Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://star.arm.ac.uk/~ambn/375.ps Äàòà èçìåíåíèÿ: Thu Dec 13 19:52:10 2001 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 20:17:09 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: molecular cloud

The proposed connection between
clouds and cosmic rays: Cloud
behaviour during the past 50­120
years
E. Pall'e & C.J. Butler
Armagh Observatory, College Hill, BT61 9DG. Armagh, N. Ireland;
epb@star.arm.ac.uk, cjb@star.arm.ac.uk
Abstract
Several authors have suggested that a link exists between the flux of galactic cosmic rays
(GCR) and cloudiness. Here we review the evidence for such a connection from studies
of cloud factors using both satellite and ground­based data. In particular, we search
for evidence for the low cloud decrease predicted by the rising levels of solar activity
and the low cloud­cosmic ray flux correlation indicated by satellite data. Sunshine and
synoptic cloud records both indicate that the global total cloud cover has increased
during the past century. This increase in total cloud cover argues against a dominating
role by solar activity (via GCR) over cloud formation on centennial time scales. Either
the predicted low cloud decrease has not occurred or the medium­high level cloud has
increased to a greater extent than low cloud has decreased.
As there is no accurate long term data available on low cloud behaviour during the last
century, we are not able to totally dismiss the link between GCR and cloudiness, but we
list a number of arguments for and against the proposed cosmic ray­cloud connection.
Keywords: Clouds, Cosmic Rays, Climate, Solar Activity, Sunshine, Global Warm­
ing.
1

1 Introduction
Many studies have shown correlations between solar activity and climate records, how­
ever, the physical mechanisms responsible remain obscure. One of the more straight­
forward mechanisms through which solar activity could influence climate is through
changes in solar irradiance, but, whereas the solar constant is known to be variable, the
amount of change detected in two decades of solar satellite observation is insufficient to
account for more than a third of the global warming since the beginning of the century,
and only a fifth over the most recent decades (Lean and Rind, 1998; Soon et al., 1996).
The contribution of the Sun to global warming has therefore been frequently overlooked
compared to that produced from the increased concentration of greenhouse gases in the
atmosphere.
Several mechanisms have been proposed in the literature through which solar activity
could alter climate including the influence of the extreme ultraviolet on the upper
atmosphere leading to changes in the global circulation system (Haigh, 1994) and the
influence of galactic cosmic rays (GCR) in the troposphere (Dickinson, 1975; Tinsley,
1989). In this review we are concerned with the suggestion that GCR may influence
cloud formation processes.
Galactic cosmic rays are very energetic particles from outside the solar system which
penetrate to the bottom of the troposphere, and are the dominant source of ionization
above 1km and below 60 km altitude. Below 1km, over land, radon is the primary
ionising agent.
Both the flux and energy spectrum of GCR reaching the Earth's surroundings are
modulated by the interplanetary magnetic field, which in turn is strongly influenced by
the magnetic field of the Sun via the solar wind. Because GCR are ionized particles,
they are deflected by the interplanetary magnetic field with the result that only some
of the more energetic particles can penetrate to the inner solar system. When solar
activity is at a maximum, the Sun's magnetic field is strong and fewer particles can
2

reach the Earth; at solar minima the Sun's magnetic field is weaker and the particle
flux grows. Thus modulation of the strength of the interplanetary magnetic field by
solar activity gives rise to the well known inverse correlation between sunspot number
(and other activity indicators) and the flux of GCR.
Dickinson (1975) suggested the possibility of a link between ionization produced by
GCR and cloudiness. He stated that ``solar­related fluctuations in some aspects of
cloudiness could be important. Any such variations in cloudiness are likely to be re­
lated to variations in production of ionization near the tropopause by galactic cosmic
rays,...''. Moreover he added that ``A 5­10% change in thick cloud cover or a 0.5 km
change in cloud heights with solar activity would probably have gone undiscovered up to
now''. Tinsley (1989) and Tinsley and Deen (1991) proposed a mechanism by which
stratospheric ionization changes produced by cosmic rays would affect microphysical
processes such as nucleation and growth of ice particles in high­level clouds, however
this mechanism remains unproven. Following Dickinson and Tinsley's proposals, Pu­
dovkin and Veretenenko (1995) found cloudiness decreases associated with Forbush
decreases of galactic cosmic ray flux (sudden drops due to geomagnetic storms). The
results indicated a reduction in the total cloud amount over the former USSR as ob­
served from the ground in the 1­2 days period following strong Forbush decrease events.
The results were found to be confined to a latitude belt of about 60­64 degrees North.
More recently, Svensmark and Friis­Christensen (1997) suggested a possible relation­
ship between the cloud cover over the oceans and the galactic cosmic ray flux. For their
study, Svensmark and Friis­Christensen (1997) used the satellite cloud cover data from
the ISCCP C2 dataset for the period 1983 to 1991. They found a correlation between
the GCR and cloudiness over oceans excluding the tropics.
Clouds have two different effects on climate. They have a cooling effect due to an
increase in the albedo and the reflection of short­wave solar radiation back to space
and they trap the long­wave radiation emitted from the ground, thus also warming the
climate. In general the clouds are believed to have a net cooling effect (Ramanathan
et al., 1989), but the amount and sign of the cloud feedback will strongly depend on
3

the cloud type and altitude.
Thus there are two possible mechanisms involving cloudiness change that could lead to
a warming of the troposphere, (1) a reduction of low cooling clouds or (2) an increase
in high warming clouds. If either or both of these possibilities resulted from the known
increase in solar activity then a smaller residual warming would remain to be explained
by the greenhouse effect.
Our aim in this paper is twofold: firstly, we review the credibility of the cosmic ray­
cloud relationship and its predictions for climate, and secondly, we try to establish the
cloud trends over the last 50 to 120 years from synoptic cloud data and cloud proxies
to test the galactic cosmic ray­cloud connection (hereafter as GCR­CC) hypothesis.
Information on past cloud changes is also valuable to test global circulation models,
since the implications for the global climate can be of dramatic importance.
In the next section we examine the evidence for a GCR­CC as derived from satellite
cloud data. In Section 3 we discuss the available ground­based synoptic observations
and proxies for clouds. In section 4 we discuss the conclusions reached from examination
of these meteorological records, and in section 5 we suggest some of the possible drivers
of change. In section 6 the arguments for and against a GCR­CC, as well as the
implication of such a connection for the climate change will be discussed.
2 The GCR­CC from satellite cloud data
The study of decadal to centennial climate variability requires long and reliable datasets.
Unfortunately global cloud datasets are scarce and of relatively short duration. Satellite
global coverage has only been available for the last 20­30 years, and has discontinuities
due to the difficulty of combining data from satellites which have used different instru­
mentation.
In their initial study, Svensmark and Friis­Christensen (1997) used data from the
4

