Anthropogenic aerosols impact cirrus clouds through ice nucleation, thereby
changing the Earth's radiation budget. However, the magnitude and sign of
anthropogenic forcing in cirrus clouds is still very uncertain depending on
the treatments for ice-nucleating particles (INPs), the treatments for haze particle
freezing, and the ice nucleation scheme. In this study, a new ice nucleation
scheme (hereafter the HYBRID scheme) is developed to combine the best
features of two previous ice nucleation schemes, so that global models are
able to calculate the ice number concentration in both updrafts and
downdrafts associated with gravity waves, and it has a robust sensitivity to the
change of aerosol number. The scheme is applied in a box model, and the ice
number concentrations (
Atmospheric aerosol loading has increased significantly since the preindustrial (PI) period, mainly due to anthropogenic emissions associated with the burning of fossil fuels and biomass. Most studies to date have focused on how the increase in anthropogenic aerosols impacts climate via warm clouds, thereby exerting a net cooling effect (Wang and Penner, 2009; Zhu et al., 2019; Gordon et al., 2016; IPCC, 2013). Compared to warm clouds, there has been much less attention paid to anthropogenic forcing as a result of changes to cirrus clouds, which is one of the least understood processes in the climate system (Fan et al., 2016). Cirrus clouds cover about 30 % of the Earth's area (Wang et al., 1996) and play an important role in the Earth's radiation budget and also influence global precipitation and the hydrologic cycle (Waliser et al., 2009; Hong et al., 2016; Matus and L'Ecuyer 2017). Ice particles in cirrus clouds are nucleated on aerosol particles, so that changes to the aerosol composition and loading may alter cirrus clouds by altering cloud microphysics, resulting in a cirrus cloud radiative forcing.
There are major uncertainties in calculating the radiative forcing of cirrus
clouds using global climate models, in terms of both its magnitude (since
PI) and its sign (Storelvmo, 2017). The ice particles in the cirrus clouds
can form either by homogeneous freezing of solution droplets (or haze
particles) (Koop et al., 2000) or by heterogeneous nucleation of ice-nucleating particles (INPs;
Cantrell and Heymsfield, 2005). Supercooled aqueous solutions such as
sulfate haze particles can form ice through homogeneous nucleation when the
relative humidity with respect to ice (RH
Despite the relatively low level of understanding of ice nucleation, a few
physically based parameterizations that treat the competition between
homogeneous and heterogeneous nucleation have been developed in order to
study the effect of ice nucleation in global climate models (Liu and Penner,
2005; Kärcher et al., 2006; Barahona and Nenes, 2008). The Liu and
Penner (2005) parameterization (hereafter LP) is derived from fitting the
simulation results of an adiabatically rising cloud parcel (Liu and Penner,
2005). The LP parameterization is only able to treat cases for which the
updraft velocity is positive, so the evaporation of drops during downdrafts
is neglected. The parameterization developed by Barahona and Nenes (2008)
(hereafter BN) is derived from an analytical solution of the cloud parcel
equations (Barahona and Nenes, 2008). The LP and BN parameterizations always
show a similar trend when there is an increase in either the haze aerosol
number concentrations or INPs (Shi and Liu, 2018), and they result in very
similar ice number concentrations when the water vapor accommodation
coefficient is set to 0.1 (Zhou et al., 2016). The Kärcher et al. (2006)
parameterization (hereafter KL) explicitly calculates the evolution of ice
supersaturation in a rising cloud parcel when different aerosol types freeze
(Kärcher et al., 2006). The KL parameterization was used in previous
studies on the effect of aerosol particles on cirrus clouds because it
includes an explicit representation of the relevant physics (Penner et al.,
2009, 2018), and Penner et al. (2018) added the capability to
represent evaporation of water in downdrafts. However, in the KL
parameterization, aerosol particles in different size bins will freeze
chronologically from the largest size bin until the rate at which RH
Global numerical simulation experiments of aerosol effects on cirrus cloud
formation have been carried out in a limited number of studies with
different ice nucleation parameterizations and updraft treatments. Penner et al. (2009) used the KL and LP parameterizations to estimate the radiative
forcing of aerosols on cirrus clouds using an offline ice nucleation and
radiative transfer model. They found a negligible forcing from sulfate but a
significant cooling (ranging from
In addition to assumptions of the extent to which soot might act as an INP,
a second source of uncertainty in the calculation of aerosol forcing in
cirrus clouds is the treatment of the sub-grid-scale updraft velocity used
in the nucleation scheme (Zhou et al., 2016). Penner et al. (2009) used a
normal probability distribution with a standard deviation of 0.33 m s
Secondary organic aerosols (SOAs) have been shown to have a highly viscous
semisolid or even glassy state at low temperatures and low RH
In this study, we combined the best features of the LP and KL
parameterizations to develop a hybrid ice nucleation scheme that accounts
for the changes in ice number concentrations in both the updrafts and
downdrafts associated with a spectrum of gravity waves. Using a global
climate model coupled with the new ice nucleation scheme, the radiative
forcing of aircraft soot and sulfate was examined. Furthermore, the
radiative forcing of anthropogenic aerosols on cirrus clouds since the PI
time period was estimated both including and excluding the effect of changes
in SOA. A global average negative anthropogenic forcing of
Summary of assumptions for aerosols to be effective INPs in the model.
