Content uploaded by Oleg Pokrovsky
Author content
All content in this area was uploaded by Oleg Pokrovsky on Jun 11, 2021
Content may be subject to copyright.
Content uploaded by Oleg Pokrovsky
Author content
All content in this area was uploaded by Oleg Pokrovsky on May 19, 2021
Content may be subject to copyright.
1189
ISSN 0001-4338, Izvestiya, Atmospheric and Oceanic Physics, 2019, Vol. 55, No. 9, pp. 1189–1197. © Pleiades Publishing, Ltd., 2019.
Russian Text © The Author(s), 2019, published in Issledovanie Zemli iz Kosmosa, 2019, No. 1, pp. 3–13.
Cloud Changes in the Period of Global Warming:
The Results of the International Satellite Project
O. M. Pokrovsky*
Russian State Hydrometeorological University, St. Petersburg, Russia
*e-mail: pokrov_06@mail.ru
Received September 5, 2018
Abstract—We present the results of an analysis of climatic series of global and regional cloudiness for 1983–
2009. The data were obtained as part of the ISCCP international satellite project. A technique of statistical time
series analysis that includes a smoothing algorithm and wavelet analysis is described. Both methods are intended
for the analysis of nonstationary series. According to the results of analysis, both global and regional cloudiness
show a decrease of 2–6%. The greatest decrease is observed in the tropics and over the oceans, while the
decrease is minimal over land. The correlation coefficient between the global cloud series on the one hand and
the global air and ocean surface temperature series on the other hand reaches values between –0.84 and –0.86.
The coefficient of determination that characterizes the regression accuracy for the prediction of global tempera-
ture variations based on the variations in the lower cloud cover in this case is 0.316.
Keywords: climatology, global and regional cloudiness, climate series, ISCCP, linear and nonlinear trends,
wavelet analysis
DOI: 10.1134/S00 01433819 0 9 03 66
INTRODUCTION
Cloudiness is a phenomenon that everyone
encounters every day of their life. It is also one of the
key meteorological elements. Cloudiness has a deci-
sive influence on the energy balance of the Earth,
since it determines the arrival of solar radiation and
regulates the outgoing thermal radiation. The global
water cycle is initiated by precipitation from the
clouds. Nevertheless, the physics of the formation and
evolution of clouds still remains an understudied field
of knowledge.
The total reflectivity (albedo) of the planet Earth is
approximately 30%, which means that approximately
30% of the incoming shortwave solar radiation is
reflected back into space. If all clouds were removed,
the global albedo would decrease to 15% and the
amount of shortwave energy available to heat the sur-
face of the planet would increase from 239 to 288 W/m2
(Hartmann, 1994). In this hypothetical case, long-
wave radiation would also be affected, from which
266 W/m2 would escape into space, as compared to
the present 234 W/m2 (Hartmann, 1994). Thus, the
total effect of removing all clouds would still lead to an
increase in the heat influx, characterized by a value of
approximately 17 W/m2. Thus, the global cloud cover
has a distinct overall cooling effect on the planet,
although the pure effect of high and low clouds is the
opposite.
Lower layer clouds, as a rule, have a cooling effect
on the global climate. They often have a significant
optical thickness and reflect most of the incoming
shortwave radiation. In addition, because of their low
altitude and high temperature, they generate a large
amount of longwave radiation that goes into space and
into higher levels of the atmosphere. Conversely, the
upper layer clouds, as a rule, provide a warming effect,
since they, due to their great altitude and low tempera-
ture, emit less longwave radiation into space. In addi-
tion, these clouds are usually thin and only slightly
reflect the incoming shortwave radiation. This consid-
eration is not purely theoretical, but demonstrated by
observations, which will be discussed further.
Observations of clouds from meteorological sta-
tions have a number of serious disadvantages. First,
the cloud point is assessed subjectively by the observer.
In addition, given that the scale of horizontal cloudi-
ness variability fluctuates from kilometers to tens of
kilometers, the effective coverage of the Earth’s terri-
tory by means of the ground-based meteorological
network amounts to hundredths of one percent.
The appearance of meteorological satellites funda-
mentally changed the situation regarding the collec-
tion of information on the spatial and temporal distri-
bution of clouds on a global scale. The continuous
operation of the Meteosat geostationary satellites was
especially important, since it provided continuous
time tracking of each cloud cover fragment with a high
USE OF SPACE INFORMATION ABOUT THE EARTH
ИССЛЕДОВАНИЕ АТМОСФЕРНЫХ ПРОЦЕССОВ
И ИЗМЕНЕНИЙ КЛИМАТА ПО КОСМИЧЕСКИМ ДАННЫМ
1190
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
POKROVSKY
spatial resolution. The use of multichannel satellite
remote sensing (RS) equipment provided a more
informed analysis of the vertical distribution of clouds.
For the first time, it became possible to begin climate
generalizations of cloud-cover data on a global scale.
However, a comparison of the results of analyzing
cloud fields obtained from different satellites and
using different processing algorithms showed signifi-
cant discrepancies, which were especially noticeable
in the polar region (Chernokul’skii, 2012; Cher-
nokulsky and Mokhov, 2012).