ISCCP C2 (International Satellite Cloud Climatology Project) dataset from 1983­1991.
They used satellite data only from geostationary satellites, discarding polar­orbiting
satellite data in the belief that these were degraded by calibration problems. Cloud
coverage over land areas was also discarded as the behaviour differed significantly from
the clouds over oceans. They found a good correlation between the galactic cosmic
ray flux and the total cloud cover over oceans which strengthened further when the
tropical belt (\Gamma22:5 o ! latitude ! +22:5 o ) was excluded. Because of the more effective
shielding of galactic cosmic rays by the Earth's magnetic field at low latitudes, the
improved correlation after removing the tropical areas was interpreted as additional
evidence for a GCR­CC.
Svensmark and Friis­Christensen (1997) and Svensmark (1998) also used data from the
Nimbus­7 matrix and the Defense Meteorological Satellite Program (DMSP). Plotted to
scale with the ISCCP C2 data, these two datasets apparently confirmed the hypothesis
of a cosmic ray­cloud connection. However the use of different inter­calibrated satel­
lite datasets was strongly criticized by Kernthaler et al. (1999) and Kristj'ansson and
Kristiansen (2000), who showed that the ISCCP and DMSP datasets disagreed when
overlapped.
On the other hand, the hypothesis of Svensmark and Friis­Christensen (1997) received
some support from a paper by Menzel et al. (1996) which found a similarity in the
variation of cirrus clouds and the GCR flux using data from the polar orbiting satellite
HIRS (High­Resolution Infrared Radiometer Sounder). However later studies using
this dataset suggest that the correlations of HIRS cloud products with cosmic rays and
other solar activity parameters are not well understood (Hinton, 1999). Using 12­month
running means (as the majority of authors in this subject have done) Hinton (1999),
found a larger number of significant negative correlations between cloud and magnetic
indices and sunspot number than positive correlations with cosmic rays, whereas the
opposite is true for the ISCCP series. He also pointed out the difficulty of assessing the
significance of the correlations for such short or filtered data series and the crucial role
of the determination of the number of degrees of freedom.
5

Serious doubts about the validity of the results by Svensmark and Friis­Christensen
(1997) and apparent contradictions were also mentioned by Kernthaler et al. (1999)
who tried to relate the different cloud types in the ISCCP C2 dataset with the galactic
cosmic ray flux, particularly during the four years 1985­1988 during which the ISCCP
data was believed to be well calibrated. They did not find any correspondence with a
particular cloud type. They argued that, given that ionization produced by comic rays
peaks at an altitude around 15 to 20 Km, high cirrus clouds should be the most affected.
They included in their correlations the data from polar satellites which they believed
were well calibrated with respect to the remaining satellites (Rossow et al., 1996). In
the case of a GCR­CC, the inclusion of this data would be expected to improve the
correlations; in fact the correlations worsened. Since the ionization produced by GCR is
larger at high latitudes, this was taken as an argument against a possible cloud­cosmic
ray relationship. The conclusions of Kernthaler et al. (1999) were in turn questioned
by Svensmark and Friis­Christensen (1999) who pointed out that the former results are
not reproduced with the later D2 data.
Kuang et al. (1998) also analysed the ISCCP C2 dataset, this time comparing the
cloud amount with time against cloud optical thickness. They showed that these two
parameters had a negative correlation, and argued that the combined effect of the two
opposing variations may be a null effect on the cloud reflectivity.
Kristj'ansson and Kristiansen (2000) compared the ISCCP C2 (1983­91) and D2 (1989­
93) datasets, together with synoptic cloud cover over the oceans from Norris (1999,
discussed in the next section) and the DMSP data, with the variation of the galactic
cosmic ray flux. Their main finding was that the correlation coefficient between total
cloud cover and cosmic ray flux diminished when data from the new ISCCP D dataset
was included. They also rejected the suggestion that the correlation was maintained
over any particular cloud type except possibly for the low marine clouds. The synoptic
data indicated only a slight negative correlation with the cosmic ray flux. The DMSP
data had a similar behaviour to the cosmic ray flux but, as the DMSP satellite detects
only liquid cloud types over oceans, Kristj'ansson and Kristiansen (2000) demonstrated
6

that its data is not directly comparable with the ISCCP data for the time during which
they overlap.
Contemporaneous with the Kristj'ansson and Kristiansen (2000) paper, two further
studies which looked for evidence of a GCR­CC were published by Pall'e and Butler
(2000) and Marsh and Svensmark (2000). They used the new re­calibrated ISCCP D2
dataset, which was believed to be superior and more reliable than the old C series and,
more importantly, cover an additional three years. They also have a better agreement
with ground­based cloud observations. From the cloud top temperature given by this
dataset, Marsh and Svensmark (2000) found a correspondence between low cloud types
and cosmic ray flux. No correlation was found for middle or high level clouds. The fact
that the surface temperature and cloud top temperature also correlated with the GCR
was taken as additional evidence for low cloud properties being affected by GCR. A
correlation between low cloud factor and GCR was found independently by Pall'e and
Butler (2000). These authors also showed that the total cloud cover did not correlate
with the cosmic ray flux after 1991/2. The analysis by Pall'e and Butler (2000) covered
different areas of the globe and found no strong latitudinal variation in the GCR­
CC, except for the polar regions where the correlation dropped. Furthermore they
argued, from the behaviour of clouds of liquid or ice droplets, that almost all variability
in ISCCP D2 low cloud cover at low and middle altitudes can be accounted for by
liquid phase clouds, and it is these alone which were responsible for the correlation.
These results came from averaging over broad latitude bands, however, more recent
results on the detailed geographical distribution of the correlations (Pall'e and Butler,
unpublished), using the ISCCP D1 data series, indicate that there exists a longitudinal
as well as latitudinal dependence.
In brief the satellite data indicate the following:
fflA strong correlation between low clouds as defined by the ISCCP D2 (1983­1994)
dataset and cosmic ray fluxes (99.8% significance level when using global averaged
yearly means). A correlation between total cloud cover and cosmic rays occurs only
7

