Access to this full-text is provided by Wiley.
Content available from Journal of Geophysical Research: Atmospheres
This content is subject to copyright. Terms and conditions apply.
Aitken Mode Aerosols Buffer Decoupled Mid‐Latitude
Boundary Layer Clouds Against Precipitation Depletion
Isabel L. McCoy
1,2,3,4
, Matthew C. Wyant
1
, Peter N. Blossey
1
,
Christopher S. Bretherton
5
, and Robert Wood
1
1
Department of Atmospheric Sciences, University of Washington, Seattle, WA, USA,
2
Cooperative Programs for the
Advancement of Earth System Science, University Corporation for Atmospheric Research, Boulder, CO, USA,
3
Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, CO, USA,
4
Chemical
Sciences Laboratory, National Oceanic and Atmospheric Administration, Boulder, CO, USA,
5
Allen Institute for Artificial
Intelligence, Seattle, WA, USA
Abstract Aerosol‐cloud‐precipitation interactions are a leading source of uncertainty in estimating climate
sensitivity. Remote marine boundary layers where accumulation mode (∼100–400 nm diameter) aerosol
concentrations are relatively low are very susceptible to aerosol changes. These regions also experience
heightened Aitken mode aerosol (∼10–100 nm) concentrations associated with ocean biology. Aitken aerosols
may significantly influence cloud properties and evolution by replenishing cloud condensation nuclei and
droplet number lost through precipitation (i.e., Aitken buffering). We use a large‐eddy simulation with an
Aitken‐mode enabled microphysics scheme to examine the role of Aitken buffering in a mid‐latitude decoupled
boundary layer cloud regime observed on 15 July 2017 during the Aerosol and Cloud Experiments in the Eastern
North Atlantic flight campaign: cumulus rising into stratocumulus under elevated Aitken concentrations (∼100–
200 mg
1
). In situ measurements are used to constrain and evaluate this case study. Our simulation accurately
captures observed aerosol‐cloud‐precipitation interactions and reveals time‐evolving processes driving regime
development and evolution. Aitken activation into the accumulation mode in the cumulus layer provides a
reservoir for turbulence and convection to carry accumulation aerosols into the drizzling stratocumulus layer
above. Further Aitken activation occurs aloft in the stratocumulus layer. Together, these activation events buffer
this cloud regime against precipitation removal, reducing cloud break‐up and associated increases in
heterogeneity. We examine cloud evolution sensitivity to initial aerosol conditions. With halved accumulation
number, Aitken aerosols restore accumulation concentrations, maintain droplet number similar to original
values, and prevent cloud break‐up. Without Aitken aerosols, precipitation‐driven cloud break‐up occurs
rapidly. In this regime, Aitken buffering sustains brighter, more homogeneous clouds for longer.
Plain Language Summary Aerosols, small particles in the atmosphere associated with ocean
biology, sea spray, land, and human‐produced emissions, influence cloud brightness and, by suppressing
precipitation and subsequent break up, cloud lifetime. Understanding aerosol‐cloud‐precipitation interactions is
critical in understanding how aerosols influence the climate system. This study examines how the very smallest
aerosol particles modify cloud formation, brightness, and lifetime over the North Atlantic ocean. We utilize a
recent set of aircraft and satellite observations from a dedicated field campaign as well as a detailed model that
resolves fine‐scale interactions important to cloud development. After comparing the model to real‐world
observations, we test how modifying the amount of small particles impacts the cloud brightness and lifetime.
We find that the small particles are able to offset precipitation removal of larger particles, helping clouds to last
longer and stay brighter.
1. Introduction
Recently, liquid cloud aerosol‐cloud interactions (aci) have been identified as a key, remaining source of un-
certainty in accurately estimating climate sensitivity (Bellouin et al., 2020). Aci impacts the climate system in two
ways (Bellouin et al., 2020; Boucher et al., 2013; Christensen et al., 2022; Wood, 2012). The first is through
radiative forcing (RF
aci
), which manifests as a change in cloud droplet number concentration (N
d
) in response to a
change in aerosol while other macrophysical characteristics (e.g., liquid water content, LWC) remain constant
(Twomey, 1977): increasing aerosol amount leads to increasing N
d
and an increase in the fraction of shortwave
reflected back to space (i.e., albedo) associated with the accompanying reduction in surface area per droplet. The
RESEARCH ARTICLE
10.1029/2023JD039572
Key Points:
•Observations of mid‐latitude decou-
pled low clouds constrain a large‐eddy
simulation investigating aerosol‐cloud‐
precipitation interactions
•Boundary layer Aitken activation and
turbulent and convective fluxes restore
accumulation mode aerosols against
precipitation losses
•Aitken buffering acts to sustain
brighter, more homogeneous clouds for
longer
Supporting Information:
Supporting Information may be found in
the online version of this article.
Correspondence to:
I. L. McCoy,
isabel.mccoy@noaa.gov
Citation:
McCoy, I. L., Wyant, M. C., Blossey, P.
N., Bretherton, C. S., & Wood, R. (2024).
Aitken mode aerosols buffer decoupled
mid‐latitude boundary layer clouds against
precipitation depletion. Journal of
Geophysical Research: Atmospheres,129,
e2023JD039572. https://doi.org/10.1029/
2023JD039572
Received 5 JULY 2023
Accepted 10 JUN 2024
Author Contributions:
Conceptualization: Isabel L. McCoy,
Matthew C. Wyant, Peter N. Blossey,
Christopher S. Bretherton, Robert Wood
Data curation: Isabel L. McCoy, Matthew
C. Wyant, Peter N. Blossey
Formal analysis: Isabel L. McCoy,
Matthew C. Wyant, Peter N. Blossey
Funding acquisition: Isabel L. McCoy,
Peter N. Blossey, Christopher
S. Bretherton, Robert Wood
Investigation: Isabel L. McCoy, Matthew
C. Wyant, Peter N. Blossey, Christopher
S. Bretherton, Robert Wood
© 2024. The Author(s).
This is an open access article under the
terms of the Creative Commons
Attribution‐NonCommercial‐NoDerivs
License, which permits use and
distribution in any medium, provided the
original work is properly cited, the use is
non‐commercial and no modifications or
adaptations are made.
MCCOY ET AL. 1 of 26
second is through cloud adjustments, which manifest as a change in cloud macrophysical characteristics (e.g.,
cloud liquid, amount, thickness, etc.) through changes in cloud microphysics (e.g., precipitation, evaporation,
etc.) (e.g., Ackerman et al., 2004; B. A. Albrecht, 1989; Bretherton et al., 2007; S. Wang et al., 2003; Xue &
Feingold, 2006). These combined effects in response to a change in aerosol (i.e., from the pre‐industrial aerosol
state to the present day) are known as the effective radiative forcing (ERF
aci
).
Global climate models (GCMs) have particular difficulty in capturing aci in cloud regimes that are biologically
active with little anthropogenic influence (e.g., Carslaw et al., 2013; McCoy et al., 2020). Some of this is likely
due to incomplete representation of Aitken aerosol production and its contribution to aci (Gordon et al., 2017;
McCoy et al., 2021). Aitken aerosols (∼10–100 nm in diameter) form through various processes including gas to
particle conversion from ocean biology emissions (Seinfeld & Pandis, 2016) which can occur at cloud edges (e.g.,
Clarke et al., 1998; Kazil et al., 2011), continental anthropogenic emissions (e.g., Twohy et al., 2002), and, in
recent studies, from sea spray production (Lawler et al., 2021; Xu et al., 2022). They have been observed in high
concentrations in the free troposphere (FT) intermittently across the globe (Williamson et al., 2019). In the
boundary layer (BL), where they are sometimes generated (Zheng et al., 2021), Aitken particles act as a key
source of accumulation mode aerosol (∼100–400 nm) (e.g., Covert et al., 1996; Sanchez et al., 2018; Zheng
et al., 2018). Accumulation mode aerosols are then activated in moist updrafts into cloud condensation nuclei
(CCN). Enhanced supersaturation (Kaufman & Tanré, 1994) and updraft strength (Pöhlker et al., 2021),
particularly in the absence of accumulation mode particles, can facilitate activation of smaller, Aitken particles
into CCN as well (Fan et al., 2018).
Aitken mode‐aerosols may have an additional role to play in aerosol‐cloud‐precipitation interactions. Drawing on
Southern Ocean observations, McCoy et al. (2021) recently hypothesized that Aitken particles buffer precipi-
tating BL clouds against cloud droplet depletion: as precipitation removes accumulation mode aerosol, peak
supersaturation increases in updrafts, and larger aerosols in the Aitken mode are able to activate into CCN with the
potential to grow larger, restoring N
d
. This buffering mechanism is consistent with the idea that changes in cloud‐
active aerosol can be partially compensated when changes in aerosol composition and size distributions lead to
increased supersaturation and thus increased activation of smaller condensation nuclei (e.g., “microphysical
buffering,” Stevens & Feingold, 2009; Twomey, 1959). During the biologically‐active Southern Ocean Austral
summer, Aitken aerosol are plentiful both in a substantial FT reservoir developed through synoptic‐scale uplift
and in the BL as a result of synoptic‐scale descent (Covert et al., 1996; McCoy et al., 2021). Southern Ocean
clouds have been observed to have many fewer optically‐thin cloud features than in similar clouds observed in the
Northeast Pacific stratocumulus (Sc) to cumulus (Cu) transitions that were experiencing similar accumulation
mode concentrations but lower Aitken mode concentrations (McCoy et al., 2021; O, Wood, & Tseng, 2018). In
the sub‐tropics, these features are generated in association with precipitation‐driven depletion of the cloud droplet
and accumulation mode aerosol populations (O, Wood, & Bretherton, 2018; O, Wood, & Tseng, 2018; Wood
et al., 2018). Less frequent occurrence of optically‐thin cloud features in the Southern Ocean is thus consistent
with a damping of precipitation processes by Aitken‐buffering.
Recent large‐eddy simulation (LES) and observational studies have found Aitken aerosols impact cloud
microphysical and radiative properties in pristine environments (Pöhlker et al., 2021; Wyant et al., 2022),
although their influence is modulated by cloud phase (Bulatovic et al., 2021). In particular, Wyant et al. (2022,
hereafter W22) developed an Aitken‐mode enabled microphysics scheme that predicts time evolution of aerosol‐
cloud‐precipitation interactions by including aerosol sinks and sources (albeit neglecting new particle formation).
W22 utilized an idealized Southeast Pacific case study of deep, precipitating Sc informed by in situ observations
to directly evaluate the Aitken‐buffering hypothesis. They simulated this case over several days, finding a gradual
loss of accumulation mode aerosol to drizzle formation led to a transition to an ultra‐clean, low cloud fraction, and
strongly precipitating Cu state. This transition could be delayed by increasing Aitken concentrations above the
inversion or through fluxes from the surface.
The Aitken‐buffering mechanism, which has both observational (McCoy et al., 2021) and modeling (Wyant
et al., 2022) support, has important implications for our understanding of aci as well as past and future climates.
Konsta et al. (2022) recently found that the “too few, too bright” bias in GCMs has persisted in many state‐of‐the‐
art models largely due to GCMs' difficulty in capturing the heterogeneity of clouds at lower cloud fractions.
Specifically, GCMs fail to represent the wide‐spread occurrence of optically‐thin cloud features (Konsta
et al., 2022) that occur across a variety of mesoscale cloud morphology patterns (Leahy et al., 2012; McCoy,
Methodology: Isabel L. McCoy, Matthew
C. Wyant, Peter N. Blossey, Christopher
S. Bretherton, Robert Wood
Project administration: Isabel L. McCoy,
Peter N. Blossey, Christopher
S. Bretherton, Robert Wood
Resources: Isabel L. McCoy, Peter
N. Blossey, Christopher S. Bretherton,
Robert Wood
Software: Isabel L. McCoy, Matthew
C. Wyant, Peter N. Blossey
Supervision: Isabel L. McCoy, Peter
N. Blossey, Christopher S. Bretherton,
Robert Wood
Validation: Isabel L. McCoy, Matthew
C. Wyant, Peter N. Blossey, Robert Wood
Visualization: Isabel L. McCoy, Matthew
C. Wyant, Peter N. Blossey
Writing – original draft: Isabel
L. McCoy
Writing – review & editing: Isabel
L. McCoy, Matthew C. Wyant, Peter
N. Blossey, Christopher S. Bretherton,
Robert Wood
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 2 of 26
McCoy, et al., 2023; Mieslinger et al., 2021; O, Wood, & Tseng, 2018) and may depend in part on the absence of
Aitken aerosols (McCoy et al., 2021). Variations in optically‐thin cloud amount across morphology patterns
contributes to differences in their cloud radiative impact and how we expect them to feed back on the climate
system under climate change (McCoy, McCoy, et al., 2023). Incomplete representation of Aitken aerosol pro-
cesses in GCMs may also influence our estimation of RF
aci
and therefore ERF
aci
as Aitken aerosols may play a
critical role in regulating N
d
in pristine, pre‐industrial environments (Gordon et al., 2016,2017; McCoy
et al., 2020). Thus, identifying the key processes involved in aerosol‐cloud‐precipitation interactions driven by
Aitken aerosols and understanding their nuances has utility in improving both our knowledge of the climate
system and the representation of cloud‐aerosol interactions in models used for climate prediction.
In this study, we build on the work of W22 by utilizing their Aitken‐enabled microphysics scheme in large eddy
simulations (Section 2.2) to examine the influence of Aitken aerosols on an observationally‐constrained case
study sampled during the recent Aerosol and Cloud Experiments in the Eastern North Atlantic (ACE‐ENA) flight
campaign in the Northeast Atlantic (J. Wang et al., 2022). Specifically, we examine a case of Cu rising into Sc
under substantial Aitken aerosol concentrations that was sampled by aircraft on 15 July 2017. We extend W22 by
using these in situ observations (Section 2.1) to constrain the LES control simulation (Section 3.1). Successful
simulation of this case allows us to identify the key processes involved in the evolution of clouds in such a regime
(Section 3.2). Aerosol sensitivity studies are conducted (Section 4) to examine the dependence on initial aerosol
state and subsequent nuances of rapid aerosol processing, changes in cloud microphysics, radiative properties,
and heterogeneity (as measured by the development of optically‐thin cloud features). We especially focus on the
influence of Aitken aerosols on cloud properties under this meteorologically‐forced regime. We conclude with a
discussion (Section 5) and summary (Section 6).