To implement the most important task of coordi-
nation and uniform processing of cloud observations
from satellites, the International Satellite Cloud Cli-
matology Project (ISCCP) was launched in 1982 in
the United States (Schiffer and Rossow, 1983). This
project was initiated as part of the WMO World Cli-
mate Research Program (WCRP). The goal of the
project was to collect and analyze radiation satellite
measurement data in order to monitor the global dis-
tribution of clouds and obtain information on the
physical properties of clouds, as well as their daily, sea-
sonal, and interannual variability. Data collection and
analysis was started on July 1, 1983, and continues to
the present time. It should be noted that the author of
(Norris, 2000) criticized the cloud-processing algo-
rithm of the ISCCP project regarding the angular cor-
rection of observational data.
To date, a number of interacting computer and sci-
entific experimental centers have been formed within
the ISCCP. The primary data-processing center, the
Satellite Processing Center (SPC), sends its results to
the Global Processing Center (GPC), where the data
are integrated according to spatial and temporal
dependences. At the next stage, the data are georefer-
enced in the Correlative Data Center (CDC). Further,
the data are calibrated and verified in the Satellite Cal-
ibration Center (SCC). At the final stage, the data are
archived in the ISCCP Central Archive (ICA) and in
the NASA Langley Research Center (LARC). At pres-
ent, the observational data are published on the web-
sites of ISCCP, NCEP, and NOAA, as well as a num-
ber of worldwide research centers, and are available to
all scientists and specialists. By 2011, a compilation of
global cloud data for climate studies with a monthly
time resolution in the period of 1983–2009 had been
performed.
The present study is dedicated to an analysis of cli-
matic cloud series on a global and regional scale that
were obtained within the ISCCP project. The first part
of the paper describes the original method for analyz-
ing climatic series, including a smoothing procedure
and the study of the spectral components of the series.
Further, the results of studying the trends in the global
cloud field are presented, as well as the series for a
number of regions and latitudinal zones. In conclu-
sion, a discussion of the results in connection with cli-
mate change over the past three decades is presented.
INITIAL DATA AND METHODS OF ANALYSIS
As part of the project, global cloud data were col-
lected from 30 US meteorological satellites: GMS 1–5;
GOES 5–12; METEOSAT 2–9; and NOAA 7, 10–12,
14–18. A specialized processing of the measurement
data was carried out in accordance with the scheme
indicated above. The publicly available climate data
arrays on clouds are presented as fractions of cloud
cover (%) at regular grid nodes with a step of 280 km
and a time resolution of 3 h. The monthly data array
size is 216 MB. Each file of average monthly values is
7.5 MB. It should be noted that there are also initial
cloud fields that have a spatial resolution of 30 km. In
that case, each monthly data set takes 1.1 GB. All data
and their description are freely available at https://
isccp.giss.nasa.gov/products/onlineData.html. Along
with the cloud data, the ISCCP archives contain
related data on air temperature and humidity.
An important feature of any climatic series is their
nonstationarity, i.e., variations not only in the charac-
ter of the behavior, but also in the statistical structure
depending on the time coordinate.
This circumstance requires the development of new
methods for analyzing such series. Traditional methods
allowed one to work with stationary series. Below, we
will consider some alternative approaches that allow
extracting more meaningful information from climate
series. Special attention is paid to filtering the interan-
nual variability called “climate noise” and identifying
slow fluctuations commonly referred to as trends. So
far, there has usually been a discussion of linear trends
that characterize a monotonous decrease or increase in
climatic characteristics. If variations in the climate indi-
cators deviate from monotonous behavior, the linear
trend technique becomes ineffective. For this reason,
we propose a method of nonlinear smoothing, which
allows one to track variations in the trends of climate
series more accurately. Another traditional method for
analyzing the spectral composition of series, Fourier
analysis, is also intended for the study of stationary pro-
cesses. For this reason, we propose using a more mod-
ern method of wavelet analysis, which provides infor-
mation on the spectral characteristics of the climatic
series that vary over time.
The analysis of climatic series usually involves
using a linear trend technique to estimate the general
trend of variations over a given time interval and the
sliding average method for filtering high-frequency
fluctuations associated with interannual variability,
which is considered climate noise. The linear trend
technique is effective when the process develops more
or less monotonously. If the tendency of the process is
broken, the estimates and even the trend sign become
dependent on the choice of the base time interval. In
the analysis of long nonstationary series, the trends
and trend signs change repeatedly. In such circum-
stances, the use of a linear trend is clearly insufficient
for a meaningful analysis of the series.
