Atmos. Chem. Phys., 7, 1693–1706, 2007
www.atmos-chem-phys.net/7/1693/2007/
© Author(s) 2007. This work is licensed
under a Creative Commons License.
Atmospheric
Chemistry
and Physics
Solar cycle variations of stratospheric ozone and temperature in
simulations of a coupled chemistry-climate model
J. Austin1 , L. L. Hood2 , and B. E. Soukharev2
1 NOAA
2 Lunar
Geophysical Fluid dynamics Laboratory, PO Box 308, Princeton, NJ 08542-0308, USA
and Planetary Laboratory, University of Arizona, Tucson, AZ 85721, USA
Received: 30 October 2006 – Published in Atmos. Chem. Phys. Discuss.: 27 November 2006
Revised: 31 January 2007 – Accepted: 21 March 2007 – Published: 28 March 2007
Abstract. The results from three 45-year simulations of a
coupled chemistry climate model are analysed for solar cycle influences on ozone and temperature. The simulations
include UV forcing at the top of the atmosphere, which includes a generic 27-day solar rotation effect as well as the
observed monthly values of the solar fluxes. The results are
analysed for the 27-day and 11-year cycles in temperature
and ozone. In accordance with previous results, the 27-day
cycle results are in good qualitative agreement with observations, particularly for ozone. However, the results show
significant variations, typically a factor of two or more in
sensitivity to solar flux, depending on the solar cycle.
In the lower and middle stratosphere we show good agreement also between the modelled and observed 11-year cycle results for the ozone vertical profile averaged over low
latitudes. In particular, the minimum in solar response near
20 hPa is well simulated. In comparison, experiments of the
model with fixed solar phase (solar maximum/solar mean)
and climatological sea surface temperatures lead to a poorer
simulation of the solar response in the ozone vertical profile,
indicating the need for variable phase simulations in solar
sensitivity experiments. The role of sea surface temperatures
and tropical upwelling in simulating the ozone minimum response are also discussed.
1
Introduction
Variations in the solar output have long been thought to influence climate (e.g. Hershel, 1801) but rigorous links to the troposphere have only relatively recently been established (e.g.
Baldwin and Dunkerton, 1989; Coughlin and Tung, 2004).
While the energy output of the sun (total solar irradiance,
TSI) clearly has a significant impact on the mean climate
Correspondence to: J. Austin
([email protected])
on energetic grounds, variations in TSI are too small to lead
to a large signal in tropospheric processes by direct radiative forcing alone. Solar variability is strongly dependent on
wavelength so that even though TSI variability is a fraction
of a % (e.g. Frölich, 2000), the variability increases substantially at the shorter wavelengths which affect ozone production and loss. A plausible mechanism for solar influences on
the atmosphere, therefore, might invoke changes in stratospheric temperature and ozone due to ultraviolet (UV) variations. Such mechanisms have been used by Haigh (1994)
for example to demonstrate the possible existence of a solar
impact on the troposphere using model simulations.
Solar cycle variations in both ozone and temperature are
now well established in the stratosphere from observations
(e.g. Hood and McCormack, 1992) and from model simulations (e.g. Brasseur, 1993). The most well-known solar variation is probably the 11-year Schwabe cycle, which
has been observed in the sunspot record for over a century (Schwabe, 1843). Solar variations also occur on much
shorter timescales, particularly in the range 13–80 days due
to harmonics and subharmonics of the 27-day solar rotation
period (e.g. Zhou et al., 1997; Hood and Zhou, 1998).
To observe small changes associated with the solar cycle requires a high density of data points in space and time.
Away from the Earth’s surface, satellite data provide the best
opportunity of detecting such signals, but only two complete
11-year solar cycles are available from this source. Many
more cycles exist for the solar rotation periods. These have
now been analysed in detail and reliable information exists on both temperature and ozone (Hood, 1986; Hood and
Cantrell, 1988; Chandra et al., 1994; Chen et al., 1997).
Modelling studies have shown good agreement with observations for the 27-day cycle (Brasseur, 1993; Fleming et
al., 1995; Williams et al., 2001), which has similar spectral
changes in UV flux than the Schwabe cycle. For the 11-year
cycle, even though ozone column amounts are correctly simulated (e.g. Zerefos et al., 1997), models have been consistent
Published by Copernicus GmbH on behalf of the European Geosciences Union.
1694
J. Austin et al.: Solar cycle variations of ozone and temperature
Table 1. List of model experiments with AMTRAC, simulating
the past. For each simulation, the date indicated refers to the
corresponding amounts for all the external forcings, including the
WMGHGs and halogen amounts.
Name
Dates
Solar Forcing
Description
TRANSA
TRANSB
TRANSC
1960-2005
1960-2005
1960-2005
Transient
Transient
Transient
Transient forcings, A
Transient forcings, B
Transient forcings, C
SL2000
2000
F10.7 =147
2000
F10.7 =222
Fixed mid-cycle
Control run, 31 years
Fixed solar max.
Control run, 26 years
SL2000B
in simulating too much ozone in the middle stratosphere, and
not enough in the lower and upper stratosphere (Brasseur,
1993; Haigh, 1994; Shindell et al. 1999; Tourpali et al.,
2003; Egorova et al., 2004). Despite improvements in models, including the use of 3-D coupled chemistry-climate models (e.g. Labitzke et al., 2002) these differences persist. A
recent synthesis of observations and comparison with model
results is given by Soukharev and Hood (2006). To date, the
simulations have typically been run in solar maximum versus
solar minimum mode, which is of course not how the atmosphere behaves. This is to minimise the computational expense of multi-decadal simulations with 3-D models. Therefore, there is a possibility either that the full solar cycle needs
to be represented, or that there are missing processes in many
of the simulations completed. Early work (Callis et al., 2001,
and references therein) have suggested that NOx generation
in the upper mesosphere due to energetic electron precipitation (EEP) may be one such missing process and recent
works (Langematz, et al., 2005; Rozanov et al., 2005a) have
addressed this using updated models. These events are expected to occur more during solar minimum and in the model
simulations the extra NOx is advected to the middle stratosphere where ozone can be depleted. As a consequence of
the ozone “self healing” effect additional ozone is generated
in the lower stratosphere. Thus, the ozone total column could
be relatively unaffected by this process but the vertical profile
of the ozone change would be improved in comparison with
measurements. However, the amount of NOx inserted into
the mesosphere in some model studies appears to be more
than can be expected by EEP events alone, and questions remain as to how sufficient NOx can be advected to the tropical middle stratosphere. Indeed, recent estimates (Hood and
Soukharev, 2006) tend not to support the presence of a significant impact of solar cycle NOx on tropical ozone. In this
work, 27-day and 11-year solar cycle variations of ozone and
temperature are investigated using transient simulations of a
coupled chemistry-climate model with observed forcings.
Atmos. Chem. Phys., 7, 1693–1706, 2007
2
Model simulations
Results are presented from simulations of the coupled chemistry climate model AMTRAC (Atmospheric Model with
TRansport And Chemistry). The simulations have been described and results have been presented previously for ozone,
water and age of air (Austin and Wilson, 2006; Austin et
al., 2007), and contributed to the model intercomparison experiment REF1 of Eyring et al. (2006). The model includes
comprehensive stratospheric chemistry but simplified tropospheric chemistry, including mainly methane chemistry. The
concentrations of well mixed greenhouse gases (WMGGs)
and organic halogen molecules have been specified from observations for the period 1960–2005, and provide radiative
forcings to the model climate. Sea Surface Temperatures and
Sea Ice (collectively referred to as SSTs) are specified at the
model lower boundary as a function of time from the Hurrell dataset (Hurrell et al., personal communication, 2005).
Observed aerosol extinctions are included to represent the
impact of volcanic eruptions. Monthly averaged solar forcing is specified from observations and a 27-day variation in
solar forcing is superimposed as a sine wave. The model
does not simulate a quasi-biennial oscillation (QBO), nor
is one forced externally. Twelve complete solar cycles of
model results are available from the three ensemble runs, together with 26 years or more each of fixed phase solar maximum/solar mean timeslice experiments. See Table 1 for the
list of experiments completed.
Solar cycles are included in the model in both the radiation and the photochemistry, but in different ways. In the radiation, the spectrally varying solar flux changes monthly in
accordance with the calculations of Lean et al. (2005) but the
27-day solar rotation period was not included, because the
large changes in the radiation code could not be included on a
timely basis. The radiation scheme consists of 18 short wave
bands, as described in more detail in Anderson et al. (2004)
and references therein. In the photochemistry, solar variability is included by a linear parameterisation of the 10.7 cm
flux, including a monthly mean term and a term representing
the 27-day solar rotation period.
The photolysis rates are calculated using a lookup table for
solar maximum and solar minimum conditions. For each of
the two solar phases, the 27-day amplitude was calculated using Upper Atmosphere Research Satellite SOLar STellar Irradiance Comparison Experiment (UARS SOLSTICE) data
(Rottman et al., 1993; Rottman, 1999), thus requiring four
data points for each photolysis coefficient, and for each of the
independent parameters of the lookup table (pressure, column ozone, solar zenith angle). Computation of the lookup
table required solar flux data for the 158 bands of the photolysis rate code for periods close to solar maximum and solar minimum. The corresponding 10.7 cm flux values were
then used in the parameterisation (below). Model sensitivity to the 11-year solar cycle is later analysed in terms of the
10.7 cm flux, and the 27-day oscillation is analysed for solar
www.atmos-chem-phys.net/7/1693/2007/
the photolysis rate and increment respectively for the 10.7 cm
flux and the 27-day oscillation, which are determined by
del simulations
J10.7 = Jmin + (F10.7 − 69.6)(Jmax − Jmin )/154.3
(2)
Solar flux at 10.7cm
but the vertical profile of the ozone change would
model results are available from the three ensemble runs, toroved in comparison with measurements. However,
gether with 26 years or more each of fixed phase solar maxount of N Ox inserted into the mesosphere in some
imum/solar mean timeslice experiments. See Table 11695
for the
J. Austin
Solar than
cycle variations
of ozone and
studies appears
toetbeal.:more
can be expected
by temperature
list of experiments completed.
ents alone, and questions remain as to how sufficient
maximum only in terms of the 205 nm flux implied in the
300
an be advected
to thedata.
tropical middle stratosphere. InSOLSTICE
ecent estimates
andrate
Soukharev,
Each(Hood
photolysis
is given by2006) tend not
250
ort the presence of a significant impact of solar cycle
J = c(J10.7 + J27 sin φ)
(1)
n tropical ozone. In this work, 27-day and 11-year soWhere
c is a factor
to correct for seasonal
changes in the sune variations
of ozone
and temperature
are investigated
200
earth distance and φ is the phase of the 27-day oscillation,
ransient simulations of a coupled chemistry-climate
assumed to be a sine wave with largest amplitude at solar
with observed
forcings.
maximum
in accordance with observations. J10.7 and J27 are
150
100
J27 =from
J27 min
+ (F10.7 − 69.6)(J
max − J27
min )/154.3 (3)
are presented
simulations
of the27coupled
chem50
imate model
AMTRAC
(Atmospheric
Model
with
Jmax and Jmin are the photolysis rates for solar maximum
and solar minimum
respectively. have
F10.7 is
the 10.7
port And Chemistry).
The simulations
been
de- cm solar
0
flux
corresponding
to
the
model
time,
which
is
linearly
in1950
1960 1970 1980 1990 2000
and results have been presented previously for ozone,
terpolated between consecutive monthly mean observed valYear
nd age of air (Austin and Wilson, 2006; Austin et al.,
ues. J27 max and J27 min are the photolysis rate increment for
and contributed
to the
model intercomparison
the 27-day
oscillation.
The quantities 69.6 experand 154.3 in the
Fig. 1. Observed monthly mean solar flux at 10.7 cm wavelength
REF1 of Eyring
et
al.
