in SST from the late 1800 s to 2100 is 0.7% which equates to a 2.1 K absolute increase in temperature. Comparing results from the eight ocean sub-regions, it can be noted that the warming trends in the northern and tropical regions are in general stronger than for the Southern oceans.

Figure 5.

Percentage change in (a)-(b) sea surface temperature (SST) and (c)-(d) 10 metre wind speed (U10) calculated from the model ensembles. The asterix signifies to the level of statistical significance of the calculated change based on a paired two-tailed Students t-test.

[25] The centennial changes in SST and U10 were tested for statistical significance using a paired two-tailed Students t-test (visualized by asterisks in Figure 5). The positive trends in SST for the previous century and the 21st century are all significant at the 99 % confidence level for all regions. In addition, both ensembles project statistically significant increases in near global as well as southern ocean U10 (Figures 5b and 5d). The SRES A1B ensemble projects an increase in U10 for the southern oceans of approximately 2% over the 21st century (Figure 5d) which is consistent with other modelling studies [Yin, 2005]. Note however that although the same SRES A1B scenario was considered in [Yin, 2005], the model ensemble differs from the current work. Due to the highly non-linear dependency of the sea salt aerosol emissions on U10 (cf. Equation (2)), this increase is expected to strongly influence sea salt aerosol emissions in these oceanic regions. The U10 trends for northern and tropical oceans are less consistent in sign and generally exhibit a lower level of statistical significance due in part to the large inter-model differences in U10 within the CMIP3 ensembles.

3.2 Sea Salt Aerosol Number Emissions

[26] Figure 6 shows ensemble- and time-averaged (1980 - 1999) integrated sea salt aerosol number emissions for the three particle size ranges shown in Figure 1. Although the magnitude of the number emissions varies between the different aerosol size categories, the geographical distribution is similar and resembles the spatial features of the U10 fields (Figures 3c and 3d). Even so, the SST dependence is also evident in Figure 6. For example, over the tropical oceans it is clear that the relatively high emissions of coarse mode particles are caused by the warm tropical surface waters. The emission of the ultrafine particles (Figure 6a) has the opposite temperature dependence, with relatively higher number emissions at higher latitudes where the surface water generally colder.

Figure 6.

c20c3m model ensemble average climatologies (1980-1999) of the estimated sea salt aerosol emissions for the three size classes depicted in Figure 1. Note the change in scale between the different panels.

[27] Table 2 provides average sea salt aerosol number emissions (×106) per square metre per second for the c20c3m and SRES A1B model ensembles for the periods 1881-1900 and 2001-2020, respectively. The results represent integrated number emissions over the full size range of the MN03 (20 nm to 2.8 μm) along with the three size classes from Figure 1. Note that the number emissions shown in Table 2 were calculated using monthly averaged U10 and will therefore be biased low compared to emissions calculated from U10 data with higher temporal resolution (Section 2.2). The same features as shown in Figure 6 can also be discerned in the results presented in Table 2; namely the highest emissions occur at higher latitudes in the southern hemisphere and the ultrafine particles dominate the number emissions. The uncertainties listed in Table 2 represent the 1σ variations in the model emission estimates, clearly showing the large disparity in the number emissions calculated from the different CMIP3 model outputs. This large variation in modelled emissions is consistent with the large disparity in the representation of sea salt aerosol emissions from the AeroCom models [Textor et al., 2006] although it must be noted that a number of different sea salt aerosol source functions were used in the AeroCom models which almost certainly contributed to the disparity in the sea salt aerosol emission estimates in the study by Textor et al. [2006]. Liu et al. [2007] used three different meteorological data sets to drive a single global aerosol model, which allowed them to quantify the contribution of differences in meteorology to uncertainties in the simulation of sea salt aerosol and corresponding radiative impacts. Using a fixed sea salt aerosol source function, Liu et al. [2007] estimated ranges of global sea salt aerosol burden and residence times: 2.4 to 5.3 Tg and 0.33 and 0.53 days, respectively. These relatively wide ranges are estimated solely as a consequence of differences in the input meteorology.