2. Data and Methods
2.1. Observations for the ACE‐ENA Case Study
In situ observations from the 15 July 2017 flight (Figure 1) during the summer phase of the ACE‐ENA campaign
(J. Wang et al., 2022) form the basis for our LES case study. This research flight by the Department of Energy G‐1
aircraft (hereafter RF16 of the campaign) sampled a system of Cu (bases at ∼500 m) rising into Sc (∼1,000–
1,500 m) to the northwest of Graciosa Island (Figure 1). This system gradually advected to the southwest over the
day (e.g., Figure 2b in J. Wang et al. (2022)). The G‐1 aircraft utilized a Lagrangian‐drift sampling pattern
consisting of multiple stacked level legs ∼60 km in length. Each leg followed a straight, crosswind line at altitudes
set to sample above, in, and below cloud and ended in a vertical ascent profile to the next level leg altitude
(Figure 1b). ERA5 reanalysis extracted for the ACE‐ENA campaign region show that, over the course of the day,
the atmosphere experienced increasing large‐scale uplift (Figure S3a in Supporting Information S1) and an
associated cooling and moistening by large‐scale vertical advection (Figures S3b and S3c in Supporting Infor-
mation S1). Mesoscale moisture convergence (e.g., Bretherton & Blossey, 2017) can be encouraged by large‐
scale uplift (e.g., as seen in trade‐wind clouds, Narenpitak et al., 2021), and may contribute to the deepening
and moistening of clouds observed in this case.
The G‐1 aircraft was outfitted with a suite of instruments, a subset of which we utilize to both develop and
compare with our LES case study. The Fast Integrated Mobility Spectrometer (FIMS, J. Wang et al., 2017) and the
Passive Cavity Aerosol Spectrometer Probe (PCASP) provide size distributions and number concentrations for
Aitken (∼10–100 nm) and accumulation (∼100–400 nm) mode size ranges, respectively. The FIMS resolves the
full 10–400 nm size range while the PCASP resolves the larger, accumulation sizes only (e.g., Figure 4). Total
aerosol number concentrations for this study are calculated as the sum of the Aitken and accumulation number
concentrations from these specified size ranges, which is found to be similar to the observations from the
Condensation Particle Counter (CPC, sizes ≥10 nm, not shown). The Fast Cloud Droplet Probe (FCDP) is used
for cloud N
d
and LWC. Precipitation flux, which includes cloud droplet sedimentation, is calculated from droplet
spectra measurements assuming terminal fall‐speeds from Rogers and Yau (1989). Spectra are based on two
instruments that optimally sample different drop size ranges (results are not sensitive to the diameter cutoff): the
FCDP (selecting diameters ≤50 μm) and the Two‐Dimensional Stereo Particle Imaging Probe (2DS, diameters
≥50 μm).
For the model‐observation comparison, we focus on the second half of the flight period (12:00–14:30 UTC). Note
that local solar time is ∼2 hr behind UTC. This portion of the flight was cloud‐rich and generally moister than the
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 3 of 26
first, drier half (see blue outlines, Figure 1b). Aitken and accumulation aerosol size ranges are simultaneously
sampled more consistently in the second half as well (shown separately in Figure S1 of Supporting Information S1
and as a sum, when both are sampled, in Figure 1b). Aerosol comparison levels are selected to be further from
clouds (where observations are sparse). The profile at ∼12:15 p.m. (profile 2, P2) sampled the depth of the BL and
is used as the initial LES aerosol profile (star in Figure 1and Figure S1 in Supporting Information S1, discussed
further in Section 2.2).
We also compare our results against cloud liquid water path (LWP), cloud optical depth (τ), and broadband albedo
retrievals for the ACE‐ENA campaign (ARM Data Center, 2017) from the NASA SATCORPS (Satellite Cloud
Observations and Radiative Property retrieval System) product which applies the VISST (Visible Infrared Solar‐
infrared Split‐Window Technique) algorithm to Meteosat‐10 satellite channels (Minnis et al., 2001,2008,2011).
The broadband albedo retrieval product includes a correction based on converting to shortwave flux and
regionally (5° ×5°) normalizing to Edition 4 of the CERES (Clouds and the Earth's Radiant Energy System) Aqua
SSF1deg product for the corresponding month. We utilize these satellite products to provide further insight into
the aerosol‐cloud‐precipitation system sampled by the aircraft. Thus, model‐satellite comparisons are also
restricted to 12:00–14:30 UTC. LES output is coarsened to the SATCORPS temporal (0.5 hr) and spatial (∼3 km)
resolutions. In order to capture a representative sample of this case's cloud heterogeneity while restricted to the
coarser satellite resolution, we use a 2° ×2° domain overlapping the flight region (27°–29°W and 39–41°N,
Figure 1. (a) MODIS Aqua visual imagery on 15 July 2017 at 14:30 UTC or 13:30 local time with the RF16 flight path
colored by time. (b) Flight altitude versus UTC time (gray) with color overlay of observations (where available) of total
aerosol from simultaneously sampled N
ait
+N
acc
or, in cloud, N
d
. In cloud sampling, where liquid water content
≥0.01 g kg
1
, is outlined in blue. The dark gray background from 12:00 to 14:30 UTC is the observational comparison period
used in model evaluation. Separately plotted N
ait
and N
acc
versions are shown in Figure S1 of Supporting Information S1. The
profile used for initializing aerosol is marked with a star in (a), (b) and outlined in white in (b).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 4 of 26
Figure 1a and Figure S2 in Supporting Information S1). We sub‐sample this into 49 sub‐domains of comparable
area to the LES simulation domain (∼0.25° ×0.25°, see Figure S2b in Supporting Information S1 for an
example). This is a sufficient sample size to facilitate statistical comparisons between the LES and the various
cloud system realizations captured by the satellite subdomains.
2.2. Aitken‐Aerosol‐Enabled Large‐Eddy Simulations
We utilize W22's novel two‐mode aerosol microphysics scheme for the System for Atmospheric Modeling
(SAM) LES simulations (25.6 ×25.6 km
2
domain with 100 m resolution). This Hoppel Transfer scheme extends
the single‐mode, two‐moment prognostic aerosol scheme of Berner et al. (2013) by including Aitken aerosol
evolution and a simple representation of sulfur chemistry. Seven prognostic variables represent accumulation and
Aitken log‐normal aerosol modes in air and droplets as well as three gas species (H
2
SO
4
, SO
2
, and DMS).
Scavenging of interstitial and other unactivated aerosol by cloud and rain drops are treated as in Berner
et al. (2013), while coagulation of unactivated aerosols follow Binkowski and Shankar (1995). A simplified
scheme for capturing basic influences of sulfur chemistry on model aerosols is also included, but new particle
formation (e.g., aerosols nucleating from gas‐phase H
2
SO
4
) is neglected for simplicity (unlike in Kazil
et al. (2011)). The only sources of Aitken aerosols considered in the scheme are from surface fluxes and
entrainment from the FT. Two aerosol modes are used to approximately capture the Aitken (∼10–100 nm) and
accumulation (∼100–400 nm) modes, though it should be noted that the characteristic modal diameter of each
aerosol mode can evolve in response to aerosol and chemical processes.
The premise of the W22 Hoppel Transfer scheme is to allow activation of Aitken mode particles in saturated
updrafts so that they can act as CCN in the model. When — during activation — the number of Aitken particles at
the critical diameter exceeds the number of accumulation mode particles, aerosols are shifted from the Aitken to
the accumulation mode to enforce equality between the Aitken and accumulation mode concentrations at the
critical diameter. Conceptually, this should place the Hoppel minimum (D
Hoppel
) at the critical diameter (D
c
) in
strong updrafts. In weak updrafts, where D
c
is larger than D
Hoppel
, no Aitken particles are moved into the
accumulation mode. For simplicity, we assume that all cloud droplets are associated with an accumulation mode
aerosol, so the “Aitken” mode is composed of unactivated aerosols. Supersaturation, which helps to determine D
c
,
is diagnostic and computed within the Morrison microphysics scheme (Morrison et al., 2005; Wyant et al., 2022).
Typical supersaturation values experienced by CCN upon activation (i.e., mean supersaturation weighted by local
activation rate) range across the BL from 0.13% to 0.26% for accumulation mode aerosols and 0.23%–0.35% for
Aitken mode aerosols transferred into the accumulation mode, comparable to the range of maximum supersat-
urations (0.1%–0.3%) observed in this region during the summer of 2017 (Gong et al., 2023). The typical updraft
strengths during activation range from ∼0.2 to 0.6 m s
1
for accumulation and ∼0.5–1.0 m s
1
for Aitken mode
aerosols.
As in W22, all aerosols are assumed soluble with the dry density and hygroscopicity of ammonium sulfate (i.e.,
B=0.51 in Eq. 3 of Abdul‐Razzak and Ghan (2000)). This is broadly consistent with the ACE‐ENA time‐of‐
flight aerosol mass spectrometer observations of bulk non‐refractory aerosol composition showing the sum-
mertime marine BL was dominated by sulfate mass (J. Wang et al., 2022). We additionally note that in the W22
formulation, transferred Aitken mode aerosols are instantaneously shifted to larger sizes (due to refitting the log‐
normal accumulation mode after transfer). While the instantaneous growth of the transferred aerosols is unre-
alistic, one would expect such particles to grow over time due to aqueous sulfate deposition (Kaufman &
Tanré, 1994). Unlike other schemes that explicitly capture sulfate growth through allowing Aitken sized particles
to grow into cloud and return to Aitken mode sizes (e.g., Feingold et al., 1996; Ivanova & Leighton, 2008a;
Ivanova & Leighton, 2008b), for computational efficiency we approximate Aitken activation as a transfer to the
accumulation mode as soon as activation occurs. Following Kaufman and Tanré (1994), it is likely sulfate uptake
depends on cloud droplet size (not on embedded aerosol size) and initially smaller aerosols will experience greater
growth than larger aerosols. Due to our bulk treatment of aerosol chemistry and use of a fixed shape parameter,
differential growth of smaller aerosols is not possible. However, this growth effect appears in the transfer albeit at
a shorter timescale and for somewhat artificial reasons.
A key distinction between this study and the more idealized W22 study is that our LES case is more tightly
constrained by in situ observations in an effort to simulate aerosol‐cloud‐precipitation interactions in a context as
similar to the real world as possible. Initial thermodynamic profiles of temperature and moisture are developed
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 5 of 26
from a combination of the Graciosa Island soundings, ERA5 reanalysis soundings extracted to the campaign
region (Figure S3 in Supporting Information S1), and in situ flight profiles. As initial simulations produced
thinner‐than‐observed clouds, the moisture profiles were slightly enhanced to better correspond with the second,
comparison portion of the observations (12:00–14:30 UTC) (Figure 1b; Figures S1 and S5b in Supporting In-
formation S1). The initial Aitken and accumulation mode aerosol number and mass mixing ratios (Figure S4a,
S4b, S5c, and S5d in Supporting Information S1) follow the P2 reference profile from RF16 (star in Figure 1).
Modal Aitken and accumulation widths (as defined by geometric standard deviation, σ
ait
=1.3 and σ
acc
=1.4 μm)
and initial diameters (Table S1 in Supporting Information S1) are selected to correspond to case observations
(Figure S4c in Supporting Information S1). While the characteristic diameter of each aerosol mode may evolve,
the modal widths are fixed in time. The initial SAM modes and the observed size distributions for P2 are shown in
Figure S6 of Supporting Information S1, initial values are detailed in Table S1, Figures S5c, and S5d in Sup-
porting Information S1.N
d
is initialized at 35 mg
1
based on the median in situ observations for the upper cloud
layer (Figure S4d in Supporting Information S1).
Simulations are initialized with a small, random moisture and temperature perturbation and run for 12 hr to allow
the development of mesoscale variability. During this period, the domain‐mean profiles of temperature, specific
humidity, aerosol number and mass mixing ratios below cloud top (∼1,175 m) are nudged to the previously
discussed, initial profiles (Figure S5 in Supporting Information S1) that capture key elements of the RF16
environment. A 10 min nudging timescale is used for all nudging above cloud top. Below cloud top, aerosols are
nudged with a 10 min timescale, while the moisture and temperature fields are nudged on a longer 1 hr timescale
to allow for the development of mesoscale organization. An exception is made below 500 m where the inverse
nudging timescale applied to temperature and moisture decreases gradually to zero at the surface. Weaker
nudging in the subcloud layer allows surface‐flux‐driven turbulence and convection to develop during the spinup
period. Afterward, nudging within the BL and the inversion layer is switched off so that the simulations are
released to run freely at 9:00 UTC and throughout the remaining 12 hr duration of the simulation (ending at 21:00
UTC). Following Blossey et al. (2021), after release each simulation is forced by the large‐scale vertical velocity
as well as moisture and temperature tendencies from ERA5 to maintain meteorology at real world conditions
throughout the simulation, while nudging to the initial profiles only in the FT at a timescale of 30 min starting
500 m above the inversion. Although aerosols are affected by large‐scale vertical motion, no large‐scale hori-
zontal advective tendencies are applied to the aerosol, so that, after a simulation is released, the aerosol evolves as
a net balance between sources and sinks as in, for example, Wood (2006).
Surface fluxes of momentum, heat and moisture are computed interactively using bulk schemes in SAM. Surface
aerosol fluxes follow the approach in W22, which depend on the 10 m wind speed (u
10
∼2 m s
1
, Figure S3d in
Supporting Information S1) raised to the 3.41 power and are thus very small in this case. A damping layer is
applied above 2 km altitude, and the model top is set at 2.9 km. Radiative heating is computed by RRTMG
(Mlawer et al., 1997) over the full atmospheric column by joining the model profiles with those from the forcing
data set above the mode. While the case study is Eulerian, the domain is translated with a fixed velocity (u,
v= 2 m s
1
) to minimize cross‐grid flow (Wyant et al., 2018).
For model‐observation comparisons, SAM aerosol number concentrations are calculated as in Zender (2001)
using aerosol size distributions truncated to specific instrument observation size ranges for Aitken (10–100 nm),
accumulation (100–400 nm), and total (combined Aitken and accumulation ranges, 10–400 nm) aerosol. Where
necessary, SAM profiles compared to observations are subset to in‐cloud (LWC ≥0.01 g kg
1
) and out‐of‐cloud
(<0.01 g kg
1
) samples (e.g., observed aerosol concentrations are only reported out‐of‐cloud while droplet
number concentrations are only reported in‐cloud). All size distributions from SAM are computed for the
combined in‐ and out‐of‐cloud aerosol across the xydomain for each time and height level. For comparisons
with observed size distributions, we have selected relatively cloud‐free altitudes (i.e., the lower BL at 300 m, the
transition layer between Cu and Sc cloud layers at 700 m, and the FT at 1.6 km). When comparing across
sensitivity studies, distributions at altitudes dominated by cloud (i.e., 0.5, 1, and 1.4 km) and aerosol budgets are
also included in order to directly examine aerosol‐cloud processing. Precipitation fluxes are calculated as the
integral of sedimentation fluxes over cloud and rain droplet sizes, equivalent to observations.