(2006).
includes
line). The broken line shows an approximate sinusoidal fit
above equations refer to theThe
10.7 model
cm flux value
at solar miniFig. 1. (solid
Observed
monthly mean solar flux at 10.7 cm wavethrough the data with an 11-year period. The data were obtained
mum, and the chemistry
amplitude ofbut
the simplified
solar cycle appropriate
the
hensive stratospheric
tropo- to length
(solid
line).
brokenData
lineCenter
shows
an approximate sifrom the NOAA The
Geophysical
(http://www.ngdc.noaa.
ratemainly
computations.
Aschemistry.
indicated byThe
e.g. Rottman
chemistry,photolysis
including
methane
nusoidalgov/stp/SOLAR/ftpsolarradio.html).
fit through the data with an 11-year period. The
the amplitude of the 27-day oscillation varies subtrations of(1999)
well mixed
greenhouse gases (WMGGs)
data were obtained from the NOAA Geophysical Data Center
stantially over the course of the 11-year Schwabe cycle and
anic halogen
molecules
have
been
specified
from
ob(http://www.ngdc.noaa.gov/stp/SOLAR/ftpsolarradio.html).
such a variation is incorporated into Eq. (3). We do not ex3 Analysis of model data
ons for theplore
period
1960-2005,
the physics
of this,and
but provide
point out radiative
that the oscillation
arisesclimate.
from movements
of sunspots
around theand
rotating sunSolar cycles are included in the model in both the radis to the model
Sea Surface
Temperatures
In analysing of the model results, we use the linear regression
which
peak during
maximum
the Schwabe
cycle, when
ation and
the photochemistry,
but in different
ways.1999),
In the
(collectively
referred
to astheSSTs)
are of
specified
at the
algorithms
of the National Algorithm
Group (NAG,
higher numbers of sunspots are present. The 10.7 cm radio
flux is a useful proxy as detailed measurements exist for over
5 solar
cycles.
In this work,
Chem. Phys.,
0000,
0001–15,
2007we investigate the 27-day cycle only at the times of the maxima in the Schwabe cycle.
Figure 1 shows the observations of 10.7 cm flux (F10.7 ) since
1950 at monthly resolution and for comparison a sinusoidal
function with 11-year period, mean 145 and amplitude 75
units is drawn in the figure. While the sinusoidal function
fits the general behaviour of the flux values, some cycles are
much stronger than others, and in general, the flux is slightly
squarer in shape than the sinusoid. No additional solar impacts such as EEP effects are included in the model.
In addition to the above experiments, a 31-year control run
(SL2000) was completed with fixed seasonally varying SSTs
and fixed WMGGs, no volcanic aerosol and a fixed mid-cycle
solar forcing. An additional 26 year simulation (SL2000B)
was also completed for the year 2000 identical to the control but with the 10.7 cm flux increased by 75 units, corresponding to the flux in January 2002. To reduce the model
spinup required, Experiment SL2000B was initialised with
the results five years after the start of Experiment SL2000.
Therefore, results for Experiments SL2000 and SL2000B are
compared for only 26 years. For comparison with the transient runs, results for the difference SL2000B–SL2000 are
rescaled to 100 units of F10.7 .
www.atmos-chem-phys.net/7/1693/2007/
implemented on the GFDL high performance computing system. For the 27-day oscillation, the regression equation assumed was www.atmos-chem-phys.org/acp/0000/0001/
O3 =a0 + a1 t + a2 σ + a3 F10.7 + a4 sin(φ−λ) + (t)
(4)
where t is time in days, σ is the aerosol surface area estimated from the optical depth and λ is an assumed phase lag.
is the residual term. Model results from one year periods
were used. To minimise seasonal effects, the ozone values
in Eq. (4) were treated as a time series of 13.5 27-day periods, which were averaged to give a single 27-day sequence
of values. Calculations were performed for the 61 values of
the phase lag −30 to +30 days. This method was applied to
the model results for one calendar year corresponding to the
peaks in each of the four Schwabe cycles – 1970, 1981, 1991
and 2001, for each of the three ensemble members.
For the Schwabe cycle, a similar regression equation was
assumed, where the ozone values were regressed separately
for each month and for the annual average.
O3 =b0 + b1 t + b2 σ + b3 F10.7 + (t)
(5)
where O3 is the monthly averaged ozone amount from the
model and t is now in months. Equation (5) was applied for
Atmos. Chem. Phys., 7, 1693–1706, 2007
John Austin et al.: Solar cycle variations of ozone and temperature
J. Austin
5
et al.: Solar cycle variations of ozone and temperature
27-day lag correlation1981
27-day lag correlation1970
0
0.
0
-0.4
0.0
0.2
0.6
0.6
0.6
0.6
10.0
0.00
-0.4
3.0
0.
2
10.0
0
00.0
30.0
30.0
-30
-20
-10
0
10
20
(UV Lags O3)
(UV leads O3)
30
0
3.0
30.0
-30
00
.0
10.0
-20
-10
0
10
20
(UV Lags O3)
(UV leads O3)
0
0
30.0
30
-30
0.
00
0
0.6 0.20.0
Pressure hPa
0.6
0.2
1.0
0
0
0.2
0.6
10.0
0.2
30
-0.4
1.0
0.6
-20
-10
0
10
20
(UV Lags O3)
(UV leads O3)
-0.40 0.0
0.2 .6
0
0.3
-0.8
-0.4
0.3
0.2
0
27-day lag correlation2001
0.1
-0.8
.4
-0.0
00.2
Pressure hPa
27-day lag correlation1991
0
0
0.0
0
-30
0.1
3.0
0.6
.2
00.0
3.0
1.0
0.6
1.0
0.0
0.0 0
0
0.3
Pressure hPa
4
Pressure hPa
00.2
-0.
0
0.
0.
2
0.3
0
0.1
0.0
0.1
0.6
1696
-20
-10
0
10
20
(UV Lags O3)
(UV leads O3)
30
◦ to 30
◦
Fig. 2. Lag correlations
ozone
andozone
UV and
averaged
over the
range
27 oscillation
day oscillation
for the years
Fig. 2. Lagbetween
correlations
between
UV averaged
over latitude
the latitude
range30
30◦S
S to 30◦ N N
forfor
the the
27 day
for the years
indicated for theindicated
four solar
cycles.
Thecycles.
contour
is 0.2.is 0.2.
for the
four solar
Theinterval
contour interval
and the 1981 values lower than observed. The values at zero
the full 45 yearlagmodel
period,
usinga wide
monthly
(Figuresimulation
4, right panel)
also covers
rangeavand are
erage values for
eachin of
the three
members.
similar
magnitude.
Theensemble
values for 1970
are very The
different
to the
years,toparticularly
near 30 hPa. and
This Hood
is likely to
regression model
isother
similar
that of Soukharev
due todifference
an interfering
rather
thancm
a solar
(2006), but thebemain
is dynamical
that heresignal
we use
10.7
impact itself.
flux as the independent solar forcing term, as the model itA similar
for temperature
yieldsAlso,
the results
shown
self was driven by
these analysis
flux values
(Eqs. 1–3).
as the
Figurea5tropical
(cf. Williams
et al., 2001,
5).term
The tempermodel does notinhave
oscillation,
theFigure
QBO
of
ature response is very small, typically less than 0.1K per %
Soukharev andchange
Hoodin(2006)
is not included.
205 nm flux. It is questionable whether the signal
Finally, the analysis
wassignificant
repeatedinusing
simulated
tem- it
is statistically
muchthe
of the
domain, although
perature as theappears
dependent
in thefor
regression
to be variable
slightly positive
maximum analysis.
correlation. A
4
4.1
Results
clear difficulty faced by the modelling technique used here,
in which all the known processes are included, is separating small signals from a dataset rich in features. This is of
course the problem also faced in the analysis of observations.
In contrast, the simulations of Williams et al. were focused
The 27-day oscillation
www.atmos-chem-phys.org/acp/0000/0001/
Results are presented of the sensitivity coefficient a4 /O3 for
a one percent change in solar flux at 205 nm wavelength.
Results are also presented for the correlation coefficient between O3 and sin φ, i.e. the correlation between O3 and the
27-day oscillation. We repeat the presentation of Williams
et al. (2001) with the AMTRAC results and a somewhat
more extensive result database. Williams et al. used an early
version of the Unified Model with Eulerian TRansport and
Chemistry (UMETRAC) (Austin et al., 2000). Since that
Atmos. Chem. Phys., 7, 1693–1706, 2007
more precisely on the 27-day oscillation since background
time,
thein chemistry
has undergone
severalaverage
iterations
trends
WMGGs, halogens
or indeed monthly
so- before
reaching
the current
version
which
was devellar flux were
not imposed.
Also,(AMTRAC)
inclusion of the
radiative
effect
the 27-day
enhanced
tempera- et al.,
oped
forof the
GFDLoscillation
climatelikely
model
AM2the(Anderson
ture signal in Williams et al. Rozanov et al. (2006) also have
2004).
a much larger signal, by about a factor of two compared with
Figure
2 shows
that obtained
here. the 27-day lag correlation for ozone for the
four cycles, averaged for the region 30◦ S to 30◦ N. The reFor 1991 the model results below 5 hPa are in good agreesults
arewith
similar
to thosefrom
presented
by and
Williams
et al., Fig. 2.
ment
measurements
MLS (Hood
Zhou, 1998).
Williams
noted
significant
differences
At 2 hPaet
andal.above
the that
observed
signal increases
to aboutoccurred
0.08%
1%change
change inwas
205applied
nm flux from
MLS.radiation
This is theand phowhen
theperUV
to both
closest
year
to
the
analysis
year,
but
there
is
a
large
spread in
tochemistry, as opposed to just the photochemistry
as applied
the model results from the different solar cycles. For the late
here.
However,
we
also
find
substantial
variations
between
1970s, Hood (1986) and Hood and Cantrell (1988) suggest
onethat
solar
cycle temperature
and the next.
the lag
at the corthe 27-day
solar For
cycleexample
peaks at about
0.06%
per 1%peak
change
in 205 nmdownwards
flux. The upper
stratospheric
peak
relation
increases
between
1 and
10 hPa for
the observations
averages
aboutnot
twice
the model
value
theinyear
1981 and 2001
butatdoes
change
substantially
for
1970 and 1991.
Atmos. Chem. Phys., 0000, 0001–15, 2007
The results for 2001 are in good agreement with Zhou
et al. (1997), their Fig. 7, which covers the period October
1991–September 1994. However in our results, the region
of high correlation extends above the stratopause. Zhou et
al. note that the correlation between ozone and UV changes
during the period. This is explored further in Hood and Zhou
(1998) who separate the results into two 500 day periods.
The first half of the period yields a lag correlation more similar to our results for 1991 while the second half yields results
www.atmos-chem-phys.net/7/1693/2007/
J. Austin et al.:6 Solar cycle variations of ozone and temperatureJohn Austin et al.: Solar cycle variations of ozone and temperature
27-day lag correlation1981
27-day lag correlation1970
0.1
0.1
-0.4
1.0
3.0
0
0.2
-0. 0.0
4
0.0
Pressure hPa
0 .2
0 .2
0
0.2
-0.4
00.
0
-20
-10
0
10
20
(UV Lags T)
(UV leads T)
30
27-day lag correlation1991
27-day lag correlation2001
0.2
-0.4
-30
-20
-10
0
10
20
(UV Lags T)
(UV leads T)
Pressure hPa
0.2
00
0.
