Volodin2008 Article RelationBetweenTemperatureSens

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

ISSN 0001-4338, Izvestiya, Atmospheric and Oceanic Physics, 2008, Vol. 44, No. 3, pp. 288–299.

© Pleiades Publishing, Ltd., 2007.


Original Russian Text © E.M. Volodin, 2008, published in Izvestiya AN. Fizika Atmosfery i Okeana, 2008, Vol. 44, No. 3, pp. 311–323.

Relation between Temperature Sensitivity to Doubled Carbon


Dioxide and the Distribution of Clouds in Current
Climate Models
E. M. Volodin
Institute of Numerical Mathematics, Russian Academy of Sciences, ul. Gubkina 8, Moscow, 119991 Russia
e-mail: volodin@inm.ras.ru
Received February 22, 2007; in final form, May 24, 2007

Abstract—The paper considers a relation between equilibrium global warming at doubled carbon dioxide (cli-
mate sensitivity) and the distribution of clouds and relative humidity in 18 state-of-the-art climate models.
There is a strong correlation among three indices: (1) model climate sensitivity, (2) mean cloud amount change
due to global warming, and (3) the difference in cloud amount between the tropics and midlatitudes. In the sim-
ulation of the present-day current, models with high sensitivity produce smaller clouds amounts in the tropics
and larger cloud amounts over midlatitude oceans than models with low sensitivity. The relative humidity in the
tropics is smaller in models with high sensitivity than in models with low sensitivity. There is a similarity
between vertical profiles of cloud amount and relative humidity under global warming and vertical profiles of
the difference in these quantities averaged over the tropics and midlatitudes. Based on the correlations obtained
and observations of cloud amount and relative humidity, an estimate is made of the sensitivity of a real climate
system.
DOI: 10.1134/S0001433808030043

INTRODUCTION its sensitivity is impossible. Therefore, attempts are


One of the key problems in the modern science on being made to determine sensitivity indirectly. One of
climate is that of determining the equilibrium sensitiv- the approaches to this problem is to find a correspon-
ity of a real climate system. To determine equilibrium dence between the model sensitivity and a simulation
sensitivity, numerical experiments, as a rule, are per- in the model of certain parameters of the current cli-
formed with an atmospheric model and a homoge- mate. In [3], for example, a correlation is established
neous 50-m layer of the ocean. The equilibrium cli- between the amplitude of the annual temperature
mate response to a doubling of ëé2 is achieved in cycle and climate sensitivity in models with a 50-m
such models during 15 to 20 years, in contrast to layer. However, for the IPCC AR4 models, the coeffi-
atmospheric models with a deep ocean, where the cient of correlation between the amplitude of the
adjustment to the ëé2 concentration change may last annual temperature cycle and climate sensitivity does
for hundreds of years. The magnitude of global warm- not exceed 0.3 [3]. The surface short-wave radiation
ing in response to a doubling of ëé2 in 18 such mod- budget is the another quantity that could be related to
els is listed in the table. Numerical experiments with climate sensitivity [4]. However, the data used in [4]
the models were performed in 2004–2005. Results of were those from the CMIP-2 [5], where the nonequi-
these experiments were used in preparing the Inter- librium response to doubled ëé2 was determined for
governmental Panel on Climate Change (IPCC) models with a deep ocean, that is, a response to a
Fourth Assessment Report [1]. The models them- 1% year–1 ëé2 increase until ëé2 doubling. In addi-
selves will be referred to as the IPCC AR4 models. tion, such a response depends on which effective layer
According to the table, the sensitivity of climate in the of the ocean participates in global warming. Further-
different models varies from 2.1 to 4.4 K. This interval more, since a major contribution to the discrepancy in
has remained nearly unchanged relative to the IPCC sensitivity comes from a different cloud response to
Third Assessment results (2.0–4.5 K). This brings up global warming, it is logical to consider a correspon-
the question of what sensitivity a real climate system dence between climate sensitivity and the distribution
should have. of clouds in the model, but the CMIP-2 data archive
A different cloud response to global warming pro- does not contain cloud data. The goal of this paper is
vides the largest contribution to the scatter of model to determine a relation between the model distribution
estimates of climate sensitivity [2]. A direct experi- of clouds derived from the simulation of the current
ment with a real climate system on the estimation of climate and the relative humidity linked to clouds and