The evolution of aerosol‐cloud precipitation interactions are examined using number and mass budgets for Aitken
and accumulation modes over several atmospheric layers. These budgets are formulated following W22. The
accumulation mode in this context is composed of unactivated accumulation, in‐cloud droplet, and in‐rain
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 6 of 26
aerosols. Thus, activation and droplet evaporation (which leaves behind unactivated accumulation mode aerosols)
do not have a net impact on the budget. For each aerosol category, the number tendencies evolve following a rate
equation:
˙
NTot.=
˙
NAit.Trans.+
˙
NTop Flux +
˙
NBot.Flux +
˙
NWet Scav.+
˙
NScav.
+
˙
NCoag.+
˙
NLarge‐Scale Vert.Mot.+
˙
NSed.+
˙
NNudge.+
˙
NRes.
(1)
This can be further simplified as:
˙
NTot.=
˙
NAit.Trans.+
˙
NTop Flux +
˙
NBot.Flux +
˙
NWet Scav.+
˙
NOther +
˙
NRes.(2)
The leading terms are activation or transfer of Aitken aerosol into the accumulation mode (Aitken Transfer),
movement of aerosol through turbulent fluxes (Top Flux and Bottom Flux relative to the layer the budget is
computed over), and removal of aerosol through autoconversion, accretion, and limiters (as in Berner et al., 2013,
Wet Scavenging). Tendency terms with small contributions are gathered for analysis purposes into the Other term.
These are scavenging (in‐cloud removal of interstitial and unactivated aerosol), coagulation (removal of aerosol
through coalescence or aggregation of aerosols via Brownian motion), aerosol transport by large‐scale vertical
motion, sedimentation of aerosols out of the atmosphere, and nudging tendencies applied during the spin‐up phase
of the model (before 9:00 UTC). The residual captures the remaining behavior of the total aerosol tendencies and,
when small, indicates that these equations capture the majority of the aerosol behavior. Note that the meaning of
the turbulent fluxes changes depending on the layer they are computed over (i.e., surface source, exchange be-
tween layers). The mass budgets have a similar formulation with an additional term for chemistry (particle growth
through chemical processing):
˙
MTot.=
˙
MChem.+
˙
MAit.Trans.+
˙
MTop Flux +
˙
MBot.Flux +
˙
MWet Scav.
+
˙
MScav.+
˙
MCoag.+
˙
MLarge‐Scale Vert.Mot.+
˙
MSed.+
˙
MNudge.+
˙
MRes.
(3)
Time evolution for all number and mass budget terms are shown in the supplement (Figures S7 and S8 in
Supporting Information S1, respectively).
Aerosol sensitivity studies, described in Section 4, adjust the initial number concentration profiles. In each case,
corresponding changes are made to the initial mass profiles so that the initial diameter and width of modes are
identical across all simulations (Table S1 in Supporting Information S1). These changes to the initial aerosol
profiles include halving the accumulation number while leaving the Aitken mode unchanged (HfAc), eliminating
the Aitken mode while leaving the accumulation mode unchanged (NoAit), and halving the accumulation mode
number while eliminating the Aitken mode (HfAcNoAit). To avoid computational issues, when Aitken aerosol is
removed in the NoAit and HfAcNoAit simulations, Aitken number and mass are set to small, non‐zero values. In
the HfAc and HfAcNoAit simulations, both accumulation mass and number are halved relative to the vertically
resolved Ctrl initial profile.
Three additional sensitivity studies are conducted that do not alter the initial aerosol profiles but instead test the
sensitivity to other aspects of the simulation set up. These are examined in Appendix A.Ctrl LD examines the
dependence of the Ctrl simulation on domain by doubling its size. Additionally, we conduct two studies,
AltHopV1 and AltHopV2, to examine the sensitivity of the Ctrl to variations in the W22 Hoppel Transfer scheme.
3. Simulating the RF16 ACE‐ENA Case Study
We first present the general behavior of the standard SAM simulation for RF16 (hereafter Ctrl, Section 2.2).
Figure 2shows the evolution of aerosol and cloud droplet number along with the corresponding changes to
horizontal variations in τ. Consistent with observations, two cloud layers form: an upper, Sc layer between ∼1–
1.5 km altitude and a shallow, Cu layer near the surface (∼400–600 m). N
ait
remains high in the FT and is initially
(∼10:00 UTC) large throughout the BL (Figure 2a). N
acc
has the opposite structure (d). Over time, the Cu in the
lower layer intensifies and drives local changes in aerosol size distributions (N
ait
reduces while N
acc
increases, b
and e), increases in N
d
(h) as larger cumuli connect (e.g., 12:00, b, e, h) and rise into the upper Sc layer (e.g., 14:00
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 7 of 26
UTC, c, f, i). The upper layer deepens with time as Sc clouds grow larger and are fed by the Cu below. Sc cells
become more well defined (e.g., 14:00 UTC) with increased core N
d
(h, i) and τ(k, l). Large‐scale uplift (Figure
S3a in Supporting Information S1) may influence this cloud evolution as well. The Ctrl simulation captures the
stratified aerosol vertical distribution evident in the RF16 observations. N
ait
is largest in the FT (Figures S1a and
S4a in Supporting Information S1) and contributes the most to the total aerosol magnitude (Figure 1b), signifi-
cantly exceeding N
acc
at most heights in the marine BL (Figures S1b and S4a in Supporting Information S1) as
will be discussed further in Section 3.1. Our simulations will facilitate further examination of essential aerosol‐
cloud‐precipitation processes at work in this decoupled low cloud regime (Section 3.2).
3.1. Observational Evaluation
Interrogating the Ctrl simulation with observations (Section 2.1 and 2.2) informs us of the capabilities and
limitations of our case study and model. Skill in reproducing the net behavior sampled during RF16 will give us
confidence in our ability to capture the complex interplay of aerosol‐cloud‐precipitation processes driving the
cloud system evolution in this regime.
Our first evaluation utilizes vertical profiles of several key quantities observed from the G‐1 aircraft over the
comparison period (12:00–14:30 UTC, Figure 3). The Ctrl median and interquartile range are compared with the
observed median and interquartile range in 10 altitude bins (which are used to account for differences in aircraft
sampling frequency across the BL). Generally, median Ctrl profiles fall within the interquartile range of observed
profiles, demonstrating the skill of our simulation. Because the simulated aerosol state, which has very good
agreement with observed out‐of‐cloud aerosol (Figures 3a–3c) is critical in our study, it is worth examining in
more detail.
Figure 2. Vertical cross sections of SAM Ctrl simulation for Aitken (a–c), accumulation (d–f), and cloud droplet (g–i) number concentrations at (a, d, g) 10:00, (b, e, h)
12:00, and (c, f, i) 14:00 UTC. Corresponding cloud optical depth spatial x‐ysnapshots for (j) 10:00, (k) 12:00, and (l) 14:00 UTC. The black lines in (j–l) shows the
location of vertical cross sections shown in (a–i).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 8 of 26
Total aerosol number concentrations (Figure 3c), the sum of Aitken and accumulation number (Figures 3a and
3b), coincide with observations across all altitudes and have one of the best overall agreements of all the variables
examined. When aerosol number is separated into its individual modes, good agreement is still found across most
altitudes for accumulation and across all altitudes for Aitken concentrations. Opposing deviations away from
observed behaviors are seen in the Aitken and accumulation size ranges in the BL near ∼500 m in the Cu layer:
simulated Aitken aerosol number (Figure 3a) is slightly depleted but still within the interquartile range while
accumulation number (Figure 3b) is augmented relative to observations. In addition to their slight deviation near
the Cu (∼0–500 m), accumulation concentrations deviate by a factor of ∼2×in the Sc layer (∼1–1.5 km, e.g.,
Figures 2e and 2f), falling outside of the observed interquartile ranges in both cloud layers. Note that there are
large observational uncertainties in the upper cloud layer as observations at this altitude are impacted by limited
opportunities for in situ aerosol sampling in cloud‐free air.
Even though the Cu and Sc cloud layers occur in both observations and SAM (e.g., Figure 1b, sparse quantiles
near ∼500 m and numerous quantiles between 1 and 1.5 km, Figures 3d and 3e), the modal aerosol partitioning
does not exactly match. The slight depletion of Aitken aerosol (Figures 2a–2c) and the accompanying increase in
accumulation mode aerosol (Figures 2d–2f) indicates sufficient updraft strength and supersaturation occurs in the
nearby, thin Cu layer to enable Aitken activation through the W22 Hoppel Transfer scheme. This can also be seen
in the evolution of the simulated aerosol size distribution at 300 m from its initial shape to its median behavior
over the observation‐comparison period (Figure 4c). The characteristic modal diameter for the Aitken mode
Figure 3. Model‐observation comparison of median vertical profiles for select variables over 12:00–14:30 UTC: number
concentrations of (a) Aitken mode, (b) accumulation mode, (c) total aerosol (the sum of accumulation and Aitken aerosol
modes), and (e) cloud droplets; and (d) cloud liquid water content (LWC). Aerosol comparisons are computed only for out‐
of‐cloud samples while cloud LWC and droplets are only for in‐cloud samples. Observations (red) are binned into 10
quantiles by altitude and shown as a median (dashed line with dots) and an interquartile range (shading with horizontal lines)
for each bin. SAM Ctrl (purple) is similarly shown as a median (line) with interquartile range (shading). Initial estimates are
included for the observations (10‐s running mean for profile 2, dashed gray) and SAM (simulation profile after spinup period
at 9:00 UTC as in Figure S5 of Supporting Information S1, solid gray). The cloud layer edges in SAM, where median cloud
fraction is 1%, are included for reference as dark gray, dashed horizontal lines in (d) and (e).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 9 of 26
moves toward the characteristic modal diameter for the accumulation mode,
mainly due to chemical processing, with the Hoppel Transfer reducing overall
Aitken aerosol number and the depth of the Hoppel minimum.
Aitken transfer also occurs in the Sc layer, as will be examined later in more
detail (e.g., Section 3.2), and contributes to the increase in accumulation
number in this layer. Comparing with the size distributions available in the
transition layer below the Sc (700 m, Figure 4b), it is apparent that the
accumulation mode shifts further to the right than observed. This suggests that
generation of accumulation mode aerosols in the lower BL also contributes to
the large accumulation concentrations in the Sc layer. Note that the observed
and simulated distributions in the FT are assisted in their overlap by nudging
above the inversion (Figure 4a). While some activation likely occurs in both
the Sc and Cu layers, the magnitude of the deviation from observations
suggests the simulated transfer is slightly too efficient. This may, in part,
reflect the complexity of initializing a rapidly evolving BL (Section 5).
The majority of the median aerosol size distributions over the observation
comparison period are within the observed interquartile range. The relative
changes in modal location in the layers below the Cu and Sc indicate distinct
underlying causes for these shifts away from the initial distributions, which
may be exaggerated by the overly active transfer. Processes driving these
disparate shifts will be examined in Section 3.2. Overall, however, simulated
aerosol partitioning agrees very well with observations across the sampled
depth (particularly for Aitken aerosols, Figures 3a and 3b), suggesting these
discrepancies do not negatively impact the total BL behavior.
Ctrl cloud microphysical properties are also in good agreement with obser-
vations. In‐cloud LWC (Figure 3d) is within the observed interquartile range
across the BL, with a slightly dryer Cu layer. N
d
(Figure 3e) also agrees well
with observations. Note that the <1% cloud fraction portion of the profiles
above ∼1.5 km are from the overshooting Cu tops above the Sc layer (e.g.,
Figures 2g–2i). There is a significant difference in the lowest observed
quantile near ∼500 m. This may be associated with sparse sampling of the Cu
clouds in this layer, as evident from the large observational uncertainties
(Figures 3d and 3e). Overall, however, the good agreement in N
d
across the
majority of the BL suggests that the net balance between the evolving aerosol
sources and sinks generated in Ctrl is realistic.
We can assess the simulated aerosol sinks in more detail using vertical pre-
cipitation profile comparisons (Figure 5, Section 2.2). More heavily precip-
itating rain events are defined where precipitation exceeds a conditional threshold (P≥0.1 mm day
1
), allowing
us to examine in‐rain precipitation flux (a), rain occurrence fraction (b), and total precipitation flux (c) separately.
Cloud drop sedimentation is also included in the simulated and observed precipitation rate estimates. Ctrl tends to
produce a smaller amount of rain but is still within the interquartile range (a). It tends to rain over a similar portion
of the BL as observed (b). Total precipitation flux is also within the observed interquartile range throughout the
BL (c). Overshooting Cu tops are apparent in the in‐rain precipitation flux (a) but not in the total (c) or rain
fraction (b), indicating their contribution is very small. In both Ctrl and observations, meaningful precipitation
peaks at similar heights.
The separation between the mean and median estimates of both in‐rain (a) and total precipitation flux (c) suggests
that Ctrl simulates a more consistently drizzling cloud system than the infrequent but heavily raining system
observed during RF16. The mean and median are significantly separated in observations: RF16 sampled a few
heavily precipitating clouds (∼30 mm day
1
) but fewer lightly precipitating clouds (a, c). Ctrl has slightly closer
mean and median behaviors and a generally lower mean than observed: there are fewer heavily precipitating
clouds produced compared to observations but more frequent drizzling clouds producing a median within the
observed interquartile range without pushing the mean toward the higher, observed values (a, c).