during the restricted interval 1983­1991 (Pall'e and Butler, 2000).
fflThere is a correspondence between water clouds measured by the DMSP and galactic
cosmic ray flux. No other cloud type data is available from this source. (Svensmark
and Friis­Christensen, 1997).
fflThere is a correspondence between cirrus clouds from the HIRS data (June 1989­May
1996) and cosmic ray flux (Menzel et al., 1996), however the significance level of this
correlation has not been well established.
fflData from the Nimbus­7 cloud matrix for total cloud cover has been studied by
Svensmark and Friis­Christensen (1997) with positive results for a cosmic ray­cloud
relationship.
2.1 The GCR­Cloud prediction
If we suppose that the GCR­CC holds for low clouds over long periods of time then,
due to an increase in solar activity and a more effective shielding of galactic cosmic rays
in the heliosphere over the past 100 years (see Lockwood and Stamper, 1999), Pall'e
and Butler (2000) predicted a decrease in the low cloud cover during the past century.
As there was no evidence to support a correlation with solar activity for other cloud
types, it was assumed that they remained constant. Thus, if only low cloud varied with
time due to a GCR­CC and all else were constant, even though total cloud cover does
not correlate with GCR, a decrease in the total cloud since the late 19th century would
be expected. The actual trend in low cloud factor for the last 100 years predicted by
the correlation between GCR and low cloud is small (¸1%), however, they estimated
that it could have caused a positive climate forcing resulting in a rise in the global
mean temperature of 0.27 o C temperature increase. The cloud variability observed by
satellite over one solar cycle is about 3­4%.
The above analysis (Pall'e and Butler, 2000), which was based on the observed cor­
relation of low cloud factor with cosmic rays and solar activity indices has serious
8

implications for our understanding of the causes of climate change as it could imply
that most of the global warming during this period can be attributed to the combined
direct (irradiance) and indirect (low cloud) effects of solar activity. However the pre­
dicted long­term trend in low cloud factor is small (¸1%) and could easily be swamped
by other factors affecting cloud cover not related to solar activity (e.g. a cloud redis­
tribution towards more convective cloud types, or increase in cloud cover due to a rise
in temperature, etc..). Reported trends in synoptic cloud observations and sunshine
records are typically of the order of a few percentage points (1­3%) change per decade,
in this sense the cloud trends proposed by the GCR­CC are not unreasonable. A com­
prehensive forcing computation would require knowledge of the variability of clouds at
all levels, and its geographical distribution. It is important, therefore, to examine any
measured cloud factors over the past century.
3 Long­term cloud data
Since satellite­based cloud records do not extend for more than a couple of decades and
calibration problems between existing datasets prevent a straight forward comparison,
we have examined synoptic cloud observations and cloud proxy records from ground
stations to study long­term cloud cover behaviour. The geographical and temporal
coverage and brief details of any trends identified are given in Table 1.
So far, only negative results towards a GCR­CC have been reported when comparing
ground based cloud observations and solar activity (Kristj'ansson and Kristiansen, 2000;
Wagner et al, 2001; Pall'e and Butler, 2001). Only Udelhoffen and Cess (2001) found
a correspondence between solar activity and total cloud cover over USA, but cloud
changes were in phase with the sunspot number and not the GCR.
There is a fundamental difference between cloud cover as seen from the ground and
observed by satellite. Ground observations are naturally weighted towards low clouds
which are more easily observed from below, whereas satellite observations from above
9

are weighted towards high cloud. Satellites cannot measure different cloud layers at
the same time, so low clouds can be obscured by higher cloud in a similar way to
the obscuration of high cloud when observed from the ground. For total cloud, the
ISCCP cloud data agrees quite well with ground observations (Seze et al., 1986; Rossow,
1996, Pall'e and Butler 2001), however, there are important differences when considering
specific cloud types (Weare, 2000). For instance, the cloud classification scheme for the
ISCCP data is not identical to that used by synoptic observers even though the same
labels are used (Rossow, 1996). Thus discrepancies between satellite and ground based
cloud observations are likely. Sunshine records are also of little help in this respect
as they cannot distinguish between cloud types and total cloud is not correlated with
GCR.
3.1 Synoptic observations of cloud cover
Synoptic observations of total cloud cover are made in many meteorological stations
worldwide. The procedure is simple; at specified times of the day, the observer estimates
the number of oktas (1/8 th of the sky) covered by clouds. A totally overcast sky is
registered as 8 and a clear sky as 0. At some stations, coverage by different cloud types
is also recorded. It is evident that, to a greater or lesser extent, such estimates are
observer dependent and therefore subject to systematic errors.
Observations of synoptic cloud cover over the oceans from volunteer ships since 1952
have been compiled and analysed by Norris (1999). Norris found that the global mean
cloud cover over the oceans increased by 1.9% (sky cover) between 1952 and 1995.
Global mean low cloud cover was also observed to increase by 3.6% during the same
period and trends in zonal mean cloud cover in 10 o ­lat bands between 60 o N and 40 o S
were all found to be positive. Several possible artifacts were examined but it was con­
sidered unlikely that they could explain the observed inter­decadal variability. Norris
concluded, however, that, in view of possible observer dependent differences, the trends
cannot be accepted as definitive until they have been corroborated by related meteoro­
10

logical parameters and satellite­based measurements.
Sun and Groisman (2000), have studied synoptic cloud cover variations over the former
USSR from 1936 to 1990. They find high cloud to be increasing and low cloud decreasing
with the total cloud increasing over this period. Though for a continental, as opposed
to a maritime climate region, these findings are pretty much in line with the trends
suggested earlier by the the cosmic ray ­ low cloud correlation. On the other hand,
though the total cloud trends are similar, it is difficult to reconcile the results by Norris
(1999) and by Sun and Groisman (2000) on low clouds, other than to point out that
they refer to very different regions. It should be noted that the Sun and Groisman
(2000) trends are very statistically significant whereas those by Norris (1999) are prone
to many systematic effects such as might arise from a change in the latitude of preferred
shipping lanes.
Similar reports to those of Norris (1999) and Sun and Groisman (2000), using cloud
synoptic observations are listed in Table 1. McGuffie and Henderson­Sellers (1988)
reported a synoptic cloud increase over Canada and the Canadian Arctic. Henderson­
Sellers (1992) reported similar results over N. America, Australia, Europe and India.
Only China (Kaiser, 1998) seems to differ from this pattern.
Whilst the behaviour of individual stations may differ strongly from the mean trends
over a large area, it is notable that almost all studies of synoptic cloud cover show
an increasing trend during the past century. No other general studies (involving more
than one station) regarding synoptic clouds or cloud type variability are known to the
authors.
3.2 Sunshine data ­ a proxy for cloud cover
One relevant proxy for cloud cover is sunshine duration. Daily records of the duration
of bright sunshine have been obtained at four stations in Ireland since May 1880 using
a standard Campbell­Stokes sunshine recorder (Observer's Handbook, 1982). Two are
11