288
RELATION BETWEEN TEMPERATURE SENSITIVITY 289

the sensitivity of climate of the given model to a dou- Climate sensitivity to CO2 doubling for the IPCC AR4 models
bling of carbon dioxide concentration.
No. Model ∆T(ä)
1 UKMO-HadGEM1, UK 4.4
DATA
2 IPSL-CM4, France 4.4
Calculation is performed with data from the 18 3 MIROC3.2(hires), Japan 4.3
IPCC AR4 models for which the equilibrium sensitiv-
4 MIROC3.2(medres), Japan 4.0
ity is reported in [1]. The data used are total cloud
amounts on a model longitude–latitude grid; three- 5 CGCM3.1(T47), Canada 3.4
dimensional monthly mean cloud amounts on a model 6 CGCM3.1(T63), Canada 3.4
grid in longitude, latitude, and the vertical; and three- 7 ECHAM5/MPI-OM, Germany 3.4
dimensional monthly mean relative humidity on a 8 GFDL-CM2.1, USA 3.4
model grid in longitude and latitude and, as a rule, at
17 standard ê-levels in the vertical. The model data 9 UKMO-HadCM3, UK 3.3
were linearly interpolated to a 5° longitude 4° latitude 10 ECHO-G, Germany/South Korea 3.2
grid and, for three-dimensional fields, at 17 standard 11 MRI-CGCM2.3.2, Japan 3.2
P-levels in the vertical, where data were possibly 12 CSIRO-Mk3.0, Australia 3.1
taken from a control experiment with the atmospheric
model and a 50-m ocean layer and from an analogous 13 GFDL-CM2.0, USA 2.9
experiment in which a doubled ëé2 concentration 14 CCSM3, USA 2.7
was specified. For those models for which data from 15 GISS-EH, USA 2.7
the experiments with a 50-m ocean were not available, 16 GISS-ER, USA 2.7
the data were derived from a numerical experiment on 17 INM-CM3.0, Russia 2.1
the simulation of a preindustrial climate with the
atmospheric and deep-ocean model. In addition, the 18 PCM, USA 2.1
data were used that were obtained from an experiment Note: The models are listed in descending order of climate sensi-
with a 1% year–1 ëé2 increase relative to the prein- tivity. More details on the models can be found at
http://www.pcmdi-llnl.gov.
dustrial value until ëé2 doubling (70 years) and for
the next 150 years with a doubled ëé2 concentration.
To examine the mean state and response to CO2 dou- near-average sensitivity. This, in turn, provides a statis-
bling, calculations in this case used the last 20 years tically more significant calculation of D.
of this numerical experiment and the corresponding For comparison with model data, the ISCCP D2
years of the preindustrial experiment. data set [6] and ERA-15 three-dimensional relative
To determine discrepancies in climate models with humidity data [7] have been used.
different sensitivity, composites were constructed for
the given high- and low-sensitivity models. The com-
posite of a quantity ï for high-sensitivity models, XH, RESULTS
was calculated in the following way: Consider the change in total cloud amount with a
NH NH doubling of ëé2 for high- and low-sensitivity models
XH = ∑ X ( ∆T
i i – [ ∆T ] ) ∑ ( ∆T i – [ ∆T ] ), (Fig. 1). Models with high sensitivity produce a
decrease in cloud amount of 4–6% over most of the
i=1 i=1
tropics and midlatitudes. At high latitudes, on the con-
where NH is the number of models with above-average trary, the cloud amount increases by 4–8%. Models
sensitivity, ∆Ti is the sensitivity of the model i, and [∆T] with low sensitivity predict only small changes in
is the sensitivity averaged over all models (about 3.2 K). cloud amount. In most of the region, the cloud amount
Likewise, the composite of ï for models with low sen- decreases by 1–2%. The change in cloud radiative
sitivity, XL, was calculated as forcing caused by a doubling of ëé2 in temperate and
NL NL
tropical latitudes (52° S–52° N) is 2.7 W m–2 for mod-
els with high sensitivity and 0.2 W m–2 for models
XL = ∑ X ( [ ∆T ] – ∆T ) ∑ ( [ ∆T ] – ∆T ),
i i i with low sensitivity. The standard deviation (SD) of
i=1 i=1 this quantity for all models is 1.25 W m–2, i.e., half the
where NL is the number of models with below-average difference between models with high and low sensi-
sensitivity. The residual of ï for high- and low-sensi- tivity. Given that the radiative forcing at the top of the
tivity models is D = XH – XL. This method provides a atmosphere due to ëé2 doubling is about 3.7 W m–2,
larger contribution to XH and XL of models with a sensi- the radiative forcing caused by ëé2 doubling and by
tivity differing from the average relative to models with cloud variation is 6.4 W m–2 for models with high sen-