Table 2. Average sea salt aerosol emissions (×106 particles m-2 s-1) for the nine oceanic regions. The indicated uncertainty is the one standard deviation of the model ensembles. Note that the total emissions refers to the integrated emissions from 20 nm to 2.8 μm, not the sum of the ultrafine, accumulation and coarse modes (see also Figure 1)
 1881-19002001-2020
TotalultrafineaccumulationcoarseTotalultrafineaccumulationcoarse
Near global (NG)25 ± 1416 ± 9.14.3 ± 2.30.43 ± 0.2324 ± 1516 ± 104.2 ± 2.60.43 ± 0.26
North Pacific (NP)18 ± 1013 ± 7.22.7 ± 1.60.21 ± 0.1318 ± 1113 ± 7.52.8 ± 1.70.23 ± 0.15
North Atlantic (NA)15 ± 9.711 ± 6.62.4 ± 1.60.21 ± 0.1415 ± 1111 ± 7.22.4 ± 1.70.21 ± 0.16
Tropical Pacific (TP)22 ± 1212 ± 6.84.7 ± 2.50.64 ± 0.3320 ± 1211 ± 6.94.3 ± 2.70.62 ± 0.37
Tropical Atlantic (TA)23 ± 1513 ± 8.64.7 ± 2.90.61 ± 0.3720 ± 1511 ± 8.24.2 ± 3.00.57 ± 0.40
Tropical Indian (TI)26 ± 1014 ± 5.65.4 ± 2.10.74 ± 0.2922 ± 1212 ± 6.14.8 ± 2.50.68 ± 0.35
South Pacific (SP)22 ± 1216 ± 8.93.1 ± 1.70.22 ± 0.1123 ± 1417 ± 103.4 ± 2.10.24 ± 0.14
South Atlantic (SA)44 ± 2833 ± 216.2 ± 3.80.37 ± 0.2145 ± 2933 ± 226.3 ± 4.10.40 ± 0.24
South Indian (SI)54 ± 2641 ± 207.6 ± 3.70.46 ± 0.2256 ± 3342 ± 258.0 ± 4.70.50 ± 0.28

[28] In Figure 2, the percentage changes in number emissions are displayed separately for each of the three aerosol size ranges (Figure 1). There is a consistent pattern of larger positive trends in the emissions for the coarse mode compared to the smaller sizes. This is due to the temperature dependence of the MN03 source function which shifts the size distribution of the number emissions to larger particle sizes as SST increases. For the near global (NG) results shown in Figure 2, the increase in the coarse mode particle emission over the full time period (1870-2100) is approximately 12%. However, it should be kept in mind that the coarse mode particles constitute only a small fraction (1 to 2 %) of the total number emissions (Figure 1). By contrast, the near global number emissions of the ultrafine particles do not change substantially over the same time period. This is attributed to a competition between the wind speed effect (which tends to increase the number emissions for all particle sizes) and the SST effect (which tends to decrease the ultrafine emissions). In the case of the tropical emissions, U10 does not increase significantly with time (Figure 5), which means that the ultrafine particle number emissions decrease whereas the coarse mode particles increase. The large U10 increase at high southern latitudes offsets the tendency of the ultrafine particle emissions to decrease with increasing SST. Regional trends in accumulation mode number emissions are generally small and/or not statistically significant for both centuries except for the Southern oceans (~4 % and ~10 % for the 20th and 21st centuries respectively).

