Introduction
Current understanding of the global climate is underpinned by the concept of
the radiation budget, the balance of energy entering and leaving the
Earth's atmosphere. Aerosols play an important role in this budget, having
direct and indirect effects via cloud processes
. Aerosols produce a net cooling effect at the
surface, with the total aerosol effective radiative forcing estimated as
-0.9 Wm-2 by the most recent Intergovernmental Panel on Climate
Change (IPCC) report. The aerosol radiative forcing substantially offsets the
effect of well mixed greenhouse gases' effective radiative forcing of
2.8 Wm-2 . However, large uncertainty in
aerosol radiative forcing remains (±0.5 Wm-2 in the 2013
IPCC report), and is in fact the largest source of uncertainty in the overall
radiation budget for the current climate .
Uncertainties due to aerosols affect not only the radiation budget, but also
chemical and meteorological parameters such as ozone concentration and
photolysis , cloud formation, albedo, temperature and
precipitation .
Natural aerosol sources account for the largest portion of this uncertainty,
explaining up to 45 % of the variance of aerosol forcing, compared to 34 %
from anthropogenic aerosol emissions . Dimethyl sulfide (DMS) produced by
marine organisms makes up approximately 19 % of global sulfur emissions,
producing a DMS flux (fluxDMS) of 17.6 Tgyr-1
, though estimates range from 9 to 35 Tgyr-1 of
sulfur . Global DMS
concentrations and fluxes remain poorly constrained by observations
, and its role in the climate system is subject
to debate .
Charlson, Lovelock, Andreae and Warren (CLAW) proposed a hypothesis by which
marine organisms, primarily phytoplankton, regulate their environment via the
increased production of dimethyl sulfonium propionate (DMSP) when stressed,
for example due to warm sea surface temperatures (SSTs) .
DMSP is degraded via bacterial processes to DMS in the ocean
, some of which is vented into the atmosphere
. Once in the atmosphere DMS has a lifetime of 1–2 days
, and oxidises to form sulfuric acid and
ultimately contributes to the aerosol burden. This additional source of
aerosol can directly or indirectly influence the radiation budget and
potentially cool local SSTs (although this has not been proven in the
literature), hence reducing phytoplankton stress. The DMS–climate system is
summarised in Fig. .
Current understanding of the DMS–climate system implies that no
bio-regulatory feedback exists as proposed by the CLAW hypothesis
. However, observations show that seasonal
cloud condensation nuclei (CCN) variability cannot be explained without a
contribution from DMS , implying that DMS is
important for the longer term climate. Complicating this problem is our
limited understanding of the global distribution of DMS, ultimately relying
on the collection of observations collated and interpolated by
, which may not capture local DMS concentrations in certain
regions such as over coral reefs and at the poles
.
A number of studies have parameterised global oceanic DMS (or DMSP)
concentrations, using primary productivity, insolation, SSTs and other fields
as predictors . However,
numerous issues arise when trying to parameterise oceanic DMS, including the
lack of observations as mentioned, but also that DMS production is species-dependent and predictors are not uniformly relevant across marine biota.
find that two older parameterisations of DMS
concentration perform little better than in the
climatology. Further uncertainty in fluxDMS arises from the
parameterisation of the sea–air flux mechanism. Several parameterisations of
fluxDMS exist e.g., resulting in a large range of
annual global fluxDMS estimates, for example, 15–35 Tgyr-1
of sulfur or 9–35 Tgyr-1 of
sulfur . Both these uncertainties (in climatology and
flux) can have significant consequences for our understanding of climate.
The importance of DMS in large-scale climate has been highlighted by numerous
global modelling studies. (using the DMS
climatology) and (using the DMS
climatology) found DMS to have a radiative effect of -1.79
and -2.03 Wm-2 at the top of the atmosphere (TOA) respectively.
doubled surface water DMS concentrations, (DMSw),
finding a TOA radiative effect of -3.42 Wm-2. These studies
perturbed DMS in the climate system in order to quantify the effect on
climate, and noted that the largest changes in radiation and cloud
microphysics occurred in the Southern Ocean, South Pacific Ocean and South Indian
Ocean.
Other modelling studies have explored the impact of anthropogenic climate
change on marine DMS production, often with opposing conclusions, making it
unclear whether marine DMS production would increase or decrease with warming
temperatures e.g.. found DMS emissions were reduced by 17 %
by the end of the century, primarily due to decreasing ocean pH (caused by
anthropogenic CO2 emissions). The reduced fluxDMS was found
to cause an additional 0.23–0.48 K of warming by the end of the
century . Reduced fluxDMS due to ocean acidification
is also modelled by , who found, under the Representative
Concentration Pathway (RCP) 8.5 to the year 2200, that DMS production decreases by
48 %, assuming a strong sensitivity of DMS production to changes in pH.
calculated a DMS temperature sensitivity of -0.041 K
per Tgyr-1 of sulfur.
Laboratory experiments have found that under ocean acidification, marine
organisms produce significantly less DMS . However, more
recently reported that polar planktonic communities show
resilience to ocean acidification. The study found an
increased fluxDMS in the polar regions, while the
study did not.
attempted to determine if a significant artificial
increase of marine DMS production (due to, for example, ocean fertilisation)
in the oceanic ecosystem could offset future warming trends. Under a scenario
where fluxDMS is increased to 46.1 Tgyr-1 of sulfur,
found that global temperature increases due to
anthropogenic climate change under RCP4.5 were partially offset, primarily
due to low- and mid-level cloud feedbacks, resulting in a radiative flux
perturbation of -2.0 Wm-2. Regional changes in precipitation
(both increases and decreases) were also noted, up to as much as 30 %.
The direct aerosol effect can be approximately linearly related to aerosol
concentration . By contrast, aerosol–cloud processes, or the
secondary aerosol indirect effects, have large uncertainties, with
implications for the radiation budget . Global climate
models are currently unable to capture many key physical and chemical
processes and interactions in the aerosol–cloud system .
These shortcomings add further uncertainty to quantification of the
DMS–climate system. Quantification of model performance is essential in
providing context and perspective to any modelling experiment.
Many DMS–climate modelling studies consider DMS under future scenarios
. However, it is clear
that our understanding of DMS in the current climate is not yet fully
established, considering both modelling and observational uncertainties.
Studies exploring DMS changes under current climate conditions
have completed short simulations
(approximately 1 year), which are too short to be indicative of a true
climatological response. Furthermore, uncertainties related to DMS emission
and fate in the atmosphere are not the only barriers to the DMS–climate
question. Climate model uncertainties and biases must also be considered,
which have not previously been provided.