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


290 VOLODIN

(‡)
4 4
2 8 6
64 2 8 6 2 4
4 2
60N 0 0 0
–2 –2 –2 –4 –2
30N –6 –4 –2 –4
–4–6 –2
–6 –6 –4 –4 –2
0
EQ –2 –2 0
–4 –2 –6
–4
–2 –4
30S –4 –4 –2
–2 –2 –4
–6 –6
–4 –4
–2 –4 –2 –2
4
60S 846 20
4 4 86 0
8
2 4 6 2 6 8
4 4
0 60E 120E 180 120W 60W 0
(b)
0
0 0
0
60N 0 0
0
0
30N –2 –2
0
0 –2
0 0
2 0
–2 –2
EQ –2
0
0 0
–2
30S 0 0 0
–2
–2
60S 0
0 0
0
0 2
2
0 60E 120E 180 120W 60W 0

–6 –4 –2 2 4

Fig. 1. Change in cloud amount (%) due to CO2 doubling in models with (a) high and (b) low sensitivity.

sitivity and 3.9 W m–2 for models with low sensitivity; oceans and fewer clouds over the tropics. Hence, the
that is, the value of the latter is a factor of 1.65 less difference in cloud amount between temperate and
than the former, which explains why the sensitivity is tropical latitudes in the present-day climate can be
lower by approximately the same amount. Consider treated as an index that must be correlated with the
now how the simulation of cloudiness in the present- model climate sensitivity. Then, as can be seen in
day climate is related to model sensitivity. Figure 2 Fig. 2, it is most convenient to use averaging either
shows the difference in cloud amount between models over temperate oceans of both hemispheres or over the
with high and low sensitivity. Models with high sensi- entire area of temperate latitudes, but only in the
tivity tend to have more clouds over midlatitude Southern Hemisphere. Figure 3 shows that three quan-

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


RELATION BETWEEN TEMPERATURE SENSITIVITY 291

–15 –15
–5
–15 –5–10 10
60N –10 –15 5
–10 –5 0
–10 5
0 –5 –5
5
30N 0 –5 –50
0 –10 –10 5
–5 0
–15 –5
–15
–5 –10 –10 –10
EQ –10 –10 –5
–15 –10
–20 –10
–5 –10 –10
–10 –5 –25 –10
0 –50 0 –10 –5
30S –5
5 5
5 10
10 –5
10 10 0
10 5
60S 5 05 0
0 –5
–10 5
–15 –20 –15 –20 –5 –5–10
–10
–25 –25 –15 –25 –15
–20 –25 –20
0 60E 120E 180 120W 60W 0

Fig. 2. Residual of cloud amount (%) between high- and low-sensitivity models. The residual exceeding the model SD is shaded
gray.

tities, (1) the difference in cloud amount between the doubled CO2 by a factor of 2. Given that the magni-
tropics and southern temperate latitudes, (2) climate tude of warming in the 52° S–52° N belt, where the
sensitivity, and (3) the change in cloud amount due to variation in cloud amount is considered, is below the
a doubling of CO2, are strongly correlated with one global average, the difference will be even smaller
another in all of the models. The coefficient of corre- than a factor of 2.
lation between the difference of cloud amount over the On the basis of the correlation between the differ-
tropics and southern temperate latitudes and climate ence of cloud amount in the tropics and southern tem-
sensitivity is –0.82. The coefficient of correlation perate latitudes and climate sensitivity and under the
between climate sensitivity and the change in cloud assumption that the data of a real climate system obey
amount due to CO2 doubling in the models is –0.80. the same law as the model-simulated data, it is possi-
Finally, the correlation between the change in cloud ble to estimate the sensitivity of a real climate from
amount due to CO2 doubling and the difference of the observed cloud data. According to observational
cloud amount over the tropics and southern temperate data, the difference in cloud amount between the trop-
latitudes is 0.75. It may be hypothesized that such a ics and southern temperate latitudes is about –25%,
high correlation is due to the fact that the change in which, as can be seen in Fig. 3a, corresponds to a cli-
cloud amount in the climate system owing to temper- mate sensitivity of 3.6 ± 0.3 K, where 0.3 K is the SD
ature variation is governed by approximately the same of model sensitivity from the best curve drawn
mechanism as that caused by CO2 doubling and occur- through the model points in Fig. 3a. The uncertainty
ring from temperate latitudes to the tropics. Indeed, range of this estimate can be increased owing to an
cloudiness changes normalized by the temperature uncertainty in the observed cloud amount, which is no
variation coincide in order of magnitude in both cases. more than 0.05 at each grid point in [8]. The corre-
For example, in the four most sensitive models, the sponding uncertainty in the estimate of climate sensi-
temperature change is about 4 K, while the change in tivity is about 0.4 K. However, since the calculation of
cloud amount caused by global warming is about –4% the difference in cloud amount between tropical and
(Fig. 3c); that is, the normalized value is about temperate latitudes involves averaging over large
−1%/K. The change in cloud amount from the tropics areas, the uncertainty in the observed value of sensi-
to temperate latitudes for these models is –35%, and tivity appears to be well below 0.05.
the change in surface temperature from the tropics to Consider how the vertical distribution of clouds
southern temperate latitudes is approximately 1.7 K; differs in models with high and low sensitivity
that is, the normalized change in cloud amount is (Fig. 4). Statistically significant differences are con-
−2%/K, which differs from an analogous quantity at centrated mainly in the tropics and subtropics. Models