Figure 4. Model‐observation comparison of median aerosol size
distributions over 12:00–14:30 UTC at three levels: (a) 1,600 m, (b) 700 m,
and (c) 300 m. Observations within 100 m of the labeled SAM altitude level
are included. SAM Ctrl size distributions (purple) for the initial (dashed,
Section 2.2) and comparison period median (solid) are shown. Inclusion of
the initial size distribution illustrates evolution of aerosols over the
intervening time. Observations are included from two instruments on the G‐
1: the FIMS (orange) and the PCASP (pink) which, respectfully, resolve the
majority of the Aitken and accumulation mode sizes (Section 2.1). In situ
values are shown as median (solid) and interquartile range (shading) over the
comparison period and can be contrasted with the comparable SAM median
(solid).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 10 of 26
Finally, we can evaluate the Ctrl radiative properties over the comparison period using satellite observations. In
order to capture the spread in cloud behavior across this case when using the coarser satellite data, we utilize 49
samples of comparable size to the simulation domain (0.25° ×0.25°) sampled within a 2° ×2° box overlapping
the flight sampling region (Section 2.1). Subdomain means and PDFs indicate the spread in behavior seen in all
three satellite retrievals for this region and time (Figure 6). After coarsening the SAM output to the satellite
resolution (Figure S2 in Supporting Information S1), we see that the Ctrl simulation (purple) tends toward the
upper end of the observed behavior for LWP and albedo but matches the mean τbehavior.
Because Ctrl N
d
corresponds well with in situ observations (Figure 3e), we expected this good agreement in the
optical properties. However, this comparison also demonstrates the contribution of column liquid water to overall
radiative properties. Despite simulating similar τas observed, Ctrl tends to simulate moister clouds (Figure 6a)
and a slightly higher area‐mean albedo (c). The tendency toward moister clouds is consistent with the higher LWC
(Figure 3d) and higher N
d
(e) at the top of the Sc layer and in the overshooting Cu, which was above the aircraft‐
sampled flight level.
Is the Ctrl simulation statistically likely to fall within observed subdomain variability? To test this, we use a PDF
constructed from the subdomain means (red) and apply Welch's unequal variances t‐test to compare the popu-
lation means of the satellite samples and the simulation. In all cases, the Ctrl mean is the same as the satellite
subdomain aggregate mean at 95% confidence. The Ctrl mean (circle) falls within the 5th–95th range from the
Figure 5. Model‐observation vertical profile comparisons for precipitation measures over 12:00–14:30 UTC. Total profiles
are computed (as in Figure 3) for: (a) in‐rain precipitation flux, (b) rain fraction, and (c) total precipitation flux. Rain is
determined based on the conditional threshold, P≥0.1 mm day
1
. In addition to median and interquartile range, means are
also shown (diamonds and dotted line for observations, thin line for SAM). The cloud layer edges in SAM, where median
cloud fraction is 1%, are included for reference as dark gray, dashed horizontal lines.
Figure 6. Model‐observation comparison for satellite observations over 12:00–14:30 UTC: (a) cloud liquid water path,
(b) cloud optical depth, and (c) albedo. PDFs for each satellite subdomain (thin lines, colored by subdomain mean value) are
shown along with their means (colored circles). A PDF constructed from the subdomain means (red) is shown along with
descriptive statistics: mean (dot), median (diamond), standard deviation (thick line), and 5th–95th range (thin line). A PDF
comparable to the subdomain PDFs is constructed from the Ctrl simulation coarsened to the satellite resolution (purple, thin
dashed line). Model statistics for comparison to the observations are also included. See Section 2.1 for further sampling
details. In (b), the optically thin cloud threshold (τ=3) is shown as a gray line (O, Wood, & Tseng, 2018).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 11 of 26
satellite subdomain aggregate (thin line). Thus, the Ctrl simulation is consistent with satellite observed cloud
property variability despite a small tendency toward moister and brighter observed cloud behaviors.
We conclude that the W22 configuration of SAM with this specified initialization method captures the majority of
the key features of the decoupled low cloud regime sampled in RF16. Ctrl exhibits skill in generating and
maintaining aerosol across the FT and the majority of the BL in both number and size distributions. The main
exception is near cloud layers where there are small discrepancies in aerosol partitioning between modes due to
Aitken particles being too readily activated and increasing the accumulation mode beyond the observed range.
Ctrl also tends to produce a cloud with slightly higher liquid water amount and fewer heavily precipitating clouds
than observed. Some of the forcings for this case study (i.e., imposed large scale uplift) and necessary initiali-
zation choices (i.e., BL moistening of reanalysis to resemble observations) may encourage this macro‐physical
response. However, neither the small differences from observations in aerosol nor cloud appear to negatively
effect N
d
or the net balance of aerosol sources and sinks. Ctrl produces τand domain‐mean albedo within sta-
tistical agreement of satellite observed ranges, suggesting SAM has sufficient skill to accurately analyze radiative
property sensitivity to aerosol changes. The small differences between observed and Ctrl behaviors will be
revisited in Section 5. However, the fidelity of the Ctrl simulation in capturing aerosol and aerosol‐cloud‐
precipitation interactions is sufficiently robust to justify further analysis: (a) identifying key aerosol‐cloud‐
precipitation processes in this morphology regime (Section 3.2) and (b) evaluating regime sensitivity to
changes in aerosol conditions (Section 4).
3.2. Identifying Key Aerosol‐Cloud‐Precipitation Processes
Satisfied with the agreement between observations and our constrained aerosol‐coupled LES for this case, we can
examine the time evolution of aerosol‐cloud‐precipitation interactions and identify which processes dominate the
behavior of this decoupled low cloud regime. Figure 7shows the vertical, time evolving profile of N
acc
(a) along
with the Aitken transfer tendency (b), updraft strength (i.e., vertical velocity variance, c), and N
ait
(d). After
release at 9:00 UTC, Aitken aerosol is transferred gradually to the accumulation mode in many small Cu‐layer
updrafts between ∼250–500 m, depleting the initial Aitken aerosol. Sub‐cloud layer convection combines with
large‐scale advective forcing to erode a weak stable layer (evident in Figure S5 of Supporting Information S1) and
enables enhanced shallow moist convection and an associated increase in Aitken transfer to accumulation mode
after 11:00 UTC (Figures 7b and 7c). As seen in Figure 2, these Cu contribute to transport and mixing between the
lower and upper cloud and aerosol layers. After 14:30 UTC, Aitken activation and transfer occurs mostly in the Sc
layer updrafts, which become more robust by the end of the simulation (after 16:30 UTC) and appear to be driven
by increasing cloud top radiative cooling in the evening hours (not shown). However, N
acc
does not increase
simultaneously, suggesting that this N
ait
activation is buffering aerosol and droplet number concentrations against
precipitation depletion.
To delve further into aerosol processes affecting these two cloud layers and their interchanges, we calculate three
atmospheric layer number budgets (Section 2.2, Figure S7 in Supporting Information S1) examining: (a) the total
depth, including the BL and lower FT (0–1.6 km, Figure 8a), (b) the upper layer (0.8–1.6 km, Figure 8b), and (c)
the lower layer (0–0.8 km, Figure 8c) tendencies. A corresponding mass budget (Figure S8 in Supporting In-
formation S1) is also computed and will be discussed where relevant. Figure 8presents the mean tendencies of the
leading terms (Equation 2) contributing to the Aitken and accumulation number evolution. To aid in interpretation,
we focus on the mean tendencies over three reference periods (highlighted in Figure 7d): after release (9:00–
12:00), during the observation comparison (12:00–14:30), and when accumulation sources and sinks are in quasi‐
balance (14:30–21:00 UTC). We also include a summary schematic illustrating these key processes (Figure 9).
Aitken transfer dominates the number tendencies in the total budget for both the Aitken (as a sink) and accu-
mulation modes (as a source). Aitken transfer peaks during the observational comparison period and then de-
creases over time, with the fewest particles transferred to the accumulation mode during the final period. The
removal of accumulation mode aerosol through wet scavenging mirrors these tendencies but with a smaller
magnitude. The result is a relatively constant, positive total accumulation number tendency across the BL and
throughout the simulation due to the Aitken transfer out competing the wet scavenging accumulation number
tendencies (
˙
NTot.acc >0). Aitken number tends to be lost throughout the simulation and BL (
˙
NTot.ait <0), peaking
in loss the same period that there is the largest transfer (12:00–14:30 UTC).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 12 of 26
In all layers, the mass budget (Figure S8 in Supporting Information S1) is dominated by accumulation mode
tendencies. Once activated, chemical processes quickly grow solute mass and diameter, committing the particles
irreversibly to the accumulation mode and likely continuing to increase their size over time (e.g., Feingold &
Kreidenweis, 2002) with potential assistance from collision‐coalescence (e.g., Hoffmann & Feingold, 2023).
Mass increases through chemical processing are partly offset by sedimentation removal. Mass is gained
throughout the first two periods before the sources and sinks begin to balance during the final period.
Dividing the total budget into layers encapsulating the Cu (lower) and Sc (upper) clouds adds additional nuance to
this story. Turbulent and convective fluxes act to mix and redistribute particles between layers. The net flux of
Aitken aerosols is downward, from the upper layer (bottom flux) into the lower layer (top flux) partly due to Cu
updrafts carrying anomalously small Aitken aerosol concentrations (Figures 2a–2c). Accumulation aerosols are
simultaneously fluxed in the opposite direction: exported from the lower layer into the upper layer. Flux ten-
dencies peak in both layers during the 12:00–14:30 UTC period.
Figure 7. Time versus altitude profiles showing the Ctrl evolution of (a) accumulation number concentration (in and out of
cloud), (b) Aitken transfer rate, (c) vertical velocity variance, and (d) Aitken number concentration. Contours of 0.05, 0.1,
and 0.2 g kg
1
liquid water (thin white lines) and 10% cloud cover (thick purple line) are included for reference. Three time
periods are marked in (d) for future reference: 9:00–12:00 (pink), 12:00–14:30 (red), and 14:30–21:00 UTC (dark red).
Vertical velocity variance <0.003 m
2
s
2
and Aitken transfer <1 mg
1
day
1
are not shown.
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 13 of 26
Aitken transfer dominates the lower layer, leading to a negative
˙
NTot.ait and positive
˙
NTot.acc at all times despite
exchanges with the upper layer. Both
˙
NTot.acc and
˙
NTot.ait magnitudes decrease with time. This is due to the
combined effect of fluxes peaking at 12:00–14:30 and decreasing Aitken transfer tendency during 14:30–21:00
UTC, reducing the amount of aerosols being both generated and brought into the lower layer.
In the upper layer,
˙
NTot.acc and
˙
NTot.ait have the same signs as in the lower layer but evolve very differently.
˙
NTot.acc increases with time while
˙
NTot.ait has the largest loss at 12:00–14:30 UTC. Wet scavenging is a significant
sink of accumulation aerosols in this layer, peaking in the 12:00–14:30 period but unable to overcome the sources
of accumulation aerosols.
˙
NTot.ait peaks in the 12:00–14:30 period, an apparent lag from the lower layer ten-
dencies. This lag between layers is likely due to the delay while Aitken aerosols are transferred in the lower layer
and fluxed upwards. It may also be related to entrainment of Aitken aerosols as the BL deepens over time. Local
Aitken transfer rates also increase with time in the upper layer, assisting in resisting the precipitation depletion
effects on accumulation aerosols.
Note that in‐cloud scavenging weakly reduces Aitken in the upper layer (9:00–14:30) and, in the final period,
large‐scale subsidence of Aitken from the FT weakly increases Aitken number in the upper layer (Figure S7c in
Figure 8. Evolution of leading number budget terms for Aitken (hatched) and accumulation modes (solid bars) computed
over the total depth (a, 0–1.6 km), over the upper layer (b, 0.8–1.6 km), and over the lower layer (c, 0–0.8 km). Mean number
tendencies are computed for the three time periods (left to right in each term category) highlighted in Figure 7d: 9:00–12:00
(pink), 12:00–14:30 (red), and 14:30–21:00 UTC (dark red). Bars show Ctrl mean tendencies while dots show equivalent
values for the sensitivity studies (Section 4). Total tendency is to the left of the gray division line and contributions from
individual terms are to the right. Companion plots for all number and mass tendency terms versus time for the Ctrl and
aerosol sensitivity simulations are in Figures S7 and S8 of Supporting Information S1.
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 14 of 26
Supporting Information S1). A longer entrainment period or a larger FT
Aitken concentration would be necessary to restore BL Aitken aerosol (see
W22 sensitivity studies, discussion in Section 5). Weak winds lead to
insignificant surface aerosol fluxes (Section 2.2, Figures 8a and 8c). In this
case, Aitken is lost at all levels and times (
˙
NTot.ait <0).
To summarize (Figure 9), in this decoupled low cloud regime Aitken mode
aerosol is activated into the accumulation mode and grown through chemical
processing in both Cu and Sc cloud layers. Turbulent and convective eddies
mix Aitken aerosols down from the upper layer into the lower layer where
they are transferred to the accumulation mode. Simultaneously, eddies export
accumulation particles up into the transition and Sc layers where they are
activated into droplets in updrafts. Precipitation depletion through wet
scavenging removes accumulation aerosol in Sc (and weakly in Cu), partially
off‐setting the increase in BL accumulation particles from Aitken transfer.
This resistance to precipitation loss and the accompanying maintenance of N
d
associated with Aitken activation into the precipitation‐depleted accumula-
tion mode is a hallmark of the Aitken‐buffering mechanism (McCoy
et al., 2021).
4. Sensitivity Studies
In this section, we build on the Ctrl simulation with three additional simu-
lations that examine the sensitivity of the RF16 cloud system and its aerosol‐
cloud‐precipitation processes to changes in accumulation and Aitken aerosol
number concentrations (Section 2.2, Table S1 in Supporting Information S1). HfAc reduces the initial Ctrl
accumulation number by half throughout the entire profile. It asks whether the amount of Aitken aerosol in Ctrl
can still buffer the aerosol‐cloud‐precipitation system against precipitation depletion in a reduced accumulation‐
mode environment. NoAit removes Aitken aerosol throughout the entire initial profile and tests whether Aitken
aerosol is important to sustaining the accumulation profile against precipitation depletion. HfAcNoAit uses the
accumulation profile of HfAc and the Aitken profile of NoAit to evaluate whether the aerosol and cloud profiles
can be sustained against precipitation removal in the reduced accumulation case without the help of Aitken
aerosol.
Differences between these simulations are immediately apparent from the time series of certain key parameters
(Figure 10). Two types of behavior are encapsulated by these simulations: Aitken‐buffered (Ctrl,HfAc) and
Aitken‐deficient (NoAit,HfAcNoAit) systems. When Aitken aerosols are present, the Aitken‐buffering mecha-
nism helps clouds to maintain coverage (a) despite depletion of LWP (c) through persistent precipitation (b).