The interactions between DMS-derived sulfur, its oxidation products and the
atmosphere can be highly non-linear, vary regionally and have far-reaching
impacts on multiple processes in the climate system . Many
of the studies noted above focused on one or two aspects of the DMS–climate
system, commonly reporting on the fluxDMS and its radiative and
temperature effects. In this study we evaluate the whole system, examining
chemical, aerosol and meteorological changes, including cloud and
precipitation effects.
This study has two aims, the first of which is to assess the suitability of
the ACCESS (Australian Community Climate and Earth System Simulator) UKCA
(United Kingdom Chemistry and Aerosol) model for examining the role of DMS in
the Earth's climate in terms of low, medium and high cloud cover, outgoing
TOA shortwave (SW) and longwave (LW) radiation, incoming surface SW
radiation, and precipitation. Secondly, ACCESS-UKCA is used to test the
large-scale sensitivity of the present-day climate to prescribed changes in
DMSw. We aim to discuss these sensitivities not only within the
specific context of the DMS–climate system, as mentioned above, over a 10-year time period, but also in the broader context of the current
uncertainties in the DMS–climate system and climate modelling.
Schematic diagram showing the ocean–atmosphere sulfur life cycle and
climate-relevant processes. Acronyms are defined as follows: sea surface
temperatures (SSTs), methane sulfonic acid (MSA), dimethylsulfoniopropionate
(DMSP), dimethyl sulfoxide (DMSO), dimethyl sulfide (DMS), cloud condensation
nuclei (CCN) and cloud droplet number (CDN).
Three simulations are performed to explore the chemical, aerosol and
meteorological implications of large DMSw perturbations. In the first
experiment, a control simulation is compared to a simulation in which all
DMSw is removed from the system to determine its current contribution
to the climate. In the second experiment, the control simulation is compared
to a simulation in which DMSw is significantly increased, and the
results are compared to that of the work by .
This paper is organised as follows: Sect. 2 outlines the methodology used
in this study, Sect. 3 evaluates how well ACCESS-UKCA performs with respect
to certain satellite products, Sect. 4 analyses the sensitivity of the
ACCESS-UKCA climate to large perturbations in DMS and Sect. 5 provides some
discussion and concluding remarks.
Methods
Model description and set-up
ACCESS-UKCA
The ACCESS-UKCA coupled climate–chemistry model is a platform from which the
influences of DMS on the large-scale climate can be evaluated. The physical
atmosphere in the ACCESS model is the United Kingdom Met Office's Unified
Model (UM). In this case, UM version 8.4 is used, in conjunction with the
UKCA chemistry model , which includes the GLObal Model of
Aerosol Processes (GLOMAP)-mode aerosol scheme described in Sect. .
Horizontal grid resolution is 1.25∘ latitude × 1.85∘ longitude,
with 85 vertical levels, where the model top is located at 85 km.
Anthropogenic emissions are prescribed pre-2000 from the Atmospheric
Chemistry and Climate Model Intercomparison Project (ACCMIP)
, and post-2000 from RCP6.0 . Biomass burning emissions are from the
GFED4s database . Emissions of other species required
by ACCESS-UKCA, and their original sources, including biogenic emissions,
chemical precursors and primary aerosol, are described in detail in
. DMS emissions are calculated within UKCA, and are
described in Sect. . Long-lived greenhouse gas
concentrations (e.g. CO2, CH4 and N2O) are
prescribed from the Coupled Model Intercomparison Project Phase 5 (CMIP5) and
RCP6.0 recommendations. Monthly mean SST and sea ice coverage are prescribed
as per the Atmospheric Model Intercomparison Project (AMIP)
. UKCA is coupled to the ACCESS radiation scheme via
O3, CH4, N2O and aerosol (direct scattering and
absorption). Aerosols further influence the large-scale cloud and
precipitation schemes via the cloud droplet number (CDN) concentration,
allowing changes in the chemical/aerosol fields to affect the
meteorology.
For this study, ACCESS-UKCA is run for the years 2000–2009, with a 1-year
spin-up, 1999. The simulations are nudged to ERA-Interim ,
using the horizontal wind component and potential temperature, at 6-hourly
intervals in the free troposphere. The use of nudging does not allow aerosol
and cloud responses to perturbed DMS to affect synoptic-scale meteorology;
hence the results here represent instantaneous responses in the climate
system. Nudging was deemed desirable for this study to limit computational
expense, allowing single runs of 10 years. Due to nudging, responses in the
simulation may be dampened, but can be attributed directly to the DMS
perturbations. The complicating effect of internal variability within the
modelled system is also avoided in nudged simulations.
GLOMAP
The GLOMAP-mode aerosol scheme uses two-moment pseudo-modal aerosol dynamics
to simulate aerosol size distributions . GLOMAP-mode
simulates particle compositions with sulfate, sea salt, elemental and organic
carbon in internally mixed modes . Dust is treated outside of
GLOMAP-mode according to the scheme detailed in .
Processes simulated within GLOMAP-mode include primary emissions, new
particle formation, particle growth by coagulation, condensation and cloud
processing and removal by dry deposition, and in-cloud and below-cloud
scavenging . New particle formation occurs via two mechanisms
in ACCESS-UKCA: free tropospheric binary homogeneous nucleation
and organic-mediated boundary layer nucleation
. The aerosol size distribution is represented in four
soluble modes (corresponding to nucleation, Aitken, accumulation and coarse
size modes) and one insoluble mode (Aitken). A full description of the scheme
can be found in , with improvements detailed in
. compare GLOMAP-mode with an older
generation aerosol scheme, finding significant differences in aerosol
response to perturbations between the two schemes.
DMSw climatology and flux parameterisation
The number of DMSw observations have increased dramatically over the
last 3 decades , although significant gaps,
both spatially and temporally, remain. use observations to
derive a gridded DMSw climatology, which is used in this study and
shown in Fig. a. The climatology shows
that high-latitude regions have the highest DMSw concentrations.
Significant sampling biases exist within the data set, with
approximately half of observations collected in late spring through summer,
and more than two-thirds of the data collected in the Northern Hemisphere.
The annual mean concentrations of (a) the
DMSw climatology and (b) the DMSw field of the
second experimental run, zonal maximum DMSw. Additionally, three
regions of interest are shown in (a) by red boxes: the Australian
region, 45–10∘ S, 110–160∘ E, the Southern Ocean (SO),
ocean grid points south of 40∘ S, and the south-eastern Pacific
(SEP), 240–270∘ E, 25∘ S–0∘.
