A convection-permitting limited area model with periodic lateral
boundary conditions and prognostic aerosol microphysics is applied to
investigate how concentrations of cloud condensation nuclei (CCN) in the
marine boundary layer are affected by high-resolution dynamical and
thermodynamic fields. The high-resolution aerosol microphysics–dynamics
model, which resolves differential particle growth and aerosol composition
across the particle size range, is applied to a domain designed to match
approximately a single grid square of a climate model. We find that, during
strongly convective conditions with high wind-speed conditions, CCN
concentrations vary by more than a factor of 8 across the domain
(5–95th percentile range), and a factor of
Aerosol particles affect the Earth's climate system directly by scattering and absorbing shortwave and long-wave radiation and indirectly by influencing the albedo and lifetime of clouds (e.g. Lohmann and Feichter, 2005). Successive IPCC climate assessment reports (e.g. Forster et al., 2007; Myhre et al., 2013) have identified the radiative forcing due to aerosol–cloud interactions as having a high level of uncertainty that needs to be better constrained for improved prediction of anthropogenic climate change.
Atmospheric aerosols, whether natural or anthropogenic, originate from two different pathways: directly emitted “primary particles” (e.g. sea spray, in marine environments) and secondary particles, which are formed by nucleation, often first requiring oxidation of gaseous precursors such as dimethyl sulfide (DMS). In general, the primary particle population can be straightforwardly classified into natural (dust, sea spray, primary biogenic) or anthropogenic (e.g. carbonaceous particles from fossil-fuel combustion sources). However, this classification is not possible for secondary particles because of the complex interactions and influences of gases with both natural and anthropogenic sources (such as sulfur dioxide) and the moderating influence of additional semi-volatile species such as ammonia and nitric acid. In the marine boundary layer however, the dominant two sources of cloud condensation nuclei (CCN) are DMS and sea spray (e.g. Raes et al., 1993; O'Dowd and de Leeuw, 2007; Boucher et al., 2013) and the relative simplicity of this particular compartment of the atmosphere allows for the systematic assessment of how two types of natural particles, primary sea spray and secondary sulfate particles from DMS, influence aerosol–cloud interactions. Carslaw et al. (2013) highlight the importance of quantifying such natural aerosols in order to accurately characterise the anthropogenic radiative forcing via aerosol–cloud interactions.
Until recently, computational costs have tended to constrain most climate models participating in international climate assessment reports to treat aerosol–cloud interactions in a simplified way, with only the mass of several aerosol types transported. With this conventional approach, CCN (number) concentrations are derived from the transported masses based on an assumed size distribution for each type, often taken to be globally uniform (e.g. Jones et al., 2011). The need to represent aerosol–cloud interactions more realistically has been a major reason for the development of a new generation of composition–climate models with interactive aerosol microphysics. The models transport both particle number concentrations and component masses (e.g. sulfate, black carbon) in multiple size classes (e.g. Mann et al., 2014), and allows for the representation of sources of primary and secondary CCN explicitly. For example, the UK's Earth System Model for CMIP6 (Coupled Model Intercomparison Project phase 6) includes the GLOMAP (Global Model of Aerosol Processes) aerosol microphysics module (Mann et al., 2010; 2012), which resolves differential particle growth and aerosol composition across the particle size range including internal mixtures via the computationally efficient modal aerosol dynamics approach.
In order to understand how aerosols and clouds interact, it is important to
assess how aerosol properties vary at finer spatial scales than are resolved
in climate models, where both convective–dynamical and aerosol microphysical
effects are likely to cause non-linear CCN variations. Whereas many
modelling studies have assessed the main features of global variations in
the aerosol particle size distribution (e.g. Ghan et al., 2001; Adams and
Seinfeld, 2002; Spracklen et al., 2005) and several have explored
aerosol–cloud interactions in regional-scale models (e.g. Bangert et al.,
2011; Zubler et al., 2011; Yang et al., 2012), only a few studies (Ekman et
al., 2004, 2006; Wang et al., 2011; Archer-Nicholls et al., 2016; Possner et
al., 2016; Weigum et al., 2016) have explored the microphysical properties
of aerosols, and their potential interactions with clouds, at resolutions of
It is known that deep convection can lead to transport of aerosols (e.g. Yin et al., 2012). In arid environments, cold-pool outflows from convection can be a major source of dust uplift, which is missed by large-scale models that parameterise moist convection (Marsham et al., 2011, 2013; Pope et al., 2016). Similarly, it has been shown that over oceans such convectively generated flows can both increase gaseous DMS emission and transport, since the convection generates locally strong winds leading to high emissions that are then preferentially transported by the convection (Devine et al., 2006). There are, however, few model studies of aerosols in ocean environments with deep convection (e.g. Cui et al., 2011) or shallow convection (e.g. Kaufman et al., 2005).