0.0
0.2
30.0
30
-30
-0.4
10.0
0
0
0
3.0
0.0
30.0
1.0
0.2
0.0
-0.4
10.0
0. 0
-0.4 0 .02
0.6
-0.4
3.0
0 0.2
0.3
0.2
00.0
.2
1.0
30
0.1
0
0
0.3
0.0 0
-20
-10
0
10
20
(UV Lags T)
(UV leads T)
0
0.1
0.2
0
-30
0.0
-30
30.0
0
30.0
0
-0
00.0.4
10.0
0.0
-0.4
Pressure hPa
10.0
0.2
0
.0
3.0
0
0.3
0.0
1.0
00
0
0.3
Pressure hPa
1697
-20
-10
0
10
20
(UV Lags T)
(UV leads T)
30
◦
◦ ◦ N for the 27 day oscillation for the years
Fig. 3. Lag correlations
temperature
and UVand
averaged
over over
the latitude
range
Fig. 3. Lagbetween
correlations
between temperature
UV averaged
the latitude
range30
30◦ SStoto3030
N for the 27 day oscillation for the years
for the
four solar
The interval
contour interval
indicated for theindicated
four solar
cycles.
Thecycles.
contour
is 0.2.is 0.2.
averaged over all the solar cycles. This is probably because
more similar toof our
resultsofforthe2001.
Asforcing,
noted by
Hood
the neglect
radiative
as well
as and
possibly
Zhou, the lackother
of aunidentified
significant
27-day
solar
signal response
impacts,
on the
temperature
signal.
above 2 hPa in Microwave Limb Sounder (MLS) data (Waters, 1989), may
related
to Schwabe
the fact cycle
that the majority of the
4.2beThe
11-year
ozone soundings were taken at night and therefore cannot reFigurevariations.
6 shows theInannual
meanthe
response
of ozone
spond to day time
contrast,
3-D model
is toa the
solar cycle
in % UT,
per 100
unitsisofvery
F10.7nearly
, together
with the
strict zonal average
at 00:00
which
a diur2 σ uncertainty. A statistically significant solar ozone signal average. Observations
of the solar signal in 1979–1980
nal is computed only in the low and middle latitude middle
(Hood, 1986; Hood
and
Cantrell,
1988) are
to lower
that of
and upper stratosphere, extending
to similar
the tropical
stratothe MLS data,sphere.
exceptAnthat
in thisfeature
case of
a significant
important
these results iscorrelathe low latitude minimum
hPa. Generally,
the results
tion between ozone
and theresponse
27-daynear
UV20signal
is present
in
are similarwhere
to thosethe
observed,
example,
the SBUV data
the lower mesosphere
SBUVfor
data
vary diurnally.
from Soukharev and Hood (2006), or the results of Randel
In general, though,
the model response is more dependent on
and Wu (2007), but there is too much model variability to
the period analysed
the observations.
For example,
the
discern than
a definitive
solar cycle in a large
part of the domain.
1981 results are about 50% less than observed for the period
November 1980
to September
analysed2007
by Hood and
Atmos.
Chem. Phys.,1982
0000, 0001–15,
Cantrell.
Lag correlation results for temperature are shown in Fig. 3,
which may be compared with Williams et al. (2001), Fig. 4.
The results are different in many respects from their Run B,
which like our simulations also excludes the 27-day cycle in
the radiative terms, and indeed look more similar to their Run
A, which includes the radiative term. AMTRAC has larger
correlations than UMETRAC (Williams et al.), possibly bewww.atmos-chem-phys.net/7/1693/2007/
The annual mean temperature response as a function of
cause
the and
temperature
more 7.difficult
to detect
and
pressure
latitude is signal
shown inis Figure
The signal
is
requires
more
27 day above
cycles70than
were simulated
statistically
significant
hPa throughout
the low with
and UMEmiddle latitudes. In high latitudes the lower stratospheric sigTRAC.
nal is opposite to the upper stratospheric signal, although not
The
agree significant.
well with Within
the observations
of
all theresults
featuresdo
are not
statistically
the latHood
Zhou
This
mayresults
be due
the small size
itudeand
range
30◦ S (1998).
to 30◦ N the
model
are to
in reasonwithsignal,
the solarmaking
cycle computed
from Stratoof able
the agreement
temperature
the detection
of a coher(SSU) dataand
(Scaife
et al.,
2000),particularly
but
entspheric
signalSounding
in both Unit
observations
model
results
outside this latitude range, observations are less reliable and
challenging. The temperature would be expected to lag the
the model diverges from the observations. For the remainder
UVofby
greater
amount than
the ozone,
because
of the time
thisapaper,
we concentrate
primarily
on the tropical
regions
taken
fortheasolar
radiative
to the and
ozone.
This
where
signal inresponse
both temperature
ozone can
be is more
more clearly
clearly
visibledefined.
in the results for 1970 and 1991 but not present
in 1981
and cycle
2001,response
and not
clearly
The solar
in the
modelapparent
is analysedin
bythe
con-observations.
sidering the term b3 /O3 from the regression analysis. Figure
8 shows
the vertical
profiletoofthe
these
results,
for each
The
ozone
sensitivity
205
nm flux
as a season,
function of altitude is presented in Fig. 4 (cf. Williams et al., 2001, Fig. 3).
Each individual www.atmos-chem-phys.org/acp/0000/0001/
Schwabe cycle is very different, although at
the phase of maximum correlation, each cycle indicates a
peak in the upper stratosphere with a minimum, near zero
response in the lower stratosphere. This is similar to the
results of Williams et al. for maximum correlation, but the
AMTRAC results decrease more substantially with altitude
in the upper stratosphere. The peak in the very low stratosphere, below 50 hPa in the results of Williams et al. (2001),
is not simulated. In the AMTRAC results, the signal to noise
Atmos. Chem. Phys., 7, 1693–1706, 2007
1698
John Austin et al.: Solar cycle variations of ozone and temperature
J. Austin
Ozone Sensitivity at maximum correlation
Ozone Sensitivity at zero lag
0.3
3.0
0.3
1970
1981
1991
2001
MLS
1.0
Pressure hPa
Pressure hPa
1.0
7
et al.: Solar cycle variations of ozone and temperature
3.0
10.0
10.0
30.0
30.0
-0.4 -0.2
0.0 0.2 0.4 0.6
% O3/% 205 nm flux
0.8
1.0
1970
1981
1991
2001
-0.4 -0.2
0.0 0.2 0.4 0.6
% O3/% 205 nm flux
0.8
1.0
Fig. 4. Ozone sensitivity
in % for each % change in model 205 nm flux for the 27-day solar rotation period for the time at which the
Fig. 4. Ozone sensitivity in % for each % change in model 205 nm flux for the 27-day solar rotation period for the time at which the
correlation between
ozonebetween
and UV
is aand
maximum
correlation
(left(left
panel)
and
lag(right
(right
panel).
Derived
from
correlation
ozone
UV is a maximum
correlation
panel)
andatatzero
zero lag
panel).
Derived
values values
from MLS
dataMLS
(Hooddata (Hood
1998) for
period 201991
Oct. to
1991
15 July
1994
given by
by the
TheThe
results
have been
over the latitude
and Zhou, 1998)and
forZhou,
the period
20the
October
15toJuly
1994
arearegiven
theblack
blackline.
line.
results
haveaveraged
been averaged
over the latitude
◦
◦ N30and
range
S toover
30◦ Nalland
over ensemble
all three ensemble
members.
errorbars
barsindicate
indicate 95%
intervals
from the
linearthe
regression
range 30◦ S to 30
three
members.
TheThe
error
95%confidence
confidence
intervals
from
linear regression
analysis.
analysis.
Temp Sensitivity at zero lag
Temp Sensitivity at maximum correlation
Pressure hPa
Pressure hPa
0.3 not imposed. Also, inclusion of the radiative effect of
were
the 27-day oscillation likely enhanced the temperature signal
in Williams
et al. Rozanov et al. (2006) also have a much
1970
1.0
1981by about a factor of two compared with that
larger signal,
obtained 1991
here.
2001
3.0
For
1991 the model results below 5 hPa are in good agreement with measurements from MLS (Hood and Zhou, 1998).
At10.0
2 hPa and above the observed signal increases to about
0.08% per 1% change in 205 nm flux from MLS. This is the
closest year to the analysis year, but there is a large spread in
30.0
the30.0
model results from the different solar cycles. For the late
The observed ozone sensitivity from Hood and Zhou
1970s,
Hood
(1986)
and Hood
and
Cantrell (1988) suggest
-0.2
0.24).
-0.2
-0.1
0.0
0.1
0.2
(1998) peaks at about 0.4% at
5 hPa-0.1
(black0.0curve0.1in Fig.
Temp K/% 205 nm flux
Temp
K/% 205 nm solar
flux cycle peaks at about 0.06%
that
the
27-day
temperature
The model simulations for 1991 and 2001 are in good agreeper 1% change in 205 nm flux. The upper stratospheric peak
ment with these values, with the 1970 values too high and
in
observations
at about
twice
the model
value
Fig.
5.
Temperature
sensitivity
in
K
for
each
%
change
in
model
205
nm the
flux for
the 27-day solaraverages
rotation period
for the time
at which
the
the 1981 values lower than observed. The values at zero lag
correlation between ozone and UV is a maximum correlation (left panel)
and at zeroover
lag (right
panel).
Derived
values from
MLS
data
(Hood because
averaged
all
the
solar
cycles.
This
is
probably
(Fig. 4, right panel)
also
covers
wide20range
andto are
similar
and Zhou,
1998)
for theaperiod
Oct. 1991
15 July
1994 are given by the black line. The results have been averaged over the latitude
ofbars
theindicate
neglect
of the radiative forcing, as well as possibly
◦
◦
S to 30for
N and
overare
all three
members.
95% confidence intervals from the linear regression
in magnitude. range
The 30values
1970
veryensemble
different
to theThe error
other
unidentified
impacts, on the temperature signal.
analysis.
other years, particularly near 30 hPa. This is likely to be due
0.3 this part of the atmosphere, so the
ratio became very small in
results are not plotted. The likelihood is that the Williams et
al. (2001) result there is1.0not a1970
real effect for the same rea1981
son. Rozanov et al. (2006) also
found a marked variation
in the sensitivity to the year of1991
simulation in a sequence of
2001
3.0
individual 1-year simulations.
Their mean results are also
MLS
in good agreement with observations, but in the mesosphere
their ozone sensitivity is10.0
larger than the results obtained here.
This could be a result of the simplified mesospheric chemistry in the current work.
to an interfering dynamical signal rather than a solar impact
4.2 The 11-year Schwabe cycle
averaged over all three ensemble members and averaged over
mum, so the values in Figure 8 should be multiplied by 1.25 itself.
the latitude range 25◦ S to 25◦ N. For a typical solar cycle,
A similar analysis
for temperature
the results
the difference
between solaryields
maximum
and solarshown
minimum
is 125-150et
units,
on 5).
the definition
of solar maxiin Fig. 5 (cf. Williams
al., depending
2001, Fig.
The temperature
response is very small, typically less than 0.1 K per % change
in 205 nm flux.www.atmos-chem-phys.org/acp/0000/0001/
It is questionable whether the signal is statistically significant in much of the domain, although it appears
to be slightly positive for maximum correlation. A clear difficulty faced by the modelling technique used here, in which
all the known processes are included, is separating small signals from a dataset rich in features. This is of course the
problem also faced in the analysis of observations. In contrast, the simulations of Williams et al. were focused more
precisely on the 27-day oscillation since background trends
in WMGGs, halogens or indeed monthly average solar flux
Atmos. Chem. Phys., 7, 1693–1706, 2007
1.50 for a full solar cycle. However, even on a monthly mean
Figure
6 cm
shows
thesolar
annual
mean
of ozone to the
the 10.7
flux at
maximum
canresponse
vary substantially.
Eachcycle
season in
has %
qualitatively
providing
confi- with the
solar
per 100similar
unitsresults
of F10.7
, together
2 σ uncertainty. A statistically significant solar ozone signal is computedAtmos.
only in
the Phys.,
low and
Chem.
0000,middle
0001–15,latitude
2007 middle
and upper stratosphere, extending to the tropical lower stratosphere. An important feature of these results is the low latitude minimum response near 20 hPa. Generally, the results
are similar to those observed, for example, the SBUV data
from Soukharev and Hood (2006), or the results of Randel
and Wu (2007), but there is too much model variability to
discern a definitive solar cycle in a large part of the domain.