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
CLOUD CHANGES IN THE PERIOD OF GLOBAL WARMING 1191
To identify statistically significant variations in the
trends of climatic series, we developed and tested a
method for estimating a nonlinear trend (Pokrovsky,
2004; Pokrovsky, 2009a; 2009b; 2009c). The pro-
posed approach is based on a combination of three
well-known methods: (1) Cleveland local polynomial
smoothing (Cleveland, 1979), (2) Tikhonov regular-
ization (Tikhonov, 1963), and (3) optimization of the
smoothing procedure based on the Wahba cross-vali-
dation criterion (Wahba, 1985). Instead of the usual
minimization of deviations of the yi measurements
from a smoothed curve obtained using spline (local
polynomial) f by means of the least-square method, it
is proposed to use a smoothing functional
(1)
which depends on the smoothing parameter λ and on
the square of the second derivative of the smoothing
function f. This, as a rule, includes the use of local
polynomials of the minimum degree with a choice of
“influencing” nodes ti, which are selected using cross
validation. The choice of λ and the approximating
polynomial (spline) is based on minimizing the Wahba
cross-validation criterion:
(2)
Here, is the approximation obtained by excluding
the results of measurements of yi at the ith instant of
time. Thus, the smoothing procedure based on solving
the (n + 1)th equation with the (n + 1)th unknown
determines the set of local polynomials of the minimal
degree, including “influencing nodes” ti and the λ
value, which give the minimum error when recovering
“missing” data.
The principal difference of the described proce-
dure (1)–(2) from the sliding average method is that
here, in addition to smoothing using (1), the “influ-
encing nodes” are selected based on criterion (2). Let
us recall that, in the case of a sliding average, all nodes
are used, regardless of the behavior of the time series.
Thus, the proposed method turns out to be more flex-
ible with respect to changes in the statistical structure
of the time series.
The standard Fourier analysis of time series (like
the sliding average method) is focused on application
to stationary series. Deviation from stationarity entails
the dependence of the Fourier spectra on the basic
interval of the analysis. The Fourier spectra depend on
the phase of the process, which varies in the nonsta-
tionary case. In order to fully cover all the features of a
nonstationary process, a method was developed to
obtain wavelet spectra, which, unlike one-dimen-
sional Fourier spectra, turn out to be two-dimen-
sional. For each value of the time coordinate t, the
wavelet spectrum gives an idea of the usual Fourier
λ= − +λ
∑∫
1
22
(,) { ()} ''() ,
n
t
ii
it
Sf y ft f t dt
−
=
λ= −
∑
v
() 2
1
1ˆ
() { () }.
n
i
ii
i
ft y
n
−
()
ˆ
i
f
analysis. Let us provide a brief description of this
method of spectral analysis of time series.
Let us consider the time series x(t). The wavelet
transform of this series has the following general form:
(3)
(* is the complex conjugation symbol) with the spec-
tral transform function
(4)
which depends on two variables: regular spectral vari-
able τ and scaling variable s. The most common is the
Morley spectral transform function (4), which looks as
follows:
(5)
Formulas (3)–(5) provide the algorithm for calcu-
lating the wavelet spectra of the time series. The corre-
sponding finite-difference approximation of the inte-
gral in (3) allows obtaining a method for calculating the
wavelet spectra for the climatic series. There is a wide
range of literature on this subject. We can recommend,
for example, the paper (Goupillaud et al., 1984), which
features a detailed bibliography on this subject.
Now that we have the above mathematical appara-
tus for the study of nonstationary processes, let us pro-
ceed to applying it to the study of climatic series.
DYNAMICS OF THE GLOBAL CLOUD COVER
As part of the ISCCP international project, the
data on the distribution of cloudiness by the lower,
middle, and upper levels were obtained (Rossow and
Schiffer, 1991). In addition, spatial and temporal dis-
tributions of total cloudiness were obtained. All esti-
mates are given in terms of cloud coverage, i.e., as a
percentage of the spatial coverage of a given territory at
a certain point in time (Rossow and Garder, 1993a). In
the case of a monthly generalization, this means the
average monthly fraction of cloud cover for a given ter-
ritory (Rossow and Garder, 1993b). A number of for-
eign studies on the analysis of the ISCCP cloud data
and identification of the links between cloud dynamics
and major climate indices should be noted (Lindzen
et al., 2001; Lindzen and Choi; 2009, 2010; Pielke et al.,
2007; Spencer et al., 2007). In the present study, we
will focus on analyzing the total cloudiness data
(https://isccp.giss.nasa.gov/products/onlineData.html).
In this section, we consider the data on the dynamics
of the fraction of global cloud coverage, which are
averaged over monthly and annual intervals and are
expressed as a percentage.
Figure 1a presents the climatic series of annual
global total cloudiness for 1983–2009 (indicated by
crosses). In addition, the curve of the regression
ψτ
ψτ = ψ
∫,
*
(, ) () ()
xs
sxttdt
(
)
τ
−τ
ψ= ψ
,
1
,
s
t
s
s
−
σ
⎛⎞
⎛⎞
ψ=
⎜⎟
−
⎜⎟
πσ
⎝⎝ ⎠⎠
σ
2
22
2
32
1
() .