located in the east of Ireland; at Armagh Observatory and The Ordnance Survey Office,
Phoenix Park, Dublin; one in the extreme west at Valentia Island/Cahirciveen, Co
Kerry; and the fourth, in the midlands, at Birr Castle, Co Offaly. More information
about the Irish sunshine data can be obtained at Pall'e and Butler (2001).
The most prominent feature of the data, for all four sites, is a gradual decline in the
total annual sunshine hours over much, if not all, of the 118 year period during which
records have been obtained (see Figure 1). The effect is particularly conspicuous at the
most westerly site of Valentia Island/Cahirciveen, on the County Kerry coast, where
the number of sunshine hours has dropped by ¸ 20% since the end of the last century.
If we plot the seasonal averages, the gradual decrease is seen in all stations in most
seasons.
The sunshine data has been shown to be a very good proxy for total cloud cover over
monthly to yearly time scales, at least over the Irish region (Pall'e and Butler, 2001).
Unfortunately the sunshine records are related only to the total cloud factor and do not
give us any information on cloud type. It can also be argued that the sunshine records
are a local measurement and that the observed trends are not necessarily similar for
other areas. However, Pall'e and Butler (2001) compared the variability of the ISCCP
D2 satellite cloud records over Ireland and other areas of the globe and found that,
over the period comprised by the data (July 1983­August 1994), cloudiness over Ireland
behaves similarly, not only to the North Atlantic region as a whole, but to mid­latitude
oceanic regions generally. Thus it seems that the sunshine factor over Ireland could be
particularly useful in indicating wider trends.
A detailed comparison between the Irish sunshine records and the synoptic cloud cover
over oceans near Ireland remains to be done. However, preliminary examination sug­
gests an approximate agreement in percentage variability of sunshine and synoptic cloud
cover over the last 50 years, of around 14% (Pall'e and Butler, 2001).
Though they cover a shorter time than the Irish records, similar results were reported
by Stanhill (1998b) for sunshine records from Israel. Also, Angell (1990) and Angell
12

et al. (1984) reported sunshine decreases and synoptic cloud increases during the period
1950­1988 for the United States. Changnon (1981) also reported sunshine decreases for
mid­western United Sates during the more extended period 1901­1977. Both, Angell
et al. (1984) and Changnon (1981), found the increasing trend in synoptic cloud to
be larger than the decreasing trend in sunshine records, and this was attributed to
an increase in the amount of thin cirrus which, though included in the synoptic cloud
record, would be insufficiently opaque to reduce the instrumental sunshine duration.
3.3 Radiation Decreases
In Table 1 we refer also to the reduction in ground solar radiation levels reported at a
world­wide distribution of sites during the last 40 years (Stanhill, 1998a and references
therein), suggesting a total cloud cover increase. However, we must be very careful when
using such measurements to establish cloud trends, since radiation decreases have not
been always accompanied by an increase in cloud cover and it is possible that changes in
cloud type or atmospheric or cloud transparency could be responsible (Liepert, 1997).
4 Discussion ­ the long­term cloud behaviour.
If Irish and similar sunshine data are indeed relevant to the global trend they would
indicate a rise in the total cloud factor since the late 19th century, in agreement with
the synoptic cloud observations. The evidence for low cloud trends is less clear though
a decrease in line with predictions from solar activity levels cannot be ruled out.
To estimate the possible contribution to global warming implied by the cosmic ray­
cloud interaction, Svensmark (1998) predicted a total cloud cover decrease during the
last century, which is not seen in the ground­based records. However it is now known
from the new ISCCP D cloud products that only the low clouds vary in phase with the
GCR. Pall'e and Butler (2000) predicted a low cloud cover decrease since 1900 from the
observed correlation of low clouds and cosmic rays in the interval 1983­1994 and known
13

changes in solar parameters. If other cloud types remained constant this would imply
a total cloud cover decrease over the last century. It is clear from Table 1 that, since
total cloud has increased over the past century, either the decrease in low cloud cover
has not occurred, or the supposition that the rest of the cloud types remain constant
is not true.
A summary of the principal characteristics of two sets of satellite cloud data and two
types of ground­based data (synoptic and sunshine) is given in Table 2. It is clear
in the table how each series has its own characteristics. Parameters such as the time
of measurement, the geographical coverage or even the problems the data present are
not equivalent. This makes comparisons between the datasets difficult. Data obtained
by one satellite, for example, does not always correspond to that from other satellites
or from the ground, because they are not always looking at the same phenomena. In
respect to the GCR­CC, we seem to be confronted by a contradiction. On one hand,
satellite measurements indicate a correlation between low cloudiness and the GCR. This
is based on global measurements for which the data is highly reliable. Unfortunately,
these datasets are available for only a few years, and the significance of the correlations
problematical and it is difficult to distinguish statistically between GRC and other
solar­related parameters. Whereas ground­based observations have the requisite long
duration, they do not correspond with the GCR and the trends argue against the GCR­
CC. In general ground­based data are less reliable than satellite observations since many
factors (observer changes, station relocation,..) can introduce spurious trends in the
series. In fact the only two `clear' trends are: increasing liquid phase clouds over oceans
from the DMSP (Defence Meteorological Satellite Project) satellites, and increasing
cirrus clouds detected by the HIRS (High­resolution Infrared Radiation Sounder), and
either or both of those trends may be a result of the short duration of the records.
The conclusions from Table 1 and 2 do not support the cosmic ray­cloud connection
and its effects on the climate system. However, neither is there sufficient evidence of
global low cloud trends over the past 100 years to refute it. All we can say is that, for
the GCR­CC to be true, the amount of mid­level and high cloud would have to have
14