The main objective of the current study is to assess spatial and temporal
variations in aerosol properties in a convection-permitting resolution model
(grid spacing
The UKCA sub-model of the UK Met Office Unified Model (MetUM) is used (hereafter UM-UKCA), including the GLOMAP-mode aerosol microphysics scheme (Mann et al., 2010), which calculates the evolution of aerosol mass and number in several log-normal size modes. The scheme represents each size mode as an internal mixture, with several aerosol components able to be simulated including sulfate (SU), sea salt (SS), dust (DU), black carbon (BC), and particulate organic matter (POM; including primary and biogenic secondary POM). Any number of modes (with a fixed standard deviation) and possible components can be tracked, but the simulations here apply the “standard” configuration used in UM-UKCA (e.g. as in Bellouin et al., 2013) with four components (SU, SS, BC, POM) in five modes (Table 1) and dust transported separately in the existing 6-bin MetUM scheme (Woodward, 2001). The aerosol processes are simulated in a size-resolved manner and include primary emissions, secondary particle formation by binary homogeneous nucleation of sulfuric acid and water, growth by coagulation, hygroscopic growth, ageing, condensation and cloud-processing and removal by dry deposition, nucleation scavenging, impaction scavenging, and sedimentation. All the details about the description of the different aerosol processes and the size distributions in UKCA are available in Mann et al. (2010, 2012).
Standard aerosol configuration for GLOMAP-mode. The size
distribution is described by log-normal modes with varying geometric mean
diameter
The standard tropospheric chemistry configuration of UM-UKCA is used
(O'Connor et al., 2014), which includes O
The simulations are carried out with UM-UKCA applied in a high-resolution
limited area model with periodic lateral boundary conditions, specifically
applying the numerical weather prediction (NWP) configuration of MetUM GA4.0
(Walters et al., 2014). MetUM GA4.0 provides tracer transport,
boundary-layer mixing, and large-scale cloud and precipitation, with UKCA
simulating atmospheric chemistry and aerosol processes. The limited area
domain is centred close to the Equator (1.32
Emissions of DMS and sea spray are interactive in the model, with their flux
into the atmosphere primarily driven by variations in the model wind speed
(using the same approaches described in Bellouin et al., 2013).
Anthropogenic emissions of SO Global Emissions InitiAtive: Global Fire Emissions Database, Version2:
Tephigram representing the vertical profile of the initial dew point temperature (dashed line) and the temperature (solid line). The thick dotted line represents the adiabatic parcel ascent and the circles indicate the specific levels of the parcel such as the lifted condensation level (LCL), the convective condensation level (CCL) and the level of free convection (LFC).
The thermodynamic (temperature and humidity) and dynamic (horizontal wind)
variables are initialised from a single model profile (Fig. 1) taken from
a global aqua-planet configuration of a MetUM operational run (where all
land points are removed). The profiles are deliberately chosen to be
strongly unstable so that the model will experience a sudden deep convective
instability in the early phase of its evolution. The convective perturbation
can clearly be seen in Fig. 2, with deep convective clouds forming after a
few hours, reaching up to a cloud top height of approximately 18 km (Fig. 2a).
The precipitation onset is after
Temporal evolution of
Snapshot spatial variations in the number concentrations
To aid interpretation and inference from the assessment of aerosol
properties in subsequent sections, in this first part of the results we
explore how the substantial emission of DMS during the intense storm period
propagates through to simulated concentrations of its oxidised forms, sulfur
dioxide (SO
Temporal evolution of the mean mass mixing ratios of the
gas precursors to aerosols. The DMS, SO
Hereafter, the analysis focuses on separately assessing the aerosol particles in the different size modes, investigating how the identified driver sources and processes are influencing simulated CCN variations at this convection permitting resolution. The analysis is restricted to the last 12 h of simulation with an emphasis on the results obtained after 18 h of integration, by which time the model has fully spun-up. Indeed, according to the extreme convective instability that induces intense updrafts the spin-up time lasts approximately 6 h of simulation.
In this section, the focus is on quantifying variations in aerosol
properties in the three different particle size ranges: Aitken, accumulation,
and coarse modes. The analysis begins (Fig. 3) with instantaneous
snapshots of surface aerosol particle concentration and size at two
different times in the simulation. Figure 3a–c present a snapshot of
spatial variability at 6 h of integration, when a dynamics intense storm
period occurred. Figure 3d–i show the snapshot spatial variation at 18 h
of integration, in more modest and representative wind-speed conditions
but with intense rain rates. The coarse mode consists entirely of sea-spray
particles, so the highest particle concentrations are expected to generally
indicate regions where simulated horizontal wind speeds are highest.