The fluxDMS from the ocean to the atmosphere remains poorly
parameterised, has large variability in space and time and cannot easily be
measured. This subsequently causes large uncertainties in the
fluxDMS parameterisation. The most common flux parameterisations
exhibit considerable ranges in fluxDMS, from 15–35 Tgyr-1 of sulfur in to 9–34 Tgyr-1 of
sulfur found in , who recommend a range of
18–24 Tgyr-1 of sulfur as a reasonable estimate.
and show that current parameterisations
overestimate the fluxDMS at high wind speeds and suggest that
annual global fluxDMS is likely to be in the lower range of current
estimates. Of the fluxDMS parameterisations available in
ACCESS-UKCA, the method yields a low to moderate flux
comparable to those calculated in and ,
and is used in this study.
The DMSflux parameterisation is described in
Eq. (), where k, the piston velocity, is parameterised under
three wind-induced sea surface regimes: smooth (Eq. ) and rough
(Eq. ) gas transfer, and a wave-breaking/bubble-bursting regime
(Eq. ).
DMSflux=kDMSw-DMSaα=kDMSwα-DMSa
For w10 < 3.6 m s-1:
k=0.17w10SCDMS60023.
For 3.6 m s-1 < w10 < 13 m s-1:
k=2.85w10-3.6SCDMS60012+0.612SCDMS60023.
For w10 > 13 m s-1:
k=5.9w10-13SCDMS60012+26.79w10-3.6SCDMS60012+0.612SCDMS60023.
Here DMSw solubility α=11.4 at 26 ∘C, w10=10 m wind speed (m s-1) and the Schmidt number of DMS SCDMS, a
measure of viscosity/diffusion and a function of sea surface temperature, is
determined following the method of . The denominator in
this function is the Schmidt number of CO2 in fresh water at 20 ∘C, SCCO2=600, which is used to normalise the
numerator (SCDMS). We assume that the concentration of
DMSa is negligible, as it is orders of magnitude smaller than that
of seawater.
Model evaluation
In order to provide a climatological evaluation of ACCESS-UKCA, and to put
the sensitivity testing of DMS into a real-world context, a comparison to
observational data sets is presented. Global means at the surface are
calculated over the 2000–2009 period, except for the cloud climatologies
which were only available from 2006 to 2009.
The following global data sets were compared to the model output: low, medium
and high cloud fractions from the GCM-Oriented Cloud-Aerosol Lidar and
Infrared Pathfinder Satellite Observation Cloud Product (CALIPSO-GOCCP)
, radiation fluxes from the Clouds and the Earth's Radiant
Energy System (CERES) Energy Balanced and Filled (EBAF) TOA Edition 4.0
and CERES EBAF Surface Edition 4.0 and
precipitation from the Tropical Rainfall Measuring Mission (TRMM)
. Cloud fraction is defined according to the
CALIPSO-GOCCP: high between 50 and 440 hPa, medium between 440 and 680 hPa and low
between 680 and 1000 hPa. Direct comparison of cloud fractions between model
output and satellites cannot take into account satellite measurement biases,
which can be resolved using a cloud satellite simulator such as the Cloud
Feedback Model Intercomparison Program (CFMIP) Observation Simulator Package
(COSP). COSP was not available in the version of ACCESS-UKCA used here,
limiting the comparison. Nevertheless, a useful comparison is still possible.
DMS sensitivity testing
To explore the sensitivity of the global climate to large perturbations in
DMSw concentrations, two experimental simulations were performed and
compared to the control run (Ctl). As described above, the Ctl simulation
used the DMSw climatology, which is shown in Fig. a.
In Experiment 1 (Exp.1), DMSw was set to zero, leaving a flux of 0.72 Tg yr-1 of sulfur derived from terrestrial sources
for example. From this we can attribute what role
ocean-derived DMS plays in shaping our current climate and enhance our
understanding of how the physical processes underpinning the DMS–climate
system operate. This may further aid our understanding of how natural
aerosols interact with the global radiation budget. In Experiment 2 (Exp.2),
the DMSw field was set to each latitude's (at the model resolution of
1.25∘) monthly zonal maximum value, following a similar method to
, and shown in Fig. b. This
simulation allows further exploration of the physical processes by which DMS
can influence the climate, when the perturbations are exaggerated.
Three regions of interest are defined for their relevance to the broader
Australian community (for which ACCESS is purposed) or are of particular
interest in the DMS–climate system. They are the Australian region,
45–10∘ S, 110–160∘ E, the Southern
Ocean (SO), ocean grid points south of 40∘ S, and the south-eastern
Pacific (SEP) that represents an area of significant stratiform cloud decks,
240–270∘ E, 25∘ S–0∘.
Global energy budget
Due to the nudging of the model to ERA-Interim and the prescribed SSTs, a
direct estimate of how global temperatures might respond to DMS perturbations
is not possible. For this reason, a simple energy balance model has been used
to estimate the effects of the DMS perturbations on global mean temperatures,
a useful metric for comparison to some previous studies.
We have used the climate component of the Finite Amplitude Impulse Response
(FAIR) model. This model is based on the one first proposed by
and subsequently used in the most recent IPCC Assessment
Report 5 for equivalent emission metric calculations .
FAIR's climate component is a simple impulse response model which emulates
the behaviour of more complex Earth system models, given a certain radiative
forcing (in this case due to DMS). FAIR has been designed to determine
temperature responses to radiative forcing of similar magnitudes to the DMS
radiative effect . FAIR's temperature response is
calculated as the sum of two components, approximately representing the
response of the upper mixed layer and deep ocean to a change in radiative
forcing . Due to its simplicity, FAIR cannot capture the
non-linearities and feedbacks in the climate system, and hence the
temperature response calculated must be taken as an estimate only.
Furthermore, in this work we consider only a single mid-range estimate of
climate sensitivity.
For each experimental run, the radiative effect (RDMS) due to increasing
or decreasing DMSw is defined as the difference between the TOA
energy balance (Q*) of an experimental run from the Ctl, which can be taken
directly from ACCESS-UKCA. By providing this radiative effect to FAIR's
climate component, we can estimate the difference in temperature expected
across the 10-year period under zero DMSw or enhanced DMSw
conditions. In ACCESS-UKCA, an ensemble experiment would be required to
provide equivalent temperature difference estimates, which would be
computationally expensive.
To provide a measure of model uncertainty of the change in Q*, we have used a
moving block bootstrap . By selecting, with replacement,
blocks of size 2 (as determined by the time series auto-correlation) to
create 1000 alternate time series, we are able to provide the 10th and 90th
percentile confidence intervals of mean change in Q*. This is subsequently
translated into an uncertainty range for the change in temperature and flux
sensitivity estimates.