[29] To provide an estimate of how the sea salt aerosol emission depends on SST, the calculations were repeated using fixed monthly SST climatologies (1881-1900 for the c20c3m calculations and 2001-2020 for the SRES A1B 21st century projections) (see Figure 7). Differences in the trends between the full MN03 and the fixed SST cases clearly follow the SST dependence in the MN03 source function (Section 2.1), i.e. ultrafine emissions tend to be lower, accumulation mode emissions almost unchanged and coarse mode emissions higher when SST changes are included in the calculations (compare Figure 2 and Figure 7). When SST is ignored in the calculation of future emissions, ultrafine particle emission trends are overestimated by up to 10 % whereas coarse mode emissions are underestimated by between 4 and 11 %. The largest impact of SST changes appears in the tropics for ultrafine particle emissions and at mid-latitudes for the coarse mode particle emissions.

Figure 7.

Percentage change in the ensemble averaged sea salt aerosol emissions for the three aerosol size classes (cf. Figure 1) calculated using fixed monthly SST climatologies (1881-1900 for the c20c3m calculations and 2001-2020 for the SRES A1B 21st century projections). The asterix signifies to the level of statistical significance of the calculated change based on a paired two-tailed Students t-test.

3.3 Discussion

[30] To have confidence in past and future projections of sea salt aerosol number emissions it is important to verify the emission calculations using real-world measurements. Unfortunately, no global measurement data-sets of sea spray emissions exist. Direct flux measurements using the eddy correlation or eddy relaxation methods are available on a campaign basis [Nilsson et al., 2010; Norris et al., 2012] however both spatial and temporal coverage are limited. Multi-annual databases of sea salt aerosol concentration measurements have been compiled [Heintzenberg et al., 2000] however this type of data is of limited value for evaluating emissions, since an atmospheric model is required for direct comparisons of concentrations. The use of a model in turn introduces a range of additional uncertainties associated with the representation of in-situ atmospheric processes [Textor et al., 2006].

[31] The increase in integrated number emissions of sea salt aerosol in a warmer climate is consistent with the previous work by Latham and Smith [1990], Penner et al. [2001] and Mahowald et al. [2006]. These earlier studies indicated that sea salt aerosol emissions constitute a negative climate feedback. The results presented here suggest that increasing SSTs have a significant effect on the size distribution of the aerosol emissions and this may in turn impact on the strength of the climate feedback (further discussed in Section 4 below).

[32] A few caveats in the calculation of the sea salt aerosol emissions should be noted:

  1. The number emissions are calculated using monthly averaged model output. Neglecting the full wind gust spectrum will result in a low bias, as the sea salt emissions are non-linearly dependent on U10 (see Equation (2)). Nevertheless, Penner et al. [2001] demonstrated that using higher temporal resolution wind fields did not alter their conclusions regarding the long-term emissions of sea salt aerosols.
  2. Changes in sea salt aerosol emissions are expected to feed back on climate change (cf. Section 4) yet these feedbacks are not included in the CMIP3 models.
  3. The stratospheric ozone boundary conditions used in the SRES A1B simulations were not applied consistently within the CMIP3 models. As discussed by Korhonen et al. [2010], the southern mid-latitude emissions of sea salt aerosols are sensitive to stratospheric ozone depletion via changes in the sub-tropical jet and so the projected changes in this region must be treated with some caution [see also Son et al., 2008].
  4. The emissions presented in Section 3.2 are calculated using a single source function (the MN03 source function). A diverse set of sea salt aerosol source functions have been proposed in the literature which contribute to the large uncertainties in the global and regional estimates of sea salt aerosol emissions [Textor et al., 2006; de Leeuw et al., 2011]. The emission changes and the sensitivity with respect to SST discussed here should therefore be regarded with some caution. Nevertheless, our results indicate that centennial changes in SST could appreciably impact global and regional sea salt aerosol emissions.

4 Implications for the Global Sea Salt Burden and Radiative Forcing

[33] Because the physico-chemical properties of freshly emitted sea salt aerosol may be altered due to in-situ atmospheric processes and the distribution of the aerosol may change with a changing climate, it is not necessarily the case that changes in aerosol emissions can be directly translated into a corresponding change in radiative forcing. Applying the sea salt aerosol emissions calculated in Section 3.2 to the CAM-Oslo global climate model, allows us to explore how the changes in sea salt aerosol emissions relate to simulated changes in atmospheric concentrations and radiative balance.