However, during the initial storm period, and at this high spatial
resolution, there are also regions of intense localised precipitation
(greater than 10 mm h
Temporal evolution of the aerosol concentration from the
Aitken
Despite the fact that particles in Aitken mode can be affected by the
emission of sea spray (e.g. Salter et al., 2015), in this remote marine
domain, particles in the Aitken mode are almost exclusively secondary in
nature, being originally formed via nucleation in the free troposphere. Over
the initial 12 h of integration, free troposphere concentrations of the driver gas for
nucleation, H
Normalised probability density function (PDF) of the
geometric radius
In Fig. 5 we show Hovmöller diagrams to further explore the temporal
evolution in surface concentrations of Aitken-, accumulation-, and coarse-mode particles during the last 12 h of integration (at
Assessing how each of the size modes spins up reveals how temporal variations in marine CCN concentrations are actually reflecting the very different time profiles of the two dominant CCN production pathways: primary emissions of sea-spray particles and entrainment of DMS-derived secondary particles formed in the free troposphere. The analysis illustrates the way a diverse community of processes (dynamical, chemical, and microphysical) together determine CCN variations in the marine boundary layer. Figure 5a shows an Aitken mode emerging after 17 h of integration which also explains the dip in accumulation-mode size (contour lines), as a substantial number of smaller secondary particles are “mode-merged in” from the Aitken mode at that time. For the coarse mode, as particle concentrations decrease, there is also a progression to smaller particles, which can be explained by that fact that, in the model, sedimentation (the dominant removal process for this mode) removes both number and mass, enabling the simulation to reflect the fact that larger particles fall faster even when they are in the same mode.
Time series of the mean
A more quantitative analysis of the simulated aerosol properties is presented hereafter, with Fig. 6 showing probability density functions (PDFs) of the geometric dry radius (a–c) and particle concentrations (d–f) for the Aitken, accumulation, and coarse modes at different times in the second 12 h of the integration. The analysis shows that, for the accumulation and coarse modes, as seen in Fig. 5, as time progresses, the particle size PDFs shift to smaller sizes, with the accumulation-mode PDFs becoming much wider in the evening as the source of smaller particles from the Aitken mode becomes significant. By contrast, as Aitken-mode concentrations increase, the particles are clearly also larger, reflecting that growth processes are acting on the particles, with this size increase ceasing at about 18 h of integration, while particle concentrations continue to increase (likely due to entrainment). For the accumulation- and coarse-mode particles, this quantitative approach is consistent with sedimentation causing the shift in size distribution as the larger particles sediment out faster than the smaller ones. Figure 7 shows the temporal evolution of the mean and standard deviation of the geometric mean radius values and number concentration (over grid boxes in the domain) for Aitken, accumulation, and coarse modes at the surface. The accumulation- and coarse-mode concentration and radius fields have largest spatial variations between 5 and 8 h of simulation as the model adjusts to the very strong sea-salt emission and quite efficient wet removal induced by the precipitation onset during the peak convective activity, whereas Aitken-mode concentrations, and their variations, stay approximately constant through that period. During the simulations, the mean radius and particle concentration values from the coarse mode, on average, decrease (Fig. 7c, f), but the mean size variations show the opposite evolution, with greater variability in the calmer second half of the day, reflecting the strengthening influence of sedimentation as sea-spray emissions decrease. For the accumulation mode, the mean particle size displays remarkably little variation over the domain between 9 and 14 h of simulation (as seen in Fig. 6a), with the variation increasing as the source of secondary CCN from the Aitken mode becomes significant later in the day.
In this marine domain, sea-salt particles represent a major component of the
CCN population (e.g. O'Dowd et al., 1997; O'Dowd and de Leeuw, 2007). Models
parameterise sea-spray emission fluxes as a function of the 10 m wind speed
(
2D distribution of the surface sea-salt emission flux
In light of the inference of sea-spray emissions fluxes from measurements of
particle concentrations, Fig. 8 presents several snapshot variation
box plots for simulated sea-salt emission flux, sea-salt mass mixing ratio
(MMR), and the CCN number concentration as a function of the
Altitude-dependent probability density function (a-PDF)
in percent of the CCN concentration at different integration times. The
a-PDFs are obtained calculating the PDF for each different level. A
resolution of 0.1 is used for quantify the logarithm of the concentration.