Model evaluation
This section compares selected ACCESS-UKCA fields to satellite-derived
observations. In order to give context to this evaluation, the ACCESS-UKCA
output is also compared to that of the CMIP5 (general circulation models, GCMs).
Cloud fraction
The 2006–2009 annual mean of the ACCESS-UKCA Ctl (a, d, g)
compared to the CALIPSO-GOCCP climatology (b, e, h), with the relative differences between the two shown in (c),
(f) and (i) (model – observations). Panels (a), (b) and (c) show
the low cloud fraction, (d), (e) and (f) show the middle cloud fraction and (g), (h) and (i) show the high cloud
fraction.
As for Fig. but for annual means for 2000–2009,
where (a), (b) and (c) show the TOA outgoing LW radiation, (d), (e) and (f) show TOA
outgoing SW radiation and (g), (h) and (i) show surface incoming SW radiation. The
observations are from the CERES EBAF TOA and Surface Ed. 4.0
; all units are in Wm-2.
The cloud fraction comparison is performed for the years 2006–2009, aligning
with the availability of CALIPSO-GOCCP data. ACCESS-UKCA simulates too little
low cloud fraction (Fig. a–c) over the majority of
the globe (mean bias of -0.16), which is consistent with findings for the
CMIP5 GCMs . Areas of large stratiform
cloud decks in eastern ocean basins are significantly underestimated, by a
fraction larger than 0.5, consistent with other CMIP5 and CFMIP Phase 1 and 2
findings . These low-level marine clouds
have an important impact on the global radiation budget and
have been identified as the primary source of uncertainty in tropical
cloud–climate feedbacks (e.g. the effects of the cloud albedo) in GCMs
. These biases have been attributed to poor vertical
distribution of clouds in the models, difficulty capturing overlapping cloud
layers, the misrepresentation of cloud structures, deficiencies with the
statistical parameterisation of clouds and likely problems in the cloud
microphysics . Low clouds over the polar regions and some
areas of northern Asia and North America are slightly overestimated. The
ACCESS-UKCA low-cloud biases over the Arctic are within the range of biases
found for the CMIP5 GCMs studied in . It should be noted
that satellite observations are subject to biases in detecting low clouds,
particularly over the Southern Ocean.
ACCESS-UKCA reproduces medium cloud fraction (Fig. d–f)
reasonably well, within ±0.1 in most regions (global mean bias of -0.01).
The largest discrepancies are overestimated medium cloud fraction over the
Southern Ocean and Antarctica, where the simulated medium cloud fraction is
at its highest globally. The Antarctic bias is of opposite sign to the CMIP5
models compared in . note that
issues within GCMs around distinguishing between clouds with tops at actual
mid-level and low-level clouds contribute to such biases. The biases in high
cloud fractions (Fig. g–i) show similar spatial patterns
to that of the low cloud fraction, where an underestimate occurs over most of
the tropics and mid-latitudes. The global mean bias is 0.05. The largest
negative biases, of up to 0.3, occur over the Maritime Continent. Moderate
overestimation is noted over the polar regions. These biases are within the
range of those found for the CMIP5 models studied in .
Interestingly, noted that due to underestimated low clouds in
the tropics, the CMIP5 models overcompensated by producing low clouds that are optically thick
and too bright and more high clouds, impacting the radiation
budget. Here, an underestimation of low clouds is also found, although there
is no evidence of an overcompensation of high clouds. Predominantly a small
underestimation of high cloud fraction is found in this simulation at
tropical to mid-latitudes (Fig. i).
Radiation
The remaining analyses consider means over the period of 2000–2009. The
comparison of the observed and simulated TOA outgoing LW radiation is shown
in Fig. a–c. The observed global mean of
239.7 Wm-2 is closely matched by the simulated
241.0 Wm-2. Compared to the CMIP5 ensemble, which tends to
underestimate TOA outgoing LW radiation, 238.6 Wm-2 from
and 238.9 Wm-2 from , TOA
outgoing LW radiation in ACCESS-UKCA is slightly overestimated. The regions
with the largest biases (both positive and negative) occur in regions of deep
convection (Fig. c), and align well spatially with
the biases in high cloud fractions shown in Fig. c.
Underestimation by -3 Wm-2 of TOA outgoing LW radiation occurs
over the polar oceans, which may partly be explained by an overestimation of
cloud fraction at all levels, and especially the mid-level clouds
(Fig. f) in this region.
Spatial biases in the TOA outgoing SW radiation
(Fig. d–f) are of greater magnitude than that of the
LW radiation. In most regions the sign of the outgoing SW radiation bias is
opposite to that of the LW radiation. The same processes as described above
that block LW radiation from escaping the atmosphere prevent SW radiation
reaching the surface, hence reflecting more sunlight and enhancing the
albedo. Globally, ACCESS-UKCA performs reasonably well, simulating the global
mean TOA outgoing SW radiation of 101.8 Wm-2 compared to the
observed 99.6 Wm-2, consistent with the multi-model mean of GCM
ensembles from previous studies . In
Fig. f, an abrupt change in sign of TOA outgoing SW
radiation at 60∘ S is found, which is also present in the CFMIP
comparisons . In the Southern Ocean, wrongly
assigned mid-level cloud types have been found to be a leading cause of the
model underestimation of TOA outgoing SW radiation .
In addition, poor representation of the physical processes surrounding
supercooled liquid water in the Southern Ocean has been found to account for
27–38 % of the total reflected solar radiation .
Over the Antarctic ice sheets, both TOA outgoing and surface incoming SW
radiation are overestimated, due to an underestimation of low clouds, which
allows the high albedo to reflect too much incoming SW radiation back out to
space.
The mean (2000–2009) annual total precipitation of (a) the
ACCESS-UKCA climatology (b) the satellite climatology from TRMM
and (c) the difference between ACCESS-UKCA and
the TRMM product (model – observations). Units are in mmyr-1.
Mean (2000–2009) values for the Ctl (a, d), the difference between Exp.1 (zero DMSw) and the Ctl
(b, e) and the difference between Exp.2 (zonally enhanced DMSw) and the Ctl (c, f).
Panels (a), (b) and (c) show the volume mixing ratio of SO2 in ppb; (d), (e) and (f) show the volume mixing ratio of H2SO4 in ppt.