Aitken aerosol is steadily lost over time in Ctrl and HfAc, mostly through transfer to the accumulation mode,
whose concentration increases steadily throughout these two simulations (e) (from ∼50 to ∼70 mg
1
in Ctrl, and
∼25–55 mg
1
in HfAc). In contrast, the Aitken‐deficient simulations (NoAit,HfAcNoAit) have significantly
different cloud fraction evolution, beginning to break up at ∼12:00 and 10:00 UTC respectively (a). They steadily
lose accumulation aerosols (e) due to precipitation depletion (b, particularly strong in HfAcNoAit) that is un-
compensated by Aitken transfer. The trend in accumulation mode for all simulations is reflected by N
d
in the
upper Sc and lower Cu cloud layers (Figure 10f).
Snapshots at 14:00 UTC (Figure 11) highlight the differences in aerosol and cloud morphology across the
simulations. Compared to Ctrl, all studies exhibit more separated upper level mesoscale cells generated from Cu
rising into Sc with a larger proportion of intervening optically thin cloud layers. HfAc maintains cells very similar
to those in Ctrl, albeit with smaller convective structures and slightly lower τacross the domain. Clouds in the
Aitken‐deficient cases are much more heterogeneous than in the Aitken‐buffered cases. NoAit has much smaller
and more broken cells with a much reduced cloud cover and lower τ. The Sc in HfAcNoAit has already collapsed
by 14:00 UTC, leaving the BL dominated by broken cumuliform structures with a few optically thin layers
remaining from precipitation‐depleted clouds.
Figure 9. Key aerosol‐cloud‐precipitation processes involved in the
evolution of the morphology regime observed during the RF16 research
flight on 15 July 2017. The key terms shown are Aitken activation/transfer
(occurring at all cloud layers), turbulent fluxes (eddies moving particles
between layers), wet scavenging (aerosol depletion in rain), and mass growth
through chemical processing. Scavenging (aerosol depletion in cloud), large‐
scale vertical motion (including subsidence and ascent, which may
encourage cloud moistening and increased organization) and entrainment of
aerosols are included for completeness but do not contribute as strongly over
this simulation's duration.
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 15 of 26
Notably, the Aitken‐buffered HfAc case loses less N
d
, LWP, cloud amount, and NetCRE compared to the Aitken‐
deficient NoAit. The reason for this is apparent from Figure 10e: Aitken transfer restores the halved accumulation
number to the initial Ctrl levels by 14:30 UTC, forestalling further precipitation depletion as N
acc
continues to be
enhanced. Ultimately, the total aerosol number is what matters for the system, which is why the Aitken‐buffering
Figure 10. Time evolution of key SAM parameters for Ctrl,HfAc,NoAit, and HfAcNoAit simulations: (a) cloud fraction, (b) precipitation flux at the surface, (c) liquid
water path, (d) net cloud radiative effect, (e) Aitken (dashed) and accumulation (solid) aerosol number concentrations (≤1.6 km), and (f) mean droplet number
concentration in the upper, Sc (≥0.8 km, solid) and lower, Cu (≤0.8 km, dash dot) cloud layers. Observation comparison period shown in dark gray (12:00–14:30 UTC).
Top of atmosphere incoming solar radiation (F, W m
2
) scaled by 1
3is included on (d) for diurnal cycle reference.
Figure 11. Cross sections of SAM sensitivity studies at14:00 UTC as in Figure 2:Ctrl (a, e, i, m), HfAc (b, f, j, n), NoAit (c, g, k, o), and HfAcNoAit (d, h, l, p). Parameters
shown are Aitken (a–d), accumulation (e–h), and cloud droplet (i–l) number concentrations and cloud optical depth (m–p).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 16 of 26
mechanism is effective. Access to Aitken aerosol in HfAc is a larger deterrent
against precipitation depletion of N
acc
and N
d
than larger initial values of N
acc
and N
d
in NoAit. This is taken to the extreme in HfAcNoAit, where accu-
mulation number is halved and the sustaining Aitken influence is removed.
Without Aitken aerosol to restore accumulation aerosol, the HfAcNoAit cloud
system cannot resist precipitation depletion and has the largest loss signatures
of all the simulations.
For all cases, N
d
is higher in the Cu than the Sc layers. This is not unexpected
based on the initially larger net gain in Cu accumulation aerosols compared to
the Sc layer, as seen in the Ctrl budgets and in varying degrees across the
sensitivity studies (Figure 8c and Figure S7e in Supporting Information S1).
We can examine the sensitivity study tendencies in more detail by contrasting
their budget results (dots in Figure 8) and size distribution evolution
(Figure 12).
In general, HfAc budgets behave the same as the Ctrl. However, it has a much
larger initial Aitken transfer and stronger fluxes (9:00–12:00 UTC in
Figure 8), which is consistent with the system compensating for the smaller
initial accumulation number, as discussed previously. Aitken tendencies re-
turn to similar levels as Ctrl in subsequent periods, but
˙
NTot.acc is slightly
larger than Ctrl during 12:00–14:30 UTC thanks to weaker wet scavenging
and fluxes. The size distribution evolution in the Sc layer has similar ten-
dencies for Ctrl and HfAc over 9:00–14:00 UTC (Figures 12a and 12b). As the
Sc layer (a, 1.4 km) grows into the FT (Figure S9a in Supporting Informa-
tion S1), the Aitken mode is depleted and the accumulation mode shifts
strongly to the right, creating a well defined Hoppel minimum associated with
cloud (e.g., Hoffmann & Feingold, 2023) and chemistry processing. Aitken is
transferred similarly in Ctrl and HfAc, reducing the Aitken number in both
and marginally shifting the mode left in HfAc. The matching evolution in their
accumulation modes suggests the larger Ctrl wet scavenging depletes its
larger initial N
acc
, bringing it in line with HfAc.
Near the bottom of the Sc layer (b, 1 km), which subsequently experiences
rising Cu towers, the accumulation mode grows more in HfAc than Ctrl such
that their distribution matches at 14:00 UTC. HfAc's evolution is likely assisted by stronger fluxes of accumu-
lation aerosols from the transition layer (Figures 8b and 8c). A rightward shift of the accumulation mode, and
depletion of Aitken, happens in the transition layer for both HfAc and Ctrl (Figure S9b in Supporting Informa-
tion S1). This is consistent with augmentation of accumulation through Aitken transfer and fluxes from the Cu
layer as well as modal shifts due to increasing mass by chemical processing.
In the Cu layer (Figure 12c), Aitken aerosols are transferred locally to the accumulation mode, reducing the
Aitken number. Again, chemical processing shifts modes to the right such that the two distinct, initial aerosol
modes are brought together, reducing the Hoppel minimum. Stronger HfAc turbulent fluxes export accumulation
particles, likely helping to redistribute particles between layers and maintain the accumulation mode. The HfAc
budget and distribution tendencies are confirmation of both where the largest transfer occurs in the system (in the
cloud layers before turbulent fluxes redistribute particles) and how the transfer and fluxes readjust in order to
restore depleted accumulation aerosol and buffer the cloud system.
Under Aitken‐deficient conditions (NoAit,HfAcNoAit), precipitation depletion leads to a loss of accumulation
number at all levels (Figure 8). Turbulent fluxes still move accumulation number from the Cu to the Sc layer but
the import of accumulation number is insufficient to offset removal through wet scavenging in the Sc. N
d
decays
at a similar rate in the Cu and Sc layers (Figure 10f) as a result of precipitation‐driven cloud breakup (dominating
the Sc layer) and accumulation export through turbulent fluxes (dominating the Cu layer) in the absence of
restorative Aitken aerosols. Over 9:00–14:00 UTC, the accumulation mode (Figure 12) shrinks at the top of the Sc
layer (a) and grows at its base (b) under the influence of precipitation depletion (dominating a) and flux transport.
Figure 12. SAM size distributions evolving (shading) from 9:00 (dotted) to
14:00 UTC (solid line) for sensitivity studies: Ctrl,HfAc,NoAit, and
HfAcNoAit. Distributions are shown at three levels: (a) the FT (9:00 UTC)
that the Sc cloud layer grows into (14:00 UTC, 1.4 km), (b) the Sc cloud base
(9:00 UTC) and, later, Cu columns (14:00 UTC, 1 km), and (c) the lower Cu
cloud layer (9:00 UTC) and later base of Cu columns (14:00 UTC, 0.5 km).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 17 of 26
Fluxes are weaker in these simulations, helping to grow the accumulation mode in the Sc layer before depletion
(b, Figure S9b in Supporting Information S1) but fluxing fewer accumulation aerosols from the Cu (c) and sub‐
cloud layers (Figures S9b and S9c in Supporting Information S1). Chemical processing is still a significant in-
fluence on the accumulation mode before too many aerosols are lost (Figure S8 in Supporting Information S1),
contributing to the small modal shift near the Sc (Figure 12a and Figure S9b in Supporting Information S1) and Cu
(Figure 12c and Figure S9c in Supporting Information S1) layers. The NoAit and HfAcNoAit tendencies confirm
that without Aitken aerosols, the cloud system undergoes a more rapid collapse driven by both precipitation
depletion (even larger in HfAcNoAit, Figure 10b) and redistribution of particles through turbulent fluxes helping
to accelerate collapse rather than resupplying new accumulation aerosols as in the Aitken‐buffered system.
The impact of Aitken buffering on radiation is seen in the diurnally varying net cloud radiative effect (NetCRE)
(Figure 10d). HfAc produces a similar albeit marginally weaker radiative response than Ctrl. In contrast, NetCRE
for NoAit and HfAcNoAit are considerably smaller in magnitude with shapes dictated by their cloud break‐up (a).
NoAit and HfAcNoAit NetCRE peak just before precipitation flux substantially increases (b) and dissipates their
cloud layers (a, c, f).
We can examine this radiative evolution, and its contributing factors, in more detail with time evolving PDFs of
key variables over 10:00–14:00 UTC (Figures 13 and 14). Because the HfAcNoAit simulation is already fairly
collapsed at 12:00 UTC, we focus on contrasting the still evolving simulation behaviors of the Ctrl with NoAit and
HfAc over this period. Figure 13 highlights the aerosol‐cloud‐precipitation evolution we expect from these three
sensitivity studies (e.g., Figures 8and 10). While precipitation increases over time (b, e, h), HfAc N
acc
(a, d, g) and
N
d
(c, f, i) PDFs shift toward Ctrl PDFs as Aitken aerosols transfer to the accumulation mode and, despite their
initial decrease, restore N
acc
,N
d
back toward the unperturbed Ctrl simulation. By 14:00 UTC, the mean and
median of N
acc
,N
d
for HfAc have been buffered to within the interquartile range of Ctrl. In contrast, NoAit PDFs
for N
acc
,N
d
shift to the left, away from the Ctrl, in response to increasing precipitation depletion in the absence of
Aitken buffering. Thus, Aitken aerosol presence is critical for sustaining N
d
in the Ctrl,HfAc simulations, as
evidenced by the swap in PDF location between NoAit and HfAc by 14:00 UTC.
Aerosol behavior controls the ability to sustain cloud homogeneity (Figure 11) and NetCRE (Figure 14). One can
also measure the inverse of this, or the amount of optically thin cloud layers (larger percentage with τ≤3,
Figures 14b, 14e, and 14h). There are far more of these generated in the NoAit case by 14:00 UTC compared to
Ctrl and HfAc, consistent with the more visually heterogeneous clouds of that simulation (Figures 11c, 11g, 11k,
and 11o). It is also worth noting that the increasing magnitude and PDF broadening that occurs for NetCRE
Figure 13. PDFs at native model resolution for the Ctrl,HfAc, and NoAit simulations at (a–c) 10:00, (d–f) 12:00, and (g–i) 14:00 UTC. N
acc
(a, d, g) and N
d
(c, f, i) are for
all values in the upper cloud layer (≥0.8 km) while precipitation flux (b, e, h) is through the bottom edge (0.8 km).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 18 of 26
happens at the same time that the LWP decreases in Ctrl,HfAc (Figures 14f and 14i). Generally, the time of largest
wet scavenging (12:00–14:30 UTC, Figure S7 in Supporting Information S1) occurs when large‐scale ascent is
peaking (12:00–15:00 UTC, Figure S3 in Supporting Information S1). This suggests that the influence of large‐
scale meteorology and aerosol processing are fundamentally linked, an important topic for future investigations.
5. Discussion
5.1. Buffering Timescales
In this decoupled Cu rising into Sc case, processes influencing cloud and aerosol evolution operate over hours.
Rapid aerosol evolution is driven by Aitken activation, turbulence, precipitation depletion, and chemical pro-
cessing (Figure 9). Precipitation loss begins to impact accumulation number and mass almost immediately in the
Ctrl simulation (Figures 8and 10). The system is still buffered against cloud break‐up, however, as depletion is
prevented by Aitken transfer and turbulent fluxes throughout the simulation (Figure 8).
The role of the elevated FT Aitken concentrations in the maintenance of BL cloud and aerosols is worth exploring
in this case study. Peaks in Aitken concentration near cloud top (e.g., Figures 2b and 2c) and in Aitken transfer
profiles (Figure 7b) as well as the Aitken mode depletion and accumulation mode growth at Sc top (Figure 12a)
indicate that entrained Aitken aerosol can directly buffer precipitation‐depleted clouds (i.e., CCN‐depleted su-
persaturated updrafts may be sufficient to activate locally entrained Aitken particles at cloud top).
The W22 10‐day simulations found that large sources of Aitken particles, either from FT import (FT Aitken set to
1,000 mg
1
) or surface production (10×surface source) could prevent BL cloud collapse in a subtropical,
meteorologically quiescent regime. Our Ctrl simulation, particularly the exponential Aitken number depletion
signature (Figure 10d), resembles the W22 BL1000 sensitivity study where BL Aitken concentrations were set to
1,000 mg
1
while FT and surface sources were kept small. Cloud breakup was delayed in BL1000 for twice as
long as the control (8 vs. 4 days), suggesting that cloud breakup will be delayed in our case too, even without large
FT or surface Aitken sources. Our initial BL Aitken aerosol (∼100 mg
1
, averaged over surface and transition
values from the initial Aitken profile, Table S1 in Supporting Information S1) was either brought in from the FT
over the past few days, generated from sea spray production (Lawler et al., 2021; Xu et al., 2022), or formed via
new particle formation in the ultra‐clean outflow at cloud edges (Kazil et al., 2011) or within the BL (Zheng
et al., 2021). Since our model neglects the new particle production mechanism and has a very small surface source
of Aitken aerosols in this weak‐wind case, it is best used to quantify FT influence.