The annual mean temperature response as a function of
pressure and latitude is shown in Fig. 7. The signal is
www.atmos-chem-phys.net/7/1693/2007/
J. Austin et
Fig. 4. Ozone sensitivity in % for each % change in model 205 nm flux for the 27-day solar rotation period for the time at which the
correlation between ozone and UV is a maximum correlation (left panel) and at zero lag (right panel). Derived values from MLS data (Hood
and Zhou, 1998) for the period 20 Oct. 1991 to 15 July 1994 are given by the black line. The results have been averaged over the latitude
range 30◦ S to 30◦ N and over all three ensemble members. The error bars indicate 95% confidence intervals from the linear regression
al.:analysis.
Solar cycle variations of ozone and temperature
Temp Sensitivity at zero lag
Temp Sensitivity at maximum correlation
0.3
3.0
0.3
1970
1981
1991
2001
MLS
1.0
Pressure hPa
Pressure hPa
1.0
1699
3.0
10.0
10.0
30.0
30.0
-0.2
-0.1
0.0
0.1
Temp K/% 205 nm flux
0.2
1970
1981
1991
2001
-0.2
-0.1
0.0
0.1
Temp K/% 205 nm flux
0.2
Fig. 5. Temperature
K for each
in model
205 205
nm nm
flux
forforthe
solarrotation
rotation
period
theat time
Fig. 5.sensitivity
Temperatureinsensitivity
in K %
for change
each % change
in model
flux
the 27-day
27-day solar
period
for thefor
time
whichat
thewhich the
correlation between
ozonebetween
and UV
is aand
maximum
correlation
(left panel)
andand
at atzero
(rightpanel).
panel).
Derived
values
fromdata
MLS
data (Hood
correlation
ozone
UV is a maximum
correlation
(left panel)
zerolag
lag (right
Derived
values
from MLS
(Hood
1998) for
period 20
Oct. to
1991
15 July
1994
givenby
bythe
the black
black line.
results
havehave
been been
averaged
over theover
latitude
and Zhou, 1998)and
forZhou,
the period
20the
October
1991
15 toJuly
1994
areare
given
line.The
The
results
averaged
the latitude
◦
◦ N30and
range
S toover
30◦ Nalland
over ensemble
all three ensemble
members.
error
barsindicate
indicate 95%
95% confidence
intervals
from from
the linear
range 30◦ S to 30
three
members.
The The
error
bars
confidence
intervals
theregression
linear regression
8 analysis.
John Austin et al.: Solar cycle variations of ozone and temperature
analysis.
3.0
10.0
0
1.00 0.5
1.0
30.0
00
Atmos. Chem. Phys.,1.0000, 0001–15, 2007
3.0
1.50
Pressure hPa
www.atmos-chem-phys.org/acp/0000/0001/
0
1.0
1.0
mum, so the values in Figure 8 should be multiplied by 1.25 0.3a full solar cycle. However, even on a monthly mean
1.50 for
the 10.7 cm flux at solar maximum can vary substantially.
.50 providing confiEach season has qualitatively similar0results
Pressure hPa
averaged over all three ensemble members and averaged over
0.3
the latitude range
25◦ S to 25◦ N. For a typical solar cycle,
the difference between solar maximum and solar minimum
0.50
is 125-150 units, depending
0 the definition of solar maxi0.5on
10.0
30.0
2.
00
1..5
00
0
0
100.0
-90 -60 -30 0 30 60 90
Latitude
100.0
-90 -60 -30 0 30 60 90
Latitude
Fig. 6. Simulated
Left panel: Simulated
annual ozone
mean ozone
responseinin%
% per
per 100
of of
10.710.7
cm flux
a function
of pressure
latitude.and latitude.
Fig. 6. Left panel:
annual mean
solarsolar
response
100units
units
cm as
flux
as a function
of and
pressure
panel: 2 σ uncertainty
in the derived
solar response.
contourinterval
interval isis0.25%
in both
panels.
Right panel: 2Right
σ uncertainty
in the derived
solar response.
TheThe
contour
0.25%
in both
panels.
Pressure hPa
0.1
0.1
0.1
3
0.1
3
ure 8 shows the vertical profile of these results, for each
season,0.3
averaged over all three ensemble members and averaged over the latitude range 25◦ S to 25◦ N. For a typical
1.0 the difference between solar maximum and sosolar cycle,
lar minimum is 125–150 units, depending on the definition
of solar3.0
maximum, so the values in Fig. 8 should be multiplied by 1.25–1.50 for a full solar cycle. However, even
on a monthly
mean the 10.7 cm flux at solar maximum can
10.0
vary substantially. Each season has qualitatively similar results providing confidence in the results, but the most strik30.0
ing feature is the local minimum in the solar response near
20 hPa. The demonstration of a statistically significant mini0.1
mum100.0
response does not appear
in any other published model
-90 (e.g.
-60 -30
0 et
30al.,60
90 Tourpali et al., 2003;
results to date
Shindell
1999;
Latitude
Egorova et al., 2004).
Rozanov et al. (2005b) simulated
a slight minimum response in this region, but they did not
0.3
Pressure hPa
0.5
0.3
0.1
0.3
statistically significant above
70 hPa throughout the low and
middle latitudes. In high latitudes the lower stratospheric sig1.0
nal is opposite to the upper stratospheric
signal, although not
all the features are statistically significant.0.5Within the latitude range 30◦ S to 30◦ N 3.0
the model results are in reasonable agreement with the solar cycle computed from Stratospheric Sounding Unit (SSU)
10.0data (Scaife et al., 2000), but
outside this latitude range, observations are less reliable and
the model diverges from the observations. For the remainder
30.0
of this paper, we concentrate primarily on the tropical regions
0.3 ozone can be
where the solar signal in both temperature and
more clearly defined. 100.0
-90 model
-60 -30
0 30 by
60con90
The solar cycle response in the
is analysed
Latitude
sidering the term b /O from the regression
analysis. Fig-
Fig. 7. Left panel: Simulated annual mean temperature solar response in K per 100 units of 10.7 cm flux as a function of pressure and
latitude. Right panel: 2 σ uncertainty in the derived solar response. The contour interval is 0.1K in both panels.
www.atmos-chem-phys.net/7/1693/2007/
dence in the results, but the most striking feature is the local
minimum in the solar response near 20 hPa. The demonstration of a statistically significant minimum response does
Atmos. Chem. Phys., 7, 1693–1706, 2007
The annual mean ozone response (Figure 9) compares well
with the observations of Soukharev and Hood (2006). While
the observational analysis implies quite large uncertainties,
Latitude
Latitude
Fig. 6. Left panel: Simulated annual mean ozone solar response in % per 100 units of 10.7 cm flux as a function of pressure and latitude.
Right panel: 2 σ uncertainty in the derived solar response. The contour
interval iset
0.25%
both panels.
J. Austin
al.: inSolar
cycle variations of ozone and temperature
Pressure hPa
0.5
3.0
10.0
10.0
0.1
30.0
0.1
30.0
3.0
0.3
Pressure hPa
1.0
0.5
0.3
0.1
1.0
0.3
0.1
0.3
0.1
1700
0.3
100.0
-90 -60 -30 0 30 60 90
Latitude
0.1
100.0
-90 -60 -30 0 30 60 90
Latitude
Fig. 7. Left
panel: Simulated
annual mean
temperature
solarresponse
response in
units
of 10.7
a function
of pressureof
andpressure and
Fig. 7. Left panel:
Simulated
annual mean
temperature
solar
in KKper
per100100
units
of cm
10.7flux
cmasflux
as a function
2 σ uncertainty
in the derived
response.The
The contour
contour interval
is 0.1K
in K
both
latitude. Rightlatitude.
panel:Right
2 σ panel:
uncertainty
in the derived
solarsolar
response.
interval
is 0.1
inpanels.
both panels.
JJA
Atmos.
Chem. Phys., 0000, 0001–15, 2007
10.0
SON
30.0
100.0
-2
-1
0
1
O3 %/100 units F107
2
3
Fig. 8. Simulated seasonal mean ozone solar response in % per
100 units of 10.7 cm flux. The results have been averaged over the
latitude range 25◦ S to 25◦ N and over all three ensemble members.
The
error bars seasonal
indicate 95%
confidence
intervals
from theinlinear
Fig. 8. Simulated
mean
ozone solar
response
% per
regression
analysis.
100 units of 10.7 cm flux. The results have been averaged over the
latitude range 25◦ S to 25◦ N and over all three ensemble members.
The error bars indicate 95% confidence intervals from the linear
comment
on the feature or provide any analysis to confirm
regression
analysis.
whether it was realistic or an aspect of random interannual
variability.
The annual
ozone response
(Fig.in9)the
compares
well
nal variations
startmean
to become
important
upper stratowith
the
observations
of
Soukharev
and
Hood
(2006).
While
sphere and mesosphere and the observations represent a biobservational
quite
uncertainties,
ased, the
day-time
solar analysis
impact, implies
whereas
thelarge
model
results inthe satellite instruments all yield qualitatively similar results
clude with
night
time
values.
Generally,
there
are
significant
disa minimum in the pressure range 10–25 hPa. Its precise
crepancies with SBUV data, but the low vertical resolution
of those instruments may have contributed. There may be
Atmos. Chem. Phys., 7, 1693–1706, 2007
some sensitivity to the period analysed but in further investigations it was found that the error bars in the measurements
were larger than any difference between solar cycles. Tour-
Pressure hPa
3.0
9
The annual mean ozone response (Figure 9) compares well
location
depends on
factors such
as data
resolution.
In genwith the observations
of Soukharev
and Hood
(2006).
While
the observational
quite large
eral,
the modelanalysis
agreesimplies
reasonably
welluncertainties,
with HALOE data,
the satellite
all yield
qualitatively
similar results
but
doesn’tinstruments
reproduce
the O3
upper
stratospheric
results
of the
solar
cycle
%/100
units
with a minimum
in the pressure range 10-25 hPa. Its precise
0.3
SAGE
data.
As
in
the
case
of
the
27-day
cycle,
ozone
diurlocation depends on factors such as data resolution. In gennal
start to
becomewell
important
in thedata,
upper stratoeral,variations
the model agrees
reasonably
with HALOE
1.0reproduce
sphere
and
mesosphere
andstratospheric
the observations
a bibut doesn’t
the upper
results ofrepresent
the
SBUV
1979-2003
SAGE
data.
As in
the case
of the 27-day
cycle,the
ozone
diur- results inased, day-time
solar
impact,
whereas
model
3.0
clude night time values. Generally, there are significant discrepancies with
SBUV data, but the low vertical resolution
10.0 www.atmos-chem-phys.org/acp/0000/0001/
of those instruments may have contributed. There may be
some sensitivity
to the period analysed but in further investi30.0
gations it was found that the error bars in the measurements
were larger
100.0 than any difference between solar cycles. Tourpali et al. (2007)
show
between
SBUV
data
-2
-1 comparisons
0
1
2
3 and
%/100 units
F10.7
Umkehr data. The two datasetsO3
generally
agree
well and in
most cases suggest that theO3
minimum
tropical
response
solar in
cycle
%/100
unitsoc0.3 lower than in the satellite data, at approximately
curs slightly
25 km (20 hPa), close to the minimum in the model solar re1.0
sponse. Tourpali
et al. also show small horizontal movements
SAGE 1985-2003
in the ozone features,
as a function of season of the year, al3.0
though in
view of the small statistical significance of these
results we have not analysed the model results in this detail.