Globally, ACCESS-UKCA overestimates incoming surface SW radiation
(Fig. g–i), with 202.4 Wm-2 compared to
observations of 198.3 Wm-2. This overestimation is slightly
greater than that found for CMIP5 GCMs of 2 ± 6 Wm-2
, though within their uncertainty. Nevertheless, large
regional biases of over ±30 Wm-2 exist. The most notable
features, apart from those discussed above, are too much incoming SW
radiation over the continents and the tropical regions, which can be
attributed in part to the underestimated cloud cover. The North Pacific
and North Atlantic oceans, the Arctic Ocean and parts of the Southern Ocean all
receive too little incoming SW radiation, consistent with overestimated cloud
cover.
Precipitation
Precipitation in ACCESS-UKCA has large positive biases in regions that
receive the most annual rainfall and align with the intertropical and South
Pacific convergence zones (ITCZ/SPCZ). These regions overestimate
precipitation by over 2000 mmyr-1. Poor performance of GCMs in
this region is not unusual however , with the current
CMIP5 GCM ensemble overestimating precipitation in a similar region by more
than 1000 mmyr-1 . found
that models in these regions produce light rain too frequently, indicating
that convective processes are poorly simulated. Two of Australia's CMIP5
GCMs, ACCESS 1.0 and 1.3, both overestimate precipitation in this region by
similar amounts to that of the ACCESS-UKCA model . If biases of
precipitation are considered as a percentage (not shown), the largest
differences occur in the eastern basins of the South Pacific (493 % over
the SEP region) and South Atlantic oceans (275 % from 0–25∘ S,
330–10∘ E).
DMS perturbations
This section aims to quantify the role of DMS in the large-scale climate
system. Two experimental simulations have been performed, described in
Sect. and Table , which
involve removing all DMSw (Exp.1) and setting the
DMSw to the zonal maximum (Exp.2).
Summary of the three global simulations presented in this study, the DMSw climatology used and the annual mean (2000–2009) total global fluxDMS.
Simulation
DMS climatology
FluxDMS (Tg yr-1 of sulfur)
Ctl
17.41
Exp.1
Zero marine DMS
0.72
Exp.2
Zonal maximum DMS from
37.05
Exp.1: zero DMSw
Chemistry response
The 2000–2009 annual mean ocean fluxDMS from ACCESS-UKCA is
17.41 Tgyr-1 of sulfur, resulting in an atmospheric DMS
(DMSa) annual mean surface concentration of 81.9 ppt. Taking all
marine DMS out of the model (but retaining the terrestrial source of
0.72 Tgyr-1 of sulfur) results in a 94 % reduction in
DMSa at the surface; throughout the troposphere, it results in a
98 % reduction of DMSa.
The vertical profiles of (a) SO2,
(b) H2SO4, (c) nucleation-mode number density,
(d) Aitken-mode number density, (e) accumulation-mode number density, (f) coarse-mode number density, (g) N3
nuclei number, (h) cloud condensation nuclei number and
(i) cloud droplet number. The solid lines represent the Ctl, dashed
lines shows Exp.1 and dotted lines show Exp.2. Blue lines show the SO (SO)
mean, red the Australian (Aus) region mean and green the south-eastern Pacific (SEP). All units are cm-3, apart from (a) ppb and
(b) ppt × 10-3.
The impact of this reduced fluxDMS on atmospheric sulfur can be
seen in Fig. a–b and d–e. Globally, there is a net decrease
of 15 % of SO2 at the surface. The largest absolute differences are
in the tropics and mid-latitudes over the oceans. Large relative decreases in
SO2 occur in the SO and SEP, of 84 and 94 % respectively.
Figure a shows the vertical profile of SO2 for the
Australian region (ref), the SO (blue) and the SEP (green). The large peak in
concentration at approximately 500 m occurring in the Australian
profile is attributable to industrial and energy-related emissions of
SO2, which is due to lofting by chimneys and smokestacks. The
SO2 in Exp.1 is consistently lower than that of the Ctl throughout
the troposphere, though for the regional means, the difference begins to
decrease closer to the tropopause.
Surface H2SO4 (Fig. d–e) shows significant loss
in predominantly clean marine areas; the SO has a 79 % decrease and the SEP
an 84 % decrease, compared to a 49 % global mean decrease. Interestingly,
heavily polluted regions, especially busy shipping lanes, undergo an increase
in H2SO4. H2SO4 is a precursor gas, which can participate
in the formation of secondary sulfate aerosol, or it can condense onto
pre-existing particles. The increased H2SO4 concentration in heavily
polluted regions results from a decreased condensational sink (not shown).
Similar non-linearities have been described in .
The vertical profiles of H2SO4 in Fig. b show
that the largest differences between Exp.1 and the Ctl occur in the free
troposphere (between 1 and 10 km) for all regions. In all three regions (each
considered a clean atmospheric environment), net decreases of H2SO4
occur.
Aerosol response
The majority of gaseous H2SO4 is taken up by aerosol formation
(99.99 %) as opposed to being removed by dry deposition (0.01 %)
. The peak in nucleation-mode number density in the free
troposphere in Fig. c coincides with the peak concentration
of H2SO4. Surface global nucleation-mode number concentration
decreases by 9 % between Exp.1 and the Ctl (see Fig. c).
While in absolute terms, clean terrestrial regions have the largest
decreases, the Australian region only has a relative decrease of 18 % in
nucleation-mode particles. Over the oceans, although few nucleation-mode particles exist, there are large relative differences of both signs.
In absolute terms, the differences in the aerosol number concentration are
greatest in the smaller aerosol modes, particularly the nucleation mode
described above. Figure d–f show the number concentrations
for the Aitken mode, accumulation mode and coarse mode (global maps not
shown). The Aitken mode (Fig. d) shows some differences
between the two simulations, with profiles reflecting reduced new particle
formation in the free troposphere and reduced condensation growth of
H2SO4 onto pre-existing particles in the boundary layer. The largest
differences are seen over the Australian region. Similar boundary layer
differences are also present in the accumulation mode, with the differences
between Exp.1 and the Ctl consistent below 1 km (Fig. e).
Little difference is seen in the coarse mode throughout the troposphere
(Fig. f), which in marine regions is dominated by sea salt.
As for Fig. , but where (a), (b) and (c) the
number concentration of N3 (condensation nuclei), (d), (e) and (f) show the number
of cloud condensation nuclei greater than 70 nm dry diameter and (g), (h) and (i) show the cloud droplet number concentration. All units are in
cm-3.
As aerosols grow towards the larger end of the Aitken mode, they become
relevant to cloud processes. Figure a shows the Ctl's
N3 (condensation) number concentration (N3 signifies particles with
a dry diameter greater than 3 nm). The difference in surface N3
number concentration between the Ctl and Exp.1 shows the largest relative
decreases occur in clean, coastal regions, predominantly in the Southern
Hemisphere, as well as some regions of the SO. In heavily polluted
terrestrial regions a small increase in the N3 number concentration
occurs. A decrease of 8 % is found globally. For the Australian region
(representative of a clean, terrestrial region), a decrease of 17 % is
found.