1
2
t
t
te
1192
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
POKROVSKY
approximation of the interannual variability of global
cloudiness by a third-degree polynomial and the cor-
responding confidence intervals for the results of the
regression analysis is given here. The fraction of the
average monthly global cloud coverage varies, on aver-
age, from 63 to 70%. At the same time, the linear trend
shows a decrease in the total cloudiness from 68 to
64.6% over the specified period. The decrease
amounts to 3.4%. Figure 1b shows the results of
smoothing using the original method outlined above
(solid line).
Figure 1 and several following illustrations show
the varying confidence intervals of scatter in the
monthly mean values with a 10% level of statistical sig-
nificance. The curve of the nonlinear trend makes it
possible to distinguish a period of increase in cloudi-
ness in 1983–1986 and a subsequent monotonous
decrease until 2000. This is followed by a slight growth
until 2004, which turns into a decrease in the period of
2005–2009. In 1986–2000, nonlinear approximation
demonstrates a decrease in the total cloudiness by 4%,
which is slightly higher than the results for the linear
trend. The width of the confidence intervals allows
one to judge the intra-annual scatter of monthly mean
values for a given year.
Thus, using the example of comparing the tradi-
tional regression method with smoothing by the
author’s method, it is shown that we achieve two
improvements to the standard approach: (1) a descrip-
tion of the localization of statistically significant min-
imums and maximums and (2) a reduction of the
width of the confidence intervals.
The calculations in (Scafetta, 2009) show that the
observed 4% decrease in cloudiness over the 13-year
period of 1987–2000 is equivalent to an increase in the
flux of incoming solar radiation by 0.9 W/m2. It is inter-
esting to note that, according to the IРСС-2007 report,
the growth of incoming solar radiation for the period
1750–2006 after the Little Ice Age was 1.6 W/m2.
Thus, over the past three decades, there has been
an increase in incoming solar radiation comparable to
the period of increase in solar activity from its histori-
cal minimum values. The question arises of how the
mentioned change in the Earth’s cloud cover was dis-
tributed across regions.
CHANGES IN THE REGIONAL
CLOUD COVER
The data obtained on the ground-based actinomet-
ric network demonstrate quite a varied picture in ana-
lyzing the trends of climatic series of incoming solar
radiation. So far, there has been no systematization of
such studies. The main reason is that devices that carry
out actinometric observations in different countries of
the world are still not standardized. This hinders the
development of a unified approach to solving the
problem of standardization of databases. In addition,
the actinometric network is mainly located on land,
while most of the Earth’s territory is covered by
oceans. Therefore, satellite observations become the
only option in studying global climate changes (Ros-
sow et al., 2002). Given that the scale of the cloud-
field variability averages at no more than 30 km, the
ground meteorological network of Roshydromet, con-
sisting of approximately 1600 stations, can cover no
more than a few hundredths of one percent of the ter-
ritory of Russia (Pokrovsky, 2004).
Figure 2 presents summarized data on the dynam-
ics of the cloud cover over land. The linear trend shows
a decline in the fraction of the cloud coverage from
58.3 to 56%. Thus, the decrease in the cloud cover
over land is smaller than on a global scale. The nonlin-
ear trend shows that the tendencies of variations in the
cloud cover over land have the most pronounced time
Fig. 1. Analysis of the series of the total global cloudiness (%)
from the ISCCP data: (a) the crosses are the initial data and
the curve is the nonlinear trend (the result of smoothing the
series by the cubic regression); (b) nonlinear trend (the result
of smoothing the series by the nonlinear algorithm) and con-
fidence intervals for the 10% level of significance.
63
69
68
67
66
65
64
1980 19901985 1995 2000 20102005
%
Years
(b)
62
63
69
70
68
67
66
65
64
199 01985 1995 2000 2005
%
(a)
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
CLOUD CHANGES IN THE PERIOD OF GLOBAL WARMING 1193
dependence in the 1990s. The decrease in the cloud
coverage from 57 to 54% occurred over the 5-year
period in 1989–1994. This was the period of the great-
est pace of global warming. In the rest of the time, a
decrease in the cloud cover over land occurs more
evenly in time. The values of the cloud cover over land
is noticeably lower than the global values. However,
the interannual variability of the width of the confi-
dence intervals over land exceeds the corresponding
global variability. This is probably due to the interan-
nual variability of cyclonic activity over the continents.
Figure 3 shows the corresponding values of the
total cloudiness for all seas and oceans. Here, the
cloudiness values are higher than over land by approx-
imately 15%. Above the water areas, a decrease in
cloudiness is also more noticeable. According to the
linear trend, the total cloudiness is reduced by 4%:
from 72.5 to 68.5%. The confidence intervals of varia-
tions in temporal trends in the dynamics of cloudiness
over water areas appear to be similar to the dynamics
of the global trends (Fig. 1a).
The tropical belt is of increased interest, since it is
here that the moisture exchange between the atmo-
sphere and the ocean is realized most intensively.
Cloudiness in this case plays the role of a mediator.