4.1 Direct Sea Salt Aerosol Effect

[34] In the following discussion regarding the direct radiative effect of sea salt aerosols, we focus on accumulation and coarse mode particles as ultrafine particles, with Dp<~ 200 nm, do not efficiently scatter solar radiation. Table 3 lists the prescribed changes in coarse and accumulation mode emissions and the simulated atmospheric burdens (i.e. vertically integrated column) from the CAM-Oslo simulations. There are clear differences between the prescribed emission changes and the corresponding atmospheric burden simulated by the model. This means by definition, that the atmospheric residence time (the total burden divided by the total emission, see Table 3) changes in response to the simulated change in climate. Note that for the fixed log-normal sea salt aerosol modes considered here, the residence time calculated using the number burden and number emissions is equivalent to the residence time based on the mass. The global, 1980-2000 average residence time calculated from the CAM-Oslo output is 45 hours and 5.8 hours for accumulation and coarse mode particles, respectively.

Table 3. Changes in number emission, atmospheric burden (number) and sea salt aerosol optical thickness (AOT) at 550 nm and total precipitation rate as simulated by CAM-Oslo. The atmospheric residence times are calculated by dividing the atmospheric burden by the emissions. Changes that are significant at the 95% and 99% level based on a two-tailed Students t-test are indicated (p>95 and p>99 respectively)
%1981 - 2000 relative to 1881 - 19002081-2100 relative to 2001 - 2020
GlobalOceans onlyGlobalOceans only
∆ # emission (coarse) 2.9 (p>0.99) 9.1 (p>0.99)
∆ atmos. # burden (coarse)1.91.64.5 (p>0.99)5.0 (p>0.99)
∆ Residence time (coarse)-1.1-1.3 (p>0.95)-3.7 (p>0.99)-4.1 (p>0.99)
∆ # emission (accumulation) 2.0 (p>0.99) 4.6 (p>0.99)
∆ atmos. # burden (accumulation)3.9 (p>0.99)3.3 (p>0.95)4.1 (p>0.99)3.3 (p>0.99)
∆ Residence time (accumulation)1.9 (p>0.95)1.3-0.35-1.1
∆ SS AOT (550 nm)3.6 (p>0.99)3.1 (p>0.99)4.2 (p>0.99)3.6 (p>0.99)
∆ Total precipitaton rate0.101.2 (p>0.99)2.9 (p>0.99)5.0 (p>0.99)

[35] CAM-Oslo simulates a decrease in the coarse mode residence time for both the 20th and 21st centuries but an increase in the accumulation mode sea salt aerosol residence time over the 20th century (statistically significant at the 95 % confidence level when averaged globally) followed by a decrease over the 21st century. The residence time is sensitive to the efficiency of removal via deposition processes which is dominated by wet removal for accumulation mode particles. Table 3 also lists the relative change in total precipitation rate from CAM-Oslo. The general increase in global and oceanic total precipitation over both the 20th and 21st centuries is in agreement with other climate models [Meehl et al., 2007b]. The response in the residence time to changes in precipitation is not straight-forward and depends on the spatial patterns of both emission and precipitation changes. For example, the increase in emissions of coarse mode aerosols is largest in the tropics where SSTs are warmest (cf. Section 2.1). However, this region is also subjected to an increase in precipitation in a warmer climate due to a strengthening of the Hadley circulation. The net result is a decrease in the residence time for coarse mode particles. For accumulation mode particles, it is more difficult to assess the reasons for the simulated change in residence times as there is no clear correlation between the spatial patterns of accumulation mode emission changes and precipitation changes.