We used the Community Earth System Model (CESM) version 1.2.2 (refer to
The LP parameterization is only able to calculate the ice nucleation in a
rising parcel, but it is not able to predict the changes in the supersaturation
or simulate the evaporation of ice in downdrafts. As a result, the scheme
used by Penner et al. (2018) to treat gravity waves cannot be used with the
LP parameterization as it was originally formulated. The KL scheme
calculates the changes in the sub-grid-scale variation in RH
In the HYBRID scheme, the supersaturation (
A series of updraft velocities at each grid point were generated based on a fitted wave spectrum to the observed equatorial gravity waves from Podglajen et al. (2016). The standard deviation of this wave spectrum was extended to other latitudes and seasons by using the parameterization proposed by Gary (2006, 2008). It was extended vertically based on the static stability, atmospheric density, and topography. This parameterization of the wave spectrum associated with gravity waves is described in Penner et al. (2018).
When the updraft velocity is positive, the LP parameterization is used to
calculate the increase in the ice number from homogeneous and/or
heterogeneous freezing, so that the HYBRID scheme avoids the lack of
sensitivity to changes in aerosol number in the KL parameterization when
calculating the number of new ice particles. The LP parameterization is
derived by fitting the results of a large set of parcel model simulations
covering different conditions in the upper troposphere (Liu and Penner,
2005). Two separate regimes are identified by the sign of
Based on the method outlined above, the HYBRID scheme calculates the increase in
We define cirrus clouds for which the effects of aerosols are
defined or calculated as all large-scale clouds formed at temperatures
<
Histogram of predicted ice number concentration for 10 000 simulations using an adiabatic parcel model (blue dashed line) and a box model using the HYBRID scheme (red dashed line). Wstd is the standard deviation of the assumed pdf of updraft velocities, while Ndust and Sulfate are the assumed dust and sulfate number concentrations.
In order to examine the ability of the HYBRID ice nucleation scheme to simulate ice number concentration, the results from a box model using the HYBRID scheme are compared with those from an adiabatic parcel model under the same simulation conditions. The adiabatic parcel model was that used to generate the LP parameterization and was introduced in Liu and Penner (2005). The two models are run for 30 min for each simulation, which is the time step used in the CESM. During the 30 min, the updraft velocity is updated every 2.2 min as recommended by Podglajen et al. (2016). The ice number concentrations after the 30 min simulation from the two models are compared. We ran both the adiabatic parcel model and the HYBRID box model using a constant updraft velocity for each 2.2 min interval. When the velocity is positive ice crystals form and grow in the HYBRID box model as described above. For downdrafts, if the supersaturation is below 100 %, the two models use the same method to simulate the evaporation of any existing ice particles, which is also the same method used in the global model as suggested in Kärcher et al. (2006). The scheme used here for the HYBRID box model is the same as the HYBRID scheme used in the global model. Note that the final interval within each 30 min simulation is shortened in order to match the treatment used in the CESM/IMPACT model which uses a 30 min time step. When run within the CESM model, this final ice number concentration is then passed back to the CESM model after this 30 min interval.
The updraft- and downdraft-associated gravity waves are determined from a
Laplace distribution as suggested by the fit to the observed gravity waves
by Podglajen et al. (2016). We conducted 10 000 simulations using a random
selection of the updrafts and downdrafts for each model. Both models use the
same selection of updrafts and downdrafts in each simulation. We set up an
initial condition with a temperature of 230 K, the standard deviation for the
updraft velocities was 0.5 m s
Figure 1 shows a histogram of the predicted ice number concentration (
The box model using the HYBRID scheme predicts larger
Description of cases.
Note: SOA: secondary organic aerosol; INP: ice-nucleating particle.
We performed a series of model experiments in which different emissions of
aerosols and aerosol precursors are used. Table 2 provides a summary of
these experiments. The base case (PD_Base) uses emissions for
the present day (PD, for the year 2000) for anthropogenic sulfur and soot
from fossil fuel adopted from the Community Emission Data System (CEDS) (Hoesly
et al., 2018). Emissions from van Marle et al. (2017) are adopted for
biomass burning emissions. These emissions are the same as the emissions
data sets used for the CMIP6 simulations. The year 2000 emissions are used
for all years of our simulation. We included soot from aircraft for 2006
based on the Aviation Environment Design Tool (AEDT) data set (Barrett et
al., 2010). The AEDT data are presumed to be more accurate than the CMIP6
aircraft emissions since they were developed based on the original flight
tracks of each of 31 million commercial flights worldwide (Wilkerson et al.,
2010). The dimethylsulfide (DMS) emissions from the ocean are assumed
constant in the PD and PI periods (Tilmes et al., 2016). The emission of
dust uses the scheme from Zender et al. (2003). In a sensitivity experiment
(PI_cSoot), the emission of cSoot is removed from
PD_Base to examine its impact on ice number concentration and
radiative forcing. We also set a case (PI_SO4) with the
emission of anthropogenic sulfur changed to PI (
The changes in aerosol emissions only influence the number concentration of ice nuclei in the CESM and thereby give the indirect radiative effect in cirrus clouds. The direct radiative effect caused by the change of aerosols is not included. Clear-sky radiative forcing in this study is not associated with direct aerosol radiative forcing but is rather mainly due to changes in water vapor which leads to changes in the clear-sky longwave radiation (Wang and Penner, 2010). All cases were run with prescribed sea surface temperatures for the present day, and winds were nudged towards ECMWF reanalysis data for the years 2005–2011 using a nudging time of 6 h (Zhang et al., 2014). The data for the last 6 years were used for analysis in this study.