Figure 14. As in Figure 13 but for the: liquid water path in the upper Sc layer (a, d, g, for ≥0.8 km), cloud optical depth (b, e, h), and net cloud radiative effect (c, f, i). A
gray line in (b, e, h) references the optical depth threshold (τ=3) for optically thin clouds (O, Wood, & Tseng, 2018).
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 19 of 26
Using a marine BL mass weighted average, we estimate that ∼60 mg
1
of Aitken mode aerosol particles were
transferred to the accumulation mode over the full duration of the Ctrl simulation (9:00–21:00 UTC). This
represents more than half the initial Aitken value within the BL. The HfAc simulation experiences an even larger
transfer, ∼70 mg
1
from 9:00 to 21:00 UTC. We estimate the entrainment flux of aerosols, determining whether
FT entrainment is able to balance these losses and, if not, how many days worth of FT Aitken entrainment are
consumed during this event. Our aerosol entrainment source calculation is based on an estimation of the
entrainment rate and the jump of aerosols across the inversion (Text S1 in Supporting Information S1).
Entrainment from the FT increases Aitken aerosol in the BL by ∼55 mg
1
in Ctrl and ∼50 mg
1
in HfAc over
9:00–21:00 UTC. Aitken transfer consumes approximately the same amount of Aitken aerosols as are entrained
over the full 12 hr simulation in Ctrl (∼0.5 days of FT Aitken entrainment) and ∼1.4×in HfAc (∼0.7 days). Note
that a similar accumulation mode estimation shows FT entrainment very weakly dilutes BL accumulation con-
centrations (∼ 5 mg
1
lost over 9:00–21:00 UTC in Ctrl and ∼ 10 mg
1
in HfAc). This is consistent, albeit
much smaller in magnitude, with FT accumulation dilution found in similar cloud structures in marine cold air
outbreak outflows (Tornow et al., 2022). However, Arctic FT entrainment of Aitken (Igel et al., 2017; Price
et al., 2023) and accumulation aerosols (Sterzinger & Igel, 2024) can sustain clouds despite no surface sources
and under low BL concentrations, pointing to the complex variation in aerosol influence across cloud regimes.
One can imagine that an air mass might experience increasing Aitken aerosol concentrations during non‐
precipitating periods which might be consumed during periods of stronger forcing and precipitation. In this
way, Aitken buffering of marine BL clouds may be accomplished, in part, with pre‐existing Aitken mode aerosols
that were entrained from the FT in the preceding days, which may help cases like ours. However, we also note
that, at this latitude, the FT Aitken number is observed to have concentrations of ∼210 cm
3
with accumulation
number concentrations of ∼250 cm
3
(Heintzenberg et al., 2000) suggesting there is an additional source of
Aitken aerosols that assists in balancing this BL sink (e.g., new particle formation, Zheng et al., 2021).
Expanding on this idea, we note that the above estimation assumes FT Aitken import only occurs locally,
neglecting the substantial particle import that occurs with the passage of mid‐latitude cyclones (e.g., Covert
et al., 1996). Zheng et al. (2021) estimate that in post‐frontal open cellular clouds occurring in the ACE‐ENA
region, which are most likely to experience FT import after the passage of a cyclone, it takes 30–45 hr for FT
air to replace the air in a 2 km deep BL. Assuming that the FT concentration of our initial profile (∼250 mg
1
,
similar to Heintzenberg et al. (2000)) is somewhat representative over this region for this season, we estimate a
post frontal entrainment rate of 130–200 mg
1
day
1
which is, respectively, a factor of ∼1 to 1.7×greater than the
Aitken transfer rate during our case (∼ 120 mg
1
day
1
). Aitken transfer during our Ctrl case would consume
∼0.3–0.5 days of Aitken aerosol entrained under post frontal conditions. This, even excluding new particle
formation at cloud edges (e.g., Kazil et al., 2011) or in the BL (e.g., Zheng et al., 2020), emphasizes that Aitken
aerosol can be frequently replenished and that Aitken buffering is likely to be both feasible and important in this
region.
5.2. Challenges of Simulating Real‐World Case Studies
We encountered a few challenges in simulating this case, in part due to a unique combination of factors. First, the
detailed aerosol‐cloud‐precipitation observations for this morphology regime were taken over a relatively short
time period. Second, this regime was rapidly evolving, in part due to the non‐trivial meteorological forcing
experienced throughout. This made using observations to both initialize and interrogate our simulation
complicated, which leads us to an important question about our model construction and its limitations: is our
initialization appropriate?
Aitken transfer occurs even in the spinup period of our simulation (e.g., Figure S5 in Supporting Information S1),
contributing to the slight mis‐partitioning of Aitken aerosol into the accumulation mode near cloud layers
compared to observations (Figures 3a and 3b). We found that this behavior was sensitive to the nudging meth-
odology utilized and were able to reduce this issue greatly with the current setup by allowing subcloud turbulence
and convection to develop during the spinup period when the spuriously large Aitken transfer associated with the
initiation of the cumulus layer could be offset by nudging. As a result, relatively weak transients occur after
nudging is switched off, and model‐observation aerosol discrepancies are largely restricted to the accumulation
mode. Comparisons with satellite observations further benefited from the development of mesoscale organization
during the spinup period. Another approach would be to follow Neggers et al. (2019) who select initial values
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 20 of 26
using a technique that could be successful in our case. They initialize many short‐duration, Lagrangian simu-
lations with varying initial states upwind of an observation platform and select conditions producing the smallest
biases.
We expect that a more carefully tuned initialization method would have a minor impact on our results, however,
and the main improvement would be in reducing the small model‐observation aerosol biases in the lower and
upper BL. Model‐observation consistency elsewhere and for other parameters (especially N
d
and the net balance
between aerosol sources and sinks) suggests that the model is credible and has skill. Thus, we expect the key
mechanisms driving aerosol‐cloud‐precipitation evolution in this regime and their sensitivity to large changes in
the initial aerosol profile (i.e., no Aitken, halved accumulation) are robust.
Our results provide additional (e.g., McCoy et al., 2021) guidance for future observational investigations into
Aitken buffering. Aitken activation will be more likely to occur during periods of peak precipitation in the diurnal
cycle, when precipitation helps to remove larger aerosols and increase peak updraft supersaturations. Cloud
regimes with decoupled boundary layers that include more developed, broken clouds and synoptic or meso‐scale
ascent are also likely to have more favorable updraft characteristics. Changes in maximum supersaturation
deduced from the Hoppel diameter in cloud‐free air (e.g., Gong et al., 2023), trends in Aitken or total aerosol
number concentrations (e.g., McCoy et al., 2021), cloud droplet number concentration relationships with size‐
resolved aerosol concentrations (e.g., McCoy et al., 2021) and supersaturations (e.g., Sanchez et al., 2021),
and re‐sampling of air masses over multiple timescales (e.g., B. Albrecht et al., 2019; Mohrmann et al., 2019) are
some methods that would help to identify occurrences of Aitken activation and buffering. As demonstrated by the
ACE‐ENA instrumental suite (J. Wang et al., 2017,2022), observations of the full aerosol size distribution, from
the Aitken through to the coarse mode, along with cloud and precipitation measurements are critical for any such
investigation.
6. Summary
We utilize the System for Atmospheric Modeling (SAM) large eddy scale (LES) model with a novel Aitken‐mode
enabled microphysics scheme (Wyant et al., 2022, hereafter W22) to investigate a summertime mid‐latitude
decoupled low cloud regime observed during the ACE‐ENA flight campaign (J. Wang et al., 2022). On 15
July 2017, the G‐1 aircraft sampled an evolving cloud system composed of cumulus (Cu) rising into stratocu-
mulus (Sc) under heightened Aitken aerosol concentrations (100–200 mg
1
) (Figure 1). In situ aircraft obser-
vations, reanalysis and satellite retrievals were used to develop and evaluate our case study.
We examined whether a large concentration of BL Aitken aerosols impacted the evolution, radiative properties,
and heterogeneity of this cloud system. Using observations to constrain our case study as well as realistic
meteorological forcing, we found that the W22 aerosol‐coupled SAM captured key time‐evolving processes
driving BL cloud and aerosol evolution. Profiles of total aerosol number matched observed evolution throughout
the BL depth. Aerosols tended to be slightly over‐partitioned into the accumulation mode in the cloud layers due
to too many Aitken particles being transferred into the accumulation mode in supersaturated updrafts, but were
within the observed interquartile range elsewhere. Simulated cloud liquid water was also within the upper end of
the observed range, leading to slightly brighter clouds with reasonable optical thickness relative to satellite re-
trievals. SAM simulated more light precipitation than observed, likely due to aircraft sampling being dominated
by a few heavily precipitating clouds. SAM cloud droplet number concentrations matched observations, indi-
cating the minor aerosol and microphysical discrepancies did not ultimately skew the net balance of cloud
condensation nuclei (CCN) sources and sinks.
We identified the key aerosol‐cloud‐precipitation processes driving the evolution of this morphology regime
(Figure 9). Aitken activation in the Cu and Sc layers generates accumulation aerosols that are grown by chemical
processing throughout the BL. These accumulation aerosols are carried up from the Cu layer to the drizzling Sc
layer by turbulent and convective motions. Simultaneously, eddies bring Aitken aerosols down from the transition
layer below the Sc to the Cu layer where they can be activated and grown. The continuous transfer of Aitken
aerosol to the accumulation mode via activation in cloud droplets in the Cu and Sc layers buffers CCN against
precipitation loss. Entrainment of Aitken aerosol from the FT roughly balances the consumption of Aitken
aerosols by transfer to the accumulation mode in this case. However, given stronger forcing, precipitation, and
wet scavenging, the reserves of Aitken aerosols might be depleted, requiring entrainment of FT Aitken aerosols to
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 21 of 26
replenish those in the BL. In particular, we estimate that BL Aitken concentrations can be restored between 0.3
and 0.5 days depending on their environment.
Aerosol sensitivity studies illustrate that Aitken buffering is essential in maintaining more homogeneous clouds
and preventing their break‐up over the 12‐hr duration of our simulation. Precipitation‐driven break up can be
delayed if BL Aitken is present, even under halved accumulation concentrations. Even with the significant
meteorological forcings present in the mid‐latitudes, the processes driving cloud morphology evolution, het-
erogeneity, and radiative properties are sensitive to Aitken aerosols. Maintaining more reflective clouds for longer
in this environment can be facilitated through Aitken buffering. Accounting for this influence in these pristine
environments will be important for reducing aerosol‐cloud interaction uncertainty in climate sensitivity.
Appendix A: Sensitivity of Aitken Activation to Simulation Design
We design two studies to test the sensitivity of the Ctrl simulation to the formulation of the W22 Hoppel Transfer
Scheme. Specifically, we focus on adjusting how the Aitken aerosol is transferred to the accumulation mode and
the influence that has on the size of the aerosol that is transferred. In AltHopV1, we apply the W22 scheme but
transfer mass as an average of aerosols with the critical diameter, D
c
, and the integrated mass of the Aitken
distribution with D>D
c
. With this approach, the characteristic diameter of transferred aerosols is larger than in
W22, where all transferred aerosols are assumed to have D=D
c
. In addition, we require that the Hoppel transfer
will never make the Aitken modal diameter larger by enforcing:
Mtransfer ≥Ntransfer ×(MAitken
NAitken )(A1)
Rarely, the Aitken mode can be larger than the accumulation mode so that the D
c
is smaller than the Aitken modal
diameter. In AltHopV2, activated Aitken aerosols (i.e., those with D>D
c
) are transferred only if D
c
<D
Hoppel
.
The mean size of transferred aerosol increases monotonically from the W22 Hoppel transfer to AltHopV1 to
AltHopV2.
We contrast the AltHopV studies with the Ctrl,Ctrl LD, and NoAit studies. The AltHopV simulations are bracketed
by the W22 scheme as originally published (Ctrl) and the opposite extreme where no Aitken are allowed to
transfer (NoAit). In our microphysics scheme, the NoAit simulation is the closest possible equivalence to pre-
venting Aitken transfer because activated Aitken aerosols do not exist as a separate category from activated
accumulation mode aerosols. We additionally include the Ctrl LD study, where the domain has been doubled to
51.2 ×51.2 km
2
, to evaluate the relative influence that the domain size has on the Ctrl simulation. We use three
time evolution frameworks for our comparisons: mean behavior (Figure A1, like Figure 10), size distributions
(Figure S10 in Supporting Information S1, like Figure 12 and Figure S9 in Supporting Information S1) and PDFs
(Figures S11 and S12 in Supporting Information S1, like Figures 13 and 14).
Ctrl LD behaves very similarly to the Ctrl and is well separated from the extreme NoAit simulation. Compared to
Ctrl,Ctrl LD tends to have slightly more cloud breakup before recovery (Figure A1a) and initially more persistent
but less prolonged precipitation (b) resulting in a larger eventual depletion in LWP (c) and NetCRE (d). Between
10:00 and 14:00 UTC, Ctrl LD is the same as Ctrl except for slightly higher LWP prior to the previously noted
depletion (Figures S11a, S11d, and S11g in Supporting Information S1). Notably, the accumulation, Aitken, and
N
d
concentrations match the Ctrl behaviors. Ctrl LD size distribution evolution matches the Ctrl as well except for
a small shift to larger accumulation mode particles at the top of the Sc layer (Figure S10b in Supporting Infor-
mation S1). We conclude that our Aitken buffering results are not sensitive to domain size. The domain size used
for the Ctrl and all other sensitivity studies is sufficient to capture the characteristics of the mesoscale structures
examined in this case (which are smaller than the domain, e.g., Figure 2).
Similarly, the small variations in behavior from the AltHopV studies cluster around the Ctrl and never present
differences as large as between the Ctrl and NoAit simulations. AltHopV1 behaves more like the Ctrl while
AltHopV2 is slightly more separated. This is consistent with the expectation that the transfer is slowed more in
AltHopV2 than AltHopV1 relative to the Ctrl as the transferred particle size increases. Accordingly, the accu-
mulation mode shifts to larger sizes for AltHopV1 and, more so, AltHopV2. The AltHopV Aitken and accumu-
lation modes are slightly more separated over time compared to the Ctrl, with a more clearly defined Hoppel
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 22 of 26
minimum in the cloud layers (Figure S10b, S10d, S10f in Supporting Information S1) and, to a lesser degree, in
the less cloudy levels (Figures S10a, S10c, and S10e in Supporting Information S1).