Therefore, it is very important to know how it varies in
the tropics (Sud et al., 1999). The linear trend in the
tropics shows the maximum decrease in cloudiness
(6%) over the period in question: from 62 to 56%. This
means that the influx of solar radiation in the tropics
increases faster than the average for the planet, and this
increase exceeds 1 W/m2. Since the tropics are domi-
nated by water areas, this fact indicates that the increas-
ing influx of solar radiation primarily leads to an
increase in sea surface temperature (SST). It is not sur-
prising that the cloud cover values and their temporal
trends are close to the global characteristics (Fig. 1).
The ground-based network is most developed in
Europe. For this reason, the overall cloud cover over
Europe was considered as the next example. The linear
trend shows a significantly more modest 2.1%
decrease in cloudiness: from 72.1 to 70%. The nonlin-
ear trend shows a decrease in cloudiness until 1991 and
a subsequent slow increase, which only partially com-
pensates for the previous cloudiness decay. Thus, no
significant variations in the incoming solar radiation
should be expected on the territory of Europe.
Another region under study is the Arctic. Figure 4
shows the smoothed time variation of the annual
cloudiness values in the polar zone located north of
70° N. The linear trend does not show any noticeable
decrease in cloudiness. However, nonlinear analysis
allows us to note significant interannual and intra-
annual variability. Several local minimums and maxi-
mums are detected. Starting from 2005, the most
noticeable decrease in cloudiness is observed, which is
consistent in time with the occurrence of the most sig-
nificant minimum of the ice cover in 2007.
Considering other regions, it should be noted that
the greatest decrease in cloudiness is found in Africa
and South America. Over the Antarctic, the cloudi-
ness increases, which supports the current trend of
increasing ice cover in this part of the world.
WAVELET ANALYSIS OF CLOUDINESS SERIES
Cloudiness series, like other climatic series, are
nonstationary. Therefore, to identify explicit and hid-
Fig. 2. Analysis of the series of the total cloudiness over
land (%) from the ISCCP data: the curve is the nonlinear
trend (the result of smoothing the series by the nonlinear
algorithm); confidence intervals for the 10% level of signif-
icance.
53.5
58.5
58.0
54.0
54.5
55.0
55.5
56.0
56.5
57.5
57.0
1980 19901985 1995 2000 20102005
%
Years
Fig. 3. Analysis of the series of the total cloudiness over
water areas (%) from the ISCCP data: the curve is the non-
linear trend (the result of smoothing the series by the non-
linear algorithm); confidence intervals for the 10% level of
significance.
67
74
73
72
71
70
69
68
1980 19901985 1995 2000 20102005
%
Years
1194
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
POKROVSKY
den periodicities, it is necessary to use an adequate
apparatus of spectral analysis. Currently, the most com-
mon method of wavelet analysis is the one that we
described in the first part of this paper and used in our
previous climatic studies (Pokrovsky, 2010; Pokrovsky,
2009a; 2009b; 2009c; 2009d).
Figure 5 illustrates the wavelet spectrum of the
series of the global total cloudiness (Fig. 1). The
abscissa indicates years, and the ordinate indicates
periodicity. The field of the figure shows the spectral
density values that correspond simultaneously to peri-
odicities and years. Among the extended (in time)
periodicities, an annual variation with a frequency
equal to 1 is distinguished. In addition, an extended
and intensive periodicity refers to a scale of approxi-
mately 32 years, which corresponds to the general pre-
dominant trend of decreasing cloudiness. The next
anomaly of the spectral density refers to a periodicity
of approximately 12 years and is tied to the period of
1990–2005. This periodicity describes a change in the
monotony of a decrease in the series after 2000.
Finally, the 16-year periodicity is found in the range of
2004–2009.
A similar analysis was carried out for the regional
cloudiness series. Figure 6 shows the results of the
wavelet analysis for the course of cloudiness over land
(Fig. 2). In this case, large values of the spectral den-
sity correspond to the annual variation. The “gap” in
large values corresponds to 1998, when the anomalous
El Niño phenomenon occurred. The 16-year compo-
nent is slightly weaker than that of the global distribu-
tion; however, the 32-year component has similar
(with the case of the global distribution) spectral den-
sity values.
The wavelet spectrum for the polar zone stands apart
(Fig. 7). There are no anomalies responsible for the 16-
year component, and the values of the 32-year compo-
nent a re twice as low. Th is is expla ine d by the weakeni ng
of statistically significant cloudiness trends. The annual
component also appears to be less intense when com-
pared to what occurs at other latitudes.
Similar results have been obtained indicating the
fundamental nature of the properties of the cloud
series during the warming period. The wavelet analysis
tool can be very effective in prediction of time series.
Unfortunately, so far, its effectiveness has not been
assessed in predicting climate series, and this is a sub-
ject for further research.
Fig. 4. Analysis of the series of the total cloudiness in the
polar zone (%) from the ISCCP data: the curve is the non-
linear trend (the result of smoothing the series by the non-
linear algorithm); confidence intervals for the 10% level of
significance.
54
72
70
68
66
58
56
64
62
60
1980 19901985 1995 2000 20102005
%
Years
Fig. 5. Wavelet analysis of the series of the total global
cloudiness from the ISCCP data (spectral density values).