The lines represent the mean (solid lines)
Figure 9 presents the vertical variation of the simulated CCN concentration
using an altitude PDF profile (a-PDF) for the same periods as mentioned
above. As expected, on average, the CCN concentration drops off with
increasing altitude reflecting a balance between turbulence and convection
lifting the particles vertically and gravitational settling transporting
larger particles back towards the surface. In the atmospheric surface layer
(lowest 100 m or so) the profile of mean CCN follows a power-law
profile but the spatial CCN variance (standard deviation over grid boxes in
the domain) decreases much less rapidly with altitude. As a consequence, the
CV increases with increasing altitude from
We have analysed spatial and temporal sea-spray and CCN variations in a convection-permitting model with interactive sea-spray emissions, sulfur chemistry and aerosol microphysics over an idealised marine tropical domain. In this marine atmosphere the two dominant CCN sources are both natural: primary sea-spray particles and secondary sulfate particles. However, even in this relatively simple two-component CCN system, our analysis has revealed that there is a diverse community of processes: dynamical, chemical, and microphysical, that combine to determine the number of particles which can activate to cloud droplets.
First, the dynamics strongly influences the sea-spray emissions since highest particle concentrations occur where wind speeds are highest, and there is a cubic wind-speed dependence for sea-salt emission. The emitted sea-spray particles have a range of sizes, being directly emitted in both the accumulation (sub-micron) and coarse (super-micron) modes. After their emission into the atmosphere, sea-salt aerosols are transported vertically by turbulent diffusion and convective updrafts, with larger particles also being influenced by sedimentation. We show that the covariation of sea-salt mass mixing ratio with wind speed is fundamentally different than that for sea-salt emission, with implications for derivations that treat the two synonymously. In particular, since sub-micron sea-spray has much longer atmospheric residence time (days) than super-micron sea-spray (hours), care must be taken when relating measured sea-spray concentrations to emissions. Intense localised precipitation during strong convection also impacts aerosol concentrations at the climate grid scale with removal effects introducing strong variations (e.g. via the impaction scavenging process). The combination of these processes impacts the particle concentration properties, which become extremely variable in space (about a factor of 8 over the entire domain, one climate model grid square) and time. We acknowledge that if the aerosol had been initialised with a background profile, the variability described here might have been lower.
Moreover, the emissions of DMS strongly vary spatially and temporally
according to wind speeds and become substantial during intense storm period
(as in Devine et al., 2006). There is a requirement for gas-phase species
SO
Sea-spray particles are highly soluble and, in most cases, are directly emitted at sizes where they are already effective CCN. In contrast, a different component of the CCN population comprises nucleated sulfate particles which need more time to grow large enough to be CCN-active. The variations in the CCN concentrations are strong and can attain a factor of 8 in strongly convective conditions, mostly reflecting the properties of larger CCN. Smaller (sub-micron) CCN, from the accumulation mode, tend to have less variation, which in part is due to their source having a significant contribution from the steady formation of secondary sulfate particles in the free troposphere. We have seen how dynamics and microphysical processes also affect CCN, in particular with a 2nd CCN peak at the top of the boundary layer during the strongly convective period before the secondary particles emerged. These effects combine to determine how the CV in CCN concentration changes with altitude, our results suggesting an increase from around 10 % at the surface to more than 20 % at the top of the marine boundary layer. Whereas CCN concentration fields close to the surface are mainly influenced by the emissions, at higher altitudes they are in general older, and inherit influences propagated via transport.
We also examine spatial and temporal variations in aerosol particle size, finding that the geometric radius of the Aitken and coarse modes are particularly variable, which will introduce further variability in cloud droplet number concentrations and cloud brightness. The different influences on the two CCN types (primary and secondary), and the diverse community of processes involved (microphysical, chemical and dynamical), makes sub-grid parameterisation of the CCN variations difficult. This study provides valuable results on the impact of the local dynamics and aerosol sources on the CCN population and on the aerosol–cloud interactions occurring at these fine spatial scales. We have applied the UM-UKCA model for non-idealised case studies with a nesting procedure to retain the larger-scale influences, as is the capability to allow these aerosol variations to couple with a new cloud microphysics scheme in MetUM (Shipway and Hill, 2012).
The authors declare that they have no conflict of interest.
This project was made possible through the financial support of the Leeds-Met Office Academic Partnership (ASCI project). The authors acknowledge use of the MONSooN system, a collaborative facility supplied under the Joint Weather and Climate Research Programme, which is a strategic partnership between the Met Office and the Natural Environment Research Council. The lead author wishes to thank Douglas Parker for useful discussions. Graham W. Mann is funded by the UK National Centre for Atmospheric Science, one of the UK Natural Environment Research Council (NERC) research centres. John H. Marsham acknowledges funding from the SAMMBA project (NE/J009822/1). Edited by: C. Hoose Reviewed by: two anonymous referees