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


292 VOLODIN

(‡) it is impossible to decide with confidence which class


2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 of models reproduce the observed vertical distribution
–5 of clouds more accurately. For example, the ISCCP
–10 data set [6] contains the distribution of (a) high
clouds, (b) middle clouds in the absence of high
–15 clouds, and (c) low clouds in the absence of middle
–20 and high clouds. To calculate the amount of, for exam-
ple, low clouds from these data, an assumption should
–25 be additionally made about the overlap of clouds at
–30 different levels. In turn, to convert model data to high,
middle, and low cloud amounts, an assumption should
–35
(b) also be made about the overlap of clouds at different
0.5 model levels. Attempts to do this demonstrate that the
0 result of comparison of the model data with observa-
–0.5 tions will depend on the assumption about the overlap
–1.0 of clouds in different layers.
–1.5
–2.0 However, the cloud amount in models is usually
–2.5 calculated from relative humidity, the observations of
–3.0 which are now available. Therefore, we compare rela-
–3.5
–4.0 tive humidity data for models with high and low sen-
–4.5 (c) sitivity with observational data.
–4.5 –4.0–3.5–3.0–2.5–2.0–1.5–1.0 –0.5 0 0.5 1.0 Figure 5 shows the latitude-circle-mean relative
humidity averaged over all models, the mean (over all
–5 models) error in relative humidity, and the difference
–10 in relative humidity between models with high and
–15 low sensitivity. In the tropics and subtropics, the dif-
ference in relative humidity between models with high
–20 and low sensitivity (Fig. 5c) is almost entirely analo-
–25 gous to the corresponding residual for clouds
–30 (Fig. 4c). According to Fig. 5b, the models in the trop-
–35 ics and subtropics tend to underestimate relative
–40 humidity in the lower troposphere and overestimate it
in the upper troposphere. This means that models with
Fig. 3. Data for 18 IPCC AR4 models: (a) climate sensitivity low sensitivity simulate relative humidity better in the
to CO2 doubling, K (abscissa) and the residual of cloud lower troposphere and models with high sensitivity
amount (%) over tropics (28° S–28° N) and southern temper- reproduce relative humidity more accurately in the
ate latitudes (56° S−36° S) (ordinate). Correlation coefficient =
–0.82. The circle shows an analogous difference from obser-
upper troposphere. In temperate latitudes, the models
vations (–25%) and the sensitivity estimate (3.6 ± 0.3 K), tend to overestimate relative humidity at almost all
where 0.3 K is the SD of model sensitivity from the best tropospheric levels.
curve drawn through the points; (b) climate sensitivity to
ëé2 doubling, K (abscissa) and cloud change (%) due to For each model, the projection P of the relative-
ëé2 doubling in the region 52° S–52 °N (ordinate). Correla- humidity error R onto the difference D of relative
tion coefficient = –0.80; (c) cloud change (%) due to ëé2 humidity for models with high and low sensitivity,
doubling (abscissa) and the difference in cloud amount (%) which is shown in Fig. 5c, is
between the tropics (28° S–28° N) and temperate latitudes