[36] On global scales, CAM-Oslo predicts increases in accumulation and coarse mode aerosol burdens for both the 20th and 21st centuries indicating a negative climate feedback associated with the direct scattering of solar radiation. However, when extrapolating the negative trend in sea salt aerosol residence time over the 21st century it is likely that the strength of the sea salt aerosol climate feedback will diminish by 2100. In summary, since the atmospheric residence time changes in response to climate change, the predicted change in sea salt aerosol emission by itself is not a reliable indicator of the change in direct radiative forcing by sea salt aerosols.

[37] The aerosol module within CAM-Oslo was applied in an on-line configuration in the simulations completed for this study, such that the simulated aerosol fields influenced and were influenced by the modeled cloud properties and radiative fluxes (cf. Seland et al. [e.g. 2008]). This means that it was not possible to calculate the instantaneous aerosol radiative forcing as defined by the IPCC [Forster et al., 2007] from the CAM-Oslo simulation output. Instead, as a proxy for the direct radiative forcing we use the sea salt aerosol optical thickness (AOT) at 550 nm. The AOT at 550 nm is expected to be most sensitive to the atmospheric burden of accumulation mode aerosol,

[38] Globally, CAM-Oslo predicts increases in the sea salt AOT over both the 20th and 21st centuries consistent with the increase in accumulation mode atmospheric burden. A large, statistically significant increase in accumulation mode aerosol emission is predicted for the Southern oceans: ~4 % for the 20th century and ~10 % for the 21st century (Figures 2e and 2f). The average change in accumulation mode number burden for the three Southern ocean regions simulated by CAM-Oslo is 5.4 % (p>0.99) for the 20th century and 13.3 % (p>0.99) for the 21st century. The corresponding AOT changes are 5.6 % (p>0.99) and 13.8 % (p>0.99) which is consistent with the strong negative feedback in the Southern oceans found by Korhonen et al. [2010].

4.2 Indirect Sea Salt Aerosol Effect in Marine Stratocumulus Regions

[39] It is well-known that changes in the aerosol concentration in the atmosphere affect the microphysical and macro-physical properties of clouds [Lohmann and Feichter, 2005], although there are large uncertainties in the overall magnitude of these effects. Marine stratocumuli (Sc) are generally believed to be highly susceptible to changes in the atmospheric aerosol concentration (including sea salt aerosol) and these clouds also have a strong influence on local and global radiation budgets [Slingo, 1990]. Marine Sc are most commonly found in stable subsidence air masses and over the cold ocean currents formed along the eastern regions of the major ocean basins on Earth. The main marine Sc regions are defined here using the low level cloud fraction from CAM-Oslo (cloud fraction between the surface and 700 hPa assuming maximum random overlap: see Figure 8). The regions reasonably agree with satellite based estimates [Stubenrauch et al., 2008, 2010], particularly outside the polar regions. It is also noted that the location of the the marine Sc regions simulated by the model do not change significantly between the 1880, 1990, 2010 and 2090 simulations.

Figure 8.

Low cloud fraction climatology (1990) from the CAM-Oslo model. Marine stratocumulus regions are highlighted by the boxes.

[40] For marine Sc clouds, cloud condensation nuclei (CCN) activation typically occurs for particles with a diameter greater than 80 nm [Hegg et al., 1991; Glantz et al., 2003]. The simulated sea salt aerosol cloud condensation nuclei (CCN) concentration is not available as diagnostic output from the model. Instead, we use the total number concentration of accumulation and coarse mode sea salt aerosols at the top of the planetary boundary layer as a proxy. Thus, we may miss a fraction of smaller sea salt aerosol particles that can act as CCN in marine stratus regions. However, for the qualitative discussion presented here we believe this approximation is still adequate. The change in accumulation plus coarse mode number concentration, is compared with the change total CCN concentration, the low cloud fraction and total cloud liquid water path in Table 4 (Note: all quantities are averaged over the marine Sc regions indicated in Figure 8).