10.0
The temperature response to the solar cycle is shown in
Fig. 10.30.0
The signal is very similar in all four seasons, and
generally increases to the upper stratosphere. The signal also
reflects100.0
the ozone signal with a local minimum in the signal
-2 This is-1consistent
0 with other
1
3
near 20 hPa.
work2 (for example
O3showed
%/100 units
F10.7
Shibata and Kodera, 2005), which
that the
UV effect
on temperature was dominant
hPa, but
the ozone
O3 above
solar 5cycle
%/100
unitseffect was 0.3
dominant below 5 hPa. Model results obtained here
agree well with Stratospheric SSU data (reanalysis of results
1.0 et al., 2000) throughout the observed range, but
from Scaife
1992-2003
are generally HALOE
smaller than
observed. Analysis of observed
3.0
temperatures to try to extract the small solar signal is extremely challenging, and other observational analyses give a
10.0
different estimate for the solar signal (e.g. Labitzke et al.,
4
5
4
5
4
5
Pressure hPa
dence in the results, but the most striking feature is the local
minimum
the solarcycle
response%/100
near 20 hPa.
The demonO3insolar
units
0.3 stration of a statistically significant minimum response does
not appear in any other published model results to date (e.g.
Shindell et al., 1999, Tourpali et al., 2003, Egorova et al.,
2004). Rozanov et al. (2005b) simulated a slight minimum
DJF
in this region, but they did not comment on the fea1.0 response
ture or provide any analysis to confirm whether it was realistic or MAM
an aspect of random interannual variability.
Pressure hPa
Pressure hPa
John Austin et al.: Solar cycle variations of ozone and temperature
30.0
www.atmos-chem-phys.net/7/1693/2007/
100.0
-2
-1
0
1
2
O3 %/100 units F10.7
3
Pressure hPa
Pressure hPa
JohnetAustin
et al.:cycle
Solarvariations
cycle variations
of ozone
and temperature
J. Austin
al.: Solar
of ozone
and temperature
O3 solar cycle %/100 units
0.3
2002; Crooks and Gray, 2005). However, these analyses
used data assimilation fields which are unreliable for low
DJF
frequency variability
Randel, personal communication,
1.0 (W.
2006). The model simulations presented in Labitzke et al.
MAM
also gave a larger temperature solar signal than the results
3.0
presented here, by aboutJJA
a factor of 2. However, none of
the simulations show the tropical ozone minimum response
SONthe temperature response from
which would tend
10.0to reduce
ozone in the lower to middle stratosphere.
30.0
5
Analysis and results for fixed solar phase
www.atmos-chem-phys.net/7/1693/2007/
O3 solar cycle %/100 units
0.3
1.0
SBUV 1979-2003
3.0
10.0
30.0
100.0
-2
-1
0
1
2
O3 %/100 units F10.7
3
4
5
4
5
4
5
O3 solar cycle %/100 units
0.3
100.0
-2
that the
Pressure hPa
1.0
SAGE 1985-2003
3.0
10.0
30.0
100.0
-2
-1
0
1
2
O3 %/100 units F10.7
3
O3 solar cycle %/100 units
0.3
1.0
Pressure hPa
-1 response
0
2 of the
3
It is plausible
tropical
is 1an artifact
O3 %/100 units F107
statistical analysis or that the ozone sensitivity may be a nonlinear function of the solar phase. Hence data from solar
maxima and minima may give extreme or different results.
To consider this, we repeat the previous model ozone analFig. 8.9) Simulated
seasonal
in % per
ysis (Fig.
using model
datamean
fromozone
just solar
thoseresponse
years corre100
units
of
10.7
cm
flux.
The
results
have
been
averaged
over the
sponding to solar maxima
and minima (1964, 1970, 1976,
latitude range 25◦ S to 25◦ N and over all three ensemble members.
1981, 1986, 1991, 1996, 2001). Considering all three ensemThe error bars indicate 95% confidence intervals from the linear
ble members
12 years for solar maximum and 12 years
regressiongives
analysis.
for solar minimum. Results for the domain 25◦ S to 25◦ N
are shown in Fig. 11 for the annual average, together with
the results
using allstart
solar
phases. important
The results
areupper
in agreenal variations
to become
in the
stratoment sphere
below and
20 hPa,
and
above
2
hPa
but
diverge
from
each
mesosphere and the observations represent
a bisolar impact,Despite
whereaslarger
the model
results inother ased,
in theday-time
middle stratosphere.
uncertainties
clude night
time values.
Generally,
there are
significant
disin examining
a limited
sample
of the period,
it would
seem
crepancies
with SBUV data,
but the
vertical
resolution
that the
lower stratospheric
minimum
is low
a robust
feature
of
thoseresults.
instruments
may have
contributed.
may be
these of
model
Nonetheless,
with
the analysisThere
presented
sensitivity
to the
analysed
furthertoinvestihere, some
it would
seem that
theperiod
differences
in but
the in
middle
upgations it was
that the significant,
error bars inbut
theitmeasurements
per stratosphere
arefound
statistically
should be
were that
largerthe
than
any difference
cycles.
Tourcautioned
analysis
is basedbetween
on justsolar
4 solar
cycles.
pali et al. (2007) show comparisons between SBUV data and
Processes not included in the regression analysis, such as
Umkehr data. The two datasets generally agree well and in
transport and chemical changes that cannot be represented
most cases suggest that the minimum in tropical response ocas simple trend terms, may have contributed to these differcurs slightly lower than in the satellite data, at approximately
ences.25Tokmobtain
a clearer
the in
solar
(20 hPa),
closesignal
to the during
minimum
the cycle,
model we
solar
examine
results
from
fixed
phase
simulations
(SL2000
and
response. Tourpali et al. also show small horizontal moveSL2000B)
completed
other workers
ments as
in previously
the ozone features,
as aby
function
of season(e.g.
of the
Brasseur,
1993;
Haigh,
1994;
Shindell
et
al.,
1999;
Tourpali of
year, although in view of the small statistical significance
et al.,these
2003;results
Egorova
al.,not
2004).
we et
have
analysed the model results in this
Thedetail.
ozone difference between solar maximum and solar meanThe
fortemperature
timeslice runs
for the
yearsolar
2000cycle
(experiments
response
to the
is shown in
SL2000B
and
of Table
1)similar
is shown
in four
Fig. seasons,
12. In theand
Figure
10.SL2000
The signal
is very
in all
increases
the very
upperdifferent
stratosphere.
Thefixed
signalsoalso
annualgenerally
mean, the
resultstoare
for the
reflects
the ozone
signal with
localtransient
minimum
in the signal
lar phase
results
as compared
witha the
simulations
20 hPa.
ThisIn
is consistent
work
(for example
(Fig. near
12, red
curve).
particular,with
the other
middle
stratospheric
Shibata
andlarger
Kodera,
2005),
which
showed
that the UV and
effect
peak is
slightly
than
in the
transient
simulations,
on temperature
wasoccurs
dominant
above
5 hPa,Abut
the ozone
a minimum
no longer
near
20 hPa.
slight
mini-efwasnear
dominant
hPa. Model
results obtained
mum fect
occurs
50 hPabelow
in the5 fixed
phase results,
but this here
is
agree
well
with
Stratospheric
SSU
data
(reanalysis
of with
results
not statistically significant. These results are consistent
from Scaife et al., 2000) throughout the observed range, but
previous simulations of other models with fixed phase solar
forcing (see e.g. Soukharev and Hood, 2006) regarding their
www.atmos-chem-phys.org/acp/0000/0001/
inability
to simulate the lower stratospheric minimum. Below about 50 hPa, the model values have a large uncertainty,
9
1701
HALOE 1992-2003
3.0
10.0
30.0
100.0
-2
-1
0
1
2
O3 %/100 units F10.7
3
Fig.
in %
% per
per100
100units
unitsofof
Fig.9. 9.Annual
Annualmean
meanozone
ozone solar
solar response
response in
10.7
cm
flux
in
comparison
with
satellite
measurements.
Therere10.7 cm flux in comparison with satellite measurements. The
◦ to 25◦◦ N and
sults
have
sults
havebeen
beenaveraged
averagedover
overthe
thelatitude
latitude range
range 25
25◦ SS to
25 N and
over
allallthree
bars indicate
indicate95%
95%conconover
threeensemble
ensemblemembers.
members. The
The error
error bars
fidence
analysis. Black
Blacklines:
lines:
fidenceintervals
intervalsfrom
fromthe
the linear
linear regression
regression analysis.
model
results as
as indicated,
indicated,from
from
modelresults;
results;coloured
coloured lines:
lines: satellite
satellite results
Soukharev
Soukharevand
andHood
Hood(2006).
(2006).
presumably due to model dynamical variability. Most seaare generally smaller than observed. Analysis of observed
sons
produced similar results, although during the Northern
temperatures to try to extract the small solar signal is exsummer and autumn periods the results are more similar to
tremely challenging, and other observational analyses give
the annual mean transient results in producing a smaller peak
a different estimate for the solar signal (e.g. Labitzke et
response
near
3 hPa.and Gray, 2005). However, these analyal., 2002,
Crooks
ses used data assimilation fields which are unreliable for low
frequency variability (William Randel, pers. comm., 2006).
6 The solar cycle impact on tropical upwelling
The model simulations presented in Labitzke et al. also gave
a
larger temperature
solarofsignal
than on
thethe
results
presented
To investigate
the impact
dynamics
results,
we exhere, by about a factor of 2. However, none of the sim-
amine the solar cycle in tropical upwelling. The tropical
mass upwelling is being increasingly recognised as an imAtmos.
Chem.
0000, 0001–15,circula2007
portant proxy for the
strength
of Phys.,
the Brewer-Dobson
tion, which has been simulated to increase over time in most
Atmos. Chem. Phys., 7, 1693–1706, 2007
3.0
10
1702
Austinetetal.:
al.: Solar
variations
of ozone
and temperature
J.John
Austin
Solarcycle
cycle
variations
of ozone
and temperature
Temp. solar cycle K/100 units
SSU
1.0
Annual
30.0
0.3
DJF
1.0
MAM
Pressure hPa
10.0
Temp. solar cycle K/100 units
0.3
Pressure hPa
Pressure hPa
1.0
3.0
JJA
10.0
SON
3.0
10.0
100.0
-0.4 -0.2
0.0 0.2 30.0
0.4 0.6
K/100 units F10.7
0.8
SSU
Annual
1.0
30.0
100.0
100.0
-0.4 -0.2
0.0 0.2 0.4 0.6
K/100 units F10.7
0.8
1.0
-0.4 -0.2
0.0 0.2 0.4 0.6
K/100 units F10.7
0.8
1.0
sponse to the solar cycle in % per 100 units of 10.7 cm flux. The results
Fig. 10. Seasonal
(left panel)
annual
(right
panel)95%
temperature
response to the solar cycle in % per 100 units of 10.7 cm flux. The results
all three ensemble
members.
The and
error
barsand
indicate
Fig. 10. Seasonal
(left panel)
annual (right
panel)confidence
temperature response to the solar cycle in % per 100 units of 10.7 cm flux. The results
◦ S to 25◦◦ N and
◦ over all three ensemble members. The error bars indicate 95% confidence
have
been
averaged
over
the
latitude
range
25
have
been
averaged
over
the
latitude
range
25
S
to
25
N
and
over5all three ensemble members. The error bars indicate 95% confidence
caife et al. (2000) were reprocessed by using the regression equation
the linearanalysis.
regression The
analysis.
SSUfrom
data from
Scaife
et al.
(2000) were
were reprocessed
by using
the regression
equation 5Eq. (5) and
intervals from theintervals
linear from
regression
SSUThe
data
Scaife
et al.
(2000)
reprocessed
by using
the regression
andto
cover
January 1980
to December 1997.
cover January 1980
December
1997.
O3 solar cycle %/100 units
0.3
5 Analysis and results for fixed solar phase
O3 solar cycle %/100 units
creased. As in
0.3the above works, the tropical upward mass
flux is determined from the mass streamfunction by integrating between the
1.0 latitudes over which the flux is upwards, approximately 30◦ S to 30◦ N.