∑R
(56° S–36° S) (ordinate). Correlation coefficient = 0.75.
P= j, k D j, k cos ( ϕ j )∆P k w j, k ,
j, k
with high sensitivity tend to have a maximum of high where j and k are the latitude and pressure indices,
clouds at 100–200 hPa, and models with low sensitiv- respectively; ϕ is the latitude; ∆Pk is the thickness layer
ity have a maximum at 200–300 hPa. In the lower tro- in pressure coordinates; and wj, k is a weight function,
posphere, models with high sensitivity produce half equal to 1 for a shaded area in Fig. 5c and 0 for an
the amount of clouds of the model with low sensitiv- unshaded area. The projection of the error of relative
ity. At southern temperate latitudes in the lower tropo- humidity and climate sensitivity for 18 IPCC AR4
sphere and in the equatorial middle troposphere, mod- models are shown in Fig. 6. The correlation coefficient
els with high sensitivity predict more clouds than between the projection and climate sensitivity is 0.75.
those with low sensitivity. However, data on the verti- The observed relative humidity, or a zero error projec-
cal distribution of clouds that are analogous to the tion, corresponds to a sensitivity of 3.4 ± 0.3 K, where
model data shown in Fig. 4 are unavailable; therefore, 0.3 K is the SD of model sensitivity from the best curve

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


RELATION BETWEEN TEMPERATURE SENSITIVITY 293

(‡)

100 102530
5 15
2000 15 10
20 10
300
5 5
400 25
500 20 20
15
600
20
700 20
800 25 20
900
30 5 25
1000
0 60S 30S EQ 30N 60N
(b)
35
100 1510 20
2000 5 25
10 15 30
300 15
Height, hPa

20 10 15 20
400 25
5
500
600 20
20
700
25
800 5 15
25
900 10 10
1000 30 10 10
0 60S 30S EQ 30N 60N
(c) 5
0 10 10 0
100 18 20
6 6
2000 –2 6
–6 2
300 8 4 6 4
400 6 2 –2
–4 0
500 0 –4
–2 0
600 0
2 –6
700 –2 4
–4 0 2
800
–4 –2
900 0
6 –4 –6
1000
0 60S 30S EQ 30N 60N

Fig. 4. Cloud amount (%) averaged over a circle of latitude in models with (a) high and (b) low sensitivity and (c) their residual.
The difference exceeding the model SD is shaded gray.

drawn through the model points in Fig. 6. Such an esti- the tropical and temperate troposphere, particularly in
mate corresponds to that obtained from the observed the upper tropical troposphere, where the decrease
distribution of clouds. reaches 5–6%. At the same time, there is a growth of
Consider how the vertical profile of cloud amount cloud amount at the tropopause. In models with low
changes with global warming in models with high and sensitivity, cloud changes are mostly similar, but less
low sensitivity (Fig. 7). In models with high sensitiv- in magnitude, while cloud changes in the lower tropi-
ity, the cloud amount decreases almost everywhere in cal and subtropical troposphere are close to zero. The

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


294 VOLODIN

(‡)
5 15
100 505545 5
15 10
2000 25 20 40 35
45 45 90
300 55 30 50 35 40 50
60
400 30
80
500
600 70
700 35 65
65 4045 70
800 5055 45 60
75 50
80 60 65 55
900 60
70 50
1000
0 60S 30S EQ 30N 60N
(b) 40
0 2 0
100 –10 0 –8–6 2 6
14 6 2 0 1481216
2000 2418 20 22 30
8 2428
26
28
26 6 8
300 18
16
14
10 10
8 12 10 8 12 20
Height, hPa

400 6 6
4 2 10
500 0
2 –2 –2
600 –4 8
700 4 6
–4
10
800 4
2
900 0 2 –6
6 5
1000 0 –2
0 60S 30S EQ 30N 60N
(c) –5
0 20
46 –2
100 –4680 –10
4
2000 –8
–6 –2–4 8 2
–20 0
300 2
4
400
–10 0
500 –8 –8
4 –6 –2
600 –4
700 –2
4
800 2 0
–2
–6
900 –4 –2 –2
–4 0
1000
0 60S 30S EQ 30N 60N

Fig. 5. (a) Relative humidity averaged over all models, (b) error from the ERA-15 data, and (c) the difference in relative humidity
between high- and low-sensitivity models. The difference exceeding the model SD is shaded gray.

discrepancy between models with high and low sensi- similar to cloud changes. The distinction is that the
tivity is statistically significant (Fig. 8c) in the upper relative humidity increases slightly in the tropical
and lower tropical and subtropical troposphere, where middle troposphere, especially in high-sensitivity
models with high sensitivity predict a larger decrease models, whereas cloud changes are close to zero there.
in cloud amount, and at the tropopause and in the With warming induced by a doubling of ëé2 and
lower stratosphere, where models with high sensitiv- warming in the tropics relative to midlatitudes, not
ity produce a larger increase in cloud amount. The only do similar changes occur in the total cloud
analogous changes in relative humidity are generally amount, but they also occur in the vertical profile of