Table 4. Changes in sea salt aerosol number emission and concentration at the top of the planetary boundary layer along with a range of cloud diagnostics (CCN: cloud condensation nuclei, LWP: total cloud liquid water path) averaged over the marine Sc regions as shown in Figure 8. Changes that are significant at the 95% and 99% level based on a two-tailed Students t-test are indicated (p>95 and p>99 respectively)
 1881 - 19002001 - 2020
SS # conc. (accumulation + coarse) (cm-3)106120
%1981 - 2000 relative to 1881 - 19002081-2100 relative to 2001 - 2020
∆ # emission (accumulation + coarse)0.953.3 (p>0.99)
∆ SS # conc. (accumulation + coarse)2.4 (p>0.95)5.2 (p>0.99)
∆ CCN conc. (total)28 (p>0.99)0.78
∆ Low cloud fraction0.654.7 (p>0.99)
∆ LWP3.1 (p>0.99)2.7 (p>0.99)

[41] Relatively large increases in the concentration of potential sea salt CCN are simulated by the model for both for the 20th and 21st centuries (2.4 % and 5.2 % respectively). Via the first aerosol indirect effect [Twomey, 1977], this increase in CCN concentration could potentially lead to a decrease in the cloud droplet effective radius and increase in the marine Sc cloud albedo. However the basic description of the first indirect effect assumes a fixed liquid water path whereas the CAM-Oslo results suggest that along with changes in sea salt aerosol concentrations, significant changes in meteorology (which affect the liquid water path and low cloud fraction) occur with the changing climate. Note also the large increase in total CCN concentration from 1890 to 1990 (28 %) associated with large increases in anthropogenic aerosol and precursor gas emissions. Anthropogenic aerosol emissions are projected to decline over the 21st century [Lamarque et al., 2011] which is reflected in the small increase in total CCN concentration (0.78 %) despite an increase of 5.2 % in the potential sea salt CCN concentration. All in all, the model results emphasize that attempting to quantify any indirect effects on marine Sc clouds due to changes in sea salt aerosol number emissions would be complicated by associated changes in anthropogenic aerosol emissions and changes in meteorology.

5 Summary and Conclusions

[42] Global and regional sea salt aerosol number emissions have been calculated for the period 1870 to 2100 by combining ensemble model output from the CMIP3 data archive with the sea salt aerosol source function of Mårtensson et al. [2003] (MN03). Results from this study suggest that the global integrated number emission of sea salt aerosol particles has increased by between 1.7 % and 2.9 % for ultrafine and coarse mode particles, respectively, over the 20th century and will increase by between 0.3 % (ultrafine) and 8.4 % (coarse) by 2100. The general increasing trends are in agreement with previous studies [Latham and Smith, 1990; Penner et al., 2001; Mahowald et al., 2006]. However, the present study shows that the calculated changes in emission are not homogeneous over all ocean regions. Sea salt number emissions increase in the Southern oceans (between 4 and 6 % for the 20th century and between 6 to 18 % for the 21st century), a change that is driven by statistically significant increases in U10 (~1 and 2 % for the 20th and 21st centuries respectively). The calculated emission changes in the tropics and northern mid-latitude exhibit both positive and negative trends that are related to changes in U10 and the SST dependencies described in the MN03 source function.

[43] In agreement with the SST dependence of the MN03 source function, which yields a decrease in the number emissions of ultrafine particles (26 nm<Dp<76 nm) with increasing SST, the size distribution of the sea salt aerosol emissions is also projected to change. Combined with the U10 trends provided by the CMIP3 ensemble over the 21st century, decreases in ultrafine sea salt aerosol number emissions are projected to occur in the Northern Pacific and Atlantic regions and large decreases (>10 %) are predicted in the tropical oceans. Over the Southern ocean regions, the relatively large increases in U10 overrides the temperature effect meaning that ultrafine particle emissions increase during both the 20th and 21st centuries.