Over the SO a relative decrease of 39 % occurs at the surface. The SO and
the SEP have far fewer aerosols in all modes except the coarse mode (see
Fig. c–f), where sea salt dominates. This decrease in
number concentration in small aerosol modes represents a large portion of the
aerosol loading in the region. The increase in nucleation-mode particles is
reflected in the N3 for the SEP region, via a more moderate decrease of
20 %.
Global and hemispheric means of the CCN sensitivity to the
fluxDMS (as defined by ) in
both absolute (cm-3/mgm-2day-1) and relative terms, for Exp.1 and Exp.2.
Region
Exp.1 absolute
Exp.2 absolute
Exp.1 relative
Exp.2 relative
Global
16.9
12.4
0.048
0.036
SH
15.8
11.2
0.090
0.063
NH
18.6
14.5
0.029
0.023
Figure d–e show the number concentration of CCN with dry
diameters greater than 70 nm (CCN70) for the Ctl and the
differences resulting from Exp.1. The largest absolute differences are in the
tropics, which, similarly to the N3, have the highest concentration.
Relatively, there is a global decrease of 5 %, while decreases of 7 %
were found over the Australian region, decreases of 8 % over the SO and
decreases of 20 % over the SEP. Differences in CDN are shown in Fig. g–h. The relative differences in CDN
show a similar spatial pattern to that of the CCN. Global mean CDN decreases
by 5 %. A decrease of 5 % is also found for the Australian region,
whereas the SO shows an 8 % decrease, and the SEP shows an 18 % decrease.
In both the CCN70 and CDN, the marine Southern Hemisphere
mid-latitudes have the largest decreases of 14 % (averaged between
5 and 35∘ S) despite the SO having some of the larger decreases in
SO2 and H2SO4.
The larger differences in concentration of both CCN and CDN in the oceanic
Southern Hemisphere tropics and mid-latitudes, compared to the SO, warrant
further investigation of how sulfate aerosols are interacting with their
background environments, for example cloud processes and pre-existing
aerosols. The SO has large concentrations of sea salt particles, which, like
more polluted regions of the Northern Hemisphere, may provide a buffering
effect to reduced DMS-derived aerosols. Additionally, in areas of persistent
low cloud formation, in-cloud aqueous sulfate oxidation is the dominant
reaction (over gaseous nucleation), which allows almost instantaneous
condensational growth of existing aerosols, and is temperature-dependent. We
speculate that poor representation of low clouds in the SO may be having
further impacts on atmospheric composition modelling than currently realised.
A cloud resolving modelling study may be better suited to gain understanding
of the complex system described here.
Following the method of , global and
hemispheric sensitivities of CCN to fluxDMS have been calculated
(Table ). The results presented here suggest a
lower CCN sensitivity to fluxDMS compared to the
study where absolute sensitivities of 94 and
63 cm-3/mgm-2day-1 of sulfur were found globally for June and December
respectively. Similar CCN sensitivities are reported in the
study (63 cm-3/mgm-2day-1
global average). The lower sensitivities in our study are likely the result
of the large (near 100 %) changes in fluxDMS (the denominator).
Relative CCN sensitivities calculated here compare well with the
studies. For example
finds mean hemispheric relative CCN sensitivities of
0.02 for the Northern Hemisphere and 0.07 for the Southern Hemisphere. These
results highlight the greater relative importance of DMS in the Southern
Hemisphere.
Cloud, radiation and precipitation response
Meteorological responses to the DMS perturbations must be considered
carefully. As detailed in the methods section, the ACCESS-UKCA simulations
are nudged to ERA-Interim potential temperature and horizontal winds,
preserving synoptic-scale meteorology and limiting any feedbacks. While
performing a non-nudged simulation would allow the meteorology to respond to
changes in the chemistry and aerosol more freely, it would make comparison of
the aerosol and meteorological responses more difficult. Within ACCESS-UKCA,
GLOMAP-mode is directly coupled to the large-scale cloud and precipitation
schemes via the CDN , as well as the radiation scheme via
aerosols and some gases (see Sect. ).
Convective rainfall and cloud formation are not directly coupled to the
aerosol scheme, but can be indirectly influenced via changes in radiation
(which can act on temperature and moisture, etc.).
Comparisons of the low cloud fraction (as a percentage)
(a–b) and incoming SW radiation at surface (Wm-2)
(c–d) over the 2000–2009 period for Exp.1 (first column) and Exp.2
(second column) minus the Ctl. The absolute values for the Ctl of these
fields can be seen in Figs. and
.
Differences in low cloud fraction occur predominantly in areas with large
stratiform cloud decks (Fig. a). The largest differences
occur in eastern basins of the Southern Hemisphere's oceans. The SEP region
shows an annual mean decrease in low cloud fraction of 9 %. In the Northern
Hemisphere (including the north-eastern Pacific where significant stratiform
cloud decks are found) and the SO (where persistent low cloud formation
occurs) only small differences are evident, which may in part be due to the
modest differences in CCN and CDN concentrations discussed in
Sect. . Stratiform cloud deck fractions are
consistently underestimated by ACCESS-UKCA and other GCMs (see
Sect. ) in comparison to other areas of significant low
cloud formation such as the SO. The mechanism behind the different responses
(between the SO and cloud deck regions), and whether the long-standing model
biases, especially those around the formation of supercooled liquid water,
have contributed to the differing responses requires further investigation.
The decrease in low cloud fraction and aerosol number concentration discussed
above leads to an increase in surface incoming SW radiation
(Fig. c). This increase in surface SW radiation is highest
in the regions of stratiform cloud deck formation. In the SEP region there is
a mean increase of 7 Wm-2.
Comparisons of the (a–c) total liquid water at 1700 m height (Qcl), (d–f) large-scale rainfall and (g–i) convective rainfall over the 2000–2009 period.
The Ctl absolute values are shown in the first column, and respectively Exp.1 and Exp.2 minus the Ctl in the second and third columns. Units are in kgkg-1 and mmyr-1.
Decreases of total liquid water (Qcl) at 1700 m height shown in
Fig. a–b are found in the stratiform cloud deck
regions. Little difference in Qcl occurs at the surface. The
decrease in Qcl is coincident with increases in large-scale
precipitation in the stratiform cloud decks, regions with very little
precipitation (Fig. d–e). In the SEP region
large-scale rainfall increases by 17 mmyr-1 (15 %) over the
Ctl mean of 111 mmyr-1.