increased in this period to a greater extent than low cloud decreased. At the present
moment, there is no firm evidence for this, except possibly in the FUSSR (Sun and
Groisman, 1999) and China (Kaiser, 1998).
The effect that an increasing trend in total cloud cover would have on the Earth's
radiation balance is not clear. Overall clouds are believed to cool the climate and thus,
to a first approximation, increasing cloudiness would result in a negative forcing to
global temperatures. In other words, clouds could reduce global warming, acting as a
thermostat. However, the net effect would depend on the altitude, the geographical
distribution and physical properties of the cloud changes. Though, on average, thin
high clouds warm and low clouds (or more specifically optically thick clouds) cool the
climate, their combined effects are not necessarily additive; e.g. high thin clouds over
low thick clouds will not warm. Even the temporal change in cloud occurrence can
be important, as at night time all clouds have a warming effect. During the past
century, much of the global warming has been attributed to milder temperatures at
night (Houghton et al., 1995), as would be expected if night­time cloud cover alone had
increased.
It is feasible, nonetheless, that if high clouds were to have increased due to some un­
known cause over the past century to a greater extent than low cloud has decreased
due to the operation of a GCR­CC, then the radiative forcing of the two changes would
be compounded with the additional positive radiative forcing from solar irradiance
changes, thereby all contributing to enhanced global warming. However, if the increase
in total cloud reflected an increase in deep convective clouds, these would have an oppo­
site forcing to any downward trend in low clouds and upward trend in solar irradiance.
Until such time as reliable datasets can be assembled with information on the trends of
different cloud types, these possibilities remain pure speculation and an encouragement
to further research.
15

5 What is causing the total cloud increase?
We have seen in the previous section how both synoptic cloud and sunshine data indicate
that total cloud cover has been increasing over the last century. But what are the reasons
for the change? Certainly the proposed connection between energetic particles entering
the atmosphere and cloud formation would explain a decrease, but not an increase. Can
the long term trend in total cloud result from a mean air and sea surface temperature
increase? Little is known about the effects that a change in temperature will have on
clouds. Global circulation models predict a cloud factor decrease when climate warms
(Cess et al., 1996) however, even though the models predict an uneven reduction in
cloudiness and clouds are still poorly parameterised, such a reduction has not been
reported so far. On the other hand, following the rise in sea­surface temperatures
which have accompanied global warming (Reid, 1987), an increase in cloud formation
might follow from an increase in evaporation rates. A progressive moistening of the
atmosphere has been reported by Wentz and Schabel (2000), however, warmer air is
able to hold more water vapour before saturation and thus relative humidity could
have remained constant. As cloud formation depends critically on relative humidity, an
increase in the water vapour content of the atmosphere (absolute humidity) would not
necessarily lead to an increase in cloud factor.
Cloud amounts over Canada have been reported to increase over practically all the re­
gions where temperatures have risen (McGuffie and Henderson­Sellers, 1988). However
station records from the Canadian Arctic showed cloud changes which seem decoupled
from temperature changes. For the United States, Angell et al. (1984) found an appre­
ciable tendency for cloudiness to be above average and sunshine to be below average
in years of warm sea surface temperature in the equatorial eastern pacific (El Ni~no).
Henderson­Sellers (1992) also noticed a tendency for total cloud amount to be greater
in warmer conditions.
An increase in tropospheric aerosols could also give rise to increased cloud formation.
However (Norris, 1999) stated that the 10 o ­lat band trends in synoptic cloud observa­
16

tions between 60 o N and 40 o S, are generally larger for the Southern Hemisphere and
Tropics than for the mid­latitude Northern Hemisphere. For Norris (1999), this argued
against attribution of increased cloud cover to increased anthropogenic aerosols, and
suggested that it is possible that global cloud cover is responding to some other global
parameter, perhaps temperature.
Pall'e and Butler (2001) found a correlation between the sunshine records over Ireland
and the solar cycle length. The solar cycle length is defined as the time from minima to
minima (or maxima to maxima) of the sunspot number, it is related to solar activity in
the sense that generally shorter solar cycles are accompanied by higher levels of activity
(higher sunspot maxima). Since the sunspot cycle length was shown to vary in phase
with the NH air temperature (Friis­Christensen and Lassen, 1991), and that relationship
was confirmed for an Irish (Armagh) temperature dataset by Butler and Johnson (1996),
they concluded that in the vicinity of Ireland at least, it seems likely that decreased
sunshine hours (increased cloudiness) result from the increased temperatures associated
with global warming, rather than to any kind of direct solar influence on clouds.
Another effect to take into consideration is the effect of increased aircraft traffic.
Changnon (1981) indicated that localized shifts to more cloudiness in zones of increased
air traffic since 1960, suggest anomalous changes related to jet­induced cirrus. How­
ever, Menzel (1996) found that the increasing trend in cirrus cloud cover detected by
the HIRS was no more noticeable over areas with heavy air traffic than over areas with
low air traffic.
In many of the sites where a decrease in total solar irradiance has been found, an
increasing trend of cirrus cloud formation has also been detected (Liepert, 1997). A
shift from stratiform to higher frequencies of convective clouds has also been observed
(Liepert, 1997; Stanhill, 1998a). Finally, another possible explanation for the long­term
cloud change could be through shifts in weather patterns such as the location of storm
tracks.
Even though none of the sunshine or synoptic cloud records are exempt from the
17

possibility of systematic errors, the agreement between the different and diverse data
series in their apparent trends reinforces our opinion that the trend towards increasing
cloudiness is real and, probably, global.
6 Does a cosmic ray ­ cloud cover connection exist?
6.1 Arguments in favour
In light of the arguments explored in previous sections it seems clear that a direct
relationship between total cloud cover and cosmic ray flux is unlikely. The most reliable
satellite dataset, the ISCCP D2 (Pall'e and Butler, 2000), the synoptic cloud cover
(Norris, 1999; Kristj'ansson and Kristiansen, 2000) and the sunshine records (Pall'e and
Butler, 2001; Wagner, 2001) support that conclusion. For different cloud types the
situation is different. Clearly the favourite candidate for a GCR­CC is the low cloud
cover. Even though discarded at first by Kernthaler et al. (1999) in the ISCCP C2
dataset, the correlation is clear in the new ISCCP D2 dataset (Pall'e and Butler, 2000;
Marsh and Svensmark, 2000).
The correspondence between the DMSP cloud records and the galactic cosmic ray
flux roughly agrees with the findings of the ISCCP D2 data. The DMSP satellite
measures only liquid clouds over the oceans, which consist mainly of low­level clouds.
However during the period of overlap the two series do not totally correspond in all
respects (Kristj'ansson and Kristiansen, 2000). The correspondence between cloud top
temperature and surface temperature with the GCR (Marsh and Svensmark, 2000) also
supports the GCR­CC.
The Earthshine project which monitors the Earth's albedo and is currently underway
at Big Bear Solar Observatory may impact on our discussion. A global and absolutely
calibrated albedo for the Earth can be determined by measuring the amount of sunlight
reflected from the Earth to the Moon and back. Preliminary results seem to support
18