The influence from the transfer slow down from the Ctrl to AltHopV1 to AltHopV2 is also evident. Like Ctrl LD,
AltHopV2 experiences more cloud breakup (Figure A1a), persistent precipitation (b), depletion of LWP (c), and
loss of NetCRE (d) than AltHopV1 and Ctrl. Unlike Ctrl LD,AltHopV2 and, to a lesser degree, AltHopV1 differ in
their number concentrations. AltHopV2 has less Aitken transfer than AltHopV1, which, in turn, has slightly less
than Ctrl. This eventually leads to a noticeable loss of accumulation aerosols and reduces the Sc layer N
d
. The
effective slow down of Aitken transfer is particularly apparent in the evolution from 10:00 to 14:00 UTC where
accumulation number, N
d
(Figure S11 in Supporting Information S1), and LWP (Figure S12 in Supporting In-
formation S1) PDFs and statistics shift to smaller values with time for AltHopV1 and, even more so, for AltHopV2.
A brief increase in precipitation at 12:00 UTC (Figure S11e in Supporting Information S1) and a slight decrease in
τand NetCRE magnitude by 14:00 UTC (Figure S12h and S12i in Supporting Information S1) suggests slower
Aitken buffering in AltHopV1 and AltHopV2 compared to the Ctrl, which eventually catches up and limits
precipitation afterward (Figure A1b).
The difference between Ctrl and NoAit is substantially larger than any of the variations in the Ctrl simulation from
the Hoppel Transfer variants or domain size. We thus conclude that the influence of the Aitken aerosols, and
particularly the absence of their activation, is a much larger signal than from any variations in the simulation setup
that were examined here. These results support the robustness of our conclusions regarding Aitken Buffering and
its importance in mid‐latitude cloud systems.
Data Availability Statement
All the ACE‐ENA campaign observations (J. Wang et al., 2022) used in this study are available at the ARM
Climate Research Facility archive at https://www.arm.gov/research/campaigns/aaf2017ace‐ena. ECMWF ERA5
Reanalysis profiles developed for the campaign (Tao & Xie, 2011) were used in our study as were the NASA
SATCORPS VISST products for ARM (ARM Data Center, 2017). The SAM model (Khairoutdinov & Ran-
dall, 2003) is publicly available at http://rossby.msrc.sunysb.edu/SAM.html. Simulation output, forcings, and
source code (including the aerosol‐enabled microphysics scheme) used in this study are all archived at Zenodo
(McCoy, Blossey, et al., 2023).
Figure A1. As in Figure 10 but for Ctrl and NoAit with the Ctrl LD,AltHopV1, and AltHopV2 sensitivity studies.
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 23 of 26
References
Abdul‐Razzak, H., & Ghan, S. J. (2000). A parameterization of aerosol activation: 2. Multiple aerosol types. Journal of Geophysical Research,
105(D5), 6837–6844. https://doi.org/10.1029/1999jd901161
Ackerman, A. S., Kirkpatrick, M. P., Stevens, D. E., & Toon, O. B. (2004). The impact of humidity above stratiform clouds on indirect aerosol
climate forcing. Nature,432(7020), 1014–1017. https://doi.org/10.1038/nature03174
Albrecht, B., Ghate, V., Mohrmann, J., Wood, R., Zuidema, P., Bretherton, C., et al. (2019). Cloud System Evolution in the Trades (CSET):
Following the evolution of boundary layer cloud systems with the NSF/NCAR GV. Bulletin of the American Meteorological Society,100(1),
93–121. https://doi.org/10.1175/BAMS‐D‐17‐0180.1
Albrecht, B. A. (1989). Aerosols, cloud microphysics, and fractional cloudiness. Science,245(4923), 1227–1230. https://doi.org/10.1126/science.
245.4923.1227
ARM Data Center. (2017). Minnis Cloud Products Using Visst Algorithm (VISSTPX2DM10MINNIS) [Dataset]. Atmospheric Radiation
Measurement (ARM) user facility, Eastern North Atlantic (ENA) External Data (satellites and others),11. Retrieved from https://www.arm.
gov/capabilities/science‐data‐products/vaps/visst/xds
Bellouin, N., Quaas, J., Gryspeerdt, E., Kinne, S., Stier, P., Watson‐Parris, D., et al. (2020). Bounding global aerosol radiative forcing of climate
change. Reviews of Geophysics,58(1), e2019RG000660. https://doi.org/10.1029/2019RG000660
Berner, A. H., Bretherton, C. S., Wood, R., & Muhlbauer, A. (2013). Marine boundary layer cloud regimes and POC formation in a CRM coupled
to a bulk aerosol scheme. Atmospheric Chemistry and Physics,13(24), 12549–12572. https://doi.org/10.5194/acp‐13‐12549‐2013
Binkowski, F. S., & Shankar, U. (1995). The Regional Particulate Matter Model: 1. Model description and preliminary results. Journal of
Geophysical Research,100(D12), 26191–26209. https://doi.org/10.1029/95JD02093
Blossey, P. N., Bretherton, C. S., & Mohrmann, J. (2021). Simulating observed cloud transitions in the northeast Pacific during CSET. Monthly
Weather Review.https://doi.org/10.1175/MWR‐D‐20‐0328.1
Boerner, T. J., Deems, S., Furlani, T. R., Knuth, S. L., & Towns, J. (2023). ACCESS: AdvancingInnovation: NSF’s Advanced Cyberinfrastructure
Coordination Ecosystem: Services & Support. In Practice and Experience in Advanced Research Computing (PEARC23), July 23–27, 2023,
Portland, OR, USA (p. 4). ACM. https://doi.org/10.1145/3569951.3597559
Boucher, O., Randall, D., Artaxo, P., Bretherton, C., Feingold, G., Forster, P., et al. (2013). Clouds and aerosols. In T. F. Stocker, D. Qin, G.‐K.
Plattner, M. Tignor, S. K. Allen, J. Doschung, et al. (Eds.), Climate change 2013: The physical science basis. Contribution of working group I to
the fifth assessment report of the Intergovernmental Panel on Climate Change (pp. 571–657). Cambridge University Press. https://doi.org/10.
1017/CBO9781107415324.016
Bretherton, C. S., & Blossey, P. N. (2017). Understanding mesoscale aggregation of shallow cumulus convection using large‐eddy simulation.
Journal of Advances in Modeling Earth Systems,9(8), 2798–2821. https://doi.org/10.1002/2017ms000981
Bretherton, C. S., Blossey, P. N., & Uchida, J. (2007). Cloud droplet sedimentation, entrainment efficiency, and subtropical stratocumulus albedo.
Geophysical Research Letters,34(3), L03813. https://doi.org/10.1029/2006gl027648
Brown, S. T., Buitrago, P., Hanna, E., Sanielevici, S., Scibek, R., & Nystrom, N. A. (2021). Bridges‐2: A platform for rapidly‐evolving and data
intensive research. In Practice and Experience in Advanced Research Computing (pp. 1–4). https://doi.org/10.1145/3437359.3465593
Bulatovic, I., Igel, A. L., Leck, C., Heintzenberg, J., Riipinen, I., & Ekman, A. M. L. (2021). The importance of Aitken mode aerosol particles for
cloud sustenance in the summertime high Arctic – A simulation study supported by observational data. Atmospheric Chemistry and Physics,
21(5), 3871–3897. https://doi.org/10.5194/acp‐21‐3871‐2021
Carslaw, K. S., Lee, L. A., Reddington, C. L., Pringle, K. J., Rap, A., Forster, P. M., et al. (2013). Large contribution of natural aerosols to
uncertainty in indirect forcing. Nature,503(7474), 67–71. https://doi.org/10.1038/nature12674
Christensen, M. W., Gettelman, A., Cermak, J., Dagan, G., Diamond, M., Douglas, A., et al. (2022). Opportunistic experiments to constrain
aerosol effective radiative forcing. Atmospheric Chemistry and Physics,22(1), 641–674. https://doi.org/10.5194/acp‐22‐641‐2022
Clarke, A. D., Varner, J. L., Eisele, F., Mauldin, R. L., Tanner, D., & Litchy, M. (1998). Particle production in the remote marine atmosphere:
Cloud outflow and subsidence during ACE 1. Journal of Geophysical Research,103(D13), 16397–16409. https://doi.org/10.1029/97jd02987
Covert, D. S., Kapustin, V. N., Bates, T. S., & Quinn, P. K. (1996). Physical properties of marine boundary layer aerosol particles of the mid‐
Pacific in relation to sources and meteorological transport. Journal of Geophysical Research,101(D3), 6919–6930. https://doi.org/10.1029/
95jd03068
Fan, J., Rosenfeld, D., Zhang, Y., Giangrande, S. E., Li, Z., Machado, L. A. T., et al. (2018). Substantial convection and precipitation en-
hancements by ultrafineaerosol particles. Science,359(6374), 411–418. https://doi.org/10.1126/science.aan8461
Feingold, G., & Kreidenweis, S. M. (2002). Cloud processing of aerosol as modeled by a large eddy simulation with coupled microphysics and
aqueous chemistry. Journal of Geophysical Research,107(D23). AAC 6–1–AAC 6–15. https://doi.org/10.1029/2002JD002054
Feingold, G., Kreidenweis, S. M., Stevens, B., & Cotton, W. R. (1996). Numerical simulations of stratocumulus processing of cloud condensation
nuclei through collision‐coalescence. Journal of Geophysical Research,101(D16), 21391–21402. https://doi.org/10.1029/96JD01552
Gong, X., Wang, Y., Xie, H., Zhang, J., Lu, Z., Wood, R., et al. (2023). Maximum supersaturation in the marine boundary layer clouds over the
North Atlantic. AGU Advances,4(6), e2022AV000855. https://doi.org/10.1029/2022AV000855
Gordon, H., Kirkby, J., Baltensperger, U., Bianchi, F., Breitenlechner, M., Curtius, J., et al. (2017). Causes and importance of new particle
formation in the present‐day and preindustrial atmospheres. Journal of Geophysical Research: Atmospheres,122(16), 8739–8760. https://doi.
org/10.1002/2017jd026844
Gordon, H., Sengupta, K., Rap, A., Duplissy, J., Frege, C., Williamson, C., et al. (2016). Reduced anthropogenic aerosol radiative forcing caused
by biogenic new particle formation. Proceedings of the National Academy of Sciences,113(43), 12053–12058. https://doi.org/10.1073/pnas.
1602360113
Heintzenberg, J., Covert, D. C., & Van Dingenen, R. (2000). Size distribution and chemical composition of marine aerosols: A compilation and
review. Tellus B,52(4), 1104–1122. https://doi.org/10.1034/j.1600‐0889.2000.00136.x
Hoffmann, F., & Feingold, G. (2023). A note on aerosol processing by droplet collision‐coalescence. Geophysical Research Letters,50(11),
e2023GL103716. https://doi.org/10.1029/2023GL103716
Igel, A. L., Ekman, A. M. L., Leck, C., Tjernström, M., Savre, J., & Sedlar, J. (2017). The free troposphere as a potential source of arctic boundary
layer aerosol particles. Geophysical Research Letters,44(13), 7053–7060. https://doi.org/10.1002/2017GL073808
Ivanova, I. T., & Leighton, H. G. (2008a). Aerosol–cloud interactions in a mesoscale model. Part I: Sensitivity to activation and collision–
coalescence. Journal of the Atmospheric Sciences,65(2), 289–308. https://doi.org/10.1175/2007JAS2207.1
Ivanova, I. T., & Leighton, H. G. (2008b). Aerosol–cloud interactions in a mesoscale model. Part II: Sensitivity to aqueous‐phase chemistry.
Journal of the Atmospheric Sciences,65(2), 309–330. https://doi.org/10.1175/2007JAS2276.1
Acknowledgments
We thank all those who gathered, worked
with, and provided data from the ACE‐
ENA field campaign and the
accompanying satellite retrievals. Data
were obtained from the Atmospheric
Radiation Measurement (ARM) Program
sponsored by the U.S. Department of
Energy, Office of Science, Office of
Biological and Environmental Research,
Climate and Environmental Sciences
Division. We acknowledge support from
the U. S. Department of Energy
Atmospheric System Research (DOE
ASR) through grants DE‐SC0020134 and
DE‐SC0021103. Research by ILM was
supported by the NOAA Climate and
Global Change Postdoctoral Fellowship
Program, administered by UCAR's
Cooperative Programs for the
Advancement of Earth System Science
(CPAESS) under award
NA18NWS4620043B and by the NOAA
cooperative agreements
NA17OAR4320101 and
NA22OAR4320151. CSB acknowledges
support from the Allen Institute for AI.
This work used Bridges‐2 (Brown
et al., 2021) at Pittsburgh Supercomputing
Center through allocation EES210037
from the Advanced Cyberinfrastructure
Coordination Ecosystem: Services &
Support (ACCESS) program (Boerner
et al., 2023), which is supported by
National Science Foundation Grants
2138259, 2138286, 2138307, 2137603,
and 2138296, and also through allocation
TG‐EES210037 as part of the Extreme
Science and Engineering Discovery
Environment (XSEDE) (Towns
et al., 2014), which is supported by NSF
ACI‐1548562. We thank our editor, Yun
Qian; two anonymous reviewers; and
Graham Feingold for their suggestions for
improving our manuscript. ILM thanks
Amy, Daniel, John, and Laura for their
insights and support. Finally, we thank
Marat Khairoutdinov for developing,
maintaining, and sharing SAM.
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 24 of 26
Kaufman, Y. J., & Tanré, D. (1994). Effect of variations in super‐saturation on the formation of cloud condensation nuclei. Nature,369(6475), 45–
48. https://doi.org/10.1038/369045a0
Kazil, J., Wang, H., Feingold, G., Clarke, A. D., Snider, J. R., & Bandy, A. R. (2011). Modeling chemical and aerosol processes in the transition
from closed to open cells during VOCALS‐REx. Atmospheric Chemistry and Physics,11(15), 7491–7514. https://doi.org/10.5194/acp‐11‐
7491‐2011
Khairoutdinov, M. F., & Randall, D. A. (2003). Cloud Resolving modeling of the ARM summer 1997 IOP: Model formulation, results, un-
certainties, and sensitivities. Journal of the Atmospheric Sciences,60(4), 607–625. https://doi.org/10.1175/1520-0469(2003)060<0607:
CRMOTA>2.0.CO;2
Konsta, D., Dufresne, J., Chepfer, H., Vial, J., Koshiro, T., Kawai, H., et al. (2022). Low‐level marine tropical clouds in six CMIP6 models are too
few, too bright but also too compact and too homogeneous. Geophysical Research Letters,49(11), e2021GL097593. https://doi.org/10.1029/
2021GL097593
Lawler, M. J., Saltzman, E. S., Karlsson, L., Zieger, P., Salter, M., Baccarini, A., et al. (2021). New insights into the composition and origins of
ultrafine aerosol in the summertime high Arctic. Geophysical Research Letters,48(21), e2021GL094395. https://doi.org/10.1029/
2021GL094395
Leahy, L. V., Wood, R., Charlson, R. J., Hostetler, C. A., Rogers, R. R., Vaughan, M. A., & Winker, D. M. (2012). On the nature and extent of
optically thin marine low clouds. Journal of Geophysical Research,117(D22), D22201. https://doi.org/10.1029/2012JD017929
McCoy, I. L., Blossey, P. N., Wyant, M. C., Bretherton, C. S., & Wood, R. (2023). Supporting data for the manuscript: “Aitken‐mode aerosols
buffer decoupled mid‐latitude boundary layer clouds against precipitation depletion” [Software and Dataset]. Zenodo, 0.1.https://doi.org/10.