32.00
0.25
0.50
1.00
8.00
16.00
2.00
4.00
199 01985 1995 2000 2005
Period, years
Years
‒4
‒3
‒2
‒1
0
1
2
3
4
2
2
2
3
4
4
4
444
4
44
3
3
3
2‒1
‒1
‒1
‒1
‒1
‒1
1
‒2
‒2
‒2 ‒2
‒3
‒2
222
2
‒1
‒2
‒2
‒2
‒2
‒4
‒4
‒4
‒4
‒4 ‒1
‒4
‒4
‒4
‒4
‒2
‒2
‒3
‒3
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
0
0
0
0
Fig. 6. Wavelet analysis of the series of the total cloudiness
for the land from the ISCCP data (spectral density values).
32.00
0.25
0.50
1.00
8.00
16.00
2.00
4.00
199 01985 1995 2000 2005
Period, years
Years
‒4
‒3
‒2
‒1
0
1
2
3
4
2
1‒1
‒1
1
1
0
0
0
2
2
3
3
3
3
32
4
444
‒1
‒2
‒2
‒3
‒2
‒3
‒1
‒1
‒1
‒1
‒1
‒1 ‒1
‒1 ‒2
‒2 ‒2
‒1 ‒1
‒2
‒2
‒2
‒2
‒1
‒3
‒4 ‒4
‒4
‒4 ‒3
‒2
1
44
0
0
0
0
1
1
1
02
2
3
3
23
4
0
0
0
0
0
3
3
3
332
2
1
1
1
0
2
2
2
1
11
1
1
1
1
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
CLOUD CHANGES IN THE PERIOD OF GLOBAL WARMING 1195
The results imply that the overall linear trend for
decreasing cloud cover is confirmed by several inde-
pendent methods of analysis.
DISCUSSION OF THE RESULTS
The presented results should be compared with the
main climate indicators. Let us consider two major
parameters of the climate system: land air temperature
(LAT) and sea surface temperature (SST). In our study,
we used the CRUTEM v3 data on the global LAT and
SST presented by the Hadley Climate Centre under the
British Meteorological Service on their official website
(https://crudata.uea.ac.uk/cru/data/crutem3/).
From general considerations, it should be expected
that a decrease in cloudiness should lead to an increase
in both LAT and SST. It is known that the global LAT
increased at the end of the 20th century and showed a
significant interannual variability in the first decade of
the current century. Figure 8 shows the course of the
global average annual temperature for the same period
of time for which the global cloudiness data are avail-
able. The linear trend indicates an increase in global
LAT by 0.6°C. The nonlinear trend shows that, in the
beginning of the 21st century, the growth of the LAT
ceased. The maximum LAT values were observed in
1998–1999, when the strongest El Niño in recent
decades occurred.
Comparing Figs. 1a and 8, we can see the inverse
relationship: a decrease in cloudiness corresponds to
an increase in temperature, while a halt in decrease in
cloudiness corresponds to a weakened temperature
increase. Naturally, this prompts the desire to calcu-
late the cross-correlation coefficient (CCC) between
the series. One can expect that it has a negative sign,
but a significant absolute value. Indeed, this is the
case. The CCC for the initial values is –0.62. The pro-
cedure of smoothing the series allows filtering out the
“noise” that characterizes random interannual vari-
ability. The CCC for the smoothed series reaches the
value of –0.86.
We have also analyzed the values of the global aver-
age annual SST for 1983–2009. The SST, according to
the graph of the linear trend, increases by 0.4°C, i.e.,
1.5 times less than the LAT. The SST, like the LAT,
increased at the end of the 20th century and ceased
growing at the beginning of this century. Therefore, it
is not surprising that the CCC between the cloudiness
and SST also reaches a high value of –0.84.
These data suggest that the effect of the cloudiness
on climate change has so far been underestimated
(IPCC, 2007). One reason for this is that climate sum-
maries of the ISCCP data have only recently become
available. Another reason is that our knowledge of the
mechanisms of cloud formation and evolution is
clearly insufficient, and the methods and parameter-
ization used in climate models are inaccurate (see
Rossow et al., 2005). The fact that a decrease in cloud-
iness, first and foremost, over the oceans, is observed
simultaneously with the increase in the ocean tem-
perature, which entails increased evaporation from the
water surface, requires a critical analysis of the existing
views regarding the mechanisms of cloud formation at
different latitudes and altitudes.
RELATION BETWEEN THE VARIATIONS
IN THE CLOUDINESS
AND GLOBAL TEMPERATURE
The above results indicate the presence of at least a
qualitative connection between the variations in the
Fig. 7. Wavelet analysis of the series of the total cloudiness
for the polar zone from the ISCCP data (spectral density
values).