In the Southern Hemisphere stratiform cloud decks, and in particular the SEP
region, the model demonstrates a distinct cloud lifetime effect in response
to removing DMS in Exp.1. Decreased CDN concentration and the associated
increase in cloud droplet size and increased liquid water lead to increased
autoconversion and large-scale rainfall. The overall impact is to reduce low
cloud fraction.
Figure g–h show the differences in convective
rainfall. While the convective rainfall scheme is not coupled directly to
GLOMAP-mode, there are differences between the simulations. Convective
rainfall decreases in Exp.1 compared to the Ctl along the ITCZ (a mean difference of 11 mmyr-1 between
20∘ S and 20∘ N). This difference represents a small fraction
(less than 1 %) of the total convective rainfall. Relatively, (not shown)
the largest differences (a 5 % decrease in the SEP) are found once again in
eastern basins of Southern Hemisphere stratiform cloud decks.
note that even when convection schemes are coupled to an
aerosol scheme, the effects of CCN on convection, and the resultant
precipitation and associated maximum updrafts, differ significantly
depending on the cell type and size, making these effects difficult to
quantify. Large differences in convective rainfall would not be expected in
these results, due to the meteorological nudging used in the experiments.
Exp.2: zonally increased DMSw
This section considers the response to zonally enhanced DMSw,
resulting in a fluxDMS of 37.05 Tgyr-1 of sulfur
(relative to 17.41 Tgyr-1 of sulfur in the Ctl simulation). For
comparison the study used a zonally enhanced
fluxDMS of 46.1 Tgyr-1 of sulfur (up from
18.2 Tgyr-1 of sulfur) under global warming scenarios. Many of
the differences resulting from zonally enhancing DMSw show
similar spatial patterns, with a similar magnitude but reversed sign compared
to Exp.1.
Globally, the differences in SO2 (Fig. c) are of
comparable magnitude to Exp.1. Increased SO2 concentrations occur
over the Australian region, the SO and the SEP: increases of 42, 172 and 89 %
respectively. There is a net decrease in H2SO4 of 14 % in
Australia, and a larger decrease over the tropical oceans. Over the SO there
is an increase of 9 %, while in the SEP a decrease of 37 % occurs.
Similar non-linearities are discussed in terms of doubled DMS in the
study. These differences in SO2 and H2SO4
are also clear in the vertical profiles shown in Fig. a–b.
Differences in the aerosol modes (see Fig. c–f) are of
a similar magnitude but opposite sign to those noted in
Sect. . Global mean N3, CCN70 and CDN
increase by 6, 4 and 5 % respectively (Fig. c, f, i).
Larger differences are seen over the SO of 27, 15 and 13 % and the SEP of
14, 19 and 17 % for N3, CCN70 and CDN respectively.
Globally, there is little difference in low cloud fraction or
Qcl, though increases are noted in regions of large stratiform
cloud decks (Fig. d), which show similar spatial patterns
to that of Exp.1. Incoming surface SW radiation has a global mean decrease of
-1.75 Wm-2. This decrease is comparable to the
finding of -2.2 Wm-2 (noting the larger
DMS perturbation by ). Lastly, decreases in large-scale
precipitation are found, again in regions of stratiform cloud decks
(Fig. f), while general increases in convective
precipitation over the tropical oceans occur (Fig. i).
The study, which analysed a warming climate, also found
large relative decreases in precipitation rate predominantly in eastern ocean
basins. We find, under the current climate, the largest relative increase in
total precipitation (not shown) in the southeast basins of the Pacific and
Atlantic oceans; however these results presented here are much nosier than the
results. find that artificial
enhancement of DMS may offset global warming, which is supported by this
study as implied by the decreases in incoming SW radiation at the surface;
however the precipitation responses warrant further study.
Temperature response
Summary of the global mean (2000–2009) radiation fields: absolute
Ctl values for the TOA shortwave (SW) and longwave (LW) outgoing and Q*
and the differences in these quantities resulting from Exp.1 and Exp.2 (from
the Ctl) as well as the FAIR temperature response. The ranges shown for
Exp.1 Ctl and Exp.2 Ctl indicate the 10th and 90th percentile confidence
intervals. n/a – not applicable
Simulation
TOA SW ↑ (Wm-2)
TOA LW ↑ (Wm-2)
Q* (Wm-2)
FAIR response (K)
Ctl absolute values
101.79
241.04
-1.35
n/a
Exp.1-Ctl
-1.82 (-2.62 to -1.23)
0.13 (0.04 to 0.20)
1.69 (1.13 to 2.43)
0.45 (0.30 to 0.64)
Exp.2-Ctl
1.57 (1.08 to 2.48)
-0.12 (-0.20 to -0.04)
-1.45 (-2.33 to -0.98)
-0.38 (-0.61 to -0.26)
The global 2000–2009 mean of the TOA radiation budget (Q*) and its main
components are provided in Table , along with the
relevant confidence intervals derived from the bootstrapping technique. Due
to the nudging used in the simulations, we do not expect the TOA Q* to be
balanced (i.e. Q*=0). The differences in Q* seen in Exp.1 and Exp.2,
1.69 Wm-2 (1.13 to 2.43, 10th and 90th confidence intervals) and
-1.48 Wm-2 (-2.33 to -0.98, 10th and 90th confidence
intervals) respectively, show a substantial radiative effect of DMS on the
energy budget. The Q* response found for Exp.1 is consistent with the
findings of 1.79 Wm-2. Using the FAIR
model's climate component, the 2000–2009 mean temperature response is
calculated to be 0.45 K (0.30 to 0.64, 10th and 90th percentile
range) for Exp.1 and -0.38 K (-0.26 to -0.61, 10th and 90th
percentile range) for Exp.2.
Other studies generally consider DMS changes under global warming and we can
make comparisons via the sensitivity of the estimated global temperature
response to changes in the fluxDMS (see
Table ). In this study, we find a response of
0.027 K per Tgyr-1 of sulfur in Exp.1, and
0.019 K per Tgyr-1 of sulfur in Exp.2. These results are
of similar magnitude to the study (0.029 K per
Tgyr-1 of sulfur) and in the range of the lowest impact scenario
of (0.03–0.060 K per Tgyr-1 of sulfur).
The other scenarios in the study (0.046–0.11 K per
Tgyr-1 of sulfur) suggest much higher temperature sensitivities
to changes in fluxDMS, as does the study
(0.041 K per Tgyr-1 of sulfur).