1.0
Pressure hPa
ulations show the tropical ozone minimum response which
would tend to reduce the temperature response from ozone
in the lower to middle stratosphere.
3.0
Pressure hPa
The response of the tropical upwelling to the solar cycle in
It is plausible that the tropical response is an artifact of the
the
simulations
statistical analysis or that the ozone sensitivity may be a non10.0is shown in Fig. 13. In both sets of simulalinear function of the solar phase. Hence data from solar tions, the theoretical uncertainties in the solar forcing terms
3.0 maxima and minima may give extreme or different results. are very large,
30.0precluding definitive statements. For examTo consider this, we repeat the previous model ozone analple,
neither
results differ significantly from zero while for
ysis (Figure 9) using model data from just those years cor100.0
the
seasonal
variation, both results are similar below 3 hPa
to solar maxima and minima (1964, 1970, 1976,
-2
-1
0
1
2
3
10.0 responding
1981, 1986, 1991, 1996, 2001). Considering all three ensem- in showing a generally
upwelling
O3positive
%/100 units
F10.7 solar signal during
ble members gives 12 years for solar maximum and 12 years Northern autumn and a negative upwelling solar signal durfor solar minimum. Results for the domain 25◦ S to 25◦ N
summer
(not
shown).
An
measure of
11. Annual
mean ozone
solar
response in
% indirect
per 100 units
30.0 are shown in Figure 11 for the annual average, together with ingFig.Northern
of
10.7
cm
flux.
The
black
line
indicates
the
results
using
model
transport
is
the
concentration
of
water
vapour
(e.g.
Mote et
the results using all solar phases. The results are in agreefrom all Again,
solar phases,
the red line
the results
obtained we show
ment below 20 hPa, and above 2 hPa but diverge from each al.,data
1996).
because
ofindicates
the large
variability
when only the calendar years of solar maximum and solar minimum
100.0 other in the middle stratosphere. Despite larger uncertainties only
mean analysis.
solar cycle
in the
two sets
werethe
usedannual
in the regression
The results
are averaged
overof simulain examining a limited sample of the period, it would seem
◦
◦
◦ ensemble
◦ N. In this
the
latitude
range
25
S
to
25
N
and
over
all
three
memtions
(Fig.
14),
averaged
between
30
S
and
30
-2 that the-1
0
1
2 a robust3feature of
lower stratospheric
minimum
is
bers. The error bars indicate the 95% confidence intervals obtained
the fixed phase runs indicate a slight increase in wathese model
results. Nonetheless,
with the analysis presented case,
O3 %/100
units F10.7
from the regression analysis.
here, it would seem that the differences in the middle to up- ter vapour concentrations and the transient runs indicate a
stratosphere are statistically significant, but it should be decrease. Results in the mesosphere should be ignored as
Fig. 11. Annualper
mean
ozone solar response in % per 100 units
cautioned that the analysis is based on just 4 solar cycles.
theences.
model
chemistry was not designed to be realistic there.
of
10.7
cm
flux.
The
black
line
indicates
the results
Fig. 11. Annual mean
ozone
in %using
per model
100
Processes
not solar
includedresponse
in the regression
analysis,
suchunits
as
To obtain a clearer signal during the solar cycle, we
data from all solartransport
phases, and
the chemical
red line indicates
the
results
obtained
Nonetheless,
below
3 hPa,
the sensitivities
to the solar
changes
cannot be
represented
from about
fixed phase
simulations
(SL2000 and
of 10.7 cm flux. The black line indicates
thethatresults
using
model examine results
when only the calendar
years
of terms,
solar maximum
and solar minimum
as simple
trend
may have contributed
to these differ- cycle
SL2000B)
as previously
completed bysignificant
other workers
(e.g.95% conare marginally
statistically
at the
data from
solar
phases,
the analysis.
red line The
indicates
theaveraged
results obtained
wereall
used
in the
regression
results are
over
fidence level.
◦ S to 25
when only
the calendar
of◦ solar
maximum
solar memminimum
the latitude
rangeAtmos.
25years
and0000,
over 0001–15,
all threeand
ensemble
Chem. N
Phys.,
2007
www.atmos-chem-phys.org/acp/0000/0001/
bers.in
The
error
bars indicate
the 95% The
confidence
intervals
obtained over Interpretation of these results is complex. The change in
were used
the
regression
analysis.
results
are averaged
from the
regression
the latitude
range
25◦ Sanalysis.
to 25◦ N and over all three ensemble mem-tropopause temperature (Fig. 10) results in a saturated vapor pressure increase of about 3% but at the hygropause, the
bers. The error bars indicate the 95% confidence intervals obtained
model results show little sensitivity to the solar cycle. Howfrom the
regression
analysis.
ever, a reduction in water vapour due to the solar cycle ocmodels
(Butchart
and Scaife, 2001; Butchart et al., 2006). In
the transient simulations presented here, the mass upwelling
curs at slightly higher altitudes, implying increased upward
motion, since the water vapour gradient is upwards. Below
also increases and there is also a simple relationship with the
10 hPa, upward motion decreases ozone and above 10 hPa
inverse age of air (Austin and Li, 2006). The consequences
are
that
over
time
chemical
constituents
had
a
shorter
stratoences. To obtain a clearer signal during the solar cycle, weupward motion increases ozone slightly, because of the dispheric
timescale
the large
scale
tropical ascent
rate in- andrection of the ozone gradient. The photochemical timescale
examine
results
fromand
fixed
phase
simulations
(SL2000
SL2000B) as previously completed by other workers (e.g.
Atmos. Chem. Phys., 7, 1693–1706, 2007
www.atmos-chem-phys.org/acp/0000/0001/
www.atmos-chem-phys.net/7/1693/2007/
0.3
O3
0.3
O3 solar cycle %/100 units
solar cycle %/100 units
O3 solar cycle %/100 units
O3
0.3 solar cycle %/100 units
0.3
J. Austin
et al.: Solar
DJFcycle variations of ozone and temperature
1.0
JJA
10.0
10.0
SON
30.0
30.0
0.3
JJA
1.0
SON
DJF
MAM
3.0
JJA
10.0
3.0
10.0
SON
30.0
hPa
3.0
John Austin et al.: Solar cycle variations of ozone and temperature
O3 solar cycle %/100 units
O3 solar cycle %/100 units
Pressure hPa
Pressure
12
MAM MAM
30.0
100.0 100.0
-2
-2
-1
1703
1.0
1.0
Pressure hPa
3.0
DJF
Pressure hPa
Pressure hPa
Pressure hPa
1.0
0.3
3.0
1.0
Transient
10.0
Transient
3.0
Annual
Annual
Transient
10.0
30.0
Annual
30.0
-1
0 100.0-2 10 -1
210
32
1
O3
%/100
units
F10.7
O3 %/100 units F10.7
O3 %/100 units F10.7
2
100.0
3
-2
3
100.0
100.0 -2
-1
-2
0-10
10 2 2 1 3
32
1
O3
%/100
units
F10.7
O3 %/100
F10.7
O3 %/100units
units F10.7
3
-1
Fig. 12. Ozone solar response in % per 100 units of 10.7 cm flux for solar maximum – solar mean from the timeslice runs SL2000B and
Fig. 12. Ozone solar response in % per 100 units of 10.7
for
maximum - solar mean from the timeslice runs SL2000B and
◦ Scmtoflux
◦ Nsolar
◦ 25for
SL2000.
The
results
an results
average
over
the10.7
latitude
range
25
for
different
seasons
(left
panel)
and
the
annual
average
12.solar
Ozone
solar
response
in
%are
per
100
units
oflatitude
10.7for
cm
flux
maximum
- solar
mean
from
the
timeslice
runs and
SL2000B an
SL2000.
an average
over cm
the
range
25
S to
25◦solar
N forthe
the
seasons
(left
panel)
and
for
thefor
annual
average
. 12. Fig.
Ozone
response
in are
% The
per
100
units
of
flux
solar
maximum
-different
solar mean
from
the
timeslice
runs
SL2000B
◦results
◦
(right
panel).
Also
included
in
the
right
hand
panel
are
the
results
for
the
transient
experiments
in
the
annual
average.
All
the
error
bars
(right
panel).
Also
included
in
the
right
hand
panel
are
the
for
the
transient
experiments
in
the
annual
average.
All
the
error
bars
◦
◦
Theare
results
are an over
average
over therange
latitude
range
25 NSfor
to 25
N
for theseasons
different(left
seasons
(left
panel)
and
for the
annual averag
2000.SL2000.
The results
an average
the
latitude
25
S
to
25
the
different
panel)
and
for
the
annual
average
denote the 95%
confidence intervals.
denote the 95% confidence
intervals.
(right Also
panel).
Also included
in the
right
hand
are the
for the experiments
transient experiments
in the
annualAll
average.
Allbars
the error ba
ht panel).
included
in the right
hand
panel
are panel
the results
forresults
the transient
in the annual
average.
the error
Upwelling solar cycle %/100 units
H2O solar cycle %/100 units
0.3
denote
the
95%
confidence
intervals.
0.3
ote the 95% confidence intervals.
Upwelling
solar%/100
cycleunits
%/100 units
Upwelling
solar cycle
30.0
3.0
10.0
30.0
TRANSA
0.3
SL2000/B
TRANSA
TRANSA
10.0
30.0
100.0
-4
...B/C ...B/C
SL2000/B
SL2000/B
-2
0
2
4
Upwelling change %/100 units F10.7
Fig. 13. Annual mean solar response in the tropical upwelling in %
per 100 units of 10.7 cm flux in the transient simulations TRANSA,
TRANSB and TRANSC (black) in comparison with the solar signal
in simulations SL2000 and SL2000B (red).
100.0
100.0
-4
-2-4 perature
0-2 sensitivity
20
4 2 more than
4 the ozone between sovaried
Upwelling
change
%/100
units
F10.7
Upwelling
change
%/100
units
F10.7
lar cycles, the temperature
response
for 1991
agreed surpris-
H2O
solar
cycleunits
%/100 units
H2O solar
cycle
%/100
TRANSA
0.3
...B/C
Pressure hPa
...B/C
3.0
1.0
3.0
1.0
1.0
SL2000/B
TRANSA
TRANSA
10.0
Pressure hPa
10.0
Pressure hPa
Pressure hPa
3.0
Pressure hPa
1.0
1.0
1.0
Pressure hPa
0.3
0.3
3.0 30.0
100.0
-4
10.0
...B/C
3.0
...B/C
SL2000/B
SL2000/B
0
2
4
10.0-2 H2O %/100
units F10.7
6
Fig. 14.
As Figure 13, but for the water vapour concentration.
30.0
30.0
istry.
100.0
100.0
For the 11-year
solar cycle in the lower and middle strato-4
-2-4
0-2
20
42
64
6
sphere, we show good agreement
between model
results and
H2O
%/100
units
F10.7
H2O
%/100
units
F10.7
observations in the transient simulations, without including
energetic electron precipitation effects. In particular, the
ingly well with observations for 1991-1994 (Hood and Zhou,
However,
for other
years
for which
measurements
averaged
results
demonstrated
a statistically
sig14. As Fig.
13, but
for the
water vapour
concentration.
Fig. 13. Annual 1998).
mean solar
response
in the
tropical
upwelling
in % ex- Fig.tropically
ist
(Hood,
1986;
Hood
and
Cantrell,
1988)
the
model
results
nificant
minimum
in
ozone
response
near
20
hPa,
which
has
. 13. Fig.
Annual
mean
solar
in the
tropical
in TRANSA,
%
Fig.
14. As
Figure
13,Figure
but for13,
thebut
water
vapour
concentration.
100 units
ofresponse
10.7
cm flux
in the
transient
simulations
13.per
Annual
mean
solar
in with
theupwelling
tropical
upwelling
in
%
Fig.
14. As
the water
agree
onlyresponse
qualitatively
observations.