32.00
0.25
0.50
1.00
8.00
16.00
2.00
4.00
199 01985 1995 2000 2005
Period, years
Years
‒4
‒3
‒2
‒1
0
1
2
3
4
0
‒2
‒2
‒2
‒2
‒1 2
‒1
‒1
‒1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
‒3
‒3 ‒3
‒3
‒1
‒1
‒1
‒4
‒2
‒2
‒2
‒1
‒2
‒2
‒1
‒3
‒3
‒2
‒1
1
1
3
3
3
3
22
2
2
2
‒2
‒4
‒3
‒4
‒2
‒4
‒1
‒1
‒3
‒3
‒3
‒3
‒3 ‒2
‒1
‒3
‒4
‒4
‒4 1
2
0
1
41
1
1
Fig. 8. Analysis of the series of the global temperature
anomaly CRUTEM3 (°C) f rom t he d ata o f Br itis h Cl imat e
Center: the curve is the nonlinear trend (the result of
smoothing the series by the nonlinear algorithm); confi-
dence intervals for the 10% level of significance.
‒0.4
1.2
1.0
0.8
0
‒0.2
0.6
0.4
0.2
1980 19901985 1995 2000 20102005
%
Years
1196
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
POKROVSKY
cloudiness and global temperature in the considered
period of global warming. Establishing quantitative
relationships through climate models is rather difficult
due to the occurrence of certain errors in the parame-
terization of the cloudiness properties and calculations
of the shortwave and longwave radiation fluxes. How-
ever, it is possible to establish statistical relationships
by means of a correlation analysis of global cloudiness
and land air temperature.
Figure 9 presents the corresponding regression
analysis results. The CRUTEM 3 data (University of
East Anglia, United Kingdom, http://www.uea.ac.uk)
were used as global temperatures. The number of
points for statistical analysis is 318. The regression
equation is Y = –0.0659 X + 19.637. The coeff icient of
determination that characterizes the accuracy of the
regression is 0.277. The latter means that this model
explains approximately 28% of the observed disper-
sion of land air temperature. High values of the global
cloud cover are associated with low global tempera-
tures, which demonstrates the cooling effect of clouds.
The regression model of linear approximation suggests
that a 1% increase in the global cloud cover corre-
sponds to a global decrease in the temperature of
about approximately 0.07°C and vice versa.
In the case of the global cloudiness of the lower
layer, the regression equation changes slightly: Y =
‒0.062X + 16.962. The coefficient of determination
characterizing the accuracy of the regression increases
and becomes in this case 0.316. From the statistical
point of view, this model explains approximately 31%
of the observed dispersion of the land air temperature.
High values of the lower cloud cover are associated
with low global temperatures, demonstrating the cool-
ing effect of the lower clouds. The simple linear
regression model suggests that a 1% increase in low
cloud cover corresponds to a global decrease in tem-
perature of approximately 0.06°C and vice versa.
Thus, the variations in the cloud cover over three
decades in the period of global warming can explain
not only the linear trend of the global temperature, but
also a certain interannual variability. However, the
inclusion of the segment that describes the temporal
evolution of cloudiness into climate models remains a
problem due to the stochastic nature of cloudiness
variability. Climate models are deterministic and can-
not be directly combined with stochastic segments of
cloudiness. However, the effect of cloudiness on the
climate change cannot be ignored due to the signifi-
cant contribution of this climate-forming parameter
and should be studied more carefully to improve cli-
mate predictions.
REFERENCES
Chernokulsky, A.V., Climatology of cloudiness in the arctic
and subarctic latitudes using satellite and ground-based
observations and reanalysis data, Soln.–Zemnaya Fiz.,
2012, vol. 21, pp. 73–78.
Chernokulsky, A.V. and Mokhov, I.I., Climatology of total
cloudiness in the Arctic: An intercomparison of obser-
vations and reanalyses, Adv. Meteorol., 2012, vol. 2012,
id 542093.
Cleveland, W.S., Robust locally weighted regression and
smoothing scatterplots, J. Am. Stat. Assoc., 1979,
vol. 74, pp. 829–836.
Goupillaud, P., Grossman, A., and Morlet, J., Cycle-oc-
tave and related transforms in seismic signal analysis,
Geoexploration, 1984, vol. 23, pp. 85–102.
Hartmann, D.L., Global Physical Climatology, Academic
Press, 1994.
IPCC (Intergovernmental Panel on Climate Change). The
Fourth Assessment Report (AR4), Geneva, Switzerland:
WMO, 2007.
Fig. 9. Results of the regression analysis of the series of global cloudiness (ISCCP) and land air temperature (CRUTEM3).
14.6
15.8
15.0
15.1
15.2
15.3
15.4
15.5
15.6
15.7
14.9
14.8
14.7
63 666564 67 69 7068
Global air temperature, °C
14.6
15.8
15.0
15.1
15.2
15.3
15.4
15.5
15.6
15.7
14.9
14.8
14.7
Global total cloudiness, %
IZVESTIYA, ATMOSPHERIC AND OCEANIC PHYSICS Vol. 55 No. 9 2019
CLOUD CHANGES IN THE PERIOD OF GLOBAL WARMING 1197
Lindzen, R.S. and Choi, Y.-S., On the determination of cli-
mate feedbacks from ERBE data, Geophys. Res. Lett.,
2009, vol. 36, L16705.
https://doi.org/10.1029/2009GL039628
Lindzen, R.S. and Choi, Y.-S., On the observational deter-
mination of climate sensitivity and its implications, Asia-
Pac. J. Atmos. Sci., 2011, vol. 47, no. 4, pp. 377–390.
