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