5281/zenodo.8088443
McCoy, I. L., Bretherton, C. S., Wood, R., Twohy, C. H., Gettelman, A., Bardeen, C. G., & Toohey, D. W. (2021). Influences of recent particle
formation on Southern Ocean aerosol variability and low cloud properties. Journal of Geophysical Research‐Atmospheres,126(8),
e2020JD033529. https://doi.org/10.1029/2020JD033529
McCoy, I. L., McCoy, D. T., Wood, R., Regayre, L., Watson‐Parris, D., Grosvenor, D. P., et al. (2020). The hemispheric contrast in cloud
microphysical properties constrains aerosol forcing. Proceedings of the National Academy of Sciences,117(32), 18998–19006. https://doi.org/
10.1073/pnas.1922502117
McCoy, I. L., McCoy, D. T., Wood, R., Zuidema, P., & Bender, F. A. (2023). The role of mesoscale cloud morphology in the shortwave cloud
feedback. Geophysical Research Letters,50(2), e2022GL101042. https://doi.org/10.1029/2022GL101042
Mieslinger, T., Stevens, B., Kölling, T., Brath, M., Wirth, M., & Buehler, S. A. (2021). Optically thin clouds in the trades. Atmospheric Chemistry
and Physics Discussions,2021, 1–33. https://doi.org/10.5194/acp‐2021‐453
Minnis, P., Nguyen, L., Palikonda, R., Heck, P. W., Spangenberg, D. A., Doelling, D. R., et al. (2008). Near‐real time cloud retrievals from
operational and research meteorological satellites. In R. H. Picard, A. Comeron, K. Schäfer, A. Amodeo, & M. V. Weele (Eds.), Remote
Sensing of Clouds and the Atmosphere XIII (Vol. 7107). SPIE. (Backup Publisher: International Society for Optics and Photonics). https://doi.
org/10.1117/12.800344.710703
Minnis, P., Smith, W. L., Jr., Young, D. F., Nguyen, L., Rapp, A. D., Heck, P. W., et al. (2001). A near‐real time method for deriving cloud and
radiation properties from satellites for weather and climate studies. In Proceedings of AMS 11th Conference Satellite Meteorology and
Oceanography, Madison, WI (October 15–18) (pp. 477–480).
Minnis, P., Sun‐Mack, S., Young, D. F., Heck, P. W., Garber, D. P., Chen, Y., et al. (2011). CERES edition‐2 cloud property retrievals using
TRMM VIRS and Terra and Aqua MODIS data—Part I: Algorithms. IEEE Transactions on Geoscience and Remote Sensing,49(11), 4374–
4400. https://doi.org/10.1109/TGRS.2011.2144601
Mlawer, E. J., Taubman, S. J., Brown, P. D., Iacono, M. J., & Clough, S. A. (1997). Radiative transfer for inhomogeneous atmospheres: RRTM, a
validated correlated‐k model for the longwave. Journal of Geophysical Research,102(D14), 16663–16682. https://doi.org/10.1029/97JD00237
Mohrmann, J., Bretherton, C. S., McCoy, I. L., McGibbon, J., Wood, R., Ghate, V., et al. (2019). Lagrangian evolution of the Northeast Pacific
marine boundary layer structure and cloud during CSET. Monthly Weather Review,147(12), 4681–4700. https://doi.org/10.1175/mwr‐d‐19‐
0053.1
Morrison, H., Curry, J. A., & Khvorostyanov, V. I. (2005). A new double‐moment microphysics parameterization for application in cloud and
climate models. Part I: Description. Journal of the Atmospheric Sciences,62(6), 1665–1677. https://doi.org/10.1175/JAS3446.1
Narenpitak, P., Kazil, J., Yamaguchi, T., Quinn, P., & Feingold, G. (2021). From sugar to flowers: A transition of shallow cumulus organization
during ATOMIC. Journal of Advances in Modeling Earth Systems,13(10), e2021MS002619. https://doi.org/10.1029/2021ms002619
Neggers, R. A. J., Chylik, J., Egerer, U., Griesche, H., Schemann, V., Seifert, P., et al. (2019). Local and remote controls on arctic mixed‐layer
evolution. Journal of Advances in Modeling Earth Systems,11(7), 2214–2237. https://doi.org/10.1029/2019MS001671
O, K.‐T., Wood, R., & Bretherton, C. S. (2018). Ultraclean layers and optically thin clouds in the stratocumulus‐to‐cumulus transition. Part II:
Depletion of cloud droplets and cloud condensation nuclei through collision–coalescence. Journal of the Atmospheric Sciences,75(5), 1653–
1673. https://doi.org/10.1175/jas‐d‐17‐0218.1
O, K.‐T., Wood, R., & Tseng, H.‐H. (2018). Deeper, precipitating PBLs associated with optically thin veil clouds in the Sc‐Cu transition.
Geophysical Research Letters,45(10), 5177–5184. https://doi.org/10.1029/2018gl077084
Pöhlker, M. L., Zhang, M., Campos Braga, R., Krüger, O. O., Pöschl, U., & Ervens, B. (2021). Aitken mode particles as CCN in aerosol‐ and
updraft‐sensitive regimes of cloud droplet formation. Atmospheric Chemistry and Physics,21(15), 11723–11740. https://doi.org/10.5194/acp‐
21‐11723‐2021
Price, R., Baccarini, A., Schmale, J., Zieger, P., Brooks, I. M., Field, P., & Carslaw, K. S. (2023). Late summer transition from a free‐tropospheric
to boundary layer source of Aitken mode aerosol in the high Arctic. Atmospheric Chemistry and Physics,23(5), 2927–2961. https://doi.org/10.
5194/acp‐23‐2927‐2023
Rogers, R. R., & Yau, M. K. (1989). A short course in cloud physics (3rd ed.). Elsevier, Butterworth‐Heinemann. ISBN: 978‐0‐7506‐3215‐7
Sanchez, K. J., Chen, C. L., Russell, L. M., Betha, R., Liu, J., Price, D. J., et al. (2018). Substantial seasonal contribution of observed biogenic
sulfate particles to cloud condensation nuclei. Scientific Reports,8(1), 3235. https://doi.org/10.1038/s41598‐018‐21590‐9
Sanchez, K. J., Roberts, G. C., Saliba, G., Russell, L. M., Twohy, C., Reeves, M. J., et al. (2021). Measurement report: Cloud processes and the
transport of biological emissions affect southern ocean particle and cloud condensation nuclei concentrations. Atmospheric Chemistry and
Physics,21(5), 3427–3446. https://doi.org/10.5194/acp‐21‐3427‐2021
Seinfeld, J. H., & Pandis, S. N. (2016). Atmospheric chemistry and physics: From air pollution to climate change (3rd ed.). John Wiley & Sons,
Incorporated. ISBN: 978‐1‐118‐94740‐1
Sterzinger, L. J., & Igel, A. L. (2024). Above‐cloud concentrations of cloud condensation nuclei help to sustain some Arctic low‐level clouds.
Atmospheric Chemistry and Physics,24(6), 3529–3540. https://doi.org/10.5194/acp‐24‐3529‐2024
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 25 of 26
Stevens, B., & Feingold, G. (2009). Untangling aerosol effects on clouds and precipitation in a buffered system. Nature,461(7264), 607–613.
https://doi.org/10.1038/nature08281
Tao, C., & Xie, S. (2011). 3‐hourly VARANAL for ACE‐ENA June 2017 to February 2018 [Dataset]. https://doi.org/10.5439/1602289
Tornow, F., Ackerman, A. S., Fridlind, A. M., Cairns, B., Crosbie, E. C., Kirschler, S., et al. (2022). Dilution of boundary layer cloud condensation
nucleus concentrations by free tropospheric entrainment during marine cold air outbreaks. Geophysical Research Letters,49(11),
e2022GL098444. https://doi.org/10.1029/2022GL098444
Towns, J., Cockerill, T., Dahan, M., Foster, I., Gaither, K., Grimshaw, A., et al. (2014). XSEDE: Accelerating scientific discovery. Computing in
Science & Engineering,16(5), 62–74. https://doi.org/10.1109/MCSE.2014.80
Twohy, C. H., Clement, C. F., Gandrud, B. W., Weinheimer, A. J., Campos, T. L., Baumgardner, D., et al. (2002). Deep convection as a source of
new particles in the midlatitude upper troposphere. Journal of Geophysical Research,107(D21). AAC 6–1–AAC 6–10. https://doi.org/10.1029/
2001jd000323
Twomey, S. (1959). The nuclei of natural cloud formation part I: The chemical diffusion method and its application to atmospheric nuclei.
Geofisica Pura e Applicata,43(1), 227–242. https://doi.org/10.1007/BF01993559
Twomey, S. (1977). The influence of pollution on the shortwave albedo of clouds. Journal of the Atmospheric Sciences,34(7), 1149–1152. https://
doi.org/10.1175/1520‐0469(1977)034<1149:TIOPOT>2.0.CO;2
Wang, J., Pikridas, M., Spielman, S. R., & Pinterich, T. (2017). A fast integrated mobility spectrometer for rapid measurement of sub‐micrometer
aerosol size distribution, Part I: Design and model evaluation. Journal of Aerosol Science,108, 44–55. https://doi.org/10.1016/j.jaerosci.2017.
02.012
Wang, J., Wood, R., Jensen, M. P., Chiu, J. C., Liu, Y., Lamer, K., et al. (2022). Aerosol and Cloud Experiments in the Eastern North Atlantic
(ACE‐ENA). Bulletin of the American Meteorological Society,103(2), E619–E641. https://doi.org/10.1175/bams‐d‐19‐0220.1
Wang, S., Wang, Q., & Feingold, G. (2003). Turbulence, condensation, and liquid water transport in numerically simulated nonprecipitating
stratocumulus clouds. Journal of the Atmospheric Sciences,60(2), 262–278. https://doi.org/10.1175/1520‐0469(2003)060<0262:
TCALWT>2.0.CO;2
Williamson, C. J., Kupc, A., Axisa, D., Bilsback, K. R., Bui, T., Campuzano‐Jost, P., et al. (2019). A large source of cloud condensation nuclei
from new particle formation in the tropics. Nature,574(7778), 399–403. https://doi.org/10.1038/s41586‐019‐1638‐9
Wood, R. (2006). Rate of loss of cloud droplets by coalescence in warm clouds. Journal of Geophysical Research,111(D21), D21205. https://doi.
org/10.1029/2006jd007553
Wood, R. (2012). Stratocumulus clouds. Monthly Weather Review,140(8), 2373–2423. https://doi.org/10.1175/mwr‐d‐11‐00121.1
Wood, R., O, K.‐T., Bretherton, C. S., Mohrmann, J., Albrecht, B. A., Zuidema, P., et al. (2018). Ultraclean layers and optically thin clouds in the
stratocumulus‐to‐cumulus transition. Part I: Observations. Journal of the Atmospheric Sciences,75(5), 1631–1652. https://doi.org/10.1175/jas‐
d‐17‐0213.1
Wyant, M. C., Bretherton, C. S., & Blossey, P. N. (2018). The sensitivity of numerical simulations of cloud‐topped boundary layers to cross‐grid
flow. Journal of Advances in Modeling Earth Systems,10(2), 466–480. https://doi.org/10.1002/2017ms001241
Wyant, M. C., Bretherton, C. S., Wood, R., Blossey, P. N., & McCoy, I. L. (2022). High free‐tropospheric Aitken‐mode aerosol concentrations
buffer cloud droplet concentrations in large‐eddy simulations of precipitating stratocumulus. Journal of Advances in Modeling Earth Systems,
14(6), e2021MS002930. https://doi.org/10.1029/2021MS002930
Xu, W., Ovadnevaite, J., Fossum, K. N., Lin, C., Huang, R.‐J., Ceburnis, D., & O’Dowd, C. (2022). Sea spray as an obscured source for marine
cloud nuclei. Nature Geoscience,15(4), 282–286. https://doi.org/10.1038/s41561‐022‐00917‐2
Xue, H., & Feingold, G. (2006). Large‐eddy simulations of trade wind cumuli: Investigation of aerosol indirect effects. Journal of the Atmospheric
Sciences,63(6), 1605–1622. https://doi.org/10.1175/JAS3706.1
Zender, C. (2001). Particle size distributions: Theory and application to aerosols, clouds, and soils.
Zheng, G., Kuang, C., Uin, J., Watson, T., & Wang, J. (2020). Large contribution of organics to condensational growth and formation of cloud
condensation nuclei (CCN) in the remote marine boundary layer. Atmospheric Chemistry and Physics,20(21), 12515–12525. https://doi.org/
10.5194/acp‐20‐12515‐2020
Zheng, G., Wang, Y., Aiken, A. C., Gallo, F., Jensen, M. P., Kollias, P., et al. (2018). Marine boundary layer aerosol in the eastern North Atlantic:
Seasonal variations and key controlling processes. Atmospheric Chemistry and Physics,18(23), 17615–17635. https://doi.org/10.5194/acp‐18‐
17615‐2018
Zheng, G., Wang, Y., Wood, R., Jensen, M. P., Kuang, C., McCoy, I. L., et al. (2021). New particle formation in the remote marine boundary layer.
Nature Communications,12(1), 527. https://doi.org/10.1038/s41467‐020‐20773‐1
Journal of Geophysical Research: Atmospheres
10.1029/2023JD039572
MCCOY ET AL. 26 of 26