a lower planetary albedo at solar activity maximum (Goode et al., 2001) which would
support the GCR­CC, however it is too early and the time span too short to reach any
definitive conclusion.
6.2 Arguments against
The main argument against the GCR­CC, is the lack of an established physical mecha­
nism. Turco et al. (1998) and Yu an Turco (2001) have proposed a mechanism ``ion­ion
recombination'', by which ionization produced by GCR could facilitate aerosol produc­
tion and growth in the lower troposphere. Svensmark and Friis­Christensen (1997) and
Marsh and Svensmark (2000) have speculated that it is not unreasonable to imagine
that systematic variations in cosmic ray ionization could influence cloud properties via
this mechanism.
An alternative mechanism to explain a cosmic ray influence in ice clouds has been
proposed by Tinsley et al. (2000) and references therein. However his mechanism refers
only to ice clouds (mainly cirrus), which are not seen to vary with the cosmic ray flux,
except for the HIRS data which covers only a very short time. It is difficult also to
explain a cloud increase during the past century using this mechanism.
Concern arises also from the temperature records. If an indirect solar forcing com­
ing from the GCR­CC acts on the climate system, a stronger variability of the mean
temperature over the ¸11 year cycle would be expected. Though ocean thermal inertia
damping is likely to be important here, a small amplitude signal would still be expected.
In addition, there is the unexplained latitudinal distribution of the low cloud­cosmic
ray correlation. On account of the latitude­differential shielding of the Earth's magnetic
field, the correlation would be expected to increase poleward, whereas in fact it remains
almost constant, except for the poles where it has a small drop (Pall'e and Butler, 2000).
Pall'e and Butler (2000) suggested that the observed latitude dependence could result
if only liquid phase clouds were amenable to the influence of cosmic rays. Ice­phase
19

clouds, more abundant at the poles or high altitudes, would be unaffected by cosmic
rays, thereby leading to a lack of correlation with clouds at both high altitudes and high
latitudes. However in a recent study, a longitudinal as well as latitudinal dependence
of the correlation has been found by Pall'e and Butler (unpublished) which no longer
supports the suggestion that only liquid clouds phase clouds are affected by GCR.
Finally, we note possible changes in cloud reflectivity and opacity. Kuang et al. (1998)
analysed the cloud optical thickness and reflectivity changes for the period 1983­1991
and found that, even though the cloud amount changed with the cosmic ray flux, the
cloud optical properties changed too, compensating for the changes in cloud amount.
As a consequence, even though the cloud factor may change, the effect on the radiation
budget could be negligible.
A further complication in any cloud influence on climate was recently raised by Del
Genio and Wolf (2000). Based on three years of low cloud data, Del Genio and Wolf
(2000) found that low clouds thin when temperatures rise. As a result, if the temper­
ature increases over the next decades, low clouds could become thinner, reduce their
albedo and thus contribute a positive feedback to global warming. If this were true,
part of the global warming experienced during the last century might have come from
a thinning of low cloud, further reducing the contribution required from the greenhouse
effect.
7 Conclusions
There appears to be an statistically significant (99.8 % significance level) correspondence
between the low cloud cover as seen from modern satellite data and the galactic cosmic
ray flux. However the duration of the dataset is short and leaves margin for many
uncertainties when trying to establish long­term behaviour of the cloud cover.
In order to assess the validity of prediction over long time scales, some synoptic cloud
observations and cloud proxies have been compared. The sunshine records and the
20

synoptic cloud cover over many areas of the Earth agree that the total cloud cover has
been increasing over the last century. This trend goes in the opposite direction to the
trend predicted for low clouds by the cosmic ray­low cloud correlation. Though the
accuracy of some of the datasets is open to question, the agreement in the observed
trends suggests that they are real.
The GCR­CC cannot be positively discounted at this time, at least for the last 20 years,
for which satellite data is available. The reason is that no accurate data for low cloud
cover is available to ascertain whether or not the correlation is maintained further back
in time (see Table 2). However, two facts: (1) the lack of correlation between total cloud
cover and GCR and (2) the general increase in total cloud cover over the last century
showed by several studies of cloud and cloud­proxy data, indicate that it is likely that
some mechanism(s) other than GCR are the main drivers of cloud cover changes on
century time scales. Some possible mechanisms for the cloud increase are discussed;
the most promising of which seems to be a rise in evaporation rates following the rise
in sea­surface temperatures. However, this will not necessarily cause an increase in
cloudiness if air temperatures warm sufficiently to prevent a rise in relative humidity.
It seems likely therefore, that the simplified scenario of cloud amount governed by
solar activity (GCR) alone, is not valid as trends in cloudiness go the opposite way to
predicted. If real, the radiative effects of clouds affected by the GCR­CC, will have
to be combined with the radiative effects of clouds influenced by other mechanism(s).
A net increase in total cloud cover (not taking into account a shift in cloud types)
during the last 100 years would lead, to a first approximation, to a negative radiative
forcing. With a net radiative forcing by clouds between 3 and 5 times larger than that
from a doubling of CO 2 concentration in the atmosphere (Henderson­Sellers, 1992),
the importance of cloud factor changes will depend on the variability of the different
cloud types and their geographical distribution. It is also clear that cloud factors have
been changing during the recent past, and until the cloud behaviour on temporal and
geographical scales is understood (via long and reliable datasets) and assimilated into
global circulation models, predictions will be seriously handicapped.
21