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


RELATION BETWEEN TEMPERATURE SENSITIVITY 295

relative humidity (Fig. 9) and cloud amount (Fig. 10). 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4
One cannot infer with confidence that changes in 80
cloud amount and relative humidity are entirely iden- 60
tical in both cases, but there is some similarity between 40
them. In models with high sensitivity, under global 20
0
warming and under warming in the tropics relative to –20
midlatitudes, the cloud amount decreases most substan- –40
tially in the upper troposphere (300–400 hPa) and at the –60
top of the boundary layer (850–900 hPa). The decrease –80
is not as large in the middle troposphere (600– –100
700 hPa). There is an increase in cloud amount at the –120
tropopause, near 100 hPa. In models with high sensi-
tivity, the changes in relative humidity are similar to Fig. 6. Climate sensitivity (abscissa) and the error projec-
those in cloud amount. At the same time, models with tion of zonally averaged relative humidity onto the differ-
low sensitivity predict the largest decrease in cloud ence shown in Fig. 5c (ordinate). Data for 18 IPCC AR4
amount and relative humidity in the middle tropo- models. The circle shows observational data (zero error pro-
jection and sensitivity 3.4 ± 0.3 K, where 0.3 K is the SD of
sphere (400–700 hPa) and an increase at 100–200 hPa. model sensitivity from the best curve drawn in the figure).
The absolute cloud and relative-humidity changes
normalized by temperature changes, as in the case of
total cloudiness, are approximately identical under mate sensitivity is associated with smaller cloud
global warming and from midlatitudes to the tropics. amounts in the tropics than in southern temperate lat-
As in the case of total cloudiness, changes from mid- itudes and with low relative humidity in the tropics. In
latitudes to the tropics are about 1.5 times as large as models with high sensitivity, the largest decrease in
those induced by global warming. cloud amount and relative humidity in the tropics in
We now compare the vertical profiles of relative comparison with temperate latitudes and under global
humidity from midlatitudes to the tropics according to warming on the entire planet occurs in the upper tro-
the model data and observations. Interestingly, the posphere (300–400 hPa) and in the lower troposphere
observational data differ from the data of both high- (850–900 hPa). On the basis of this result and the
and low-sensitivity models. The model-simulated rel- observed distribution of clouds and relative humidity,
ative humidity in the lower troposphere is smaller in an estimate is obtained for the observed sensitivity of
the tropics than at temperate latitudes (Fig. 9b). This climate. It is 3.6 ± 0.3 K from the cloud data and 3.4 ±
is true for both the composite of models with high sen- 0.3 K from the relative humidity data.
sitivity and the composite of models with low sensitivity,
However, these estimates should be used with cau-
as well as for each model separately. At the same time,
tion for the following reasons. In models with high
according to the ERA-15 data, the relative humidity
sensitivity, the cloud amount under global warming
below 600 hPa has approximately the same value in the
decreases most substantially in the tropical upper tro-
tropics as in temperate latitudes and is lower than in tem-
posphere and less in the boundary layer. However, as
perate latitudes only in the 300–500 hPa layer. The
shown by numerical experiments with the INMCM3.0
increase in relative humidity in the 100–200 hPa layer
model, a decrease in the tropical high-level cloud
from the reanalysis data is generally consistent with
amount has no large impact on model sensitivity,
results from both high- and low-sensitivity models.
because the radiative forcing induced by this decrease
is close to zero and is composed of a positive short-
DISCUSSION AND CONCLUSIONS wave forcing and a negative longwave forcing of
about the same value. At the same time, models with
It is shown that there is a similarity in the IPCC high sensitivity differ from those with low sensitivity
AR4 models between the change in cloud amount due primarily in a change in shortwave radiative forcing.
to ëé2 doubling and the difference in cloud amount
This change due to a doubling of ëé2 is 3.7 W m–2 for
between the tropics and southern temperate latitudes.
This is true for total cloudiness and for vertical pro- high-sensitivity models and 0.8 W m–2 for low-sensitiv-
files of cloudiness and relative humidity. Because it is ity models. The SD of this quantity for all models is
a different response of cloudiness that largely deter- 1.55 W m–2, or almost half the difference for high- and
mines the discrepancy in climate sensitivity to a dou- low-sensitivity models. The change in longwave radia-
bling of ëé2 in various models, a correlation exists tive forcing through a doubling of ëé2 is –1.0 W m–2 in
between the difference of cloud amount over southern high-sensitivity models and –0.6 W m–2 in low-sensitiv-
temperate latitudes and the tropics in the present-day ity models, with the SD for all models of 0.5 W m–2.
climate and the sensitivity of climate. An analogous Thus, the discrepancy of variations in cloud radiative
correlation exists between the distribution of relative forcing for high- and low-sensitivity models is due
humidity and climate sensitivity. Namely, the high cli- mainly to a shortwave component. This is indicative