[44] Changes in accumulation mode (151 nm<Dp<449 nm) sea salt number emissions are relatively small in the Northern and tropical ocean regions but statistically significant increases are calculated for the Southern oceans, again driven by increases in U10. Coarse mode (655 nm<Dp<2.8 μm) number emissions increase in all ocean regions for both the 20th and 21st centuries, partly because of the positive correlation between coarse mode emissions and SST (as stated by the MN03 source function) and partly due to increasing U10. For the 21st century, large coarse mode emission increases (>10 %) are projected for the North Pacific and all of the Southern ocean regions. However, the coarse mode emissions only contribute between 1 and 2 % to the total (20 nm to 2.8 μm) integrated number emissions.

[45] The emission changes discussed in this study were calculated using only one out of a number of available sea salt aerosol source functions currently in use in global models. Previous studies show evident discrepancies in the size-resolved sea salt aerosol production fluxes calculated using these different source functions [de Leeuw et al., 2011]. Therefore, the calculated emission trends in this study should be regarded with some caution. Nevertheless, our results indicate that SST changes may be important in altering the size distribution of sea salt aerosol emissions over centennial time scales. More research is required to better quantify the sensitivity of sea salt aerosol emissions to wind speed, SST and other properties of ocean surface water (e.g. salinity, particulate and dissolved organic content) to reduce the uncertainties in centennial scale simulations of marine aerosol.

[46] Results from the CAM-Oslo GCM suggest that the residence time of sea salt aerosol in the atmosphere changes in response to anthropogenic climate change. This is primarily attributed to trends in the sea salt emission and precipitation rates over the oceans. The projected reduction in residence time over the 21st century means that the changes in atmospheric burden and sea salt aerosol AOT are lower than what would be expected considering changes in sea salt emissions alone. Based on these results we conclude that changes in sea salt aerosol number emissions are poor proxy for changes in direct radiative forcing by sea salt aerosols.

[47] CAM-Oslo predicts a global average increase in sea salt aerosol AOT of 3.6 % (p>0.99) for the 20th and 4.2 % (p>0.99) for the 21st century, suggesting a negative climate feedback associated with the direct scattering of solar radiation in agreement with previous studies [Latham and Smith, 1990; Penner et al., 2001; Mahowald et al., 2006]. However, the strength of this feedback may weaken over the 21st century due to a reduction in the residence time of sea salt aerosol. Simulated changes in the emission and concentration of sea salt aerosol within the major marine Sc regions were also investigated. It was found that any impact of changes in sea salt aerosol number emissions on the radiative properties of marine Sc clouds is likely to be difficult to separate from commensurate changes in CCN concentrations due to anthropogenic aerosol emissions and changes in meteorology.

[48] Our study clearly illustrates that fully coupled aerosol-climate modelling is required to understand the relative importance of changes in sea salt aerosol emissions on the direct scattering of solar radiation as well as the potential impacts on the radiative properties of marine Sc clouds. However, there are large differences in the description of aerosol processes in global climate models. It is important to extend this work by studying output from several models in order to reduce the uncertainties and to better quantify the strength of the potential sea salt aerosol-climate feedback.

Acknowledgments

[49] We acknowledge the modelling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) and the WCRP's Working Group on Coupled Modelling (WGCM) for their roles in making available the WCRP CMIP3 multi-model dataset. Support of this dataset is provided by the Office of Science, U.S. Department of Energy. A. Ekman would like to thank the Vinnova VINNMER program for financial support. The authors at Stockholm University wish to thank the Swedish Research Council (Project GRACE) and the Bert Bolin Centre for Climate Research (http://www.bbcc.su.se/). The research by T. Iversen, A. Kirkevåg and Ø. Seland was supported by the projects NorClim (Norwegian Research Council Grant 178246), POLARCAT (Norwegian Research Council Grant 175916) and EarthClim (207711/E10). Hamish Struthers would like to acknowledge the support of the CRAICC (Cryosphere-atmosphere interactions in a changing Arctic climate, http://www.atm.helsinki.fi/craicc/) project.

Ancillary