Acknowledgements
We would like to thank Dr. J. Haigh and an anonymous referee for their comments on
an earlier version of this manuscript. The sunshine data for Dublin, Birr and Valentia
Island/Cahirciveen were kindly provided by D. Fitzgerald of Met Eireann, Dublin.
Research at Armagh Observatory is grant­aided by the Department of Culture, Arts
and Leisure for Northern Ireland.
References
Angell, J.K., 1990. Variation in United States cloudiness and sunshine duration be­
tween 1950 and the drought year of 1988. J. Clim. 3, 256­308.
Angell, J.K., Korshover, J., Cotton, G.F., 1984. Variation in United States cloudiness
and sunshine, 1950­82. J. Clim. & App. Met. 23, 752­761.
Butler, C.J., Johnson, D.J., 1996. A provisional long mean air temperature series for
Armagh Observatory. JASTP 58, 1657­1672.
Cess, R.D., et al.(34), 1996. Cloud feedback in atmospheric circulation models: an
update. J. Geophys. Res. 101, 12791­12794.
Changnon, S.A., 1981. Midwestern cloud, sunshine and temperature trends since 1901:
possible evidences of jet contrail effects. J. App. Met. 20, 496­508.
Del Genio, A.D., Wolf, A.B., 2000. The temperature dependence of the liquid water
path of low clouds in the southern Great Plains. J. Clim. 13, No. 9, 3465­3486.
Dickinson, R.E., 1975. Solar variability and the lower atmosphere. Bulletin American
Meteorological Society 56, 1240­1248.
Friis­Christensen, E., Lassen, K., 1991. Length of the solar cycle, an indicator of solar
activity closely associated with climate. Science 254, 698­700.
Goode, P.R., Qiu, J., Yurchyshyn, V., Hickey, J., Chu, M­C., Kolbe, E., Brown, C.T.,
22

Koonin, S.E., 2001. Earthshine observations of the Earth's reflectance. Geophys. Res.
Letts. 28, 1671­1674.
Haigh, J.D, 1994. The role of stratospheric ozone in modulating the solar radiative
forcing. Nature 370, 544­546.
Henderson­Sellers, A., 1992. Continental cloudiness changes this century. GeoJournal
27:3, 255­262.
Hinton, B.B., 1999. Solar Cycle Cloud Variations. Space Science and Engineering
Center Internal Report. University of Wisconsin­Madison, 9 Mach 1999.
Houghton, J.T., Meira Filho, L.G., Callander, B.A., Harris, N., Kattenberg, A.,
Maskell, K., (Eds). Climate Change 1995. The science of climate change, 1995. Cam­
bridge University Press, p 437.
Kaiser, D.P, 1998. Analysis of total cloud amount over China, 1951­1994. Geophys.
Res. Letts. 25, 3599­3602.
Kernthaler, S.C., Toumi, R., Haigh, J.D., 1999. Some doubts concerning a link between
cosmic ray fluxes and global cloudiness. Geophys. Res. Letts. 26, 863­865.
Kuang, Z., Jiang, Y., Yung, Y.L., 1998. Cloud optical thickness variations during
1983­1991: solar cycle or ENSO? Geophys. Res. Letts. 25, 1415­1417.
Kristj'ansson, J.E., Kristiansen, J., 2000. Is there a cosmic ray signal in recent varia­
tions in global cloudiness and cloud radiative forcing? J. Geophys. Res. 105, 11851­
11863.
Liepert, B.G., 1997. Recent changes in solar radiation under cloudy conditions in
Germany. Int. J. Clim. 17, 1581­1593.
Lean, J., Rind, D., 1998. Climate forcing by changing solar radiation. J. Clim. 11,
3069­3094.
Lockwood, M., Stamper, R., 1999. Long­term drift of the coronal source magnetic flux
23

and the total solar irradiance. Geophys. Res. Letts. 26 (16), 2461­2464.
Marsh, N.D., Svensmark, H., 2000. Cosmic rays, clouds and climate. Space Science
Review (in press)
McGuffie, K., Henderson­Sellers, A., 1988. Is Canadian cloudiness increasing? Atmosphere­
Ocean 26, 608­633.
Menzel, W.P., Wylie, D.P., Strabala, K.I., 1996. Seven years of global cirrus cloud
statistics using HIRS. IRS'96: Current problems in atmospheric radiation, Smith and
Stammes (Eds), ISBN 0­937194­39­5.
Norris, J.R., 1999. On trends and possible artifacts in global ocean cloud cover between
1952 and 1995. J. Clim. 12, 1864­1870.
Observers Handbook, 4th Edition, 1982. UK Meteorological Office (Eds). Her Majesty's
Stationery Office, London, p 153.
Pall'e, E., Butler, C.J., 2000. The influence of cosmic rays on terrestrial clouds and
global warming. Astron. & Geophys. 41, 18­22.
Pall'e, E., Butler, C.J., 2001. Sunshine records from Ireland, cloud factors and possible
links to solar activity and cosmic rays. Int. J. Clim. 21, 709­729.
Pudovkin, M.I., Veretenenko, S.V., 1995. Cloudiness decreases associated with Forbush­
decreases of galactic cosmic rays. J. Geophys. Res. 57, 1349­1355.
Ramanathan, V., Cess, R.D., Harrison, E.F., Minnis, P., Barkstrom, B.R., Ahmad,
E., Hartmann, D., 1989. Cloud­radiative forcing and climate. Results from the Earth
Radiation Budget Experiment. Science 243, 52.
Reid, G.C., 1987. Influence of solar variability on global sea surface temperatures.
Nature 129: 142­143.
Rossow, W.B., Walker, A.W., Beuschel, D.E., Roiter, M.D., 1996. 1996: International
Satellite Cloud Climatology Project (ISCCP) Documentation of New Cloud Datasets.
24