The predicted
The vertically integrated total
The global average
The annual average change in column number concentration
The competition between the heterogeneous INPs and homogeneous haze
particles determines the change in
As in Fig. 3 but for the difference between the
PD_Base and PI_SO4 cases. Differences
significant at the 90 % level according to a Student's
In contrast to the case of aircraft soot, the increase in the sulfur
emissions from PI to PD leads to a large increase in
As in Fig. 3 but for the difference between the
PD_Base and PI_ALL cases. Differences
significant at the 90 % level according to a Student's
The change in
It is conspicuous that sulfur and aircraft soot emissions have effects with
different signs for the change in
Annual mean plots of the change in vertically integrated averaged
ice water path for the difference between the PD_Base and
PI_cSoot cases
Forcing and cloud changes associated with changes in aircraft soot, sulfur, and all anthropogenic aerosols.
Note: Differences significant at the 90 % level according to a Student's
The decrease in
Annual mean plots of all-sky shortwave radiative forcing
The changes in
As in Fig. 7 but for the difference between the
PD_Base and PI_SO4 cases. Differences
significant at the 90 % level according to a Student's
As in Fig. 7 but for the difference between the
PD_Base and PI_ALL cases. Differences
significant at the 90 % level according to a Student's
The total all-sky net forcing (NRF) is determined by the balance between SRF
and LRF. The radiative effects in cirrus clouds are dominated by longwave
radiative effects. However, it is still possible that the radiative forcing
due to changes in
Although the increase in sulfur emissions from PI to PD leads to an increase
in the global average
Compared to the NRF due to the increased sulfur emissions, the NRF is less
positive in South Asia and more negative over the middle to high latitudes of
the NH and the north Indian Ocean when including the increased emissions of
surface and aircraft soot as well as sulfur together (Fig. 9e). In the middle
to high latitudes of the Southern Hemisphere (SH), SRF and LRF always cancel
each other so that NRF is negligible. As a result, the global average NRF
due to all the increased anthropogenic emissions from PI to PD is
The vertically integrated number of INPs
The annual average change in column number concentration
SOA particles have a strong potential to act as INPs and therefore influence
the formation of cirrus clouds. We examined the radiative forcing of all
anthropogenic aerosols on cirrus clouds when including the INPs from
newly nucleated SOA particles. Figure 10 shows the column number
concentration of INPs and zonal average of INPs from SOA in the PD and PI
atmosphere. Changes in natural SOA precursors between the PI and PD
atmospheres (i.e., isoprene,
Annual mean plots of the changes in ice water path
Change in vertically integrated ice number concentration
The increased
This work develops a new ice nucleation parameterization, HYBRID, which is a
combination of the LP and KL parameterizations. The global model using this
new scheme is able to simulate the growth and decay of ice particles in the
updrafts and downdrafts associated with gravity waves as in the modified KL
scheme (Penner et al., 2018) and is able to treat the changes in aerosol
number concentration from freezing haze particles with fidelity in the sign
of the change as in the LP scheme. The HYBRID scheme overcomes some of the
deficiencies in previous ice nucleation schemes. We evaluated the HYBRID ice
nucleation scheme by comparing the scheme with the Liu and Penner (2005)
adiabatic parcel model and by comparing its global predictions using
observed
We performed a series of model experiments using the HYBRID ice nucleation
scheme to explore the forcing and cloud changes associated with changes in
aircraft soot, sulfur emissions and all anthropogenic emissions from the PI
to PD. Results are summarized in Table 3. The INPs from aircraft soot
usually decrease
The changes in
The influence of SOA on the anthropogenic forcing of aerosols in large-scale
cirrus clouds was examined. The additional INPs from SOA increase the
The radiative forcing of anthropogenic aerosols' effect on cirrus clouds
estimated in this study is less negative than the result indicated in Penner
et al. (2009) (
The CESM1.2.2 model and related documentation can be download from
The supplement related to this article is available online at:
JZ developed the model, performed the simulations, and analyzed all data. JEP guided the model development and data analysis. Both authors contributed to writing the paper.
The authors declare that they have no conflict of interest.
Computer time was provided by the NCAR CISL.
This research has been supported by the National Aeronautics and Space Administration (grant no. NNX15AE34G), the National Science Foundation, Directorate for Geosciences (grant no. 1540954), and the Startup Foundation of Tianjin University (grant no. 390/0903061032).
This paper was edited by Martina Krämer and reviewed by two anonymous referees.