The estimated temperature response to perturbations in the
fluxDMS (K per Tgyr-1 of sulfur) for the
current study's experiments (Exp.1 and Exp.2) and those found in the
literature. The ranges shown for Exp.1 and Exp.2 indicate the 10th and 90th
percentile confidence intervals.
Experiment
K per Tgyr-1 of sulfur
Exp.1
0.027 (0.018 to 0.038)
Exp.2
0.019 (0.013 to 0.031)
0.041
– low pH impact scenario
0.03–0.060
– medium pH impact scenario
0.046–0.096
– high pH impact scenario
0.051–0.11
0.029
Discussion and conclusions
The ACCESS-UKCA chemistry–climate model, which includes a detailed
microphysical aerosol module, has been evaluated against satellite
observations of cloud fraction, radiation and precipitation and subsequently
used to conduct sensitivity experiments to determine the role of DMS in
several aspects of the climate system.
Important considerations when using climate models include the inherent
uncertainties associated with all climate simulations, e.g. emissions
uncertainties (both natural and anthropogenic), parameterisations and
physical representations of atmospheric processes. Nevertheless, where clear
shortcomings have been found in comparison to the satellite-derived
observations, the ACCESS-UKCA model has been found to perform with comparable
skill to current CMIP5 GCMs. Additionally, it is important to note biases in
the satellite products themselves, for example in cloud fraction retrievals
as noted in or .
Of particular interest, our evaluation of ACCESS-UKCA shows an
underestimation of large stratiform cloud decks located in the eastern
mid-latitude basins of the Earth's oceans. These regions of extensive low
cloud produce little rainfall (that is overestimated by the model) and are
often regions of high primary productivity. These biases have not been
attributed to a single cause (multiple theories have been proposed, as
discussed in Sect. ), indicating a gap in
understanding of atmospheric processes in these regions .
Globally, removing or enhancing DMSw from the climate system
leads to significant responses in chemistry and aerosol concentrations. While
changes in meteorological parameters (low cloud fraction, large-scale
precipitation, moisture, radiation) are largest in the Southern Hemisphere
stratiform cloud decks, global mean differences were small. We find that DMS
in these stratiform regions plays an important role in cloud processes and
precipitation suppression (as discussed in or with
regards to anthropogenic pollution in ). Furthermore,
we have demonstrated that marine DMS is responsible for increasing low cloud
fraction in stratiform cloud deck regions, a demonstration of the second
aerosol indirect (or lifetime) effect . These results
indicate that a greater understanding of natural aerosols and their
interaction with cloud processes (both via observations and modelling
studies) in these regions may improve model representation, as it is these
regions that show considerable model bias in comparison to observations.
In other regions of significant low cloud formation (SO, Northern Hemisphere
cloud decks), aerosol sources such as sea salt and anthropogenic aerosols may
buffer the regions from changes in DMS-derived aerosols. Additionally, in the
SO, representation of cloud processes in global climate models is poor,
especially with respect to supercooled liquid water
. It is likely that these biases are misrepresenting
the DMS–climate interactions in these regions.
By nudging these simulations, the model response to the DMS perturbations is
limited to fast (aerosol and cloud) changes. We suggest that free-running
ensemble experiments are performed to gain insight into the
aerosol–cloud–climate processes occurring in regions of significant DMS
influence. Such experiments should focus on improving microphysical aspects
of aerosol–cloud interaction in these regions (and how it differs among
regions) or improving the representation of aerosols, in particular natural
aerosols.
Previous studies examining the role of DMS in the climate system have not
identified stratiform cloud decks as regions of particular importance.
Instead, these studies focused on cloud feedbacks in the SO
. estimated the global TOA
radiative effect of DMS to be 1.79 Wm-2, which is consistent
with our results (1.69 Wm-2), but slightly lower than the
estimation of 2.03 Wm-2 estimated by (who
used the previous DMSw climatology).
In this study, we find the estimated temperature responses per unit change in
DMS-derived sulfur flux (0.027 or 0.019 K per Tgyr-1 of
sulfur) are lower than those reported in the
(0.046–0.096 K per Tgyr-1 of sulfur, medium-impact
scenario) and (0.041 K per Tgyr-1
of sulfur) studies. The temperature response sensitivities calculated here
are comparable to those given in (0.029 K per
Tgyr-1 of sulfur). Without further information, it is difficult
to speculate on the cause of the discrepancy between the results presented
here and those in and . However, the
discrepancy between these results suggests the need for better observational
constraints, and highlights the complexity of the DMS–aerosol–cloud system.
Natural aerosols account for a significant source of uncertainty in climate
modelling and radiation budgets . Our study uses the
DMSw climatology with the flux
parameterisation. Though this data set and method are commonplace for
DMS–climate studies, both are limited by sparse observations and
uncertainties . For example, recent studies have indicated
that coral reefs produce significant amounts of DMS, and are an unaccounted
for source of sulfur . Furthermore,
larger concentrations and/or fluxes of DMS than what we currently consider
have also been found at the poles, especially around sea ice and polynyas
.
Observational deficiencies become particularly relevant when considering the
stratiform cloud deck regions. In the data set, the SEP
region contains only two ship campaigns collecting measurements in January
and February. The cloud deck in the Southern Hemisphere eastern basin of the
Indian Ocean has no DMSw observations. The higher susceptibility of
cloud and precipitation to changes in DMS in these regions suggest that they
should be a priority for future atmospheric composition field campaigns.
To place the conclusions of this study into a broader perspective, we must
consider the DMS–climate system within the context of anthropogenic climate
change despite the uncertainties discussed above and in
Sect. . As discussed in the introduction and above, a better
understanding of current global DMS is essential before future scenarios can
be considered with certainty. Nevertheless, ,
and have suggested that global
production of DMS by marine phytoplankton is vulnerable to ocean
acidification, amongst other oceanic changes expected with global warming,
for example impacts on nutrient availability, salinity, mixed layer depths
and light penetration . While both the and
temperature responses are much larger than those found
here, our results imply a 25 % decrease in fluxDMS would result
in an increase of 0.12 K (0.07 to 0.16, 10th and 90th confidence intervals)
globally. Considering the current Paris Agreement target of limiting global
warming to 1.5–2.0 K, the sensitivity of ocean-derived sulfate aerosol to
warming temperatures and ocean acidification becomes important. Strategies to
mitigate anthropogenic climate change must consider not only the effect of
increased CO2 on temperatures, but also on ocean pH. Mitigating only
temperature increases (e.g. via solar radiation management) may have short-term
cooling effects; however, without removing CO2 from the
atmosphere, ocean acidification will continue to impact marine life, and as
demonstrated here, the climate.