WMO/TD­No. 737, World Meteorological Organization.
Seze, G., Drake, F., Desbois, M., Henderson­Sellers, A., 1986. Total and low cloud
amounts over France and southern Britain in the summer of 1983: comparison of
surface­observed and satellite retrieved values. Int. J. Rem. Sens. 7, 1031­1050.
Stahle, D.W., Cleaveland, M.K., Cerveny, R.S., 1991. Tree­ring reconstructed sunshine
duration over central USA. Int. J. Clim. 11, 285­295.
Stanhill, G., 1998a. Estimation of direct solar beam irradiance from measurements of
the duration of bright sunshine. Int. J. Clim. 18, 347­354.
Stanhill, G., 1998b. Long­term trends in, and spatial variations of, solar irradiance in
Ireland. Int. J. Clim. 18, 1015­1030.
Sun, B., Groisman, P.Y., 2000. Cloudiness variations over the Former Soviet Union.
Int. J. Clim. 20, 1097­1111.
Soon, W.H., Posmentier, E.S., Baliunas, S.L., 1996. Inference of solar irradiance vari­
ability from terrestrial temperature changes, 1880­1993: An astrophysical application
of the Sun­Climate connection. Astrophys. J. 472, 891­902.
Svensmark, H., Friis­Christensen, E., 1999. Reply to comments on ``Variation of cosmic
ray flux and global cloud coverage ­ a missing link in solar­terrestrial relationships''. J.
Atmos. Solar­Terrest. Phys. 62, 79­80.
Svensmark, H., Friis­Christensen, E., 1997. Variations of Cosmic Ray Flux and global
cloud coverage. A missing link in solar­climate relationships. J. Atmos. Solar­Terrest.
Phys. 59, 1225­1232.
Svensmark, H., 1998. Influence of cosmic rays on Earth's climate. Phys. Rev. Letts.
81, 5027­5030.
Tinsley, B.A., Rohrbaugh, R.P., Hei, M., Beard, K.V., 2000. Effects of Image Charges
on the Scavenging of Aerosol Particles by Cloud Droplets, and on Droplet Charging
25

and Possible Ice Nucleation Processes. J. Atmos. Sci. 57, 2118­2134.
Tinsley, B.A., Brown, G.M., Scherrer, P.H., 1989. Solar variability influences on
weather and climate ­ possible connections through cosmic­ray fluxes and storm in­
tensification. J. Atmos. Solar­Terrest. Phys. 94, 14783­14792.
Turco, R.P., Zhao, J., Yu, F., 1998. A new source of tropospheric aerosols: Ion­ion
recombination. Geophys. Res. Letts. 25, 635­638.
Udelhofen, P.M., Cess, R.D., 2001. Cloud cover variations over the United States: An
influence of cosmic rays or solar activity? Geophys. Res. Letts. 28, 2617.
Wagner, G., Livingstone, D.M., Masarik, J., Muscheler, R., Bee, J., 2001. Some results
relevant to the discussion of a possible link between cosmic rays and the Earth's climate.
J. Geoph. Res. 106(D4), 3381­3387.
Weare, B.C., 2000. Near­Global observations of low clouds.J. Clim. 13, 1255­1268.
Wentz, F.J., Schabel, M., 2000. Precise climate monitoring using complementary
satellite data sets. Nature 403, 414­416.
Yu, F.Q., Turco, R.P., 2001. From molecular clusters to nanoparticles: Role of ambient
ionization in tropospheric aerosol formation. J. Geoph. Res. 106(D5), 4797­4814.
26

Figure 1: Mean annual sunshine factor for the four Irish sites (1881­
1998).
27

Dataset Total Cloud Low Cloud High Cloud Period
Ground Data Trends
1 Sunshine in Ireland (decrease) Increase -- -- 1881­1998
2 Sunshine in Israel (decrease) Increase -- -- 1979­1995
3 Sunshine over midwestern USA (decrease) Increase -- -- 1901­1977
4 Sunshine over USA (decrease) Increase -- -- 1950­1988
5 Sunshine Central USA (tree­rings) Stable -- -- 1700­1980
6 Synoptic Clouds over FUSSR Increase Decrease Increase 1936­1990
7 Synoptic Cloud over Oceans Increase Increase -- 1952­1995
8 Synoptic Cloud over Australia Increase -- -- 1900­1990
9 Synoptic Cloud over North America Increase -- -- 1900­1990
10 Synoptic Cloud over India Increase -- -- 1900­1990
11 Synoptic Cloud over Europe Increase -- -- 1900­1990
12 Synoptic Cloud over Canada Increase -- -- 1900­1954
13 Synoptic Cloud over Arctic Canada Increase -- -- 1900­1980
14 Synoptic Cloud over China Decrease -- -- 1954­1994
15 Ground­based solar radiation (various) Increase -- -- 1960­2000
Satellite Data Trends
16 ISCCP C2 Stable -- Increase 1985­1988
17 ISCCP D2 Stable Stable Stable 1983­1994
18 DMSP (water clouds over Oceans) Increase -- -- 1988­1998
19 HIRS (only Cirrus) -- -- Increase 1989­1996
Table 1: Sunshine records and synoptic cloud data compilations from the ground. Satel­
lite cloud measurements from different satellites are also displayed, however the short
duration of those records makes them unsuitable for long­term studies. Cloud types
reported as stable, as in the case of ISCCP C2 and D2, are in fact quite variable over
the duration of the data, and a definite trend cannot be clearly established. Refer­
ences: 1 Pall'e and Butler (2001); 2 Stanhill (1998a); 3 Changnon (1981); 4 Angell (1990);
5 Stahle et al. (1991); 6 Sun and Groisman (2000); 7 Norris (1999); 8 Henderson­Sellers
(1992); 9 Henderson­Sellers (1992); 10 Henderson­Sellers (1992); 11 Henderson­Sellers
(1992); 12 McGuffie and Henderson­Sellers (1988); 13 McGuffie and Henderson­Sellers
(1988); 14 Kaiser (1998); 15 Stanhill, 1998b and therein; Liepert, 1997; 16 Kernthaler et al.,
1999; 17 Pall'e and Butler, 2000; 18 Kristj'ansson and Kristiansen, 2000; 19 Menzel et al.,
1996.
28

Question ISCCP­D2 DMSP Synoptic Clouds Sunshine
Data reliability Good Good Not Instrumental Good
Time of observations Day+Night Day only Day only Day only
Geographical Coverage Global Oceans Local Local
Main Problem Duration Duration Spurious trends? --
Is there a trend in cloud cover? -- -- Increase Increase
Can we distinguish cloud types? Yes Water clouds Total/Low No
Is total cloud correlated with GCR? No -- No No
Is high cloud correlated with GCR? No -- -- --
Is mid cloud correlated with GCR? No -- -- --
Is low cloud correlated with GCR? Yes Yes? No --
Table 2: A brief summary of the distinguishing features of two sets of satellite cloud
data and the two types of ground­based cloud data. See text for details.
29

Figure 1:
30