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


296 VOLODIN

(‡)
2 76
100 30 1 –2 1 –10 2
2000 3 3
–2 –5 –3 1
300 2 1
400 0 –1
500 –2
600
700 –1
0
800 –1 4
900 –2
1000 0 2
0 60S 30S EQ 30N 60N
(b) 1
100
1 –
2000 1 1 0 1
–1 1
300 0
Height, hPa

–2
400 –1 –1
1 –1
500 –4
600
700
800 0 0 –1
–2
900
1000
0 60S 30S EQ 30N 60N
(c)
1 76 0
100 2 1 –1 1
–4–3
3 1
2000 2
1 –2 –5 10
300 0 1
0 –1 –2
400 –1
500 0
600 0
700 1 0
800 2
–1 –1
900 –1
–1
1000
0 60S 30S EQ 30N 60N

Fig. 7. Cloud change (%) due to ëé2 doubling averaged over a circle of latitude for models with (a) high and (b) low sensitivity
and (c) the residual of these fields. The difference exceeding the model SD is shaded gray.

of a large role of the surface cloudiness variation in a the relative humidity in the tropical upper troposphere
different change of cloud radiative forcing in high- from observational data is overestimated in all mod-
and low-sensitivity models. els. The same can evidently be said of cloudiness. In
the lower troposphere, in contrast, the relative humid-
The estimates of climate sensitivity from observa- ity in the tropics is underestimated in most of the mod-
tional data are somewhat overstated in comparison els, the difference in relative humidity between the
with the average over all models, primarily because tropics and southern temperate latitudes is negative,

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


RELATION BETWEEN TEMPERATURE SENSITIVITY 297

(‡)
3 2 81010
14
4 0 123
100 7 65 81 –1 56 8 3
9 10
–3 –6 11
12
2000 87
645
300 3 2
1 2
–1 –4
400 0 01
2 –1
500 –3
600 –1 –4
0 3
700 0
–3
800 1
–1 –1 –2
–2
900
0
1000
0 60S 30S EQ 30N 60N
(b)
1
100 32 11 4
5 4 –1 5
2000 32 1
300 0 0 2
0
–1
Height, hPa

400 1 –1 0
1 0 1
500
600
–1
700 0 –2
800 0 –2
–1 –1 0
900
1000 –4
0 60S 30S EQ 30N 60N
(c)
1 2
100 1 11 3 11 2
4 3 –2 345
5 5 43 6
2000 –5 7
34 2 –4 –6
300 1 –2
0 01
400 –3
500 –1 –3 2 –2
600 –1
700 0 0
1
800 0
900 1 –1
0 –1 –1
1000
0 60S 30S EQ 30N 60N

Fig. 8. Relative humidity change (%) due to ëé2 doubling averaged over a circle of latitude for models with (a) high and (b) low
sensitivity and (c) the difference of these fields. The difference exceeding the model SD is shaded gray.

and the reanalysis data are close to zero. This circum- In [4], it is concluded that high-sensitivity models
stance means that, if the sensitivity of a real climate tend to have small amounts of subinversion clouds at
system is estimated only from data for the lower tro- the top of the boundary layer, while low-sensitivity
posphere, where cloud variations have the largest models should have large amounts of subinversion
effect on sensitivity, then the sensitivity of a real cli- clouds. This result does not contradict the main find-
mate must be below the average over all models. ing of the given paper, because, as shown by numeri-

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


298 VOLODIN

Pressure, mb (‡) Pressure, mb (‡)

100 100
200 200
300 300
400 400
500 500
600 600
700 700
800 800
900 900
1000 1000
(b) (b)
100 100
200 200
300 300
400 400
500 500
600 600
700 700
800 800
900 900
1000 1000
–2.0 –1.5 –1.0 –0.5 0 0.5 1.0 0.5 2.0 2.5 3.0 –1.2 –0.9 –0.6 –0.3 0 0.3 0.6 0.9 1.2