This suggests
the
not
previously
been published
(e.g.forShindell
et al.,vapour
1999; concentration.
TRANSC
insimulations
comparison
with
the
signal
100 units
ofTRANSB
10.7
flux
in
theflux
TRANSA,
need
fortransient
a(black)
further
of thesimulations
27-day
solarsolar
cycle
in models
Soukharev and Hood, 2006). Examination of the ozone valper 100
unitscm
ofand
10.7
cm
in theanalysis
transient
TRANSA,
in simulations
SL2000
and
SL2000B
(red).
ANSB
and
TRANSC
(black)
with
the
signal
and in
thecomparison
factors
contributing
to
it. solar
It is likely
that the different
ues just for the solar maxima and solar minima during the
TRANSB and TRANSC (black) in comparison with the solar signal
strengths of
the Schwabe cycle may be a contributing factor
135 years of simulations (’fixed solar phase analysis’) reimulations
SL2000
and
SL2000B
(red).
in simulations SL2000
SL2000B
dueand
to the
presence of(red).
possible nonlinearities in the chemistry. sulted in a higher ozone response in the middle and up-
istry.
For the 11-year
cycle
in the
lower
andlower
middle
stratodecreases rapidly
with
altitude
so that
vertical
For thesolar
11-year
solar
cycle
in the
and
middle strato
Atmos.
Chem.
Phys., 0000,
0001–15,
2007motion bewww.atmos-chem-phys.org/acp/0000/0001/
atureperature
sensitivity
varied
more
than
the
ozone
between
sosphere,
we
show
good
agreement
between
model
results
andresults an
comes
less
effective.
Consequently,
the
decrease
in
ozone
with
the
dynamical
viewpoint
of
Kodera
and
Kuroda
(2002),
sphere, we show good agreement between model
sensitivity varied more than the ozone between sodue
to
the
solar
cycle
(Fig.
9)
and
its
absence
in
the
fixed
based
on
a
simplified
model.
The
implications
on
tempercycles,
temperature
responseresponse
for 1991for
agreed
surprisobservations
in the transient
simulations,
without including
lar the
cycles,
the
temperature
1991decrease
agreed surprisobservations
in the transient
simulations,
without includin
ature electron
of upwardprecipitation
motion would effects.
be to increase
the adiabaticthe
runs (Fig. for
12) 1991-1994
may be related
to the
ly well withphase
observations
(Hood
and Zhou, in wa-energetic
In particular,
ingly well
with seen
observations
for
1991-1994
(Hood and
energetic
precipitation
effects.
particular, th
ter vapour
in Fig.
There
are two problems
with Zhou,
thistropically
cooling
relative electron
to
the direct
radiative effect.
There In
would
averaged
results
demonstrated
a statistically
sig98). However,
for other
years
for14.which
measurements
exaveraged
results
a statistically sig
1998).argument,
However,
other
for which
measurements
(i)for
there
is noyears
clear increase
in tropical
upwellingex- alsotropically
be an indirect
radiative
effect demonstrated
from the solar induced
(Hood, 1986;
Hood
and
Cantrell,
1988)
the
model
results
nificant
minimum
in
ozone
response
near
20
hPa,
which
has
itself 1986;
(Fig. 13)
and (ii)
proposed 1988)
dynamical
– an
ozone
change.minimum
These impacts
on ozone
and temperature
are which ha
ist (Hood,
Hood
andtheCantrell,
the changes
model results
nificant
in ozone
response
near 20 hPa,
ee only qualitatively
with
observations.
This
suggests
the
not
previously
been
published
(e.g.
Shindell
et
al.,
1999;
increased
in upwelling
forobservations.
high solar flux This
– is inconsistent
with the model
results
shown in(e.g.
Figs. 9Shindell
and 10. et al., 199
agree only
qualitatively
with
suggests the consistent
not previously
been
published
d for a further analysis of the 27-day solar cycle in models
Soukharev and Hood, 2006). Examination of the ozone valneed for a further analysis of the 27-day solar cycle in models
Soukharev and Hood, 2006). Examination of the ozone va
the factors contributing to it. It is likely that the different
ues just for the solar maxima and solar minima during the
Chem.
Phys., and
7, 1693–1706,
2007 during th
and thewww.atmos-chem-phys.net/7/1693/2007/
factors contributing to it. It is likely that the different
ues just for Atmos.
the solar
maxima
solar minima
ngths of the Schwabe cycle may be a contributing factor
135 years of simulations (’fixed solar phase analysis’) restrengths of the Schwabe cycle may be a contributing factor
135 years of simulations (’fixed solar phase analysis’) r
to the presence of possible nonlinearities in the chemsulted in a higher ozone response in the middle and updue to the presence of possible nonlinearities in the chemsulted in a higher ozone response in the middle and up
1704
7
Conclusions
A coupled chemistry climate model has been used to simulate the impacts of the 27-day solar rotation cycle and the
11-year Schwabe cycle on ozone and temperature.
The 27-day results were analysed for a period of one calendar year at the maximum of each of the 11-year Schwabe
cycles when the response was largest. The results were found
to vary from one solar cycle to the next. Good agreement
was found particularly in ozone between the model results
and observations of Zhou et al. (1997) and Hood and Zhou
(1998). Depending on the year chosen, the ozone results
demonstrated the downward propagation of the 27-day oscillation phase in some cases but with little phase propagation in other years. The ozone sensitivity to the 27-day oscillation was also quantitatively well reproduced, peaking at
about 0.4% per 1% change in 205 nm flux. In the model, the
peak occurred at higher altitude and in the upper stratosphere
the model diverged from the observations, but this may be an
error in the observations, which are not strict diurnal averages. The temperature response to the 27-day cycle is small,
typically less than 0.5 K, and is difficult to extract from both
model and observations. Although as a percentage, the temperature sensitivity varied more than the ozone between solar cycles, the temperature response for 1991 agreed surprisingly well with observations for 1991–1994 (Hood and Zhou,
1998). However, for other years for which measurements exist (Hood, 1986; Hood and Cantrell, 1988) the model results
agree only qualitatively with observations. This suggests the
need for a further analysis of the 27-day solar cycle in models
and the factors contributing to it. It is likely that the different
strengths of the Schwabe cycle may be a contributing factor
due to the presence of possible nonlinearities in the chemistry.
For the 11-year solar cycle in the lower and middle stratosphere, we show good agreement between model results and
observations in the transient simulations, without including
energetic electron precipitation effects. In particular, the
tropically averaged results demonstrated a statistically significant minimum in ozone response near 20 hPa, which has
not previously been published (e.g. Shindell et al., 1999;
Soukharev and Hood, 2006). Examination of the ozone values just for the solar maxima and solar minima during the
135 years of simulations (“fixed solar phase analysis”) resulted in a higher ozone response in the middle and upper stratosphere while the minimum in ozone response in
the lower stratosphere was largely unaffected. In a final set
of experiments, with fixed solar phase applied continuously
during the runs, as well as fixed climatological SSTs, the
ozone response was quite different to the results of the transient runs with the absence of the lower stratospheric minimum. These results are generally consistent with the transient simulations reported by Eyring et al. (2006) which have
at the time of writing been analysed for a solar cycle signal (K. Matthes, R. Garcia, K. Shibata, E. Rozanov, personal
Atmos. Chem. Phys., 7, 1693–1706, 2007
J. Austin et al.: Solar cycle variations of ozone and temperature
communications, 2006).
The temperature response to the 11-year solar cycle was
determined in large part by the ozone. Comparison with measurements proved to give mixed results as it is hampered by
the need to obtain instrument stability of better than a few
tenths of a K over a decade or more. Nonetheless, the results
agree reasonably well over the pressure range 1 to 30 hPa
with the results obtained from one of the satellite datasets
that has arguably been best scrutinised (see Nash and Brownscombe, 1983; Scaife et al., 2000).
Although solar cycle effects are small, about an order
of magnitude smaller than ozone changes over the last few
decades due to chlorine change, accurate simulation of these
processes provides confidence in the predictions of the models for the future. One of the significant changes introduced
in AMTRAC which is not present in many previous simulations exploring the solar cycle, is the fact that the observed
monthly varying Schwabe and temporally varying 27-day cycles are imposed, whereas previous simulations have been
completed with a fixed solar phase (Labitzke et al., 2002;
Tourpali et al. 2003; Williams et al., 2000 etc.).
It has been suggested that interactions with the quasibiennial oscillation (QBO) may be responsible for the tropical minimum in the ozone solar cycle response. This has
been attributed either to a simple interference of the QBO
in the statistical analysis (Lee and Smith, 2003) or to a
real solar cycle modulation of the QBO itself (e.g., McCormack, 2003). The results obtained here without a model
QBO would suggest that this is not essential. Also, observational studies (Soukharev and Hood, 2006), indicate a broad
minimum over a wide latitude range consistently occurring
over several solar cycles, which is not easily explained by
the QBO alone. Nevertheless, a pronounced local minimum
ozone solar cycle response is often obtained at 10 hPa centered on the equator (see Figs. 6 and 7 of Soukharev and
Hood). This pronounced minimum does suggest a partial
role for the QBO in producing at least some of the observed
ozone response minimum. This possibility should be investigated in the future using models that simulate a realistic
QBO.
It should also be recognised that the solar cycle is by no
means sinusoidal in shape (Fig. 1). This may give rise to
assymetries also in the dynamics, as well as additional processes with timescales shorter than the 11-year period. Examination of the tropical upwelling, which is a strong candidate for a dynamical response to the solar cycle, proved
to be inconclusive in the simulations presented here due at
least in part to the large uncertainties in determining the solar signal from a somewhat noisy field. However, simulated
water vapour values contained a solar cycle indicative of enhanced upward motion during high solar fluxes. There is a
possibility that this increase may have been driven by SSTs,
since as noted by White et al. (2003), global SSTs have been
phase locked to the solar cycle. Solar induced changes in
tropical SSTs could affect the tropical tropopause temperwww.atmos-chem-phys.net/7/1693/2007/
J. Austin et al.: Solar cycle variations of ozone and temperature
ature (T. Reichler, personal communication, 2006), as detected in observations on the 27-day time scale (Hood, 2003).
Nonetheless, it is not clear that the model is simulating the
correct magnitude or even sign of this effect, since theoretical and observational evidence favors relative downwelling
in the tropics near solar maxima (Kodera and Kuroda, 2002;
Hood and Soukharev, 2003). Further long simulations are
therefore needed both with different models as well as longer
and more simulations with the same model to extract more
reliable statistics from chaotic fields.
In summary, the improved agreement with observations
in our simulations compared with previous work has arisen
from the specification of all phases of the 11-year solar cycle
as well as time varying SSTs. These two components were
absent from control runs in which simulated ozone was consistent with previous simulations. A plausible argument has
been presented as to why the SSTs could be playing a major role, an issue which could be examined by completing
further simulations with a full solar cycle but climatological
SSTs.
Acknowledgements. J. Austin’s research was supported by the
Visiting Scientist Program at the NOAA Geophysical Fluid Dynamics Laboratory, administered by the University Corporation for
Atmospheric Research. We would like to thank D. Schwarzkopf,
L. Horowitz and two anonymous reviewers for their useful comments on earlier drafts of the paper.
Edited by: V. Fomichev
References
Anderson, J. L., Balaji, V., Broccoli, A. J., et al.: The new GFDL
global atmosphere and land model AM2/LM2: Evaluation with
prescribed SST simulations, J. Climate, 17, 4641–4673, 2004.
Austin J.: A three-dimensional coupled chemistry-climate model
simulation of past stratospheric trends, J. Atmos. Sci., 59, 218–
232, 2002.
Austin, J. and Li, F.: On the relationship between the strength of
the Brewer-Dobson circulation and the age of stratospheric air,
Geophys. Res. Lett., 33, L17807, doi:10.1029/2006GL026867,
2006.