Lindzen, R.S., Chou, M.-D., and Hou, A.Y., Does the
earth have an adaptive infrared iris?, Bull. Am. Meteorol.
Soc., 2001, vol. 82, pp. 417–432.
Norris, J.R., What can cloud observations tell us about cli-
mate variability?, Space Sci. Rev., 2000, vol. 94,
pp. 375–380.
Pielke Sr., R.A., Davey, C.A., Niyogi, D., et al. Unresolved
issues with the assessment of multi-decadal global land
surface temperature trends, J. Geophys. Res., 2007,
vol. 112, p. 08.
https://doi.org/10.1029/2006JD008229
Pokrovsky, O.M., Kompozitsiya nablyudenii atmosfery i
okeana (Composition of Observations of the Atmo-
sphere and Ocean), St. Petersburg: Gidrometeoizdat,
2004.
Pokrovsky, O.M., A coherency between the North Atlantic-
temperature nonlinear trend, the eastern Arctic ice ex-
tent driftand change in the atmospheric circulation re-
gimes over thenorthern Eurasia, Influence of Climate
Change and Sub-Arctic Conditions on the Changing Arc-
tic, Nihoul, J.C.J. and Kostianoy, A.G., Eds., Springer,
2009a, pp. 25–36.
Pokrovsky, O.M., Relationship between the SST trends in
the North Pacific, the ice extent changes in Pacific sec-
tor of Arctic and corresponding shifts in the marine
ecosystem in subarctic area, Proceedings of the Arctic
Frontiers Science Conf., Trom sø, Nor way: Trom sø Uni-
versity, 2009b, pp. 72–73.
Pokrovsky, O.M., Coherence between the winter pacific
decadal oscillation and the surface air temperature
trends in the continental regions adjoining the North
Pacific, CLIVAR Exchanges, Southampton, UK,
2009c, vol. 49–50, pp. 32–35.
Pokrovsky, O.M., The North Atlantic SST impact on the
ice extent in the Kara and Barents seas, Sea Technology,
2009d, vol. 50, no. 9, pp. 27–32.
Pokrovsky, O.M., Analysis of climate change factors from
remote and contact measurements, Issled. Zemli Kos-
mosa, 2010, no. 5, pp. 11–24.
Rossow, W.B. and Garder, L.C., Cloud detection using sat-
ellite measurements of infrared and visible radiances for
ISCCP, J. Clim., 1993a, vol. 6, pp. 2341–2369.
Rossow, W.B. and Garder, L.C., Validation of ISCCP cloud
detections, J. Clim., 1993b, vol. 6, pp. 2370–2393.
Rossow, W.B. and Schiffer, R.A., ISCCP cloud data prod-
ucts, Bull. Am. Meteorol. Soc., 1991, vol. 72, pp. 2–20.
Rossow, W.B., Delo, C., and Cairns, B., Implications of the
observed mesoscale variations of clouds for Earth’s ra-
diation budget, J. Clim., 2002, vol. 15, pp. 557–585.
Rossow, W.B., Zhang, Y-C., and Wang, J-H., A statistical
model of cloud vertical structure based on reconciling
cloud layer amounts inferred from satellites and radio-
sonde humidity profiles, J. Clim., 2005, vol. 18,
pp. 3587–3605.
Scafetta, N., Empirical analysis of the solar contribution to
global mean air surface temperature change, J. Atmos.
Sol.-Terr. Phys., 200 9, vol. 71, pp. 1916–1923.
https:/ /d oi.org/10.1016/j.jastp.2009.07.007
Schiffer, R.A. and Rossow, W.B., The international satellite
cloud climatology project (ISCCP): The first project of
the world climate research programme, Bull. Am. Mete-
orol. Soc., 1983, vol. 64, pp. 779–784.
Spencer, R.W., Braswell, W.D., Christy, J.R., and Hnilo, J.,
Cloud and radiation budget changes associated with
tropical intraseasonal oscillations, Geophys. Res. Lett.,
2007, vol. 34, L15707.
https://doi.org/10.1029/2007GL029698
Sud, Y.C., Walker, G.K., and Lau, K.-M., Mechanisms
regulating deep moist convection and sea-surface tem-
peratures in the tropics, Geophys. Res. Lett., 1999,
vol. 26, no. 8, pp. 1019–1022.
Tikhonov, A.N., Solution of the inverse problem by the reg-
ularization method, Dokl. Akad. Nauk SSSR, 1963,
vol. 153, no. 1, pp. 49–53.
Wahba, G., A comparison of GCV and GML for choosing the
smoothing parameter in the generalized spline smoothing
problem, Ann. Stat., 1985, vol. 13, pp. 1378–1402.
Translated by M. Chubarova
SPELL: OK