Fig. 9. (a) Change in the vertical profile of relative humidity


due to ëé2 doubling in the region 52° S–52° N normalized Fig. 10. (a) Change in the vertical profile of cloud amount
by the surface temperature change (%/K). (b) Difference in due to ëé2 doubling in the region 52° S–52° N normalized
relative humidity between the tropics (24° S–24° N) and by the surface temperature change (%/K). (b) Difference in
southern temperate latitudes (56° S–36° S) normalized by cloud amount between the tropics (24° S−24° N) and south-
the surface temperature difference between the tropics and ern temperate latitudes (56° S−36° N) over oceans normal-
temperate latitudes (%/K) for high-sensitivity models (thin ized by the surface temperature difference between the tropics
solid line), low-sensitivity models (dashed line), and obser- and southern temperate latitudes (%/K) from (thin solid line)
vations (thick solid line). high-sensitivity and (dashed line) low-sensitivity models.

cal experiments with the INMCM3.0 model, the intro- vection intensity, etc. It is this class of models that
duction of a cloud increase under inversion conditions includes the INMCM3.0, which has one of the lowest
provides larger total cloud amounts in the tropics and sensitivities. A considerable decrease in relative
subtropics than in temperate latitudes. However, vary- humidity owing to warming can be observed, for
ing the subinversion cloud amount is not the only example, in models that use a mass flux scheme for
mechanism of formation of a different sensitivity in convection parametrization [9]. In such models, the
models. For example, neglecting subinversion clouds cell is split into a cloudy part, with a 100% relative
in the INMCM3.0 model increases sensitivity from humidity, and a cloudless part, in which the relative
2.1 to 2.5 K; that is, the model sensitivity is still below humidity is the lower, the more intense downdrafts in
the average over all models even after additional sub- the cell. Subgrid-scale updrafts in the cloudy part and
inversion clouds are eliminated. downdrafts in the cloudless part must cancel each
Let us explain qualitatively what could induce a other, so that the total vertical subgrid-scale mass flux
significant decrease in relative humidity under global would be zero. If we assume that the share of the
warming in high-sensitivity models and a small cloudy and cloudless part is independent of tempera-
change in relative humidity and cloudiness in low- ture and that the intensity of the updrafts in the cloudy
sensitivity models. Small changes in relative humidity part of the cell increases with global warming, the
during global warming can take place in models total relative humidity after the introduction of con-
where, after the implementation of convection param- vection parametrization must be reduced owing to a
etrization, the relative humidity is prescribed explic- decrease in relative humidity in the cloudless part with
itly and is independent of temperature, moisture, con- increasing intensity of the downdrafts.

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008


RELATION BETWEEN TEMPERATURE SENSITIVITY 299

ACKNOWLEDGMENTS Balance on the Earth’s Surface at Increased Contents of


Carbon Dioxide,” Izv. Akad. Nauk, Fiz. Atmos. Okeana
This study was supported by the Russian Founda- 40, 306–313 (2004) [Izv., Atmos. Ocean. Phys. 40, 269–
tion for Basic Research, project nos. 06-05-64331 and 275 (2004)].
06-05-64246.
5. G. A. Meehl, G. J. Boer, C. Covey, et al., “The Coupled
Model Intercomparison Project (CMIP),” Bull. Am.
REFERENCES Meteorol. Soc. 81, 313–318 (2000).
1. D. Randall, A Wood, S. Bony, et al., “Climate Models 6. W. B. Rossow and R. A. Schiffef, “ISCCP Cloud Data
and Their Evaluation,” in IPCC WG1 Assessment Report Product,” Bull. Am. Meteorol. Soc. 72, 2–20 (1991).
(Cambridge Univ. Press, Cambridge, 2007) (in press). 7. S. M. Uppala, P. W. Kallberg, A. J. Simmons, et al., “The
2. B. J. Soden and I. M. Held, “An Assessment of Climate ERA-40 Re-Analysis,” Q. J. R. Meteorol. Soc. 131
Feedbacks in Coupled Ocean–Atmosphere Models,” J. (612), 2961–3012 (2005).
Clim. 19, 3354–3360 (2006). 8. W. B. Rossow and R. A. Schiffer, “Advances in Under-
3. R. Knutti, G. A. Meehl, M. R. Allen, et al., “Constraining standing Clouds from ISCCP,” Bull. Am. Meteorol. Soc.
Climate Sensitivity from the Seasonal Cycle in Surface 80, 2261–2288 (1999).
Temperature,” J. Clim. 19, 4224–4233 (2006). 9. M. Tiedtke, “A Comprehensive Mass Flux Scheme for
4. E. M. Volodin and CMIP Member Group, “Relation Cumulus Parameterisation in Large-Scale Models,”
between the Global-Warming Parameter and the Heat Mon. Weather Rev. 117, 1779–1800 (1989).

IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 44 No. 3 2008

You might also like