Austin, J. and Wilson, R. J.: Ensemble simulations of the decline and recovery of stratospheric ozone, J. Geophys. Res., 111,
D16314, doi:10.1029/2005JD006907, 2006.
Austin, J., Wilson, R. J., Li, F., and Vömel, H.: Evolution of water vapor concentrations and stratospheric age of air in coupled
chemistry-climate model simulations, J. Atmos. Sci., 64, 905–
921, 2007.
Baldwin, M. P. and Dunkerton, T. J.: Observations and statistical
simulations of a proposed solar cycle/QBO/weather relationship,
Geophys. Res. Lett., 16, 863–866, 1989.
Brasseur, G.: The response of the middle atmosphere to long-term
and short-term solar variability: A two-dimensional model, J.
Geophys. Res., 98, 23 079–23 090, 1993.
www.atmos-chem-phys.net/7/1693/2007/
1705
Butchart, N. and Scaife, A. A.: Removal of chlorofluorocarbons
by increased mass exchange between the stratosphere and the
troposphere in a changing climate, 410, 799–802, 2001.
Butchart, N., Scaife, A. A., Bourqui, M., de Grandpré, J., Hare, S.
H. E., Kettlebororough, J., Langematz, U., Manzini, E., Sassi,
F., Shibata, K., Shindell, D., and Sigmond, M.: Simulations of
anthropogenic change in the strength of the Brewer-Dobson circulation, Clim. Dyn., 27(7–8), 727–741, 2006.
Callis, L. B., Natarajan, M., and Lambeth, J. D.: Solar-atmosphere
coupling by electrons (SOLACE) 3. Comparisons of simulations
and observations, 1979–1997, issues and implications, J. Geophys. Res., 106, 7523–7539, 2001.
Chandra, S., McPeters, R. D., Planet, W., and Nagatani, R. M.: The
27 day solar UV response of stratospheric ozone: solar cycle 21
vs. solar cycle 22, J. Atmos. Terr. Phys., 56, 1057–1065, 1994.
Chen, L., London, J., and Brasseur, G.: Middle atmosphere ozone
and temperature responses to solar irradiance variations over 27day periods, J. Geophys. Res., 102, 29 957–29 979, 1997.
Coughlin, K. and Tung, K. K.: Eleven-year solar cycle signal throughout the lower atmosphere, J. Geophys. Res., 109,
D21105, doi:10.0129/2004JD004873, 2004.
Crooks, S. A. and Gray, L. J.: Characterization of the 11-year solar signal using a multiple regression analysis of the ERA-40
dataset, J. Climate, 18, 996–1015, 2005.
Egorova, T., Rozanov, E., Manzini, E., Haberreiter, M., Schmutz,
W., Zubov, V., and Peter, T.: Chemical and dynamical response
to the 11-year variability of the solar irradiance simulated with
a chemistry-climate model, Geophys. Res. Lett., 31, L06119,
doi:10.1029/2003GL019294, 2004.
Eyring, V., Butchart, N., Waugh, D. W., et al.: Assessment of temperature, trace species and ozone in chemistry-climate model
simulations of the recent past, J. Geophys. Res., 111, D22308,
doi:10.0129/2006JD007327, 2006.
Fleming, E. L., Chandra, S., Jackman, C. H., Considine, D. B., and
Douglass, A. R.: The middle atmospheric response to short and
long term solar UV variations: analysis of observations and 2D
model results, J. Atmos. Terr. Phys., 57, 333–365, 1995.
Frölich, C.: Observations of irradiance measurements, Space Sci.
Rev., 94, 15–24, 2000.
Haigh, J. D.: The role of stratospheric ozone in modulating the solar
radiative forcing of climate, Nature, 370, 544–546, 1994.
Hershel, W.: Observations tending to investigatethe nature of the
sun, in order to find the causes or symptoms of its variable emission of light and heat, with the remarks on the use that may possibly be drawn from solar observations, Phil. Trans, Roy. Soc.
London, 91, 265–318, 1801.
Hood, L. L.: Coupled stratospheric ozone responses to short-term
changes in solar ultraviolet flux: an analysis of Nimbus 7 SBUV
and SAMS data, J. Geophys. Res., 91, 5264–5276, 1986.
Hood, L. L.: Thermal response of the tropical tropopause region
to solar ultraviolet variations, Geophys. Res. Lett., 23, 2215,
doi:10.1029/2003GL018364, 2003.
Hood, L. L. and Cantrell, S.: Stratospheric ozone and temperature
responses to short term solar ultraviolet variations: reproducibility of low-latitude response measurements, Ann. Geophys., 6,
525–530, 1988,
http://www.ann-geophys.net/6/525/1988/.
Hood, L. L. and McCormack, J. P.: Components of interannual
ozone change based on Nimbus 7 TOMS data, Geophys. Res.
Atmos. Chem. Phys., 7, 1693–1706, 2007
1706
Lett., 19, 2309–2312, 1992.
Hood, L. L. and Soukharev, B. E.: Quasi-decadal variability of the
tropical lower stratosphere: The role of extratropical wave forcing, J. Atmos. Sci., 60, 2389–2403, 2003.
Hood, L. L. and Soukharev, B. E.: Solar induced variations of odd
nitrogen: multiple regression analysis of UARS HALOE data,
Geophys. Res. Lett., 33, L22805, doi:10.1029/2006GL028122,
2006.
Hood, L. L. and Zhou, S.: Stratospheric effects of 27-day solar ultraviolet variations: an analysis of UARS MLS ozone and temperature data, J. Geophys. Res., 103, 3629–3638, 1998.
Kodera, K. and Kuroda, Y.: Dynamical response to the solar cycle,
J. Geophys. Res., 107(D24), 4749, doi:10.1029/2002JD002224,
2002.
Labitzke, K., Austin, J., Butchart, N., Knight, J., Takahashi, M.,
Nakamoto, M., Nagashima, T., Haigh, J., and Williams, V.: The
global signal of the 11-year solar cycle in the stratosphere: observations and model results, J. Atmos. Solar Terr. Phys., 64, 203–
210, 2002.
Langematz, U., Grenfell, J. K., Matthes, K., Mieth, P., Kunze, M.,
Steil, B., and Brühl, C.: Chemical effects in 11-year solar cycle
simulations with the Freie Universität Berlin Climate Middle Atmosphere Model with online chemistry (FUB-CMAM-CHEM),
Geophys. Res. Lett., 32, L13803, doi:10.1029/2005GL022686,
2005.
Lean, J., Rottman, G., Harder, J., and Kopp, G.: SORCE contributions to new understanding of global change and solar variability,
Solar Phys., 203, 27–53, 2005.
Lee, H. and Smith, A. K.: Simulation of the combined effects of
solar cycle, quasi-biennial oscillation, and volcanic forcing on
stratospheric ozone changes in recent decades, J. Geophys. Res.,
108(D2), 4049, doi:10.1029/2001JD001503, 2003.
McCormack, J. P.: The influence of the 11-year solar cycle on
the quasi-biennial oscillation, Geophys. Res. Lett., 30, 2162,
doi:10.1029/2003GL018314, 2003.
Mote, P. W., Rosenlof, K. H., McIntyre, M. E., Carr, E. S., Gille, J.
C., Holton, J. R., Kinnersley, J. S., Pumphrey, H. C., Russell III,
J. M., and Waters, J. W.: An atmospheric tape recorder: the imprint of tropical tropopause temperatures on stratospheric water
vapor, J. Geophys. Res., 101, 3989–4006, 1996.
NAG: NAG Library, Mark 19, Copyright National Algorithms
Group Limited, 1999.
Nash, J. and Brownscombe, J. L.: Validation of the stratospheric
sounding unit, Adv. Space Res., 2, 59–62, 1983.
Randel, W. and Wu, F.: A stratospheric ozone profile data
set for 1979-2005: variability, trends, and comparisons
with ozone column data, J. Geophys. Res., 111, in press,
doi:10.1029/2006JD007339, 2007.
Rottman, G.: Solar ultraviolet irradiance and its temporal variation,
J. Atmos. Sol.-Terr. Phys., 61, 37–44, 1999.
Rottman, G. J., Woods, T. N., and Sparn, T. P.: SOLar STellar Irradiance Comparison Experiment I: 1. instrument design and operation, J. Geophys. Res., 98, 10 667–10 677, 1993.
Atmos. Chem. Phys., 7, 1693–1706, 2007
J. Austin et al.: Solar cycle variations of ozone and temperature
Rozanov, E., Callis, L., Schlesinger, M., Yang, F., Andronova, N.,
and Zubov, V.: Atmospheric response to NOy source due to energetic electron precipitation, Geophys. Res. Lett, 32, L14811,
doi:10.1029/2005GL023041, 2005a.
Rozanov, E., Schraner, M., Egorova, T., Ohmura, A., Wild, M.,
Schmutz, W., and Peter, T.: Solar signal in atmospheric ozone,
temperature and dynamics simulated with CCM SOCOL in transient mode, Mem. S.A. It., 76, 876–879, 2005b.
Rozanov, E., Egorova, T., Schmutz, W., and Peter, Th.: Simulation
of the stratospheric ozone and temperature response to the solar
irradiance variability during sun rotation cycle, J. Atmos. SolarTerr. Phys., 68, 2203–2213, 2006.
Scaife, A. A., Austin, J., Butchart, N., Pawson, S., Keil, M., Nash,
J., and James, I. N.: Seasonal and interannual variability of the
stratosphere diagnosed from UKMO TOVS analyses, Quart. J.
Roy. Meteorol. Soc., 126, 2585–2604, 2000.
Schwabe, H.: Solar Observations during 1843, Astronomische
Nachrichten, 20, no. 495, 1843.
Shibata, K. and Kodera, K.: Simulation of radiative and dynamical
responses of the middle atmosphere to the 11-year solar cycle, J.
Atmos. Sol. Terr. Phys., 67, 125–143, 2005.
Shindell, D., Rind, D., Balachandran, N., Lean, J., and Lonergan,
P.: Solar cycle variability, ozone, and climate, Science, 284, 305–
308, 1999.
Soukharev, B. E. and Hood, L. L.: The solar cycle variation of
stratospheric ozone: multiple regression analysis of long-term
satellite data sets and comparisons with models, J. Geophys.
Res., 111, D20314, doi:10.1029/2006JD007107, 2006.
Tourpali, K., Schurmans, C. J. E., von Dorland, R., Steil, B., and
Brühl, C.: Stratospheric and tropospheric response to enhanced
solar UV radiation: A model study, Geophys. Res. Lett., 30,
1231, doi:10.1029/2002GL016650, 2003.
Tourpali, K., Zerefos, C. S., Balis, D. S., and Bais, A. F.: The
11-year solar cycle in stratospheric ozone: comparison between
Umkehr and SBUV v8 and effects on surface erythemal irradiance, J. Geophys. Res., 111, doi:10.1029/2006JD007760, 2007.
Waters, J.: Microwave limb sounding of Earth’s upper atmosphere,
Atmos. Res., 23, 391–410, 1989.
White, W. B., Dettinger, M. D., and Cayan, D. R.: Sources of global
warming of the upper ocean on decadal period scales, J. Geophys. Res., 108(C8), 3248, doi:10.1029/2002JC001396, 2003.
Williams, J., Austin, J., and Haigh, J.: Model simulations of the
impact of the 27-day solar rotation period on the stratosphere,
Adv. Space Res., 27, 1933–1942, 2001.
Zerefos, C. S., Tourpali, K., Bojkov, B. R., and Balis, D. S.:
Solaractivity-total ozone relationships: observations and model
studies with heterogeneous chemistry, J. Geophys., 102, 1561–
1569, 1997.
Zhou, S., Rottman, G. J., and Miller, A. J.: Stratospheric ozone
response to short and intermediate-term variations of solar UV
flux, J. Geophys. Res., 102, 9003–9011, 1997.
www.atmos-chem-phys.net/7/1693/2007/